am335x g_audio gadget cannot be enumerated properly on windows XP

Hi Sam,

I will ask the factory team for support on this.

I never got UAC1 enumeration working correctly since kernel 3.2.

Samming Wang said:

I got a "!" symbol in the windows device manager after USB enumeration.

How to debug it?

It does not tell much with Windows host. You might want to connect to a Linux host, which gives more enumeration information in kernel logs. Even better if you have a protocol analyzer, with debugging the UAC1 driver, you might get some clue of the failure.

Hi Bin,

Thanks for your information!

Another question, how about the UAC2 support on SDK7.0?

Thanks

Samming,

I can build and load the UAC2 module without any issue, but have not got time to play with it yet. I read somewhere that the UAC2 driver is different from UAC1, that UAC2 does not link to the real audio module for audio input/output as in UAC1, so I guess we have to do something more to test UAC2, I may be wrong.

Hi Bin,

I tried to play a wave file to UAC2 alsa device and receive it on USB host; it is successful with USB max packet size of value "64".

But when I replace the max packet size with 512, it fails.

What happens?

Thanks,

Sam

Hi Bin,

If UAC2 test is not convenient to you, UAC1 will be helpful for me.

Could you help verify UAC1 to play 48k audio? 

I doubt whether the USB UDC/DMA driver (of AM335x SDK7) can achieve max packet size of 1024 bytes in isochronous mode.

Thanks,

Sam

Sam,

I am busy on other things and will not have time on UAC.

Samming Wang said:

Could you help verify UAC1 to play 48k audio? 

I tried a few times before, but never got UAC1 enumeration working.

Samming Wang said:

I doubt whether the USB UDC/DMA driver (of AM335x SDK7) can achieve max packet size of 1024 bytes in isochronous mode.

I doubt this has been validated, but I don't think 1024 bytes ISO should have issue in gadget mode. In host mode, webcam with 1020 bytes ISO has been validated. It is the same DMA driver for both host and gadget mode.

My question is UAC does not require much data, a few hundred bytes should be sufficient, why you want to use 1024 bytes?

Hi Bin,

1024-byte is just for confirming the driver performance; in fact, 512-byte is enough for my applications.

As I mentioned before, it fails when I play a 48k wave file by UAC2 with 512-byte packet, but it successes playing a 16k wave file with 64-byte packet.

I would like to make sure whether original Linux f_uac1 and f_uac2 modules are workable in AM335x SDK7 or not?

Thanks for your effort,

Sam

Sam,

I guess you tested 512-byte packets with CPPI DMA enabled. Have to tried to disable CPPI with 512-byte packets to see if it works?

Hi Bin,

Could you kindly tell me how to disable CPPI DMA?

Thanks,

Sam

SDK70 disable CPPI DMA in menuconfig:

Device Drivers  --->
    [*] USB support  --->
        MUSB DMA mode (Disable DMA (always use PIO))  --->

Hi Bin,

I got the same result when using "Disable CPPI"  config.

Any other suggestion?

Thanks,

Sam

Sam,

If UAC still does not work even with CPPI disabled, I would think this is the UAC driver issue. But I did not play much with UAC, and not able to give your further comments.

Bin,

AM335x TRM states double packet buffering for USB fifo can improve iso transactions bandwidth and avoid underrun and overrun errors.

When would you release the related patch for activating the double packet buffering mechanism?

Thanks,

Sam

Sam, we don't have a plan yet to implement the double buffering.