[linux.git] / fs / ext4 / dir.c

/*
 *  linux/fs/ext4/dir.c
 *
 * Copyright (C) 1992, 1993, 1994, 1995
 * Remy Card ([email protected])
 * Laboratoire MASI - Institut Blaise Pascal
 * Universite Pierre et Marie Curie (Paris VI)
 *
 *  from
 *
 *  linux/fs/minix/dir.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  ext4 directory handling functions
 *
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller ([email protected]), 1995
 *
 * Hash Tree Directory indexing (c) 2001  Daniel Phillips
 *
 */

#include <linux/fs.h>
#include <linux/jbd2.h>
#include <linux/buffer_head.h>
#include <linux/slab.h>
#include <linux/rbtree.h>
#include "ext4.h"

static unsigned char ext4_filetype_table[] = {
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

static int ext4_dx_readdir(struct file *filp,
			   void *dirent, filldir_t filldir);

static unsigned char get_dtype(struct super_block *sb, int filetype)
{
	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
	    (filetype >= EXT4_FT_MAX))
		return DT_UNKNOWN;

	return (ext4_filetype_table[filetype]);
}

/**
 * Check if the given dir-inode refers to an htree-indexed directory
 * (or a directory which chould potentially get coverted to use htree
 * indexing).
 *
 * Return 1 if it is a dx dir, 0 if not
 */
static int is_dx_dir(struct inode *inode)
{
	struct super_block *sb = inode->i_sb;

	if (EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
		     EXT4_FEATURE_COMPAT_DIR_INDEX) &&
	    ((ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) ||
	     ((inode->i_size >> sb->s_blocksize_bits) == 1)))
		return 1;

	return 0;
}

/*
 * Return 0 if the directory entry is OK, and 1 if there is a problem
 *
 * Note: this is the opposite of what ext2 and ext3 historically returned...
 */
int __ext4_check_dir_entry(const char *function, unsigned int line,
			   struct inode *dir, struct file *filp,
			   struct ext4_dir_entry_2 *de,
			   struct buffer_head *bh,
			   unsigned int offset)
{
	const char *error_msg = NULL;
	const int rlen = ext4_rec_len_from_disk(de->rec_len,
						dir->i_sb->s_blocksize);

	if (unlikely(rlen < EXT4_DIR_REC_LEN(1)))
		error_msg = "rec_len is smaller than minimal";
	else if (unlikely(rlen % 4 != 0))
		error_msg = "rec_len % 4 != 0";
	else if (unlikely(rlen < EXT4_DIR_REC_LEN(de->name_len)))
		error_msg = "rec_len is too small for name_len";
	else if (unlikely(((char *) de - bh->b_data) + rlen >
			  dir->i_sb->s_blocksize))
		error_msg = "directory entry across blocks";
	else if (unlikely(le32_to_cpu(de->inode) >
			le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count)))
		error_msg = "inode out of bounds";
	else
		return 0;

	if (filp)
		ext4_error_file(filp, function, line, bh->b_blocknr,
				"bad entry in directory: %s - offset=%u(%u), "
				"inode=%u, rec_len=%d, name_len=%d",
				error_msg, (unsigned) (offset % bh->b_size),
				offset, le32_to_cpu(de->inode),
				rlen, de->name_len);
	else
		ext4_error_inode(dir, function, line, bh->b_blocknr,
				"bad entry in directory: %s - offset=%u(%u), "
				"inode=%u, rec_len=%d, name_len=%d",
				error_msg, (unsigned) (offset % bh->b_size),
				offset, le32_to_cpu(de->inode),
				rlen, de->name_len);

	return 1;
}

static int ext4_readdir(struct file *filp,
			 void *dirent, filldir_t filldir)
{
	int error = 0;
	unsigned int offset;
	int i, stored;
	struct ext4_dir_entry_2 *de;
	int err;
	struct inode *inode = filp->f_path.dentry->d_inode;
	struct super_block *sb = inode->i_sb;
	int ret = 0;
	int dir_has_error = 0;

	if (is_dx_dir(inode)) {
		err = ext4_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.
		 */
		ext4_clear_inode_flag(filp->f_path.dentry->d_inode,
				      EXT4_INODE_INDEX);
	}
	stored = 0;
	offset = filp->f_pos & (sb->s_blocksize - 1);

	while (!error && !stored && filp->f_pos < inode->i_size) {
		struct ext4_map_blocks map;
		struct buffer_head *bh = NULL;

		map.m_lblk = filp->f_pos >> EXT4_BLOCK_SIZE_BITS(sb);
		map.m_len = 1;
		err = ext4_map_blocks(NULL, inode, &map, 0);
		if (err > 0) {
			pgoff_t index = map.m_pblk >>
					(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 = ext4_bread(NULL, inode, map.m_lblk, 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) {
				EXT4_ERROR_FILE(filp, 0,
						"directory contains a "
						"hole at offset %llu",
					   (unsigned long long) 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;
		}

		/* Check the checksum */
		if (!buffer_verified(bh) &&
		    !ext4_dirent_csum_verify(inode,
				(struct ext4_dir_entry *)bh->b_data)) {
			EXT4_ERROR_FILE(filp, 0, "directory fails checksum "
					"at offset %llu",
					(unsigned long long)filp->f_pos);
			filp->f_pos += sb->s_blocksize - offset;
			continue;
		}
		set_buffer_verified(bh);

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 ext4_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 (ext4_rec_len_from_disk(de->rec_len,
					sb->s_blocksize) < EXT4_DIR_REC_LEN(1))
					break;
				i += ext4_rec_len_from_disk(de->rec_len,
							    sb->s_blocksize);
			}
			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 ext4_dir_entry_2 *) (bh->b_data + offset);
			if (ext4_check_dir_entry(inode, filp, 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 += ext4_rec_len_from_disk(de->rec_len,
					sb->s_blocksize);
			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 += ext4_rec_len_from_disk(de->rec_len,
						sb->s_blocksize);
		}
		offset = 0;
		brelse(bh);
	}
out:
	return ret;
}

static inline int is_32bit_api(void)
{
#ifdef CONFIG_COMPAT
	return is_compat_task();
#else
	return (BITS_PER_LONG == 32);
#endif
}

/*
 * These functions convert from the major/minor hash to an f_pos
 * value for dx directories
 *
 * Upper layer (for example NFS) should specify FMODE_32BITHASH or
 * FMODE_64BITHASH explicitly. On the other hand, we allow ext4 to be mounted
 * directly on both 32-bit and 64-bit nodes, under such case, neither
 * FMODE_32BITHASH nor FMODE_64BITHASH is specified.
 */
static inline loff_t hash2pos(struct file *filp, __u32 major, __u32 minor)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return major >> 1;
	else
		return ((__u64)(major >> 1) << 32) | (__u64)minor;
}

static inline __u32 pos2maj_hash(struct file *filp, loff_t pos)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return (pos << 1) & 0xffffffff;
	else
		return ((pos >> 32) << 1) & 0xffffffff;
}

static inline __u32 pos2min_hash(struct file *filp, loff_t pos)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return 0;
	else
		return pos & 0xffffffff;
}

/*
 * Return 32- or 64-bit end-of-file for dx directories
 */
static inline loff_t ext4_get_htree_eof(struct file *filp)
{
	if ((filp->f_mode & FMODE_32BITHASH) ||
	    (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
		return EXT4_HTREE_EOF_32BIT;
	else
		return EXT4_HTREE_EOF_64BIT;
}


/*
 * ext4_dir_llseek() calls generic_file_llseek_size to handle htree
 * directories, where the "offset" is in terms of the filename hash
 * value instead of the byte offset.
 *
 * Because we may return a 64-bit hash that is well beyond offset limits,
 * we need to pass the max hash as the maximum allowable offset in
 * the htree directory case.
 *
 * For non-htree, ext4_llseek already chooses the proper max offset.
 */
loff_t ext4_dir_llseek(struct file *file, loff_t offset, int origin)
{
	struct inode *inode = file->f_mapping->host;
	int dx_dir = is_dx_dir(inode);
	loff_t htree_max = ext4_get_htree_eof(file);

	if (likely(dx_dir))
		return generic_file_llseek_size(file, offset, origin,
						    htree_max, htree_max);
	else
		return ext4_llseek(file, offset, origin);
}

/*
 * 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 *ext4_htree_create_dir_info(struct file *filp,
							   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(filp, pos);
	p->curr_minor_hash = pos2min_hash(filp, pos);
	return p;
}

void ext4_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 ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
			     __u32 minor_hash,
			     struct ext4_dir_entry_2 *dirent)
{
	struct rb_node **p, *parent = NULL;
	struct fname *fname, *new_fn;
	struct dir_private_info *info;
	int len;

	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 ext4_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) {
		ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: comm %s: "
			 "called with null fname?!?", __func__, __LINE__,
			 inode->i_ino, current->comm);
		return 0;
	}
	curr_pos = hash2pos(filp, 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 ext4_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 = ext4_htree_create_dir_info(filp, filp->f_pos);
		if (!info)
			return -ENOMEM;
		filp->private_data = info;
	}

	if (filp->f_pos == ext4_get_htree_eof(filp))
		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, filp->f_pos);
		info->curr_minor_hash = pos2min_hash(filp, 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 = ext4_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 = ext4_get_htree_eof(filp);
				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 = ext4_get_htree_eof(filp);
				break;
			}
			info->curr_hash = info->next_hash;
			info->curr_minor_hash = 0;
		}
	}
finished:
	info->last_pos = filp->f_pos;
	return 0;
}

static int ext4_release_dir(struct inode *inode, struct file *filp)
{
	if (filp->private_data)
		ext4_htree_free_dir_info(filp->private_data);

	return 0;
}

const struct file_operations ext4_dir_operations = {
	.llseek		= ext4_dir_llseek,
	.read		= generic_read_dir,
	.readdir	= ext4_readdir,
	.unlocked_ioctl = ext4_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl	= ext4_compat_ioctl,
#endif
	.fsync		= ext4_sync_file,
	.release	= ext4_release_dir,
};
Commit	Line	Data
ac27a0ec	1	/*
617ba13b	2	* linux/fs/ext4/dir.c
ac27a0ec DK	3	*
	4	* Copyright (C) 1992, 1993, 1994, 1995
	5	* Remy Card ([email protected])
	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	*
617ba13b	15	* ext4 directory handling functions
ac27a0ec DK	16	*
	17	* Big-endian to little-endian byte-swapping/bitmaps by
	18	* David S. Miller ([email protected]), 1995
	19	*
	20	* Hash Tree Directory indexing (c) 2001 Daniel Phillips
	21	*
	22	*/
	23
	24	#include <linux/fs.h>
dab291af	25	#include <linux/jbd2.h>
ac27a0ec	26	#include <linux/buffer_head.h>
ac27a0ec DK	27	#include <linux/slab.h>
ac27a0ec DK	28	#include <linux/rbtree.h>
3dcf5451	29	#include "ext4.h"
ac27a0ec	30
617ba13b	31	static unsigned char ext4_filetype_table[] = {
ac27a0ec DK	32	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
	33	};
	34
af5bc92d TT	35	static int ext4_dx_readdir(struct file *filp,
af5bc92d TT	36	void *dirent, filldir_t filldir);
ac27a0ec DK	37
	38	static unsigned char get_dtype(struct super_block *sb, int filetype)
	39	{
617ba13b MC	40	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) \|\|
617ba13b MC	41	(filetype >= EXT4_FT_MAX))
ac27a0ec DK	42	return DT_UNKNOWN;
ac27a0ec DK	43
617ba13b	44	return (ext4_filetype_table[filetype]);
ac27a0ec DK	45	}
ac27a0ec DK	46
d1f5273e FY	47	/**
	48	* Check if the given dir-inode refers to an htree-indexed directory
	49	* (or a directory which chould potentially get coverted to use htree
	50	* indexing).
	51	*
	52	* Return 1 if it is a dx dir, 0 if not
	53	*/
	54	static int is_dx_dir(struct inode *inode)
	55	{
	56	struct super_block *sb = inode->i_sb;
	57
	58	if (EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
	59	EXT4_FEATURE_COMPAT_DIR_INDEX) &&
	60	((ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) \|\|
	61	((inode->i_size >> sb->s_blocksize_bits) == 1)))
	62	return 1;
	63
	64	return 0;
	65	}
	66
cad3f007 TT	67	/*
	68	* Return 0 if the directory entry is OK, and 1 if there is a problem
	69	*
	70	* Note: this is the opposite of what ext2 and ext3 historically returned...
	71	*/
60fd4da3	72	int __ext4_check_dir_entry(const char *function, unsigned int line,
f7c21177	73	struct inode dir, struct file filp,
60fd4da3 TT	74	struct ext4_dir_entry_2 *de,
	75	struct buffer_head *bh,
	76	unsigned int offset)
ac27a0ec	77	{
af5bc92d	78	const char *error_msg = NULL;
3d0518f4 WY	79	const int rlen = ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	80	dir->i_sb->s_blocksize);
ac27a0ec	81
cad3f007	82	if (unlikely(rlen < EXT4_DIR_REC_LEN(1)))
ac27a0ec	83	error_msg = "rec_len is smaller than minimal";
cad3f007	84	else if (unlikely(rlen % 4 != 0))
ac27a0ec	85	error_msg = "rec_len % 4 != 0";
cad3f007	86	else if (unlikely(rlen < EXT4_DIR_REC_LEN(de->name_len)))
ac27a0ec	87	error_msg = "rec_len is too small for name_len";
cad3f007 TT	88	else if (unlikely(((char *) de - bh->b_data) + rlen >
cad3f007 TT	89	dir->i_sb->s_blocksize))
ac27a0ec	90	error_msg = "directory entry across blocks";
cad3f007 TT	91	else if (unlikely(le32_to_cpu(de->inode) >
cad3f007 TT	92	le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count)))
ac27a0ec	93	error_msg = "inode out of bounds";
cad3f007 TT	94	else
	95	return 0;
	96
f7c21177	97	if (filp)
4fda4003	98	ext4_error_file(filp, function, line, bh->b_blocknr,
f7c21177 TT	99	"bad entry in directory: %s - offset=%u(%u), "
f7c21177 TT	100	"inode=%u, rec_len=%d, name_len=%d",
4fda4003	101	error_msg, (unsigned) (offset % bh->b_size),
f7c21177 TT	102	offset, le32_to_cpu(de->inode),
	103	rlen, de->name_len);
	104	else
4fda4003	105	ext4_error_inode(dir, function, line, bh->b_blocknr,
f7c21177 TT	106	"bad entry in directory: %s - offset=%u(%u), "
f7c21177 TT	107	"inode=%u, rec_len=%d, name_len=%d",
4fda4003	108	error_msg, (unsigned) (offset % bh->b_size),
f7c21177 TT	109	offset, le32_to_cpu(de->inode),
	110	rlen, de->name_len);
	111
cad3f007	112	return 1;
ac27a0ec DK	113	}
ac27a0ec DK	114
af5bc92d TT	115	static int ext4_readdir(struct file *filp,
af5bc92d TT	116	void *dirent, filldir_t filldir)
ac27a0ec DK	117	{
ac27a0ec DK	118	int error = 0;
498e5f24	119	unsigned int offset;
ac27a0ec	120	int i, stored;
617ba13b	121	struct ext4_dir_entry_2 *de;
ac27a0ec	122	int err;
9d549890	123	struct inode *inode = filp->f_path.dentry->d_inode;
d1f5273e	124	struct super_block *sb = inode->i_sb;
ac27a0ec	125	int ret = 0;
9d9f1775	126	int dir_has_error = 0;
ac27a0ec	127
d1f5273e	128	if (is_dx_dir(inode)) {
617ba13b	129	err = ext4_dx_readdir(filp, dirent, filldir);
ac27a0ec DK	130	if (err != ERR_BAD_DX_DIR) {
	131	ret = err;
	132	goto out;
	133	}
	134	/*
	135	* We don't set the inode dirty flag since it's not
	136	* critical that it get flushed back to the disk.
	137	*/
07a03824 TT	138	ext4_clear_inode_flag(filp->f_path.dentry->d_inode,
07a03824 TT	139	EXT4_INODE_INDEX);
ac27a0ec	140	}
ac27a0ec DK	141	stored = 0;
	142	offset = filp->f_pos & (sb->s_blocksize - 1);
	143
	144	while (!error && !stored && filp->f_pos < inode->i_size) {
2ed88685	145	struct ext4_map_blocks map;
ac27a0ec DK	146	struct buffer_head *bh = NULL;
ac27a0ec DK	147
2ed88685 TT	148	map.m_lblk = filp->f_pos >> EXT4_BLOCK_SIZE_BITS(sb);
	149	map.m_len = 1;
	150	err = ext4_map_blocks(NULL, inode, &map, 0);
ac27a0ec	151	if (err > 0) {
2ed88685	152	pgoff_t index = map.m_pblk >>
dc7868fc FW	153	(PAGE_CACHE_SHIFT - inode->i_blkbits);
dc7868fc FW	154	if (!ra_has_index(&filp->f_ra, index))
cf914a7d	155	page_cache_sync_readahead(
dc7868fc FW	156	sb->s_bdev->bd_inode->i_mapping,
dc7868fc FW	157	&filp->f_ra, filp,
cf914a7d	158	index, 1);
f4e6b498	159	filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
2ed88685	160	bh = ext4_bread(NULL, inode, map.m_lblk, 0, &err);
ac27a0ec DK	161	}
	162
	163	/*
	164	* We ignore I/O errors on directories so users have a chance
	165	* of recovering data when there's a bad sector
	166	*/
	167	if (!bh) {
9d9f1775	168	if (!dir_has_error) {
f7c21177 TT	169	EXT4_ERROR_FILE(filp, 0,
	170	"directory contains a "
	171	"hole at offset %llu",
9d9f1775 ES	172	(unsigned long long) filp->f_pos);
	173	dir_has_error = 1;
	174	}
e6c40211 ES	175	/* corrupt size? Maybe no more blocks to read */
	176	if (filp->f_pos > inode->i_blocks << 9)
	177	break;
ac27a0ec DK	178	filp->f_pos += sb->s_blocksize - offset;
	179	continue;
	180	}
	181
b0336e8d DW	182	/* Check the checksum */
	183	if (!buffer_verified(bh) &&
	184	!ext4_dirent_csum_verify(inode,
	185	(struct ext4_dir_entry *)bh->b_data)) {
	186	EXT4_ERROR_FILE(filp, 0, "directory fails checksum "
	187	"at offset %llu",
	188	(unsigned long long)filp->f_pos);
	189	filp->f_pos += sb->s_blocksize - offset;
	190	continue;
	191	}
	192	set_buffer_verified(bh);
	193
ac27a0ec DK	194	revalidate:
	195	/* If the dir block has changed since the last call to
	196	* readdir(2), then we might be pointing to an invalid
	197	* dirent right now. Scan from the start of the block
	198	* to make sure. */
	199	if (filp->f_version != inode->i_version) {
	200	for (i = 0; i < sb->s_blocksize && i < offset; ) {
617ba13b	201	de = (struct ext4_dir_entry_2 *)
ac27a0ec DK	202	(bh->b_data + i);
	203	/* It's too expensive to do a full
	204	* dirent test each time round this
	205	* loop, but we do have to test at
	206	* least that it is non-zero. A
	207	* failure will be detected in the
	208	* dirent test below. */
3d0518f4 WY	209	if (ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	210	sb->s_blocksize) < EXT4_DIR_REC_LEN(1))
ac27a0ec	211	break;
3d0518f4 WY	212	i += ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	213	sb->s_blocksize);
ac27a0ec DK	214	}
	215	offset = i;
	216	filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
	217	\| offset;
	218	filp->f_version = inode->i_version;
	219	}
	220
	221	while (!error && filp->f_pos < inode->i_size
	222	&& offset < sb->s_blocksize) {
617ba13b	223	de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
f7c21177	224	if (ext4_check_dir_entry(inode, filp, de,
cad3f007	225	bh, offset)) {
63f57933 AM	226	/*
	227	* On error, skip the f_pos to the next block
	228	*/
ac27a0ec DK	229	filp->f_pos = (filp->f_pos \|
ac27a0ec DK	230	(sb->s_blocksize - 1)) + 1;
af5bc92d	231	brelse(bh);
ac27a0ec DK	232	ret = stored;
	233	goto out;
	234	}
3d0518f4 WY	235	offset += ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	236	sb->s_blocksize);
ac27a0ec DK	237	if (le32_to_cpu(de->inode)) {
	238	/* We might block in the next section
	239	* if the data destination is
	240	* currently swapped out. So, use a
	241	* version stamp to detect whether or
	242	* not the directory has been modified
	243	* during the copy operation.
	244	*/
2b47c361	245	u64 version = filp->f_version;
ac27a0ec DK	246
	247	error = filldir(dirent, de->name,
	248	de->name_len,
	249	filp->f_pos,
	250	le32_to_cpu(de->inode),
	251	get_dtype(sb, de->file_type));
	252	if (error)
	253	break;
	254	if (version != filp->f_version)
	255	goto revalidate;
af5bc92d	256	stored++;
ac27a0ec	257	}
3d0518f4 WY	258	filp->f_pos += ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	259	sb->s_blocksize);
ac27a0ec DK	260	}
ac27a0ec DK	261	offset = 0;
af5bc92d	262	brelse(bh);
ac27a0ec DK	263	}
	264	out:
	265	return ret;
	266	}
	267
d1f5273e FY	268	static inline int is_32bit_api(void)
	269	{
	270	#ifdef CONFIG_COMPAT
	271	return is_compat_task();
	272	#else
	273	return (BITS_PER_LONG == 32);
	274	#endif
	275	}
	276
ac27a0ec DK	277	/*
ac27a0ec DK	278	* These functions convert from the major/minor hash to an f_pos
d1f5273e FY	279	* value for dx directories
	280	*
	281	* Upper layer (for example NFS) should specify FMODE_32BITHASH or
	282	* FMODE_64BITHASH explicitly. On the other hand, we allow ext4 to be mounted
	283	* directly on both 32-bit and 64-bit nodes, under such case, neither
	284	* FMODE_32BITHASH nor FMODE_64BITHASH is specified.
	285	*/
	286	static inline loff_t hash2pos(struct file *filp, __u32 major, __u32 minor)
	287	{
	288	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	289	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	290	return major >> 1;
	291	else
	292	return ((__u64)(major >> 1) << 32) \| (__u64)minor;
	293	}
	294
	295	static inline __u32 pos2maj_hash(struct file *filp, loff_t pos)
	296	{
	297	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	298	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	299	return (pos << 1) & 0xffffffff;
	300	else
	301	return ((pos >> 32) << 1) & 0xffffffff;
	302	}
	303
	304	static inline __u32 pos2min_hash(struct file *filp, loff_t pos)
	305	{
	306	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	307	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	308	return 0;
	309	else
	310	return pos & 0xffffffff;
	311	}
	312
	313	/*
	314	* Return 32- or 64-bit end-of-file for dx directories
	315	*/
	316	static inline loff_t ext4_get_htree_eof(struct file *filp)
	317	{
	318	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	319	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	320	return EXT4_HTREE_EOF_32BIT;
	321	else
	322	return EXT4_HTREE_EOF_64BIT;
	323	}
	324
	325
	326	/*
ec7268ce ES	327	* ext4_dir_llseek() calls generic_file_llseek_size to handle htree
	328	* directories, where the "offset" is in terms of the filename hash
	329	* value instead of the byte offset.
ac27a0ec	330	*
ec7268ce ES	331	* Because we may return a 64-bit hash that is well beyond offset limits,
	332	* we need to pass the max hash as the maximum allowable offset in
	333	* the htree directory case.
	334	*
	335	* For non-htree, ext4_llseek already chooses the proper max offset.
ac27a0ec	336	*/
d1f5273e FY	337	loff_t ext4_dir_llseek(struct file *file, loff_t offset, int origin)
	338	{
	339	struct inode *inode = file->f_mapping->host;
d1f5273e	340	int dx_dir = is_dx_dir(inode);
ec7268ce	341	loff_t htree_max = ext4_get_htree_eof(file);
d1f5273e	342
ec7268ce ES	343	if (likely(dx_dir))
	344	return generic_file_llseek_size(file, offset, origin,
	345	htree_max, htree_max);
	346	else
	347	return ext4_llseek(file, offset, origin);
d1f5273e	348	}
ac27a0ec DK	349
	350	/*
	351	* This structure holds the nodes of the red-black tree used to store
	352	* the directory entry in hash order.
	353	*/
	354	struct fname {
	355	__u32 hash;
	356	__u32 minor_hash;
	357	struct rb_node rb_hash;
	358	struct fname *next;
	359	__u32 inode;
	360	__u8 name_len;
	361	__u8 file_type;
	362	char name[0];
	363	};
	364
	365	/*
	366	* This functoin implements a non-recursive way of freeing all of the
	367	* nodes in the red-black tree.
	368	*/
	369	static void free_rb_tree_fname(struct rb_root *root)
	370	{
	371	struct rb_node *n = root->rb_node;
	372	struct rb_node *parent;
	373	struct fname *fname;
	374
	375	while (n) {
	376	/* Do the node's children first */
69baee06	377	if (n->rb_left) {
ac27a0ec DK	378	n = n->rb_left;
	379	continue;
	380	}
	381	if (n->rb_right) {
	382	n = n->rb_right;
	383	continue;
	384	}
	385	/*
	386	* The node has no children; free it, and then zero
	387	* out parent's link to it. Finally go to the
	388	* beginning of the loop and try to free the parent
	389	* node.
	390	*/
	391	parent = rb_parent(n);
	392	fname = rb_entry(n, struct fname, rb_hash);
	393	while (fname) {
af5bc92d	394	struct fname *old = fname;
ac27a0ec	395	fname = fname->next;
af5bc92d	396	kfree(old);
ac27a0ec DK	397	}
ac27a0ec DK	398	if (!parent)
64e290ec	399	*root = RB_ROOT;
ac27a0ec DK	400	else if (parent->rb_left == n)
	401	parent->rb_left = NULL;
	402	else if (parent->rb_right == n)
	403	parent->rb_right = NULL;
	404	n = parent;
	405	}
ac27a0ec DK	406	}
	407
	408
d1f5273e FY	409	static struct dir_private_info ext4_htree_create_dir_info(struct file filp,
d1f5273e FY	410	loff_t pos)
ac27a0ec DK	411	{
	412	struct dir_private_info *p;
	413
69baee06	414	p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
ac27a0ec DK	415	if (!p)
ac27a0ec DK	416	return NULL;
d1f5273e FY	417	p->curr_hash = pos2maj_hash(filp, pos);
d1f5273e FY	418	p->curr_minor_hash = pos2min_hash(filp, pos);
ac27a0ec DK	419	return p;
	420	}
	421
617ba13b	422	void ext4_htree_free_dir_info(struct dir_private_info *p)
ac27a0ec DK	423	{
	424	free_rb_tree_fname(&p->root);
	425	kfree(p);
	426	}
	427
	428	/*
	429	* Given a directory entry, enter it into the fname rb tree.
	430	*/
617ba13b	431	int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
ac27a0ec	432	__u32 minor_hash,
617ba13b	433	struct ext4_dir_entry_2 *dirent)
ac27a0ec DK	434	{
ac27a0ec DK	435	struct rb_node *p, parent = NULL;
af5bc92d	436	struct fname fname, new_fn;
ac27a0ec DK	437	struct dir_private_info *info;
	438	int len;
	439
a271fe85	440	info = dir_file->private_data;
ac27a0ec DK	441	p = &info->root.rb_node;
	442
	443	/* Create and allocate the fname structure */
	444	len = sizeof(struct fname) + dirent->name_len + 1;
	445	new_fn = kzalloc(len, GFP_KERNEL);
	446	if (!new_fn)
	447	return -ENOMEM;
	448	new_fn->hash = hash;
	449	new_fn->minor_hash = minor_hash;
	450	new_fn->inode = le32_to_cpu(dirent->inode);
	451	new_fn->name_len = dirent->name_len;
	452	new_fn->file_type = dirent->file_type;
	453	memcpy(new_fn->name, dirent->name, dirent->name_len);
	454	new_fn->name[dirent->name_len] = 0;
	455
	456	while (*p) {
	457	parent = *p;
	458	fname = rb_entry(parent, struct fname, rb_hash);
	459
	460	/*
	461	* If the hash and minor hash match up, then we put
	462	* them on a linked list. This rarely happens...
	463	*/
	464	if ((new_fn->hash == fname->hash) &&
	465	(new_fn->minor_hash == fname->minor_hash)) {
	466	new_fn->next = fname->next;
	467	fname->next = new_fn;
	468	return 0;
	469	}
	470
	471	if (new_fn->hash < fname->hash)
	472	p = &(*p)->rb_left;
	473	else if (new_fn->hash > fname->hash)
	474	p = &(*p)->rb_right;
	475	else if (new_fn->minor_hash < fname->minor_hash)
	476	p = &(*p)->rb_left;
	477	else /* if (new_fn->minor_hash > fname->minor_hash) */
	478	p = &(*p)->rb_right;
	479	}
	480
	481	rb_link_node(&new_fn->rb_hash, parent, p);
	482	rb_insert_color(&new_fn->rb_hash, &info->root);
	483	return 0;
	484	}
	485
	486
	487
	488	/*
617ba13b	489	* This is a helper function for ext4_dx_readdir. It calls filldir
ac27a0ec DK	490	* for all entres on the fname linked list. (Normally there is only
	491	* one entry on the linked list, unless there are 62 bit hash collisions.)
	492	*/
af5bc92d	493	static int call_filldir(struct file filp, void dirent,
ac27a0ec DK	494	filldir_t filldir, struct fname *fname)
	495	{
	496	struct dir_private_info *info = filp->private_data;
	497	loff_t curr_pos;
9d549890	498	struct inode *inode = filp->f_path.dentry->d_inode;
af5bc92d	499	struct super_block *sb;
ac27a0ec DK	500	int error;
	501
	502	sb = inode->i_sb;
	503
	504	if (!fname) {
92b97816 TT	505	ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: comm %s: "
	506	"called with null fname?!?", __func__, __LINE__,
	507	inode->i_ino, current->comm);
ac27a0ec DK	508	return 0;
ac27a0ec DK	509	}
d1f5273e	510	curr_pos = hash2pos(filp, fname->hash, fname->minor_hash);
ac27a0ec DK	511	while (fname) {
	512	error = filldir(dirent, fname->name,
	513	fname->name_len, curr_pos,
	514	fname->inode,
	515	get_dtype(sb, fname->file_type));
	516	if (error) {
	517	filp->f_pos = curr_pos;
d0156417	518	info->extra_fname = fname;
ac27a0ec DK	519	return error;
	520	}
	521	fname = fname->next;
	522	}
	523	return 0;
	524	}
	525
af5bc92d TT	526	static int ext4_dx_readdir(struct file *filp,
af5bc92d TT	527	void *dirent, filldir_t filldir)
ac27a0ec DK	528	{
ac27a0ec DK	529	struct dir_private_info *info = filp->private_data;
9d549890	530	struct inode *inode = filp->f_path.dentry->d_inode;
ac27a0ec DK	531	struct fname *fname;
	532	int ret;
	533
	534	if (!info) {
d1f5273e	535	info = ext4_htree_create_dir_info(filp, filp->f_pos);
ac27a0ec DK	536	if (!info)
	537	return -ENOMEM;
	538	filp->private_data = info;
	539	}
	540
d1f5273e	541	if (filp->f_pos == ext4_get_htree_eof(filp))
ac27a0ec DK	542	return 0; /* EOF */
	543
	544	/* Some one has messed with f_pos; reset the world */
	545	if (info->last_pos != filp->f_pos) {
	546	free_rb_tree_fname(&info->root);
	547	info->curr_node = NULL;
	548	info->extra_fname = NULL;
d1f5273e FY	549	info->curr_hash = pos2maj_hash(filp, filp->f_pos);
d1f5273e FY	550	info->curr_minor_hash = pos2min_hash(filp, filp->f_pos);
ac27a0ec DK	551	}
	552
	553	/*
	554	* If there are any leftover names on the hash collision
	555	* chain, return them first.
	556	*/
d0156417 TT	557	if (info->extra_fname) {
	558	if (call_filldir(filp, dirent, filldir, info->extra_fname))
	559	goto finished;
d0156417	560	info->extra_fname = NULL;
3c37fc86	561	goto next_node;
d0156417	562	} else if (!info->curr_node)
ac27a0ec DK	563	info->curr_node = rb_first(&info->root);
	564
	565	while (1) {
	566	/*
	567	* Fill the rbtree if we have no more entries,
	568	* or the inode has changed since we last read in the
	569	* cached entries.
	570	*/
	571	if ((!info->curr_node) \|\|
	572	(filp->f_version != inode->i_version)) {
	573	info->curr_node = NULL;
	574	free_rb_tree_fname(&info->root);
	575	filp->f_version = inode->i_version;
617ba13b	576	ret = ext4_htree_fill_tree(filp, info->curr_hash,
ac27a0ec DK	577	info->curr_minor_hash,
	578	&info->next_hash);
	579	if (ret < 0)
	580	return ret;
	581	if (ret == 0) {
d1f5273e	582	filp->f_pos = ext4_get_htree_eof(filp);
ac27a0ec DK	583	break;
	584	}
	585	info->curr_node = rb_first(&info->root);
	586	}
	587
	588	fname = rb_entry(info->curr_node, struct fname, rb_hash);
	589	info->curr_hash = fname->hash;
	590	info->curr_minor_hash = fname->minor_hash;
	591	if (call_filldir(filp, dirent, filldir, fname))
	592	break;
3c37fc86	593	next_node:
ac27a0ec	594	info->curr_node = rb_next(info->curr_node);
3c37fc86 TT	595	if (info->curr_node) {
	596	fname = rb_entry(info->curr_node, struct fname,
	597	rb_hash);
	598	info->curr_hash = fname->hash;
	599	info->curr_minor_hash = fname->minor_hash;
	600	} else {
ac27a0ec	601	if (info->next_hash == ~0) {
d1f5273e	602	filp->f_pos = ext4_get_htree_eof(filp);
ac27a0ec DK	603	break;
	604	}
	605	info->curr_hash = info->next_hash;
	606	info->curr_minor_hash = 0;
	607	}
	608	}
	609	finished:
	610	info->last_pos = filp->f_pos;
	611	return 0;
	612	}
	613
af5bc92d	614	static int ext4_release_dir(struct inode inode, struct file filp)
ac27a0ec	615	{
63f57933	616	if (filp->private_data)
617ba13b	617	ext4_htree_free_dir_info(filp->private_data);
ac27a0ec DK	618
	619	return 0;
	620	}
d1f5273e FY	621
	622	const struct file_operations ext4_dir_operations = {
	623	.llseek = ext4_dir_llseek,
	624	.read = generic_read_dir,
	625	.readdir = ext4_readdir,
	626	.unlocked_ioctl = ext4_ioctl,
	627	#ifdef CONFIG_COMPAT
	628	.compat_ioctl = ext4_compat_ioctl,
	629	#endif
	630	.fsync = ext4_sync_file,
	631	.release = ext4_release_dir,
	632	};