On Sunday 12 December 2004 4:28 pm, Rob Browning wrote: > David Brownell <[EMAIL PROTECTED]> writes: > > > That last point says to me there are some strange timing > > issues lurking. But transfer rates under 10 Mbyte/sec > > have so far seemed best explained by drive adapter issues. > > In case it might help diagnose, I can provide one more bit of > information. > > If I place an older 5400rpm 20GB Maxtor in the enclosure, create one > large partition, and then execute "mke2fs -j /dev/uba1", the mke2fs > command takes 5 minutes or more to run, and the CPU remains pegged at > 100% throughout. This is on an Athlon 2500 with 1GB RAM.
Sounds like you should use usb-storage instead of "ub". When I make time to test ehci with disks, one of the basic tests is mkfs with badblocks. That's even more I/O intensive, and I've seen hundreds of Kbytes queued at once. FYI, 2.6.0 did not handle the "mkfs -cc" well, but "mkfs -c" behaved; while lately I've seen no issues with "mkfs -cc" either. The way "ub" does I/O is page-at-a-time ... one IRQ per page, and probably not the max of eight per millisecond. (There's an IRQ latency module parameter for EHCI that could shrink that.) No I/O queue to speak of. The way "usb-storage" does it is scatterlist-at-a-time .. one IRQ per scatterlist, which for a (small in that context) 128 KB scatterlist means one every 2-10 ms; 3-4 ms is "good", older drives may be nearer the high end. For a 512 Kbyte scatterlist, the CPU load would be even less. That'd explain why the CPU load is higher with "ub". Also consider that USB I/O wait time can be accounted wierdly, especially by "uptime"; I'm not sure what you mean by "100% throughput". - Dave ------------------------------------------------------- SF email is sponsored by - The IT Product Guide Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now. http://productguide.itmanagersjournal.com/ _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
