Re: [PATCH v3 06/25] ASoC: renesas: rz-ssi: Terminate all the DMA transactions

[Claudiu Beznea]

Hi, Geert,

On 09.12.2024 15:15, Geert Uytterhoeven wrote:
> Hi Claudiu,
> 
> On Wed, Nov 13, 2024 at 2:35 PM Claudiu <claudiu.beznea@xxxxxxxxx> wrote:
>> From: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
>>
>> In case of full duplex the 1st closed stream doesn't benefit from the
>> dmaengine_terminate_async(). Call it after the companion stream is
>> closed.
>>
>> Fixes: 4f8cd05a4305 ("ASoC: sh: rz-ssi: Add full duplex support")
>> Cc: stable@xxxxxxxxxxxxxxx
>> Reviewed-by: Biju Das <biju.das.jz@xxxxxxxxxxxxxx>
>> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
> 
> Thanks for your patch!
> 
>> Changes in v3:
>> - collected tags
>> - use proper fixes commit SHA1 and description
> 
> I am not sure which one is the correct one: the above, or commit> 26ac471c5354583c ("ASoC: sh: rz-ssi: Add SSI DMAC support")...

IIRC, I had this one on the previous version but in the review process it
has been proposed to used 4f8cd05a4305 ("ASoC: sh: rz-ssi: Add full duplex
support"). I think 4f8cd05a4305 ("ASoC: sh: rz-ssi: Add full duplex
support") is the right one as the issue is related to full duplex.


> 
>> --- a/sound/soc/renesas/rz-ssi.c
>> +++ b/sound/soc/renesas/rz-ssi.c
>> @@ -415,8 +415,12 @@ static int rz_ssi_stop(struct rz_ssi_priv *ssi, struct rz_ssi_stream *strm)
>>         rz_ssi_reg_mask_setl(ssi, SSICR, SSICR_TEN | SSICR_REN, 0);
>>
>>         /* Cancel all remaining DMA transactions */
>> -       if (rz_ssi_is_dma_enabled(ssi))
>> -               dmaengine_terminate_async(strm->dma_ch);
>> +       if (rz_ssi_is_dma_enabled(ssi)) {
>> +               if (ssi->playback.dma_ch)
>> +                       dmaengine_terminate_async(ssi->playback.dma_ch);
>> +               if (ssi->capture.dma_ch)
>> +                       dmaengine_terminate_async(ssi->capture.dma_ch);
>> +       }
> 
> rz_ssi_stop() is called twice: once for capture, and a second time for
> playback. How come that doesn't stop both?

It is called from this path:

static int rz_ssi_dai_trigger(struct snd_pcm_substream *substream, int cmd,

                              struct snd_soc_dai *dai)

{

        // ...
        case SNDRV_PCM_TRIGGER_STOP:

                rz_ssi_stop(ssi, strm);

                rz_ssi_stream_quit(ssi, strm);


        // ...
}

rz_ssi_stop() is as follow:

static int rz_ssi_stop(struct rz_ssi_priv *ssi, struct rz_ssi_stream *strm)

{

        strm->running = 0;



        if (rz_ssi_is_stream_running(&ssi->playback) ||

            rz_ssi_is_stream_running(&ssi->capture))

                return 0;

        // ...
}

rz_ssi_is_stream_running() is as follows:

static inline bool rz_ssi_is_stream_running(struct rz_ssi_stream *strm)

{

        return strm->substream && strm->running;

}


The strm->substream is set to NULL in:

static void rz_ssi_stream_quit(struct rz_ssi_priv *ssi,

                               struct rz_ssi_stream *strm)

{

        struct snd_soc_dai *dai = rz_ssi_get_dai(strm->substream);



        rz_ssi_set_substream(strm, NULL);


        // ...
}

Thus, when the 1st full duplex stream is closed, as the companion stream is
still running it doesn't benefit from dmaengine_terminate_async().

I'll update the commit description in the next version.

Thank you for your review,
Claudiu

> Perhaps the checks at the top of rz_ssi_stop() are not correct?
> Disclaimer: I am no sound expert, so I may be missing something...
> 
>>
>>         rz_ssi_set_idle(ssi);
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
>