On Mon, Nov 23, 2009 at 7:12 AM, Andy Walls <awalls@xxxxxxxxx> wrote: > 5. If you don't give an MDL back to the firmware, it never uses it > again. That's why you see the sweep-up log messages. As soon as an MDL > is skipped *on the order of the depth* of q_busy times, when looking for > the currently DMA_DONE'd MDL, that skipped MDL must have been dropped. > It is picked up and put back into rotation then. Perhaps I am misinterpreting the definition of "sweep-up" in this context. Don't the buffers get forcefully returned to the pool at that point? If so, why would I see the same error over and over long after the CPU utilization has dropped back to a reasonable level. I feel like I must be missing something here. 1. CPU load goes up (ok) 2. Packets start to get dropped (expected) 3. CPU load goes back down (ok) 4. Packets continue to get dropped and never recycled, even after minutes of virtually no CPU load? I can totally appreciate the notion that the video would look choppy when the system is otherwise under high load, but my expectation would be that once the load drops back to 0, the video should look fine (perhaps with some small window of time where it is still recovering). Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
