Re: [PATCH v24 17/34] ASoC: qcom: qdsp6: Add USB backend ASoC driver for Q6

[Wesley Cheng <quic_wcheng@xxxxxxxxxxx>]

Hi Pierre,

On 8/1/2024 11:32 PM, Pierre-Louis Bossart wrote:
>
> On 8/2/24 01:10, Wesley Cheng wrote:
>> Hi Pierre,
>>
>> On 8/1/2024 1:40 AM, Pierre-Louis Bossart wrote:
>>>
>>>> +static int q6usb_hw_params(struct snd_pcm_substream *substream,
>>>> +			   struct snd_pcm_hw_params *params,
>>>> +			   struct snd_soc_dai *dai)
>>>> +{
>>>> +	struct q6usb_port_data *data = dev_get_drvdata(dai->dev);
>>>> +	struct snd_soc_pcm_runtime *rtd = substream->private_data;
>>>> +	struct snd_soc_dai *cpu_dai = snd_soc_rtd_to_cpu(rtd, 0);
>>>> +	struct q6afe_port *q6usb_afe;
>>>> +	struct snd_soc_usb_device *sdev;
>>>> +	int ret;
>>>> +
>>>> +	/* No active chip index */
>>>> +	if (list_empty(&data->devices))
>>>> +		return -EINVAL;
>>>> +
>>>> +	mutex_lock(&data->mutex);
>>>> +	sdev = list_last_entry(&data->devices, struct snd_soc_usb_device, list);
>>>> +
>>>> +	q6usb_afe = q6afe_port_get_from_id(cpu_dai->dev, USB_RX);
>>>> +	if (IS_ERR(q6usb_afe))
>>>> +		goto out;
>>>> +
>>>> +	/* Notify audio DSP about the devices being offloaded */
>>>> +	ret = afe_port_send_usb_dev_param(q6usb_afe, sdev->card_idx,
>>>> +						sdev->pcm_idx);
>>>> +
>>>> +out:
>>>> +	mutex_unlock(&data->mutex);
>>>> +
>>>> +	return ret;
>>>> +}
>>> Humm, multiple questions here
>>>
>>> a) is this intentional that the params are not used in a hw_params routine?
>> Think this was answered in patch#34.
> yes, but that really begs the question if the format check shouldn't be
> added here.

Sure, I'm not against squashing the format check to this patch.

Thanks

Wesley Cheng

>>> b) if yes, could this be replaced by a .prepare callback
>>>
>>> c) along the same lines as b), is suspend-resume during playback
>>> supported? Usually this is handled with a .prepare callback to restore
>>> connections.
>> I don't see us supporting that throughout any of the QC based DAI drivers, so this probably isn't implemented yet.  In terms of supporting system PM suspend for this USB offload path, we're going to explicitly stop the audio stream from the USB offload driver (qc_audio_offload) before we suspend the usb device. (refer to qc_usb_audio_offload_suspend()
> The system suspend-resume during playback is not enabled in all
> platforms indeed, it mostly depends on what userspace does. IIRC this is
> required for Chrome/CRAS and it's supported by aplay.