MySQL Bugs: #112328: build_cache_of_ndb_users running from slave thread doesn't work

Bug #112328	build_cache_of_ndb_users running from slave thread doesn't work
Submitted:	12 Sep 2023 22:50	Modified:	8 Nov 2023 11:14
Reporter:	Mikael Ronström	Email Updates:
Status:	Duplicate	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S3 (Non-critical)
Version:	8.0.23	OS:	Any
Assigned to:		CPU Architecture:	Any

Description:
When combining setting up MySQL Cluster Replication and using
NDB stored users a segmentation fault is hit in build_cache_of_ndb_users.
The reasons is that it calls exec_sql that executes a function that
eventually ends up in mysql_execute_command. However since it is executed
from the slave thread it reports success with no result set.
This doesn't play with the functionality of build_cache_of_ndb_users and
a seg fault occurs.

How to repeat:
Run a set up of replication with ndb stored users in the cluster.

Suggested fix:
Ensure that exec_sql somehow ignores that it is currently running as a slave thread during execution of exec_sql.

Hello Mikael,

Thank you for the report and feedback.

Sincerely,
Umesh

Posted by developer:
 
Hi Mikael,

There is an MTR test called ndb_rpl_stored_grants that seems to test the situation you are describing. And in build_cache_of_ndb_users(), it is expected that the query can succeed but return 0 rows. So I think there must be something slightly different happening in this crash. Can you supply the stack trace? Or possibly something more to help reproduce the crash?

Thanks,

JD

Added a code contribution with a test case and the fix for this bug:
https://github.com/mysql/mysql-server/pull/498

Turns out that the PR was a duplicate of a code contribution already made.
See BUG#112775.

Posted by developer:
 
As stated by reporter this is duplicate and has been fixed in upcoming version of MySQL. Closing as duplicate.