[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: d-i status wrt i386 & amd64 EFI machines



On Wednesday 28 May 2008, Julien BLACHE wrote:
> > There already is code to recognize Intel Macs as a separate subarch
> > which allows to use separate partitioning recipes that could include a
> > EFI system partition, but currently that means the existing one would
> > probably be lost.
>
> Which is not a good idea :) So if Mac && 1st partition is FAT32 ->
> mark the partition as being the EFI partition (can be generalized to
> all the EFI machines I think).

The real problem here is that partman-auto will always start from scratch 
when partitioning a drive (except when "largest free space" is used, but 
that only works if the free space is already there [1]).
The only way to keep an existing partition is by using a custom recipe and 
setting its size to _exactly_ the current size and doing the same for all 
partitions that come before it. Any deviation will mean that start of 
partition and actual start of file system will no longer be the same.

This is further complicated by the size calculations and rounding done by 
partman and libparted. Throw LVM and crypto overheads in the mix and you 
have a recipe for disaster...

There is no provision to say "partition the drive, but leave the first one 
alone". Especially for the first partition (if it is #1 and at the 
beginning of the disk) that should be possible to implement though by 
coding what I described in [1] (the recipe used should then of course not 
include an EFI partition).

I'm now going to move this thread to /dev/null and just let you get on with 
things :-)

Good luck,
FJP

[1] A trick I sometimes use is to use manual partitioning to delete unwanted
partitions and then start guided partitioning from partman's main menu to 
partition the freed up space.

Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: