On Fri, 23 Sep 2011, Pitt, Jason N wrote: > Is there something I could look for in the usbmon dump to track down > what is causing this? I'd really like to figure out if there is a > way around this limitation (or what is limiting the system) without > dividing the problem among multiple machines in order to conquer it. > This is an 8 core machine with 32 Gb of RAM and a huge number of USB > controllers, it seems nuts that it can't record 12, 640x480 video > streams at once.... You're right, this does not look like a hardware limitation at all. It looks more like a driver bug. If you're interested, this part of the trace is the smoking gun: ffff8804128f4800 409271004 C Zi:6:002:1 0:1:1848:0 32 0:0:0 0:2460:0 0:4920:0 0:7380:0 0:9840:0 0 ffff8804128f4800 409271010 S Zi:6:002:1 -115:1:1848 32 -18:0:2460 -18:2460:2460 -18:4920:2460 -18:7380:2460 -18:9840:2460 78720 < ffff8804128f7800 409271021 C Zi:6:002:1 0:1:1880:0 32 -18:0:0 -18:2460:0 -18:4920:0 -18:7380:0 -18:9840:0 0 It shows an URB completing immediately after the previous URB, before any of its transfers could have taken place over the bus. Evidently the driver thought the URB was complete at a time when it was still active. There have been a few changes in that part of the ehci-hcd driver recently. Can you try running with a current 3.1-rc kernel? It may well make a difference. If the problem still occurs, I can send you a diagnostic patch for the current kernel to try and find out what's going wrong. BTW, your lsusb output shows that one of the cameras is different from the others. In the lsusb listing it's the camera on bus 13 (the one made by Feiya), but in the usbmon trace it appears to have moved to bus 4. As it happens, that camera has a bug in the descriptor for endpoint 3; the bInterval value uses the full-speed encoding instead of the proper high-speed encoding. However it is not contributing the problem you see -- the problem in the usbmon trace affects the camera on bus 6. There are no apparent problems with bus 4. 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
