Fwd: IDE HPA

[Greg Felix]

---------- Forwarded message ----------
From: Alan Cox <alan at lxorguk.ukuu.org.uk>
Date: 30.08.2005 11:05
Subject: Re: IDE HPA
To: Bartlomiej Zolnierkiewicz <bzolnier at gmail.com>
Cc: Greg Felix <greg.felix at gmail.com>, Oliver Tennert
<O.Tennert at science-computing.de>, linux-kernel at vger.kernel.org


On Maw, 2005-08-30 at 18:16 +0200, Bartlomiej Zolnierkiewicz wrote:
> HPA shouldn't be disabled by default and new kernel parameter ("hdx=hpa")
> should be added for disabling HPA (yep, people with buggy BIOS-es will
> have to add this parameter to their kernel command line, sorry).

Thats large numbers of systems. Large numbers of disks as strapped for
32GB and other clipping arrangements. With a vendor hat on thats
unworkable because

a) It will stop thousands of people installing their systems
b) Many users will get horrible corruption when they update the kernel
and their box explodes as the fs tries to write to areas of disk that
have vanished mysteriously.

(and we know all about this because ancient kernels had options for
doing this in the compile that burned people)

So its a very bad idea indeed. A boot option for not disabling the hpa
is possibly sensible for a few users who want that, or simply getting
them to fix their buggy user space app would be even simpler.

The only way I can see to truely automate it for most cases would be to
snoop the partition table if its MSDOS format and see if the table
matches the HPA clipped disk or the non-HPA clipped disk. If it matches
the HPA clipped disk then you know not to fiddle. Otherwise its either a
new disk, clipped by the 32GB jumper, non-x86 disk etc in which case you
might as well disable any HPA.

Alan