[Arm-dev] Gigabyte MP30-AR0, nomodeset, and retaining output on serial console

Mon Jun 6 08:23:03 UTC 2016
Richard W.M. Jones <rjones at redhat.com>

On Thu, Jun 02, 2016 at 11:07:25PM +0100, Gordan Bobic wrote:
> 4.4.12 kernel introduces a regression that seems to break this. When
> booting 4.4.12 with:
> 
> earlycon=uart8250,mmio32,0x1c020000
> 
> it locks up solid before anything appears on the console output.
[...]
> Has anybody got any ideas on any extra troubleshooting or options to
> be passed for initializing the UART that would produce a more
> meaningful bug report? The cynic in me, supported by past
> experience, suspects that if I file a bug on kernel.org it won't
> even get triaged for months, if ever, especially since the problem
> is manifesting on hardware few have access to.

Although I don't think the underlying bug is related, I'll just note
there is a similar thing happening with RHEL 7.3 kernels (which are
4.5.0 plus lots of backports from upstream):

  https://bugzilla.redhat.com/show_bug.cgi?id=1337705

The difference being I was able to get a stack trace using the same
earlycon line you are using, and it turns out to be caused by PCI
probing, I think reading from a non-existent config space address or
device.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-builder quickly builds VMs from scratch
http://libguestfs.org/virt-builder.1.html