ocfs2: make room for unwritten extents flag

Due to the size of our group bitmaps, we'll never have a leaf node extent
record with more than 16 bits worth of clusters. Split e_clusters up so that
leaf nodes can get a flags field where we can mark unwritten extents.
Interior nodes whose length references all the child nodes beneath it can't
split their e_clusters field, so we use a union to preserve sizing there.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>

fs/ocfs2/alloc.h

@@ -85,4 +85,23 @@ int ocfs2_commit_truncate(struct ocfs2_super *osb,
 int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el,
 		    u32 cpos, struct buffer_head **leaf_bh);
 
+/*
+ * Helper function to look at the # of clusters in an extent record.
+ */
+static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
+					      struct ocfs2_extent_rec *rec)
+{
+	/*
+	 * Cluster count in extent records is slightly different
+	 * between interior nodes and leaf nodes. This is to support
+	 * unwritten extents which need a flags field in leaf node
+	 * records, thus shrinking the available space for a clusters
+	 * field.
+	 */
+	if (el->l_tree_depth)
+		return le32_to_cpu(rec->e_int_clusters);
+	else
+		return le16_to_cpu(rec->e_leaf_clusters);
+}
+
 #endif /* OCFS2_ALLOC_H */

fs/ocfs2/extent_map.c

@@ -50,13 +50,15 @@ static int ocfs2_search_extent_list(struct ocfs2_extent_list *el,
 	int ret = -1;
 	int i;
 	struct ocfs2_extent_rec *rec;
-	u32 rec_end, rec_start;
+	u32 rec_end, rec_start, clusters;
 
 	for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
 		rec = &el->l_recs[i];
 
 		rec_start = le32_to_cpu(rec->e_cpos);
-		rec_end = rec_start + le32_to_cpu(rec->e_clusters);
+		clusters = ocfs2_rec_clusters(el, rec);
+
+		rec_end = rec_start + clusters;
 
 		if (v_cluster >= rec_start && v_cluster < rec_end) {
 			ret = i;
@@ -98,6 +100,15 @@ int ocfs2_get_clusters(struct inode *inode, u32 v_cluster,
 
 		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
 		el = &eb->h_list;
+
+		if (el->l_tree_depth) {
+			ocfs2_error(inode->i_sb,
+				    "Inode %lu has non zero tree depth in "
+				    "leaf block %llu\n", inode->i_ino,
+				    (unsigned long long)eb_bh->b_blocknr);
+			ret = -EROFS;
+			goto out;
+		}
 	}
 
 	i = ocfs2_search_extent_list(el, v_cluster);
@@ -118,7 +129,7 @@ int ocfs2_get_clusters(struct inode *inode, u32 v_cluster,
 			ocfs2_error(inode->i_sb, "Inode %lu has bad extent "
 				    "record (%u, %u, 0)", inode->i_ino,
 				    le32_to_cpu(rec->e_cpos),
-				    le32_to_cpu(rec->e_clusters));
+				    ocfs2_rec_clusters(el, rec));
 			ret = -EROFS;
 			goto out;
 		}
@@ -130,7 +141,7 @@ int ocfs2_get_clusters(struct inode *inode, u32 v_cluster,
 		*p_cluster = *p_cluster + coff;
 
 		if (num_clusters)
-			*num_clusters = le32_to_cpu(rec->e_clusters) - coff;
+			*num_clusters = ocfs2_rec_clusters(el, rec) - coff;
 	}
 
 out:

fs/ocfs2/file.c

@@ -1127,7 +1127,6 @@ static int ocfs2_check_range_for_holes(struct inode *inode, loff_t pos,
 				       size_t count)
 {
 	int ret = 0;
-	unsigned int extent_flags;
 	u32 cpos, clusters, extent_len, phys_cpos;
 	struct super_block *sb = inode->i_sb;
 
@@ -1135,14 +1134,13 @@ static int ocfs2_check_range_for_holes(struct inode *inode, loff_t pos,
 	clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
 
 	while (clusters) {
-		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
-					 &extent_flags);
+		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len);
 		if (ret < 0) {
 			mlog_errno(ret);
 			goto out;
 		}
 
-		if (phys_cpos == 0 || (extent_flags & OCFS2_EXT_UNWRITTEN)) {
+		if (phys_cpos == 0) {
 			ret = 1;
 			break;
 		}

fs/ocfs2/journal.h

@@ -390,7 +390,7 @@ static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
 	/* We may be deleting metadata blocks, so metadata alloc dinode +
 	   one desc. block for each possible delete. */
 	if (tree_depth && next_free == 1 &&
-	    le32_to_cpu(last_el->l_recs[i].e_clusters) == clusters_to_del)
+	    ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del)
 		credits += 1 + tree_depth;
 
 	/* update to the truncate log. */

fs/ocfs2/ocfs2_fs.h

@@ -155,6 +155,12 @@
 #define OCFS2_FL_VISIBLE	(0x000100FF)	/* User visible flags */
 #define OCFS2_FL_MODIFIABLE	(0x000100FF)	/* User modifiable flags */
 
+/*
+ * Extent record flags (e_node.leaf.flags)
+ */
+#define OCFS2_EXT_UNWRITTEN	(0x01)	/* Extent is allocated but
+					 * unwritten */
+
 /*
  * ioctl commands
  */
@@ -283,10 +289,21 @@ static unsigned char ocfs2_type_by_mode[S_IFMT >> S_SHIFT] = {
 /*
  * On disk extent record for OCFS2
  * It describes a range of clusters on disk.
+ *
+ * Length fields are divided into interior and leaf node versions.
+ * This leaves room for a flags field (OCFS2_EXT_*) in the leaf nodes.
  */
 struct ocfs2_extent_rec {
 /*00*/	__le32 e_cpos;		/* Offset into the file, in clusters */
-	__le32 e_clusters;	/* Clusters covered by this extent */
+	union {
+		__le32 e_int_clusters; /* Clusters covered by all children */
+		struct {
+			__le16 e_leaf_clusters; /* Clusters covered by this
+						   extent */
+			__u8 e_reserved1;
+			__u8 e_flags; /* Extent flags */
+		};
+	};
 	__le64 e_blkno;		/* Physical disk offset, in blocks */
 /*10*/
 };