[U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs

[Ben Warren]

Hi Prafulla,

Prafulla Wadaskar wrote:
> Hi Ben,
>
> How are you?
> Seems to be very busy as usual.... :-)
>
> Can you pls kindly take out some time to review these patches.
> hopefully they are as per your requirements since we have invested considerable time to make them up to the mark.
> May be-you can accept them :-) 
> http://lists.denx.de/pipermail/u-boot/2009-May/053111.html
>   
We don't currently have a proper PHY framework in place, although over 
the years there's been much talk in this area.  For now I'll pass on 
this one.
> http://lists.denx.de/pipermail/u-boot/2009-May/052905.html
>   
I've pulled this into net/next.  Thanks for spinning it so many times.
> Just to share with you,
> in netdev.h- though struct and enums defined looks similar-
> I have used separate struct and enums for both mv88e61xx and mv88e1116 because only basic init support is available in those drives, but in coming time there will be more features added to them.
> I think since these are initial drivers in net/phy/ we should provide common/phy.h and keep netdev.h untouched
> What do you think?
>
>   
As I mentioned above, the existing PHY infrastructure is weak.  The 
switch driver is fine here, but I don't want to add any regular PHYs 
without a properly designed interface.  A lot of work has been done in 
this area but hasn't been published yet.  Stay tuned...
> Also you might be ready with review comments for this
> http://lists.denx.de/pipermail/u-boot/2009-May/053113.html
> Nothing much changed in v4 compared to patch version v1.
> few global comments that I received for this patch from you are-
> 1. the code is too big.
> 2. can you strip down features to minimum
> Just to reply these: 
> the entire egiga driver is kept as separate driver .c and .h,
> I am in progress to reuse the code from other drivers.
> The features are kept minimal required but I can cut down some of features like "update network statistics" so as to make code size small and acceptable :-)
> Still your comments will help me to provide better code...
>
>   
The driver is really big, which isn't necessarily a show stopper.  I'd 
like to see you #ifdef CONFIG_xxx out the non-essential stuff 
(statistics etc.) so that users aren't forced to compile it in.  Also, 
please put the changes to cpu.c into a different patch.
> Last but not least
> Shall I resend above patches if you feel this is more comfortable for you?
>  
> Regards..
> Prafulla. . .  
>
>   
Thanks for being so flexible.

regards,
Ben