On Sat, 5 Apr 2014, Beralt Meppelink wrote: > Since an update to kernel 3.13 my system suffers from a bug, probably > in the EHCI scheduler. I have an em28xx based DVB USB stick which is > connected to a EHCI host and uses an isochronous endpoint to transfer > data from the device to the host. Initially I thought this to be related > to the em28xx module itself, and posted a bug report there [1]. > > Enabling some additional debug information showed the EHCI scheduler > overflow: > kernel: ehci-pci 0000:00:1d.0: request ffff8800ccc59000 would overflow > (8136+63 >= 8192) > > My best guess is that some URB's are not removed from the queue, but my > device driver keeps submitting them when opening a new stream. It does > not seem to happen during a stream (i.e. when resubmitting a completed > URB). It only occurs when opening and closing streams repeatedly (i.e. > when stopping and submitting URB's). So maybe kill()ing or stopping > URB's does not work properly. > > My hardware information is attached to the bug report. I'm using an > Intel NUC (Celeron), with a QS77 chipset. I cannot reproduce the bug on > a 3.12 kernel (Arch Linux; kernel 3.12.8-1). Looking at the recent > kernel I cannot find any updates since then, and nobody reported this as > a bug as far as I know. So it might be related to my specific hardware, > although someone else reported the same problem in 3.13 kernels in the > bug report [1]. > > [1] https://bugzilla.kernel.org/show_bug.cgi?id=72891 There were quite a few changes to the EHCI scheduling code between 3.12 and 3.13, although the affect of those changes should have been minimal. Can you perform a git bisection to find the particular commit causing the problem? Also, can you post a usbmon trace showing the problem? (See Documentation/usb/usbmon.txt.) 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
