Bug#624268: [squeeze] include important changes from 2.6.32.39

[Ben Hutchings <ben@decadent.org.uk>]

On Tue, Apr 26, 2011 at 06:17:08PM -0600, dann frazier wrote:
[...]
> > 6f396d4a x86, cpu: Fix regression in AMD errata checking code
> 
> fixes hangs on some K8 systems

I believe this is a fixup for these:

[...]
> > d4274252 x86, AMD: Set ARAT feature on AMD processors
> > 7a3b25c0 x86, cpu: Clean up AMD erratum 400 workaround
> > bba4804e x86, cpu: AMD errata checking framework
> 
> These avoid switching to HPET timers in deep C states on AMD
> CPUs. Though obviously an improvement, I'm not really sure what makes
> these candidates for stable - unless people have been seeing hangs w/
> the broadcast timer code on AMD systems?

I'm dubious about the value of these quite large changes, as you may
have seen in discussion on stable-review.  Given that no-one was able
to explain what bug is fixed, and there is a functional change hidden
in the 'cleanup' patch which again no-one was able to explain, I would
favour reverting these.

> > 17ebcafe x86: Fix a bogus unwind annotation in lib/semaphore_32.S
> 
> Not sure what the end user value is here... better kernel backtracing?
[...]

I have no idea either.

Ben.

-- 
Ben Hutchings
We get into the habit of living before acquiring the habit of thinking.
                                                              - Albert Camus