On 03/04/2014 10:27 AM, Mark Brown wrote:
On Tue, Mar 04, 2014 at 10:20:16AM +0100, Lars-Peter Clausen wrote:
codec->write = hw_write;
- codec->read = hw_read;
But this part will break snd_soc_read() and with it basically every driver that has any controls or DAPM widgets.
Bah, so it does.
I do have a set of patches that moves IO to the component level that will clean this all up quite a bit. I'm just waiting for Xiubo Li snd_soc_codec_set_cache_io() cleanup to be merged first, since it will make things a bit simpler.
That seems to be going really slowly...
Yea, but we are not in a hurry. My plan was to get all the small independent driver cleanups necessary for componetization merged for 3.15 and then for 3.16 do the componetnization in the ASoC core.
in any case the real fix here is to get all the users converted over to regmap. We could also just do something similar to what we've done with update_bits() (where we did it for locking) which is going to be much quicker.
So what the patchset does is add a regmap pointer to the component struct. For now it also adds snd_soc_component_{read,write,update_bits, update_bits_async} wrapper functions which have the same signature as the matching regmap function. Internally it does check if the regmap field of the component is set, if yes, just forwards the call to the matching regmap function. If not it falls back to the read, write callbacks setup by CODEC or platform drivers. Once every driver has been converted to regmap we can remove the fallback and make the function static inline.
- Lars