On Thu, Aug 21, 2014 at 03:55:13PM -0400, Alan Stern wrote: > On Thu, 21 Aug 2014, Paul Zimmerman wrote: > > > > From: Alan Stern [mailto:st...@rowland.harvard.edu] > > > Sent: Thursday, August 21, 2014 12:16 PM > > > > > > On Thu, 21 Aug 2014, Paul Zimmerman wrote: > > > > > > > AFAIK, it has always been a requirement that usb_ep_dequeue() release > > > > the data buffer and call the urb's completion routine before it returns. > > > > > > Hmmm. I don't know if that really has been a requirement, but it does > > > seem to be true in the UDC drivers I'm familiar with. > > > > > > > Quite a few of the function drivers depend on that, after the call to > > > > usb_ep_dequeue() they immediately free the data buffer. So if it's not > > > > explicitly documented like that, it should be. > > > > > > Okay, would you like to write a documentation patch? > > > > Something like this? I didn't find any references to usp_ep_dequeue() > > under Documentation/, so I guess the only place is the kerneldoc in the > > header. > > Yes, that's where it is. > > > I also fixed up some strange capitalization in the existing > > text while I was there. If this looks OK to you I will send a proper > > patch submission to Felipe. > > > > [PATCH] usb: gadget: document a usb_ep_dequeue() requirement > > > > Document the requirement that the request be dequeued and its > > completion routine called before usb_ep_dequeue() returns. > > > > diff --git a/include/linux/usb/gadget.h b/include/linux/usb/gadget.h > > index c3a6185..c540557 100644 > > --- a/include/linux/usb/gadget.h > > +++ b/include/linux/usb/gadget.h > > @@ -345,12 +345,13 @@ static inline int usb_ep_queue(struct usb_ep *ep, > > * @ep:the endpoint associated with the request > > * @req:the request being canceled > > * > > - * if the request is still active on the endpoint, it is dequeued and its > > + * If the request is still active on the endpoint, it is dequeued and its > > * completion routine is called (with status -ECONNRESET); else a negative > > - * error code is returned. > > + * error code is returned. This is guaranteed to happen before the call to > > + * usb_ep_dequeue() returns. > > * > > - * note that some hardware can't clear out write fifos (to unlink the > > request > > - * at the head of the queue) except as part of disconnecting from usb. > > such > > + * Note that some hardware can't clear out write fifos (to unlink the > > request > > + * at the head of the queue) except as part of disconnecting from usb. Such > > * restrictions prevent drivers from supporting configuration changes, > > * even to configuration zero (a "chapter 9" requirement). > > */ > > Looks good to me. > > Acked-by: Alan Stern <st...@rowland.harvard.edu>
Thanks Alan and Paul :-) Paul, can you just send this as a formal patch with Alan's Acked-by ? Thank you -- balbi
signature.asc
Description: Digital signature
