> On Tue, Jun 30, 2015 at 12:33 AM, Steve Calfee <stevecalfee@xxxxxxxxx>
> wrote:
> > On Mon, Jun 29, 2015 at 7:16 AM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>
> wrote:
> >> On Mon, 29 Jun 2015, Peter Chen wrote:
> >>
> >>> Just like Steve pointed, it should be a ZLT problem, do you have
> >>> below patch in your tree, and the host may not send zlt, but you may
> >>> queue an zero-length request, the f_hid does not set req->zero flag
> >>> either.
> >>>
> >>> commit 953c66469735aed8d2ada639a72b150f01dae605
> >>> Author: Abbas Raza <Abbas_Raza@xxxxxxxxxx>
> >>> Date: Thu Jul 17 19:34:31 2014 +0800
> >>>
> >>> usb: chipidea: udc: Disable auto ZLP generation on ep0
> >>>
> >>>
> >>> If it still has problem, send me apps if possible (with needed
> >>> kernel patches).
> >>
> >> Note that a UDC should _never_ queue an extra zero-length packet for
> >> an OUT transfer, no matter how req->zero is set. In other words,
> >> req->zero is supposed to affect only IN transfers.
> >>
> > Yes UDC gadgets receive packets sent by the host on OUT transfers. If
> > ZLT transfers are needed for the protocol, a short transfer is needed
> > to end an OUT sequence and it must be sent by the sender, in this case
> > the host. The UDC should detect receiving a zero length EP0 OUT and
> > that signals the end of the 64 byte transfer from host to gadget.
> >
> > Regards, Steve
>
> Thanks Alan, Peter and Steve for your help and quick response.
>
> In fact, today I tested with a similar patch i.e. have QH_IOS and QH_ZLT set for
> control endpoint in udc.c. The 64 byte transfer appears to work fine.
>
> I am not sure whether this is the ideal solution, Testing some more to ensure
> that there are no regressions. I will send out an update in about a day.
>
The most probably is the QH_ZLT bit, your tree may not the patch I mentioned
above. The QH_IOS only affects a setup being received.
Peter
��.n��������+%������w��{.n�����{���)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
