On Tue, Sep 04, 2018 at 03:24:19PM -0300, Martin Pieuchot wrote: > On 31/08/18(Fri) 09:19, Tom Murphy wrote: > > [...] > > Here's the dmesg from a XHCI_DEBUG kernel before it crashes (it appears > > to loop quite a few times before the kernel protection fault kicks in.) > > You're now hitting a bug a ugen(4). Because you're detaching a device > while ugen_do_read() is sleeping for a transfer to finish. This is > definitively something that needs to be fixed. > > However... > > > xhci0: timeout aborting transfer > > xhci_abort_xfer: xfer=0xffffff047d8afe10 status=IN_PROGRESS err=CANCELLED > > actlen=0 len=24 idx=0 > > How do you generate this message?
I plug in the Android phone, I start up adb with 'adb start-server'. For whatever reason, the phone comes back as error: device offline (I can only actually connect via adb when the phone is booted into recovery mode, so not sure if this is a bug with LineageOS, anyway, that's not really relevant to the bug here) I physically unplug the phone and the kernel starts generating the xhci0: timeout aborting transfer messages in a loop. After a while the protection fault drops me into ddb. I don't think the transfer ever finishes because the device doesn't respond to it in the first place. So unplugging the device is what causes it to error. If I keep the phone plugged in, the kernel protection fault doesn't show up until I unplug it.
