> On Mon, 23 Aug 2010, David Brownell wrote: > > > t it seems > > wasteful (and > > possibly flat-out wrong in the presence of frame-list > > wraparound) > > ISTR comments about the dubiousness of wraparound behaviors... Here's a scenario to consider. Suppose there are no isochronous transfers going on, but there is a long-period interrupt transfer and the completion handler always resubmits. Suppose two successive URBs complete exactly one second apart (since the period is large this is not unlikely), and let N be the frame list entry in which the two transactions occurred. Finally, suppose that IRQ latency causes the interrupt handler for the second completion to be delayed by 1 ms. Then during the first IRQ the CPU will read N from the frame counter and during the second it will read N+1. As a result, while processing the second IRQ, the big loop in scan_periodic() will look only for events occurring during frame N+1, so it won't see the URB that completed during frame N. It seems we should insure that scan_periodic() checks at least the last SCHEDULE_SLOP frames if the frame list has wrapped since the last scan. Even worse things can happen with isochronous transfers if too many of them are scheduled with the URB_NO_INTERRUPT flag. We could end up in a situation where new transfers can't be added because the frame list is full, and the frame list doesn't empty out because we never get a completion IRQ. We need to have at least one such IRQ before the frame list wraps. I'm not sure what the best solution is, other than simply to ignore URB_NO_INTERRUPT for isochronous URBs. Alan Stern -- 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
