Hi Daniel,

I've never met such an behavior and I've no idea what may be the root cause. Did you check how generic the problem is (topology/platform dependent)?

Karol

W dniu 21.07.2020 o 20:03, Pierre-Louis Bossart pisze:

On 7/21/20 8:50 AM, Daniel Baluta wrote:

...

Hi Tomasz, Karol, Ranjani,

I wonder if you have any clues on why $subject happens.

We have a simple pipeline like this:

PCM0  --> B0--+                           |--mixer-->B1-->DAI PCM1 --> B1--+

Subdevices are created like this:

oot@imx8qxpc0mek:~# aplay -l **** List of PLAYBACK Hardware Devices **** card 0: sofwm8960audio [sof-wm8960-audio], device 0: Port0 (*) []    Subdevices: 1/1    Subdevice #0: subdevice #0 card 0: sofwm8960audio [sof-wm8960-audio], device 1: Passthrough Playback 1 (*) []    Subdevices: 1/1    Subdevice #0: subdevice #0

But, when we use aplay in the following order:

$ (1) aplay -Dhw:0,1 -f S32_LE -c 2 -r 48000 file1.wav $ (2) aplay -Dhw:0,0 -f S32_LE -c 2 -r 48000 file0.wav

everything is OK, mixing works.

But when the order is reversed:

$ (2) aplay -Dhw:0,0 -f S32_LE -c 2 -r 48000 file0.wav $ (1) aplay -Dhw:0,1 -f S32_LE -c 2 -r 48000 file1.wav

(1) aplay exists with an error: aplay: pcm_write:2057: write error: Input/output error

I wonder if you met this. Topology looks to be symmetric with respect to PCM0/PCM1.

Maybe an issue with the scheduling dependencies?

This looks similar to what I recently changed for Baytrail/CherryTrail, and I had to go through some hoops to make sure both 'host' pipelines were scheduled based the DAI pipeline:

https://github.com/thesofproject/sof/blob/master/tools/topology/sof-byt-code...