Btrfs: Add inode map, and the start of file extent items

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

fs/btrfs/Makefile

@@ -3,7 +3,8 @@ CFLAGS = -g -Wall -Werror
 headers = radix-tree.h ctree.h disk-io.h kerncompat.h print-tree.h list.h \
 	  transaction.h
 objects = ctree.o disk-io.o radix-tree.o mkfs.o extent-tree.o print-tree.o \
-	  root-tree.o dir-item.o hash.o file-item.o inode-item.o
+	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
+	  inode-map.o \
 
 # if you don't have sparse installed, use ls instead
 CHECKFLAGS=-D__linux__ -Dlinux -D__STDC__ -Dunix -D__unix__ -Wbitwise \

fs/btrfs/ctree.h

@@ -10,7 +10,8 @@ struct btrfs_trans_handle;
 
 #define BTRFS_ROOT_TREE_OBJECTID 1
 #define BTRFS_EXTENT_TREE_OBJECTID 2
-#define BTRFS_FS_TREE_OBJECTID 3
+#define BTRFS_INODE_MAP_OBJECTID 3
+#define BTRFS_FS_TREE_OBJECTID 4
 
 /*
  * the key defines the order in the tree, and so it also defines (optimal)
@@ -178,31 +179,65 @@ struct btrfs_root_item {
 	__le64 block_limit;
 	__le64 blocks_used;
 	__le32 refs;
-};
+} __attribute__ ((__packed__));
 
-/*
- * in ram representation of the tree.  extent_root is used for all allocations
- * and for the extent tree extent_root root.  current_insert is used
- * only for the extent tree.
- */
-struct btrfs_root {
-	struct btrfs_buffer *node;
-	struct btrfs_buffer *commit_root;
+struct btrfs_file_extent_item {
+	/*
+	 * disk space consumed by the extent, checksum blocks are included
+	 * in these numbers
+	 */
+	__le64 disk_blocknr;
+	__le64 disk_num_blocks;
+	/*
+	 * the logical offset in file bytes (no csums)
+	 * this extent record is for.  This allows a file extent to point
+	 * into the middle of an existing extent on disk, sharing it
+	 * between two snapshots (useful if some bytes in the middle of the
+	 * extent have changed
+	 */
+	__le64 offset;
+	/*
+	 * the logical number of file blocks (no csums included)
+	 */
+	__le64 num_blocks;
+} __attribute__ ((__packed__));
+
+struct btrfs_inode_map_item {
+	struct btrfs_disk_key key;
+} __attribute__ ((__packed__));
+
+struct btrfs_fs_info {
+	struct btrfs_root *fs_root;
 	struct btrfs_root *extent_root;
 	struct btrfs_root *tree_root;
+	struct btrfs_root *inode_root;
 	struct btrfs_key current_insert;
 	struct btrfs_key last_insert;
-	int fp;
 	struct radix_tree_root cache_radix;
 	struct radix_tree_root pinned_radix;
 	struct list_head trans;
 	struct list_head cache;
+	u64 last_inode_alloc;
+	u64 last_inode_alloc_dirid;
 	int cache_size;
-	int ref_cows;
+	int fp;
+	struct btrfs_trans_handle *running_transaction;
+};
+
+/*
+ * in ram representation of the tree.  extent_root is used for all allocations
+ * and for the extent tree extent_root root.  current_insert is used
+ * only for the extent tree.
+ */
+struct btrfs_root {
+	struct btrfs_buffer *node;
+	struct btrfs_buffer *commit_root;
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
+	struct btrfs_fs_info *fs_info;
 	u32 blocksize;
-	struct btrfs_trans_handle *running_transaction;
+	int ref_cows;
+	u32 type;
 };
 
 /* the lower bits in the key flags defines the item type */
@@ -240,11 +275,17 @@ struct btrfs_root {
  * are used, and how many references there are to each block
  */
 #define BTRFS_EXTENT_ITEM_KEY	6
+
+/*
+ * the inode map records which inode numbers are in use and where
+ * they actually live on disk
+ */
+#define BTRFS_INODE_MAP_ITEM_KEY 7
 /*
  * string items are for debugging.  They just store a short string of
  * data in the FS
  */
-#define BTRFS_STRING_ITEM_KEY	7
+#define BTRFS_STRING_ITEM_KEY	8
 
 static inline u64 btrfs_inode_generation(struct btrfs_inode_item *i)
 {
@@ -654,6 +695,57 @@ static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l)
 {
 	return (u8 *)l->items;
 }
+
+static inline u64 btrfs_file_extent_disk_blocknr(struct btrfs_file_extent_item
+						 *e)
+{
+	return le64_to_cpu(e->disk_blocknr);
+}
+
+static inline void btrfs_set_file_extent_disk_blocknr(struct
+						      btrfs_file_extent_item
+						      *e, u64 val)
+{
+	e->disk_blocknr = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_file_extent_disk_num_blocks(struct
+						    btrfs_file_extent_item *e)
+{
+	return le64_to_cpu(e->disk_num_blocks);
+}
+
+static inline void btrfs_set_file_extent_disk_num_blocks(struct
+							 btrfs_file_extent_item
+							 *e, u64 val)
+{
+	e->disk_num_blocks = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_file_extent_offset(struct btrfs_file_extent_item *e)
+{
+	return le64_to_cpu(e->offset);
+}
+
+static inline void btrfs_set_file_extent_offset(struct btrfs_file_extent_item
+						*e, u64 val)
+{
+	e->offset = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_file_extent_num_blocks(struct btrfs_file_extent_item
+					       *e)
+{
+	return le64_to_cpu(e->num_blocks);
+}
+
+static inline void btrfs_set_file_extent_num_blocks(struct
+						    btrfs_file_extent_item *e,
+						    u64 val)
+{
+	e->num_blocks = cpu_to_le64(val);
+}
+
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
@@ -701,4 +793,13 @@ int btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 			  int name_len, int mod);
 int btrfs_match_dir_item_name(struct btrfs_root *root, struct btrfs_path *path,
 			      char *name, int name_len);
+int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *fs_root,
+			     u64 dirid, u64 *objectid);
+int btrfs_insert_inode_map(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   u64 objectid, struct btrfs_key *location);
+int btrfs_lookup_inode_map(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root, struct btrfs_path *path,
+			   u64 objectid, int mod);
 #endif

fs/btrfs/debug-tree.c

@@ -15,8 +15,13 @@ int main(int ac, char **av) {
 	printf("fs tree\n");
 	btrfs_print_tree(root, root->node);
 	printf("map tree\n");
-	btrfs_print_tree(root->extent_root, root->extent_root->node);
+	btrfs_print_tree(root->fs_info->extent_root,
+			 root->fs_info->extent_root->node);
+	printf("inode tree\n");
+	btrfs_print_tree(root->fs_info->inode_root,
+			 root->fs_info->inode_root->node);
 	printf("root tree\n");
-	btrfs_print_tree(root->tree_root, root->tree_root->node);
+	btrfs_print_tree(root->fs_info->tree_root,
+			 root->fs_info->tree_root->node);
 	return 0;
 }

fs/btrfs/dir-test.c

@@ -45,13 +45,26 @@ static int ins_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	int ret;
 	char buf[128];
 	unsigned long oid;
+	u64 objectid;
 	struct btrfs_path path;
+	struct btrfs_key inode_map;
 
 	find_num(radix, &oid, 0);
 	sprintf(buf, "str-%lu", oid);
 
+	ret = btrfs_find_free_objectid(trans, root, dir_oid + 1, &objectid);
+	if (ret)
+		goto error;
+
+	inode_map.objectid = objectid;
+	inode_map.flags = 0;
+	inode_map.offset = 0;
+
+	ret = btrfs_insert_inode_map(trans, root, objectid, &inode_map);
+	if (ret)
+		goto error;
 	ret = btrfs_insert_dir_item(trans, root, buf, strlen(buf), dir_oid,
-				    file_oid, 1);
+				    objectid, 1);
 	if (ret)
 		goto error;
 
@@ -120,6 +133,53 @@ static int insert_dup(struct btrfs_trans_handle *trans, struct btrfs_root
 	return 0;
 }
 
+static int del_dir_item(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root,
+			struct radix_tree_root *radix,
+			unsigned long radix_index,
+			struct btrfs_path *path)
+{
+	int ret;
+	unsigned long *ptr;
+	u64 file_objectid;
+	struct btrfs_dir_item *di;
+	struct btrfs_path map_path;
+
+	/* find the inode number of the file */
+	di = btrfs_item_ptr(&path->nodes[0]->leaf, path->slots[0],
+			    struct btrfs_dir_item);
+	file_objectid = btrfs_dir_objectid(di);
+
+	/* delete the directory item */
+	ret = btrfs_del_item(trans, root, path);
+	if (ret)
+		goto out;
+
+	/* delete the inode mapping */
+	btrfs_init_path(&map_path);
+	ret = btrfs_lookup_inode_map(trans, root, &map_path, file_objectid, -1);
+	if (ret)
+		goto out_release;
+	ret = btrfs_del_item(trans, root->fs_info->inode_root, &map_path);
+	if (ret)
+		goto out_release;
+
+	if (root->fs_info->last_inode_alloc > file_objectid)
+		root->fs_info->last_inode_alloc = file_objectid;
+	btrfs_release_path(root, &map_path);
+	ptr = radix_tree_delete(radix, radix_index);
+	if (!ptr) {
+		ret = -5555;
+		goto out;
+	}
+	return 0;
+out_release:
+	btrfs_release_path(root, &map_path);
+out:
+	printf("failed to delete %lu %d\n", radix_index, ret);
+	return -1;
+}
+
 static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct radix_tree_root *radix)
 {
@@ -127,7 +187,6 @@ static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	char buf[128];
 	unsigned long oid;
 	struct btrfs_path path;
-	unsigned long *ptr;
 
 	ret = find_num(radix, &oid, 1);
 	if (ret < 0)
@@ -138,19 +197,14 @@ static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 				    strlen(buf), -1);
 	if (ret)
 		goto out_release;
-	ret = btrfs_del_item(trans, root, &path);
+
+	ret = del_dir_item(trans, root, radix, oid, &path);
 	if (ret)
 		goto out_release;
 	btrfs_release_path(root, &path);
-	ptr = radix_tree_delete(radix, oid);
-	if (!ptr) {
-		ret = -5555;
-		goto out;
-	}
-	return 0;
+	return ret;
 out_release:
 	btrfs_release_path(root, &path);
-out:
 	printf("failed to delete %lu %d\n", oid, ret);
 	return -1;
 }
@@ -162,6 +216,8 @@ static int lookup_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	char buf[128];
 	int ret;
 	unsigned long oid;
+	u64 objectid;
+	struct btrfs_dir_item *di;
 
 	ret = find_num(radix, &oid, 1);
 	if (ret < 0)
@@ -170,6 +226,14 @@ static int lookup_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_init_path(&path);
 	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
 				    strlen(buf), 0);
+	if (!ret) {
+		di = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
+				    struct btrfs_dir_item);
+		objectid = btrfs_dir_objectid(di);
+		btrfs_release_path(root, &path);
+		btrfs_init_path(&path);
+		ret = btrfs_lookup_inode_map(trans, root, &path, objectid, 0);
+	}
 	btrfs_release_path(root, &path);
 	if (ret) {
 		printf("unable to find key %lu\n", oid);
@@ -210,7 +274,6 @@ static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 	u32 found_len;
 	int ret;
 	int slot;
-	int *ptr;
 	int count = 0;
 	char buf[128];
 	struct btrfs_dir_item *di;
@@ -241,7 +304,7 @@ static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 		BUG_ON(found_len > 128);
 		buf[found_len] = '\0';
 		found = atoi(buf + 4);
-		ret = btrfs_del_item(trans, root, &path);
+		ret = del_dir_item(trans, root, radix, found, &path);
 		count++;
 		if (ret) {
 			fprintf(stderr,
@@ -250,14 +313,10 @@ static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			return -1;
 		}
 		btrfs_release_path(root, &path);
-		ptr = radix_tree_delete(radix, found);
-		if (!ptr)
-			goto error;
 		if (!keep_running)
 			break;
 	}
 	return 0;
-error:
 	fprintf(stderr, "failed to delete from the radix %lu\n", found);
 	return -1;
 }

fs/btrfs/extent-tree.c

@@ -35,13 +35,15 @@ static int inc_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_key ins;
 	u32 refs;
 
-	find_free_extent(trans, root->extent_root, 0, 0, (u64)-1, &ins);
+	find_free_extent(trans, root->fs_info->extent_root, 0, 0, (u64)-1,
+			 &ins);
 	btrfs_init_path(&path);
 	key.objectid = blocknr;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = 1;
-	ret = btrfs_search_slot(trans, root->extent_root, &key, &path, 0, 1);
+	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, &path,
+				0, 1);
 	if (ret != 0)
 		BUG();
 	BUG_ON(ret != 0);
@@ -51,9 +53,9 @@ static int inc_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_extent_refs(item, refs + 1);
 
 	BUG_ON(list_empty(&path.nodes[0]->dirty));
-	btrfs_release_path(root->extent_root, &path);
-	finish_current_insert(trans, root->extent_root);
-	run_pending(trans, root->extent_root);
+	btrfs_release_path(root->fs_info->extent_root, &path);
+	finish_current_insert(trans, root->fs_info->extent_root);
+	run_pending(trans, root->fs_info->extent_root);
 	return 0;
 }
 
@@ -70,13 +72,14 @@ static int lookup_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 	key.offset = 1;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
-	ret = btrfs_search_slot(trans, root->extent_root, &key, &path, 0, 0);
+	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, &path,
+				0, 0);
 	if (ret != 0)
 		BUG();
 	l = &path.nodes[0]->leaf;
 	item = btrfs_item_ptr(l, path.slots[0], struct btrfs_extent_item);
 	*refs = btrfs_extent_refs(item);
-	btrfs_release_path(root->extent_root, &path);
+	btrfs_release_path(root->fs_info->extent_root, &path);
 	return 0;
 }
 
@@ -107,19 +110,20 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 	int i;
 
 	while(1) {
-		ret = radix_tree_gang_lookup(&root->pinned_radix,
-						 (void **)gang, 0,
-						 ARRAY_SIZE(gang));
+		ret = radix_tree_gang_lookup(&root->fs_info->pinned_radix,
+					     (void **)gang, 0,
+					     ARRAY_SIZE(gang));
 		if (!ret)
 			break;
 		if (!first)
 			first = gang[0];
 		for (i = 0; i < ret; i++) {
-			radix_tree_delete(&root->pinned_radix, gang[i]);
+			radix_tree_delete(&root->fs_info->pinned_radix,
+					  gang[i]);
 		}
 	}
-	root->last_insert.objectid = first;
-	root->last_insert.offset = 0;
+	root->fs_info->last_insert.objectid = first;
+	root->fs_info->last_insert.offset = 0;
 	return 0;
 }
 
@@ -138,13 +142,14 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 	ins.flags = 0;
 	btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
 
-	for (i = 0; i < extent_root->current_insert.flags; i++) {
-		ins.objectid = extent_root->current_insert.objectid + i;
+	for (i = 0; i < extent_root->fs_info->current_insert.flags; i++) {
+		ins.objectid = extent_root->fs_info->current_insert.objectid +
+				i;
 		ret = btrfs_insert_item(trans, extent_root, &ins, &extent_item,
 					sizeof(extent_item));
 		BUG_ON(ret);
 	}
-	extent_root->current_insert.offset = 0;
+	extent_root->fs_info->current_insert.offset = 0;
 	return 0;
 }
 
@@ -156,7 +161,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
-	struct btrfs_root *extent_root = root->extent_root;
+	struct btrfs_root *extent_root = root->fs_info->extent_root;
 	int ret;
 	struct btrfs_extent_item *ei;
 	struct btrfs_key ins;
@@ -186,14 +191,16 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (pin) {
 			int err;
 			radix_tree_preload(GFP_KERNEL);
-			err = radix_tree_insert(&extent_root->pinned_radix,
-					  blocknr, (void *)blocknr);
+			err = radix_tree_insert(
+					&extent_root->fs_info->pinned_radix,
+					blocknr, (void *)blocknr);
 			BUG_ON(err);
 			radix_tree_preload_end();
 		}
 		ret = btrfs_del_item(trans, extent_root, &path);
-		if (!pin && extent_root->last_insert.objectid > blocknr)
-			extent_root->last_insert.objectid = blocknr;
+		if (!pin && extent_root->fs_info->last_insert.objectid >
+		    blocknr)
+			extent_root->fs_info->last_insert.objectid = blocknr;
 		if (ret)
 			BUG();
 	}
@@ -214,18 +221,19 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 	int i;
 
 	while(1) {
-		ret = radix_tree_gang_lookup_tag(&extent_root->cache_radix,
-						 (void **)gang, 0,
-						 ARRAY_SIZE(gang),
-						 CTREE_EXTENT_PENDING_DEL);
+		ret = radix_tree_gang_lookup_tag(
+					&extent_root->fs_info->cache_radix,
+					(void **)gang, 0,
+					ARRAY_SIZE(gang),
+					CTREE_EXTENT_PENDING_DEL);
 		if (!ret)
 			break;
 		for (i = 0; i < ret; i++) {
 			ret = __free_extent(trans, extent_root,
 					    gang[i]->blocknr, 1, 1);
-			radix_tree_tag_clear(&extent_root->cache_radix,
-						gang[i]->blocknr,
-						CTREE_EXTENT_PENDING_DEL);
+			radix_tree_tag_clear(&extent_root->fs_info->cache_radix,
+					     gang[i]->blocknr,
+					     CTREE_EXTENT_PENDING_DEL);
 			btrfs_block_release(extent_root, gang[i]);
 		}
 	}
@@ -235,8 +243,8 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
 		       *extent_root)
 {
-	while(radix_tree_tagged(&extent_root->cache_radix,
-			        CTREE_EXTENT_PENDING_DEL))
+	while(radix_tree_tagged(&extent_root->fs_info->cache_radix,
+				CTREE_EXTENT_PENDING_DEL))
 		del_pending_extents(trans, extent_root);
 	return 0;
 }
@@ -248,19 +256,19 @@ static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, u64 blocknr, u64 num_blocks, int pin)
 {
-	struct btrfs_root *extent_root = root->extent_root;
+	struct btrfs_root *extent_root = root->fs_info->extent_root;
 	struct btrfs_buffer *t;
 	int pending_ret;
 	int ret;
 
 	if (root == extent_root) {
 		t = find_tree_block(root, blocknr);
-		radix_tree_tag_set(&root->cache_radix, blocknr,
+		radix_tree_tag_set(&root->fs_info->cache_radix, blocknr,
 				   CTREE_EXTENT_PENDING_DEL);
 		return 0;
 	}
 	ret = __free_extent(trans, root, blocknr, num_blocks, pin);
-	pending_ret = run_pending(trans, root->extent_root);
+	pending_ret = run_pending(trans, root->fs_info->extent_root);
 	return ret ? ret : pending_ret;
 }
 
@@ -285,12 +293,12 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	u64 test_block;
 	int start_found;
 	struct btrfs_leaf *l;
-	struct btrfs_root * root = orig_root->extent_root;
+	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	int total_needed = num_blocks;
 
 	total_needed += (btrfs_header_level(&root->node->node.header) + 1) * 3;
-	if (root->last_insert.objectid > search_start)
-		search_start = root->last_insert.objectid;
+	if (root->fs_info->last_insert.objectid > search_start)
+		search_start = root->fs_info->last_insert.objectid;
 
 	ins->flags = 0;
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
@@ -353,16 +361,17 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	BUG_ON(ins->objectid < search_start);
 	for (test_block = ins->objectid;
 	     test_block < ins->objectid + total_needed; test_block++) {
-		if (radix_tree_lookup(&root->pinned_radix, test_block)) {
+		if (radix_tree_lookup(&root->fs_info->pinned_radix,
+				      test_block)) {
 			search_start = test_block + 1;
 			goto check_failed;
 		}
 	}
-	BUG_ON(root->current_insert.offset);
-	root->current_insert.offset = total_needed - num_blocks;
-	root->current_insert.objectid = ins->objectid + num_blocks;
-	root->current_insert.flags = 0;
-	root->last_insert.objectid = ins->objectid;
+	BUG_ON(root->fs_info->current_insert.offset);
+	root->fs_info->current_insert.offset = total_needed - num_blocks;
+	root->fs_info->current_insert.objectid = ins->objectid + num_blocks;
+	root->fs_info->current_insert.flags = 0;
+	root->fs_info->last_insert.objectid = ins->objectid;
 	ins->offset = num_blocks;
 	return 0;
 error:
@@ -383,20 +392,20 @@ static int alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 {
 	int ret;
 	int pending_ret;
-	struct btrfs_root *extent_root = root->extent_root;
+	struct btrfs_root *extent_root = root->fs_info->extent_root;
 	struct btrfs_extent_item extent_item;
 
 	btrfs_set_extent_refs(&extent_item, 1);
 	btrfs_set_extent_owner(&extent_item, owner);
 
 	if (root == extent_root) {
-		BUG_ON(extent_root->current_insert.offset == 0);
+		BUG_ON(extent_root->fs_info->current_insert.offset == 0);
 		BUG_ON(num_blocks != 1);
-		BUG_ON(extent_root->current_insert.flags ==
-		       extent_root->current_insert.offset);
+		BUG_ON(extent_root->fs_info->current_insert.flags ==
+		       extent_root->fs_info->current_insert.offset);
 		ins->offset = 1;
-		ins->objectid = extent_root->current_insert.objectid +
-				extent_root->current_insert.flags++;
+		ins->objectid = extent_root->fs_info->current_insert.objectid +
+				extent_root->fs_info->current_insert.flags++;
 		return 0;
 	}
 	ret = find_free_extent(trans, root, num_blocks, search_start,

fs/btrfs/file-item.c

@@ -4,4 +4,10 @@
 #include "radix-tree.h"
 #include "ctree.h"
 #include "disk-io.h"
+#include "transaction.h"
 
+int btrfs_create_file(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root, u64 dirid, u64 *objectid)
+{
+	return 0;
+}

fs/btrfs/inode-map.c

+#include <stdio.h>
+#include <stdlib.h>
+#include "kerncompat.h"
+#include "radix-tree.h"
+#include "ctree.h"
+#include "disk-io.h"
+#include "transaction.h"
+
+/*
+ * walks the btree of allocated inodes and find a hole.
+ */
+int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *fs_root,
+			     u64 dirid, u64 *objectid)
+{
+	struct btrfs_path path;
+	struct btrfs_key key;
+	int ret;
+	u64 hole_size = 0;
+	int slot = 0;
+	u64 last_ino;
+	int start_found;
+	struct btrfs_leaf *l;
+	struct btrfs_root *root = fs_root->fs_info->inode_root;
+	struct btrfs_key search_key;
+	u64 search_start = dirid;
+
+	if (fs_root->fs_info->last_inode_alloc_dirid == dirid)
+		search_start = fs_root->fs_info->last_inode_alloc;
+
+	search_key.objectid = search_start;
+	search_key.flags = 0;
+	btrfs_set_key_type(&search_key, BTRFS_INODE_MAP_ITEM_KEY);
+	search_key.offset = 0;
+
+	btrfs_init_path(&path);
+	start_found = 0;
+	ret = btrfs_search_slot(trans, root, &search_key, &path, 0, 0);
+	if (ret < 0)
+		goto error;
+
+	if (path.slots[0] > 0)
+		path.slots[0]--;
+
+	while (1) {
+		l = &path.nodes[0]->leaf;
+		slot = path.slots[0];
+		if (slot >= btrfs_header_nritems(&l->header)) {
+			ret = btrfs_next_leaf(root, &path);
+			if (ret == 0)
+				continue;
+			if (ret < 0)
+				goto error;
+			if (!start_found) {
+				*objectid = search_start;
+				start_found = 1;
+				goto found;
+			}
+			*objectid = last_ino > search_start ?
+				last_ino : search_start;
+			goto found;
+		}
+		btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
+		if (key.objectid >= search_start) {
+			if (start_found) {
+				if (last_ino < search_start)
+					last_ino = search_start;
+				hole_size = key.objectid - last_ino;
+				if (hole_size > 0) {
+					*objectid = last_ino;
+					goto found;
+				}
+			}
+		}
+		start_found = 1;
+		last_ino = key.objectid + 1;
+		path.slots[0]++;
+	}
+	// FIXME -ENOSPC
+found:
+	root->fs_info->last_inode_alloc = *objectid;
+	root->fs_info->last_inode_alloc_dirid = dirid;
+	btrfs_release_path(root, &path);
+	BUG_ON(*objectid < search_start);
+	return 0;
+error:
+	btrfs_release_path(root, &path);
+	return ret;
+}
+
+int btrfs_insert_inode_map(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *fs_root,
+			   u64 objectid, struct btrfs_key *location)
+{
+	int ret = 0;
+	struct btrfs_path path;
+	struct btrfs_inode_map_item *inode_item;
+	struct btrfs_key key;
+	struct btrfs_root *inode_root = fs_root->fs_info->inode_root;
+
+	key.objectid = objectid;
+	key.flags = 0;
+	btrfs_set_key_type(&key, BTRFS_INODE_MAP_ITEM_KEY);
+	key.offset = 0;
+	btrfs_init_path(&path);
+	ret = btrfs_insert_empty_item(trans, inode_root, &path, &key,
+				      sizeof(struct btrfs_inode_map_item));
+	if (ret)
+		goto out;
+
+	inode_item = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
+				  struct btrfs_inode_map_item);
+	btrfs_cpu_key_to_disk(&inode_item->key, location);
+out:
+	btrfs_release_path(inode_root, &path);
+	return ret;
+}
+
+int btrfs_lookup_inode_map(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *fs_root, struct btrfs_path *path,
+			   u64 objectid, int mod)
+{
+	int ret;
+	struct btrfs_key key;
+	int ins_len = mod < 0 ? -1 : 0;
+	int cow = mod != 0;
+	struct btrfs_root *inode_root = fs_root->fs_info->inode_root;
+
+	key.objectid = objectid;
+	key.flags = 0;
+	key.offset = 0;
+	btrfs_set_key_type(&key, BTRFS_INODE_MAP_ITEM_KEY);
+	ret = btrfs_search_slot(trans, inode_root, &key, path, ins_len, cow);
+	return ret;
+}
+

fs/btrfs/mkfs.c

@@ -42,7 +42,7 @@ int mkfs(int fd, u64 num_blocks, u32 blocksize)
 	btrfs_set_header_parentid(&empty_leaf->header,
 				  BTRFS_ROOT_TREE_OBJECTID);
 	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 1);
-	btrfs_set_header_nritems(&empty_leaf->header, 2);
+	btrfs_set_header_nritems(&empty_leaf->header, 3);
 
 	/* create the items for the root tree */
 	btrfs_set_root_blocknr(&root_item, start_block + 2);
@@ -61,8 +61,16 @@ int mkfs(int fd, u64 num_blocks, u32 blocksize)
 	btrfs_set_root_blocknr(&root_item, start_block + 3);
 	itemoff = itemoff - sizeof(root_item);
 	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_disk_key_objectid(&item.key, BTRFS_FS_TREE_OBJECTID);
+	btrfs_set_disk_key_objectid(&item.key, BTRFS_INODE_MAP_OBJECTID);
 	memcpy(empty_leaf->items + 1, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
+		&root_item, sizeof(root_item));
+
+	btrfs_set_root_blocknr(&root_item, start_block + 4);
+	itemoff = itemoff - sizeof(root_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_disk_key_objectid(&item.key, BTRFS_FS_TREE_OBJECTID);
+	memcpy(empty_leaf->items + 2, &item, sizeof(item));
 	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
 		&root_item, sizeof(root_item));
 	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 1) * blocksize);
@@ -71,7 +79,7 @@ int mkfs(int fd, u64 num_blocks, u32 blocksize)
 	btrfs_set_header_parentid(&empty_leaf->header,
 				  BTRFS_EXTENT_TREE_OBJECTID);
 	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 2);
-	btrfs_set_header_nritems(&empty_leaf->header, 4);
+	btrfs_set_header_nritems(&empty_leaf->header, 5);
 
 	/* item1, reserve blocks 0-16 */
 	btrfs_set_disk_key_objectid(&item.key, 0);
@@ -108,12 +116,12 @@ int mkfs(int fd, u64 num_blocks, u32 blocksize)
 	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
 		&extent_item, btrfs_item_size(&item));
 
-	/* item4, give block 19 to the FS root */
+	/* item4, give block 19 to the inode map */
 	btrfs_set_disk_key_objectid(&item.key, start_block + 3);
 	btrfs_set_disk_key_offset(&item.key, 1);
 	itemoff = itemoff - sizeof(struct btrfs_extent_item);
 	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_extent_owner(&extent_item, BTRFS_FS_TREE_OBJECTID);
+	btrfs_set_extent_owner(&extent_item, BTRFS_INODE_MAP_OBJECTID);
 	memcpy(empty_leaf->items + 3, &item, sizeof(item));
 	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
 		&extent_item, btrfs_item_size(&item));
@@ -121,11 +129,33 @@ int mkfs(int fd, u64 num_blocks, u32 blocksize)
 	if (ret != blocksize)
 		return -1;
 
-	/* finally create the FS root */
-	btrfs_set_header_parentid(&empty_leaf->header, BTRFS_FS_TREE_OBJECTID);
+	/* item5, give block 20 to the FS root */
+	btrfs_set_disk_key_objectid(&item.key, start_block + 4);
+	btrfs_set_disk_key_offset(&item.key, 1);
+	itemoff = itemoff - sizeof(struct btrfs_extent_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_extent_owner(&extent_item, BTRFS_FS_TREE_OBJECTID);
+	memcpy(empty_leaf->items + 4, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
+		&extent_item, btrfs_item_size(&item));
+	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 2) * blocksize);
+	if (ret != blocksize)
+		return -1;
+
+	/* create the inode map */
+	btrfs_set_header_parentid(&empty_leaf->header,
+				  BTRFS_INODE_MAP_OBJECTID);
 	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 3);
 	btrfs_set_header_nritems(&empty_leaf->header, 0);
 	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 3) * blocksize);
+	if (ret != blocksize)
+		return -1;
+
+	/* finally create the FS root */
+	btrfs_set_header_parentid(&empty_leaf->header, BTRFS_FS_TREE_OBJECTID);
+	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 4);
+	btrfs_set_header_nritems(&empty_leaf->header, 0);
+	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 4) * blocksize);
 	if (ret != blocksize)
 		return -1;
 	return 0;

fs/btrfs/print-tree.c

@@ -13,8 +13,8 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 	struct btrfs_extent_item *ei;
 	struct btrfs_root_item *ri;
 	struct btrfs_dir_item *di;
+	struct btrfs_inode_map_item *mi;
 	u32 type;
-	u32 namelen;
 
 	printf("leaf %Lu total ptrs %d free space %d\n",
 		btrfs_header_blocknr(&l->header), nr,
@@ -34,15 +34,13 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 		case BTRFS_INODE_ITEM_KEY:
 			break;
 		case BTRFS_DIR_ITEM_KEY:
-			namelen = btrfs_item_size(l->items + i) - sizeof(*di);
 			di = btrfs_item_ptr(l, i, struct btrfs_dir_item);
 			printf("\t\tdir oid %Lu flags %u type %u\n",
 				btrfs_dir_objectid(di),
 				btrfs_dir_flags(di),
 				btrfs_dir_type(di));
 			printf("\t\tname %.*s\n",
-				namelen, (char *)(di + 1));
-
+			       btrfs_dir_name_len(di),(char *)(di + 1));
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
@@ -54,6 +52,13 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 			printf("\t\textent data refs %u owner %Lu\n",
 				btrfs_extent_refs(ei), btrfs_extent_owner(ei));
 			break;
+		case BTRFS_INODE_MAP_ITEM_KEY:
+			mi = btrfs_item_ptr(l, i, struct btrfs_inode_map_item);
+			printf("\t\tinode map key %Lu %u %Lu\n",
+			       btrfs_disk_key_objectid(&mi->key),
+			       btrfs_disk_key_flags(&mi->key),
+			       btrfs_disk_key_offset(&mi->key));
+			break;
 		case BTRFS_STRING_ITEM_KEY:
 			printf("\t\titem data %.*s\n", btrfs_item_size(item),
 				btrfs_leaf_data(l) + btrfs_item_offset(item));

fs/btrfs/quick-test.c

@@ -172,7 +172,8 @@ int main(int ac, char **av) {
 	printf("tree size is now %d\n", tree_size);
 	printf("root %p commit root %p\n", root->node, root->commit_root);
 	printf("map tree\n");
-	btrfs_print_tree(root->extent_root, root->extent_root->node);
+	btrfs_print_tree(root->fs_info->extent_root,
+			 root->fs_info->extent_root->node);
 	close_ctree(root, &super);
 	return 0;
 }