Description of problem: Errors with unmount ignore_paths. Patch is in attachment. This patch added also additional paths for ignoring for distributions MagOS, MageiaOS. But this patch is needed for Mageia also.
Created attachment 3207 [details] patch with fix
Keywords: (none) => PATCH
Hardware: i586 => All
CC: (none) => mageia
Can you describe the problem in more depth? I don't like the idea of patching this file continuously and would rather find an alternative solution. I'd rather that these file systems are attempted to be unmounted and fail in systemd but then are handed back over the initramfs which should deconstruct them properly - there are already appropriate mechanisms in place to do that, so would rather use them if possible.
/mnt/live, /mnt/livemedia, /mnt/liveloop are used by Mageia-based distributions. Filesystems: aufs or unionfs. When systemd are unmounting this directories (when poweroff or restart), then there are many-many errors on display. That's why we needs patch for systemd. But you can find an alternative solution.
Created attachment 3214 [details] File system
On picture I show how file system is used. I hope it will help to you.
The errors from systemd shouldn't really be errors as such, just information stating that certain paths cannot be unmounted. Does this result in any actual delays or anything or is it just for the reasons of cosmetics that you want to hide such warnings? As far as I know, the same messages/warnings would be shown for /usr if you opt to have that on a separate partition.
Please, hide messages for cosmetics.
So I've had a think about this and I don't really like the idea of patching systemd to ignore these specific paths. In more recent systemd's there is a mount option to specify that certain mount points be handled in the initrd (x-initrd.mount). This is currently only used in the fstab generator, but IMO it should probably be extended to the unmounting logic also. Using this system (assuming it's fixed slightly upstream) should allow you to specify certain mount points that are basically not handled in this logic and thus delegated to something further up the stack. I'll mention the lack of x-initrd.mount in the umount logic to upstream guys, but in the mean time this is a wontfix I'm afraid. All the best.
Status: NEW => RESOLVEDResolution: (none) => WONTFIX