swapon01 doesn't take EPERM into account

[richiejp]

Test fails when swapon returns EPERM. This is the correct behavior when we don't have privs.

[Martchus]

On my system the test fails with libswap.c:57: TCONF: Swapfile on tmpfs not implemented. That it tries to use /tmp which is using a tmpfs on many systems is likely not ideal as well. The same counts for the other swapon tests.

This could be easily fixed by adding:

    .needs_root = 1,
    .mntpoint = MNTPOINT,
    .mount_device = 1,

[pevik]

Well, EPERM is on other filesystems as well, tested on btrfs and ext4. I guess we would need root for the permission anyway.

.needs_root = 1,
.mntpoint = MNTPOINT,
.mount_device = 1,

How about use .all_filesystems = 1. tmpfs is already detected on root, but I'd maybe check add for EPERM in libswap.c:59 => TCONF, just for sure.

[Martchus]

Does it really make sense to use .all_filesystems = 1? This would run the test on multiple filesystems but it suppose we just need one that works.

[Martchus]

@richiejp I must also say that the ticket description is not very clear.

To me the test code looks like it would produce TFAIL if swapon01 returns EPERM. I say "looks like" because I'm not sure how to reproduce the problem. Are you asking to change the test code so it would produce TCONF in that case (as @pevik suggested)?

[pevik]

FYI working on this, more fixes are needed, eg. in tst_ioctl.c

[Martchus]

Ok, then I'll leave it to you. (By the way, is there some way to assign oneself to tickets. This would allow us to avoid having multiple people work on the same ticket in parallel.)

[metan-ucw]

Also I'm not sure that all_filesystems would work here, since unless user has passed a real block device, we would end up with a block device in a file on a tmpfs, that probably still does not work with swap. There is more problems with LTP tests with /tmp on tmpfs, the only proper solution is to pass TMPDIR env variable to LTP that points to a real filesystem.

[metan-ucw]

Also I guess that all that is actually needed is to add '.needs_root = 1,' to swapon01.c

[pevik]

@Martchus @richiejp Feel free to do it as you started on it. The cleanup I do is not relevant to this.

[Martchus]

Supposedly using .needs_root = 1 is the simplest approach, indeed. I created a patch: https://patchwork.ozlabs.org/project/ltp/patch/20231011152900.4274-1-mkittler@suse.de/