On Thu, Apr 19, 2018 at 06:00:52PM +0530, Naresh Kamboju wrote:
> On 19 April 2018 at 17:39, Ben Hutchings <ben.hutchings@codethink.co.uk> wrote:
> > On Thu, 2018-04-19 at 16:42 +0530, Naresh Kamboju wrote:
> >> >
> >> > Can you try 'git bisect'?  I'll hold off on releasing 4.9.y until this
> >> > gets figured out.
> >>
> >> After reverting this patch, network started works on arm32 x15 device.
> >> d7ba3c00047d ("net: phy: micrel: Restore led_mode and clk_sel on resume")
> >
> > Does that also fix the hang during 'ip l'?  (If the hang only occurred
> > on the same systems that oops'd in the micrel driver, it could well be
> > because the task that oops'd was holding the rtnetlink lock.)
>
> Now 'ip l' works and the oops message is gone.
>
> - Naresh

There may be some confusion since this patch was in 4.9.94 and we did
not report this issue then. The way these failures presented themselves
made them a bit subtle (we didn't get any hard failures, just some
incomplete jobs - not an unusual occurrence on embedded boards) so we
didn't notice them in the last release. It was only after Naresh
manually investigated that this was discovered.

We'll think about how we can improve our process to make these much more
obvious in an automated way.

Dan