2008-05-27 20:56:45

by Frank Filz

[permalink] [raw]
Subject: [NFS] Corrupted stateid

I'm helping debug a problem where a Linux client (2.6.18 kernel) is
talking to an AIX server. We see an issue where Linux is happily using
the stateid issued by the server (other part only, does not include the
sequence number):


and then occasionally uses a different stateid:


This stateid is the correct stateid + 20000, however, AIX accepted it
(very odd).

Interestingly, this alternate stateid is used in a close, which the AIX
server also accepts, returning the following stateid:


which is the original stateid +20000 - 8000

It appears that somewhere the code is overlaying some flag bits onto the

The weird part is that way before this, AIX suddenly complains about the
original stateid with
NFS4ERR_BAD_STATEID. The sequence number has not changed at all during
this period.

Apparently another trace does show the Linux client making this same
corruption of the stateid, but in that case, the AIX server DOES

Any thoughts what might be going on here from a Linux perspective?



This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
NFS maillist - [email protected]
Please note that [email protected] is being discontinued.
Please subscribe to [email protected] instead.