Hi,
The commit 043b8daa578f ("ASoC: sun4i-i2s: Update global enable with bitmask") has broken the A33 codec since it's been merged as part as 4.14-rc1.
The reason has been that while that commit was obviously right, one of its side effect was to clear the bit that would set the I2S controller in slave mode, effectively forcing it to act as master.
This was actually working by accident since the codec also had its condition on whether to act as a master or a slave backward, meaning that even though we were setting the i2s controller as slave and the codec as master, we were ending up with the exact opposite.
The commit mentionned above (rightfully) broke that combination, and we ended up with the two devices in slave mode, which obviously didn't work.
The way to fix that is in two parts. The first one is obviously to make the codec act as its proper role. However, that's not sufficient because some logic was missing in the initial driver that was merged to act as such. Indeed the BCLK divider was never programmed, meaning that we would generate a BCLK running at the PLL rate, which is way too fast. The second patch addresses that.
The final fix is here to address an issue that has been there for quite some time too and would invert the two channels. It appears that the codec and I2S drivers don't share the same polarity for the LRCK signal. The fix is quite obvious and is to invert whatever value has been programmed in ALSA.
All these commits should go as fix in 4.14 if possible, as without them the audio is completely broken.
Let me know what you think, Maxime
Maxime Ripard (3): ASoC: sun8i-codec: Invert Master / Slave condition ASoC: sun8i-codec: Set the BCLK divider ASoC: sun8i-codec: Fix left and right channels inversion
sound/soc/sunxi/sun8i-codec.c | 63 +++++++++++++++++++++++++++++++++++++++---- 1 file changed, 58 insertions(+), 5 deletions(-)