Fix WAL format incompatibility introduced by backpatching of 52ac6cd

akorotkov · akorotkov · commit d1166af192ed · 2019-03-24T16:31:23.000+03:00
52ac6cd added new field to ginxlogDeletePage and was backpatched to 9.4. That led to problems when patched postgres instance applies WAL records generated by non-patched one. WAL records generated by non-patched instance don't contain new field, which patched one is expecting to see. Thankfully, we can distinguish patched and non-patched WAL records by their data size. If we see that WAL record is generated by non-patched instance, we skip processing of new field. This commit comes with some assertions. In particular, if it appears that on some platform struct data size didn't change then static assertion will trigger. Reported-by: Simon Riggs Discussion: https://postgr.es/m/CANP8%2Bj%2BK4whxf7ET7%2BgO%2BG-baC3-WxqqH%3DnV4X2CgfEPA3Yu3g%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Simon Riggs, Alvaro Herrera Backpatch-through: 9.4
diff --git a/src/backend/access/gin/ginxlog.c b/src/backend/access/gin/ginxlog.c
@@ -720,12 +720,30 @@ ginRedoDeletePage(XLogRecPtr lsn, XLogRecord *record)
 	else
 	{
 		dbuffer = XLogReadBuffer(data->node, data->blkno, false);
+
+		/*
+		 * deleteXid field of ginxlogDeletePage was added during backpatching.
+		 * But, non-backpatched instances will continue generate WAL without
+		 * this field.  We should be able to correctly apply that.  We can
+		 * distinguish new WAL records by size their data, because
+		 * ginxlogDeletePage changes its size on both 32-bit and 64-bit
+		 * platforms.
+		 */
+		StaticAssertStmt(sizeof(ginxlogDeletePage) !=
+						 sizeof(ginxlogDeletePageOld),
+						 "ginxlogDeletePage size should be changed "
+						 "with addition of deleteXid field");
+		Assert(record->xl_len == sizeof(ginxlogDeletePage) ||
+			   record->xl_len == sizeof(ginxlogDeletePageOld));
+
 		if (BufferIsValid(dbuffer))
 		{
 			page = BufferGetPage(dbuffer);
 			if (lsn > PageGetLSN(page))
 			{
 				Assert(GinPageIsData(page));
+				if (record->xl_len == sizeof(ginxlogDeletePage))
+					GinPageSetDeleteXid(page, data->deleteXid);
 				GinPageGetOpaque(page)->flags = GIN_DELETED;
 				PageSetLSN(page, lsn);
 				MarkBufferDirty(dbuffer);
diff --git a/src/include/access/gin_private.h b/src/include/access/gin_private.h
@@ -586,6 +586,20 @@ typedef struct ginxlogDeletePage
 	TransactionId deleteXid;	/* last Xid which could see this page in scan */
 } ginxlogDeletePage;
 
+/*
+ * Previous version of ginxlogDeletePage struct, which didn't have deleteXid
+ * field.  Used for size comparison (see ginRedoDeletePage()).
+ */
+typedef struct ginxlogDeletePageOld
+{
+	RelFileNode node;
+	BlockNumber blkno;
+	BlockNumber parentBlkno;
+	OffsetNumber parentOffset;
+	BlockNumber leftBlkno;
+	BlockNumber rightLink;
+} ginxlogDeletePageOld;
+
 #define XLOG_GIN_UPDATE_META_PAGE 0x60
 
 typedef struct ginxlogUpdateMeta
