On Mon, Jun 06, 2011 at 11:10:23AM -0400, Alan Stern wrote: > How about just unconditionally calling usb_hcd_alloc_bandwidth() before > looping through the endpoint arrays? I don't think that's a good idea, because there could be URBs pending before the loop calling usb_disable_endpoint() completes. The bandwidth allocation functions assume that all URB have been canceled or completed before an endpoint is dropped. Otherwise the command to remove the endpoint rings can race with transfer completion. > Alternatively, how about making > xhci-hcd's endpoint_disable method responsible for releasing the > endpoint's bandwidth reservation? The problem is that the USB core calls the endpoint_disable method too often, for all endpoints, without checking whether the endpoints were actually active. The xHCI driver doesn't keep track of which endpoints are added; it expects the USB core to do so. If the USB core unconditionally drops all endpoints, the command to drop them will be rejected by the host controller. Besides, I'm not too keen on re-writing the whole bandwidth checking code just for this special case. :) Any other thoughts? Is there anything technically wrong with the patch, or it just looks ugly? 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
