mount.nfs: Fix retry= to handle lack of reserved port situation
In the case of several (>500) mounts running at the same time
with -o tcp, the number of attempts that succeed is about 300-500
because it run out of priviledged port (they are busy in TIME_WAIT
state).
Signed-off-by: Flavio Leitner <flavio.leitner@gmail.com> Signed-off-by: Neil Brown <neilb@suse.de>
Prior to David Howell's mount changes in 2.6.18, users who mounted
different directories which happened to be from the same filesystem on the
server would get different super blocks, and hence could choose different
mount options. As long as there were no hard linked files that crossed from
one subtree to another, this was quite safe.
Post the changes, if the two directories are on the same filesystem (have
the same 'fsid'), they will share the same super block, and hence the same
mount options.
Add a flag to allow users to elect not to share the NFS super block with
another mount point, even if the fsids are the same. This will allow
users to set different mount options for the two different super blocks, as
was previously possible. It is still up to the user to ensure that there
are no cache coherency issues when doing this, however the default
behaviour will be to share super blocks whenever two paths result in
the same fsid.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Signed-off-by: Neil Brown <neilb@suse.de>
Steve Dickson [Thu, 17 May 2007 00:04:19 +0000 (10:04 +1000)]
mount.nfs - NFSv4 mounts give wrong error message when server denies the mount
When nfs4 mount fail because the exported directory does
not exist, the mount command claims the local mount point
does not exist which is wrong. This patch fixes that problem
as well as makes the v4 mount failures look like v3/v2 failures.
Signed-off-by: Steve Dickson <steved@redhat.com> Signed-off-by: Neil Brown <neilb@suse.de>
In client_compose(), free() the hostent structure returned before
exiting. Normally, gethostbyaddr() returns a pointer to a static
struct, but this hostent comes from either get_reliable_hostbyaddr() or
get_hostent(), both which return a pointer they privately xmalloc()ed,
which thus can and should be free()d.
Signed-Off-By: Steinar H. Gunderson <sesse@debian.org>
Jeff Layton [Wed, 9 May 2007 14:19:33 +0000 (10:19 -0400)]
nfs-utils: have mountd hold open etab file to force inode number to change
This patch changes mountd to hold the etab file open so that when it's
changed by exportfs, the inode number should change. We then change
auth_reload to reload the file based on whether st_ino is different
from the last time it was checked. It also changes auth_reload to
maintain a static counter value and return it instead of a timestamp
and fixes up get_exportlist accordingly. Finally, it adds some
comments to xtab_write to warn people about editing the etab in place.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: NeilBrown <neilb@suse.de>
Neil Brown [Tue, 8 May 2007 04:43:28 +0000 (14:43 +1000)]
Remove rquotad
rquotad isn't really used by anyone - as you can tell by the fact that
it only works for ext2 and ext3 (if those).
The 'quota' package contains a working and maintained rquota and all
distros appear to use that one. So remove rquotad from this package
to avoid confusion.
Jeff Layton [Mon, 7 May 2007 11:53:32 +0000 (07:53 -0400)]
update manpages for showmount and mountd
This patch updates the manpages for showmount and mountd. It adds a
description of the new mountd -r option, and a caveat about the unreliability
of showmount -a.
Neil Brown [Thu, 3 May 2007 00:46:02 +0000 (10:46 +1000)]
Detect version of libblkid and act accordingly.
libblkid earlier than 1.40 has a memory leak bug that make it unsuitable
for use in mountd.
So detect the version and default to not using it if too old. Give appropriate
warnings in various cases.
Kevin Coffman [Thu, 19 Apr 2007 21:39:27 +0000 (17:39 -0400)]
Update libgssapi requirements
Update the required version of libgssapi from 0.9 to 0.11.
(Working with Heimdal requires 0.11. Symbol versioning was
introduced in 0.10 and should be used everywhere, although
not absolutely required.)
Neil Brown [Mon, 16 Apr 2007 01:15:40 +0000 (11:15 +1000)]
Be more cautious about use for privilege ports (<1024).
Ports < 1024 are a scarce resource and should not be used
carelessly. Technically they should be not used at all without
registration with IANA, but sometimes we need them despite that.
So: for the socket that RPC services listen on, don't use a <1024 port
by default. There is no need.
For sockets that we send messages on, that are long-lived, and that might
need to appear 'privileged', avoid using a number that is registered in
/etc/services if possible.
Neil Brown [Mon, 2 Apr 2007 03:26:05 +0000 (13:26 +1000)]
exportfs - test exportability of filesystems when exportfs is run.
When exporting a filesystem test to see if the kernel is likely
to accept the export and print suitable warning message if not.
Don't actually fail the 'exportfs' as by the time a MOUNT request
arrives, the filesystem might be exportable.
Neil Brown [Mon, 2 Apr 2007 03:26:01 +0000 (13:26 +1000)]
mountd - improve checks and error messages for export failure.
If an attempt is made to export a non-(dir|file), just ignore it.
This should get caught by exportfs.
If an attempt is made to export a non-exportable filesystem, report
an error. Hopefully exportfs can trap some these as well, but
catching them in mountd as well is good.
Neil Brown [Mon, 2 Apr 2007 03:25:40 +0000 (13:25 +1000)]
Tell NFS/lockd client what that local state number is.
Both SM_STAT and SM_MON can return the state of an NSM, but it is
unclear which NSM they return the state of, so the value cannot be
used, and lockd doesn't use it.
Document this confusion, and give the current state to the kernel
via a sysctl if that sysctl is available (since about 2.6.19).
This should make is possible for the NFS server to detect a small
class of bad SM_NOTIFY packets and not flush locks in that case.
Kevin Coffman [Fri, 30 Mar 2007 22:32:14 +0000 (18:32 -0400)]
Use newly added keytab functions
Use the new functions added in the previous patch.
Obtain machine credentials in a pre-determined order
Look for appropriate machine credentials in the following order:
root/<fqdn>@REALM
nfs/<fqdn>@REALM
host/<fqdn>@REALM
root/<any-name>@REALM
nfs/<any-name>@REALM
host/<any-name>@REALM
The first matching credential will be used.
Also, the machine credentials to be used are now determined
"on-demand" rather than at gssd startup. This allows keytab
additions to be noticed and used without requiring a restart of gssd.
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu> Signed-off-by: Neil Brown <neilb@suse.de>
Kevin Coffman [Fri, 30 Mar 2007 22:32:10 +0000 (18:32 -0400)]
Hide differences between MIT and Heimdal in macros
Clean up a lot of #ifdef'd code using macros, masking
the differences between MIT and Heimdal implementations.
The currently unused macros will be used in later patches.
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu> Signed-off-by: Neil Brown <neilb@suse.de>
Neil Brown [Thu, 29 Mar 2007 03:20:20 +0000 (13:20 +1000)]
Change default to use system rpcgen.
If system-installed rpcgen if such exists.
If none is found, build our own.
Override with
./configure --with-rpcgen=internal
for internal rpcgen or
./configure --with-rpcgen=/local/rpcgen
for a non-standard location.
Neil Brown [Thu, 29 Mar 2007 03:08:08 +0000 (13:08 +1000)]
statd - use dnsname to ensure correct matching of NOTIFY requests.
When lockd asks to monitor a host, we find the FQDN from the DNS
and remember that, both internally and in the /var/lib/nfs/sm/*
file.
When we receive an SM_NOTIFY request, we compare both the
mon_name and the source IP address against that DNS name to find
a match.
If a DNS name is not available, we fall back to the name provided by
lockd, which at least is known to map to an IP address via
gethostbyname.
Neil Brown [Thu, 29 Mar 2007 03:08:07 +0000 (13:08 +1000)]
statd - check for 'priv' when looking for duplicate registrations.
From the point of view of the client (lockd), the 'priv' blob is probably
the most important key, so make sure to not throw away requests with
new 'priv' information.
Neil Brown [Thu, 29 Mar 2007 03:08:01 +0000 (13:08 +1000)]
mountd - better response to failed attempts to export filesystems
If the kernel rejects an attempt to export a filesystem - e.g. because
it is not exportable, we shouldn't just ignore the error, but rather
should tell the kernel that the relevant filehandle or path cannot be supported.
We should really print out some error messages too.
Neil Brown [Fri, 23 Mar 2007 01:36:36 +0000 (12:36 +1100)]
statd - fix bug so statd can talk to kernel again.
We need to call statd_get_socket before dropping privileges so that we
have a privileged port. We use to do that when initialising
notification as the same socket was used for reboot notication as for
callbacks to the kernel. Now it is a different socket..
Neil Brown [Thu, 22 Mar 2007 05:56:20 +0000 (16:56 +1100)]
mount.nfs: Fix issue with -o user,exec
It would seem to make sense for mount.nfs to impose the
"-o user" => "-o noexec,nodev,nosuid"
rule. However if you give "user,exec" to /sbin/mount,
it will pass down
nodev,nosuid,user
with the 'exec' flag :-(
So we have to leave that handling of that particular rule to
/sbin/mount.
Neil Brown [Thu, 22 Mar 2007 05:06:22 +0000 (16:06 +1100)]
Never set SO_REUSEADDR on a UDP socket.
The effect is quite different from TCP sockets.
For TCP, it allows you to listen for new connections even if there
are outstanding old connections with the same local address.
For UDP, it allows other people to steal your packets by
binding to the same address.
Kevin Coffman [Tue, 20 Mar 2007 22:52:17 +0000 (18:52 -0400)]
Fix a couple of problems that crept into mount
Commit 6facb22402a0bd8cd49be2ed1a0856b24fef42f4 changed the allocation
of len to no longer get 20 extra bytes. It needs to get at least one
extra byte for a null character, otherwise a single extra option such
as "sec=krb5" is never copied in parse_opt() and is dropped.
Neil Brown [Tue, 20 Mar 2007 05:17:27 +0000 (16:17 +1100)]
statd - remove try_to_resolve
try_to_resolve is used to resolve a hostname when sending a notification.
But we now only send notifications to localhost, so name resolution is not
needed.
Neil Brown [Tue, 20 Mar 2007 00:47:28 +0000 (11:47 +1100)]
Allow disabling of libblkid usage.
Some versions of libblkid have a terrible memory leak which makes
mounted grow toooo big. So support
--disable-uuid
to remove the uuid functionality and liblkid with it.
Neil Brown [Tue, 20 Mar 2007 00:15:29 +0000 (11:15 +1100)]
Add --with-rpcgen= for configure so that the system rpcgen can be used.
Ultimately it makes sense to remove remove rpcgen from the nfs-utils
release as it is already in the glibc release. With this patch
you can use the system rpcgen to make sure it works.
It is not default yet, but it might be in a future release.
Neil Brown [Tue, 20 Mar 2007 00:13:42 +0000 (11:13 +1100)]
Handle -o remount better
On -o remount, we need to update the entry in mtab rather than
add a new one. update_mtab does this so use that.
However it might free some strings that shouldn't be freed, so
stop it from calling free - the program will exit soon anyway
so no exit is needed.
Neil Brown [Mon, 19 Mar 2007 22:50:33 +0000 (09:50 +1100)]
Prevent sm-notify from being run multiple times per reboot.
As "mount.nfs" can start statd, and as statd can start sm-notify,
the risk of sm-notify being run multiple times increases.
As this is not normally appropriate, sm-notify now creates a
file in /var/run which will stop future instances from being
run (though ofcourse this behaviour can be controlled by a
new command line option).
Karel Zak [Mon, 19 Mar 2007 19:33:17 +0000 (20:33 +0100)]
Correctly handle -f (fake) mount option.
The fake option has to write to mtab like a normal mount. Read mount(8) man
page for more details. It's very important for system init scripts that use
"-f" as a way how write info about mount points to /etc/mtab.
Signed-off-by: Karel Zak <kzak@redhat.com> Signed-off-by: Neil Brown <neilb@suse.de>
Neil Brown [Mon, 19 Mar 2007 00:52:30 +0000 (11:52 +1100)]
Handle 'crossmnt' when kernel asked for export options for a directory.
If the kernel finds a mountpoint below a 'crossmnt' export, it will ask
mounted what export options to use. With this patch it will return
the same export options as for the top 'crossmnt' export unless
more specific options have been given.
Neil Brown [Mon, 19 Mar 2007 00:41:21 +0000 (11:41 +1100)]
Allow exporting of a subfilesystem when 'crossmnt' is set.
If we are asked to export a filesystem which is not explicitly
exported, but an ancestor is exported as 'crossmnt', export the
filesystems with the same options as the ancestor.
This is the first step to making 'crossmnt' effectively export
a tree of filesystems.
Kevin Coffman [Fri, 16 Mar 2007 14:27:48 +0000 (10:27 -0400)]
Create two separate paths for pipefs_dir and pipefs_nfsdir.
Future work needs access to the base pipefs directory rather than
the nfs subdirectory. Create two separate paths called
pipefs_dir and pipefs_nfsdir with the name of each.
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu> Signed-off-by: Neil Brown <neilb@suse.de>
Kevin Coffman [Fri, 16 Mar 2007 14:27:46 +0000 (10:27 -0400)]
Allow any credential to be used for machine credentials
Don't restrict machine credentials to be "nfs/<machine.name>".
Use any usable credentials contained in the keytab file.
[We actually attempt to use the first entry found for each
realm, not every entry, in the keytab.]
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu> Signed-off-by: Neil Brown <neilb@suse.de>
Kevin Coffman [Fri, 16 Mar 2007 14:27:44 +0000 (10:27 -0400)]
Add option to allow root to use credentials other than machine credentials
Add a new option ("-n") to rpc.gssd to indicate that accesses as root
(uid 0) should not use machine credentials, but should instead use
"normal" Kerberos credentials obtained by root.
This change was prompted by a suggestion and patch from Daniel
Muntz <Dan.Muntz@netapp.com>. That patch suggested trying "normal"
credentials first and falling back to using machine creds for
uid 0 if normal creds failed.
This opens up the case where root may have credentials as "foo@REALM"
and begins accessing files. Then the context using those credentials
expires and must be renewed. If the credentials are now expired, then
root's new context would fall back and be created with the machine
credentials.
Instead, this patch insists that the administrator choose to use either
machine credentials for accesses by uid 0 (the default behavior, as
it was before) or "normal" credentials. In the latter case, arrangements
must be made to obtain credentials before attempting a mount. There
should be no doubts which credentials are used for uid 0.
Signed-off-by: Kevin Coffman <kwc@citi.umich.edu> Signed-off-by: Neil Brown <neilb@suse.de>
Neil Brown [Fri, 16 Mar 2007 07:07:41 +0000 (18:07 +1100)]
Correctly handle "user" and "users" mount options.
If "user" or "users" is given, then allow mount.nfs to be run
by a non-root user providing that the mountpoint, filesystem, and options
exactly match what is found in fstab.
For "user", record the user name in mtab so they can unmount the
filesystem later.
Also alwasys ignore auto, owner, group and their negations as well
as "_netdev", "comment" and "loop".
projects / nfs-utils.git / log