On 01/13/2012 10:05 AM, Sarah Sharp wrote: > On Thu, Jan 12, 2012 at 10:35:28AM -0500, Alan Stern wrote: >> If that was the main problem, there's an easy way around it. Simply >> never advance the enqueue pointer into the segment currently occupied >> by the dequeue pointer. Of course, in practice this means every ring >> will end up using a minimum of two segments; the tradeoff ought to be >> worthwhile. >> >> But maybe you already know all this... > > It's true, we could make sure there's always one link TRB free to expand > the ring. That solution was just too simple for me to think of, I > guess. :-P > > Andiry, what do you think? > Yeah... I think Alan's suggestion is OK. There is still a waiting-for-dequeue-pointer mechanism for URB enqueue, but this waiting is safe and when expand the ring, the current enqueue segment is always OK to be expanded. I'll try to modify the last patch with this mechanism. Thanks for the suggestions. Thanks, Andiry -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
