[Sound-open-firmware] Questions about the pause-release flow

[Paul Olaru]

Alright, removed some of the state checks from the EDMA driver. This changes
the problem to the one of the hardware interaction between EDMA and ESAI.
Channel states are now not preventing any flow from happening, although I'm
not 100% certain they are correct.

Guess I need to focus on them now.

Jerome, Guido: Since I cannot access the DMA hardware from the ESAI driver
or the ESAI hardware from the DAI driver (separation of concerns) how can I
ensure that changing the ERQ/EARQ bits in the TCD will work correctly? Do I
need to, on resume, fill the pipeline with silence? Do I need to do something
else? At the end of my esai_start() function the TFE condition is signaled and
afterwards dma_start is called in the resume flow.

-----Original Message-----
From: Sound-open-firmware <sound-open-firmware-bounces at alsa-project.org> On Behalf Of Paul Olaru
Sent: Monday, September 16, 2019 1:12 PM
To: Liam Girdwood <liam.r.girdwood at linux.intel.com>; sound-open-firmware at alsa-project.org
Cc: Jerome Laclavere <jerome.laclavere at nxp.com>; Daniel Baluta <daniel.baluta at nxp.com>; Guido Roncarolo <guido.roncarolo at nxp.com>
Subject: Re: [Sound-open-firmware] Questions about the pause-release flow

Replies inline.

> > Shouldn't unpausing actually send COMP_TRIGGER_RELEASE?
> Yeah, it should do - it's designed to mirror the ALSA trigger commands.

Somehow something is weirded out as in the IPC I now see that the trigger
command does differ between pause and release. Funny enough,
dai_comp_trigger receives command 3 for resume but then, without any
extra error trace I get error -22. Will add more traces.

> > I am doing my own state checking but I suspect it may be wrong.
> The FW does check pipeline and component state here, so please make
> sure any DMA has stopped and set the correct state before the stream is
> restarted.

I believe one of those checks fails even though I didn't even interact with
the code in dai.c (AFAIK, I will check)

> > What should I look into, and is this actually correct behavior?
> There is a state transition diagram in component.h iirc, so best to
> make sure that pause puts your pipeline in the correct state. I guess
> that STOP and START work fine ?

STOP and START work fine, barring the occasional IPC timeout issue
which is caused by an unrelated problem on my platform.

Will do further checks and remove my manual state management
(I trust the state management made by component_set_state even
though it probably fails right there). Will reply if I uncover anything
else.
_______________________________________________
Sound-open-firmware mailing list
Sound-open-firmware at alsa-project.org
https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.alsa-project.org%2Fmailman%2Flistinfo%2Fsound-open-firmware&data=02%7C01%7Cpaul.olaru%40nxp.com%7Cd4cac4ce2737464e1cbc08d73a8e705d%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C1%7C637042255758865968&sdata=5WAb0I6Fg1nTXyYD%2BLkuMYXjBXpe7TzYkTUQ1Gn0jT4%3D&reserved=0