[PATCH] ext3: add extent map support

On disk extents format:
/*
* this is extent on-disk structure
* it's used at the bottom of the tree
*/
struct ext3_extent {
__le32  ee_block;       /* first logical block extent covers */
__le16  ee_len;         /* number of blocks covered by extent */
__le16  ee_start_hi;    /* high 16 bits of physical block */
__le32  ee_start;       /* low 32 bigs of physical block */
};

Signed-off-by: Alex Tomas <alex@clusterfs.com>
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

7 files changed, 2108 insertions, 11 deletions

diff --git a/fs/ext4/Makefile b/fs/ext4/Makefile
index 09c487893e4a..a6acb96ebeb9 100644
--- a/fs/ext4/Makefile
+++ b/fs/ext4/Makefile
@@ -5,7 +5,7 @@
 obj-$(CONFIG_EXT4DEV_FS) += ext4dev.o
 
 ext4dev-y	:= balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \
-	   ioctl.o namei.o super.o symlink.o hash.o resize.o
+	   ioctl.o namei.o super.o symlink.o hash.o resize.o extents.o
 
 ext4dev-$(CONFIG_EXT4DEV_FS_XATTR)	+= xattr.o xattr_user.o xattr_trusted.o
 ext4dev-$(CONFIG_EXT4DEV_FS_POSIX_ACL)	+= acl.o
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index 9833d5d00c46..18ac173af575 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -134,8 +134,7 @@ static int ext4_readdir(struct file * filp,
 		struct buffer_head *bh = NULL;
 
 		map_bh.b_state = 0;
-		err = ext4_get_blocks_handle(NULL, inode, blk, 1,
-						&map_bh, 0, 0);
+		err = ext4_get_blocks_wrap(NULL, inode, blk, 1, &map_bh, 0, 0);
 		if (err > 0) {
 			page_cache_readahead(sb->s_bdev->bd_inode->i_mapping,
 				&filp->f_ra,
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
new file mode 100644
index 000000000000..f67b2ef6a71f
--- /dev/null
+++ b/fs/ext4/extents.c
@@ -0,0 +1,2075 @@
+/*
+ * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
+ * Written by Alex Tomas <alex@clusterfs.com>
+ *
+ * Architecture independence:
+ *   Copyright (c) 2005, Bull S.A.
+ *   Written by Pierre Peiffer <pierre.peiffer@bull.net>
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public Licens
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
+ */
+
+/*
+ * Extents support for EXT4
+ *
+ * TODO:
+ *   - ext4*_error() should be used in some situations
+ *   - analyze all BUG()/BUG_ON(), use -EIO where appropriate
+ *   - smart tree reduction
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/time.h>
+#include <linux/ext4_jbd2.h>
+#include <linux/jbd.h>
+#include <linux/smp_lock.h>
+#include <linux/highuid.h>
+#include <linux/pagemap.h>
+#include <linux/quotaops.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+#include <linux/ext4_fs_extents.h>
+#include <asm/uaccess.h>
+
+
+static int ext4_ext_check_header(const char *function, struct inode *inode,
+				struct ext4_extent_header *eh)
+{
+	const char *error_msg = NULL;
+
+	if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
+		error_msg = "invalid magic";
+		goto corrupted;
+	}
+	if (unlikely(eh->eh_max == 0)) {
+		error_msg = "invalid eh_max";
+		goto corrupted;
+	}
+	if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
+		error_msg = "invalid eh_entries";
+		goto corrupted;
+	}
+	return 0;
+
+corrupted:
+	ext4_error(inode->i_sb, function,
+			"bad header in inode #%lu: %s - magic %x, "
+			"entries %u, max %u, depth %u",
+			inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic),
+			le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
+			le16_to_cpu(eh->eh_depth));
+
+	return -EIO;
+}
+
+static handle_t *ext4_ext_journal_restart(handle_t *handle, int needed)
+{
+	int err;
+
+	if (handle->h_buffer_credits > needed)
+		return handle;
+	if (!ext4_journal_extend(handle, needed))
+		return handle;
+	err = ext4_journal_restart(handle, needed);
+
+	return handle;
+}
+
+/*
+ * could return:
+ *  - EROFS
+ *  - ENOMEM
+ */
+static int ext4_ext_get_access(handle_t *handle, struct inode *inode,
+				struct ext4_ext_path *path)
+{
+	if (path->p_bh) {
+		/* path points to block */
+		return ext4_journal_get_write_access(handle, path->p_bh);
+	}
+	/* path points to leaf/index in inode body */
+	/* we use in-core data, no need to protect them */
+	return 0;
+}
+
+/*
+ * could return:
+ *  - EROFS
+ *  - ENOMEM
+ *  - EIO
+ */
+static int ext4_ext_dirty(handle_t *handle, struct inode *inode,
+				struct ext4_ext_path *path)
+{
+	int err;
+	if (path->p_bh) {
+		/* path points to block */
+		err = ext4_journal_dirty_metadata(handle, path->p_bh);
+	} else {
+		/* path points to leaf/index in inode body */
+		err = ext4_mark_inode_dirty(handle, inode);
+	}
+	return err;
+}
+
+static int ext4_ext_find_goal(struct inode *inode,
+			      struct ext4_ext_path *path,
+			      unsigned long block)
+{
+	struct ext4_inode_info *ei = EXT4_I(inode);
+	unsigned long bg_start;
+	unsigned long colour;
+	int depth;
+
+	if (path) {
+		struct ext4_extent *ex;
+		depth = path->p_depth;
+
+		/* try to predict block placement */
+		if ((ex = path[depth].p_ext))
+			return le32_to_cpu(ex->ee_start)
+					+ (block - le32_to_cpu(ex->ee_block));
+
+		/* it looks index is empty
+		 * try to find starting from index itself */
+		if (path[depth].p_bh)
+			return path[depth].p_bh->b_blocknr;
+	}
+
+	/* OK. use inode's group */
+	bg_start = (ei->i_block_group * EXT4_BLOCKS_PER_GROUP(inode->i_sb)) +
+		le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_first_data_block);
+	colour = (current->pid % 16) *
+			(EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16);
+	return bg_start + colour + block;
+}
+
+static int
+ext4_ext_new_block(handle_t *handle, struct inode *inode,
+			struct ext4_ext_path *path,
+			struct ext4_extent *ex, int *err)
+{
+	int goal, newblock;
+
+	goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
+	newblock = ext4_new_block(handle, inode, goal, err);
+	return newblock;
+}
+
+static inline int ext4_ext_space_block(struct inode *inode)
+{
+	int size;
+
+	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
+			/ sizeof(struct ext4_extent);
+#ifdef AGRESSIVE_TEST
+	if (size > 6)
+		size = 6;
+#endif
+	return size;
+}
+
+static inline int ext4_ext_space_block_idx(struct inode *inode)
+{
+	int size;
+
+	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
+			/ sizeof(struct ext4_extent_idx);
+#ifdef AGRESSIVE_TEST
+	if (size > 5)
+		size = 5;
+#endif
+	return size;
+}
+
+static inline int ext4_ext_space_root(struct inode *inode)
+{
+	int size;
+
+	size = sizeof(EXT4_I(inode)->i_data);
+	size -= sizeof(struct ext4_extent_header);
+	size /= sizeof(struct ext4_extent);
+#ifdef AGRESSIVE_TEST
+	if (size > 3)
+		size = 3;
+#endif
+	return size;
+}
+
+static inline int ext4_ext_space_root_idx(struct inode *inode)
+{
+	int size;
+
+	size = sizeof(EXT4_I(inode)->i_data);
+	size -= sizeof(struct ext4_extent_header);
+	size /= sizeof(struct ext4_extent_idx);
+#ifdef AGRESSIVE_TEST
+	if (size > 4)
+		size = 4;
+#endif
+	return size;
+}
+
+#ifdef EXT_DEBUG
+static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
+{
+	int k, l = path->p_depth;
+
+	ext_debug("path:");
+	for (k = 0; k <= l; k++, path++) {
+		if (path->p_idx) {
+		  ext_debug("  %d->%d", le32_to_cpu(path->p_idx->ei_block),
+			    le32_to_cpu(path->p_idx->ei_leaf));
+		} else if (path->p_ext) {
+			ext_debug("  %d:%d:%d",
+				  le32_to_cpu(path->p_ext->ee_block),
+				  le16_to_cpu(path->p_ext->ee_len),
+				  le32_to_cpu(path->p_ext->ee_start));
+		} else
+			ext_debug("  []");
+	}
+	ext_debug("\n");
+}
+
+static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
+{
+	int depth = ext_depth(inode);
+	struct ext4_extent_header *eh;
+	struct ext4_extent *ex;
+	int i;
+
+	if (!path)
+		return;
+
+	eh = path[depth].p_hdr;
+	ex = EXT_FIRST_EXTENT(eh);
+
+	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
+		ext_debug("%d:%d:%d ", le32_to_cpu(ex->ee_block),
+			  le16_to_cpu(ex->ee_len),
+			  le32_to_cpu(ex->ee_start));
+	}
+	ext_debug("\n");
+}
+#else
+#define ext4_ext_show_path(inode,path)
+#define ext4_ext_show_leaf(inode,path)
+#endif
+
+static void ext4_ext_drop_refs(struct ext4_ext_path *path)
+{
+	int depth = path->p_depth;
+	int i;
+
+	for (i = 0; i <= depth; i++, path++)
+		if (path->p_bh) {
+			brelse(path->p_bh);
+			path->p_bh = NULL;
+		}
+}
+
+/*
+ * binary search for closest index by given block
+ */
+static void
+ext4_ext_binsearch_idx(struct inode *inode, struct ext4_ext_path *path, int block)
+{
+	struct ext4_extent_header *eh = path->p_hdr;
+	struct ext4_extent_idx *r, *l, *m;
+
+	BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC);
+	BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max));
+	BUG_ON(le16_to_cpu(eh->eh_entries) <= 0);
+
+	ext_debug("binsearch for %d(idx):  ", block);
+
+	l = EXT_FIRST_INDEX(eh) + 1;
+	r = EXT_FIRST_INDEX(eh) + le16_to_cpu(eh->eh_entries) - 1;
+	while (l <= r) {
+		m = l + (r - l) / 2;
+		if (block < le32_to_cpu(m->ei_block))
+			r = m - 1;
+		else
+			l = m + 1;
+		ext_debug("%p(%u):%p(%u):%p(%u) ", l, l->ei_block,
+				m, m->ei_block, r, r->ei_block);
+	}
+
+	path->p_idx = l - 1;
+	ext_debug("  -> %d->%d ", le32_to_cpu(path->p_idx->ei_block),
+		  le32_to_cpu(path->p_idx->ei_leaf));
+
+#ifdef CHECK_BINSEARCH
+	{
+		struct ext4_extent_idx *chix, *ix;
+		int k;
+
+		chix = ix = EXT_FIRST_INDEX(eh);
+		for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) {
+		  if (k != 0 &&
+		      le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) {
+				printk("k=%d, ix=0x%p, first=0x%p\n", k,
+					ix, EXT_FIRST_INDEX(eh));
+				printk("%u <= %u\n",
+				       le32_to_cpu(ix->ei_block),
+				       le32_to_cpu(ix[-1].ei_block));
+			}
+			BUG_ON(k && le32_to_cpu(ix->ei_block)
+				           <= le32_to_cpu(ix[-1].ei_block));
+			if (block < le32_to_cpu(ix->ei_block))
+				break;
+			chix = ix;
+		}
+		BUG_ON(chix != path->p_idx);
+	}
+#endif
+
+}
+
+/*
+ * binary search for closest extent by given block
+ */
+static void
+ext4_ext_binsearch(struct inode *inode, struct ext4_ext_path *path, int block)
+{
+	struct ext4_extent_header *eh = path->p_hdr;
+	struct ext4_extent *r, *l, *m;
+
+	BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC);
+	BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max));
+
+	if (eh->eh_entries == 0) {
+		/*
+		 * this leaf is empty yet:
+		 *  we get such a leaf in split/add case
+		 */
+		return;
+	}
+
+	ext_debug("binsearch for %d:  ", block);
+
+	l = EXT_FIRST_EXTENT(eh) + 1;
+	r = EXT_FIRST_EXTENT(eh) + le16_to_cpu(eh->eh_entries) - 1;
+
+	while (l <= r) {
+		m = l + (r - l) / 2;
+		if (block < le32_to_cpu(m->ee_block))
+			r = m - 1;
+		else
+			l = m + 1;
+		ext_debug("%p(%u):%p(%u):%p(%u) ", l, l->ee_block,
+				m, m->ee_block, r, r->ee_block);
+	}
+
+	path->p_ext = l - 1;
+	ext_debug("  -> %d:%d:%d ",
+		        le32_to_cpu(path->p_ext->ee_block),
+		        le32_to_cpu(path->p_ext->ee_start),
+		        le16_to_cpu(path->p_ext->ee_len));
+
+#ifdef CHECK_BINSEARCH
+	{
+		struct ext4_extent *chex, *ex;
+		int k;
+
+		chex = ex = EXT_FIRST_EXTENT(eh);
+		for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) {
+			BUG_ON(k && le32_to_cpu(ex->ee_block)
+				          <= le32_to_cpu(ex[-1].ee_block));
+			if (block < le32_to_cpu(ex->ee_block))
+				break;
+			chex = ex;
+		}
+		BUG_ON(chex != path->p_ext);
+	}
+#endif
+
+}
+
+int ext4_ext_tree_init(handle_t *handle, struct inode *inode)
+{
+	struct ext4_extent_header *eh;
+
+	eh = ext_inode_hdr(inode);
+	eh->eh_depth = 0;
+	eh->eh_entries = 0;
+	eh->eh_magic = EXT4_EXT_MAGIC;
+	eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode));
+	ext4_mark_inode_dirty(handle, inode);
+	ext4_ext_invalidate_cache(inode);
+	return 0;
+}
+
+struct ext4_ext_path *
+ext4_ext_find_extent(struct inode *inode, int block, struct ext4_ext_path *path)
+{
+	struct ext4_extent_header *eh;
+	struct buffer_head *bh;
+	short int depth, i, ppos = 0, alloc = 0;
+
+	eh = ext_inode_hdr(inode);
+	BUG_ON(eh == NULL);
+	if (ext4_ext_check_header(__FUNCTION__, inode, eh))
+		return ERR_PTR(-EIO);
+
+	i = depth = ext_depth(inode);
+
+	/* account possible depth increase */
+	if (!path) {
+		path = kmalloc(sizeof(struct ext4_ext_path) * (depth + 2),
+				GFP_NOFS);
+		if (!path)
+			return ERR_PTR(-ENOMEM);
+		alloc = 1;
+	}
+	memset(path, 0, sizeof(struct ext4_ext_path) * (depth + 1));
+	path[0].p_hdr = eh;
+
+	/* walk through the tree */
+	while (i) {
+		ext_debug("depth %d: num %d, max %d\n",
+			  ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
+		ext4_ext_binsearch_idx(inode, path + ppos, block);
+		path[ppos].p_block = le32_to_cpu(path[ppos].p_idx->ei_leaf);
+		path[ppos].p_depth = i;
+		path[ppos].p_ext = NULL;
+
+		bh = sb_bread(inode->i_sb, path[ppos].p_block);
+		if (!bh)
+			goto err;
+
+		eh = ext_block_hdr(bh);
+		ppos++;
+		BUG_ON(ppos > depth);
+		path[ppos].p_bh = bh;
+		path[ppos].p_hdr = eh;
+		i--;
+
+		if (ext4_ext_check_header(__FUNCTION__, inode, eh))
+			goto err;
+	}
+
+	path[ppos].p_depth = i;
+	path[ppos].p_hdr = eh;
+	path[ppos].p_ext = NULL;
+	path[ppos].p_idx = NULL;
+
+	if (ext4_ext_check_header(__FUNCTION__, inode, eh))
+		goto err;
+
+	/* find extent */
+	ext4_ext_binsearch(inode, path + ppos, block);
+
+	ext4_ext_show_path(inode, path);
+
+	return path;
+
+err:
+	ext4_ext_drop_refs(path);
+	if (alloc)
+		kfree(path);
+	return ERR_PTR(-EIO);
+}
+
+/*
+ * insert new index [logical;ptr] into the block at cupr
+ * it check where to insert: before curp or after curp
+ */
+static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
+				struct ext4_ext_path *curp,
+				int logical, int ptr)
+{
+	struct ext4_extent_idx *ix;
+	int len, err;
+
+	if ((err = ext4_ext_get_access(handle, inode, curp)))
+		return err;
+
+	BUG_ON(logical == le32_to_cpu(curp->p_idx->ei_block));
+	len = EXT_MAX_INDEX(curp->p_hdr) - curp->p_idx;
+	if (logical > le32_to_cpu(curp->p_idx->ei_block)) {
+		/* insert after */
+		if (curp->p_idx != EXT_LAST_INDEX(curp->p_hdr)) {
+			len = (len - 1) * sizeof(struct ext4_extent_idx);
+			len = len < 0 ? 0 : len;
+			ext_debug("insert new index %d after: %d. "
+					"move %d from 0x%p to 0x%p\n",
+					logical, ptr, len,
+					(curp->p_idx + 1), (curp->p_idx + 2));
+			memmove(curp->p_idx + 2, curp->p_idx + 1, len);
+		}
+		ix = curp->p_idx + 1;
+	} else {
+		/* insert before */
+		len = len * sizeof(struct ext4_extent_idx);
+		len = len < 0 ? 0 : len;
+		ext_debug("insert new index %d before: %d. "
+				"move %d from 0x%p to 0x%p\n",
+				logical, ptr, len,
+				curp->p_idx, (curp->p_idx + 1));
+		memmove(curp->p_idx + 1, curp->p_idx, len);
+		ix = curp->p_idx;
+	}
+
+	ix->ei_block = cpu_to_le32(logical);
+	ix->ei_leaf = cpu_to_le32(ptr);
+	curp->p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(curp->p_hdr->eh_entries)+1);
+
+	BUG_ON(le16_to_cpu(curp->p_hdr->eh_entries)
+	                     > le16_to_cpu(curp->p_hdr->eh_max));
+	BUG_ON(ix > EXT_LAST_INDEX(curp->p_hdr));
+
+	err = ext4_ext_dirty(handle, inode, curp);
+	ext4_std_error(inode->i_sb, err);
+
+	return err;
+}
+
+/*
+ * routine inserts new subtree into the path, using free index entry
+ * at depth 'at:
+ *  - allocates all needed blocks (new leaf and all intermediate index blocks)
+ *  - makes decision where to split
+ *  - moves remaining extens and index entries (right to the split point)
+ *    into the newly allocated blocks
+ *  - initialize subtree
+ */
+static int ext4_ext_split(handle_t *handle, struct inode *inode,
+				struct ext4_ext_path *path,
+				struct ext4_extent *newext, int at)
+{
+	struct buffer_head *bh = NULL;
+	int depth = ext_depth(inode);
+	struct ext4_extent_header *neh;
+	struct ext4_extent_idx *fidx;
+	struct ext4_extent *ex;
+	int i = at, k, m, a;
+	unsigned long newblock, oldblock;
+	__le32 border;
+	int *ablocks = NULL; /* array of allocated blocks */
+	int err = 0;
+
+	/* make decision: where to split? */
+	/* FIXME: now desicion is simplest: at current extent */
+
+	/* if current leaf will be splitted, then we should use
+	 * border from split point */
+	BUG_ON(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr));
+	if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) {
+		border = path[depth].p_ext[1].ee_block;
+		ext_debug("leaf will be splitted."
+				" next leaf starts at %d\n",
+			          le32_to_cpu(border));
+	} else {
+		border = newext->ee_block;
+		ext_debug("leaf will be added."
+				" next leaf starts at %d\n",
+			        le32_to_cpu(border));
+	}
+
+	/*
+	 * if error occurs, then we break processing
+	 * and turn filesystem read-only. so, index won't
+	 * be inserted and tree will be in consistent
+	 * state. next mount will repair buffers too
+	 */
+
+	/*
+	 * get array to track all allocated blocks
+	 * we need this to handle errors and free blocks
+	 * upon them
+	 */
+	ablocks = kmalloc(sizeof(unsigned long) * depth, GFP_NOFS);
+	if (!ablocks)
+		return -ENOMEM;
+	memset(ablocks, 0, sizeof(unsigned long) * depth);
+
+	/* allocate all needed blocks */
+	ext_debug("allocate %d blocks for indexes/leaf\n", depth - at);
+	for (a = 0; a < depth - at; a++) {
+		newblock = ext4_ext_new_block(handle, inode, path, newext, &err);
+		if (newblock == 0)
+			goto cleanup;
+		ablocks[a] = newblock;
+	}
+
+	/* initialize new leaf */
+	newblock = ablocks[--a];
+	BUG_ON(newblock == 0);
+	bh = sb_getblk(inode->i_sb, newblock);
+	if (!bh) {
+		err = -EIO;
+		goto cleanup;
+	}
+	lock_buffer(bh);
+
+	if ((err = ext4_journal_get_create_access(handle, bh)))
+		goto cleanup;
+
+	neh = ext_block_hdr(bh);
+	neh->eh_entries = 0;
+	neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode));
+	neh->eh_magic = EXT4_EXT_MAGIC;
+	neh->eh_depth = 0;
+	ex = EXT_FIRST_EXTENT(neh);
+
+	/* move remain of path[depth] to the new leaf */
+	BUG_ON(path[depth].p_hdr->eh_entries != path[depth].p_hdr->eh_max);
+	/* start copy from next extent */
+	/* TODO: we could do it by single memmove */
+	m = 0;
+	path[depth].p_ext++;
+	while (path[depth].p_ext <=
+			EXT_MAX_EXTENT(path[depth].p_hdr)) {
+		ext_debug("move %d:%d:%d in new leaf %lu\n",
+			        le32_to_cpu(path[depth].p_ext->ee_block),
+			        le32_to_cpu(path[depth].p_ext->ee_start),
+			        le16_to_cpu(path[depth].p_ext->ee_len),
+				newblock);
+		/*memmove(ex++, path[depth].p_ext++,
+				sizeof(struct ext4_extent));
+		neh->eh_entries++;*/
+		path[depth].p_ext++;
+		m++;
+	}
+	if (m) {
+		memmove(ex, path[depth].p_ext-m, sizeof(struct ext4_extent)*m);
+		neh->eh_entries = cpu_to_le16(le16_to_cpu(neh->eh_entries)+m);
+	}
+
+	set_buffer_uptodate(bh);
+	unlock_buffer(bh);
+
+	if ((err = ext4_journal_dirty_metadata(handle, bh)))
+		goto cleanup;
+	brelse(bh);
+	bh = NULL;
+
+	/* correct old leaf */
+	if (m) {
+		if ((err = ext4_ext_get_access(handle, inode, path + depth)))
+			goto cleanup;
+		path[depth].p_hdr->eh_entries =
+		     cpu_to_le16(le16_to_cpu(path[depth].p_hdr->eh_entries)-m);
+		if ((err = ext4_ext_dirty(handle, inode, path + depth)))
+			goto cleanup;
+
+	}
+
+	/* create intermediate indexes */
+	k = depth - at - 1;
+	BUG_ON(k < 0);
+	if (k)
+		ext_debug("create %d intermediate indices\n", k);
+	/* insert new index into current index block */
+	/* current depth stored in i var */
+	i = depth - 1;
+	while (k--) {
+		oldblock = newblock;
+		newblock = ablocks[--a];
+		bh = sb_getblk(inode->i_sb, newblock);
+		if (!bh) {
+			err = -EIO;
+			goto cleanup;
+		}
+		lock_buffer(bh);
+
+		if ((err = ext4_journal_get_create_access(handle, bh)))
+			goto cleanup;
+
+		neh = ext_block_hdr(bh);
+		neh->eh_entries = cpu_to_le16(1);
+		neh->eh_magic = EXT4_EXT_MAGIC;
+		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode));
+		neh->eh_depth = cpu_to_le16(depth - i);
+		fidx = EXT_FIRST_INDEX(neh);
+		fidx->ei_block = border;
+		fidx->ei_leaf = cpu_to_le32(oldblock);
+
+		ext_debug("int.index at %d (block %lu): %lu -> %lu\n", i,
+				newblock, (unsigned long) le32_to_cpu(border),
+				oldblock);
+		/* copy indexes */
+		m = 0;
+		path[i].p_idx++;
+
+		ext_debug("cur 0x%p, last 0x%p\n", path[i].p_idx,
+				EXT_MAX_INDEX(path[i].p_hdr));
+		BUG_ON(EXT_MAX_INDEX(path[i].p_hdr) !=
+				EXT_LAST_INDEX(path[i].p_hdr));
+		while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) {
+			ext_debug("%d: move %d:%d in new index %lu\n", i,
+				        le32_to_cpu(path[i].p_idx->ei_block),
+				        le32_to_cpu(path[i].p_idx->ei_leaf),
+				        newblock);
+			/*memmove(++fidx, path[i].p_idx++,
+					sizeof(struct ext4_extent_idx));
+			neh->eh_entries++;
+			BUG_ON(neh->eh_entries > neh->eh_max);*/
+			path[i].p_idx++;
+			m++;
+		}
+		if (m) {
+			memmove(++fidx, path[i].p_idx - m,
+				sizeof(struct ext4_extent_idx) * m);
+			neh->eh_entries =
+				cpu_to_le16(le16_to_cpu(neh->eh_entries) + m);
+		}
+		set_buffer_uptodate(bh);
+		unlock_buffer(bh);
+
+		if ((err = ext4_journal_dirty_metadata(handle, bh)))
+			goto cleanup;
+		brelse(bh);
+		bh = NULL;
+
+		/* correct old index */
+		if (m) {
+			err = ext4_ext_get_access(handle, inode, path + i);
+			if (err)
+				goto cleanup;
+			path[i].p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(path[i].p_hdr->eh_entries)-m);
+			err = ext4_ext_dirty(handle, inode, path + i);
+			if (err)
+				goto cleanup;
+		}
+
+		i--;
+	}
+
+	/* insert new index */
+	if (err)
+		goto cleanup;
+
+	err = ext4_ext_insert_index(handle, inode, path + at,
+				    le32_to_cpu(border), newblock);
+
+cleanup:
+	if (bh) {
+		if (buffer_locked(bh))
+			unlock_buffer(bh);
+		brelse(bh);
+	}
+
+	if (err) {
+		/* free all allocated blocks in error case */
+		for (i = 0; i < depth; i++) {
+			if (!ablocks[i])
+				continue;
+			ext4_free_blocks(handle, inode, ablocks[i], 1);
+		}
+	}
+	kfree(ablocks);
+
+	return err;
+}
+
+/*
+ * routine implements tree growing procedure:
+ *  - allocates new block
+ *  - moves top-level data (index block or leaf) into the new block
+ *  - initialize new top-level, creating index that points to the
+ *    just created block
+ */
+static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
+					struct ext4_ext_path *path,
+					struct ext4_extent *newext)
+{
+	struct ext4_ext_path *curp = path;
+	struct ext4_extent_header *neh;
+	struct ext4_extent_idx *fidx;
+	struct buffer_head *bh;
+	unsigned long newblock;
+	int err = 0;
+
+	newblock = ext4_ext_new_block(handle, inode, path, newext, &err);
+	if (newblock == 0)
+		return err;
+
+	bh = sb_getblk(inode->i_sb, newblock);
+	if (!bh) {
+		err = -EIO;
+		ext4_std_error(inode->i_sb, err);
+		return err;
+	}
+	lock_buffer(bh);
+
+	if ((err = ext4_journal_get_create_access(handle, bh))) {
+		unlock_buffer(bh);
+		goto out;
+	}
+
+	/* move top-level index/leaf into new block */
+	memmove(bh->b_data, curp->p_hdr, sizeof(EXT4_I(inode)->i_data));
+
+	/* set size of new block */
+	neh = ext_block_hdr(bh);
+	/* old root could have indexes or leaves
+	 * so calculate e_max right way */
+	if (ext_depth(inode))
+	  neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode));
+	else
+	  neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode));
+	neh->eh_magic = EXT4_EXT_MAGIC;
+	set_buffer_uptodate(bh);
+	unlock_buffer(bh);
+
+	if ((err = ext4_journal_dirty_metadata(handle, bh)))
+		goto out;
+
+	/* create index in new top-level index: num,max,pointer */
+	if ((err = ext4_ext_get_access(handle, inode, curp)))
+		goto out;
+
+	curp->p_hdr->eh_magic = EXT4_EXT_MAGIC;
+	curp->p_hdr->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode));
+	curp->p_hdr->eh_entries = cpu_to_le16(1);
+	curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr);
+	/* FIXME: it works, but actually path[0] can be index */
+	curp->p_idx->ei_block = EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block;
+	curp->p_idx->ei_leaf = cpu_to_le32(newblock);
+
+	neh = ext_inode_hdr(inode);
+	fidx = EXT_FIRST_INDEX(neh);
+	ext_debug("new root: num %d(%d), lblock %d, ptr %d\n",
+		  le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
+		  le32_to_cpu(fidx->ei_block), le32_to_cpu(fidx->ei_leaf));
+
+	neh->eh_depth = cpu_to_le16(path->p_depth + 1);
+	err = ext4_ext_dirty(handle, inode, curp);
+out:
+	brelse(bh);
+
+	return err;
+}
+
+/*
+ * routine finds empty index and adds new leaf. if no free index found
+ * then it requests in-depth growing
+ */
+static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode,
+					struct ext4_ext_path *path,
+					struct ext4_extent *newext)
+{
+	struct ext4_ext_path *curp;
+	int depth, i, err = 0;
+
+repeat:
+	i = depth = ext_depth(inode);
+
+	/* walk up to the tree and look for free index entry */
+	curp = path + depth;
+	while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) {
+		i--;
+		curp--;
+	}
+
+	/* we use already allocated block for index block
+	 * so, subsequent data blocks should be contigoues */
+	if (EXT_HAS_FREE_INDEX(curp)) {
+		/* if we found index with free entry, then use that
+		 * entry: create all needed subtree and add new leaf */
+		err = ext4_ext_split(handle, inode, path, newext, i);
+
+		/* refill path */
+		ext4_ext_drop_refs(path);
+		path = ext4_ext_find_extent(inode,
+					    le32_to_cpu(newext->ee_block),
+					    path);
+		if (IS_ERR(path))
+			err = PTR_ERR(path);
+	} else {
+		/* tree is full, time to grow in depth */
+		err = ext4_ext_grow_indepth(handle, inode, path, newext);
+		if (err)
+			goto out;
+
+		/* refill path */
+		ext4_ext_drop_refs(path);
+		path = ext4_ext_find_extent(inode,
+					    le32_to_cpu(newext->ee_block),
+					    path);
+		if (IS_ERR(path)) {
+			err = PTR_ERR(path);
+			goto out;
+		}
+
+		/*
+		 * only first (depth 0 -> 1) produces free space
+		 * in all other cases we have to split growed tree
+		 */
+		depth = ext_depth(inode);
+		if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) {
+			/* now we need split */
+			goto repeat;
+		}
+	}
+
+out:
+	return err;
+}
+
+/*
+ * returns allocated block in subsequent extent or EXT_MAX_BLOCK
+ * NOTE: it consider block number from index entry as
+ * allocated block. thus, index entries have to be consistent
+ * with leafs
+ */
+static unsigned long
+ext4_ext_next_allocated_block(struct ext4_ext_path *path)
+{
+	int depth;
+
+	BUG_ON(path == NULL);
+	depth = path->p_depth;
+
+	if (depth == 0 && path->p_ext == NULL)
+		return EXT_MAX_BLOCK;
+
+	while (depth >= 0) {
+		if (depth == path->p_depth) {
+			/* leaf */
+			if (path[depth].p_ext !=
+					EXT_LAST_EXTENT(path[depth].p_hdr))
+			  return le32_to_cpu(path[depth].p_ext[1].ee_block);
+		} else {
+			/* index */
+			if (path[depth].p_idx !=
+					EXT_LAST_INDEX(path[depth].p_hdr))
+			  return le32_to_cpu(path[depth].p_idx[1].ei_block);
+		}
+		depth--;
+	}
+
+	return EXT_MAX_BLOCK;
+}
+
+/*
+ * returns first allocated block from next leaf or EXT_MAX_BLOCK
+ */
+static unsigned ext4_ext_next_leaf_block(struct inode *inode,
+                                               struct ext4_ext_path *path)
+{
+	int depth;
+
+	BUG_ON(path == NULL);
+	depth = path->p_depth;
+
+	/* zero-tree has no leaf blocks at all */
+	if (depth == 0)
+		return EXT_MAX_BLOCK;
+
+	/* go to index block */
+	depth--;
+
+	while (depth >= 0) {
+		if (path[depth].p_idx !=
+				EXT_LAST_INDEX(path[depth].p_hdr))
+		  return le32_to_cpu(path[depth].p_idx[1].ei_block);
+		depth--;
+	}
+
+	return EXT_MAX_BLOCK;
+}
+
+/*
+ * if leaf gets modified and modified extent is first in the leaf
+ * then we have to correct all indexes above
+ * TODO: do we need to correct tree in all cases?
+ */
+int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode,
+				struct ext4_ext_path *path)
+{
+	struct ext4_extent_header *eh;
+	int depth = ext_depth(inode);
+	struct ext4_extent *ex;
+	__le32 border;
+	int k, err = 0;
+
+	eh = path[depth].p_hdr;
+	ex = path[depth].p_ext;
+	BUG_ON(ex == NULL);
+	BUG_ON(eh == NULL);
+
+	if (depth == 0) {
+		/* there is no tree at all */
+		return 0;
+	}
+
+	if (ex != EXT_FIRST_EXTENT(eh)) {
+		/* we correct tree if first leaf got modified only */
+		return 0;
+	}
+
+	/*
+	 * TODO: we need correction if border is smaller then current one
+	 */
+	k = depth - 1;
+	border = path[depth].p_ext->ee_block;
+	if ((err = ext4_ext_get_access(handle, inode, path + k)))
+		return err;
+	path[k].p_idx->ei_block = border;
+	if ((err = ext4_ext_dirty(handle, inode, path + k)))
+		return err;
+
+	while (k--) {
+		/* change all left-side indexes */
+		if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr))
+			break;
+		if ((err = ext4_ext_get_access(handle, inode, path + k)))
+			break;
+		path[k].p_idx->ei_block = border;
+		if ((err = ext4_ext_dirty(handle, inode, path + k)))
+			break;
+	}
+
+	return err;
+}
+
+static int inline
+ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
+				struct ext4_extent *ex2)
+{
+	/* FIXME: 48bit support */
+        if (le32_to_cpu(ex1->ee_block) + le16_to_cpu(ex1->ee_len)
+	    != le32_to_cpu(ex2->ee_block))
+		return 0;
+
+#ifdef AGRESSIVE_TEST
+	if (le16_to_cpu(ex1->ee_len) >= 4)
+		return 0;
+#endif
+
+        if (le32_to_cpu(ex1->ee_start) + le16_to_cpu(ex1->ee_len)
+			== le32_to_cpu(ex2->ee_start))
+		return 1;
+	return 0;
+}
+
+/*
+ * this routine tries to merge requsted extent into the existing
+ * extent or inserts requested extent as new one into the tree,
+ * creating new leaf in no-space case
+ */
+int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
+				struct ext4_ext_path *path,
+				struct ext4_extent *newext)
+{
+	struct ext4_extent_header * eh;
+	struct ext4_extent *ex, *fex;
+	struct ext4_extent *nearex; /* nearest extent */
+	struct ext4_ext_path *npath = NULL;
+	int depth, len, err, next;
+
+	BUG_ON(newext->ee_len == 0);
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	BUG_ON(path[depth].p_hdr == NULL);
+
+	/* try to insert block into found extent and return */
+	if (ex && ext4_can_extents_be_merged(inode, ex, newext)) {
+		ext_debug("append %d block to %d:%d (from %d)\n",
+				le16_to_cpu(newext->ee_len),
+				le32_to_cpu(ex->ee_block),
+				le16_to_cpu(ex->ee_len),
+				le32_to_cpu(ex->ee_start));
+		if ((err = ext4_ext_get_access(handle, inode, path + depth)))
+			return err;