use of phy_connect() in drivers/net/gianfar.c ?

[Andy Fleming]

On Oct 30, 2005, at 09:18, Wolfgang Denk wrote:

> Dear Matthew,
>
> in message <1130632707.7470.16.camel at localhost.localdomain> you wrote:
> >
> > As far as I understand there are patches in the -mm tree with  
> contain
> > the updates for the Gianfar driver. The issue is due to a
>
> Do you happen to know of a specific (set of) patch(es)? URL?
>
> > synchronization issue with the new PHY abstraction layer that was  
> added.
>
> Yes, obviously. But I am still so naïve to expect that  such  changes
> would not make it untested (the code is not even compile clean) to an
> official kernel release like 2.6.14 ?

Sorry about that.  The problem was that the offending changes needed  
to be broken into two patches.  The PHY Layer (a separate patch)  
required changes to the driver, which required concurrent changes in  
the platform init code (the MDIO bus was set up as its own device).   
Unfortunately, the driver changes needed to be handled through  
netdev, while the platform changes were through -mm.  We tried to  
sync it up so that the changes got into Linus' tree at the same time,  
but that did not happen.  If you pull the tree today, you will find  
that patch is now in the main tree.