In GIN recompression code, use mmemove rather than memcpy, for vacuum.

When vacuuming a data leaf page, any compressed posting lists that are not
modified, are copied back to the buffer from a later location in the same
buffer rather than from  a palloc'd copy. IOW, they are just moved
downwards in the same buffer. Because the source and destination addresses
can overlap, we must use memmove rather than memcpy.

Report and fix by Alexander Korotkov.

1 files changed, 14 insertions, 1 deletions

diff --git a/src/backend/access/gin/gindatapage.c b/src/backend/access/gin/gindatapage.c
index 8504f4c7afc..ebdacd40c55 100644
--- a/src/backend/access/gin/gindatapage.c
+++ b/src/backend/access/gin/gindatapage.c
@@ -753,6 +753,13 @@ ginVacuumPostingTreeLeaf(Relation indexrel, Buffer buffer, GinVacuumState *gvs)
  * *prdata is filled with WAL information about this operation. The caller
  * is responsible for inserting to the WAL, along with any other information
  * about the operation that triggered this recompression.
+ *
+ * NOTE: The segment pointers can point directly to the same buffer, with
+ * the limitation that any earlier segment must not overlap with an original,
+ * later segment. In other words, some segments may point the original buffer
+ * as long as you don't make any segments larger. Currently, leafRepackItems
+ * satisies this rule because it rewrites all segments after the first
+ * modified one, and vacuum can only make segments shorter.
  */
 static void
 dataPlaceToPageLeafRecompress(Buffer buf, disassembledLeaf *leaf,
@@ -798,7 +805,13 @@ dataPlaceToPageLeafRecompress(Buffer buf, disassembledLeaf *leaf,
 		if (!modified)
 			unmodifiedsize += segsize;
 		else
-			memcpy(ptr, seginfo->seg, segsize);
+		{
+			/*
+			 * Use memmove rather than memcpy, in case the segment points
+			 * to the same buffer
+			 */
+			memmove(ptr, seginfo->seg, segsize);
+		}
 		ptr += segsize;
 		newsize += segsize;
 	}