Re: [RFC PATCH v2 20/22] sound: usb: Prevent starting of audio stream if in use

From: Pierre-Louis Bossart
Date: Tue Feb 07 2023 - 17:44:31 EST




On 2/6/23 19:15, Wesley Cheng wrote:
> Hi Pierre,
>
> On 1/26/2023 8:12 AM, Pierre-Louis Bossart wrote:
>>
>>
>> On 1/25/23 21:14, Wesley Cheng wrote:
>>> With USB audio offloading, an audio session is started from the ASoC
>>> platform sound card and PCM devices.  Likewise, the USB SND path is
>>> still
>>> readily available for use, in case the non-offload path is desired.  In
>>> order to prevent the two entities from attempting to use the USB bus,
>>> introduce a flag that determines when either paths are in use.
>>>
>>> If a PCM device is already in use, the check will return an error to
>>> userspace notifying that the stream is currently busy.  This ensures
>>> that
>>> only one path is using the USB substream.
>>
>> It's good to maintain mutual exclusion, but it's still very hard for an
>> application to figure out which card can be used when.
>>
>> Returning -EBUSY is not super helpful. There should be something like a
>> notification or connection status so that routing decisions can be made
>> without trial-and-error.
>>
>
> The USB offload driver does have access to the USB substream that is
> being utilized/offloaded.  Maybe in addition to this check, we can also
> set the PCM runtime state as well (for that particular substream)?  That
> way userspace can fetch information about if the stream is busy or not.

You're missing the point. When a card is exposed but the PCM devices may
or may not be usable (consuming data with no sound rendered or returning
an error), it's much better to provide a clear connection status to
userspace.

Let me give you an example. Intel drivers can expose 3 HDMI/DP PCM
devices. Userspace has no idea which one to use, so there's a jack
control that tells userspace whether there is a receiver connected so
that the audio server can use the relevant PCM device.

Audio routing based on trial and error is really problematic, errors can
happen but they should be exceptional (e.g. xruns), not a means of
driver-userspace communication on the device status.