My Kernel bug is celebrating 2 years. Can you help me fix it?

[Thomas Petazzoni]

Dear Peter Senna Tschudin,

On Tue, 4 Mar 2014 22:26:37 +0100, Peter Senna Tschudin wrote:
> I have reported a bug more than two years ago and it is still
> affecting me. The bug report gives some information:
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=787299
> 
> I have tried basic debug instructions from:
> https://www.kernel.org/doc/Documentation/power/basic-pm-debugging.txt
> 
> And everything works as expected when:
> # echo freeze > /sys/power/state
> # echo disk > /sys/power/state
> 
> I have asked for help for fixing it:
> https://lkml.org/lkml/2013/11/1/186
> 
> But I don't have a serial port. How can I debug this issue without a
> "real" serial port? Or what else can I try? How can I explore the hint
> about the problem only happening with VT-d enabled in BIOS? How can I
> explore the hint about the problem not happening if the option
> nox2apic is passed to the Kernel?

Something I would try in this situation is to boot with mem=<some value
smaller than the amount of RAM>, and then have the kernel write some
debugging informations manually at a fixed location in RAM that has
been reserved by lowering the amount of RAM using mem=. Then, when you
reboot, you can dump what has been left in this memory location. Of
course this requires that 1/ this memory location is not overwritten by
the BIOS/bootloader and 2/ that you can do a warm reset to not loose
the contents of the memory. Since I don't do much x86 kernel hacking,
I never had to do that on x86, but I've used this trick a few times on
ARM platforms.

If that works, then it means you can put some debugging details all
over the kernel to find where things hand exactly during the resume
process.

Another embedded trick is to find a LED that you can easily turn
on/off. It is very useful to see if you reach a given portion of code
or not.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com