[parted-devel] [PATCH] GPT & BIOS Boot partition

[H. Peter Anvin]

Robert Millan wrote:
> 
> A dedicated partition is, in fact, the only way to get back the static
> 63-sector area we used to have with DOS labels.
> 
> Of course, you have other options;  you can put your code in a filesystem
> and hardcode its offset (expect some fun with xfs), or you can assume your
> filesystem of choice reserves space at the beginning of its partition.  GRUB
> can do both but neither of them is as good as embedding the bootstrap code
> in sectors 1 to 63 like it used to (AFAICT).
> 

Sectors 1-62 were never a good idea to use, simply because there was no 
management of it.  Plus, you couldn't assume they were there.

> So, let's summarize that we have two options:
> 
>  1- We can just assume this partition type is "for use by BIOS-based
>     bootloaders as they see fit".  This is what GRUB already does, but also
>     what other bootloaders can do if they find that useful.
> 
>     But, of course, if you previously ignored the 63-sector gap and just jumped
>     from MBR to PBR, this option doesn't provide any advantage to you, only the
>     second one does.
> 
>  2- We can try to standarize a boot mechanism on top of it.  My GPT-MBR work
>     is aimed in that direction.  Although there are limitations as you pointed
>     out (and others you didn't mention), but overall I see this as a bonus
>     once you have a partition type that bootloaders can use in their own
>     specific way.
> 
>> especially with Redmond OSes seemingly going a different way.
> 
> What do you mean?  Last I heard, their way was not supporting GPT on BIOS at
> all :-)

They do; you have to have MBR entries for your boot partitions, but then 
they pick up other partitions from the GPT.

Boot partition is obviously limited to 2 TB.

	-hpa