ocfs2: Abstract ocfs2_extent_tree in b-tree operations.
authorTao Ma <tao.ma@oracle.com>
Mon, 18 Aug 2008 09:38:44 +0000 (17:38 +0800)
committerMark Fasheh <mfasheh@suse.com>
Mon, 13 Oct 2008 20:57:58 +0000 (13:57 -0700)
In the old extent tree operation, we take the hypothesis that we
are using the ocfs2_extent_list in ocfs2_dinode as the tree root.
As xattr will also use ocfs2_extent_list to store large value
for a xattr entry, we refactor the tree operation so that xattr
can use it directly.

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

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

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
fs/ocfs2/alloc.c
fs/ocfs2/alloc.h
fs/ocfs2/aops.c
fs/ocfs2/dir.c
fs/ocfs2/file.c
fs/ocfs2/file.h
fs/ocfs2/suballoc.c
fs/ocfs2/suballoc.h

index dc36cd140754c6979bd4e4ddcfeb52da473361d2..579659bae6c5b64486eae170b77be538c28abb26 100644 (file)
 
 #include "buffer_head_io.h"
 
+/*
+ * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract
+ * the b-tree operations in ocfs2. Now all the b-tree operations are not
+ * limited to ocfs2_dinode only. Any data which need to allocate clusters
+ * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree
+ * and operation.
+ *
+ * ocfs2_extent_tree contains info for the root of the b-tree, it must have a
+ * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree
+ * functions.
+ * ocfs2_extent_tree_operations abstract the normal operations we do for
+ * the root of extent b-tree.
+ */
+struct ocfs2_extent_tree;
+
+struct ocfs2_extent_tree_operations {
+       void (*set_last_eb_blk) (struct ocfs2_extent_tree *et, u64 blkno);
+       u64 (*get_last_eb_blk) (struct ocfs2_extent_tree *et);
+       void (*update_clusters) (struct inode *inode,
+                                struct ocfs2_extent_tree *et,
+                                u32 new_clusters);
+       int (*sanity_check) (struct inode *inode, struct ocfs2_extent_tree *et);
+};
+
+struct ocfs2_extent_tree {
+       enum ocfs2_extent_tree_type type;
+       struct ocfs2_extent_tree_operations *eops;
+       struct buffer_head *root_bh;
+       struct ocfs2_extent_list *root_el;
+};
+
+static void ocfs2_dinode_set_last_eb_blk(struct ocfs2_extent_tree *et,
+                                        u64 blkno)
+{
+       struct ocfs2_dinode *di = (struct ocfs2_dinode *)et->root_bh->b_data;
+
+       BUG_ON(et->type != OCFS2_DINODE_EXTENT);
+       di->i_last_eb_blk = cpu_to_le64(blkno);
+}
+
+static u64 ocfs2_dinode_get_last_eb_blk(struct ocfs2_extent_tree *et)
+{
+       struct ocfs2_dinode *di = (struct ocfs2_dinode *)et->root_bh->b_data;
+
+       BUG_ON(et->type != OCFS2_DINODE_EXTENT);
+       return le64_to_cpu(di->i_last_eb_blk);
+}
+
+static void ocfs2_dinode_update_clusters(struct inode *inode,
+                                        struct ocfs2_extent_tree *et,
+                                        u32 clusters)
+{
+       struct ocfs2_dinode *di =
+                       (struct ocfs2_dinode *)et->root_bh->b_data;
+
+       le32_add_cpu(&di->i_clusters, clusters);
+       spin_lock(&OCFS2_I(inode)->ip_lock);
+       OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters);
+       spin_unlock(&OCFS2_I(inode)->ip_lock);
+}
+
+static int ocfs2_dinode_sanity_check(struct inode *inode,
+                                    struct ocfs2_extent_tree *et)
+{
+       int ret = 0;
+       struct ocfs2_dinode *di;
+
+       BUG_ON(et->type != OCFS2_DINODE_EXTENT);
+
+       di = (struct ocfs2_dinode *)et->root_bh->b_data;
+       if (!OCFS2_IS_VALID_DINODE(di)) {
+               ret = -EIO;
+               ocfs2_error(inode->i_sb,
+                       "Inode %llu has invalid path root",
+                       (unsigned long long)OCFS2_I(inode)->ip_blkno);
+       }
+
+       return ret;
+}
+
+static struct ocfs2_extent_tree_operations ocfs2_dinode_et_ops = {
+       .set_last_eb_blk        = ocfs2_dinode_set_last_eb_blk,
+       .get_last_eb_blk        = ocfs2_dinode_get_last_eb_blk,
+       .update_clusters        = ocfs2_dinode_update_clusters,
+       .sanity_check           = ocfs2_dinode_sanity_check,
+};
+
+static struct ocfs2_extent_tree*
+        ocfs2_new_extent_tree(struct buffer_head *bh,
+                              enum ocfs2_extent_tree_type et_type)
+{
+       struct ocfs2_extent_tree *et;
+
+       et = kzalloc(sizeof(*et), GFP_NOFS);
+       if (!et)
+               return NULL;
+
+       et->type = et_type;
+       get_bh(bh);
+       et->root_bh = bh;
+
+       /* current we only support dinode extent. */
+       BUG_ON(et->type != OCFS2_DINODE_EXTENT);
+       if (et_type == OCFS2_DINODE_EXTENT) {
+               et->root_el = &((struct ocfs2_dinode *)bh->b_data)->id2.i_list;
+               et->eops = &ocfs2_dinode_et_ops;
+       }
+
+       return et;
+}
+
+static void ocfs2_free_extent_tree(struct ocfs2_extent_tree *et)
+{
+       if (et) {
+               brelse(et->root_bh);
+               kfree(et);
+       }
+}
+
+static inline void ocfs2_set_last_eb_blk(struct ocfs2_extent_tree *et,
+                                        u64 new_last_eb_blk)
+{
+       et->eops->set_last_eb_blk(et, new_last_eb_blk);
+}
+
+static inline u64 ocfs2_get_last_eb_blk(struct ocfs2_extent_tree *et)
+{
+       return et->eops->get_last_eb_blk(et);
+}
+
+static inline void ocfs2_update_clusters(struct inode *inode,
+                                        struct ocfs2_extent_tree *et,
+                                        u32 clusters)
+{
+       et->eops->update_clusters(inode, et, clusters);
+}
+
 static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
 static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
                                         struct ocfs2_extent_block *eb);
@@ -204,17 +341,6 @@ static struct ocfs2_path *ocfs2_new_path(struct buffer_head *root_bh,
        return path;
 }
 
-/*
- * Allocate and initialize a new path based on a disk inode tree.
- */
-static struct ocfs2_path *ocfs2_new_inode_path(struct buffer_head *di_bh)
-{
-       struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
-       struct ocfs2_extent_list *el = &di->id2.i_list;
-
-       return ocfs2_new_path(di_bh, el);
-}
-
 /*
  * Convenience function to journal all components in a path.
  */
@@ -368,24 +494,33 @@ struct ocfs2_merge_ctxt {
  */
 int ocfs2_num_free_extents(struct ocfs2_super *osb,
                           struct inode *inode,
-                          struct buffer_head *bh)
+                          struct buffer_head *root_bh,
+                          enum ocfs2_extent_tree_type type)
 {
        int retval;
-       struct ocfs2_extent_list *el;
+       struct ocfs2_extent_list *el = NULL;
        struct ocfs2_extent_block *eb;
        struct buffer_head *eb_bh = NULL;
-       struct ocfs2_dinode *fe = (struct ocfs2_dinode *)bh->b_data;
+       u64 last_eb_blk = 0;
 
        mlog_entry_void();
 
-       if (!OCFS2_IS_VALID_DINODE(fe)) {
-               OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
-               retval = -EIO;
-               goto bail;
+       if (type == OCFS2_DINODE_EXTENT) {
+               struct ocfs2_dinode *fe =
+                               (struct ocfs2_dinode *)root_bh->b_data;
+               if (!OCFS2_IS_VALID_DINODE(fe)) {
+                       OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
+                       retval = -EIO;
+                       goto bail;
+               }
+
+               if (fe->i_last_eb_blk)
+                       last_eb_blk = le64_to_cpu(fe->i_last_eb_blk);
+               el = &fe->id2.i_list;
        }
 
-       if (fe->i_last_eb_blk) {
-               retval = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
+       if (last_eb_blk) {
+               retval = ocfs2_read_block(osb, last_eb_blk,
                                          &eb_bh, OCFS2_BH_CACHED, inode);
                if (retval < 0) {
                        mlog_errno(retval);
@@ -393,8 +528,7 @@ int ocfs2_num_free_extents(struct ocfs2_super *osb,
                }
                eb = (struct ocfs2_extent_block *) eb_bh->b_data;
                el = &eb->h_list;
-       } else
-               el = &fe->id2.i_list;
+       }
 
        BUG_ON(el->l_tree_depth != 0);
 
@@ -532,7 +666,7 @@ static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list  *el)
 static int ocfs2_add_branch(struct ocfs2_super *osb,
                            handle_t *handle,
                            struct inode *inode,
-                           struct buffer_head *fe_bh,
+                           struct ocfs2_extent_tree *et,
                            struct buffer_head *eb_bh,
                            struct buffer_head **last_eb_bh,
                            struct ocfs2_alloc_context *meta_ac)
@@ -541,7 +675,6 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
        u64 next_blkno, new_last_eb_blk;
        struct buffer_head *bh;
        struct buffer_head **new_eb_bhs = NULL;
-       struct ocfs2_dinode *fe;
        struct ocfs2_extent_block *eb;
        struct ocfs2_extent_list  *eb_el;
        struct ocfs2_extent_list  *el;
@@ -551,13 +684,11 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
 
        BUG_ON(!last_eb_bh || !*last_eb_bh);
 
-       fe = (struct ocfs2_dinode *) fe_bh->b_data;
-
        if (eb_bh) {
                eb = (struct ocfs2_extent_block *) eb_bh->b_data;
                el = &eb->h_list;
        } else
-               el = &fe->id2.i_list;
+               el = et->root_el;
 
        /* we never add a branch to a leaf. */
        BUG_ON(!el->l_tree_depth);
@@ -647,7 +778,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
                mlog_errno(status);
                goto bail;
        }
-       status = ocfs2_journal_access(handle, inode, fe_bh,
+       status = ocfs2_journal_access(handle, inode, et->root_bh,
                                      OCFS2_JOURNAL_ACCESS_WRITE);
        if (status < 0) {
                mlog_errno(status);
@@ -663,7 +794,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
        }
 
        /* Link the new branch into the rest of the tree (el will
-        * either be on the fe, or the extent block passed in. */
+        * either be on the root_bh, or the extent block passed in. */
        i = le16_to_cpu(el->l_next_free_rec);
        el->l_recs[i].e_blkno = cpu_to_le64(next_blkno);
        el->l_recs[i].e_cpos = cpu_to_le32(new_cpos);
@@ -672,7 +803,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
 
        /* fe needs a new last extent block pointer, as does the
         * next_leaf on the previously last-extent-block. */
-       fe->i_last_eb_blk = cpu_to_le64(new_last_eb_blk);
+       ocfs2_set_last_eb_blk(et, new_last_eb_blk);
 
        eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
        eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);
@@ -680,7 +811,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb,
        status = ocfs2_journal_dirty(handle, *last_eb_bh);
        if (status < 0)
                mlog_errno(status);
-       status = ocfs2_journal_dirty(handle, fe_bh);
+       status = ocfs2_journal_dirty(handle, et->root_bh);
        if (status < 0)
                mlog_errno(status);
        if (eb_bh) {
@@ -718,16 +849,15 @@ bail:
 static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
                                  handle_t *handle,
                                  struct inode *inode,
-                                 struct buffer_head *fe_bh,
+                                 struct ocfs2_extent_tree *et,
                                  struct ocfs2_alloc_context *meta_ac,
                                  struct buffer_head **ret_new_eb_bh)
 {
        int status, i;
        u32 new_clusters;
        struct buffer_head *new_eb_bh = NULL;
-       struct ocfs2_dinode *fe;
        struct ocfs2_extent_block *eb;
-       struct ocfs2_extent_list  *fe_el;
+       struct ocfs2_extent_list  *root_el;
        struct ocfs2_extent_list  *eb_el;
 
        mlog_entry_void();
@@ -747,8 +877,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
        }
 
        eb_el = &eb->h_list;
-       fe = (struct ocfs2_dinode *) fe_bh->b_data;
-       fe_el = &fe->id2.i_list;
+       root_el = et->root_el;
 
        status = ocfs2_journal_access(handle, inode, new_eb_bh,
                                      OCFS2_JOURNAL_ACCESS_CREATE);
@@ -757,11 +886,11 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
                goto bail;
        }
 
-       /* copy the fe data into the new extent block */
-       eb_el->l_tree_depth = fe_el->l_tree_depth;
-       eb_el->l_next_free_rec = fe_el->l_next_free_rec;
-       for(i = 0; i < le16_to_cpu(fe_el->l_next_free_rec); i++)
-               eb_el->l_recs[i] = fe_el->l_recs[i];
+       /* copy the root extent list data into the new extent block */
+       eb_el->l_tree_depth = root_el->l_tree_depth;
+       eb_el->l_next_free_rec = root_el->l_next_free_rec;
+       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) {
@@ -769,7 +898,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
                goto bail;
        }
 
-       status = ocfs2_journal_access(handle, inode, fe_bh,
+       status = ocfs2_journal_access(handle, inode, et->root_bh,
                                      OCFS2_JOURNAL_ACCESS_WRITE);
        if (status < 0) {
                mlog_errno(status);
@@ -778,21 +907,21 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
 
        new_clusters = ocfs2_sum_rightmost_rec(eb_el);
 
-       /* update fe now */
-       le16_add_cpu(&fe_el->l_tree_depth, 1);
-       fe_el->l_recs[0].e_cpos = 0;
-       fe_el->l_recs[0].e_blkno = eb->h_blkno;
-       fe_el->l_recs[0].e_int_clusters = cpu_to_le32(new_clusters);
-       for(i = 1; i < le16_to_cpu(fe_el->l_next_free_rec); i++)
-               memset(&fe_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec));
-       fe_el->l_next_free_rec = cpu_to_le16(1);
+       /* update root_bh now */
+       le16_add_cpu(&root_el->l_tree_depth, 1);
+       root_el->l_recs[0].e_cpos = 0;
+       root_el->l_recs[0].e_blkno = eb->h_blkno;
+       root_el->l_recs[0].e_int_clusters = cpu_to_le32(new_clusters);
+       for (i = 1; i < le16_to_cpu(root_el->l_next_free_rec); i++)
+               memset(&root_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec));
+       root_el->l_next_free_rec = cpu_to_le16(1);
 
        /* If this is our 1st tree depth shift, then last_eb_blk
         * becomes the allocated extent block */
-       if (fe_el->l_tree_depth == cpu_to_le16(1))
-               fe->i_last_eb_blk = eb->h_blkno;
+       if (root_el->l_tree_depth == cpu_to_le16(1))
+               ocfs2_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
 
-       status = ocfs2_journal_dirty(handle, fe_bh);
+       status = ocfs2_journal_dirty(handle, et->root_bh);
        if (status < 0) {
                mlog_errno(status);
                goto bail;
@@ -818,22 +947,21 @@ bail:
  * 1) a lowest extent block is found, then we pass it back in
  *    *lowest_eb_bh and return '0'
  *
- * 2) the search fails to find anything, but the dinode has room. We
+ * 2) the search fails to find anything, but the root_el has room. We
  *    pass NULL back in *lowest_eb_bh, but still return '0'
  *
- * 3) the search fails to find anything AND the dinode is full, in
+ * 3) the search fails to find anything AND the root_el is full, in
  *    which case we return > 0
  *
  * return status < 0 indicates an error.
  */
 static int ocfs2_find_branch_target(struct ocfs2_super *osb,
                                    struct inode *inode,
-                                   struct buffer_head *fe_bh,
+                                   struct ocfs2_extent_tree *et,
                                    struct buffer_head **target_bh)
 {
        int status = 0, i;
        u64 blkno;
-       struct ocfs2_dinode *fe;
        struct ocfs2_extent_block *eb;
        struct ocfs2_extent_list  *el;
        struct buffer_head *bh = NULL;
@@ -843,8 +971,7 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb,
 
        *target_bh = NULL;
 
-       fe = (struct ocfs2_dinode *) fe_bh->b_data;
-       el = &fe->id2.i_list;
+       el = et->root_el;
 
        while(le16_to_cpu(el->l_tree_depth) > 1) {
                if (le16_to_cpu(el->l_next_free_rec) == 0) {
@@ -896,8 +1023,8 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb,
 
        /* If we didn't find one and the fe doesn't have any room,
         * then return '1' */
-       if (!lowest_bh
-           && (fe->id2.i_list.l_next_free_rec == fe->id2.i_list.l_count))
+       el = et->root_el;
+       if (!lowest_bh && (el->l_next_free_rec == el->l_count))
                status = 1;
 
        *target_bh = lowest_bh;
@@ -920,19 +1047,19 @@ bail:
  * *last_eb_bh will be updated by ocfs2_add_branch().
  */
 static int ocfs2_grow_tree(struct inode *inode, handle_t *handle,
-                          struct buffer_head *di_bh, int *final_depth,
+                          struct ocfs2_extent_tree *et, int *final_depth,
                           struct buffer_head **last_eb_bh,
                           struct ocfs2_alloc_context *meta_ac)
 {
        int ret, shift;
-       struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
-       int depth = le16_to_cpu(di->id2.i_list.l_tree_depth);
+       struct ocfs2_extent_list *el = et->root_el;
+       int depth = le16_to_cpu(el->l_tree_depth);
        struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
        struct buffer_head *bh = NULL;
 
        BUG_ON(meta_ac == NULL);
 
-       shift = ocfs2_find_branch_target(osb, inode, di_bh, &bh);
+       shift = ocfs2_find_branch_target(osb, inode, et, &bh);
        if (shift < 0) {
                ret = shift;
                mlog_errno(ret);
@@ -949,7 +1076,7 @@ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle,
                /* ocfs2_shift_tree_depth will return us a buffer with
                 * the new extent block (so we can pass that to
                 * ocfs2_add_branch). */
-               ret = ocfs2_shift_tree_depth(osb, handle, inode, di_bh,
+               ret = ocfs2_shift_tree_depth(osb, handle, inode, et,
                                             meta_ac, &bh);
                if (ret < 0) {
                        mlog_errno(ret);
@@ -976,7 +1103,7 @@ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle,
        /* call ocfs2_add_branch to add the final part of the tree with
         * the new data. */
        mlog(0, "add branch. bh = %p\n", bh);
-       ret = ocfs2_add_branch(osb, handle, inode, di_bh, bh, last_eb_bh,
+       ret = ocfs2_add_branch(osb, handle, inode, et, bh, last_eb_bh,
                               meta_ac);
        if (ret < 0) {
                mlog_errno(ret);
@@ -2059,11 +2186,11 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle,
                                     struct ocfs2_path *right_path,
                                     int subtree_index,
                                     struct ocfs2_cached_dealloc_ctxt *dealloc,
-                                    int *deleted)
+                                    int *deleted,
+                                    struct ocfs2_extent_tree *et)
 {
        int ret, i, del_right_subtree = 0, right_has_empty = 0;
-       struct buffer_head *root_bh, *di_bh = path_root_bh(right_path);
-       struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+       struct buffer_head *root_bh, *et_root_bh = path_root_bh(right_path);
        struct ocfs2_extent_list *right_leaf_el, *left_leaf_el;
        struct ocfs2_extent_block *eb;
 
@@ -2115,7 +2242,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle,
                 * We have to update i_last_eb_blk during the meta
                 * data delete.
                 */
-               ret = ocfs2_journal_access(handle, inode, di_bh,
+               ret = ocfs2_journal_access(handle, inode, et_root_bh,
                                           OCFS2_JOURNAL_ACCESS_WRITE);
                if (ret) {
                        mlog_errno(ret);
@@ -2190,7 +2317,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle,
                ocfs2_update_edge_lengths(inode, handle, left_path);
 
                eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
-               di->i_last_eb_blk = eb->h_blkno;
+               ocfs2_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
 
                /*
                 * Removal of the extent in the left leaf was skipped
@@ -2200,7 +2327,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle,
                if (right_has_empty)
                        ocfs2_remove_empty_extent(left_leaf_el);
 
-               ret = ocfs2_journal_dirty(handle, di_bh);
+               ret = ocfs2_journal_dirty(handle, et_root_bh);
                if (ret)
                        mlog_errno(ret);
 
@@ -2323,7 +2450,8 @@ static int __ocfs2_rotate_tree_left(struct inode *inode,
                                    handle_t *handle, int orig_credits,
                                    struct ocfs2_path *path,
                                    struct ocfs2_cached_dealloc_ctxt *dealloc,
-                                   struct ocfs2_path **empty_extent_path)
+                                   struct ocfs2_path **empty_extent_path,
+                                   struct ocfs2_extent_tree *et)
 {
        int ret, subtree_root, deleted;
        u32 right_cpos;
@@ -2396,7 +2524,7 @@ static int __ocfs2_rotate_tree_left(struct inode *inode,
 
                ret = ocfs2_rotate_subtree_left(inode, handle, left_path,
                                                right_path, subtree_root,
-                                               dealloc, &deleted);
+                                               dealloc, &deleted, et);
                if (ret == -EAGAIN) {
                        /*
                         * The rotation has to temporarily stop due to
@@ -2439,29 +2567,20 @@ out:
 }
 
 static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle,
-                                      struct ocfs2_path *path,
-                                      struct ocfs2_cached_dealloc_ctxt *dealloc)
+                               struct ocfs2_path *path,
+                               struct ocfs2_cached_dealloc_ctxt *dealloc,
+                               struct ocfs2_extent_tree *et)
 {
        int ret, subtree_index;
        u32 cpos;
        struct ocfs2_path *left_path = NULL;
-       struct ocfs2_dinode *di;
        struct ocfs2_extent_block *eb;
        struct ocfs2_extent_list *el;
 
-       /*
-        * XXX: This code assumes that the root is an inode, which is
-        * true for now but may change as tree code gets generic.
-        */
-       di = (struct ocfs2_dinode *)path_root_bh(path)->b_data;
-       if (!OCFS2_IS_VALID_DINODE(di)) {
-               ret = -EIO;
-               ocfs2_error(inode->i_sb,
-                           "Inode %llu has invalid path root",
-                           (unsigned long long)OCFS2_I(inode)->ip_blkno);
-               goto out;
-       }
 
+       ret = et->eops->sanity_check(inode, et);
+       if (ret)
+               goto out;
        /*
         * There's two ways we handle this depending on
         * whether path is the only existing one.
@@ -2518,7 +2637,7 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle,
                ocfs2_update_edge_lengths(inode, handle, left_path);
 
                eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
-               di->i_last_eb_blk = eb->h_blkno;
+               ocfs2_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
        } else {
                /*
                 * 'path' is also the leftmost path which
@@ -2529,12 +2648,12 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle,
                 */
                ocfs2_unlink_path(inode, handle, dealloc, path, 1);
 
-               el = &di->id2.i_list;
+               el = et->root_el;
                el->l_tree_depth = 0;
                el->l_next_free_rec = 0;
                memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
 
-               di->i_last_eb_blk = 0;
+               ocfs2_set_last_eb_blk(et, 0);
        }
 
        ocfs2_journal_dirty(handle, path_root_bh(path));
@@ -2562,7 +2681,8 @@ out:
  */
 static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle,
                                  struct ocfs2_path *path,
-                                 struct ocfs2_cached_dealloc_ctxt *dealloc)
+                                 struct ocfs2_cached_dealloc_ctxt *dealloc,
+                                 struct ocfs2_extent_tree *et)
 {
        int ret, orig_credits = handle->h_buffer_credits;
        struct ocfs2_path *tmp_path = NULL, *restart_path = NULL;
@@ -2576,7 +2696,7 @@ static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle,
        if (path->p_tree_depth == 0) {
 rightmost_no_delete:
                /*
-                * In-inode extents. This is trivially handled, so do
+                * Inline extents. This is trivially handled, so do
                 * it up front.
                 */
                ret = ocfs2_rotate_rightmost_leaf_left(inode, handle,
@@ -2630,7 +2750,7 @@ rightmost_no_delete:
                 */
 
                ret = ocfs2_remove_rightmost_path(inode, handle, path,
-                                                 dealloc);
+                                                 dealloc, et);
                if (ret)
                        mlog_errno(ret);
                goto out;
@@ -2642,7 +2762,7 @@ rightmost_no_delete:
         */
 try_rotate:
        ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, path,
-                                      dealloc, &restart_path);
+                                      dealloc, &restart_path, et);
        if (ret && ret != -EAGAIN) {
                mlog_errno(ret);
                goto out;
@@ -2654,7 +2774,7 @@ try_rotate:
 
                ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits,
                                               tmp_path, dealloc,
-                                              &restart_path);
+                                              &restart_path, et);
                if (ret && ret != -EAGAIN) {
                        mlog_errno(ret);
                        goto out;
@@ -2940,6 +3060,7 @@ static int ocfs2_merge_rec_left(struct inode *inode,
                                handle_t *handle,
                                struct ocfs2_extent_rec *split_rec,
                                struct ocfs2_cached_dealloc_ctxt *dealloc,
+                               struct ocfs2_extent_tree *et,
                                int index)
 {
        int ret, i, subtree_index = 0, has_empty_extent = 0;
@@ -3060,7 +3181,8 @@ static int ocfs2_merge_rec_left(struct inode *inode,
                    le16_to_cpu(el->l_next_free_rec) == 1) {
 
                        ret = ocfs2_remove_rightmost_path(inode, handle,
-                                                         right_path, dealloc);
+                                                         right_path,
+                                                         dealloc, et);
                        if (ret) {
                                mlog_errno(ret);
                                goto out;
@@ -3087,7 +3209,8 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                                     int split_index,
                                     struct ocfs2_extent_rec *split_rec,
                                     struct ocfs2_cached_dealloc_ctxt *dealloc,
-                                    struct ocfs2_merge_ctxt *ctxt)
+                                    struct ocfs2_merge_ctxt *ctxt,
+                                    struct ocfs2_extent_tree *et)
 
 {
        int ret = 0;
@@ -3105,7 +3228,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                 * illegal.
                 */
                ret = ocfs2_rotate_tree_left(inode, handle, path,
-                                            dealloc);
+                                            dealloc, et);
                if (ret) {
                        mlog_errno(ret);
                        goto out;
@@ -3148,7 +3271,8 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0]));
 
                /* The merge left us with an empty extent, remove it. */
-               ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc);
+               ret = ocfs2_rotate_tree_left(inode, handle, path,
+                                            dealloc, et);
                if (ret) {
                        mlog_errno(ret);
                        goto out;
@@ -3162,7 +3286,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                 */
                ret = ocfs2_merge_rec_left(inode, path,
                                           handle, rec,
-                                          dealloc,
+                                          dealloc, et,
                                           split_index);
 
                if (ret) {
@@ -3171,7 +3295,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                }
 
                ret = ocfs2_rotate_tree_left(inode, handle, path,
-                                            dealloc);
+                                            dealloc, et);
                /*
                 * Error from this last rotate is not critical, so
                 * print but don't bubble it up.
@@ -3191,7 +3315,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                        ret = ocfs2_merge_rec_left(inode,
                                                   path,
                                                   handle, split_rec,
-                                                  dealloc,
+                                                  dealloc, et,
                                                   split_index);
                        if (ret) {
                                mlog_errno(ret);
@@ -3214,7 +3338,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode,
                         * our leaf. Try to rotate it away.
                         */
                        ret = ocfs2_rotate_tree_left(inode, handle, path,
-                                                    dealloc);
+                                                    dealloc, et);
                        if (ret)
                                mlog_errno(ret);
                        ret = 0;
@@ -3348,16 +3472,6 @@ rotate:
        ocfs2_rotate_leaf(el, insert_rec);
 }
 
-static inline void ocfs2_update_dinode_clusters(struct inode *inode,
-                                               struct ocfs2_dinode *di,
-                                               u32 clusters)
-{
-       le32_add_cpu(&di->i_clusters, clusters);
-       spin_lock(&OCFS2_I(inode)->ip_lock);
-       OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters);
-       spin_unlock(&OCFS2_I(inode)->ip_lock);
-}
-
 static void ocfs2_adjust_rightmost_records(struct inode *inode,
                                           handle_t *handle,
                                           struct ocfs2_path *path,
@@ -3559,8 +3673,8 @@ static void ocfs2_split_record(struct inode *inode,
 }
 
 /*
- * This function only does inserts on an allocation b-tree. For dinode
- * lists, ocfs2_insert_at_leaf() is called directly.
+ * This function only does inserts on an allocation b-tree. For tree
+ * depth = 0, ocfs2_insert_at_leaf() is called directly.
  *
  * right_path is the path we want to do the actual insert
  * in. left_path should only be passed in if we need to update that
@@ -3657,7 +3771,7 @@ out:
 
 static int ocfs2_do_insert_extent(struct inode *inode,
                                  handle_t *handle,
-                                 struct buffer_head *di_bh,
+                                 struct ocfs2_extent_tree *et,
                                  struct ocfs2_extent_rec *insert_rec,
                                  struct ocfs2_insert_type *type)
 {
@@ -3665,13 +3779,11 @@ static int ocfs2_do_insert_extent(struct inode *inode,
        u32 cpos;
        struct ocfs2_path *right_path = NULL;
        struct ocfs2_path *left_path = NULL;
-       struct ocfs2_dinode *di;
        struct ocfs2_extent_list *el;
 
-       di = (struct ocfs2_dinode *) di_bh->b_data;
-       el = &di->id2.i_list;
+       el = et->root_el;
 
-       ret = ocfs2_journal_access(handle, inode, di_bh,
+       ret = ocfs2_journal_access(handle, inode, et->root_bh,
                                   OCFS2_JOURNAL_ACCESS_WRITE);
        if (ret) {
                mlog_errno(ret);
@@ -3683,7 +3795,7 @@ static int ocfs2_do_insert_extent(struct inode *inode,
                goto out_update_clusters;
        }
 
-       right_path = ocfs2_new_inode_path(di_bh);
+       right_path = ocfs2_new_path(et->root_bh, et->root_el);
        if (!right_path) {
                ret = -ENOMEM;
                mlog_errno(ret);
@@ -3733,7 +3845,7 @@ static int ocfs2_do_insert_extent(struct inode *inode,
                 * ocfs2_rotate_tree_right() might have extended the
                 * transaction without re-journaling our tree root.
                 */
-               ret = ocfs2_journal_access(handle, inode, di_bh,
+               ret = ocfs2_journal_access(handle, inode, et->root_bh,
                                           OCFS2_JOURNAL_ACCESS_WRITE);
                if (ret) {
                        mlog_errno(ret);
@@ -3758,10 +3870,10 @@ static int ocfs2_do_insert_extent(struct inode *inode,
 
 out_update_clusters:
        if (type->ins_split == SPLIT_NONE)
-               ocfs2_update_dinode_clusters(inode, di,
-                                            le16_to_cpu(insert_rec->e_leaf_clusters));
+               ocfs2_update_clusters(inode, et,
+                                     le16_to_cpu(insert_rec->e_leaf_clusters));
 
-       ret = ocfs2_journal_dirty(handle, di_bh);
+       ret = ocfs2_journal_dirty(handle, et->root_bh);
        if (ret)
                mlog_errno(ret);
 
@@ -3915,8 +4027,8 @@ static void ocfs2_figure_contig_type(struct inode *inode,
  * ocfs2_figure_appending_type() will figure out whether we'll have to
  * insert at the tail of the rightmost leaf.
  *
- * This should also work against the dinode list for tree's with 0
- * depth. If we consider the dinode list to be the rightmost leaf node
+ * This should also work against the root extent list for tree's with 0
+ * depth. If we consider the root extent list to be the rightmost leaf node
  * then the logic here makes sense.
  */
 static void ocfs2_figure_appending_type(struct ocfs2_insert_type *insert,
@@ -3967,14 +4079,13 @@ set_tail_append:
  * structure.
  */
 static int ocfs2_figure_insert_type(struct inode *inode,
-                                   struct buffer_head *di_bh,
+                                   struct ocfs2_extent_tree *et,
                                    struct buffer_head **last_eb_bh,
                                    struct ocfs2_extent_rec *insert_rec,
                                    int *free_records,
                                    struct ocfs2_insert_type *insert)
 {
        int ret;
-       struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
        struct ocfs2_extent_block *eb;
        struct ocfs2_extent_list *el;
        struct ocfs2_path *path = NULL;
@@ -3982,7 +4093,7 @@ static int ocfs2_figure_insert_type(struct inode *inode,
 
        insert->ins_split = SPLIT_NONE;
 
-       el = &di->id2.i_list;
+       el = et->root_el;
        insert->ins_tree_depth = le16_to_cpu(el->l_tree_depth);
 
        if (el->l_tree_depth) {
@@ -3993,7 +4104,7 @@ static int ocfs2_figure_insert_type(struct inode *inode,
                 * may want it later.
                 */
                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
-                                      le64_to_cpu(di->i_last_eb_blk), &bh,
+                                      ocfs2_get_last_eb_blk(et), &bh,
                                       OCFS2_BH_CACHED, inode);
                if (ret) {
                        mlog_exit(ret);
@@ -4020,7 +4131,7 @@ static int ocfs2_figure_insert_type(struct inode *inode,
                return 0;
        }
 
-       path = ocfs2_new_inode_path(di_bh);
+       path = ocfs2_new_path(et->root_bh, et->root_el);
        if (!path) {
                ret = -ENOMEM;
                mlog_errno(ret);
@@ -4070,7 +4181,8 @@ static int ocfs2_figure_insert_type(struct inode *inode,
         * the case that we're doing a tail append, so maybe we can
         * take advantage of that information somehow.
         */
-       if (le64_to_cpu(di->i_last_eb_blk) == path_leaf_bh(path)->b_blocknr) {
+       if (ocfs2_get_last_eb_blk(et) ==
+           path_leaf_bh(path)->b_blocknr) {
                /*
                 * Ok, ocfs2_find_path() returned us the rightmost
                 * tree path. This might be an appending insert. There are
@@ -4100,21 +4212,30 @@ out:
 int ocfs2_insert_extent(struct ocfs2_super *osb,
                        handle_t *handle,
                        struct inode *inode,
-                       struct buffer_head *fe_bh,
+                       struct buffer_head *root_bh,
                        u32 cpos,
                        u64 start_blk,
                        u32 new_clusters,
                        u8 flags,
-                       struct ocfs2_alloc_context *meta_ac)
+                       struct ocfs2_alloc_context *meta_ac,
+                       enum ocfs2_extent_tree_type et_type)
 {
        int status;
        int uninitialized_var(free_records);
        struct buffer_head *last_eb_bh = NULL;
        struct ocfs2_insert_type insert = {0, };
        struct ocfs2_extent_rec rec;
+       struct ocfs2_extent_tree *et = NULL;
 
        BUG_ON(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL);
 
+       et = ocfs2_new_extent_tree(root_bh, et_type);
+       if (!et) {
+               status = -ENOMEM;
+               mlog_errno(status);
+               goto bail;
+       }
+
        mlog(0, "add %u clusters at position %u to inode %llu\n",
             new_clusters, cpos, (unsigned long long)OCFS2_I(inode)->ip_blkno);
 
@@ -4132,7 +4253,7 @@ int ocfs2_insert_extent(struct ocfs2_super *osb,
        rec.e_leaf_clusters = cpu_to_le16(new_clusters);
        rec.e_flags = flags;
 
-       status = ocfs2_figure_insert_type(inode, fe_bh, &last_eb_bh, &rec,
+       status = ocfs2_figure_insert_type(inode, et, &last_eb_bh, &rec,
                                          &free_records, &insert);
        if (status < 0) {
                mlog_errno(status);
@@ -4146,7 +4267,7 @@ int ocfs2_insert_extent(struct ocfs2_super *osb,
             free_records, insert.ins_tree_depth);
 
        if (insert.ins_contig == CONTIG_NONE && free_records == 0) {
-               status = ocfs2_grow_tree(inode, handle, fe_bh,
+               status = ocfs2_grow_tree(inode, handle, et,
                                         &insert.ins_tree_depth, &last_eb_bh,
                                         meta_ac);
                if (status) {
@@ -4156,16 +4277,18 @@ int ocfs2_insert_extent(struct ocfs2_super *osb,
        }
 
        /* Finally, we can add clusters. This might rotate the tree for us. */
-       status = ocfs2_do_insert_extent(inode, handle, fe_bh, &rec, &insert);
+       status = ocfs2_do_insert_extent(inode, handle, et, &rec, &insert);
        if (status < 0)
                mlog_errno(status);
-       else
+       else if (et->type == OCFS2_DINODE_EXTENT)
                ocfs2_extent_map_insert_rec(inode, &rec);
 
 bail:
        if (last_eb_bh)
                brelse(last_eb_bh);
 
+       if (et)
+               ocfs2_free_extent_tree(et);
        mlog_exit(status);
        return status;
 }
@@ -4193,7 +4316,7 @@ static void ocfs2_make_right_split_rec(struct super_block *sb,
 static int ocfs2_split_and_insert(struct inode *inode,
                                  handle_t *handle,
                                  struct ocfs2_path *path,
-                                 struct buffer_head *di_bh,
+                                 struct ocfs2_extent_tree *et,
                                  struct buffer_head **last_eb_bh,
                                  int split_index,
                                  struct ocfs2_extent_rec *orig_split_rec,
@@ -4207,7 +4330,6 @@ static int ocfs2_split_and_insert(struct inode *inode,
        struct ocfs2_extent_rec split_rec = *orig_split_rec;
        struct ocfs2_insert_type insert;
        struct ocfs2_extent_block *eb;
-       struct ocfs2_dinode *di;
 
 leftright:
        /*
@@ -4216,8 +4338,7 @@ leftright:
         */
        rec = path_leaf_el(path)->l_recs[split_index];
 
-       di = (struct ocfs2_dinode *)di_bh->b_data;
-       rightmost_el = &di->id2.i_list;
+       rightmost_el = et->root_el;
 
        depth = le16_to_cpu(rightmost_el->l_tree_depth);
        if (depth) {
@@ -4228,8 +4349,8 @@ leftright:
 
        if (le16_to_cpu(rightmost_el->l_next_free_rec) ==
            le16_to_cpu(rightmost_el->l_count)) {
-               ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, last_eb_bh,
-                                     meta_ac);
+               ret = ocfs2_grow_tree(inode, handle, et,
+                                     &depth, last_eb_bh, meta_ac);
                if (ret) {
                        mlog_errno(ret);
                        goto out;
@@ -4266,8 +4387,7 @@ leftright:
                do_leftright = 1;
        }
 
-       ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec,
-                                    &insert);
+       ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert);
        if (ret) {
                mlog_errno(ret);
                goto out;
@@ -4309,8 +4429,9 @@ out:
  * of the tree is required. All other cases will degrade into a less
  * optimal tree layout.
  *
- * last_eb_bh should be the rightmost leaf block for any inode with a
- * btree. Since a split may grow the tree or a merge might shrink it, the caller cannot trust the contents of that buffer after this call.
+ * last_eb_bh should be the rightmost leaf block for any extent
+ * btree. Since a split may grow the tree or a merge might shrink it,
+ * the caller cannot trust the contents of that buffer after this call.
  *
  * This code is optimized for readability - several passes might be
  * made over certain portions of the tree. All of those blocks will
@@ -4318,7 +4439,7 @@ out:
  * extra overhead is not expressed in terms of disk reads.
  */
 static int __ocfs2_mark_extent_written(struct inode *inode,
-                                      struct buffer_head *di_bh,
+                                      struct ocfs2_extent_tree *et,
                                       handle_t *handle,
                                       struct ocfs2_path *path,
                                       int split_index,
@@ -4358,10 +4479,9 @@ static int __ocfs2_mark_extent_written(struct inode *inode,
         */
        if (path->p_tree_depth) {
                struct ocfs2_extent_block *eb;
-               struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
 
                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
-                                      le64_to_cpu(di->i_last_eb_blk),
+                                      ocfs2_get_last_eb_blk(et),
                                       &last_eb_bh, OCFS2_BH_CACHED, inode);
                if (ret) {
                        mlog_exit(ret);
@@ -4395,7 +4515,7 @@ static int __ocfs2_mark_extent_written(struct inode *inode,
                if (ctxt.c_split_covers_rec)
                        el->l_recs[split_index] = *split_rec;
                else
-                       ret = ocfs2_split_and_insert(inode, handle, path, di_bh,
+                       ret = ocfs2_split_and_insert(inode, handle, path, et,
                                                     &last_eb_bh, split_index,
                                                     split_rec, meta_ac);
                if (ret)
@@ -4403,7 +4523,7 @@ static int __ocfs2_mark_extent_written(struct inode *inode,
        } else {
                ret = ocfs2_try_to_merge_extent(inode, handle, path,
                                                split_index, split_rec,
-                                               dealloc, &ctxt);
+                                               dealloc, &ctxt, et);
                if (ret)
                        mlog_errno(ret);
        }
@@ -4421,16 +4541,18 @@ out:
  *
  * The caller is responsible for passing down meta_ac if we'll need it.
  */
-int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh,
+int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *root_bh,
                              handle_t *handle, u32 cpos, u32 len, u32 phys,
                              struct ocfs2_alloc_context *meta_ac,
-                             struct ocfs2_cached_dealloc_ctxt *dealloc)
+                             struct ocfs2_cached_dealloc_ctxt *dealloc,
+                             enum ocfs2_extent_tree_type et_type)
 {
        int ret, index;
        u64 start_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys);
        struct ocfs2_extent_rec split_rec;
        struct ocfs2_path *left_path = NULL;
        struct ocfs2_extent_list *el;
+       struct ocfs2_extent_tree *et = NULL;
 
        mlog(0, "Inode %lu cpos %u, len %u, phys %u (%llu)\n",
             inode->i_ino, cpos, len, phys, (unsigned long long)start_blkno);
@@ -4444,13 +4566,21 @@ int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh,
                goto out;
        }
 
+       et = ocfs2_new_extent_tree(root_bh, et_type);
+       if (!et) {
+               ret = -ENOMEM;
+               mlog_errno(ret);
+               goto out;
+       }
+
        /*
         * XXX: This should be fixed up so that we just re-insert the
         * next extent records.
         */
-       ocfs2_extent_map_trunc(inode, 0);
+       if (et_type == OCFS2_DINODE_EXTENT)
+               ocfs2_extent_map_trunc(inode, 0);
 
-       left_path = ocfs2_new_inode_path(di_bh);
+       left_path = ocfs2_new_path(et->root_bh, et->root_el);
        if (!left_path) {
                ret = -ENOMEM;
                mlog_errno(ret);
@@ -4481,23 +4611,25 @@ int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh,
        split_rec.e_flags = path_leaf_el(left_path)->l_recs[index].e_flags;
        split_rec.e_flags &= ~OCFS2_EXT_UNWRITTEN;
 
-       ret = __ocfs2_mark_extent_written(inode, di_bh, handle, left_path,
-                                         index, &split_rec, meta_ac, dealloc);
+       ret = __ocfs2_mark_extent_written(inode, et, handle, left_path,
+                                         index, &split_rec, meta_ac,
+                                         dealloc);
        if (ret)
                mlog_errno(ret);
 
 out:
        ocfs2_free_path(left_path);
+       if (et)
+               ocfs2_free_extent_tree(et);
        return ret;
 }
 
-static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh,
+static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et,
                            handle_t *handle, struct ocfs2_path *path,
                            int index, u32 new_range,
                            struct ocfs2_alloc_context *meta_ac)
 {
        int ret, depth, credits = handle->h_buffer_credits;
-       struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
        struct buffer_head *last_eb_bh = NULL;
        struct ocfs2_extent_block *eb;
        struct ocfs2_extent_list *rightmost_el, *el;
@@ -4515,7 +4647,7 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh,
        depth = path->p_tree_depth;
        if (depth > 0) {
                ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
-                                      le64_to_cpu(di->i_last_eb_blk),
+                                      ocfs2_get_last_eb_blk(et),
                                       &last_eb_bh, OCFS2_BH_CACHED, inode);
                if (ret < 0) {
                        mlog_errno(ret);
@@ -4528,7 +4660,7 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh,
                rightmost_el = path_leaf_el(path);
 
        credits += path->p_tree_depth +
-                  ocfs2_extend_meta_needed(&di->id2.i_list);
+                  ocfs2_extend_meta_needed(et->root_el);
        ret = ocfs2_extend_trans(handle, credits);
        if (ret) {
                mlog_errno(ret);
@@ -4537,7 +4669,7 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh,
 
        if (le16_to_cpu(rightmost_el->l_next_free_rec) ==
            le16_to_cpu(rightmost_el->l_count)) {
-               ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, &last_eb_bh,
+               ret = ocfs2_grow_tree(inode, handle, et, &depth, &last_eb_bh,
                                      meta_ac);
                if (ret) {
                        mlog_errno(ret);
@@ -4551,7 +4683,7 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh,
        insert.ins_split = SPLIT_RIGHT;
        insert.ins_tree_depth = depth;
 
-       ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec, &insert);
+       ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert);
        if (ret)
                mlog_errno(ret);
 
@@ -4563,7 +4695,8 @@ out:
 static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle,
                              struct ocfs2_path *path, int index,
                              struct ocfs2_cached_dealloc_ctxt *dealloc,
-                             u32 cpos, u32 len)
+                             u32 cpos, u32 len,
+                             struct ocfs2_extent_tree *et)
 {
        int ret;
        u32 left_cpos, rec_range, trunc_range;
@@ -4575,7 +4708,7 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle,
        struct ocfs2_extent_block *eb;
 
        if (ocfs2_is_empty_extent(&el->l_recs[0]) && index > 0) {
-               ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc);
+               ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et);
                if (ret) {
                        mlog_errno(ret);
                        goto out;
@@ -4706,7 +4839,7 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle,
 
        ocfs2_journal_dirty(handle, path_leaf_bh(path));
 
-       ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc);
+       ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et);
        if (ret) {
                mlog_errno(ret);
                goto out;
@@ -4717,20 +4850,29 @@ out:
        return ret;
 }
 
-int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
+int ocfs2_remove_extent(struct inode *inode, struct buffer_head *root_bh,
                        u32 cpos, u32 len, handle_t *handle,
                        struct ocfs2_alloc_context *meta_ac,
-                       struct ocfs2_cached_dealloc_ctxt *dealloc)
+                       struct ocfs2_cached_dealloc_ctxt *dealloc,
+                       enum ocfs2_extent_tree_type et_type)
 {
        int ret, index;
        u32 rec_range, trunc_range;
        struct ocfs2_extent_rec *rec;
        struct ocfs2_extent_list *el;
-       struct ocfs2_path *path;
+       struct ocfs2_path *path = NULL;
+       struct ocfs2_extent_tree *et = NULL;
+
+       et = ocfs2_new_extent_tree(root_bh, et_type);
+       if (!et) {
+               ret = -ENOMEM;
+               mlog_errno(ret);
+               goto out;
+       }
 
        ocfs2_extent_map_trunc(inode, 0);
 
-       path = ocfs2_new_inode_path(di_bh);
+       path = ocfs2_new_path(et->root_bh, et->root_el);
        if (!path) {
                ret = -ENOMEM;
                mlog_errno(ret);
@@ -4783,13 +4925,13 @@ int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
 
        if (le32_to_cpu(rec->e_cpos) == cpos || rec_range == trunc_range) {
                ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc,
-                                        cpos, len);
+                                        cpos, len, et);
                if (ret) {
                        mlog_errno(ret);
                        goto out;
                }
        } else {
-               ret = ocfs2_split_tree(inode, di_bh, handle, path, index,
+               ret = ocfs2_split_tree(inode, et, handle, path, index,
                                       trunc_range, meta_ac);
                if (ret) {
                        mlog_errno(ret);
@@ -4838,7 +4980,7 @@ int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
                }
 
                ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc,
-                                        cpos, len);
+                                        cpos, len, et);
                if (ret) {
                        mlog_errno(ret);
                        goto out;
@@ -4847,6 +4989,8 @@ int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
 
 out:
        ocfs2_free_path(path);
+       if (et)
+               ocfs2_free_extent_tree(et);
        return ret;
 }
 
@@ -6355,7 +6499,8 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode,
                 * the in-inode data from our pages.
                 */
                ret = ocfs2_insert_extent(osb, handle, inode, di_bh,
-                                         0, block, 1, 0, NULL);
+                                         0, block, 1, 0,
+                                         NULL, OCFS2_DINODE_EXTENT);
                if (ret) {
                        mlog_errno(ret);
                        goto out_commit;
@@ -6397,13 +6542,14 @@ int ocfs2_commit_truncate(struct ocfs2_super *osb,
        handle_t *handle = NULL;
        struct inode *tl_inode = osb->osb_tl_inode;
        struct ocfs2_path *path = NULL;
+       struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data;
 
        mlog_entry_void();
 
        new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb,
                                                     i_size_read(inode));
 
-       path = ocfs2_new_inode_path(fe_bh);
+       path = ocfs2_new_path(fe_bh, &di->id2.i_list);
        if (!path) {
                status = -ENOMEM;
                mlog_errno(status);
index a0e334f10cd136edbe4c2616d3ecf19832237895..473c8bcc62fbc097d44dfb7e7730bd59b5c9df78 100644 (file)
 #ifndef OCFS2_ALLOC_H
 #define OCFS2_ALLOC_H
 
+enum ocfs2_extent_tree_type {
+       OCFS2_DINODE_EXTENT = 0,
+};
+
 struct ocfs2_alloc_context;
 int ocfs2_insert_extent(struct ocfs2_super *osb,
                        handle_t *handle,
                        struct inode *inode,
-                       struct buffer_head *fe_bh,
+                       struct buffer_head *root_bh,
                        u32 cpos,
                        u64 start_blk,
                        u32 new_clusters,
                        u8 flags,
-                       struct ocfs2_alloc_context *meta_ac);
+                       struct ocfs2_alloc_context *meta_ac,
+                       enum ocfs2_extent_tree_type et_type);
 struct ocfs2_cached_dealloc_ctxt;
-int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh,
+int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *root_bh,
                              handle_t *handle, u32 cpos, u32 len, u32 phys,
                              struct ocfs2_alloc_context *meta_ac,
-                             struct ocfs2_cached_dealloc_ctxt *dealloc);
-int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh,
+                             struct ocfs2_cached_dealloc_ctxt *dealloc,
+                             enum ocfs2_extent_tree_type et_type);
+int ocfs2_remove_extent(struct inode *inode, struct buffer_head *root_bh,
                        u32 cpos, u32 len, handle_t *handle,
                        struct ocfs2_alloc_context *meta_ac,
-                       struct ocfs2_cached_dealloc_ctxt *dealloc);
+                       struct ocfs2_cached_dealloc_ctxt *dealloc,
+                       enum ocfs2_extent_tree_type et_type);
 int ocfs2_num_free_extents(struct ocfs2_super *osb,
                           struct inode *inode,
-                          struct buffer_head *bh);
+                          struct buffer_head *root_bh,
+                          enum ocfs2_extent_tree_type et_type);
+
 /*
  * how many new metadata chunks would an allocation need at maximum?
  *
index bbe3f8b2d0e2058fbe19793507bf415b2593870e..44ea5eb3fdc4e678ccb9369df52c8b5ec6430cdb 100644 (file)
@@ -1278,7 +1278,8 @@ static int ocfs2_write_cluster(struct address_space *mapping,
        } else if (unwritten) {
                ret = ocfs2_mark_extent_written(inode, wc->w_di_bh,
                                                wc->w_handle, cpos, 1, phys,
-                                               meta_ac, &wc->w_dealloc);
+                                               meta_ac, &wc->w_dealloc,
+                                               OCFS2_DINODE_EXTENT);
                if (ret < 0) {
                        mlog_errno(ret);
                        goto out;
@@ -1712,7 +1713,13 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
                 * ocfs2_lock_allocators(). It greatly over-estimates
                 * the work to be done.
                 */
-               ret = ocfs2_lock_allocators(inode, wc->w_di_bh,
+               mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u,"
+                    " clusters_to_add = %u, extents_to_split = %u\n",
+                    (unsigned long long)OCFS2_I(inode)->ip_blkno,
+                    (long long)i_size_read(inode), le32_to_cpu(di->i_clusters),
+                    clusters_to_alloc, extents_to_split);
+
+               ret = ocfs2_lock_allocators(inode, wc->w_di_bh, &di->id2.i_list,
                                            clusters_to_alloc, extents_to_split,
                                            &data_ac, &meta_ac);
                if (ret) {
index 126aa219c0c1b207383beee5fe654cddb54d6603..ba0fb9e16264a8d16053e7adab86910960f4e4a9 100644 (file)
@@ -1306,7 +1306,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
         * related blocks have been journaled already.
         */
        ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 0, blkno, len, 0,
-                                 NULL);
+                                 NULL, OCFS2_DINODE_EXTENT);
        if (ret) {
                mlog_errno(ret);
                goto out_commit;
@@ -1338,7 +1338,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
                blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off);
 
                ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 1, blkno,
-                                         len, 0, NULL);
+                                         len, 0, NULL, OCFS2_DINODE_EXTENT);
                if (ret) {
                        mlog_errno(ret);
                        goto out_commit;
@@ -1481,7 +1481,8 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
        if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) {
                spin_unlock(&OCFS2_I(dir)->ip_lock);
                num_free_extents = ocfs2_num_free_extents(osb, dir,
-                                                         parent_fe_bh);
+                                                         parent_fe_bh,
+                                                         OCFS2_DINODE_EXTENT);
                if (num_free_extents < 0) {
                        status = num_free_extents;
                        mlog_errno(status);
index a31bba6c5575d7aa0aef03fdaeb18fc60d73d2bc..f567cc53d9bc6fa69339067f691376885bac4714 100644 (file)
@@ -521,7 +521,8 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
        if (mark_unwritten)
                flags = OCFS2_EXT_UNWRITTEN;
 
-       free_extents = ocfs2_num_free_extents(osb, inode, fe_bh);
+       free_extents = ocfs2_num_free_extents(osb, inode, fe_bh,
+                                             OCFS2_DINODE_EXTENT);
        if (free_extents < 0) {
                status = free_extents;
                mlog_errno(status);
@@ -570,7 +571,7 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
             num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno);
        status = ocfs2_insert_extent(osb, handle, inode, fe_bh,
                                     *logical_offset, block, num_bits,
-                                    flags, meta_ac);
+                                    flags, meta_ac, OCFS2_DINODE_EXTENT);
        if (status < 0) {
                mlog_errno(status);
                goto leave;
@@ -599,92 +600,6 @@ leave:
        return status;
 }
 
-/*
- * For a given allocation, determine which allocators will need to be
- * accessed, and lock them, reserving the appropriate number of bits.
- *
- * Sparse file systems call this from ocfs2_write_begin_nolock()
- * and ocfs2_allocate_unwritten_extents().
- *
- * File systems which don't support holes call this from
- * ocfs2_extend_allocation().
- */
-int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *di_bh,
-                         u32 clusters_to_add, u32 extents_to_split,
-                         struct ocfs2_alloc_context **data_ac,
-                         struct ocfs2_alloc_context **meta_ac)
-{
-       int ret = 0, num_free_extents;
-       unsigned int max_recs_needed = clusters_to_add + 2 * extents_to_split;
-       struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-       struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
-
-       *meta_ac = NULL;
-       if (data_ac)
-               *data_ac = NULL;
-
-       BUG_ON(clusters_to_add != 0 && data_ac == NULL);
-
-       mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u, "
-            "clusters_to_add = %u, extents_to_split = %u\n",
-            (unsigned long long)OCFS2_I(inode)->ip_blkno, (long long)i_size_read(inode),
-            le32_to_cpu(di->i_clusters), clusters_to_add, extents_to_split);
-
-       num_free_extents = ocfs2_num_free_extents(osb, inode, di_bh);
-       if (num_free_extents < 0) {
-               ret = num_free_extents;
-               mlog_errno(ret);
-               goto out;
-       }
-
-       /*
-        * Sparse allocation file systems need to be more conservative
-        * with reserving room for expansion - the actual allocation
-        * happens while we've got a journal handle open so re-taking
-        * a cluster lock (because we ran out of room for another
-        * extent) will violate ordering rules.
-        *
-        * Most of the time we'll only be seeing this 1 cluster at a time
-        * anyway.
-        *
-        * Always lock for any unwritten extents - we might want to
-        * add blocks during a split.
-        */
-       if (!num_free_extents ||
-           (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) {
-               ret = ocfs2_reserve_new_metadata(osb, &di->id2.i_list, meta_ac);
-               if (ret < 0) {
-                       if (ret != -ENOSPC)
-                               mlog_errno(ret);
-                       goto out;
-               }
-       }
-
-       if (clusters_to_add == 0)
-               goto out;
-
-       ret = ocfs2_reserve_clusters(osb, clusters_to_add, data_ac);
-       if (ret < 0) {
-               if (ret != -ENOSPC)
-                       mlog_errno(ret);
-               goto out;
-       }
-
-out:
-       if (ret) {
-               if (*meta_ac) {
-                       ocfs2_free_alloc_context(*meta_ac);
-                       *meta_ac = NULL;
-               }
-
-               /*
-                * We cannot have an error and a non null *data_ac.
-                */
-       }
-
-       return ret;
-}
-
 static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
                                     u32 clusters_to_add, int mark_unwritten)
 {
@@ -725,7 +640,13 @@ static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
 restart_all:
        BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
 
-       status = ocfs2_lock_allocators(inode, bh, clusters_to_add, 0, &data_ac,
+       mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u, "
+            "clusters_to_add = %u\n",
+            (unsigned long long)OCFS2_I(inode)->ip_blkno,
+            (long long)i_size_read(inode), le32_to_cpu(fe->i_clusters),
+            clusters_to_add);
+       status = ocfs2_lock_allocators(inode, bh, &fe->id2.i_list,
+                                      clusters_to_add, 0, &data_ac,
                                       &meta_ac);
        if (status) {
                mlog_errno(status);
@@ -1397,7 +1318,8 @@ static int __ocfs2_remove_inode_range(struct inode *inode,
        struct ocfs2_alloc_context *meta_ac = NULL;
        struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
 
-       ret = ocfs2_lock_allocators(inode, di_bh, 0, 1, NULL, &meta_ac);
+       ret = ocfs2_lock_allocators(inode, di_bh, &di->id2.i_list,
+                                   0, 1, NULL, &meta_ac);
        if (ret) {
                mlog_errno(ret);
                return ret;
@@ -1428,7 +1350,7 @@ static int __ocfs2_remove_inode_range(struct inode *inode,
        }
 
        ret = ocfs2_remove_extent(inode, di_bh, cpos, len, handle, meta_ac,
-                                 dealloc);
+                                 dealloc, OCFS2_DINODE_EXTENT);
        if (ret) {
                mlog_errno(ret);
                goto out_commit;
index c96b8054fbe750114d41e5892c04a7715d67a440..18e5c80cc7377c5bcc8e8df15eee3a82e2feca8f 100644 (file)
@@ -57,10 +57,6 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
                               enum ocfs2_alloc_restarted *reason_ret);
 int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size,
                          u64 zero_to);
-int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *fe,
-                         u32 clusters_to_add, u32 extents_to_split,
-                         struct ocfs2_alloc_context **data_ac,
-                         struct ocfs2_alloc_context **meta_ac);
 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
 int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
                  struct kstat *stat);
index 2a817bca1ddb14ad03907dee8ed9dca50969e351..b642c825fb7c810934a20b5fc51ab2f540f134bd 100644 (file)
@@ -1894,3 +1894,85 @@ static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe)
                       (unsigned long long)fe->id2.i_chain.cl_recs[i].c_blkno);
        }
 }
+
+/*
+ * For a given allocation, determine which allocators will need to be
+ * accessed, and lock them, reserving the appropriate number of bits.
+ *
+ * Sparse file systems call this from ocfs2_write_begin_nolock()
+ * and ocfs2_allocate_unwritten_extents().
+ *
+ * File systems which don't support holes call this from
+ * ocfs2_extend_allocation().
+ */
+int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *root_bh,
+                         struct ocfs2_extent_list *root_el,
+                         u32 clusters_to_add, u32 extents_to_split,
+                         struct ocfs2_alloc_context **data_ac,
+                         struct ocfs2_alloc_context **meta_ac)
+{
+       int ret = 0, num_free_extents;
+       unsigned int max_recs_needed = clusters_to_add + 2 * extents_to_split;
+       struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+       *meta_ac = NULL;
+       if (data_ac)
+               *data_ac = NULL;
+
+       BUG_ON(clusters_to_add != 0 && data_ac == NULL);
+
+       num_free_extents = ocfs2_num_free_extents(osb, inode, root_bh,
+                                                 OCFS2_DINODE_EXTENT);
+       if (num_free_extents < 0) {
+               ret = num_free_extents;
+               mlog_errno(ret);
+               goto out;
+       }
+
+       /*
+        * Sparse allocation file systems need to be more conservative
+        * with reserving room for expansion - the actual allocation
+        * happens while we've got a journal handle open so re-taking
+        * a cluster lock (because we ran out of room for another
+        * extent) will violate ordering rules.
+        *
+        * Most of the time we'll only be seeing this 1 cluster at a time
+        * anyway.
+        *
+        * Always lock for any unwritten extents - we might want to
+        * add blocks during a split.
+        */
+       if (!num_free_extents ||
+           (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) {
+               ret = ocfs2_reserve_new_metadata(osb, root_el, meta_ac);
+               if (ret < 0) {
+                       if (ret != -ENOSPC)
+                               mlog_errno(ret);
+                       goto out;
+               }
+       }
+
+       if (clusters_to_add == 0)
+               goto out;
+
+       ret = ocfs2_reserve_clusters(osb, clusters_to_add, data_ac);
+       if (ret < 0) {
+               if (ret != -ENOSPC)
+                       mlog_errno(ret);
+               goto out;
+       }
+
+out:
+       if (ret) {
+               if (*meta_ac) {
+                       ocfs2_free_alloc_context(*meta_ac);
+                       *meta_ac = NULL;
+               }
+
+               /*
+                * We cannot have an error and a non null *data_ac.
+                */
+       }
+
+       return ret;
+}
index 3f96c875bcf750d3d393e9952732030d8fc76428..a3e531e62df27cc19825497c33dd63ebed909ff7 100644 (file)
@@ -162,4 +162,9 @@ u64 ocfs2_which_cluster_group(struct inode *inode, u32 cluster);
 int ocfs2_check_group_descriptor(struct super_block *sb,
                                 struct ocfs2_dinode *di,
                                 struct ocfs2_group_desc *gd);
+int ocfs2_lock_allocators(struct inode *inode, struct buffer_head *root_bh,
+                         struct ocfs2_extent_list *root_el,
+                         u32 clusters_to_add, u32 extents_to_split,
+                         struct ocfs2_alloc_context **data_ac,
+                         struct ocfs2_alloc_context **meta_ac);
 #endif /* _CHAINALLOC_H_ */