d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2b519b1707
freebsd-dev/usr.sbin/mountd/mountd.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

4018 lines
97 KiB
C
Raw Normal View History

General further adoption of SPDX licensing ID tags. Mainly focus on files that use BSD 3-Clause license. The Software Package Data Exchange (SPDX) group provides a specification to make it easier for automated tools to detect and summarize well known opensource licenses. We are gradually adopting the specification, noting that the tags are considered only advisory and do not, in any way, superceed or replace the license texts. Special thanks to Wind River for providing access to "The Duke of Highlander" tool: an older (2014) run over FreeBSD tree was useful as a starting point.
2017-11-20 19:49:47 +00:00
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Herb Hasler and Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
Renumber copyright clause 4 Renumber cluase 4 to 3, per what everybody else did when BSD granted them permission to remove clause 3. My insistance on keeping the same numbering for legal reasons is too pedantic, so give up on that point. Submitted by: Jan Schaumann <jschauma@stevens.edu> Pull Request: https://github.com/freebsd/freebsd/pull/96
2017-02-28 23:42:47 +00:00
* 3. Neither the name of the University nor the names of its contributors
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#ifndef lint
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
static const char copyright[] =
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
"@(#) Copyright (c) 1989, 1993\n\
The Regents of the University of California. All rights reserved.\n";
Automatically load NFS and a bevy of other filesystems.
1994-09-22 22:17:02 +00:00
#endif /*not lint*/
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
#if 0
s/rpc.mountd/mountd/ Add FBSDID Reorder #if/#endif around sccsid to conform style(9)
2002-10-16 16:04:50 +00:00
#ifndef lint
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
static char sccsid[] = "@(#)mountd.c 8.15 (Berkeley) 5/1/95";
Automatically load NFS and a bevy of other filesystems.
1994-09-22 22:17:02 +00:00
#endif /*not lint*/
s/rpc.mountd/mountd/ Add FBSDID Reorder #if/#endif around sccsid to conform style(9)
2002-10-16 16:04:50 +00:00
#endif
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#include <sys/param.h>
Make use of the UID_NOBODY and GID_NOGROUP definitions in sys/conf.h. r362214 exposed UID_NOBODY and GID_NOGROUP to userspace, so use them instead of the numbers. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D25281
2020-06-16 02:35:30 +00:00
#include <sys/conf.h>
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
#include <sys/fcntl.h>
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
#include <sys/fnv_hash.h>
Use fgetln to remove the static limit on the length of lines in /etc/exports. Oversized lines were unlikely due to the large 10k limit, but any found would cause mountd to exit with an error. Also fix one or two compiler warnings.
2002-05-14 23:24:28 +00:00
#include <sys/linker.h>
#include <sys/module.h>
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
#include <sys/mount.h>
mountd: Convert exportlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for the exportlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12502
2017-10-04 08:43:56 +00:00
#include <sys/queue.h>
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
#include <sys/stat.h>
#include <sys/sysctl.h>
#include <sys/syslog.h>
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#include <rpc/rpc.h>
Implement nonblocking tpc-connections. rpcgen -m does still produce backcompatible code. Reviewed by: rwatson Obtained from: NetBSD MFC after: 1 day
2003-01-16 07:27:30 +00:00
#include <rpc/rpc_com.h>
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#include <rpc/pmap_clnt.h>
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
#include <rpc/pmap_prot.h>
#include <rpcsvc/mount.h>
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
#include <nfs/nfsproto.h>
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
#include <nfs/nfssvc.h>
Userland part of nfs client/server split and cleanup.
2001-09-18 23:34:44 +00:00
#include <nfsserver/nfs.h>
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
#include <fs/nfs/nfsport.h>
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#include <arpa/inet.h>
#include <ctype.h>
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
#include <err.h>
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#include <errno.h>
#include <grp.h>
Use pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock anymore.
2005-08-24 19:17:06 +00:00
#include <libutil.h>
Use the standardized CHAR_BIT constant instead of NBBY in userland.
2002-09-25 04:06:37 +00:00
#include <limits.h>
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#include <netdb.h>
#include <pwd.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "pathnames.h"
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
#include "mntopts.h"
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#ifdef DEBUG
#include <stdarg.h>
#endif
/*
* Structures for keeping the mount list and export list
*/
struct mountlist {
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
char ml_host[MNTNAMLEN+1];
char ml_dirp[MNTPATHLEN+1];
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_ENTRY(mountlist) next;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
};
struct dirlist {
struct dirlist *dp_left;
struct dirlist *dp_right;
int dp_flag;
struct hostlist *dp_hosts; /* List of hosts this dir exported to */
mountd: Replace malloc+strcpy to strdup Reviewed by: bapt MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12503
2017-09-26 09:18:18 +00:00
char *dp_dirp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
};
/* dp_flag bits */
#define DP_DEFSET 0x1
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
#define DP_HOSTSET 0x2
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
/*
* maproot/mapall credentials.
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
* cr_smallgrps can be used for a group list up to SMALLNGROUPS in size.
* Larger group lists are malloc'd/free'd.
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
*/
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
#define SMALLNGROUPS 32
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct expcred {
uid_t cr_uid;
int cr_ngroups;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
gid_t cr_smallgrps[SMALLNGROUPS];
gid_t *cr_groups;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
};
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct exportlist {
struct dirlist *ex_dirl;
struct dirlist *ex_defdir;
Replace global list for grouplist with list(s) for each exportlist element. In mountd.c, the grouplist structures are linked into a single global linked list headed by "grphead". The only use of this linked list is to free all list elements when the exportlist elements are also all being free'd at the time the exports are being reloaded. This patch replaces this one global linked list head with a list head in each exportlist structure, where the grouplist elements for that exported file system are linked. The only change is that now the grouplist elements are free'd with the associated exportlist element as they are free'd instead of all grouplist elements being free'd after the exportlist elements are free'd. This change should have no effect in practice. This is being done, since a future patch that will add a "-I" option for incrementally updating the exports in the kernel needs to know which grouplist elements are associated with each exported file system and having them linked into a list headed by the exportlist element does that. MFC after: 1 month
2019-05-14 22:00:47 +00:00
struct grouplist *ex_grphead;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
int ex_flag;
fsid_t ex_fs;
char *ex_fsdir;
Merge WebNFS support from NetBSD. Obtained from: NetBSD
1997-07-16 09:27:53 +00:00
char *ex_indexfile;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct expcred ex_defanon;
uint64_t ex_defexflags;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int ex_numsecflavors;
int ex_secflavors[MAXSECFLAVORS];
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
int ex_defnumsecflavors;
int ex_defsecflavors[MAXSECFLAVORS];
mountd: Convert exportlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for the exportlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12502
2017-10-04 08:43:56 +00:00
SLIST_ENTRY(exportlist) entries;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
};
/* ex_flag bits */
#define EX_LINKED 0x1
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
#define EX_DONE 0x2
#define EX_DEFSET 0x4
#define EX_PUBLICFH 0x8
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
SLIST_HEAD(exportlisthead, exportlist);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct netmsk {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
struct sockaddr_storage nt_net;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
struct sockaddr_storage nt_mask;
Here is a patch to make mountd work. It just replace u_long with u_int32_t and shouldn't affect on i386. Without this patch, - unaligned accesses occur - permission denied randomly Submitted by: Hidetoshi Shimokawa <simokawa@sat.t.u-tokyo.ac.jp>
1998-12-29 09:38:49 +00:00
char *nt_name;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
};
union grouptypes {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
struct addrinfo *gt_addrinfo;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct netmsk gt_net;
};
struct grouplist {
int gr_type;
union grouptypes gr_ptr;
struct grouplist *gr_next;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct expcred gr_anon;
uint64_t gr_exflags;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
int gr_flag;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
int gr_numsecflavors;
int gr_secflavors[MAXSECFLAVORS];
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
};
/* Group types */
#define GT_NULL 0x0
#define GT_HOST 0x1
#define GT_NET 0x2
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
#define GT_DEFAULT 0x3
Fixed stupid bug in mountd: it would seem that the kernel doesn't allow you to push the same host into its NFS export lists twice, but mountd tries to do it anyway. This means that putting: /some_file_system -ro host1 host1 in your /etc/exports file causes an error. This is bogus: mountd should be smart enough to ignore the second instance of host1. This can be a problem in some configurations that use netgroups. For example, each host in my netgroups database is has two entries: startide (startide,-,) (startide.ctr.columbia.edu,-,) When mountd sees this, it tries to put startide.ctr.columbia.edu into the export list *twice*. Just listing 'startide' /etc/exports list will also screw up because mountd will try to resolve the netgroup 'startide' instead of the hostname 'startide.' My solution is watch for duplicate entries in get_host() and mark them as grouptype GT_IGNORE, which do_mount() will now cheefully throw away. This is a bit of a kludge, but it was the least obtrusive fix I could come up with. Also silenced a compiler warning: arguments passwd to xdr_long() should be u_long, not int. :)
1995-03-26 23:36:38 +00:00
#define GT_IGNORE 0x5
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/* Group flags */
#define GR_FND 0x1
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct hostlist {
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
int ht_flag; /* Uses DP_xx bits */
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct grouplist *ht_grp;
struct hostlist *ht_next;
};
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
struct fhreturn {
int fhr_flag;
int fhr_vers;
nfsfh_t fhr_fh;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int fhr_numsecflavors;
int *fhr_secflavors;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
};
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
#define GETPORT_MAXTRY 20 /* Max tries to get a port # */
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
/*
* How long to delay a reload of exports when there are RPC request(s)
* to process, in usec. Must be less than 1second.
*/
#define RELOADDELAY 250000
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/* Global defs */
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static char *add_expdir(struct dirlist **, char *, int);
static void add_dlist(struct dirlist **, struct dirlist *,
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
struct grouplist *, int, struct exportlist *,
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct expcred *, uint64_t);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void add_mlist(char *, char *);
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
static int check_path_component(const char *, char **);
static int check_dirpath(char *, char **);
static int check_statfs(const char *, struct statfs *, char **);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int check_options(struct dirlist *);
static int checkmask(struct sockaddr *sa);
static int chk_host(struct dirlist *, struct sockaddr *, int *, int *,
int *, int **);
Allow /etc/exports to contain usernames/groups with spaces in them. If you are getting your users/groups from a directory service such as LDAP or AD it's possible for those usernames or groupnames to contain spaces. Submitted by: Sean E. Fagan Reviewed by: rmacklem MFC after: 1 week Sponsored by: iXsystems
2016-01-07 05:34:39 +00:00
static char *strsep_quote(char **stringp, const char *delim);
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
static int create_service(struct netconfig *nconf);
static void complete_service(struct netconfig *nconf, char *port_str);
static void clearout_service(void);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void del_mlist(char *hostp, char *dirp);
static struct dirlist *dirp_search(struct dirlist *, char *);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static int do_export_mount(struct exportlist *, struct statfs *);
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
static int do_mount(struct exportlist *, struct grouplist *, uint64_t,
struct expcred *, char *, int, struct statfs *, int, int *);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int do_opt(char **, char **, struct exportlist *,
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct grouplist *, int *, uint64_t *, struct expcred *);
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
static struct exportlist *ex_search(fsid_t *, struct exportlisthead *);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct exportlist *get_exp(void);
static void free_dir(struct dirlist *);
static void free_exp(struct exportlist *);
static void free_grp(struct grouplist *);
static void free_host(struct hostlist *);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static void free_v4rootexp(void);
static void get_exportlist_one(int);
static void get_exportlist(int);
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
static void insert_exports(struct exportlist *, struct exportlisthead *);
static void free_exports(struct exportlisthead *);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static void read_exportfile(int);
static int compare_nmount_exportlist(struct iovec *, int, char *);
static int compare_export(struct exportlist *, struct exportlist *);
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
static int compare_cred(struct expcred *, struct expcred *);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static int compare_secflavor(int *, int *, int);
Factor code into two new functions in preparation for a future commit. Factor code into two functions. read_exportfile() a functon which reads the exports file(s) and calls get_exportlist_one() to process each of them. delete_export() a function which deletes the exports in the kernel for a file system. The contents of these functions is just the same code as was used to do the operations, moved into separate functions. As such, there is no semantic change. This is being done in preparation for a future commit that will add an option to do incremental changes of kernel exports upon receiving SIGHUP. MFC after: 1 month
2019-05-11 22:41:58 +00:00
static void delete_export(struct iovec *, int, struct statfs *, char *);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int get_host(char *, struct grouplist *, struct grouplist *);
static struct hostlist *get_ht(void);
static int get_line(void);
static void get_mountlist(void);
static int get_net(char *, struct netmsk *, int);
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
static void getexp_err(struct exportlist *, struct grouplist *, const char *);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct grouplist *get_grp(void);
static void hang_dirp(struct dirlist *, struct grouplist *,
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct exportlist *, int, struct expcred *, uint64_t);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void huphandler(int sig);
static int makemask(struct sockaddr_storage *ssp, int bitlen);
static void mntsrv(struct svc_req *, SVCXPRT *);
static void nextfield(char **, char **);
static void out_of_mem(void);
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
static void parsecred(char *, struct expcred *);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int parsesec(char *, struct exportlist *);
static int put_exlist(struct dirlist *, XDR *, struct dirlist *,
int *, int);
static void *sa_rawaddr(struct sockaddr *sa, int *nbytes);
static int sacmp(struct sockaddr *sa1, struct sockaddr *sa2,
struct sockaddr *samask);
static int scan_tree(struct dirlist *, struct sockaddr *);
static void usage(void);
static int xdr_dir(XDR *, char *);
static int xdr_explist(XDR *, caddr_t);
static int xdr_explist_brief(XDR *, caddr_t);
static int xdr_explist_common(XDR *, caddr_t, int);
static int xdr_fhs(XDR *, caddr_t);
static int xdr_mlist(XDR *, caddr_t);
static void terminate(int);
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
static void cp_cred(struct expcred *, struct expcred *);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
#define EXPHASH(f) (fnv_32_buf((f), sizeof(fsid_t), 0) % exphashsize)
static struct exportlisthead *exphead = NULL;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static struct exportlisthead *oldexphead = NULL;
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
static int exphashsize = 0;
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
static SLIST_HEAD(, mountlist) mlhead = SLIST_HEAD_INITIALIZER(&mlhead);
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static char *exnames_default[2] = { _PATH_EXPORTS, NULL };
static char **exnames;
static char **hosts = NULL;
static int force_v2 = 0;
static int resvport_only = 1;
static int nhosts = 0;
static int dir_only = 1;
static int dolog = 0;
static int got_sighup = 0;
static int xcreated = 0;
static char *svcport_str = NULL;
static int mallocd_svcport = 0;
static int *sock_fd;
static int sock_fdcnt;
static int sock_fdpos;
static int suspend_nfsd = 0;
static int opt_flags;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
static int have_v6 = 1;
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int v4root_phase = 0;
static char v4root_dirpath[PATH_MAX + 1];
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static struct exportlist *v4root_ep = NULL;
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int has_publicfh = 0;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
static int has_set_publicfh = 0;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct pidfh *pfh = NULL;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/* Bits for opt_flags above */
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#define OP_MAPROOT 0x01
#define OP_MAPALL 0x02
Userland part of nfs client/server split and cleanup.
2001-09-18 23:34:44 +00:00
/* 0x4 free */
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#define OP_MASK 0x08
#define OP_NET 0x10
#define OP_ALLDIRS 0x40
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
#define OP_HAVEMASK 0x80 /* A mask was specified or inferred. */
Try to give a more descriptive error message for the pilot error of attempting to export the non-root of a filesystem with -alldirs. This pilot error seems to be very common, and the "could not remount" error message doesn't give much hints about the real reason. See the old PR below for an example. While i was at it, make it possible to entirely omit the often annoying error message in that case by specifying the "quiet" exports flag. This allows to specify something like /cdrom -alldirs,ro,quiet <where to export to> which will silently fail if nothing is mounted under /cdrom, but do the rigth thing as soon as you mount something. While doing this, i've put the embedded example in the exports(5) man page into a subsection of its own as it ought to be. Thanks for Paul Southworth for reminding me about this problem. PR: bin/4448 MFC after: 1 month
2002-07-18 20:52:17 +00:00
#define OP_QUIET 0x100
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
#define OP_MASKLEN 0x200
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#define OP_SEC 0x400
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#ifdef DEBUG
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int debug = 1;
static void SYSLOG(int, const char *, ...) __printflike(2, 3);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#define syslog SYSLOG
#else
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int debug = 0;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
#endif
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/*
* The LOGDEBUG() syslog() calls are always compiled into the daemon.
* To enable them, create a file at _PATH_MOUNTDDEBUG. This file can be empty.
* To disable the logging, just delete the file at _PATH_MOUNTDDEBUG.
*/
static int logdebug = 0;
#define LOGDEBUG(format, ...) \
(logdebug ? syslog(LOG_DEBUG, format, ## __VA_ARGS__) : 0)
Allow /etc/exports to contain usernames/groups with spaces in them. If you are getting your users/groups from a directory service such as LDAP or AD it's possible for those usernames or groupnames to contain spaces. Submitted by: Sean E. Fagan Reviewed by: rmacklem MFC after: 1 week Sponsored by: iXsystems
2016-01-07 05:34:39 +00:00
/*
* Similar to strsep(), but it allows for quoted strings
* and escaped characters.
*
* It returns the string (or NULL, if *stringp is NULL),
* which is a de-quoted version of the string if necessary.
*
* It modifies *stringp in place.
*/
static char *
strsep_quote(char **stringp, const char *delim)
{
char *srcptr, *dstptr, *retval;
char quot = 0;
if (stringp == NULL || *stringp == NULL)
return (NULL);
srcptr = dstptr = retval = *stringp;
while (*srcptr) {
/*
* We're looking for several edge cases here.
* First: if we're in quote state (quot != 0),
* then we ignore the delim characters, but otherwise
* process as normal, unless it is the quote character.
* Second: if the current character is a backslash,
* we take the next character as-is, without checking
* for delim, quote, or backslash. Exception: if the
* next character is a NUL, that's the end of the string.
* Third: if the character is a quote character, we toggle
* quote state.
* Otherwise: check the current character for NUL, or
* being in delim, and end the string if either is true.
*/
if (*srcptr == '\\') {
srcptr++;
/*
* The edge case here is if the next character
* is NUL, we want to stop processing. But if
* it's not NUL, then we simply want to copy it.
*/
if (*srcptr) {
*dstptr++ = *srcptr++;
}
continue;
}
if (quot == 0 && (*srcptr == '\'' || *srcptr == '"')) {
quot = *srcptr++;
continue;
}
if (quot && *srcptr == quot) {
/* End of the quoted part */
quot = 0;
srcptr++;
continue;
}
if (!quot && strchr(delim, *srcptr))
break;
*dstptr++ = *srcptr++;
}
*stringp = (*srcptr == '\0') ? NULL : srcptr + 1;
Fix strsep_quote() on strings without quotes. For strings without quotes and escapes dstptr and srcptr are equal, so zeroing *dstptr before checking *srcptr is not a good idea. In practice it means that in -maproot=65534:65533 everything after the colon is lost. The problem was there since r293305, but before r346976 it was covered by improper strsep_quote() usage. PR: 238725 MFC after: 3 days Sponsored by: iXsystems, Inc.
2019-06-25 17:00:53 +00:00
*dstptr = 0; /* Terminate the string */
Allow /etc/exports to contain usernames/groups with spaces in them. If you are getting your users/groups from a directory service such as LDAP or AD it's possible for those usernames or groupnames to contain spaces. Submitted by: Sean E. Fagan Reviewed by: rmacklem MFC after: 1 week Sponsored by: iXsystems
2016-01-07 05:34:39 +00:00
return (retval);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Mountd server for NFS mount protocol as described in:
* NFS: Network File System Protocol Specification, RFC1094, Appendix A
* The optional arguments are the exports file name
* default: _PATH_EXPORTS
* and "-n" to allow nonroot mount.
*/
int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
main(int argc, char **argv)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
fd_set readfds;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
struct netconfig *nconf;
char *endptr, **hosts_bak;
void *nc_handle;
Use pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock anymore.
2005-08-24 19:17:06 +00:00
pid_t otherpid;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
in_port_t svcport;
int c, k, s;
Implement nonblocking tpc-connections. rpcgen -m does still produce backcompatible code. Reviewed by: rwatson Obtained from: NetBSD MFC after: 1 day
2003-01-16 07:27:30 +00:00
int maxrec = RPC_MAXDATASIZE;
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
int attempt_cnt, port_len, port_pos, ret;
char **port_list;
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
uint64_t curtime, nexttime;
struct timeval tv;
struct timespec tp;
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
sigset_t sig_mask, sighup_mask;
int enable_rpcbind;
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
enable_rpcbind = 1;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
/* Check that another mountd isn't already running. */
Pidfiles should be created with permission preventing users from opening them for reading. When user can open file for reading, he can also flock(2) it, which can lead to confusions. Pointed out by: green
2005-09-16 11:24:28 +00:00
pfh = pidfile_open(_PATH_MOUNTDPID, 0600, &otherpid);
Use pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock anymore.
2005-08-24 19:17:06 +00:00
if (pfh == NULL) {
if (errno == EEXIST)
errx(1, "mountd already running, pid: %d.", otherpid);
warn("cannot open or create pidfile");
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
s = socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP);
if (s < 0)
have_v6 = 0;
else
close(s);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
while ((c = getopt(argc, argv, "2deh:lnp:RrS")) != -1)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
switch (c) {
Add a -2 flag to mountd to allow an admin to disable NFSv3 services. This is handy for testing and possibly to work around busted v3 clients.
1997-04-22 10:37:27 +00:00
case '2':
force_v2 = 1;
break;
Change the "-4" argument for nfsd and mountd to "-e" to avoid confusion, since it does not refer to IPv4 nor NFSv4, but to running the experimental server instead of the regular one. Approved by: kib (mentor)
2009-05-28 20:28:13 +00:00
case 'e':
This patch changes the default NFS server to the new one, which was referred to as the experimental server. It also adds a new command line option "-o" to both mountd and nfsd that forces them to use the old/regular NFS server. The "-e" option for these commands is now a no-op, since the new server is the default. I will be committing rc script and man changes soon. Discussed on freebsd-fs@.
2011-04-24 00:06:56 +00:00
/* now a no-op, since this is the default */
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
break;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
case 'n':
resvport_only = 0;
break;
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
case 'R':
/* Do not support Mount protocol */
enable_rpcbind = 0;
break;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
case 'r':
dir_only = 0;
break;
Add a undocumented '-d' flag to set debugging.
1995-05-21 19:31:09 +00:00
case 'd':
debug = debug ? 0 : 1;
break;
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
case 'l':
Rename the 'log' variable rather than -fno-builtin-log
2003-10-30 22:57:43 +00:00
dolog = 1;
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
break;
Add a new option to mountd(8), -p <port>. This allows the user to specify a known port for use in firewall rulesets; otherwise the port is chosen at run-time by bindresvport(). MFC after: 1 week
2004-03-04 04:42:52 +00:00
case 'p':
endptr = NULL;
svcport = (in_port_t)strtoul(optarg, &endptr, 10);
if (endptr == NULL || *endptr != '\0' ||
svcport == 0 || svcport >= IPPORT_MAX)
usage();
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
svcport_str = strdup(optarg);
break;
case 'h':
++nhosts;
hosts_bak = hosts;
hosts_bak = realloc(hosts, nhosts * sizeof(char *));
if (hosts_bak == NULL) {
if (hosts != NULL) {
for (k = 0; k < nhosts; k++)
free(hosts[k]);
free(hosts);
out_of_mem();
}
}
hosts = hosts_bak;
hosts[nhosts - 1] = strdup(optarg);
if (hosts[nhosts - 1] == NULL) {
for (k = 0; k < (nhosts - 1); k++)
free(hosts[k]);
free(hosts);
out_of_mem();
}
Add a new option to mountd(8), -p <port>. This allows the user to specify a known port for use in firewall rulesets; otherwise the port is chosen at run-time by bindresvport(). MFC after: 1 week
2004-03-04 04:42:52 +00:00
break;
Add a new '-S' option to mountd, which tells it to suspend execution of the nfsd threads while it is reloading the exports. This avoids clients from getting intermittent access errors when the exports are being reloaded non-atomically. It is not an ideal solution, since requests will back up while the nfsd threads are suspended. Also, when this option is used, if mountd crashes while reloading exports, mountd will have to be restarted to get the nfsd threads to resume execution. This has been tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). The nfse patch offers a more comprehensive solution for this issue. PR: kern/9619, kern/131342 Reviewed by: kib MFC after: 2 weeks
2012-10-15 00:17:16 +00:00
case 'S':
suspend_nfsd = 1;
break;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
default:
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
usage();
Cleanup unnecessary semicolons from utilities we all love.
2016-04-15 22:31:22 +00:00
}
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
if (enable_rpcbind == 0) {
if (svcport_str != NULL) {
warnx("-p option not compatible with -R, ignored");
free(svcport_str);
svcport_str = NULL;
}
if (nhosts > 0) {
warnx("-h option not compatible with -R, ignored");
for (k = 0; k < nhosts; k++)
free(hosts[k]);
free(hosts);
hosts = NULL;
nhosts = 0;
}
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
Remove oldnfs leftovers from mountd(8). Reviewed by: rmacklem@ Sponsored by: The FreeBSD Foundation
2015-04-29 12:33:00 +00:00
if (modfind("nfsd") < 0) {
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
/* Not present in kernel, try loading it */
Remove oldnfs leftovers from mountd(8). Reviewed by: rmacklem@ Sponsored by: The FreeBSD Foundation
2015-04-29 12:33:00 +00:00
if (kldload("nfsd") < 0 || modfind("nfsd") < 0)
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
errx(1, "NFS server is not available");
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
argc -= optind;
argv += optind;
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
if (argc > 0)
exnames = argv;
else
exnames = exnames_default;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
openlog("mountd", LOG_PID, LOG_DAEMON);
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("getting export list");
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
get_exportlist(0);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("getting mount list");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
get_mountlist();
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("here we go");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (debug == 0) {
daemon(0, 0);
signal(SIGINT, SIG_IGN);
signal(SIGQUIT, SIG_IGN);
}
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
signal(SIGHUP, huphandler);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
signal(SIGTERM, terminate);
Ignore SIGPIPE, instead of terminating process. PR: 81230 Submitted by: Dmitrij Tejblum <tejblum yandex-team ru> MFC after: 1 month
2006-11-18 22:43:20 +00:00
signal(SIGPIPE, SIG_IGN);
Use pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock anymore.
2005-08-24 19:17:06 +00:00
pidfile_write(pfh);
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
if (enable_rpcbind != 0) {
rpcb_unset(MOUNTPROG, MOUNTVERS, NULL);
rpcb_unset(MOUNTPROG, MOUNTVERS3, NULL);
rpc_control(RPC_SVC_CONNMAXREC_SET, &maxrec);
if (!resvport_only) {
if (sysctlbyname("vfs.nfsd.nfs_privport", NULL, NULL,
&resvport_only, sizeof(resvport_only)) != 0 &&
errno != ENOENT) {
syslog(LOG_ERR, "sysctl: %m");
exit(1);
}
Make a sysconfig variable controlling if teh kernel should accept nfs requests from non-privileged ports. Change mountd such that it does never set this variable, but only clears it when run with -n. Also document this in the man page.
1997-04-09 20:17:15 +00:00
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
/*
* If no hosts were specified, add a wildcard entry to bind to
* INADDR_ANY. Otherwise make sure 127.0.0.1 and ::1 are added
* to the list.
*/
if (nhosts == 0) {
hosts = malloc(sizeof(char *));
if (hosts == NULL)
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
out_of_mem();
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
hosts[0] = "*";
nhosts = 1;
} else {
hosts_bak = hosts;
if (have_v6) {
hosts_bak = realloc(hosts, (nhosts + 2) *
sizeof(char *));
if (hosts_bak == NULL) {
for (k = 0; k < nhosts; k++)
free(hosts[k]);
free(hosts);
out_of_mem();
} else
hosts = hosts_bak;
nhosts += 2;
hosts[nhosts - 2] = "::1";
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
} else {
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
hosts_bak = realloc(hosts, (nhosts + 1) *
sizeof(char *));
if (hosts_bak == NULL) {
for (k = 0; k < nhosts; k++)
free(hosts[k]);
free(hosts);
out_of_mem();
} else {
nhosts += 1;
hosts = hosts_bak;
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
hosts[nhosts - 1] = "127.0.0.1";
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
attempt_cnt = 1;
sock_fdcnt = 0;
sock_fd = NULL;
port_list = NULL;
port_len = 0;
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
if (enable_rpcbind != 0) {
nc_handle = setnetconfig();
while ((nconf = getnetconfig(nc_handle))) {
if (nconf->nc_flag & NC_VISIBLE) {
if (have_v6 == 0 && strcmp(nconf->nc_protofmly,
"inet6") == 0) {
/* DO NOTHING */
} else {
ret = create_service(nconf);
if (ret == 1)
/* Ignore this call */
continue;
if (ret < 0) {
/*
* Failed to bind port, so close
* off all sockets created and
* try again if the port# was
* dynamically assigned via
* bind(2).
*/
clearout_service();
if (mallocd_svcport != 0 &&
attempt_cnt <
GETPORT_MAXTRY) {
free(svcport_str);
svcport_str = NULL;
mallocd_svcport = 0;
} else {
errno = EADDRINUSE;
syslog(LOG_ERR,
"bindresvport_sa:"
" %m");
exit(1);
}
/*
* Start over at the first
* service.
*/
free(sock_fd);
sock_fdcnt = 0;
sock_fd = NULL;
nc_handle = setnetconfig();
attempt_cnt++;
} else if (mallocd_svcport != 0 &&
attempt_cnt == GETPORT_MAXTRY) {
/*
* For the last attempt, allow
* different port #s for each
* nconf by saving the
* svcport_str setting it back
* to NULL.
*/
port_list = realloc(port_list,
(port_len + 1) *
sizeof(char *));
if (port_list == NULL)
out_of_mem();
port_list[port_len++] =
svcport_str;
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
svcport_str = NULL;
mallocd_svcport = 0;
}
}
}
}
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
/*
* Successfully bound the ports, so call complete_service() to
* do the rest of the setup on the service(s).
*/
sock_fdpos = 0;
port_pos = 0;
nc_handle = setnetconfig();
while ((nconf = getnetconfig(nc_handle))) {
if (nconf->nc_flag & NC_VISIBLE) {
if (have_v6 == 0 && strcmp(nconf->nc_protofmly,
"inet6") == 0) {
/* DO NOTHING */
} else if (port_list != NULL) {
if (port_pos >= port_len) {
syslog(LOG_ERR, "too many"
" port#s");
exit(1);
}
complete_service(nconf,
port_list[port_pos++]);
} else
complete_service(nconf, svcport_str);
}
}
endnetconfig(nc_handle);
free(sock_fd);
if (port_list != NULL) {
for (port_pos = 0; port_pos < port_len; port_pos++)
free(port_list[port_pos]);
free(port_list);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
if (xcreated == 0) {
syslog(LOG_ERR, "could not create any services");
exit(1);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
/* Expand svc_run() here so that we can call get_exportlist(). */
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
curtime = nexttime = 0;
sigemptyset(&sighup_mask);
sigaddset(&sighup_mask, SIGHUP);
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
for (;;) {
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
clock_gettime(CLOCK_MONOTONIC, &tp);
curtime = tp.tv_sec;
curtime = curtime * 1000000 + tp.tv_nsec / 1000;
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
sigprocmask(SIG_BLOCK, &sighup_mask, &sig_mask);
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
if (got_sighup && curtime >= nexttime) {
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
got_sighup = 0;
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
get_exportlist(1);
clock_gettime(CLOCK_MONOTONIC, &tp);
nexttime = tp.tv_sec;
nexttime = nexttime * 1000000 + tp.tv_nsec / 1000 +
RELOADDELAY;
} else
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
/*
* If a reload is pending, poll for received request(s),
* otherwise set a RELOADDELAY timeout, since a SIGHUP
* could be processed between the got_sighup test and
* the select() system call.
*/
tv.tv_sec = 0;
if (got_sighup)
tv.tv_usec = 0;
else
tv.tv_usec = RELOADDELAY;
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
if (enable_rpcbind != 0) {
readfds = svc_fdset;
switch (select(svc_maxfd + 1, &readfds, NULL, NULL,
&tv)) {
case -1:
if (errno == EINTR) {
/* Allow a reload now. */
nexttime = 0;
continue;
}
syslog(LOG_ERR, "mountd died: select: %m");
exit(1);
case 0:
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
/* Allow a reload now. */
nexttime = 0;
continue;
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
default:
svc_getreqset(&readfds);
Fix mountd so that it will not lose SIGHUPs that indicate "reload exports". Without this patch, if a SIGHUP is handled while the process is executing get_exportlist(), that SIGHUP is essentially ignored because the got_sighup variable is reset to 0 after get_exportlist(). This results in the exports file(s) not being reloaded until another SIGHUP signal is sent to mountd. This patch fixes this by resetting got_sighup to zero before the get_exportlist() call while SIGHUP is blocked. It also defines a delay time of 250msec before doing another exports reload if there are RPC request(s) to process. This prevents repeated exports reloads from delaying handling of RPC requests significantly. PR: 246597 Reported by: patrykkotlowski@gmail.com Tested by: patrykkotlowski@gmail.com Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D25127
2020-06-06 00:40:02 +00:00
}
Add "-R" option to tell mountd to not register with rpcbind. rpcbind is now considered a security risk for some sites. Since an NFSv4 only NFS server does not need rpcbind, it makes sense to have an option that implements this. This patch adds a "-R" option that disables the Mount protocol (not used by NFSv4) and does not register with rpcbind. Changes are required to /etc/rc.d/mountd and /etc/rc.d/nfsd. Those will be in a separate commit. Reviewed by: freqlabs, asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D26746
2020-10-24 22:48:28 +00:00
} else {
/* Simply wait for a signal. */
sigsuspend(&sig_mask);
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
}
}
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
}
/*
* This routine creates and binds sockets on the appropriate
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
* addresses. It gets called one time for each transport.
mountd(8): Fix a typo in a source code comment - s/ingore/ignore/ MFC after: 3 days
2023-01-19 13:23:51 +00:00
* It returns 0 upon success, 1 for ignore the call and -1 to indicate
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
* bind failed with EADDRINUSE.
* Any file descriptors that have been created are stored in sock_fd and
* the total count of them is maintained in sock_fdcnt.
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
*/
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
static int
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
create_service(struct netconfig *nconf)
{
struct addrinfo hints, *res = NULL;
struct sockaddr_in *sin;
struct sockaddr_in6 *sin6;
struct __rpc_sockinfo si;
int aicode;
int fd;
int nhostsbak;
int one = 1;
int r;
u_int32_t host_addr[4]; /* IPv4 or IPv6 */
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
int mallocd_res;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
if ((nconf->nc_semantics != NC_TPI_CLTS) &&
(nconf->nc_semantics != NC_TPI_COTS) &&
(nconf->nc_semantics != NC_TPI_COTS_ORD))
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
return (1); /* not my type */
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
/*
* XXX - using RPC library internal functions.
*/
if (!__rpc_nconf2sockinfo(nconf, &si)) {
syslog(LOG_ERR, "cannot get information for %s",
nconf->nc_netid);
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
return (1);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
}
/* Get mountd's address on this transport */
memset(&hints, 0, sizeof hints);
hints.ai_family = si.si_af;
hints.ai_socktype = si.si_socktype;
hints.ai_protocol = si.si_proto;
/*
* Bind to specific IPs if asked to
*/
nhostsbak = nhosts;
while (nhostsbak > 0) {
--nhostsbak;
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
sock_fd = realloc(sock_fd, (sock_fdcnt + 1) * sizeof(int));
if (sock_fd == NULL)
out_of_mem();
sock_fd[sock_fdcnt++] = -1; /* Set invalid for now. */
mallocd_res = 0;
When mountd is creating sockets, it iterates over all addresses specified in the "hosts" array and eventually looks up the network address with getaddrinfo(). At one point it checks for a numeric address and if it sees one, it sets a hint parameter to force getaddrinfo to interpret the host as a numeric address. However that hint is not cleared for subsequent iterations of the loop and if any hosts seen after this point are host names, getaddrinfo will fail on the name. The result of this bug is that you cannot pass a host name to the -h flag. Unfortunately, the first iteration will either process ::1 or 127.0.0.1, so the flag is set on the first iteration and all host names will fail to be processed. The same bug applies to rpc.lockd and rpc.statd, so fix them too. Differential Revision: https://reviews.freebsd.org/D1507 Reported by: Dylan Martin MFC after: 1 week Sponsored by: Sandvine Inc.
2015-01-19 00:33:32 +00:00
hints.ai_flags = AI_PASSIVE;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
/*
* XXX - using RPC library internal functions.
*/
if ((fd = __rpc_nconf2fd(nconf)) < 0) {
int non_fatal = 0;
Fix socket calls on error post-r243965. Submitted by: Garrett Cooper
2012-12-21 15:54:13 +00:00
if (errno == EAFNOSUPPORT &&
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
nconf->nc_semantics != NC_TPI_CLTS)
non_fatal = 1;
syslog(non_fatal ? LOG_DEBUG : LOG_ERR,
"cannot create socket for %s", nconf->nc_netid);
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
if (non_fatal != 0)
continue;
exit(1);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
}
switch (hints.ai_family) {
case AF_INET:
if (inet_pton(AF_INET, hosts[nhostsbak],
host_addr) == 1) {
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
hints.ai_flags |= AI_NUMERICHOST;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
} else {
/*
* Skip if we have an AF_INET6 address.
*/
if (inet_pton(AF_INET6, hosts[nhostsbak],
host_addr) == 1) {
close(fd);
continue;
}
}
break;
case AF_INET6:
if (inet_pton(AF_INET6, hosts[nhostsbak],
host_addr) == 1) {
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
hints.ai_flags |= AI_NUMERICHOST;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
} else {
/*
* Skip if we have an AF_INET address.
*/
if (inet_pton(AF_INET, hosts[nhostsbak],
host_addr) == 1) {
close(fd);
continue;
}
}
/*
* We're doing host-based access checks here, so don't
* allow v4-in-v6 to confuse things. The kernel will
* disable it by default on NFS sockets too.
*/
if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, &one,
sizeof one) < 0) {
syslog(LOG_ERR,
"can't disable v4-in-v6 on IPv6 socket");
exit(1);
}
break;
default:
break;
}
/*
* If no hosts were specified, just bind to INADDR_ANY
*/
if (strcmp("*", hosts[nhostsbak]) == 0) {
if (svcport_str == NULL) {
res = malloc(sizeof(struct addrinfo));
if (res == NULL)
out_of_mem();
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
mallocd_res = 1;
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
res->ai_flags = hints.ai_flags;
res->ai_family = hints.ai_family;
res->ai_protocol = hints.ai_protocol;
switch (res->ai_family) {
case AF_INET:
sin = malloc(sizeof(struct sockaddr_in));
if (sin == NULL)
out_of_mem();
sin->sin_family = AF_INET;
sin->sin_port = htons(0);
sin->sin_addr.s_addr = htonl(INADDR_ANY);
res->ai_addr = (struct sockaddr*) sin;
res->ai_addrlen = (socklen_t)
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
sizeof(struct sockaddr_in);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
break;
case AF_INET6:
sin6 = malloc(sizeof(struct sockaddr_in6));
Use the correct variable to check for a malloc failing. This fixes a case where mountd would fail to start with "out of memory" logged to syslog.
2007-10-27 12:24:47 +00:00
if (sin6 == NULL)
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
out_of_mem();
sin6->sin6_family = AF_INET6;
sin6->sin6_port = htons(0);
sin6->sin6_addr = in6addr_any;
res->ai_addr = (struct sockaddr*) sin6;
res->ai_addrlen = (socklen_t)
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
sizeof(struct sockaddr_in6);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
break;
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
default:
syslog(LOG_ERR, "bad addr fam %d",
res->ai_family);
exit(1);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
}
} else {
if ((aicode = getaddrinfo(NULL, svcport_str,
&hints, &res)) != 0) {
syslog(LOG_ERR,
"cannot get local address for %s: %s",
nconf->nc_netid,
gai_strerror(aicode));
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
close(fd);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
continue;
}
}
} else {
if ((aicode = getaddrinfo(hosts[nhostsbak], svcport_str,
&hints, &res)) != 0) {
syslog(LOG_ERR,
"cannot get local address for %s: %s",
nconf->nc_netid, gai_strerror(aicode));
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
close(fd);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
continue;
}
}
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
/* Store the fd. */
sock_fd[sock_fdcnt - 1] = fd;
/* Now, attempt the bind. */
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
r = bindresvport_sa(fd, res->ai_addr);
if (r != 0) {
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
if (errno == EADDRINUSE && mallocd_svcport != 0) {
if (mallocd_res != 0) {
free(res->ai_addr);
free(res);
} else
freeaddrinfo(res);
return (-1);
}
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
syslog(LOG_ERR, "bindresvport_sa: %m");
exit(1);
}
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
if (svcport_str == NULL) {
svcport_str = malloc(NI_MAXSERV * sizeof(char));
if (svcport_str == NULL)
out_of_mem();
mallocd_svcport = 1;
if (getnameinfo(res->ai_addr,
res->ai_addr->sa_len, NULL, NI_MAXHOST,
svcport_str, NI_MAXSERV * sizeof(char),
NI_NUMERICHOST | NI_NUMERICSERV))
errx(1, "Cannot get port number");
}
if (mallocd_res != 0) {
free(res->ai_addr);
free(res);
} else
freeaddrinfo(res);
res = NULL;
}
return (0);
}
/*
* Called after all the create_service() calls have succeeded, to complete
* the setup and registration.
*/
static void
complete_service(struct netconfig *nconf, char *port_str)
{
struct addrinfo hints, *res = NULL;
struct __rpc_sockinfo si;
struct netbuf servaddr;
SVCXPRT *transp = NULL;
int aicode, fd, nhostsbak;
int registered = 0;
if ((nconf->nc_semantics != NC_TPI_CLTS) &&
(nconf->nc_semantics != NC_TPI_COTS) &&
(nconf->nc_semantics != NC_TPI_COTS_ORD))
return; /* not my type */
/*
* XXX - using RPC library internal functions.
*/
if (!__rpc_nconf2sockinfo(nconf, &si)) {
syslog(LOG_ERR, "cannot get information for %s",
nconf->nc_netid);
return;
}
nhostsbak = nhosts;
while (nhostsbak > 0) {
--nhostsbak;
if (sock_fdpos >= sock_fdcnt) {
/* Should never happen. */
syslog(LOG_ERR, "Ran out of socket fd's");
return;
}
fd = sock_fd[sock_fdpos++];
if (fd < 0)
continue;
mountd has no way to configure the listen queue depth; rather than add a new option, we pass -1 down to listen, which causes it to use the kern.ipc.soacceptqueue sysctl. Approved by: mav MFC after: 2 weeks Sponsored by: iXsystems Inc
2018-11-14 19:06:43 +00:00
/*
* Using -1 tells listen(2) to use
* kern.ipc.soacceptqueue for the backlog.
*/
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
if (nconf->nc_semantics != NC_TPI_CLTS)
mountd has no way to configure the listen queue depth; rather than add a new option, we pass -1 down to listen, which causes it to use the kern.ipc.soacceptqueue sysctl. Approved by: mav MFC after: 2 weeks Sponsored by: iXsystems Inc
2018-11-14 19:06:43 +00:00
listen(fd, -1);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
if (nconf->nc_semantics == NC_TPI_CLTS )
transp = svc_dg_create(fd, 0, 0);
else
transp = svc_vc_create(fd, RPC_MAXDATASIZE,
RPC_MAXDATASIZE);
if (transp != (SVCXPRT *) NULL) {
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (!svc_reg(transp, MOUNTPROG, MOUNTVERS, mntsrv,
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
NULL))
syslog(LOG_ERR,
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
"can't register %s MOUNTVERS service",
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
nconf->nc_netid);
if (!force_v2) {
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (!svc_reg(transp, MOUNTPROG, MOUNTVERS3,
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
mntsrv, NULL))
syslog(LOG_ERR,
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
"can't register %s MOUNTVERS3 service",
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
nconf->nc_netid);
}
} else
syslog(LOG_WARNING, "can't create %s services",
nconf->nc_netid);
if (registered == 0) {
registered = 1;
memset(&hints, 0, sizeof hints);
hints.ai_flags = AI_PASSIVE;
hints.ai_family = si.si_af;
hints.ai_socktype = si.si_socktype;
hints.ai_protocol = si.si_proto;
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
if ((aicode = getaddrinfo(NULL, port_str, &hints,
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
&res)) != 0) {
syslog(LOG_ERR, "cannot get local address: %s",
gai_strerror(aicode));
exit(1);
}
servaddr.buf = malloc(res->ai_addrlen);
memcpy(servaddr.buf, res->ai_addr, res->ai_addrlen);
servaddr.len = res->ai_addrlen;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
rpcb_set(MOUNTPROG, MOUNTVERS, nconf, &servaddr);
rpcb_set(MOUNTPROG, MOUNTVERS3, nconf, &servaddr);
Add the -h <bindip> option to mountd, similar to the one in nfsd(8) -h bindip Specify specific IP addresses to bind to for TCP and UDP requests. This option may be specified multiple times. If no -h option is specified, mountd will bind to INADDR_ANY. Note that when specifying IP addresses with -h, mountd will automatically add 127.0.0.1 and if IPv6 is enabled, ::1 to the list. PR: bin/114097 Reviewed by: pjd (an eariler version of the patch) MFC after: 1 week
2007-10-20 11:25:34 +00:00
xcreated++;
freeaddrinfo(res);
}
} /* end while */
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
}
Fix the nfs related daemons so that they don't intermittently fail with "bind: address already in use". This problem was reported to the freebsd-stable@ mailing list on Feb. 19 under the subject heading "statd/lockd startup failure" by george+freebsd at m5p dot com. The problem is that the first combination of {udp,tcp X ipv4,ipv6} would select a port# dynamically, but one of the other three combinations would have that port# already in use. The patch is somewhat involved because it was requested by dougb@ that the four combinations use the same port# wherever possible. The patch splits the create_service() function into two functions. The first goes as far as bind(2) in a loop for up to GETPORT_MAXTRY - 1 times, attempting to use the same port# for all four cases. If these attempts fail, the last attempt allows the 4 cases to use different port #s. After this function has succeeded, the second function, called complete_service(), does the rest of what create_service() did. The three daemons mountd, rpc.lockd and rpc.statd all have a create_service() function that is patched in a similar way. However, create_service() has non-trivial differences for the three daemons that made it impractical to share the same functions between them. Reviewed by: jhb MFC after: 2 weeks
2011-06-02 19:33:33 +00:00
/*
* Clear out sockets after a failure to bind one of them, so that the
* cycle of socket creation/binding can start anew.
*/
static void
clearout_service(void)
{
int i;
for (i = 0; i < sock_fdcnt; i++) {
if (sock_fd[i] >= 0) {
shutdown(sock_fd[i], SHUT_RDWR);
close(sock_fd[i]);
}
}
}
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
usage(void)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
{
fprintf(stderr,
Change the "-4" argument for nfsd and mountd to "-e" to avoid confusion, since it does not refer to IPv4 nor NFSv4, but to running the experimental server instead of the regular one. Approved by: kib (mentor)
2009-05-28 20:28:13 +00:00
"usage: mountd [-2] [-d] [-e] [-l] [-n] [-p <port>] [-r] "
Add a new '-S' option to mountd, which tells it to suspend execution of the nfsd threads while it is reloading the exports. This avoids clients from getting intermittent access errors when the exports are being reloaded non-atomically. It is not an ideal solution, since requests will back up while the nfsd threads are suspended. Also, when this option is used, if mountd crashes while reloading exports, mountd will have to be restarted to get the nfsd threads to resume execution. This has been tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). The nfse patch offers a more comprehensive solution for this issue. PR: kern/9619, kern/131342 Reviewed by: kib MFC after: 2 weeks
2012-10-15 00:17:16 +00:00
"[-S] [-h <bindip>] [export_file ...]\n");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
exit(1);
}
/*
* The mount rpc service
*/
void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
mntsrv(struct svc_req *rqstp, SVCXPRT *transp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct exportlist *ep;
struct dirlist *dp;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
struct fhreturn fhr;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct stat stb;
struct statfs fsb;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
char host[NI_MAXHOST], numerichost[NI_MAXHOST];
int lookup_failed = 1;
struct sockaddr *saddr;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
u_short sport;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
char rpcpath[MNTPATHLEN + 1], dirpath[MAXPATHLEN];
Correct a type-mismatch between xdr_long and the variable "bad". Way back in r28911 (August 1997, CVS rev 1.22) we imported a NetBSD information leak fix via OpenBSD. Unfortunatly we failed to track the followup commit that fixed the type of the error code. Apply the change from int to long now. Reviewed by: emaste Found by: CHERI Obtained from: CheriBSD MFC after: 3 days Sponsored by: DARPA Differential Revision: https://reviews.freebsd.org/D25779
2020-07-22 23:39:58 +00:00
int defset, hostset;
long bad = 0;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigset_t sighup_mask;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
int numsecflavors, *secflavorsp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigemptyset(&sighup_mask);
sigaddset(&sighup_mask, SIGHUP);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
saddr = svc_getrpccaller(transp)->buf;
switch (saddr->sa_family) {
case AF_INET6:
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
sport = ntohs(((struct sockaddr_in6 *)saddr)->sin6_port);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
break;
case AF_INET:
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
sport = ntohs(((struct sockaddr_in *)saddr)->sin_port);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
break;
default:
syslog(LOG_ERR, "request from unknown address family");
return;
}
Reduce number of DNS queries in mountd. As reported by a FreeNAS user (see https://redmine.ixsystems.com/issues/55728), mountd does more calls to getnameinfo() than it needs to; this changes it to only call it for the RPC calls it needs the name information for. Reported by: Dave Flowers Reviewed by: imp, mav Approved by: mav (mentor) MFC after: 2 weeks Sponsored by: iXsystems Inc Differential Revision: https://reviews.freebsd.org/D18430
2018-12-06 18:21:48 +00:00
switch (rqstp->rq_proc) {
case MOUNTPROC_MNT:
case MOUNTPROC_UMNT:
case MOUNTPROC_UMNTALL:
lookup_failed = getnameinfo(saddr, saddr->sa_len, host,
sizeof host, NULL, 0, 0);
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
getnameinfo(saddr, saddr->sa_len, numerichost,
sizeof numerichost, NULL, 0, NI_NUMERICHOST);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
switch (rqstp->rq_proc) {
case NULLPROC:
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_void, NULL))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
case MOUNTPROC_MNT:
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
if (sport >= IPPORT_RESERVED && resvport_only) {
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"mount request from %s from unprivileged port",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
svcerr_weakauth(transp);
return;
}
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_getargs(transp, (xdrproc_t)xdr_dir, rpcpath)) {
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE, "undecodable mount request from %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
svcerr_decode(transp);
return;
}
/*
* Get the real pathname and make sure it is a directory
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
* or a regular file if the -r option was specified
* and it exists.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Put the umount request through the same translation as the mount request. Fix some uninitialized warnings and a style bug while we are here. Problem still exists where a failed umount can be misreported as success. Submitted by: Martin Blapp <mb@imp.ch>
1999-10-06 18:20:44 +00:00
if (realpath(rpcpath, dirpath) == NULL ||
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
stat(dirpath, &stb) < 0 ||
statfs(dirpath, &fsb) < 0) {
chdir("/"); /* Just in case realpath doesn't */
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
"mount request from %s for non existent path %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost, dirpath);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("stat failed on %s", dirpath);
Plug info agthering attack as pointed out on Bugtraq recently. Obtained from: OpenBSD
1997-08-29 19:22:28 +00:00
bad = ENOENT; /* We will send error reply later */
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
mountd: Return proper errno values in a few error paths When attempting to mount a non-directory which exists, return ENOTDIR instead of ENOENT. If stat() or statfs() failed, don't pass part of the invalid (struct statfs) to ex_search(). In that same case, preserve the value of "bad" rather than overwriting with EACCES. Submitted by: Bruce Leverett (Panasas) Reviewed by: rmacklem MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D14438
2018-02-21 00:19:02 +00:00
if (!bad &&
!S_ISDIR(stb.st_mode) &&
(dir_only || !S_ISREG(stb.st_mode))) {
syslog(LOG_NOTICE,
"mount request from %s for non-directory path %s",
numerichost, dirpath);
if (debug)
warnx("mounting non-directory %s", dirpath);
bad = ENOTDIR; /* We will send error reply later */
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/* Check in the exports list */
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigprocmask(SIG_BLOCK, &sighup_mask, NULL);
mountd: Return proper errno values in a few error paths When attempting to mount a non-directory which exists, return ENOTDIR instead of ENOENT. If stat() or statfs() failed, don't pass part of the invalid (struct statfs) to ex_search(). In that same case, preserve the value of "bad" rather than overwriting with EACCES. Submitted by: Bruce Leverett (Panasas) Reviewed by: rmacklem MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D14438
2018-02-21 00:19:02 +00:00
if (bad)
ep = NULL;
else
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
ep = ex_search(&fsb.f_fsid, exphead);
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
hostset = defset = 0;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
if (ep && (chk_host(ep->ex_defdir, saddr, &defset, &hostset,
&numsecflavors, &secflavorsp) ||
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
((dp = dirp_search(ep->ex_dirl, dirpath)) &&
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
chk_host(dp, saddr, &defset, &hostset, &numsecflavors,
&secflavorsp)) ||
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
(defset && scan_tree(ep->ex_defdir, saddr) == 0 &&
scan_tree(ep->ex_dirl, saddr) == 0))) {
Plug info agthering attack as pointed out on Bugtraq recently. Obtained from: OpenBSD
1997-08-29 19:22:28 +00:00
if (bad) {
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_long,
Plug info agthering attack as pointed out on Bugtraq recently. Obtained from: OpenBSD
1997-08-29 19:22:28 +00:00
(caddr_t)&bad))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Plug info agthering attack as pointed out on Bugtraq recently. Obtained from: OpenBSD
1997-08-29 19:22:28 +00:00
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
return;
}
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
if (hostset & DP_HOSTSET) {
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
fhr.fhr_flag = hostset;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
fhr.fhr_numsecflavors = numsecflavors;
fhr.fhr_secflavors = secflavorsp;
} else {
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
fhr.fhr_flag = defset;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
fhr.fhr_numsecflavors = ep->ex_defnumsecflavors;
fhr.fhr_secflavors = ep->ex_defsecflavors;
}
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
fhr.fhr_vers = rqstp->rq_vers;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/* Get the file handle */
Merge from Lite2 (use new getvfsbyname() and mount(2) interface, cleanup)
1997-03-11 12:43:45 +00:00
memset(&fhr.fhr_fh, 0, sizeof(nfsfh_t));
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
if (getfh(dirpath, (fhandle_t *)&fhr.fhr_fh) < 0) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
bad = errno;
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't get fh for %s", dirpath);
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_long,
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
(caddr_t)&bad))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
}
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_fhs,
(caddr_t)&fhr))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (!lookup_failed)
add_mlist(host, dirpath);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
else
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
add_mlist(numerichost, dirpath);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("mount successful");
Rename the 'log' variable rather than -fno-builtin-log
2003-10-30 22:57:43 +00:00
if (dolog)
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"mount request succeeded from %s for %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost, dirpath);
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
} else {
mountd: Return proper errno values in a few error paths When attempting to mount a non-directory which exists, return ENOTDIR instead of ENOENT. If stat() or statfs() failed, don't pass part of the invalid (struct statfs) to ex_search(). In that same case, preserve the value of "bad" rather than overwriting with EACCES. Submitted by: Bruce Leverett (Panasas) Reviewed by: rmacklem MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D14438
2018-02-21 00:19:02 +00:00
if (!bad)
bad = EACCES;
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"mount request denied from %s for %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost, dirpath);
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
}
Plug info agthering attack as pointed out on Bugtraq recently. Obtained from: OpenBSD
1997-08-29 19:22:28 +00:00
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (bad && !svc_sendreply(transp, (xdrproc_t)xdr_long,
(caddr_t)&bad))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
case MOUNTPROC_DUMP:
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_mlist, (caddr_t)NULL))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Rename the 'log' variable rather than -fno-builtin-log
2003-10-30 22:57:43 +00:00
else if (dolog)
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"dump request succeeded from %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
case MOUNTPROC_UMNT:
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
if (sport >= IPPORT_RESERVED && resvport_only) {
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"umount request from %s from unprivileged port",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
svcerr_weakauth(transp);
return;
}
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_getargs(transp, (xdrproc_t)xdr_dir, rpcpath)) {
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE, "undecodable umount request from %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
svcerr_decode(transp);
return;
}
Put the umount request through the same translation as the mount request. Fix some uninitialized warnings and a style bug while we are here. Problem still exists where a failed umount can be misreported as success. Submitted by: Martin Blapp <mb@imp.ch>
1999-10-06 18:20:44 +00:00
if (realpath(rpcpath, dirpath) == NULL) {
syslog(LOG_NOTICE, "umount request from %s "
"for non existent path %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost, dirpath);
Put the umount request through the same translation as the mount request. Fix some uninitialized warnings and a style bug while we are here. Problem still exists where a failed umount can be misreported as success. Submitted by: Martin Blapp <mb@imp.ch>
1999-10-06 18:20:44 +00:00
}
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_void, (caddr_t)NULL))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (!lookup_failed)
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
del_mlist(host, dirpath);
del_mlist(numerichost, dirpath);
Rename the 'log' variable rather than -fno-builtin-log
2003-10-30 22:57:43 +00:00
if (dolog)
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"umount request succeeded from %s for %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost, dirpath);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
case MOUNTPROC_UMNTALL:
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
if (sport >= IPPORT_RESERVED && resvport_only) {
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"umountall request from %s from unprivileged port",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
svcerr_weakauth(transp);
return;
}
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_void, (caddr_t)NULL))
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't send reply");
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (!lookup_failed)
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
del_mlist(host, NULL);
del_mlist(numerichost, NULL);
Rename the 'log' variable rather than -fno-builtin-log
2003-10-30 22:57:43 +00:00
if (dolog)
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"umountall request succeeded from %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
case MOUNTPROC_EXPORT:
Use -fno-builtin-log. Fix svc_sendreply() missing required casts. It is now WARNS=2 clean again.
2003-10-26 05:58:21 +00:00
if (!svc_sendreply(transp, (xdrproc_t)xdr_explist, (caddr_t)NULL))
if (!svc_sendreply(transp, (xdrproc_t)xdr_explist_brief,
(caddr_t)NULL))
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
syslog(LOG_ERR, "can't send reply");
Rename the 'log' variable rather than -fno-builtin-log
2003-10-30 22:57:43 +00:00
if (dolog)
Log all failed mount attempts. Also add a flag (-l) so mountd will also log all succeeded requests to mountd.
1997-12-09 18:43:44 +00:00
syslog(LOG_NOTICE,
"export request succeeded from %s",
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
numerichost);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
default:
svcerr_noproc(transp);
return;
}
}
/*
* Xdr conversion for a dirpath string
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
xdr_dir(XDR *xdrsp, char *dirp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
return (xdr_string(xdrsp, &dirp, MNTPATHLEN));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
/*
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
* Xdr routine to generate file handle reply
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
xdr_fhs(XDR *xdrsp, caddr_t cp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Remove 'register' keyword. It does not help modern compilers, and some may take some hit from it. (I also found several functions that listed *every* of its 10 local vars with "register" -- just how many free registers do people think machines have?)
2002-03-20 17:55:10 +00:00
struct fhreturn *fhrp = (struct fhreturn *)cp;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
u_long ok = 0, len, auth;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int i;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (!xdr_long(xdrsp, &ok))
return (0);
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
switch (fhrp->fhr_vers) {
case 1:
return (xdr_opaque(xdrsp, (caddr_t)&fhrp->fhr_fh, NFSX_V2FH));
case 3:
len = NFSX_V3FH;
if (!xdr_long(xdrsp, &len))
return (0);
if (!xdr_opaque(xdrsp, (caddr_t)&fhrp->fhr_fh, len))
return (0);
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
if (fhrp->fhr_numsecflavors) {
if (!xdr_int(xdrsp, &fhrp->fhr_numsecflavors))
return (0);
for (i = 0; i < fhrp->fhr_numsecflavors; i++)
if (!xdr_int(xdrsp, &fhrp->fhr_secflavors[i]))
return (0);
return (1);
} else {
auth = AUTH_SYS;
len = 1;
if (!xdr_long(xdrsp, &len))
return (0);
return (xdr_long(xdrsp, &auth));
}
Cleanup unnecessary semicolons from utilities we all love.
2016-04-15 22:31:22 +00:00
}
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
return (0);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
xdr_mlist(XDR *xdrsp, caddr_t cp __unused)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct mountlist *mlp;
int true = 1;
int false = 0;
char *strp;
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_FOREACH(mlp, &mlhead, next) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (!xdr_bool(xdrsp, &true))
return (0);
strp = &mlp->ml_host[0];
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (!xdr_string(xdrsp, &strp, MNTNAMLEN))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (0);
strp = &mlp->ml_dirp[0];
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (!xdr_string(xdrsp, &strp, MNTPATHLEN))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (0);
}
if (!xdr_bool(xdrsp, &false))
return (0);
return (1);
}
/*
* Xdr conversion for export list
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
xdr_explist_common(XDR *xdrsp, caddr_t cp __unused, int brief)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct exportlist *ep;
int false = 0;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
int putdef;
sigset_t sighup_mask;
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
int i;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigemptyset(&sighup_mask);
sigaddset(&sighup_mask, SIGHUP);
sigprocmask(SIG_BLOCK, &sighup_mask, NULL);
mountd: Convert exportlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for the exportlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12502
2017-10-04 08:43:56 +00:00
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
for (i = 0; i < exphashsize; i++)
SLIST_FOREACH(ep, &exphead[i], entries) {
putdef = 0;
if (put_exlist(ep->ex_dirl, xdrsp, ep->ex_defdir,
&putdef, brief))
goto errout;
if (ep->ex_defdir && putdef == 0 &&
put_exlist(ep->ex_defdir, xdrsp, NULL,
&putdef, brief))
goto errout;
}
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (!xdr_bool(xdrsp, &false))
return (0);
return (1);
errout:
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (0);
}
/*
* Called from xdr_explist() to traverse the tree and export the
* directory paths.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
put_exlist(struct dirlist *dp, XDR *xdrsp, struct dirlist *adp, int *putdefp,
int brief)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct grouplist *grp;
struct hostlist *hp;
int true = 1;
int false = 0;
int gotalldir = 0;
char *strp;
if (dp) {
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
if (put_exlist(dp->dp_left, xdrsp, adp, putdefp, brief))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
if (!xdr_bool(xdrsp, &true))
return (1);
strp = dp->dp_dirp;
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (!xdr_string(xdrsp, &strp, MNTPATHLEN))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
if (adp && !strcmp(dp->dp_dirp, adp->dp_dirp)) {
gotalldir = 1;
*putdefp = 1;
}
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
if (brief) {
if (!xdr_bool(xdrsp, &true))
return (1);
strp = "(...)";
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (!xdr_string(xdrsp, &strp, MNTPATHLEN))
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
return (1);
} else if ((dp->dp_flag & DP_DEFSET) == 0 &&
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
(gotalldir == 0 || (adp->dp_flag & DP_DEFSET) == 0)) {
hp = dp->dp_hosts;
while (hp) {
grp = hp->ht_grp;
if (grp->gr_type == GT_HOST) {
if (!xdr_bool(xdrsp, &true))
return (1);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
strp = grp->gr_ptr.gt_addrinfo->ai_canonname;
Remove trailing whitespace.
1995-05-30 06:12:45 +00:00
if (!xdr_string(xdrsp, &strp,
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
MNTNAMLEN))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
} else if (grp->gr_type == GT_NET) {
if (!xdr_bool(xdrsp, &true))
return (1);
strp = grp->gr_ptr.gt_net.nt_name;
Remove trailing whitespace.
1995-05-30 06:12:45 +00:00
if (!xdr_string(xdrsp, &strp,
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
MNTNAMLEN))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
hp = hp->ht_next;
if (gotalldir && hp == (struct hostlist *)NULL) {
hp = adp->dp_hosts;
gotalldir = 0;
}
}
}
if (!xdr_bool(xdrsp, &false))
return (1);
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
if (put_exlist(dp->dp_right, xdrsp, adp, putdefp, brief))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
return (0);
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
xdr_explist(XDR *xdrsp, caddr_t cp)
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
{
return xdr_explist_common(xdrsp, cp, 0);
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
xdr_explist_brief(XDR *xdrsp, caddr_t cp)
If an RPC fails, try a shorter reply in case the fault lies with the size of the response. PR: misc/26320 MFC After: 2 weeks
2002-07-15 18:34:27 +00:00
{
return xdr_explist_common(xdrsp, cp, 1);
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static char *line;
static size_t linesize;
static FILE *exp_file;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
* Get the export list from one, currently open file
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
static void
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
get_exportlist_one(int passno)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
mountd: Convert exportlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for the exportlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12502
2017-10-04 08:43:56 +00:00
struct exportlist *ep;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
struct grouplist *grp, *tgrp, *savgrp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
struct dirlist *dirhead;
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
struct statfs fsb;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
struct expcred anon;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
char *cp, *endcp, *dirp, *hst, *usr, *dom, savedc;
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
char *err_msg = NULL;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
int len, has_host, got_nondir, dirplen, netgrp;
uint64_t exflags;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
v4root_phase = 0;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
anon.cr_groups = NULL;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
dirhead = (struct dirlist *)NULL;
while (get_line()) {
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("got line %s", line);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
cp = line;
nextfield(&cp, &endcp);
if (*cp == '#')
goto nextline;
/*
* Set defaults.
*/
has_host = FALSE;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
anon.cr_groups = anon.cr_smallgrps;
Make use of the UID_NOBODY and GID_NOGROUP definitions in sys/conf.h. r362214 exposed UID_NOBODY and GID_NOGROUP to userspace, so use them instead of the numbers. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D25281
2020-06-16 02:35:30 +00:00
anon.cr_uid = UID_NOBODY;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
anon.cr_ngroups = 1;
Make use of the UID_NOBODY and GID_NOGROUP definitions in sys/conf.h. r362214 exposed UID_NOBODY and GID_NOGROUP to userspace, so use them instead of the numbers. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D25281
2020-06-16 02:35:30 +00:00
anon.cr_groups[0] = GID_NOGROUP;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
exflags = MNT_EXPORTED;
got_nondir = 0;
opt_flags = 0;
ep = (struct exportlist *)NULL;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
dirp = NULL;
/*
* Handle the V4 root dir.
*/
if (*cp == 'V' && *(cp + 1) == '4' && *(cp + 2) == ':') {
/*
* V4: just indicates that it is the v4 root point,
* so skip over that and set v4root_phase.
*/
if (v4root_phase > 0) {
syslog(LOG_ERR, "V4:duplicate line, ignored");
goto nextline;
}
v4root_phase = 1;
cp += 3;
nextfield(&cp, &endcp);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Create new exports list entry
*/
len = endcp-cp;
tgrp = grp = get_grp();
while (len > 0) {
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
if (len > MNTNAMLEN) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "mountpoint too long");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
if (*cp == '-') {
if (ep == (struct exportlist *)NULL) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp,
"flag before export path definition");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("doing opt %s", cp);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
got_nondir = 1;
if (do_opt(&cp, &endcp, ep, grp, &has_host,
&exflags, &anon)) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
} else if (*cp == '/') {
savedc = *endcp;
*endcp = '\0';
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase > 1) {
if (dirp != NULL) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "Multiple V4 dirs");
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
goto nextline;
}
}
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
if (check_dirpath(cp, &err_msg) &&
check_statfs(cp, &fsb, &err_msg)) {
Add a warning message to mountd for exported file systems that are automounted, since that configuration isn't supported. This still allows the export, since two emails I received felt that this should not be disabled. It sends the message to syslog(LOG_ERR..), so that it goes to the same places as the other messages related to /etc/exports problems, even though it is a warning and not an error. Reviewed by: trasz MFC after: 2 weeks
2015-05-16 12:05:26 +00:00
if ((fsb.f_flags & MNT_AUTOMOUNTED) != 0)
syslog(LOG_ERR, "Warning: exporting of "
"automounted fs %s not supported", cp);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (got_nondir) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "dirs must be first");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase == 1) {
if (dirp != NULL) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "Multiple V4 dirs");
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
goto nextline;
}
if (strlen(v4root_dirpath) == 0) {
strlcpy(v4root_dirpath, cp,
sizeof (v4root_dirpath));
} else if (strcmp(v4root_dirpath, cp)
!= 0) {
syslog(LOG_ERR,
"different V4 dirpath %s", cp);
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
dirp = cp;
v4root_phase = 2;
got_nondir = 1;
ep = get_exp();
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
} else {
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (ep) {
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
if (fsidcmp(&ep->ex_fs, &fsb.f_fsid)
!= 0) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp,
"fsid mismatch");
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
goto nextline;
}
} else {
/*
* See if this directory is already
* in the list.
*/
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
ep = ex_search(&fsb.f_fsid, exphead);
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (ep == (struct exportlist *)NULL) {
ep = get_exp();
ep->ex_fs = fsb.f_fsid;
mountd: Replace malloc+strcpy to strdup Reviewed by: bapt MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12503
2017-09-26 09:18:18 +00:00
ep->ex_fsdir = strdup(fsb.f_mntonname);
if (ep->ex_fsdir == NULL)
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
out_of_mem();
if (debug)
warnx(
"making new ep fs=0x%x,0x%x",
fsb.f_fsid.val[0],
fsb.f_fsid.val[1]);
} else if (debug)
warnx("found ep fs=0x%x,0x%x",
fsb.f_fsid.val[0],
fsb.f_fsid.val[1]);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
* Add dirpath to export mount point.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
dirp = add_expdir(&dirhead, cp, len);
dirplen = len;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
} else {
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
if (err_msg != NULL) {
getexp_err(ep, tgrp, err_msg);
free(err_msg);
err_msg = NULL;
} else {
getexp_err(ep, tgrp,
"symbolic link in export path or "
"statfs failed");
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
*endcp = savedc;
} else {
savedc = *endcp;
*endcp = '\0';
got_nondir = 1;
if (ep == (struct exportlist *)NULL) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp,
"host(s) before export path definition");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
/*
* Get the host or netgroup.
*/
setnetgrent(cp);
netgrp = getnetgrent(&hst, &usr, &dom);
do {
if (has_host) {
grp->gr_next = get_grp();
grp = grp->gr_next;
}
if (netgrp) {
Watch out for null hostnames in netgroup entries, to avoid dumping core. This happens if someone tries to export to a netgroup like: mygroup (,,mynisdomain)
1998-06-15 15:41:41 +00:00
if (hst == 0) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR,
"null hostname in netgroup %s, skipping", cp);
Oops, the previous commit missed one line of code.
1998-06-15 15:43:13 +00:00
grp->gr_type = GT_IGNORE;
Watch out for null hostnames in netgroup entries, to avoid dumping core. This happens if someone tries to export to a netgroup like: mygroup (,,mynisdomain)
1998-06-15 15:41:41 +00:00
} else if (get_host(hst, grp, tgrp)) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR,
"bad host %s in netgroup %s, skipping", hst, cp);
If a host in an export line in /etc/exports (or within a netgroup on an export line) is unresolvable, make a note of it via syslog and skip that individual host instead of skipping the entire line. PR: 1981, 815 Perused by: joerg
1997-09-12 16:25:24 +00:00
grp->gr_type = GT_IGNORE;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Fixed stupid bug in mountd: it would seem that the kernel doesn't allow you to push the same host into its NFS export lists twice, but mountd tries to do it anyway. This means that putting: /some_file_system -ro host1 host1 in your /etc/exports file causes an error. This is bogus: mountd should be smart enough to ignore the second instance of host1. This can be a problem in some configurations that use netgroups. For example, each host in my netgroups database is has two entries: startide (startide,-,) (startide.ctr.columbia.edu,-,) When mountd sees this, it tries to put startide.ctr.columbia.edu into the export list *twice*. Just listing 'startide' /etc/exports list will also screw up because mountd will try to resolve the netgroup 'startide' instead of the hostname 'startide.' My solution is watch for duplicate entries in get_host() and mark them as grouptype GT_IGNORE, which do_mount() will now cheefully throw away. This is a bit of a kludge, but it was the least obtrusive fix I could come up with. Also silenced a compiler warning: arguments passwd to xdr_long() should be u_long, not int. :)
1995-03-26 23:36:38 +00:00
} else if (get_host(cp, grp, tgrp)) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "bad host %s, skipping", cp);
If a host in an export line in /etc/exports (or within a netgroup on an export line) is unresolvable, make a note of it via syslog and skip that individual host instead of skipping the entire line. PR: 1981, 815 Perused by: joerg
1997-09-12 16:25:24 +00:00
grp->gr_type = GT_IGNORE;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
has_host = TRUE;
} while (netgrp && getnetgrent(&hst, &usr, &dom));
endnetgrent();
*endcp = savedc;
}
cp = endcp;
nextfield(&cp, &endcp);
len = endcp - cp;
}
if (check_options(dirhead)) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, NULL);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
}
if (!has_host) {
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
grp->gr_type = GT_DEFAULT;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (debug)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
warnx("adding a default entry");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Don't allow a network export coincide with a list of
* host(s) on the same line.
*/
} else if ((opt_flags & OP_NET) && tgrp->gr_next) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "network/host conflict");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
goto nextline;
If a host in an export line in /etc/exports (or within a netgroup on an export line) is unresolvable, make a note of it via syslog and skip that individual host instead of skipping the entire line. PR: 1981, 815 Perused by: joerg
1997-09-12 16:25:24 +00:00
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
/*
* If an export list was specified on this line, make sure
If a host in an export line in /etc/exports (or within a netgroup on an export line) is unresolvable, make a note of it via syslog and skip that individual host instead of skipping the entire line. PR: 1981, 815 Perused by: joerg
1997-09-12 16:25:24 +00:00
* that we have at least one valid entry, otherwise skip it.
*/
} else {
grp = tgrp;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
while (grp && grp->gr_type == GT_IGNORE)
If a host in an export line in /etc/exports (or within a netgroup on an export line) is unresolvable, make a note of it via syslog and skip that individual host instead of skipping the entire line. PR: 1981, 815 Perused by: joerg
1997-09-12 16:25:24 +00:00
grp = grp->gr_next;
if (! grp) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "no valid entries");
If a host in an export line in /etc/exports (or within a netgroup on an export line) is unresolvable, make a note of it via syslog and skip that individual host instead of skipping the entire line. PR: 1981, 815 Perused by: joerg
1997-09-12 16:25:24 +00:00
goto nextline;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase == 1) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, "V4:root, no dirp, ignored");
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
goto nextline;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Loop through hosts, pushing the exports into the kernel.
* After loop, tgrp points to the start of the list and
* grp points to the last entry in the list.
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
* Do not do the do_mount() for passno == 1, since the
* second pass will do it, as required.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
grp = tgrp;
do {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
grp->gr_exflags = exflags;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
cp_cred(&grp->gr_anon, &anon);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
if (v4root_phase == 2 && passno == 0)
LOGDEBUG("do_mount v4root");
if (passno == 0 && do_mount(ep, grp, exflags, &anon,
dirp, dirplen, &fsb, ep->ex_numsecflavors,
ep->ex_secflavors)) {
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(ep, tgrp, NULL);
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
goto nextline;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
} while (grp->gr_next && (grp = grp->gr_next));
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
/*
* For V4: don't enter in mount lists.
*/
When mountd.c parses the nfsv4 root line(s) in /etc/exports, it allocates data structures that are never linked into the tree or free'd. As such, mountd would leak memory every time it parsed an nfsv4 root line. This patch frees up those structures to plug the leak. Approved by: kib (mentor)
2009-06-23 21:48:04 +00:00
if (v4root_phase > 0 && v4root_phase <= 2) {
/*
r348590 had mention of "-I" in a comment that no longer applied to the patch. Take "-I" out of the comment line, since the patch no longer uses the "-I" option. MFC after: 1 month
2019-06-03 23:07:46 +00:00
* These structures are used for the reload,
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
* so save them for that case. Otherwise, just
When mountd.c parses the nfsv4 root line(s) in /etc/exports, it allocates data structures that are never linked into the tree or free'd. As such, mountd would leak memory every time it parsed an nfsv4 root line. This patch frees up those structures to plug the leak. Approved by: kib (mentor)
2009-06-23 21:48:04 +00:00
* free them up now.
*/
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
if (passno == 1 && ep != NULL) {
savgrp = tgrp;
while (tgrp != NULL) {
/*
* Save the security flavors and exflags
* for this host set in the groups.
*/
tgrp->gr_numsecflavors =
ep->ex_numsecflavors;
if (ep->ex_numsecflavors > 0)
memcpy(tgrp->gr_secflavors,
ep->ex_secflavors,
sizeof(ep->ex_secflavors));
tgrp = tgrp->gr_next;
}
if (v4root_ep == NULL) {
v4root_ep = ep;
ep = NULL; /* Don't free below. */
}
grp->gr_next = v4root_ep->ex_grphead;
v4root_ep->ex_grphead = savgrp;
}
When mountd.c parses the nfsv4 root line(s) in /etc/exports, it allocates data structures that are never linked into the tree or free'd. As such, mountd would leak memory every time it parsed an nfsv4 root line. This patch frees up those structures to plug the leak. Approved by: kib (mentor)
2009-06-23 21:48:04 +00:00
if (ep != NULL)
free_exp(ep);
while (tgrp != NULL) {
grp = tgrp;
tgrp = tgrp->gr_next;
free_grp(grp);
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
goto nextline;
When mountd.c parses the nfsv4 root line(s) in /etc/exports, it allocates data structures that are never linked into the tree or free'd. As such, mountd would leak memory every time it parsed an nfsv4 root line. This patch frees up those structures to plug the leak. Approved by: kib (mentor)
2009-06-23 21:48:04 +00:00
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Success. Update the data structures.
*/
if (has_host) {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
hang_dirp(dirhead, tgrp, ep, opt_flags, &anon, exflags);
Replace global list for grouplist with list(s) for each exportlist element. In mountd.c, the grouplist structures are linked into a single global linked list headed by "grphead". The only use of this linked list is to free all list elements when the exportlist elements are also all being free'd at the time the exports are being reloaded. This patch replaces this one global linked list head with a list head in each exportlist structure, where the grouplist elements for that exported file system are linked. The only change is that now the grouplist elements are free'd with the associated exportlist element as they are free'd instead of all grouplist elements being free'd after the exportlist elements are free'd. This change should have no effect in practice. This is being done, since a future patch that will add a "-I" option for incrementally updating the exports in the kernel needs to know which grouplist elements are associated with each exported file system and having them linked into a list headed by the exportlist element does that. MFC after: 1 month
2019-05-14 22:00:47 +00:00
grp->gr_next = ep->ex_grphead;
ep->ex_grphead = tgrp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
} else {
hang_dirp(dirhead, (struct grouplist *)NULL, ep,
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
opt_flags, &anon, exflags);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
free_grp(grp);
}
dirhead = (struct dirlist *)NULL;
if ((ep->ex_flag & EX_LINKED) == 0) {
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
insert_exports(ep, exphead);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
ep->ex_flag |= EX_LINKED;
}
nextline:
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
v4root_phase = 0;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (dirhead) {
free_dir(dirhead);
dirhead = (struct dirlist *)NULL;
}
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (anon.cr_groups != anon.cr_smallgrps) {
free(anon.cr_groups);
anon.cr_groups = NULL;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
}
/*
* Get the export list from all specified files
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
get_exportlist(int passno)
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
{
struct export_args export;
struct iovec *iov;
Factor code into two new functions in preparation for a future commit. Factor code into two functions. read_exportfile() a functon which reads the exports file(s) and calls get_exportlist_one() to process each of them. delete_export() a function which deletes the exports in the kernel for a file system. The contents of these functions is just the same code as was used to do the operations, moved into separate functions. As such, there is no semantic change. This is being done in preparation for a future commit that will add an option to do incremental changes of kernel exports upon receiving SIGHUP. MFC after: 1 month
2019-05-11 22:41:58 +00:00
struct statfs *mntbufp;
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
char errmsg[255];
mountd(8): generate a syslog message when the "V4:" line is missing Daniel reported that NFSv4 mounts were not working despite having set "nfsv4_server_enable=YES" in /etc/rc.conf. Mountd was logging a message that there was no /etc/exports file. He noted that creating a /etc/exports file with a "V4:" line in it was needed make NFSv4 mounts work. At least one "V4:" line in one of the exports(5) file(s) is needed to make NFSv4 mounts work. This patch fixes mountd.c so that it logs a message indicting that there is no "V4:" line in any exports(5) file when NFSv4 mounts are enabled. To avoid this message being generated erroneously, /etc/rc.d/mountd is updated to make sure vfs.nfsd.server_max_nfsvers is properly set before mountd(8) is started. Reported by: debdrup PR: 253901 MFC after: 2 weeks
2021-03-09 00:08:02 +00:00
int error, i, nfs_maxvers, num;
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
int iovlen;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
struct nfsex_args eargs;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
FILE *debug_file;
mountd(8): generate a syslog message when the "V4:" line is missing Daniel reported that NFSv4 mounts were not working despite having set "nfsv4_server_enable=YES" in /etc/rc.conf. Mountd was logging a message that there was no /etc/exports file. He noted that creating a /etc/exports file with a "V4:" line in it was needed make NFSv4 mounts work. At least one "V4:" line in one of the exports(5) file(s) is needed to make NFSv4 mounts work. This patch fixes mountd.c so that it logs a message indicting that there is no "V4:" line in any exports(5) file when NFSv4 mounts are enabled. To avoid this message being generated erroneously, /etc/rc.d/mountd is updated to make sure vfs.nfsd.server_max_nfsvers is properly set before mountd(8) is started. Reported by: debdrup PR: 253901 MFC after: 2 weeks
2021-03-09 00:08:02 +00:00
size_t nfs_maxvers_size;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
if ((debug_file = fopen(_PATH_MOUNTDDEBUG, "r")) != NULL) {
fclose(debug_file);
logdebug = 1;
} else
logdebug = 0;
LOGDEBUG("passno=%d", passno);
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
v4root_dirpath[0] = '\0';
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
free_v4rootexp();
if (passno == 1) {
/*
* Save the current lists as old ones, so that the new lists
* can be compared with the old ones in the 2nd pass.
*/
for (i = 0; i < exphashsize; i++) {
SLIST_FIRST(&oldexphead[i]) = SLIST_FIRST(&exphead[i]);
SLIST_INIT(&exphead[i]);
}
/* Note that the public fh has not yet been set. */
has_set_publicfh = 0;
/* Read the export file(s) and process them */
read_exportfile(passno);
} else {
/*
* Just make the old lists empty.
* exphashsize == 0 for the first call, before oldexphead
* has been initialized-->loop won't be executed.
*/
for (i = 0; i < exphashsize; i++)
SLIST_INIT(&oldexphead[i]);
}
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
bzero(&export, sizeof(export));
export.ex_flags = MNT_DELEXPORT;
iov = NULL;
iovlen = 0;
bzero(errmsg, sizeof(errmsg));
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
if (suspend_nfsd != 0)
(void)nfssvc(NFSSVC_SUSPENDNFSD, NULL);
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
/*
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
* Delete the old V4 root dir.
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
*/
bzero(&eargs, sizeof (eargs));
eargs.export.ex_flags = MNT_DELEXPORT;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
if (nfssvc(NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT, (caddr_t)&eargs) < 0 &&
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
errno != ENOENT)
syslog(LOG_ERR, "Can't delete exports for V4:");
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
build_iovec(&iov, &iovlen, "fstype", NULL, 0);
build_iovec(&iov, &iovlen, "fspath", NULL, 0);
build_iovec(&iov, &iovlen, "from", NULL, 0);
build_iovec(&iov, &iovlen, "update", NULL, 0);
build_iovec(&iov, &iovlen, "export", &export,
sizeof(export));
build_iovec(&iov, &iovlen, "errmsg", errmsg,
sizeof(errmsg));
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
/*
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
* For passno == 1, compare the old and new lists updating the kernel
* exports for any cases that have changed.
* This call is doing the second pass through the lists.
* If it fails, fall back on the bulk reload.
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
*/
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
if (passno == 1 && compare_nmount_exportlist(iov, iovlen, errmsg) ==
0) {
LOGDEBUG("compareok");
/* Free up the old lists. */
free_exports(oldexphead);
} else {
LOGDEBUG("doing passno=0");
/*
* Clear flag that notes if a public fh has been exported.
* It is set by do_mount() if MNT_EXPUBLIC is set for the entry.
*/
has_publicfh = 0;
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/* exphead == NULL if not yet allocated (first call). */
if (exphead != NULL) {
/*
* First, get rid of the old lists.
*/
free_exports(exphead);
free_exports(oldexphead);
}
/*
* And delete exports that are in the kernel for all local
* filesystems.
* XXX: Should know how to handle all local exportable
* filesystems.
*/
num = getmntinfo(&mntbufp, MNT_NOWAIT);
/* Allocate hash tables, for first call. */
if (exphead == NULL) {
/* Target an average linked list length of 10. */
exphashsize = num / 10;
if (exphashsize < 1)
exphashsize = 1;
else if (exphashsize > 100000)
exphashsize = 100000;
exphead = malloc(exphashsize * sizeof(*exphead));
oldexphead = malloc(exphashsize * sizeof(*oldexphead));
if (exphead == NULL || oldexphead == NULL)
errx(1, "Can't malloc hash tables");
for (i = 0; i < exphashsize; i++) {
SLIST_INIT(&exphead[i]);
SLIST_INIT(&oldexphead[i]);
}
}
for (i = 0; i < num; i++)
delete_export(iov, iovlen, &mntbufp[i], errmsg);
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/* Read the export file(s) and process them */
read_exportfile(0);
}
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
mountd(8): generate a syslog message when the "V4:" line is missing Daniel reported that NFSv4 mounts were not working despite having set "nfsv4_server_enable=YES" in /etc/rc.conf. Mountd was logging a message that there was no /etc/exports file. He noted that creating a /etc/exports file with a "V4:" line in it was needed make NFSv4 mounts work. At least one "V4:" line in one of the exports(5) file(s) is needed to make NFSv4 mounts work. This patch fixes mountd.c so that it logs a message indicting that there is no "V4:" line in any exports(5) file when NFSv4 mounts are enabled. To avoid this message being generated erroneously, /etc/rc.d/mountd is updated to make sure vfs.nfsd.server_max_nfsvers is properly set before mountd(8) is started. Reported by: debdrup PR: 253901 MFC after: 2 weeks
2021-03-09 00:08:02 +00:00
if (strlen(v4root_dirpath) == 0) {
/* Check to see if a V4: line is needed. */
nfs_maxvers_size = sizeof(nfs_maxvers);
error = sysctlbyname("vfs.nfsd.server_max_nfsvers",
&nfs_maxvers, &nfs_maxvers_size, NULL, 0);
if (error != 0 || nfs_maxvers < NFS_VER2 || nfs_maxvers >
NFS_VER4) {
syslog(LOG_ERR, "sysctlbyname(vfs.nfsd."
"server_max_nfsvers) failed, defaulting to NFSv3");
nfs_maxvers = NFS_VER3;
}
if (nfs_maxvers == NFS_VER4)
syslog(LOG_ERR, "NFSv4 requires at least one V4: line");
}
Add support for multiple exports files. This will be useful for example for ZFS, where we have automatically generated /etc/zfs/exports file, which should not be edited directly. Discussed with: rwatson
2007-02-03 00:12:18 +00:00
if (iov != NULL) {
/* Free strings allocated by strdup() in getmntopts.c */
free(iov[0].iov_base); /* fstype */
free(iov[2].iov_base); /* fspath */
free(iov[4].iov_base); /* from */
free(iov[6].iov_base); /* update */
free(iov[8].iov_base); /* export */
free(iov[10].iov_base); /* errmsg */
/* free iov, allocated by realloc() */
free(iov);
iovlen = 0;
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
/*
* If there was no public fh, clear any previous one set.
*/
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
if (has_publicfh == 0) {
LOGDEBUG("clear public fh");
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
(void) nfssvc(NFSSVC_NOPUBLICFH, NULL);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
}
Add a new '-S' option to mountd, which tells it to suspend execution of the nfsd threads while it is reloading the exports. This avoids clients from getting intermittent access errors when the exports are being reloaded non-atomically. It is not an ideal solution, since requests will back up while the nfsd threads are suspended. Also, when this option is used, if mountd crashes while reloading exports, mountd will have to be restarted to get the nfsd threads to resume execution. This has been tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). The nfse patch offers a more comprehensive solution for this issue. PR: kern/9619, kern/131342 Reviewed by: kib MFC after: 2 weeks
2012-10-15 00:17:16 +00:00
/* Resume the nfsd. If they weren't suspended, this is harmless. */
(void)nfssvc(NFSSVC_RESUMENFSD, NULL);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
LOGDEBUG("eo get_exportlist");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
/*
* Insert an export entry in the appropriate list.
*/
static void
insert_exports(struct exportlist *ep, struct exportlisthead *exhp)
{
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
uint32_t i;
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
i = EXPHASH(&ep->ex_fs);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
LOGDEBUG("fs=%s hash=%i", ep->ex_fsdir, i);
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
SLIST_INSERT_HEAD(&exhp[i], ep, entries);
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
}
/*
* Free up the exports lists passed in as arguments.
*/
static void
free_exports(struct exportlisthead *exhp)
{
struct exportlist *ep, *ep2;
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
int i;
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
for (i = 0; i < exphashsize; i++) {
SLIST_FOREACH_SAFE(ep, &exhp[i], entries, ep2) {
SLIST_REMOVE(&exhp[i], ep, exportlist, entries);
free_exp(ep);
}
SLIST_INIT(&exhp[i]);
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
}
}
Factor code into two new functions in preparation for a future commit. Factor code into two functions. read_exportfile() a functon which reads the exports file(s) and calls get_exportlist_one() to process each of them. delete_export() a function which deletes the exports in the kernel for a file system. The contents of these functions is just the same code as was used to do the operations, moved into separate functions. As such, there is no semantic change. This is being done in preparation for a future commit that will add an option to do incremental changes of kernel exports upon receiving SIGHUP. MFC after: 1 month
2019-05-11 22:41:58 +00:00
/*
* Read the exports file(s) and call get_exportlist_one() for each line.
*/
static void
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
read_exportfile(int passno)
Factor code into two new functions in preparation for a future commit. Factor code into two functions. read_exportfile() a functon which reads the exports file(s) and calls get_exportlist_one() to process each of them. delete_export() a function which deletes the exports in the kernel for a file system. The contents of these functions is just the same code as was used to do the operations, moved into separate functions. As such, there is no semantic change. This is being done in preparation for a future commit that will add an option to do incremental changes of kernel exports upon receiving SIGHUP. MFC after: 1 month
2019-05-11 22:41:58 +00:00
{
int done, i;
/*
* Read in the exports file and build the list, calling
* nmount() as we go along to push the export rules into the kernel.
*/
done = 0;
for (i = 0; exnames[i] != NULL; i++) {
if (debug)
warnx("reading exports from %s", exnames[i]);
if ((exp_file = fopen(exnames[i], "r")) == NULL) {
syslog(LOG_WARNING, "can't open %s", exnames[i]);
continue;
}
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
get_exportlist_one(passno);
Factor code into two new functions in preparation for a future commit. Factor code into two functions. read_exportfile() a functon which reads the exports file(s) and calls get_exportlist_one() to process each of them. delete_export() a function which deletes the exports in the kernel for a file system. The contents of these functions is just the same code as was used to do the operations, moved into separate functions. As such, there is no semantic change. This is being done in preparation for a future commit that will add an option to do incremental changes of kernel exports upon receiving SIGHUP. MFC after: 1 month
2019-05-11 22:41:58 +00:00
fclose(exp_file);
done++;
}
if (done == 0) {
syslog(LOG_ERR, "can't open any exports file");
exit(2);
}
}
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/*
* Compare the export lists against the old ones and do nmount() operations
* for any cases that have changed. This avoids doing nmount() for entries
* that have not changed.
* Return 0 upon success, 1 otherwise.
*/
static int
compare_nmount_exportlist(struct iovec *iov, int iovlen, char *errmsg)
{
struct exportlist *ep, *oep;
struct grouplist *grp;
struct statfs fs, ofs;
int i, ret;
/*
* Loop through the current list and look for an entry in the old
* list.
* If found, check to see if it the same.
* If it is not the same, delete and re-export.
* Then mark it done on the old list.
* else (not found)
* export it.
* Any entries left in the old list after processing must have their
* exports deleted.
*/
for (i = 0; i < exphashsize; i++)
SLIST_FOREACH(ep, &exphead[i], entries) {
LOGDEBUG("foreach ep=%s", ep->ex_fsdir);
oep = ex_search(&ep->ex_fs, oldexphead);
if (oep != NULL) {
/*
* Check the mount paths are the same.
* If not, return 1 so that the reload of the
* exports will be done in bulk, the
* passno == 0 way.
*/
LOGDEBUG("found old exp");
if (strcmp(ep->ex_fsdir, oep->ex_fsdir) != 0)
return (1);
LOGDEBUG("same fsdir");
/*
* Test to see if the entry is the same.
* If not the same delete exports and
* re-export.
*/
if (compare_export(ep, oep) != 0) {
/*
* Clear has_publicfh if if was set
* in the old exports, but only if it
* has not been set during processing of
* the exports for this pass, as
* indicated by has_set_publicfh.
*/
if (has_set_publicfh == 0 &&
(oep->ex_flag & EX_PUBLICFH) != 0)
has_publicfh = 0;
/* Delete and re-export. */
if (statfs(ep->ex_fsdir, &fs) < 0)
return (1);
delete_export(iov, iovlen, &fs, errmsg);
ret = do_export_mount(ep, &fs);
if (ret != 0)
return (ret);
}
oep->ex_flag |= EX_DONE;
LOGDEBUG("exdone");
} else {
LOGDEBUG("not found so export");
/* Not found, so do export. */
if (statfs(ep->ex_fsdir, &fs) < 0)
return (1);
ret = do_export_mount(ep, &fs);
if (ret != 0)
return (ret);
}
}
/* Delete exports not done. */
for (i = 0; i < exphashsize; i++)
SLIST_FOREACH(oep, &oldexphead[i], entries) {
if ((oep->ex_flag & EX_DONE) == 0) {
LOGDEBUG("not done delete=%s", oep->ex_fsdir);
if (statfs(oep->ex_fsdir, &ofs) >= 0 &&
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
fsidcmp(&oep->ex_fs, &ofs.f_fsid) == 0) {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
LOGDEBUG("do delete");
/*
* Clear has_publicfh if if was set
* in the old exports, but only if it
* has not been set during processing of
* the exports for this pass, as
* indicated by has_set_publicfh.
*/
if (has_set_publicfh == 0 &&
(oep->ex_flag & EX_PUBLICFH) != 0)
has_publicfh = 0;
delete_export(iov, iovlen, &ofs,
errmsg);
}
}
}
/* Do the V4 root exports, as required. */
grp = NULL;
if (v4root_ep != NULL)
grp = v4root_ep->ex_grphead;
v4root_phase = 2;
while (v4root_ep != NULL && grp != NULL) {
LOGDEBUG("v4root expath=%s", v4root_dirpath);
ret = do_mount(v4root_ep, grp, grp->gr_exflags, &grp->gr_anon,
v4root_dirpath, strlen(v4root_dirpath), &fs,
grp->gr_numsecflavors, grp->gr_secflavors);
if (ret != 0) {
v4root_phase = 0;
return (ret);
}
grp = grp->gr_next;
}
v4root_phase = 0;
free_v4rootexp();
return (0);
}
/*
* Compare old and current exportlist entries for the fsid and return 0
* if they are the same, 1 otherwise.
*/
static int
compare_export(struct exportlist *ep, struct exportlist *oep)
{
struct grouplist *grp, *ogrp;
if (strcmp(ep->ex_fsdir, oep->ex_fsdir) != 0)
return (1);
if ((ep->ex_flag & EX_DEFSET) != (oep->ex_flag & EX_DEFSET))
return (1);
if ((ep->ex_defdir != NULL && oep->ex_defdir == NULL) ||
(ep->ex_defdir == NULL && oep->ex_defdir != NULL))
return (1);
if (ep->ex_defdir != NULL && (ep->ex_defdir->dp_flag & DP_DEFSET) !=
(oep->ex_defdir->dp_flag & DP_DEFSET))
return (1);
if ((ep->ex_flag & EX_DEFSET) != 0 && (ep->ex_defnumsecflavors !=
oep->ex_defnumsecflavors || ep->ex_defexflags !=
oep->ex_defexflags || compare_cred(&ep->ex_defanon,
&oep->ex_defanon) != 0 || compare_secflavor(ep->ex_defsecflavors,
oep->ex_defsecflavors, ep->ex_defnumsecflavors) != 0))
return (1);
/* Now, check all the groups. */
for (ogrp = oep->ex_grphead; ogrp != NULL; ogrp = ogrp->gr_next)
ogrp->gr_flag = 0;
for (grp = ep->ex_grphead; grp != NULL; grp = grp->gr_next) {
for (ogrp = oep->ex_grphead; ogrp != NULL; ogrp =
ogrp->gr_next)
if ((ogrp->gr_flag & GR_FND) == 0 &&
grp->gr_numsecflavors == ogrp->gr_numsecflavors &&
grp->gr_exflags == ogrp->gr_exflags &&
compare_cred(&grp->gr_anon, &ogrp->gr_anon) == 0 &&
compare_secflavor(grp->gr_secflavors,
ogrp->gr_secflavors, grp->gr_numsecflavors) == 0)
break;
if (ogrp != NULL)
ogrp->gr_flag |= GR_FND;
else
return (1);
}
for (ogrp = oep->ex_grphead; ogrp != NULL; ogrp = ogrp->gr_next)
if ((ogrp->gr_flag & GR_FND) == 0)
return (1);
return (0);
}
/*
* This algorithm compares two arrays of "n" items. It returns 0 if they are
* the "same" and 1 otherwise. Although suboptimal, it is always safe to
* return 1, which makes compare_nmount_export() reload the exports entry.
* "same" refers to having the same set of values in the two arrays.
* The arrays are in no particular order and duplicates (multiple entries
* in an array with the same value) is allowed.
* The algorithm is inefficient, but the common case of indentical arrays is
* handled first and "n" is normally fairly small.
* Since the two functions need the same algorithm but for arrays of
* different types (gid_t vs int), this is done as a macro.
*/
#define COMPARE_ARRAYS(a1, a2, n) \
do { \
int fnd, fndarray[(n)], i, j; \
/* Handle common case of identical arrays. */ \
for (i = 0; i < (n); i++) \
if ((a1)[i] != (a2)[i]) \
break; \
if (i == (n)) \
return (0); \
for (i = 0; i < (n); i++) \
fndarray[i] = 0; \
for (i = 0; i < (n); i++) { \
fnd = 0; \
for (j = 0; j < (n); j++) { \
if ((a1)[i] == (a2)[j]) { \
fndarray[j] = 1; \
fnd = 1; \
} \
} \
if (fnd == 0) \
return (1); \
} \
for (i = 0; i < (n); i++) \
if (fndarray[i] == 0) \
return (1); \
return (0); \
} while (0)
/*
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
* Compare two struct expcred's. Return 0 if the same and 1 otherwise.
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
*/
static int
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
compare_cred(struct expcred *cr0, struct expcred *cr1)
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
{
if (cr0->cr_uid != cr1->cr_uid || cr0->cr_ngroups != cr1->cr_ngroups)
return (1);
COMPARE_ARRAYS(cr0->cr_groups, cr1->cr_groups, cr0->cr_ngroups);
}
/*
* Compare two lists of security flavors. Return 0 if the same and 1 otherwise.
*/
static int
compare_secflavor(int *sec1, int *sec2, int nsec)
{
COMPARE_ARRAYS(sec1, sec2, nsec);
}
Factor code into two new functions in preparation for a future commit. Factor code into two functions. read_exportfile() a functon which reads the exports file(s) and calls get_exportlist_one() to process each of them. delete_export() a function which deletes the exports in the kernel for a file system. The contents of these functions is just the same code as was used to do the operations, moved into separate functions. As such, there is no semantic change. This is being done in preparation for a future commit that will add an option to do incremental changes of kernel exports upon receiving SIGHUP. MFC after: 1 month
2019-05-11 22:41:58 +00:00
/*
* Delete an exports entry.
*/
static void
delete_export(struct iovec *iov, int iovlen, struct statfs *fsp, char *errmsg)
{
struct xvfsconf vfc;
if (getvfsbyname(fsp->f_fstypename, &vfc) != 0) {
syslog(LOG_ERR, "getvfsbyname() failed for %s",
fsp->f_fstypename);
return;
}
/*
* We do not need to delete "export" flag from
* filesystems that do not have it set.
*/
if (!(fsp->f_flags & MNT_EXPORTED))
return;
/*
* Do not delete export for network filesystem by
* passing "export" arg to nmount().
* It only makes sense to do this for local filesystems.
*/
if (vfc.vfc_flags & VFCF_NETWORK)
return;
iov[1].iov_base = fsp->f_fstypename;
iov[1].iov_len = strlen(fsp->f_fstypename) + 1;
iov[3].iov_base = fsp->f_mntonname;
iov[3].iov_len = strlen(fsp->f_mntonname) + 1;
iov[5].iov_base = fsp->f_mntfromname;
iov[5].iov_len = strlen(fsp->f_mntfromname) + 1;
errmsg[0] = '\0';
/*
* EXDEV is returned when path exists but is not a
* mount point. May happens if raced with unmount.
*/
if (nmount(iov, iovlen, fsp->f_flags) < 0 && errno != ENOENT &&
errno != ENOTSUP && errno != EXDEV) {
syslog(LOG_ERR,
"can't delete exports for %s: %m %s",
fsp->f_mntonname, errmsg);
}
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Allocate an export list element
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct exportlist *
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_exp(void)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct exportlist *ep;
Use calloc() instead of an explicit memset. MFC after: 2 weeks
2011-07-14 07:35:28 +00:00
ep = (struct exportlist *)calloc(1, sizeof (struct exportlist));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (ep == (struct exportlist *)NULL)
out_of_mem();
return (ep);
}
/*
* Allocate a group list element
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct grouplist *
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_grp(void)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct grouplist *gp;
Use calloc() instead of an explicit memset. MFC after: 2 weeks
2011-07-14 07:35:28 +00:00
gp = (struct grouplist *)calloc(1, sizeof (struct grouplist));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (gp == (struct grouplist *)NULL)
out_of_mem();
return (gp);
}
/*
* Clean up upon an error in get_exportlist().
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
getexp_err(struct exportlist *ep, struct grouplist *grp, const char *reason)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct grouplist *tgrp;
mountd(8): Produce vaguely meaningful error messages Sponsored by: Dell EMC Isilon
2018-02-08 01:34:35 +00:00
if (!(opt_flags & OP_QUIET)) {
if (reason != NULL)
syslog(LOG_ERR, "bad exports list line '%s': %s", line,
reason);
else
syslog(LOG_ERR, "bad exports list line '%s'", line);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (ep && (ep->ex_flag & EX_LINKED) == 0)
free_exp(ep);
while (grp) {
tgrp = grp;
grp = grp->gr_next;
free_grp(tgrp);
}
}
/*
* Search the export list for a matching fs.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct exportlist *
Factor out some exportlist list operations into separate functions. This patch moves the code that removes and frees all exportlist elements out into a separate function called free_exports(). It does the same for the insertion of a new exportlist entry into a list. It also adds a second argument to ex_search() for the list to use. None of these changes have any semantic effect. They are being done to prepare the code for future patches that convert the single linked list for the exportlist to a hash table of lists and a patch that will do incremental changes of exports in the kernel. And it fixes the argument for SLIST_HEAD_INITIALIZER() to be a pointer, which doesn't really matter, since SLIST_HEAD_INITIALIZER() doesn't use the argument. MFC after: 1 month
2019-05-10 23:52:17 +00:00
ex_search(fsid_t *fsid, struct exportlisthead *exhp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct exportlist *ep;
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
uint32_t i;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Replace a single linked list with a hash table of lists. mountd.c uses a single linked list of "struct exportlist" structures, where there is one of these for each exported file system on the NFS server. This list gets long if there are a large number of file systems exported and the list must be searched for each line in the exports file(s) when SIGHUP causes the exports file(s) to be reloaded. A simple benchmark that traverses SLIST() elements and compares two 32bit fields in the structure for equal (which is what the search is) appears to take a couple of nsec. So, for a server with 72000 exported file systems, this can take about 5sec during reload of the exports file(s). By replacing the single linked list with a hash table with a target of 10 elements per list, the time should be reduced to less than 1msec. Peter Errikson (who has a server with 72000+ exported file systems) ran a test program using 5 hashes to see how they worked. fnv_32_buf(fsid,..., 0) fnv_32_buf(fsid,..., FNV1_32_INIT) hash32_buf(fsid,..., 0) hash32_buf(fsid,..., HASHINIT) - plus simply using the low order bits of fsid.val[0]. The first three behaved about equally well, with the first one being slightly better than the others. It has an average variation of about 4.5% about the target list length and that is what this patch uses. Peter Errikson also tested this hash table version and found that the performance wasn't measurably improved by a larger hash table, so a load factor of 10 appears adequate. Tested by: pen@lysator.liu.se (with other patches) PR: 237860 MFC after: 1 month
2019-05-31 01:28:48 +00:00
i = EXPHASH(fsid);
SLIST_FOREACH(ep, &exhp[i], entries) {
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
if (fsidcmp(&ep->ex_fs, fsid) == 0)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (ep);
}
mountd: Convert exportlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for the exportlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12502
2017-10-04 08:43:56 +00:00
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (ep);
}
/*
* Add a directory path to the list.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static char *
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
add_expdir(struct dirlist **dpp, char *cp, int len)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct dirlist *dp;
mountd: Remove unneeded cast Reported by: kib MFC after: 1 week X MFC With: r324007
2017-09-26 11:11:17 +00:00
dp = malloc(sizeof (struct dirlist));
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
if (dp == (struct dirlist *)NULL)
out_of_mem();
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
dp->dp_left = *dpp;
dp->dp_right = (struct dirlist *)NULL;
dp->dp_flag = 0;
dp->dp_hosts = (struct hostlist *)NULL;
mountd: Replace malloc+strcpy to strdup Reviewed by: bapt MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12503
2017-09-26 09:18:18 +00:00
dp->dp_dirp = strndup(cp, len);
if (dp->dp_dirp == NULL)
out_of_mem();
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*dpp = dp;
return (dp->dp_dirp);
}
/*
* Hang the dir list element off the dirpath binary tree as required
* and update the entry for host.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
hang_dirp(struct dirlist *dp, struct grouplist *grp, struct exportlist *ep,
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
int flags, struct expcred *anoncrp, uint64_t exflags)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct hostlist *hp;
struct dirlist *dp2;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
if (flags & OP_ALLDIRS) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (ep->ex_defdir)
free((caddr_t)dp);
else
ep->ex_defdir = dp;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
if (grp == (struct grouplist *)NULL) {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
ep->ex_flag |= EX_DEFSET;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
ep->ex_defdir->dp_flag |= DP_DEFSET;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
/* Save the default security flavors list. */
ep->ex_defnumsecflavors = ep->ex_numsecflavors;
if (ep->ex_numsecflavors > 0)
memcpy(ep->ex_defsecflavors, ep->ex_secflavors,
sizeof(ep->ex_secflavors));
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
cp_cred(&ep->ex_defanon, anoncrp);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
ep->ex_defexflags = exflags;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
} else while (grp) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
hp = get_ht();
hp->ht_grp = grp;
hp->ht_next = ep->ex_defdir->dp_hosts;
ep->ex_defdir->dp_hosts = hp;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
/* Save the security flavors list for this host set. */
grp->gr_numsecflavors = ep->ex_numsecflavors;
if (ep->ex_numsecflavors > 0)
memcpy(grp->gr_secflavors, ep->ex_secflavors,
sizeof(ep->ex_secflavors));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
grp = grp->gr_next;
}
} else {
/*
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
* Loop through the directories adding them to the tree.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
while (dp) {
dp2 = dp->dp_left;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
add_dlist(&ep->ex_dirl, dp, grp, flags, ep, anoncrp,
exflags);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
dp = dp2;
}
}
}
/*
* Traverse the binary tree either updating a node that is already there
* for the new directory or adding the new node.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
add_dlist(struct dirlist **dpp, struct dirlist *newdp, struct grouplist *grp,
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
int flags, struct exportlist *ep, struct expcred *anoncrp,
uint64_t exflags)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct dirlist *dp;
struct hostlist *hp;
int cmp;
dp = *dpp;
if (dp) {
cmp = strcmp(dp->dp_dirp, newdp->dp_dirp);
if (cmp > 0) {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
add_dlist(&dp->dp_left, newdp, grp, flags, ep, anoncrp,
exflags);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
} else if (cmp < 0) {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
add_dlist(&dp->dp_right, newdp, grp, flags, ep, anoncrp,
exflags);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
} else
free((caddr_t)newdp);
} else {
dp = newdp;
dp->dp_left = (struct dirlist *)NULL;
*dpp = dp;
}
if (grp) {
/*
* Hang all of the host(s) off of the directory point.
*/
do {
hp = get_ht();
hp->ht_grp = grp;
hp->ht_next = dp->dp_hosts;
dp->dp_hosts = hp;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
/* Save the security flavors list for this host set. */
grp->gr_numsecflavors = ep->ex_numsecflavors;
if (ep->ex_numsecflavors > 0)
memcpy(grp->gr_secflavors, ep->ex_secflavors,
sizeof(ep->ex_secflavors));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
grp = grp->gr_next;
} while (grp);
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
} else {
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
ep->ex_flag |= EX_DEFSET;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
dp->dp_flag |= DP_DEFSET;
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
/* Save the default security flavors list. */
ep->ex_defnumsecflavors = ep->ex_numsecflavors;
if (ep->ex_numsecflavors > 0)
memcpy(ep->ex_defsecflavors, ep->ex_secflavors,
sizeof(ep->ex_secflavors));
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
cp_cred(&ep->ex_defanon, anoncrp);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
ep->ex_defexflags = exflags;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
/*
* Search for a dirpath on the export point.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct dirlist *
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
dirp_search(struct dirlist *dp, char *dirp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
int cmp;
if (dp) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
cmp = strcmp(dp->dp_dirp, dirp);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (cmp > 0)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
return (dirp_search(dp->dp_left, dirp));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
else if (cmp < 0)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
return (dirp_search(dp->dp_right, dirp));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
else
return (dp);
}
return (dp);
}
/*
* Scan for a host match in a directory tree.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
chk_host(struct dirlist *dp, struct sockaddr *saddr, int *defsetp,
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
int *hostsetp, int *numsecflavors, int **secflavorsp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct hostlist *hp;
struct grouplist *grp;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
struct addrinfo *ai;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (dp) {
if (dp->dp_flag & DP_DEFSET)
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
*defsetp = dp->dp_flag;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
hp = dp->dp_hosts;
while (hp) {
grp = hp->ht_grp;
switch (grp->gr_type) {
case GT_HOST:
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
ai = grp->gr_ptr.gt_addrinfo;
for (; ai; ai = ai->ai_next) {
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (!sacmp(ai->ai_addr, saddr, NULL)) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
*hostsetp =
(hp->ht_flag | DP_HOSTSET);
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
if (numsecflavors != NULL) {
*numsecflavors =
grp->gr_numsecflavors;
*secflavorsp =
grp->gr_secflavors;
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
return (1);
}
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
break;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
case GT_NET:
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (!sacmp(saddr, (struct sockaddr *)
&grp->gr_ptr.gt_net.nt_net,
(struct sockaddr *)
&grp->gr_ptr.gt_net.nt_mask)) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
*hostsetp = (hp->ht_flag | DP_HOSTSET);
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
if (numsecflavors != NULL) {
*numsecflavors =
grp->gr_numsecflavors;
*secflavorsp =
grp->gr_secflavors;
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
return (1);
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
break;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
hp = hp->ht_next;
}
}
return (0);
}
/*
* Scan tree for a host that matches the address.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
scan_tree(struct dirlist *dp, struct sockaddr *saddr)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
int defset, hostset;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (dp) {
if (scan_tree(dp->dp_left, saddr))
return (1);
Attila Bogar reported a bug in mountd when multiple export entries with different security flavors are in the exports(5) file. For that case, mountd replies with the security flavors of the last entry and not the correct one for the client host. This patch fixes that by storing separate copies of the flavors for each host/net case, plus a default one for the case where no hosts/nets are specified on an entry in the exports(5) file. Unlike the patch in the PR, it replies with the security flavors for the entry instead of merging the security flavors for all the entries and replying with that. Tested by: attila.bogar at linguamatics.com PR: kern/164933 MFC after: 2 weeks
2012-09-24 23:57:16 +00:00
if (chk_host(dp, saddr, &defset, &hostset, NULL, NULL))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
if (scan_tree(dp->dp_right, saddr))
return (1);
}
return (0);
}
/*
* Traverse the dirlist tree and free it up.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
free_dir(struct dirlist *dp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
if (dp) {
free_dir(dp->dp_left);
free_dir(dp->dp_right);
free_host(dp->dp_hosts);
mountd: Avoid memory leak by freeing dp_dirp Introduced in r324007, the data alloced by strdup was never free'ed. While here, remove cast to caddr_t when freeing dp. Reported by: bde MFC after: 1 week X MFC With: r324007
2017-09-26 12:15:13 +00:00
free(dp->dp_dirp);
free(dp);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
}
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
/*
* Parse a colon separated list of security flavors
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
parsesec(char *seclist, struct exportlist *ep)
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
{
char *cp, savedc;
int flavor;
ep->ex_numsecflavors = 0;
for (;;) {
cp = strchr(seclist, ':');
if (cp) {
savedc = *cp;
*cp = '\0';
}
if (!strcmp(seclist, "sys"))
flavor = AUTH_SYS;
else if (!strcmp(seclist, "krb5"))
flavor = RPCSEC_GSS_KRB5;
else if (!strcmp(seclist, "krb5i"))
flavor = RPCSEC_GSS_KRB5I;
else if (!strcmp(seclist, "krb5p"))
flavor = RPCSEC_GSS_KRB5P;
else {
if (cp)
*cp = savedc;
syslog(LOG_ERR, "bad sec flavor: %s", seclist);
return (1);
}
if (ep->ex_numsecflavors == MAXSECFLAVORS) {
if (cp)
*cp = savedc;
syslog(LOG_ERR, "too many sec flavors: %s", seclist);
return (1);
}
ep->ex_secflavors[ep->ex_numsecflavors] = flavor;
ep->ex_numsecflavors++;
if (cp) {
*cp = savedc;
seclist = cp + 1;
} else {
break;
}
}
return (0);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Parse the option string and update fields.
* Option arguments may either be -<option>=<value> or
* -<option> <value>
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
do_opt(char **cpp, char **endcpp, struct exportlist *ep, struct grouplist *grp,
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
int *has_hostp, uint64_t *exflagsp, struct expcred *cr)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
char *cpoptarg, *cpoptend;
char *cp, *endcp, *cpopt, savedc, savedc2;
int allflag, usedarg;
Put the umount request through the same translation as the mount request. Fix some uninitialized warnings and a style bug while we are here. Problem still exists where a failed umount can be misreported as success. Submitted by: Martin Blapp <mb@imp.ch>
1999-10-06 18:20:44 +00:00
savedc2 = '\0';
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
cpopt = *cpp;
cpopt++;
cp = *endcpp;
savedc = *cp;
*cp = '\0';
while (cpopt && *cpopt) {
allflag = 1;
usedarg = -2;
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
if ((cpoptend = strchr(cpopt, ','))) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*cpoptend++ = '\0';
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
if ((cpoptarg = strchr(cpopt, '=')))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*cpoptarg++ = '\0';
} else {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
if ((cpoptarg = strchr(cpopt, '=')))
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*cpoptarg++ = '\0';
else {
*cp = savedc;
nextfield(&cp, &endcp);
**endcpp = '\0';
if (endcp > cp && *cp != '-') {
cpoptarg = cp;
savedc2 = *endcp;
*endcp = '\0';
usedarg = 0;
}
}
}
if (!strcmp(cpopt, "ro") || !strcmp(cpopt, "o")) {
*exflagsp |= MNT_EXRDONLY;
} else if (cpoptarg && (!strcmp(cpopt, "maproot") ||
!(allflag = strcmp(cpopt, "mapall")) ||
!strcmp(cpopt, "root") || !strcmp(cpopt, "r"))) {
usedarg++;
parsecred(cpoptarg, cr);
if (allflag == 0) {
*exflagsp |= MNT_EXPORTANON;
opt_flags |= OP_MAPALL;
} else
opt_flags |= OP_MAPROOT;
} else if (cpoptarg && (!strcmp(cpopt, "mask") ||
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
!strcmp(cpopt, "m"))) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (get_net(cpoptarg, &grp->gr_ptr.gt_net, 1)) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "bad mask: %s", cpoptarg);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
usedarg++;
opt_flags |= OP_MASK;
} else if (cpoptarg && (!strcmp(cpopt, "network") ||
!strcmp(cpopt, "n"))) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (strchr(cpoptarg, '/') != NULL) {
if (debug)
fprintf(stderr, "setting OP_MASKLEN\n");
opt_flags |= OP_MASKLEN;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (grp->gr_type != GT_NULL) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "network/host conflict");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
} else if (get_net(cpoptarg, &grp->gr_ptr.gt_net, 0)) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "bad net: %s", cpoptarg);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
grp->gr_type = GT_NET;
*has_hostp = 1;
usedarg++;
opt_flags |= OP_NET;
} else if (!strcmp(cpopt, "alldirs")) {
opt_flags |= OP_ALLDIRS;
Merge WebNFS support from NetBSD. Obtained from: NetBSD
1997-07-16 09:27:53 +00:00
} else if (!strcmp(cpopt, "public")) {
*exflagsp |= MNT_EXPUBLIC;
} else if (!strcmp(cpopt, "webnfs")) {
*exflagsp |= (MNT_EXPUBLIC|MNT_EXRDONLY|MNT_EXPORTANON);
opt_flags |= OP_MAPALL;
} else if (cpoptarg && !strcmp(cpopt, "index")) {
ep->ex_indexfile = strdup(cpoptarg);
Try to give a more descriptive error message for the pilot error of attempting to export the non-root of a filesystem with -alldirs. This pilot error seems to be very common, and the "could not remount" error message doesn't give much hints about the real reason. See the old PR below for an example. While i was at it, make it possible to entirely omit the often annoying error message in that case by specifying the "quiet" exports flag. This allows to specify something like /cdrom -alldirs,ro,quiet <where to export to> which will silently fail if nothing is mounted under /cdrom, but do the rigth thing as soon as you mount something. While doing this, i've put the embedded example in the exports(5) man page into a subsection of its own as it ought to be. Thanks for Paul Southworth for reminding me about this problem. PR: bin/4448 MFC after: 1 month
2002-07-18 20:52:17 +00:00
} else if (!strcmp(cpopt, "quiet")) {
opt_flags |= OP_QUIET;
Check if the -sec option is given without an argument. PR: bin/170413 Submitted by: Andrey Simonenko <simon@comsys.ntu-kpi.kiev.ua> MFC after: 1 week
2013-02-20 12:40:26 +00:00
} else if (cpoptarg && !strcmp(cpopt, "sec")) {
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
if (parsesec(cpoptarg, ep))
return (1);
opt_flags |= OP_SEC;
usedarg++;
Add support for the NFS over TLS exports to mountd. Three new export flags are added to mountd that will restrict exported file system mounts to use TLS. Without these flags, TLS is allowed, but not required. The exports(5) man page will be updated in a future commit.
2020-08-30 21:21:58 +00:00
} else if (!strcmp(cpopt, "tls")) {
*exflagsp |= MNT_EXTLS;
} else if (!strcmp(cpopt, "tlscert")) {
*exflagsp |= (MNT_EXTLS | MNT_EXTLSCERT);
} else if (!strcmp(cpopt, "tlscertuser")) {
*exflagsp |= (MNT_EXTLS | MNT_EXTLSCERT |
MNT_EXTLSCERTUSER);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
} else {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "bad opt %s", cpopt);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
if (usedarg >= 0) {
*endcp = savedc2;
**endcpp = savedc;
if (usedarg > 0) {
*cpp = cp;
*endcpp = endcp;
}
return (0);
}
cpopt = cpoptend;
}
**endcpp = savedc;
return (0);
}
/*
* Translate a character string to the corresponding list of network
* addresses for a hostname.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_host(char *cp, struct grouplist *grp, struct grouplist *tgrp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Fixed stupid bug in mountd: it would seem that the kernel doesn't allow you to push the same host into its NFS export lists twice, but mountd tries to do it anyway. This means that putting: /some_file_system -ro host1 host1 in your /etc/exports file causes an error. This is bogus: mountd should be smart enough to ignore the second instance of host1. This can be a problem in some configurations that use netgroups. For example, each host in my netgroups database is has two entries: startide (startide,-,) (startide.ctr.columbia.edu,-,) When mountd sees this, it tries to put startide.ctr.columbia.edu into the export list *twice*. Just listing 'startide' /etc/exports list will also screw up because mountd will try to resolve the netgroup 'startide' instead of the hostname 'startide.' My solution is watch for duplicate entries in get_host() and mark them as grouptype GT_IGNORE, which do_mount() will now cheefully throw away. This is a bit of a kludge, but it was the least obtrusive fix I could come up with. Also silenced a compiler warning: arguments passwd to xdr_long() should be u_long, not int. :)
1995-03-26 23:36:38 +00:00
struct grouplist *checkgrp;
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
struct addrinfo *ai, *tai, hints;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
int ecode;
char host[NI_MAXHOST];
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (grp->gr_type != GT_NULL) {
syslog(LOG_ERR, "Bad netgroup type for ip host %s", cp);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
memset(&hints, 0, sizeof hints);
hints.ai_flags = AI_CANONNAME;
hints.ai_protocol = IPPROTO_UDP;
ecode = getaddrinfo(cp, NULL, &hints, &ai);
if (ecode != 0) {
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
syslog(LOG_ERR,"can't get address info for host %s", cp);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
return 1;
}
grp->gr_ptr.gt_addrinfo = ai;
while (ai != NULL) {
if (ai->ai_canonname == NULL) {
if (getnameinfo(ai->ai_addr, ai->ai_addrlen, host,
NI_WITHSCOPEID cleanup. Neither RFC 2553 nor RFC 3493 defines NI_WITHSCOPEID, and our getaddrinfo(3) does nothing special for it, now.
2005-05-13 16:31:11 +00:00
sizeof host, NULL, 0, NI_NUMERICHOST) != 0)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
strlcpy(host, "?", sizeof(host));
ai->ai_canonname = strdup(host);
ai->ai_flags |= AI_CANONNAME;
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (debug)
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
fprintf(stderr, "got host %s\n", ai->ai_canonname);
/*
* Sanity check: make sure we don't already have an entry
* for this host in the grouplist.
*/
for (checkgrp = tgrp; checkgrp != NULL;
checkgrp = checkgrp->gr_next) {
if (checkgrp->gr_type != GT_HOST)
continue;
for (tai = checkgrp->gr_ptr.gt_addrinfo; tai != NULL;
tai = tai->ai_next) {
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (sacmp(tai->ai_addr, ai->ai_addr, NULL) != 0)
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
continue;
if (debug)
fprintf(stderr,
"ignoring duplicate host %s\n",
ai->ai_canonname);
grp->gr_type = GT_IGNORE;
return (0);
}
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
ai = ai->ai_next;
}
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
grp->gr_type = GT_HOST;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (0);
}
/*
* Free up an exports list component
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
free_exp(struct exportlist *ep)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Replace global list for grouplist with list(s) for each exportlist element. In mountd.c, the grouplist structures are linked into a single global linked list headed by "grphead". The only use of this linked list is to free all list elements when the exportlist elements are also all being free'd at the time the exports are being reloaded. This patch replaces this one global linked list head with a list head in each exportlist structure, where the grouplist elements for that exported file system are linked. The only change is that now the grouplist elements are free'd with the associated exportlist element as they are free'd instead of all grouplist elements being free'd after the exportlist elements are free'd. This change should have no effect in practice. This is being done, since a future patch that will add a "-I" option for incrementally updating the exports in the kernel needs to know which grouplist elements are associated with each exported file system and having them linked into a list headed by the exportlist element does that. MFC after: 1 month
2019-05-14 22:00:47 +00:00
struct grouplist *grp, *tgrp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (ep->ex_defdir) {
free_host(ep->ex_defdir->dp_hosts);
free((caddr_t)ep->ex_defdir);
}
if (ep->ex_fsdir)
free(ep->ex_fsdir);
Merge WebNFS support from NetBSD. Obtained from: NetBSD
1997-07-16 09:27:53 +00:00
if (ep->ex_indexfile)
free(ep->ex_indexfile);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
free_dir(ep->ex_dirl);
Replace global list for grouplist with list(s) for each exportlist element. In mountd.c, the grouplist structures are linked into a single global linked list headed by "grphead". The only use of this linked list is to free all list elements when the exportlist elements are also all being free'd at the time the exports are being reloaded. This patch replaces this one global linked list head with a list head in each exportlist structure, where the grouplist elements for that exported file system are linked. The only change is that now the grouplist elements are free'd with the associated exportlist element as they are free'd instead of all grouplist elements being free'd after the exportlist elements are free'd. This change should have no effect in practice. This is being done, since a future patch that will add a "-I" option for incrementally updating the exports in the kernel needs to know which grouplist elements are associated with each exported file system and having them linked into a list headed by the exportlist element does that. MFC after: 1 month
2019-05-14 22:00:47 +00:00
grp = ep->ex_grphead;
while (grp) {
tgrp = grp;
grp = grp->gr_next;
free_grp(tgrp);
}
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (ep->ex_defanon.cr_groups != ep->ex_defanon.cr_smallgrps)
free(ep->ex_defanon.cr_groups);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
free((caddr_t)ep);
}
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/*
* Free up the v4root exports.
*/
static void
free_v4rootexp(void)
{
if (v4root_ep != NULL) {
free_exp(v4root_ep);
v4root_ep = NULL;
}
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Free hosts.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
free_host(struct hostlist *hp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct hostlist *hp2;
while (hp) {
hp2 = hp;
hp = hp->ht_next;
free((caddr_t)hp2);
}
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static struct hostlist *
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_ht(void)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
struct hostlist *hp;
hp = (struct hostlist *)malloc(sizeof (struct hostlist));
if (hp == (struct hostlist *)NULL)
out_of_mem();
hp->ht_next = (struct hostlist *)NULL;
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
hp->ht_flag = 0;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (hp);
}
/*
* Out of memory, fatal
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
out_of_mem(void)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "out of memory");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
exit(2);
}
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
/*
* Call do_mount() from the struct exportlist, for each case needed.
*/
static int
do_export_mount(struct exportlist *ep, struct statfs *fsp)
{
struct grouplist *grp, defgrp;
int ret;
size_t dirlen;
LOGDEBUG("do_export_mount=%s", ep->ex_fsdir);
dirlen = strlen(ep->ex_fsdir);
if ((ep->ex_flag & EX_DEFSET) != 0) {
defgrp.gr_type = GT_DEFAULT;
defgrp.gr_next = NULL;
/* We have an entry for all other hosts/nets. */
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
LOGDEBUG("ex_defexflags=0x%jx", (uintmax_t)ep->ex_defexflags);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
ret = do_mount(ep, &defgrp, ep->ex_defexflags, &ep->ex_defanon,
ep->ex_fsdir, dirlen, fsp, ep->ex_defnumsecflavors,
ep->ex_defsecflavors);
if (ret != 0)
return (ret);
}
/* Do a mount for each group. */
grp = ep->ex_grphead;
while (grp != NULL) {
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
LOGDEBUG("do mount gr_type=0x%x gr_exflags=0x%jx",
grp->gr_type, (uintmax_t)grp->gr_exflags);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
ret = do_mount(ep, grp, grp->gr_exflags, &grp->gr_anon,
ep->ex_fsdir, dirlen, fsp, grp->gr_numsecflavors,
grp->gr_secflavors);
if (ret != 0)
return (ret);
grp = grp->gr_next;
}
return (0);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
* Do the nmount() syscall with the update flag to push the export info into
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
* the kernel.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
do_mount(struct exportlist *ep, struct grouplist *grp, uint64_t exflags,
struct expcred *anoncrp, char *dirp, int dirplen, struct statfs *fsb,
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
int numsecflavors, int *secflavors)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
When exporting a directory that is not a mountpoint, mountd repeatedly removes the last path component until the mount() succeeds. However, the code never checks if it has passed the mountpoint, so in some cases where the mount() never succeeds, it can end up applying the flags from a mounted filesystem to the underlying one. Add a sanity check to the code which removes the last path component: test that the fsid associated with the new path is the same as that of the old one. PR: bin/7872
2001-04-22 21:50:30 +00:00
struct statfs fsb1;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
struct addrinfo *ai;
Remove oldnfs leftovers from mountd(8). Reviewed by: rmacklem@ Sponsored by: The FreeBSD Foundation
2015-04-29 12:33:00 +00:00
struct export_args *eap;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
char errmsg[255];
char *cp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
int done;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
char savedc;
struct iovec *iov;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int i, iovlen;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
int ret;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
struct nfsex_args nfsea;
Remove oldnfs leftovers from mountd(8). Reviewed by: rmacklem@ Sponsored by: The FreeBSD Foundation
2015-04-29 12:33:00 +00:00
eap = &nfsea.export;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
cp = NULL;
savedc = '\0';
iov = NULL;
iovlen = 0;
ret = 0;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
bzero(eap, sizeof (struct export_args));
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
bzero(errmsg, sizeof(errmsg));
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_flags = exflags;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
eap->ex_uid = anoncrp->cr_uid;
eap->ex_ngroups = anoncrp->cr_ngroups;
if (eap->ex_ngroups > 0) {
eap->ex_groups = malloc(eap->ex_ngroups * sizeof(gid_t));
memcpy(eap->ex_groups, anoncrp->cr_groups, eap->ex_ngroups *
sizeof(gid_t));
}
LOGDEBUG("do_mount exflags=0x%jx", (uintmax_t)exflags);
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_indexfile = ep->ex_indexfile;
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
if (grp->gr_type == GT_HOST)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
ai = grp->gr_ptr.gt_addrinfo;
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
else
ai = NULL;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
eap->ex_numsecflavors = numsecflavors;
LOGDEBUG("do_mount numsec=%d", numsecflavors);
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
for (i = 0; i < eap->ex_numsecflavors; i++)
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
eap->ex_secflavors[i] = secflavors[i];
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (eap->ex_numsecflavors == 0) {
eap->ex_numsecflavors = 1;
eap->ex_secflavors[0] = AUTH_SYS;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
done = FALSE;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase == 0) {
build_iovec(&iov, &iovlen, "fstype", NULL, 0);
build_iovec(&iov, &iovlen, "fspath", NULL, 0);
build_iovec(&iov, &iovlen, "from", NULL, 0);
build_iovec(&iov, &iovlen, "update", NULL, 0);
build_iovec(&iov, &iovlen, "export", eap,
sizeof (struct export_args));
build_iovec(&iov, &iovlen, "errmsg", errmsg, sizeof(errmsg));
}
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
while (!done) {
switch (grp->gr_type) {
case GT_HOST:
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
if (ai->ai_addr->sa_family == AF_INET6 && have_v6 == 0)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
goto skip;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_addr = ai->ai_addr;
eap->ex_addrlen = ai->ai_addrlen;
eap->ex_masklen = 0;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
break;
case GT_NET:
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (grp->gr_ptr.gt_net.nt_net.ss_family == AF_INET6 &&
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
have_v6 == 0)
goto skip;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_addr =
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
(struct sockaddr *)&grp->gr_ptr.gt_net.nt_net;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_addrlen =
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
((struct sockaddr *)&grp->gr_ptr.gt_net.nt_net)->sa_len;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_mask =
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
(struct sockaddr *)&grp->gr_ptr.gt_net.nt_mask;
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_masklen = ((struct sockaddr *)&grp->gr_ptr.gt_net.nt_mask)->sa_len;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
break;
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
case GT_DEFAULT:
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
eap->ex_addr = NULL;
eap->ex_addrlen = 0;
eap->ex_mask = NULL;
eap->ex_masklen = 0;
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
break;
Fixed stupid bug in mountd: it would seem that the kernel doesn't allow you to push the same host into its NFS export lists twice, but mountd tries to do it anyway. This means that putting: /some_file_system -ro host1 host1 in your /etc/exports file causes an error. This is bogus: mountd should be smart enough to ignore the second instance of host1. This can be a problem in some configurations that use netgroups. For example, each host in my netgroups database is has two entries: startide (startide,-,) (startide.ctr.columbia.edu,-,) When mountd sees this, it tries to put startide.ctr.columbia.edu into the export list *twice*. Just listing 'startide' /etc/exports list will also screw up because mountd will try to resolve the netgroup 'startide' instead of the hostname 'startide.' My solution is watch for duplicate entries in get_host() and mark them as grouptype GT_IGNORE, which do_mount() will now cheefully throw away. This is a bit of a kludge, but it was the least obtrusive fix I could come up with. Also silenced a compiler warning: arguments passwd to xdr_long() should be u_long, not int. :)
1995-03-26 23:36:38 +00:00
case GT_IGNORE:
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
ret = 0;
goto error_exit;
Fixed stupid bug in mountd: it would seem that the kernel doesn't allow you to push the same host into its NFS export lists twice, but mountd tries to do it anyway. This means that putting: /some_file_system -ro host1 host1 in your /etc/exports file causes an error. This is bogus: mountd should be smart enough to ignore the second instance of host1. This can be a problem in some configurations that use netgroups. For example, each host in my netgroups database is has two entries: startide (startide,-,) (startide.ctr.columbia.edu,-,) When mountd sees this, it tries to put startide.ctr.columbia.edu into the export list *twice*. Just listing 'startide' /etc/exports list will also screw up because mountd will try to resolve the netgroup 'startide' instead of the hostname 'startide.' My solution is watch for duplicate entries in get_host() and mark them as grouptype GT_IGNORE, which do_mount() will now cheefully throw away. This is a bit of a kludge, but it was the least obtrusive fix I could come up with. Also silenced a compiler warning: arguments passwd to xdr_long() should be u_long, not int. :)
1995-03-26 23:36:38 +00:00
break;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
default:
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "bad grouptype");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (cp)
*cp = savedc;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
ret = 1;
goto error_exit;
Cleanup unnecessary semicolons from utilities we all love.
2016-04-15 22:31:22 +00:00
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
* For V4:, use the nfssvc() syscall, instead of mount().
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase == 2) {
nfsea.fspec = v4root_dirpath;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
if (nfssvc(NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT,
(caddr_t)&nfsea) < 0) {
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
syslog(LOG_ERR, "Exporting V4: failed");
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
ret = 2;
goto error_exit;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
} else {
/*
* XXX:
* Maybe I should just use the fsb->f_mntonname path
* instead of looping back up the dirp to the mount
* point??
* Also, needs to know how to export all types of local
* exportable filesystems and not just "ufs".
*/
iov[1].iov_base = fsb->f_fstypename; /* "fstype" */
iov[1].iov_len = strlen(fsb->f_fstypename) + 1;
iov[3].iov_base = fsb->f_mntonname; /* "fspath" */
iov[3].iov_len = strlen(fsb->f_mntonname) + 1;
iov[5].iov_base = fsb->f_mntfromname; /* "from" */
iov[5].iov_len = strlen(fsb->f_mntfromname) + 1;
Avoid showing stale errors when nmount(2) fails. Sometimes nmount(2) will fail without setting errmsg. The previous (ignored) error would then be shown as the reason for the failed call if the next nmount(2) also fails without [ENOENT,ENOTSUP]. An example is when there is a tmpfs mounted with -o size. vfs_filteropt() adds 'size' as an error in errmsg due to 'size' not being in tmpfs_updateopts. Then tmpfs_mount returns [ENOTSUP] from nmount(2), which is then ignored. The next call may race with an unmount causing an invalid [EINVAL] that then does log an error, with the tmpfs errmsg. The race itself is a separate issue to fix as it is expected to have an [ENOENT] returned instead. In this example the mount being shown is actually nullfs, not tmpfs that the error is from. mountd[740]: can't delete exports for /poudriere/data/.m/exp-head-commit-test-devel/04/.npkg: Invalid argument mount option <size> is unknown It should only show: mountd[740]: can't delete exports for /poudriere/data/.m/exp-head-commit-test-devel/04/.npkg: Invalid argument MFC after: 2 weeks
2014-08-19 21:04:31 +00:00
errmsg[0] = '\0';
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
while (nmount(iov, iovlen, fsb->f_flags) < 0) {
if (cp)
*cp-- = savedc;
else
cp = dirp + dirplen - 1;
if (opt_flags & OP_QUIET) {
ret = 1;
goto error_exit;
}
if (errno == EPERM) {
if (debug)
Show error messages if nmount() failed. MFC after: 1 month
2012-08-27 19:08:47 +00:00
warnx("can't change attributes for %s: %s",
dirp, errmsg);
Try to give a more descriptive error message for the pilot error of attempting to export the non-root of a filesystem with -alldirs. This pilot error seems to be very common, and the "could not remount" error message doesn't give much hints about the real reason. See the old PR below for an example. While i was at it, make it possible to entirely omit the often annoying error message in that case by specifying the "quiet" exports flag. This allows to specify something like /cdrom -alldirs,ro,quiet <where to export to> which will silently fail if nothing is mounted under /cdrom, but do the rigth thing as soon as you mount something. While doing this, i've put the embedded example in the exports(5) man page into a subsection of its own as it ought to be. Thanks for Paul Southworth for reminding me about this problem. PR: bin/4448 MFC after: 1 month
2002-07-18 20:52:17 +00:00
syslog(LOG_ERR,
Show error messages if nmount() failed. MFC after: 1 month
2012-08-27 19:08:47 +00:00
"can't change attributes for %s: %s",
dirp, errmsg);
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
ret = 1;
goto error_exit;
}
if (opt_flags & OP_ALLDIRS) {
if (errno == EINVAL)
syslog(LOG_ERR,
Try to give a more descriptive error message for the pilot error of attempting to export the non-root of a filesystem with -alldirs. This pilot error seems to be very common, and the "could not remount" error message doesn't give much hints about the real reason. See the old PR below for an example. While i was at it, make it possible to entirely omit the often annoying error message in that case by specifying the "quiet" exports flag. This allows to specify something like /cdrom -alldirs,ro,quiet <where to export to> which will silently fail if nothing is mounted under /cdrom, but do the rigth thing as soon as you mount something. While doing this, i've put the embedded example in the exports(5) man page into a subsection of its own as it ought to be. Thanks for Paul Southworth for reminding me about this problem. PR: bin/4448 MFC after: 1 month
2002-07-18 20:52:17 +00:00
"-alldirs requested but %s is not a filesystem mountpoint",
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
dirp);
else
syslog(LOG_ERR,
"could not remount %s: %m",
dirp);
ret = 1;
goto error_exit;
}
/* back up over the last component */
Avoid reading one byte before the path buffer. This happens when there's only one component (e.g. "/foo"). This (mostly-harmless) bug has been present since June 1990 when it was commited to mountd.c SCCS version 5.9. Note: the bug is on the second changed line, the first line is changed for visual consistency. Reviewed by: cem, emaste, mckusick, rmacklem Found with: CHERI Obtained from: CheriBSD MFC after: 1 week Sponsored by: DARPA Differential Revision: https://reviews.freebsd.org/D25759
2020-07-22 21:44:51 +00:00
while (cp > dirp && *cp == '/')
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
cp--;
Avoid reading one byte before the path buffer. This happens when there's only one component (e.g. "/foo"). This (mostly-harmless) bug has been present since June 1990 when it was commited to mountd.c SCCS version 5.9. Note: the bug is on the second changed line, the first line is changed for visual consistency. Reviewed by: cem, emaste, mckusick, rmacklem Found with: CHERI Obtained from: CheriBSD MFC after: 1 week Sponsored by: DARPA Differential Revision: https://reviews.freebsd.org/D25759
2020-07-22 21:44:51 +00:00
while (cp > dirp && *(cp - 1) != '/')
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
cp--;
if (cp == dirp) {
if (debug)
warnx("mnt unsucc");
syslog(LOG_ERR, "can't export %s %s",
dirp, errmsg);
ret = 1;
goto error_exit;
}
savedc = *cp;
*cp = '\0';
/*
* Check that we're still on the same
* filesystem.
*/
if (statfs(dirp, &fsb1) != 0 ||
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
fsidcmp(&fsb1.f_fsid, &fsb->f_fsid) != 0) {
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
*cp = savedc;
Try to give a more descriptive error message for the pilot error of attempting to export the non-root of a filesystem with -alldirs. This pilot error seems to be very common, and the "could not remount" error message doesn't give much hints about the real reason. See the old PR below for an example. While i was at it, make it possible to entirely omit the often annoying error message in that case by specifying the "quiet" exports flag. This allows to specify something like /cdrom -alldirs,ro,quiet <where to export to> which will silently fail if nothing is mounted under /cdrom, but do the rigth thing as soon as you mount something. While doing this, i've put the embedded example in the exports(5) man page into a subsection of its own as it ought to be. Thanks for Paul Southworth for reminding me about this problem. PR: bin/4448 MFC after: 1 month
2002-07-18 20:52:17 +00:00
syslog(LOG_ERR,
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
"can't export %s %s", dirp,
errmsg);
ret = 1;
goto error_exit;
}
When exporting a directory that is not a mountpoint, mountd repeatedly removes the last path component until the mount() succeeds. However, the code never checks if it has passed the mountpoint, so in some cases where the mount() never succeeds, it can end up applying the flags from a mounted filesystem to the underlying one. Add a sanity check to the code which removes the last path component: test that the fsid associated with the new path is the same as that of the old one. PR: bin/7872
2001-04-22 21:50:30 +00:00
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
/*
* For the experimental server:
* If this is the public directory, get the file handle
* and load it into the kernel via the nfssvc() syscall.
*/
Remove oldnfs leftovers from mountd(8). Reviewed by: rmacklem@ Sponsored by: The FreeBSD Foundation
2015-04-29 12:33:00 +00:00
if ((exflags & MNT_EXPUBLIC) != 0) {
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
fhandle_t fh;
char *public_name;
if (eap->ex_indexfile != NULL)
public_name = eap->ex_indexfile;
else
public_name = dirp;
if (getfh(public_name, &fh) < 0)
syslog(LOG_ERR,
"Can't get public fh for %s", public_name);
else if (nfssvc(NFSSVC_PUBLICFH, (caddr_t)&fh) < 0)
syslog(LOG_ERR,
"Can't set public fh for %s", public_name);
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
else {
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
has_publicfh = 1;
Modify mountd so that it incrementally updates the kernel exports upon a reload. Without this patch, mountd would delete/load all exports from the exports file(s) when it receives a SIGHUP. This works fine for small exports file(s), but can take several seconds to do when there are large numbers (10000+) of exported file systems. Most of this time is spent doing the system calls that delete/export each of these file systems. When the "-S" option has been specified (the default these days), the nfsd threads are suspended for several seconds while the reload is done. This patch changes mountd so that it only does system calls for file systems where the exports have been changed/added/deleted as compared to the exports done for the previous load/reload of the exports file(s). Basically, when SIGHUP is posted to mountd, it saves the exportlist structures from the previous load and creates a new set of structures from the current exports file(s). Then it compares the current with the previous and only does system calls for cases that have been changed/added/deleted. The nfsd threads do not need to be suspended until the comparison step is being done. This results in a suspension period of milliseconds for a server with 10000+ exported file systems. There is some code using a LOGDEBUG() macro that allow runtime debugging output via syslog(LOG_DEBUG,...) that can be enabled by creating a file called /var/log/mountd.debug. This code is expected to be replaced with code that uses dtrace by cy@ in the near future, once issues w.r.t. dtrace in stable/12 have been resolved. The patch should not change the usage of the exports file(s), but improves the performance of reloading large exports file(s) where there are only a small number of changes done to the file(s). Tested by: pen@lysator.liu.se PR: 237860 Reviewed by: kib MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D20487
2019-06-03 22:58:51 +00:00
has_set_publicfh = 1;
ep->ex_flag |= EX_PUBLICFH;
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
skip:
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
if (ai != NULL)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
ai = ai->ai_next;
A few more mountd cleanups: - Remove some horrible code that faked a "struct addrinfo" to be later passed to freeaddrinfo(). Instead, add a new group type "GT_DEFAULT" used to denote that the filesystem is exported to the world, and treat this case separately. - Don't clear the AI_CANONNAME flag in a struct addrinfo returned by getaddrinfo. There's still a bit more struct addrinfo abuse left in here. - Simplify do_mount() slightly by using an addrinfo pointer to keep track of the current address.
2001-04-18 00:28:37 +00:00
if (ai == NULL)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
done = TRUE;
}
if (cp)
*cp = savedc;
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
error_exit:
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
free(eap->ex_groups);
Convert mountd to nmount(). Remove some hardcoded dependencies on ufs, cd9660, msdosfs, and ntfs, but not all dependencies.
2006-05-23 17:10:17 +00:00
/* free strings allocated by strdup() in getmntopts.c */
if (iov != NULL) {
free(iov[0].iov_base); /* fstype */
free(iov[2].iov_base); /* fspath */
free(iov[4].iov_base); /* from */
free(iov[6].iov_base); /* update */
free(iov[8].iov_base); /* export */
free(iov[10].iov_base); /* errmsg */
/* free iov, allocated by realloc() */
free(iov);
}
return (ret);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
/*
* Translate a net address.
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
*
* If `maskflg' is nonzero, then `cp' is a netmask, not a network address.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_net(char *cp, struct netmsk *net, int maskflg)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Reinstate one more old bugfix that got lost in the tirpc commit: always look up -network and -mask addresses numerically before trying getnetbyname(). Without this, we may end up attempting DNS queries on silly names such as "127.0.0.0.my-domain.com". See the commit log from revisions 1.21 and 1.20 for further details.
2001-04-23 10:12:31 +00:00
struct netent *np = NULL;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
char *name, *p, *prefp;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
struct sockaddr_in sin;
Reinstate one more old bugfix that got lost in the tirpc commit: always look up -network and -mask addresses numerically before trying getnetbyname(). Without this, we may end up attempting DNS queries on silly names such as "127.0.0.0.my-domain.com". See the commit log from revisions 1.21 and 1.20 for further details.
2001-04-23 10:12:31 +00:00
struct sockaddr *sa = NULL;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
struct addrinfo hints, *ai = NULL;
char netname[NI_MAXHOST];
long preflen;
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
p = prefp = NULL;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if ((opt_flags & OP_MASKLEN) && !maskflg) {
p = strchr(cp, '/');
*p = '\0';
prefp = p + 1;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
Reinstate one more old bugfix that got lost in the tirpc commit: always look up -network and -mask addresses numerically before trying getnetbyname(). Without this, we may end up attempting DNS queries on silly names such as "127.0.0.0.my-domain.com". See the commit log from revisions 1.21 and 1.20 for further details.
2001-04-23 10:12:31 +00:00
/*
* Check for a numeric address first. We wish to avoid
* possible DNS lookups in getnetbyname().
*/
if (isxdigit(*cp) || *cp == ':') {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
memset(&hints, 0, sizeof hints);
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/* Ensure the mask and the network have the same family. */
if (maskflg && (opt_flags & OP_NET))
hints.ai_family = net->nt_net.ss_family;
else if (!maskflg && (opt_flags & OP_HAVEMASK))
hints.ai_family = net->nt_mask.ss_family;
else
hints.ai_family = AF_UNSPEC;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
hints.ai_flags = AI_NUMERICHOST;
Reinstate one more old bugfix that got lost in the tirpc commit: always look up -network and -mask addresses numerically before trying getnetbyname(). Without this, we may end up attempting DNS queries on silly names such as "127.0.0.0.my-domain.com". See the commit log from revisions 1.21 and 1.20 for further details.
2001-04-23 10:12:31 +00:00
if (getaddrinfo(cp, NULL, &hints, &ai) == 0)
sa = ai->ai_addr;
if (sa != NULL && ai->ai_family == AF_INET) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
/*
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
* The address in `cp' is really a network address, so
* use inet_network() to re-interpret this correctly.
* e.g. "127.1" means 127.1.0.0, not 127.0.0.1.
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
*/
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
bzero(&sin, sizeof sin);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
sin.sin_family = AF_INET;
sin.sin_len = sizeof sin;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
sin.sin_addr = inet_makeaddr(inet_network(cp), 0);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (debug)
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
fprintf(stderr, "get_net: v4 addr %s\n",
inet_ntoa(sin.sin_addr));
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
sa = (struct sockaddr *)&sin;
Reinstate one more old bugfix that got lost in the tirpc commit: always look up -network and -mask addresses numerically before trying getnetbyname(). Without this, we may end up attempting DNS queries on silly names such as "127.0.0.0.my-domain.com". See the commit log from revisions 1.21 and 1.20 for further details.
2001-04-23 10:12:31 +00:00
}
}
if (sa == NULL && (np = getnetbyname(cp)) != NULL) {
bzero(&sin, sizeof sin);
sin.sin_family = AF_INET;
sin.sin_len = sizeof sin;
sin.sin_addr = inet_makeaddr(np->n_net, 0);
sa = (struct sockaddr *)&sin;
}
if (sa == NULL)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
goto fail;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (maskflg) {
/* The specified sockaddr is a mask. */
if (checkmask(sa) != 0)
goto fail;
bcopy(sa, &net->nt_mask, sa->sa_len);
opt_flags |= OP_HAVEMASK;
} else {
/* The specified sockaddr is a network address. */
bcopy(sa, &net->nt_net, sa->sa_len);
Back out msmith's recent commit which breaks using symbolic names for netmasks. Instead, reverse the order of the testing, so if a symbolic name starts with a digit, we'll see if we can make a network address out of it first. If that fails, then we'll call getnet...
1997-04-30 18:40:12 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/* Get a network name for the export list. */
if (np) {
name = np->n_name;
} else if (getnameinfo(sa, sa->sa_len, netname, sizeof netname,
NI_WITHSCOPEID cleanup. Neither RFC 2553 nor RFC 3493 defines NI_WITHSCOPEID, and our getaddrinfo(3) does nothing special for it, now.
2005-05-13 16:31:11 +00:00
NULL, 0, NI_NUMERICHOST) == 0) {
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
name = netname;
} else {
goto fail;
}
if ((net->nt_name = strdup(name)) == NULL)
out_of_mem();
/*
* Extract a mask from either a "/<masklen>" suffix, or
* from the class of an IPv4 address.
*/
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (opt_flags & OP_MASKLEN) {
preflen = strtol(prefp, NULL, 10);
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (preflen < 0L || preflen == LONG_MAX)
goto fail;
bcopy(sa, &net->nt_mask, sa->sa_len);
if (makemask(&net->nt_mask, (int)preflen) != 0)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
goto fail;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
opt_flags |= OP_HAVEMASK;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
*p = '/';
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
} else if (sa->sa_family == AF_INET &&
(opt_flags & OP_MASK) == 0) {
in_addr_t addr;
mountd: deprecate exports to a network without mask The exports file format allows export to a network using an explicit mask or prefix length (CIDR). It also allows a network with just a dotted address, in which case the historical mask was used. Deprecate this usage, and warn when it is used. Document that this is deprecated. MFC after: 1 month Reviewed by: rmacklem, bcr, #manpages Differential Revision: https://reviews.freebsd.org/D32713
2021-10-27 03:48:23 +00:00
syslog(LOG_WARNING,
"WARNING: No mask specified for %s, "
"using out-of-date default", name);
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
addr = ((struct sockaddr_in *)sa)->sin_addr.s_addr;
if (IN_CLASSA(addr))
preflen = 8;
else if (IN_CLASSB(addr))
preflen = 16;
else if (IN_CLASSC(addr))
preflen = 24;
mountd: deprecate exports to a network without mask The exports file format allows export to a network using an explicit mask or prefix length (CIDR). It also allows a network with just a dotted address, in which case the historical mask was used. Deprecate this usage, and warn when it is used. Document that this is deprecated. MFC after: 1 month Reviewed by: rmacklem, bcr, #manpages Differential Revision: https://reviews.freebsd.org/D32713
2021-10-27 03:48:23 +00:00
else if (IN_CLASSD(addr)) /* XXX Multicast??? */
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
preflen = 28;
else
preflen = 32; /* XXX */
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
bcopy(sa, &net->nt_mask, sa->sa_len);
makemask(&net->nt_mask, (int)preflen);
opt_flags |= OP_HAVEMASK;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (ai)
freeaddrinfo(ai);
return 0;
fail:
if (ai)
freeaddrinfo(ai);
return 1;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
/*
* Parse out the next white space separated field
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
nextfield(char **cp, char **endcp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
char *p;
Respect quotes and escapes when splitting exports fields. Without this r293305 was still unable to handle names with spaces. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-04-30 21:38:38 +00:00
char quot = 0;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
p = *cp;
while (*p == ' ' || *p == '\t')
p++;
Respect quotes and escapes when splitting exports fields. Without this r293305 was still unable to handle names with spaces. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-04-30 21:38:38 +00:00
*cp = p;
while (*p != '\0') {
if (quot) {
if (*p == quot)
quot = 0;
} else {
if (*p == '\\' && *(p + 1) != '\0')
p++;
else if (*p == '\'' || *p == '"')
quot = *p;
else if (*p == ' ' || *p == '\t')
break;
}
p++;
};
*endcp = p;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
/*
* Get an exports file line. Skip over blank lines and handle line
* continuations.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_line(void)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
char *p, *cp;
Use fgetln to remove the static limit on the length of lines in /etc/exports. Oversized lines were unlikely due to the large 10k limit, but any found would cause mountd to exit with an error. Also fix one or two compiler warnings.
2002-05-14 23:24:28 +00:00
size_t len;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
int totlen, cont_line;
/*
* Loop around ignoring blank lines and getting all continuation lines.
*/
p = line;
totlen = 0;
do {
Use fgetln to remove the static limit on the length of lines in /etc/exports. Oversized lines were unlikely due to the large 10k limit, but any found would cause mountd to exit with an error. Also fix one or two compiler warnings.
2002-05-14 23:24:28 +00:00
if ((p = fgetln(exp_file, &len)) == NULL)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (0);
cp = p + len - 1;
cont_line = 0;
while (cp >= p &&
(*cp == ' ' || *cp == '\t' || *cp == '\n' || *cp == '\\')) {
if (*cp == '\\')
cont_line = 1;
cp--;
len--;
}
Correct handling of continuation lines. Instead of treating the backslash as nothing, treat it like a space so that adjacent lines aren't glued together. PR: 8479 Submitted by: Adrian Filipi-Martin <adrian@ubergeeks.com>
2001-07-02 23:56:01 +00:00
if (cont_line) {
*++cp = ' ';
len++;
}
Use fgetln to remove the static limit on the length of lines in /etc/exports. Oversized lines were unlikely due to the large 10k limit, but any found would cause mountd to exit with an error. Also fix one or two compiler warnings.
2002-05-14 23:24:28 +00:00
if (linesize < len + totlen + 1) {
linesize = len + totlen + 1;
line = realloc(line, linesize);
if (line == NULL)
out_of_mem();
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Use fgetln to remove the static limit on the length of lines in /etc/exports. Oversized lines were unlikely due to the large 10k limit, but any found would cause mountd to exit with an error. Also fix one or two compiler warnings.
2002-05-14 23:24:28 +00:00
memcpy(line + totlen, p, len);
totlen += len;
line[totlen] = '\0';
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
} while (totlen == 0 || cont_line);
return (1);
}
/*
* Parse a description of a credential.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
parsecred(char *namelist, struct expcred *cr)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
char *name;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
int inpos;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
char *names;
struct passwd *pw;
struct group *gr;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
gid_t groups[NGROUPS_MAX + 1];
int ngroups;
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
unsigned long name_ul;
char *end = NULL;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
* Set up the unprivileged user.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
cr->cr_groups = cr->cr_smallgrps;
Make use of the UID_NOBODY and GID_NOGROUP definitions in sys/conf.h. r362214 exposed UID_NOBODY and GID_NOGROUP to userspace, so use them instead of the numbers. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D25281
2020-06-16 02:35:30 +00:00
cr->cr_uid = UID_NOBODY;
cr->cr_groups[0] = GID_NOGROUP;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
cr->cr_ngroups = 1;
/*
* Get the user's password table entry.
*/
Respect quotes and escapes when splitting exports fields. Without this r293305 was still unable to handle names with spaces. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-04-30 21:38:38 +00:00
names = namelist;
name = strsep_quote(&names, ":");
Allow /etc/exports to contain usernames/groups with spaces in them. If you are getting your users/groups from a directory service such as LDAP or AD it's possible for those usernames or groupnames to contain spaces. Submitted by: Sean E. Fagan Reviewed by: rmacklem MFC after: 1 week Sponsored by: iXsystems
2016-01-07 05:34:39 +00:00
/* Bug? name could be NULL here */
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
name_ul = strtoul(name, &end, 10);
if (*end != '\0' || end == name)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
pw = getpwnam(name);
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
else
pw = getpwuid((uid_t)name_ul);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Credentials specified as those of a user.
*/
if (names == NULL) {
if (pw == NULL) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "unknown user: %s", name);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
}
cr->cr_uid = pw->pw_uid;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
ngroups = NGROUPS_MAX + 1;
if (getgrouplist(pw->pw_name, pw->pw_gid, groups, &ngroups)) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "too many groups");
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
ngroups = NGROUPS_MAX + 1;
mountd: fix a crash when getgrouplist reports too many groups Previously the code only warned about the condition and then happily proceeded to use the too large value resulting in the array out-of-bounds access. Obtained from: Panzura (Chuanbo Zheng) MFC after: 10 days Sponsored by: Panzura
2018-04-16 09:17:36 +00:00
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
Pass an array of gid_t rather than an array of int to getgroups(). PR: 56646
2004-10-02 11:40:48 +00:00
* Compress out duplicate.
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (ngroups > 1 && groups[0] == groups[1]) {
ngroups--;
inpos = 2;
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
} else {
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
inpos = 1;
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
}
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (ngroups > NGROUPS_MAX)
ngroups = NGROUPS_MAX;
if (ngroups > SMALLNGROUPS)
cr->cr_groups = malloc(ngroups * sizeof(gid_t));
cr->cr_ngroups = ngroups;
cr->cr_groups[0] = groups[0];
memcpy(&cr->cr_groups[1], &groups[inpos], (ngroups - 1) *
sizeof(gid_t));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
}
/*
* Explicit credential specified as a colon separated list:
* uid:gid:gid:...
*/
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
if (pw != NULL) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
cr->cr_uid = pw->pw_uid;
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
} else if (*end != '\0' || end == name) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "unknown user: %s", name);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
} else {
cr->cr_uid = name_ul;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
cr->cr_ngroups = 0;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
while (names != NULL && *names != '\0' && cr->cr_ngroups < NGROUPS_MAX) {
Respect quotes and escapes when splitting exports fields. Without this r293305 was still unable to handle names with spaces. MFC after: 1 week Sponsored by: iXsystems, Inc.
2019-04-30 21:38:38 +00:00
name = strsep_quote(&names, ":");
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
name_ul = strtoul(name, &end, 10);
if (*end != '\0' || end == name) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if ((gr = getgrnam(name)) == NULL) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "unknown group: %s", name);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
continue;
}
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
groups[cr->cr_ngroups++] = gr->gr_gid;
mountd: Fix handling of usernames that start with a digit yocalebo_gmail.com submitted a patch for mountd.c that fixes the case where a username starts with a digit. Without this patch, the username that starts with a digit is misinterpreted as a numeric uid. With this patch, any string that does not entirely convert to a decimal number via strtoul() is considered a user/group name. Submitted by: yocalebo_gmail.com Reviewed by: rmacklem MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D32993
2021-11-18 21:35:25 +00:00
} else {
groups[cr->cr_ngroups++] = name_ul;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
}
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
if (names != NULL && *names != '\0' && cr->cr_ngroups == NGROUPS_MAX)
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "too many groups");
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (cr->cr_ngroups > SMALLNGROUPS)
cr->cr_groups = malloc(cr->cr_ngroups * sizeof(gid_t));
memcpy(cr->cr_groups, groups, cr->cr_ngroups * sizeof(gid_t));
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
#define STRSIZ (MNTNAMLEN+MNTPATHLEN+50)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
* Routines that maintain the remote mounttab
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
get_mountlist(void)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
struct mountlist *mlp;
Merge from Lite2 (use new getvfsbyname() and mount(2) interface, cleanup)
1997-03-11 12:43:45 +00:00
char *host, *dirp, *cp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
char str[STRSIZ];
FILE *mlfile;
if ((mlfile = fopen(_PATH_RMOUNTLIST, "r")) == NULL) {
(1) Remove ISO support, it's dead in the kernel anyway. (2) Check for ENOENT when checking for /var/db/mountdtab (3) Remove a signal handler that called broken functions. (4) Remove the broken functions. Submitted by: Martin Blapp <mb@imp.ch> Reviewed by: bde (1), billf ([234])
1999-11-12 21:52:10 +00:00
if (errno == ENOENT)
return;
else {
syslog(LOG_ERR, "can't open %s", _PATH_RMOUNTLIST);
return;
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
while (fgets(str, STRSIZ, mlfile) != NULL) {
Merge from Lite2 (use new getvfsbyname() and mount(2) interface, cleanup)
1997-03-11 12:43:45 +00:00
cp = str;
host = strsep(&cp, " \t\n");
dirp = strsep(&cp, " \t\n");
if (host == NULL || dirp == NULL)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
continue;
mlp = (struct mountlist *)malloc(sizeof (*mlp));
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
if (mlp == (struct mountlist *)NULL)
out_of_mem();
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
strncpy(mlp->ml_host, host, MNTNAMLEN);
mlp->ml_host[MNTNAMLEN] = '\0';
strncpy(mlp->ml_dirp, dirp, MNTPATHLEN);
mlp->ml_dirp[MNTPATHLEN] = '\0';
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_INSERT_HEAD(&mlhead, mlp, next);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
fclose(mlfile);
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Various bugfixes and cleanups, mainly from Martin Blapp: - Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version, ours lets the caller generate any syslog() messages, so that it can include the service name in the message. - Initialise a few local variables to clarify the logic and avoid some compiler warnings. - Remove a few unused functions and local variables, and fix some whitespace issues. - Reinstate the logic for avoiding duplicate host entries that got removed accidentally in revision 1.41 (added in r1.5). This bit was submitted in a slightly different form by Thomas Quinot. Submitted by: Martin Blapp <mb@imp.ch>, Thomas Quinot <quinot@inf.enst.fr> PR: bin/26148
2001-04-17 22:25:48 +00:00
del_mlist(char *hostp, char *dirp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
struct mountlist *mlp, *mlp2;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
FILE *mlfile;
int fnd = 0;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_FOREACH_SAFE(mlp, &mlhead, next, mlp2) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (!strcmp(mlp->ml_host, hostp) &&
(!dirp || !strcmp(mlp->ml_dirp, dirp))) {
fnd = 1;
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_REMOVE(&mlhead, mlp, mountlist, next);
free((caddr_t)mlp);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
}
if (fnd) {
if ((mlfile = fopen(_PATH_RMOUNTLIST, "w")) == NULL) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR,"can't update %s", _PATH_RMOUNTLIST);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
}
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_FOREACH(mlp, &mlhead, next) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
fprintf(mlfile, "%s %s\n", mlp->ml_host, mlp->ml_dirp);
}
fclose(mlfile);
}
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
add_mlist(char *hostp, char *dirp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
struct mountlist *mlp;
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
FILE *mlfile;
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_FOREACH(mlp, &mlhead, next) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (!strcmp(mlp->ml_host, hostp) && !strcmp(mlp->ml_dirp, dirp))
return;
}
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
mlp = (struct mountlist *)malloc(sizeof (*mlp));
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
if (mlp == (struct mountlist *)NULL)
out_of_mem();
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
strncpy(mlp->ml_host, hostp, MNTNAMLEN);
mlp->ml_host[MNTNAMLEN] = '\0';
strncpy(mlp->ml_dirp, dirp, MNTPATHLEN);
mlp->ml_dirp[MNTPATHLEN] = '\0';
mountd: Convert mountlist to SLIST Use SLIST from sys/queue.h instead of homebrew linked list for mountlist. Reviewed by: bapt, rmacklem MFC after: 1 week Sponsored by: Gandi.net Differential Revision: https://reviews.freebsd.org/D12504
2017-10-04 08:48:05 +00:00
SLIST_INSERT_HEAD(&mlhead, mlp, next);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if ((mlfile = fopen(_PATH_RMOUNTLIST, "a")) == NULL) {
Correct improper use of .Sm. Document -d flag. Correct use of .Nm. Remove unused #includes. Add usage(). Use warnx(). Correct spelling. Abort when malloc() fails.
1998-07-15 06:21:41 +00:00
syslog(LOG_ERR, "can't update %s", _PATH_RMOUNTLIST);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return;
}
fprintf(mlfile, "%s %s\n", mlp->ml_host, mlp->ml_dirp);
fclose(mlfile);
}
/*
* Free up a group list.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
free_grp(struct grouplist *grp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
if (grp->gr_type == GT_HOST) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
if (grp->gr_ptr.gt_addrinfo != NULL)
freeaddrinfo(grp->gr_ptr.gt_addrinfo);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
} else if (grp->gr_type == GT_NET) {
if (grp->gr_ptr.gt_net.nt_name)
free(grp->gr_ptr.gt_net.nt_name);
}
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (grp->gr_anon.cr_groups != grp->gr_anon.cr_smallgrps)
free(grp->gr_anon.cr_groups);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
free((caddr_t)grp);
}
#ifdef DEBUG
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
SYSLOG(int pri, const char *fmt, ...)
{
va_list ap;
va_start(ap, fmt);
vfprintf(stderr, fmt, ap);
va_end(ap);
}
#endif /* DEBUG */
/*
* Check options for consistency.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
check_options(struct dirlist *dp)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase == 0 && dp == NULL)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
Userland part of nfs client/server split and cleanup.
2001-09-18 23:34:44 +00:00
if ((opt_flags & (OP_MAPROOT | OP_MAPALL)) == (OP_MAPROOT | OP_MAPALL)) {
syslog(LOG_ERR, "-mapall and -maproot mutually exclusive");
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (1);
}
if ((opt_flags & OP_MASK) && (opt_flags & OP_NET) == 0) {
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
syslog(LOG_ERR, "-mask requires -network");
return (1);
}
if ((opt_flags & OP_NET) && (opt_flags & OP_HAVEMASK) == 0) {
syslog(LOG_ERR, "-network requires mask specification");
return (1);
}
if ((opt_flags & OP_MASK) && (opt_flags & OP_MASKLEN)) {
syslog(LOG_ERR, "-mask and /masklen are mutually exclusive");
return (1);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Modify mountd to handle the experimental nfs server as well as the regular one. It now takes a "-4" command line argument to force it to use the experimental server. Otherwise it will use the regular server unless the experimental server is the only one linked into the kernel. A third kind of line has been added to /etc/exports, which is specific to NFSv4 and defines where the NFSv4 tree root is and can be used to limit access to NFSv4 state handling operations that do not use any file handle. Approved by: kib (mentor)
2009-05-27 22:02:54 +00:00
if (v4root_phase > 0 &&
(opt_flags &
~(OP_SEC | OP_MASK | OP_NET | OP_HAVEMASK | OP_MASKLEN)) != 0) {
syslog(LOG_ERR,"only -sec,-net,-mask options allowed on V4:");
return (1);
}
If the "-alldirs" export option was used for the V4: line, mountd would crash in check_options() since dp == NULL for the V4: line. This patch moves the check for options allowed on the V4: line to ahead of where dp is used to avoid this crash. Reported by: mamalos AT eng.auth.gr MFC after: 1 week
2010-05-06 00:24:08 +00:00
if ((opt_flags & OP_ALLDIRS) && dp->dp_left) {
syslog(LOG_ERR, "-alldirs has multiple directories");
return (1);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
return (0);
}
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
static int
check_path_component(const char *path, char **err)
{
struct stat sb;
if (lstat(path, &sb)) {
asprintf(err, "%s: lstat() failed: %s.\n",
path, strerror(errno));
return (0);
}
switch (sb.st_mode & S_IFMT) {
case S_IFDIR:
return (1);
case S_IFLNK:
asprintf(err, "%s: path is a symbolic link.\n", path);
break;
case S_IFREG:
asprintf(err, "%s: path is a file rather than a directory.\n",
path);
break;
default:
asprintf(err, "%s: path is not a directory.\n", path);
}
return (0);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
/*
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
* Check each path component for the presence of symbolic links. Return true
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
check_dirpath(char *dirp, char **err)
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
{
char *cp;
cp = dirp + 1;
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
while (*cp) {
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
if (*cp == '/') {
*cp = '\0';
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
if (!check_path_component(dirp, err)) {
*cp = '/';
return (0);
}
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
*cp = '/';
}
cp++;
}
mountd: Improve error message for exports lines Currently mountd print error message "symbolic link in export path or statfs failed" in case some path component in an exports line fails validation. This revision improves the error message by giving more information about the precise error as well as the path component that caused the issue. Submitted by: Andrew Walker <awalker@ixsystems.com> Reviewed by: mav, rmacklem Differential Revision: https://reviews.freebsd.org/D39840
2023-05-06 18:57:14 +00:00
if (!check_path_component(dirp, err))
return (0);
return (1);
}
/*
* Populate statfs information. Return true on success.
*/
static int
check_statfs(const char *dirp, struct statfs *fsb, char **err)
{
if (statfs(dirp, fsb)) {
asprintf(err, "%s: statfs() failed: %s\n", dirp,
strerror(errno));
return (0);
}
return (1);
BSD 4.4 Lite sbin Sources Note: XNSrouted and routed NOT imported here, they shall be imported with usr.sbin.
1994-05-26 06:35:07 +00:00
}
Changes to support version 3 of the NFS protocol. The version 2 support has been tested (client+server) against FreeBSD-2.0, IRIX 5.3 and FreeBSD-current (using a loopback mount). The version 2 support is stable AFAIK. The version 3 support has been tested with a loopback mount and minimally against an IRIX 5.3 server. It needs more testing and may have problems. I have patched amd to support the new variable length filehandles although it will still only use version 2 of the protocol. Before booting a kernel with these changes, nfs clients will need to at least build and install /usr/sbin/mount_nfs. Servers will need to build and install /usr/sbin/mountd. NFS diskless support is untested. Obtained from: Rick Macklem <rick@snowhite.cis.uoguelph.ca>
1995-06-27 11:07:30 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/*
* Make a netmask according to the specified prefix length. The ss_family
* and other non-address fields must be initialised before calling this.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
makemask(struct sockaddr_storage *ssp, int bitlen)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
{
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
u_char *p;
int bits, i, len;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if ((p = sa_rawaddr((struct sockaddr *)ssp, &len)) == NULL)
return (-1);
Use the standardized CHAR_BIT constant instead of NBBY in userland.
2002-09-25 04:06:37 +00:00
if (bitlen > len * CHAR_BIT)
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
return (-1);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
for (i = 0; i < len; i++) {
Use MIN macro from sys/param.h. MFC after: 2 weeks.
2016-05-02 01:49:42 +00:00
bits = MIN(CHAR_BIT, bitlen);
Fixed conversion of prefix length to a netmask.
2011-03-01 11:47:16 +00:00
*p++ = (u_char)~0 << (CHAR_BIT - bits);
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
bitlen -= bits;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
return 0;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/*
* Check that the sockaddr is a valid netmask. Returns 0 if the mask
* is acceptable (i.e. of the form 1...10....0).
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
checkmask(struct sockaddr *sa)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
{
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
u_char *mask;
int i, len;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if ((mask = sa_rawaddr(sa, &len)) == NULL)
return (-1);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
for (i = 0; i < len; i++)
if (mask[i] != 0xff)
break;
if (i < len) {
if (~mask[i] & (u_char)(~mask[i] + 1))
return (-1);
i++;
}
for (; i < len; i++)
if (mask[i] != 0)
return (-1);
return (0);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/*
* Compare two sockaddrs according to a specified mask. Return zero if
* `sa1' matches `sa2' when filtered by the netmask in `samask'.
Spelling fixes for usr.sbin/
2011-12-30 10:58:14 +00:00
* If samask is NULL, perform a full comparison.
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static int
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
sacmp(struct sockaddr *sa1, struct sockaddr *sa2, struct sockaddr *samask)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
{
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
unsigned char *p1, *p2, *mask;
int len, i;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (sa1->sa_family != sa2->sa_family ||
(p1 = sa_rawaddr(sa1, &len)) == NULL ||
(p2 = sa_rawaddr(sa2, NULL)) == NULL)
return (1);
switch (sa1->sa_family) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
case AF_INET6:
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (((struct sockaddr_in6 *)sa1)->sin6_scope_id !=
((struct sockaddr_in6 *)sa2)->sin6_scope_id)
return (1);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
break;
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/* Simple binary comparison if no mask specified. */
if (samask == NULL)
return (memcmp(p1, p2, len));
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/* Set up the mask, and do a mask-based comparison. */
if (sa1->sa_family != samask->sa_family ||
(mask = sa_rawaddr(samask, NULL)) == NULL)
return (1);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
for (i = 0; i < len; i++)
if ((p1[i] & mask[i]) != (p2[i] & mask[i]))
return (1);
return (0);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
/*
* Return a pointer to the part of the sockaddr that contains the
* raw address, and set *nbytes to its length in bytes. Returns
* NULL if the address family is unknown.
*/
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void *
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
sa_rawaddr(struct sockaddr *sa, int *nbytes) {
void *p;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
int len;
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
switch (sa->sa_family) {
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
case AF_INET:
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
len = sizeof(((struct sockaddr_in *)sa)->sin_addr);
p = &((struct sockaddr_in *)sa)->sin_addr;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
break;
case AF_INET6:
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
len = sizeof(((struct sockaddr_in6 *)sa)->sin6_addr);
p = &((struct sockaddr_in6 *)sa)->sin6_addr;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
break;
default:
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
p = NULL;
len = 0;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
The introduction of IPv6 support from NetBSD's mountd invalidated a number of assumptions related to the parsing of options in /etc/exports, and missed a few necessary new error checks. The main problems related to netmasks: an IPv6 network address missing a netmask would result in the filesystem being exported to the whole IPv6 world, non-continuous netmasks would be made continuous without any warnings, and nothing prevented you specifying an IPv4 mask with an IPv6 address. This change addresses these issues. As a side-effect we now store netmasks in sockaddr structs (this matches the kernel interface, and is closer to the way it used to be). Add a flag OP_HAVEMASK to keep track of whether or not we have successfully got a mask from any source. Replace some mask-related helper functions with versions that use the sockaddr-based masks. Also tidy up get_net() and fix the code that interprets IPv4 partial networks such as "127.1" as network rather than host addresses. Properly zero out some structures that were ending up partially containing junk from the stack, fix a few formatting issues, and add a comment noting some assumptions about export arguments.
2001-04-21 20:06:18 +00:00
if (nbytes != NULL)
*nbytes = len;
return (p);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
Add __unused. Ansi prototypes.
2010-12-20 09:28:28 +00:00
huphandler(int sig __unused)
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
{
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
Fix a long-standing bug relating to the handling of SIGHUP: mountd would call malloc, stdio and other library functions from the signal handler which is not safe due to reentrancy problems. Instead, add a simple handler that just sets a flag, and call the more complex function from main() when necessary. Unfortunately to be able to check this flag, we must expand the svc_run() call, but the RPC library makes that relatively easy to do.
2001-04-21 00:55:17 +00:00
got_sighup = 1;
}
Staticize some stuff in mountd(8); no functional changes. MFC after: 1 month Sponsored by: The FreeBSD Foundation
2015-07-04 08:40:48 +00:00
static void
terminate(int sig __unused)
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
{
Use pidfile(3) in mountd(8). There is no need to use /var/run/mountd.lock anymore.
2005-08-24 19:17:06 +00:00
pidfile_remove(pfh);
Don't use sys/nfs/rpcv2.h - it is part of the old kernel RPC implementation and will be removed.
2009-06-24 18:42:21 +00:00
rpcb_unset(MOUNTPROG, MOUNTVERS, NULL);
rpcb_unset(MOUNTPROG, MOUNTVERS3, NULL);
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
exit (0);
}
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
static void
cp_cred(struct expcred *outcr, struct expcred *incr)
{
outcr->cr_uid = incr->cr_uid;
outcr->cr_ngroups = incr->cr_ngroups;
Modify mountd.c so that it does not always malloc 4K for the map credentials. r362163 upgraded mountd so that it could handle MAX_NGROUPS groups for the anonymous user credentials (the ones provided by -maproot and -mapall exports options). The problem is that this resulted in every export structure growing by about 4Kbytes, because the cr_groups field went from 16->MAX_NGROUPS. This patch fixes this by only including a small 32 element cr_groups in the structure and then malloc()'ng cr_groups when a larger one is needed. The value of SMALLNGROUPS is arbitrarily set to 32, assuming most users used by -maproot or -mapall will be in <= 32 groups. Reviewed by: kib, freqlabs Differential Revision: https://reviews.freebsd.org/D26521
2020-10-10 00:01:40 +00:00
if (outcr->cr_ngroups > SMALLNGROUPS)
outcr->cr_groups = malloc(outcr->cr_ngroups * sizeof(gid_t));
else
outcr->cr_groups = outcr->cr_smallgrps;
Modify mountd to use the new struct export_args committed by r362158. r362158 modified struct export_args for make the ex_flags field 64bits and also changed the anonymous credentials to allow more than 16 groups. This patch fixes mountd.c to use the new structure. It does allocate larger exportlist and grouplist structures now. That will be fixed in a future commit. The only visible change will be that the credentials provided for the -maproot and -mapall exports options can now have more than 16 groups. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:40:00 +00:00
memcpy(outcr->cr_groups, incr->cr_groups, incr->cr_ngroups *
sizeof(gid_t));
}
Reference in New Issue Copy Permalink