lkml.org 
[lkml]   [2003]   [Mar]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: 2.5.65-mm1: eth0: Transmit error, Tx status register 90
----- Original Message -----  
From: Andrew Morton <akpm@digeo.com>
Date: Wed, 19 Mar 2003 03:32:35 -0800
To: "Felipe Alfaro Solana" <felipe_alfaro@linuxmail.org>
Subject: Re: 2.5.65-mm1: eth0: Transmit error, Tx status register 90

> "Felipe Alfaro Solana" <felipe_alfaro@linuxmail.org> wrote:
> > So it seems a problem with the network transport.
>
> Ah. You're using cardbus. Please send lspci output.

I have attached lspci -vvvxxx from 2.5.65-mm2. Maybe
you can see something strange here (cause I must
admit I don't understand a thing of it).

> The NIC driver is basically unchanged from 2.4. It will be a
> cardbus/PCI/northbrige/southbridge setup problem.

Well, I've investigated a little more and this is what I found
until date:

1. The benchmarks I did with 2.4.20 were *not* using
kernel built-in PCMCIA support, but SF's "pcmcia-cs".
This means that the 3Com driver is different (and I also
suspect that yenta socket driver is also different): when
using 2.4.20's built-in PCMCIA support, my 3CCFE575CT
uses 3c59x kernel module, but "pcmcia-cs" uses a
different driver: 3c575_cb.

2. I have tried doing benchmarks with RedHat's own
kernel (based on 2.4.20 with many Alan Cox patches),
using built-in PCMCIA and 3c59x driver and it *does*
have the same problems that I'm seeing with 2.5, that
is, it takes 4 times more time to send than to receive.

3. I've been unable to use "lspci" on my 2.4.20 kernel
with SourceForge's pcmcia-cs. It complains that
"/proc/bus/pci/06" does not exist. For some reason,
when using "pcmcia-cs", CardBus is not mapped
as a PCI bus.

So, it seems it's not a 2.5 specific problem, as I can
reproduce it with a 2.4 kernel. It seems more a problem
with the CardBus/PCMCIA and 3Com modules.

> Look for suspicious dmesg output. Compare the
>`lspci -vvxx' output for 2.5 and 2.4, see if any of the
> registers look different. This will be hard.

Well, I haven't done this comparison as the problem
is reproducible either with 2.5 or 2.4 kernels. I do only
get full bandwidth when using SourceForge's
pcmcia-cs proyect.

I have to do some more tests, but meanwhile, do you
have any ideas? What's next?

Thank you very much!

Felipe

--
______________________________________________
http://www.linuxmail.org/
Now with e-mail forwarding for only US$5.95/yr

Powered by Outblaze
[unhandled content-type:application/octet-stream]
\
 
 \ /
  Last update: 2005-03-22 13:34    [W:0.032 / U:0.892 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site