[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?*/
	.ioctl		= ext3_ioctl,		/* BKL held */
#ifdef CONFIG_COMPAT
	.compat_ioctl	= ext3_compat_ioctl,
#endif
	.fsync		= ext3_sync_file,	/* BKL held */
#ifdef CONFIG_EXT3_INDEX
	.release	= ext3_release_dir,
#endif
};


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 = le16_to_cpu(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;

	sb = inode->i_sb;

#ifdef CONFIG_EXT3_INDEX
	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;
	}
#endif
	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, 0);
		if (err > 0) {
			page_cache_readahead(sb->s_bdev->bd_inode->i_mapping,
				&filp->f_ra,
				filp,
				map_bh.b_blocknr >>
					(PAGE_CACHE_SHIFT - inode->i_blkbits),
				1);
			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) {
			ext3_error (sb, "ext3_readdir",
				"directory #%lu contains a hole at offset %lu",
				inode->i_ino, (unsigned long)filp->f_pos);
			/* 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 (le16_to_cpu(de->rec_len) <
						EXT3_DIR_REC_LEN(1))
					break;
				i += le16_to_cpu(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 += le16_to_cpu(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.
				 */
				unsigned long 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 += le16_to_cpu(de->rec_len);
		}
		offset = 0;
		brelse (bh);
	}
out:
	return ret;
}

#ifdef CONFIG_EXT3_INDEX
/*
 * 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_node = NULL;
		else if (parent->rb_left == n)
			parent->rb_left = NULL;
		else if (parent->rb_right == n)
			parent->rb_right = NULL;
		n = parent;
	}
	root->rb_node = NULL;
}


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

	p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL);
	if (!p)
		return NULL;
	p->root.rb_node = NULL;
	p->curr_node = NULL;
	p->extra_fname = NULL;
	p->last_pos = 0;
	p->curr_hash = pos2maj_hash(pos);
	p->curr_minor_hash = pos2min_hash(pos);
	p->next_hash = 0;
	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->next;
			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 = 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 &&
	    call_filldir(filp, dirent, filldir, info->extra_fname))
		goto finished;

	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;

		info->curr_node = rb_next(info->curr_node);
		if (!info->curr_node) {
			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;
}

#endif
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?*/
	45	.ioctl = ext3_ioctl, /* BKL held */
52a700c5 DH	46	#ifdef CONFIG_COMPAT
	47	.compat_ioctl = ext3_compat_ioctl,
	48	#endif
1da177e4 LT	49	.fsync = ext3_sync_file, /* BKL held */
	50	#ifdef CONFIG_EXT3_INDEX
	51	.release = ext3_release_dir,
	52	#endif
	53	};
	54
	55
	56	static unsigned char get_dtype(struct super_block *sb, int filetype)
	57	{
	58	if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) \|\|
	59	(filetype >= EXT3_FT_MAX))
	60	return DT_UNKNOWN;
	61
	62	return (ext3_filetype_table[filetype]);
	63	}
ae6ddcc5	64
1da177e4 LT	65
	66	int ext3_check_dir_entry (const char * function, struct inode * dir,
	67	struct ext3_dir_entry_2 * de,
	68	struct buffer_head * bh,
	69	unsigned long offset)
	70	{
	71	const char * error_msg = NULL;
e9ad5620	72	const int rlen = le16_to_cpu(de->rec_len);
1da177e4 LT	73
	74	if (rlen < EXT3_DIR_REC_LEN(1))
	75	error_msg = "rec_len is smaller than minimal";
	76	else if (rlen % 4 != 0)
	77	error_msg = "rec_len % 4 != 0";
	78	else if (rlen < EXT3_DIR_REC_LEN(de->name_len))
	79	error_msg = "rec_len is too small for name_len";
	80	else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)
	81	error_msg = "directory entry across blocks";
	82	else if (le32_to_cpu(de->inode) >
	83	le32_to_cpu(EXT3_SB(dir->i_sb)->s_es->s_inodes_count))
	84	error_msg = "inode out of bounds";
	85
	86	if (error_msg != NULL)
	87	ext3_error (dir->i_sb, function,
	88	"bad entry in directory #%lu: %s - "
	89	"offset=%lu, inode=%lu, rec_len=%d, name_len=%d",
	90	dir->i_ino, error_msg, offset,
	91	(unsigned long) le32_to_cpu(de->inode),
	92	rlen, de->name_len);
	93	return error_msg == NULL ? 1 : 0;
	94	}
	95
	96	static int ext3_readdir(struct file * filp,
	97	void * dirent, filldir_t filldir)
	98	{
	99	int error = 0;
d8733c29 AM	100	unsigned long offset;
	101	int i, stored;
	102	struct ext3_dir_entry_2 *de;
	103	struct super_block *sb;
1da177e4	104	int err;
fe21a693	105	struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4 LT	106	int ret = 0;
	107
	108	sb = inode->i_sb;
	109
	110	#ifdef CONFIG_EXT3_INDEX
	111	if (EXT3_HAS_COMPAT_FEATURE(inode->i_sb,
	112	EXT3_FEATURE_COMPAT_DIR_INDEX) &&
	113	((EXT3_I(inode)->i_flags & EXT3_INDEX_FL) \|\|
	114	((inode->i_size >> sb->s_blocksize_bits) == 1))) {
	115	err = ext3_dx_readdir(filp, dirent, filldir);
	116	if (err != ERR_BAD_DX_DIR) {
	117	ret = err;
	118	goto out;
	119	}
	120	/*
	121	* We don't set the inode dirty flag since it's not
	122	* critical that it get flushed back to the disk.
	123	*/
fe21a693	124	EXT3_I(filp->f_path.dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL;
1da177e4 LT	125	}
	126	#endif
	127	stored = 0;
1da177e4 LT	128	offset = filp->f_pos & (sb->s_blocksize - 1);
	129
	130	while (!error && !stored && filp->f_pos < inode->i_size) {
d8733c29 AM	131	unsigned long blk = filp->f_pos >> EXT3_BLOCK_SIZE_BITS(sb);
	132	struct buffer_head map_bh;
	133	struct buffer_head *bh = NULL;
	134
	135	map_bh.b_state = 0;
89747d36 MC	136	err = ext3_get_blocks_handle(NULL, inode, blk, 1,
	137	&map_bh, 0, 0);
	138	if (err > 0) {
d8733c29 AM	139	page_cache_readahead(sb->s_bdev->bd_inode->i_mapping,
	140	&filp->f_ra,
	141	filp,
	142	map_bh.b_blocknr >>
	143	(PAGE_CACHE_SHIFT - inode->i_blkbits),
	144	1);
	145	bh = ext3_bread(NULL, inode, blk, 0, &err);
	146	}
	147
	148	/*
	149	* We ignore I/O errors on directories so users have a chance
	150	* of recovering data when there's a bad sector
	151	*/
1da177e4 LT	152	if (!bh) {
	153	ext3_error (sb, "ext3_readdir",
	154	"directory #%lu contains a hole at offset %lu",
	155	inode->i_ino, (unsigned long)filp->f_pos);
40b85134 ES	156	/* corrupt size? Maybe no more blocks to read */
	157	if (filp->f_pos > inode->i_blocks << 9)
	158	break;
1da177e4 LT	159	filp->f_pos += sb->s_blocksize - offset;
	160	continue;
	161	}
	162
1da177e4 LT	163	revalidate:
	164	/* If the dir block has changed since the last call to
	165	* readdir(2), then we might be pointing to an invalid
	166	* dirent right now. Scan from the start of the block
	167	* to make sure. */
	168	if (filp->f_version != inode->i_version) {
	169	for (i = 0; i < sb->s_blocksize && i < offset; ) {
ae6ddcc5	170	de = (struct ext3_dir_entry_2 *)
1da177e4 LT	171	(bh->b_data + i);
	172	/* It's too expensive to do a full
	173	* dirent test each time round this
	174	* loop, but we do have to test at
	175	* least that it is non-zero. A
	176	* failure will be detected in the
	177	* dirent test below. */
	178	if (le16_to_cpu(de->rec_len) <
	179	EXT3_DIR_REC_LEN(1))
	180	break;
	181	i += le16_to_cpu(de->rec_len);
	182	}
	183	offset = i;
	184	filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
	185	\| offset;
	186	filp->f_version = inode->i_version;
	187	}
	188
ae6ddcc5	189	while (!error && filp->f_pos < inode->i_size
1da177e4 LT	190	&& offset < sb->s_blocksize) {
	191	de = (struct ext3_dir_entry_2 *) (bh->b_data + offset);
	192	if (!ext3_check_dir_entry ("ext3_readdir", inode, de,
	193	bh, offset)) {
	194	/* On error, skip the f_pos to the
	195	next block. */
	196	filp->f_pos = (filp->f_pos \|
	197	(sb->s_blocksize - 1)) + 1;
	198	brelse (bh);
	199	ret = stored;
	200	goto out;
	201	}
	202	offset += le16_to_cpu(de->rec_len);
	203	if (le32_to_cpu(de->inode)) {
	204	/* We might block in the next section
	205	* if the data destination is
	206	* currently swapped out. So, use a
	207	* version stamp to detect whether or
	208	* not the directory has been modified
	209	* during the copy operation.
	210	*/
	211	unsigned long version = filp->f_version;
	212
	213	error = filldir(dirent, de->name,
	214	de->name_len,
	215	filp->f_pos,
	216	le32_to_cpu(de->inode),
	217	get_dtype(sb, de->file_type));
	218	if (error)
	219	break;
	220	if (version != filp->f_version)
	221	goto revalidate;
	222	stored ++;
	223	}
	224	filp->f_pos += le16_to_cpu(de->rec_len);
	225	}
	226	offset = 0;
	227	brelse (bh);
	228	}
	229	out:
	230	return ret;
	231	}
	232
	233	#ifdef CONFIG_EXT3_INDEX
	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 */
	278	if ((n)->rb_left) {
	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)
	300	root->rb_node = NULL;
	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	}
	307	root->rb_node = NULL;
	308	}
	309
	310
	311	static struct dir_private_info *create_dir_info(loff_t pos)
	312	{
	313	struct dir_private_info *p;
	314
	315	p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL);
	316	if (!p)
	317	return NULL;
	318	p->root.rb_node = NULL;
	319	p->curr_node = NULL;
	320	p->extra_fname = NULL;
	321	p->last_pos = 0;
	322	p->curr_hash = pos2maj_hash(pos);
	323	p->curr_minor_hash = pos2min_hash(pos);
	324	p->next_hash = 0;
	325	return p;
	326	}
	327
	328	void ext3_htree_free_dir_info(struct dir_private_info *p)
	329	{
	330	free_rb_tree_fname(&p->root);
	331	kfree(p);
	332	}
	333
	334	/*
	335	* Given a directory entry, enter it into the fname rb tree.
	336	*/
	337	int ext3_htree_store_dirent(struct file *dir_file, __u32 hash,
	338	__u32 minor_hash,
	339	struct ext3_dir_entry_2 *dirent)
	340	{
	341	struct rb_node *p, parent = NULL;
	342	struct fname * fname, *new_fn;
	343	struct dir_private_info *info;
	344	int len;
	345
	346	info = (struct dir_private_info *) dir_file->private_data;
	347	p = &info->root.rb_node;
	348
	349	/* Create and allocate the fname structure */
	350	len = sizeof(struct fname) + dirent->name_len + 1;
f8314dc6	351	new_fn = kzalloc(len, GFP_KERNEL);
1da177e4 LT	352	if (!new_fn)
1da177e4 LT	353	return -ENOMEM;
1da177e4 LT	354	new_fn->hash = hash;
	355	new_fn->minor_hash = minor_hash;
	356	new_fn->inode = le32_to_cpu(dirent->inode);
	357	new_fn->name_len = dirent->name_len;
	358	new_fn->file_type = dirent->file_type;
	359	memcpy(new_fn->name, dirent->name, dirent->name_len);
	360	new_fn->name[dirent->name_len] = 0;
	361
	362	while (*p) {
	363	parent = *p;
	364	fname = rb_entry(parent, struct fname, rb_hash);
	365
	366	/*
	367	* If the hash and minor hash match up, then we put
	368	* them on a linked list. This rarely happens...
	369	*/
	370	if ((new_fn->hash == fname->hash) &&
	371	(new_fn->minor_hash == fname->minor_hash)) {
	372	new_fn->next = fname->next;
	373	fname->next = new_fn;
	374	return 0;
	375	}
	376
	377	if (new_fn->hash < fname->hash)
	378	p = &(*p)->rb_left;
	379	else if (new_fn->hash > fname->hash)
	380	p = &(*p)->rb_right;
	381	else if (new_fn->minor_hash < fname->minor_hash)
	382	p = &(*p)->rb_left;
	383	else /* if (new_fn->minor_hash > fname->minor_hash) */
	384	p = &(*p)->rb_right;
	385	}
	386
	387	rb_link_node(&new_fn->rb_hash, parent, p);
	388	rb_insert_color(&new_fn->rb_hash, &info->root);
	389	return 0;
	390	}
	391
	392
	393
	394	/*
	395	* This is a helper function for ext3_dx_readdir. It calls filldir
	396	* for all entres on the fname linked list. (Normally there is only
	397	* one entry on the linked list, unless there are 62 bit hash collisions.)
	398	*/
	399	static int call_filldir(struct file * filp, void * dirent,
	400	filldir_t filldir, struct fname *fname)
	401	{
	402	struct dir_private_info *info = filp->private_data;
	403	loff_t curr_pos;
fe21a693	404	struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4 LT	405	struct super_block * sb;
	406	int error;
	407
	408	sb = inode->i_sb;
	409
	410	if (!fname) {
	411	printk("call_filldir: called with null fname?!?\n");
	412	return 0;
	413	}
	414	curr_pos = hash2pos(fname->hash, fname->minor_hash);
	415	while (fname) {
	416	error = filldir(dirent, fname->name,
ae6ddcc5	417	fname->name_len, curr_pos,
1da177e4 LT	418	fname->inode,
	419	get_dtype(sb, fname->file_type));
	420	if (error) {
	421	filp->f_pos = curr_pos;
	422	info->extra_fname = fname->next;
	423	return error;
	424	}
	425	fname = fname->next;
	426	}
	427	return 0;
	428	}
	429
	430	static int ext3_dx_readdir(struct file * filp,
	431	void * dirent, filldir_t filldir)
	432	{
	433	struct dir_private_info *info = filp->private_data;
fe21a693	434	struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4 LT	435	struct fname *fname;
	436	int ret;
	437
	438	if (!info) {
	439	info = create_dir_info(filp->f_pos);
	440	if (!info)
	441	return -ENOMEM;
	442	filp->private_data = info;
	443	}
	444
	445	if (filp->f_pos == EXT3_HTREE_EOF)
	446	return 0; /* EOF */
	447
	448	/* Some one has messed with f_pos; reset the world */
	449	if (info->last_pos != filp->f_pos) {
	450	free_rb_tree_fname(&info->root);
	451	info->curr_node = NULL;
	452	info->extra_fname = NULL;
	453	info->curr_hash = pos2maj_hash(filp->f_pos);
	454	info->curr_minor_hash = pos2min_hash(filp->f_pos);
	455	}
	456
	457	/*
	458	* If there are any leftover names on the hash collision
	459	* chain, return them first.
	460	*/
	461	if (info->extra_fname &&
	462	call_filldir(filp, dirent, filldir, info->extra_fname))
	463	goto finished;
	464
	465	if (!info->curr_node)
	466	info->curr_node = rb_first(&info->root);
	467
	468	while (1) {
	469	/*
	470	* Fill the rbtree if we have no more entries,
	471	* or the inode has changed since we last read in the
ae6ddcc5	472	* cached entries.
1da177e4 LT	473	*/
	474	if ((!info->curr_node) \|\|
	475	(filp->f_version != inode->i_version)) {
	476	info->curr_node = NULL;
	477	free_rb_tree_fname(&info->root);
	478	filp->f_version = inode->i_version;
	479	ret = ext3_htree_fill_tree(filp, info->curr_hash,
	480	info->curr_minor_hash,
	481	&info->next_hash);
	482	if (ret < 0)
	483	return ret;
	484	if (ret == 0) {
	485	filp->f_pos = EXT3_HTREE_EOF;
	486	break;
	487	}
	488	info->curr_node = rb_first(&info->root);
	489	}
	490
	491	fname = rb_entry(info->curr_node, struct fname, rb_hash);
	492	info->curr_hash = fname->hash;
	493	info->curr_minor_hash = fname->minor_hash;
	494	if (call_filldir(filp, dirent, filldir, fname))
	495	break;
	496
	497	info->curr_node = rb_next(info->curr_node);
	498	if (!info->curr_node) {
	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	}
	519
	520	#endif