Hi, On Tue, Feb 07, 2012 at 11:47:50PM -0800, Ido Shayevitz wrote: > > One thing that strikes me though, is that whatever Tanya did, will now > > become quite messy because now we have the gadget driver setting > > ep->desc (through config_ep_by_speed()) but the UDC needs the clear that > > pointer on ep_disable. This smeels like disaster to me, but go ahead. > > Split your patch and resend. > > Yes... harmful, but the bug is already there... :) > Moreover, the config_ep_by_speed() sets the ep->desc, and when enabling > the endpoint (gadget.h::usb_ep_enable) the ep->desc is sent to the udc. > So hopefully, just clearing the ep->desc on disable is the only problem. > > So as first bugfix I will just set the endpoint descriptor to NULL on > endpoint disable. > Should I do it to all the UDCs ? all in one patch ? sure, no need to split that part. One patch fixing the bug all over the place is fine. -- balbi
Attachment:
signature.asc
Description: Digital signature
