After some playing around with (reiser4 on top of raid0 on top of 2xSCSI on top of aic7xxx) as rootfs for new ubuntu installation, I decided to give up reiser4, seems like non-including it into kernel tree is the right idea, this filesystem may be cool and stuff, but it is buggy, and bugs are not fixed. I ended up with something that would be too expensive for me to fix in terms of time:
“kernel bug in inode.c:252, invalid operand 000”, is caused by inode deletion, as the traceback indicates, some googling revealed that this is because of non-proper truncating or something like that.
The funny thing is that it only appears during the boot, when I use normally (not as rootfs) it works just fine.
At this point I wanted to convert already created rootfs to old-good-well-working-yeas-i-know-slow-it-is ext3 w/o relaunching all the deboostrap stuff. Giving a try to convertfs (i hate this narod.ru ….), and getting an error:
============== Copying files ============== ls: /tmp/convertfs/fs2root: No such file or directory df: `/tmp/convertfs/fs2root': No such file or directory umount: /dev/loop7: not mounted
it appears that convertfs does some mount/unmount operations while it goes “copying files”, and i was trying to see what it does by ls’ing it’s /tmp/convertfs, blocking for some milliseconds all mount/umount operations.
now I am waiting for it to finish w/o ls’ing, so far so good.
Well, while I was typing it failed again with the same error, so I am not as smart as I thought, to the hell with it, I already wanted to launch debootstrap again, since I have to run, but after mounting the device I’ve seen there is only one fsimage file, an ext3 image, mounted, seems like convertfs did copy of everything there, but cleared the original… great, since the consistency of the copy is unknown, I launch a clean debootstrap…
P.S. there is a bug report on debian bugzilla which IMO is the same problem