libre #smartos

4 Dec 2023

SmartOS -- A Modern, Open Source, Datacenter Operating System | smartos.org | github.com/TritonDataCenter/smartos-live | github.com/TritonDataCenter/triton | Security -- TritonDataCenter.com/security-issues | Channel logs - log.omnios.org/smartos

• 20:39
  bahamat
  xmerlin: We have been able to reproduce it with the reproducer script, but it's much less frequent than it seems like other systems are encountering. The most recent release from last week includes the fix.
• 20:59
  xmerlin
  bahamat as I can see you have integrate the patch without mutex but the patch in illumos and openzfs use mutex ...why?
• 21:36
  bahamat
  xmerlin: We imported it early. Was there an additional issue found that necessitated the mutex, or is it just extra precaution?
• 21:37
  bahamat
  Let me check with Dan...we might respin the release.
• 21:51
  bahamat
  Ok, so the mutex was put into openzfs for a completely unrelated issue, over two years ago where lseek(2) would fail, seemingly at random.
• 21:52
  bahamat
  But it doesn't seem to be related to the data corruption issue.
• 21:53
  bahamat
  So we're not respinning the release, but it will be in our next release (and in the current dev build, assuming it finished).
• 21:54
  bahamat
  And yeah, that build has finished.
• 21:55
  bahamat
  That image is master-20231204T135840Z
• 22:05
  danmcd
  Worst case xmerlin is that the window of 16087 gets only much smaller without the mutex. With the mutex it should be eliminated.
• 22:06
  danmcd
  I ran 100 runs of the brute-force parallel reproducer on the release-build and it didn't trigger at all.
• 22:06
  danmcd
  (it == the bug)