Re: AW: 3c59x slow; debug=5 removes the error; why?

On Tue, 18 May 2004, Christian Ostheimer wrote:

> not with a dell bios... - I checked with windows xp: the interrupts
> are shared the same way: grafics adapter (and usb and cardbus, both
> unused during tests) share irq 11 with the nic :-(

Sorry, I have no idea what to do to improve the situation. I don't 
think that it has to do with the driver at all, but with how the 
kernel calls the driver's interrupt routine.

There is an ugly workaround: increase the size of the Rx ring in the 
driver and recompile it. This might reduce the Rx overruns and so 
improve the transfer rate. However, you have to find an appropriate 
value by guessing or trial-and-error, there is no rule. You might then 
start seeing "too much work in interrupt" messages, so you might need 
to adjust max_interrupt_work accordingly.

I just saw your additional message that mentioned 2.6.6 as an 
improvement. How much improvement are we talking about ? Is there 
other difference in /proc/interrupts or the output of lspci -v between 
2.6.6 and the previous ones ?

-- 
Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: [email protected]


_______________________________________________
vortex mailing list
[email protected]
http://www.scyld.com/mailman/listinfo/vortex