Check for unlinked znodes after igrab()

[mfoliveira]

Motivation and Context

The changes in commit 41e1aa2a / PR#9583 introduced a regression on
tmpfile_001_pos: fsetxattr() on a O_TMPFILE file descriptor started
to fail with errno ENODATA:

    openat(AT_FDCWD, "/test", O_RDWR|O_TMPFILE, 0666) = 3
    <...>
    fsetxattr(3, "user.test", <...>, 64, 0) = -1 ENODATA (No data available)

Description

The originally proposed change on PR#9583 is not susceptible to it,
so just move the code/if-checks around back in that way, to fix it.

Before that commit:

    $ cat /sys/module/zfs/version
    0.8.0-401_gcc1a1e17

    $ TESTDIR="/test" /usr/share/zfs/zfs-tests/tests/functional/tmpfile/tmpfile_001_pos ; echo $?
    Verify O_TMPFILE is working properly.
    0

After that commit:

    $ cat /sys/module/zfs/version
    0.8.0-402_g41e1aa2a

    $ TESTDIR="/test" /usr/share/zfs/zfs-tests/tests/functional/tmpfile/tmpfile_001_pos ; echo $?
    Verify O_TMPFILE is working properly.
    fsetxattr: No data available
    6

With this commit:

    $ cat /sys/module/zfs/version
    0.8.0-405_g...

    $ TESTDIR="/test" /usr/share/zfs/zfs-tests/tests/functional/tmpfile/tmpfile_001_pos ; echo $?
    Verify O_TMPFILE is working properly.
    0

How Has This Been Tested?

The zfs test-suite (functional) has been run before/after commit 41e1aa2a / PR#9583, and only tmpfile_001_pos had a consistent, repeatable regression.
The patch has been verified to fix it / make this particular test-case pass again.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the ZFS on Linux code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• All new and existing tests passed.
• All commit messages are properly formatted and contain Signed-off-by.

[behlendorf]

Thank you for the quick fix which is correct. I noticed this myself yesterday but hadn't yet started to investigate. I can verify it does correctly resolve the O_TMPFILE failures.

The root cause here is that new files created with O_TMPFILE are immediately added to the unlinked list and have z_unlinked set to ensure they're not leaked during a crash. Which means we need to perform the igrab() before checking z_unlinked. It's possible this might have also caused an issue when manipulating open but unlinked files, though I didn't see any issue with some quick testing.

The other OpenZFS platforms do not support the O_TMPFILE option so this wouldn't be an issue there.

[codecov]

Codecov Report

Merging #9602 into master will decrease coverage by 0.18%.
The diff coverage is 25%.

@@            Coverage Diff             @@
##           master    #9602      +/-   ##
==========================================
- Coverage   79.22%   79.04%   -0.19%     
==========================================
  Files         419      419              
  Lines      123704   123704              
==========================================
- Hits        98009    97785     -224     
- Misses      25695    25919     +224

Flag	Coverage Δ
#kernel	79.73% <25%> (-0.19%)	⬇️
#user	66.5% <ø> (-0.22%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 540312d...010df42. Read the comment docs.