On Wed, 22 Jun 2022 22:23:50 +0200, John Veness wrote:
Hello,
This is my first ever kernel work, so I hope I am doing everything correctly.
I have a USB analogue AV capture dongle with a MacroSilicon MS2100E chip inside. PID 0x534d, VID 0x0021 (534d:0021). Apparently the MS2106 uses the same PID/VID, so hopefully acts the same, but I don't have one to check.
For audio, the MS2100E has the same problems as the MS2109 USB HDMI capture dongle, namely claiming 96kHz mono sound when it's actually 48kHz stereo, and with left/right channel problems.
Luckily, the fix is the exact same as that for the MS2109, as first implemented by Marcan (Hector Martin) in July 2020. Below is a patch which is basically a copy and paste of his, with a different VID.
I have tested this patch in 5.19.0 RC2. I haven't recruited any other testers.
The suggested code change looks OK. So far, so good...
Even with this patch, there is a remaining problem, which is not present in the MS2109. The sound sample values range from 0x0000 to 0x7fff, with silence around 0x4000, i.e. 15-bit-ish audio. This actually sounds OK to the ear (although half as loud as it should be), but looks odd when looking at the waveform, and makes volume meters always think the sound is very loud.
To convert to s16le, I can bitshift one bit left, and subtract 32768. I'm told that this isn't something that can or should be done in the kernel, but should be in userspace. Any more advice on how to fix this remaining quirk would be very welcome.
Ouch, this is painful. We haven't had any devices that require a 15 bit unsigned format, and maybe we don't want to add it to the common standard format just for one funky device, either. Such data processing could be done in alsa-lib, but for the proper interaction with the user-space, the kernel should provide some information so that user-space can process the data accordingly. However, we have no proper way defined for it generically, so far.
Maybe an easy way would be to create an alsa-lib external plugin, and apply it per device. Jaroslav, could it be done via UCM?
Nonetheless, as I say, with the following kernel patch the captured audio certainly sounds right, so is a big improvement and makes these dongles usable:
So as a first step, we can merge the patch as is. The rest needs more consideration.
In anyway, for merging the patch, it has to be submitted properly in a right format, especially including your Signed-off-by tag. Please refer to Documentation/process/submitting-patches.rst for details.
thanks,
Takashi