Re: [REGRESSION] usb: gadget: f_fs: Allow scatter-gather buffers

[John Stultz]

On Mon, May 13, 2019 at 7:08 AM Andrzej Pietrasiewicz
<andrzej.p@xxxxxxxxxxxxx> wrote:
> W dniu 09.05.2019 o 23:23, John Stultz pisze:
> > So yes, the kzalloc/memset patch is a clear improvement, as it avoids
> > the bootup crash on dwc2, and seems like it should go in.
> >
> > However, there is still the outstanding issue w/  functionfs sg
> > support stalling on larger transfers.
>
> Do you get "functionfs read size 512 > requested size 24, splitting
> request into multiple reads" message when problems happen?

Unfortunately no.

> Is there anything in the kernel log?

Nope. Just the transfers stall/hang and further attempts at adb
connections fail until the USB cable is unplugged and replugged.

> I'm unable to reproduce your problems. I thought I was able, but
> it was another problem, which is fixed with:
>
> 5acb4b970184d189d901192d075997c933b82260
> dwc2: gadget: Fix completed transfer size calculation in DDMA
>
> (or you can simply take upstream drivers/usb/dwc2).

Yea, I'm able to test against mainline. I can give this a whirl, but
since it affects multiple drivers, I suspect the trouble is in the
f_fs code, not just dwc2.

> Do your problems happen on dwc2 or dwc3?

The transfer hangs happen on *both* dwc2 and dwc3. And on both we can
use a similar workaround of disabling sg_supported to get by.

https://git.linaro.org/people/john.stultz/android-dev.git/commit/?h=dev/hikey-mainline-WIP&id=21dfaac615f1f287377897804cbfe69450d489e3
https://git.linaro.org/people/john.stultz/android-dev.git/commit/?h=dev/hikey960-mainline-WIP&id=5b70ec4ae1c06ae700fcca7141130c71e205fa1c


> Is there a way to try your adb without building and running the
> whole Android?

Maybe? I have heard of folks doing it in the past, though I don't
really know a quick path to get there.

Is there anything else I can try for you?

thanks
-john