[541] in SIPB_Linux_Development
[becker@super.org (Donald J. Becker): Re: Linux NFS slow?]
daemon@ATHENA.MIT.EDU (Erik Nygren)
Mon Apr 4 21:41:22 1994
To: linux-dev@MIT.EDU
Date: Mon, 04 Apr 94 21:41:04 EDT
From: Erik Nygren <nygren@MIT.EDU>
Just a reminder to people using keesh's exported source on non-Linux machines.
--- Erik
------- Forwarded Message
Replied: Mon, 04 Apr 94 21:39:51 EDT
Replied: "becker@super.org (Donald J. Becker) "
Received: from MIT.MIT.EDU by po7.MIT.EDU (5.61/4.7) id AA15255; Mon, 4 Apr 94 20:29:53 EDT
Received: from super.super.org by MIT.EDU with SMTP
id AA03714; Mon, 4 Apr 94 20:29:50 EDT
Received: from descartes.super.org by super.super.org (4.1/SMI-4.1)
id AA01537; Mon, 4 Apr 94 20:29:44 EDT
Date: Mon, 4 Apr 94 20:29:44 EDT
From: becker@super.org (Donald J. Becker)
Message-Id: <9404050029.AA01537@super.super.org>
To: nygren@MIT.EDU
Subject: Re: Linux NFS slow?
>For me, it is the other way around. I don't really have a problem with NFS
>client (being served by Ultrix Dec 5000's and BSD Vaxen), but I find the
>Linux NFS server to be very unstable when serving to non-Linux clients.
You must set the client's read size and write size (rsize and wsize) to
1024. This is usually done as an option to mount.
>I've seen this problem on two separate machines: mine (foundation.mit.edu)
>and the SIPB office Linux machine (keesh.mit.edu).
Did the SIPB office ever get another '486?
>Mar 6 06:14:38 foundation kernel: kmalloc: I refuse to allocate 8504 bytes (for
> now max = 4096).
These are messages from the networking layer, before the packets get to the
nfsd. Some client is trying to do 8K reads or writes.
Donald Becker becker@super.org
IDA Supercomputing Research Center
17100 Science Drive, Bowie MD 20715 301-805-7482
------- End of Forwarded Message
