High-bandwidth/multiple transactions per microframe in usb audio?

High-bandwidth/multiple transactions per microframe in usb audio?

[Pavel Hofman]

Hi,

Please does the current USB audio driver support multiple transactions
per microframe (high-bandwidth mode)? I could not find it in endpoint.c
but it may be there or somewhere else :-)

Thanks a lot,

Pavel.

Re: High-bandwidth/multiple transactions per microframe in usb audio?

[Alexander Tsoy]

В Пт, 24/04/2020 в 11:42 +0200, Pavel Hofman пишет:
> Hi,
> 
> Please does the current USB audio driver support multiple
> transactions
> per microframe (high-bandwidth mode)? I could not find it in
> endpoint.c
> but it may be there or somewhere else :-)

This should be transparent for the driver. See usb_submit_urb()
function from drivers/usb/core/urb.c. It allows packets larger than
1024 for isochronous transfers over High Speed and Super Speed buses.

Re: High-bandwidth/multiple transactions per microframe in usb audio?

[Pavel Hofman]

Dne 25. 04. 20 v 14:06 Alexander Tsoy napsal(a):
> В Пт, 24/04/2020 в 11:42 +0200, Pavel Hofman пишет:
>> Hi,
>>
>> Please does the current USB audio driver support multiple
>> transactions
>> per microframe (high-bandwidth mode)? I could not find it in
>> endpoint.c
>> but it may be there or somewhere else :-)
> 
> This should be transparent for the driver. See usb_submit_urb()
> function from drivers/usb/core/urb.c. It allows packets larger than
> 1024 for isochronous transfers over High Speed and Super Speed buses.

Alexander, thanks a lot for the info and hint, indeed at
https://github.com/torvalds/linux/blob/master/drivers/usb/core/urb.c#L444

Best regards,

Pavel.

Re: High-bandwidth/multiple transactions per microframe in usb audio?

[Alexander Tsoy]

В Сб, 25/04/2020 в 20:06 +0200, Pavel Hofman пишет:
> Dne 25. 04. 20 v 14:06 Alexander Tsoy napsal(a):
> > В Пт, 24/04/2020 в 11:42 +0200, Pavel Hofman пишет:
> > > Hi,
> > > 
> > > Please does the current USB audio driver support multiple
> > > transactions
> > > per microframe (high-bandwidth mode)? I could not find it in
> > > endpoint.c
> > > but it may be there or somewhere else :-)
> > 
> > This should be transparent for the driver. See usb_submit_urb()
> > function from drivers/usb/core/urb.c. It allows packets larger than
> > 1024 for isochronous transfers over High Speed and Super Speed
> > buses.
> 
> Alexander, thanks a lot for the info and hint, indeed at
> https://github.com/torvalds/linux/blob/master/drivers/usb/core/urb.c#L444
> 

And I just noticed that USB-audio driver can't properly parse max
packet size for SuperSpeed and SuperSpeed+ devices yet. See
audio_format_alloc_init().

Re: High-bandwidth/multiple transactions per microframe in usb audio?

[Pavel Hofman]

Dne 25. 04. 20 v 21:50 Alexander Tsoy napsal(a):
> В Сб, 25/04/2020 в 20:06 +0200, Pavel Hofman пишет:
> 
> And I just noticed that USB-audio driver can't properly parse max
> packet size for SuperSpeed and SuperSpeed+ devices yet. See
> audio_format_alloc_init().

Very good catch. IIUC there are two basically equivalent calculations
for HighSpeed -

https://elixir.bootlin.com/linux/v5.6.7/source/drivers/usb/core/urb.c#L445


https://elixir.bootlin.com/linux/v5.6.7/source/sound/usb/stream.c#L691


And audio_format_alloc_init missing the sections for SS, SS+ in
usb_submit_urb . Perhpas a common method to make the calculation for all
cases would make it more robust?

Thanks,

Pavel.

Re: High-bandwidth/multiple transactions per microframe in usb audio?

[Takashi Iwai]

On Sun, 26 Apr 2020 10:55:32 +0200,
Pavel Hofman wrote:
> 
> Dne 25. 04. 20 v 21:50 Alexander Tsoy napsal(a):
> > В Сб, 25/04/2020 в 20:06 +0200, Pavel Hofman пишет:
> > 
> > And I just noticed that USB-audio driver can't properly parse max
> > packet size for SuperSpeed and SuperSpeed+ devices yet. See
> > audio_format_alloc_init().
> 
> Very good catch. IIUC there are two basically equivalent calculations
> for HighSpeed -
> 
> https://elixir.bootlin.com/linux/v5.6.7/source/drivers/usb/core/urb.c#L445
> 
> 
> https://elixir.bootlin.com/linux/v5.6.7/source/sound/usb/stream.c#L691
> 
> 
> And audio_format_alloc_init missing the sections for SS, SS+ in
> usb_submit_urb . Perhpas a common method to make the calculation for all
> cases would make it more robust?

Yes, a common USB helper function sounds like a good idea.


thanks,

Takashi