Resolver crashes caused by destroying locked mutexes
Since yesterday, all three RPZ-mode "stress" test jobs failed for main
due to named
crashing during the test:
- https://gitlab.isc.org/isc-projects/bind9/-/jobs/2953945
- https://gitlab.isc.org/isc-projects/bind9/-/jobs/2953948
- https://gitlab.isc.org/isc-projects/bind9/-/jobs/2953951
The error logged is:
01-Dec-2022 02:43:32.062 general: critical: adb.c:1188:destroy_adbentry(): fatal error:
01-Dec-2022 02:43:32.062 general: critical: pthread_mutex_destroy(): Device or resource busy (16)
01-Dec-2022 02:43:32.062 general: critical: exiting (due to fatal error in library)
This most likely means that an dns_adbentry_t
's lock
mutex was still
locked while its destruction was attempted.
These crashes did not happen for v9_18
, so !7138 (merged) is the most likely
culprit.
Job artifacts (including core dumps and binaries) were preserved.
Edited by Michał Kępień