On Sun, 2 Dec 2001, Mark Burazin wrote: > usb_submit_urb() can fail, but it will call the completion handler, while > before it was not so, so I manually freed his urb and data when submit failed
Do I understand you correctly, you've seen situations where usb_submit_urb returns some error (which one) but the completion handler for this very urb gets called later on? Wouldn't this create a whole bunch of races like you've seen for freeing vs. completion handler? And how to solve this - blocking on urb->complete would sleep forever if the completion handler would not always be called. IMHO the returning of usb_submit_urb() should (and is, AFAICS) a clear barrier indicating whether the completion handler will get called or not. Martin _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
