Hi, On Tue, Sep 02, 2014 at 12:02:03PM -0400, Alan Stern wrote: > On Tue, 2 Sep 2014, Felipe Balbi wrote: > > > On Tue, Sep 02, 2014 at 11:32:52AM -0400, Alan Stern wrote: > > > On Fri, 29 Aug 2014, Felipe Balbi wrote: > > > > > > > > I still have some old patch files lying around, adding reset callback > > > > > support to dummy-hcd, net2280, and net2272. Would you like me to post > > > > > them? > > > > > > > > Please do :-) let's get all of that sorted out soon. > > > > > > Patches coming up. These were written about two years ago, and > > > although they have been forward-ported, I haven't tested them since > > > they were written. They are based on a patch you posted on August 16, > > > 2012 (usb: gadget: add reset method to struct usb_gadget_driver). > > > > alright, I still have my branch with that patch together with musb and > > dwc3 implementation. The problem I see, though, is that all three of > > your patches and my dwc3 and musb implementation conditionally calls > > ->disconnect() if the gadget driver doesn't driver doesn't implement > > ->reset(). If we're talking about usb_gadget_disconnect() from > > ->disconnect(), than all 5 patches will cause regressions. > > That needn't be a problem. If Peter updates the four gadget drivers, > adding reset callbacks, then we can remove the parts of our patches > that invoke the disconnect callback if there is no reset callback. In > other words, we can make reset callbacks mandatory for gadget drivers. alright, but we need to do this in steps to avoid regressions or a non-bisectable tree. So maybe we add ->reset as an optional method, implement support for it to all UDC drivers, patch all gadget drivers to implement reset, make reset mandatory. Does that work ? -- balbi
Attachment:
signature.asc
Description: Digital signature
