diff options
| author | Marko Mäkelä <marko.makela@mariadb.com> | 2018-05-16 16:35:33 +0300 |
|---|---|---|
| committer | Marko Mäkelä <marko.makela@mariadb.com> | 2018-05-16 16:35:33 +0300 |
| commit | a4e7800701d0764fe4cbb85b81d7c7cb54677334 (patch) | |
| tree | 7db65106dfe46274dd4a2042bf202cb8b1dbda82 | |
| parent | ebc24950e6d107b2958887cc03fe67db1073ccf2 (diff) | |
| download | mariadb-git-a4e7800701d0764fe4cbb85b81d7c7cb54677334.tar.gz | |
MDEV-13779 InnoDB fails to shut down purge workers, causing hang
srv_purge_coordinator_thread(): Wait for all purge worker threads
to actually exit. An analysis of a core dump of a hung 10.3 server
revealed that one srv_worker_thread did not exit, even though the
purge coordinator had exited. This caused kill_server_thread and
mysqld_main to wait indefinitely. The main InnoDB shutdown was
never called, because unireg_end() was never called.
| -rw-r--r-- | storage/innobase/srv/srv0srv.cc | 14 |
1 files changed, 12 insertions, 2 deletions
diff --git a/storage/innobase/srv/srv0srv.cc b/storage/innobase/srv/srv0srv.cc index 5f93ff7cd7c..802271bb96f 100644 --- a/storage/innobase/srv/srv0srv.cc +++ b/storage/innobase/srv/srv0srv.cc @@ -2876,8 +2876,18 @@ DECLARE_THREAD(srv_purge_coordinator_thread)( #endif /* UNIV_DEBUG_THREAD_CREATION */ /* Ensure that all the worker threads quit. */ - if (srv_n_purge_threads > 1) { - srv_release_threads(SRV_WORKER, srv_n_purge_threads - 1); + if (ulint n_workers = srv_n_purge_threads - 1) { + const srv_slot_t* slot; + const srv_slot_t* const end = &srv_sys.sys_threads[ + srv_sys.n_sys_threads]; + + do { + srv_release_threads(SRV_WORKER, n_workers); + srv_sys_mutex_enter(); + for (slot = &srv_sys.sys_threads[2]; + !slot++->in_use && slot < end; ); + srv_sys_mutex_exit(); + } while (slot < end); } innobase_destroy_background_thd(thd); |