tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Call for testing "Attansic/Atheros L1 gigabit ethernet"

On Tue, Feb 24, 2009 at 01:12:17PM +0100, Christoph Egger wrote:

> Further, Jochen Kunz reports me network problems with age(4),
> which he can't reproduce with fxp(4).
> When he uses age(4) with nfs over udp, then processes accessing
> nfs hang in state 'D'. With nfs over tcp, transfer is very slow.
> In both cases, he gets 'watchdog timeout's.
Just to add: If I cause lots of network trafic with nc(1) (netcat),
without any NFS involved, I get the same results: age(4) watchdog
timeouts and massive packet loss while ping(8)ing.
The difference NFS TCP vs. UDB may be related to the biger NFS
read- / write-sizes in UDP mode, causing endless reties in UDP
mode, where TCP can recover earlier by narrowing the window.

> I have no idea why and I also don't know if the bug is in nfs
> or in age(4).
Well. If it works fine with fxp(4) at full load...

Today my /home disk died a suden death. At the momet I am restoring
my /home from an old DLT backup. Anything since Jan. 6. is lost... :-(
Looks like I will be bussy with data recovery and not able to do 
any testing the next days. Please resend any Emails if you wane
get an answer...



Home | Main Index | Thread Index | Old Index