Discussed with Yves on Slack. Raw PMP in comment #7 corresponds to processed PMP in comment #9.
Upon further analysis found the missing bit of comment #8: thread 1984 also has locked one of thd->LOCK_thd_data, and so the picture is
Thread 2266: tries to lock thd->LOCK_thd_data
has LOCK_index
Thread 1984: tries to lock LOCK_log
has thd->LOCK_thd_data
Thread 357: tries to lock LOCK_index
has LOCK_log
1984 -> 357 -> 2266 -> 1984. Deadlock. Will move this to separate bug as symptoms are different from a transient COMMIT stall with concurrent SHOW STATUS and will fix together.
Discussed with Yves on Slack. Raw PMP in comment #7 corresponds to processed PMP in comment #9.
Upon further analysis found the missing bit of comment #8: thread 1984 also has locked one of thd->LOCK_thd_data, and so the picture is
Thread 2266: tries to lock thd->LOCK_thd_data
has LOCK_index
Thread 1984: tries to lock LOCK_log
has thd->LOCK_thd_data
Thread 357: tries to lock LOCK_index
has LOCK_log
1984 -> 357 -> 2266 -> 1984. Deadlock. Will move this to separate bug as symptoms are different from a transient COMMIT stall with concurrent SHOW STATUS and will fix together.