summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* mountstats: Fix spurious I/O errorsChuck Lever2014-11-171-0/+4
| | | | | | | | | | | | | | | | | When running mountstats under "watch," occassionally the output shows "close failed in file object desctructor: sys.excepthook is missing" and the data display is messed up. This seems to be a common problem when Python script output is piped to another program. Ensure stdout/stderr is completely flushed before mountstats exits, and add an IOError exception handler to catch these exceptions gracefully. Solution suggested by: http://bugs.python.org/issue11380 Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* rpc.mountd: set libtirpc nonblocking mode to avoid DOSBodo Stroesser2014-11-121-0/+17
| | | | | | | | | | | | | | | This patch is experimental. In works fine in that it removes the vulnerability against a DOS attack. rpc.mountd can be blocked by a bad client, that sends many RPC requests but never reads the responses. This might happen intentionally or caused by a wrong network config (MTU). The patch switches on the nonblocking mode of libtirpc. In that mode writes can block for a max of 2 seconds. Attackers are forced to send requests slower, as libtirpc will close a connection if it finds two requests to read at the same time. Reviewed-by: NeilBrown <neilb@suse.de> Signed-off-by: Bodo Stroesser <bstroesser@ts.fujitsu.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* rpc.mountd: set nonblocking mode with libtirpcBodo Stroesser2014-11-121-0/+7
| | | | | | | | | | | | If mountd is built with libtirpc the tcp listeners and the sockets waiting for UDP messages are not in non-blocking mode. Thus if running with multiple threads (-t XX), all threads will wake up from select on a connection request or a UDP message, but only one thread will succeed. All others will wait on accept() or read() for the next event. Reviewed-by: NeilBrown <neilb@suse.de> Signed-off-by: Bodo Stroesser <bstroesser@ts.fujitsu.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* rpc.mountd: set nonblocking mode if no libtirpcBodo Stroesser2014-11-123-33/+36
| | | | | | | | | | | | | If mountd is built without libtirpc and it is started using "-p XXX" option, the tcp listeners and the sockets waiting for UDP messages are not in non-blocking mode. Thus if running with multiple threads (-t XX), all threads will wake up from select on a connection request or a UDP message, but only one thread will succeed. All others will wait on accept() or read() for the next event. Reviewed-by: NeilBrown <neilb@suse.de> Signed-off-by: Bodo Stroesser <bstroesser@ts.fujitsu.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* configure: Fixed typo in configure.acSteve Dickson2014-11-061-1/+1
| | | | Signed-off-by: Steve Dickson <steved@redhat.com>
* configure: Fixed logic around $enable_gss and $enable_svcgssSteve Dickson2014-11-051-4/+2
| | | | Signed-off-by: Steve Dickson <steved@redhat.com>
* exportfs: changes handling of unresolvable entriesHenrique Martins2014-11-043-5/+12
| | | | | | | | | The patch to nfs/exportfs to allow nfsd to start when there are some, but not all, unresolvable entries in /etc/exports. Signed-off-by: Henrique Martins <linux@martins.cc> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsidmap: Correct a failure to set key timeout valuesBenjamin Coddington2014-11-041-0/+3
| | | | | | | | | The caller must be a possesor of the key to set the attributes, so link the destination keyring to the current thread's keyring before instantiation so that after instantiation the timeout can be set. Signed-off-by: Benjamin Coddington <bcodding@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsidmap: set the correct DEFAULT_KEYRING stringBenjamin Coddington2014-11-041-1/+1
| | | | | Signed-off-by: Benjamin Coddington <bcodding@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsidmap: keyctl_invalidate kernel compatibilityBenjamin Coddington2014-11-041-4/+17
| | | | | | | | | | Change the keyctl_invalidate call to use the syscall interface directly so that when building with libkeyutils missing keyctl_invalidate the build succeeds. Attempt to use _invalidate and fall back to _revoke if the current kernel is missing _invalidate. Signed-off-by: Benjamin Coddington <bcodding@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* mountstats: Don't display retrans stats unless there are re-xmitSigned-off-by: Chuck Lever2014-11-041-2/+5
| | | | | | | | Clean up display of RPC statistics by omitting retransmit statistics when there have been no retransmissions. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* umount: make it sensitive to -vKarel Zak2014-10-241-1/+13
| | | | | | | Let's print verbose messages like original non-libmount version. Signed-off-by: Karel Zak <kzak@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsstat: fix server rpc stats headerMarko Myllynen2014-10-221-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | linux/net/sunrpc/stats.c has: /* * Get RPC server stats */ void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp) { ... seq_printf(seq, "rpc %u %u %u %u %u\n", statp->rpccnt, statp->rpcbadfmt+statp->rpcbadauth+statp->rpcbadclnt, statp->rpcbadfmt, statp->rpcbadauth, statp->rpcbadclnt); ... But when /proc/net/rpc/nfsd contains rpc 1 2 3 4 5 nfsstat -sv prints: Server rpc stats: calls badcalls badclnt badauth xdrcall 1 2 3 4 5 Change this to match the kernel: Server rpc stats: calls badcalls badfmt badauth badclnt 1 2 3 4 5 Fixes https://bugzilla.redhat.com/show_bug.cgi?id=1142842. Signed-off-by: Marko Myllynen <myllynen@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Steve Dickson <steved@redhat.com>
* Centralize dependencies on the auth unit.Simo Sorce2014-09-303-6/+12
| | | | | | | | | | | | | | | | | | With this patch either gssproxy or rpc.svcgssd are started only if the auth module is requested, and it finds a keytab. If the wants are in the main nfs-client or nfs-server unit files then the two deamons are started unconditionally and would require conditions which we can test once and for all in a single unit file instead. Change also Before and After statments accordingly to properly serialize loading modules and starting daemons in 3 steps 1. load kernel GSS auth module 2. start GSS handling daemons 3. start NFS client/server daemons Signed-off-by: Simo Sorce <simo@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfs-client.target: Removed a ordering cycle with nfs-server.serviceSteve Dickson2014-09-291-1/+1
| | | | | | | | | | | | | | On a VM that boots very fast and with out kerberos enabled a systemd ordering cycle happens between nfs-client and nfs-server units when both were enabled. Having nfs-client start the gssd daemons in the same matter as the nfs-server does ('After=') seemed to eliminated the ordering cycle. Acked-by: J. Bruce Fields <bfields@fieldses.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* Release 1.3.1Steve Dickson2014-09-251-1/+1
| | | | Signed-off-by: Steve Dickson <steved@redhat.com>
* nfs-service: Added gssproxy supportSteve Dickson2014-09-251-3/+3
| | | | | | | | | | | | | | | | | | | When kernel have gssproxy support the the gssproxy daemon should be used to manage the GSSAPI creds. So this patch adds "calls" to the gssproxy daemon from the NFS server systemd unit file. When gssproxy is installed, gssproxy will be start and rpc.svcgssd will not be. When gssproxy is not installed the rpc.svcgssd daemon will be started. Note, there are already existing hooks in the rpc-svcgssd service file that will ensure the gssproxy will be started before rpc.svcgssd which allows the script not to start rpc.svcsdd when gssproxy is installed and running. Signed-off-by: Steve Dickson <steved@redhat.com>
* rpc.svcgssd: Add a configure switch to disable building the daemonSteve Dickson2014-09-252-6/+28
| | | | | | | | Now that gssproxy is supported on modern kernels, the svcgssd is no longer needed. This switch disables the building of the daemon. Signed-off-by: Steve Dickson <steved@redhat.com>
* systemd: manually insert auth_rpcgss module.J. Bruce Fields2014-09-243-1/+16
| | | | | | | | | | | | | | | | | | | We need to insert the auth_rpcgss module before starting rpc.svcgssd or gss-proxy, for two reasons: - gss-proxy needs access to the /proc/net/rpc/use-gss-proxy file to set up communication with knfsd. - the unit files need to able to test for the existance of the same path in order to decide whether the kernel supports gss-proxy or not. Currently we're using dependencies on proc-fs-nfsd.mount for this, but that works only because of the nfsd kernel module references some symbols in auth_rpcgss, which is an odd implementation detail we're likely to fix some day. Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsstat: document nfsstat io/ra server facilitiesMarko Myllynen2014-09-221-0/+8
| | | | | | | Document the newly added nfsstat io/ra server facilities. Signed-off-by: Marko Myllynen <myllynen@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsstat: add server ra cache statsMarko Myllynen2014-09-221-4/+25
| | | | | | | Add server read ahead cache statistics to nfsstat. Signed-off-by: Marko Myllynen <myllynen@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsstat: add server io statsMarko Myllynen2014-09-221-3/+23
| | | | | | | Add server io statistics to nfsstat. Signed-off-by: Marko Myllynen <myllynen@redhat.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: fetch NFSDCLTRACK_GRACE_START out of environmentJeff Layton2014-09-193-1/+112
| | | | | | | | | | | | | Allow the fetching of NFSDCLTRACK_GRACE_START out of environment variables. If it's present in the "create" or "init" upcalls, then we can use that to query the database to see whether there are any clients that have not issued a RECLAIM_COMPLETE since that time. If there aren't any, then we know that all reclaim activity is now done and we can then cue the kernel to lift the grace period. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: grab the NFSDCLTRACK_CLIENT_HAS_SESSION env var if it's presentJeff Layton2014-09-193-12/+55
| | | | | | | | | | ...and set the has_session field in the DB based on whether it's true or not. Since we no longer set the timestamp for v4.1+ clients on a check operation, we must be careful to set the timestamp to zero for v4.1+ clients found via the legacy tracker. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: update schema to v2Jeff Layton2014-09-191-8/+93
| | | | | | | | | | | | | | | | | | | | From: Jeff Layton <jlayton@poochiereds.net> In order to allow knfsd's lock manager to lift its grace period early, we need to figure out whether all clients have finished reclaiming their state not. Unfortunately, the current code doesn't allow us to ascertain this. All we track for each client is a timestamp that tells us when the last "check" or "create" operation came in. Not only is this insufficient with clients that use sessions, it's also wrong. We only want to update the timestamp on v4.1 clients when the "create" operation comes in or we can leave the server susceptible to edge condition #2 in RFC5661, section 8.4.3. Once the grace period is lifted, we disallow reclaim on subsequent reboots for clients that have not sent a RECLAIM_COMPLETE. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: overhaul database initializtionJeff Layton2014-09-193-85/+141
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | We have some possibility for races with nfsdcltrack when the DB schema is upgraded. Suppose we update the nfs-utils package on a machine after the DB has been initialized. With the current scheme of initializing the DB only during the "init" phase, we could end up with a new program that expects a new schema with an old database. We could try to do a one-time update when the package is installed, but that could be racy. We could get an upcall between when the program is installed and when we run the update. Also, relying on packaging to get that right is tricky at best. To fix this, change how the database initialization and checking of the schema revision works. On every upcall, attempt to open the db as we normally would. If that fails, then try to create the directory if it doesn't exist and then retry the open. If it fails again, then give up. If we get a successful open, then query the DB for the schema version. If it matches what we expect, then declare success and move on. If the query fails then assume that the DB isn't set up yet. Start an exclusive transaction, check the schema version again and then set up the DB if no one raced in to create it in the meantime. This should only add a tiny bit of overhead on most upcalls (just an extra select of the parameters table), and should improve the performance of the "init" upcall. It'll also make it possible to handle DB schema changes sanely. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: rename CLD_* constants with CLTRACK_* prefixesJeff Layton2014-09-191-6/+6
| | | | | | | | Since nfsdcld has been dead for a few years now, clean up the prefixes on the constants. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: update comments in sqlite.cJeff Layton2014-09-191-6/+4
| | | | | | | Clean up and fix some inaccuracies. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* sm-notify: inform the kernel if there were no hosts to notifyJeff Layton2014-09-191-0/+25
| | | | | | | | | | | | | | In the event that there no hosts to be notified after a reboot, there's no real reason to force lockd to wait the entire grace period before handing out locks. We're not expecting any reclaim requests to come in that situation. Have sm-notify do a write to /proc/fs/lockd/nlm_end_grace if that file is present. That informs the kernel that it's OK to go ahead and lift lockd's grace period early. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsdcltrack: fix some minor nits in the manpageJeff Layton2014-09-191-2/+3
| | | | | | | | nfsdcltrack isn't a daemon, and we should make mention of the section in RFC5661. Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* exportfs: Properly parse IPv6 literal strings with null terminationTodd Vierling2014-09-191-10/+8
| | | | | | | | | | | | | | | | | The original implementation was using strncpy() with a truncation length to an uninitialized stack buffer, leaving a string that was only null terminated by luck. While here, change to use no-copy semantics (no extra buffer) to avoid buffer overflows altogether. exportfs already modifies argv contents elsewhere, so this doesn't break anything anew. Fixes: 4663c648 (exportfs: Support raw IPv6 addresses with "client:/path") Signed-off-by: Todd Vierling <todd.vierling@oracle.com> Reviewed-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* Fixed typo in nfsstat -sv outputSteve Dickson2014-09-161-1/+1
| | | | Signed-off-by: Steve Dickson <steved@redhat.com>
* nfs_get_tcpclient/nfs_get_udpclient - make bind(2) implicitChris Perl2014-09-161-48/+20
| | | | | | | | | | | | | | | | | | | | | | | | | When attempting to establish a local ephemeral endpoint for a TCP or UDP socket, do not explicitly call bind(2), instead let it happen implicilty when the socket is first used. The main motivating factor for this change is when TCP runs out of unique ephemeral ports (i.e. cannot find any ephemeral ports which are not a part of *any* TCP connection). In this situation if you explicitly call bind(2), then the call will fail with EADDRINUSE. However, if you allow the allocation of an ephemeral port to happen implicitly as part of connect(2) (or other functions), then ephemeral ports can be reused, so long as the combination of (local_ip, local_port, remote_ip, remote_port) is unique for TCP sockets on the system. This doesn't matter for UDP sockets, but it seemed easiest to treat TCP and UDP sockets the same. This can allow mount.nfs(8) to continue to function successfully, even in the face of misbehaving applications which are creating a large number of TCP connections. Signed-off-by: Chris Perl <chris.perl@gmail.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* gssd: ensure that preferred_realm is non-NULL before passing it to strcmpJeff Layton2014-09-151-1/+1
| | | | | | | | | | | | | | It's possible for "preferred_realm" to be NULL, in which case we don't want to pass it to strcmp. Other places that use this variable test whether it's NULL first -- we need to do the same here. This should fix the gssd crash reported here: https://bugzilla.redhat.com/show_bug.cgi?id=1108615 Reported-by: Brian J. Murrell <brian@interlinx.bc.ca> Signed-off-by: Jeff Layton <jlayton@primarydata.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* exportfs: fix test of NULL pointer in host_pton()Natanael Copa2014-09-151-5/+7
| | | | | | | | This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=1083018 Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* Include libgen.h for basenameNatanael Copa2014-09-155-0/+5
| | | | | | | | | | According POSIX basename(3) should have an #include <libgen.h> There are a different GNU implementation too, that can be used with _GNU_SOURCE, but the POSIX version is good enough and more portable. Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* mountd: Only work around glibc bugs on glibcNatanael Copa2014-09-151-1/+1
| | | | | Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* Allow usage of getrpcbynumber() when getrpcbynumber_r() is unavailableNatanael Copa2014-09-152-8/+13
| | | | | Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* nfsstat: replace the legacy SA_ONESHOT with standard SA_RESETHANDNatanael Copa2014-09-151-1/+1
| | | | | Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* mountd: use standard dev_t instead of glibc internalsNatanael Copa2014-09-151-1/+1
| | | | | | | | The __dev_t is a GNU libc internal. Use the standard dev_t instead, which is specified in POSIX. Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* replace __attribute_malloc__ with the more portable __attribute__((__malloc__))Natanael Copa2014-09-156-21/+21
| | | | | Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* configure.ac: enable GNU_SOURCE for stat64/statfs64Natanael Copa2014-09-151-0/+1
| | | | | | | | | | Use AC_USE_SYSTEM_EXTENSIONS to enable GNU_SOURCE, which is needed for: - stat64 in utils/exportfs/exportfs.c - statfs64 in utils/mountd/cache.c Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* Fix header include for definition of NULLNatanael Copa2014-09-151-0/+1
| | | | | | | NULL is defined in stdlib.h so we need to include that. Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* conffile: use standard uint*_t and unsigned charNatanael Copa2014-09-152-8/+9
| | | | | | | Use the standard integer types. This fixes compiling errors with musl libc. Signed-off-by: Natanael Copa <ncopa@alpinelinux.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* exportfs: taught exportfs -v about the nordirplus optionSteve Dickson2014-08-241-0/+2
| | | | Signed-off-by: Steve Dickson <steved@redhat.com>
* nfs-utils: Allow turning off nfsv3 readdir_plusRajesh Ghanekar2014-08-203-1/+11
| | | | | | | | | | | | | | | | | | | | | | | | | | | One of our customer's application only needs file names, not file attributes. With directories having 10K+ inodes (assuming buffer cache has directory blocks cached having file names, but inode cache is limited and hence need eviction of older cached inodes), older inodes are evicted periodically. So if they keep on doing readdir(2) from NSF client on multiple directories, some directory's files are periodically removed from inode cache and hence new readdir(2) on same directory requires disk access to bring back inodes again to inode cache. As READDIRPLUS request fetches attributes also, doing getattr on each file on server, it causes unnecessary disk accesses. If READDIRPLUS on NFS client is returned with -ENOTSUPP, NFS client uses READDIR request which just gets the names of the files in a directory, not attributes, hence avoiding disk accesses on server. There's already a corresponding client-side mount option, but an export option reduces the need for configuration across multiple clients. This flag affects NFSv3 only. If it turns out it's needed for NFSv4 as well then we may have to figure out how to extend the behavior to NFSv4, but it's not currently obvious how to do that. Signed-off-by: Rajesh Ghanekar <rajesh_ghanekar@symantec.com> Signed-off-by: Steve Dickson <steved@redhat.com>
* start-statd: clean up output when systemd is not installedMike Frysinger2014-08-191-5/+9
| | | | | | | | | | | | If you don't have systemd, then this script dumps: /usr/sbin/start-statd: line 8: systemctl: command not found This isn't terribly useful since we ultimately fall back to running the daemon ourselves, so probe for systemd's existence before we try to use it. Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Steve Dickson <steved@redhat.com>
* gssd: select non-conventional principalJurjen Bokma2014-08-181-6/+19
| | | | | | | | | | | | | | | | This enable kerberized NFS mounts to succeed even if the principal is not <HOSTNAME>$. It works by reading another principal name from the [appdefaults] section of krb5.conf: [appdefaults] nfs = { ad_principal_name = 129.125.39.115$ } Signed-off-by: Jurjen Bokma <j.bokma@rug.nl> Signed-off-by: Steve Dickson <steved@redhat.com>
* blkmapd: dump useful device information to syslogChristoph Hellwig2014-08-133-1/+11
| | | | | | | | Dump some information about used devices to syslog so that an admin can troubleshoot failing blocklayout mounts. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Steve Dickson <steved@redhat.com>
* blkmapd: fix broken multipath handlingChristoph Hellwig2014-08-131-23/+4
| | | | | | | | | | | | | | | | | We do want to use the dm-multipath device if it exists, which the code is generally prepared for, except that this check excludes them early. In addition this will also add the passive path to the device list, which is harmless if an active one exists as that or the multipath device will be preferred, and at least allows us to work if it doesn't. Also fix up the check if an path needs to be updated to remove the silly partition check - pNFS block offset are relative to the device so partion should never match it instead of the full device. On the other hand the simplistic check easily creates false positives, e.g. dm-10 is considered a partition of dm-1. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Steve Dickson <steved@redhat.com>
generated by cgit (git 2.34.1) at 2025-09-22 08:25:44 +0000