Unset MyBEEntry, making elog.c's call to pgstat_get_my_query_id() safe.

Previously log messages late during shutdown could end up using either another
backend's PgBackendStatus (multi user) or segfault (single user) because
pgstat_get_my_query_id()'s check for !MyBEEntry didn't filter out use after
pgstat_beshutdown_hook().

This became a bug in 4f0b0966c86, but was a bit fishy before. But given
there's no known problematic cases before 14, it doesn't seem worth
backpatching further.

Also fixes a wrong filename in a comment, introduced in e1025044.

Reported-By: Andres Freund <andres@anarazel.de>
Reviewed-By: Julien Rouhaud <rjuju123@gmail.com>
Discussion: https://postgr.es/m/Julien Rouhaud <rjuju123@gmail.com>
Backpatch: 14-

diff --git a/src/backend/utils/activity/backend_status.c b/src/backend/utils/activity/backend_status.c
index 2901f9f5a9fa255f813ad066ebc12af1c18c3361..e19c4506efa4851e6104fe0d272d3d7aa39ff206 100644 (file)
--- a/src/backend/utils/activity/backend_status.c
+++ b/src/backend/utils/activity/backend_status.c
@@ -46,7 +46,7 @@ bool      pgstat_track_activities = false;
 int            pgstat_track_activity_query_size = 1024;
 
 
-/* exposed so that progress.c can access it */
+/* exposed so that backend_progress.c can access it */
 PgBackendStatus *MyBEEntry = NULL;
 
 
@@ -469,6 +469,9 @@ pgstat_beshutdown_hook(int code, Datum arg)
    beentry->st_procpid = 0;    /* mark invalid */
 
    PGSTAT_END_WRITE_ACTIVITY(beentry);
+
+   /* so that functions can check if backend_status.c is up via MyBEEntry */
+   MyBEEntry = NULL;
 }
 
 /*