[BUG] server-state-recovery branch fails to mount on x86_64

[J. Bruce Fields]

On Fri, Oct 20, 2006 at 12:54:28PM -0700, Bryce Harrington wrote:
> We're seeing a new failure in recent git branches, that results in our
> 64-bit xeons being unable to mount with nfsv4.  Mounting with nfsv3
> seems to work okay, and nfsv4 on our 32-bit x86 boxes are okay.
> 
> Here is one of the runs exhibiting this problem:
> 
>    http://crucible.osdl.org/runs/2543/
> 
> You can see in the nfs13.log file, when it starts running newpynfs it
> gets the error "mount: nfs12:/: can't read superblock".
> 
> Looking at other runs, I suspect this problem may also be present on the
> server-cluster-locking-api and now also on the latest nfs-client-stable
> branch (linux-2.6.19-rc2-ebfa15f-nfs-client-stable) from 10/19/2006.

If that's the case then it's probably also true in 2.6.19-rc2.

Would it be possible to get a trace of the network traffic between
client and server during the time of the failed mount?

So that'd mean just running

	tcpdump -wTMP -s0

then trying the mount, then killing the tcpdump, then sending me TMP.

--b.