Re: Neptune regression...

[Matheos Worku <Matheos.Worku@Sun.COM>]

David Miller wrote:
> Matheos, this patch of your's:
>
> commit 7f7c4072ea552f97a0898331322f71986a97299c
> Author: Matheos Worku <matheos.worku@sun.com>
> Date:   Thu Apr 24 21:02:37 2008 -0700
>
>     niu: Determine the # of ports from the card's VPD data
>     
>     [ Fix minor whitespace and coding style stuff... -DaveM ]
>     
>     Signed-off-by: Matheos Worku <matheos.worku@sun.com>
>     Signed-off-by: David S. Miller <davem@davemloft.net>
>
> Breaks on-board Neptune chips on Maramba systems.
>
> Previously the code in niu_get_and_validate_port() used a default of
> "4" ports if the various probing methods failed.  Your code removed
> that fallback.
>
> This case seems to be what triggers for on-board Maramba chips.
>
> So, for now, I'm going to put the:
>
> 	if (!parent->num_ports)
> 		parent->num_ports = 4;
>
> back in there to fix the regression.  I've validated that this fixes
> the problem.
>
> Is there some other mechanism we should be using here to detect the
> port count in the Maramba on-board case?
>   
Dave,

Sorry I broke the driver.
I think the clean way to handle this would be to use the "board" 
property which exists for all Neptune variants, both in OBP properties  
as well as PCI VPD. I will work on the fix based on that approach  and 
will generate a patch as soon possible.
Regards
Matheos