On Sun, Jul 20, 2008 at 09:26:12PM +0200, Frans Pop wrote:
> On Saturday 19 July 2008, Jérémy Bobbio wrote:
> > I have quickly reviewed all bug reports against these two packages and
> > no bugs seem to mention this issue. Either the relevant installation
> > reports were not sorted correctly or we can assume that it is a pretty
> > minor issue and that our users did work around it themselves.
>
> This is not about users complaining about it, but about internal
> consistency. The locking infrastructure was added by David as part of
> p-crypto and then also added for new LVM devices.
Oh… Ok. So devices and partitions locking are a feature that was added during
Etch development cycle. As partman-md has not really received the attention it
should have since, the code has not been updated to use locking as well.
Did I understood you correctly?
Do you see the absence of proper locking infrastructure as a blocker for the
changes in the initialization model? For partman-md in Lenny?
Let's agree that partman-md should really be reworked on during our next
release cycle ; as I have read many time, by its complete inclusion in
partman as a first task.
> I think I extended it to be used for pre-existing LVM volumes (but I may be
> wrong).
Quoting do_initial_setup() in partman-lvm/choose_partition/lvm/do_option:
[ "$RET" = true ] && log-output -t partman-lvm vgchange -a y
# TODO: We need to update and lock the devices that LVM just claimed
Despite its roughness, the patch proposed earlier locks Physical Volumes
properly after the initial device detection.
Cheers,
--
Jérémy Bobbio .''`.
lunar@debian.org : :Ⓐ : # apt-get install anarchism
`. `'`
`-
Attachment:
signature.asc
Description: Digital signature