Horton

Re: ERROR: invalid spinlock number: 0

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Fujii Masao <masao.fujii@oss.nttdata.com>
Cc: pgsql-hackers@postgresql.org
Date: 2021-02-15T08:27:21Z
Lists: pgsql-hackers 
On Thu, Feb 11, 2021 at 11:30:13PM +0900, Fujii Masao wrote:
> Yes, so what about the attached patch?

I see.  So the first error triggering the spinlock error would cause
a transaction failure because the fallback implementation of atomics
uses a spinlock for this variable, and it may not initialized in this
code path.

> We didn't notice this issue long time because no regression test checks
> pg_stat_wal_receiver. So I included such test in the patch.

Moving that behind ready_to_display is fine by me seeing where the
initialization is done.  The test case is a good addition.

+    * Read "writtenUpto" without holding a spinlock. So it may not be
+    * consistent with other WAL receiver's shared variables protected by a
+    * spinlock. This is OK because that variable is used only for
+    * informational purpose and should not be used for data integrity checks.
It seems to me that the first two sentences of this comment should be
combined together.
--
Michael

Commits

  1. Fix "invalid spinlock number: 0" error in pg_stat_wal_receiver.

  2. Display the time when the process started waiting for the lock, in pg_locks, take 2

  3. Display the time when the process started waiting for the lock, in pg_locks.

  4. Make pg_stat_wal_receiver consistent with the WAL receiver's shmem info