[LU-15259] SLES15.2 sanity test_103a test_125 test_154a: setfacl: Option -m: Invalid argument near character 3

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0, Lustre 2.15.5
• Labels: ubuntu
• Severity: 3
• Assignee: Andreas Dilger
• Reporter: Andreas Dilger

Description

This issue was created by maloo for Andreas Dilger <[e-mail]>

This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/e1e7f0ed-909c-41f4-87a2-7e69b8ef80a0

test_103a failed with the following error:

$ setfacl -m u:bin:rw f -- failed
~                                     ? setfacl: Option -m: Invalid argument near character 3
$ setfacl -R -m u:bin:rwx h -- failed
~                                     ? setfacl: Option -m: Invalid argument near character 3
run_acl_subtest cp failed

It appears that this is some change in the "setfacl" command-line parsing for SLES15.2, as it hasn't failed for any other distro in this way, and fails 100% for SLES15.2 for this test on a number of different branches, so is not related to a recent code change. It is more likely related to a new version of the acl package installed on SLES15.2, but will likely also be seen as other distros update, unless this is a change local to SLES15 itself.

VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
sanity test_103a - run_acl_subtest cp failed

Issue Links

• is blocked by: LU-17875 sanity test_154b: FAIL: setfacl /mnt/lustre/.lustre/fid/[0x240004282:0x3:0x0] failed.
• is related to: LU-16673 sanity test_125: failures with aarch64 servers
• is related to: LU-17875 sanity test_154b: FAIL: setfacl /mnt/lustre/.lustre/fid/[0x240004282:0x3:0x0] failed.
• is related to: LU-15301 kernel update [SLES15 SP2 5.3.18-24.93.1]
• is related to: LU-15781 Ubuntu 22.04 LTS release support
• is related to: LU-5641 sanity.sh test_103a "acl test" fails on el7
• is related to: LU-10334 Ubuntu1604 client sanity-103a: FAIL: ACL misc test failed

Comments

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/55434/
Subject: LU-15259 tests: use existing usernames for setfacl
Project: fs/lustre-release
Branch: b2_15
Current Patch Set:
Commit: 225e4bdf39a5f2303a0e69c4a695138160f5a17f

[Gerrit Updater]

"Frederick Dilger <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/55434
Subject: LU-15259 tests: use existing usernames for setfacl
Project: fs/lustre-release
Branch: b2_15
Current Patch Set: 1
Commit: 98c33f795854b8e01c3b6d72d596b59798e5baae

[Peter Jones]

Landed for 2.16

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/45868/
Subject: LU-15259 tests: skip acl tests if no bin/daemon users
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: da7f2848a2aae2ec4852b23e55d23d42a30205ce

[Andreas Dilger]

The 45868 patch is just skipping these subtest for now, so this ticket should not be closed if that patch lands.

I think there are two (complementary) solutions:

• configure the autotest nodes to have bin and daemon users (the test itself should add them to the appropriate groups), and the subtests will not be skipped, but this requires every developer and autotest system to be changed forever into the future
• dynamically select the usernames from the existing system (e.g. sanityusr and sanityusr1) and replace the usage of bin and daemon in the test script so that the tests will run. I have an updated patch 45627 that should be able to do this.

[Gerrit Updater]

"Andreas Dilger <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/45868
Subject: LU-15259 tests: acl tests fail if no bin/daemon users
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: a984d1405c647bfb622a6083c5bccf4e548fcc78

[Andreas Dilger]

According to the permission.test, the "daemon" user must be a member of the "bin" group, but this isn't the case for "nobody" and the "rpc" group, so a few tests fail.

[Gerrit Updater]

"Andreas Dilger <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/45627
Subject: LU-15259 tests: use existing usernames for setfacl
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: a02353c0ef6abbc183f3b687963f3e52cab7b247

[Andreas Dilger]

So it looks like "bin" is not listed in the SLES15 /etc/passwd file. It looks like "u:daemon" is also missing. Different users needs to be put into the test scripts.

[Andreas Dilger]

This needs some investigation directly on SLES15.2 to see what has changed with the setfacl command, and the tests that are failing with this problem need to be updated to use a compatible command-line format (ideally in a way that is compatible with older setfacl commands, and/or a bug filed with SuSE about this interface breakage. The working acl version on RHEL8 is acl-2.2.53, while SLES15 looks like it has acl-2.3.1.

The other possibility is that "bin" is not a valid username on SLES15? This shows as bin:x:1:1:bin:/bin:/sbin/nologin in RHEL. In that case, we'd need to find some other username that is common between the distros, or (less great) create the "bin" user when SLES15 is installed.