Bug#586143: Doesn't boot: Unaligned pointer 4c191bea

[Colin Watson]

On Sat, Jun 19, 2010 at 01:07:06AM +0200, Andreas Jacob wrote:
> Could it be that grub-install, or your patched debug version did
> remove the stage2 file? When I execute a grub-install, it seems to
> touch or to rewrite almost all files under /boot/grub. May be it also
> deletes the stage2 file?

No, it's very specific about what files it touches and it doesn't touch
stage2.

> Unfortunately i have no backup from /boot where i can look if these
> file were there before. But I did the grub update within a
> safe-upgrade, shuted down my laptop, an 2 hours later on rebooting
> there was the described behaviour. Meanwhile it's impossible for me to
> reproduce the error.
> 
> As i can read here, some other people have also no stage2 file but the
> same behavior. This couldn't be contretemps.

As a matter of fact it could well be.  There are layers to this problem.
The direct cause is that the core image and /boot/grub/*.mod have got
out of sync and can no longer call each other's code directly in the way
they expect to be able to.  However, there are a variety of reasons why
that could happen, and in fact none of them are really particularly new!
We get a batch of these bug reports each time the interface between
GRUB's core image and modules changes in any significant way.  #554790
is essentially a cluster of the same kinds of problems; the symptoms are
different, but that's not very relevant.

The fact that most people this time round are seeing this due to the
presence of /boot/grub/stage2 is very interesting, because that's a
separate and easier-to-fix problem: I have a fix in progress already,
but of course I need to construct tests for it locally before unleashing
it on the world.  I want to keep this bug report for just that
situation, because bug reports should ideally be separated by cause
rather than by symptom if there's any hope of tracking and fixing them
in an organised way.  After all, the purpose of bug reports is to
improve the software, not to be forums.

We've interpreted #554790 and its merged bugs as the situation where the
core image and modules get out of sync due to installing GRUB to the
wrong device, which can happen quite easily on multiple-disk systems.  I
have code for this but it still has known incorrect behaviours and needs
work; but I absolutely intend to get this landed for squeeze.

Sometimes it's just user error, where somebody has installed GRUB to a
partition but the BIOS is actually booting from a separate installation
in the MBR which has wound up orphaned.

But what I'm interested in is making sure there aren't any situations I
haven't thought of yet.  This bug was a brand new cause for a problem I
thought I'd fully classified, and that was fascinating and useful.
Perhaps you have a different brand new cause - or perhaps you
essentially have #554790.  Without analysis, I can't tell, and I don't
want to use this bug for an analysis of your problem because we have
already established that it seems to be qualitatively different in cause
despite having similar symptoms.

Regards,

-- 
Colin Watson                                       [cjwatson at debian.org]