Description:
Running any MTR test that uses Events with Thread Sanitizer, gives this data race:
WARNING: ThreadSanitizer: data race (pid=10571)
Write of size 1 at 0x7b3000001af9 by thread T43:
#0 Event_queue::lock_data(char const*, unsigned int) /export2/repo/mysql/sql/event_queue.cc:670 (mysqld+0x2d153c9)
#1 Event_queue::drop_schema_events(MYSQL_LEX_STRING) /export2/repo/mysql/sql/event_queue.cc:336 (mysqld+0x2d1626e)
#2 Events::drop_schema_events(THD*, dd::Schema const&) /export2/repo/mysql/sql/events.cc:740 (mysqld+0x28cc844)
#3 mysql_rm_db(THD*, MYSQL_LEX_CSTRING const&, bool) /export2/repo/mysql/sql/sql_db.cc:630 (mysqld+0x26b2239)
#4 mysql_execute_command(THD*, bool) /export2/repo/mysql/sql/sql_parse.cc:3495 (mysqld+0x2724436)
#5 mysql_parse(THD*, Parser_state*, bool) /export2/repo/mysql/sql/sql_parse.cc:5033 (mysqld+0x2720261)
#6 dispatch_command(THD*, COM_DATA const*, enum_server_command) /export2/repo/mysql/sql/sql_parse.cc:1677 (mysqld+0x271d895)
#7 do_command(THD*) /export2/repo/mysql/sql/sql_parse.cc:1233 (mysqld+0x271f3de)
#8 handle_connection(void*) /export2/repo/mysql/sql/conn_handler/connection_handler_per_thread.cc:303 (mysqld+0x28bcc63)
#9 pfs_spawn_thread(void*) /export2/repo/mysql/storage/perfschema/pfs.cc:2836 (mysqld+0x4281dc1)
Previous write of size 1 at 0x7b3000001af9 by thread T38 (mutexes: write M525367666731525320):
#0 Event_queue::lock_data(char const*, unsigned int) /export2/repo/mysql/sql/event_queue.cc:674 (mysqld+0x2d15408)
#1 Event_queue::get_top_for_execution_if_time(THD*, Event_queue_element_for_exec**) /export2/repo/mysql/sql/event_queue.cc:546 (mysqld+0x2d168bd)
#2 Event_scheduler::run(THD*) /export2/repo/mysql/sql/event_scheduler.cc:548 (mysqld+0x2d1b9b7)
#3 event_scheduler_thread(void*) /export2/repo/mysql/sql/event_scheduler.cc:268 (mysqld+0x2d1b746)
#4 pfs_spawn_thread(void*) /export2/repo/mysql/storage/perfschema/pfs.cc:2836 (mysqld+0x4281dc1)
This race can have consequences for the "Event queue status" printout during crash reporting.
How to repeat:
See above.
Suggested fix:
Use atomics.
Description: Running any MTR test that uses Events with Thread Sanitizer, gives this data race: WARNING: ThreadSanitizer: data race (pid=10571) Write of size 1 at 0x7b3000001af9 by thread T43: #0 Event_queue::lock_data(char const*, unsigned int) /export2/repo/mysql/sql/event_queue.cc:670 (mysqld+0x2d153c9) #1 Event_queue::drop_schema_events(MYSQL_LEX_STRING) /export2/repo/mysql/sql/event_queue.cc:336 (mysqld+0x2d1626e) #2 Events::drop_schema_events(THD*, dd::Schema const&) /export2/repo/mysql/sql/events.cc:740 (mysqld+0x28cc844) #3 mysql_rm_db(THD*, MYSQL_LEX_CSTRING const&, bool) /export2/repo/mysql/sql/sql_db.cc:630 (mysqld+0x26b2239) #4 mysql_execute_command(THD*, bool) /export2/repo/mysql/sql/sql_parse.cc:3495 (mysqld+0x2724436) #5 mysql_parse(THD*, Parser_state*, bool) /export2/repo/mysql/sql/sql_parse.cc:5033 (mysqld+0x2720261) #6 dispatch_command(THD*, COM_DATA const*, enum_server_command) /export2/repo/mysql/sql/sql_parse.cc:1677 (mysqld+0x271d895) #7 do_command(THD*) /export2/repo/mysql/sql/sql_parse.cc:1233 (mysqld+0x271f3de) #8 handle_connection(void*) /export2/repo/mysql/sql/conn_handler/connection_handler_per_thread.cc:303 (mysqld+0x28bcc63) #9 pfs_spawn_thread(void*) /export2/repo/mysql/storage/perfschema/pfs.cc:2836 (mysqld+0x4281dc1) Previous write of size 1 at 0x7b3000001af9 by thread T38 (mutexes: write M525367666731525320): #0 Event_queue::lock_data(char const*, unsigned int) /export2/repo/mysql/sql/event_queue.cc:674 (mysqld+0x2d15408) #1 Event_queue::get_top_for_execution_if_time(THD*, Event_queue_element_for_exec**) /export2/repo/mysql/sql/event_queue.cc:546 (mysqld+0x2d168bd) #2 Event_scheduler::run(THD*) /export2/repo/mysql/sql/event_scheduler.cc:548 (mysqld+0x2d1b9b7) #3 event_scheduler_thread(void*) /export2/repo/mysql/sql/event_scheduler.cc:268 (mysqld+0x2d1b746) #4 pfs_spawn_thread(void*) /export2/repo/mysql/storage/perfschema/pfs.cc:2836 (mysqld+0x4281dc1) This race can have consequences for the "Event queue status" printout during crash reporting. How to repeat: See above. Suggested fix: Use atomics.