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

/*
 *  linux/fs/ext4/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
 *
 *  ext4 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/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;
		}

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() based on generic_file_llseek() to handle both
 * non-htree and htree directories, where the "offset" is in terms
 * of the filename hash value instead of the byte offset.
 *
 * NOTE: offsets obtained *before* ext4_set_inode_flag(dir, EXT4_INODE_INDEX)
 *       will be invalid once the directory was converted into a dx directory
 */
loff_t ext4_dir_llseek(struct file *file, loff_t offset, int origin)
{
	struct inode *inode = file->f_mapping->host;
	loff_t ret = -EINVAL;
	int dx_dir = is_dx_dir(inode);

	mutex_lock(&inode->i_mutex);

	/* NOTE: relative offsets with dx directories might not work
	 *       as expected, as it is difficult to figure out the
	 *       correct offset between dx hashes */

	switch (origin) {
	case SEEK_END:
		if (unlikely(offset > 0))
			goto out_err; /* not supported for directories */

		/* so only negative offsets are left, does that have a
		 * meaning for directories at all? */
		if (dx_dir)
			offset += ext4_get_htree_eof(file);
		else
			offset += inode->i_size;
		break;
	case SEEK_CUR:
		/*
		 * Here we special-case the lseek(fd, 0, SEEK_CUR)
		 * position-querying operation.  Avoid rewriting the "same"
		 * f_pos value back to the file because a concurrent read(),
		 * write() or lseek() might have altered it
		 */
		if (offset == 0) {
			offset = file->f_pos;
			goto out_ok;
		}

		offset += file->f_pos;
		break;
	}

	if (unlikely(offset < 0))
		goto out_err;

	if (!dx_dir) {
		if (offset > inode->i_sb->s_maxbytes)
			goto out_err;
	} else if (offset > ext4_get_htree_eof(file))
		goto out_err;

	/* Special lock needed here? */
	if (offset != file->f_pos) {
		file->f_pos = offset;
		file->f_version = 0;
	}

out_ok:
	ret = offset;
out_err:
	mutex_unlock(&inode->i_mutex);

	return ret;
}

/*
 * 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 (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	*
617ba13b	15	* ext4 directory handling functions
ac27a0ec DK	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>
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
	182	revalidate:
	183	/* If the dir block has changed since the last call to
	184	* readdir(2), then we might be pointing to an invalid
	185	* dirent right now. Scan from the start of the block
	186	* to make sure. */
	187	if (filp->f_version != inode->i_version) {
	188	for (i = 0; i < sb->s_blocksize && i < offset; ) {
617ba13b	189	de = (struct ext4_dir_entry_2 *)
ac27a0ec DK	190	(bh->b_data + i);
	191	/* It's too expensive to do a full
	192	* dirent test each time round this
	193	* loop, but we do have to test at
	194	* least that it is non-zero. A
	195	* failure will be detected in the
	196	* dirent test below. */
3d0518f4 WY	197	if (ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	198	sb->s_blocksize) < EXT4_DIR_REC_LEN(1))
ac27a0ec	199	break;
3d0518f4 WY	200	i += ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	201	sb->s_blocksize);
ac27a0ec DK	202	}
	203	offset = i;
	204	filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
	205	\| offset;
	206	filp->f_version = inode->i_version;
	207	}
	208
	209	while (!error && filp->f_pos < inode->i_size
	210	&& offset < sb->s_blocksize) {
617ba13b	211	de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
f7c21177	212	if (ext4_check_dir_entry(inode, filp, de,
cad3f007	213	bh, offset)) {
63f57933 AM	214	/*
	215	* On error, skip the f_pos to the next block
	216	*/
ac27a0ec DK	217	filp->f_pos = (filp->f_pos \|
ac27a0ec DK	218	(sb->s_blocksize - 1)) + 1;
af5bc92d	219	brelse(bh);
ac27a0ec DK	220	ret = stored;
	221	goto out;
	222	}
3d0518f4 WY	223	offset += ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	224	sb->s_blocksize);
ac27a0ec DK	225	if (le32_to_cpu(de->inode)) {
	226	/* We might block in the next section
	227	* if the data destination is
	228	* currently swapped out. So, use a
	229	* version stamp to detect whether or
	230	* not the directory has been modified
	231	* during the copy operation.
	232	*/
2b47c361	233	u64 version = filp->f_version;
ac27a0ec DK	234
	235	error = filldir(dirent, de->name,
	236	de->name_len,
	237	filp->f_pos,
	238	le32_to_cpu(de->inode),
	239	get_dtype(sb, de->file_type));
	240	if (error)
	241	break;
	242	if (version != filp->f_version)
	243	goto revalidate;
af5bc92d	244	stored++;
ac27a0ec	245	}
3d0518f4 WY	246	filp->f_pos += ext4_rec_len_from_disk(de->rec_len,
3d0518f4 WY	247	sb->s_blocksize);
ac27a0ec DK	248	}
ac27a0ec DK	249	offset = 0;
af5bc92d	250	brelse(bh);
ac27a0ec DK	251	}
	252	out:
	253	return ret;
	254	}
	255
d1f5273e FY	256	static inline int is_32bit_api(void)
	257	{
	258	#ifdef CONFIG_COMPAT
	259	return is_compat_task();
	260	#else
	261	return (BITS_PER_LONG == 32);
	262	#endif
	263	}
	264
ac27a0ec DK	265	/*
ac27a0ec DK	266	* These functions convert from the major/minor hash to an f_pos
d1f5273e FY	267	* value for dx directories
	268	*
	269	* Upper layer (for example NFS) should specify FMODE_32BITHASH or
	270	* FMODE_64BITHASH explicitly. On the other hand, we allow ext4 to be mounted
	271	* directly on both 32-bit and 64-bit nodes, under such case, neither
	272	* FMODE_32BITHASH nor FMODE_64BITHASH is specified.
	273	*/
	274	static inline loff_t hash2pos(struct file *filp, __u32 major, __u32 minor)
	275	{
	276	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	277	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	278	return major >> 1;
	279	else
	280	return ((__u64)(major >> 1) << 32) \| (__u64)minor;
	281	}
	282
	283	static inline __u32 pos2maj_hash(struct file *filp, loff_t pos)
	284	{
	285	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	286	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	287	return (pos << 1) & 0xffffffff;
	288	else
	289	return ((pos >> 32) << 1) & 0xffffffff;
	290	}
	291
	292	static inline __u32 pos2min_hash(struct file *filp, loff_t pos)
	293	{
	294	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	295	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	296	return 0;
	297	else
	298	return pos & 0xffffffff;
	299	}
	300
	301	/*
	302	* Return 32- or 64-bit end-of-file for dx directories
	303	*/
	304	static inline loff_t ext4_get_htree_eof(struct file *filp)
	305	{
	306	if ((filp->f_mode & FMODE_32BITHASH) \|\|
	307	(!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
	308	return EXT4_HTREE_EOF_32BIT;
	309	else
	310	return EXT4_HTREE_EOF_64BIT;
	311	}
	312
	313
	314	/*
	315	* ext4_dir_llseek() based on generic_file_llseek() to handle both
	316	* non-htree and htree directories, where the "offset" is in terms
	317	* of the filename hash value instead of the byte offset.
ac27a0ec	318	*
d1f5273e FY	319	* NOTE: offsets obtained before ext4_set_inode_flag(dir, EXT4_INODE_INDEX)
d1f5273e FY	320	* will be invalid once the directory was converted into a dx directory
ac27a0ec	321	*/
d1f5273e FY	322	loff_t ext4_dir_llseek(struct file *file, loff_t offset, int origin)
	323	{
	324	struct inode *inode = file->f_mapping->host;
	325	loff_t ret = -EINVAL;
	326	int dx_dir = is_dx_dir(inode);
	327
	328	mutex_lock(&inode->i_mutex);
	329
	330	/* NOTE: relative offsets with dx directories might not work
	331	* as expected, as it is difficult to figure out the
	332	* correct offset between dx hashes */
	333
	334	switch (origin) {
	335	case SEEK_END:
	336	if (unlikely(offset > 0))
	337	goto out_err; /* not supported for directories */
	338
	339	/* so only negative offsets are left, does that have a
	340	* meaning for directories at all? */
	341	if (dx_dir)
	342	offset += ext4_get_htree_eof(file);
	343	else
	344	offset += inode->i_size;
	345	break;
	346	case SEEK_CUR:
	347	/*
	348	* Here we special-case the lseek(fd, 0, SEEK_CUR)
	349	* position-querying operation. Avoid rewriting the "same"
	350	* f_pos value back to the file because a concurrent read(),
	351	* write() or lseek() might have altered it
	352	*/
	353	if (offset == 0) {
	354	offset = file->f_pos;
	355	goto out_ok;
	356	}
	357
	358	offset += file->f_pos;
	359	break;
	360	}
	361
	362	if (unlikely(offset < 0))
	363	goto out_err;
	364
	365	if (!dx_dir) {
	366	if (offset > inode->i_sb->s_maxbytes)
	367	goto out_err;
	368	} else if (offset > ext4_get_htree_eof(file))
	369	goto out_err;
	370
	371	/* Special lock needed here? */
	372	if (offset != file->f_pos) {
	373	file->f_pos = offset;
	374	file->f_version = 0;
	375	}
	376
	377	out_ok:
	378	ret = offset;
	379	out_err:
	380	mutex_unlock(&inode->i_mutex);
	381
	382	return ret;
	383	}
ac27a0ec DK	384
	385	/*
	386	* This structure holds the nodes of the red-black tree used to store
	387	* the directory entry in hash order.
	388	*/
	389	struct fname {
	390	__u32 hash;
	391	__u32 minor_hash;
	392	struct rb_node rb_hash;
	393	struct fname *next;
	394	__u32 inode;
	395	__u8 name_len;
	396	__u8 file_type;
	397	char name[0];
	398	};
	399
	400	/*
	401	* This functoin implements a non-recursive way of freeing all of the
	402	* nodes in the red-black tree.
	403	*/
	404	static void free_rb_tree_fname(struct rb_root *root)
	405	{
	406	struct rb_node *n = root->rb_node;
	407	struct rb_node *parent;
	408	struct fname *fname;
	409
	410	while (n) {
	411	/* Do the node's children first */
69baee06	412	if (n->rb_left) {
ac27a0ec DK	413	n = n->rb_left;
	414	continue;
	415	}
	416	if (n->rb_right) {
	417	n = n->rb_right;
	418	continue;
	419	}
	420	/*
	421	* The node has no children; free it, and then zero
	422	* out parent's link to it. Finally go to the
	423	* beginning of the loop and try to free the parent
	424	* node.
	425	*/
	426	parent = rb_parent(n);
	427	fname = rb_entry(n, struct fname, rb_hash);
	428	while (fname) {
af5bc92d	429	struct fname *old = fname;
ac27a0ec	430	fname = fname->next;
af5bc92d	431	kfree(old);
ac27a0ec DK	432	}
ac27a0ec DK	433	if (!parent)
64e290ec	434	*root = RB_ROOT;
ac27a0ec DK	435	else if (parent->rb_left == n)
	436	parent->rb_left = NULL;
	437	else if (parent->rb_right == n)
	438	parent->rb_right = NULL;
	439	n = parent;
	440	}
ac27a0ec DK	441	}
	442
	443
d1f5273e FY	444	static struct dir_private_info ext4_htree_create_dir_info(struct file filp,
d1f5273e FY	445	loff_t pos)
ac27a0ec DK	446	{
	447	struct dir_private_info *p;
	448
69baee06	449	p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
ac27a0ec DK	450	if (!p)
ac27a0ec DK	451	return NULL;
d1f5273e FY	452	p->curr_hash = pos2maj_hash(filp, pos);
d1f5273e FY	453	p->curr_minor_hash = pos2min_hash(filp, pos);
ac27a0ec DK	454	return p;
	455	}
	456
617ba13b	457	void ext4_htree_free_dir_info(struct dir_private_info *p)
ac27a0ec DK	458	{
	459	free_rb_tree_fname(&p->root);
	460	kfree(p);
	461	}
	462
	463	/*
	464	* Given a directory entry, enter it into the fname rb tree.
	465	*/
617ba13b	466	int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
ac27a0ec	467	__u32 minor_hash,
617ba13b	468	struct ext4_dir_entry_2 *dirent)
ac27a0ec DK	469	{
ac27a0ec DK	470	struct rb_node *p, parent = NULL;
af5bc92d	471	struct fname fname, new_fn;
ac27a0ec DK	472	struct dir_private_info *info;
	473	int len;
	474
a271fe85	475	info = dir_file->private_data;
ac27a0ec DK	476	p = &info->root.rb_node;
	477
	478	/* Create and allocate the fname structure */
	479	len = sizeof(struct fname) + dirent->name_len + 1;
	480	new_fn = kzalloc(len, GFP_KERNEL);
	481	if (!new_fn)
	482	return -ENOMEM;
	483	new_fn->hash = hash;
	484	new_fn->minor_hash = minor_hash;
	485	new_fn->inode = le32_to_cpu(dirent->inode);
	486	new_fn->name_len = dirent->name_len;
	487	new_fn->file_type = dirent->file_type;
	488	memcpy(new_fn->name, dirent->name, dirent->name_len);
	489	new_fn->name[dirent->name_len] = 0;
	490
	491	while (*p) {
	492	parent = *p;
	493	fname = rb_entry(parent, struct fname, rb_hash);
	494
	495	/*
	496	* If the hash and minor hash match up, then we put
	497	* them on a linked list. This rarely happens...
	498	*/
	499	if ((new_fn->hash == fname->hash) &&
	500	(new_fn->minor_hash == fname->minor_hash)) {
	501	new_fn->next = fname->next;
	502	fname->next = new_fn;
	503	return 0;
	504	}
	505
	506	if (new_fn->hash < fname->hash)
	507	p = &(*p)->rb_left;
	508	else if (new_fn->hash > fname->hash)
	509	p = &(*p)->rb_right;
	510	else if (new_fn->minor_hash < fname->minor_hash)
	511	p = &(*p)->rb_left;
	512	else /* if (new_fn->minor_hash > fname->minor_hash) */
	513	p = &(*p)->rb_right;
	514	}
	515
	516	rb_link_node(&new_fn->rb_hash, parent, p);
	517	rb_insert_color(&new_fn->rb_hash, &info->root);
	518	return 0;
	519	}
	520
	521
	522
	523	/*
617ba13b	524	* This is a helper function for ext4_dx_readdir. It calls filldir
ac27a0ec DK	525	* for all entres on the fname linked list. (Normally there is only
	526	* one entry on the linked list, unless there are 62 bit hash collisions.)
	527	*/
af5bc92d	528	static int call_filldir(struct file filp, void dirent,
ac27a0ec DK	529	filldir_t filldir, struct fname *fname)
	530	{
	531	struct dir_private_info *info = filp->private_data;
	532	loff_t curr_pos;
9d549890	533	struct inode *inode = filp->f_path.dentry->d_inode;
af5bc92d	534	struct super_block *sb;
ac27a0ec DK	535	int error;
	536
	537	sb = inode->i_sb;
	538
	539	if (!fname) {
92b97816 TT	540	ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: comm %s: "
	541	"called with null fname?!?", __func__, __LINE__,
	542	inode->i_ino, current->comm);
ac27a0ec DK	543	return 0;
ac27a0ec DK	544	}
d1f5273e	545	curr_pos = hash2pos(filp, fname->hash, fname->minor_hash);
ac27a0ec DK	546	while (fname) {
	547	error = filldir(dirent, fname->name,
	548	fname->name_len, curr_pos,
	549	fname->inode,
	550	get_dtype(sb, fname->file_type));
	551	if (error) {
	552	filp->f_pos = curr_pos;
d0156417	553	info->extra_fname = fname;
ac27a0ec DK	554	return error;
	555	}
	556	fname = fname->next;
	557	}
	558	return 0;
	559	}
	560
af5bc92d TT	561	static int ext4_dx_readdir(struct file *filp,
af5bc92d TT	562	void *dirent, filldir_t filldir)
ac27a0ec DK	563	{
ac27a0ec DK	564	struct dir_private_info *info = filp->private_data;
9d549890	565	struct inode *inode = filp->f_path.dentry->d_inode;
ac27a0ec DK	566	struct fname *fname;
	567	int ret;
	568
	569	if (!info) {
d1f5273e	570	info = ext4_htree_create_dir_info(filp, filp->f_pos);
ac27a0ec DK	571	if (!info)
	572	return -ENOMEM;
	573	filp->private_data = info;
	574	}
	575
d1f5273e	576	if (filp->f_pos == ext4_get_htree_eof(filp))
ac27a0ec DK	577	return 0; /* EOF */
	578
	579	/* Some one has messed with f_pos; reset the world */
	580	if (info->last_pos != filp->f_pos) {
	581	free_rb_tree_fname(&info->root);
	582	info->curr_node = NULL;
	583	info->extra_fname = NULL;
d1f5273e FY	584	info->curr_hash = pos2maj_hash(filp, filp->f_pos);
d1f5273e FY	585	info->curr_minor_hash = pos2min_hash(filp, filp->f_pos);
ac27a0ec DK	586	}
	587
	588	/*
	589	* If there are any leftover names on the hash collision
	590	* chain, return them first.
	591	*/
d0156417 TT	592	if (info->extra_fname) {
	593	if (call_filldir(filp, dirent, filldir, info->extra_fname))
	594	goto finished;
d0156417	595	info->extra_fname = NULL;
3c37fc86	596	goto next_node;
d0156417	597	} else if (!info->curr_node)
ac27a0ec DK	598	info->curr_node = rb_first(&info->root);
	599
	600	while (1) {
	601	/*
	602	* Fill the rbtree if we have no more entries,
	603	* or the inode has changed since we last read in the
	604	* cached entries.
	605	*/
	606	if ((!info->curr_node) \|\|
	607	(filp->f_version != inode->i_version)) {
	608	info->curr_node = NULL;
	609	free_rb_tree_fname(&info->root);
	610	filp->f_version = inode->i_version;
617ba13b	611	ret = ext4_htree_fill_tree(filp, info->curr_hash,
ac27a0ec DK	612	info->curr_minor_hash,
	613	&info->next_hash);
	614	if (ret < 0)
	615	return ret;
	616	if (ret == 0) {
d1f5273e	617	filp->f_pos = ext4_get_htree_eof(filp);
ac27a0ec DK	618	break;
	619	}
	620	info->curr_node = rb_first(&info->root);
	621	}
	622
	623	fname = rb_entry(info->curr_node, struct fname, rb_hash);
	624	info->curr_hash = fname->hash;
	625	info->curr_minor_hash = fname->minor_hash;
	626	if (call_filldir(filp, dirent, filldir, fname))
	627	break;
3c37fc86	628	next_node:
ac27a0ec	629	info->curr_node = rb_next(info->curr_node);
3c37fc86 TT	630	if (info->curr_node) {
	631	fname = rb_entry(info->curr_node, struct fname,
	632	rb_hash);
	633	info->curr_hash = fname->hash;
	634	info->curr_minor_hash = fname->minor_hash;
	635	} else {
ac27a0ec	636	if (info->next_hash == ~0) {
d1f5273e	637	filp->f_pos = ext4_get_htree_eof(filp);
ac27a0ec DK	638	break;
	639	}
	640	info->curr_hash = info->next_hash;
	641	info->curr_minor_hash = 0;
	642	}
	643	}
	644	finished:
	645	info->last_pos = filp->f_pos;
	646	return 0;
	647	}
	648
af5bc92d	649	static int ext4_release_dir(struct inode inode, struct file filp)
ac27a0ec	650	{
63f57933	651	if (filp->private_data)
617ba13b	652	ext4_htree_free_dir_info(filp->private_data);
ac27a0ec DK	653
	654	return 0;
	655	}
d1f5273e FY	656
	657	const struct file_operations ext4_dir_operations = {
	658	.llseek = ext4_dir_llseek,
	659	.read = generic_read_dir,
	660	.readdir = ext4_readdir,
	661	.unlocked_ioctl = ext4_ioctl,
	662	#ifdef CONFIG_COMPAT
	663	.compat_ioctl = ext4_compat_ioctl,
	664	#endif
	665	.fsync = ext4_sync_file,
	666	.release = ext4_release_dir,
	667	};