[3106] in linux-net channel archive
Re: whats rsize supposed to be now?
daemon@ATHENA.MIT.EDU (Jon Peatfield)
Tue Jun 4 16:42:33 1996
To: "Marty Leisner" <leisner@sdsp.mc.xerox.com>
cc: jp107@damtp.cam.ac.uk, linux-net@vger.rutgers.edu
In-reply-to: Your message of "Mon, 03 Jun 1996 15:26:37 PDT."
<9606032226.AA06894@gnu.mc.xerox.com>
Date: Tue, 04 Jun 1996 13:30:40 +0100
From: Jon Peatfield <J.S.Peatfield@damtp.cam.ac.uk>
Since about 1.1.30 8K NFS has worked fine. Before that there were problems
with memory allocation in the nfs client side which could cause problems with
over 1K NFS, and there was a bug in the packet retry until about 1.1.45 such
that the NFS could get hung trying to retry a request (but the packet it was
sending out was corrupted).
Without 8K nfs we would probably have had to move to a BSD system instead of
Linux by now...
I've not looked at the NFS code for a very long time, but I gather that the
new kernels (1.3.80+ ?) have the NFS code using the normal buffercache so the
performance is even better (especially when running executables over NFS).
Still no biod/nfsiod like stuff, but now that the memory model is tidy it
ought to be a bit easier than before. I'd like to do this myself but as
always don't have the time.
This reminds me I must check if my mount and amd patches ever made it into the
standard sources, (I know that the kernel patch is in 1.3.x).
-- Jon
Jon Peatfield, Computer Officer, the DAMTP, University of Cambridge
Telephone: +44 1223 3 37852 Mail: J.S.Peatfield@damtp.cam.ac.uk
