Swapfile in fstab disables GPT automount of swap partition

[cvlc12]

systemd version the issue has been seen with

256.8

Used distribution

Arch

Linux kernel version used

6.12

CPU architectures issue was seen on

x86_64

Component

systemd-gpt-auto-generator

Expected behaviour you didn't see

Swap partition (encrypted) properly configured for automount:

# swapon
NAME      TYPE      SIZE USED PRIO
/dev/dm-2 partition  16G   0B   -2

Add an additional swapfile in fstab:

# cat /etc/fstab
/swap/swapfile none swap defaults 0 0

# swapon --all
# swapon
NAME           TYPE      SIZE USED PRIO
/dev/dm-2      partition  16G   0B   -2
/swap/swapfile file        4G   0B   -3

Unexpected behaviour you saw

Reboot, and the swap partition is not automounted anymore:

$ swapon
NAME           TYPE SIZE USED PRIO
/swap/swapfile file   4G   0B   -2

Why does adding a swapfile disable swap partition automounting ? If this is the desired behavior, I believe it's not documented anywhere?

Steps to reproduce the problem

No response

Additional program output to the terminal or log subsystem illustrating the issue

No response

[poettering]

So the philosophy so far was that if swap is manually configured we won#t interfere with that, and not insist on adding the auto-detected one.

I am not convinced we really should change that, the logic behind it makes a ton of sense to me: pick what you want: manual or automatic, but no hybrid.

[poettering]

or in other words, we should probably document this behaviour but not change it, and instead recommend that if people configure one swap they should configure all swap.

[cvlc12]

On the other hand, couldn't it make sense to have a regular automounted swap partition, and then being able to augment it with a swapfile when needed?
The problem with the partition being the fixed size I guess.

[poettering]

well, people wan't all kinds of things, but i find this an niche case, that is very easily fixed by the folks: just make your fstab complete and the issue is gone.

[cvlc12]

Sure. In that case just better docs would do I guess.

[YHNdnzj]

But this would induce a great inconsistency in the future with my plan for #27247. For the auto resume to be fully automated, gpt auto swap must be decrypted in initrd already. It's very weird if we do that, but then refrain from mounting it after switch-root because other swaps are specified.