Ethernet breakdown on a Yosemite evaluation board

[Josh Boyer]

On Tue, 19 Feb 2008 14:00:19 +0100
"Nicolas Genevrier" <Nicolas.Genevrier at jdsu.com> wrote:

> Hello all,
> 
> I noticed an Ethernet blocking on a Yosemite evaluation board when we
> are using both Ethernet devices (eth0 and eth1).
> To reproduce the issue, I just telnet the board from a PC (PC1) through
> eth0 and then telnet another PC (PC2) from the first telnet session
> through eth1. Then I display a lot of characters on the last telnet
> session:
> 	$ for i in `seq 1 100000`; do echo "$i
> aaaaaaaaaaa..(x160)..aaa"; done
> 
> After a while, transmission is blocked and leads to an rxde interruption
> from the MAL.
> 
> Yosemite evaluation kit runs Linux 2.16.13 but I reproduced the issue
> with our board running Linux-2.6.16 on a 440EP.
> In Linux 2.6.16, the new ibm_emac driver tries to recover the error but
> Ethernet remains very very slow even one day after the end of the
> transmission.
>  I didn't notice any evolution related to this neither any discussion on
> the topic. Sorry if I missed them.
> 
> I am working on the subject but an idea or related information would be
> helpful.

There was a patch from Wolfgang Ocker recently that fixed some interrupt
settings for 440EP in arch/ppc.  You might want to look at that.  I
still need to evaluate if the arch/powerpc port needs fixing.

http://patchwork.ozlabs.org/linuxppc/patch?person=1210&id=16789

josh