Parted CHS news...

[Andrew Clausen]

Hi all,

I've been working on a new CHS fix.  It's a fair bit of work, and the
end isn't really in sight yet.

Firstly, does anyone know anything about sector sizes != 512?  I've
never seen one in the wild.  How do we find out what the BIOS sector
size is?  Is it correlated to the hardware sector size reported by
the BLKSSZGET ioctl?  Should we be dragging this into the CHS mess?

Secondly, I found someone who could reproduce the CHS problem.  I
suspect this is what happened:

(1) Windows was installed to use CHS.  There is some UseBIOSToBoot
option in the Windows mass-installer that OEMs use.  It's buggy, and
I suspect some OEMs thought they were disabling this CHS rubbish, but
in fact were enabling it.

        http://support.microsoft.com/default.aspx?scid=kb;en-us;282191

(2) The Windows partition started at the start, and ended beyond the
1024 cylinder mark.  There are multiple consistent geometries with this
limited information - Parted isn't smart enough to find one.  Therefore,
Parted was unable to deduce the BIOS geometry, and didn't find a
consistent one.

(3) Since Parted had an inconsistent geometry, it changed the partition
table and broke Windows.

I don't know why I didn't think of this before.  Now I can see what
causes the problem for many people.  In any case, the current CVS
solution (which I will replace soon) does not break anything in this
situation.

Cheers,
Andrew