Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2: (47 commits)
  ocfs2: Silence a gcc warning.
  ocfs2: Don't retry xattr set in case value extension fails.
  ocfs2:dlm: avoid dlm->ast_lock lockres->spinlock dependency break
  ocfs2: Reset xattr value size after xa_cleanup_value_truncate().
  fs/ocfs2/dlm: Use kstrdup
  fs/ocfs2/dlm: Drop memory allocation cast
  Ocfs2: Optimize punching-hole code.
  Ocfs2: Make ocfs2_find_cpos_for_left_leaf() public.
  Ocfs2: Fix hole punching to correctly do CoW during cluster zeroing.
  Ocfs2: Optimize ocfs2 truncate to use ocfs2_remove_btree_range() instead.
  ocfs2: Block signals for mkdir/link/symlink/O_CREAT.
  ocfs2: Wrap signal blocking in void functions.
  ocfs2/dlm: Increase o2dlm lockres hash size
  ocfs2: Make ocfs2_extend_trans() really extend.
  ocfs2/trivial: Code cleanup for allocation reservation.
  ocfs2: make ocfs2_adjust_resv_from_alloc simple.
  ocfs2: Make nointr a default mount option
  ocfs2/dlm: Make o2dlm domain join/leave messages KERN_NOTICE
  o2net: log socket state changes
  ocfs2: print node # when tcp fails
  ...

Documentation/filesystems/ocfs2.txt

@@ -80,3 +80,10 @@ user_xattr	(*)	Enables Extended User Attributes.
 nouser_xattr		Disables Extended User Attributes.
 acl			Enables POSIX Access Control Lists support.
 noacl		(*)	Disables POSIX Access Control Lists support.
+resv_level=2	(*)	Set how agressive allocation reservations will be.
+			Valid values are between 0 (reservations off) to 8
+			(maximum space for reservations).
+dir_resv_level=	(*)	By default, directory reservations will scale with file
+			reservations - users should rarely need to change this
+			value. If allocation reservations are turned off, this
+			option will have no effect.

fs/ocfs2/Makefile

@@ -29,6 +29,7 @@ ocfs2-objs := \
 	mmap.o 			\
 	namei.o 		\
 	refcounttree.o		\
+	reservations.o		\
 	resize.o		\
 	slot_map.o 		\
 	suballoc.o 		\

fs/ocfs2/alloc.c

@@ -1006,7 +1006,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 	int count, status, i;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 first_blkno;
+	u64 suballoc_loc, first_blkno;
 	struct ocfs2_super *osb =
 		OCFS2_SB(ocfs2_metadata_cache_get_super(et->et_ci));
 	struct ocfs2_extent_block *eb;
@@ -1015,10 +1015,10 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 
 	count = 0;
 	while (count < wanted) {
-		status = ocfs2_claim_metadata(osb,
-					      handle,
+		status = ocfs2_claim_metadata(handle,
 					      meta_ac,
 					      wanted - count,
+					      &suballoc_loc,
 					      &suballoc_bit_start,
 					      &num_got,
 					      &first_blkno);
@@ -1052,6 +1052,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 			eb->h_fs_generation = cpu_to_le32(osb->fs_generation);
 			eb->h_suballoc_slot =
 				cpu_to_le16(meta_ac->ac_alloc_slot);
+			eb->h_suballoc_loc = cpu_to_le64(suballoc_loc);
 			eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 			eb->h_list.l_count =
 				cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));
@@ -1061,11 +1062,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 
 			/* We'll also be dirtied by the caller, so
 			 * this isn't absolutely necessary. */
-			status = ocfs2_journal_dirty(handle, bhs[i]);
-			if (status < 0) {
-				mlog_errno(status);
-				goto bail;
-			}
+			ocfs2_journal_dirty(handle, bhs[i]);
 		}
 
 		count += num_got;
@@ -1129,8 +1126,7 @@ static int ocfs2_adjust_rightmost_branch(handle_t *handle,
 		goto out;
 	}
 
-	status = ocfs2_extend_trans(handle, path_num_items(path) +
-				    handle->h_buffer_credits);
+	status = ocfs2_extend_trans(handle, path_num_items(path));
 	if (status < 0) {
 		mlog_errno(status);
 		goto out;
@@ -1270,12 +1266,7 @@ static int ocfs2_add_branch(handle_t *handle,
 		if (!eb_el->l_tree_depth)
 			new_last_eb_blk = le64_to_cpu(eb->h_blkno);
 
-		status = ocfs2_journal_dirty(handle, bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-
+		ocfs2_journal_dirty(handle, bh);
 		next_blkno = le64_to_cpu(eb->h_blkno);
 	}
 
@@ -1321,17 +1312,10 @@ static int ocfs2_add_branch(handle_t *handle,
 	eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
 	eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);
 
-	status = ocfs2_journal_dirty(handle, *last_eb_bh);
-	if (status < 0)
-		mlog_errno(status);
-	status = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (status < 0)
-		mlog_errno(status);
-	if (eb_bh) {
-		status = ocfs2_journal_dirty(handle, eb_bh);
-		if (status < 0)
-			mlog_errno(status);
-	}
+	ocfs2_journal_dirty(handle, *last_eb_bh);
+	ocfs2_journal_dirty(handle, et->et_root_bh);
+	if (eb_bh)
+		ocfs2_journal_dirty(handle, eb_bh);
 
 	/*
 	 * Some callers want to track the rightmost leaf so pass it
@@ -1399,11 +1383,7 @@ static int ocfs2_shift_tree_depth(handle_t *handle,
 	for (i = 0; i < le16_to_cpu(root_el->l_next_free_rec); i++)
 		eb_el->l_recs[i] = root_el->l_recs[i];
 
-	status = ocfs2_journal_dirty(handle, new_eb_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, new_eb_bh);
 
 	status = ocfs2_et_root_journal_access(handle, et,
 					      OCFS2_JOURNAL_ACCESS_WRITE);
@@ -1428,11 +1408,7 @@ static int ocfs2_shift_tree_depth(handle_t *handle,
 	if (root_el->l_tree_depth == cpu_to_le16(1))
 		ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
 
-	status = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
 	*ret_new_eb_bh = new_eb_bh;
 	new_eb_bh = NULL;
@@ -2064,7 +2040,7 @@ static void ocfs2_complete_edge_insert(handle_t *handle,
 				       struct ocfs2_path *right_path,
 				       int subtree_index)
 {
-	int ret, i, idx;
+	int i, idx;
 	struct ocfs2_extent_list *el, *left_el, *right_el;
 	struct ocfs2_extent_rec *left_rec, *right_rec;
 	struct buffer_head *root_bh = left_path->p_node[subtree_index].bh;
@@ -2102,13 +2078,8 @@ static void ocfs2_complete_edge_insert(handle_t *handle,
 		ocfs2_adjust_adjacent_records(left_rec, left_el, right_rec,
 					      right_el);
 
-		ret = ocfs2_journal_dirty(handle, left_path->p_node[i].bh);
-		if (ret)
-			mlog_errno(ret);
-
-		ret = ocfs2_journal_dirty(handle, right_path->p_node[i].bh);
-		if (ret)
-			mlog_errno(ret);
+		ocfs2_journal_dirty(handle, left_path->p_node[i].bh);
+		ocfs2_journal_dirty(handle, right_path->p_node[i].bh);
 
 		/*
 		 * Setup our list pointers now so that the current
@@ -2132,9 +2103,7 @@ static void ocfs2_complete_edge_insert(handle_t *handle,
 
 	root_bh = left_path->p_node[subtree_index].bh;
 
-	ret = ocfs2_journal_dirty(handle, root_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, root_bh);
 }
 
 static int ocfs2_rotate_subtree_right(handle_t *handle,
@@ -2207,11 +2176,7 @@ static int ocfs2_rotate_subtree_right(handle_t *handle,
 
 	ocfs2_create_empty_extent(right_el);
 
-	ret = ocfs2_journal_dirty(handle, right_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, right_leaf_bh);
 
 	/* Do the copy now. */
 	i = le16_to_cpu(left_el->l_next_free_rec) - 1;
@@ -2230,11 +2195,7 @@ static int ocfs2_rotate_subtree_right(handle_t *handle,
 	memset(&left_el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
 	le16_add_cpu(&left_el->l_next_free_rec, 1);
 
-	ret = ocfs2_journal_dirty(handle, left_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, left_leaf_bh);
 
 	ocfs2_complete_edge_insert(handle, left_path, right_path,
 				   subtree_index);
@@ -2249,8 +2210,8 @@ out:
  *
  * Will return zero if the path passed in is already the leftmost path.
  */
-static int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
-					 struct ocfs2_path *path, u32 *cpos)
+int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
+				  struct ocfs2_path *path, u32 *cpos)
 {
 	int i, j, ret = 0;
 	u64 blkno;
@@ -2327,20 +2288,14 @@ static int ocfs2_extend_rotate_transaction(handle_t *handle, int subtree_depth,
 					   int op_credits,
 					   struct ocfs2_path *path)
 {
-	int ret;
+	int ret = 0;
 	int credits = (path->p_tree_depth - subtree_depth) * 2 + 1 + op_credits;
 
-	if (handle->h_buffer_credits < credits) {
+	if (handle->h_buffer_credits < credits)
 		ret = ocfs2_extend_trans(handle,
 					 credits - handle->h_buffer_credits);
-		if (ret)
-			return ret;
 
-		if (unlikely(handle->h_buffer_credits < credits))
-			return ocfs2_extend_trans(handle, credits);
-	}
-
-	return 0;
+	return ret;
 }
 
 /*
@@ -2584,8 +2539,7 @@ static int ocfs2_update_edge_lengths(handle_t *handle,
 	 * records for all the bh in the path.
 	 * So we have to allocate extra credits and access them.
 	 */
-	ret = ocfs2_extend_trans(handle,
-				 handle->h_buffer_credits + subtree_index);
+	ret = ocfs2_extend_trans(handle, subtree_index);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -2823,12 +2777,8 @@ static int ocfs2_rotate_subtree_left(handle_t *handle,
 		ocfs2_remove_empty_extent(right_leaf_el);
 	}
 
-	ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
-	if (ret)
-		mlog_errno(ret);
-	ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
+	ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
 
 	if (del_right_subtree) {
 		ocfs2_unlink_subtree(handle, et, left_path, right_path,
@@ -2851,9 +2801,7 @@ static int ocfs2_rotate_subtree_left(handle_t *handle,
 		if (right_has_empty)
 			ocfs2_remove_empty_extent(left_leaf_el);
 
-		ret = ocfs2_journal_dirty(handle, et_root_bh);
-		if (ret)
-			mlog_errno(ret);
+		ocfs2_journal_dirty(handle, et_root_bh);
 
 		*deleted = 1;
 	} else
@@ -2962,10 +2910,7 @@ static int ocfs2_rotate_rightmost_leaf_left(handle_t *handle,
 	}
 
 	ocfs2_remove_empty_extent(el);
-
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, bh);
 
 out:
 	return ret;
@@ -3506,15 +3451,9 @@ static int ocfs2_merge_rec_right(struct ocfs2_path *left_path,
 
 	ocfs2_cleanup_merge(el, index);
 
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret)
-		mlog_errno(ret);
-
+	ocfs2_journal_dirty(handle, bh);
 	if (right_path) {
-		ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
-		if (ret)
-			mlog_errno(ret);
-
+		ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
 		ocfs2_complete_edge_insert(handle, left_path, right_path,
 					   subtree_index);
 	}
@@ -3683,14 +3622,9 @@ static int ocfs2_merge_rec_left(struct ocfs2_path *right_path,
 
 	ocfs2_cleanup_merge(el, index);
 
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret)
-		mlog_errno(ret);
-
+	ocfs2_journal_dirty(handle, bh);
 	if (left_path) {
-		ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
-		if (ret)
-			mlog_errno(ret);
+		ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
 
 		/*
 		 * In the situation that the right_rec is empty and the extent
@@ -4016,10 +3950,7 @@ static void ocfs2_adjust_rightmost_records(handle_t *handle,
 		le32_add_cpu(&rec->e_int_clusters,
 			     -le32_to_cpu(rec->e_cpos));
 
-		ret = ocfs2_journal_dirty(handle, bh);
-		if (ret)
-			mlog_errno(ret);
-
+		ocfs2_journal_dirty(handle, bh);
 	}
 }
 
@@ -4203,17 +4134,13 @@ static int ocfs2_insert_path(handle_t *handle,
 	struct buffer_head *leaf_bh = path_leaf_bh(right_path);
 
 	if (left_path) {
-		int credits = handle->h_buffer_credits;
-
 		/*
 		 * There's a chance that left_path got passed back to
 		 * us without being accounted for in the
 		 * journal. Extend our transaction here to be sure we
 		 * can change those blocks.
 		 */
-		credits += left_path->p_tree_depth;
-
-		ret = ocfs2_extend_trans(handle, credits);
+		ret = ocfs2_extend_trans(handle, left_path->p_tree_depth);
 		if (ret < 0) {
 			mlog_errno(ret);
 			goto out;
@@ -4251,17 +4178,13 @@ static int ocfs2_insert_path(handle_t *handle,
 		 * dirty this for us.
 		 */
 		if (left_path)
-			ret = ocfs2_journal_dirty(handle,
-						  path_leaf_bh(left_path));
-			if (ret)
-				mlog_errno(ret);
+			ocfs2_journal_dirty(handle,
+					    path_leaf_bh(left_path));
 	} else
 		ocfs2_insert_at_leaf(et, insert_rec, path_leaf_el(right_path),
 				     insert);
 
-	ret = ocfs2_journal_dirty(handle, leaf_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, leaf_bh);
 
 	if (left_path) {
 		/*
@@ -4384,9 +4307,7 @@ out_update_clusters:
 		ocfs2_et_update_clusters(et,
 					 le16_to_cpu(insert_rec->e_leaf_clusters));
 
-	ret = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
 out:
 	ocfs2_free_path(left_path);
@@ -4866,7 +4787,7 @@ int ocfs2_add_clusters_in_btree(handle_t *handle,
 		goto leave;
 	}
 
-	status = __ocfs2_claim_clusters(osb, handle, data_ac, 1,
+	status = __ocfs2_claim_clusters(handle, data_ac, 1,
 					clusters_to_add, &bit_off, &num_bits);
 	if (status < 0) {
 		if (status != -ENOSPC)
@@ -4895,11 +4816,7 @@ int ocfs2_add_clusters_in_btree(handle_t *handle,
 		goto leave;
 	}
 
-	status = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
 	clusters_to_add -= num_bits;
 	*logical_offset += num_bits;
@@ -5309,7 +5226,7 @@ static int ocfs2_split_tree(handle_t *handle, struct ocfs2_extent_tree *et,
 			    int index, u32 new_range,
 			    struct ocfs2_alloc_context *meta_ac)
 {
-	int ret, depth, credits = handle->h_buffer_credits;
+	int ret, depth, credits;
 	struct buffer_head *last_eb_bh = NULL;
 	struct ocfs2_extent_block *eb;
 	struct ocfs2_extent_list *rightmost_el, *el;
@@ -5340,8 +5257,8 @@ static int ocfs2_split_tree(handle_t *handle, struct ocfs2_extent_tree *et,
 	} else
 		rightmost_el = path_leaf_el(path);
 
-	credits += path->p_tree_depth +
-		   ocfs2_extend_meta_needed(et->et_root_el);
+	credits = path->p_tree_depth +
+		  ocfs2_extend_meta_needed(et->et_root_el);
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
@@ -5671,19 +5588,97 @@ out:
 	return ret;
 }
 
+/*
+ * ocfs2_reserve_blocks_for_rec_trunc() would look basically the
+ * same as ocfs2_lock_alloctors(), except for it accepts a blocks
+ * number to reserve some extra blocks, and it only handles meta
+ * data allocations.
+ *
+ * Currently, only ocfs2_remove_btree_range() uses it for truncating
+ * and punching holes.
+ */
+static int ocfs2_reserve_blocks_for_rec_trunc(struct inode *inode,
+					      struct ocfs2_extent_tree *et,
+					      u32 extents_to_split,
+					      struct ocfs2_alloc_context **ac,
+					      int extra_blocks)
+{
+	int ret = 0, num_free_extents;
+	unsigned int max_recs_needed = 2 * extents_to_split;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	*ac = NULL;
+
+	num_free_extents = ocfs2_num_free_extents(osb, et);
+	if (num_free_extents < 0) {
+		ret = num_free_extents;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	if (!num_free_extents ||
+	    (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed))
+		extra_blocks += ocfs2_extend_meta_needed(et->et_root_el);
+
+	if (extra_blocks) {
+		ret = ocfs2_reserve_new_metadata_blocks(osb, extra_blocks, ac);
+		if (ret < 0) {
+			if (ret != -ENOSPC)
+				mlog_errno(ret);
+			goto out;
+		}
+	}
+
+out:
+	if (ret) {
+		if (*ac) {
+			ocfs2_free_alloc_context(*ac);
+			*ac = NULL;
+		}
+	}
+
+	return ret;
+}
+
 int ocfs2_remove_btree_range(struct inode *inode,
 			     struct ocfs2_extent_tree *et,
-			     u32 cpos, u32 phys_cpos, u32 len,
-			     struct ocfs2_cached_dealloc_ctxt *dealloc)
+			     u32 cpos, u32 phys_cpos, u32 len, int flags,
+			     struct ocfs2_cached_dealloc_ctxt *dealloc,
+			     u64 refcount_loc)
 {
-	int ret;
+	int ret, credits = 0, extra_blocks = 0;
 	u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct inode *tl_inode = osb->osb_tl_inode;
 	handle_t *handle;
 	struct ocfs2_alloc_context *meta_ac = NULL;
+	struct ocfs2_refcount_tree *ref_tree = NULL;
+
+	if ((flags & OCFS2_EXT_REFCOUNTED) && len) {
+		BUG_ON(!(OCFS2_I(inode)->ip_dyn_features &
+			 OCFS2_HAS_REFCOUNT_FL));
+
+		ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1,
+					       &ref_tree, NULL);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
 
-	ret = ocfs2_lock_allocators(inode, et, 0, 1, NULL, &meta_ac);
+		ret = ocfs2_prepare_refcount_change_for_del(inode,
+							    refcount_loc,
+							    phys_blkno,
+							    len,
+							    &credits,
+							    &extra_blocks);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+	}
+
+	ret = ocfs2_reserve_blocks_for_rec_trunc(inode, et, 1, &meta_ac,
+						 extra_blocks);
 	if (ret) {
 		mlog_errno(ret);
 		return ret;
@@ -5699,7 +5694,8 @@ int ocfs2_remove_btree_range(struct inode *inode,
 		}
 	}
 
-	handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb));
+	handle = ocfs2_start_trans(osb,
+			ocfs2_remove_extent_credits(osb->sb) + credits);
 	if (IS_ERR(handle)) {
 		ret = PTR_ERR(handle);
 		mlog_errno(ret);
@@ -5724,15 +5720,22 @@ int ocfs2_remove_btree_range(struct inode *inode,
 
 	ocfs2_et_update_clusters(et, -len);
 
-	ret = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
-	ret = ocfs2_truncate_log_append(osb, handle, phys_blkno, len);
-	if (ret)
-		mlog_errno(ret);
+	if (phys_blkno) {
+		if (flags & OCFS2_EXT_REFCOUNTED)
+			ret = ocfs2_decrease_refcount(inode, handle,
+					ocfs2_blocks_to_clusters(osb->sb,
+								 phys_blkno),
+					len, meta_ac,
+					dealloc, 1);
+		else
+			ret = ocfs2_truncate_log_append(osb, handle,
+							phys_blkno, len);
+		if (ret)
+			mlog_errno(ret);
+
+	}
 
 out_commit:
 	ocfs2_commit_trans(osb, handle);
@@ -5742,6 +5745,9 @@ out:
 	if (meta_ac)
 		ocfs2_free_alloc_context(meta_ac);
 
+	if (ref_tree)
+		ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
+
 	return ret;
 }
 
@@ -5850,11 +5856,7 @@ int ocfs2_truncate_log_append(struct ocfs2_super *osb,
 	}
 	tl->tl_recs[index].t_clusters = cpu_to_le32(num_clusters);
 
-	status = ocfs2_journal_dirty(handle, tl_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, tl_bh);
 
 bail:
 	mlog_exit(status);
@@ -5893,11 +5895,7 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
 
 		tl->tl_used = cpu_to_le16(i);
 
-		status = ocfs2_journal_dirty(handle, tl_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
+		ocfs2_journal_dirty(handle, tl_bh);
 
 		/* TODO: Perhaps we can calculate the bulk of the
 		 * credits up front rather than extending like
@@ -6298,6 +6296,7 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb)
  */
 struct ocfs2_cached_block_free {
 	struct ocfs2_cached_block_free		*free_next;
+	u64					free_bg;
 	u64					free_blk;
 	unsigned int				free_bit;
 };
@@ -6344,8 +6343,11 @@ static int ocfs2_free_cached_blocks(struct ocfs2_super *osb,
 	}
 
 	while (head) {
-		bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
-						      head->free_bit);
+		if (head->free_bg)
+			bg_blkno = head->free_bg;
+		else
+			bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
+							      head->free_bit);
 		mlog(0, "Free bit: (bit %u, blkno %llu)\n",
 		     head->free_bit, (unsigned long long)head->free_blk);
 
@@ -6393,7 +6395,7 @@ int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 	int ret = 0;
 	struct ocfs2_cached_block_free *item;
 
-	item = kmalloc(sizeof(*item), GFP_NOFS);
+	item = kzalloc(sizeof(*item), GFP_NOFS);
 	if (item == NULL) {
 		ret = -ENOMEM;
 		mlog_errno(ret);
@@ -6533,8 +6535,8 @@ ocfs2_find_per_slot_free_list(int type,
 }
 
 int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
-			      int type, int slot, u64 blkno,
-			      unsigned int bit)
+			      int type, int slot, u64 suballoc,
+			      u64 blkno, unsigned int bit)
 {
 	int ret;
 	struct ocfs2_per_slot_free_list *fl;
@@ -6547,7 +6549,7 @@ int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 		goto out;
 	}
 
-	item = kmalloc(sizeof(*item), GFP_NOFS);
+	item = kzalloc(sizeof(*item), GFP_NOFS);
 	if (item == NULL) {
 		ret = -ENOMEM;
 		mlog_errno(ret);
@@ -6557,6 +6559,7 @@ int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 	mlog(0, "Insert: (type %d, slot %u, bit %u, blk %llu)\n",
 	     type, slot, bit, (unsigned long long)blkno);
 
+	item->free_bg = suballoc;
 	item->free_blk = blkno;
 	item->free_bit = bit;
 	item->free_next = fl->f_first;
@@ -6573,433 +6576,11 @@ static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
 {
 	return ocfs2_cache_block_dealloc(ctxt, EXTENT_ALLOC_SYSTEM_INODE,
 					 le16_to_cpu(eb->h_suballoc_slot),
+					 le64_to_cpu(eb->h_suballoc_loc),
 					 le64_to_cpu(eb->h_blkno),
 					 le16_to_cpu(eb->h_suballoc_bit));
 }
 
-/* This function will figure out whether the currently last extent
- * block will be deleted, and if it will, what the new last extent
- * block will be so we can update his h_next_leaf_blk field, as well
- * as the dinodes i_last_eb_blk */
-static int ocfs2_find_new_last_ext_blk(struct inode *inode,
-				       unsigned int clusters_to_del,
-				       struct ocfs2_path *path,
-				       struct buffer_head **new_last_eb)
-{
-	int next_free, ret = 0;
-	u32 cpos;
-	struct ocfs2_extent_rec *rec;
-	struct ocfs2_extent_block *eb;
-	struct ocfs2_extent_list *el;
-	struct buffer_head *bh = NULL;
-
-	*new_last_eb = NULL;
-
-	/* we have no tree, so of course, no last_eb. */
-	if (!path->p_tree_depth)
-		goto out;
-
-	/* trunc to zero special case - this makes tree_depth = 0
-	 * regardless of what it is.  */
-	if (OCFS2_I(inode)->ip_clusters == clusters_to_del)
-		goto out;
-
-	el = path_leaf_el(path);
-	BUG_ON(!el->l_next_free_rec);
-
-	/*
-	 * Make sure that this extent list will actually be empty
-	 * after we clear away the data. We can shortcut out if
-	 * there's more than one non-empty extent in the
-	 * list. Otherwise, a check of the remaining extent is
-	 * necessary.
-	 */
-	next_free = le16_to_cpu(el->l_next_free_rec);
-	rec = NULL;
-	if (ocfs2_is_empty_extent(&el->l_recs[0])) {
-		if (next_free > 2)
-			goto out;
-
-		/* We may have a valid extent in index 1, check it. */
-		if (next_free == 2)
-			rec = &el->l_recs[1];
-
-		/*
-		 * Fall through - no more nonempty extents, so we want
-		 * to delete this leaf.
-		 */
-	} else {
-		if (next_free > 1)
-			goto out;
-
-		rec = &el->l_recs[0];
-	}
-
-	if (rec) {
-		/*
-		 * Check it we'll only be trimming off the end of this
-		 * cluster.
-		 */
-		if (le16_to_cpu(rec->e_leaf_clusters) > clusters_to_del)
-			goto out;
-	}
-
-	ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	ret = ocfs2_find_leaf(INODE_CACHE(inode), path_root_el(path), cpos, &bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	eb = (struct ocfs2_extent_block *) bh->b_data;
-	el = &eb->h_list;
-
-	/* ocfs2_find_leaf() gets the eb from ocfs2_read_extent_block().
-	 * Any corruption is a code bug. */
-	BUG_ON(!OCFS2_IS_VALID_EXTENT_BLOCK(eb));
-
-	*new_last_eb = bh;
-	get_bh(*new_last_eb);
-	mlog(0, "returning block %llu, (cpos: %u)\n",
-	     (unsigned long long)le64_to_cpu(eb->h_blkno), cpos);
-out:
-	brelse(bh);
-
-	return ret;
-}
-
-/*
- * Trim some clusters off the rightmost edge of a tree. Only called
- * during truncate.
- *
- * The caller needs to:
- *   - start journaling of each path component.
- *   - compute and fully set up any new last ext block
- */
-static int ocfs2_trim_tree(struct inode *inode, struct ocfs2_path *path,
-			   handle_t *handle, struct ocfs2_truncate_context *tc,
-			   u32 clusters_to_del, u64 *delete_start, u8 *flags)
-{
-	int ret, i, index = path->p_tree_depth;
-	u32 new_edge = 0;
-	u64 deleted_eb = 0;
-	struct buffer_head *bh;
-	struct ocfs2_extent_list *el;
-	struct ocfs2_extent_rec *rec;
-
-	*delete_start = 0;
-	*flags = 0;
-
-	while (index >= 0) {
-		bh = path->p_node[index].bh;
-		el = path->p_node[index].el;
-
-		mlog(0, "traveling tree (index = %d, block = %llu)\n",
-		     index,  (unsigned long long)bh->b_blocknr);
-
-		BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
-
-		if (index !=
-		    (path->p_tree_depth - le16_to_cpu(el->l_tree_depth))) {
-			ocfs2_error(inode->i_sb,
-				    "Inode %lu has invalid ext. block %llu",
-				    inode->i_ino,
-				    (unsigned long long)bh->b_blocknr);
-			ret = -EROFS;
-			goto out;
-		}
-
-find_tail_record:
-		i = le16_to_cpu(el->l_next_free_rec) - 1;
-		rec = &el->l_recs[i];
-
-		mlog(0, "Extent list before: record %d: (%u, %u, %llu), "
-		     "next = %u\n", i, le32_to_cpu(rec->e_cpos),
-		     ocfs2_rec_clusters(el, rec),
-		     (unsigned long long)le64_to_cpu(rec->e_blkno),
-		     le16_to_cpu(el->l_next_free_rec));
-
-		BUG_ON(ocfs2_rec_clusters(el, rec) < clusters_to_del);
-
-		if (le16_to_cpu(el->l_tree_depth) == 0) {
-			/*
-			 * If the leaf block contains a single empty
-			 * extent and no records, we can just remove
-			 * the block.
-			 */
-			if (i == 0 && ocfs2_is_empty_extent(rec)) {
-				memset(rec, 0,
-				       sizeof(struct ocfs2_extent_rec));
-				el->l_next_free_rec = cpu_to_le16(0);
-
-				goto delete;
-			}
-
-			/*
-			 * Remove any empty extents by shifting things
-			 * left. That should make life much easier on
-			 * the code below. This condition is rare
-			 * enough that we shouldn't see a performance
-			 * hit.
-			 */
-			if (ocfs2_is_empty_extent(&el->l_recs[0])) {
-				le16_add_cpu(&el->l_next_free_rec, -1);
-
-				for(i = 0;
-				    i < le16_to_cpu(el->l_next_free_rec); i++)
-					el->l_recs[i] = el->l_recs[i + 1];
-
-				memset(&el->l_recs[i], 0,
-				       sizeof(struct ocfs2_extent_rec));
-
-				/*
-				 * We've modified our extent list. The
-				 * simplest way to handle this change
-				 * is to being the search from the
-				 * start again.
-				 */
-				goto find_tail_record;
-			}
-
-			le16_add_cpu(&rec->e_leaf_clusters, -clusters_to_del);
-
-			/*
-			 * We'll use "new_edge" on our way back up the
-			 * tree to know what our rightmost cpos is.
-			 */
-			new_edge = le16_to_cpu(rec->e_leaf_clusters);
-			new_edge += le32_to_cpu(rec->e_cpos);
-
-			/*
-			 * The caller will use this to delete data blocks.
-			 */
-			*delete_start = le64_to_cpu(rec->e_blkno)
-				+ ocfs2_clusters_to_blocks(inode->i_sb,
-					le16_to_cpu(rec->e_leaf_clusters));
-			*flags = rec->e_flags;
-
-			/*
-			 * If it's now empty, remove this record.
-			 */
-			if (le16_to_cpu(rec->e_leaf_clusters) == 0) {
-				memset(rec, 0,
-				       sizeof(struct ocfs2_extent_rec));
-				le16_add_cpu(&el->l_next_free_rec, -1);
-			}
-		} else {
-			if (le64_to_cpu(rec->e_blkno) == deleted_eb) {
-				memset(rec, 0,
-				       sizeof(struct ocfs2_extent_rec));
-				le16_add_cpu(&el->l_next_free_rec, -1);
-
-				goto delete;
-			}
-
-			/* Can this actually happen? */
-			if (le16_to_cpu(el->l_next_free_rec) == 0)
-				goto delete;
-
-			/*
-			 * We never actually deleted any clusters
-			 * because our leaf was empty. There's no
-			 * reason to adjust the rightmost edge then.
-			 */
-			if (new_edge == 0)
-				goto delete;
-
-			rec->e_int_clusters = cpu_to_le32(new_edge);
-			le32_add_cpu(&rec->e_int_clusters,
-				     -le32_to_cpu(rec->e_cpos));
-
-			 /*
-			  * A deleted child record should have been
-			  * caught above.
-			  */
-			 BUG_ON(le32_to_cpu(rec->e_int_clusters) == 0);
-		}
-
-delete:
-		ret = ocfs2_journal_dirty(handle, bh);
-		if (ret) {
-			mlog_errno(ret);
-			goto out;
-		}
-
-		mlog(0, "extent list container %llu, after: record %d: "
-		     "(%u, %u, %llu), next = %u.\n",
-		     (unsigned long long)bh->b_blocknr, i,
-		     le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec),
-		     (unsigned long long)le64_to_cpu(rec->e_blkno),
-		     le16_to_cpu(el->l_next_free_rec));
-
-		/*
-		 * We must be careful to only attempt delete of an
-		 * extent block (and not the root inode block).
-		 */
-		if (index > 0 && le16_to_cpu(el->l_next_free_rec) == 0) {
-			struct ocfs2_extent_block *eb =
-				(struct ocfs2_extent_block *)bh->b_data;
-
-			/*
-			 * Save this for use when processing the
-			 * parent block.
-			 */
-			deleted_eb = le64_to_cpu(eb->h_blkno);
-
-			mlog(0, "deleting this extent block.\n");
-
-			ocfs2_remove_from_cache(INODE_CACHE(inode), bh);
-
-			BUG_ON(ocfs2_rec_clusters(el, &el->l_recs[0]));
-			BUG_ON(le32_to_cpu(el->l_recs[0].e_cpos));
-			BUG_ON(le64_to_cpu(el->l_recs[0].e_blkno));
-
-			ret = ocfs2_cache_extent_block_free(&tc->tc_dealloc, eb);
-			/* An error here is not fatal. */
-			if (ret < 0)
-				mlog_errno(ret);
-		} else {
-			deleted_eb = 0;
-		}
-
-		index--;
-	}
-
-	ret = 0;
-out:
-	return ret;
-}
-
-static int ocfs2_do_truncate(struct ocfs2_super *osb,
-			     unsigned int clusters_to_del,
-			     struct inode *inode,
-			     struct buffer_head *fe_bh,
-			     handle_t *handle,
-			     struct ocfs2_truncate_context *tc,
-			     struct ocfs2_path *path,
-			     struct ocfs2_alloc_context *meta_ac)
-{
-	int status;
-	struct ocfs2_dinode *fe;
-	struct ocfs2_extent_block *last_eb = NULL;
-	struct ocfs2_extent_list *el;
-	struct buffer_head *last_eb_bh = NULL;
-	u64 delete_blk = 0;
-	u8 rec_flags;
-
-	fe = (struct ocfs2_dinode *) fe_bh->b_data;
-
-	status = ocfs2_find_new_last_ext_blk(inode, clusters_to_del,
-					     path, &last_eb_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	/*
-	 * Each component will be touched, so we might as well journal
-	 * here to avoid having to handle errors later.
-	 */
-	status = ocfs2_journal_access_path(INODE_CACHE(inode), handle, path);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	if (last_eb_bh) {
-		status = ocfs2_journal_access_eb(handle, INODE_CACHE(inode), last_eb_bh,
-						 OCFS2_JOURNAL_ACCESS_WRITE);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-
-		last_eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
-	}
-
-	el = &(fe->id2.i_list);
-
-	/*
-	 * Lower levels depend on this never happening, but it's best
-	 * to check it up here before changing the tree.
-	 */
-	if (el->l_tree_depth && el->l_recs[0].e_int_clusters == 0) {
-		ocfs2_error(inode->i_sb,
-			    "Inode %lu has an empty extent record, depth %u\n",
-			    inode->i_ino, le16_to_cpu(el->l_tree_depth));
-		status = -EROFS;
-		goto bail;
-	}
-
-	dquot_free_space_nodirty(inode,
-			ocfs2_clusters_to_bytes(osb->sb, clusters_to_del));
-	spin_lock(&OCFS2_I(inode)->ip_lock);
-	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters) -
-				      clusters_to_del;
-	spin_unlock(&OCFS2_I(inode)->ip_lock);
-	le32_add_cpu(&fe->i_clusters, -clusters_to_del);
-	inode->i_blocks = ocfs2_inode_sector_count(inode);
-
-	status = ocfs2_trim_tree(inode, path, handle, tc,
-				 clusters_to_del, &delete_blk, &rec_flags);
-	if (status) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	if (le32_to_cpu(fe->i_clusters) == 0) {
-		/* trunc to zero is a special case. */
-		el->l_tree_depth = 0;
-		fe->i_last_eb_blk = 0;
-	} else if (last_eb)
-		fe->i_last_eb_blk = last_eb->h_blkno;
-
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	if (last_eb) {
-		/* If there will be a new last extent block, then by
-		 * definition, there cannot be any leaves to the right of
-		 * him. */
-		last_eb->h_next_leaf_blk = 0;
-		status = ocfs2_journal_dirty(handle, last_eb_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-
-	if (delete_blk) {
-		if (rec_flags & OCFS2_EXT_REFCOUNTED)
-			status = ocfs2_decrease_refcount(inode, handle,
-					ocfs2_blocks_to_clusters(osb->sb,
-								 delete_blk),
-					clusters_to_del, meta_ac,
-					&tc->tc_dealloc, 1);
-		else
-			status = ocfs2_truncate_log_append(osb, handle,
-							   delete_blk,
-							   clusters_to_del);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-	status = 0;
-bail:
-	brelse(last_eb_bh);
-	mlog_exit(status);
-	return status;
-}
-
 static int ocfs2_zero_func(handle_t *handle, struct buffer_head *bh)
 {
 	set_buffer_uptodate(bh);
@@ -7307,7 +6888,9 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode,
 			goto out_commit;
 		did_quota = 1;
 
-		ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
+		data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv;
+
+		ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off,
 					   &num);
 		if (ret) {
 			mlog_errno(ret);
@@ -7406,26 +6989,29 @@ out:
  */
 int ocfs2_commit_truncate(struct ocfs2_super *osb,
 			  struct inode *inode,
-			  struct buffer_head *fe_bh,
-			  struct ocfs2_truncate_context *tc)
+			  struct buffer_head *di_bh)
 {
-	int status, i, credits, tl_sem = 0;
-	u32 clusters_to_del, new_highest_cpos, range;
+	int status = 0, i, flags = 0;
+	u32 new_highest_cpos, range, trunc_cpos, trunc_len, phys_cpos, coff;
 	u64 blkno = 0;
 	struct ocfs2_extent_list *el;
-	handle_t *handle = NULL;
-	struct inode *tl_inode = osb->osb_tl_inode;
+	struct ocfs2_extent_rec *rec;
 	struct ocfs2_path *path = NULL;
-	struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data;
-	struct ocfs2_alloc_context *meta_ac = NULL;
-	struct ocfs2_refcount_tree *ref_tree = NULL;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	struct ocfs2_extent_list *root_el = &(di->id2.i_list);
+	u64 refcount_loc = le64_to_cpu(di->i_refcount_loc);
+	struct ocfs2_extent_tree et;
+	struct ocfs2_cached_dealloc_ctxt dealloc;
 
 	mlog_entry_void();
 
+	ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
+	ocfs2_init_dealloc_ctxt(&dealloc);
+
 	new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb,
 						     i_size_read(inode));
 
-	path = ocfs2_new_path(fe_bh, &di->id2.i_list,
+	path = ocfs2_new_path(di_bh, &di->id2.i_list,
 			      ocfs2_journal_access_di);
 	if (!path) {
 		status = -ENOMEM;
@@ -7444,8 +7030,6 @@ start:
 		goto bail;
 	}
 
-	credits = 0;
-
 	/*
 	 * Truncate always works against the rightmost tree branch.
 	 */
@@ -7480,101 +7064,62 @@ start:
 	}
 
 	i = le16_to_cpu(el->l_next_free_rec) - 1;
-	range = le32_to_cpu(el->l_recs[i].e_cpos) +
-		ocfs2_rec_clusters(el, &el->l_recs[i]);
-	if (i == 0 && ocfs2_is_empty_extent(&el->l_recs[i])) {
-		clusters_to_del = 0;
-	} else if (le32_to_cpu(el->l_recs[i].e_cpos) >= new_highest_cpos) {
-		clusters_to_del = ocfs2_rec_clusters(el, &el->l_recs[i]);
-		blkno = le64_to_cpu(el->l_recs[i].e_blkno);
+	rec = &el->l_recs[i];
+	flags = rec->e_flags;
+	range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
+
+	if (i == 0 && ocfs2_is_empty_extent(rec)) {
+		/*
+		 * Lower levels depend on this never happening, but it's best
+		 * to check it up here before changing the tree.
+		*/
+		if (root_el->l_tree_depth && rec->e_int_clusters == 0) {
+			ocfs2_error(inode->i_sb, "Inode %lu has an empty "
+				    "extent record, depth %u\n", inode->i_ino,
+				    le16_to_cpu(root_el->l_tree_depth));
+			status = -EROFS;
+			goto bail;
+		}
+		trunc_cpos = le32_to_cpu(rec->e_cpos);
+		trunc_len = 0;
+		blkno = 0;
+	} else if (le32_to_cpu(rec->e_cpos) >= new_highest_cpos) {
+		/*
+		 * Truncate entire record.
+		 */
+		trunc_cpos = le32_to_cpu(rec->e_cpos);
+		trunc_len = ocfs2_rec_clusters(el, rec);
+		blkno = le64_to_cpu(rec->e_blkno);
 	} else if (range > new_highest_cpos) {
-		clusters_to_del = (ocfs2_rec_clusters(el, &el->l_recs[i]) +
-				   le32_to_cpu(el->l_recs[i].e_cpos)) -
-				  new_highest_cpos;
-		blkno = le64_to_cpu(el->l_recs[i].e_blkno) +
-			ocfs2_clusters_to_blocks(inode->i_sb,
-				ocfs2_rec_clusters(el, &el->l_recs[i]) -
-				clusters_to_del);
+		/*
+		 * Partial truncate. it also should be
+		 * the last truncate we're doing.
+		 */
+		trunc_cpos = new_highest_cpos;
+		trunc_len = range - new_highest_cpos;
+		coff = new_highest_cpos - le32_to_cpu(rec->e_cpos);
+		blkno = le64_to_cpu(rec->e_blkno) +
+				ocfs2_clusters_to_blocks(inode->i_sb, coff);
 	} else {
+		/*
+		 * Truncate completed, leave happily.
+		 */
 		status = 0;
 		goto bail;
 	}
 
-	mlog(0, "clusters_to_del = %u in this pass, tail blk=%llu\n",
-	     clusters_to_del, (unsigned long long)path_leaf_bh(path)->b_blocknr);
-
-	if (el->l_recs[i].e_flags & OCFS2_EXT_REFCOUNTED && clusters_to_del) {
-		BUG_ON(!(OCFS2_I(inode)->ip_dyn_features &
-			 OCFS2_HAS_REFCOUNT_FL));
-
-		status = ocfs2_lock_refcount_tree(osb,
-						le64_to_cpu(di->i_refcount_loc),
-						1, &ref_tree, NULL);
-		if (status) {
-			mlog_errno(status);
-			goto bail;
-		}
-
-		status = ocfs2_prepare_refcount_change_for_del(inode, fe_bh,
-							       blkno,
-							       clusters_to_del,
-							       &credits,
-							       &meta_ac);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-
-	mutex_lock(&tl_inode->i_mutex);
-	tl_sem = 1;
-	/* ocfs2_truncate_log_needs_flush guarantees us at least one
-	 * record is free for use. If there isn't any, we flush to get
-	 * an empty truncate log.  */
-	if (ocfs2_truncate_log_needs_flush(osb)) {
-		status = __ocfs2_flush_truncate_log(osb);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
+	phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
 
-	credits += ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
-						(struct ocfs2_dinode *)fe_bh->b_data,
-						el);
-	handle = ocfs2_start_trans(osb, credits);
-	if (IS_ERR(handle)) {
-		status = PTR_ERR(handle);
-		handle = NULL;
-		mlog_errno(status);
-		goto bail;
-	}
-
-	status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh, handle,
-				   tc, path, meta_ac);
+	status = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
+					  phys_cpos, trunc_len, flags, &dealloc,
+					  refcount_loc);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
 
-	mutex_unlock(&tl_inode->i_mutex);
-	tl_sem = 0;
-
-	ocfs2_commit_trans(osb, handle);
-	handle = NULL;
-
 	ocfs2_reinit_path(path, 1);
 
-	if (meta_ac) {
-		ocfs2_free_alloc_context(meta_ac);
-		meta_ac = NULL;
-	}
-
-	if (ref_tree) {
-		ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
-		ref_tree = NULL;
-	}
-
 	/*
 	 * The check above will catch the case where we've truncated
 	 * away all allocation.
@@ -7585,25 +7130,10 @@ bail:
 
 	ocfs2_schedule_truncate_log_flush(osb, 1);
 
-	if (tl_sem)
-		mutex_unlock(&tl_inode->i_mutex);
-
-	if (handle)
-		ocfs2_commit_trans(osb, handle);
-
-	if (meta_ac)
-		ocfs2_free_alloc_context(meta_ac);
-
-	if (ref_tree)
-		ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
-
-	ocfs2_run_deallocs(osb, &tc->tc_dealloc);
+	ocfs2_run_deallocs(osb, &dealloc);
 
 	ocfs2_free_path(path);
 
-	/* This will drop the ext_alloc cluster lock for us */
-	ocfs2_free_truncate_context(tc);
-
 	mlog_exit(status);
 	return status;
 }

fs/ocfs2/alloc.h

@@ -140,8 +140,9 @@ int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
 			struct ocfs2_cached_dealloc_ctxt *dealloc);
 int ocfs2_remove_btree_range(struct inode *inode,
 			     struct ocfs2_extent_tree *et,
-			     u32 cpos, u32 phys_cpos, u32 len,
-			     struct ocfs2_cached_dealloc_ctxt *dealloc);
+			     u32 cpos, u32 phys_cpos, u32 len, int flags,
+			     struct ocfs2_cached_dealloc_ctxt *dealloc,
+			     u64 refcount_loc);
 
 int ocfs2_num_free_extents(struct ocfs2_super *osb,
 			   struct ocfs2_extent_tree *et);
@@ -209,7 +210,7 @@ static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
 int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 				u64 blkno, unsigned int bit);
 int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
-			      int type, int slot, u64 blkno,
+			      int type, int slot, u64 suballoc, u64 blkno,
 			      unsigned int bit);
 static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
 {
@@ -233,8 +234,7 @@ int ocfs2_prepare_truncate(struct ocfs2_super *osb,
 			   struct ocfs2_truncate_context **tc);
 int ocfs2_commit_truncate(struct ocfs2_super *osb,
 			  struct inode *inode,
-			  struct buffer_head *fe_bh,
-			  struct ocfs2_truncate_context *tc);
+			  struct buffer_head *di_bh);
 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 			  unsigned int start, unsigned int end, int trunc);
 
@@ -319,6 +319,8 @@ int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
 			      struct ocfs2_path *path);
 int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
 				   struct ocfs2_path *path, u32 *cpos);
+int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
+				  struct ocfs2_path *path, u32 *cpos);
 int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
 			    struct ocfs2_path *left,
 			    struct ocfs2_path *right);

fs/ocfs2/aops.c

@@ -1735,6 +1735,9 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
 			goto out;
 		}
 
+		if (data_ac)
+			data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv;
+
 		credits = ocfs2_calc_extend_credits(inode->i_sb,
 						    &di->id2.i_list,
 						    clusters_to_alloc);

fs/ocfs2/cluster/masklog.c

@@ -116,6 +116,7 @@ static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = {
 	define_mask(ERROR),
 	define_mask(NOTICE),
 	define_mask(KTHREAD),
+	define_mask(RESERVATIONS),
 };
 
 static struct attribute *mlog_attr_ptrs[MLOG_MAX_BITS] = {NULL, };

fs/ocfs2/cluster/masklog.h

@@ -119,6 +119,7 @@
 #define ML_ERROR	0x0000000100000000ULL /* sent to KERN_ERR */
 #define ML_NOTICE	0x0000000200000000ULL /* setn to KERN_NOTICE */
 #define ML_KTHREAD	0x0000000400000000ULL /* kernel thread activity */
+#define	ML_RESERVATIONS	0x0000000800000000ULL /* ocfs2 alloc reservations */
 
 #define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
 #define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT)

fs/ocfs2/cluster/tcp.c

@@ -583,6 +583,9 @@ static void o2net_state_change(struct sock *sk)
 			o2net_sc_queue_work(sc, &sc->sc_connect_work);
 			break;
 		default:
+			printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT
+			      " shutdown, state %d\n",
+			      SC_NODEF_ARGS(sc), sk->sk_state);
 			o2net_sc_queue_work(sc, &sc->sc_shutdown_work);
 			break;
 	}

fs/ocfs2/dir.c

@@ -1194,7 +1194,7 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir,
 			else
 				de->inode = 0;
 			dir->i_version++;
-			status = ocfs2_journal_dirty(handle, bh);
+			ocfs2_journal_dirty(handle, bh);
 			goto bail;
 		}
 		i += le16_to_cpu(de->rec_len);
@@ -1752,7 +1752,7 @@ int __ocfs2_add_entry(handle_t *handle,
 				ocfs2_recalc_free_list(dir, handle, lookup);
 
 			dir->i_version++;
-			status = ocfs2_journal_dirty(handle, insert_bh);
+			ocfs2_journal_dirty(handle, insert_bh);
 			retval = 0;
 			goto bail;
 		}
@@ -2297,12 +2297,7 @@ static int ocfs2_fill_new_dir_id(struct ocfs2_super *osb,
 	}
 
 	ocfs2_fill_initial_dirents(inode, parent, data->id_data, size);
-
 	ocfs2_journal_dirty(handle, di_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
 
 	i_size_write(inode, size);
 	inode->i_nlink = 2;
@@ -2366,11 +2361,7 @@ static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb,
 		ocfs2_init_dir_trailer(inode, new_bh, size);
 	}
 
-	status = ocfs2_journal_dirty(handle, new_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, new_bh);
 
 	i_size_write(inode, inode->i_sb->s_blocksize);
 	inode->i_nlink = 2;
@@ -2404,15 +2395,15 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 	int ret;
 	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
 	u16 dr_suballoc_bit;
-	u64 dr_blkno;
+	u64 suballoc_loc, dr_blkno;
 	unsigned int num_bits;
 	struct buffer_head *dx_root_bh = NULL;
 	struct ocfs2_dx_root_block *dx_root;
 	struct ocfs2_dir_block_trailer *trailer =
 		ocfs2_trailer_from_bh(dirdata_bh, dir->i_sb);
 
-	ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1, &dr_suballoc_bit,
-				   &num_bits, &dr_blkno);
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
+				   &dr_suballoc_bit, &num_bits, &dr_blkno);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -2440,6 +2431,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 	memset(dx_root, 0, osb->sb->s_blocksize);
 	strcpy(dx_root->dr_signature, OCFS2_DX_ROOT_SIGNATURE);
 	dx_root->dr_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	dx_root->dr_suballoc_loc = cpu_to_le64(suballoc_loc);
 	dx_root->dr_suballoc_bit = cpu_to_le16(dr_suballoc_bit);
 	dx_root->dr_fs_generation = cpu_to_le32(osb->fs_generation);
 	dx_root->dr_blkno = cpu_to_le64(dr_blkno);
@@ -2458,10 +2450,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 		dx_root->dr_list.l_count =
 			cpu_to_le16(ocfs2_extent_recs_per_dx_root(osb->sb));
 	}
-
-	ret = ocfs2_journal_dirty(handle, dx_root_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, dx_root_bh);
 
 	ret = ocfs2_journal_access_di(handle, INODE_CACHE(dir), di_bh,
 				      OCFS2_JOURNAL_ACCESS_CREATE);
@@ -2475,9 +2464,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 	OCFS2_I(dir)->ip_dyn_features |= OCFS2_INDEXED_DIR_FL;
 	di->i_dyn_features = cpu_to_le16(OCFS2_I(dir)->ip_dyn_features);
 
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, di_bh);
 
 	*ret_dx_root_bh = dx_root_bh;
 	dx_root_bh = NULL;
@@ -2558,7 +2545,7 @@ static int __ocfs2_dx_dir_new_cluster(struct inode *dir,
 	 * chance of contiguousness as the directory grows in number
 	 * of entries.
 	 */
-	ret = __ocfs2_claim_clusters(osb, handle, data_ac, 1, 1, &phys, &num);
+	ret = __ocfs2_claim_clusters(handle, data_ac, 1, 1, &phys, &num);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -2991,7 +2978,9 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 * if we only get one now, that's enough to continue. The rest
 	 * will be claimed after the conversion to extents.
 	 */
-	ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, &len);
+	if (ocfs2_dir_resv_allowed(osb))
+		data_ac->ac_resv = &oi->ip_la_data_resv;
+	ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off, &len);
 	if (ret) {
 		mlog_errno(ret);
 		goto out_commit;
@@ -3034,11 +3023,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 		ocfs2_init_dir_trailer(dir, dirdata_bh, i);
 	}
 
-	ret = ocfs2_journal_dirty(handle, dirdata_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, dirdata_bh);
 
 	if (ocfs2_supports_indexed_dirs(osb) && !dx_inline) {
 		/*
@@ -3104,11 +3089,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 */
 	dir->i_blocks = ocfs2_inode_sector_count(dir);
 
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, di_bh);
 
 	if (ocfs2_supports_indexed_dirs(osb)) {
 		ret = ocfs2_dx_dir_attach_index(osb, handle, dir, di_bh,
@@ -3138,7 +3119,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 * pass. Claim the 2nd cluster as a separate extent.
 	 */
 	if (alloc > len) {
-		ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
+		ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off,
 					   &len);
 		if (ret) {
 			mlog_errno(ret);
@@ -3369,6 +3350,9 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
 			goto bail;
 		}
 
+		if (ocfs2_dir_resv_allowed(osb))
+			data_ac->ac_resv = &OCFS2_I(dir)->ip_la_data_resv;
+
 		credits = ocfs2_calc_extend_credits(sb, el, 1);
 	} else {
 		spin_unlock(&OCFS2_I(dir)->ip_lock);
@@ -3423,11 +3407,7 @@ do_extend:
 	} else {
 		de->rec_len = cpu_to_le16(sb->s_blocksize);
 	}
-	status = ocfs2_journal_dirty(handle, new_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, new_bh);
 
 	dir_i_size += dir->i_sb->s_blocksize;
 	i_size_write(dir, dir_i_size);
@@ -3906,11 +3886,7 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir,
 	     sizeof(struct ocfs2_dx_entry), dx_leaf_sort_cmp,
 	     dx_leaf_sort_swap);
 
-	ret = ocfs2_journal_dirty(handle, dx_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, dx_leaf_bh);
 
 	ret = ocfs2_dx_dir_find_leaf_split(dx_leaf, leaf_cpos, insert_hash,
 					   &split_hash);
@@ -4490,7 +4466,10 @@ static int ocfs2_dx_dir_remove_index(struct inode *dir,
 
 	blk = le64_to_cpu(dx_root->dr_blkno);
 	bit = le16_to_cpu(dx_root->dr_suballoc_bit);
-	bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+	if (dx_root->dr_suballoc_loc)
+		bg_blkno = le64_to_cpu(dx_root->dr_suballoc_loc);
+	else
+		bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 	ret = ocfs2_free_suballoc_bits(handle, dx_alloc_inode, dx_alloc_bh,
 				       bit, bg_blkno, 1);
 	if (ret)
@@ -4551,8 +4530,8 @@ int ocfs2_dx_dir_truncate(struct inode *dir, struct buffer_head *di_bh)
 
 		p_cpos = ocfs2_blocks_to_clusters(dir->i_sb, blkno);
 
-		ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen,
-					       &dealloc);
+		ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen, 0,
+					       &dealloc, 0);
 		if (ret) {
 			mlog_errno(ret);
 			goto out;

fs/ocfs2/dlm/dlmast.c

@@ -88,7 +88,7 @@ static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 	return 0;
 }
 
-static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 {
 	mlog_entry_void();
 
@@ -145,7 +145,7 @@ void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 }
 
 
-static void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 {
 	mlog_entry_void();
 
@@ -451,7 +451,9 @@ int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
 	ret = o2net_send_message_vec(DLM_PROXY_AST_MSG, dlm->key, vec, veclen,
 				     lock->ml.node, &status);
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", ret, DLM_PROXY_AST_MSG, dlm->key,
+		     lock->ml.node);
 	else {
 		if (status == DLM_RECOVERING) {
 			mlog(ML_ERROR, "sent AST to node %u, it thinks this "

fs/ocfs2/dlm/dlmcommon.h

@@ -37,7 +37,7 @@
 #define DLM_THREAD_SHUFFLE_INTERVAL    5     // flush everything every 5 passes
 #define DLM_THREAD_MS                  200   // flush at least every 200 ms
 
-#define DLM_HASH_SIZE_DEFAULT	(1 << 14)
+#define DLM_HASH_SIZE_DEFAULT	(1 << 17)
 #if DLM_HASH_SIZE_DEFAULT < PAGE_SIZE
 # define DLM_HASH_PAGES		1
 #else
@@ -904,6 +904,8 @@ void __dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm,
 
 void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
 void dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
+void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
+void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
 void dlm_do_local_ast(struct dlm_ctxt *dlm,
 		      struct dlm_lock_resource *res,
 		      struct dlm_lock *lock);

fs/ocfs2/dlm/dlmconvert.c

@@ -390,7 +390,9 @@ static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm,
 		} else if (ret != DLM_NORMAL && ret != DLM_NOTQUEUED)
 			dlm_error(ret);
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_CONVERT_LOCK_MSG, dlm->key,
+		     res->owner);
 		if (dlm_is_host_down(tmpret)) {
 			/* instead of logging the same network error over
 			 * and over, sleep here and wait for the heartbeat

fs/ocfs2/dlm/dlmdomain.c

@@ -511,7 +511,7 @@ static void __dlm_print_nodes(struct dlm_ctxt *dlm)
 
 	assert_spin_locked(&dlm->spinlock);
 
-	printk(KERN_INFO "ocfs2_dlm: Nodes in domain (\"%s\"): ", dlm->name);
+	printk(KERN_NOTICE "o2dlm: Nodes in domain %s: ", dlm->name);
 
 	while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES,
 				     node + 1)) < O2NM_MAX_NODES) {
@@ -534,7 +534,7 @@ static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
 
 	node = exit_msg->node_idx;
 
-	printk(KERN_INFO "ocfs2_dlm: Node %u leaves domain %s\n", node, dlm->name);
+	printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s\n", node, dlm->name);
 
 	spin_lock(&dlm->spinlock);
 	clear_bit(node, dlm->domain_map);
@@ -565,7 +565,9 @@ static int dlm_send_one_domain_exit(struct dlm_ctxt *dlm,
 	status = o2net_send_message(DLM_EXIT_DOMAIN_MSG, dlm->key,
 				    &leave_msg, sizeof(leave_msg), node,
 				    NULL);
-
+	if (status < 0)
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_EXIT_DOMAIN_MSG, dlm->key, node);
 	mlog(0, "status return %d from o2net_send_message\n", status);
 
 	return status;
@@ -904,7 +906,7 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
 		set_bit(assert->node_idx, dlm->domain_map);
 		__dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
 
-		printk(KERN_INFO "ocfs2_dlm: Node %u joins domain %s\n",
+		printk(KERN_NOTICE "o2dlm: Node %u joins domain %s\n",
 		       assert->node_idx, dlm->name);
 		__dlm_print_nodes(dlm);
 
@@ -962,7 +964,9 @@ static int dlm_send_one_join_cancel(struct dlm_ctxt *dlm,
 				    &cancel_msg, sizeof(cancel_msg), node,
 				    NULL);
 	if (status < 0) {
-		mlog_errno(status);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY,
+		     node);
 		goto bail;
 	}
 
@@ -1029,10 +1033,11 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
 	byte_copymap(join_msg.node_map, dlm->live_nodes_map, O2NM_MAX_NODES);
 
 	status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg,
-				    sizeof(join_msg), node,
-				    &join_resp);
+				    sizeof(join_msg), node, &join_resp);
 	if (status < 0 && status != -ENOPROTOOPT) {
-		mlog_errno(status);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_QUERY_JOIN_MSG, DLM_MOD_KEY,
+		     node);
 		goto bail;
 	}
 	dlm_query_join_wire_to_packet(join_resp, &packet);
@@ -1103,7 +1108,9 @@ static int dlm_send_one_join_assert(struct dlm_ctxt *dlm,
 				    &assert_msg, sizeof(assert_msg), node,
 				    NULL);
 	if (status < 0)
-		mlog_errno(status);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
+		     node);
 
 	return status;
 }
@@ -1516,7 +1523,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
 		goto leave;
 	}
 
-	dlm->name = kmalloc(strlen(domain) + 1, GFP_KERNEL);
+	dlm->name = kstrdup(domain, GFP_KERNEL);
 	if (dlm->name == NULL) {
 		mlog_errno(-ENOMEM);
 		kfree(dlm);
@@ -1550,7 +1557,6 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
 	for (i = 0; i < DLM_HASH_BUCKETS; i++)
 		INIT_HLIST_HEAD(dlm_master_hash(dlm, i));
 
-	strcpy(dlm->name, domain);
 	dlm->key = key;
 	dlm->node_num = o2nm_this_node();
 

fs/ocfs2/dlm/dlmlock.c

@@ -329,7 +329,9 @@ static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
 			BUG();
 		}
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_CREATE_LOCK_MSG, dlm->key,
+		     res->owner);
 		if (dlm_is_host_down(tmpret)) {
 			ret = DLM_RECOVERING;
 			mlog(0, "node %u died so returning DLM_RECOVERING "
@@ -429,7 +431,7 @@ struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie,
 	struct dlm_lock *lock;
 	int kernel_allocated = 0;
 
-	lock = (struct dlm_lock *) kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
+	lock = kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
 	if (!lock)
 		return NULL;
 

fs/ocfs2/dlm/dlmmaster.c

@@ -617,13 +617,11 @@ struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm,
 {
 	struct dlm_lock_resource *res = NULL;
 
-	res = (struct dlm_lock_resource *)
-				kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
+	res = kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
 	if (!res)
 		goto error;
 
-	res->lockname.name = (char *)
-				kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
+	res->lockname.name = kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
 	if (!res->lockname.name)
 		goto error;
 
@@ -757,8 +755,7 @@ lookup:
 		spin_unlock(&dlm->spinlock);
 		mlog(0, "allocating a new resource\n");
 		/* nothing found and we need to allocate one. */
-		alloc_mle = (struct dlm_master_list_entry *)
-			kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
+		alloc_mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 		if (!alloc_mle)
 			goto leave;
 		res = dlm_new_lockres(dlm, lockid, namelen);
@@ -1542,8 +1539,7 @@ way_up_top:
 			spin_unlock(&dlm->master_lock);
 			spin_unlock(&dlm->spinlock);
 
-			mle = (struct dlm_master_list_entry *)
-				kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
+			mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 			if (!mle) {
 				response = DLM_MASTER_RESP_ERROR;
 				mlog_errno(-ENOMEM);
@@ -1666,7 +1662,9 @@ again:
 		tmpret = o2net_send_message(DLM_ASSERT_MASTER_MSG, dlm->key,
 					    &assert, sizeof(assert), to, &r);
 		if (tmpret < 0) {
-			mlog(0, "assert_master returned %d!\n", tmpret);
+			mlog(ML_ERROR, "Error %d when sending message %u (key "
+			     "0x%x) to node %u\n", tmpret,
+			     DLM_ASSERT_MASTER_MSG, dlm->key, to);
 			if (!dlm_is_host_down(tmpret)) {
 				mlog(ML_ERROR, "unhandled error=%d!\n", tmpret);
 				BUG();
@@ -2205,7 +2203,9 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 	ret = o2net_send_message(DLM_DEREF_LOCKRES_MSG, dlm->key,
 				 &deref, sizeof(deref), res->owner, &r);
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", ret, DLM_DEREF_LOCKRES_MSG, dlm->key,
+		     res->owner);
 	else if (r < 0) {
 		/* BAD.  other node says I did not have a ref. */
 		mlog(ML_ERROR,"while dropping ref on %s:%.*s "
@@ -2452,8 +2452,7 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
 		goto leave;
 	}
 
-	mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache,
-								GFP_NOFS);
+	mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 	if (!mle) {
 		mlog_errno(ret);
 		goto leave;
@@ -2975,7 +2974,9 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
 					 &migrate, sizeof(migrate), nodenum,
 					 &status);
 		if (ret < 0) {
-			mlog(0, "migrate_request returned %d!\n", ret);
+			mlog(ML_ERROR, "Error %d when sending message %u (key "
+			     "0x%x) to node %u\n", ret, DLM_MIGRATE_REQUEST_MSG,
+			     dlm->key, nodenum);
 			if (!dlm_is_host_down(ret)) {
 				mlog(ML_ERROR, "unhandled error=%d!\n", ret);
 				BUG();
@@ -3033,8 +3034,7 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data,
 	hash = dlm_lockid_hash(name, namelen);
 
 	/* preallocate.. if this fails, abort */
-	mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache,
-							 GFP_NOFS);
+	mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 
 	if (!mle) {
 		ret = -ENOMEM;

fs/ocfs2/dlm/dlmrecovery.c

@@ -803,7 +803,9 @@ static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from,
 
 	/* negative status is handled by caller */
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_LOCK_REQUEST_MSG,
+		     dlm->key, request_from);
 
 	// return from here, then
 	// sleep until all received or error
@@ -955,10 +957,10 @@ static int dlm_send_all_done_msg(struct dlm_ctxt *dlm, u8 dead_node, u8 send_to)
 	ret = o2net_send_message(DLM_RECO_DATA_DONE_MSG, dlm->key, &done_msg,
 				 sizeof(done_msg), send_to, &tmpret);
 	if (ret < 0) {
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_RECO_DATA_DONE_MSG,
+		     dlm->key, send_to);
 		if (!dlm_is_host_down(ret)) {
-			mlog_errno(ret);
-			mlog(ML_ERROR, "%s: unknown error sending data-done "
-			     "to %u\n", dlm->name, send_to);
 			BUG();
 		}
 	} else
@@ -1126,7 +1128,9 @@ static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm,
 	if (ret < 0) {
 		/* XXX: negative status is not handled.
 		 * this will end up killing this node. */
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_MIG_LOCKRES_MSG,
+		     dlm->key, send_to);
 	} else {
 		/* might get an -ENOMEM back here */
 		ret = status;
@@ -1642,7 +1646,9 @@ int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
 				 &req, sizeof(req), nodenum, &status);
 	/* XXX: negative status not handled properly here. */
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_MASTER_REQUERY_MSG,
+		     dlm->key, nodenum);
 	else {
 		BUG_ON(status < 0);
 		BUG_ON(status > DLM_LOCK_RES_OWNER_UNKNOWN);
@@ -2640,7 +2646,7 @@ retry:
 		if (dlm_is_host_down(ret)) {
 			/* node is down.  not involved in recovery
 			 * so just keep going */
-			mlog(0, "%s: node %u was down when sending "
+			mlog(ML_NOTICE, "%s: node %u was down when sending "
 			     "begin reco msg (%d)\n", dlm->name, nodenum, ret);
 			ret = 0;
 		}
@@ -2660,11 +2666,12 @@ retry:
 		}
 		if (ret < 0) {
 			struct dlm_lock_resource *res;
+
 			/* this is now a serious problem, possibly ENOMEM
 			 * in the network stack.  must retry */
 			mlog_errno(ret);
 			mlog(ML_ERROR, "begin reco of dlm %s to node %u "
-			    " returned %d\n", dlm->name, nodenum, ret);
+			     "returned %d\n", dlm->name, nodenum, ret);
 			res = dlm_lookup_lockres(dlm, DLM_RECOVERY_LOCK_NAME,
 						 DLM_RECOVERY_LOCK_NAME_LEN);
 			if (res) {
@@ -2789,7 +2796,9 @@ stage2:
 		if (ret >= 0)
 			ret = status;
 		if (ret < 0) {
-			mlog_errno(ret);
+			mlog(ML_ERROR, "Error %d when sending message %u (key "
+			     "0x%x) to node %u\n", ret, DLM_FINALIZE_RECO_MSG,
+			     dlm->key, nodenum);
 			if (dlm_is_host_down(ret)) {
 				/* this has no effect on this recovery
 				 * session, so set the status to zero to

fs/ocfs2/dlm/dlmthread.c

@@ -309,6 +309,7 @@ static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
 	 * spinlock, and because we know that it is not migrating/
 	 * recovering/in-progress, it is fine to reserve asts and
 	 * basts right before queueing them all throughout */
+	assert_spin_locked(&dlm->ast_lock);
 	assert_spin_locked(&res->spinlock);
 	BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
 			      DLM_LOCK_RES_RECOVERING|
@@ -337,7 +338,7 @@ converting:
 			/* queue the BAST if not already */
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			/* update the highest_blocked if needed */
 			if (lock->ml.highest_blocked < target->ml.convert_type)
@@ -355,7 +356,7 @@ converting:
 			can_grant = 0;
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			if (lock->ml.highest_blocked < target->ml.convert_type)
 				lock->ml.highest_blocked =
@@ -383,7 +384,7 @@ converting:
 		spin_unlock(&target->spinlock);
 
 		__dlm_lockres_reserve_ast(res);
-		dlm_queue_ast(dlm, target);
+		__dlm_queue_ast(dlm, target);
 		/* go back and check for more */
 		goto converting;
 	}
@@ -402,7 +403,7 @@ blocked:
 			can_grant = 0;
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			if (lock->ml.highest_blocked < target->ml.type)
 				lock->ml.highest_blocked = target->ml.type;
@@ -418,7 +419,7 @@ blocked:
 			can_grant = 0;
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			if (lock->ml.highest_blocked < target->ml.type)
 				lock->ml.highest_blocked = target->ml.type;
@@ -444,7 +445,7 @@ blocked:
 		spin_unlock(&target->spinlock);
 
 		__dlm_lockres_reserve_ast(res);
-		dlm_queue_ast(dlm, target);
+		__dlm_queue_ast(dlm, target);
 		/* go back and check for more */
 		goto converting;
 	}
@@ -674,6 +675,7 @@ static int dlm_thread(void *data)
 		 	/* lockres can be re-dirtied/re-added to the
 			 * dirty_list in this gap, but that is ok */
 
+			spin_lock(&dlm->ast_lock);
 			spin_lock(&res->spinlock);
 			if (res->owner != dlm->node_num) {
 				__dlm_print_one_lock_resource(res);
@@ -694,6 +696,7 @@ static int dlm_thread(void *data)
 				/* move it to the tail and keep going */
 				res->state &= ~DLM_LOCK_RES_DIRTY;
 				spin_unlock(&res->spinlock);
+				spin_unlock(&dlm->ast_lock);
 				mlog(0, "delaying list shuffling for in-"
 				     "progress lockres %.*s, state=%d\n",
 				     res->lockname.len, res->lockname.name,
@@ -715,6 +718,7 @@ static int dlm_thread(void *data)
 			dlm_shuffle_lists(dlm, res);
 			res->state &= ~DLM_LOCK_RES_DIRTY;
 			spin_unlock(&res->spinlock);
+			spin_unlock(&dlm->ast_lock);
 
 			dlm_lockres_calc_usage(dlm, res);
 

fs/ocfs2/dlm/dlmunlock.c

@@ -354,7 +354,8 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
 			mlog(0, "master was in-progress.  retry\n");
 		ret = status;
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_UNLOCK_LOCK_MSG, dlm->key, owner);
 		if (dlm_is_host_down(tmpret)) {
 			/* NOTE: this seems strange, but it is what we want.
 			 * when the master goes down during a cancel or

fs/ocfs2/file.c

@@ -278,10 +278,7 @@ int ocfs2_update_inode_atime(struct inode *inode,
 	inode->i_atime = CURRENT_TIME;
 	di->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
 	di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
-
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, bh);
 
 out_commit:
 	ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
@@ -430,9 +427,7 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
 	di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
 	di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
 
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_journal_dirty(handle, fe_bh);
 
 out_commit:
 	ocfs2_commit_trans(osb, handle);
@@ -449,7 +444,6 @@ static int ocfs2_truncate_file(struct inode *inode,
 	int status = 0;
 	struct ocfs2_dinode *fe = NULL;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	struct ocfs2_truncate_context *tc = NULL;
 
 	mlog_entry("(inode = %llu, new_i_size = %llu\n",
 		   (unsigned long long)OCFS2_I(inode)->ip_blkno,
@@ -488,6 +482,9 @@ static int ocfs2_truncate_file(struct inode *inode,
 
 	down_write(&OCFS2_I(inode)->ip_alloc_sem);
 
+	ocfs2_resv_discard(&osb->osb_la_resmap,
+			   &OCFS2_I(inode)->ip_la_data_resv);
+
 	/*
 	 * The inode lock forced other nodes to sync and drop their
 	 * pages, which (correctly) happens even if we have a truncate
@@ -517,13 +514,7 @@ static int ocfs2_truncate_file(struct inode *inode,
 		goto bail_unlock_sem;
 	}
 
-	status = ocfs2_prepare_truncate(osb, inode, di_bh, &tc);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail_unlock_sem;
-	}
-
-	status = ocfs2_commit_truncate(osb, inode, di_bh, tc);
+	status = ocfs2_commit_truncate(osb, inode, di_bh);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail_unlock_sem;
@@ -666,11 +657,7 @@ restarted_transaction:
 		goto leave;
 	}
 
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	spin_lock(&OCFS2_I(inode)->ip_lock);
 	clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
@@ -1195,9 +1182,7 @@ static int __ocfs2_write_remove_suid(struct inode *inode,
 	di = (struct ocfs2_dinode *) bh->b_data;
 	di->i_mode = cpu_to_le16(inode->i_mode);
 
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, bh);
 
 out_trans:
 	ocfs2_commit_trans(osb, handle);
@@ -1434,16 +1419,90 @@ out:
 	return ret;
 }
 
+static int ocfs2_find_rec(struct ocfs2_extent_list *el, u32 pos)
+{
+	int i;
+	struct ocfs2_extent_rec *rec = NULL;
+
+	for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
+
+		rec = &el->l_recs[i];
+
+		if (le32_to_cpu(rec->e_cpos) < pos)
+			break;
+	}
+
+	return i;
+}
+
+/*
+ * Helper to calculate the punching pos and length in one run, we handle the
+ * following three cases in order:
+ *
+ * - remove the entire record
+ * - remove a partial record
+ * - no record needs to be removed (hole-punching completed)
+*/
+static void ocfs2_calc_trunc_pos(struct inode *inode,
+				 struct ocfs2_extent_list *el,
+				 struct ocfs2_extent_rec *rec,
+				 u32 trunc_start, u32 *trunc_cpos,
+				 u32 *trunc_len, u32 *trunc_end,
+				 u64 *blkno, int *done)
+{
+	int ret = 0;
+	u32 coff, range;
+
+	range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
+
+	if (le32_to_cpu(rec->e_cpos) >= trunc_start) {
+		*trunc_cpos = le32_to_cpu(rec->e_cpos);
+		/*
+		 * Skip holes if any.
+		 */
+		if (range < *trunc_end)
+			*trunc_end = range;
+		*trunc_len = *trunc_end - le32_to_cpu(rec->e_cpos);
+		*blkno = le64_to_cpu(rec->e_blkno);
+		*trunc_end = le32_to_cpu(rec->e_cpos);
+	} else if (range > trunc_start) {
+		*trunc_cpos = trunc_start;
+		*trunc_len = *trunc_end - trunc_start;
+		coff = trunc_start - le32_to_cpu(rec->e_cpos);
+		*blkno = le64_to_cpu(rec->e_blkno) +
+				ocfs2_clusters_to_blocks(inode->i_sb, coff);
+		*trunc_end = trunc_start;
+	} else {
+		/*
+		 * It may have two following possibilities:
+		 *
+		 * - last record has been removed
+		 * - trunc_start was within a hole
+		 *
+		 * both two cases mean the completion of hole punching.
+		 */
+		ret = 1;
+	}
+
+	*done = ret;
+}
+
 static int ocfs2_remove_inode_range(struct inode *inode,
 				    struct buffer_head *di_bh, u64 byte_start,
 				    u64 byte_len)
 {
-	int ret = 0;
-	u32 trunc_start, trunc_len, cpos, phys_cpos, alloc_size;
+	int ret = 0, flags = 0, done = 0, i;
+	u32 trunc_start, trunc_len, trunc_end, trunc_cpos, phys_cpos;
+	u32 cluster_in_el;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct ocfs2_cached_dealloc_ctxt dealloc;
 	struct address_space *mapping = inode->i_mapping;
 	struct ocfs2_extent_tree et;
+	struct ocfs2_path *path = NULL;
+	struct ocfs2_extent_list *el = NULL;
+	struct ocfs2_extent_rec *rec = NULL;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	u64 blkno, refcount_loc = le64_to_cpu(di->i_refcount_loc);
 
 	ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
 	ocfs2_init_dealloc_ctxt(&dealloc);
@@ -1469,17 +1528,35 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 		goto out;
 	}
 
+	/*
+	 * For reflinks, we may need to CoW 2 clusters which might be
+	 * partially zero'd later, if hole's start and end offset were
+	 * within one cluster(means is not exactly aligned to clustersize).
+	 */
+
+	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) {
+
+		ret = ocfs2_cow_file_pos(inode, di_bh, byte_start);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		ret = ocfs2_cow_file_pos(inode, di_bh, byte_start + byte_len);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+	}
+
 	trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start);
-	trunc_len = (byte_start + byte_len) >> osb->s_clustersize_bits;
-	if (trunc_len >= trunc_start)
-		trunc_len -= trunc_start;
-	else
-		trunc_len = 0;
+	trunc_end = (byte_start + byte_len) >> osb->s_clustersize_bits;
+	cluster_in_el = trunc_end;
 
-	mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, clen: %u\n",
+	mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, cend: %u\n",
 	     (unsigned long long)OCFS2_I(inode)->ip_blkno,
 	     (unsigned long long)byte_start,
-	     (unsigned long long)byte_len, trunc_start, trunc_len);
+	     (unsigned long long)byte_len, trunc_start, trunc_end);
 
 	ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len);
 	if (ret) {
@@ -1487,31 +1564,79 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 		goto out;
 	}
 
-	cpos = trunc_start;
-	while (trunc_len) {
-		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos,
-					 &alloc_size, NULL);
+	path = ocfs2_new_path_from_et(&et);
+	if (!path) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	while (trunc_end > trunc_start) {
+
+		ret = ocfs2_find_path(INODE_CACHE(inode), path,
+				      cluster_in_el);
 		if (ret) {
 			mlog_errno(ret);
 			goto out;
 		}
 
-		if (alloc_size > trunc_len)
-			alloc_size = trunc_len;
+		el = path_leaf_el(path);
 
-		/* Only do work for non-holes */
-		if (phys_cpos != 0) {
-			ret = ocfs2_remove_btree_range(inode, &et, cpos,
-						       phys_cpos, alloc_size,
-						       &dealloc);
+		i = ocfs2_find_rec(el, trunc_end);
+		/*
+		 * Need to go to previous extent block.
+		 */
+		if (i < 0) {
+			if (path->p_tree_depth == 0)
+				break;
+
+			ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
+							    path,
+							    &cluster_in_el);
 			if (ret) {
 				mlog_errno(ret);
 				goto out;
 			}
+
+			/*
+			 * We've reached the leftmost extent block,
+			 * it's safe to leave.
+			 */
+			if (cluster_in_el == 0)
+				break;
+
+			/*
+			 * The 'pos' searched for previous extent block is
+			 * always one cluster less than actual trunc_end.
+			 */
+			trunc_end = cluster_in_el + 1;
+
+			ocfs2_reinit_path(path, 1);
+
+			continue;
+
+		} else
+			rec = &el->l_recs[i];
+
+		ocfs2_calc_trunc_pos(inode, el, rec, trunc_start, &trunc_cpos,
+				     &trunc_len, &trunc_end, &blkno, &done);
+		if (done)
+			break;
+
+		flags = rec->e_flags;
+		phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
+
+		ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
+					       phys_cpos, trunc_len, flags,
+					       &dealloc, refcount_loc);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
 		}
 
-		cpos += alloc_size;
-		trunc_len -= alloc_size;
+		cluster_in_el = trunc_end;
+
+		ocfs2_reinit_path(path, 1);
 	}
 
 	ocfs2_truncate_cluster_pages(inode, byte_start, byte_len);

fs/ocfs2/inode.c

@@ -376,6 +376,10 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
 
 	OCFS2_I(inode)->ip_last_used_slot = 0;
 	OCFS2_I(inode)->ip_last_used_group = 0;
+
+	if (S_ISDIR(inode->i_mode))
+		ocfs2_resv_set_type(&OCFS2_I(inode)->ip_la_data_resv,
+				    OCFS2_RESV_FLAG_DIR);
 	mlog_exit_void();
 }
 
@@ -539,7 +543,6 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
 				     struct buffer_head *fe_bh)
 {
 	int status = 0;
-	struct ocfs2_truncate_context *tc = NULL;
 	struct ocfs2_dinode *fe;
 	handle_t *handle = NULL;
 
@@ -582,13 +585,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
 		ocfs2_commit_trans(osb, handle);
 		handle = NULL;
 
-		status = ocfs2_prepare_truncate(osb, inode, fe_bh, &tc);
-		if (status < 0) {
-			mlog_errno(status);
-			goto out;
-		}
-
-		status = ocfs2_commit_truncate(osb, inode, fe_bh, tc);
+		status = ocfs2_commit_truncate(osb, inode, fe_bh);
 		if (status < 0) {
 			mlog_errno(status);
 			goto out;
@@ -659,12 +656,7 @@ static int ocfs2_remove_inode(struct inode *inode,
 
 	di->i_dtime = cpu_to_le64(CURRENT_TIME.tv_sec);
 	di->i_flags &= cpu_to_le32(~(OCFS2_VALID_FL | OCFS2_ORPHANED_FL));
-
-	status = ocfs2_journal_dirty(handle, di_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail_commit;
-	}
+	ocfs2_journal_dirty(handle, di_bh);
 
 	ocfs2_remove_from_cache(INODE_CACHE(inode), di_bh);
 	dquot_free_inode(inode);
@@ -980,7 +972,7 @@ static void ocfs2_cleanup_delete_inode(struct inode *inode,
 void ocfs2_delete_inode(struct inode *inode)
 {
 	int wipe, status;
-	sigset_t blocked, oldset;
+	sigset_t oldset;
 	struct buffer_head *di_bh = NULL;
 
 	mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
@@ -1007,13 +999,7 @@ void ocfs2_delete_inode(struct inode *inode)
 	 * messaging paths may return us -ERESTARTSYS. Which would
 	 * cause us to exit early, resulting in inodes being orphaned
 	 * forever. */
-	sigfillset(&blocked);
-	status = sigprocmask(SIG_BLOCK, &blocked, &oldset);
-	if (status < 0) {
-		mlog_errno(status);
-		ocfs2_cleanup_delete_inode(inode, 1);
-		goto bail;
-	}
+	ocfs2_block_signals(&oldset);
 
 	/*
 	 * Synchronize us against ocfs2_get_dentry. We take this in
@@ -1087,9 +1073,7 @@ bail_unlock_nfs_sync:
 	ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0);
 
 bail_unblock:
-	status = sigprocmask(SIG_SETMASK, &oldset, NULL);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_unblock_signals(&oldset);
 bail:
 	clear_inode(inode);
 	mlog_exit_void();
@@ -1123,6 +1107,10 @@ void ocfs2_clear_inode(struct inode *inode)
 	ocfs2_mark_lockres_freeing(&oi->ip_inode_lockres);
 	ocfs2_mark_lockres_freeing(&oi->ip_open_lockres);
 
+	ocfs2_resv_discard(&OCFS2_SB(inode->i_sb)->osb_la_resmap,
+			   &oi->ip_la_data_resv);
+	ocfs2_resv_init_once(&oi->ip_la_data_resv);
+
 	/* We very well may get a clear_inode before all an inodes
 	 * metadata has hit disk. Of course, we can't drop any cluster
 	 * locks until the journal has finished with it. The only
@@ -1298,13 +1286,8 @@ int ocfs2_mark_inode_dirty(handle_t *handle,
 	fe->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
 	fe->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
 
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0)
-		mlog_errno(status);
-
-	status = 0;
+	ocfs2_journal_dirty(handle, bh);
 leave:
-
 	mlog_exit(status);
 	return status;
 }

fs/ocfs2/inode.h

@@ -70,6 +70,8 @@ struct ocfs2_inode_info
 	/* Only valid if the inode is the dir. */
 	u32				ip_last_used_slot;
 	u64				ip_last_used_group;
+
+	struct ocfs2_alloc_reservation	ip_la_data_resv;
 };
 
 /*

fs/ocfs2/journal.c

@@ -402,9 +402,7 @@ int ocfs2_commit_trans(struct ocfs2_super *osb,
 }
 
 /*
- * 'nblocks' is what you want to add to the current
- * transaction. extend_trans will either extend the current handle by
- * nblocks, or commit it and start a new one with nblocks credits.
+ * 'nblocks' is what you want to add to the current transaction.
  *
  * This might call jbd2_journal_restart() which will commit dirty buffers
  * and then restart the transaction. Before calling
@@ -422,11 +420,15 @@ int ocfs2_commit_trans(struct ocfs2_super *osb,
  */
 int ocfs2_extend_trans(handle_t *handle, int nblocks)
 {
-	int status;
+	int status, old_nblocks;
 
 	BUG_ON(!handle);
-	BUG_ON(!nblocks);
+	BUG_ON(nblocks < 0);
+
+	if (!nblocks)
+		return 0;
 
+	old_nblocks = handle->h_buffer_credits;
 	mlog_entry_void();
 
 	mlog(0, "Trying to extend transaction by %d blocks\n", nblocks);
@@ -445,7 +447,8 @@ int ocfs2_extend_trans(handle_t *handle, int nblocks)
 		mlog(0,
 		     "jbd2_journal_extend failed, trying "
 		     "jbd2_journal_restart\n");
-		status = jbd2_journal_restart(handle, nblocks);
+		status = jbd2_journal_restart(handle,
+					      old_nblocks + nblocks);
 		if (status < 0) {
 			mlog_errno(status);
 			goto bail;
@@ -734,8 +737,7 @@ int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
 	return __ocfs2_journal_access(handle, ci, bh, NULL, type);
 }
 
-int ocfs2_journal_dirty(handle_t *handle,
-			struct buffer_head *bh)
+void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh)
 {
 	int status;
 
@@ -743,13 +745,9 @@ int ocfs2_journal_dirty(handle_t *handle,
 		   (unsigned long long)bh->b_blocknr);
 
 	status = jbd2_journal_dirty_metadata(handle, bh);
-	if (status < 0)
-		mlog(ML_ERROR, "Could not dirty metadata buffer. "
-		     "(bh->b_blocknr=%llu)\n",
-		     (unsigned long long)bh->b_blocknr);
+	BUG_ON(status);
 
-	mlog_exit(status);
-	return status;
+	mlog_exit_void();
 }
 
 #define OCFS2_DEFAULT_COMMIT_INTERVAL	(HZ * JBD2_DEFAULT_MAX_COMMIT_AGE)

fs/ocfs2/journal.h

@@ -325,8 +325,7 @@ int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
  *	<modify the bh>
  * 	ocfs2_journal_dirty(handle, bh);
  */
-int                  ocfs2_journal_dirty(handle_t *handle,
-					 struct buffer_head *bh);
+void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh);
 
 /*
  *  Credit Macros:
@@ -562,6 +561,18 @@ static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
 	return blocks;
 }
 
+/*
+ * Allocating a discontiguous block group requires the credits from
+ * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
+ * the group descriptor's extent list.  The caller already has started
+ * the transaction with ocfs2_calc_group_alloc_credits().  They extend
+ * it with these credits.
+ */
+static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
+{
+	return ocfs2_extent_recs_per_gd(sb);
+}
+
 static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
 						unsigned int clusters_to_del,
 						struct ocfs2_dinode *fe,

fs/ocfs2/localalloc.c

@@ -52,7 +52,8 @@ static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc);
 
 static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 					     struct ocfs2_dinode *alloc,
-					     u32 numbits);
+					     u32 *numbits,
+					     struct ocfs2_alloc_reservation *resv);
 
 static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc);
 
@@ -74,6 +75,144 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
 static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
 					  struct inode *local_alloc_inode);
 
+/*
+ * ocfs2_la_default_mb() - determine a default size, in megabytes of
+ * the local alloc.
+ *
+ * Generally, we'd like to pick as large a local alloc as
+ * possible. Performance on large workloads tends to scale
+ * proportionally to la size. In addition to that, the reservations
+ * code functions more efficiently as it can reserve more windows for
+ * write.
+ *
+ * Some things work against us when trying to choose a large local alloc:
+ *
+ * - We need to ensure our sizing is picked to leave enough space in
+ *   group descriptors for other allocations (such as block groups,
+ *   etc). Picking default sizes which are a multiple of 4 could help
+ *   - block groups are allocated in 2mb and 4mb chunks.
+ *
+ * - Likewise, we don't want to starve other nodes of bits on small
+ *   file systems. This can easily be taken care of by limiting our
+ *   default to a reasonable size (256M) on larger cluster sizes.
+ *
+ * - Some file systems can't support very large sizes - 4k and 8k in
+ *   particular are limited to less than 128 and 256 megabytes respectively.
+ *
+ * The following reference table shows group descriptor and local
+ * alloc maximums at various cluster sizes (4k blocksize)
+ *
+ * csize: 4K	group: 126M	la: 121M
+ * csize: 8K	group: 252M	la: 243M
+ * csize: 16K	group: 504M	la: 486M
+ * csize: 32K	group: 1008M	la: 972M
+ * csize: 64K	group: 2016M	la: 1944M
+ * csize: 128K	group: 4032M	la: 3888M
+ * csize: 256K	group: 8064M	la: 7776M
+ * csize: 512K	group: 16128M	la: 15552M
+ * csize: 1024K	group: 32256M	la: 31104M
+ */
+#define	OCFS2_LA_MAX_DEFAULT_MB	256
+#define	OCFS2_LA_OLD_DEFAULT	8
+unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb)
+{
+	unsigned int la_mb;
+	unsigned int gd_mb;
+	unsigned int megs_per_slot;
+	struct super_block *sb = osb->sb;
+
+	gd_mb = ocfs2_clusters_to_megabytes(osb->sb,
+		8 * ocfs2_group_bitmap_size(sb, 0, osb->s_feature_incompat));
+
+	/*
+	 * This takes care of files systems with very small group
+	 * descriptors - 512 byte blocksize at cluster sizes lower
+	 * than 16K and also 1k blocksize with 4k cluster size.
+	 */
+	if ((sb->s_blocksize == 512 && osb->s_clustersize <= 8192)
+	    || (sb->s_blocksize == 1024 && osb->s_clustersize == 4096))
+		return OCFS2_LA_OLD_DEFAULT;
+
+	/*
+	 * Leave enough room for some block groups and make the final
+	 * value we work from a multiple of 4.
+	 */
+	gd_mb -= 16;
+	gd_mb &= 0xFFFFFFFB;
+
+	la_mb = gd_mb;
+
+	/*
+	 * Keep window sizes down to a reasonable default
+	 */
+	if (la_mb > OCFS2_LA_MAX_DEFAULT_MB) {
+		/*
+		 * Some clustersize / blocksize combinations will have
+		 * given us a larger than OCFS2_LA_MAX_DEFAULT_MB
+		 * default size, but get poor distribution when
+		 * limited to exactly 256 megabytes.
+		 *
+		 * As an example, 16K clustersize at 4K blocksize
+		 * gives us a cluster group size of 504M. Paring the
+		 * local alloc size down to 256 however, would give us
+		 * only one window and around 200MB left in the
+		 * cluster group. Instead, find the first size below
+		 * 256 which would give us an even distribution.
+		 *
+		 * Larger cluster group sizes actually work out pretty
+		 * well when pared to 256, so we don't have to do this
+		 * for any group that fits more than two
+		 * OCFS2_LA_MAX_DEFAULT_MB windows.
+		 */
+		if (gd_mb > (2 * OCFS2_LA_MAX_DEFAULT_MB))
+			la_mb = 256;
+		else {
+			unsigned int gd_mult = gd_mb;
+
+			while (gd_mult > 256)
+				gd_mult = gd_mult >> 1;
+
+			la_mb = gd_mult;
+		}
+	}
+
+	megs_per_slot = osb->osb_clusters_at_boot / osb->max_slots;
+	megs_per_slot = ocfs2_clusters_to_megabytes(osb->sb, megs_per_slot);
+	/* Too many nodes, too few disk clusters. */
+	if (megs_per_slot < la_mb)
+		la_mb = megs_per_slot;
+
+	return la_mb;
+}
+
+void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb)
+{
+	struct super_block *sb = osb->sb;
+	unsigned int la_default_mb = ocfs2_la_default_mb(osb);
+	unsigned int la_max_mb;
+
+	la_max_mb = ocfs2_clusters_to_megabytes(sb,
+						ocfs2_local_alloc_size(sb) * 8);
+
+	mlog(0, "requested: %dM, max: %uM, default: %uM\n",
+	     requested_mb, la_max_mb, la_default_mb);
+
+	if (requested_mb == -1) {
+		/* No user request - use defaults */
+		osb->local_alloc_default_bits =
+			ocfs2_megabytes_to_clusters(sb, la_default_mb);
+	} else if (requested_mb > la_max_mb) {
+		/* Request is too big, we give the maximum available */
+		osb->local_alloc_default_bits =
+			ocfs2_megabytes_to_clusters(sb, la_max_mb);
+	} else {
+		osb->local_alloc_default_bits =
+			ocfs2_megabytes_to_clusters(sb, requested_mb);
+	}
+
+	osb->local_alloc_bits = osb->local_alloc_default_bits;
+}
+
 static inline int ocfs2_la_state_enabled(struct ocfs2_super *osb)
 {
 	return (osb->local_alloc_state == OCFS2_LA_THROTTLED ||
@@ -156,7 +295,7 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb)
 		     osb->local_alloc_bits, (osb->bitmap_cpg - 1));
 		osb->local_alloc_bits =
 			ocfs2_megabytes_to_clusters(osb->sb,
-						    OCFS2_DEFAULT_LOCAL_ALLOC_SIZE);
+						    ocfs2_la_default_mb(osb));
 	}
 
 	/* read the alloc off disk */
@@ -262,6 +401,8 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
 
 	osb->local_alloc_state = OCFS2_LA_DISABLED;
 
+	ocfs2_resmap_uninit(&osb->osb_la_resmap);
+
 	main_bm_inode = ocfs2_get_system_file_inode(osb,
 						    GLOBAL_BITMAP_SYSTEM_INODE,
 						    OCFS2_INVALID_SLOT);
@@ -305,12 +446,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
 	}
 
 	ocfs2_clear_local_alloc(alloc);
-
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	brelse(bh);
 	osb->local_alloc_bh = NULL;
@@ -481,46 +617,6 @@ out:
 	return status;
 }
 
-/* Check to see if the local alloc window is within ac->ac_max_block */
-static int ocfs2_local_alloc_in_range(struct inode *inode,
-				      struct ocfs2_alloc_context *ac,
-				      u32 bits_wanted)
-{
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	struct ocfs2_dinode *alloc;
-	struct ocfs2_local_alloc *la;
-	int start;
-	u64 block_off;
-
-	if (!ac->ac_max_block)
-		return 1;
-
-	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
-	la = OCFS2_LOCAL_ALLOC(alloc);
-
-	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted);
-	if (start == -1) {
-		mlog_errno(-ENOSPC);
-		return 0;
-	}
-
-	/*
-	 * Converting (bm_off + start + bits_wanted) to blocks gives us
-	 * the blkno just past our actual allocation.  This is perfect
-	 * to compare with ac_max_block.
-	 */
-	block_off = ocfs2_clusters_to_blocks(inode->i_sb,
-					     le32_to_cpu(la->la_bm_off) +
-					     start + bits_wanted);
-	mlog(0, "Checking %llu against %llu\n",
-	     (unsigned long long)block_off,
-	     (unsigned long long)ac->ac_max_block);
-	if (block_off > ac->ac_max_block)
-		return 0;
-
-	return 1;
-}
-
 /*
  * make sure we've got at least bits_wanted contiguous bits in the
  * local alloc. You lose them when you drop i_mutex.
@@ -613,17 +709,6 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
 		mlog(0, "Calling in_range for max block %llu\n",
 		     (unsigned long long)ac->ac_max_block);
 
-	if (!ocfs2_local_alloc_in_range(local_alloc_inode, ac,
-					bits_wanted)) {
-		/*
-		 * The window is outside ac->ac_max_block.
-		 * This errno tells the caller to keep localalloc enabled
-		 * but to get the allocation from the main bitmap.
-		 */
-		status = -EFBIG;
-		goto bail;
-	}
-
 	ac->ac_inode = local_alloc_inode;
 	/* We should never use localalloc from another slot */
 	ac->ac_alloc_slot = osb->slot_num;
@@ -664,7 +749,8 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
 	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
 	la = OCFS2_LOCAL_ALLOC(alloc);
 
-	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted);
+	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, &bits_wanted,
+						  ac->ac_resv);
 	if (start == -1) {
 		/* TODO: Shouldn't we just BUG here? */
 		status = -ENOSPC;
@@ -674,8 +760,6 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
 
 	bitmap = la->la_bitmap;
 	*bit_off = le32_to_cpu(la->la_bm_off) + start;
-	/* local alloc is always contiguous by nature -- we never
-	 * delete bits from it! */
 	*num_bits = bits_wanted;
 
 	status = ocfs2_journal_access_di(handle,
@@ -687,18 +771,15 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
 		goto bail;
 	}
 
+	ocfs2_resmap_claimed_bits(&osb->osb_la_resmap, ac->ac_resv, start,
+				  bits_wanted);
+
 	while(bits_wanted--)
 		ocfs2_set_bit(start++, bitmap);
 
 	le32_add_cpu(&alloc->id1.bitmap1.i_used, *num_bits);
+	ocfs2_journal_dirty(handle, osb->local_alloc_bh);
 
-	status = ocfs2_journal_dirty(handle, osb->local_alloc_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	status = 0;
 bail:
 	mlog_exit(status);
 	return status;
@@ -722,13 +803,17 @@ static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc)
 }
 
 static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
-					     struct ocfs2_dinode *alloc,
-					     u32 numbits)
+				     struct ocfs2_dinode *alloc,
+				     u32 *numbits,
+				     struct ocfs2_alloc_reservation *resv)
 {
 	int numfound, bitoff, left, startoff, lastzero;
+	int local_resv = 0;
+	struct ocfs2_alloc_reservation r;
 	void *bitmap = NULL;
+	struct ocfs2_reservation_map *resmap = &osb->osb_la_resmap;
 
-	mlog_entry("(numbits wanted = %u)\n", numbits);
+	mlog_entry("(numbits wanted = %u)\n", *numbits);
 
 	if (!alloc->id1.bitmap1.i_total) {
 		mlog(0, "No bits in my window!\n");
@@ -736,6 +821,30 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 		goto bail;
 	}
 
+	if (!resv) {
+		local_resv = 1;
+		ocfs2_resv_init_once(&r);
+		ocfs2_resv_set_type(&r, OCFS2_RESV_FLAG_TMP);
+		resv = &r;
+	}
+
+	numfound = *numbits;
+	if (ocfs2_resmap_resv_bits(resmap, resv, &bitoff, &numfound) == 0) {
+		if (numfound < *numbits)
+			*numbits = numfound;
+		goto bail;
+	}
+
+	/*
+	 * Code error. While reservations are enabled, local
+	 * allocation should _always_ go through them.
+	 */
+	BUG_ON(osb->osb_resv_level != 0);
+
+	/*
+	 * Reservations are disabled. Handle this the old way.
+	 */
+
 	bitmap = OCFS2_LOCAL_ALLOC(alloc)->la_bitmap;
 
 	numfound = bitoff = startoff = 0;
@@ -761,7 +870,7 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 			startoff = bitoff+1;
 		}
 		/* we got everything we needed */
-		if (numfound == numbits) {
+		if (numfound == *numbits) {
 			/* mlog(0, "Found it all!\n"); */
 			break;
 		}
@@ -770,12 +879,15 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 	mlog(0, "Exiting loop, bitoff = %d, numfound = %d\n", bitoff,
 	     numfound);
 
-	if (numfound == numbits)
+	if (numfound == *numbits)
 		bitoff = startoff - numfound;
 	else
 		bitoff = -1;
 
 bail:
+	if (local_resv)
+		ocfs2_resv_discard(resmap, resv);
+
 	mlog_exit(bitoff);
 	return bitoff;
 }
@@ -1049,7 +1161,7 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
 	/* we used the generic suballoc reserve function, but we set
 	 * everything up nicely, so there's no reason why we can't use
 	 * the more specific cluster api to claim bits. */
-	status = ocfs2_claim_clusters(osb, handle, ac, osb->local_alloc_bits,
+	status = ocfs2_claim_clusters(handle, ac, osb->local_alloc_bits,
 				      &cluster_off, &cluster_count);
 	if (status == -ENOSPC) {
 retry_enospc:
@@ -1063,7 +1175,7 @@ retry_enospc:
 			goto bail;
 
 		ac->ac_bits_wanted = osb->local_alloc_default_bits;
-		status = ocfs2_claim_clusters(osb, handle, ac,
+		status = ocfs2_claim_clusters(handle, ac,
 					      osb->local_alloc_bits,
 					      &cluster_off,
 					      &cluster_count);
@@ -1098,6 +1210,9 @@ retry_enospc:
 	memset(OCFS2_LOCAL_ALLOC(alloc)->la_bitmap, 0,
 	       le16_to_cpu(la->la_size));
 
+	ocfs2_resmap_restart(&osb->osb_la_resmap, cluster_count,
+			     OCFS2_LOCAL_ALLOC(alloc)->la_bitmap);
+
 	mlog(0, "New window allocated:\n");
 	mlog(0, "window la_bm_off = %u\n",
 	     OCFS2_LOCAL_ALLOC(alloc)->la_bm_off);
@@ -1169,12 +1284,7 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
 	}
 
 	ocfs2_clear_local_alloc(alloc);
-
-	status = ocfs2_journal_dirty(handle, osb->local_alloc_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, osb->local_alloc_bh);
 
 	status = ocfs2_sync_local_to_main(osb, handle, alloc_copy,
 					  main_bm_inode, main_bm_bh);
@@ -1192,7 +1302,6 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
 
 	atomic_inc(&osb->alloc_stats.moves);
 
-	status = 0;
 bail:
 	if (handle)
 		ocfs2_commit_trans(osb, handle);

fs/ocfs2/localalloc.h

@@ -30,6 +30,9 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb);
 
 void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb);
 
+void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb);
+unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb);
+
 int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
 				     int node_num,
 				     struct ocfs2_dinode **alloc_copy);

fs/ocfs2/mmap.c

@@ -41,44 +41,20 @@
 #include "file.h"
 #include "inode.h"
 #include "mmap.h"
+#include "super.h"
 
-static inline int ocfs2_vm_op_block_sigs(sigset_t *blocked, sigset_t *oldset)
-{
-	/* The best way to deal with signals in the vm path is
-	 * to block them upfront, rather than allowing the
-	 * locking paths to return -ERESTARTSYS. */
-	sigfillset(blocked);
-
-	/* We should technically never get a bad return value
-	 * from sigprocmask */
-	return sigprocmask(SIG_BLOCK, blocked, oldset);
-}
-
-static inline int ocfs2_vm_op_unblock_sigs(sigset_t *oldset)
-{
-	return sigprocmask(SIG_SETMASK, oldset, NULL);
-}
 
 static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf)
 {
-	sigset_t blocked, oldset;
-	int error, ret;
+	sigset_t oldset;
+	int ret;
 
 	mlog_entry("(area=%p, page offset=%lu)\n", area, vmf->pgoff);
 
-	error = ocfs2_vm_op_block_sigs(&blocked, &oldset);
-	if (error < 0) {
-		mlog_errno(error);
-		ret = VM_FAULT_SIGBUS;
-		goto out;
-	}
-
+	ocfs2_block_signals(&oldset);
 	ret = filemap_fault(area, vmf);
+	ocfs2_unblock_signals(&oldset);
 
-	error = ocfs2_vm_op_unblock_sigs(&oldset);
-	if (error < 0)
-		mlog_errno(error);
-out:
 	mlog_exit_ptr(vmf->page);
 	return ret;
 }
@@ -158,14 +134,10 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	struct page *page = vmf->page;
 	struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
 	struct buffer_head *di_bh = NULL;
-	sigset_t blocked, oldset;
-	int ret, ret2;
+	sigset_t oldset;
+	int ret;
 
-	ret = ocfs2_vm_op_block_sigs(&blocked, &oldset);
-	if (ret < 0) {
-		mlog_errno(ret);
-		return ret;
-	}
+	ocfs2_block_signals(&oldset);
 
 	/*
 	 * The cluster locks taken will block a truncate from another
@@ -193,9 +165,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	ocfs2_inode_unlock(inode, 1);
 
 out:
-	ret2 = ocfs2_vm_op_unblock_sigs(&oldset);
-	if (ret2 < 0)
-		mlog_errno(ret2);
+	ocfs2_unblock_signals(&oldset);
 	if (ret)
 		ret = VM_FAULT_SIGBUS;
 	return ret;

fs/ocfs2/namei.c

@@ -239,6 +239,8 @@ static int ocfs2_mknod(struct inode *dir,
 	};
 	int did_quota_inode = 0;
 	struct ocfs2_dir_lookup_result lookup = { NULL, };
+	sigset_t oldset;
+	int did_block_signals = 0;
 
 	mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, mode,
 		   (unsigned long)dev, dentry->d_name.len,
@@ -350,6 +352,10 @@ static int ocfs2_mknod(struct inode *dir,
 		goto leave;
 	}
 
+	/* Starting to change things, restart is no longer possible. */
+	ocfs2_block_signals(&oldset);
+	did_block_signals = 1;
+
 	status = dquot_alloc_inode(inode);
 	if (status)
 		goto leave;
@@ -384,11 +390,7 @@ static int ocfs2_mknod(struct inode *dir,
 			goto leave;
 		}
 		ocfs2_add_links_count(dirfe, 1);
-		status = ocfs2_journal_dirty(handle, parent_fe_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto leave;
-		}
+		ocfs2_journal_dirty(handle, parent_fe_bh);
 		inc_nlink(dir);
 	}
 
@@ -439,6 +441,8 @@ leave:
 		ocfs2_commit_trans(osb, handle);
 
 	ocfs2_inode_unlock(dir, 1);
+	if (did_block_signals)
+		ocfs2_unblock_signals(&oldset);
 
 	if (status == -ENOSPC)
 		mlog(0, "Disk is full\n");
@@ -487,14 +491,15 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
 	int status = 0;
 	struct ocfs2_dinode *fe = NULL;
 	struct ocfs2_extent_list *fel;
-	u64 fe_blkno = 0;
+	u64 suballoc_loc, fe_blkno = 0;
 	u16 suballoc_bit;
 	u16 feat;
 
 	*new_fe_bh = NULL;
 
-	status = ocfs2_claim_new_inode(osb, handle, dir, parent_fe_bh,
-				       inode_ac, &suballoc_bit, &fe_blkno);
+	status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
+				       inode_ac, &suballoc_loc,
+				       &suballoc_bit, &fe_blkno);
 	if (status < 0) {
 		mlog_errno(status);
 		goto leave;
@@ -531,6 +536,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
 	fe->i_generation = cpu_to_le32(inode->i_generation);
 	fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
 	fe->i_blkno = cpu_to_le64(fe_blkno);
+	fe->i_suballoc_loc = cpu_to_le64(suballoc_loc);
 	fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
 	fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot);
 	fe->i_uid = cpu_to_le32(inode->i_uid);
@@ -567,11 +573,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
 		fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb));
 	}
 
-	status = ocfs2_journal_dirty(handle, *new_fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, *new_fe_bh);
 
 	ocfs2_populate_inode(inode, fe, 1);
 	ocfs2_ci_set_new(osb, INODE_CACHE(inode));
@@ -637,6 +639,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 	struct ocfs2_dinode *fe = NULL;
 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
 	struct ocfs2_dir_lookup_result lookup = { NULL, };
+	sigset_t oldset;
 
 	mlog_entry("(inode=%lu, old='%.*s' new='%.*s')\n", inode->i_ino,
 		   old_dentry->d_name.len, old_dentry->d_name.name,
@@ -693,6 +696,9 @@ static int ocfs2_link(struct dentry *old_dentry,
 		goto out_unlock_inode;
 	}
 
+	/* Starting to change things, restart is no longer possible. */
+	ocfs2_block_signals(&oldset);
+
 	err = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
 				      OCFS2_JOURNAL_ACCESS_WRITE);
 	if (err < 0) {
@@ -705,14 +711,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 	ocfs2_set_links_count(fe, inode->i_nlink);
 	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
 	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
-
-	err = ocfs2_journal_dirty(handle, fe_bh);
-	if (err < 0) {
-		ocfs2_add_links_count(fe, -1);
-		drop_nlink(inode);
-		mlog_errno(err);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, fe_bh);
 
 	err = ocfs2_add_entry(handle, dentry, inode,
 			      OCFS2_I(inode)->ip_blkno,
@@ -736,6 +735,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 
 out_commit:
 	ocfs2_commit_trans(osb, handle);
+	ocfs2_unblock_signals(&oldset);
 out_unlock_inode:
 	ocfs2_inode_unlock(inode, 1);
 
@@ -909,12 +909,7 @@ static int ocfs2_unlink(struct inode *dir,
 		drop_nlink(inode);
 	drop_nlink(inode);
 	ocfs2_set_links_count(fe, inode->i_nlink);
-
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, fe_bh);
 
 	dir->i_ctime = dir->i_mtime = CURRENT_TIME;
 	if (S_ISDIR(inode->i_mode))
@@ -1332,12 +1327,7 @@ static int ocfs2_rename(struct inode *old_dir,
 			ocfs2_set_links_count(newfe, 0);
 		else
 			ocfs2_add_links_count(newfe, -1);
-
-		status = ocfs2_journal_dirty(handle, newfe_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
+		ocfs2_journal_dirty(handle, newfe_bh);
 	} else {
 		/* if the name was not found in new_dir, add it now */
 		status = ocfs2_add_entry(handle, new_dentry, old_inode,
@@ -1356,10 +1346,7 @@ static int ocfs2_rename(struct inode *old_dir,
 
 		old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec);
 		old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec);
-
-		status = ocfs2_journal_dirty(handle, old_inode_bh);
-		if (status < 0)
-			mlog_errno(status);
+		ocfs2_journal_dirty(handle, old_inode_bh);
 	} else
 		mlog_errno(status);
 
@@ -1431,7 +1418,7 @@ static int ocfs2_rename(struct inode *old_dir,
 							 OCFS2_JOURNAL_ACCESS_WRITE);
 			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
 			ocfs2_set_links_count(fe, old_dir->i_nlink);
-			status = ocfs2_journal_dirty(handle, old_dir_bh);
+			ocfs2_journal_dirty(handle, old_dir_bh);
 		}
 	}
 	ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir);
@@ -1563,11 +1550,7 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
 		       (bytes_left > sb->s_blocksize) ? sb->s_blocksize :
 		       bytes_left);
 
-		status = ocfs2_journal_dirty(handle, bhs[virtual]);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
+		ocfs2_journal_dirty(handle, bhs[virtual]);
 
 		virtual++;
 		p_blkno++;
@@ -1611,6 +1594,8 @@ static int ocfs2_symlink(struct inode *dir,
 	};
 	int did_quota = 0, did_quota_inode = 0;
 	struct ocfs2_dir_lookup_result lookup = { NULL, };
+	sigset_t oldset;
+	int did_block_signals = 0;
 
 	mlog_entry("(0x%p, 0x%p, symname='%s' actual='%.*s')\n", dir,
 		   dentry, symname, dentry->d_name.len, dentry->d_name.name);
@@ -1706,6 +1691,10 @@ static int ocfs2_symlink(struct inode *dir,
 		goto bail;
 	}
 
+	/* Starting to change things, restart is no longer possible. */
+	ocfs2_block_signals(&oldset);
+	did_block_signals = 1;
+
 	status = dquot_alloc_inode(inode);
 	if (status)
 		goto bail;
@@ -1814,6 +1803,8 @@ bail:
 		ocfs2_commit_trans(osb, handle);
 
 	ocfs2_inode_unlock(dir, 1);
+	if (did_block_signals)
+		ocfs2_unblock_signals(&oldset);
 
 	brelse(new_fe_bh);
 	brelse(parent_fe_bh);
@@ -1961,12 +1952,7 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb,
 	if (S_ISDIR(inode->i_mode))
 		ocfs2_add_links_count(orphan_fe, 1);
 	orphan_dir_inode->i_nlink = ocfs2_read_links_count(orphan_fe);
-
-	status = ocfs2_journal_dirty(handle, orphan_dir_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, orphan_dir_bh);
 
 	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
 				   OCFS2_ORPHAN_NAMELEN, inode,
@@ -2065,12 +2051,7 @@ int ocfs2_orphan_del(struct ocfs2_super *osb,
 	if (S_ISDIR(inode->i_mode))
 		ocfs2_add_links_count(orphan_fe, -1);
 	orphan_dir_inode->i_nlink = ocfs2_read_links_count(orphan_fe);
-
-	status = ocfs2_journal_dirty(handle, orphan_dir_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, orphan_dir_bh);
 
 leave:
 	ocfs2_free_dir_lookup_result(&lookup);

fs/ocfs2/ocfs2.h

@@ -47,6 +47,7 @@
 /* For struct ocfs2_blockcheck_stats */
 #include "blockcheck.h"
 
+#include "reservations.h"
 
 /* Caching of metadata buffers */
 
@@ -341,6 +342,9 @@ struct ocfs2_super
 	 */
 	unsigned int local_alloc_bits;
 	unsigned int local_alloc_default_bits;
+	/* osb_clusters_at_boot can become stale! Do not trust it to
+	 * be up to date. */
+	unsigned int osb_clusters_at_boot;
 
 	enum ocfs2_local_alloc_state local_alloc_state; /* protected
 							 * by osb_lock */
@@ -349,6 +353,11 @@ struct ocfs2_super
 
 	u64 la_last_gd;
 
+	struct ocfs2_reservation_map	osb_la_resmap;
+
+	unsigned int	osb_resv_level;
+	unsigned int	osb_dir_resv_level;
+
 	/* Next three fields are for local node slot recovery during
 	 * mount. */
 	int dirty;
@@ -482,6 +491,13 @@ static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
 	return 0;
 }
 
+static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
+		return 1;
+	return 0;
+}
+
 static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
 {
 	if (ocfs2_supports_indexed_dirs(osb))
@@ -763,6 +779,12 @@ static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
 	return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
 }
 
+static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb,
+						       unsigned int clusters)
+{
+	return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits);
+}
+
 static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap)
 {
 	ext2_set_bit(bit, bitmap);

fs/ocfs2/ocfs2_fs.h

@@ -100,7 +100,8 @@
 					 | OCFS2_FEATURE_INCOMPAT_XATTR \
 					 | OCFS2_FEATURE_INCOMPAT_META_ECC \
 					 | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
-					 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
+					 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \
+					 | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
 #define OCFS2_FEATURE_RO_COMPAT_SUPP	(OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
 					 | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
 					 | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
@@ -165,6 +166,9 @@
 /* Refcount tree support */
 #define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE	0x1000
 
+/* Discontigous block groups */
+#define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG	0x2000
+
 /*
  * backup superblock flag is used to indicate that this volume
  * has backup superblocks.
@@ -282,14 +286,6 @@
 /* Journal limits (in bytes) */
 #define OCFS2_MIN_JOURNAL_SIZE		(4 * 1024 * 1024)
 
-/*
- * Default local alloc size (in megabytes)
- *
- * The value chosen should be such that most allocations, including new
- * block groups, use local alloc.
- */
-#define OCFS2_DEFAULT_LOCAL_ALLOC_SIZE	8
-
 /*
  * Inline extended attribute size (in bytes)
  * The value chosen should be aligned to 16 byte boundaries.
@@ -512,7 +508,10 @@ struct ocfs2_extent_block
 					   block group */
 	__le32 h_fs_generation;		/* Must match super block */
 	__le64 h_blkno;			/* Offset on disk, in blocks */
-/*20*/	__le64 h_reserved3;
+/*20*/	__le64 h_suballoc_loc;		/* Suballocator block group this
+					   eb belongs to.  Only valid
+					   if allocated from a
+					   discontiguous block group */
 	__le64 h_next_leaf_blk;		/* Offset on disk, in blocks,
 					   of next leaf header pointing
 					   to data */
@@ -679,7 +678,11 @@ struct ocfs2_dinode {
 /*80*/	struct ocfs2_block_check i_check;	/* Error checking */
 /*88*/	__le64 i_dx_root;		/* Pointer to dir index root block */
 /*90*/	__le64 i_refcount_loc;
-	__le64 i_reserved2[4];
+	__le64 i_suballoc_loc;		/* Suballocator block group this
+					   inode belongs to.  Only valid
+					   if allocated from a
+					   discontiguous block group */
+/*A0*/	__le64 i_reserved2[3];
 /*B8*/	union {
 		__le64 i_pad1;		/* Generic way to refer to this
 					   64bit union */
@@ -814,7 +817,12 @@ struct ocfs2_dx_root_block {
 	__le32		dr_reserved2;
 	__le64		dr_free_blk;		/* Pointer to head of free
 						 * unindexed block list. */
-	__le64		dr_reserved3[15];
+	__le64		dr_suballoc_loc;	/* Suballocator block group
+						   this root belongs to.
+						   Only valid if allocated
+						   from a discontiguous
+						   block group */
+	__le64		dr_reserved3[14];
 	union {
 		struct ocfs2_extent_list dr_list; /* Keep this aligned to 128
 						   * bits for maximum space
@@ -839,6 +847,13 @@ struct ocfs2_dx_leaf {
 	struct ocfs2_dx_entry_list	dl_list;
 };
 
+/*
+ * Largest bitmap for a block (suballocator) group in bytes.  This limit
+ * does not affect cluster groups (global allocator).  Cluster group
+ * bitmaps run to the end of the block.
+ */
+#define OCFS2_MAX_BG_BITMAP_SIZE	256
+
 /*
  * On disk allocator group structure for OCFS2
  */
@@ -860,7 +875,29 @@ struct ocfs2_group_desc
 	__le64   bg_blkno;               /* Offset on disk, in blocks */
 /*30*/	struct ocfs2_block_check bg_check;	/* Error checking */
 	__le64   bg_reserved2;
-/*40*/	__u8    bg_bitmap[0];
+/*40*/	union {
+		__u8    bg_bitmap[0];
+		struct {
+			/*
+			 * Block groups may be discontiguous when
+			 * OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG is set.
+			 * The extents of a discontigous block group are
+			 * stored in bg_list.  It is a flat list.
+			 * l_tree_depth must always be zero.  A
+			 * discontiguous group is signified by a non-zero
+			 * bg_list->l_next_free_rec.  Only block groups
+			 * can be discontiguous; Cluster groups cannot.
+			 * We've never made a block group with more than
+			 * 2048 blocks (256 bytes of bg_bitmap).  This
+			 * codifies that limit so that we can fit bg_list.
+			 * bg_size of a discontiguous block group will
+			 * be 256 to match bg_bitmap_filler.
+			 */
+			__u8 bg_bitmap_filler[OCFS2_MAX_BG_BITMAP_SIZE];
+/*140*/			struct ocfs2_extent_list bg_list;
+		};
+	};
+/* Actual on-disk size is one block */
 };
 
 struct ocfs2_refcount_rec {
@@ -905,7 +942,11 @@ struct ocfs2_refcount_block {
 /*40*/	__le32 rf_generation;		/* generation number. all be the same
 					 * for the same refcount tree. */
 	__le32 rf_reserved0;
-	__le64 rf_reserved1[7];
+	__le64 rf_suballoc_loc;		/* Suballocator block group this
+					   refcount block belongs to. Only
+					   valid if allocated from a
+					   discontiguous block group */
+/*50*/	__le64 rf_reserved1[6];
 /*80*/	union {
 		struct ocfs2_refcount_list rf_records;  /* List of refcount
 							  records */
@@ -1017,7 +1058,10 @@ struct ocfs2_xattr_block {
 					real xattr or a xattr tree. */
 	__le16	xb_reserved0;
 	__le32  xb_reserved1;
-	__le64	xb_reserved2;
+	__le64	xb_suballoc_loc;	/* Suballocator block group this
+					   xattr block belongs to. Only
+					   valid if allocated from a
+					   discontiguous block group */
 /*30*/	union {
 		struct ocfs2_xattr_header xb_header; /* xattr header if this
 							block contains xattr */
@@ -1254,6 +1298,16 @@ static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
 	return size / sizeof(struct ocfs2_extent_rec);
 }
 
+static inline u16 ocfs2_extent_recs_per_gd(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
+
+	return size / sizeof(struct ocfs2_extent_rec);
+}
+
 static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb)
 {
 	int size;
@@ -1284,13 +1338,23 @@ static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
 	return size;
 }
 
-static inline int ocfs2_group_bitmap_size(struct super_block *sb)
+static inline int ocfs2_group_bitmap_size(struct super_block *sb,
+					  int suballocator,
+					  u32 feature_incompat)
 {
-	int size;
-
-	size = sb->s_blocksize -
+	int size = sb->s_blocksize -
 		offsetof(struct ocfs2_group_desc, bg_bitmap);
 
+	/*
+	 * The cluster allocator uses the entire block.  Suballocators have
+	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
+	 * code expects bg_size set to the maximum.  Thus we must keep
+	 * bg_size as-is unless discontig_bg is enabled.
+	 */
+	if (suballocator &&
+	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
+		size = OCFS2_MAX_BG_BITMAP_SIZE;
+
 	return size;
 }
 
@@ -1402,23 +1466,43 @@ static inline int ocfs2_extent_recs_per_eb(int blocksize)
 	return size / sizeof(struct ocfs2_extent_rec);
 }
 
-static inline int ocfs2_local_alloc_size(int blocksize)
+static inline int ocfs2_extent_recs_per_gd(int blocksize)
 {
 	int size;
 
 	size = blocksize -
-		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
+		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
 
-	return size;
+	return size / sizeof(struct ocfs2_extent_rec);
 }
 
-static inline int ocfs2_group_bitmap_size(int blocksize)
+static inline int ocfs2_local_alloc_size(int blocksize)
 {
 	int size;
 
 	size = blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
+
+	return size;
+}
+
+static inline int ocfs2_group_bitmap_size(int blocksize,
+					  int suballocator,
+					  uint32_t feature_incompat)
+{
+	int size = sb->s_blocksize -
 		offsetof(struct ocfs2_group_desc, bg_bitmap);
 
+	/*
+	 * The cluster allocator uses the entire block.  Suballocators have
+	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
+	 * code expects bg_size set to the maximum.  Thus we must keep
+	 * bg_size as-is unless discontig_bg is enabled.
+	 */
+	if (suballocator &&
+	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
+		size = OCFS2_MAX_BG_BITMAP_SIZE;
+
 	return size;
 }
 
@@ -1491,5 +1575,19 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
 	de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
 }
 
+static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
+{
+	if ((offsetof(struct ocfs2_group_desc, bg_bitmap) +
+	     le16_to_cpu(gd->bg_size)) !=
+	    offsetof(struct ocfs2_group_desc, bg_list))
+		return 0;
+	/*
+	 * Only valid to check l_next_free_rec if
+	 * bg_bitmap + bg_size == bg_list.
+	 */
+	if (!gd->bg_list.l_next_free_rec)
+		return 0;
+	return 1;
+}
 #endif  /* _OCFS2_FS_H */
 

fs/ocfs2/quota_global.c

@@ -261,10 +261,8 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type,
 		brelse(bh);
 		goto out;
 	}
-	err = ocfs2_journal_dirty(handle, bh);
+	ocfs2_journal_dirty(handle, bh);
 	brelse(bh);
-	if (err < 0)
-		goto out;
 out:
 	if (err) {
 		mutex_unlock(&gqinode->i_mutex);

fs/ocfs2/quota_local.c

@@ -119,12 +119,8 @@ static int ocfs2_modify_bh(struct inode *inode, struct buffer_head *bh,
 	lock_buffer(bh);
 	modify(bh, private);
 	unlock_buffer(bh);
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		ocfs2_commit_trans(OCFS2_SB(sb), handle);
-		return status;
-	}
+	ocfs2_journal_dirty(handle, bh);
+
 	status = ocfs2_commit_trans(OCFS2_SB(sb), handle);
 	if (status < 0) {
 		mlog_errno(status);
@@ -523,9 +519,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode,
 			ocfs2_clear_bit(bit, dchunk->dqc_bitmap);
 			le32_add_cpu(&dchunk->dqc_free, 1);
 			unlock_buffer(qbh);
-			status = ocfs2_journal_dirty(handle, qbh);
-			if (status < 0)
-				mlog_errno(status);
+			ocfs2_journal_dirty(handle, qbh);
 out_commit:
 			mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
 			ocfs2_commit_trans(OCFS2_SB(sb), handle);
@@ -631,9 +625,7 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb,
 		lock_buffer(bh);
 		ldinfo->dqi_flags = cpu_to_le32(flags | OLQF_CLEAN);
 		unlock_buffer(bh);
-		status = ocfs2_journal_dirty(handle, bh);
-		if (status < 0)
-			mlog_errno(status);
+		ocfs2_journal_dirty(handle, bh);
 out_trans:
 		ocfs2_commit_trans(osb, handle);
 out_bh:
@@ -1009,11 +1001,7 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk(
 	       sb->s_blocksize - sizeof(struct ocfs2_local_disk_chunk) -
 	       OCFS2_QBLK_RESERVED_SPACE);
 	unlock_buffer(bh);
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	/* Initialize new block with structures */
 	down_read(&OCFS2_I(lqinode)->ip_alloc_sem);
@@ -1040,11 +1028,7 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk(
 	lock_buffer(dbh);
 	memset(dbh->b_data, 0, sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE);
 	unlock_buffer(dbh);
-	status = ocfs2_journal_dirty(handle, dbh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, dbh);
 
 	/* Update local quotafile info */
 	oinfo->dqi_blocks += 2;
@@ -1155,11 +1139,8 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file(
 	lock_buffer(bh);
 	memset(bh->b_data, 0, sb->s_blocksize);
 	unlock_buffer(bh);
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, bh);
+
 	/* Update chunk header */
 	status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode),
 					 chunk->qc_headerbh,
@@ -1173,11 +1154,8 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file(
 	lock_buffer(chunk->qc_headerbh);
 	le32_add_cpu(&dchunk->dqc_free, ol_quota_entries_per_block(sb));
 	unlock_buffer(chunk->qc_headerbh);
-	status = ocfs2_journal_dirty(handle, chunk->qc_headerbh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, chunk->qc_headerbh);
+
 	/* Update file header */
 	oinfo->dqi_blocks++;
 	status = ocfs2_local_write_info(sb, type);
@@ -1312,12 +1290,8 @@ static int ocfs2_local_release_dquot(struct dquot *dquot)
 	ocfs2_clear_bit(offset, dchunk->dqc_bitmap);
 	le32_add_cpu(&dchunk->dqc_free, 1);
 	unlock_buffer(od->dq_chunk->qc_headerbh);
-	status = ocfs2_journal_dirty(handle, od->dq_chunk->qc_headerbh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out;
-	}
-	status = 0;
+	ocfs2_journal_dirty(handle, od->dq_chunk->qc_headerbh);
+
 out:
 	/* Clear the read bit so that next time someone uses this
 	 * dquot he reads fresh info from disk and allocates local

fs/ocfs2/refcounttree.c

@@ -570,7 +570,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
 	struct ocfs2_refcount_tree *new_tree = NULL, *tree = NULL;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 first_blkno;
+	u64 suballoc_loc, first_blkno;
 
 	BUG_ON(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL);
 
@@ -596,7 +596,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
 		goto out_commit;
 	}
 
-	ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1,
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
 				   &suballoc_bit_start, &num_got,
 				   &first_blkno);
 	if (ret) {
@@ -626,6 +626,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
 	memset(rb, 0, inode->i_sb->s_blocksize);
 	strcpy((void *)rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
 	rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
 	rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 	rb->rf_fs_generation = cpu_to_le32(osb->fs_generation);
 	rb->rf_blkno = cpu_to_le64(first_blkno);
@@ -790,7 +791,10 @@ int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh)
 	if (le32_to_cpu(rb->rf_count) == 1) {
 		blk = le64_to_cpu(rb->rf_blkno);
 		bit = le16_to_cpu(rb->rf_suballoc_bit);
-		bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+		if (rb->rf_suballoc_loc)
+			bg_blkno = le64_to_cpu(rb->rf_suballoc_loc);
+		else
+			bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 
 		alloc_inode = ocfs2_get_system_file_inode(osb,
 					EXTENT_ALLOC_SYSTEM_INODE,
@@ -1268,9 +1272,7 @@ static int ocfs2_change_refcount_rec(handle_t *handle,
 	} else if (merge)
 		ocfs2_refcount_rec_merge(rb, index);
 
-	ret = ocfs2_journal_dirty(handle, ref_leaf_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, ref_leaf_bh);
 out:
 	return ret;
 }
@@ -1284,7 +1286,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
 	int ret;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 blkno;
+	u64 suballoc_loc, blkno;
 	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
 	struct buffer_head *new_bh = NULL;
 	struct ocfs2_refcount_block *new_rb;
@@ -1298,7 +1300,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
 		goto out;
 	}
 
-	ret = ocfs2_claim_metadata(OCFS2_SB(sb), handle, meta_ac, 1,
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
 				   &suballoc_bit_start, &num_got,
 				   &blkno);
 	if (ret) {
@@ -1330,6 +1332,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
 
 	new_rb = (struct ocfs2_refcount_block *)new_bh->b_data;
 	new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
 	new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 	new_rb->rf_blkno = cpu_to_le64(blkno);
 	new_rb->rf_cpos = cpu_to_le32(0);
@@ -1524,7 +1527,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
 	int ret;
 	u16 suballoc_bit_start;
 	u32 num_got, new_cpos;
-	u64 blkno;
+	u64 suballoc_loc, blkno;
 	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
 	struct ocfs2_refcount_block *root_rb =
 			(struct ocfs2_refcount_block *)ref_root_bh->b_data;
@@ -1548,7 +1551,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
 		goto out;
 	}
 
-	ret = ocfs2_claim_metadata(OCFS2_SB(sb), handle, meta_ac, 1,
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
 				   &suballoc_bit_start, &num_got,
 				   &blkno);
 	if (ret) {
@@ -1576,6 +1579,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
 	memset(new_rb, 0, sb->s_blocksize);
 	strcpy((void *)new_rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
 	new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
 	new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 	new_rb->rf_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
 	new_rb->rf_blkno = cpu_to_le64(blkno);
@@ -1694,7 +1698,7 @@ static int ocfs2_adjust_refcount_rec(handle_t *handle,
 	 * 2 more credits, one for the leaf refcount block, one for
 	 * the extent block contains the extent rec.
 	 */
-	ret = ocfs2_extend_trans(handle, handle->h_buffer_credits + 2);
+	ret = ocfs2_extend_trans(handle, 2);
 	if (ret < 0) {
 		mlog_errno(ret);
 		goto out;
@@ -1802,11 +1806,7 @@ static int ocfs2_insert_refcount_rec(handle_t *handle,
 	if (merge)
 		ocfs2_refcount_rec_merge(rb, index);
 
-	ret = ocfs2_journal_dirty(handle, ref_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, ref_leaf_bh);
 
 	if (index == 0) {
 		ret = ocfs2_adjust_refcount_rec(handle, ci,
@@ -1977,9 +1977,7 @@ static int ocfs2_split_refcount_rec(handle_t *handle,
 			ocfs2_refcount_rec_merge(rb, index);
 	}
 
-	ret = ocfs2_journal_dirty(handle, ref_leaf_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, ref_leaf_bh);
 
 out:
 	brelse(new_bh);
@@ -2112,6 +2110,7 @@ static int ocfs2_remove_refcount_extent(handle_t *handle,
 	 */
 	ret = ocfs2_cache_block_dealloc(dealloc, EXTENT_ALLOC_SYSTEM_INODE,
 					le16_to_cpu(rb->rf_suballoc_slot),
+					le64_to_cpu(rb->rf_suballoc_loc),
 					le64_to_cpu(rb->rf_blkno),
 					le16_to_cpu(rb->rf_suballoc_bit));
 	if (ret) {
@@ -2516,20 +2515,19 @@ out:
  *
  * Normally the refcount blocks store these refcount should be
  * contiguous also, so that we can get the number easily.
- * As for meta_ac, we will at most add split 2 refcount record and
- * 2 more refcount block, so just check it in a rough way.
+ * We will at most add split 2 refcount records and 2 more
+ * refcount blocks, so just check it in a rough way.
  *
  * Caller must hold refcount tree lock.
  */
 int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
-					  struct buffer_head *di_bh,
+					  u64 refcount_loc,
 					  u64 phys_blkno,
 					  u32 clusters,
 					  int *credits,
-					  struct ocfs2_alloc_context **meta_ac)
+					  int *ref_blocks)
 {
-	int ret, ref_blocks = 0;
-	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	int ret;
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
 	struct buffer_head *ref_root_bh = NULL;
 	struct ocfs2_refcount_tree *tree;
@@ -2546,14 +2544,13 @@ int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
 	BUG_ON(!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL));
 
 	ret = ocfs2_get_refcount_tree(OCFS2_SB(inode->i_sb),
-				      le64_to_cpu(di->i_refcount_loc), &tree);
+				      refcount_loc, &tree);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	ret = ocfs2_read_refcount_block(&tree->rf_ci,
-					le64_to_cpu(di->i_refcount_loc),
+	ret = ocfs2_read_refcount_block(&tree->rf_ci, refcount_loc,
 					&ref_root_bh);
 	if (ret) {
 		mlog_errno(ret);
@@ -2564,21 +2561,14 @@ int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
 					       &tree->rf_ci,
 					       ref_root_bh,
 					       start_cpos, clusters,
-					       &ref_blocks, credits);
+					       ref_blocks, credits);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	mlog(0, "reserve new metadata %d, credits = %d\n",
-	     ref_blocks, *credits);
-
-	if (ref_blocks) {
-		ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb),
-							ref_blocks, meta_ac);
-		if (ret)
-			mlog_errno(ret);
-	}
+	mlog(0, "reserve new metadata %d blocks, credits = %d\n",
+	     *ref_blocks, *credits);
 
 out:
 	brelse(ref_root_bh);
@@ -3040,11 +3030,7 @@ static int ocfs2_duplicate_clusters_by_jbd(handle_t *handle,
 		}
 
 		memcpy(new_bh->b_data, old_bh->b_data, sb->s_blocksize);
-		ret = ocfs2_journal_dirty(handle, new_bh);
-		if (ret) {
-			mlog_errno(ret);
-			break;
-		}
+		ocfs2_journal_dirty(handle, new_bh);
 
 		brelse(new_bh);
 		brelse(old_bh);
@@ -3282,7 +3268,7 @@ static int ocfs2_make_clusters_writable(struct super_block *sb,
 		} else {
 			delete = 1;
 
-			ret = __ocfs2_claim_clusters(osb, handle,
+			ret = __ocfs2_claim_clusters(handle,
 						     context->data_ac,
 						     1, set_len,
 						     &new_bit, &new_len);

fs/ocfs2/refcounttree.h

@@ -47,11 +47,11 @@ int ocfs2_decrease_refcount(struct inode *inode,
 			    struct ocfs2_cached_dealloc_ctxt *dealloc,
 			    int delete);
 int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
-					  struct buffer_head *di_bh,
+					  u64 refcount_loc,
 					  u64 phys_blkno,
 					  u32 clusters,
 					  int *credits,
-					  struct ocfs2_alloc_context **meta_ac);
+					  int *ref_blocks);
 int ocfs2_refcount_cow(struct inode *inode, struct buffer_head *di_bh,
 		       u32 cpos, u32 write_len, u32 max_cpos);
 

fs/ocfs2/reservations.c

@@ -0,0 +1,847 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * reservations.c
+ *
+ * Allocation reservations implementation
+ *
+ * Some code borrowed from fs/ext3/balloc.c and is:
+ *
+ * Copyright (C) 1992, 1993, 1994, 1995
+ * Remy Card (card@masi.ibp.fr)
+ * Laboratoire MASI - Institut Blaise Pascal
+ * Universite Pierre et Marie Curie (Paris VI)
+ *
+ * The rest is copyright (C) 2010 Novell.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/bitops.h>
+#include <linux/list.h>
+
+#define MLOG_MASK_PREFIX ML_RESERVATIONS
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#ifdef CONFIG_OCFS2_DEBUG_FS
+#define OCFS2_CHECK_RESERVATIONS
+#endif
+
+DEFINE_SPINLOCK(resv_lock);
+
+#define	OCFS2_MIN_RESV_WINDOW_BITS	8
+#define	OCFS2_MAX_RESV_WINDOW_BITS	1024
+
+int ocfs2_dir_resv_allowed(struct ocfs2_super *osb)
+{
+	return (osb->osb_resv_level && osb->osb_dir_resv_level);
+}
+
+static unsigned int ocfs2_resv_window_bits(struct ocfs2_reservation_map *resmap,
+					   struct ocfs2_alloc_reservation *resv)
+{
+	struct ocfs2_super *osb = resmap->m_osb;
+	unsigned int bits;
+
+	if (!(resv->r_flags & OCFS2_RESV_FLAG_DIR)) {
+		/* 8, 16, 32, 64, 128, 256, 512, 1024 */
+		bits = 4 << osb->osb_resv_level;
+	} else {
+		bits = 4 << osb->osb_dir_resv_level;
+	}
+	return bits;
+}
+
+static inline unsigned int ocfs2_resv_end(struct ocfs2_alloc_reservation *resv)
+{
+	if (resv->r_len)
+		return resv->r_start + resv->r_len - 1;
+	return resv->r_start;
+}
+
+static inline int ocfs2_resv_empty(struct ocfs2_alloc_reservation *resv)
+{
+	return !!(resv->r_len == 0);
+}
+
+static inline int ocfs2_resmap_disabled(struct ocfs2_reservation_map *resmap)
+{
+	if (resmap->m_osb->osb_resv_level == 0)
+		return 1;
+	return 0;
+}
+
+static void ocfs2_dump_resv(struct ocfs2_reservation_map *resmap)
+{
+	struct ocfs2_super *osb = resmap->m_osb;
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+	int i = 0;
+
+	mlog(ML_NOTICE, "Dumping resmap for device %s. Bitmap length: %u\n",
+	     osb->dev_str, resmap->m_bitmap_len);
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		mlog(ML_NOTICE, "start: %u\tend: %u\tlen: %u\tlast_start: %u"
+		     "\tlast_len: %u\n", resv->r_start,
+		     ocfs2_resv_end(resv), resv->r_len, resv->r_last_start,
+		     resv->r_last_len);
+
+		node = rb_next(node);
+		i++;
+	}
+
+	mlog(ML_NOTICE, "%d reservations found. LRU follows\n", i);
+
+	i = 0;
+	list_for_each_entry(resv, &resmap->m_lru, r_lru) {
+		mlog(ML_NOTICE, "LRU(%d) start: %u\tend: %u\tlen: %u\t"
+		     "last_start: %u\tlast_len: %u\n", i, resv->r_start,
+		     ocfs2_resv_end(resv), resv->r_len, resv->r_last_start,
+		     resv->r_last_len);
+
+		i++;
+	}
+}
+
+#ifdef OCFS2_CHECK_RESERVATIONS
+static int ocfs2_validate_resmap_bits(struct ocfs2_reservation_map *resmap,
+				      int i,
+				      struct ocfs2_alloc_reservation *resv)
+{
+	char *disk_bitmap = resmap->m_disk_bitmap;
+	unsigned int start = resv->r_start;
+	unsigned int end = ocfs2_resv_end(resv);
+
+	while (start <= end) {
+		if (ocfs2_test_bit(start, disk_bitmap)) {
+			mlog(ML_ERROR,
+			     "reservation %d covers an allocated area "
+			     "starting at bit %u!\n", i, start);
+			return 1;
+		}
+
+		start++;
+	}
+	return 0;
+}
+
+static void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap)
+{
+	unsigned int off = 0;
+	int i = 0;
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		if (i > 0 && resv->r_start <= off) {
+			mlog(ML_ERROR, "reservation %d has bad start off!\n",
+			     i);
+			goto bad;
+		}
+
+		if (resv->r_len == 0) {
+			mlog(ML_ERROR, "reservation %d has no length!\n",
+			     i);
+			goto bad;
+		}
+
+		if (resv->r_start > ocfs2_resv_end(resv)) {
+			mlog(ML_ERROR, "reservation %d has invalid range!\n",
+			     i);
+			goto bad;
+		}
+
+		if (ocfs2_resv_end(resv) >= resmap->m_bitmap_len) {
+			mlog(ML_ERROR, "reservation %d extends past bitmap!\n",
+			     i);
+			goto bad;
+		}
+
+		if (ocfs2_validate_resmap_bits(resmap, i, resv))
+			goto bad;
+
+		off = ocfs2_resv_end(resv);
+		node = rb_next(node);
+
+		i++;
+	}
+	return;
+
+bad:
+	ocfs2_dump_resv(resmap);
+	BUG();
+}
+#else
+static inline void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap)
+{
+
+}
+#endif
+
+void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv)
+{
+	memset(resv, 0, sizeof(*resv));
+	INIT_LIST_HEAD(&resv->r_lru);
+}
+
+void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv,
+			 unsigned int flags)
+{
+	BUG_ON(flags & ~OCFS2_RESV_TYPES);
+
+	resv->r_flags |= flags;
+}
+
+int ocfs2_resmap_init(struct ocfs2_super *osb,
+		      struct ocfs2_reservation_map *resmap)
+{
+	memset(resmap, 0, sizeof(*resmap));
+
+	resmap->m_osb = osb;
+	resmap->m_reservations = RB_ROOT;
+	/* m_bitmap_len is initialized to zero by the above memset. */
+	INIT_LIST_HEAD(&resmap->m_lru);
+
+	return 0;
+}
+
+static void ocfs2_resv_mark_lru(struct ocfs2_reservation_map *resmap,
+				struct ocfs2_alloc_reservation *resv)
+{
+	assert_spin_locked(&resv_lock);
+
+	if (!list_empty(&resv->r_lru))
+		list_del_init(&resv->r_lru);
+
+	list_add_tail(&resv->r_lru, &resmap->m_lru);
+}
+
+static void __ocfs2_resv_trunc(struct ocfs2_alloc_reservation *resv)
+{
+	resv->r_len = 0;
+	resv->r_start = 0;
+}
+
+static void ocfs2_resv_remove(struct ocfs2_reservation_map *resmap,
+			      struct ocfs2_alloc_reservation *resv)
+{
+	if (resv->r_flags & OCFS2_RESV_FLAG_INUSE) {
+		list_del_init(&resv->r_lru);
+		rb_erase(&resv->r_node, &resmap->m_reservations);
+		resv->r_flags &= ~OCFS2_RESV_FLAG_INUSE;
+	}
+}
+
+static void __ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+				 struct ocfs2_alloc_reservation *resv)
+{
+	assert_spin_locked(&resv_lock);
+
+	__ocfs2_resv_trunc(resv);
+	/*
+	 * last_len and last_start no longer make sense if
+	 * we're changing the range of our allocations.
+	 */
+	resv->r_last_len = resv->r_last_start = 0;
+
+	ocfs2_resv_remove(resmap, resv);
+}
+
+/* does nothing if 'resv' is null */
+void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+			struct ocfs2_alloc_reservation *resv)
+{
+	if (resv) {
+		spin_lock(&resv_lock);
+		__ocfs2_resv_discard(resmap, resv);
+		spin_unlock(&resv_lock);
+	}
+}
+
+static void ocfs2_resmap_clear_all_resv(struct ocfs2_reservation_map *resmap)
+{
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+
+	assert_spin_locked(&resv_lock);
+
+	while ((node = rb_last(&resmap->m_reservations)) != NULL) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		__ocfs2_resv_discard(resmap, resv);
+	}
+}
+
+void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap,
+			  unsigned int clen, char *disk_bitmap)
+{
+	if (ocfs2_resmap_disabled(resmap))
+		return;
+
+	spin_lock(&resv_lock);
+
+	ocfs2_resmap_clear_all_resv(resmap);
+	resmap->m_bitmap_len = clen;
+	resmap->m_disk_bitmap = disk_bitmap;
+
+	spin_unlock(&resv_lock);
+}
+
+void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap)
+{
+	/* Does nothing for now. Keep this around for API symmetry */
+}
+
+static void ocfs2_resv_insert(struct ocfs2_reservation_map *resmap,
+			      struct ocfs2_alloc_reservation *new)
+{
+	struct rb_root *root = &resmap->m_reservations;
+	struct rb_node *parent = NULL;
+	struct rb_node **p = &root->rb_node;
+	struct ocfs2_alloc_reservation *tmp;
+
+	assert_spin_locked(&resv_lock);
+
+	mlog(0, "Insert reservation start: %u len: %u\n", new->r_start,
+	     new->r_len);
+
+	while (*p) {
+		parent = *p;
+
+		tmp = rb_entry(parent, struct ocfs2_alloc_reservation, r_node);
+
+		if (new->r_start < tmp->r_start) {
+			p = &(*p)->rb_left;
+
+			/*
+			 * This is a good place to check for
+			 * overlapping reservations.
+			 */
+			BUG_ON(ocfs2_resv_end(new) >= tmp->r_start);
+		} else if (new->r_start > ocfs2_resv_end(tmp)) {
+			p = &(*p)->rb_right;
+		} else {
+			/* This should never happen! */
+			mlog(ML_ERROR, "Duplicate reservation window!\n");
+			BUG();
+		}
+	}
+
+	rb_link_node(&new->r_node, parent, p);
+	rb_insert_color(&new->r_node, root);
+	new->r_flags |= OCFS2_RESV_FLAG_INUSE;
+
+	ocfs2_resv_mark_lru(resmap, new);
+
+	ocfs2_check_resmap(resmap);
+}
+
+/**
+ * ocfs2_find_resv_lhs() - find the window which contains goal
+ * @resmap: reservation map to search
+ * @goal: which bit to search for
+ *
+ * If a window containing that goal is not found, we return the window
+ * which comes before goal. Returns NULL on empty rbtree or no window
+ * before goal.
+ */
+static struct ocfs2_alloc_reservation *
+ocfs2_find_resv_lhs(struct ocfs2_reservation_map *resmap, unsigned int goal)
+{
+	struct ocfs2_alloc_reservation *resv = NULL;
+	struct ocfs2_alloc_reservation *prev_resv = NULL;
+	struct rb_node *node = resmap->m_reservations.rb_node;
+
+	assert_spin_locked(&resv_lock);
+
+	if (!node)
+		return NULL;
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		if (resv->r_start <= goal && ocfs2_resv_end(resv) >= goal)
+			break;
+
+		/* Check if we overshot the reservation just before goal? */
+		if (resv->r_start > goal) {
+			resv = prev_resv;
+			break;
+		}
+
+		prev_resv = resv;
+		node = rb_next(node);
+	}
+
+	return resv;
+}
+
+/*
+ * We are given a range within the bitmap, which corresponds to a gap
+ * inside the reservations tree (search_start, search_len). The range
+ * can be anything from the whole bitmap, to a gap between
+ * reservations.
+ *
+ * The start value of *rstart is insignificant.
+ *
+ * This function searches the bitmap range starting at search_start
+ * with length search_len for a set of contiguous free bits. We try
+ * to find up to 'wanted' bits, but can sometimes return less.
+ *
+ * Returns the length of allocation, 0 if no free bits are found.
+ *
+ * *cstart and *clen will also be populated with the result.
+ */
+static int ocfs2_resmap_find_free_bits(struct ocfs2_reservation_map *resmap,
+				       unsigned int wanted,
+				       unsigned int search_start,
+				       unsigned int search_len,
+				       unsigned int *rstart,
+				       unsigned int *rlen)
+{
+	void *bitmap = resmap->m_disk_bitmap;
+	unsigned int best_start, best_len = 0;
+	int offset, start, found;
+
+	mlog(0, "Find %u bits within range (%u, len %u) resmap len: %u\n",
+	     wanted, search_start, search_len, resmap->m_bitmap_len);
+
+	found = best_start = best_len = 0;
+
+	start = search_start;
+	while ((offset = ocfs2_find_next_zero_bit(bitmap, resmap->m_bitmap_len,
+						 start)) != -1) {
+		/* Search reached end of the region */
+		if (offset >= (search_start + search_len))
+			break;
+
+		if (offset == start) {
+			/* we found a zero */
+			found++;
+			/* move start to the next bit to test */
+			start++;
+		} else {
+			/* got a zero after some ones */
+			found = 1;
+			start = offset + 1;
+		}
+		if (found > best_len) {
+			best_len = found;
+			best_start = start - found;
+		}
+
+		if (found >= wanted)
+			break;
+	}
+
+	if (best_len == 0)
+		return 0;
+
+	if (best_len >= wanted)
+		best_len = wanted;
+
+	*rlen = best_len;
+	*rstart = best_start;
+
+	mlog(0, "Found start: %u len: %u\n", best_start, best_len);
+
+	return *rlen;
+}
+
+static void __ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap,
+				     struct ocfs2_alloc_reservation *resv,
+				     unsigned int goal, unsigned int wanted)
+{
+	struct rb_root *root = &resmap->m_reservations;
+	unsigned int gap_start, gap_end, gap_len;
+	struct ocfs2_alloc_reservation *prev_resv, *next_resv;
+	struct rb_node *prev, *next;
+	unsigned int cstart, clen;
+	unsigned int best_start = 0, best_len = 0;
+
+	/*
+	 * Nasty cases to consider:
+	 *
+	 * - rbtree is empty
+	 * - our window should be first in all reservations
+	 * - our window should be last in all reservations
+	 * - need to make sure we don't go past end of bitmap
+	 */
+
+	mlog(0, "resv start: %u resv end: %u goal: %u wanted: %u\n",
+	     resv->r_start, ocfs2_resv_end(resv), goal, wanted);
+
+	assert_spin_locked(&resv_lock);
+
+	if (RB_EMPTY_ROOT(root)) {
+		/*
+		 * Easiest case - empty tree. We can just take
+		 * whatever window of free bits we want.
+		 */
+
+		mlog(0, "Empty root\n");
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal,
+						   resmap->m_bitmap_len - goal,
+						   &cstart, &clen);
+
+		/*
+		 * This should never happen - the local alloc window
+		 * will always have free bits when we're called.
+		 */
+		BUG_ON(goal == 0 && clen == 0);
+
+		if (clen == 0)
+			return;
+
+		resv->r_start = cstart;
+		resv->r_len = clen;
+
+		ocfs2_resv_insert(resmap, resv);
+		return;
+	}
+
+	prev_resv = ocfs2_find_resv_lhs(resmap, goal);
+
+	if (prev_resv == NULL) {
+		mlog(0, "Goal on LHS of leftmost window\n");
+
+		/*
+		 * A NULL here means that the search code couldn't
+		 * find a window that starts before goal.
+		 *
+		 * However, we can take the first window after goal,
+		 * which is also by definition, the leftmost window in
+		 * the entire tree. If we can find free bits in the
+		 * gap between goal and the LHS window, then the
+		 * reservation can safely be placed there.
+		 *
+		 * Otherwise we fall back to a linear search, checking
+		 * the gaps in between windows for a place to
+		 * allocate.
+		 */
+
+		next = rb_first(root);
+		next_resv = rb_entry(next, struct ocfs2_alloc_reservation,
+				     r_node);
+
+		/*
+		 * The search should never return such a window. (see
+		 * comment above
+		 */
+		if (next_resv->r_start <= goal) {
+			mlog(ML_ERROR, "goal: %u next_resv: start %u len %u\n",
+			     goal, next_resv->r_start, next_resv->r_len);
+			ocfs2_dump_resv(resmap);
+			BUG();
+		}
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal,
+						   next_resv->r_start - goal,
+						   &cstart, &clen);
+		if (clen) {
+			best_len = clen;
+			best_start = cstart;
+			if (best_len == wanted)
+				goto out_insert;
+		}
+
+		prev_resv = next_resv;
+		next_resv = NULL;
+	}
+
+	prev = &prev_resv->r_node;
+
+	/* Now we do a linear search for a window, starting at 'prev_rsv' */
+	while (1) {
+		next = rb_next(prev);
+		if (next) {
+			mlog(0, "One more resv found in linear search\n");
+			next_resv = rb_entry(next,
+					     struct ocfs2_alloc_reservation,
+					     r_node);
+
+			gap_start = ocfs2_resv_end(prev_resv) + 1;
+			gap_end = next_resv->r_start - 1;
+			gap_len = gap_end - gap_start + 1;
+		} else {
+			mlog(0, "No next node\n");
+			/*
+			 * We're at the rightmost edge of the
+			 * tree. See if a reservation between this
+			 * window and the end of the bitmap will work.
+			 */
+			gap_start = ocfs2_resv_end(prev_resv) + 1;
+			gap_len = resmap->m_bitmap_len - gap_start;
+			gap_end = resmap->m_bitmap_len - 1;
+		}
+
+		/*
+		 * No need to check this gap if we have already found
+		 * a larger region of free bits.
+		 */
+		if (gap_len <= best_len)
+			goto next_resv;
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, gap_start,
+						   gap_len, &cstart, &clen);
+		if (clen == wanted) {
+			best_len = clen;
+			best_start = cstart;
+			goto out_insert;
+		} else if (clen > best_len) {
+			best_len = clen;
+			best_start = cstart;
+		}
+
+next_resv:
+		if (!next)
+			break;
+
+		prev = next;
+		prev_resv = rb_entry(prev, struct ocfs2_alloc_reservation,
+				     r_node);
+	}
+
+out_insert:
+	if (best_len) {
+		resv->r_start = best_start;
+		resv->r_len = best_len;
+		ocfs2_resv_insert(resmap, resv);
+	}
+}
+
+static void ocfs2_cannibalize_resv(struct ocfs2_reservation_map *resmap,
+				   struct ocfs2_alloc_reservation *resv,
+				   unsigned int wanted)
+{
+	struct ocfs2_alloc_reservation *lru_resv;
+	int tmpwindow = !!(resv->r_flags & OCFS2_RESV_FLAG_TMP);
+	unsigned int min_bits;
+
+	if (!tmpwindow)
+		min_bits = ocfs2_resv_window_bits(resmap, resv) >> 1;
+	else
+		min_bits = wanted; /* We at know the temp window will use all
+				    * of these bits */
+
+	/*
+	 * Take the first reservation off the LRU as our 'target'. We
+	 * don't try to be smart about it. There might be a case for
+	 * searching based on size but I don't have enough data to be
+	 * sure. --Mark (3/16/2010)
+	 */
+	lru_resv = list_first_entry(&resmap->m_lru,
+				    struct ocfs2_alloc_reservation, r_lru);
+
+	mlog(0, "lru resv: start: %u len: %u end: %u\n", lru_resv->r_start,
+	     lru_resv->r_len, ocfs2_resv_end(lru_resv));
+
+	/*
+	 * Cannibalize (some or all) of the target reservation and
+	 * feed it to the current window.
+	 */
+	if (lru_resv->r_len <= min_bits) {
+		/*
+		 * Discard completely if size is less than or equal to a
+		 * reasonable threshold - 50% of window bits for non temporary
+		 * windows.
+		 */
+		resv->r_start = lru_resv->r_start;
+		resv->r_len = lru_resv->r_len;
+
+		__ocfs2_resv_discard(resmap, lru_resv);
+	} else {
+		unsigned int shrink;
+		if (tmpwindow)
+			shrink = min_bits;
+		else
+			shrink = lru_resv->r_len / 2;
+
+		lru_resv->r_len -= shrink;
+
+		resv->r_start = ocfs2_resv_end(lru_resv) + 1;
+		resv->r_len = shrink;
+	}
+
+	mlog(0, "Reservation now looks like: r_start: %u r_end: %u "
+	     "r_len: %u r_last_start: %u r_last_len: %u\n",
+	     resv->r_start, ocfs2_resv_end(resv), resv->r_len,
+	     resv->r_last_start, resv->r_last_len);
+
+	ocfs2_resv_insert(resmap, resv);
+}
+
+static void ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap,
+				   struct ocfs2_alloc_reservation *resv,
+				   unsigned int wanted)
+{
+	unsigned int goal = 0;
+
+	BUG_ON(!ocfs2_resv_empty(resv));
+
+	/*
+	 * Begin by trying to get a window as close to the previous
+	 * one as possible. Using the most recent allocation as a
+	 * start goal makes sense.
+	 */
+	if (resv->r_last_len) {
+		goal = resv->r_last_start + resv->r_last_len;
+		if (goal >= resmap->m_bitmap_len)
+			goal = 0;
+	}
+
+	__ocfs2_resv_find_window(resmap, resv, goal, wanted);
+
+	/* Search from last alloc didn't work, try once more from beginning. */
+	if (ocfs2_resv_empty(resv) && goal != 0)
+		__ocfs2_resv_find_window(resmap, resv, 0, wanted);
+
+	if (ocfs2_resv_empty(resv)) {
+		/*
+		 * Still empty? Pull oldest one off the LRU, remove it from
+		 * tree, put this one in it's place.
+		 */
+		ocfs2_cannibalize_resv(resmap, resv, wanted);
+	}
+
+	BUG_ON(ocfs2_resv_empty(resv));
+}
+
+int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap,
+			   struct ocfs2_alloc_reservation *resv,
+			   int *cstart, int *clen)
+{
+	unsigned int wanted = *clen;
+
+	if (resv == NULL || ocfs2_resmap_disabled(resmap))
+		return -ENOSPC;
+
+	spin_lock(&resv_lock);
+
+	/*
+	 * We don't want to over-allocate for temporary
+	 * windows. Otherwise, we run the risk of fragmenting the
+	 * allocation space.
+	 */
+	wanted = ocfs2_resv_window_bits(resmap, resv);
+	if ((resv->r_flags & OCFS2_RESV_FLAG_TMP) || wanted < *clen)
+		wanted = *clen;
+
+	if (ocfs2_resv_empty(resv)) {
+		mlog(0, "empty reservation, find new window\n");
+
+		/*
+		 * Try to get a window here. If it works, we must fall
+		 * through and test the bitmap . This avoids some
+		 * ping-ponging of windows due to non-reserved space
+		 * being allocation before we initialize a window for
+		 * that inode.
+		 */
+		ocfs2_resv_find_window(resmap, resv, wanted);
+	}
+
+	BUG_ON(ocfs2_resv_empty(resv));
+
+	*cstart = resv->r_start;
+	*clen = resv->r_len;
+
+	spin_unlock(&resv_lock);
+	return 0;
+}
+
+static void
+	ocfs2_adjust_resv_from_alloc(struct ocfs2_reservation_map *resmap,
+				     struct ocfs2_alloc_reservation *resv,
+				     unsigned int start, unsigned int end)
+{
+	unsigned int rhs = 0;
+	unsigned int old_end = ocfs2_resv_end(resv);
+
+	BUG_ON(start != resv->r_start || old_end < end);
+
+	/*
+	 * Completely used? We can remove it then.
+	 */
+	if (old_end == end) {
+		__ocfs2_resv_discard(resmap, resv);
+		return;
+	}
+
+	rhs = old_end - end;
+
+	/*
+	 * This should have been trapped above.
+	 */
+	BUG_ON(rhs == 0);
+
+	resv->r_start = end + 1;
+	resv->r_len = old_end - resv->r_start + 1;
+}
+
+void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap,
+			       struct ocfs2_alloc_reservation *resv,
+			       u32 cstart, u32 clen)
+{
+	unsigned int cend = cstart + clen - 1;
+
+	if (resmap == NULL || ocfs2_resmap_disabled(resmap))
+		return;
+
+	if (resv == NULL)
+		return;
+
+	BUG_ON(cstart != resv->r_start);
+
+	spin_lock(&resv_lock);
+
+	mlog(0, "claim bits: cstart: %u cend: %u clen: %u r_start: %u "
+	     "r_end: %u r_len: %u, r_last_start: %u r_last_len: %u\n",
+	     cstart, cend, clen, resv->r_start, ocfs2_resv_end(resv),
+	     resv->r_len, resv->r_last_start, resv->r_last_len);
+
+	BUG_ON(cstart < resv->r_start);
+	BUG_ON(cstart > ocfs2_resv_end(resv));
+	BUG_ON(cend > ocfs2_resv_end(resv));
+
+	ocfs2_adjust_resv_from_alloc(resmap, resv, cstart, cend);
+	resv->r_last_start = cstart;
+	resv->r_last_len = clen;
+
+	/*
+	 * May have been discarded above from
+	 * ocfs2_adjust_resv_from_alloc().
+	 */
+	if (!ocfs2_resv_empty(resv))
+		ocfs2_resv_mark_lru(resmap, resv);
+
+	mlog(0, "Reservation now looks like: r_start: %u r_end: %u "
+	     "r_len: %u r_last_start: %u r_last_len: %u\n",
+	     resv->r_start, ocfs2_resv_end(resv), resv->r_len,
+	     resv->r_last_start, resv->r_last_len);
+
+	ocfs2_check_resmap(resmap);
+
+	spin_unlock(&resv_lock);
+}

fs/ocfs2/reservations.h

@@ -0,0 +1,159 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * reservations.h
+ *
+ * Allocation reservations function prototypes and structures.
+ *
+ * Copyright (C) 2010 Novell.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef	OCFS2_RESERVATIONS_H
+#define	OCFS2_RESERVATIONS_H
+
+#include <linux/rbtree.h>
+
+#define OCFS2_DEFAULT_RESV_LEVEL	2
+#define OCFS2_MAX_RESV_LEVEL	9
+#define OCFS2_MIN_RESV_LEVEL	0
+
+struct ocfs2_alloc_reservation {
+	struct rb_node	r_node;
+
+	unsigned int	r_start;	/* Begining of current window */
+	unsigned int	r_len;		/* Length of the window */
+
+	unsigned int	r_last_len;	/* Length of most recent alloc */
+	unsigned int	r_last_start;	/* Start of most recent alloc */
+	struct list_head	r_lru;	/* LRU list head */
+
+	unsigned int	r_flags;
+};
+
+#define	OCFS2_RESV_FLAG_INUSE	0x01	/* Set when r_node is part of a btree */
+#define	OCFS2_RESV_FLAG_TMP	0x02	/* Temporary reservation, will be
+					 * destroyed immedately after use */
+#define	OCFS2_RESV_FLAG_DIR	0x04	/* Reservation is for an unindexed
+					 * directory btree */
+
+struct ocfs2_reservation_map {
+	struct rb_root		m_reservations;
+	char			*m_disk_bitmap;
+
+	struct ocfs2_super	*m_osb;
+
+	/* The following are not initialized to meaningful values until a disk
+	 * bitmap is provided. */
+	u32			m_bitmap_len;	/* Number of valid
+						 * bits available */
+
+	struct list_head	m_lru;		/* LRU of reservations
+						 * structures. */
+
+};
+
+void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv);
+
+#define OCFS2_RESV_TYPES	(OCFS2_RESV_FLAG_TMP|OCFS2_RESV_FLAG_DIR)
+void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv,
+			 unsigned int flags);
+
+int ocfs2_dir_resv_allowed(struct ocfs2_super *osb);
+
+/**
+ * ocfs2_resv_discard() - truncate a reservation
+ * @resmap:
+ * @resv: the reservation to truncate.
+ *
+ * After this function is called, the reservation will be empty, and
+ * unlinked from the rbtree.
+ */
+void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+			struct ocfs2_alloc_reservation *resv);
+
+
+/**
+ * ocfs2_resmap_init() - Initialize fields of a reservations bitmap
+ * @resmap: struct ocfs2_reservation_map to initialize
+ * @obj: unused for now
+ * @ops: unused for now
+ * @max_bitmap_bytes: Maximum size of the bitmap (typically blocksize)
+ *
+ * Only possible return value other than '0' is -ENOMEM for failure to
+ * allocation mirror bitmap.
+ */
+int ocfs2_resmap_init(struct ocfs2_super *osb,
+		      struct ocfs2_reservation_map *resmap);
+
+/**
+ * ocfs2_resmap_restart() - "restart" a reservation bitmap
+ * @resmap: reservations bitmap
+ * @clen: Number of valid bits in the bitmap
+ * @disk_bitmap: the disk bitmap this resmap should refer to.
+ *
+ * Re-initialize the parameters of a reservation bitmap. This is
+ * useful for local alloc window slides.
+ *
+ * This function will call ocfs2_trunc_resv against all existing
+ * reservations. A future version will recalculate existing
+ * reservations based on the new bitmap.
+ */
+void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap,
+			  unsigned int clen, char *disk_bitmap);
+
+/**
+ * ocfs2_resmap_uninit() - uninitialize a reservation bitmap structure
+ * @resmap: the struct ocfs2_reservation_map to uninitialize
+ */
+void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap);
+
+/**
+ * ocfs2_resmap_resv_bits() - Return still-valid reservation bits
+ * @resmap: reservations bitmap
+ * @resv: reservation to base search from
+ * @cstart: start of proposed allocation
+ * @clen: length (in clusters) of proposed allocation
+ *
+ * Using the reservation data from resv, this function will compare
+ * resmap and resmap->m_disk_bitmap to determine what part (if any) of
+ * the reservation window is still clear to use. If resv is empty,
+ * this function will try to allocate a window for it.
+ *
+ * On success, zero is returned and the valid allocation area is set in cstart
+ * and clen.
+ *
+ * Returns -ENOSPC if reservations are disabled.
+ */
+int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap,
+			   struct ocfs2_alloc_reservation *resv,
+			   int *cstart, int *clen);
+
+/**
+ * ocfs2_resmap_claimed_bits() - Tell the reservation code that bits were used.
+ * @resmap: reservations bitmap
+ * @resv: optional reservation to recalulate based on new bitmap
+ * @cstart: start of allocation in clusters
+ * @clen: end of allocation in clusters.
+ *
+ * Tell the reservation code that bits were used to fulfill allocation in
+ * resmap. The bits don't have to have been part of any existing
+ * reservation. But we must always call this function when bits are claimed.
+ * Internally, the reservations code will use this information to mark the
+ * reservations bitmap. If resv is passed, it's next allocation window will be
+ * calculated. It also expects that 'cstart' is the same as we passed back
+ * from ocfs2_resmap_resv_bits().
+ */
+void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap,
+			       struct ocfs2_alloc_reservation *resv,
+			       u32 cstart, u32 clen);
+
+#endif	/* OCFS2_RESERVATIONS_H */

fs/ocfs2/resize.c

@@ -134,11 +134,7 @@ static int ocfs2_update_last_group_and_inode(handle_t *handle,
 		le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
 	}
 
-	ret = ocfs2_journal_dirty(handle, group_bh);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto out_rollback;
-	}
+	ocfs2_journal_dirty(handle, group_bh);
 
 	/* update the inode accordingly. */
 	ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh,
@@ -319,7 +315,8 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters)
 	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
 
 	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
-				 ocfs2_group_bitmap_size(osb->sb) * 8) {
+		ocfs2_group_bitmap_size(osb->sb, 0,
+					osb->s_feature_incompat) * 8) {
 		mlog(ML_ERROR, "The disk is too old and small. "
 		     "Force to do offline resize.");
 		ret = -EINVAL;
@@ -500,7 +497,8 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
 	fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
 
 	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
-				 ocfs2_group_bitmap_size(osb->sb) * 8) {
+		ocfs2_group_bitmap_size(osb->sb, 0,
+					osb->s_feature_incompat) * 8) {
 		mlog(ML_ERROR, "The disk is too old and small."
 		     " Force to do offline resize.");
 		ret = -EINVAL;
@@ -545,12 +543,7 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
 
 	group = (struct ocfs2_group_desc *)group_bh->b_data;
 	group->bg_next_group = cr->c_blkno;
-
-	ret = ocfs2_journal_dirty(handle, group_bh);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, group_bh);
 
 	ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode),
 				      main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE);

fs/ocfs2/suballoc.c

@@ -53,6 +53,15 @@
 
 #define OCFS2_MAX_TO_STEAL		1024
 
+struct ocfs2_suballoc_result {
+	u64		sr_bg_blkno;	/* The bg we allocated from.  Set
+					   to 0 when a block group is
+					   contiguous. */
+	u64		sr_blkno;	/* The first allocated block */
+	unsigned int	sr_bit_offset;	/* The bit in the bg */
+	unsigned int	sr_bits;	/* How many bits we claimed */
+};
+
 static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg);
 static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe);
 static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl);
@@ -60,6 +69,7 @@ static int ocfs2_block_group_fill(handle_t *handle,
 				  struct inode *alloc_inode,
 				  struct buffer_head *bg_bh,
 				  u64 group_blkno,
+				  unsigned int group_clusters,
 				  u16 my_chain,
 				  struct ocfs2_chain_list *cl);
 static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
@@ -73,20 +83,17 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 				      struct buffer_head *group_bh,
 				      u32 bits_wanted, u32 min_bits,
 				      u64 max_block,
-				      u16 *bit_off, u16 *bits_found);
+				      struct ocfs2_suballoc_result *res);
 static int ocfs2_block_group_search(struct inode *inode,
 				    struct buffer_head *group_bh,
 				    u32 bits_wanted, u32 min_bits,
 				    u64 max_block,
-				    u16 *bit_off, u16 *bits_found);
-static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
-				     struct ocfs2_alloc_context *ac,
+				    struct ocfs2_suballoc_result *res);
+static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
 				     handle_t *handle,
 				     u32 bits_wanted,
 				     u32 min_bits,
-				     u16 *bit_off,
-				     unsigned int *num_bits,
-				     u64 *bg_blkno);
+				     struct ocfs2_suballoc_result *res);
 static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
 					 int nr);
 static inline int ocfs2_block_group_set_bits(handle_t *handle,
@@ -130,6 +137,7 @@ void ocfs2_free_ac_resource(struct ocfs2_alloc_context *ac)
 	}
 	brelse(ac->ac_bh);
 	ac->ac_bh = NULL;
+	ac->ac_resv = NULL;
 }
 
 void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac)
@@ -325,14 +333,38 @@ out:
 	return rc;
 }
 
+static void ocfs2_bg_discontig_add_extent(struct ocfs2_super *osb,
+					  struct ocfs2_group_desc *bg,
+					  struct ocfs2_chain_list *cl,
+					  u64 p_blkno, u32 clusters)
+{
+	struct ocfs2_extent_list *el = &bg->bg_list;
+	struct ocfs2_extent_rec *rec;
+
+	BUG_ON(!ocfs2_supports_discontig_bg(osb));
+	if (!el->l_next_free_rec)
+		el->l_count = cpu_to_le16(ocfs2_extent_recs_per_gd(osb->sb));
+	rec = &el->l_recs[le16_to_cpu(el->l_next_free_rec)];
+	rec->e_blkno = cpu_to_le64(p_blkno);
+	rec->e_cpos = cpu_to_le32(le16_to_cpu(bg->bg_bits) /
+				  le16_to_cpu(cl->cl_bpc));
+	rec->e_leaf_clusters = cpu_to_le32(clusters);
+	le16_add_cpu(&bg->bg_bits, clusters * le16_to_cpu(cl->cl_bpc));
+	le16_add_cpu(&bg->bg_free_bits_count,
+		     clusters * le16_to_cpu(cl->cl_bpc));
+	le16_add_cpu(&el->l_next_free_rec, 1);
+}
+
 static int ocfs2_block_group_fill(handle_t *handle,
 				  struct inode *alloc_inode,
 				  struct buffer_head *bg_bh,
 				  u64 group_blkno,
+				  unsigned int group_clusters,
 				  u16 my_chain,
 				  struct ocfs2_chain_list *cl)
 {
 	int status = 0;
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
 	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
 	struct super_block * sb = alloc_inode->i_sb;
 
@@ -359,19 +391,23 @@ static int ocfs2_block_group_fill(handle_t *handle,
 	memset(bg, 0, sb->s_blocksize);
 	strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
 	bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
-	bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb));
-	bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
+	bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1,
+						osb->s_feature_incompat));
 	bg->bg_chain = cpu_to_le16(my_chain);
 	bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
 	bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno);
 	bg->bg_blkno = cpu_to_le64(group_blkno);
+	if (group_clusters == le16_to_cpu(cl->cl_cpg))
+		bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
+	else
+		ocfs2_bg_discontig_add_extent(osb, bg, cl, group_blkno,
+					      group_clusters);
+
 	/* set the 1st bit in the bitmap to account for the descriptor block */
 	ocfs2_set_bit(0, (unsigned long *)bg->bg_bitmap);
 	bg->bg_free_bits_count = cpu_to_le16(le16_to_cpu(bg->bg_bits) - 1);
 
-	status = ocfs2_journal_dirty(handle, bg_bh);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_journal_dirty(handle, bg_bh);
 
 	/* There is no need to zero out or otherwise initialize the
 	 * other blocks in a group - All valid FS metadata in a block
@@ -397,6 +433,238 @@ static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl)
 	return best;
 }
 
+static struct buffer_head *
+ocfs2_block_group_alloc_contig(struct ocfs2_super *osb, handle_t *handle,
+			       struct inode *alloc_inode,
+			       struct ocfs2_alloc_context *ac,
+			       struct ocfs2_chain_list *cl)
+{
+	int status;
+	u32 bit_off, num_bits;
+	u64 bg_blkno;
+	struct buffer_head *bg_bh;
+	unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
+
+	status = ocfs2_claim_clusters(handle, ac,
+				      le16_to_cpu(cl->cl_cpg), &bit_off,
+				      &num_bits);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	/* setup the group */
+	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+	mlog(0, "new descriptor, record %u, at block %llu\n",
+	     alloc_rec, (unsigned long long)bg_blkno);
+
+	bg_bh = sb_getblk(osb->sb, bg_blkno);
+	if (!bg_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
+
+	status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh,
+					bg_blkno, num_bits, alloc_rec, cl);
+	if (status < 0) {
+		brelse(bg_bh);
+		mlog_errno(status);
+	}
+
+bail:
+	return status ? ERR_PTR(status) : bg_bh;
+}
+
+static int ocfs2_block_group_claim_bits(struct ocfs2_super *osb,
+					handle_t *handle,
+					struct ocfs2_alloc_context *ac,
+					unsigned int min_bits,
+					u32 *bit_off, u32 *num_bits)
+{
+	int status = 0;
+
+	while (min_bits) {
+		status = ocfs2_claim_clusters(handle, ac, min_bits,
+					      bit_off, num_bits);
+		if (status != -ENOSPC)
+			break;
+
+		min_bits >>= 1;
+	}
+
+	return status;
+}
+
+static int ocfs2_block_group_grow_discontig(handle_t *handle,
+					    struct inode *alloc_inode,
+					    struct buffer_head *bg_bh,
+					    struct ocfs2_alloc_context *ac,
+					    struct ocfs2_chain_list *cl,
+					    unsigned int min_bits)
+{
+	int status;
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+	struct ocfs2_group_desc *bg =
+		(struct ocfs2_group_desc *)bg_bh->b_data;
+	unsigned int needed = le16_to_cpu(cl->cl_cpg) -
+			 le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
+	u32 p_cpos, clusters;
+	u64 p_blkno;
+	struct ocfs2_extent_list *el = &bg->bg_list;
+
+	status = ocfs2_journal_access_gd(handle,
+					 INODE_CACHE(alloc_inode),
+					 bg_bh,
+					 OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	while ((needed > 0) && (le16_to_cpu(el->l_next_free_rec) <
+				le16_to_cpu(el->l_count))) {
+		if (min_bits > needed)
+			min_bits = needed;
+		status = ocfs2_block_group_claim_bits(osb, handle, ac,
+						      min_bits, &p_cpos,
+						      &clusters);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+		p_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cpos);
+		ocfs2_bg_discontig_add_extent(osb, bg, cl, p_blkno,
+					      clusters);
+
+		min_bits = clusters;
+		needed = le16_to_cpu(cl->cl_cpg) -
+			 le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
+	}
+
+	if (needed > 0) {
+		/*
+		 * We have used up all the extent rec but can't fill up
+		 * the cpg. So bail out.
+		 */
+		status = -ENOSPC;
+		goto bail;
+	}
+
+	ocfs2_journal_dirty(handle, bg_bh);
+
+bail:
+	return status;
+}
+
+static void ocfs2_bg_alloc_cleanup(handle_t *handle,
+				   struct ocfs2_alloc_context *cluster_ac,
+				   struct inode *alloc_inode,
+				   struct buffer_head *bg_bh)
+{
+	int i, ret;
+	struct ocfs2_group_desc *bg;
+	struct ocfs2_extent_list *el;
+	struct ocfs2_extent_rec *rec;
+
+	if (!bg_bh)
+		return;
+
+	bg = (struct ocfs2_group_desc *)bg_bh->b_data;
+	el = &bg->bg_list;
+	for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
+		rec = &el->l_recs[i];
+		ret = ocfs2_free_clusters(handle, cluster_ac->ac_inode,
+					  cluster_ac->ac_bh,
+					  le64_to_cpu(rec->e_blkno),
+					  le32_to_cpu(rec->e_leaf_clusters));
+		if (ret)
+			mlog_errno(ret);
+		/* Try all the clusters to free */
+	}
+
+	ocfs2_remove_from_cache(INODE_CACHE(alloc_inode), bg_bh);
+	brelse(bg_bh);
+}
+
+static struct buffer_head *
+ocfs2_block_group_alloc_discontig(handle_t *handle,
+				  struct inode *alloc_inode,
+				  struct ocfs2_alloc_context *ac,
+				  struct ocfs2_chain_list *cl)
+{
+	int status;
+	u32 bit_off, num_bits;
+	u64 bg_blkno;
+	unsigned int min_bits = le16_to_cpu(cl->cl_cpg) >> 1;
+	struct buffer_head *bg_bh = NULL;
+	unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+
+	if (!ocfs2_supports_discontig_bg(osb)) {
+		status = -ENOSPC;
+		goto bail;
+	}
+
+	status = ocfs2_extend_trans(handle,
+				    ocfs2_calc_bg_discontig_credits(osb->sb));
+	if (status) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/*
+	 * We're going to be grabbing from multiple cluster groups.
+	 * We don't have enough credits to relink them all, and the
+	 * cluster groups will be staying in cache for the duration of
+	 * this operation.
+	 */
+	ac->ac_allow_chain_relink = 0;
+
+	/* Claim the first region */
+	status = ocfs2_block_group_claim_bits(osb, handle, ac, min_bits,
+					      &bit_off, &num_bits);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+	min_bits = num_bits;
+
+	/* setup the group */
+	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+	mlog(0, "new descriptor, record %u, at block %llu\n",
+	     alloc_rec, (unsigned long long)bg_blkno);
+
+	bg_bh = sb_getblk(osb->sb, bg_blkno);
+	if (!bg_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
+
+	status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh,
+					bg_blkno, num_bits, alloc_rec, cl);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_block_group_grow_discontig(handle, alloc_inode,
+						  bg_bh, ac, cl, min_bits);
+	if (status)
+		mlog_errno(status);
+
+bail:
+	if (status)
+		ocfs2_bg_alloc_cleanup(handle, ac, alloc_inode, bg_bh);
+	return status ? ERR_PTR(status) : bg_bh;
+}
+
 /*
  * We expect the block group allocator to already be locked.
  */
@@ -412,9 +680,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 	struct ocfs2_chain_list *cl;
 	struct ocfs2_alloc_context *ac = NULL;
 	handle_t *handle = NULL;
-	u32 bit_off, num_bits;
 	u16 alloc_rec;
-	u64 bg_blkno;
 	struct buffer_head *bg_bh = NULL;
 	struct ocfs2_group_desc *bg;
 
@@ -447,44 +713,20 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 		     (unsigned long long)*last_alloc_group);
 		ac->ac_last_group = *last_alloc_group;
 	}
-	status = ocfs2_claim_clusters(osb,
-				      handle,
-				      ac,
-				      le16_to_cpu(cl->cl_cpg),
-				      &bit_off,
-				      &num_bits);
-	if (status < 0) {
+
+	bg_bh = ocfs2_block_group_alloc_contig(osb, handle, alloc_inode,
+					       ac, cl);
+	if (IS_ERR(bg_bh) && (PTR_ERR(bg_bh) == -ENOSPC))
+		bg_bh = ocfs2_block_group_alloc_discontig(handle,
+							  alloc_inode,
+							  ac, cl);
+	if (IS_ERR(bg_bh)) {
+		status = PTR_ERR(bg_bh);
+		bg_bh = NULL;
 		if (status != -ENOSPC)
 			mlog_errno(status);
 		goto bail;
 	}
-
-	alloc_rec = ocfs2_find_smallest_chain(cl);
-
-	/* setup the group */
-	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
-	mlog(0, "new descriptor, record %u, at block %llu\n",
-	     alloc_rec, (unsigned long long)bg_blkno);
-
-	bg_bh = sb_getblk(osb->sb, bg_blkno);
-	if (!bg_bh) {
-		status = -EIO;
-		mlog_errno(status);
-		goto bail;
-	}
-	ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
-
-	status = ocfs2_block_group_fill(handle,
-					alloc_inode,
-					bg_bh,
-					bg_blkno,
-					alloc_rec,
-					cl);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
 	bg = (struct ocfs2_group_desc *) bg_bh->b_data;
 
 	status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode),
@@ -494,10 +736,12 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 		goto bail;
 	}
 
+	alloc_rec = le16_to_cpu(bg->bg_chain);
 	le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
 		     le16_to_cpu(bg->bg_free_bits_count));
-	le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, le16_to_cpu(bg->bg_bits));
-	cl->cl_recs[alloc_rec].c_blkno  = cpu_to_le64(bg_blkno);
+	le32_add_cpu(&cl->cl_recs[alloc_rec].c_total,
+		     le16_to_cpu(bg->bg_bits));
+	cl->cl_recs[alloc_rec].c_blkno  = cpu_to_le64(bg->bg_blkno);
 	if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
 		le16_add_cpu(&cl->cl_next_free_rec, 1);
 
@@ -506,11 +750,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 	le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
 	le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
 
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	spin_lock(&OCFS2_I(alloc_inode)->ip_lock);
 	OCFS2_I(alloc_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
@@ -760,7 +1000,7 @@ int ocfs2_reserve_new_metadata_blocks(struct ocfs2_super *osb,
 	status = ocfs2_reserve_suballoc_bits(osb, (*ac),
 					     EXTENT_ALLOC_SYSTEM_INODE,
 					     (u32)osb->slot_num, NULL,
-					     ALLOC_NEW_GROUP);
+					     ALLOC_GROUPS_FROM_GLOBAL|ALLOC_NEW_GROUP);
 
 
 	if (status >= 0) {
@@ -946,11 +1186,7 @@ static int ocfs2_reserve_clusters_with_limit(struct ocfs2_super *osb,
 		status = ocfs2_reserve_local_alloc_bits(osb,
 							bits_wanted,
 							*ac);
-		if (status == -EFBIG) {
-			/* The local alloc window is outside ac_max_block.
-			 * use the main bitmap. */
-			status = -ENOSPC;
-		} else if ((status < 0) && (status != -ENOSPC)) {
+		if ((status < 0) && (status != -ENOSPC)) {
 			mlog_errno(status);
 			goto bail;
 		}
@@ -1033,8 +1269,7 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
 					     struct buffer_head *bg_bh,
 					     unsigned int bits_wanted,
 					     unsigned int total_bits,
-					     u16 *bit_off,
-					     u16 *bits_found)
+					     struct ocfs2_suballoc_result *res)
 {
 	void *bitmap;
 	u16 best_offset, best_size;
@@ -1078,14 +1313,9 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
 		}
 	}
 
-	/* XXX: I think the first clause is equivalent to the second
-	 * 	- jlbec */
-	if (found == bits_wanted) {
-		*bit_off = start - found;
-		*bits_found = found;
-	} else if (best_size) {
-		*bit_off = best_offset;
-		*bits_found = best_size;
+	if (best_size) {
+		res->sr_bit_offset = best_offset;
+		res->sr_bits = best_size;
 	} else {
 		status = -ENOSPC;
 		/* No error log here -- see the comment above
@@ -1129,16 +1359,10 @@ static inline int ocfs2_block_group_set_bits(handle_t *handle,
 	}
 
 	le16_add_cpu(&bg->bg_free_bits_count, -num_bits);
-
 	while(num_bits--)
 		ocfs2_set_bit(bit_off++, bitmap);
 
-	status = ocfs2_journal_dirty(handle,
-				     group_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, group_bh);
 
 bail:
 	mlog_exit(status);
@@ -1202,12 +1426,7 @@ static int ocfs2_relink_block_group(handle_t *handle,
 	}
 
 	prev_bg->bg_next_group = bg->bg_next_group;
-
-	status = ocfs2_journal_dirty(handle, prev_bg_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_rollback;
-	}
+	ocfs2_journal_dirty(handle, prev_bg_bh);
 
 	status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode),
 					 bg_bh, OCFS2_JOURNAL_ACCESS_WRITE);
@@ -1217,12 +1436,7 @@ static int ocfs2_relink_block_group(handle_t *handle,
 	}
 
 	bg->bg_next_group = fe->id2.i_chain.cl_recs[chain].c_blkno;
-
-	status = ocfs2_journal_dirty(handle, bg_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_rollback;
-	}
+	ocfs2_journal_dirty(handle, bg_bh);
 
 	status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode),
 					 fe_bh, OCFS2_JOURNAL_ACCESS_WRITE);
@@ -1232,14 +1446,8 @@ static int ocfs2_relink_block_group(handle_t *handle,
 	}
 
 	fe->id2.i_chain.cl_recs[chain].c_blkno = bg->bg_blkno;
+	ocfs2_journal_dirty(handle, fe_bh);
 
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_rollback;
-	}
-
-	status = 0;
 out_rollback:
 	if (status < 0) {
 		fe->id2.i_chain.cl_recs[chain].c_blkno = cpu_to_le64(fe_ptr);
@@ -1263,14 +1471,13 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 				      struct buffer_head *group_bh,
 				      u32 bits_wanted, u32 min_bits,
 				      u64 max_block,
-				      u16 *bit_off, u16 *bits_found)
+				      struct ocfs2_suballoc_result *res)
 {
 	int search = -ENOSPC;
 	int ret;
 	u64 blkoff;
 	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *) group_bh->b_data;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	u16 tmp_off, tmp_found;
 	unsigned int max_bits, gd_cluster_off;
 
 	BUG_ON(!ocfs2_is_cluster_bitmap(inode));
@@ -1297,15 +1504,15 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 
 		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
 							group_bh, bits_wanted,
-							max_bits,
-							&tmp_off, &tmp_found);
+							max_bits, res);
 		if (ret)
 			return ret;
 
 		if (max_block) {
 			blkoff = ocfs2_clusters_to_blocks(inode->i_sb,
 							  gd_cluster_off +
-							  tmp_off + tmp_found);
+							  res->sr_bit_offset +
+							  res->sr_bits);
 			mlog(0, "Checking %llu against %llu\n",
 			     (unsigned long long)blkoff,
 			     (unsigned long long)max_block);
@@ -1317,16 +1524,14 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 		 * return success, but we still want to return
 		 * -ENOSPC unless it found the minimum number
 		 * of bits. */
-		if (min_bits <= tmp_found) {
-			*bit_off = tmp_off;
-			*bits_found = tmp_found;
+		if (min_bits <= res->sr_bits)
 			search = 0; /* success */
-		} else if (tmp_found) {
+		else if (res->sr_bits) {
 			/*
 			 * Don't show bits which we'll be returning
 			 * for allocation to the local alloc bitmap.
 			 */
-			ocfs2_local_alloc_seen_free_bits(osb, tmp_found);
+			ocfs2_local_alloc_seen_free_bits(osb, res->sr_bits);
 		}
 	}
 
@@ -1337,7 +1542,7 @@ static int ocfs2_block_group_search(struct inode *inode,
 				    struct buffer_head *group_bh,
 				    u32 bits_wanted, u32 min_bits,
 				    u64 max_block,
-				    u16 *bit_off, u16 *bits_found)
+				    struct ocfs2_suballoc_result *res)
 {
 	int ret = -ENOSPC;
 	u64 blkoff;
@@ -1350,10 +1555,10 @@ static int ocfs2_block_group_search(struct inode *inode,
 		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
 							group_bh, bits_wanted,
 							le16_to_cpu(bg->bg_bits),
-							bit_off, bits_found);
+							res);
 		if (!ret && max_block) {
-			blkoff = le64_to_cpu(bg->bg_blkno) + *bit_off +
-				*bits_found;
+			blkoff = le64_to_cpu(bg->bg_blkno) +
+				res->sr_bit_offset + res->sr_bits;
 			mlog(0, "Checking %llu against %llu\n",
 			     (unsigned long long)blkoff,
 			     (unsigned long long)max_block);
@@ -1386,33 +1591,76 @@ static int ocfs2_alloc_dinode_update_counts(struct inode *inode,
 	tmp_used = le32_to_cpu(di->id1.bitmap1.i_used);
 	di->id1.bitmap1.i_used = cpu_to_le32(num_bits + tmp_used);
 	le32_add_cpu(&cl->cl_recs[chain].c_free, -num_bits);
-
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, di_bh);
 
 out:
 	return ret;
 }
 
+static int ocfs2_bg_discontig_fix_by_rec(struct ocfs2_suballoc_result *res,
+					 struct ocfs2_extent_rec *rec,
+					 struct ocfs2_chain_list *cl)
+{
+	unsigned int bpc = le16_to_cpu(cl->cl_bpc);
+	unsigned int bitoff = le32_to_cpu(rec->e_cpos) * bpc;
+	unsigned int bitcount = le32_to_cpu(rec->e_leaf_clusters) * bpc;
+
+	if (res->sr_bit_offset < bitoff)
+		return 0;
+	if (res->sr_bit_offset >= (bitoff + bitcount))
+		return 0;
+	res->sr_blkno = le64_to_cpu(rec->e_blkno) +
+		(res->sr_bit_offset - bitoff);
+	if ((res->sr_bit_offset + res->sr_bits) > (bitoff + bitcount))
+		res->sr_bits = (bitoff + bitcount) - res->sr_bit_offset;
+	return 1;
+}
+
+static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac,
+					  struct ocfs2_group_desc *bg,
+					  struct ocfs2_suballoc_result *res)
+{
+	int i;
+	u64 bg_blkno = res->sr_bg_blkno;  /* Save off */
+	struct ocfs2_extent_rec *rec;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
+	struct ocfs2_chain_list *cl = &di->id2.i_chain;
+
+	if (ocfs2_is_cluster_bitmap(ac->ac_inode)) {
+		res->sr_blkno = 0;
+		return;
+	}
+
+	res->sr_blkno = res->sr_bg_blkno + res->sr_bit_offset;
+	res->sr_bg_blkno = 0;  /* Clear it for contig block groups */
+	if (!ocfs2_supports_discontig_bg(OCFS2_SB(ac->ac_inode->i_sb)) ||
+	    !bg->bg_list.l_next_free_rec)
+		return;
+
+	for (i = 0; i < le16_to_cpu(bg->bg_list.l_next_free_rec); i++) {
+		rec = &bg->bg_list.l_recs[i];
+		if (ocfs2_bg_discontig_fix_by_rec(res, rec, cl)) {
+			res->sr_bg_blkno = bg_blkno;  /* Restore */
+			break;
+		}
+	}
+}
+
 static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
 				  handle_t *handle,
 				  u32 bits_wanted,
 				  u32 min_bits,
-				  u16 *bit_off,
-				  unsigned int *num_bits,
-				  u64 gd_blkno,
+				  struct ocfs2_suballoc_result *res,
 				  u16 *bits_left)
 {
 	int ret;
-	u16 found;
 	struct buffer_head *group_bh = NULL;
 	struct ocfs2_group_desc *gd;
 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
 	struct inode *alloc_inode = ac->ac_inode;
 
-	ret = ocfs2_read_group_descriptor(alloc_inode, di, gd_blkno,
-					  &group_bh);
+	ret = ocfs2_read_group_descriptor(alloc_inode, di,
+					  res->sr_bg_blkno, &group_bh);
 	if (ret < 0) {
 		mlog_errno(ret);
 		return ret;
@@ -1420,17 +1668,18 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
 
 	gd = (struct ocfs2_group_desc *) group_bh->b_data;
 	ret = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits,
-				  ac->ac_max_block, bit_off, &found);
+				  ac->ac_max_block, res);
 	if (ret < 0) {
 		if (ret != -ENOSPC)
 			mlog_errno(ret);
 		goto out;
 	}
 
-	*num_bits = found;
+	if (!ret)
+		ocfs2_bg_discontig_fix_result(ac, gd, res);
 
 	ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
-					       *num_bits,
+					       res->sr_bits,
 					       le16_to_cpu(gd->bg_chain));
 	if (ret < 0) {
 		mlog_errno(ret);
@@ -1438,7 +1687,7 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
 	}
 
 	ret = ocfs2_block_group_set_bits(handle, alloc_inode, gd, group_bh,
-					 *bit_off, *num_bits);
+					 res->sr_bit_offset, res->sr_bits);
 	if (ret < 0)
 		mlog_errno(ret);
 
@@ -1454,13 +1703,11 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 			      handle_t *handle,
 			      u32 bits_wanted,
 			      u32 min_bits,
-			      u16 *bit_off,
-			      unsigned int *num_bits,
-			      u64 *bg_blkno,
+			      struct ocfs2_suballoc_result *res,
 			      u16 *bits_left)
 {
 	int status;
-	u16 chain, tmp_bits;
+	u16 chain;
 	u32 tmp_used;
 	u64 next_group;
 	struct inode *alloc_inode = ac->ac_inode;
@@ -1489,8 +1736,8 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	 * the 1st group with any empty bits. */
 	while ((status = ac->ac_group_search(alloc_inode, group_bh,
 					     bits_wanted, min_bits,
-					     ac->ac_max_block, bit_off,
-					     &tmp_bits)) == -ENOSPC) {
+					     ac->ac_max_block,
+					     res)) == -ENOSPC) {
 		if (!bg->bg_next_group)
 			break;
 
@@ -1515,11 +1762,14 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	}
 
 	mlog(0, "alloc succeeds: we give %u bits from block group %llu\n",
-	     tmp_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
+	     res->sr_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
 
-	*num_bits = tmp_bits;
+	res->sr_bg_blkno = le64_to_cpu(bg->bg_blkno);
+
+	BUG_ON(res->sr_bits == 0);
+	if (!status)
+		ocfs2_bg_discontig_fix_result(ac, bg, res);
 
-	BUG_ON(*num_bits == 0);
 
 	/*
 	 * Keep track of previous block descriptor read. When
@@ -1536,7 +1786,7 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	 */
 	if (ac->ac_allow_chain_relink &&
 	    (prev_group_bh) &&
-	    (ocfs2_block_group_reasonably_empty(bg, *num_bits))) {
+	    (ocfs2_block_group_reasonably_empty(bg, res->sr_bits))) {
 		status = ocfs2_relink_block_group(handle, alloc_inode,
 						  ac->ac_bh, group_bh,
 						  prev_group_bh, chain);
@@ -1558,31 +1808,24 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	}
 
 	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
-	fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used);
-	le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits));
-
-	status = ocfs2_journal_dirty(handle,
-				     ac->ac_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	fe->id1.bitmap1.i_used = cpu_to_le32(res->sr_bits + tmp_used);
+	le32_add_cpu(&cl->cl_recs[chain].c_free, -res->sr_bits);
+	ocfs2_journal_dirty(handle, ac->ac_bh);
 
 	status = ocfs2_block_group_set_bits(handle,
 					    alloc_inode,
 					    bg,
 					    group_bh,
-					    *bit_off,
-					    *num_bits);
+					    res->sr_bit_offset,
+					    res->sr_bits);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
 
-	mlog(0, "Allocated %u bits from suballocator %llu\n", *num_bits,
+	mlog(0, "Allocated %u bits from suballocator %llu\n", res->sr_bits,
 	     (unsigned long long)le64_to_cpu(fe->i_blkno));
 
-	*bg_blkno = le64_to_cpu(bg->bg_blkno);
 	*bits_left = le16_to_cpu(bg->bg_free_bits_count);
 bail:
 	brelse(group_bh);
@@ -1593,19 +1836,15 @@ bail:
 }
 
 /* will give out up to bits_wanted contiguous bits. */
-static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
-				     struct ocfs2_alloc_context *ac,
+static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
 				     handle_t *handle,
 				     u32 bits_wanted,
 				     u32 min_bits,
-				     u16 *bit_off,
-				     unsigned int *num_bits,
-				     u64 *bg_blkno)
+				     struct ocfs2_suballoc_result *res)
 {
 	int status;
 	u16 victim, i;
 	u16 bits_left = 0;
-	u64 hint_blkno = ac->ac_last_group;
 	struct ocfs2_chain_list *cl;
 	struct ocfs2_dinode *fe;
 
@@ -1623,7 +1862,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 
 	if (le32_to_cpu(fe->id1.bitmap1.i_used) >=
 	    le32_to_cpu(fe->id1.bitmap1.i_total)) {
-		ocfs2_error(osb->sb, "Chain allocator dinode %llu has %u used "
+		ocfs2_error(ac->ac_inode->i_sb,
+			    "Chain allocator dinode %llu has %u used "
 			    "bits but only %u total.",
 			    (unsigned long long)le64_to_cpu(fe->i_blkno),
 			    le32_to_cpu(fe->id1.bitmap1.i_used),
@@ -1632,22 +1872,16 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 		goto bail;
 	}
 
-	if (hint_blkno) {
+	res->sr_bg_blkno = ac->ac_last_group;
+	if (res->sr_bg_blkno) {
 		/* Attempt to short-circuit the usual search mechanism
 		 * by jumping straight to the most recently used
 		 * allocation group. This helps us mantain some
 		 * contiguousness across allocations. */
 		status = ocfs2_search_one_group(ac, handle, bits_wanted,
-						min_bits, bit_off, num_bits,
-						hint_blkno, &bits_left);
-		if (!status) {
-			/* Be careful to update *bg_blkno here as the
-			 * caller is expecting it to be filled in, and
-			 * ocfs2_search_one_group() won't do that for
-			 * us. */
-			*bg_blkno = hint_blkno;
+						min_bits, res, &bits_left);
+		if (!status)
 			goto set_hint;
-		}
 		if (status < 0 && status != -ENOSPC) {
 			mlog_errno(status);
 			goto bail;
@@ -1660,8 +1894,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 	ac->ac_chain = victim;
 	ac->ac_allow_chain_relink = 1;
 
-	status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, bit_off,
-				    num_bits, bg_blkno, &bits_left);
+	status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
+				    res, &bits_left);
 	if (!status)
 		goto set_hint;
 	if (status < 0 && status != -ENOSPC) {
@@ -1685,8 +1919,7 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 
 		ac->ac_chain = i;
 		status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
-					    bit_off, num_bits, bg_blkno,
-					    &bits_left);
+					    res, &bits_left);
 		if (!status)
 			break;
 		if (status < 0 && status != -ENOSPC) {
@@ -1703,7 +1936,7 @@ set_hint:
 		if (bits_left < min_bits)
 			ac->ac_last_group = 0;
 		else
-			ac->ac_last_group = *bg_blkno;
+			ac->ac_last_group = res->sr_bg_blkno;
 	}
 
 bail:
@@ -1711,37 +1944,37 @@ bail:
 	return status;
 }
 
-int ocfs2_claim_metadata(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_metadata(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 bits_wanted,
+			 u64 *suballoc_loc,
 			 u16 *suballoc_bit_start,
 			 unsigned int *num_bits,
 			 u64 *blkno_start)
 {
 	int status;
-	u64 bg_blkno;
+	struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
 
 	BUG_ON(!ac);
 	BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
 	BUG_ON(ac->ac_which != OCFS2_AC_USE_META);
 
-	status = ocfs2_claim_suballoc_bits(osb,
-					   ac,
+	status = ocfs2_claim_suballoc_bits(ac,
 					   handle,
 					   bits_wanted,
 					   1,
-					   suballoc_bit_start,
-					   num_bits,
-					   &bg_blkno);
+					   &res);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
-	atomic_inc(&osb->alloc_stats.bg_allocs);
+	atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
 
-	*blkno_start = bg_blkno + (u64) *suballoc_bit_start;
-	ac->ac_bits_given += (*num_bits);
+	*suballoc_loc = res.sr_bg_blkno;
+	*suballoc_bit_start = res.sr_bit_offset;
+	*blkno_start = res.sr_blkno;
+	ac->ac_bits_given += res.sr_bits;
+	*num_bits = res.sr_bits;
 	status = 0;
 bail:
 	mlog_exit(status);
@@ -1749,10 +1982,10 @@ bail:
 }
 
 static void ocfs2_init_inode_ac_group(struct inode *dir,
-				      struct buffer_head *parent_fe_bh,
+				      struct buffer_head *parent_di_bh,
 				      struct ocfs2_alloc_context *ac)
 {
-	struct ocfs2_dinode *fe = (struct ocfs2_dinode *)parent_fe_bh->b_data;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)parent_di_bh->b_data;
 	/*
 	 * Try to allocate inodes from some specific group.
 	 *
@@ -1766,10 +1999,14 @@ static void ocfs2_init_inode_ac_group(struct inode *dir,
 	if (OCFS2_I(dir)->ip_last_used_group &&
 	    OCFS2_I(dir)->ip_last_used_slot == ac->ac_alloc_slot)
 		ac->ac_last_group = OCFS2_I(dir)->ip_last_used_group;
-	else if (le16_to_cpu(fe->i_suballoc_slot) == ac->ac_alloc_slot)
-		ac->ac_last_group = ocfs2_which_suballoc_group(
-					le64_to_cpu(fe->i_blkno),
-					le16_to_cpu(fe->i_suballoc_bit));
+	else if (le16_to_cpu(di->i_suballoc_slot) == ac->ac_alloc_slot) {
+		if (di->i_suballoc_loc)
+			ac->ac_last_group = le64_to_cpu(di->i_suballoc_loc);
+		else
+			ac->ac_last_group = ocfs2_which_suballoc_group(
+					le64_to_cpu(di->i_blkno),
+					le16_to_cpu(di->i_suballoc_bit));
+	}
 }
 
 static inline void ocfs2_save_inode_ac_group(struct inode *dir,
@@ -1779,17 +2016,16 @@ static inline void ocfs2_save_inode_ac_group(struct inode *dir,
 	OCFS2_I(dir)->ip_last_used_slot = ac->ac_alloc_slot;
 }
 
-int ocfs2_claim_new_inode(struct ocfs2_super *osb,
-			  handle_t *handle,
+int ocfs2_claim_new_inode(handle_t *handle,
 			  struct inode *dir,
 			  struct buffer_head *parent_fe_bh,
 			  struct ocfs2_alloc_context *ac,
+			  u64 *suballoc_loc,
 			  u16 *suballoc_bit,
 			  u64 *fe_blkno)
 {
 	int status;
-	unsigned int num_bits;
-	u64 bg_blkno;
+	struct ocfs2_suballoc_result res;
 
 	mlog_entry_void();
 
@@ -1800,23 +2036,22 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
 
 	ocfs2_init_inode_ac_group(dir, parent_fe_bh, ac);
 
-	status = ocfs2_claim_suballoc_bits(osb,
-					   ac,
+	status = ocfs2_claim_suballoc_bits(ac,
 					   handle,
 					   1,
 					   1,
-					   suballoc_bit,
-					   &num_bits,
-					   &bg_blkno);
+					   &res);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
-	atomic_inc(&osb->alloc_stats.bg_allocs);
+	atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
 
-	BUG_ON(num_bits != 1);
+	BUG_ON(res.sr_bits != 1);
 
-	*fe_blkno = bg_blkno + (u64) (*suballoc_bit);
+	*suballoc_loc = res.sr_bg_blkno;
+	*suballoc_bit = res.sr_bit_offset;
+	*fe_blkno = res.sr_blkno;
 	ac->ac_bits_given++;
 	ocfs2_save_inode_ac_group(dir, ac);
 	status = 0;
@@ -1886,8 +2121,7 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode,
  * contig. allocation, set to '1' to indicate we can deal with extents
  * of any size.
  */
-int __ocfs2_claim_clusters(struct ocfs2_super *osb,
-			   handle_t *handle,
+int __ocfs2_claim_clusters(handle_t *handle,
 			   struct ocfs2_alloc_context *ac,
 			   u32 min_clusters,
 			   u32 max_clusters,
@@ -1896,8 +2130,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
 {
 	int status;
 	unsigned int bits_wanted = max_clusters;
-	u64 bg_blkno = 0;
-	u16 bg_bit_off;
+	struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
+	struct ocfs2_super *osb = OCFS2_SB(ac->ac_inode->i_sb);
 
 	mlog_entry_void();
 
@@ -1907,6 +2141,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
 	       && ac->ac_which != OCFS2_AC_USE_MAIN);
 
 	if (ac->ac_which == OCFS2_AC_USE_LOCAL) {
+		WARN_ON(min_clusters > 1);
+
 		status = ocfs2_claim_local_alloc_bits(osb,
 						      handle,
 						      ac,
@@ -1929,20 +2165,19 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
 		if (bits_wanted > (osb->bitmap_cpg - 1))
 			bits_wanted = osb->bitmap_cpg - 1;
 
-		status = ocfs2_claim_suballoc_bits(osb,
-						   ac,
+		status = ocfs2_claim_suballoc_bits(ac,
 						   handle,
 						   bits_wanted,
 						   min_clusters,
-						   &bg_bit_off,
-						   num_clusters,
-						   &bg_blkno);
+						   &res);
 		if (!status) {
+			BUG_ON(res.sr_blkno); /* cluster alloc can't set */
 			*cluster_start =
 				ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode,
-								 bg_blkno,
-								 bg_bit_off);
+								 res.sr_bg_blkno,
+								 res.sr_bit_offset);
 			atomic_inc(&osb->alloc_stats.bitmap_data);
+			*num_clusters = res.sr_bits;
 		}
 	}
 	if (status < 0) {
@@ -1958,8 +2193,7 @@ bail:
 	return status;
 }
 
-int ocfs2_claim_clusters(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_clusters(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 min_clusters,
 			 u32 *cluster_start,
@@ -1967,7 +2201,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
 {
 	unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given;
 
-	return __ocfs2_claim_clusters(osb, handle, ac, min_clusters,
+	return __ocfs2_claim_clusters(handle, ac, min_clusters,
 				      bits_wanted, cluster_start, num_clusters);
 }
 
@@ -2023,9 +2257,7 @@ static int ocfs2_block_group_clear_bits(handle_t *handle,
 	if (undo_fn)
 		jbd_unlock_bh_state(group_bh);
 
-	status = ocfs2_journal_dirty(handle, group_bh);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_journal_dirty(handle, group_bh);
 bail:
 	return status;
 }
@@ -2092,12 +2324,7 @@ static int _ocfs2_free_suballoc_bits(handle_t *handle,
 		     count);
 	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
 	fe->id1.bitmap1.i_used = cpu_to_le32(tmp_used - count);
-
-	status = ocfs2_journal_dirty(handle, alloc_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, alloc_bh);
 
 bail:
 	brelse(group_bh);
@@ -2126,6 +2353,8 @@ int ocfs2_free_dinode(handle_t *handle,
 	u16 bit = le16_to_cpu(di->i_suballoc_bit);
 	u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 
+	if (di->i_suballoc_loc)
+		bg_blkno = le64_to_cpu(di->i_suballoc_loc);
 	return ocfs2_free_suballoc_bits(handle, inode_alloc_inode,
 					inode_alloc_bh, bit, bg_blkno, 1);
 }
@@ -2395,7 +2624,7 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
 				   struct buffer_head *alloc_bh, u64 blkno,
 				   u16 bit, int *res)
 {
-	struct ocfs2_dinode *alloc_fe;
+	struct ocfs2_dinode *alloc_di;
 	struct ocfs2_group_desc *group;
 	struct buffer_head *group_bh = NULL;
 	u64 bg_blkno;
@@ -2404,17 +2633,20 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
 	mlog_entry("blkno: %llu bit: %u\n", (unsigned long long)blkno,
 		   (unsigned int)bit);
 
-	alloc_fe = (struct ocfs2_dinode *)alloc_bh->b_data;
-	if ((bit + 1) > ocfs2_bits_per_group(&alloc_fe->id2.i_chain)) {
+	alloc_di = (struct ocfs2_dinode *)alloc_bh->b_data;
+	if ((bit + 1) > ocfs2_bits_per_group(&alloc_di->id2.i_chain)) {
 		mlog(ML_ERROR, "suballoc bit %u out of range of %u\n",
 		     (unsigned int)bit,
-		     ocfs2_bits_per_group(&alloc_fe->id2.i_chain));
+		     ocfs2_bits_per_group(&alloc_di->id2.i_chain));
 		status = -EINVAL;
 		goto bail;
 	}
 
-	bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
-	status = ocfs2_read_group_descriptor(suballoc, alloc_fe, bg_blkno,
+	if (alloc_di->i_suballoc_loc)
+		bg_blkno = le64_to_cpu(alloc_di->i_suballoc_loc);
+	else
+		bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
+	status = ocfs2_read_group_descriptor(suballoc, alloc_di, bg_blkno,
 					     &group_bh);
 	if (status < 0) {
 		mlog(ML_ERROR, "read group %llu failed %d\n",

fs/ocfs2/suballoc.h

@@ -26,13 +26,14 @@
 #ifndef _CHAINALLOC_H_
 #define _CHAINALLOC_H_
 
+struct ocfs2_suballoc_result;
 typedef int (group_search_t)(struct inode *,
 			     struct buffer_head *,
 			     u32,			/* bits_wanted */
 			     u32,			/* min_bits */
 			     u64,			/* max_block */
-			     u16 *,			/* *bit_off */
-			     u16 *);			/* *bits_found */
+			     struct ocfs2_suballoc_result *);
+							/* found bits */
 
 struct ocfs2_alloc_context {
 	struct inode *ac_inode;    /* which bitmap are we allocating from? */
@@ -54,6 +55,8 @@ struct ocfs2_alloc_context {
 	u64    ac_last_group;
 	u64    ac_max_block;  /* Highest block number to allocate. 0 is
 				 is the same as ~0 - unlimited */
+
+	struct ocfs2_alloc_reservation	*ac_resv;
 };
 
 void ocfs2_init_steal_slots(struct ocfs2_super *osb);
@@ -80,22 +83,21 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb,
 			   u32 bits_wanted,
 			   struct ocfs2_alloc_context **ac);
 
-int ocfs2_claim_metadata(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_metadata(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 bits_wanted,
+			 u64 *suballoc_loc,
 			 u16 *suballoc_bit_start,
 			 u32 *num_bits,
 			 u64 *blkno_start);
-int ocfs2_claim_new_inode(struct ocfs2_super *osb,
-			  handle_t *handle,
+int ocfs2_claim_new_inode(handle_t *handle,
 			  struct inode *dir,
 			  struct buffer_head *parent_fe_bh,
 			  struct ocfs2_alloc_context *ac,
+			  u64 *suballoc_loc,
 			  u16 *suballoc_bit,
 			  u64 *fe_blkno);
-int ocfs2_claim_clusters(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_clusters(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 min_clusters,
 			 u32 *cluster_start,
@@ -104,8 +106,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
  * Use this variant of ocfs2_claim_clusters to specify a maxiumum
  * number of clusters smaller than the allocation reserved.
  */
-int __ocfs2_claim_clusters(struct ocfs2_super *osb,
-			   handle_t *handle,
+int __ocfs2_claim_clusters(handle_t *handle,
 			   struct ocfs2_alloc_context *ac,
 			   u32 min_clusters,
 			   u32 max_clusters,

fs/ocfs2/super.c

@@ -94,7 +94,9 @@ struct mount_options
 	unsigned long	mount_opt;
 	unsigned int	atime_quantum;
 	signed short	slot;
-	unsigned int	localalloc_opt;
+	int		localalloc_opt;
+	unsigned int	resv_level;
+	int		dir_resv_level;
 	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 };
 
@@ -176,6 +178,8 @@ enum {
 	Opt_noacl,
 	Opt_usrquota,
 	Opt_grpquota,
+	Opt_resv_level,
+	Opt_dir_resv_level,
 	Opt_err,
 };
 
@@ -202,6 +206,8 @@ static const match_table_t tokens = {
 	{Opt_noacl, "noacl"},
 	{Opt_usrquota, "usrquota"},
 	{Opt_grpquota, "grpquota"},
+	{Opt_resv_level, "resv_level=%u"},
+	{Opt_dir_resv_level, "dir_resv_level=%u"},
 	{Opt_err, NULL}
 };
 
@@ -1028,8 +1034,14 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 	osb->s_atime_quantum = parsed_options.atime_quantum;
 	osb->preferred_slot = parsed_options.slot;
 	osb->osb_commit_interval = parsed_options.commit_interval;
-	osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
-	osb->local_alloc_bits = osb->local_alloc_default_bits;
+
+	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
+	osb->osb_resv_level = parsed_options.resv_level;
+	osb->osb_dir_resv_level = parsed_options.resv_level;
+	if (parsed_options.dir_resv_level == -1)
+		osb->osb_dir_resv_level = parsed_options.resv_level;
+	else
+		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
 
 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
 	if (status)
@@ -1285,11 +1297,13 @@ static int ocfs2_parse_options(struct super_block *sb,
 		   options ? options : "(none)");
 
 	mopt->commit_interval = 0;
-	mopt->mount_opt = 0;
+	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
 	mopt->slot = OCFS2_INVALID_SLOT;
-	mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
+	mopt->localalloc_opt = -1;
 	mopt->cluster_stack[0] = '\0';
+	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
+	mopt->dir_resv_level = -1;
 
 	if (!options) {
 		status = 1;
@@ -1380,7 +1394,7 @@ static int ocfs2_parse_options(struct super_block *sb,
 				status = 0;
 				goto bail;
 			}
-			if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
+			if (option >= 0)
 				mopt->localalloc_opt = option;
 			break;
 		case Opt_localflocks:
@@ -1433,6 +1447,28 @@ static int ocfs2_parse_options(struct super_block *sb,
 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
 			break;
+		case Opt_resv_level:
+			if (is_remount)
+				break;
+			if (match_int(&args[0], &option)) {
+				status = 0;
+				goto bail;
+			}
+			if (option >= OCFS2_MIN_RESV_LEVEL &&
+			    option < OCFS2_MAX_RESV_LEVEL)
+				mopt->resv_level = option;
+			break;
+		case Opt_dir_resv_level:
+			if (is_remount)
+				break;
+			if (match_int(&args[0], &option)) {
+				status = 0;
+				goto bail;
+			}
+			if (option >= OCFS2_MIN_RESV_LEVEL &&
+			    option < OCFS2_MAX_RESV_LEVEL)
+				mopt->dir_resv_level = option;
+			break;
 		default:
 			mlog(ML_ERROR,
 			     "Unrecognized mount option \"%s\" "
@@ -1487,7 +1523,7 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
 			   (unsigned) (osb->osb_commit_interval / HZ));
 
 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
-	if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
+	if (local_alloc_megs != ocfs2_la_default_mb(osb))
 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
 
 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
@@ -1514,6 +1550,12 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
 	else
 		seq_printf(s, ",noacl");
 
+	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
+		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
+
+	if (osb->osb_dir_resv_level != osb->osb_resv_level)
+		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
+
 	return 0;
 }
 
@@ -1688,6 +1730,8 @@ static void ocfs2_inode_init_once(void *data)
 	oi->ip_blkno = 0ULL;
 	oi->ip_clusters = 0;
 
+	ocfs2_resv_init_once(&oi->ip_la_data_resv);
+
 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
@@ -2042,6 +2086,12 @@ static int ocfs2_initialize_super(struct super_block *sb,
 
 	init_waitqueue_head(&osb->osb_mount_event);
 
+	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
+	if (status) {
+		mlog_errno(status);
+		goto bail;
+	}
+
 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
 	if (!osb->vol_label) {
 		mlog(ML_ERROR, "unable to alloc vol label\n");
@@ -2224,9 +2274,11 @@ static int ocfs2_initialize_super(struct super_block *sb,
 	}
 
 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
+	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
 	iput(inode);
 
-	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
+	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
+				 osb->s_feature_incompat) * 8;
 
 	status = ocfs2_init_slot_info(osb);
 	if (status < 0) {
@@ -2509,5 +2561,25 @@ void __ocfs2_abort(struct super_block* sb,
 	ocfs2_handle_error(sb);
 }
 
+/*
+ * Void signal blockers, because in-kernel sigprocmask() only fails
+ * when SIG_* is wrong.
+ */
+void ocfs2_block_signals(sigset_t *oldset)
+{
+	int rc;
+	sigset_t blocked;
+
+	sigfillset(&blocked);
+	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
+	BUG_ON(rc);
+}
+
+void ocfs2_unblock_signals(sigset_t *oldset)
+{
+	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
+	BUG_ON(rc);
+}
+
 module_init(ocfs2_init);
 module_exit(ocfs2_exit);

fs/ocfs2/super.h

@@ -45,4 +45,11 @@ void __ocfs2_abort(struct super_block *sb,
 
 #define ocfs2_abort(sb, fmt, args...) __ocfs2_abort(sb, __PRETTY_FUNCTION__, fmt, ##args)
 
+/*
+ * Void signal blockers, because in-kernel sigprocmask() only fails
+ * when SIG_* is wrong.
+ */
+void ocfs2_block_signals(sigset_t *oldset);
+void ocfs2_unblock_signals(sigset_t *oldset);
+
 #endif /* OCFS2_SUPER_H */

fs/ocfs2/xattr.c

@@ -79,6 +79,7 @@ struct ocfs2_xattr_set_ctxt {
 	struct ocfs2_alloc_context *meta_ac;
 	struct ocfs2_alloc_context *data_ac;
 	struct ocfs2_cached_dealloc_ctxt dealloc;
+	int set_abort;
 };
 
 #define OCFS2_XATTR_ROOT_SIZE	(sizeof(struct ocfs2_xattr_def_value_root))
@@ -739,11 +740,7 @@ static int ocfs2_xattr_extend_allocation(struct inode *inode,
 		goto leave;
 	}
 
-	status = ocfs2_journal_dirty(handle, vb->vb_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, vb->vb_bh);
 
 	clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - prev_clusters;
 
@@ -786,12 +783,7 @@ static int __ocfs2_remove_xattr_range(struct inode *inode,
 	}
 
 	le32_add_cpu(&vb->vb_xv->xr_clusters, -len);
-
-	ret = ocfs2_journal_dirty(handle, vb->vb_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, vb->vb_bh);
 
 	if (ext_flags & OCFS2_EXT_REFCOUNTED)
 		ret = ocfs2_decrease_refcount(inode, handle,
@@ -1374,11 +1366,7 @@ static int __ocfs2_xattr_set_value_outside(struct inode *inode,
 				memset(bh->b_data + cp_len, 0,
 				       blocksize - cp_len);
 
-			ret = ocfs2_journal_dirty(handle, bh);
-			if (ret < 0) {
-				mlog_errno(ret);
-				goto out;
-			}
+			ocfs2_journal_dirty(handle, bh);
 			brelse(bh);
 			bh = NULL;
 
@@ -2148,15 +2136,19 @@ alloc_value:
 		orig_clusters = ocfs2_xa_value_clusters(loc);
 		rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt);
 		if (rc < 0) {
-			/*
-			 * If we tried to grow an existing external value,
-			 * ocfs2_xa_cleanuP-value_truncate() is going to
-			 * let it stand.  We have to restore its original
-			 * value size.
-			 */
-			loc->xl_entry->xe_value_size = orig_value_size;
+			ctxt->set_abort = 1;
 			ocfs2_xa_cleanup_value_truncate(loc, "growing",
 							orig_clusters);
+			/*
+			 * If we were growing an existing value,
+			 * ocfs2_xa_cleanup_value_truncate() won't remove
+			 * the entry. We need to restore the original value
+			 * size.
+			 */
+			if (loc->xl_entry) {
+				BUG_ON(!orig_value_size);
+				loc->xl_entry->xe_value_size = orig_value_size;
+			}
 			mlog_errno(rc);
 		}
 	}
@@ -2479,7 +2471,10 @@ static int ocfs2_xattr_free_block(struct inode *inode,
 	xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
 	blk = le64_to_cpu(xb->xb_blkno);
 	bit = le16_to_cpu(xb->xb_suballoc_bit);
-	bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+	if (xb->xb_suballoc_loc)
+		bg_blkno = le64_to_cpu(xb->xb_suballoc_loc);
+	else
+		bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 
 	xb_alloc_inode = ocfs2_get_system_file_inode(osb,
 				EXTENT_ALLOC_SYSTEM_INODE,
@@ -2594,9 +2589,7 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh)
 	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
 	spin_unlock(&oi->ip_lock);
 
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, di_bh);
 out_commit:
 	ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
 out:
@@ -2724,9 +2717,7 @@ static int ocfs2_xattr_ibody_init(struct inode *inode,
 	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
 	spin_unlock(&oi->ip_lock);
 
-	ret = ocfs2_journal_dirty(ctxt->handle, di_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(ctxt->handle, di_bh);
 
 out:
 	return ret;
@@ -2846,9 +2837,8 @@ static int ocfs2_create_xattr_block(struct inode *inode,
 	int ret;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 first_blkno;
+	u64 suballoc_loc, first_blkno;
 	struct ocfs2_dinode *di =  (struct ocfs2_dinode *)inode_bh->b_data;
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct buffer_head *new_bh = NULL;
 	struct ocfs2_xattr_block *xblk;
 
@@ -2859,9 +2849,9 @@ static int ocfs2_create_xattr_block(struct inode *inode,
 		goto end;
 	}
 
-	ret = ocfs2_claim_metadata(osb, ctxt->handle, ctxt->meta_ac, 1,
-				   &suballoc_bit_start, &num_got,
-				   &first_blkno);
+	ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1,
+				   &suballoc_loc, &suballoc_bit_start,
+				   &num_got, &first_blkno);
 	if (ret < 0) {
 		mlog_errno(ret);
 		goto end;
@@ -2883,8 +2873,10 @@ static int ocfs2_create_xattr_block(struct inode *inode,
 	memset(xblk, 0, inode->i_sb->s_blocksize);
 	strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE);
 	xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot);
+	xblk->xb_suballoc_loc = cpu_to_le64(suballoc_loc);
 	xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start);
-	xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation);
+	xblk->xb_fs_generation =
+		cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation);
 	xblk->xb_blkno = cpu_to_le64(first_blkno);
 	if (indexed) {
 		struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root;
@@ -2956,7 +2948,7 @@ static int ocfs2_xattr_block_set(struct inode *inode,
 		ret = ocfs2_xa_set(&loc, xi, ctxt);
 		if (!ret)
 			xs->here = loc.xl_entry;
-		else if (ret != -ENOSPC)
+		else if ((ret != -ENOSPC) || ctxt->set_abort)
 			goto end;
 		else {
 			ret = ocfs2_xattr_create_index_block(inode, xs, ctxt);
@@ -3312,14 +3304,13 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 				goto out;
 			}
 
-			ret = ocfs2_extend_trans(ctxt->handle, credits +
-					ctxt->handle->h_buffer_credits);
+			ret = ocfs2_extend_trans(ctxt->handle, credits);
 			if (ret) {
 				mlog_errno(ret);
 				goto out;
 			}
 			ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
-		} else if (ret == -ENOSPC) {
+		} else if ((ret == -ENOSPC) && !ctxt->set_abort) {
 			if (di->i_xattr_loc && !xbs->xattr_bh) {
 				ret = ocfs2_xattr_block_find(inode,
 							     xi->xi_name_index,
@@ -3343,8 +3334,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 					goto out;
 				}
 
-				ret = ocfs2_extend_trans(ctxt->handle, credits +
-					ctxt->handle->h_buffer_credits);
+				ret = ocfs2_extend_trans(ctxt->handle, credits);
 				if (ret) {
 					mlog_errno(ret);
 					goto out;
@@ -3378,8 +3368,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 					goto out;
 				}
 
-				ret = ocfs2_extend_trans(ctxt->handle, credits +
-						ctxt->handle->h_buffer_credits);
+				ret = ocfs2_extend_trans(ctxt->handle, credits);
 				if (ret) {
 					mlog_errno(ret);
 					goto out;
@@ -4249,7 +4238,6 @@ static int ocfs2_xattr_create_index_block(struct inode *inode,
 	u32 bit_off, len;
 	u64 blkno;
 	handle_t *handle = ctxt->handle;
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
 	struct buffer_head *xb_bh = xs->xattr_bh;
 	struct ocfs2_xattr_block *xb =
@@ -4277,7 +4265,7 @@ static int ocfs2_xattr_create_index_block(struct inode *inode,
 		goto out;
 	}
 
-	ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac,
+	ret = __ocfs2_claim_clusters(handle, ctxt->data_ac,
 				     1, 1, &bit_off, &len);
 	if (ret) {
 		mlog_errno(ret);
@@ -4887,8 +4875,7 @@ static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle,
 	 * We need to update the first bucket of the old extent and all
 	 * the buckets going to the new extent.
 	 */
-	credits = ((num_buckets + 1) * blks_per_bucket) +
-		handle->h_buffer_credits;
+	credits = ((num_buckets + 1) * blks_per_bucket);
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
@@ -4958,7 +4945,7 @@ static int ocfs2_divide_xattr_cluster(struct inode *inode,
 				      u32 *first_hash)
 {
 	u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
-	int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits;
+	int ret, credits = 2 * blk_per_bucket;
 
 	BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize);
 
@@ -5099,7 +5086,7 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode,
 		goto leave;
 	}
 
-	ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1,
+	ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1,
 				     clusters_to_add, &bit_off, &num_bits);
 	if (ret < 0) {
 		if (ret != -ENOSPC)
@@ -5153,9 +5140,7 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode,
 		goto leave;
 	}
 
-	ret = ocfs2_journal_dirty(handle, root_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, root_bh);
 
 leave:
 	return ret;
@@ -5200,8 +5185,7 @@ static int ocfs2_extend_xattr_bucket(struct inode *inode,
 	 * existing bucket.  Then we add the last existing bucket, the
 	 * new bucket, and the first bucket (3 * blk_per_bucket).
 	 */
-	credits = (end_blk - target_blk) + (3 * blk_per_bucket) +
-		  handle->h_buffer_credits;
+	credits = (end_blk - target_blk) + (3 * blk_per_bucket);
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
@@ -5477,12 +5461,7 @@ static int ocfs2_rm_xattr_cluster(struct inode *inode,
 	}
 
 	le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len);
-
-	ret = ocfs2_journal_dirty(handle, root_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, root_bh);
 
 	ret = ocfs2_truncate_log_append(osb, handle, blkno, len);
 	if (ret)
@@ -6935,7 +6914,7 @@ static int ocfs2_reflink_xattr_rec(struct inode *inode,
 		goto out;
 	}
 
-	ret = ocfs2_claim_clusters(osb, handle, data_ac,
+	ret = ocfs2_claim_clusters(handle, data_ac,
 				   len, &p_cluster, &num_clusters);
 	if (ret) {
 		mlog_errno(ret);