[parted-devel] RfD: Do not clobber the BootCode in GPT PMBRs

[Joel Granados]

On Wed, Feb 25, 2009 at 05:19:45PM -0800, H. Peter Anvin wrote:
> Joel Granados wrote:
>> Just read this going throuhg 3 months of backed up parted mail.  You
>> make a very good point but I think going agains the spec is not the way
>> to go.  This would increase the possibility of incompatibilities with
>> other UEFI spec implementations other than what parted has (which I'm
>> not al all sure it complies 100% :)
>>
>> For me, this is a big question mark ATM.
>
> It's not going against spec.  The UEFI spec doesn't specify what goes in  
> that area, and it is *guaranteed* to break GPT booting on BIOS  
> platforms, which is about as disastrous as you can get.
>
> 	-hpa
>

Just took a look at the code and there is nothing in the history and I
found the clobber from gpt, pretty much the same as it was from the
first commit.

Its best to change the clobber concept in gpt.c from "lets zero
everything" to "lets zero main header, backup header, and create a
minimum PMBR".  Here minimum PMBR means leave everything as is except
the partition table, which, according to UEFI spans from byte 446 to
446+(16*4)

Comments?

-- 
Joel Andres Granados
Brno, Czech Republic, Red Hat.