dir.c

linux 内核源代码

static int ocfs2_dir_foreach_blk_id(struct inode *inode,
				    u64 *f_version,
				    loff_t *f_pos, void *priv,
				    filldir_t filldir, int *filldir_err)
{
	int ret, i, filldir_ret;
	unsigned long offset = *f_pos;
	struct buffer_head *di_bh = NULL;
	struct ocfs2_dinode *di;
	struct ocfs2_inline_data *data;
	struct ocfs2_dir_entry *de;

	ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), OCFS2_I(inode)->ip_blkno,
			       &di_bh, OCFS2_BH_CACHED, inode);
	if (ret) {
		mlog(ML_ERROR, "Unable to read inode block for dir %llu\n",
		     (unsigned long long)OCFS2_I(inode)->ip_blkno);
		goto out;
	}

	di = (struct ocfs2_dinode *)di_bh->b_data;
	data = &di->id2.i_data;

	while (*f_pos < i_size_read(inode)) {
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 (*f_version != inode->i_version) {
			for (i = 0; i < i_size_read(inode) && i < offset; ) {
				de = (struct ocfs2_dir_entry *)
					(data->id_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) <
				    OCFS2_DIR_REC_LEN(1))
					break;
				i += le16_to_cpu(de->rec_len);
			}
			*f_pos = offset = i;
			*f_version = inode->i_version;
		}

		de = (struct ocfs2_dir_entry *) (data->id_data + *f_pos);
		if (!ocfs2_check_dir_entry(inode, de, di_bh, *f_pos)) {
			/* On error, skip the f_pos to the end. */
			*f_pos = i_size_read(inode);
			goto out;
		}
		offset += le16_to_cpu(de->rec_len);
		if (le64_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 = *f_version;
			unsigned char d_type = DT_UNKNOWN;

			if (de->file_type < OCFS2_FT_MAX)
				d_type = ocfs2_filetype_table[de->file_type];

			filldir_ret = filldir(priv, de->name,
					      de->name_len,
					      *f_pos,
					      le64_to_cpu(de->inode),
					      d_type);
			if (filldir_ret) {
				if (filldir_err)
					*filldir_err = filldir_ret;
				break;
			}
			if (version != *f_version)
				goto revalidate;
		}
		*f_pos += le16_to_cpu(de->rec_len);
	}

out:
	brelse(di_bh);

	return 0;
}

static int ocfs2_dir_foreach_blk_el(struct inode *inode,
				    u64 *f_version,
				    loff_t *f_pos, void *priv,
				    filldir_t filldir, int *filldir_err)
{
	int error = 0;
	unsigned long offset, blk, last_ra_blk = 0;
	int i, stored;
	struct buffer_head * bh, * tmp;
	struct ocfs2_dir_entry * de;
	int err;
	struct super_block * sb = inode->i_sb;
	unsigned int ra_sectors = 16;

	stored = 0;
	bh = NULL;

	offset = (*f_pos) & (sb->s_blocksize - 1);

	while (!error && !stored && *f_pos < i_size_read(inode)) {
		blk = (*f_pos) >> sb->s_blocksize_bits;
		bh = ocfs2_bread(inode, blk, &err, 0);
		if (!bh) {
			mlog(ML_ERROR,
			     "directory #%llu contains a hole at offset %lld\n",
			     (unsigned long long)OCFS2_I(inode)->ip_blkno,
			     *f_pos);
			*f_pos += sb->s_blocksize - offset;
			continue;
		}

		/* The idea here is to begin with 8k read-ahead and to stay
		 * 4k ahead of our current position.
		 *
		 * TODO: Use the pagecache for this. We just need to
		 * make sure it's cluster-safe... */
		if (!last_ra_blk
		    || (((last_ra_blk - blk) << 9) <= (ra_sectors / 2))) {
			for (i = ra_sectors >> (sb->s_blocksize_bits - 9);
			     i > 0; i--) {
				tmp = ocfs2_bread(inode, ++blk, &err, 1);
				if (tmp)
					brelse(tmp);
			}
			last_ra_blk = blk;
			ra_sectors = 8;
		}

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 (*f_version != inode->i_version) {
			for (i = 0; i < sb->s_blocksize && i < offset; ) {
				de = (struct ocfs2_dir_entry *) (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) <
				    OCFS2_DIR_REC_LEN(1))
					break;
				i += le16_to_cpu(de->rec_len);
			}
			offset = i;
			*f_pos = ((*f_pos) & ~(sb->s_blocksize - 1))
				| offset;
			*f_version = inode->i_version;
		}

		while (!error && *f_pos < i_size_read(inode)
		       && offset < sb->s_blocksize) {
			de = (struct ocfs2_dir_entry *) (bh->b_data + offset);
			if (!ocfs2_check_dir_entry(inode, de, bh, offset)) {
				/* On error, skip the f_pos to the
				   next block. */
				*f_pos = ((*f_pos) | (sb->s_blocksize - 1)) + 1;
				brelse(bh);
				goto out;
			}
			offset += le16_to_cpu(de->rec_len);
			if (le64_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 = *f_version;
				unsigned char d_type = DT_UNKNOWN;

				if (de->file_type < OCFS2_FT_MAX)
					d_type = ocfs2_filetype_table[de->file_type];
				error = filldir(priv, de->name,
						de->name_len,
						*f_pos,
						le64_to_cpu(de->inode),
						d_type);
				if (error) {
					if (filldir_err)
						*filldir_err = error;
					break;
				}
				if (version != *f_version)
					goto revalidate;
				stored ++;
			}
			*f_pos += le16_to_cpu(de->rec_len);
		}
		offset = 0;
		brelse(bh);
	}

	stored = 0;
out:
	return stored;
}

static int ocfs2_dir_foreach_blk(struct inode *inode, u64 *f_version,
				 loff_t *f_pos, void *priv, filldir_t filldir,
				 int *filldir_err)
{
	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
		return ocfs2_dir_foreach_blk_id(inode, f_version, f_pos, priv,
						filldir, filldir_err);

	return ocfs2_dir_foreach_blk_el(inode, f_version, f_pos, priv, filldir,
					filldir_err);
}

/*
 * This is intended to be called from inside other kernel functions,
 * so we fake some arguments.
 */
int ocfs2_dir_foreach(struct inode *inode, loff_t *f_pos, void *priv,
		      filldir_t filldir)
{
	int ret = 0, filldir_err = 0;
	u64 version = inode->i_version;

	while (*f_pos < i_size_read(inode)) {
		ret = ocfs2_dir_foreach_blk(inode, &version, f_pos, priv,
					    filldir, &filldir_err);
		if (ret || filldir_err)
			break;
	}

	if (ret > 0)
		ret = -EIO;

	return 0;
}

/*
 * ocfs2_readdir()
 *
 */
int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir)
{
	int error = 0;
	struct inode *inode = filp->f_path.dentry->d_inode;
	int lock_level = 0;

	mlog_entry("dirino=%llu\n",
		   (unsigned long long)OCFS2_I(inode)->ip_blkno);

	error = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level);
	if (lock_level && error >= 0) {
		/* We release EX lock which used to update atime
		 * and get PR lock again to reduce contention
		 * on commonly accessed directories. */
		ocfs2_meta_unlock(inode, 1);
		lock_level = 0;
		error = ocfs2_meta_lock(inode, NULL, 0);
	}
	if (error < 0) {
		if (error != -ENOENT)
			mlog_errno(error);
		/* we haven't got any yet, so propagate the error. */
		goto bail_nolock;
	}

	error = ocfs2_dir_foreach_blk(inode, &filp->f_version, &filp->f_pos,
				      dirent, filldir, NULL);

	ocfs2_meta_unlock(inode, lock_level);

bail_nolock:
	mlog_exit(error);

	return error;
}

/*
 * NOTE: this should always be called with parent dir i_mutex taken.
 */
int ocfs2_find_files_on_disk(const char *name,
			     int namelen,
			     u64 *blkno,
			     struct inode *inode,
			     struct buffer_head **dirent_bh,
			     struct ocfs2_dir_entry **dirent)
{
	int status = -ENOENT;

	mlog_entry("(name=%.*s, blkno=%p, inode=%p, dirent_bh=%p, dirent=%p)\n",
		   namelen, name, blkno, inode, dirent_bh, dirent);

	*dirent_bh = ocfs2_find_entry(name, namelen, inode, dirent);
	if (!*dirent_bh || !*dirent) {
		status = -ENOENT;
		goto leave;
	}

	*blkno = le64_to_cpu((*dirent)->inode);

	status = 0;
leave:
	if (status < 0) {
		*dirent = NULL;
		if (*dirent_bh) {
			brelse(*dirent_bh);
			*dirent_bh = NULL;
		}
	}

	mlog_exit(status);
	return status;
}

/*
 * Convenience function for callers which just want the block number
 * mapped to a name and don't require the full dirent info, etc.
 */
int ocfs2_lookup_ino_from_name(struct inode *dir, const char *name,
			       int namelen, u64 *blkno)
{
	int ret;
	struct buffer_head *bh = NULL;
	struct ocfs2_dir_entry *dirent = NULL;

	ret = ocfs2_find_files_on_disk(name, namelen, blkno, dir, &bh, &dirent);
	brelse(bh);

	return ret;
}

/* Check for a name within a directory.
 *
 * Return 0 if the name does not exist
 * Return -EEXIST if the directory contains the name
 *
 * Callers should have i_mutex + a cluster lock on dir
 */
int ocfs2_check_dir_for_entry(struct inode *dir,
			      const char *name,
			      int namelen)
{
	int ret;
	struct buffer_head *dirent_bh = NULL;
	struct ocfs2_dir_entry *dirent = NULL;

	mlog_entry("dir %llu, name '%.*s'\n",
		   (unsigned long long)OCFS2_I(dir)->ip_blkno, namelen, name);

	ret = -EEXIST;
	dirent_bh = ocfs2_find_entry(name, namelen, dir, &dirent);
	if (dirent_bh)
		goto bail;

	ret = 0;
bail:
	if (dirent_bh)
		brelse(dirent_bh);

	mlog_exit(ret);
	return ret;
}

struct ocfs2_empty_dir_priv {
	unsigned seen_dot;
	unsigned seen_dot_dot;
	unsigned seen_other;
};
static int ocfs2_empty_dir_filldir(void *priv, const char *name, int name_len,
				   loff_t pos, u64 ino, unsigned type)
{
	struct ocfs2_empty_dir_priv *p = priv;

	/*
	 * Check the positions of "." and ".." records to be sure
	 * they're in the correct place.
	 */
	if (name_len == 1 && !strncmp(".", name, 1) && pos == 0) {
		p->seen_dot = 1;
		return 0;
	}

	if (name_len == 2 && !strncmp("..", name, 2) &&
	    pos == OCFS2_DIR_REC_LEN(1)) {
		p->seen_dot_dot = 1;
		return 0;
	}

	p->seen_other = 1;
	return 1;
}
/*
 * routine to check that the specified directory is empty (for rmdir)
 *
 * Returns 1 if dir is empty, zero otherwise.
 */
int ocfs2_empty_dir(struct inode *inode)
{
	int ret;
	loff_t start = 0;
	struct ocfs2_empty_dir_priv priv;

	memset(&priv, 0, sizeof(priv));

	ret = ocfs2_dir_foreach(inode, &start, &priv, ocfs2_empty_dir_filldir);
	if (ret)
		mlog_errno(ret);

	if (!priv.seen_dot || !priv.seen_dot_dot) {
		mlog(ML_ERROR, "bad directory (dir #%llu) - no `.' or `..'\n",
		     (unsigned long long)OCFS2_I(inode)->ip_blkno);
		/*
		 * XXX: Is it really safe to allow an unlink to continue?
		 */
		return 1;
	}

	return !priv.seen_other;
}

static void ocfs2_fill_initial_dirents(struct inode *inode,
				       struct inode *parent,
				       char *start, unsigned int size)
{
	struct ocfs2_dir_entry *de = (struct ocfs2_dir_entry *)start;

	de->inode = cpu_to_le64(OCFS2_I(inode)->ip_blkno);
	de->name_len = 1;
	de->rec_len =
		cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len));
	strcpy(de->name, ".");
	ocfs2_set_de_type(de, S_IFDIR);

	de = (struct ocfs2_dir_entry *) ((char *)de + le16_to_cpu(de->rec_len));
	de->inode = cpu_to_le64(OCFS2_I(parent)->ip_blkno);
	de->rec_len = cpu_to_le16(size - OCFS2_DIR_REC_LEN(1));
	de->name_len = 2;
	strcpy(de->name, "..");
	ocfs2_set_de_type(de, S_IFDIR);
}

/*
 * This works together with code in ocfs2_mknod_locked() which sets
 * the inline-data flag and initializes the inline-data section.
 */
static int ocfs2_fill_new_dir_id(struct ocfs2_super *osb,
				 handle_t *handle,
				 struct inode *parent,
				 struct inode *inode,
				 struct buffer_head *di_bh)
{
	int ret;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	struct ocfs2_inline_data *data = &di->id2.i_data;
	unsigned int size = le16_to_cpu(data->id_count);

	ret = ocfs2_journal_access(handle, inode, di_bh,
				   OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ocfs2_fill_initial_dirents(inode, parent, data->id_data, size);

	ocfs2_journal_dirty(handle, di_bh);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	i_size_write(inode, size);
	inode->i_nlink = 2;
	inode->i_blocks = ocfs2_inode_sector_count(inode);

	ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
	if (ret < 0)
		mlog_errno(ret);

out:
	return ret;
}

static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb,
				 handle_t *handle,
				 struct inode *parent,
				 struct inode *inode,
				 struct buffer_head *fe_bh,
				 struct ocfs2_alloc_context *data_ac)
{
	int status;
	struct buffer_head *new_bh = NULL;

	mlog_entry_void();

	status = ocfs2_do_extend_dir(osb->sb, handle, inode, fe_bh,
				     data_ac, NULL, &new_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	ocfs2_set_new_buffer_uptodate(inode, new_bh);

	status = ocfs2_journal_access(handle, inode, new_bh,
				      OCFS2_JOURNAL_ACCESS_CREATE);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	memset(new_bh->b_data, 0, osb->sb->s_blocksize);

	ocfs2_fill_initial_dirents(inode, parent, new_bh->b_data,
				   osb->sb->s_blocksize);

	status = ocfs2_journal_dirty(handle, new_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	i_size_write(inode, inode->i_sb->s_blocksize);
	inode->i_nlink = 2;
	inode->i_blocks = ocfs2_inode_sector_count(inode);
	status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	status = 0;
bail:
	if (new_bh)
		brelse(new_bh);

	mlog_exit(status);
	return status;
}

int ocfs2_fill_new_dir(struct ocfs2_super *osb,
		       handle_t *handle,
		       struct inode *parent,
		       struct inode *inode,
		       struct buffer_head *fe_bh,
		       struct ocfs2_alloc_context *data_ac)
{
	BUG_ON(!ocfs2_supports_inline_data(osb) && data_ac == NULL);

	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
		return ocfs2_fill_new_dir_id(osb, handle, parent, inode, fe_bh);

	return ocfs2_fill_new_dir_el(osb, handle, parent, inode, fe_bh,
				     data_ac);
}

static void ocfs2_expand_last_dirent(char *start, unsigned int old_size,
				     unsigned int new_size)
{
	struct ocfs2_dir_entry *de;
	struct ocfs2_dir_entry *prev_de;
	char *de_buf, *limit;
	unsigned int bytes = new_size - old_size;

	limit = start + old_size;
	de_buf = start;
	de = (struct ocfs2_dir_entry *)de_buf;
	do {
		prev_de = de;
		de_buf += le16_to_cpu(de->rec_len);
		de = (struct ocfs2_dir_entry *)de_buf;
	} while (de_buf < limit);

	le16_add_cpu(&prev_de->rec_len, bytes);
}

/*