After upgrading a Mandriva 2010.2 x86 system to Mageia 2 alpha 2, I noticed the following message logged: fedora-storage-init[1225]: /sbin/kpartx: invalid option -- 'u' Sure enough, /lib/systemd/fedora-storage-init contains a call "/sbin/kpartx -u -a -p p", but kpartx-0.4.8-18.mga1 (which is what's included in alpha2) says: usage : kpartx [-a|-d|-l] [-v] wholedisk with no mention of -u. The offending invocation is a part of initscripts-9.34-2.mga2, but it's probably only kpartx that needs upgrading.
I just downloaded multipath-tools-0.4.9 (the latest release) and that doesn't support -u either. I took a look at the kpartx in Fedora 16, and it turns out the -u option (which sets the force_udev_rules flag) was added by them (in one of 107(!) patches to multipath-tools). So, the right solution isn't immediately obvious.
CC: (none) => mageiaDepends on: (none) => 2120
thomas, rtp, blino, any idea ?
CC: (none) => arnaud.patard, dmorganecAssignee: bugsquad => tmb
Seems Fedora have a few patches applied. They also roll their own tarball I think... http://pkgs.fedoraproject.org/gitweb/?p=device-mapper-multipath.git;a=tree Seems the git version of multipath-tools supports the -u option (update): http://git.kernel.org/gitweb.cgi?p=linux/storage/multipath/hare/multipath-tools.git;a=summary Perhaps we just need to update our version to git master and see if any of the Fedora patches make sense for us too?
CC: (none) => mageia
patch added for -u flag in multipath-tools-0.4.8-18.1.mga2
Can this be closed now? I know there is still a problem with softraid for intel stuff, but that's not specifically the same as this bug.
Heh, The multipath-tools-0.4.8-18.1.mga2 I pushed out never applied the patch wich I noticed when doing the upgrade to 0.4.9 :/ Anyway, I fixed so the "-u" flag now works in multipath-tools-0.4.9-1.mga2 And that is now capable of booting the system and mounting the partitions on the first raid set, so for this part it is ok. It still fails to activate the second raid set where swap is located, but that is another bug... So closing this one
Status: NEW => RESOLVEDResolution: (none) => FIXED