BUG #18805: A specific query on a hash partitioned table always causes a "signal 11: Segmentation fault" error.

The following bug has been logged on the website:

Bug reference: 18805
Logged by: weijie JL
Email address: weijie1006jl(at)gmail(dot)com
PostgreSQL version: 17.2
Operating system: Rocky Linux release 8.10 (Green Obsidian)
Description:

Executing a specific SQL query in the database consistently results in a
"signal 11: Segmentation fault" error, and other connections report the
error: FATAL: 57P03: the database system is in recovery mode.

When the error occurs, the system log shows: kernel: XFS (dm-2): Corruption
of in-memory data (0x8) detected at xfs_trans_cancel+0xc6/0x130 [xfs]
(fs/xfs/xfs_trans.c:958). Shutting down filesystem. This indicates in-memory
data corruption in the XFS system, and the issue appears after this error.

We conducted the following tests:

Yesterday, we restored a backup using pgBackRest to another instance and
configured streaming replication. Initially, the issue did not reoccur, and
we switched to using this instance as the primary database. Since the
business team had adjusted the SQL, everything worked fine. However, when we
tested the problematic SQL again today, the issue reappeared.
On the problematic virtual machine, we tried restarting, reinstalling the
database software, and repairing the filesystem with xfs_repair, but the
issue persisted.
We moved the pgdata directory to a new disk space, but the issue still
persisted after starting the database.
We reinstalled the PostgreSQL software, but the issue persisted.
We uploaded the PostgreSQL source code to a test environment for debugging
and eventually identified that the issue was caused by the
enable_partitionwise_join parameter. Disabling the enable_partitionwise_join
parameter in the database prevented the issue from recurring.