[linux-2.6-block.git] / fs / btrfs / super.c

#include <linux/module.h>
#include <linux/buffer_head.h>
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/smp_lock.h>
#include <linux/backing-dev.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"

#define BTRFS_SUPER_MAGIC 0x9123682E

static struct inode_operations btrfs_dir_inode_operations;
static struct super_operations btrfs_super_ops;
static struct file_operations btrfs_dir_file_operations;

static void btrfs_read_locked_inode(struct inode *inode)
{
	struct btrfs_path path;
	struct btrfs_inode_item *inode_item;
	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	int ret;
	btrfs_init_path(&path);
	ret = btrfs_lookup_inode(NULL, root, &path, inode->i_ino, 0);
	if (ret) {
		make_bad_inode(inode);
		btrfs_release_path(root, &path);
		return;
	}
	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);
	btrfs_release_path(root, &path);
	switch (inode->i_mode & S_IFMT) {
#if 0
	default:
		init_special_inode(inode, inode->i_mode,
				   btrfs_inode_rdev(inode_item));
		break;
#endif
	case S_IFREG:
		break;
	case S_IFDIR:
		inode->i_op = &btrfs_dir_inode_operations;
		inode->i_fop = &btrfs_dir_file_operations;
		break;
	case S_IFLNK:
		// inode->i_op = &page_symlink_inode_operations;
		break;
	}
	return;
}

static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
{
	struct btrfs_path path;
	struct btrfs_root *root;
	struct btrfs_trans_handle *trans;
	const char *name = dentry->d_name.name;
	int name_len = dentry->d_name.len;
	int ret;
	u64 objectid;
	struct btrfs_dir_item *di;

	btrfs_init_path(&path);
	root = btrfs_sb(dir->i_sb);
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);

	ret = btrfs_lookup_dir_item(trans, root, &path, dir->i_ino,
				    name, name_len, -1);
	if (ret < 0)
		goto err;
	if (ret > 0) {
		ret = -ENOENT;
		goto err;
	}
	di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
			    struct btrfs_dir_item);
	objectid = btrfs_dir_objectid(di);

	ret = btrfs_del_item(trans, root, &path);
	BUG_ON(ret);
	dentry->d_inode->i_ctime = dir->i_ctime;
err:
	btrfs_release_path(root, &path);
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);
	if (ret == 0)
		inode_dec_link_count(dentry->d_inode);
	return ret;
}

static int btrfs_free_inode(struct btrfs_trans_handle *trans,
			    struct btrfs_root *root,
			    struct inode *inode)
{
	u64 objectid = inode->i_ino;
	struct btrfs_path path;
	struct btrfs_inode_map_item *map;
	struct btrfs_key stat_data_key;
	int ret;
	clear_inode(inode);
	btrfs_init_path(&path);
	ret = btrfs_lookup_inode_map(trans, root, &path, objectid, -1);
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		btrfs_release_path(root, &path);
		goto error;
	}
	map = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
			    struct btrfs_inode_map_item);
	btrfs_disk_key_to_cpu(&stat_data_key, &map->key);
	ret = btrfs_del_item(trans, root->fs_info->inode_root, &path);
	BUG_ON(ret);
	btrfs_release_path(root, &path);
	btrfs_init_path(&path);

	ret = btrfs_lookup_inode(trans, root, &path, objectid, -1);
	BUG_ON(ret);
	ret = btrfs_del_item(trans, root, &path);
	BUG_ON(ret);
	btrfs_release_path(root, &path);
error:
	return ret;
}

static void btrfs_delete_inode(struct inode *inode)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	truncate_inode_pages(&inode->i_data, 0);
	if (is_bad_inode(inode)) {
		goto no_delete;
	}
	inode->i_size = 0;
	if (inode->i_blocks)
		WARN_ON(1);

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	btrfs_free_inode(trans, root, inode);
	btrfs_end_transaction(trans, root);
	mutex_unlock(&root->fs_info->fs_mutex);
	return;
no_delete:
	clear_inode(inode);
}


static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
			      ino_t *ino)
{
	const char *name = dentry->d_name.name;
	int namelen = dentry->d_name.len;
	struct btrfs_dir_item *di;
	struct btrfs_path path;
	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	int ret;

	btrfs_init_path(&path);
	ret = btrfs_lookup_dir_item(NULL, root, &path, dir->i_ino, name,
				    namelen, 0);
	if (ret || !btrfs_match_dir_item_name(root, &path, name, namelen)) {
		*ino = 0;
		goto out;
	}
	di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
			    struct btrfs_dir_item);
	*ino = btrfs_dir_objectid(di);
out:
	btrfs_release_path(root, &path);
	return ret;
}

static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
				   struct nameidata *nd)
{
	struct inode * inode;
	ino_t ino;
	int ret;

	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);

	ret = btrfs_inode_by_name(dir, dentry, &ino);
	if (ret < 0)
		return ERR_PTR(ret);
	inode = NULL;
	if (ino) {
		inode = iget(dir->i_sb, ino);
		if (!inode)
			return ERR_PTR(-EACCES);
	}
	return d_splice_alias(inode, dentry);
}

static void reada_leaves(struct btrfs_root *root, struct btrfs_path *path)
{
	struct buffer_head *bh;
	struct btrfs_node *node;
	int i;
	int nritems;
	u64 objectid;
	u64 item_objectid;
	u64 blocknr;
	int slot;

	if (!path->nodes[1])
		return;
	node = btrfs_buffer_node(path->nodes[1]);
	slot = path->slots[1];
	objectid = btrfs_disk_key_objectid(&node->ptrs[slot].key);
	nritems = btrfs_header_nritems(&node->header);
	for (i = slot; i < nritems; i++) {
		item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
		if (item_objectid != objectid)
			break;
		blocknr = btrfs_node_blockptr(node, i);
		bh = sb_getblk(root->fs_info->sb, blocknr);
		ll_rw_block(READ, 1, &bh);
		brelse(bh);
	}

}
static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
	struct inode *inode = filp->f_path.dentry->d_inode;
	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
	struct btrfs_path path;
	int ret;
	u32 nritems;
	struct btrfs_leaf *leaf;
	int slot;
	int advance;
	unsigned char d_type = DT_UNKNOWN;
	int over = 0;

	key.objectid = inode->i_ino;
	key.flags = 0;
	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
	key.offset = filp->f_pos;
	btrfs_init_path(&path);
	ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
	if (ret < 0) {
		goto err;
	}
	advance = 0;
	reada_leaves(root, &path);
	while(1) {
		leaf = btrfs_buffer_leaf(path.nodes[0]);
		nritems = btrfs_header_nritems(&leaf->header);
		slot = path.slots[0];
		if (advance) {
			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);
				slot = path.slots[0];
				if (path.nodes[1] && path.slots[1] == 0)
					reada_leaves(root, &path);
			} else {
				slot++;
				path.slots[0]++;
			}
		}
		advance = 1;
		item = leaf->items + slot;
		if (btrfs_disk_key_objectid(&item->key) != key.objectid)
			break;
		if (btrfs_disk_key_type(&item->key) != BTRFS_DIR_ITEM_KEY)
			continue;
		if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
			continue;
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		over = filldir(dirent, (const char *)(di + 1),
			       btrfs_dir_name_len(di),
			       btrfs_disk_key_offset(&item->key),
			       btrfs_dir_objectid(di), d_type);
		if (over) {
			filp->f_pos = btrfs_disk_key_offset(&item->key);
			break;
		}
		filp->f_pos = btrfs_disk_key_offset(&item->key) + 1;
	}
	ret = 0;
err:
	btrfs_release_path(root, &path);
	return ret;
}

static void btrfs_put_super (struct super_block * sb)
{
	struct btrfs_root *root = btrfs_sb(sb);
	int ret;

	ret = close_ctree(root);
	if (ret) {
		printk("close ctree returns %d\n", ret);
	}
	sb->s_fs_info = NULL;
}

static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
{
	struct inode * inode;
	struct dentry * root_dentry;
	struct btrfs_super_block *disk_super;
	struct buffer_head *bh;
	struct btrfs_root *root;

	sb->s_maxbytes = MAX_LFS_FILESIZE;
	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = BTRFS_SUPER_MAGIC;
	sb->s_op = &btrfs_super_ops;
	sb->s_time_gran = 1;

	bh = sb_bread(sb, BTRFS_SUPER_INFO_OFFSET / sb->s_blocksize);
	if (!bh) {
		printk("btrfs: unable to read on disk super\n");
		return -EIO;
	}
	disk_super = (struct btrfs_super_block *)bh->b_data;
	root = open_ctree(sb, bh, disk_super);
	sb->s_fs_info = root;
	if (!root) {
		printk("btrfs: open_ctree failed\n");
		return -EIO;
	}
	printk("read in super total blocks %Lu root %Lu\n",
	       btrfs_super_total_blocks(disk_super),
	       btrfs_super_root_dir(disk_super));

	inode = iget_locked(sb, btrfs_super_root_dir(disk_super));
	if (!inode)
		return -ENOMEM;
	if (inode->i_state & I_NEW) {
		btrfs_read_locked_inode(inode);
		unlock_new_inode(inode);
	}

	root_dentry = d_alloc_root(inode);
	if (!root_dentry) {
		iput(inode);
		return -ENOMEM;
	}
	sb->s_root = root_dentry;

	return 0;
}

static void fill_inode_item(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);
}

static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct inode *dir, int mode)
{
	struct inode *inode;
	struct btrfs_inode_item inode_item;
	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	struct btrfs_key key;
	int ret;
	u64 objectid;

	inode = new_inode(dir->i_sb);
	if (!inode)
		return ERR_PTR(-ENOMEM);

	ret = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
	BUG_ON(ret);

	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_SEC;
	/* FIXME do this on link */
	if (mode & S_IFDIR)
		inode->i_size = 3;
	fill_inode_item(&inode_item, inode);

	key.objectid = objectid;
	key.flags = 0;
	key.offset = 0;
	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
	ret = btrfs_insert_inode_map(trans, root, objectid, &key);
	BUG_ON(ret);

	ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
	BUG_ON(ret);

	insert_inode_hash(inode);
	// FIXME mark_inode_dirty(inode)
	return inode;
}

static int btrfs_add_link(struct btrfs_trans_handle *trans,
			    struct dentry *dentry, struct inode *inode)
{
	int ret;
	ret = btrfs_insert_dir_item(trans, btrfs_sb(inode->i_sb),
				    dentry->d_name.name, dentry->d_name.len,
				    dentry->d_parent->d_inode->i_ino,
				    inode->i_ino, 0);
	return ret;
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
			    struct dentry *dentry, struct inode *inode)
{
	int err = btrfs_add_link(trans, dentry, inode);
	if (!err) {
		d_instantiate(dentry, inode);
		return 0;
	}
	return err;
}

static int btrfs_create(struct inode *dir, struct dentry *dentry,
			int mode, struct nameidata *nd)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	struct inode *inode;
	int err;
	int drop_inode = 0;

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	inode = btrfs_new_inode(trans, dir, mode);
	err = PTR_ERR(inode);
	if (IS_ERR(inode))
		goto out_unlock;
	// FIXME mark the inode dirty
	err = btrfs_add_nondir(trans, dentry, inode);
	if (err)
		drop_inode = 1;
	dir->i_sb->s_dirt = 1;
	btrfs_end_transaction(trans, root);
out_unlock:
	mutex_unlock(&root->fs_info->fs_mutex);
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	return err;
}

static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
				struct inode *inode, struct inode *dir)
{
	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	int ret;
	char buf[2];
	buf[0] = '.';
	buf[1] = '.';

	ret = btrfs_insert_dir_item(trans, root, buf, 1, inode->i_ino,
				    inode->i_ino, 1);
	if (ret)
		goto error;
	ret = btrfs_insert_dir_item(trans, root, buf, 2, inode->i_ino,
				    dir->i_ino, 1);
error:
	return ret;
}

static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
	struct inode *inode;
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	int err = 0;
	int drop_on_err = 0;

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out_unlock;
	}
	inode = btrfs_new_inode(trans, dir, S_IFDIR | mode);
	if (IS_ERR(inode)) {
		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;

	err = btrfs_make_empty_dir(trans, inode, dir);
	if (err)
		goto out_fail;
	err = btrfs_add_link(trans, dentry, inode);
	if (err)
		goto out_fail;
	d_instantiate(dentry, inode);
	mark_inode_dirty(inode);
	drop_on_err = 0;

out_fail:
	btrfs_end_transaction(trans, root);
out_unlock:
	mutex_unlock(&root->fs_info->fs_mutex);
	if (drop_on_err)
		iput(inode);
	return err;
}

static int btrfs_sync_fs(struct super_block *sb, int wait)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root;
	int ret;

	sb->s_dirt = 0;
	if (!wait) {
		filemap_flush(sb->s_bdev->bd_inode->i_mapping);
		return 0;
	}
	filemap_write_and_wait(sb->s_bdev->bd_inode->i_mapping);

	root = btrfs_sb(sb);
	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	ret = btrfs_commit_transaction(trans, root);
	sb->s_dirt = 0;
	BUG_ON(ret);
printk("btrfs sync_fs\n");
	mutex_unlock(&root->fs_info->fs_mutex);
	return 0;
}

static void btrfs_write_super(struct super_block *sb)
{
	btrfs_sync_fs(sb, 1);
}


static int btrfs_get_sb(struct file_system_type *fs_type,
	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
{
	return get_sb_bdev(fs_type, flags, dev_name, data,
			   btrfs_fill_super, mnt);
}

static struct file_system_type btrfs_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "btrfs",
	.get_sb		= btrfs_get_sb,
	.kill_sb	= kill_block_super,
	.fs_flags	= FS_REQUIRES_DEV,
};

static struct super_operations btrfs_super_ops = {
	.statfs		= simple_statfs,
	.delete_inode	= btrfs_delete_inode,
	.put_super	= btrfs_put_super,
	.read_inode	= btrfs_read_locked_inode,
	.write_super	= btrfs_write_super,
	.sync_fs	= btrfs_sync_fs,
};

static struct inode_operations btrfs_dir_inode_operations = {
	.lookup		= btrfs_lookup,
	.create		= btrfs_create,
	.unlink		= btrfs_unlink,
	.mkdir		= btrfs_mkdir,
};

static struct file_operations btrfs_dir_file_operations = {
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
	.readdir	= btrfs_readdir,
};


static int __init init_btrfs_fs(void)
{
	printk("btrfs loaded!\n");
	return register_filesystem(&btrfs_fs_type);
}

static void __exit exit_btrfs_fs(void)
{
	unregister_filesystem(&btrfs_fs_type);
	printk("btrfs unloaded\n");
}

module_init(init_btrfs_fs)
module_exit(exit_btrfs_fs)

MODULE_LICENSE("GPL");
Commit	Line	Data
2e635a27	1	#include <linux/module.h>
e20d96d6	2	#include <linux/buffer_head.h>
2e635a27 CM	3	#include <linux/fs.h>
	4	#include <linux/pagemap.h>
	5	#include <linux/highmem.h>
	6	#include <linux/time.h>
	7	#include <linux/init.h>
	8	#include <linux/string.h>
	9	#include <linux/smp_lock.h>
	10	#include <linux/backing-dev.h>
	11	#include "ctree.h"
e20d96d6	12	#include "disk-io.h"
d5719762	13	#include "transaction.h"
2e635a27 CM	14
2e635a27 CM	15	#define BTRFS_SUPER_MAGIC 0x9123682E
e20d96d6 CM	16
	17	static struct inode_operations btrfs_dir_inode_operations;
	18	static struct super_operations btrfs_super_ops;
	19	static struct file_operations btrfs_dir_file_operations;
	20
e20d96d6	21	static void btrfs_read_locked_inode(struct inode *inode)
2e635a27	22	{
e20d96d6 CM	23	struct btrfs_path path;
	24	struct btrfs_inode_item *inode_item;
	25	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	26	int ret;
e20d96d6 CM	27	btrfs_init_path(&path);
	28	ret = btrfs_lookup_inode(NULL, root, &path, inode->i_ino, 0);
	29	if (ret) {
	30	make_bad_inode(inode);
78fae27e	31	btrfs_release_path(root, &path);
e20d96d6 CM	32	return;
	33	}
	34	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
	35	path.slots[0],
	36	struct btrfs_inode_item);
2e635a27	37
e20d96d6 CM	38	inode->i_mode = btrfs_inode_mode(inode_item);
	39	inode->i_nlink = btrfs_inode_nlink(inode_item);
	40	inode->i_uid = btrfs_inode_uid(inode_item);
	41	inode->i_gid = btrfs_inode_gid(inode_item);
	42	inode->i_size = btrfs_inode_size(inode_item);
	43	inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
	44	inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
	45	inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
	46	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
	47	inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
	48	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
	49	inode->i_blocks = btrfs_inode_nblocks(inode_item);
	50	inode->i_generation = btrfs_inode_generation(inode_item);
e20d96d6 CM	51	btrfs_release_path(root, &path);
	52	switch (inode->i_mode & S_IFMT) {
	53	#if 0
	54	default:
	55	init_special_inode(inode, inode->i_mode,
	56	btrfs_inode_rdev(inode_item));
	57	break;
	58	#endif
	59	case S_IFREG:
e20d96d6 CM	60	break;
e20d96d6 CM	61	case S_IFDIR:
e20d96d6 CM	62	inode->i_op = &btrfs_dir_inode_operations;
	63	inode->i_fop = &btrfs_dir_file_operations;
	64	break;
	65	case S_IFLNK:
e20d96d6 CM	66	// inode->i_op = &page_symlink_inode_operations;
e20d96d6 CM	67	break;
2e635a27	68	}
e20d96d6	69	return;
2e635a27 CM	70	}
2e635a27 CM	71
134e9731 CM	72	static int btrfs_unlink(struct inode dir, struct dentry dentry)
	73	{
	74	struct btrfs_path path;
	75	struct btrfs_root *root;
	76	struct btrfs_trans_handle *trans;
	77	const char *name = dentry->d_name.name;
	78	int name_len = dentry->d_name.len;
	79	int ret;
	80	u64 objectid;
	81	struct btrfs_dir_item *di;
	82
	83	btrfs_init_path(&path);
	84	root = btrfs_sb(dir->i_sb);
	85	mutex_lock(&root->fs_info->fs_mutex);
	86	trans = btrfs_start_transaction(root, 1);
	87
	88	ret = btrfs_lookup_dir_item(trans, root, &path, dir->i_ino,
	89	name, name_len, -1);
	90	if (ret < 0)
	91	goto err;
	92	if (ret > 0) {
	93	ret = -ENOENT;
	94	goto err;
	95	}
	96	di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
	97	struct btrfs_dir_item);
	98	objectid = btrfs_dir_objectid(di);
	99
	100	ret = btrfs_del_item(trans, root, &path);
	101	BUG_ON(ret);
	102	dentry->d_inode->i_ctime = dir->i_ctime;
	103	err:
	104	btrfs_release_path(root, &path);
	105	btrfs_end_transaction(trans, root);
	106	mutex_unlock(&root->fs_info->fs_mutex);
	107	if (ret == 0)
	108	inode_dec_link_count(dentry->d_inode);
	109	return ret;
	110	}
	111
	112	static int btrfs_free_inode(struct btrfs_trans_handle *trans,
	113	struct btrfs_root *root,
	114	struct inode *inode)
	115	{
	116	u64 objectid = inode->i_ino;
	117	struct btrfs_path path;
	118	struct btrfs_inode_map_item *map;
	119	struct btrfs_key stat_data_key;
	120	int ret;
	121	clear_inode(inode);
	122	btrfs_init_path(&path);
	123	ret = btrfs_lookup_inode_map(trans, root, &path, objectid, -1);
	124	if (ret) {
	125	if (ret > 0)
	126	ret = -ENOENT;
	127	btrfs_release_path(root, &path);
	128	goto error;
	129	}
	130	map = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
	131	struct btrfs_inode_map_item);
	132	btrfs_disk_key_to_cpu(&stat_data_key, &map->key);
	133	ret = btrfs_del_item(trans, root->fs_info->inode_root, &path);
	134	BUG_ON(ret);
	135	btrfs_release_path(root, &path);
136	btrfs_init_path(&path);
137
138	ret = btrfs_lookup_inode(trans, root, &path, objectid, -1);
139	BUG_ON(ret);
140	ret = btrfs_del_item(trans, root, &path);
141	BUG_ON(ret);
142	btrfs_release_path(root, &path);
143	error:
144	return ret;
145	}
146
147	static void btrfs_delete_inode(struct inode *inode)
148	{
149	struct btrfs_trans_handle *trans;
150	struct btrfs_root *root = btrfs_sb(inode->i_sb);
151	truncate_inode_pages(&inode->i_data, 0);
152	if (is_bad_inode(inode)) {
153	goto no_delete;
154	}
155	inode->i_size = 0;
156	if (inode->i_blocks)
157	WARN_ON(1);
158
159	mutex_lock(&root->fs_info->fs_mutex);
160	trans = btrfs_start_transaction(root, 1);
161	btrfs_free_inode(trans, root, inode);
162	btrfs_end_transaction(trans, root);
163	mutex_unlock(&root->fs_info->fs_mutex);
164	return;
165	no_delete:
166	clear_inode(inode);
167	}
168
169
e20d96d6 CM	170	static int btrfs_inode_by_name(struct inode dir, struct dentry dentry,
	171	ino_t *ino)
	172	{
	173	const char *name = dentry->d_name.name;
	174	int namelen = dentry->d_name.len;
	175	struct btrfs_dir_item *di;
	176	struct btrfs_path path;
	177	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	178	int ret;
	179
	180	btrfs_init_path(&path);
	181	ret = btrfs_lookup_dir_item(NULL, root, &path, dir->i_ino, name,
	182	namelen, 0);
7f5c1516	183	if (ret \|\| !btrfs_match_dir_item_name(root, &path, name, namelen)) {
e20d96d6 CM	184	*ino = 0;
	185	goto out;
	186	}
	187	di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
	188	struct btrfs_dir_item);
	189	*ino = btrfs_dir_objectid(di);
	190	out:
	191	btrfs_release_path(root, &path);
	192	return ret;
	193	}
	194
	195	static struct dentry btrfs_lookup(struct inode dir, struct dentry *dentry,
	196	struct nameidata *nd)
	197	{
	198	struct inode * inode;
	199	ino_t ino;
	200	int ret;
	201
	202	if (dentry->d_name.len > BTRFS_NAME_LEN)
	203	return ERR_PTR(-ENAMETOOLONG);
	204
	205	ret = btrfs_inode_by_name(dir, dentry, &ino);
	206	if (ret < 0)
	207	return ERR_PTR(ret);
	208	inode = NULL;
	209	if (ino) {
e20d96d6 CM	210	inode = iget(dir->i_sb, ino);
	211	if (!inode)
	212	return ERR_PTR(-EACCES);
	213	}
	214	return d_splice_alias(inode, dentry);
	215	}
	216
dcea7915 CM	217	static void reada_leaves(struct btrfs_root root, struct btrfs_path path)
	218	{
	219	struct buffer_head *bh;
	220	struct btrfs_node *node;
	221	int i;
	222	int nritems;
	223	u64 objectid;
	224	u64 item_objectid;
	225	u64 blocknr;
	226	int slot;
	227
	228	if (!path->nodes[1])
	229	return;
	230	node = btrfs_buffer_node(path->nodes[1]);
	231	slot = path->slots[1];
	232	objectid = btrfs_disk_key_objectid(&node->ptrs[slot].key);
	233	nritems = btrfs_header_nritems(&node->header);
	234	for (i = slot; i < nritems; i++) {
	235	item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
	236	if (item_objectid != objectid)
	237	break;
	238	blocknr = btrfs_node_blockptr(node, i);
	239	bh = sb_getblk(root->fs_info->sb, blocknr);
	240	ll_rw_block(READ, 1, &bh);
	241	brelse(bh);
	242	}
	243
	244	}
e20d96d6 CM	245	static int btrfs_readdir(struct file filp, void dirent, filldir_t filldir)
	246	{
	247	struct inode *inode = filp->f_path.dentry->d_inode;
	248	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	249	struct btrfs_item *item;
	250	struct btrfs_dir_item *di;
	251	struct btrfs_key key;
	252	struct btrfs_path path;
	253	int ret;
	254	u32 nritems;
	255	struct btrfs_leaf *leaf;
	256	int slot;
	257	int advance;
	258	unsigned char d_type = DT_UNKNOWN;
7f5c1516	259	int over = 0;
e20d96d6 CM	260
e20d96d6 CM	261	key.objectid = inode->i_ino;
e20d96d6 CM	262	key.flags = 0;
	263	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
	264	key.offset = filp->f_pos;
	265	btrfs_init_path(&path);
	266	ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
	267	if (ret < 0) {
	268	goto err;
	269	}
7f5c1516	270	advance = 0;
dcea7915	271	reada_leaves(root, &path);
e20d96d6 CM	272	while(1) {
	273	leaf = btrfs_buffer_leaf(path.nodes[0]);
	274	nritems = btrfs_header_nritems(&leaf->header);
	275	slot = path.slots[0];
e20d96d6	276	if (advance) {
e20d96d6 CM	277	if (slot == nritems -1) {
	278	ret = btrfs_next_leaf(root, &path);
	279	if (ret)
	280	break;
	281	leaf = btrfs_buffer_leaf(path.nodes[0]);
	282	nritems = btrfs_header_nritems(&leaf->header);
	283	slot = path.slots[0];
dcea7915 CM	284	if (path.nodes[1] && path.slots[1] == 0)
dcea7915 CM	285	reada_leaves(root, &path);
e20d96d6 CM	286	} else {
	287	slot++;
	288	path.slots[0]++;
	289	}
	290	}
	291	advance = 1;
	292	item = leaf->items + slot;
e20d96d6 CM	293	if (btrfs_disk_key_objectid(&item->key) != key.objectid)
	294	break;
	295	if (btrfs_disk_key_type(&item->key) != BTRFS_DIR_ITEM_KEY)
	296	continue;
7f5c1516 CM	297	if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
7f5c1516 CM	298	continue;
e20d96d6	299	di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
e20d96d6 CM	300	over = filldir(dirent, (const char *)(di + 1),
	301	btrfs_dir_name_len(di),
	302	btrfs_disk_key_offset(&item->key),
	303	btrfs_dir_objectid(di), d_type);
7f5c1516 CM	304	if (over) {
7f5c1516 CM	305	filp->f_pos = btrfs_disk_key_offset(&item->key);
e20d96d6	306	break;
7f5c1516	307	}
e20d96d6 CM	308	filp->f_pos = btrfs_disk_key_offset(&item->key) + 1;
e20d96d6 CM	309	}
e20d96d6 CM	310	ret = 0;
	311	err:
	312	btrfs_release_path(root, &path);
	313	return ret;
	314	}
	315
	316	static void btrfs_put_super (struct super_block * sb)
	317	{
	318	struct btrfs_root *root = btrfs_sb(sb);
	319	int ret;
	320
	321	ret = close_ctree(root);
	322	if (ret) {
	323	printk("close ctree returns %d\n", ret);
	324	}
	325	sb->s_fs_info = NULL;
	326	}
2e635a27 CM	327
	328	static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
	329	{
	330	struct inode * inode;
e20d96d6 CM	331	struct dentry * root_dentry;
	332	struct btrfs_super_block *disk_super;
	333	struct buffer_head *bh;
	334	struct btrfs_root *root;
2e635a27 CM	335
	336	sb->s_maxbytes = MAX_LFS_FILESIZE;
	337	sb->s_blocksize = PAGE_CACHE_SIZE;
	338	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	339	sb->s_magic = BTRFS_SUPER_MAGIC;
e20d96d6	340	sb->s_op = &btrfs_super_ops;
2e635a27	341	sb->s_time_gran = 1;
e20d96d6 CM	342
	343	bh = sb_bread(sb, BTRFS_SUPER_INFO_OFFSET / sb->s_blocksize);
	344	if (!bh) {
	345	printk("btrfs: unable to read on disk super\n");
	346	return -EIO;
	347	}
	348	disk_super = (struct btrfs_super_block *)bh->b_data;
	349	root = open_ctree(sb, bh, disk_super);
	350	sb->s_fs_info = root;
	351	if (!root) {
	352	printk("btrfs: open_ctree failed\n");
	353	return -EIO;
	354	}
	355	printk("read in super total blocks %Lu root %Lu\n",
	356	btrfs_super_total_blocks(disk_super),
	357	btrfs_super_root_dir(disk_super));
	358
	359	inode = iget_locked(sb, btrfs_super_root_dir(disk_super));
2e635a27 CM	360	if (!inode)
2e635a27 CM	361	return -ENOMEM;
e20d96d6 CM	362	if (inode->i_state & I_NEW) {
	363	btrfs_read_locked_inode(inode);
	364	unlock_new_inode(inode);
	365	}
2e635a27	366
e20d96d6 CM	367	root_dentry = d_alloc_root(inode);
e20d96d6 CM	368	if (!root_dentry) {
2e635a27 CM	369	iput(inode);
	370	return -ENOMEM;
	371	}
e20d96d6 CM	372	sb->s_root = root_dentry;
e20d96d6 CM	373
2e635a27 CM	374	return 0;
	375	}
	376
d5719762 CM	377	static void fill_inode_item(struct btrfs_inode_item *item,
	378	struct inode *inode)
	379	{
	380	btrfs_set_inode_uid(item, inode->i_uid);
	381	btrfs_set_inode_gid(item, inode->i_gid);
	382	btrfs_set_inode_size(item, inode->i_size);
	383	btrfs_set_inode_mode(item, inode->i_mode);
	384	btrfs_set_inode_nlink(item, inode->i_nlink);
	385	btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
	386	btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
	387	btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
	388	btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
	389	btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
	390	btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
	391	btrfs_set_inode_nblocks(item, inode->i_blocks);
	392	btrfs_set_inode_generation(item, inode->i_generation);
	393	}
	394
	395	static struct inode btrfs_new_inode(struct btrfs_trans_handle trans,
	396	struct inode *dir, int mode)
	397	{
	398	struct inode *inode;
	399	struct btrfs_inode_item inode_item;
	400	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	401	struct btrfs_key key;
	402	int ret;
	403	u64 objectid;
	404
	405	inode = new_inode(dir->i_sb);
	406	if (!inode)
	407	return ERR_PTR(-ENOMEM);
	408
	409	ret = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
	410	BUG_ON(ret);
	411
	412	inode->i_uid = current->fsuid;
	413	inode->i_gid = current->fsgid;
	414	inode->i_mode = mode;
	415	inode->i_ino = objectid;
	416	inode->i_blocks = 0;
	417	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
f7922033 CM	418	/* FIXME do this on link */
	419	if (mode & S_IFDIR)
	420	inode->i_size = 3;
d5719762 CM	421	fill_inode_item(&inode_item, inode);
d5719762 CM	422
d5719762 CM	423	key.objectid = objectid;
	424	key.flags = 0;
	425	key.offset = 0;
	426	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
	427	ret = btrfs_insert_inode_map(trans, root, objectid, &key);
	428	BUG_ON(ret);
	429
	430	ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
	431	BUG_ON(ret);
	432
	433	insert_inode_hash(inode);
	434	// FIXME mark_inode_dirty(inode)
	435	return inode;
	436	}
	437
	438	static int btrfs_add_link(struct btrfs_trans_handle *trans,
	439	struct dentry dentry, struct inode inode)
	440	{
	441	int ret;
	442	ret = btrfs_insert_dir_item(trans, btrfs_sb(inode->i_sb),
	443	dentry->d_name.name, dentry->d_name.len,
	444	dentry->d_parent->d_inode->i_ino,
	445	inode->i_ino, 0);
d5719762 CM	446	return ret;
	447	}
	448
	449	static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
	450	struct dentry dentry, struct inode inode)
	451	{
	452	int err = btrfs_add_link(trans, dentry, inode);
	453	if (!err) {
	454	d_instantiate(dentry, inode);
	455	return 0;
	456	}
d5719762 CM	457	return err;
	458	}
	459
	460	static int btrfs_create(struct inode dir, struct dentry dentry,
	461	int mode, struct nameidata *nd)
	462	{
	463	struct btrfs_trans_handle *trans;
	464	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	465	struct inode *inode;
	466	int err;
134e9731	467	int drop_inode = 0;
d5719762	468
d561c025	469	mutex_lock(&root->fs_info->fs_mutex);
d5719762 CM	470	trans = btrfs_start_transaction(root, 1);
	471	inode = btrfs_new_inode(trans, dir, mode);
	472	err = PTR_ERR(inode);
	473	if (IS_ERR(inode))
d561c025	474	goto out_unlock;
d5719762 CM	475	// FIXME mark the inode dirty
d5719762 CM	476	err = btrfs_add_nondir(trans, dentry, inode);
134e9731 CM	477	if (err)
134e9731 CM	478	drop_inode = 1;
d5719762 CM	479	dir->i_sb->s_dirt = 1;
d5719762 CM	480	btrfs_end_transaction(trans, root);
d561c025 CM	481	out_unlock:
d561c025 CM	482	mutex_unlock(&root->fs_info->fs_mutex);
134e9731 CM	483	if (drop_inode) {
	484	inode_dec_link_count(inode);
	485	iput(inode);
	486	}
d5719762 CM	487	return err;
	488	}
	489
f7922033 CM	490	static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
	491	struct inode inode, struct inode dir)
	492	{
	493	struct btrfs_root *root = btrfs_sb(inode->i_sb);
	494	int ret;
	495	char buf[2];
	496	buf[0] = '.';
	497	buf[1] = '.';
	498
	499	ret = btrfs_insert_dir_item(trans, root, buf, 1, inode->i_ino,
	500	inode->i_ino, 1);
	501	if (ret)
	502	goto error;
	503	ret = btrfs_insert_dir_item(trans, root, buf, 2, inode->i_ino,
	504	dir->i_ino, 1);
	505	error:
	506	return ret;
	507	}
	508
	509	static int btrfs_mkdir(struct inode dir, struct dentry dentry, int mode)
	510	{
	511	struct inode *inode;
	512	struct btrfs_trans_handle *trans;
	513	struct btrfs_root *root = btrfs_sb(dir->i_sb);
	514	int err = 0;
	515	int drop_on_err = 0;
	516
	517	mutex_lock(&root->fs_info->fs_mutex);
	518	trans = btrfs_start_transaction(root, 1);
	519	if (IS_ERR(trans)) {
	520	err = PTR_ERR(trans);
	521	goto out_unlock;
	522	}
	523	inode = btrfs_new_inode(trans, dir, S_IFDIR \| mode);
	524	if (IS_ERR(inode)) {
	525	err = PTR_ERR(inode);
	526	goto out_fail;
	527	}
	528	drop_on_err = 1;
	529	inode->i_op = &btrfs_dir_inode_operations;
	530	inode->i_fop = &btrfs_dir_file_operations;
	531
	532	err = btrfs_make_empty_dir(trans, inode, dir);
	533	if (err)
	534	goto out_fail;
	535	err = btrfs_add_link(trans, dentry, inode);
	536	if (err)
	537	goto out_fail;
	538	d_instantiate(dentry, inode);
	539	mark_inode_dirty(inode);
	540	drop_on_err = 0;
	541
	542	out_fail:
	543	btrfs_end_transaction(trans, root);
	544	out_unlock:
	545	mutex_unlock(&root->fs_info->fs_mutex);
	546	if (drop_on_err)
	547	iput(inode);
	548	return err;
	549	}
	550
d5719762 CM	551	static int btrfs_sync_fs(struct super_block *sb, int wait)
	552	{
	553	struct btrfs_trans_handle *trans;
	554	struct btrfs_root *root;
	555	int ret;
df2ce34c	556
d5719762	557	sb->s_dirt = 0;
d561c025 CM	558	if (!wait) {
	559	filemap_flush(sb->s_bdev->bd_inode->i_mapping);
	560	return 0;
	561	}
	562	filemap_write_and_wait(sb->s_bdev->bd_inode->i_mapping);
df2ce34c	563
d5719762	564	root = btrfs_sb(sb);
d561c025	565	mutex_lock(&root->fs_info->fs_mutex);
d5719762 CM	566	trans = btrfs_start_transaction(root, 1);
	567	ret = btrfs_commit_transaction(trans, root);
	568	sb->s_dirt = 0;
	569	BUG_ON(ret);
	570	printk("btrfs sync_fs\n");
d561c025	571	mutex_unlock(&root->fs_info->fs_mutex);
d5719762 CM	572	return 0;
	573	}
	574
d561c025 CM	575	static void btrfs_write_super(struct super_block *sb)
	576	{
	577	btrfs_sync_fs(sb, 1);
	578	}
	579
	580
2e635a27 CM	581	static int btrfs_get_sb(struct file_system_type *fs_type,
	582	int flags, const char dev_name, void data, struct vfsmount *mnt)
	583	{
	584	return get_sb_bdev(fs_type, flags, dev_name, data,
	585	btrfs_fill_super, mnt);
	586	}
	587
	588	static struct file_system_type btrfs_fs_type = {
	589	.owner = THIS_MODULE,
	590	.name = "btrfs",
	591	.get_sb = btrfs_get_sb,
	592	.kill_sb = kill_block_super,
	593	.fs_flags = FS_REQUIRES_DEV,
	594	};
	595
e20d96d6 CM	596	static struct super_operations btrfs_super_ops = {
e20d96d6 CM	597	.statfs = simple_statfs,
134e9731	598	.delete_inode = btrfs_delete_inode,
e20d96d6 CM	599	.put_super = btrfs_put_super,
e20d96d6 CM	600	.read_inode = btrfs_read_locked_inode,
d5719762 CM	601	.write_super = btrfs_write_super,
d5719762 CM	602	.sync_fs = btrfs_sync_fs,
e20d96d6 CM	603	};
	604
	605	static struct inode_operations btrfs_dir_inode_operations = {
	606	.lookup = btrfs_lookup,
d5719762	607	.create = btrfs_create,
134e9731	608	.unlink = btrfs_unlink,
f7922033	609	.mkdir = btrfs_mkdir,
e20d96d6 CM	610	};
	611
	612	static struct file_operations btrfs_dir_file_operations = {
	613	.llseek = generic_file_llseek,
	614	.read = generic_read_dir,
	615	.readdir = btrfs_readdir,
	616	};
	617
	618
2e635a27 CM	619	static int __init init_btrfs_fs(void)
	620	{
	621	printk("btrfs loaded!\n");
	622	return register_filesystem(&btrfs_fs_type);
	623	}
	624
	625	static void __exit exit_btrfs_fs(void)
	626	{
	627	unregister_filesystem(&btrfs_fs_type);
	628	printk("btrfs unloaded\n");
	629	}
	630
	631	module_init(init_btrfs_fs)
	632	module_exit(exit_btrfs_fs)
	633
	634	MODULE_LICENSE("GPL");