On Fri, Aug 05, 2011 at 01:55:21PM +0800, Xu, Andiry wrote: > In the patch description you mention "but leave the cycle bit of the > first TRB (which will show software-owned) intact", but here it leaves > both the first_trb and last_trb intact. I think that is because here the > cur_td->last_trb actually point to the trb pass the real last trb of the > td, right? From driver view, that is correct I think, but maybe this > casuses some sort of confusion because here the cur_td->last_trb does > not really belong to the td. Please correct me if I'm wrong. Yes, that is correct. last_trb points to one TRB past the last enqueued TRB. I have to do it that way, in case there is a link TRB at the very end of the last TD. The link TRB would have been marked as hardware owned in inc_enq() when the last TRB of the last TD was enqueued. If I only passed in the last TRB of the last enqueued TD, it would have pointed to the TRB before the link TRB, and the link TRB wouldn't have its cycle bit flipped back. Since inc_enq() XORs the cycle bit, the next time something was enqueued up to the link TRB, the cycle would have been flipped the opposite way we intended (to software owned). Is there some comment I can add that would make this more clear? Sarah Sharp -- 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
