I'm attempting to bring up a LVM logical volume during boot under kernel 2.6.8-gentoo-r7 w. lvm2. I get the following messages during the boot process: * Setting up the Logical Volume Manager... /sbin/rc: line 429: [proc id] Segmentation fault /sbin/vgscan >/dev/null This causes the subsequent invocation of vgchange -a in /etc/init.d/checkfs to be skipped and the target logical volume isn't activated. If I run vgscan after the boot is finished, vgscan doesn't segfault, but issues various messages, including a note re. the target logical volume needing activation, and exits normally. I've run into this in the process of trying to migrate a production desktop system from kernl v2.4 to v2.6, and it's a slightly involved process to move from my (very functional) desktop under kernel 2.4.20 to a running CLI login under 2.6.8, involving connecting a pair of test drives, which are part of a sofware RAID1 array, unmerging lvm-user and emerging lvm2, plus several other (scripted) changes, so I can't easily switch back and forth at will. But since I have to go there eventually, I'll be happy to try any tests.
I'm also running into a segfault if I set lvm.conf to scan /dev/md. I'm using devfs on this box, and have a RAID1 mirror set up, /dev/md0 -> /dev/md/0. I want to use this RAID array as an LVM physical volume, and can set it up, include it in a volume group and can create a workable logical volume in it; however, if I set up lvm.conf with the following: device { scan = "/dev/md" } and then run any of the lvm2 utilities, I get a segmentation fault. In particular, pvscan should pick up on the fact that /dev/md/0 has been set up as a LVM physical volume. Once this happens, there appears to be some sort of corruption somewhere, since if I remove the 'scan' line from lvm.conf subsequent invocation of most lvm2 utilities just hang indefinitely, and a reboot hangs on shutting down the Logical Volume Manager requiring that I use a hardware reset on the box.
Relevant information on exactly this problem can be found at: <http://www.tgharold.com/techblog/2004/06/gentoo-segmentation-fault-in-vgscan.shtml> This looks like a interaction between devfs, md and lvm2 that needs to be addessed at a kernel level.
The solution posted at <http://www.tgharold.com/techblog/2004/06/gentoo-segmentation-fault-in-vgscan.shtml> puts a band-aid on the problem here, too, although it doesn't appear to address the fundamental issue of what looks like adverse interactions between md, devfs and lvm2. At least _I_ can move on :-)
If the information on the TechBlog (URL cited above) is accurate, this bug is basically another manifestation of the problem behind Bug #61258.
*** This bug has been marked as a duplicate of 61258 ***
