ocfs2: Abstract ocfs2_extent_tree in b-tree operations.

In the old extent tree operation, we take the hypothesis that we
are using the ocfs2_extent_list in ocfs2_dinode as the tree root.
As xattr will also use ocfs2_extent_list to store large value
for a xattr entry, we refactor the tree operation so that xattr
can use it directly.

The refactoring includes 4 steps:
1. Abstract set/get of last_eb_blk and update_clusters since they may
   be stored in different location for dinode and xattr.
2. Add a new structure named ocfs2_extent_tree to indicate the
   extent tree the operation will work on.
3. Remove all the use of fe_bh and di, use root_bh and root_el in
   extent tree instead. So now all the fe_bh is replaced with
   et->root_bh, el with root_el accordingly.
4. Make ocfs2_lock_allocators generic. Now it is limited to be only used
   in file extend allocation. But the whole function is useful when we want
   to store large EAs.

Note: This patch doesn't touch ocfs2_commit_truncate() since it is not used
for anything other than truncate inode data btrees.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>

fs/ocfs2/alloc.h

@@ -26,28 +26,37 @@
 #ifndef OCFS2_ALLOC_H
 #define OCFS2_ALLOC_H
 
+enum ocfs2_extent_tree_type {
+	OCFS2_DINODE_EXTENT = 0,
+};
+
 struct ocfs2_alloc_context;
 int ocfs2_insert_extent(struct ocfs2_super *osb,
 			handle_t *handle,
 			struct inode *inode,
-			struct buffer_head *fe_bh,
+			struct buffer_head *root_bh,
 			u32 cpos,
 			u64 start_blk,
 			u32 new_clusters,
 			u8 flags,
-			struct ocfs2_alloc_context *meta_ac);
+			struct ocfs2_alloc_context *meta_ac,
+			enum ocfs2_extent_tree_type et_type);
 struct ocfs2_cached_dealloc_ctxt;
-int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh,
+int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *root_bh,
 			      handle_t *handle, u32 cpos, u32 len, u32 phys,
 			      struct ocfs2_alloc_context *meta_ac,
-			      struct ocfs2_cached_dealloc_ctxt *dealloc);
+			      struct ocfs2_cached_dealloc_ctxt *dealloc,
-int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
+			      enum ocfs2_extent_tree_type et_type);
+int ocfs2_remove_extent(struct inode *inode, struct buffer_head *root_bh,
 			u32 cpos, u32 len, handle_t *handle,
 			struct ocfs2_alloc_context *meta_ac,
-			struct ocfs2_cached_dealloc_ctxt *dealloc);
+			struct ocfs2_cached_dealloc_ctxt *dealloc,
+			enum ocfs2_extent_tree_type et_type);
 int ocfs2_num_free_extents(struct ocfs2_super *osb,
 			   struct inode *inode,
-			   struct buffer_head *bh);
+			   struct buffer_head *root_bh,
+			   enum ocfs2_extent_tree_type et_type);
+
 /*
  * how many new metadata chunks would an allocation need at maximum?
  *

fs/ocfs2/aops.c

@@ -1278,7 +1278,8 @@ static int ocfs2_write_cluster(struct address_space *mapping,
 	} else if (unwritten) {
 		ret = ocfs2_mark_extent_written(inode, wc->w_di_bh,
 						wc->w_handle, cpos, 1, phys,
-						meta_ac, &wc->w_dealloc);
+						meta_ac, &wc->w_dealloc,
+						OCFS2_DINODE_EXTENT);
 		if (ret < 0) {
 			mlog_errno(ret);
 			goto out;
@@ -1712,7 +1713,13 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
 		 * ocfs2_lock_allocators(). It greatly over-estimates
 		 * the work to be done.
 		 */
-		ret = ocfs2_lock_allocators(inode, wc->w_di_bh,
+		mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u,"
+		     " clusters_to_add = %u, extents_to_split = %u\n",
+		     (unsigned long long)OCFS2_I(inode)->ip_blkno,
+		     (long long)i_size_read(inode), le32_to_cpu(di->i_clusters),
+		     clusters_to_alloc, extents_to_split);
+
+		ret = ocfs2_lock_allocators(inode, wc->w_di_bh, &di->id2.i_list,
 					    clusters_to_alloc, extents_to_split,
 					    &data_ac, &meta_ac);
 		if (ret) {

fs/ocfs2/dir.c

@@ -1306,7 +1306,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 * related blocks have been journaled already.
 	 */
 	ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 0, blkno, len, 0,
-				  NULL);
+				  NULL, OCFS2_DINODE_EXTENT);
 	if (ret) {
 		mlog_errno(ret);
 		goto out_commit;
@@ -1338,7 +1338,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 		blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off);
 
 		ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 1, blkno,
-					  len, 0, NULL);
+					  len, 0, NULL, OCFS2_DINODE_EXTENT);
 		if (ret) {
 			mlog_errno(ret);
 			goto out_commit;
@@ -1481,7 +1481,8 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
 	if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) {
 		spin_unlock(&OCFS2_I(dir)->ip_lock);
 		num_free_extents = ocfs2_num_free_extents(osb, dir,
-							  parent_fe_bh);
+							  parent_fe_bh,
+							  OCFS2_DINODE_EXTENT);
 		if (num_free_extents < 0) {
 			status = num_free_extents;
 			mlog_errno(status);

fs/ocfs2/file.c

@@ -521,7 +521,8 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
 	if (mark_unwritten)
 		flags = OCFS2_EXT_UNWRITTEN;
 
-	free_extents = ocfs2_num_free_extents(osb, inode, fe_bh);
+	free_extents = ocfs2_num_free_extents(osb, inode, fe_bh,
+					      OCFS2_DINODE_EXTENT);
 	if (free_extents < 0) {
 		status = free_extents;
 		mlog_errno(status);
@@ -570,7 +571,7 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
 	     num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno);
 	status = ocfs2_insert_extent(osb, handle, inode, fe_bh,
 				     *logical_offset, block, num_bits,
-				     flags, meta_ac);
+				     flags, meta_ac, OCFS2_DINODE_EXTENT);
 	if (status < 0) {
 		mlog_errno(status);
 		goto leave;
@@ -599,92 +600,6 @@ leave:
 	return status;
 }
 
-/*
- * For a given allocation, determine which allocators will need to be
- * accessed, and lock them, reserving the appropriate number of bits.
- *
- * Sparse file systems call this from ocfs2_write_begin_nolock()
- * and ocfs2_allocate_unwritten_extents().
- *
- * File systems which don't support holes call this from
- * ocfs2_extend_allocation().
- */
-int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *di_bh,
-			  u32 clusters_to_add, u32 extents_to_split,
-			  struct ocfs2_alloc_context **data_ac,
-			  struct ocfs2_alloc_context **meta_ac)
-{
-	int ret = 0, num_free_extents;
-	unsigned int max_recs_needed = clusters_to_add + 2 * extents_to_split;
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
-
-	*meta_ac = NULL;
-	if (data_ac)
-		*data_ac = NULL;
-
-	BUG_ON(clusters_to_add != 0 && data_ac == NULL);
-
-	mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u, "
-	     "clusters_to_add = %u, extents_to_split = %u\n",
-	     (unsigned long long)OCFS2_I(inode)->ip_blkno, (long long)i_size_read(inode),
-	     le32_to_cpu(di->i_clusters), clusters_to_add, extents_to_split);
-
-	num_free_extents = ocfs2_num_free_extents(osb, inode, di_bh);
-	if (num_free_extents < 0) {
-		ret = num_free_extents;
-		mlog_errno(ret);
-		goto out;
-	}
-
-	/*
-	 * Sparse allocation file systems need to be more conservative
-	 * with reserving room for expansion - the actual allocation
-	 * happens while we've got a journal handle open so re-taking
-	 * a cluster lock (because we ran out of room for another
-	 * extent) will violate ordering rules.
-	 *
-	 * Most of the time we'll only be seeing this 1 cluster at a time
-	 * anyway.
-	 *
-	 * Always lock for any unwritten extents - we might want to
-	 * add blocks during a split.
-	 */
-	if (!num_free_extents ||
-	    (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) {
-		ret = ocfs2_reserve_new_metadata(osb, &di->id2.i_list, meta_ac);
-		if (ret < 0) {
-			if (ret != -ENOSPC)
-				mlog_errno(ret);
-			goto out;
-		}
-	}
-
-	if (clusters_to_add == 0)
-		goto out;
-
-	ret = ocfs2_reserve_clusters(osb, clusters_to_add, data_ac);
-	if (ret < 0) {
-		if (ret != -ENOSPC)
-			mlog_errno(ret);
-		goto out;
-	}
-
-out:
-	if (ret) {
-		if (*meta_ac) {
-			ocfs2_free_alloc_context(*meta_ac);
-			*meta_ac = NULL;
-		}
-
-		/*
-		 * We cannot have an error and a non null *data_ac.
-		 */
-	}
-
-	return ret;
-}
-
 static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
 				     u32 clusters_to_add, int mark_unwritten)
 {
@@ -725,7 +640,13 @@ static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
 restart_all:
 	BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
 
-	status = ocfs2_lock_allocators(inode, bh, clusters_to_add, 0, &data_ac,
+	mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u, "
+	     "clusters_to_add = %u\n",
+	     (unsigned long long)OCFS2_I(inode)->ip_blkno,
+	     (long long)i_size_read(inode), le32_to_cpu(fe->i_clusters),
+	     clusters_to_add);
+	status = ocfs2_lock_allocators(inode, bh, &fe->id2.i_list,
+				       clusters_to_add, 0, &data_ac,
 				       &meta_ac);
 	if (status) {
 		mlog_errno(status);
@@ -1397,7 +1318,8 @@ static int __ocfs2_remove_inode_range(struct inode *inode,
 	struct ocfs2_alloc_context *meta_ac = NULL;
 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
 
-	ret = ocfs2_lock_allocators(inode, di_bh, 0, 1, NULL, &meta_ac);
+	ret = ocfs2_lock_allocators(inode, di_bh, &di->id2.i_list,
+				    0, 1, NULL, &meta_ac);
 	if (ret) {
 		mlog_errno(ret);
 		return ret;
@@ -1428,7 +1350,7 @@ static int __ocfs2_remove_inode_range(struct inode *inode,
 	}
 
 	ret = ocfs2_remove_extent(inode, di_bh, cpos, len, handle, meta_ac,
-				  dealloc);
+				  dealloc, OCFS2_DINODE_EXTENT);
 	if (ret) {
 		mlog_errno(ret);
 		goto out_commit;

fs/ocfs2/file.h

@@ -57,10 +57,6 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
 			       enum ocfs2_alloc_restarted *reason_ret);
 int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size,
 			  u64 zero_to);
-int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *fe,
-			  u32 clusters_to_add, u32 extents_to_split,
-			  struct ocfs2_alloc_context **data_ac,
-			  struct ocfs2_alloc_context **meta_ac);
 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
 int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
 		  struct kstat *stat);

fs/ocfs2/suballoc.c

@@ -1894,3 +1894,85 @@ static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe)
 		       (unsigned long long)fe->id2.i_chain.cl_recs[i].c_blkno);
 	}
 }
+
+/*
+ * For a given allocation, determine which allocators will need to be
+ * accessed, and lock them, reserving the appropriate number of bits.
+ *
+ * Sparse file systems call this from ocfs2_write_begin_nolock()
+ * and ocfs2_allocate_unwritten_extents().
+ *
+ * File systems which don't support holes call this from
+ * ocfs2_extend_allocation().
+ */
+int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *root_bh,
+			  struct ocfs2_extent_list *root_el,
+			  u32 clusters_to_add, u32 extents_to_split,
+			  struct ocfs2_alloc_context **data_ac,
+			  struct ocfs2_alloc_context **meta_ac)
+{
+	int ret = 0, num_free_extents;
+	unsigned int max_recs_needed = clusters_to_add + 2 * extents_to_split;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	*meta_ac = NULL;
+	if (data_ac)
+		*data_ac = NULL;
+
+	BUG_ON(clusters_to_add != 0 && data_ac == NULL);
+
+	num_free_extents = ocfs2_num_free_extents(osb, inode, root_bh,
+						  OCFS2_DINODE_EXTENT);
+	if (num_free_extents < 0) {
+		ret = num_free_extents;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	/*
+	 * Sparse allocation file systems need to be more conservative
+	 * with reserving room for expansion - the actual allocation
+	 * happens while we've got a journal handle open so re-taking
+	 * a cluster lock (because we ran out of room for another
+	 * extent) will violate ordering rules.
+	 *
+	 * Most of the time we'll only be seeing this 1 cluster at a time
+	 * anyway.
+	 *
+	 * Always lock for any unwritten extents - we might want to
+	 * add blocks during a split.
+	 */
+	if (!num_free_extents ||
+	    (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) {
+		ret = ocfs2_reserve_new_metadata(osb, root_el, meta_ac);
+		if (ret < 0) {
+			if (ret != -ENOSPC)
+				mlog_errno(ret);
+			goto out;
+		}
+	}
+
+	if (clusters_to_add == 0)
+		goto out;
+
+	ret = ocfs2_reserve_clusters(osb, clusters_to_add, data_ac);
+	if (ret < 0) {
+		if (ret != -ENOSPC)
+			mlog_errno(ret);
+		goto out;
+	}
+
+out:
+	if (ret) {
+		if (*meta_ac) {
+			ocfs2_free_alloc_context(*meta_ac);
+			*meta_ac = NULL;
+		}
+
+		/*
+		 * We cannot have an error and a non null *data_ac.
+		 */
+	}
+
+	return ret;
+}

fs/ocfs2/suballoc.h

@@ -162,4 +162,9 @@ u64 ocfs2_which_cluster_group(struct inode *inode, u32 cluster);
 int ocfs2_check_group_descriptor(struct super_block *sb,
 				 struct ocfs2_dinode *di,
 				 struct ocfs2_group_desc *gd);
+int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *root_bh,
+			  struct ocfs2_extent_list *root_el,
+			  u32 clusters_to_add, u32 extents_to_split,
+			  struct ocfs2_alloc_context **data_ac,
+			  struct ocfs2_alloc_context **meta_ac);
 #endif /* _CHAINALLOC_H_ */