Hi, On Mon, Jun 25, 2012 at 07:52:01PM +0530, Pratyush Anand wrote: > On 6/25/2012 7:39 PM, Felipe Balbi wrote: > >Hi, > > > >On Mon, Jun 25, 2012 at 07:16:56PM +0530, Pratyush Anand wrote: > >>If a gadget halts control transfer stage, then that stage is stalled, > >>but core should be ready to receive next setup packet. > >> > >>This patch restarts control transfer along with other software correct > >>flag settings. > >> > >>Signed-off-by: Pratyush Anand<pratyush.anand@xxxxxx> > >>Signed-off-by: Michel Sanches<michel.sanches@xxxxxx> > > > >have you actually experienced any issues ? I ask this because if you > >look at dwc3_ep0_stall_and_restart() you will see that this is properly > >handled: > > Yes, I experienced one failing case. cool. > I receive setup packet. They are correct. So gadget goes for this > command execution. Its a lengthy command (a read from a slow device), > so dwc3 goes into DWC3_EP_PENDING_REQUEST. > > Now, gadget experiences that slow device is not working at all. So, > it needs to stall the data IN stage. Gadget calls usb_ep_set_halt. > > I do not see calling of dwc3_ep0_stall_and_restart from set_halt. Did > I miss something? I see... I'm confused with what would be the best fix here: make ->complete() return error code or re-factor stall_and_restart() so that it becomes ep0's set_halt() and set_wedge implementation. What do you think ? -- balbi
Attachment:
signature.asc
Description: Digital signature
