Subject: Re: kern/11923: new NFS client hangs with "ne0: warning - receiver ring buffer overrun"
To: ITOH Yasufumi <itohy@netbsd.org>
From: Manuel Bouyer <bouyer@antioche.lip6.fr>
List: netbsd-bugs
Date: 01/10/2001 19:42:40
On Wed, Jan 10, 2001 at 08:05:06AM +0900, ITOH Yasufumi wrote:
>
> >Number: 11923
> >Category: kern
> >Synopsis: new NFS client hangs with "ne0: warning - receiver ring buffer overrun"
> >Confidential: no
> >Severity: serious
> >Priority: high
> >Responsible: kern-bug-people
> >State: open
> >Class: sw-bug
> >Submitter-Id: net
> >Arrival-Date: Tue Jan 09 15:06:00 PST 2001
> >Closed-Date:
> >Last-Modified:
> >Originator: ITOH Yasufumi
> >Release: 1.5Q (Jan. 9, 2001)
> >Organization:
> >Environment:
> System: NetBSD acha.my.domain 1.5Q NetBSD 1.5Q (ACHA) #69: Tue Jan 9 17:33:08 JST 2001 itohy@pino.my.domain:/w/src/sys/arch/x68k/compile/ACHA x68k
> Architecture: m68k
> Machine: x68k
> machine config: X68030 (MC68030 + MC68882 25MHz, Main memory 12MB)
> with ``Neptune-X'' Ethernet board (ne0)
> object format: a.out
>
> >Description:
> NFS is totally unusable.
>
> Recent NFS code (client side) causes
> "ne0: warning - receiver ring buffer overrun"
> and hangs the process.
>
> NFS client status
> ---------- ------
> 1.4P OK (do not cause overrun or hang)
> 1.4Y OK
> 1.5A OK
> 1.5D OK
> 1.5E OK August 11, 2000
> 1.5G NG (overrun and hang) October 5, 2000
> 1.5H NG
> 1.5P NG
>
> The port-dependent code is not so much changed, and I think some
> changes of the mi NFS code cause this trouble.
Did you check a, say, 1.5E client with the 1.5Q server ?
It's possible that some optimisation in NFS made the server send data
too quickly for your client, forcing the client to drop fragments.
Did you try to reduce the read and/or write size ?
--
Manuel Bouyer <bouyer@antioche.eu.org>
--