On Thursday 01 October 2009 14:09:19 ext Mark Brown wrote:
On Thu, Oct 01, 2009 at 02:02:04PM +0300, Peter Ujfalusi wrote:
On Thursday 01 October 2009 13:51:07 ext Mark Brown wrote:
I'd rather use dev_name() for the CODEC itself if possible, that is more likely to be stable going forward and one of the immediate aims with the API refactoring is to remove socdev entirely at runtime.
At this point the codec->dev was NULL, that is why I have used the socdev->dev instead.
Right, your CODEC driver needs updating to current APIs. Omitting the device name is a good fallback here since you won't be able to have more than one of the same device anyway if a device isn't provided.
One of the codec to be blamed is the twl4030 codec, but since that does not use any _device/_driver method to load, there is no dev to be assigned to codec-
dev...
I don't know if it would make sense to use platform_driver_register (and associates) with it in the future.
The other codec, which I'm preparing for sending is using i2c probing, so I have fixed that, so now it has codec->dev initialized correctly.
your CODEC driver needs updating to current APIs
Hmm, there are quite a bit of inconsistency among the codec drivers, so I'm not sure what is the correct way of doing this. I will use the wm8993 driver for further fine tuning, since that seams to be rather new in the tree.
I'll send a patch later to fix this implementation: if the codec->dev is valid, than it is going to use dev_name(codec->dev), if it is NULL, than it will use either the codec->name or provide the same string as the current implementation does ( {dev_name(socdev->dev)}-{codec->name} ), whichever is the preferred.
Thanks, Péter