I apologize, but git itself is still over my head, and I don't know
how to get bisect. I download the kernel tar.xz files and tell
Debian's make-kpkg utility to compile them for me.
But for what it's worth, the patch from Bryan Paluch that changes
IO_WATCHDOG_DELAY in drivers/usb/host/ohci-hcd.c from 250 to 275 msecs
either fixed or masked the problem (no problem scanning with the patch
applied). This would perhaps lend support to his theory about timing
changes having an unintended effect on the ohci drivers.
Quoting Mr. Paluch:
>> I suspect that the new timer wheel based timers in 4.8 have exposed a bug
>> in the ohci driver. Before the io watchdog timer was set to 250 with 20 msecs
>> of slack time. This is more than likely just a hacky work around [...]
Sure enough, I hacked, and it worked around.
On Fri, Oct 14, 2016 at 5:11 PM, Alan Stern <st...@rowland.harvard.edu> wrote:
> On Fri, 14 Oct 2016, Benjamin Staton wrote:
>> The Issue: Scanning (with xsane 0.999) while using kernel 4.8 (I
>> tested 4.8 rc1, rc5, rc7, rc8, and 4.8.1) fails when the scanner
>> becomes disconnected from the USB bus. Every time.
>> Scanning (with xsane 0.999) while using kernel 4.7.0 or earlier 4's
>> works without fail.
>> My scanner is known to lsusb as "Bus 006 Device 002: ID 04a9:2206
>> Canon, Inc. CanoScan N650U/N656U" and to Xsane as "Canon CanoScan
>> N650U/N flatbed scanner [plustek:libusb:006:002]".
>> Relevant lines from /var/log/kern.log at the time of the disconnect:
>> Oct 14 13:30:00 quiz kernel: [ 167.746817] ohci-pci 0000:00:12.0:
>> HcDoneHead not written back; disabled
>> Oct 14 13:30:00 quiz kernel: [ 167.746829] ohci-pci 0000:00:12.0: HC
>> died; cleaning up
>> Oct 14 13:30:00 quiz kernel: [ 167.746916] usb 6-1: USB disconnect,
>> device number 2
>> Oct 14 13:30:00 quiz kernel: [ 167.747618] usb 6-3: USB disconnect,
>> device number 3
>> Oct 14 13:30:00 quiz kernel: [ 167.747624] usb 6-3.2: USB disconnect,
>> device number 4
>> Oct 14 13:30:00 quiz kernel: [ 167.803283] usb 6-3.3: USB disconnect,
>> device number 5
>> This is on an asrock 970M Pro3 motherboard w/AMD FX-8300 CPU and 32GB
>> RAM running debian testing/9. Kernels built from kernel.org without
>> further patching.
>> Apologies for posting to wrong place (bugzilla) and happy to provide more
> Can you run git bisect between 4.7 and 4.8-rc1 to find the particular
> commit that introduced this problem? There haven't been any
> significant changes to the ohci-hcd driver during that period.
> Alan Stern
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html