Subject: Re: Sharing MII and I2C code in ethernet drivers
To: Manuel Bouyer <bouyer@chassiron.ensta.fr>
From: John Nemeth <jnemeth@cue.bc.ca>
List: tech-kern
Date: 09/19/1997 11:46:07
On Sep 19, 11:05am, Manuel Bouyer wrote:
} On Sep 18, Chris G. Demetriou wrote
} > "Wrong answer."  You can do that by defining and using attributes; you
} > don't need device declarations (and the user config file hassles) at
} > all.
} 
} I meant: "The thunderland driver can uses 3 different PHYs drivers, but me,
} user, use a board with only one PHY, so I want to be able to compile a kernel
} with only the tl driver and the driver for the PHY I use". The best way
} to allow this is to attach the phys to the controller in the user config
} file, isn't it ?

     No.  One physical device -> one line in the config file.  Having
to put three lines in the config file for one physical board with a
single function is gross and absolutely rediculous.  Not to mentions
error prone.  We just went through this with the Hayes ESP.  I believe
the final decision is that there should be only one line.  An MII may
act like a bus at the programming level, but it is not a physical bus
(i.e. you can't attach multiple devices to it or change the device
that is attached to it).

}-- End of excerpt from Manuel Bouyer