Hi, I installed Gentoo Linux on a HP Proliant DL360 with a Compaq Smart Array controller configured with a mirrored disk. The full mirrored system disk was placed under the control of LVM. All went smooth, and I installed LILO successfully. (still working under the stable gentoo-sources 2.4 kernel) When I upgraded to the 2.6 kernel with device-mapper and LVM2, I found out the hard way that LILO doen't know about device-mapper. I reported the problem to the device-mapper development mailing list, and got a reply from Christophe Saout <christophe@saout.de> in which he acknowledged the problem and provided me with a patch (complete with a modified ebuild script for the lilo-22.5.8-r1 package! I applied the patch, emerged lilo-22.5.8-r1, and was finally able to install LILO under the 2.6-kernel with device-mapper. Please accept his patch into the stable LILO as soon as possible, because without it you can't boot a LVM2-based system. Thanks. I don't know how to attaches ebuild scripts, patches and stuff to a bug report. I'll first commit this entry now, and then try to find out how to do that. Regards, Toon.
Created attachment 20961 [details] This is the modified ebuild script by Christophe Saout <christophe@saout.de> Two more attachments coming up. Regards, Toon.
Created attachment 20962 [details] This is the digest file that goes with the modified ebuild script. One more attachment coming up.
Created attachment 20963 [details, diff] This is the patch to LILO That's it! All credits go to Christophe Saout <christophe@saout.de> Regards, Toon.
On more thing, you can find my original detailed bug report at: http://lists.sistina.com/pipermail/dm-devel/2003-November/000317.html
Just guessing on the maintainer.
Christophe Saout made this remark when he sent me the patch: "I'm sure someone can come up with an ebuild file that sets the define depending on the availability of libdevmapper or something." So this is probably something extra that has to be done.
max, you handle lvm2 it looks like -- is lvm2 exclusively a 2.6 thing? in other words can we just do the switch on kernel version 2.6 to apply this patch?
actually, on warpzero's suggestion, I'll use a local USE flag called devmapr to enable this patch
I use device-mapper/lvm2 on custom 2.4 with the dm patch (and some other stuff). Enabling this via a use flag is probably the best approach, though I am not entirely comfortable with a use flag named 'devmapr'. How about lvm2 or dm? But in any case, that is for the maintainer of lilo to decide.
I think this bug can be closed.
I agree, this can be closed. Device mapper functionality has been in the lilo ebuild for two release cycles now.