Re: Driver for via-velocity gigabit NIC

Timothy Legge <[email protected]> Tue, 07 Mar 2006 07:01:21 -0400
Newsgroups gmane.network.etherboot.devel
Message-ID <[email protected]>
Marty Connor wrote:
> On Mar 6, 2006, at 8:29 PM, Timothy Legge wrote:
> 
>> Hi
>> I have a working version of a driver for the via-velocity 6120  based 
>> cards.
> 
> 
> Nice.  Is this a chip that is on standalone PCI cards or something  that 
> is more on motherboards only?

Yes, they are available on Ebay for next to nothing.

>> The only problem is that it is only working on 5.2.  With 5.4, it  
>> causes a computer reset on a dual opteron machine and causes an old  
>> Pentium to freeze.
>> I am currently at a loss as to why.  I imagine it is the result of  
>> some  bug or another but I have never seen a driver work on one  
>> version and fail so badly on another.
>> The bug manifests itself on the first or second go through the  
>> transmit function and it seems to fail part way though.
>> Does anyone have any quick thought on obvious things to review?
> 
> 
> Well, the two things that come to mind are interrupts and the  
> transmit/receive buffer structure.

I had not considered Interupts.  As for the buffer structure it is 4 
transmit and 8 receive buffers with the required 64 byte alignment.  I 
did do the alignment differntly this time but it is very similar to the 
via-rhine method.

> How many buffers have you allocated, and are they in a ring structure  
> with the usual polling?

Standar ring with the ususal polling.  The receive buffers can only be 
returned in groups of 4 but that seems to be fine.

> It sounds like your code might be getting walked on, possibly by a  
> stray pointer.  It's really easy to forget an '&', and since the  memory 
> management is very different between the versions, it's very  easy to 
> see how the problem might manifest itself differently between  
> versions.  The 5.4 build systems is pretty different in the way it  uses 
> ld to make the image.
> I suspect that printing out the transmit and receive buffer pointer  
> addresses may be helpful.

Good idea, I will give it a go.

> Can you drop the driver somewhere where we can take a look at it?

I will spend a little more time cleaning it up and I will attach it to 
the patch list.

Tim


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642