Make pg_restore's identify_locking_dependencies() more bulletproof.

tglsfdc · tglsfdc · commit 97aa524f18c0 · 2018-08-28T19:46:59.000-04:00
This function had a blacklist of dump object types that it believed needed exclusive lock ... but we hadn't maintained that, so that it was missing ROW SECURITY, POLICY, and INDEX ATTACH items, all of which need (or should be treated as needing) exclusive lock. Since the same oversight seems likely in future, let's reverse the sense of the test so that the code has a whitelist of safe object types; better to wrongly assume a command can't be run in parallel than the opposite. Currently the only POST_DATA object type that's safe is CREATE INDEX ... and that list hasn't changed in a long time. Back-patch to 9.5 where RLS came in. Discussion: https://postgr.es/m/11450.1535483506@sss.pgh.pa.us
diff --git a/src/bin/pg_dump/pg_backup_archiver.c b/src/bin/pg_dump/pg_backup_archiver.c
@@ -4514,16 +4514,24 @@ identify_locking_dependencies(ArchiveHandle *AH, TocEntry *te)
 	int			nlockids;
 	int			i;
 
+	/*
+	 * We only care about this for POST_DATA items.  PRE_DATA items are not
+	 * run in parallel, and DATA items are all independent by assumption.
+	 */
+	if (te->section != SECTION_POST_DATA)
+		return;
+
 	/* Quick exit if no dependencies at all */
 	if (te->nDeps == 0)
 		return;
 
-	/* Exit if this entry doesn't need exclusive lock on other objects */
-	if (!(strcmp(te->desc, "CONSTRAINT") == 0 ||
-		  strcmp(te->desc, "CHECK CONSTRAINT") == 0 ||
-		  strcmp(te->desc, "FK CONSTRAINT") == 0 ||
-		  strcmp(te->desc, "RULE") == 0 ||
-		  strcmp(te->desc, "TRIGGER") == 0))
+	/*
+	 * Most POST_DATA items are ALTER TABLEs or some moral equivalent of that,
+	 * and hence require exclusive lock.  However, we know that CREATE INDEX
+	 * does not.  (Maybe someday index-creating CONSTRAINTs will fall in that
+	 * category too ... but today is not that day.)
+	 */
+	if (strcmp(te->desc, "INDEX") == 0)
 		return;
 
 	/*
