On 03/19/2014 09:31 PM, Lars-Peter Clausen wrote:
On 03/19/2014 08:21 PM, Mark Brown wrote:
On Wed, Mar 19, 2014 at 09:15:14PM +0200, Jyri Sarha wrote:
On 03/19/2014 08:51 PM, Lars-Peter Clausen wrote:
When does this make sense? Either the bitclock is inverted for all of them or for none.
Definition of clock signal and it's inversion varies between chip manufacturers and sometimes it may not be possible to get all the dai drivers to work identically in this respect. Because of this in some cases there may be a need to set the inversion bit only at one end of the link.
No, Linux has a definition of all the clock modes which applies to all devices regardless of what the manufacturer documents in their datasheet.
Yep. The clock properties are well defined for the different modes that can be specified in the format property. It's up to the driver to translate this to driver specific settings. If two drivers behave differently for the same mode one of them (or both) are broken.
After a little thinking it is clear to me too that only reason to have this overwrite capability is a badly written dai driver. Even with buggy HW it should always be possible to present the working modes in daifmt terms. Ok, let's remove it.
Here is an updated DT example with updated phandle for master settings:
sound { compatible = "simple-audio-card"; simple-audio-card,name = "Simple Audio"; simple-audio-card,widgets = ... simple-audio-card,routing = ...
simple-audio-card,dai-link@0 { /* I2S - codec */ format = "i2s"; bitclock-master = <&dai_link_master> frame-master = <&dai_link_master>; bitclock-inversion; simple-audio-card,cpu { sound-dai = <&audio1 0>; }; dai_link_master: simple-audio-card,codec { sound-dai = <&codec 0>; system-clock-frequency = <12000000>; }; }; ...
We should probably add the definitions for the different formats to the DT bindings. E.g. what is default, what is inverted polarity, etc.
That is a good idea.
Best regards, Jyri