> However, I have two issues with this patch. > > First, it seems like you're giving the link TRB to the hardware before > the > driver writes the TD at the top of the ring. Won't you run into the > same issue you had with streams processing? Or will the host > controller > not notice the link TRB until the driver rings the doorbell? It should be fine because of the doorbell. But even if it comes fast and the controller picks it up, the controller won't do anything until the following TRBs are activated at which point it will continue with the stream state machine. It's only when there are no following TRBs and the link TRB is activated as part of the previous TD that we have a problem. > > Second, I don't think you're handling hardware that has the link TRB > quirk correctly. For that 0.95 hardware, the xHCI driver has to make > sure never to clear the chain bit on the link TRB. (This is because > one > part of the 0.95 spec said "the chain bit shall always be set", and > then > the stop endpoint rules talked about the chain bit being cleared.) Ok I've reworked the patch. I'll send it shortly. John -- 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
