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

Re: d-i's guided LVM setup

On Sunday 24 June 2007 20:47, Mike Hommey wrote:
> I did some etch installs recently, and noticed the guided LVM setups
> lead to having the LVM physical volume setup in a logical partition.
> Is there a rationale to not use a primary partition instead ?

This is possibly related to #353257.

I also see that swap partitions are not marked as '$primary{ }' in the 
auto-partitioning recipes [1], which could also have an effect.
Hmm. For recipes where more than four partitions are created, we cannot 
mark all of them as '$primary{ }' anyway because of regular partitioning 
and as swap is created inside the VG this is probably not relevant.

There is this code in partman/partman-auto-lvm/auto-lvm_tools.sh:
   # Creating envelope
   scheme="$normalscheme${NL}100 1000 1000000000 ext3 method{ $method }"

Unless I'm very much mistaken, the '100 1000 1000000000 ext3 method{ 
$method }' defines the partition in which the LVM will be created.
This does not include a '$primary{ }' tag, but including it may be 
problematic as not all disk labels support primairy/extended partitions 
(although possibly partman may ignore such tags in that case and 
otherwise it should be possible to test if they are supported).

It could also be problematic because the recipe could in theory already 
define $MAX_PRIMARY_PARTITIONS - 1 primary partitions.

Another option could be to explicitly include the "LVM partition" in the 
recipes (to be ignored for regular partitioning of course).

Some serious analysis of this issue would be much appreciated. (And so 
would patches :-)

Note that the whole issue is slightly complicated by guided LVM-on-crypto, 
which is an extension of guided LVM.


[1] E.g. partman/partman-auto/recipes/atomic in D-I SVN.

Attachment: pgppNuRZ8ItiR.pgp
Description: PGP signature

Reply to: