feat: gated fsync kernel (matches coreos-stable)

[m2Giles]

Thank you for contributing to the Universal Blue project!

Please read the Contributor's Guide before submitting a pull request.

[bsherman]

A little nitpicky, but why is this a negative? Isn't it a bit more clear to:

if [[ "${{matrix.kernel_flavor}}" == "fsync-gated" ]]; then
    export KERNEL_IMAGE=fsync-kernel
else
    export KERNEL_IMAGE=${{ matrix.kernel_flavor }}-kernel
fi

[m2Giles]

Only reason was that's just how I was thinking. I agree that not using the negative is better

[bsherman]

My only comment is I struggle with the name 'fsync-gated' ... it does signify a delayed kernel release, but it doesn't connect much to the fact that we're matching the coreos-stable kernel version.

The thoughts I've had seem either too wordy or suffer from the same problem but just disconnecting another concept.

• fsync-coreos-stable
• fsync-coreos
• fsync-stable
  ???

Thoughts on another name, or just leave it as is.

[bsherman]

@m2Giles there was a comment made in Discord about the coreos-testing, zfs failure on fedora 40 builds failing... Just calling it out here... the fsync patches themselves don't seem to be conflicting with the zfs kmod build in 6.9. But 6.10 is failing the zfs kmod build even without fsync patches.

zfs 2.2.6 is pending release and will address 6.10 builds. openzfs/zfs#16472

[m2Giles]

For name I think fsync-coreos should get the idea across. fsync-coreos-stable would be clearer but unsure if we need that distinction

[bsherman]

The one failing build now is just for coreos-testing-zfs, which will fail until openzfs releases version 2.2.6. This failure should not block merging of this PR.