Re: Re: Re: Re: Re: Re: Re: Question: Rx Ring Buffers in the rtl8139 module

Donald Becker <[email protected]> schrieb am 30.08.02 21:42:53:
> On Fri, 30 Aug 2002 [email protected] wrote:
> 
> > > Perhaps your laptop is slowing down the PCI bus to save power, based on
> > > a lack of keyboard activity.  Or perhaps the PCI parameters are set so
> > > that PCI bursts do not occur.
> > > 
> > > If the latter, you are in one of the few situation where the PCI bus
> > > settings will have to be tuned.
> > 
> > are you talking of BIOS parameter? There is _very_ little to tune.
> > 
> > this is a linux 2.4 system. Is there anything how I can tune the PCI
> > bus settings?
> 
> 30 seconds with google...

ok, it took my a little longer. Among many others, there is a tool
called powertwek (http://powertweak.sourceforge.net/overview.html). 
With that I found out, that the PCI bandwith was At the bridge (not at
the NIC).

The VT8363/8365 Bridge was blocking/slowing down the bus. 
After allowing this device to do fast back-to-back writes it works
_much_ better.

On my setup, I need now to do this:
setpci -s 0:0.0 70=de
setpci -s 0:0.0 71=8c
at boot time of the laptop. Your milage may vary. This makes only 
sense on my Bridge, others WILL vary. You have been warned :-)

If you encounter a similar problem with the PCI bandwith, then
like powertweak (which is collected know how about some deviced with a gui)
might help to find the bottleneck.

Until now I found no negatves of this change - I will post my experience
in a few days.
 
> 
> (This is going beyond rtl819 driver issues.)
> 

right. Thanks anyway

    Joerg