Hi,
I recently installed SuSE 10.0 on a machine we have and I notice
that NFS locking seems to be broken. Whenever I try to set a lock
with fcntl F_SETLK on an NFS mounted filesystem served by a SuSE
10.0 machine I get a "Stale NFS file handle" error. This happens
from a variety of NFS clients IRIX, Solaris, SuSE 9.x and MAC OSX.
If I turn on debugging by doing "echo 65535 > /proc/sys/sunrpc/nlm_debug"
I get the following messages in dmesg/syslog:
lockd: request from 905c6e82
lockd: LOCK called
lockd: nlm_lookup_host(905c6e82, p=17, v=4, my role=server, name=uwast)
lockd: host garbage collection
lockd: nlmsvc_mark_resources
nlm_gc_hosts skipping uwast (cnt 2 use 0 exp 42936407)
lockd: nsm_monitor(uwast)
lockd: nlm_file_lookup(02000001 05000800 00000080 0d5287c8 00000001 0d5287c7)
lockd: creating file for (02000001 05000800 00000080 0d5287c8 00000001 0d5287c7)
lockd: open failed (error 1)
lockd: release host uwast, refcount is now 0
nlmsvc_retry_blocked(00000000, when=0)
nlmsvc_retry_blocked(00000000, when=0)
lockd: nlm_lookup_host(905c6e82, p=6, v=4, my role=client, name=uwast)
Anyone have an idea on how to fix this problem?
Here's a simple test program that shows this problem:
#include
On Tuesday 01 November 2005 16:59, Stephan L Jansen wrote:
I recently installed SuSE 10.0 on a machine we have and I notice that NFS locking seems to be broken.
They are experimenting with NFSv4 code? http://lists.opensuse.org/archive/opensuse/2005-Nov/0051.html -- // Janne
On Wed, 2005-11-02 at 07:38 +0200, Janne Karhunen wrote:
On Tuesday 01 November 2005 16:59, Stephan L Jansen wrote:
I recently installed SuSE 10.0 on a machine we have and I notice that NFS locking seems to be broken.
Did you see this problem in earlier 2.6 kernels? In 2.6, the nfs code went into the kernel and is no longer a daemon. I also see that the init scripts for NFS have changed across the past few SUSE releases. As I see it, the NFS startup scripts have been: SUSE 9.0: nfslock nfsserver SUSE 9.2 and later: nfsboot nfsserver So, in 9.2, the nfslock init stuff was moved into a different startup script (presumably). Maybe there are also new variables/files controlling locking. I would check these two startup scripts. -- Roger
participants (3)
-
Janne Karhunen
-
Roger Oberholtzer
-
Stephan L Jansen