Re: [PATCH v2 0/2 RESEND] usb: gadget: uac2: support explicit channel counts

[kaf@nwlink.com]

On 2026-10-17 01:05, Greg Kroah-Hartman wrote:
> On Fri, Oct 16, 2026 at 09:25:32AM -0700, Kyle Farrell wrote:
>> USB Audio Class 2.0 permits up to 255 channels in an audio channel 
>> cluster
>> via bNrChannels. Independently, a channel-location bitmap is specified 
>> via
>> bmChannelConfig for up to 27 predefined spatial locations. Setting
>> bmChannelConfig to zero indicates that the channels have no predefined
>> spatial positions.
>> 
>> The current UAC2 gadget driver derives bNrChannels from the number of 
>> set
>> bits in bmChannelConfig. Since bmChannelConfig provides 27 predefined
>> spatial locations this limits playback and capture streams to 27 
>> channels.
>> 
>> This series allows channel counts up to the UAC2 limit of 255 without
>> requiring predefined spatial channel locations in bmChannelConfig.
>> Existing ConfigFS channel-mask configuration continues to work 
>> unchanged.
>> New optional playback and capture count attributes allow the number of
>> channels to be specified separately from the channel-location bitmap.
>> 
>> The series is split so that the shared u_audio infrastructure is 
>> updated
>> first without changing behavior, followed by the UAC2 functional 
>> change.
>> 
>> Patch 1:
>>   Pass resolved playback and capture channel counts explicitly into
>>   the UAC1/UAC2 shared u_audio implementation instead of deriving them
>>   from channel masks.
>> 
>> Patch 2:
>>   Add ConfigFS playback/capture channel-count attributes to the UAC2
>>   function driver and use them when generating descriptors and stream
>>   configuration, allowing counts beyond the 27 predefined channel
>>   locations.
>> 
>> Testing
>> 
>> Testing was performed using a Raspberry Pi Zero 2 W configured as a 
>> USB
>> UAC2 gadget. Streaming and descriptor validation were performed 
>> against
>> both Linux and macOS hosts.  Windows hosts were not tested.
>> 
>> UAC1
>>   - Verified playback/capture operation via alsaloop.
>>   - No descriptor or behavioral changes observed.
>> 
>> UAC2
>>   - Verified legacy channel-mask configuration (1-27 channels).
>>   - Verified explicit channel-count mode with 2, 28, 61, 62, 96 
>> channels.
>>   - Verified invalid channel-count/channel-mask combinations are 
>> rejected.
>>   - Verified Feature Unit descriptor size limits at 61/62 channel 
>> boundary.
>>   - Verified descriptors including:
>>       * bNrChannels
>>       * bmChannelConfig
>>       * Feature Unit presence and bLength
>>       * AudioControl wTotalLength
>>       * endpoint wMaxPacketSize
>>   - Verified ALSA-reported playback and capture channel counts.
>>   - Verified streaming using alsaloop.
>> 
>> Changes in v2:
>>   - Edit commit messages for clarity and brevity.
>>   - Add Assisted-by: LLM attribution.
>>   - No code changes.
> 
> Why was RESEND needed here?  Saying that would be nice.
> 
> Also, sashiko points out a number of problems...
> 
> thanks,
> 
> greg k-h

I resent v2 as both it and v1 had received no review. I should have 
stated
that in the cover letter.

I'm reviewing the Sashiko findings and will address relevant issues 
before
submitting a revised series.

Thanks,
Kyle