Flag index metapages as standard-format in xlog.c calls.

tglsfdc · tglsfdc · commit 4c11d2c559e7 · 2017-11-03T16:31:32.000-04:00
btree, hash, and bloom indexes all set up their metapages in standard format (that is, with pd_lower and pd_upper correctly delimiting the unused area); but they mostly didn't inform the xlog routines of this. When calling log_newpage[_buffer], this is bad because it loses the opportunity to compress unused data out of the WAL record. When calling XLogRegisterBuffer, it's not such a performance problem because all of these call sites also use REGBUF_WILL_INIT, preventing an FPI image from being written. But it's still a good idea to provide the flag when relevant, because that aids WAL consistency checking. This completes the project of getting all the in-core index AMs to handle their metapage WAL operations similarly. Amit Kapila, reviewed by Michael Paquier Discussion: https://postgr.es/m/0d273805-0e9e-ec1a-cb84-d4da400b8f85@lab.ntt.co.jp
diff --git a/contrib/bloom/blinsert.c b/contrib/bloom/blinsert.c
@@ -175,7 +175,7 @@ blbuildempty(Relation index)
 	smgrwrite(index->rd_smgr, INIT_FORKNUM, BLOOM_METAPAGE_BLKNO,
 			  (char *) metapage, true);
 	log_newpage(&index->rd_smgr->smgr_rnode.node, INIT_FORKNUM,
-				BLOOM_METAPAGE_BLKNO, metapage, false);
+				BLOOM_METAPAGE_BLKNO, metapage, true);
 
 	/*
 	 * An immediate sync is required even if we xlog'd the page, because the
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
@@ -403,7 +403,7 @@ _hash_init(Relation rel, double num_tuples, ForkNumber forkNum)
 
 		XLogBeginInsert();
 		XLogRegisterData((char *) &xlrec, SizeOfHashInitMetaPage);
-		XLogRegisterBuffer(0, metabuf, REGBUF_WILL_INIT);
+		XLogRegisterBuffer(0, metabuf, REGBUF_WILL_INIT | REGBUF_STANDARD);
 
 		recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_INIT_META_PAGE);
 
@@ -592,8 +592,9 @@ _hash_init_metabuffer(Buffer buf, double num_tuples, RegProcedure procid,
 	metap->hashm_firstfree = 0;
 
 	/*
-	 * Set pd_lower just past the end of the metadata.  This is to log full
-	 * page image of metapage in xloginsert.c.
+	 * Set pd_lower just past the end of the metadata.  This is essential,
+	 * because without doing so, metadata will be lost if xlog.c compresses
+	 * the page.
 	 */
 	((PageHeader) page)->pd_lower =
 		((char *) metap + sizeof(HashMetaPageData)) - (char *) page;
diff --git a/src/backend/access/nbtree/nbtinsert.c b/src/backend/access/nbtree/nbtinsert.c
@@ -898,7 +898,7 @@ _bt_insertonpg(Relation rel,
 				xlmeta.fastroot = metad->btm_fastroot;
 				xlmeta.fastlevel = metad->btm_fastlevel;
 
-				XLogRegisterBuffer(2, metabuf, REGBUF_WILL_INIT);
+				XLogRegisterBuffer(2, metabuf, REGBUF_WILL_INIT | REGBUF_STANDARD);
 				XLogRegisterBufData(2, (char *) &xlmeta, sizeof(xl_btree_metadata));
 
 				xlinfo = XLOG_BTREE_INSERT_META;
@@ -2032,7 +2032,7 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
 
 		XLogRegisterBuffer(0, rootbuf, REGBUF_WILL_INIT);
 		XLogRegisterBuffer(1, lbuf, REGBUF_STANDARD);
-		XLogRegisterBuffer(2, metabuf, REGBUF_WILL_INIT);
+		XLogRegisterBuffer(2, metabuf, REGBUF_WILL_INIT | REGBUF_STANDARD);
 
 		md.root = rootblknum;
 		md.level = metad->btm_level;
diff --git a/src/backend/access/nbtree/nbtpage.c b/src/backend/access/nbtree/nbtpage.c
@@ -65,8 +65,9 @@ _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level)
 	metaopaque->btpo_flags = BTP_META;
 
 	/*
-	 * Set pd_lower just past the end of the metadata.  This is not essential
-	 * but it makes the page look compressible to xlog.c.
+	 * Set pd_lower just past the end of the metadata.  This is essential,
+	 * because without doing so, metadata will be lost if xlog.c compresses
+	 * the page.
 	 */
 	((PageHeader) page)->pd_lower =
 		((char *) metad + sizeof(BTMetaPageData)) - (char *) page;
@@ -241,7 +242,7 @@ _bt_getroot(Relation rel, int access)
 
 			XLogBeginInsert();
 			XLogRegisterBuffer(0, rootbuf, REGBUF_WILL_INIT);
-			XLogRegisterBuffer(2, metabuf, REGBUF_WILL_INIT);
+			XLogRegisterBuffer(2, metabuf, REGBUF_WILL_INIT | REGBUF_STANDARD);
 
 			md.root = rootblkno;
 			md.level = 0;
@@ -1827,7 +1828,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, bool *rightsib_empty)
 
 		if (BufferIsValid(metabuf))
 		{
-			XLogRegisterBuffer(4, metabuf, REGBUF_WILL_INIT);
+			XLogRegisterBuffer(4, metabuf, REGBUF_WILL_INIT | REGBUF_STANDARD);
 
 			xlmeta.root = metad->btm_root;
 			xlmeta.level = metad->btm_level;
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
@@ -298,7 +298,7 @@ btbuildempty(Relation index)
 	smgrwrite(index->rd_smgr, INIT_FORKNUM, BTREE_METAPAGE,
 			  (char *) metapage, true);
 	log_newpage(&index->rd_smgr->smgr_rnode.node, INIT_FORKNUM,
-				BTREE_METAPAGE, metapage, false);
+				BTREE_METAPAGE, metapage, true);
 
 	/*
 	 * An immediate sync is required even if we xlog'd the page, because the
diff --git a/src/backend/access/nbtree/nbtxlog.c b/src/backend/access/nbtree/nbtxlog.c
@@ -107,8 +107,9 @@ _bt_restore_meta(XLogReaderState *record, uint8 block_id)
 	pageop->btpo_flags = BTP_META;
 
 	/*
-	 * Set pd_lower just past the end of the metadata.  This is not essential
-	 * but it makes the page look compressible to xlog.c.
+	 * Set pd_lower just past the end of the metadata.  This is essential,
+	 * because without doing so, metadata will be lost if xlog.c compresses
+	 * the page.
 	 */
 	((PageHeader) metapg)->pd_lower =
 		((char *) md + sizeof(BTMetaPageData)) - (char *) metapg;
