Hi, The way "/sbin/splash-functions.sh" plays with mounting/moving of ${spl_tmpdir} and ${spl_cachedir}, combined with the "checkroot" init script, leads to a stale entry in /etc/mtab at the end of the boot sequence. What happen is the following: 1) splash_cache_prep() mounts ${spl_tmpdir}, and moves it to ${spl_cachedir}. /etc/mtab is not modified because of the "-n" mount option. 2) /etc/init.d/checkroot starts, and fill /etc/mtab with /proc/mounts entries (including ${spl_cachedir}). 3) splash_cache_cleanup() moves ${spl_cachedir} back to ${spl_tmpdir}, with the "-n" option again. From here, the /etc/mtab entry about ${spl_cachedir} is wrong. 4) splash_cache_cleanup() unmount ${spl_tmpdir}, which sure doesn't remove the wrong /etc/mtab entry about ${spl_cachedir}. It's not a big issue to have this stale mtab entry, but it's sometimes annoying, like for instance in "df" output where it adds an extra line: % df -h ... cachedir 15G 9,5G 4,3G 70% /lib/splash/cache ... (the reported sizes are actually those of my / partition) I suggest you simply remove the "-n" option from the "mount --move" in splash_cache_cleanup (i will attach a patch), because: - during a normal boot sequence, the FS is mounted rw at this step (at least, that's what happens on my system) - even if it was not, it wouldn't prevent mount to do his job and to properly move the mount point: i've just made some tests with / (and thus /etc) remounted ro, and mount didn't complain at all. It didn't even return the error code 16 documented in the man page... Weird, but a good news from the POV of this bug :) And if you're not confident with completly removing the "-n" option, you can still use something like "mount --move ... || move -n --move ...". Note: I'm using media-gfx/splashutils-1.1.9.10, but iirc, this bug has been there for a long time, and thus in some previous versions too.
Created attachment 79026 [details, diff] splash-functions_mtab_cleanup.patch
Fixed in splashutils-1.3. Thanks for the patch :)
