[net-next-2.6.git] / fs / ext3 / dir.c

/*
 *  linux/fs/ext3/dir.c
 *
 * Copyright (C) 1992, 1993, 1994, 1995
 * Remy Card (card@masi.ibp.fr)
 * Laboratoire MASI - Institut Blaise Pascal
 * Universite Pierre et Marie Curie (Paris VI)
 *
 *  from
 *
 *  linux/fs/minix/dir.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  ext3 directory handling functions
 *
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller (davem@caip.rutgers.edu), 1995
 *
 * Hash Tree Directory indexing (c) 2001  Daniel Phillips
 *
 */

#include <linux/fs.h>
#include <linux/jbd.h>
#include <linux/ext3_fs.h>
#include <linux/buffer_head.h>
#include <linux/slab.h>
#include <linux/rbtree.h>

static unsigned char ext3_filetype_table[] = {
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

static int ext3_readdir(struct file *, void *, filldir_t);
static int ext3_dx_readdir(struct file * filp,
			   void * dirent, filldir_t filldir);
static int ext3_release_dir (struct inode * inode,
				struct file * filp);

const struct file_operations ext3_dir_operations = {
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
	.readdir	= ext3_readdir,		/* we take BKL. needed?*/
	.unlocked_ioctl	= ext3_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl	= ext3_compat_ioctl,
#endif
	.fsync		= ext3_sync_file,	/* BKL held */
	.release	= ext3_release_dir,
};


static unsigned char get_dtype(struct super_block *sb, int filetype)
{
	if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) ||
	    (filetype >= EXT3_FT_MAX))
		return DT_UNKNOWN;

	return (ext3_filetype_table[filetype]);
}


int ext3_check_dir_entry (const char * function, struct inode * dir,
			  struct ext3_dir_entry_2 * de,
			  struct buffer_head * bh,
			  unsigned long offset)
{
	const char * error_msg = NULL;
	const int rlen = ext3_rec_len_from_disk(de->rec_len);

	if (rlen < EXT3_DIR_REC_LEN(1))
		error_msg = "rec_len is smaller than minimal";
	else if (rlen % 4 != 0)
		error_msg = "rec_len % 4 != 0";
	else if (rlen < EXT3_DIR_REC_LEN(de->name_len))
		error_msg = "rec_len is too small for name_len";
	else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)
		error_msg = "directory entry across blocks";
	else if (le32_to_cpu(de->inode) >
			le32_to_cpu(EXT3_SB(dir->i_sb)->s_es->s_inodes_count))
		error_msg = "inode out of bounds";

	if (error_msg != NULL)
		ext3_error (dir->i_sb, function,
			"bad entry in directory #%lu: %s - "
			"offset=%lu, inode=%lu, rec_len=%d, name_len=%d",
			dir->i_ino, error_msg, offset,
			(unsigned long) le32_to_cpu(de->inode),
			rlen, de->name_len);
	return error_msg == NULL ? 1 : 0;
}

static int ext3_readdir(struct file * filp,
			 void * dirent, filldir_t filldir)
{
	int error = 0;
	unsigned long offset;
	int i, stored;
	struct ext3_dir_entry_2 *de;
	struct super_block *sb;
	int err;
	struct inode *inode = filp->f_path.dentry->d_inode;
	int ret = 0;
	int dir_has_error = 0;

	sb = inode->i_sb;

	if (EXT3_HAS_COMPAT_FEATURE(inode->i_sb,
				    EXT3_FEATURE_COMPAT_DIR_INDEX) &&
	    ((EXT3_I(inode)->i_flags & EXT3_INDEX_FL) ||
	     ((inode->i_size >> sb->s_blocksize_bits) == 1))) {
		err = ext3_dx_readdir(filp, dirent, filldir);
		if (err != ERR_BAD_DX_DIR) {
			ret = err;
			goto out;
		}
		/*
		 * We don't set the inode dirty flag since it's not
		 * critical that it get flushed back to the disk.
		 */
		EXT3_I(filp->f_path.dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL;
	}
	stored = 0;
	offset = filp->f_pos & (sb->s_blocksize - 1);

	while (!error && !stored && filp->f_pos < inode->i_size) {
		unsigned long blk = filp->f_pos >> EXT3_BLOCK_SIZE_BITS(sb);
		struct buffer_head map_bh;
		struct buffer_head *bh = NULL;

		map_bh.b_state = 0;
		err = ext3_get_blocks_handle(NULL, inode, blk, 1, &map_bh, 0);
		if (err > 0) {
			pgoff_t index = map_bh.b_blocknr >>
					(PAGE_CACHE_SHIFT - inode->i_blkbits);
			if (!ra_has_index(&filp->f_ra, index))
				page_cache_sync_readahead(
					sb->s_bdev->bd_inode->i_mapping,
					&filp->f_ra, filp,
					index, 1);
			filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
			bh = ext3_bread(NULL, inode, blk, 0, &err);
		}

		/*
		 * We ignore I/O errors on directories so users have a chance
		 * of recovering data when there's a bad sector
		 */
		if (!bh) {
			if (!dir_has_error) {
				ext3_error(sb, __func__, "directory #%lu "
					"contains a hole at offset %lld",
					inode->i_ino, filp->f_pos);
				dir_has_error = 1;
			}
			/* corrupt size?  Maybe no more blocks to read */
			if (filp->f_pos > inode->i_blocks << 9)
				break;
			filp->f_pos += sb->s_blocksize - offset;
			continue;
		}

revalidate:
		/* If the dir block has changed since the last call to
		 * readdir(2), then we might be pointing to an invalid
		 * dirent right now.  Scan from the start of the block
		 * to make sure. */
		if (filp->f_version != inode->i_version) {
			for (i = 0; i < sb->s_blocksize && i < offset; ) {
				de = (struct ext3_dir_entry_2 *)
					(bh->b_data + i);
				/* It's too expensive to do a full
				 * dirent test each time round this
				 * loop, but we do have to test at
				 * least that it is non-zero.  A
				 * failure will be detected in the
				 * dirent test below. */
				if (ext3_rec_len_from_disk(de->rec_len) <
						EXT3_DIR_REC_LEN(1))
					break;
				i += ext3_rec_len_from_disk(de->rec_len);
			}
			offset = i;
			filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
				| offset;
			filp->f_version = inode->i_version;
		}

		while (!error && filp->f_pos < inode->i_size
		       && offset < sb->s_blocksize) {
			de = (struct ext3_dir_entry_2 *) (bh->b_data + offset);
			if (!ext3_check_dir_entry ("ext3_readdir", inode, de,
						   bh, offset)) {
				/* On error, skip the f_pos to the
                                   next block. */
				filp->f_pos = (filp->f_pos |
						(sb->s_blocksize - 1)) + 1;
				brelse (bh);
				ret = stored;
				goto out;
			}
			offset += ext3_rec_len_from_disk(de->rec_len);
			if (le32_to_cpu(de->inode)) {
				/* We might block in the next section
				 * if the data destination is
				 * currently swapped out.  So, use a
				 * version stamp to detect whether or
				 * not the directory has been modified
				 * during the copy operation.
				 */
				u64 version = filp->f_version;

				error = filldir(dirent, de->name,
						de->name_len,
						filp->f_pos,
						le32_to_cpu(de->inode),
						get_dtype(sb, de->file_type));
				if (error)
					break;
				if (version != filp->f_version)
					goto revalidate;
				stored ++;
			}
			filp->f_pos += ext3_rec_len_from_disk(de->rec_len);
		}
		offset = 0;
		brelse (bh);
	}
out:
	return ret;
}

/*
 * These functions convert from the major/minor hash to an f_pos
 * value.
 *
 * Currently we only use major hash numer.  This is unfortunate, but
 * on 32-bit machines, the same VFS interface is used for lseek and
 * llseek, so if we use the 64 bit offset, then the 32-bit versions of
 * lseek/telldir/seekdir will blow out spectacularly, and from within
 * the ext2 low-level routine, we don't know if we're being called by
 * a 64-bit version of the system call or the 32-bit version of the
 * system call.  Worse yet, NFSv2 only allows for a 32-bit readdir
 * cookie.  Sigh.
 */
#define hash2pos(major, minor)	(major >> 1)
#define pos2maj_hash(pos)	((pos << 1) & 0xffffffff)
#define pos2min_hash(pos)	(0)

/*
 * This structure holds the nodes of the red-black tree used to store
 * the directory entry in hash order.
 */
struct fname {
	__u32		hash;
	__u32		minor_hash;
	struct rb_node	rb_hash;
	struct fname	*next;
	__u32		inode;
	__u8		name_len;
	__u8		file_type;
	char		name[0];
};

/*
 * This functoin implements a non-recursive way of freeing all of the
 * nodes in the red-black tree.
 */
static void free_rb_tree_fname(struct rb_root *root)
{
	struct rb_node	*n = root->rb_node;
	struct rb_node	*parent;
	struct fname	*fname;

	while (n) {
		/* Do the node's children first */
		if (n->rb_left) {
			n = n->rb_left;
			continue;
		}
		if (n->rb_right) {
			n = n->rb_right;
			continue;
		}
		/*
		 * The node has no children; free it, and then zero
		 * out parent's link to it.  Finally go to the
		 * beginning of the loop and try to free the parent
		 * node.
		 */
		parent = rb_parent(n);
		fname = rb_entry(n, struct fname, rb_hash);
		while (fname) {
			struct fname * old = fname;
			fname = fname->next;
			kfree (old);
		}
		if (!parent)
			*root = RB_ROOT;
		else if (parent->rb_left == n)
			parent->rb_left = NULL;
		else if (parent->rb_right == n)
			parent->rb_right = NULL;
		n = parent;
	}
}


static struct dir_private_info *ext3_htree_create_dir_info(loff_t pos)
{
	struct dir_private_info *p;

	p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
	if (!p)
		return NULL;
	p->curr_hash = pos2maj_hash(pos);
	p->curr_minor_hash = pos2min_hash(pos);
	return p;
}

void ext3_htree_free_dir_info(struct dir_private_info *p)
{
	free_rb_tree_fname(&p->root);
	kfree(p);
}

/*
 * Given a directory entry, enter it into the fname rb tree.
 */
int ext3_htree_store_dirent(struct file *dir_file, __u32 hash,
			     __u32 minor_hash,
			     struct ext3_dir_entry_2 *dirent)
{
	struct rb_node **p, *parent = NULL;
	struct fname * fname, *new_fn;
	struct dir_private_info *info;
	int len;

	info = (struct dir_private_info *) dir_file->private_data;
	p = &info->root.rb_node;

	/* Create and allocate the fname structure */
	len = sizeof(struct fname) + dirent->name_len + 1;
	new_fn = kzalloc(len, GFP_KERNEL);
	if (!new_fn)
		return -ENOMEM;
	new_fn->hash = hash;
	new_fn->minor_hash = minor_hash;
	new_fn->inode = le32_to_cpu(dirent->inode);
	new_fn->name_len = dirent->name_len;
	new_fn->file_type = dirent->file_type;
	memcpy(new_fn->name, dirent->name, dirent->name_len);
	new_fn->name[dirent->name_len] = 0;

	while (*p) {
		parent = *p;
		fname = rb_entry(parent, struct fname, rb_hash);

		/*
		 * If the hash and minor hash match up, then we put
		 * them on a linked list.  This rarely happens...
		 */
		if ((new_fn->hash == fname->hash) &&
		    (new_fn->minor_hash == fname->minor_hash)) {
			new_fn->next = fname->next;
			fname->next = new_fn;
			return 0;
		}

		if (new_fn->hash < fname->hash)
			p = &(*p)->rb_left;
		else if (new_fn->hash > fname->hash)
			p = &(*p)->rb_right;
		else if (new_fn->minor_hash < fname->minor_hash)
			p = &(*p)->rb_left;
		else /* if (new_fn->minor_hash > fname->minor_hash) */
			p = &(*p)->rb_right;
	}

	rb_link_node(&new_fn->rb_hash, parent, p);
	rb_insert_color(&new_fn->rb_hash, &info->root);
	return 0;
}


/*
 * This is a helper function for ext3_dx_readdir.  It calls filldir
 * for all entres on the fname linked list.  (Normally there is only
 * one entry on the linked list, unless there are 62 bit hash collisions.)
 */
static int call_filldir(struct file * filp, void * dirent,
			filldir_t filldir, struct fname *fname)
{
	struct dir_private_info *info = filp->private_data;
	loff_t	curr_pos;
	struct inode *inode = filp->f_path.dentry->d_inode;
	struct super_block * sb;
	int error;

	sb = inode->i_sb;

	if (!fname) {
		printk("call_filldir: called with null fname?!?\n");
		return 0;
	}
	curr_pos = hash2pos(fname->hash, fname->minor_hash);
	while (fname) {
		error = filldir(dirent, fname->name,
				fname->name_len, curr_pos,
				fname->inode,
				get_dtype(sb, fname->file_type));
		if (error) {
			filp->f_pos = curr_pos;
			info->extra_fname = fname;
			return error;
		}
		fname = fname->next;
	}
	return 0;
}

static int ext3_dx_readdir(struct file * filp,
			 void * dirent, filldir_t filldir)
{
	struct dir_private_info *info = filp->private_data;
	struct inode *inode = filp->f_path.dentry->d_inode;
	struct fname *fname;
	int	ret;

	if (!info) {
		info = ext3_htree_create_dir_info(filp->f_pos);
		if (!info)
			return -ENOMEM;
		filp->private_data = info;
	}

	if (filp->f_pos == EXT3_HTREE_EOF)
		return 0;	/* EOF */

	/* Some one has messed with f_pos; reset the world */
	if (info->last_pos != filp->f_pos) {
		free_rb_tree_fname(&info->root);
		info->curr_node = NULL;
		info->extra_fname = NULL;
		info->curr_hash = pos2maj_hash(filp->f_pos);
		info->curr_minor_hash = pos2min_hash(filp->f_pos);
	}

	/*
	 * If there are any leftover names on the hash collision
	 * chain, return them first.
	 */
	if (info->extra_fname) {
		if (call_filldir(filp, dirent, filldir, info->extra_fname))
			goto finished;
		info->extra_fname = NULL;
		goto next_node;
	} else if (!info->curr_node)
		info->curr_node = rb_first(&info->root);

	while (1) {
		/*
		 * Fill the rbtree if we have no more entries,
		 * or the inode has changed since we last read in the
		 * cached entries.
		 */
		if ((!info->curr_node) ||
		    (filp->f_version != inode->i_version)) {
			info->curr_node = NULL;
			free_rb_tree_fname(&info->root);
			filp->f_version = inode->i_version;
			ret = ext3_htree_fill_tree(filp, info->curr_hash,
						   info->curr_minor_hash,
						   &info->next_hash);
			if (ret < 0)
				return ret;
			if (ret == 0) {
				filp->f_pos = EXT3_HTREE_EOF;
				break;
			}
			info->curr_node = rb_first(&info->root);
		}

		fname = rb_entry(info->curr_node, struct fname, rb_hash);
		info->curr_hash = fname->hash;
		info->curr_minor_hash = fname->minor_hash;
		if (call_filldir(filp, dirent, filldir, fname))
			break;
	next_node:
		info->curr_node = rb_next(info->curr_node);
		if (info->curr_node) {
			fname = rb_entry(info->curr_node, struct fname,
					 rb_hash);
			info->curr_hash = fname->hash;
			info->curr_minor_hash = fname->minor_hash;
		} else {
			if (info->next_hash == ~0) {
				filp->f_pos = EXT3_HTREE_EOF;
				break;
			}
			info->curr_hash = info->next_hash;
			info->curr_minor_hash = 0;
		}
	}
finished:
	info->last_pos = filp->f_pos;
	return 0;
}

static int ext3_release_dir (struct inode * inode, struct file * filp)
{
       if (filp->private_data)
		ext3_htree_free_dir_info(filp->private_data);

	return 0;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/fs/ext3/dir.c
	3	*
	4	* Copyright (C) 1992, 1993, 1994, 1995
	5	* Remy Card (card@masi.ibp.fr)
	6	* Laboratoire MASI - Institut Blaise Pascal
	7	* Universite Pierre et Marie Curie (Paris VI)
	8	*
	9	* from
	10	*
	11	* linux/fs/minix/dir.c
	12	*
	13	* Copyright (C) 1991, 1992 Linus Torvalds
	14	*
	15	* ext3 directory handling functions
	16	*
	17	* Big-endian to little-endian byte-swapping/bitmaps by
	18	* David S. Miller (davem@caip.rutgers.edu), 1995
	19	*
	20	* Hash Tree Directory indexing (c) 2001 Daniel Phillips
	21	*
	22	*/
	23
	24	#include <linux/fs.h>
	25	#include <linux/jbd.h>
	26	#include <linux/ext3_fs.h>
	27	#include <linux/buffer_head.h>
1da177e4 LT	28	#include <linux/slab.h>
	29	#include <linux/rbtree.h>
	30
	31	static unsigned char ext3_filetype_table[] = {
	32	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
	33	};
	34
	35	static int ext3_readdir(struct file , void , filldir_t);
	36	static int ext3_dx_readdir(struct file * filp,
	37	void * dirent, filldir_t filldir);
	38	static int ext3_release_dir (struct inode * inode,
	39	struct file * filp);
	40
4b6f5d20	41	const struct file_operations ext3_dir_operations = {
1da177e4 LT	42	.llseek = generic_file_llseek,
	43	.read = generic_read_dir,
	44	.readdir = ext3_readdir, /* we take BKL. needed?*/
039fd8ce	45	.unlocked_ioctl = ext3_ioctl,
52a700c5 DH	46	#ifdef CONFIG_COMPAT
	47	.compat_ioctl = ext3_compat_ioctl,
	48	#endif
1da177e4	49	.fsync = ext3_sync_file, /* BKL held */
1da177e4	50	.release = ext3_release_dir,
1da177e4 LT	51	};
	52
	53
	54	static unsigned char get_dtype(struct super_block *sb, int filetype)
	55	{
	56	if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) \|\|
	57	(filetype >= EXT3_FT_MAX))
	58	return DT_UNKNOWN;
	59
	60	return (ext3_filetype_table[filetype]);
	61	}
ae6ddcc5	62
1da177e4 LT	63
	64	int ext3_check_dir_entry (const char * function, struct inode * dir,
	65	struct ext3_dir_entry_2 * de,
	66	struct buffer_head * bh,
	67	unsigned long offset)
	68	{
	69	const char * error_msg = NULL;
7c06a8dc	70	const int rlen = ext3_rec_len_from_disk(de->rec_len);
1da177e4 LT	71
	72	if (rlen < EXT3_DIR_REC_LEN(1))
	73	error_msg = "rec_len is smaller than minimal";
	74	else if (rlen % 4 != 0)
	75	error_msg = "rec_len % 4 != 0";
	76	else if (rlen < EXT3_DIR_REC_LEN(de->name_len))
	77	error_msg = "rec_len is too small for name_len";
	78	else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)
	79	error_msg = "directory entry across blocks";
	80	else if (le32_to_cpu(de->inode) >
	81	le32_to_cpu(EXT3_SB(dir->i_sb)->s_es->s_inodes_count))
	82	error_msg = "inode out of bounds";
	83
	84	if (error_msg != NULL)
	85	ext3_error (dir->i_sb, function,
	86	"bad entry in directory #%lu: %s - "
	87	"offset=%lu, inode=%lu, rec_len=%d, name_len=%d",
	88	dir->i_ino, error_msg, offset,
	89	(unsigned long) le32_to_cpu(de->inode),
	90	rlen, de->name_len);
	91	return error_msg == NULL ? 1 : 0;
	92	}
	93
	94	static int ext3_readdir(struct file * filp,
	95	void * dirent, filldir_t filldir)
	96	{
	97	int error = 0;
d8733c29 AM	98	unsigned long offset;
	99	int i, stored;
	100	struct ext3_dir_entry_2 *de;
	101	struct super_block *sb;
1da177e4	102	int err;
fe21a693	103	struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4	104	int ret = 0;
cdbf6dba	105	int dir_has_error = 0;
1da177e4 LT	106
	107	sb = inode->i_sb;
	108
1da177e4 LT	109	if (EXT3_HAS_COMPAT_FEATURE(inode->i_sb,
	110	EXT3_FEATURE_COMPAT_DIR_INDEX) &&
	111	((EXT3_I(inode)->i_flags & EXT3_INDEX_FL) \|\|
	112	((inode->i_size >> sb->s_blocksize_bits) == 1))) {
	113	err = ext3_dx_readdir(filp, dirent, filldir);
	114	if (err != ERR_BAD_DX_DIR) {
	115	ret = err;
	116	goto out;
	117	}
	118	/*
	119	* We don't set the inode dirty flag since it's not
	120	* critical that it get flushed back to the disk.
	121	*/
fe21a693	122	EXT3_I(filp->f_path.dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL;
1da177e4	123	}
1da177e4	124	stored = 0;
1da177e4 LT	125	offset = filp->f_pos & (sb->s_blocksize - 1);
	126
	127	while (!error && !stored && filp->f_pos < inode->i_size) {
d8733c29 AM	128	unsigned long blk = filp->f_pos >> EXT3_BLOCK_SIZE_BITS(sb);
	129	struct buffer_head map_bh;
	130	struct buffer_head *bh = NULL;
	131
	132	map_bh.b_state = 0;
43237b54	133	err = ext3_get_blocks_handle(NULL, inode, blk, 1, &map_bh, 0);
89747d36	134	if (err > 0) {
dc7868fc FW	135	pgoff_t index = map_bh.b_blocknr >>
	136	(PAGE_CACHE_SHIFT - inode->i_blkbits);
	137	if (!ra_has_index(&filp->f_ra, index))
cf914a7d	138	page_cache_sync_readahead(
dc7868fc FW	139	sb->s_bdev->bd_inode->i_mapping,
dc7868fc FW	140	&filp->f_ra, filp,
cf914a7d	141	index, 1);
f4e6b498	142	filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
d8733c29 AM	143	bh = ext3_bread(NULL, inode, blk, 0, &err);
	144	}
	145
	146	/*
	147	* We ignore I/O errors on directories so users have a chance
	148	* of recovering data when there's a bad sector
	149	*/
1da177e4	150	if (!bh) {
cdbf6dba ES	151	if (!dir_has_error) {
	152	ext3_error(sb, __func__, "directory #%lu "
	153	"contains a hole at offset %lld",
	154	inode->i_ino, filp->f_pos);
	155	dir_has_error = 1;
	156	}
40b85134 ES	157	/* corrupt size? Maybe no more blocks to read */
	158	if (filp->f_pos > inode->i_blocks << 9)
	159	break;
1da177e4 LT	160	filp->f_pos += sb->s_blocksize - offset;
	161	continue;
	162	}
	163
1da177e4 LT	164	revalidate:
	165	/* If the dir block has changed since the last call to
	166	* readdir(2), then we might be pointing to an invalid
	167	* dirent right now. Scan from the start of the block
	168	* to make sure. */
	169	if (filp->f_version != inode->i_version) {
	170	for (i = 0; i < sb->s_blocksize && i < offset; ) {
ae6ddcc5	171	de = (struct ext3_dir_entry_2 *)
1da177e4 LT	172	(bh->b_data + i);
	173	/* It's too expensive to do a full
	174	* dirent test each time round this
	175	* loop, but we do have to test at
	176	* least that it is non-zero. A
	177	* failure will be detected in the
	178	* dirent test below. */
7c06a8dc	179	if (ext3_rec_len_from_disk(de->rec_len) <
1da177e4 LT	180	EXT3_DIR_REC_LEN(1))
1da177e4 LT	181	break;
7c06a8dc	182	i += ext3_rec_len_from_disk(de->rec_len);
1da177e4 LT	183	}
	184	offset = i;
	185	filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
	186	\| offset;
	187	filp->f_version = inode->i_version;
	188	}
	189
ae6ddcc5	190	while (!error && filp->f_pos < inode->i_size
1da177e4 LT	191	&& offset < sb->s_blocksize) {
	192	de = (struct ext3_dir_entry_2 *) (bh->b_data + offset);
	193	if (!ext3_check_dir_entry ("ext3_readdir", inode, de,
	194	bh, offset)) {
	195	/* On error, skip the f_pos to the
	196	next block. */
	197	filp->f_pos = (filp->f_pos \|
	198	(sb->s_blocksize - 1)) + 1;
	199	brelse (bh);
	200	ret = stored;
	201	goto out;
	202	}
7c06a8dc	203	offset += ext3_rec_len_from_disk(de->rec_len);
1da177e4 LT	204	if (le32_to_cpu(de->inode)) {
	205	/* We might block in the next section
	206	* if the data destination is
	207	* currently swapped out. So, use a
	208	* version stamp to detect whether or
	209	* not the directory has been modified
	210	* during the copy operation.
	211	*/
2b47c361	212	u64 version = filp->f_version;
1da177e4 LT	213
	214	error = filldir(dirent, de->name,
	215	de->name_len,
	216	filp->f_pos,
	217	le32_to_cpu(de->inode),
	218	get_dtype(sb, de->file_type));
	219	if (error)
	220	break;
	221	if (version != filp->f_version)
	222	goto revalidate;
	223	stored ++;
	224	}
7c06a8dc	225	filp->f_pos += ext3_rec_len_from_disk(de->rec_len);
1da177e4 LT	226	}
	227	offset = 0;
	228	brelse (bh);
	229	}
	230	out:
	231	return ret;
	232	}
	233
1da177e4 LT	234	/*
	235	* These functions convert from the major/minor hash to an f_pos
	236	* value.
ae6ddcc5	237	*
1da177e4 LT	238	* Currently we only use major hash numer. This is unfortunate, but
	239	* on 32-bit machines, the same VFS interface is used for lseek and
	240	* llseek, so if we use the 64 bit offset, then the 32-bit versions of
	241	* lseek/telldir/seekdir will blow out spectacularly, and from within
	242	* the ext2 low-level routine, we don't know if we're being called by
	243	* a 64-bit version of the system call or the 32-bit version of the
	244	* system call. Worse yet, NFSv2 only allows for a 32-bit readdir
	245	* cookie. Sigh.
	246	*/
	247	#define hash2pos(major, minor) (major >> 1)
	248	#define pos2maj_hash(pos) ((pos << 1) & 0xffffffff)
	249	#define pos2min_hash(pos) (0)
	250
	251	/*
	252	* This structure holds the nodes of the red-black tree used to store
	253	* the directory entry in hash order.
	254	*/
	255	struct fname {
	256	__u32 hash;
	257	__u32 minor_hash;
ae6ddcc5	258	struct rb_node rb_hash;
1da177e4 LT	259	struct fname *next;
	260	__u32 inode;
	261	__u8 name_len;
	262	__u8 file_type;
	263	char name[0];
	264	};
	265
	266	/*
	267	* This functoin implements a non-recursive way of freeing all of the
	268	* nodes in the red-black tree.
	269	*/
	270	static void free_rb_tree_fname(struct rb_root *root)
	271	{
	272	struct rb_node *n = root->rb_node;
	273	struct rb_node *parent;
	274	struct fname *fname;
	275
	276	while (n) {
	277	/* Do the node's children first */
9ebfbe9f	278	if (n->rb_left) {
1da177e4 LT	279	n = n->rb_left;
	280	continue;
	281	}
	282	if (n->rb_right) {
	283	n = n->rb_right;
	284	continue;
	285	}
	286	/*
	287	* The node has no children; free it, and then zero
	288	* out parent's link to it. Finally go to the
	289	* beginning of the loop and try to free the parent
	290	* node.
	291	*/
52b5108c	292	parent = rb_parent(n);
1da177e4 LT	293	fname = rb_entry(n, struct fname, rb_hash);
	294	while (fname) {
	295	struct fname * old = fname;
	296	fname = fname->next;
	297	kfree (old);
	298	}
	299	if (!parent)
1513b02c	300	*root = RB_ROOT;
1da177e4 LT	301	else if (parent->rb_left == n)
	302	parent->rb_left = NULL;
	303	else if (parent->rb_right == n)
	304	parent->rb_right = NULL;
	305	n = parent;
	306	}
1da177e4 LT	307	}
	308
	309
9ebfbe9f	310	static struct dir_private_info *ext3_htree_create_dir_info(loff_t pos)
1da177e4 LT	311	{
	312	struct dir_private_info *p;
	313
9ebfbe9f	314	p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
1da177e4 LT	315	if (!p)
1da177e4 LT	316	return NULL;
1da177e4 LT	317	p->curr_hash = pos2maj_hash(pos);
1da177e4 LT	318	p->curr_minor_hash = pos2min_hash(pos);
1da177e4 LT	319	return p;
	320	}
	321
	322	void ext3_htree_free_dir_info(struct dir_private_info *p)
	323	{
	324	free_rb_tree_fname(&p->root);
	325	kfree(p);
	326	}
	327
	328	/*
	329	* Given a directory entry, enter it into the fname rb tree.
	330	*/
	331	int ext3_htree_store_dirent(struct file *dir_file, __u32 hash,
	332	__u32 minor_hash,
	333	struct ext3_dir_entry_2 *dirent)
	334	{
	335	struct rb_node *p, parent = NULL;
	336	struct fname * fname, *new_fn;
	337	struct dir_private_info *info;
	338	int len;
	339
	340	info = (struct dir_private_info *) dir_file->private_data;
	341	p = &info->root.rb_node;
	342
	343	/* Create and allocate the fname structure */
	344	len = sizeof(struct fname) + dirent->name_len + 1;
f8314dc6	345	new_fn = kzalloc(len, GFP_KERNEL);
1da177e4 LT	346	if (!new_fn)
1da177e4 LT	347	return -ENOMEM;
1da177e4 LT	348	new_fn->hash = hash;
	349	new_fn->minor_hash = minor_hash;
	350	new_fn->inode = le32_to_cpu(dirent->inode);
	351	new_fn->name_len = dirent->name_len;
	352	new_fn->file_type = dirent->file_type;
	353	memcpy(new_fn->name, dirent->name, dirent->name_len);
	354	new_fn->name[dirent->name_len] = 0;
	355
	356	while (*p) {
	357	parent = *p;
	358	fname = rb_entry(parent, struct fname, rb_hash);
	359
	360	/*
	361	* If the hash and minor hash match up, then we put
	362	* them on a linked list. This rarely happens...
	363	*/
	364	if ((new_fn->hash == fname->hash) &&
	365	(new_fn->minor_hash == fname->minor_hash)) {
	366	new_fn->next = fname->next;
	367	fname->next = new_fn;
	368	return 0;
	369	}
	370
	371	if (new_fn->hash < fname->hash)
	372	p = &(*p)->rb_left;
	373	else if (new_fn->hash > fname->hash)
	374	p = &(*p)->rb_right;
	375	else if (new_fn->minor_hash < fname->minor_hash)
	376	p = &(*p)->rb_left;
	377	else /* if (new_fn->minor_hash > fname->minor_hash) */
	378	p = &(*p)->rb_right;
	379	}
	380
	381	rb_link_node(&new_fn->rb_hash, parent, p);
	382	rb_insert_color(&new_fn->rb_hash, &info->root);
	383	return 0;
	384	}
	385
	386
	387
	388	/*
	389	* This is a helper function for ext3_dx_readdir. It calls filldir
	390	* for all entres on the fname linked list. (Normally there is only
	391	* one entry on the linked list, unless there are 62 bit hash collisions.)
	392	*/
	393	static int call_filldir(struct file * filp, void * dirent,
	394	filldir_t filldir, struct fname *fname)
	395	{
	396	struct dir_private_info *info = filp->private_data;
	397	loff_t curr_pos;
fe21a693	398	struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4 LT	399	struct super_block * sb;
	400	int error;
	401
	402	sb = inode->i_sb;
	403
	404	if (!fname) {
	405	printk("call_filldir: called with null fname?!?\n");
	406	return 0;
	407	}
	408	curr_pos = hash2pos(fname->hash, fname->minor_hash);
	409	while (fname) {
	410	error = filldir(dirent, fname->name,
ae6ddcc5	411	fname->name_len, curr_pos,
1da177e4 LT	412	fname->inode,
	413	get_dtype(sb, fname->file_type));
	414	if (error) {
	415	filp->f_pos = curr_pos;
6a897cf4	416	info->extra_fname = fname;
1da177e4 LT	417	return error;
	418	}
	419	fname = fname->next;
	420	}
	421	return 0;
	422	}
	423
	424	static int ext3_dx_readdir(struct file * filp,
	425	void * dirent, filldir_t filldir)
	426	{
	427	struct dir_private_info *info = filp->private_data;
fe21a693	428	struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4 LT	429	struct fname *fname;
	430	int ret;
	431
	432	if (!info) {
9ebfbe9f	433	info = ext3_htree_create_dir_info(filp->f_pos);
1da177e4 LT	434	if (!info)
	435	return -ENOMEM;
	436	filp->private_data = info;
	437	}
	438
	439	if (filp->f_pos == EXT3_HTREE_EOF)
	440	return 0; /* EOF */
	441
	442	/* Some one has messed with f_pos; reset the world */
	443	if (info->last_pos != filp->f_pos) {
	444	free_rb_tree_fname(&info->root);
	445	info->curr_node = NULL;
	446	info->extra_fname = NULL;
	447	info->curr_hash = pos2maj_hash(filp->f_pos);
	448	info->curr_minor_hash = pos2min_hash(filp->f_pos);
	449	}
	450
	451	/*
	452	* If there are any leftover names on the hash collision
	453	* chain, return them first.
	454	*/
6a897cf4 ED	455	if (info->extra_fname) {
	456	if (call_filldir(filp, dirent, filldir, info->extra_fname))
	457	goto finished;
6a897cf4	458	info->extra_fname = NULL;
8c9fa93d	459	goto next_node;
6a897cf4	460	} else if (!info->curr_node)
1da177e4 LT	461	info->curr_node = rb_first(&info->root);
	462
	463	while (1) {
	464	/*
	465	* Fill the rbtree if we have no more entries,
	466	* or the inode has changed since we last read in the
ae6ddcc5	467	* cached entries.
1da177e4 LT	468	*/
	469	if ((!info->curr_node) \|\|
	470	(filp->f_version != inode->i_version)) {
	471	info->curr_node = NULL;
	472	free_rb_tree_fname(&info->root);
	473	filp->f_version = inode->i_version;
	474	ret = ext3_htree_fill_tree(filp, info->curr_hash,
	475	info->curr_minor_hash,
	476	&info->next_hash);
	477	if (ret < 0)
	478	return ret;
	479	if (ret == 0) {
	480	filp->f_pos = EXT3_HTREE_EOF;
	481	break;
	482	}
	483	info->curr_node = rb_first(&info->root);
	484	}
	485
	486	fname = rb_entry(info->curr_node, struct fname, rb_hash);
	487	info->curr_hash = fname->hash;
	488	info->curr_minor_hash = fname->minor_hash;
	489	if (call_filldir(filp, dirent, filldir, fname))
	490	break;
8c9fa93d	491	next_node:
1da177e4	492	info->curr_node = rb_next(info->curr_node);
8c9fa93d TT	493	if (info->curr_node) {
	494	fname = rb_entry(info->curr_node, struct fname,
	495	rb_hash);
	496	info->curr_hash = fname->hash;
	497	info->curr_minor_hash = fname->minor_hash;
	498	} else {
1da177e4 LT	499	if (info->next_hash == ~0) {
	500	filp->f_pos = EXT3_HTREE_EOF;
	501	break;
	502	}
	503	info->curr_hash = info->next_hash;
	504	info->curr_minor_hash = 0;
	505	}
	506	}
	507	finished:
	508	info->last_pos = filp->f_pos;
	509	return 0;
	510	}
	511
	512	static int ext3_release_dir (struct inode * inode, struct file * filp)
	513	{
	514	if (filp->private_data)
	515	ext3_htree_free_dir_info(filp->private_data);
	516
	517	return 0;
	518	}