Hi Alan, On Tue, 2010-03-30 at 11:26 -0400, Alan Stern wrote: > What version of the kernel are you using? Have you tried using the > most recent release? I was using 2.6.30, and had tried 2.6.33 in the past just to see if that fixed it (no luck). I re-tested today using the latest git tree (2.6.34-rc2). Results below. > The fact that you're doing isochronous I/O to a full-speed device > through a high-speed hub complicates matters. This is the one area > where ehci-hcd still has significant problems. I was aware of this, but somehow I convinced myself I had this working on a similar setup last year. Though now I'm worried that it was actually still dying and just fellback to OHCI and so I didn't notice. > Each of these should produce 2 seconds of silence, right? 24*8000 > bytes = 48000*2 samples (with 2 bytes/sample) = 2 seconds (at 48000 > Hz, mono). Correct! > These messages indicate that the controller failed, but they don't say > what the reason for the failure was. True -- one other clue I just verified is that "lspci -vv" on the EHCI controller *after* the HC dies reveals "<TAbort+". Bogus DMA address perhaps? > Those events are merely cancellations of input URBs (although the > start_frame and actual_length values don't make sense). According to > the interval values, similar input URBs should show up in a full usbmon > trace every 64 milliseconds. Thanks for the explanation. I do see these every ~64 milliseconds in my trace. > Use the most recent kernel version you can. Post the complete usbmon > traces, not just the last few entries. To reduce the size of the > output, you could play only 200 ms of silence instead of 2 seconds. OK, I did all of this today on the latest kernel and got similar results. One difference I noticed is that the failure happens the first time I play audio with 2.6.34-rc2, rather than failing the second time. Full usbmon trace here (for 250ms of silence): http://xes-inc.com/sources/debug/usbmon-ehci-dies-2.6.34-rc2.txt other debug output + usbmon output http://pastebin.com/DwKWihUB Thanks - Nate Case <ncase@xxxxxxxxxxx> -- 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
