On Sat, May 23, 2009 at 07:12:55PM -0400, Jon Smirl wrote:
[Jon, could you please fix the word wrapping in your MUA, there's none at all which makes it more difficult than it needs to be to respond to your e-mail.]
Mark is not enthused about soc-of-simple.c so rather than extend it for AC97 I altered the drivers to not use it. Instead they use the old way of manually binding everything. Mark would like to see OF binding more closely integrated to the core. Once a proper solution for OF
I really don't feel that this reflects the issues I had with the code well. There are two basic issues here:
- You had been trying to represent the AC97 CODEC drivers as platform devices which is not a good approach, largely since AC97 is a real bus and should be represented as such.
- soc-of-simple really should be merged into the core now, there's no need to have to add changes to individual drivers any more since drivers are already regstering with the core.
These two points are basically unrelated except in that your motiviation for doing the platform device thing was trying to fit your machine drivers into the existing soc-of-simple framework. It's the changes to the cross-platform AC97 drivers that were causing real problems, soc-of-simple is only an issue in that using it motivates these other changes.
binding is agreed upon it is easy to convert the existing drivers. A first step would be converting the existing codec drivers so that they can be dynamically loaded.
I'm not aware of any CODEC drivers which can't currently be built and used as modules. If you mean "load via the normal device model" then yes, that'd be very good (and is in progress) but it's another issue and as I explained last time AC97 poses particular problems there.