Freeze for Test3 (and string/feature freeze)

[Alan Cox]

On Sat, Mar 17, 2007 at 03:43:06PM -0400, Warren Togami wrote:
> >Fallback - I don't have one.  That is why its currently top of my problem
> >list. With no HPA support we get folk who cannot upgrade or whose upgrade
> >blows up part way (more likely) and they need to reinstall.
> 
> Is there also risk of blowing away or corrupting whatever data is in the 
> HPA?

If you have a system which is using the HPA for drive clipping to avoid
BIOS bugs with old BIOSes, or is setting it but it was cleared and then
repartitioned for Linux then an upgrade without HPA resetting will begin
happily munching away and then blow up when it touches the first sector of
an update that is in the HPA region.

You can (usually) detect this because the partition table end will point
outside the apparent disk size.

In the other direction if you remove the HPA and don't repartition then
you'll not create anything that leaks out and you can detect the HPA size
and boot size by querying the drive initial id data (which unlike drivers/ide
actually is the initial data).

My test code requires booting with libata.ignore_hpa=1 to remove the HPA,
I'm not sure what the right default is.

Alan