Debugging a hung connection

[Dave Quigley]

On Wed, 2008-05-21 at 09:59 -0400, Norman Elton wrote:
> I've got an NFSv4 writable mount that seems to be hung. I can do an
> "ls", but neither read (cat) nor write (echo) operations succeed. Both
> hang until I kill the process. Other clients are able to continue to
> read/write without a problem. I've got a packet capture, and there
> don't seem be any packets getting dropped by firewall, etc.
> 
> I'm not entirely sure how to pin down the problem. I've got PCAPs of a
> working and nonworking client. I see NFS4 calls going back and forth,
> but am not quite sure what I'm looking at. Is there a good resource to
> explain what the different operations mean, and how/where they would
> be failing?

In terms of linux implementation specifics I haven't found much other
than the code as a reference. The Umich guys have a small page on kernel
hacking/debugging NFSv4 which can be found at [1]. I have found that
turning on the extra debug statements really helps in figure out what is
going wrong. If you want to understand what each of the NFSv4 operations
does the best reference for that is the NFSv4.1 specification [2].

[1] http://www.citi.umich.edu/projects/nfsv4/linux/nfsv4-hacking.html
[2] http://www.nfsv4-editor.org/draft23/draft-ietf-nfsv4-minorversion1-23.txt

Dave