Skip to contentSkip to conversation

Back-end state transition query

DISCUSSION Patrick Lai 2 participants 3 messages Oct 25, 2012 – Oct 29, 2012

lore
Share & tools
lore
#0 lore raw

Hi Liam,

I am facing an issue related to how DPCM manages state of back-end substream. Here is description of problem:

When front-end PCM substream is in paused state, back-end PCM substream will be put in paused state as well if given front-end PCM substream is the only client of given back-end. Based on current implementation of DPCM, when application closes front-end PCM substream, DPCM framework will not allow back-end enter HW_FREE state so back-end will never get shutdown completely as well.

refer to function soc_dpcm_be_dai_hw_free(), here is the check for states

    if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_PARAMS) &&
        (be->dpcm[stream].state != SND_SOC_DPCM_STATE_PREPARE) &&
        (be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_FREE) &&
        (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP))
                        continue;

Can't DPCM also allow back-end hardware being freed in STATE_PAUSED?
Even if DPCM expects trigger stop coming to update back-end state to STATE_STOP, the following condition check will never allow back-end enter stop state anyway

Refer to soc_dpcm_be_dai_trigger()

case SNDRV_PCM_TRIGGER_STOP:

     if (be->dpcm[stream].state != SND_SOC_DPCM_STATE_START)
                                continue;

Can you please take a look and recommend the appropriate state transition or it would be okay to allow PAUSED state in soc_dpcm_be_dai_hw_free()?

Thanks
Patrick

Signature · 2 lines
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
#1 lore raw

Hi Patrick,

Quoted 14 lines · Patrick Lai

Hi Liam,

I am facing an issue related to how DPCM manages state of back-end substream. Here is description of problem:

When front-end PCM substream is in paused state, back-end PCM substream will be put in paused state as well if given front-end PCM substream is the only client of given back-end. Based on current implementation of DPCM, when application closes front-end PCM substream, DPCM framework will not allow back-end enter HW_FREE state so back-end will never get shutdown completely as well.

refer to function soc_dpcm_be_dai_hw_free(), here is the check for states

Quoted 7 lines
    if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_PARAMS) &&
        (be->dpcm[stream].state != SND_SOC_DPCM_STATE_PREPARE) &&
        (be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_FREE) &&
        (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP))
                        continue;

Can't DPCM also allow back-end hardware being freed in STATE_PAUSED?

This seems like a bug. We should hw_free() the BE here.

Quoted 9 lines

Even if DPCM expects trigger stop coming to update back-end state to STATE_STOP, the following condition check will never allow back-end enter stop state anyway

Refer to soc_dpcm_be_dai_trigger()

case SNDRV_PCM_TRIGGER_STOP:

     if (be->dpcm[stream].state != SND_SOC_DPCM_STATE_START)
                                continue;

Can you please take a look and recommend the appropriate state transition or it would be okay to allow PAUSED state in soc_dpcm_be_dai_hw_free()?

It should be OK to allow the paused state in hw_free(), but I'm thinking we should probably see the trigger stop when we close the FE (although I'm not 100% certain atm of seeing the STOP as I've not traced this PAUSE -> CLOSE sequence). I assume you have a patch that allows this ?

Thanks

Liam

#2 lore raw

On 10/25/2012 10:59 AM, Liam Girdwood wrote:
Hi Liam,

Quoted 1 line

It should be OK to allow the paused state in hw_free(), but I'm thinking we should probably see the trigger stop when we close the FE (although I'm not 100% certain atm of seeing the STOP as I've not traced this PAUSE -> CLOSE sequence).

I am also surprised to see that trigger stop never comes. I can dig in more to figure out why.

I assume you have a patch that allows this ?

Yes, I will publish to alsa-devel

Thanks
Patrick

Quoted 11 lines

Thanks

Liam

_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

Signature · 2 lines
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.