Fix mdtruncate() to close fd.c handle of deleted segments.

anarazel · anarazel · commit d2a5b2b28f24 · 2016-09-08T16:52:13.000-07:00
mdtruncate() forgot to FileClose() a segment's mdfd_vfd, when deleting it. That lead to a fd.c handle to a truncated file being kept open until backend exit. The issue appears to have been introduced way back in 1a5c450, before that the handle was closed inside FileUnlink(). The impact of this bug is limited - only VACUUM and ON COMMIT TRUNCATE for temporary tables, truncate files in place (i.e. TRUNCATE itself is not affected), and the relation has to be bigger than 1GB. The consequences of a leaked fd.c handle aren't severe either. Discussion: <20160908220748.oqh37ukwqqncbl3n@alap3.anarazel.de> Backpatch: all supported releases
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
@@ -903,6 +903,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
 			v = v->mdfd_chain;
 			Assert(ov != reln->md_fd[forknum]); /* we never drop the 1st
 												 * segment */
+			FileClose(ov->mdfd_vfd);
 			pfree(ov);
 		}
 		else if (priorblocks + ((BlockNumber) RELSEG_SIZE) > nblocks)

Original file line number	Diff line number	Diff line change
`@@ -903,6 +903,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)`
`903`	`903`	`v = v->mdfd_chain;`
`904`	`904`	`Assert(ov != reln->md_fd[forknum]); /* we never drop the 1st`
`905`	`905`	`* segment */`
	`906`	`+ FileClose(ov->mdfd_vfd);`
`906`	`907`	`pfree(ov);`
`907`	`908`	`}`
`908`	`909`	`else if (priorblocks + ((BlockNumber) RELSEG_SIZE) > nblocks)`