On Thu, 19 Jun 2014 17:44:39 -0400 (EDT), Alan Stern wrote: > On Thu, 19 Jun 2014, Dennis New wrote: > > > On Thu, 19 Jun 2014 17:03:55 -0400 (EDT), Alan Stern wrote: > > > On Tue, 17 Jun 2014, Dennis New wrote: > > > > > > > On Thu, 12 Jun 2014 10:20:54 -0400 (EDT), Alan Stern wrote: > > > > > Dennis and Matteo: > > > > > > > > > > I promised to send both of you a patch changing the way > > > > > ohci-hcd handles hardware bugs. Well, it's finally ready for > > > > > testing. There's only a limited amount I can do on my own > > > > > machine, so now it's up to you guys. > > > > > > > > > > The patch was made against an early -rc version of 3.15, but > > > > > it will apply okay to 3.14 and maybe even earlier kernels. > > > > > > > > So, it happened again, after a random few days, this time with > > > > dmesg a bit different: > > > > > > > > [210263.872368] ohci-pci 0000:00:13.0: frame counter not > > > > updating; disabled > > > > [210263.872377] ohci-pci 0000:00:13.0: HC died; cleaning up > > > > [210263.872456] usb 2-4: USB disconnect, device number 4 > > > > [210264.872344] timeout: still 2 active urbs on EP #3 > > > > > > > > The frame counter indeed was not updating. I notice that this > > > > time, it says there are "still 2 active urbs", whereas before > > > > it always said 3. I also noticed that before I would still get > > > > "USB disconnect" messages when I unplugged my device, but not > > > > this time. It looks like there was a more thorough disabling of > > > > the usb/ohci system this time. > > > > > > That "timeout:" message came from the snd-usb-audio driver. I > > > don't know if the number of URBs is especially significant, > > > probably not. The lack of "USB disconnect" messages is expected, > > > because as you realized, the OHCI driver was disabled. > > > > > > Anyway, it looks like things are working better than they did > > > before. Your system isn't crashing when the problem occurs, right? > > > This might not be the solution you wanted, but I think it's the > > > best we can do. > > > > My system didn't really crash before either -- just the ohci. I > > forget if the process that was using the snd-usb-audio driver > > remained hung in the background, but commands like "lsusb" do not > > work. I am forced to reboot to get ohci-usb working again :\. Isn't > > there a way to reset the ohci subsystem? :s > > And with the patch, lsusb and the sound processes do work after the > fault? I want to make sure, since you didn't say explicitly. > > You can restart the OHCI controller by this sequence of commands: > > echo 0000:00:13.0 >/sys/bus/pci/drivers/ohci_pci/unbind > echo 0000:00:13.0 >/sys/bus/pci/drivers/ohci_pci/bind > > However, some USB devices can't recover from this sort of event until > they have been unplugged. I tried it with two flash drives plugged > in; afterwards one of them worked and the other didn't. With the new patch, mplayer is able to close gracefully, with pcm errors like "No such device". This time, my dmesg was slightly different, with an "HcDoneHead" message: [139650.866080] ohci-pci 0000:00:13.0: HcDoneHead not written back; disabled [139650.866089] ohci-pci 0000:00:13.0: HC died; cleaning up [139650.866166] usb 2-4: USB disconnect, device number 3 [139651.866074] timeout: still 2 active urbs on EP #3 However, lsusb does not work, and I cannot kill it -- it remains in D (uninterruptible sleep) state. I am also unable to rmmod the snd_usb_audio module -- rmmod gets stuck in D state too. When I tried echoing (0000:00:13.0) to ohci_pci/unbind, although dmesg says: [140078.654462] ohci-pci 0000:00:13.0: remove, state 1 [140078.654480] usb usb2: USB disconnect, device number 1 ... the echo command (a bash internal) does not terminate, and also gets stuck in D state. Although the symlink to 0000:00:13.0 does get removed from that directory. But trying to echo this back to ohci_pci/bind doesn't work: "write error: No such device" Even though 0000:00:13.0 is still listed in /sys/kernel/debug/usb/ohci. Try to "cat" /sys/kernel/debug/usb/devices also fails, and cat gets stuck in D state too. My usb thumbdrive also no longer works after this occurs, although I'm pretty sure it uses EHCI. I'm forced to reboot to get usb working again :s. -- 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
