Re: nfs + post-2.1.36 kernels == N.G.

linux@oot.co.uk
Fri, 30 May 1997 09:02:05 EDT

>From: "Steven N. Hirsch" <shirsch@ibm.net>
>
>I can provide another data point regarding the (mis)behavior of NFS when
>the server is running 2.1.37+. Short transactions seem to work, but
>any continuous write > 1M (or thereabouts) causes a complete failure of
>networking on the server box. The only cure seems to be rebooting.
>
>No log messages. No screen messages.

I seem to get something similar; With Linux as a client I get complete
lockups under NFS for 2.1.39+ (and probably earlier) whenever medium-large
reads occur (about 20k+).
With tcpdump/snoop I can see the first three blocks of 4096 bytes
transferred, and then the machine hangs solid.

BTW the server is Solaris 2.4

Does the new nfs server code share anything with the client stuff?
I've not had a chance to dive into the sources to sort this one out...

Adrian.

---------------------------------
Adrian St.John (adrian@oot.co.uk)
Object-Oriented Technologies Ltd.