ext4: cache extent hole in extent status tree for ext4_da_map_blocks()

Currently extent status tree doesn't cache extent hole when a write
looks up in extent tree to make sure whether a block has been allocated
or not.  In this case, we don't put extent hole in extent cache because
later this extent might be removed and a new delayed extent might be
added back.  But it will cause a defect when we do a lot of writes.  If
we don't put extent hole in extent cache, the following writes also need
to access extent tree to look at whether or not a block has been
allocated.  It brings a cache miss.  This commit fixes this defect.
Also if the inode doesn't have any extent, this extent hole will be
cached as well.

Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

4 files changed, 20 insertions, 24 deletions

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 7b3f3b1..98da4cd 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -556,10 +556,8 @@ enum {
 #define EXT4_GET_BLOCKS_KEEP_SIZE		0x0080
 	/* Do not take i_data_sem locking in ext4_map_blocks */
 #define EXT4_GET_BLOCKS_NO_LOCK			0x0100
-	/* Do not put hole in extent cache */
-#define EXT4_GET_BLOCKS_NO_PUT_HOLE		0x0200
 	/* Convert written extents to unwritten */
-#define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN	0x0400
+#define EXT4_GET_BLOCKS_CONVERT_UNWRITTEN	0x0200
 
 /*
  * The bit position of these flags must not overlap with any of the
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 7ef2f11..1ee24d7 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -2306,16 +2306,16 @@ ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
 				ext4_lblk_t block)
 {
 	int depth = ext_depth(inode);
-	unsigned long len = 0;
-	ext4_lblk_t lblock = 0;
+	ext4_lblk_t len;
+	ext4_lblk_t lblock;
 	struct ext4_extent *ex;
+	struct extent_status es;
 
 	ex = path[depth].p_ext;
 	if (ex == NULL) {
-		/*
-		 * there is no extent yet, so gap is [0;-] and we
-		 * don't cache it
-		 */
+		/* there is no extent yet, so gap is [0;-] */
+		lblock = 0;
+		len = EXT_MAX_BLOCKS;
 		ext_debug("cache gap(whole file):");
 	} else if (block < le32_to_cpu(ex->ee_block)) {
 		lblock = block;
@@ -2324,9 +2324,6 @@ ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
 				block,
 				le32_to_cpu(ex->ee_block),
 				 ext4_ext_get_actual_len(ex));
-		if (!ext4_find_delalloc_range(inode, lblock, lblock + len - 1))
-			ext4_es_insert_extent(inode, lblock, len, ~0,
-					      EXTENT_STATUS_HOLE);
 	} else if (block >= le32_to_cpu(ex->ee_block)
 			+ ext4_ext_get_actual_len(ex)) {
 		ext4_lblk_t next;
@@ -2340,14 +2337,19 @@ ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
 				block);
 		BUG_ON(next == lblock);
 		len = next - lblock;
-		if (!ext4_find_delalloc_range(inode, lblock, lblock + len - 1))
-			ext4_es_insert_extent(inode, lblock, len, ~0,
-					      EXTENT_STATUS_HOLE);
 	} else {
 		BUG();
 	}
 
-	ext_debug(" -> %u:%lu\n", lblock, len);
+	ext4_es_find_delayed_extent_range(inode, lblock, lblock + len - 1, &es);
+	if (es.es_len) {
+		/* There's delayed extent containing lblock? */
+		if (es.es_lblk <= lblock)
+			return;
+		len = min(es.es_lblk - lblock, len);
+	}
+	ext_debug(" -> %u:%u\n", lblock, len);
+	ext4_es_insert_extent(inode, lblock, len, ~0, EXTENT_STATUS_HOLE);
 }
 
 /*
@@ -4368,8 +4370,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 		 * put just found gap into cache to speed up
 		 * subsequent requests
 		 */
-		if ((flags & EXT4_GET_BLOCKS_NO_PUT_HOLE) == 0)
-			ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
+		ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
 		goto out2;
 	}
 
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 2315e45..d5a46a8 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1432,11 +1432,9 @@ static int ext4_da_map_blocks(struct inode *inode, sector_t iblock,
 	if (ext4_has_inline_data(inode))
 		retval = 0;
 	else if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
-		retval = ext4_ext_map_blocks(NULL, inode, map,
-					     EXT4_GET_BLOCKS_NO_PUT_HOLE);
+		retval = ext4_ext_map_blocks(NULL, inode, map, 0);
 	else
-		retval = ext4_ind_map_blocks(NULL, inode, map,
-					     EXT4_GET_BLOCKS_NO_PUT_HOLE);
+		retval = ext4_ind_map_blocks(NULL, inode, map, 0);
 
 add_delayed:
 	if (retval == 0) {
diff --git a/include/trace/events/ext4.h b/include/trace/events/ext4.h
index bb7dcbe..cd37a58 100644
--- a/include/trace/events/ext4.h
+++ b/include/trace/events/ext4.h
@@ -43,8 +43,7 @@ struct extent_status;
 	{ EXT4_GET_BLOCKS_METADATA_NOFAIL,	"METADATA_NOFAIL" },	\
 	{ EXT4_GET_BLOCKS_NO_NORMALIZE,		"NO_NORMALIZE" },	\
 	{ EXT4_GET_BLOCKS_KEEP_SIZE,		"KEEP_SIZE" },		\
-	{ EXT4_GET_BLOCKS_NO_LOCK,		"NO_LOCK" },		\
-	{ EXT4_GET_BLOCKS_NO_PUT_HOLE,		"NO_PUT_HOLE" })
+	{ EXT4_GET_BLOCKS_NO_LOCK,		"NO_LOCK" })
 
 #define show_mflags(flags) __print_flags(flags, "",	\
 	{ EXT4_MAP_NEW,		"N" },			\