Bug#611493: kernel panics or freezes helped but not eliminated by warm reboot or "rootdelay=20"

[Jonathan Nieder <jrnieder@gmail.com>]

Hi Pascal,

Pascal BERNARD wrote:

> I forged another initramfs using one
> from a Mint distribution and replacing the modules with the one of my
> "stable" kernel.
>
> It is much more stable. It failed once. I suppose it can be with the
> normal init which has not been "upgraded".

I'm kind of confused.  What was attaching an initramfs to this bug
supposed to accomplish?  I assume you are still talking about kernel
panics/freezes; does it still produce GPFs when, and only when,
performing a cold reboot instead of a warm reboot?

A general protection fault like this is by definition a kernel or
hardware bug.  It is not a bug in the initramfs generation tool.

Learning that some factor makes things feel more stable is very useful
as a symptom, but much more useful would be to actually find and fix
the problem. :)  So ideally we want to have (1) a simple, reliable
reproduction recipe and (2) a set of symptoms that result from it, so
we can come up with (3) a working theory about what is going wrong,
and then check whether it is true and address it.

If you have such a reliable recipe, the best thing you can possibly do
is to try a 3.x kernel from sid or experimental (note: this does not
involve installing any additional packages from outside squeeze except
linux-base and initramfs-tools).  If it fails, then we can get
upstream developers' feedback more easily.  If it succeeds, then we
will know there is a fix already available and can try to track down
which patch that is in order to apply the patch to squeeze.

Sorry to be so meta.  Still, hope that helps,
Jonathan