linux-origin
[Top] [All Lists]

Re: ethernet

To: Leo Dagum <dagum@xxxxxxxxxxxxxxxxxxx>
Subject: Re: ethernet
From: Ralf Baechle <ralf@xxxxxxxxxxx>
Date: Fri, 11 Aug 2000 05:40:34 +0200
Cc: Keith Owens <kaos@xxxxxxxxxx>, linux-origin@xxxxxxxxxxx, Kanoj Sarcar <kanoj@xxxxxxxxxxxxxxxxxxx>
In-reply-to: <200008110238.TAA20056@barrel.engr.sgi.com>; from dagum@barrel.engr.sgi.com on Thu, Aug 10, 2000 at 07:38:47PM -0700
References: <200008101801.LAA50412@google.engr.sgi.com> <11989.965948096@ocs3.ocs-net> <200008110238.TAA20056@barrel.engr.sgi.com>
Sender: owner-linux-origin@xxxxxxxxxxx
On Thu, Aug 10, 2000 at 07:38:47PM -0700, Leo Dagum wrote:

> > > Doing the traffic in reverse automatically fills in the hardware
> > > address of the machine that the arp request came from.  This tends to
> > > hide the RX overrun problem.
> > 
> > And indeed we tend to have RX overrun problem with this driver.  I can
> > easily blow my Origin off the net by just floodpinging it for a few
> > seconds.  In such a case the IOC3 will deactivate RX and signal an
> > interrupt.  The driver handles this by trying to reinitialize everything
> > but somehow this often doesn't work.  
> 
> When the driver reinitializes everything, that includes the dma buffers.
> So if, e.g. the driver is also xmit'ing a packet when it gets the
> error, the xmit will fail and panic the system because the dma
> buffer disappears between getting set up and getting xmitted.
> 
> Not sure if that's the problem you see, but I've certainly seen
> that happen.

That's definately not what's going on in this case.

  Ralf

<Prev in Thread] Current Thread [Next in Thread>