Can you merge this to Marcelo's tree? It's against pre7. This patch just adds a few strategic wmb() calls, to ensure that the HC and HCD agree on just what control and bulk requests are being made.
Seems restructuring "usb-ohci" into "ohci-hcd" made that failure become more common/visible. (Faster code?) So this backports a 2.5 "ohci-hcd" fix into 2.4's "usb-ohci". The effect of not having these calls was that some hardware could fail to enumerate devices ... kernel messages would report a "usb_control/bulk_msg: timeout" happened. We've had such (infrequent) problem reports before, but until now nobody was able to confirm the wmb() calls mattered. - Dave
ohci24-0501.patch
Description: Binary data
