Description:

genfstab embeds device mapper numbers in swap lines. Since these numbers change on reboot (particularly between the install CD and booting from disk), they become incorrect on reboot, causing a difficult to track 60-second pause on every boot and resulting in swap being disabled.

Additional info:
* package version(s)

Confirmed in arch-install-scripts 8-1.

* config and/or log files etc.

Job dev-dm\x2d4.device/start timed out.
Timed out waiting for device dev-dm\x2d4.device.
Dependency failed for /dev/dm-4.
Job dev-dm\x2d4.swap/start failed with result 'dependency'.
Job dev-dm\x2d4.device/start failed with result 'timeout'.

Steps to reproduce:

Run genfstab when using lvm or dm-crypt or both for the swap partition.

For example, I am using lvm on a luks partition, and swap on a logical volume called /dev/mapper/decrypted-swap. If I run genfstab without the -U flag, I get output like the following:

# genfstab -p /
/dev/mapper/decrypted-root / ext4 rw,relatime,data=ordered 0 2
/dev/sda1 /boot ext2 rw,relatime 0 2
/dev/dm-4 none swap defaults 0 0

This is somewhat understandable in that /proc/swaps does not contain symbolic link paths for /dev/mapper/... the way /proc/mounts does, but it leads to a frustrating install experience. It took me several hours to understand why my system was stalling for 60 seconds every time I rebooted with some weird journalctl message about /dev/dm-4 timing out. (It was dm-4 when I installed because the install CD was using some device mapper numbers, but becomes /dev/dm-1 when I boot from disk.)

A fix to this would save people a lot of frustration. At a minimum, genfstab could check if the swap path matches /dev/dm-[0-9]* and if so issue a warning to stderr that device mapper numbers change and the user should either run genfstab -U or manually edit the swap line to contain a more permanent device name.

A slightly fancier fix would be to search /dev/mapper for a symbolic link pointing back to the appropriate /dev/dm-*, and to output that for the swap line.