Btrfs: Create extent_buffer interface for large blocksizes

Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/Makefile

@@ -23,4 +23,6 @@ modules_install:
 clean:
 	$(MAKE) -C $(KERNELDIR) M=`pwd` clean
 
+tester:
+	$(MAKE) -C $(KERNELDIR) M=`pwd` tree-defrag.o transaction.o sysfs.o super.o root-tree.o inode-map.o inode-item.o inode.o file-item.o file.o extent_map.o disk-io.o ctree.o dir-item.o extent-tree.o
 endif

fs/btrfs/dir-item.c

@@ -33,7 +33,7 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle
 	int ret;
 	char *ptr;
 	struct btrfs_item *item;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 
 	ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
 	if (ret == -EEXIST) {
@@ -49,11 +49,11 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle
 	if (ret < 0)
 		return ERR_PTR(ret);
 	WARN_ON(ret > 0);
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	item = leaf->items + path->slots[0];
+	leaf = path->nodes[0];
+	item = btrfs_item_nr(leaf, path->slots[0]);
 	ptr = btrfs_item_ptr(leaf, path->slots[0], char);
-	BUG_ON(data_size > btrfs_item_size(item));
-	ptr += btrfs_item_size(item) - data_size;
+	BUG_ON(data_size > btrfs_item_size(leaf, item));
+	ptr += btrfs_item_size(leaf, item) - data_size;
 	return (struct btrfs_dir_item *)ptr;
 }
 
@@ -65,12 +65,13 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret2 = 0;
 	struct btrfs_path *path;
 	struct btrfs_dir_item *dir_item;
-	char *name_ptr;
+	struct extent_buffer *leaf;
+	unsigned long name_ptr;
 	struct btrfs_key key;
+	struct btrfs_disk_key disk_key;
 	u32 data_size;
 
 	key.objectid = dir;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
 	ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
@@ -85,14 +86,16 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		goto out;
 	}
 
-	btrfs_cpu_key_to_disk(&dir_item->location, location);
-	btrfs_set_dir_type(dir_item, type);
-	btrfs_set_dir_flags(dir_item, 0);
-	btrfs_set_dir_name_len(dir_item, name_len);
-	name_ptr = (char *)(dir_item + 1);
+	leaf = path->nodes[0];
+	btrfs_cpu_key_to_disk(&disk_key, location);
+	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
+	btrfs_set_dir_type(leaf, dir_item, type);
+	btrfs_set_dir_flags(leaf, dir_item, 0);
+	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	name_ptr = (unsigned long)(dir_item + 1);
 
-	btrfs_memcpy(root, path->nodes[0]->b_data, name_ptr, name, name_len);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	write_extent_buffer(leaf, name, name_ptr, name_len);
+	btrfs_mark_buffer_dirty(leaf);
 
 second_insert:
 	/* FIXME, use some real flag for selecting the extra index */
@@ -110,13 +113,15 @@ second_insert:
 		ret2 = PTR_ERR(dir_item);
 		goto out;
 	}
-	btrfs_cpu_key_to_disk(&dir_item->location, location);
-	btrfs_set_dir_type(dir_item, type);
-	btrfs_set_dir_flags(dir_item, 0);
-	btrfs_set_dir_name_len(dir_item, name_len);
-	name_ptr = (char *)(dir_item + 1);
-	btrfs_memcpy(root, path->nodes[0]->b_data, name_ptr, name, name_len);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	leaf = path->nodes[0];
+	btrfs_cpu_key_to_disk(&disk_key, location);
+	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
+	btrfs_set_dir_type(leaf, dir_item, type);
+	btrfs_set_dir_flags(leaf, dir_item, 0);
+	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	name_ptr = (unsigned long)(dir_item + 1);
+	write_extent_buffer(leaf, name, name_ptr, name_len);
+	btrfs_mark_buffer_dirty(leaf);
 out:
 	btrfs_free_path(path);
 	if (ret)
@@ -136,14 +141,15 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	int ins_len = mod < 0 ? -1 : 0;
 	int cow = mod != 0;
-	struct btrfs_disk_key *found_key;
-	struct btrfs_leaf *leaf;
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
 
 	key.objectid = dir;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
+
 	ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
+
 	ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow);
 	if (ret < 0)
 		return ERR_PTR(ret);
@@ -152,12 +158,13 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 			return NULL;
 		path->slots[0]--;
 	}
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	found_key = &leaf->items[path->slots[0]].key;
 
-	if (btrfs_disk_key_objectid(found_key) != dir ||
-	    btrfs_disk_key_type(found_key) != BTRFS_DIR_ITEM_KEY ||
-	    btrfs_disk_key_offset(found_key) != key.offset)
+	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+
+	if (found_key.objectid != dir ||
+	    btrfs_key_type(&found_key) != BTRFS_DIR_ITEM_KEY ||
+	    found_key.offset != key.offset)
 		return NULL;
 
 	return btrfs_match_dir_item_name(root, path, name, name_len);
@@ -176,7 +183,6 @@ btrfs_lookup_dir_index_item(struct btrfs_trans_handle *trans,
 	int cow = mod != 0;
 
 	key.objectid = dir;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
 	key.offset = objectid;
 
@@ -193,21 +199,22 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
 			      const char *name, int name_len)
 {
 	struct btrfs_dir_item *dir_item;
-	char *name_ptr;
+	unsigned long name_ptr;
 	u32 total_len;
 	u32 cur = 0;
 	u32 this_len;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
+	leaf = path->nodes[0];
 	dir_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dir_item);
-	total_len = btrfs_item_size(leaf->items + path->slots[0]);
+	total_len = btrfs_item_size_nr(leaf, path->slots[0]);
 	while(cur < total_len) {
-		this_len = sizeof(*dir_item) + btrfs_dir_name_len(dir_item);
-		name_ptr = (char *)(dir_item + 1);
+		this_len = sizeof(*dir_item) +
+			btrfs_dir_name_len(leaf, dir_item);
+		name_ptr = (unsigned long)(dir_item + 1);
 
-		if (btrfs_dir_name_len(dir_item) == name_len &&
-		    memcmp(name_ptr, name, name_len) == 0)
+		if (btrfs_dir_name_len(leaf, dir_item) == name_len &&
+		    memcmp_extent_buffer(leaf, name, name_ptr, name_len) == 0)
 			return dir_item;
 
 		cur += this_len;
@@ -223,20 +230,23 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans,
 			      struct btrfs_dir_item *di)
 {
 
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u32 sub_item_len;
 	u32 item_len;
 	int ret = 0;
 
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	sub_item_len = sizeof(*di) + btrfs_dir_name_len(di);
-	item_len = btrfs_item_size(leaf->items + path->slots[0]);
-	if (sub_item_len == btrfs_item_size(leaf->items + path->slots[0])) {
+	leaf = path->nodes[0];
+	sub_item_len = sizeof(*di) + btrfs_dir_name_len(leaf, di);
+	item_len = btrfs_item_size_nr(leaf, path->slots[0]);
+	if (sub_item_len == item_len) {
 		ret = btrfs_del_item(trans, root, path);
 	} else {
-		char *ptr = (char *)di;
-		char *start = btrfs_item_ptr(leaf, path->slots[0], char);
-		btrfs_memmove(root, leaf, ptr, ptr + sub_item_len,
+		/* MARKER */
+		unsigned long ptr = (unsigned long)di;
+		unsigned long start;
+
+		start = btrfs_item_ptr_offset(leaf, path->slots[0]);
+		memmove_extent_buffer(leaf, ptr, ptr + sub_item_len,
 			item_len - (ptr + sub_item_len - start));
 		ret = btrfs_truncate_item(trans, root, path,
 					  item_len - sub_item_len);

fs/btrfs/disk-io.c

@@ -23,138 +23,132 @@
 #include <linux/swap.h>
 #include <linux/radix-tree.h>
 #include <linux/writeback.h>
+#include <linux/buffer_head.h> // for block_sync_page
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
 
-u64 bh_blocknr(struct buffer_head *bh)
+#if 0
+static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 {
-	return bh->b_blocknr;
-}
-
-static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
-{
-	struct btrfs_node *node = btrfs_buffer_node(buf);
-	if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
-		printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n",
-		       (unsigned long long)bh_blocknr(buf),
-		       (unsigned long long)btrfs_header_blocknr(&node->header));
+	if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
+		printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
+		       (unsigned long long)extent_buffer_blocknr(buf),
+		       (unsigned long long)btrfs_header_blocknr(buf));
 		return 1;
 	}
 	return 0;
 }
+#endif
 
-struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
+struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
+					    u64 blocknr)
 {
-	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
-	int blockbits = root->fs_info->sb->s_blocksize_bits;
-	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
-	struct page *page;
-	struct buffer_head *bh;
-	struct buffer_head *head;
-	struct buffer_head *ret = NULL;
-
-
-	page = find_lock_page(mapping, index);
-	if (!page)
-		return NULL;
-
-	if (!page_has_buffers(page))
-		goto out_unlock;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
+				   blocknr * root->sectorsize,
+				   root->sectorsize, GFP_NOFS);
+}
 
-	head = page_buffers(page);
-	bh = head;
-	do {
-		if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
-			ret = bh;
-			get_bh(bh);
-			goto out_unlock;
-		}
-		bh = bh->b_this_page;
-	} while (bh != head);
-out_unlock:
-	unlock_page(page);
-	page_cache_release(page);
-	return ret;
+struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
+						 u64 blocknr)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
+				   blocknr * root->sectorsize,
+				   root->sectorsize, GFP_NOFS);
 }
 
-int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
-			     u64 logical)
+struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
+				    size_t page_offset, u64 start, u64 end,
+				    int create)
 {
-	if (logical == 0) {
-		bh->b_bdev = NULL;
-		bh->b_blocknr = 0;
-		set_buffer_mapped(bh);
-	} else {
-		map_bh(bh, root->fs_info->sb, logical);
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_map *em;
+	int ret;
+
+again:
+	em = lookup_extent_mapping(em_tree, start, end);
+	if (em) {
+		goto out;
 	}
-	return 0;
+	em = alloc_extent_map(GFP_NOFS);
+	if (!em) {
+		em = ERR_PTR(-ENOMEM);
+		goto out;
+	}
+	em->start = 0;
+	em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1;
+	em->block_start = 0;
+	em->block_end = em->end;
+	em->bdev = inode->i_sb->s_bdev;
+	ret = add_extent_mapping(em_tree, em);
+	if (ret == -EEXIST) {
+		free_extent_map(em);
+		em = NULL;
+		goto again;
+	} else if (ret) {
+		em = ERR_PTR(ret);
+	}
+out:
+	return em;
 }
 
-struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 blocknr)
+static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
-	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
-	int blockbits = root->fs_info->sb->s_blocksize_bits;
-	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
-	struct page *page;
-	struct buffer_head *bh;
-	struct buffer_head *head;
-	struct buffer_head *ret = NULL;
-	int err;
-	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	return extent_write_full_page(tree, page, btree_get_extent, wbc);
+}
+int btree_readpage(struct file *file, struct page *page)
+{
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	return extent_read_full_page(tree, page, btree_get_extent);
+}
 
-	page = find_or_create_page(mapping, index, GFP_NOFS);
-	if (!page)
-		return NULL;
+static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
+{
+	struct extent_map_tree *tree;
+	int ret;
 
-	if (!page_has_buffers(page))
-		create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
-	head = page_buffers(page);
-	bh = head;
-	do {
-		if (!buffer_mapped(bh)) {
-			err = btrfs_map_bh_to_logical(root, bh, first_block);
-			BUG_ON(err);
-		}
-		if (bh_blocknr(bh) == blocknr) {
-			ret = bh;
-			get_bh(bh);
-			goto out_unlock;
-		}
-		bh = bh->b_this_page;
-		first_block++;
-	} while (bh != head);
-out_unlock:
-	unlock_page(page);
-	if (ret)
-		touch_buffer(ret);
-	page_cache_release(page);
+	BUG_ON(page->private != 1);
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	ret = try_release_extent_mapping(tree, page);
+	if (ret == 1) {
+		ClearPagePrivate(page);
+		set_page_private(page, 0);
+		page_cache_release(page);
+	}
 	return ret;
 }
 
-static int btree_get_block(struct inode *inode, sector_t iblock,
-			   struct buffer_head *bh, int create)
+static void btree_invalidatepage(struct page *page, unsigned long offset)
 {
-	int err;
-	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
-	err = btrfs_map_bh_to_logical(root, bh, iblock);
-	return err;
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	extent_invalidatepage(tree, page, offset);
+	btree_releasepage(page, GFP_NOFS);
 }
 
 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 		    char *result)
 {
+	return 0;
+#if 0
 	u32 crc;
 	crc = crc32c(0, data, len);
 	memcpy(result, &crc, BTRFS_CRC32_SIZE);
 	return 0;
+#endif
 }
 
-static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
+#if 0
+static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 			   int verify)
 {
+	return 0;
 	char result[BTRFS_CRC32_SIZE];
 	int ret;
 	struct btrfs_node *node;
@@ -176,7 +170,9 @@ static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
 	}
 	return 0;
 }
+#endif
 
+#if 0
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
 	struct buffer_head *bh;
@@ -195,87 +191,65 @@ static int btree_writepage(struct page *page, struct writeback_control *wbc)
 	} while (bh != head);
 	return block_write_full_page(page, btree_get_block, wbc);
 }
-
-static int btree_readpage(struct file * file, struct page * page)
-{
-	return block_read_full_page(page, btree_get_block);
-}
+#endif
 
 static struct address_space_operations btree_aops = {
 	.readpage	= btree_readpage,
 	.writepage	= btree_writepage,
+	.releasepage	= btree_releasepage,
+	.invalidatepage = btree_invalidatepage,
 	.sync_page	= block_sync_page,
 };
 
 int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct buffer_head *bh = NULL;
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
 	int ret = 0;
 
-	bh = btrfs_find_create_tree_block(root, blocknr);
-	if (!bh)
+	buf = btrfs_find_create_tree_block(root, blocknr);
+	if (!buf)
 		return 0;
-	if (buffer_uptodate(bh)) {
-		ret = 1;
-		goto done;
-	}
-	if (test_set_buffer_locked(bh)) {
-		ret = 1;
-		goto done;
-	}
-	if (!buffer_uptodate(bh)) {
-		get_bh(bh);
-		bh->b_end_io = end_buffer_read_sync;
-		submit_bh(READ, bh);
-	} else {
-		unlock_buffer(bh);
-		ret = 1;
-	}
-done:
-	brelse(bh);
+	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
+				 buf, 0);
+	free_extent_buffer(buf);
 	return ret;
 }
 
-struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct buffer_head *bh = NULL;
-
-	bh = btrfs_find_create_tree_block(root, blocknr);
-	if (!bh)
-		return bh;
-	if (buffer_uptodate(bh))
-		goto uptodate;
-	lock_buffer(bh);
-	if (!buffer_uptodate(bh)) {
-		get_bh(bh);
-		bh->b_end_io = end_buffer_read_sync;
-		submit_bh(READ, bh);
-		wait_on_buffer(bh);
-		if (!buffer_uptodate(bh))
-			goto fail;
-	} else {
-		unlock_buffer(bh);
-	}
-uptodate:
-	if (!buffer_checked(bh)) {
-		csum_tree_block(root, bh, 1);
-		set_buffer_checked(bh);
-	}
-	if (check_tree_block(root, bh))
-		goto fail;
-	return bh;
-fail:
-	brelse(bh);
-	return NULL;
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+
+	buf = btrfs_find_create_tree_block(root, blocknr);
+	if (!buf)
+		return NULL;
+	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
+				 buf, 1);
+	return buf;
 }
 
 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf)
+		     struct extent_buffer *buf)
 {
-	WARN_ON(atomic_read(&buf->b_count) == 0);
-	lock_buffer(buf);
-	clear_buffer_dirty(buf);
-	unlock_buffer(buf);
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
+	return 0;
+}
+
+int wait_on_tree_block_writeback(struct btrfs_root *root,
+				 struct extent_buffer *buf)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree,
+					buf);
+	return 0;
+}
+
+int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
 	return 0;
 }
 
@@ -287,7 +261,9 @@ static int __setup_root(int blocksize,
 	root->node = NULL;
 	root->inode = NULL;
 	root->commit_root = NULL;
-	root->blocksize = blocksize;
+	root->sectorsize = blocksize;
+	root->nodesize = blocksize;
+	root->leafsize = blocksize;
 	root->ref_cows = 0;
 	root->fs_info = fs_info;
 	root->objectid = objectid;
@@ -332,7 +308,7 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 	struct btrfs_root *root;
 	struct btrfs_root *tree_root = fs_info->tree_root;
 	struct btrfs_path *path;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	u64 highest_inode;
 	int ret = 0;
 
@@ -361,11 +337,10 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 			ret = -ENOENT;
 		goto out;
 	}
-	l = btrfs_buffer_leaf(path->nodes[0]);
-	memcpy(&root->root_item,
-	       btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
+	l = path->nodes[0];
+	read_extent_buffer(l, &root->root_item,
+	       btrfs_item_ptr_offset(l, path->slots[0]),
 	       sizeof(root->root_item));
-	memcpy(&root->root_key, location, sizeof(*location));
 	ret = 0;
 out:
 	btrfs_release_path(root, path);
@@ -406,21 +381,21 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				(unsigned long)root->root_key.objectid,
 				root);
 	if (ret) {
-		brelse(root->node);
+		free_extent_buffer(root->node);
 		kfree(root);
 		return ERR_PTR(ret);
 	}
 
 	ret = btrfs_set_root_name(root, name, namelen);
 	if (ret) {
-		brelse(root->node);
+		free_extent_buffer(root->node);
 		kfree(root);
 		return ERR_PTR(ret);
 	}
 
 	ret = btrfs_sysfs_add_root(root);
 	if (ret) {
-		brelse(root->node);
+		free_extent_buffer(root->node);
 		kfree(root->name);
 		kfree(root);
 		return ERR_PTR(ret);
@@ -471,6 +446,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->btree_inode->i_nlink = 1;
 	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
+	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
+			     fs_info->btree_inode->i_mapping,
+			     GFP_NOFS);
 	fs_info->do_barriers = 1;
 	fs_info->closing = 0;
 
@@ -493,10 +471,14 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 
 	if (!fs_info->sb_buffer)
 		goto fail_iput;
-	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
-	fs_info->disk_super = disk_super;
-	memcpy(&fs_info->super_copy, disk_super, sizeof(fs_info->super_copy));
 
+	read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
+			   sizeof(fs_info->super_copy));
+
+	read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
+			   (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
+			   BTRFS_FSID_SIZE);
+	disk_super = &fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
@@ -530,9 +512,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	return tree_root;
 
 fail_tree_root:
-	btrfs_block_release(tree_root, tree_root->node);
+	free_extent_buffer(tree_root->node);
 fail_sb_buffer:
-	btrfs_block_release(tree_root, fs_info->sb_buffer);
+	free_extent_buffer(fs_info->sb_buffer);
 fail_iput:
 	iput(fs_info->btree_inode);
 fail:
@@ -546,31 +528,13 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root)
 {
 	int ret;
-	struct buffer_head *bh = root->fs_info->sb_buffer;
-
-	lock_buffer(bh);
-	WARN_ON(atomic_read(&bh->b_count) < 1);
-	clear_buffer_dirty(bh);
-	csum_tree_block(root, bh, 0);
-	bh->b_end_io = end_buffer_write_sync;
-	get_bh(bh);
-	if (root->fs_info->do_barriers)
-		ret = submit_bh(WRITE_BARRIER, bh);
-	else
-		ret = submit_bh(WRITE, bh);
-	if (ret == -EOPNOTSUPP) {
-		get_bh(bh);
-		lock_buffer(bh);
-		set_buffer_uptodate(bh);
-		root->fs_info->do_barriers = 0;
-		ret = submit_bh(WRITE, bh);
-	}
-	wait_on_buffer(bh);
-	if (!buffer_uptodate(bh)) {
-		WARN_ON(1);
-		return -EIO;
-	}
-	return 0;
+	struct extent_buffer *super = root->fs_info->sb_buffer;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super);
+	ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
+				     super->start, super->len);
+	return ret;
 }
 
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
@@ -581,9 +545,9 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 	if (root->inode)
 		iput(root->inode);
 	if (root->node)
-		brelse(root->node);
+		free_extent_buffer(root->node);
 	if (root->commit_root)
-		brelse(root->commit_root);
+		free_extent_buffer(root->commit_root);
 	if (root->name)
 		kfree(root->name);
 	kfree(root);
@@ -629,12 +593,10 @@ int close_ctree(struct btrfs_root *root)
 	mutex_unlock(&fs_info->fs_mutex);
 
 	if (fs_info->extent_root->node)
-		btrfs_block_release(fs_info->extent_root,
-				    fs_info->extent_root->node);
+		free_extent_buffer(fs_info->extent_root->node);
 	if (fs_info->tree_root->node)
-		btrfs_block_release(fs_info->tree_root,
-				    fs_info->tree_root->node);
-	btrfs_block_release(root, fs_info->sb_buffer);
+		free_extent_buffer(fs_info->tree_root->node);
+	free_extent_buffer(fs_info->sb_buffer);
 	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
 	iput(fs_info->btree_inode);
 
@@ -645,25 +607,32 @@ int close_ctree(struct btrfs_root *root)
 	return 0;
 }
 
-void btrfs_mark_buffer_dirty(struct buffer_head *bh)
+int btrfs_buffer_uptodate(struct extent_buffer *buf)
+{
+	struct inode *btree_inode = buf->pages[0]->mapping->host;
+	return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf);
+}
+
+int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
 {
-	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
-	u64 transid = btrfs_header_generation(btrfs_buffer_header(bh));
+	struct inode *btree_inode = buf->pages[0]->mapping->host;
+	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree,
+					  buf);
+}
 
-	WARN_ON(!atomic_read(&bh->b_count));
+void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	u64 transid = btrfs_header_generation(buf);
+	struct inode *btree_inode = root->fs_info->btree_inode;
 
 	if (transid != root->fs_info->generation) {
 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
-			(unsigned long long)bh->b_blocknr,
+			(unsigned long long)extent_buffer_blocknr(buf),
 			transid, root->fs_info->generation);
 		WARN_ON(1);
 	}
-	mark_buffer_dirty(bh);
-}
-
-void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
-{
-	brelse(buf);
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
 }
 
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)

fs/btrfs/disk-io.h

@@ -19,68 +19,35 @@
 #ifndef __DISKIO__
 #define __DISKIO__
 
-#include <linux/buffer_head.h>
-
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 
-enum btrfs_bh_state_bits {
-	BH_Checked = BH_PrivateStart,
-	BH_Defrag,
-	BH_DefragDone,
-};
-BUFFER_FNS(Checked, checked);
-BUFFER_FNS(Defrag, defrag);
-BUFFER_FNS(DefragDone, defrag_done);
-
-static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh)
-{
-	return (struct btrfs_node *)bh->b_data;
-}
-
-static inline struct btrfs_leaf *btrfs_buffer_leaf(struct buffer_head *bh)
-{
-	return (struct btrfs_leaf *)bh->b_data;
-}
-
-static inline struct btrfs_header *btrfs_buffer_header(struct buffer_head *bh)
-{
-	return &((struct btrfs_node *)bh->b_data)->header;
-}
-
-struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr);
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
 int readahead_tree_block(struct btrfs_root *root, u64 blocknr);
-struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 blocknr);
-int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf);
-int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf);
+struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
+						   u64 blocknr);
 int clean_tree_block(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root, struct buffer_head *buf);
-int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root);
+		     struct btrfs_root *root, struct extent_buffer *buf);
 struct btrfs_root *open_ctree(struct super_block *sb);
 int close_ctree(struct btrfs_root *root);
-void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
-struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
-int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
-		    char *result);
+struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
+					    u64 blocknr);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);
 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 					       struct btrfs_key *location);
-u64 bh_blocknr(struct buffer_head *bh);
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
 			   u64 device_id,
 			   u64 block_start,
 			   u64 num_blocks);
-int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
-			     u64 logical);
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
-void btrfs_mark_buffer_dirty(struct buffer_head *bh);
+void btrfs_mark_buffer_dirty(struct extent_buffer *buf);
+int btrfs_buffer_uptodate(struct extent_buffer *buf);
+int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
+int wait_on_tree_block_writeback(struct btrfs_root *root,
+				 struct extent_buffer *buf);
 #endif

fs/btrfs/extent-tree.c

@@ -33,7 +33,7 @@ static int cache_block_group(struct btrfs_root *root,
 	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct radix_tree_root *extent_radix;
 	int slot;
 	u64 i;
@@ -56,7 +56,6 @@ static int cache_block_group(struct btrfs_root *root,
 	path->reada = 2;
 	first_free = block_group->key.objectid;
 	key.objectid = block_group->key.objectid;
-	key.flags = 0;
 	key.offset = 0;
 
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
@@ -69,9 +68,9 @@ static int cache_block_group(struct btrfs_root *root,
 		path->slots[0]--;
 
 	while(1) {
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
+		leaf = path->nodes[0];
 		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(&leaf->header)) {
+		if (slot >= btrfs_header_nritems(leaf)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret < 0)
 				goto err;
@@ -82,7 +81,7 @@ static int cache_block_group(struct btrfs_root *root,
 			}
 		}
 
-		btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (key.objectid < block_group->key.objectid) {
 			if (key.objectid + key.offset > first_free)
 				first_free = key.objectid + key.offset;
@@ -116,8 +115,7 @@ next:
 		hole_size = block_group->key.objectid +
 			block_group->key.offset - last;
 		for (i = 0; i < hole_size; i++) {
-			set_radix_bit(extent_radix,
-					last + i);
+			set_radix_bit(extent_radix, last + i);
 		}
 	}
 	block_group->cached = 1;
@@ -366,7 +364,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_extent_item *item;
 	u32 refs;
 
@@ -375,7 +373,6 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 		return -ENOMEM;
 
 	key.objectid = blocknr;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = num_blocks;
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
@@ -386,10 +383,10 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 		BUG();
 	}
 	BUG_ON(ret != 0);
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	l = path->nodes[0];
 	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
-	refs = btrfs_extent_refs(item);
-	btrfs_set_extent_refs(item, refs + 1);
+	refs = btrfs_extent_refs(l, item);
+	btrfs_set_extent_refs(l, item, refs + 1);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 
 	btrfs_release_path(root->fs_info->extent_root, path);
@@ -414,23 +411,25 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_extent_item *item;
 
 	path = btrfs_alloc_path();
 	key.objectid = blocknr;
 	key.offset = num_blocks;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 				0, 0);
 	if (ret < 0)
 		goto out;
-	if (ret != 0)
+	if (ret != 0) {
+		btrfs_print_leaf(root, path->nodes[0]);
+		printk("failed to find block number %Lu\n", blocknr);
 		BUG();
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	}
+	l = path->nodes[0];
 	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
-	*refs = btrfs_extent_refs(item);
+	*refs = btrfs_extent_refs(l, item);
 out:
 	btrfs_free_path(path);
 	return 0;
@@ -439,16 +438,16 @@ out:
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root)
 {
-	return btrfs_inc_extent_ref(trans, root, bh_blocknr(root->node), 1);
+	return btrfs_inc_extent_ref(trans, root,
+				    extent_buffer_blocknr(root->node), 1);
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct buffer_head *buf)
+		  struct extent_buffer *buf)
 {
 	u64 blocknr;
-	struct btrfs_node *buf_node;
-	struct btrfs_leaf *buf_leaf;
-	struct btrfs_disk_key *key;
+	u32 nritems;
+	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	int i;
 	int leaf;
@@ -458,31 +457,31 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 
 	if (!root->ref_cows)
 		return 0;
-	buf_node = btrfs_buffer_node(buf);
-	leaf = btrfs_is_leaf(buf_node);
-	buf_leaf = btrfs_buffer_leaf(buf);
-	for (i = 0; i < btrfs_header_nritems(&buf_node->header); i++) {
+
+	leaf = btrfs_is_leaf(buf);
+	nritems = btrfs_header_nritems(buf);
+	for (i = 0; i < nritems; i++) {
 		if (leaf) {
 			u64 disk_blocknr;
-			key = &buf_leaf->items[i].key;
-			if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+			btrfs_item_key_to_cpu(buf, &key, i);
+			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
-			fi = btrfs_item_ptr(buf_leaf, i,
+			fi = btrfs_item_ptr(buf, i,
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) ==
+			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
 			if (disk_blocknr == 0)
 				continue;
 			ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(fi));
+				    btrfs_file_extent_disk_num_blocks(buf, fi));
 			if (ret) {
 				faili = i;
 				goto fail;
 			}
 		} else {
-			blocknr = btrfs_node_blockptr(buf_node, i);
+			blocknr = btrfs_node_blockptr(buf, i);
 			ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
 			if (ret) {
 				faili = i;
@@ -496,22 +495,23 @@ fail:
 	for (i =0; i < faili; i++) {
 		if (leaf) {
 			u64 disk_blocknr;
-			key = &buf_leaf->items[i].key;
-			if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+			btrfs_item_key_to_cpu(buf, &key, i);
+			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
-			fi = btrfs_item_ptr(buf_leaf, i,
+			fi = btrfs_item_ptr(buf, i,
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) ==
+			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
 			if (disk_blocknr == 0)
 				continue;
 			err = btrfs_free_extent(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(fi), 0);
+				    btrfs_file_extent_disk_num_blocks(buf,
+								      fi), 0);
 			BUG_ON(err);
 		} else {
-			blocknr = btrfs_node_blockptr(buf_node, i);
+			blocknr = btrfs_node_blockptr(buf, i);
 			err = btrfs_free_extent(trans, root, blocknr, 1, 0);
 			BUG_ON(err);
 		}
@@ -527,16 +527,18 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans,
 	int ret;
 	int pending_ret;
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
-	struct btrfs_block_group_item *bi;
+	unsigned long bi;
+	struct extent_buffer *leaf;
 
 	ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
 	if (ret < 0)
 		goto fail;
 	BUG_ON(ret);
-	bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
-			    struct btrfs_block_group_item);
-	memcpy(bi, &cache->item, sizeof(*bi));
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+
+	leaf = path->nodes[0];
+	bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
+	write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
+	btrfs_mark_buffer_dirty(leaf);
 	btrfs_release_path(extent_root, path);
 fail:
 	finish_current_insert(trans, extent_root);
@@ -768,11 +770,11 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 	unsigned long gang[8];
 	struct btrfs_fs_info *info = extent_root->fs_info;
 
-	btrfs_set_extent_refs(&extent_item, 1);
+	btrfs_set_stack_extent_refs(&extent_item, 1);
 	ins.offset = 1;
-	ins.flags = 0;
 	btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
-	btrfs_set_extent_owner(&extent_item, extent_root->root_key.objectid);
+	btrfs_set_stack_extent_owner(&extent_item,
+				     extent_root->root_key.objectid);
 
 	while(1) {
 		ret = find_first_radix_bit(&info->extent_ins_radix, gang, 0,
@@ -795,23 +797,20 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
 {
 	int err;
-	struct btrfs_header *header;
-	struct buffer_head *bh;
+	struct extent_buffer *buf;
 
 	if (!pending) {
-		bh = btrfs_find_tree_block(root, blocknr);
-		if (bh) {
-			if (buffer_uptodate(bh)) {
+		buf = btrfs_find_tree_block(root, blocknr);
+		if (buf) {
+			if (btrfs_buffer_uptodate(buf)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
-				header = btrfs_buffer_header(bh);
-				if (btrfs_header_generation(header) ==
-				    transid) {
-					btrfs_block_release(root, bh);
+				if (btrfs_header_generation(buf) == transid) {
+					free_extent_buffer(buf);
 					return 0;
 				}
 			}
-			btrfs_block_release(root, bh);
+			free_extent_buffer(buf);
 		}
 		err = set_radix_bit(&root->fs_info->pinned_radix, blocknr);
 		if (!err) {
@@ -839,12 +838,12 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_key key;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_root *extent_root = info->extent_root;
+	struct extent_buffer *leaf;
 	int ret;
 	struct btrfs_extent_item *ei;
 	u32 refs;
 
 	key.objectid = blocknr;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = num_blocks;
 
@@ -856,12 +855,16 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret < 0)
 		return ret;
 	BUG_ON(ret);
-	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
 			    struct btrfs_extent_item);
-	BUG_ON(ei->refs == 0);
-	refs = btrfs_extent_refs(ei) - 1;
-	btrfs_set_extent_refs(ei, refs);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	refs = btrfs_extent_refs(leaf, ei);
+	BUG_ON(refs == 0);
+	refs -= 1;
+	btrfs_set_extent_refs(leaf, ei, refs);
+	btrfs_mark_buffer_dirty(leaf);
+
 	if (refs == 0) {
 		u64 super_blocks_used, root_blocks_used;
 
@@ -876,8 +879,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 					    super_blocks_used - num_blocks);
 
 		/* block accounting for root item */
-		root_blocks_used = btrfs_root_blocks_used(&root->root_item);
-		btrfs_set_root_blocks_used(&root->root_item,
+		root_blocks_used = btrfs_root_used(&root->root_item);
+		btrfs_set_root_used(&root->root_item,
 					   root_blocks_used - num_blocks);
 
 		ret = btrfs_del_item(trans, extent_root, path);
@@ -984,7 +987,7 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	u64 test_block;
 	u64 orig_search_start = search_start;
 	int start_found;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	struct btrfs_fs_info *info = root->fs_info;
 	int total_needed = num_blocks;
@@ -994,10 +997,10 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	int wrapped = 0;
 
 	WARN_ON(num_blocks < 1);
-	ins->flags = 0;
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 
-	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	level = btrfs_header_level(root->node);
+
 	if (search_end == (u64)-1)
 		search_end = btrfs_super_total_blocks(&info->super_copy);
 	if (hint_block) {
@@ -1034,8 +1037,9 @@ check_failed:
 		path->slots[0]--;
 	}
 
-	l = btrfs_buffer_leaf(path->nodes[0]);
-	btrfs_disk_key_to_cpu(&key, &l->items[path->slots[0]].key);
+	l = path->nodes[0];
+	btrfs_item_key_to_cpu(l, &key, path->slots[0]);
+
 	/*
 	 * a rare case, go back one key if we hit a block group item
 	 * instead of an extent item
@@ -1055,9 +1059,9 @@ check_failed:
 	}
 
 	while (1) {
-		l = btrfs_buffer_leaf(path->nodes[0]);
+		l = path->nodes[0];
 		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(&l->header)) {
+		if (slot >= btrfs_header_nritems(l)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret == 0)
 				continue;
@@ -1075,7 +1079,7 @@ check_failed:
 			goto check_pending;
 		}
 
-		btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
+		btrfs_item_key_to_cpu(l, &key, slot);
 		if (key.objectid >= search_start && key.objectid > last_block &&
 		    start_found) {
 			if (last_block < search_start)
@@ -1183,8 +1187,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_root *extent_root = info->extent_root;
 	struct btrfs_extent_item extent_item;
 
-	btrfs_set_extent_refs(&extent_item, 1);
-	btrfs_set_extent_owner(&extent_item, owner);
+	btrfs_set_stack_extent_refs(&extent_item, 1);
+	btrfs_set_stack_extent_owner(&extent_item, owner);
 
 	WARN_ON(num_blocks < 1);
 	ret = find_free_extent(trans, root, num_blocks, empty_size,
@@ -1201,8 +1205,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 				    num_blocks);
 
 	/* block accounting for root item */
-	root_blocks_used = btrfs_root_blocks_used(&root->root_item);
-	btrfs_set_root_blocks_used(&root->root_item, root_blocks_used +
+	root_blocks_used = btrfs_root_used(&root->root_item);
+	btrfs_set_root_used(&root->root_item, root_blocks_used +
 				   num_blocks);
 
 	if (root == extent_root) {
@@ -1241,13 +1245,13 @@ update_block:
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
  */
-struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					   struct btrfs_root *root, u64 hint,
-					   u64 empty_size)
+struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+					     struct btrfs_root *root, u64 hint,
+					     u64 empty_size)
 {
 	struct btrfs_key ins;
 	int ret;
-	struct buffer_head *buf;
+	struct extent_buffer *buf;
 
 	ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
 				 1, empty_size, hint, (u64)-1, &ins, 0);
@@ -1260,53 +1264,57 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 		btrfs_free_extent(trans, root, ins.objectid, 1, 0);
 		return ERR_PTR(-ENOMEM);
 	}
-	WARN_ON(buffer_dirty(buf));
-	set_buffer_uptodate(buf);
+	btrfs_set_buffer_uptodate(buf);
+	set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
+			 buf->start + buf->len - 1, GFP_NOFS);
+	/*
 	set_buffer_checked(buf);
 	set_buffer_defrag(buf);
-	set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
+	*/
+	/* FIXME!!!!!!!!!!!!!!!!
+	set_radix_bit(&trans->transaction->dirty_pages, buf->pages[0]->index);
+	*/
 	trans->blocks_used++;
 	return buf;
 }
 
 static int drop_leaf_ref(struct btrfs_trans_handle *trans,
-			 struct btrfs_root *root, struct buffer_head *cur)
+			 struct btrfs_root *root, struct extent_buffer *leaf)
 {
-	struct btrfs_disk_key *key;
-	struct btrfs_leaf *leaf;
+	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	int i;
 	int nritems;
 	int ret;
 
-	BUG_ON(!btrfs_is_leaf(btrfs_buffer_node(cur)));
-	leaf = btrfs_buffer_leaf(cur);
-	nritems = btrfs_header_nritems(&leaf->header);
+	BUG_ON(!btrfs_is_leaf(leaf));
+	nritems = btrfs_header_nritems(leaf);
 	for (i = 0; i < nritems; i++) {
 		u64 disk_blocknr;
-		key = &leaf->items[i].key;
-		if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+
+		btrfs_item_key_to_cpu(leaf, &key, i);
+		if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 			continue;
 		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
-		if (btrfs_file_extent_type(fi) == BTRFS_FILE_EXTENT_INLINE)
+		if (btrfs_file_extent_type(leaf, fi) ==
+		    BTRFS_FILE_EXTENT_INLINE)
 			continue;
 		/*
 		 * FIXME make sure to insert a trans record that
 		 * repeats the snapshot del on crash
 		 */
-		disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+		disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi);
 		if (disk_blocknr == 0)
 			continue;
 		ret = btrfs_free_extent(trans, root, disk_blocknr,
-					btrfs_file_extent_disk_num_blocks(fi),
-					0);
+				btrfs_file_extent_disk_num_blocks(leaf, fi), 0);
 		BUG_ON(ret);
 	}
 	return 0;
 }
 
 static void reada_walk_down(struct btrfs_root *root,
-			    struct btrfs_node *node)
+			    struct extent_buffer *node)
 {
 	int i;
 	u32 nritems;
@@ -1314,7 +1322,7 @@ static void reada_walk_down(struct btrfs_root *root,
 	int ret;
 	u32 refs;
 
-	nritems = btrfs_header_nritems(&node->header);
+	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
 		blocknr = btrfs_node_blockptr(node, i);
 		ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
@@ -1337,16 +1345,17 @@ static void reada_walk_down(struct btrfs_root *root,
 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, struct btrfs_path *path, int *level)
 {
-	struct buffer_head *next;
-	struct buffer_head *cur;
+	struct extent_buffer *next;
+	struct extent_buffer *cur;
 	u64 blocknr;
 	int ret;
 	u32 refs;
 
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
-	ret = lookup_extent_ref(trans, root, bh_blocknr(path->nodes[*level]),
-			       1, &refs);
+	ret = lookup_extent_ref(trans, root,
+				extent_buffer_blocknr(path->nodes[*level]),
+				1, &refs);
 	BUG_ON(ret);
 	if (refs > 1)
 		goto out;
@@ -1360,21 +1369,20 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 		cur = path->nodes[*level];
 
 		if (*level > 0 && path->slots[*level] == 0)
-			reada_walk_down(root, btrfs_buffer_node(cur));
+			reada_walk_down(root, cur);
 
-		if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
+		if (btrfs_header_level(cur) != *level)
 			WARN_ON(1);
 
 		if (path->slots[*level] >=
-		    btrfs_header_nritems(btrfs_buffer_header(cur)))
+		    btrfs_header_nritems(cur))
 			break;
 		if (*level == 0) {
 			ret = drop_leaf_ref(trans, root, cur);
 			BUG_ON(ret);
 			break;
 		}
-		blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur),
-					      path->slots[*level]);
+		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
 		ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
 		BUG_ON(ret);
 		if (refs != 1) {
@@ -1384,8 +1392,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			continue;
 		}
 		next = btrfs_find_tree_block(root, blocknr);
-		if (!next || !buffer_uptodate(next)) {
-			brelse(next);
+		if (!next || !btrfs_buffer_uptodate(next)) {
+			free_extent_buffer(next);
 			mutex_unlock(&root->fs_info->fs_mutex);
 			next = read_tree_block(root, blocknr);
 			mutex_lock(&root->fs_info->fs_mutex);
@@ -1395,7 +1403,7 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			BUG_ON(ret);
 			if (refs != 1) {
 				path->slots[*level]++;
-				brelse(next);
+				free_extent_buffer(next);
 				ret = btrfs_free_extent(trans, root,
 							blocknr, 1, 1);
 				BUG_ON(ret);
@@ -1404,17 +1412,17 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 		}
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
-			btrfs_block_release(root, path->nodes[*level-1]);
+			free_extent_buffer(path->nodes[*level-1]);
 		path->nodes[*level-1] = next;
-		*level = btrfs_header_level(btrfs_buffer_header(next));
+		*level = btrfs_header_level(next);
 		path->slots[*level] = 0;
 	}
 out:
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 	ret = btrfs_free_extent(trans, root,
-				bh_blocknr(path->nodes[*level]), 1, 1);
-	btrfs_block_release(root, path->nodes[*level]);
+			extent_buffer_blocknr(path->nodes[*level]), 1, 1);
+	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
 	BUG_ON(ret);
@@ -1436,24 +1444,24 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
-		if (slot < btrfs_header_nritems(
-		    btrfs_buffer_header(path->nodes[i])) - 1) {
-			struct btrfs_node *node;
-			node = btrfs_buffer_node(path->nodes[i]);
+		if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
+			struct extent_buffer *node;
+			struct btrfs_disk_key disk_key;
+			node = path->nodes[i];
 			path->slots[i]++;
 			*level = i;
 			WARN_ON(*level == 0);
+			btrfs_node_key(node, &disk_key, path->slots[i]);
 			memcpy(&root_item->drop_progress,
-			       &node->ptrs[path->slots[i]].key,
-			       sizeof(root_item->drop_progress));
+			       &disk_key, sizeof(disk_key));
 			root_item->drop_level = i;
 			return 0;
 		} else {
 			ret = btrfs_free_extent(trans, root,
-						bh_blocknr(path->nodes[*level]),
-						1, 1);
+				    extent_buffer_blocknr(path->nodes[*level]),
+				    1, 1);
 			BUG_ON(ret);
-			btrfs_block_release(root, path->nodes[*level]);
+			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
 		}
@@ -1480,15 +1488,15 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 
-	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	level = btrfs_header_level(root->node);
 	orig_level = level;
 	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
 		path->nodes[level] = root->node;
 		path->slots[level] = 0;
 	} else {
 		struct btrfs_key key;
-		struct btrfs_disk_key *found_key;
-		struct btrfs_node *node;
+		struct btrfs_disk_key found_key;
+		struct extent_buffer *node;
 
 		btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
 		level = root_item->drop_level;
@@ -1498,10 +1506,10 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 			ret = wret;
 			goto out;
 		}
-		node = btrfs_buffer_node(path->nodes[level]);
-		found_key = &node->ptrs[path->slots[level]].key;
-		WARN_ON(memcmp(found_key, &root_item->drop_progress,
-			       sizeof(*found_key)));
+		node = path->nodes[level];
+		btrfs_node_key(node, &found_key, path->slots[level]);
+		WARN_ON(memcmp(&found_key, &root_item->drop_progress,
+			       sizeof(found_key)));
 	}
 	while(1) {
 		wret = walk_down_tree(trans, root, path, &level);
@@ -1516,12 +1524,12 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret < 0)
 			ret = wret;
 		ret = -EAGAIN;
-		get_bh(root->node);
+		extent_buffer_get(root->node);
 		break;
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path->nodes[i]) {
-			btrfs_block_release(root, path->nodes[i]);
+			free_extent_buffer(path->nodes[i]);
 			path->nodes[i] = 0;
 		}
 	}
@@ -1581,13 +1589,12 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	struct btrfs_path *path;
 	int ret;
 	int err = 0;
-	struct btrfs_block_group_item *bi;
 	struct btrfs_block_group_cache *cache;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct radix_tree_root *radix;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u64 group_size_blocks;
 	u64 used;
 
@@ -1596,7 +1603,6 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	root = info->extent_root;
 	key.objectid = 0;
 	key.offset = group_size_blocks;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
 
 	path = btrfs_alloc_path();
@@ -1610,18 +1616,18 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 			err = ret;
 			break;
 		}
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		btrfs_disk_key_to_cpu(&found_key,
-				      &leaf->items[path->slots[0]].key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		cache = kmalloc(sizeof(*cache), GFP_NOFS);
 		if (!cache) {
 			err = -1;
 			break;
 		}
 
-		bi = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_block_group_item);
-		if (bi->flags & BTRFS_BLOCK_GROUP_DATA) {
+		read_extent_buffer(leaf, &cache->item,
+				   btrfs_item_ptr_offset(leaf, path->slots[0]),
+				   sizeof(cache->item));
+		if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
 			radix = &info->block_group_data_radix;
 			cache->data = 1;
 		} else {
@@ -1629,7 +1635,6 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 			cache->data = 0;
 		}
 
-		memcpy(&cache->item, bi, sizeof(*bi));
 		memcpy(&cache->key, &found_key, sizeof(found_key));
 		cache->last_alloc = cache->key.objectid;
 		cache->first_free = cache->key.objectid;
@@ -1640,11 +1645,12 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 
 		key.objectid = found_key.objectid + found_key.offset;
 		btrfs_release_path(root, path);
+
 		ret = radix_tree_insert(radix, found_key.objectid +
 					found_key.offset - 1,
 					(void *)cache);
 		BUG_ON(ret);
-		used = btrfs_block_group_used(bi);
+		used = btrfs_block_group_used(&cache->item);
 		if (used < div_factor(key.offset, 8)) {
 			radix_tree_tag_set(radix, found_key.objectid +
 					   found_key.offset - 1,

fs/btrfs/extent_map.c

@@ -26,16 +26,6 @@ struct tree_entry {
 	struct rb_node rb_node;
 };
 
-/* bits for the extent state */
-#define EXTENT_DIRTY 1
-#define EXTENT_WRITEBACK (1 << 1)
-#define EXTENT_UPTODATE (1 << 2)
-#define EXTENT_LOCKED (1 << 3)
-#define EXTENT_NEW (1 << 4)
-#define EXTENT_DELALLOC (1 << 5)
-
-#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
-
 void __init extent_map_init(void)
 {
 	extent_map_cache = btrfs_cache_create("extent_map",
@@ -223,7 +213,8 @@ int add_extent_mapping(struct extent_map_tree *tree,
 		if (rb)
 			prev = rb_entry(rb, struct extent_map, rb_node);
 		if (prev && prev->end + 1 == em->start &&
-		    ((em->block_start == 0 && prev->block_start == 0) ||
+		    ((em->block_start == EXTENT_MAP_HOLE &&
+		      prev->block_start == EXTENT_MAP_HOLE) ||
 			     (em->block_start == prev->block_end + 1))) {
 			em->start = prev->start;
 			em->block_start = prev->block_start;
@@ -926,6 +917,40 @@ int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end)
 }
 EXPORT_SYMBOL(set_range_writeback);
 
+int find_first_extent_bit(struct extent_map_tree *tree, u64 start,
+			  u64 *start_ret, u64 *end_ret, int bits)
+{
+	struct rb_node *node;
+	struct extent_state *state;
+	int ret = 1;
+
+	write_lock_irq(&tree->lock);
+	/*
+	 * this search will find all the extents that end after
+	 * our range starts.
+	 */
+	node = tree_search(&tree->state, start);
+	if (!node || IS_ERR(node)) {
+		goto out;
+	}
+
+	while(1) {
+		state = rb_entry(node, struct extent_state, rb_node);
+		if (state->state & bits) {
+			*start_ret = state->start;
+			*end_ret = state->end;
+			ret = 0;
+		}
+		node = rb_next(node);
+		if (!node)
+			break;
+	}
+out:
+	write_unlock_irq(&tree->lock);
+	return ret;
+}
+EXPORT_SYMBOL(find_first_extent_bit);
+
 u64 find_lock_delalloc_range(struct extent_map_tree *tree,
 			     u64 start, u64 lock_start, u64 *end, u64 max_bytes)
 {
@@ -1450,7 +1475,7 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
 		em = NULL;
 
 		/* we've found a hole, just zero and go on */
-		if (block_start == 0) {
+		if (block_start == EXTENT_MAP_HOLE) {
 			zero_user_page(page, page_offset, iosize, KM_USER0);
 			set_extent_uptodate(tree, cur, cur + iosize - 1,
 					    GFP_NOFS);
@@ -1593,7 +1618,8 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
 		free_extent_map(em);
 		em = NULL;
 
-		if (block_start == 0 || block_start == EXTENT_MAP_INLINE) {
+		if (block_start == EXTENT_MAP_HOLE ||
+		    block_start == EXTENT_MAP_INLINE) {
 			clear_extent_dirty(tree, cur,
 					   cur + iosize - 1, GFP_NOFS);
 			cur = cur + iosize;
@@ -1630,7 +1656,6 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
 		nr++;
 	}
 done:
-	WARN_ON(test_range_bit(tree, start, page_end, EXTENT_DIRTY, 0));
 	unlock_extent(tree, start, page_end, GFP_NOFS);
 	unlock_page(page);
 	return 0;
@@ -1827,8 +1852,623 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
 
 	// XXX(hch): block 0 is valid in some cases, e.g. XFS RT device
 	if (em->block_start == EXTENT_MAP_INLINE ||
-	    em->block_start == 0)
-	    	return 0;
+	    em->block_start == EXTENT_MAP_HOLE)
+		return 0;
 
 	return (em->block_start + start - em->start) >> inode->i_blkbits;
 }
+
+struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
+					  u64 start, unsigned long len,
+					  gfp_t mask)
+{
+	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
+				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long i;
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	struct extent_buffer *eb;
+	struct page *p;
+	struct address_space *mapping = tree->mapping;
+	int uptodate = 0;
+
+	eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask);
+	if (!eb || IS_ERR(eb))
+		return NULL;
+
+	eb->start = start;
+	eb->len = len;
+	atomic_set(&eb->refs, 1);
+
+	for (i = 0; i < num_pages; i++, index++) {
+		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
+		if (!p)
+			goto fail;
+		eb->pages[i] = p;
+		if (!PageUptodate(p))
+			uptodate = 0;
+		unlock_page(p);
+	}
+	if (uptodate)
+		eb->flags |= EXTENT_UPTODATE;
+	return eb;
+fail:
+	free_extent_buffer(eb);
+	return NULL;
+}
+EXPORT_SYMBOL(alloc_extent_buffer);
+
+struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
+					 u64 start, unsigned long len,
+					  gfp_t mask)
+{
+	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
+				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long i;
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	struct extent_buffer *eb;
+	struct page *p;
+	struct address_space *mapping = tree->mapping;
+
+	eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask);
+	if (!eb || IS_ERR(eb))
+		return NULL;
+
+	eb->start = start;
+	eb->len = len;
+	atomic_set(&eb->refs, 1);
+
+	for (i = 0; i < num_pages; i++, index++) {
+		p = find_get_page(mapping, index);
+		if (!p)
+			goto fail;
+		eb->pages[i] = p;
+	}
+	return eb;
+fail:
+	free_extent_buffer(eb);
+	return NULL;
+}
+EXPORT_SYMBOL(find_extent_buffer);
+
+void free_extent_buffer(struct extent_buffer *eb)
+{
+	unsigned long i;
+	unsigned long num_pages;
+
+	if (!eb)
+		return;
+
+	if (!atomic_dec_and_test(&eb->refs))
+		return;
+
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+
+	for (i = 0; i < num_pages; i++) {
+		if (eb->pages[i])
+			page_cache_release(eb->pages[i]);
+	}
+	kfree(eb);
+}
+EXPORT_SYMBOL(free_extent_buffer);
+
+int clear_extent_buffer_dirty(struct extent_map_tree *tree,
+			      struct extent_buffer *eb)
+{
+	int set;
+	unsigned long i;
+	unsigned long num_pages;
+	struct page *page;
+
+	u64 start = eb->start;
+	u64 end = start + eb->len - 1;
+
+	set = clear_extent_dirty(tree, start, end, GFP_NOFS);
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		lock_page(page);
+		/*
+		 * if we're on the last page or the first page and the
+		 * block isn't aligned on a page boundary, do extra checks
+		 * to make sure we don't clean page that is partially dirty
+		 */
+		if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
+		    ((i == num_pages - 1) &&
+		     ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) {
+			start = page->index << PAGE_CACHE_SHIFT;
+			end  = start + PAGE_CACHE_SIZE - 1;
+			if (test_range_bit(tree, start, end,
+					   EXTENT_DIRTY, 0)) {
+				unlock_page(page);
+				continue;
+			}
+		}
+		clear_page_dirty_for_io(page);
+		unlock_page(page);
+	}
+	return 0;
+}
+EXPORT_SYMBOL(clear_extent_buffer_dirty);
+
+int wait_on_extent_buffer_writeback(struct extent_map_tree *tree,
+				    struct extent_buffer *eb)
+{
+	return wait_on_extent_writeback(tree, eb->start,
+					eb->start + eb->len - 1);
+}
+EXPORT_SYMBOL(wait_on_extent_buffer_writeback);
+
+int set_extent_buffer_dirty(struct extent_map_tree *tree,
+			     struct extent_buffer *eb)
+{
+	return set_range_dirty(tree, eb->start, eb->start + eb->len - 1);
+}
+EXPORT_SYMBOL(set_extent_buffer_dirty);
+
+int set_extent_buffer_uptodate(struct extent_map_tree *tree,
+				struct extent_buffer *eb)
+{
+	unsigned long i;
+	struct page *page;
+	unsigned long num_pages;
+
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+
+	set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
+			    GFP_NOFS);
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
+		    ((i == num_pages - 1) &&
+		     ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) {
+			check_page_uptodate(tree, page);
+			continue;
+		}
+		SetPageUptodate(page);
+	}
+	return 0;
+}
+EXPORT_SYMBOL(set_extent_buffer_uptodate);
+
+int extent_buffer_uptodate(struct extent_map_tree *tree,
+			     struct extent_buffer *eb)
+{
+	if (eb->flags & EXTENT_UPTODATE)
+		return 1;
+	return test_range_bit(tree, eb->start, eb->start + eb->len - 1,
+			   EXTENT_UPTODATE, 1);
+}
+EXPORT_SYMBOL(extent_buffer_uptodate);
+
+int read_extent_buffer_pages(struct extent_map_tree *tree,
+			     struct extent_buffer *eb, int wait)
+{
+	unsigned long i;
+	struct page *page;
+	int err;
+	int ret = 0;
+	unsigned long num_pages;
+
+	if (eb->flags & EXTENT_UPTODATE)
+		return 0;
+
+	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
+			   EXTENT_UPTODATE, 1)) {
+		return 0;
+	}
+
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		if (PageUptodate(page)) {
+			continue;
+		}
+		if (!wait) {
+			if (TestSetPageLocked(page)) {
+				continue;
+			}
+		} else {
+			lock_page(page);
+		}
+		if (!PageUptodate(page)) {
+			err = page->mapping->a_ops->readpage(NULL, page);
+			if (err) {
+				ret = err;
+			}
+		} else {
+			unlock_page(page);
+		}
+	}
+
+	if (ret || !wait) {
+		return ret;
+	}
+
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		wait_on_page_locked(page);
+		if (!PageUptodate(page)) {
+			ret = -EIO;
+		}
+	}
+	eb->flags |= EXTENT_UPTODATE;
+	return ret;
+}
+EXPORT_SYMBOL(read_extent_buffer_pages);
+
+void read_extent_buffer(struct extent_buffer *eb, void *dstv,
+			unsigned long start,
+			unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char *dst = (char *)dstv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, (PAGE_CACHE_SIZE - offset));
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		memcpy(dst, kaddr + offset, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+
+		dst += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+}
+EXPORT_SYMBOL(read_extent_buffer);
+
+int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+		      char **token, char **map,
+		      unsigned long *map_start,
+		      unsigned long *map_len, int km)
+{
+	size_t offset;
+	char *kaddr;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+
+	if (i == 0) {
+		offset = start_offset;
+		*map_start = 0;
+	} else {
+		offset = 0;
+		*map_start = (i << PAGE_CACHE_SHIFT) - offset;
+	}
+
+	// kaddr = kmap_atomic(eb->pages[i], km);
+	kaddr = page_address(eb->pages[i]);
+	*token = kaddr;
+	*map = kaddr + offset;
+	*map_len = PAGE_CACHE_SIZE - offset;
+	return 0;
+}
+EXPORT_SYMBOL(map_extent_buffer);
+
+void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
+{
+	// kunmap_atomic(token, km);
+}
+EXPORT_SYMBOL(unmap_extent_buffer);
+
+int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
+			  unsigned long start,
+			  unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char *ptr = (char *)ptrv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+	int ret = 0;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, (PAGE_CACHE_SIZE - offset));
+
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		ret = memcmp(ptr, kaddr + offset, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+		if (ret)
+			break;
+
+		ptr += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+	return ret;
+}
+EXPORT_SYMBOL(memcmp_extent_buffer);
+
+void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
+			 unsigned long start, unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char *src = (char *)srcv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, PAGE_CACHE_SIZE - offset);
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		memcpy(kaddr + offset, src, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+
+		src += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+}
+EXPORT_SYMBOL(write_extent_buffer);
+
+void memset_extent_buffer(struct extent_buffer *eb, char c,
+			  unsigned long start, unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, PAGE_CACHE_SIZE - offset);
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		memset(kaddr + offset, c, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+}
+EXPORT_SYMBOL(memset_extent_buffer);
+
+void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
+			unsigned long dst_offset, unsigned long src_offset,
+			unsigned long len)
+{
+	u64 dst_len = dst->len;
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(src->len != dst_len);
+
+	offset = dst_offset & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		page = dst->pages[i];
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
+
+		// kaddr = kmap_atomic(page, KM_USER1);
+		kaddr = page_address(page);
+		read_extent_buffer(src, kaddr + offset, src_offset, cur);
+		// kunmap_atomic(kaddr, KM_USER1);
+
+		src_offset += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+	}
+}
+EXPORT_SYMBOL(copy_extent_buffer);
+
+static void move_pages(struct page *dst_page, struct page *src_page,
+		       unsigned long dst_off, unsigned long src_off,
+		       unsigned long len)
+{
+	// char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
+	char *dst_kaddr = page_address(dst_page);
+	if (dst_page == src_page) {
+		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
+	} else {
+		// char *src_kaddr = kmap_atomic(src_page, KM_USER1);
+		char *src_kaddr = page_address(src_page);
+		char *p = dst_kaddr + dst_off + len;
+		char *s = src_kaddr + src_off + len;
+
+		while (len--)
+			*--p = *--s;
+
+		// kunmap_atomic(src_kaddr, KM_USER1);
+	}
+	// kunmap_atomic(dst_kaddr, KM_USER0);
+}
+
+static void copy_pages(struct page *dst_page, struct page *src_page,
+		       unsigned long dst_off, unsigned long src_off,
+		       unsigned long len)
+{
+	//kmap_atomic(dst_page, KM_USER0);
+	char *dst_kaddr = page_address(dst_page);
+	char *src_kaddr;
+
+	if (dst_page != src_page)
+		src_kaddr = page_address(src_page); // kmap_atomic(src_page, KM_USER1);
+	else
+		src_kaddr = dst_kaddr;
+
+	memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
+	/*
+	kunmap_atomic(dst_kaddr, KM_USER0);
+	if (dst_page != src_page)
+		kunmap_atomic(src_kaddr, KM_USER1);
+	*/
+}
+
+void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len)
+{
+	size_t cur;
+	size_t dst_off_in_page;
+	size_t src_off_in_page;
+	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long dst_i;
+	unsigned long src_i;
+
+	if (src_offset + len > dst->len) {
+		printk("memmove bogus src_offset %lu move len %lu len %lu\n",
+		       src_offset, len, dst->len);
+		BUG_ON(1);
+	}
+	if (dst_offset + len > dst->len) {
+		printk("memmove bogus dst_offset %lu move len %lu len %lu\n",
+		       dst_offset, len, dst->len);
+		BUG_ON(1);
+	}
+
+	while(len > 0) {
+		dst_off_in_page = dst_offset &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+		src_off_in_page = src_offset &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+
+		dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
+		src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
+
+		if (src_i == 0)
+			src_off_in_page += start_offset;
+		if (dst_i == 0)
+			dst_off_in_page += start_offset;
+
+		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
+					       src_off_in_page));
+		cur = min(cur, (unsigned long)(PAGE_CACHE_SIZE -
+					       dst_off_in_page));
+
+		copy_pages(dst->pages[dst_i], dst->pages[src_i],
+			   dst_off_in_page, src_off_in_page, cur);
+
+		src_offset += cur;
+		dst_offset += cur;
+		len -= cur;
+	}
+}
+EXPORT_SYMBOL(memcpy_extent_buffer);
+
+void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len)
+{
+	size_t cur;
+	size_t dst_off_in_page;
+	size_t src_off_in_page;
+	unsigned long dst_end = dst_offset + len - 1;
+	unsigned long src_end = src_offset + len - 1;
+	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long dst_i;
+	unsigned long src_i;
+
+	if (src_offset + len > dst->len) {
+		printk("memmove bogus src_offset %lu move len %lu len %lu\n",
+		       src_offset, len, dst->len);
+		BUG_ON(1);
+	}
+	if (dst_offset + len > dst->len) {
+		printk("memmove bogus dst_offset %lu move len %lu len %lu\n",
+		       dst_offset, len, dst->len);
+		BUG_ON(1);
+	}
+	if (dst_offset < src_offset) {
+		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
+		return;
+	}
+	while(len > 0) {
+		dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
+		src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
+
+		dst_off_in_page = dst_end &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+		src_off_in_page = src_end &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+
+		if (src_i == 0)
+			src_off_in_page += start_offset;
+		if (dst_i == 0)
+			dst_off_in_page += start_offset;
+
+		cur = min(len, src_off_in_page + 1);
+		cur = min(cur, dst_off_in_page + 1);
+// printk("move pages orig dst %lu src %lu len %lu, this %lu %lu %lu\n", dst_offset, src_offset, len, dst_off_in_page - cur + 1, src_off_in_page - cur + 1, cur);
+		move_pages(dst->pages[dst_i], dst->pages[src_i],
+			   dst_off_in_page - cur + 1,
+			   src_off_in_page - cur + 1, cur);
+
+		dst_end -= cur - 1;
+		src_end -= cur - 1;
+		len -= cur;
+	}
+}
+EXPORT_SYMBOL(memmove_extent_buffer);

fs/btrfs/extent_map.h

@@ -3,9 +3,20 @@
 
 #include <linux/rbtree.h>
 
+#define EXTENT_MAP_HOLE (u64)-3
 #define EXTENT_MAP_INLINE (u64)-2
 #define EXTENT_MAP_DELALLOC (u64)-1
 
+/* bits for the extent state */
+#define EXTENT_DIRTY 1
+#define EXTENT_WRITEBACK (1 << 1)
+#define EXTENT_UPTODATE (1 << 2)
+#define EXTENT_LOCKED (1 << 3)
+#define EXTENT_NEW (1 << 4)
+#define EXTENT_DELALLOC (1 << 5)
+#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+
+
 struct extent_map_ops {
 	int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
 	int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
@@ -51,10 +62,13 @@ struct extent_state {
 	struct list_head list;
 };
 
+#define EXTENT_BUFFER_SIZE(nr) (sizeof(struct extent_buffer) + \
+			       (nr - 1) * sizeof(struct page *))
 struct extent_buffer {
 	u64 start;
-	u64 end; /* inclusive */
-	char *addr;
+	unsigned long len;
+	atomic_t refs;
+	int flags;
 	struct page *pages[];
 };
 
@@ -87,8 +101,12 @@ int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
 		   gfp_t mask);
 int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
 		     gfp_t mask);
+int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+		       gfp_t mask);
 int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end,
 		     gfp_t mask);
+int find_first_extent_bit(struct extent_map_tree *tree, u64 start,
+			  u64 *start_ret, u64 *end_ret, int bits);
 int extent_invalidatepage(struct extent_map_tree *tree,
 			  struct page *page, unsigned long offset);
 int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
@@ -106,4 +124,57 @@ int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end);
 int set_state_private(struct extent_map_tree *tree, u64 start, u64 private);
 int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private);
 void set_page_extent_mapped(struct page *page);
+
+struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
+					  u64 start, unsigned long len,
+					  gfp_t mask);
+struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
+					 u64 start, unsigned long len,
+					  gfp_t mask);
+void free_extent_buffer(struct extent_buffer *eb);
+int read_extent_buffer_pages(struct extent_map_tree *tree,
+			     struct extent_buffer *eb, int wait);
+
+static inline void extent_buffer_get(struct extent_buffer *eb)
+{
+	atomic_inc(&eb->refs);
+}
+
+static inline u64 extent_buffer_blocknr(struct extent_buffer *eb)
+{
+	return eb->start / 4096;
+}
+
+int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
+			  unsigned long start,
+			  unsigned long len);
+void read_extent_buffer(struct extent_buffer *eb, void *dst,
+			unsigned long start,
+			unsigned long len);
+void write_extent_buffer(struct extent_buffer *eb, const void *src,
+			 unsigned long start, unsigned long len);
+void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
+			unsigned long dst_offset, unsigned long src_offset,
+			unsigned long len);
+void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len);
+void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len);
+void memset_extent_buffer(struct extent_buffer *eb, char c,
+			  unsigned long start, unsigned long len);
+int wait_on_extent_buffer_writeback(struct extent_map_tree *tree,
+				    struct extent_buffer *eb);
+int clear_extent_buffer_dirty(struct extent_map_tree *tree,
+			      struct extent_buffer *eb);
+int set_extent_buffer_dirty(struct extent_map_tree *tree,
+			     struct extent_buffer *eb);
+int set_extent_buffer_uptodate(struct extent_map_tree *tree,
+			       struct extent_buffer *eb);
+int extent_buffer_uptodate(struct extent_map_tree *tree,
+			   struct extent_buffer *eb);
+int map_extent_buffer(struct extent_buffer *eb, unsigned long offset,
+		      char **token, char **map,
+		      unsigned long *map_start,
+		      unsigned long *map_len, int km);
+void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
 #endif

fs/btrfs/file-item.c

@@ -34,12 +34,12 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_file_extent_item *item;
 	struct btrfs_key file_key;
 	struct btrfs_path *path;
+	struct extent_buffer *leaf;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 	file_key.objectid = objectid;
 	file_key.offset = pos;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);
 
 	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
@@ -47,15 +47,16 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret);
-	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-	btrfs_set_file_extent_disk_blocknr(item, offset);
-	btrfs_set_file_extent_disk_num_blocks(item, disk_num_blocks);
-	btrfs_set_file_extent_offset(item, 0);
-	btrfs_set_file_extent_num_blocks(item, num_blocks);
-	btrfs_set_file_extent_generation(item, trans->transid);
-	btrfs_set_file_extent_type(item, BTRFS_FILE_EXTENT_REG);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_set_file_extent_disk_blocknr(leaf, item, offset);
+	btrfs_set_file_extent_disk_num_blocks(leaf, item, disk_num_blocks);
+	btrfs_set_file_extent_offset(leaf, item, 0);
+	btrfs_set_file_extent_num_blocks(leaf, item, num_blocks);
+	btrfs_set_file_extent_generation(leaf, item, trans->transid);
+	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
+	btrfs_mark_buffer_dirty(leaf);
 out:
 	btrfs_free_path(path);
 	return ret;
@@ -71,32 +72,30 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 	struct btrfs_key file_key;
 	struct btrfs_key found_key;
 	struct btrfs_csum_item *item;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u64 csum_offset = 0;
 	int csums_in_item;
 
 	file_key.objectid = objectid;
 	file_key.offset = offset;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY);
 	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
 	if (ret < 0)
 		goto fail;
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
+	leaf = path->nodes[0];
 	if (ret > 0) {
 		ret = 1;
 		if (path->slots[0] == 0)
 			goto fail;
 		path->slots[0]--;
-		btrfs_disk_key_to_cpu(&found_key,
-				      &leaf->items[path->slots[0]].key);
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY ||
 		    found_key.objectid != objectid) {
 			goto fail;
 		}
 		csum_offset = (offset - found_key.offset) >>
 				root->fs_info->sb->s_blocksize_bits;
-		csums_in_item = btrfs_item_size(leaf->items + path->slots[0]);
+		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
 		csums_in_item /= BTRFS_CRC32_SIZE;
 
 		if (csum_offset >= csums_in_item) {
@@ -127,7 +126,6 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 
 	file_key.objectid = objectid;
 	file_key.offset = offset;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);
 	ret = btrfs_search_slot(trans, root, &file_key, path, ins_len, cow);
 	return ret;
@@ -138,12 +136,14 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 			  u64 objectid, u64 offset,
 			  char *data, size_t len)
 {
+	return 0;
+#if 0
 	int ret;
 	struct btrfs_key file_key;
 	struct btrfs_key found_key;
 	struct btrfs_path *path;
 	struct btrfs_csum_item *item;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u64 csum_offset;
 
 	path = btrfs_alloc_path();
@@ -161,8 +161,8 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 	if (ret == -EFBIG) {
 		u32 item_size;
 		/* we found one, but it isn't big enough yet */
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		item_size = btrfs_item_size(leaf->items + path->slots[0]);
+		leaf = path->nodes[0];
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 		if ((item_size / BTRFS_CRC32_SIZE) >= MAX_CSUM_ITEMS(root)) {
 			/* already at max size, make a new one */
 			goto insert;
@@ -188,8 +188,8 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 		goto insert;
 	}
 	path->slots[0]--;
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	btrfs_disk_key_to_cpu(&found_key, &leaf->items[path->slots[0]].key);
+	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 	csum_offset = (offset - found_key.offset) >>
 			root->fs_info->sb->s_blocksize_bits;
 	if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY ||
@@ -197,10 +197,10 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 	    csum_offset >= MAX_CSUM_ITEMS(root)) {
 		goto insert;
 	}
-	if (csum_offset >= btrfs_item_size(leaf->items + path->slots[0]) /
+	if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) /
 	    BTRFS_CRC32_SIZE) {
 		u32 diff = (csum_offset + 1) * BTRFS_CRC32_SIZE;
-		diff = diff - btrfs_item_size(leaf->items + path->slots[0]);
+		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
 		if (diff != BTRFS_CRC32_SIZE)
 			goto insert;
 		ret = btrfs_extend_item(trans, root, path, diff);
@@ -220,21 +220,20 @@ insert:
 		goto fail;
 	}
 csum:
-	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
-			      struct btrfs_csum_item);
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
 	ret = 0;
 	item = (struct btrfs_csum_item *)((unsigned char *)item +
 					  csum_offset * BTRFS_CRC32_SIZE);
 found:
-	btrfs_check_bounds(&item->csum, BTRFS_CRC32_SIZE,
-			   path->nodes[0]->b_data,
-			   root->fs_info->sb->s_blocksize);
+	/* FIXME!!!!!!!!!!!! */
 	ret = btrfs_csum_data(root, data, len, &item->csum);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 fail:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
 	return ret;
+#endif
 }
 
 int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
@@ -242,21 +241,21 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
 			u64 isize)
 {
 	struct btrfs_key key;
-	struct btrfs_leaf *leaf = btrfs_buffer_leaf(path->nodes[0]);
+	struct extent_buffer *leaf = path->nodes[0];
 	int slot = path->slots[0];
 	int ret;
 	u32 new_item_size;
 	u64 new_item_span;
 	u64 blocks;
 
-	btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+	btrfs_item_key_to_cpu(leaf, &key, slot);
 	if (isize <= key.offset)
 		return 0;
 	new_item_span = isize - key.offset;
-	blocks = (new_item_span + root->blocksize - 1) >>
+	blocks = (new_item_span + root->sectorsize - 1) >>
 		root->fs_info->sb->s_blocksize_bits;
 	new_item_size = blocks * BTRFS_CRC32_SIZE;
-	if (new_item_size >= btrfs_item_size(leaf->items + slot))
+	if (new_item_size >= btrfs_item_size_nr(leaf, slot))
 		return 0;
 	ret = btrfs_truncate_item(trans, root, path, new_item_size);
 	BUG_ON(ret);

fs/btrfs/file.c

@@ -16,7 +16,6 @@
  * Boston, MA 021110-1307, USA.
  */
 
-#include <linux/buffer_head.h>
 #include <linux/fs.h>
 #include <linux/pagemap.h>
 #include <linux/highmem.h>
@@ -88,7 +87,9 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_key key;
 	struct btrfs_path *path;
-	char *ptr, *kaddr;
+	struct extent_buffer *leaf;
+	char *kaddr;
+	unsigned long ptr;
 	struct btrfs_file_extent_item *ei;
 	u32 datasize;
 	int err = 0;
@@ -102,7 +103,6 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 
 	key.objectid = inode->i_ino;
 	key.offset = offset;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
 	BUG_ON(size >= PAGE_CACHE_SIZE);
 	datasize = btrfs_file_extent_calc_inline_size(size);
@@ -113,18 +113,17 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 		err = ret;
 		goto fail;
 	}
-	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-	       path->slots[0], struct btrfs_file_extent_item);
-	btrfs_set_file_extent_generation(ei, trans->transid);
-	btrfs_set_file_extent_type(ei,
-				   BTRFS_FILE_EXTENT_INLINE);
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
+	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
 
 	kaddr = kmap_atomic(page, KM_USER0);
-	btrfs_memcpy(root, path->nodes[0]->b_data,
-		     ptr, kaddr + page_offset, size);
+	write_extent_buffer(leaf, kaddr + page_offset, ptr, size);
 	kunmap_atomic(kaddr, KM_USER0);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_mark_buffer_dirty(leaf);
 fail:
 	btrfs_free_path(path);
 	return err;
@@ -156,8 +155,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 
 	em->bdev = inode->i_sb->s_bdev;
 
-	start_pos = pos & ~((u64)root->blocksize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
+	start_pos = pos & ~((u64)root->sectorsize - 1);
+	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
 			inode->i_blkbits;
 
 	down_read(&BTRFS_I(inode)->root->snap_sem);
@@ -184,7 +183,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	if (inode->i_size < start_pos) {
 		u64 last_pos_in_file;
 		u64 hole_size;
-		u64 mask = root->blocksize - 1;
+		u64 mask = root->sectorsize - 1;
 		last_pos_in_file = (isize + mask) & ~mask;
 		hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
 
@@ -227,8 +226,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		/* step one, delete the existing extents in this range */
 		/* FIXME blocksize != pagesize */
 		err = btrfs_drop_extents(trans, root, inode, start_pos,
-			 (pos + write_bytes + root->blocksize -1) &
-			 ~((u64)root->blocksize - 1), &hint_block);
+			 (pos + write_bytes + root->sectorsize -1) &
+			 ~((u64)root->sectorsize - 1), &hint_block);
 		if (err)
 			goto failed;
 
@@ -288,7 +287,7 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 {
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	int slot;
 	struct btrfs_file_extent_item *extent;
 	u64 extent_end = 0;
@@ -327,10 +326,10 @@ next_slot:
 		found_extent = 0;
 		found_inline = 0;
 		extent = NULL;
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
+		leaf = path->nodes[0];
 		slot = path->slots[0];
 		ret = 0;
-		btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (key.offset >= end || key.objectid != inode->i_ino) {
 			goto out;
 		}
@@ -344,17 +343,18 @@ next_slot:
 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
 			extent = btrfs_item_ptr(leaf, slot,
 						struct btrfs_file_extent_item);
-			found_type = btrfs_file_extent_type(extent);
+			found_type = btrfs_file_extent_type(leaf, extent);
 			if (found_type == BTRFS_FILE_EXTENT_REG) {
 				extent_end = key.offset +
-					(btrfs_file_extent_num_blocks(extent) <<
+				 (btrfs_file_extent_num_blocks(leaf, extent) <<
 					 inode->i_blkbits);
 				found_extent = 1;
 			} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+				struct btrfs_item *item;
+				item = btrfs_item_nr(leaf, slot);
 				found_inline = 1;
 				extent_end = key.offset +
-				     btrfs_file_extent_inline_len(leaf->items +
-								  slot);
+				     btrfs_file_extent_inline_len(leaf, item);
 			}
 		} else {
 			extent_end = search_start;
@@ -365,8 +365,7 @@ next_slot:
 		    search_start >= extent_end) {
 			int nextret;
 			u32 nritems;
-			nritems = btrfs_header_nritems(
-					btrfs_buffer_header(path->nodes[0]));
+			nritems = btrfs_header_nritems(leaf);
 			if (slot >= nritems - 1) {
 				nextret = btrfs_next_leaf(root, path);
 				if (nextret)
@@ -380,7 +379,7 @@ next_slot:
 
 		/* FIXME, there's only one inline extent allowed right now */
 		if (found_inline) {
-			u64 mask = root->blocksize - 1;
+			u64 mask = root->sectorsize - 1;
 			search_start = (extent_end + mask) & ~mask;
 		} else
 			search_start = extent_end;
@@ -388,10 +387,13 @@ next_slot:
 		if (end < extent_end && end >= key.offset) {
 			if (found_extent) {
 				u64 disk_blocknr =
-					btrfs_file_extent_disk_blocknr(extent);
+				    btrfs_file_extent_disk_blocknr(leaf,extent);
 				u64 disk_num_blocks =
-				      btrfs_file_extent_disk_num_blocks(extent);
-				memcpy(&old, extent, sizeof(old));
+				    btrfs_file_extent_disk_num_blocks(leaf,
+								      extent);
+				read_extent_buffer(leaf, &old,
+						   (unsigned long)extent,
+						   sizeof(old));
 				if (disk_blocknr != 0) {
 					ret = btrfs_inc_extent_ref(trans, root,
 					         disk_blocknr, disk_num_blocks);
@@ -406,20 +408,24 @@ next_slot:
 			u64 new_num;
 			u64 old_num;
 			keep = 1;
-			WARN_ON(start & (root->blocksize - 1));
+			WARN_ON(start & (root->sectorsize - 1));
 			if (found_extent) {
 				new_num = (start - key.offset) >>
 					inode->i_blkbits;
-				old_num = btrfs_file_extent_num_blocks(extent);
+				old_num = btrfs_file_extent_num_blocks(leaf,
+								       extent);
 				*hint_block =
-					btrfs_file_extent_disk_blocknr(extent);
-				if (btrfs_file_extent_disk_blocknr(extent)) {
+					btrfs_file_extent_disk_blocknr(leaf,
+								       extent);
+				if (btrfs_file_extent_disk_blocknr(leaf,
+								   extent)) {
 					inode->i_blocks -=
 						(old_num - new_num) << 3;
 				}
-				btrfs_set_file_extent_num_blocks(extent,
+				btrfs_set_file_extent_num_blocks(leaf,
+								 extent,
 								 new_num);
-				btrfs_mark_buffer_dirty(path->nodes[0]);
+				btrfs_mark_buffer_dirty(leaf);
 			} else {
 				WARN_ON(1);
 			}
@@ -431,13 +437,17 @@ next_slot:
 			u64 extent_num_blocks = 0;
 			if (found_extent) {
 				disk_blocknr =
-				      btrfs_file_extent_disk_blocknr(extent);
+				      btrfs_file_extent_disk_blocknr(leaf,
+								     extent);
 				disk_num_blocks =
-				      btrfs_file_extent_disk_num_blocks(extent);
+				      btrfs_file_extent_disk_num_blocks(leaf,
+									extent);
 				extent_num_blocks =
-				      btrfs_file_extent_num_blocks(extent);
+				      btrfs_file_extent_num_blocks(leaf,
+								   extent);
 				*hint_block =
-					btrfs_file_extent_disk_blocknr(extent);
+					btrfs_file_extent_disk_blocknr(leaf,
+								       extent);
 			}
 			ret = btrfs_del_item(trans, root, path);
 			/* TODO update progress marker and return */
@@ -464,42 +474,37 @@ next_slot:
 			struct btrfs_key ins;
 			ins.objectid = inode->i_ino;
 			ins.offset = end;
-			ins.flags = 0;
 			btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY);
 			btrfs_release_path(root, path);
 			ret = btrfs_insert_empty_item(trans, root, path, &ins,
 						      sizeof(*extent));
 
+			leaf = path->nodes[0];
 			if (ret) {
-				btrfs_print_leaf(root, btrfs_buffer_leaf(path->nodes[0]));
-				printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.flags, ins.offset, start, end, key.offset, extent_end, keep);
+				btrfs_print_leaf(root, leaf);
+				printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.type, ins.offset, start, end, key.offset, extent_end, keep);
 			}
 			BUG_ON(ret);
-			extent = btrfs_item_ptr(
-				    btrfs_buffer_leaf(path->nodes[0]),
-				    path->slots[0],
-				    struct btrfs_file_extent_item);
-			btrfs_set_file_extent_disk_blocknr(extent,
-				    btrfs_file_extent_disk_blocknr(&old));
-			btrfs_set_file_extent_disk_num_blocks(extent,
-				    btrfs_file_extent_disk_num_blocks(&old));
-
-			btrfs_set_file_extent_offset(extent,
-				    btrfs_file_extent_offset(&old) +
+			extent = btrfs_item_ptr(leaf, path->slots[0],
+						struct btrfs_file_extent_item);
+			write_extent_buffer(leaf, &old,
+					    (unsigned long)extent, sizeof(old));
+
+			btrfs_set_file_extent_offset(leaf, extent,
+				    le64_to_cpu(old.offset) +
 				    ((end - key.offset) >> inode->i_blkbits));
-			WARN_ON(btrfs_file_extent_num_blocks(&old) <
+			WARN_ON(le64_to_cpu(old.num_blocks) <
 				(extent_end - end) >> inode->i_blkbits);
-			btrfs_set_file_extent_num_blocks(extent,
+			btrfs_set_file_extent_num_blocks(leaf, extent,
 				    (extent_end - end) >> inode->i_blkbits);
 
-			btrfs_set_file_extent_type(extent,
+			btrfs_set_file_extent_type(leaf, extent,
 						   BTRFS_FILE_EXTENT_REG);
-			btrfs_set_file_extent_generation(extent,
-				    btrfs_file_extent_generation(&old));
 			btrfs_mark_buffer_dirty(path->nodes[0]);
-			if (btrfs_file_extent_disk_blocknr(&old) != 0) {
+			if (le64_to_cpu(old.disk_blocknr) != 0) {
 				inode->i_blocks +=
-				      btrfs_file_extent_num_blocks(extent) << 3;
+				      btrfs_file_extent_num_blocks(leaf,
+								   extent) << 3;
 			}
 			ret = 0;
 			goto out;
@@ -529,8 +534,8 @@ static int prepare_pages(struct btrfs_root *root,
 	u64 num_blocks;
 	u64 start_pos;
 
-	start_pos = pos & ~((u64)root->blocksize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
+	start_pos = pos & ~((u64)root->sectorsize - 1);
+	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
 			inode->i_blkbits;
 
 	memset(pages, 0, num_pages * sizeof(struct page *));

fs/btrfs/inode-item.c

@@ -20,24 +20,18 @@
 #include "disk-io.h"
 #include "transaction.h"
 
-int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *root, u64 objectid, struct btrfs_inode_item
-		       *inode_item)
+int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
+			     struct btrfs_path *path, u64 objectid)
 {
-	struct btrfs_path *path;
 	struct btrfs_key key;
 	int ret;
 	key.objectid = objectid;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	ret = btrfs_insert_item(trans, root, &key, inode_item,
-				sizeof(*inode_item));
-	btrfs_release_path(root, path);
-	btrfs_free_path(path);
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(struct btrfs_inode_item));
 	if (ret == 0 && objectid > root->highest_inode)
 		root->highest_inode = objectid;
 	return ret;
@@ -51,15 +45,15 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 	int cow = mod != 0;
 	int ret;
 	int slot;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct btrfs_key found_key;
 
 	ret = btrfs_search_slot(trans, root, location, path, ins_len, cow);
 	if (ret > 0 && btrfs_key_type(location) == BTRFS_ROOT_ITEM_KEY &&
 	    location->offset == (u64)-1 && path->slots[0] != 0) {
 		slot = path->slots[0] - 1;
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		btrfs_disk_key_to_cpu(&found_key, &leaf->items[slot].key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 		if (found_key.objectid == location->objectid &&
 		    btrfs_key_type(&found_key) == btrfs_key_type(location)) {
 			path->slots[0]--;

fs/btrfs/inode-map.c

@@ -24,8 +24,9 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
 {
 	struct btrfs_path *path;
 	int ret;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_key search_key;
+	struct btrfs_key found_key;
 	int slot;
 
 	path = btrfs_alloc_path();
@@ -39,8 +40,9 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
 	BUG_ON(ret == 0);
 	if (path->slots[0] > 0) {
 		slot = path->slots[0] - 1;
-		l = btrfs_buffer_leaf(path->nodes[0]);
-		*objectid = btrfs_disk_key_objectid(&l->items[slot].key);
+		l = path->nodes[0];
+		btrfs_item_key_to_cpu(l, &found_key, slot);
+		*objectid = found_key.objectid;
 	} else {
 		*objectid = BTRFS_FIRST_FREE_OBJECTID;
 	}
@@ -64,13 +66,12 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 	int slot = 0;
 	u64 last_ino = 0;
 	int start_found;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_key search_key;
 	u64 search_start = dirid;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
-	search_key.flags = 0;
 	search_start = root->last_inode_alloc;
 	search_start = max(search_start, BTRFS_FIRST_FREE_OBJECTID);
 	search_key.objectid = search_start;
@@ -86,9 +87,9 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 		path->slots[0]--;
 
 	while (1) {
-		l = btrfs_buffer_leaf(path->nodes[0]);
+		l = path->nodes[0];
 		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(&l->header)) {
+		if (slot >= btrfs_header_nritems(l)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret == 0)
 				continue;
@@ -103,7 +104,7 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 				last_ino : search_start;
 			goto found;
 		}
-		btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
+		btrfs_item_key_to_cpu(l, &key, slot);
 		if (key.objectid >= search_start) {
 			if (start_found) {
 				if (last_ino < search_start)

fs/btrfs/inode.c

@@ -159,10 +159,8 @@ out:
 
 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 {
-	char csum[BTRFS_CRC32_SIZE];
 	size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
 	struct inode *inode = page->mapping->host;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	char *kaddr;
 	u64 private;
@@ -173,11 +171,15 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 	if (ret) {
 		goto zeroit;
 	}
+	/*
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	char csum[BTRFS_CRC32_SIZE];
 	ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum);
 	BUG_ON(ret);
 	if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) {
 		goto zeroit;
 	}
+	*/
 	kunmap_atomic(kaddr, KM_IRQ0);
 	return 0;
 
@@ -192,7 +194,9 @@ zeroit:
 void btrfs_read_locked_inode(struct inode *inode)
 {
 	struct btrfs_path *path;
+	struct extent_buffer *leaf;
 	struct btrfs_inode_item *inode_item;
+	struct btrfs_inode_timespec *tspec;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_key location;
 	u64 alloc_group_block;
@@ -205,29 +209,37 @@ void btrfs_read_locked_inode(struct inode *inode)
 
 	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
 	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
-	if (ret) {
+	if (ret)
 		goto make_bad;
-	}
-	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-				  path->slots[0],
-				  struct btrfs_inode_item);
 
-	inode->i_mode = btrfs_inode_mode(inode_item);
-	inode->i_nlink = btrfs_inode_nlink(inode_item);
-	inode->i_uid = btrfs_inode_uid(inode_item);
-	inode->i_gid = btrfs_inode_gid(inode_item);
-	inode->i_size = btrfs_inode_size(inode_item);
-	inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
-	inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
-	inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
-	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
-	inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
-	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
-	inode->i_blocks = btrfs_inode_nblocks(inode_item);
-	inode->i_generation = btrfs_inode_generation(inode_item);
+	leaf = path->nodes[0];
+	inode_item = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_inode_item);
+
+	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
+	inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
+	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
+	inode->i_gid = btrfs_inode_gid(leaf, inode_item);
+	inode->i_size = btrfs_inode_size(leaf, inode_item);
+
+	tspec = btrfs_inode_atime(inode_item);
+	inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
+	inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
+
+	tspec = btrfs_inode_mtime(inode_item);
+	inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
+	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
+
+	tspec = btrfs_inode_ctime(inode_item);
+	inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
+	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
+
+	inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
+	inode->i_generation = btrfs_inode_generation(leaf, inode_item);
 	inode->i_rdev = 0;
-	rdev = btrfs_inode_rdev(inode_item);
-	alloc_group_block = btrfs_inode_block_group(inode_item);
+	rdev = btrfs_inode_rdev(leaf, inode_item);
+
+	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
 	BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
 						       alloc_group_block);
 
@@ -267,24 +279,35 @@ make_bad:
 	make_bad_inode(inode);
 }
 
-static void fill_inode_item(struct btrfs_inode_item *item,
+static void fill_inode_item(struct extent_buffer *leaf,
+			    struct btrfs_inode_item *item,
 			    struct inode *inode)
 {
-	btrfs_set_inode_uid(item, inode->i_uid);
-	btrfs_set_inode_gid(item, inode->i_gid);
-	btrfs_set_inode_size(item, inode->i_size);
-	btrfs_set_inode_mode(item, inode->i_mode);
-	btrfs_set_inode_nlink(item, inode->i_nlink);
-	btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
-	btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
-	btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
-	btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
-	btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
-	btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
-	btrfs_set_inode_nblocks(item, inode->i_blocks);
-	btrfs_set_inode_generation(item, inode->i_generation);
-	btrfs_set_inode_rdev(item, inode->i_rdev);
-	btrfs_set_inode_block_group(item,
+	btrfs_set_inode_uid(leaf, item, inode->i_uid);
+	btrfs_set_inode_gid(leaf, item, inode->i_gid);
+	btrfs_set_inode_size(leaf, item, inode->i_size);
+	btrfs_set_inode_mode(leaf, item, inode->i_mode);
+	btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
+
+	btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
+			       inode->i_atime.tv_sec);
+	btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
+				inode->i_atime.tv_nsec);
+
+	btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
+			       inode->i_mtime.tv_sec);
+	btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
+				inode->i_mtime.tv_nsec);
+
+	btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
+			       inode->i_ctime.tv_sec);
+	btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
+				inode->i_ctime.tv_nsec);
+
+	btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
+	btrfs_set_inode_generation(leaf, item, inode->i_generation);
+	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
+	btrfs_set_inode_block_group(leaf, item,
 				    BTRFS_I(inode)->block_group->key.objectid);
 }
 
@@ -294,6 +317,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_inode_item *inode_item;
 	struct btrfs_path *path;
+	struct extent_buffer *leaf;
 	int ret;
 
 	path = btrfs_alloc_path();
@@ -306,12 +330,12 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 		goto failed;
 	}
 
-	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-				  path->slots[0],
+	leaf = path->nodes[0];
+	inode_item = btrfs_item_ptr(leaf, path->slots[0],
 				  struct btrfs_inode_item);
 
-	fill_inode_item(inode_item, inode);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	fill_inode_item(leaf, inode_item, inode);
+	btrfs_mark_buffer_dirty(leaf);
 	btrfs_set_inode_last_trans(trans, inode);
 	ret = 0;
 failed:
@@ -330,8 +354,9 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 	const char *name = dentry->d_name.name;
 	int name_len = dentry->d_name.len;
 	int ret = 0;
-	u64 objectid;
+	struct extent_buffer *leaf;
 	struct btrfs_dir_item *di;
+	struct btrfs_key key;
 
 	path = btrfs_alloc_path();
 	if (!path) {
@@ -349,14 +374,15 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 		ret = -ENOENT;
 		goto err;
 	}
-	objectid = btrfs_disk_key_objectid(&di->location);
+	leaf = path->nodes[0];
+	btrfs_dir_item_key_to_cpu(leaf, di, &key);
 	ret = btrfs_delete_one_dir_name(trans, root, path, di);
 	if (ret)
 		goto err;
 	btrfs_release_path(root, path);
 
 	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
-					 objectid, name, name_len, -1);
+					 key.objectid, name, name_len, -1);
 	if (IS_ERR(di)) {
 		ret = PTR_ERR(di);
 		goto err;
@@ -391,12 +417,15 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 	root = BTRFS_I(dir)->root;
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, dir);
 	ret = btrfs_unlink_trans(trans, root, dir, dentry);
 	nr = trans->blocks_used;
+
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+
 	return ret;
 }
 
@@ -411,7 +440,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key found_key;
 	int found_type;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	char *goodnames = "..";
 	unsigned long nr;
 
@@ -419,10 +448,11 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	BUG_ON(!path);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, dir);
 	key.objectid = inode->i_ino;
 	key.offset = (u64)-1;
-	key.flags = (u32)-1;
+	key.type = (u8)-1;
 	while(1) {
 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 		if (ret < 0) {
@@ -435,9 +465,8 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 			goto out;
 		}
 		path->slots[0]--;
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		btrfs_disk_key_to_cpu(&found_key,
-				      &leaf->items[path->slots[0]].key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		found_type = btrfs_key_type(&found_key);
 		if (found_key.objectid != inode->i_ino) {
 			err = -ENOENT;
@@ -513,9 +542,9 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	int ret;
 	struct btrfs_path *path;
 	struct btrfs_key key;
-	struct btrfs_disk_key *found_key;
+	struct btrfs_key found_key;
 	u32 found_type;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *fi;
 	u64 extent_start = 0;
 	u64 extent_num_blocks = 0;
@@ -527,10 +556,12 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	path = btrfs_alloc_path();
 	path->reada = -1;
 	BUG_ON(!path);
+
 	/* FIXME, add redo link to tree so we don't leak on crash */
 	key.objectid = inode->i_ino;
 	key.offset = (u64)-1;
-	key.flags = (u32)-1;
+	key.type = (u8)-1;
+
 	while(1) {
 		btrfs_init_path(path);
 		fi = NULL;
@@ -542,27 +573,28 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			BUG_ON(path->slots[0] == 0);
 			path->slots[0]--;
 		}
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		found_key = &leaf->items[path->slots[0]].key;
-		found_type = btrfs_disk_key_type(found_key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		found_type = btrfs_key_type(&found_key);
 
-		if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
+		if (found_key.objectid != inode->i_ino)
 			break;
+
 		if (found_type != BTRFS_CSUM_ITEM_KEY &&
 		    found_type != BTRFS_DIR_ITEM_KEY &&
 		    found_type != BTRFS_DIR_INDEX_KEY &&
 		    found_type != BTRFS_EXTENT_DATA_KEY)
 			break;
 
-		item_end = btrfs_disk_key_offset(found_key);
+		item_end = found_key.offset;
 		if (found_type == BTRFS_EXTENT_DATA_KEY) {
-			fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-					    path->slots[0],
+			fi = btrfs_item_ptr(leaf, path->slots[0],
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) !=
+			if (btrfs_file_extent_type(leaf, fi) !=
 			    BTRFS_FILE_EXTENT_INLINE) {
-				item_end += btrfs_file_extent_num_blocks(fi) <<
-						inode->i_blkbits;
+				item_end +=
+				    btrfs_file_extent_num_blocks(leaf, fi) <<
+				    inode->i_blkbits;
 			}
 		}
 		if (found_type == BTRFS_CSUM_ITEM_KEY) {
@@ -583,7 +615,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			btrfs_set_key_type(&key, found_type);
 			continue;
 		}
-		if (btrfs_disk_key_offset(found_key) >= inode->i_size)
+		if (found_key.offset >= inode->i_size)
 			del_item = 1;
 		else
 			del_item = 0;
@@ -591,30 +623,31 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 
 		/* FIXME, shrink the extent if the ref count is only 1 */
 		if (found_type == BTRFS_EXTENT_DATA_KEY &&
-			   btrfs_file_extent_type(fi) !=
+			   btrfs_file_extent_type(leaf, fi) !=
 			   BTRFS_FILE_EXTENT_INLINE) {
 			u64 num_dec;
-			extent_start = btrfs_file_extent_disk_blocknr(fi);
+			extent_start = btrfs_file_extent_disk_blocknr(leaf, fi);
 			if (!del_item) {
 				u64 orig_num_blocks =
-					btrfs_file_extent_num_blocks(fi);
+					btrfs_file_extent_num_blocks(leaf, fi);
 				extent_num_blocks = inode->i_size -
-					btrfs_disk_key_offset(found_key) +
-					root->blocksize - 1;
+					found_key.offset + root->sectorsize - 1;
 				extent_num_blocks >>= inode->i_blkbits;
-				btrfs_set_file_extent_num_blocks(fi,
+				btrfs_set_file_extent_num_blocks(leaf, fi,
 							 extent_num_blocks);
 				num_dec = (orig_num_blocks -
 					   extent_num_blocks) << 3;
 				if (extent_start != 0) {
 					inode->i_blocks -= num_dec;
 				}
-				btrfs_mark_buffer_dirty(path->nodes[0]);
+				btrfs_mark_buffer_dirty(leaf);
 			} else {
 				extent_num_blocks =
-					btrfs_file_extent_disk_num_blocks(fi);
+					btrfs_file_extent_disk_num_blocks(leaf,
+									  fi);
 				/* FIXME blocksize != 4096 */
-				num_dec = btrfs_file_extent_num_blocks(fi) << 3;
+				num_dec = btrfs_file_extent_num_blocks(leaf,
+								       fi) << 3;
 				if (extent_start != 0) {
 					found_extent = 1;
 					inode->i_blocks -= num_dec;
@@ -725,7 +758,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 
-		u64 mask = root->blocksize - 1;
+		u64 mask = root->sectorsize - 1;
 		u64 pos = (inode->i_size + mask) & ~mask;
 		u64 block_end = attr->ia_size | mask;
 		u64 hole_size;
@@ -771,9 +804,11 @@ void btrfs_delete_inode(struct inode *inode)
 	if (is_bad_inode(inode)) {
 		goto no_delete;
 	}
+
 	inode->i_size = 0;
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, inode);
 	ret = btrfs_truncate_in_trans(trans, root, inode);
 	if (ret)
@@ -782,6 +817,7 @@ void btrfs_delete_inode(struct inode *inode)
 	if (ret)
 		goto no_delete_lock;
 	nr = trans->blocks_used;
+
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
@@ -819,7 +855,7 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
 		ret = 0;
 		goto out;
 	}
-	btrfs_disk_key_to_cpu(location, &di->location);
+	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
 out:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
@@ -856,7 +892,6 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 
 	ri = &(*sub_root)->root_item;
 	location->objectid = btrfs_root_dirid(ri);
-	location->flags = 0;
 	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 	location->offset = 0;
 
@@ -908,11 +943,14 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_inode_by_name(dir, dentry, &location);
 	mutex_unlock(&root->fs_info->fs_mutex);
+
 	if (ret < 0)
 		return ERR_PTR(ret);
+
 	inode = NULL;
 	if (location.objectid) {
 		ret = fixup_tree_root_location(root, &location, &sub_root,
@@ -952,10 +990,11 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct btrfs_item *item;
 	struct btrfs_dir_item *di;
 	struct btrfs_key key;
+	struct btrfs_key found_key;
 	struct btrfs_path *path;
 	int ret;
 	u32 nritems;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	int slot;
 	int advance;
 	unsigned char d_type;
@@ -964,15 +1003,19 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	u32 di_total;
 	u32 di_len;
 	int key_type = BTRFS_DIR_INDEX_KEY;
+	char tmp_name[32];
+	char *name_ptr;
+	int name_len;
 
 	/* FIXME, use a real flag for deciding about the key type */
 	if (root->fs_info->tree_root == root)
 		key_type = BTRFS_DIR_ITEM_KEY;
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	key.objectid = inode->i_ino;
-	key.flags = 0;
 	btrfs_set_key_type(&key, key_type);
 	key.offset = filp->f_pos;
+
 	path = btrfs_alloc_path();
 	path->reada = 2;
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
@@ -980,16 +1023,16 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 		goto err;
 	advance = 0;
 	while(1) {
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		nritems = btrfs_header_nritems(&leaf->header);
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
 		slot = path->slots[0];
 		if (advance || slot >= nritems) {
 			if (slot >= nritems -1) {
 				ret = btrfs_next_leaf(root, path);
 				if (ret)
 					break;
-				leaf = btrfs_buffer_leaf(path->nodes[0]);
-				nritems = btrfs_header_nritems(&leaf->header);
+				leaf = path->nodes[0];
+				nritems = btrfs_header_nritems(leaf);
 				slot = path->slots[0];
 			} else {
 				slot++;
@@ -997,28 +1040,48 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 			}
 		}
 		advance = 1;
-		item = leaf->items + slot;
-		if (btrfs_disk_key_objectid(&item->key) != key.objectid)
+		item = btrfs_item_nr(leaf, slot);
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+		if (found_key.objectid != key.objectid)
 			break;
-		if (btrfs_disk_key_type(&item->key) != key_type)
+		if (btrfs_key_type(&found_key) != key_type)
 			break;
-		if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
+		if (found_key.offset < filp->f_pos)
 			continue;
-		filp->f_pos = btrfs_disk_key_offset(&item->key);
+
+		filp->f_pos = found_key.offset;
 		advance = 1;
 		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 		di_cur = 0;
-		di_total = btrfs_item_size(leaf->items + slot);
+		di_total = btrfs_item_size(leaf, item);
 		while(di_cur < di_total) {
-			d_type = btrfs_filetype_table[btrfs_dir_type(di)];
-			over = filldir(dirent, (const char *)(di + 1),
-				       btrfs_dir_name_len(di),
-				       btrfs_disk_key_offset(&item->key),
-				       btrfs_disk_key_objectid(&di->location),
+			struct btrfs_key location;
+
+			name_len = btrfs_dir_name_len(leaf, di);
+			if (name_len < 32) {
+				name_ptr = tmp_name;
+			} else {
+				name_ptr = kmalloc(name_len, GFP_NOFS);
+				BUG_ON(!name_ptr);
+			}
+			read_extent_buffer(leaf, name_ptr,
+					   (unsigned long)(di + 1), name_len);
+
+			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
+			btrfs_dir_item_key_to_cpu(leaf, di, &location);
+
+			over = filldir(dirent, name_ptr, name_len,
+				       found_key.offset,
+				       location.objectid,
 				       d_type);
+
+			if (name_ptr != tmp_name)
+				kfree(name_ptr);
+
 			if (over)
 				goto nopos;
-			di_len = btrfs_dir_name_len(di) + sizeof(*di);
+			di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
 			di_cur += di_len;
 			di = (struct btrfs_dir_item *)((char *)di + di_len);
 		}
@@ -1075,11 +1138,15 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 				     int mode)
 {
 	struct inode *inode;
-	struct btrfs_inode_item inode_item;
+	struct btrfs_inode_item *inode_item;
 	struct btrfs_key *location;
+	struct btrfs_path *path;
 	int ret;
 	int owner;
 
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
 	inode = new_inode(root->fs_info->sb);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
@@ -1095,24 +1162,32 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	group = btrfs_find_block_group(root, group, 0, 0, owner);
 	BTRFS_I(inode)->block_group = group;
 
+	ret = btrfs_insert_empty_inode(trans, root, path, objectid);
+	if (ret)
+		goto fail;
+
 	inode->i_uid = current->fsuid;
 	inode->i_gid = current->fsgid;
 	inode->i_mode = mode;
 	inode->i_ino = objectid;
 	inode->i_blocks = 0;
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
-	fill_inode_item(&inode_item, inode);
+	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				  struct btrfs_inode_item);
+	fill_inode_item(path->nodes[0], inode_item, inode);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_free_path(path);
+
 	location = &BTRFS_I(inode)->location;
 	location->objectid = objectid;
-	location->flags = 0;
 	location->offset = 0;
 	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 
-	ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
-	if (ret)
-		return ERR_PTR(ret);
 	insert_inode_hash(inode);
 	return inode;
+fail:
+	btrfs_free_path(path);
+	return ERR_PTR(ret);
 }
 
 static inline u8 btrfs_inode_type(struct inode *inode)
@@ -1127,8 +1202,8 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
 	struct inode *parent_inode;
+
 	key.objectid = inode->i_ino;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
@@ -1285,14 +1360,18 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	inc_nlink(inode);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, dir);
 	atomic_inc(&inode->i_count);
 	err = btrfs_add_nondir(trans, dentry, inode);
+
 	if (err)
 		drop_inode = 1;
+
 	dir->i_sb->s_dirt = 1;
 	btrfs_update_inode_block_group(trans, dir);
 	err = btrfs_update_inode(trans, root, inode);
+
 	if (err)
 		drop_inode = 1;
 
@@ -1321,13 +1400,13 @@ static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
 
 	key.objectid = objectid;
 	key.offset = 0;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 
 	ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
 				    &key, BTRFS_FT_DIR);
 	if (ret)
 		goto error;
+
 	key.objectid = dirid;
 	ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
 				    &key, BTRFS_FT_DIR);
@@ -1350,6 +1429,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, dir);
+
 	if (IS_ERR(trans)) {
 		err = PTR_ERR(trans);
 		goto out_unlock;
@@ -1367,6 +1447,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 		err = PTR_ERR(inode);
 		goto out_fail;
 	}
+
 	drop_on_err = 1;
 	inode->i_op = &btrfs_dir_inode_operations;
 	inode->i_fop = &btrfs_dir_file_operations;
@@ -1380,9 +1461,11 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	err = btrfs_update_inode(trans, root, inode);
 	if (err)
 		goto out_fail;
+
 	err = btrfs_add_link(trans, dentry, inode);
 	if (err)
 		goto out_fail;
+
 	d_instantiate(dentry, inode);
 	drop_on_err = 0;
 	dir->i_sb->s_dirt = 1;
@@ -1392,6 +1475,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 out_fail:
 	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
+
 out_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	if (drop_on_err)
@@ -1415,8 +1499,8 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 	struct btrfs_path *path;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_file_extent_item *item;
-	struct btrfs_leaf *leaf;
-	struct btrfs_disk_key *found_key;
+	struct extent_buffer *leaf;
+	struct btrfs_key found_key;
 	struct extent_map *em = NULL;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct btrfs_trans_handle *trans = NULL;
@@ -1436,8 +1520,8 @@ again:
 			err = -ENOMEM;
 			goto out;
 		}
-		em->start = 0;
-		em->end = 0;
+		em->start = EXTENT_MAP_HOLE;
+		em->end = EXTENT_MAP_HOLE;
 	}
 	em->bdev = inode->i_sb->s_bdev;
 	ret = btrfs_lookup_file_extent(NULL, root, path,
@@ -1453,25 +1537,27 @@ again:
 		path->slots[0]--;
 	}
 
-	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	blocknr = btrfs_file_extent_disk_blocknr(item);
-	blocknr += btrfs_file_extent_offset(item);
+
+	blocknr = btrfs_file_extent_disk_blocknr(leaf, item);
+	blocknr += btrfs_file_extent_offset(leaf, item);
 
 	/* are we inside the extent that was found? */
-	found_key = &leaf->items[path->slots[0]].key;
-	found_type = btrfs_disk_key_type(found_key);
-	if (btrfs_disk_key_objectid(found_key) != objectid ||
+	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+	found_type = btrfs_key_type(&found_key);
+	if (found_key.objectid != objectid ||
 	    found_type != BTRFS_EXTENT_DATA_KEY) {
 		goto not_found;
 	}
 
-	found_type = btrfs_file_extent_type(item);
-	extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
+	found_type = btrfs_file_extent_type(leaf, item);
+	extent_start = found_key.offset;
 	if (found_type == BTRFS_FILE_EXTENT_REG) {
 		extent_end = extent_start +
-		       (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
+		       (btrfs_file_extent_num_blocks(leaf, item) <<
+			inode->i_blkbits);
 		err = 0;
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
@@ -1484,28 +1570,29 @@ again:
 			}
 			goto not_found_em;
 		}
-		if (btrfs_file_extent_disk_blocknr(item) == 0) {
+		if (btrfs_file_extent_disk_blocknr(leaf, item) == 0) {
 			em->start = extent_start;
 			em->end = extent_end - 1;
-			em->block_start = 0;
-			em->block_end = 0;
+			em->block_start = EXTENT_MAP_HOLE;
+			em->block_end = EXTENT_MAP_HOLE;
 			goto insert;
 		}
 		em->block_start = blocknr << inode->i_blkbits;
 		em->block_end = em->block_start +
-			(btrfs_file_extent_num_blocks(item) <<
+			(btrfs_file_extent_num_blocks(leaf, item) <<
 			 inode->i_blkbits) - 1;
 		em->start = extent_start;
 		em->end = extent_end - 1;
 		goto insert;
 	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
-		char *ptr;
+		unsigned long ptr;
 		char *map;
 		u32 size;
 
-		size = btrfs_file_extent_inline_len(leaf->items +
-						    path->slots[0]);
-		extent_end = extent_start | ((u64)root->blocksize - 1);
+		size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
+						    path->slots[0]));
+
+		extent_end = extent_start | ((u64)root->sectorsize - 1);
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
 			if (start < extent_start) {
@@ -1517,18 +1604,21 @@ again:
 			}
 			goto not_found_em;
 		}
+
 		em->block_start = EXTENT_MAP_INLINE;
 		em->block_end = EXTENT_MAP_INLINE;
 		em->start = extent_start;
 		em->end = extent_end;
+
 		if (!page) {
 			goto insert;
 		}
+
 		ptr = btrfs_file_extent_inline_start(item);
 		map = kmap(page);
-		memcpy(map + page_offset, ptr, size);
+		read_extent_buffer(leaf, map + page_offset, ptr, size);
 		memset(map + page_offset + size, 0,
-		       root->blocksize - (page_offset + size));
+		       root->sectorsize - (page_offset + size));
 		flush_dcache_page(page);
 		kunmap(page);
 		set_extent_uptodate(em_tree, extent_start,
@@ -1542,8 +1632,8 @@ not_found:
 	em->start = start;
 	em->end = end;
 not_found_em:
-	em->block_start = 0;
-	em->block_end = 0;
+	em->block_start = EXTENT_MAP_HOLE;
+	em->block_end = EXTENT_MAP_HOLE;
 insert:
 	btrfs_release_path(root, path);
 	if (em->start > start || em->end < start) {
@@ -1712,6 +1802,7 @@ static void btrfs_truncate(struct inode *inode)
 	ret = btrfs_truncate_in_trans(trans, root, inode);
 	btrfs_update_inode(trans, root, inode);
 	nr = trans->blocks_used;
+
 	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
@@ -1731,8 +1822,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	struct btrfs_key key;
 	struct btrfs_root_item root_item;
 	struct btrfs_inode_item *inode_item;
-	struct buffer_head *subvol;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct btrfs_root *new_root;
 	struct inode *inode;
 	struct inode *dir;
@@ -1746,34 +1836,37 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
 
-	subvol = btrfs_alloc_free_block(trans, root, 0, 0);
-	if (IS_ERR(subvol))
-		return PTR_ERR(subvol);
-	leaf = btrfs_buffer_leaf(subvol);
-	btrfs_set_header_nritems(&leaf->header, 0);
-	btrfs_set_header_level(&leaf->header, 0);
-	btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
-	btrfs_set_header_generation(&leaf->header, trans->transid);
-	btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
-	memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
-	       sizeof(leaf->header.fsid));
-	btrfs_mark_buffer_dirty(subvol);
+	leaf = btrfs_alloc_free_block(trans, root, 0, 0);
+	if (IS_ERR(leaf))
+		return PTR_ERR(leaf);
+
+	btrfs_set_header_nritems(leaf, 0);
+	btrfs_set_header_level(leaf, 0);
+	btrfs_set_header_blocknr(leaf, extent_buffer_blocknr(leaf));
+	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_owner(leaf, root->root_key.objectid);
+	write_extent_buffer(leaf, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(leaf),
+			    BTRFS_FSID_SIZE);
+	btrfs_mark_buffer_dirty(leaf);
 
 	inode_item = &root_item.inode;
 	memset(inode_item, 0, sizeof(*inode_item));
-	btrfs_set_inode_generation(inode_item, 1);
-	btrfs_set_inode_size(inode_item, 3);
-	btrfs_set_inode_nlink(inode_item, 1);
-	btrfs_set_inode_nblocks(inode_item, 1);
-	btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
+	inode_item->generation = cpu_to_le64(1);
+	inode_item->size = cpu_to_le64(3);
+	inode_item->nlink = cpu_to_le32(1);
+	inode_item->nblocks = cpu_to_le64(1);
+	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 
-	btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
+	btrfs_set_root_blocknr(&root_item, extent_buffer_blocknr(leaf));
 	btrfs_set_root_refs(&root_item, 1);
-	btrfs_set_root_blocks_used(&root_item, 0);
+	btrfs_set_root_used(&root_item, 0);
+
 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 	root_item.drop_level = 0;
-	brelse(subvol);
-	subvol = NULL;
+
+	free_extent_buffer(leaf);
+	leaf = NULL;
 
 	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
 				       0, &objectid);
@@ -1784,7 +1877,6 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 
 	key.objectid = objectid;
 	key.offset = 1;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&root_item);
@@ -1845,7 +1937,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key key;
 	struct btrfs_root_item new_root_item;
-	struct buffer_head *tmp;
+	struct extent_buffer *tmp;
 	int ret;
 	int err;
 	u64 objectid;
@@ -1876,10 +1968,11 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 
 	key.objectid = objectid;
 	key.offset = 1;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+
 	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
-	btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
+	btrfs_set_root_blocknr(&new_root_item,
+			       extent_buffer_blocknr(root->node));
 
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);
@@ -1904,8 +1997,10 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 fail:
 	nr = trans->blocks_used;
 	err = btrfs_commit_transaction(trans, root);
+
 	if (err && !ret)
 		ret = err;
+
 	mutex_unlock(&root->fs_info->fs_mutex);
 	up_write(&root->snap_sem);
 	btrfs_btree_balance_dirty(root, nr);
@@ -1986,7 +2081,7 @@ static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
 
 	if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
 		return -EFAULT;
-	
+
 	namelen = strlen(vol_args.name);
 	if (namelen > BTRFS_VOL_NAME_MAX)
 		return -EINVAL;
@@ -2164,8 +2259,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
 		return -ENOTEMPTY;
 	}
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, new_dir);
 	path = btrfs_alloc_path();
 	if (!path) {
@@ -2177,9 +2274,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 	old_dir->i_ctime = old_dir->i_mtime = ctime;
 	new_dir->i_ctime = new_dir->i_mtime = ctime;
 	old_inode->i_ctime = ctime;
+
 	if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
 		struct btrfs_key *location = &BTRFS_I(new_dir)->location;
-		u64 old_parent_oid;
+		struct btrfs_key old_parent_key;
 		di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
 					   "..", 2, -1);
 		if (IS_ERR(di)) {
@@ -2190,7 +2288,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 			ret = -ENOENT;
 			goto out_fail;
 		}
-		old_parent_oid = btrfs_disk_key_objectid(&di->location);
+		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
 		ret = btrfs_del_item(trans, root, path);
 		if (ret) {
 			goto out_fail;
@@ -2199,7 +2297,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 
 		di = btrfs_lookup_dir_index_item(trans, root, path,
 						 old_inode->i_ino,
-						 old_parent_oid,
+						 old_parent_key.objectid,
 						 "..", 2, -1);
 		if (IS_ERR(di)) {
 			ret = PTR_ERR(di);
@@ -2257,8 +2355,9 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	u64 objectid;
 	int name_len;
 	int datasize;
-	char *ptr;
+	unsigned long ptr;
 	struct btrfs_file_extent_item *ei;
+	struct extent_buffer *leaf;
 	unsigned long nr;
 
 	name_len = strlen(symname) + 1;
@@ -2302,7 +2401,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	BUG_ON(!path);
 	key.objectid = inode->i_ino;
 	key.offset = 0;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
 	datasize = btrfs_file_extent_calc_inline_size(name_len);
 	err = btrfs_insert_empty_item(trans, root, path, &key,
@@ -2311,16 +2409,17 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 		drop_inode = 1;
 		goto out_unlock;
 	}
-	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-	       path->slots[0], struct btrfs_file_extent_item);
-	btrfs_set_file_extent_generation(ei, trans->transid);
-	btrfs_set_file_extent_type(ei,
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
+	btrfs_set_file_extent_type(leaf, ei,
 				   BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
-	btrfs_memcpy(root, path->nodes[0]->b_data,
-		     ptr, symname, name_len);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	write_extent_buffer(leaf, symname, ptr, name_len);
+	btrfs_mark_buffer_dirty(leaf);
 	btrfs_free_path(path);
+
 	inode->i_op = &btrfs_symlink_inode_operations;
 	inode->i_mapping->a_ops = &btrfs_symlink_aops;
 	inode->i_size = name_len - 1;

fs/btrfs/print-tree.c

@@ -20,10 +20,10 @@
 #include "disk-io.h"
 #include "print-tree.h"
 
-void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
+void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 {
 	int i;
-	u32 nr = btrfs_header_nritems(&l->header);
+	u32 nr = btrfs_header_nritems(l);
 	struct btrfs_item *item;
 	struct btrfs_extent_item *ei;
 	struct btrfs_root_item *ri;
@@ -31,119 +31,113 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 	struct btrfs_inode_item *ii;
 	struct btrfs_block_group_item *bi;
 	struct btrfs_file_extent_item *fi;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
 	u32 type;
 
 	printk("leaf %llu total ptrs %d free space %d\n",
-		(unsigned long long)btrfs_header_blocknr(&l->header), nr,
+		(unsigned long long)btrfs_header_blocknr(l), nr,
 		btrfs_leaf_free_space(root, l));
 	for (i = 0 ; i < nr ; i++) {
-		item = l->items + i;
-		type = btrfs_disk_key_type(&item->key);
+		item = btrfs_item_nr(l, i);
+		btrfs_item_key_to_cpu(l, &key, i);
+		type = btrfs_key_type(&key);
 		printk("\titem %d key (%llu %x %llu) itemoff %d itemsize %d\n",
 			i,
-			(unsigned long long)btrfs_disk_key_objectid(&item->key),
-			btrfs_disk_key_flags(&item->key),
-			(unsigned long long)btrfs_disk_key_offset(&item->key),
-			btrfs_item_offset(item),
-			btrfs_item_size(item));
+			(unsigned long long)key.objectid, type,
+			(unsigned long long)key.offset,
+			btrfs_item_offset(l, item), btrfs_item_size(l, item));
 		switch (type) {
 		case BTRFS_INODE_ITEM_KEY:
 			ii = btrfs_item_ptr(l, i, struct btrfs_inode_item);
 			printk("\t\tinode generation %llu size %llu mode %o\n",
-			       (unsigned long long)btrfs_inode_generation(ii),
-			       (unsigned long long)btrfs_inode_size(ii),
-			       btrfs_inode_mode(ii));
+		              (unsigned long long)btrfs_inode_generation(l, ii),
+			      (unsigned long long)btrfs_inode_size(l, ii),
+			       btrfs_inode_mode(l, ii));
 			break;
 		case BTRFS_DIR_ITEM_KEY:
 			di = btrfs_item_ptr(l, i, struct btrfs_dir_item);
+			btrfs_dir_item_key_to_cpu(l, di, &found_key);
 			printk("\t\tdir oid %llu flags %u type %u\n",
-				(unsigned long long)btrfs_disk_key_objectid(
-							    &di->location),
-				btrfs_dir_flags(di),
-				btrfs_dir_type(di));
-			printk("\t\tname %.*s\n",
-			       btrfs_dir_name_len(di),(char *)(di + 1));
+				(unsigned long long)found_key.objectid,
+				btrfs_dir_flags(l, di),
+				btrfs_dir_type(l, di));
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
 			printk("\t\troot data blocknr %llu refs %u\n",
-				(unsigned long long)btrfs_root_blocknr(ri),
-				btrfs_root_refs(ri));
+				(unsigned long long)btrfs_disk_root_blocknr(l, ri),
+				btrfs_disk_root_refs(l, ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
 			ei = btrfs_item_ptr(l, i, struct btrfs_extent_item);
 			printk("\t\textent data refs %u\n",
-				btrfs_extent_refs(ei));
+				btrfs_extent_refs(l, ei));
 			break;
 
 		case BTRFS_EXTENT_DATA_KEY:
 			fi = btrfs_item_ptr(l, i,
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) ==
+			if (btrfs_file_extent_type(l, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE) {
 				printk("\t\tinline extent data size %u\n",
-			           btrfs_file_extent_inline_len(l->items + i));
+			           btrfs_file_extent_inline_len(l, item));
 				break;
 			}
 			printk("\t\textent data disk block %llu nr %llu\n",
-			       (unsigned long long)btrfs_file_extent_disk_blocknr(fi),
-			       (unsigned long long)btrfs_file_extent_disk_num_blocks(fi));
+			       (unsigned long long)btrfs_file_extent_disk_blocknr(l, fi),
+			       (unsigned long long)btrfs_file_extent_disk_num_blocks(l, fi));
 			printk("\t\textent data offset %llu nr %llu\n",
-			  (unsigned long long)btrfs_file_extent_offset(fi),
-			  (unsigned long long)btrfs_file_extent_num_blocks(fi));
+			  (unsigned long long)btrfs_file_extent_offset(l, fi),
+			  (unsigned long long)btrfs_file_extent_num_blocks(l, fi));
 			break;
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
 					    struct btrfs_block_group_item);
 			printk("\t\tblock group used %llu\n",
-			       (unsigned long long)btrfs_block_group_used(bi));
-			break;
-		case BTRFS_STRING_ITEM_KEY:
-			printk("\t\titem data %.*s\n", btrfs_item_size(item),
-				btrfs_leaf_data(l) + btrfs_item_offset(item));
+			       (unsigned long long)btrfs_disk_block_group_used(l, bi));
 			break;
 		};
 	}
 }
 
-void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t)
+void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 {
 	int i;
 	u32 nr;
-	struct btrfs_node *c;
+	struct btrfs_key key;
 
-	if (!t)
+	if (!c)
 		return;
-	c = btrfs_buffer_node(t);
-	nr = btrfs_header_nritems(&c->header);
+	nr = btrfs_header_nritems(c);
 	if (btrfs_is_leaf(c)) {
-		btrfs_print_leaf(root, (struct btrfs_leaf *)c);
+		btrfs_print_leaf(root, c);
 		return;
 	}
 	printk("node %llu level %d total ptrs %d free spc %u\n",
-	       (unsigned long long)btrfs_header_blocknr(&c->header),
-	       btrfs_header_level(&c->header), nr,
+	       (unsigned long long)btrfs_header_blocknr(c),
+	       btrfs_header_level(c), nr,
 	       (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	for (i = 0; i < nr; i++) {
+		btrfs_node_key_to_cpu(c, &key, i);
 		printk("\tkey %d (%llu %u %llu) block %llu\n",
 		       i,
-		       (unsigned long long)c->ptrs[i].key.objectid,
-		       c->ptrs[i].key.flags,
-		       (unsigned long long)c->ptrs[i].key.offset,
+		       (unsigned long long)key.objectid,
+		       key.type,
+		       (unsigned long long)key.offset,
 		       (unsigned long long)btrfs_node_blockptr(c, i));
 	}
 	for (i = 0; i < nr; i++) {
-		struct buffer_head *next_buf = read_tree_block(root,
+		struct extent_buffer *next = read_tree_block(root,
 						btrfs_node_blockptr(c, i));
-		struct btrfs_node *next = btrfs_buffer_node(next_buf);
 		if (btrfs_is_leaf(next) &&
-		    btrfs_header_level(&c->header) != 1)
+		    btrfs_header_level(c) != 1)
 			BUG();
-		if (btrfs_header_level(&next->header) !=
-			btrfs_header_level(&c->header) - 1)
+		if (btrfs_header_level(next) !=
+			btrfs_header_level(c) - 1)
 			BUG();
-		btrfs_print_tree(root, next_buf);
-		btrfs_block_release(root, next_buf);
+		btrfs_print_tree(root, next);
+		free_extent_buffer(next);
 	}
 }
 

fs/btrfs/print-tree.h

@@ -18,6 +18,6 @@
 
 #ifndef __PRINT_TREE_
 #define __PRINT_TREE_
-void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l);
-void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t);
+void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l);
+void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *t);
 #endif

fs/btrfs/root-tree.c

@@ -26,12 +26,13 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 {
 	struct btrfs_path *path;
 	struct btrfs_key search_key;
-	struct btrfs_leaf *l;
+	struct btrfs_key found_key;
+	struct extent_buffer *l;
 	int ret;
 	int slot;
 
 	search_key.objectid = objectid;
-	search_key.flags = (u32)-1;
+	search_key.type = (u8)-1;
 	search_key.offset = (u64)-1;
 
 	path = btrfs_alloc_path();
@@ -39,17 +40,19 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
 	if (ret < 0)
 		goto out;
+
 	BUG_ON(ret == 0);
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	l = path->nodes[0];
 	BUG_ON(path->slots[0] == 0);
 	slot = path->slots[0] - 1;
-	if (btrfs_disk_key_objectid(&l->items[slot].key) != objectid) {
+	btrfs_item_key_to_cpu(l, &found_key, slot);
+	if (found_key.objectid != objectid) {
 		ret = 1;
 		goto out;
 	}
-	memcpy(item, btrfs_item_ptr(l, slot, struct btrfs_root_item),
-		sizeof(*item));
-	btrfs_disk_key_to_cpu(key, &l->items[slot].key);
+	read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot),
+			   sizeof(*item));
+	memcpy(key, &found_key, sizeof(found_key));
 	ret = 0;
 out:
 	btrfs_release_path(root, path);
@@ -62,10 +65,10 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *item)
 {
 	struct btrfs_path *path;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	int ret;
 	int slot;
-	struct btrfs_root_item *update_item;
+	unsigned long ptr;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -73,10 +76,10 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	l = path->nodes[0];
 	slot = path->slots[0];
-	update_item = btrfs_item_ptr(l, slot, struct btrfs_root_item);
-	btrfs_memcpy(root, l, update_item, item, sizeof(*item));
+	ptr = btrfs_item_ptr_offset(l, slot);
+	write_extent_buffer(l, item, ptr, sizeof(*item));
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 out:
 	btrfs_release_path(root, path);
@@ -103,11 +106,10 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
 	struct btrfs_path *path;
 	int ret;
 	u32 nritems;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	int slot;
 
 	key.objectid = objectid;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 	key.offset = 0;
 	path = btrfs_alloc_path();
@@ -117,19 +119,19 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
 	if (ret < 0)
 		goto err;
 	while(1) {
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		nritems = btrfs_header_nritems(&leaf->header);
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
 		slot = path->slots[0];
 		if (slot >= nritems) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret)
 				break;
-			leaf = btrfs_buffer_leaf(path->nodes[0]);
-			nritems = btrfs_header_nritems(&leaf->header);
+			leaf = path->nodes[0];
+			nritems = btrfs_header_nritems(leaf);
 			slot = path->slots[0];
 		}
-		item = leaf->items + slot;
-		btrfs_disk_key_to_cpu(&key, &item->key);
+		item = btrfs_item_nr(leaf, slot);
+		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY)
 			goto next;
 
@@ -140,7 +142,7 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
 			break;
 
 		ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item);
-		if (btrfs_root_refs(ri) != 0)
+		if (btrfs_disk_root_refs(leaf, ri) != 0)
 			goto next;
 
 		dead_root = btrfs_read_fs_root_no_radix(root->fs_info, &key);
@@ -170,6 +172,7 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	int ret;
 	u32 refs;
 	struct btrfs_root_item *ri;
+	struct extent_buffer *leaf;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -177,10 +180,10 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
-	ri = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-			    path->slots[0], struct btrfs_root_item);
+	leaf = path->nodes[0];
+	ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item);
 
-	refs = btrfs_root_refs(ri);
+	refs = btrfs_disk_root_refs(leaf, ri);
 	BUG_ON(refs != 0);
 	ret = btrfs_del_item(trans, root, path);
 out:

fs/btrfs/super.c

@@ -41,7 +41,7 @@
 #include "ioctl.h"
 #include "print-tree.h"
 
-#define BTRFS_SUPER_MAGIC 0x9123682E
+#define BTRFS_SUPER_MAGIC 0x9123683E
 
 static struct super_operations btrfs_super_ops;
 
@@ -115,13 +115,12 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 		return -EIO;
 	}
 	sb->s_fs_info = tree_root;
-	disk_super = tree_root->fs_info->disk_super;
+	disk_super = &tree_root->fs_info->super_copy;
 	inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 				  tree_root);
 	bi = BTRFS_I(inode);
 	bi->location.objectid = inode->i_ino;
 	bi->location.offset = 0;
-	bi->location.flags = 0;
 	bi->root = tree_root;
 
 	btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
@@ -281,6 +280,7 @@ error_s:
 error_bdev:
 	close_bdev_excl(bdev);
 error:
+printk("get_sb failed\n");
 	return error;
 }
 /* end copy & paste */
@@ -295,6 +295,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type,
 	ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data,
 			btrfs_fill_super, mnt,
 			subvol_name ? subvol_name : "default");
+printk("btrfs_get_sb returns %d\n", ret);
 	return ret;
 }
 

fs/btrfs/sysfs.c

@@ -31,31 +31,31 @@
 static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_root_blocks_used(&root->root_item));
+		(unsigned long long)btrfs_root_used(&root->root_item));
 }
 
 static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_root_block_limit(&root->root_item));
+		(unsigned long long)btrfs_root_limit(&root->root_item));
 }
 
 static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_blocks_used(fs->disk_super));
+		(unsigned long long)btrfs_super_blocks_used(&fs->super_copy));
 }
 
 static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_total_blocks(fs->disk_super));
+		(unsigned long long)btrfs_super_total_blocks(&fs->super_copy));
 }
 
 static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_blocksize(fs->disk_super));
+		(unsigned long long)btrfs_super_sectorsize(&fs->super_copy));
 }
 
 /* this is for root attrs (subvols/snapshots) */

fs/btrfs/transaction.c

@@ -19,6 +19,7 @@
 #include <linux/fs.h>
 #include <linux/sched.h>
 #include <linux/writeback.h>
+#include <linux/pagemap.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -66,7 +67,9 @@ static int join_transaction(struct btrfs_root *root)
 		cur_trans->commit_done = 0;
 		cur_trans->start_time = get_seconds();
 		list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
-		init_bit_radix(&cur_trans->dirty_pages);
+		extent_map_tree_init(&cur_trans->dirty_pages,
+				     root->fs_info->btree_inode->i_mapping,
+				     GFP_NOFS);
 	} else {
 		cur_trans->num_writers++;
 		cur_trans->num_joined++;
@@ -88,7 +91,7 @@ static int record_root_in_trans(struct btrfs_root *root)
 				   (unsigned long)root->root_key.objectid,
 				   BTRFS_ROOT_DEFRAG_TAG);
 			root->commit_root = root->node;
-			get_bh(root->node);
+			extent_buffer_get(root->node);
 		} else {
 			WARN_ON(1);
 		}
@@ -144,29 +147,30 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
 {
-	unsigned long gang[16];
 	int ret;
-	int i;
 	int err;
 	int werr = 0;
+	struct extent_map_tree *dirty_pages;
 	struct page *page;
-	struct radix_tree_root *dirty_pages;
 	struct inode *btree_inode = root->fs_info->btree_inode;
+	u64 start;
+	u64 end;
+	unsigned long index;
 
 	if (!trans || !trans->transaction) {
 		return filemap_write_and_wait(btree_inode->i_mapping);
 	}
 	dirty_pages = &trans->transaction->dirty_pages;
 	while(1) {
-		ret = find_first_radix_bit(dirty_pages, gang,
-					   0, ARRAY_SIZE(gang));
-		if (!ret)
+		ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
+					    EXTENT_DIRTY);
+		if (ret)
 			break;
-		for (i = 0; i < ret; i++) {
-			/* FIXME EIO */
-			clear_radix_bit(dirty_pages, gang[i]);
-			page = find_lock_page(btree_inode->i_mapping,
-					      gang[i]);
+		clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
+		while(start <= end) {
+			index = start >> PAGE_CACHE_SHIFT;
+			start = (index + 1) << PAGE_CACHE_SHIFT;
+			page = find_lock_page(btree_inode->i_mapping, index);
 			if (!page)
 				continue;
 			if (PageWriteback(page)) {
@@ -202,10 +206,11 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 	btrfs_write_dirty_block_groups(trans, extent_root);
 	while(1) {
 		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block == bh_blocknr(extent_root->node))
+		if (old_extent_block ==
+		    extent_buffer_blocknr(extent_root->node))
 			break;
 		btrfs_set_root_blocknr(&extent_root->root_item,
-				       bh_blocknr(extent_root->node));
+			       extent_buffer_blocknr(extent_root->node));
 		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
@@ -279,9 +284,9 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 				     (unsigned long)root->root_key.objectid,
 				     BTRFS_ROOT_TRANS_TAG);
 			if (root->commit_root == root->node) {
-				WARN_ON(bh_blocknr(root->node) !=
+				WARN_ON(extent_buffer_blocknr(root->node) !=
 					btrfs_root_blocknr(&root->root_item));
-				brelse(root->commit_root);
+				free_extent_buffer(root->commit_root);
 				root->commit_root = NULL;
 
 				/* make sure to update the root on disk
@@ -310,7 +315,7 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 
 			root->root_key.offset = root->fs_info->generation;
 			btrfs_set_root_blocknr(&root->root_item,
-					       bh_blocknr(root->node));
+				       extent_buffer_blocknr(root->node));
 			err = btrfs_insert_root(trans, root->fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
@@ -389,10 +394,10 @@ int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info)
 		for (i = 0; i < ret; i++) {
 			root = gang[i];
 			last = root->root_key.objectid + 1;
-			btrfs_defrag_root(root, 1);
+			// btrfs_defrag_root(root, 1);
 		}
 	}
-	btrfs_defrag_root(info->extent_root, 1);
+	// btrfs_defrag_root(info->extent_root, 1);
 	return err;
 }
 
@@ -414,7 +419,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		dirty = list_entry(list->next, struct dirty_root, list);
 		list_del_init(&dirty->list);
 
-		num_blocks = btrfs_root_blocks_used(&dirty->root->root_item);
+		num_blocks = btrfs_root_used(&dirty->root->root_item);
 		root = dirty->latest_root;
 
 		while(1) {
@@ -441,11 +446,11 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		}
 		BUG_ON(ret);
 
-		num_blocks -= btrfs_root_blocks_used(&dirty->root->root_item);
-		blocks_used = btrfs_root_blocks_used(&root->root_item);
+		num_blocks -= btrfs_root_used(&dirty->root->root_item);
+		blocks_used = btrfs_root_used(&root->root_item);
 		if (num_blocks) {
 			record_root_in_trans(root);
-			btrfs_set_root_blocks_used(&root->root_item,
+			btrfs_set_root_used(&root->root_item,
 						   blocks_used - num_blocks);
 		}
 		ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
@@ -553,9 +558,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_set_super_generation(&root->fs_info->super_copy,
 				   cur_trans->transid);
 	btrfs_set_super_root(&root->fs_info->super_copy,
-			     bh_blocknr(root->fs_info->tree_root->node));
-	memcpy(root->fs_info->disk_super, &root->fs_info->super_copy,
-	       sizeof(root->fs_info->super_copy));
+		     extent_buffer_blocknr(root->fs_info->tree_root->node));
+
+	write_extent_buffer(root->fs_info->sb_buffer,
+			    &root->fs_info->super_copy, 0,
+			    sizeof(root->fs_info->super_copy));
 
 	btrfs_copy_pinned(root, &pinned_copy);
 

fs/btrfs/transaction.h

@@ -28,7 +28,7 @@ struct btrfs_transaction {
 	int use_count;
 	int commit_done;
 	struct list_head list;
-	struct radix_tree_root dirty_pages;
+	struct extent_map_tree dirty_pages;
 	unsigned long start_time;
 	wait_queue_head_t writer_wait;
 	wait_queue_head_t commit_wait;
@@ -83,5 +83,6 @@ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest,
 int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info);
 int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
 int btrfs_clean_old_snapshots(struct btrfs_root *root);
-
+int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root);
 #endif

fs/btrfs/tree-defrag.c

@@ -23,14 +23,14 @@
 #include "transaction.h"
 
 static void reada_defrag(struct btrfs_root *root,
-			 struct btrfs_node *node)
+			 struct extent_buffer *node)
 {
 	int i;
 	u32 nritems;
 	u64 blocknr;
 	int ret;
 
-	nritems = btrfs_header_nritems(&node->header);
+	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
 		blocknr = btrfs_node_blockptr(node, i);
 		ret = readahead_tree_block(root, blocknr);
@@ -44,8 +44,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 			    struct btrfs_path *path, int *level,
 			    int cache_only, u64 *last_ret)
 {
-	struct buffer_head *next;
-	struct buffer_head *cur;
+	struct extent_buffer *next;
+	struct extent_buffer *cur;
 	u64 blocknr;
 	int ret = 0;
 	int is_extent = 0;
@@ -62,13 +62,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 		cur = path->nodes[*level];
 
 		if (!cache_only && *level > 1 && path->slots[*level] == 0)
-			reada_defrag(root, btrfs_buffer_node(cur));
+			reada_defrag(root, cur);
 
-		if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
+		if (btrfs_header_level(cur) != *level)
 			WARN_ON(1);
 
 		if (path->slots[*level] >=
-		    btrfs_header_nritems(btrfs_buffer_header(cur)))
+		    btrfs_header_nritems(cur))
 			break;
 
 		if (*level == 1) {
@@ -80,14 +80,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 
 			break;
 		}
-		blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur),
-					      path->slots[*level]);
+		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
 
 		if (cache_only) {
 			next = btrfs_find_tree_block(root, blocknr);
-			if (!next || !buffer_uptodate(next) ||
-			   buffer_locked(next) || !buffer_defrag(next)) {
-				brelse(next);
+			/* FIXME, test for defrag */
+			if (!next || !btrfs_buffer_uptodate(next)) {
+				free_extent_buffer(next);
 				path->slots[*level]++;
 				continue;
 			}
@@ -106,16 +105,18 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
-			btrfs_block_release(root, path->nodes[*level-1]);
+			free_extent_buffer(path->nodes[*level-1]);
 		path->nodes[*level-1] = next;
-		*level = btrfs_header_level(btrfs_buffer_header(next));
+		*level = btrfs_header_level(next);
 		path->slots[*level] = 0;
 	}
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
+#if 0
 	clear_buffer_defrag(path->nodes[*level]);
 	clear_buffer_defrag_done(path->nodes[*level]);
-	btrfs_block_release(root, path->nodes[*level]);
+#endif
+	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
 	WARN_ON(ret);
@@ -129,24 +130,25 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans,
 {
 	int i;
 	int slot;
-	struct btrfs_node *node;
+	struct extent_buffer *node;
 
 	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
-		if (slot < btrfs_header_nritems(
-		    btrfs_buffer_header(path->nodes[i])) - 1) {
+		if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
 			path->slots[i]++;
 			*level = i;
-			node = btrfs_buffer_node(path->nodes[i]);
+			node = path->nodes[i];
 			WARN_ON(i == 0);
-			btrfs_disk_key_to_cpu(&root->defrag_progress,
-					      &node->ptrs[path->slots[i]].key);
+			btrfs_node_key_to_cpu(node, &root->defrag_progress,
+					      path->slots[i]);
 			root->defrag_level = i;
 			return 0;
 		} else {
+			/*
 			clear_buffer_defrag(path->nodes[*level]);
 			clear_buffer_defrag_done(path->nodes[*level]);
-			btrfs_block_release(root, path->nodes[*level]);
+			*/
+			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
 		}
@@ -158,7 +160,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, int cache_only)
 {
 	struct btrfs_path *path = NULL;
-	struct buffer_head *tmp;
+	struct extent_buffer *tmp;
 	int ret = 0;
 	int wret;
 	int level;
@@ -172,17 +174,18 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 
 	if (root->ref_cows == 0 && !is_extent)
 		goto out;
+
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	level = btrfs_header_level(root->node);
 	orig_level = level;
 	if (level == 0) {
 		goto out;
 	}
 	if (root->defrag_progress.objectid == 0) {
-		get_bh(root->node);
+		extent_buffer_get(root->node);
 		ret = btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
 		BUG_ON(ret);
 		ret = btrfs_realloc_node(trans, root, root->node, cache_only,
@@ -200,12 +203,15 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 
 		if (is_extent)
 			btrfs_extent_post_op(trans, root);
+
 		if (wret < 0) {
 			ret = wret;
 			goto out;
 		}
+
 		while(level > 0 && !path->nodes[level])
 			level--;
+
 		if (!path->nodes[level]) {
 			ret = 0;
 			goto out;
@@ -230,7 +236,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path->nodes[i]) {
-			btrfs_block_release(root, path->nodes[i]);
+			free_extent_buffer(path->nodes[i]);
 			path->nodes[i] = 0;
 		}
 	}