https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c2 --- Comment #2 from Neil Brown 2010-10-26 20:27:07 UTC --- It might help to see /etc/exports from the server and /etc/fstab from the client - just to make the picture more concrete. If there are any messages from 'mountd' in /var/log/message on the server, they would probably be very helpful. But most helpful would be a tcpdump trace. So a client run tcpdump -w /tmp/trace -s 0 host SERVERNAME then mount the nfs filesystems and try to access them. Then kill the 'tcpdump' and attach /tmp/trace to this bug. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c4 Michael Monnerie changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW CC| |novell-web@zmi.at InfoProvider|novell-web@zmi.at | --- Comment #4 from Michael Monnerie 2010-10-26 23:04:53 UTC --- Leonardo, I can confirm again that a downgrade solves the problem, an upgrade brings it back. Neil, see attached my fstab from client and exports from server, both shortened to only show NFS stuff and contain no comments. In fstab I left two commented out lines which I keep in case NFS4 makes problems so I can mount in v3. But with the update on the server, it mounts and looses the mount after some minutes ("stale NFS handle"). # grep mountd /var/log/messages: Oct 27 00:40:39 orion mountd[3465]: authenticated mount request from 10.72.27.2:840 for / (/) Oct 27 00:40:46 orion mountd[3465]: authenticated unmount request from 10.72.27.2:805 for / (/) Oct 27 00:41:56 orion mountd[3465]: authenticated mount request from 10.72.27.2:900 for / (/) Oct 27 00:44:22 orion mountd[3465]: Caught signal 15, un-registering and exiting. Oct 27 00:44:29 orion mountd[17574]: Caught signal 15, un-registering and exiting. Oct 27 00:44:31 orion mountd[17942]: authenticated unmount request from 10.72.27.2:686 for / (/) Oct 27 00:44:33 orion mountd[17942]: /disks/nfsserver/q and /disks/daten/samba have same filehandle for *,10.0.0.0/8,10.72.27.2, using first Oct 27 00:44:33 orion mountd[17942]: /disks/nfsserver/z and /disks/daten/samba2/firma-zmi have same filehandle for *,10.0.0.0/8,10.72.27.2, using first Oct 27 00:44:33 orion mountd[17942]: authenticated mount request from 10.72.27.2:679 for / (/) Also attached the tcpdump trace from the client machine. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c6 --- Comment #6 from Michael Monnerie 2010-10-26 23:06:13 UTC --- Created an attachment (id=397196) --> (http://bugzilla.novell.com/attachment.cgi?id=397196) /etc/exports from server -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c8 Michael Monnerie changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW InfoProvider|novell-web@zmi.at | --- Comment #8 from Michael Monnerie 2010-10-27 01:33:56 UTC --- Sorry for the confusion, I've deleted that line too: /disks/daten *(rw,no_root_squash,async,nohide,no_subtree_check) but as you said: not important Yes, /q /z /s are symlinks. ls -l /q shows an empty dir. I see just the "." and ".." entries. Same for the others: empty dirs, no contents. Very simple to test: I just downgrade to nfs-kernel-server-1.2.1-2.6.6 and it works. Then upgrade to nfs-kernel-server-1.2.1-2.9.2 and it doesn't work anymore. Just have to call "rcnfsserver restart" on the server and "rcnfs restart" on the client. I've done a new dump. The first with the new, non-working 2.9.2 nfsserver. I did rcnfs restart l /q l /z and then again with the working nfs 2.6.6 version rcnfs restart l /q as this works already, I stopped here. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c10 --- Comment #10 from Michael Monnerie 2010-10-27 01:36:25 UTC --- Created an attachment (id=397205) --> (http://bugzilla.novell.com/attachment.cgi?id=397205) Trace from mount and ls with working nfsserver 2.6.6 Dump done from client -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c12 Neil Brown changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO InfoProvider| |novell-web@zmi.at --- Comment #12 from Neil Brown 2010-10-27 03:20:50 UTC --- Could I also get the contents of /var/lib/nfs/etab both when it is working and when it isn't? Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c14 --- Comment #14 from Neil Brown 2010-10-27 04:02:28 UTC --- Created an attachment (id=397208) --> (http://bugzilla.novell.com/attachment.cgi?id=397208) unofficial rpm for testing This is a fixed rpm for testing. It only needs to be installed on the server. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=648923 https://bugzilla.novell.com/show_bug.cgi?id=648923#c16 Michael Monnerie changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW InfoProvider|novell-web@zmi.at | --- Comment #16 from Michael Monnerie 2010-10-27 08:57:17 UTC --- Neil, of course /disks/nfsserver/q on the server is a bind-mount as it is suggested in docs, on the server: # grep samba /etc/fstab /disks/daten/samba /disks/nfsserver/q none bind 0 0 as well as all other subdirs of /disks/nfsserver are. ******** I can confirm than your fixed 2.9.3 nfsserver rpm works for me. Thank you very much for the premium quality support. ******** I did not try the client rpm, as my testing client is a openSUSE 11.3. I whish there would be a bugzilla for SLES as well. If I'd only have had this problem with SLES clients, I wouldn't know where to report it. BTW: on that server, I use XFS with inode64 mount option. It doesn't work when one of the "real dirs" has a 64bit inode, in above case /disks/daten/samba must have an inode within the 32bit range, otherwise it won't work. Is this a bug which I should report, or a limitation of NFS? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.