On Sun, 13 Oct 2002, Johannes Erdfelt wrote: > BTW - I did some searching through the specs, and while it doesn't > guarantee much to control transfers, it does specifically mention (5.5.4): > > "If there are control transfers pending for multiple endpoints, control > transfers for the different endpoints are selected according to a fair > access policy that is Host Controller implementation-dependent". > > It also goes on to say that control is "best effort", etc. > > However, with depth first on control, it can definately be construed as > not being fair. So, switching low speed control back to breadth is > probably the best thing to do, like in your original patch.
Another point to consider: As it was already said, EP0-control is mostly used for configuration. This implies there might be rather long delays appearing when the device has to do the internal housekeeping. It's not only the 50ms granted for every control transfer to complete - if there is a DATA stage, a single control transfer might take up to 5 seconds. Martin ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
