Hi- I found a "bug" in the nilfs2 mount command, which unfortunately will be made moot with the advent of baselayout-2. checkroot very intelligently checks to see if a filesystem *wants* to be mounted rw, rather than whether it's mounted rw. It looks at the output of 'mount -vf -o remount /' to see if it's to be mounted rw, then checks carefully to see if it can be written to in order to be sure. Ordinarily, mount outputs something like: /dev/sda3 on / type xfs (rw,noatime) even when the filesystem is mounted readonly. If a nilfs2 filesystem is mounted ro, mount.nilfs doesn't spit out the required line that awk expects, so the script fails. When a nilfs filesystem is mounted rw, mount itself will spit out the required line. The outcome of this is that the root filesystem in question never is mounted read-write. What seems to be necessary is to either hide mount.nilfs2 or change its output so that when called (by mount) it spits out the same information that mount does. Again, baselayout 2 will fix this by doing away with the check (as far as I can tell, which is absolutely fine). If I figure out how to make mount.nilfs2 output the required information, I will update this bug with a patch (and notify upstream). -Robin K.
A better fix is to specify the filesystem type of / as rootfs, not nilfs2 (mount will only call mount.nilfs2 if the filesystem is specified to be of type nilfs2 in fstab). Given both these kludges, you may mark the bug closed. It's going away anyway.
(In reply to comment #1) > A better fix is to specify the filesystem type of / as rootfs, not nilfs2 > (mount will only call mount.nilfs2 if the filesystem is specified to be of type > nilfs2 in fstab). > Given both these kludges, you may mark the bug closed. It's going away anyway. > Using "rootfs" worked perfectly for me. Can this be added to the ebuild for the uninformed? Best regards
hm, is this still valid? It seems /etc/init.d/root has changed its behavior
*** Bug 337577 has been marked as a duplicate of this bug. ***
