dir.c

linux 内核源代码

 * We allocate enough clusters to fulfill "blocks_wanted", but set
 * i_size to exactly one block. Ocfs2_extend_dir() will handle the
 * rest automatically for us.
 *
 * *first_block_bh is a pointer to the 1st data block allocated to the
 *  directory.
 */
static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
				   unsigned int blocks_wanted,
				   struct buffer_head **first_block_bh)
{
	int ret, credits = OCFS2_INLINE_TO_EXTENTS_CREDITS;
	u32 alloc, bit_off, len;
	struct super_block *sb = dir->i_sb;
	u64 blkno, bytes = blocks_wanted << sb->s_blocksize_bits;
	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
	struct ocfs2_inode_info *oi = OCFS2_I(dir);
	struct ocfs2_alloc_context *data_ac;
	struct buffer_head *dirdata_bh = NULL;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	handle_t *handle;

	alloc = ocfs2_clusters_for_bytes(sb, bytes);

	/*
	 * We should never need more than 2 clusters for this -
	 * maximum dirent size is far less than one block. In fact,
	 * the only time we'd need more than one cluster is if
	 * blocksize == clustersize and the dirent won't fit in the
	 * extra space that the expansion to a single block gives. As
	 * of today, that only happens on 4k/4k file systems.
	 */
	BUG_ON(alloc > 2);

	ret = ocfs2_reserve_clusters(osb, alloc, &data_ac);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	down_write(&oi->ip_alloc_sem);

	/*
	 * Prepare for worst case allocation scenario of two seperate
	 * extents.
	 */
	if (alloc == 2)
		credits += OCFS2_SUBALLOC_ALLOC;

	handle = ocfs2_start_trans(osb, credits);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out_sem;
	}

	/*
	 * Try to claim as many clusters as the bitmap can give though
	 * if we only get one now, that's enough to continue. The rest
	 * will be claimed after the conversion to extents.
	 */
	ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, &len);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	/*
	 * Operations are carefully ordered so that we set up the new
	 * data block first. The conversion from inline data to
	 * extents follows.
	 */
	blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off);
	dirdata_bh = sb_getblk(sb, blkno);
	if (!dirdata_bh) {
		ret = -EIO;
		mlog_errno(ret);
		goto out_commit;
	}

	ocfs2_set_new_buffer_uptodate(dir, dirdata_bh);

	ret = ocfs2_journal_access(handle, dir, dirdata_bh,
				   OCFS2_JOURNAL_ACCESS_CREATE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	memcpy(dirdata_bh->b_data, di->id2.i_data.id_data, i_size_read(dir));
	memset(dirdata_bh->b_data + i_size_read(dir), 0,
	       sb->s_blocksize - i_size_read(dir));
	ocfs2_expand_last_dirent(dirdata_bh->b_data, i_size_read(dir),
				 sb->s_blocksize);

	ret = ocfs2_journal_dirty(handle, dirdata_bh);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	/*
	 * Set extent, i_size, etc on the directory. After this, the
	 * inode should contain the same exact dirents as before and
	 * be fully accessible from system calls.
	 *
	 * We let the later dirent insert modify c/mtime - to the user
	 * the data hasn't changed.
	 */
	ret = ocfs2_journal_access(handle, dir, di_bh,
				   OCFS2_JOURNAL_ACCESS_CREATE);
	if (ret) {
		mlog_errno(ret);
		goto out_commit;
	}

	spin_lock(&oi->ip_lock);
	oi->ip_dyn_features &= ~OCFS2_INLINE_DATA_FL;
	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
	spin_unlock(&oi->ip_lock);

	ocfs2_dinode_new_extent_list(dir, di);

	i_size_write(dir, sb->s_blocksize);
	dir->i_mtime = dir->i_ctime = CURRENT_TIME;

	di->i_size = cpu_to_le64(sb->s_blocksize);
	di->i_ctime = di->i_mtime = cpu_to_le64(dir->i_ctime.tv_sec);
	di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(dir->i_ctime.tv_nsec);
	dir->i_blocks = ocfs2_inode_sector_count(dir);

	/*
	 * This should never fail as our extent list is empty and all
	 * related blocks have been journaled already.
	 */
	ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 0, blkno, len, 0,
				  NULL);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

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

	/*
	 * We asked for two clusters, but only got one in the 1st
	 * pass. Claim the 2nd cluster as a separate extent.
	 */
	if (alloc > len) {
		ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
					   &len);
		if (ret) {
			mlog_errno(ret);
			goto out_commit;
		}
		blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off);

		ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 1, blkno,
					  len, 0, NULL);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
	}

	*first_block_bh = dirdata_bh;
	dirdata_bh = NULL;

out_commit:
	ocfs2_commit_trans(osb, handle);

out_sem:
	up_write(&oi->ip_alloc_sem);

out:
	if (data_ac)
		ocfs2_free_alloc_context(data_ac);

	brelse(dirdata_bh);

	return ret;
}

/* returns a bh of the 1st new block in the allocation. */
static int ocfs2_do_extend_dir(struct super_block *sb,
			       handle_t *handle,
			       struct inode *dir,
			       struct buffer_head *parent_fe_bh,
			       struct ocfs2_alloc_context *data_ac,
			       struct ocfs2_alloc_context *meta_ac,
			       struct buffer_head **new_bh)
{
	int status;
	int extend;
	u64 p_blkno, v_blkno;

	spin_lock(&OCFS2_I(dir)->ip_lock);
	extend = (i_size_read(dir) == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters));
	spin_unlock(&OCFS2_I(dir)->ip_lock);

	if (extend) {
		u32 offset = OCFS2_I(dir)->ip_clusters;

		status = ocfs2_do_extend_allocation(OCFS2_SB(sb), dir, &offset,
						    1, 0, parent_fe_bh, handle,
						    data_ac, meta_ac, NULL);
		BUG_ON(status == -EAGAIN);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	}

	v_blkno = ocfs2_blocks_for_bytes(sb, i_size_read(dir));
	status = ocfs2_extent_map_get_blocks(dir, v_blkno, &p_blkno, NULL, NULL);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	*new_bh = sb_getblk(sb, p_blkno);
	if (!*new_bh) {
		status = -EIO;
		mlog_errno(status);
		goto bail;
	}
	status = 0;
bail:
	mlog_exit(status);
	return status;
}

/*
 * Assumes you already have a cluster lock on the directory.
 *
 * 'blocks_wanted' is only used if we have an inline directory which
 * is to be turned into an extent based one. The size of the dirent to
 * insert might be larger than the space gained by growing to just one
 * block, so we may have to grow the inode by two blocks in that case.
 */
static int ocfs2_extend_dir(struct ocfs2_super *osb,
			    struct inode *dir,
			    struct buffer_head *parent_fe_bh,
			    unsigned int blocks_wanted,
			    struct buffer_head **new_de_bh)
{
	int status = 0;
	int credits, num_free_extents, drop_alloc_sem = 0;
	loff_t dir_i_size;
	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
	struct ocfs2_alloc_context *data_ac = NULL;
	struct ocfs2_alloc_context *meta_ac = NULL;
	handle_t *handle = NULL;
	struct buffer_head *new_bh = NULL;
	struct ocfs2_dir_entry * de;
	struct super_block *sb = osb->sb;

	mlog_entry_void();

	if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
		status = ocfs2_expand_inline_dir(dir, parent_fe_bh,
						 blocks_wanted, &new_bh);
		if (status) {
			mlog_errno(status);
			goto bail;
		}

		if (blocks_wanted == 1) {
			/*
			 * If the new dirent will fit inside the space
			 * created by pushing out to one block, then
			 * we can complete the operation
			 * here. Otherwise we have to expand i_size
			 * and format the 2nd block below.
			 */
			BUG_ON(new_bh == NULL);
			goto bail_bh;
		}

		/*
		 * Get rid of 'new_bh' - we want to format the 2nd
		 * data block and return that instead.
		 */
		brelse(new_bh);
		new_bh = NULL;

		dir_i_size = i_size_read(dir);
		credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS;
		goto do_extend;
	}

	dir_i_size = i_size_read(dir);
	mlog(0, "extending dir %llu (i_size = %lld)\n",
	     (unsigned long long)OCFS2_I(dir)->ip_blkno, dir_i_size);

	/* dir->i_size is always block aligned. */
	spin_lock(&OCFS2_I(dir)->ip_lock);
	if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) {
		spin_unlock(&OCFS2_I(dir)->ip_lock);
		num_free_extents = ocfs2_num_free_extents(osb, dir, fe);
		if (num_free_extents < 0) {
			status = num_free_extents;
			mlog_errno(status);
			goto bail;
		}

		if (!num_free_extents) {
			status = ocfs2_reserve_new_metadata(osb, fe, &meta_ac);
			if (status < 0) {
				if (status != -ENOSPC)
					mlog_errno(status);
				goto bail;
			}
		}

		status = ocfs2_reserve_clusters(osb, 1, &data_ac);
		if (status < 0) {
			if (status != -ENOSPC)
				mlog_errno(status);
			goto bail;
		}

		credits = ocfs2_calc_extend_credits(sb, fe, 1);
	} else {
		spin_unlock(&OCFS2_I(dir)->ip_lock);
		credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS;
	}

do_extend:
	down_write(&OCFS2_I(dir)->ip_alloc_sem);
	drop_alloc_sem = 1;

	handle = ocfs2_start_trans(osb, credits);
	if (IS_ERR(handle)) {
		status = PTR_ERR(handle);
		handle = NULL;
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_do_extend_dir(osb->sb, handle, dir, parent_fe_bh,
				     data_ac, meta_ac, &new_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	ocfs2_set_new_buffer_uptodate(dir, new_bh);

	status = ocfs2_journal_access(handle, dir, new_bh,
				      OCFS2_JOURNAL_ACCESS_CREATE);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	memset(new_bh->b_data, 0, sb->s_blocksize);
	de = (struct ocfs2_dir_entry *) new_bh->b_data;
	de->inode = 0;
	de->rec_len = cpu_to_le16(sb->s_blocksize);
	status = ocfs2_journal_dirty(handle, new_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	dir_i_size += dir->i_sb->s_blocksize;
	i_size_write(dir, dir_i_size);
	dir->i_blocks = ocfs2_inode_sector_count(dir);
	status = ocfs2_mark_inode_dirty(handle, dir, parent_fe_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

bail_bh:
	*new_de_bh = new_bh;
	get_bh(*new_de_bh);
bail:
	if (drop_alloc_sem)
		up_write(&OCFS2_I(dir)->ip_alloc_sem);
	if (handle)
		ocfs2_commit_trans(osb, handle);

	if (data_ac)
		ocfs2_free_alloc_context(data_ac);
	if (meta_ac)
		ocfs2_free_alloc_context(meta_ac);

	if (new_bh)
		brelse(new_bh);

	mlog_exit(status);
	return status;
}

static int ocfs2_find_dir_space_id(struct inode *dir, struct buffer_head *di_bh,
				   const char *name, int namelen,
				   struct buffer_head **ret_de_bh,
				   unsigned int *blocks_wanted)
{
	int ret;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
	struct ocfs2_dir_entry *de, *last_de = NULL;
	char *de_buf, *limit;
	unsigned long offset = 0;
	unsigned int rec_len, new_rec_len;

	de_buf = di->id2.i_data.id_data;
	limit = de_buf + i_size_read(dir);
	rec_len = OCFS2_DIR_REC_LEN(namelen);

	while (de_buf < limit) {
		de = (struct ocfs2_dir_entry *)de_buf;

		if (!ocfs2_check_dir_entry(dir, de, di_bh, offset)) {
			ret = -ENOENT;
			goto out;
		}
		if (ocfs2_match(namelen, name, de)) {
			ret = -EEXIST;
			goto out;
		}
		if (ocfs2_dirent_would_fit(de, rec_len)) {
			/* Ok, we found a spot. Return this bh and let
			 * the caller actually fill it in. */
			*ret_de_bh = di_bh;
			get_bh(*ret_de_bh);
			ret = 0;
			goto out;
		}

		last_de = de;
		de_buf += le16_to_cpu(de->rec_len);
		offset += le16_to_cpu(de->rec_len);
	}

	/*
	 * We're going to require expansion of the directory - figure
	 * out how many blocks we'll need so that a place for the
	 * dirent can be found.
	 */
	*blocks_wanted = 1;
	new_rec_len = le16_to_cpu(last_de->rec_len) + (dir->i_sb->s_blocksize - i_size_read(dir));
	if (new_rec_len < (rec_len + OCFS2_DIR_REC_LEN(last_de->name_len)))
		*blocks_wanted = 2;

	ret = -ENOSPC;
out:
	return ret;
}

static int ocfs2_find_dir_space_el(struct inode *dir, const char *name,
				   int namelen, struct buffer_head **ret_de_bh)
{
	unsigned long offset;
	struct buffer_head *bh = NULL;
	unsigned short rec_len;
	struct ocfs2_dir_entry *de;
	struct super_block *sb = dir->i_sb;
	int status;

	bh = ocfs2_bread(dir, 0, &status, 0);
	if (!bh) {
		mlog_errno(status);
		goto bail;
	}

	rec_len = OCFS2_DIR_REC_LEN(namelen);
	offset = 0;
	de = (struct ocfs2_dir_entry *) bh->b_data;
	while (1) {
		if ((char *)de >= sb->s_blocksize + bh->b_data) {
			brelse(bh);
			bh = NULL;

			if (i_size_read(dir) <= offset) {
				/*
				 * Caller will have to expand this
				 * directory.
				 */
				status = -ENOSPC;
				goto bail;
			}
			bh = ocfs2_bread(dir,
					 offset >> sb->s_blocksize_bits,
					 &status,
					 0);
			if (!bh) {
				mlog_errno(status);
				goto bail;
			}
			/* move to next block */
			de = (struct ocfs2_dir_entry *) bh->b_data;
		}
		if (!ocfs2_check_dir_entry(dir, de, bh, offset)) {
			status = -ENOENT;
			goto bail;
		}
		if (ocfs2_match(namelen, name, de)) {
			status = -EEXIST;
			goto bail;
		}
		if (ocfs2_dirent_would_fit(de, rec_len)) {
			/* Ok, we found a spot. Return this bh and let
			 * the caller actually fill it in. */
			*ret_de_bh = bh;
			get_bh(*ret_de_bh);
			status = 0;
			goto bail;
		}
		offset += le16_to_cpu(de->rec_len);
		de = (struct ocfs2_dir_entry *)((char *) de + le16_to_cpu(de->rec_len));
	}

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

	mlog_exit(status);
	return status;
}

int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb,
				 struct inode *dir,
				 struct buffer_head *parent_fe_bh,
				 const char *name,
				 int namelen,
				 struct buffer_head **ret_de_bh)
{
	int ret;
	unsigned int blocks_wanted = 1;
	struct buffer_head *bh = NULL;

	mlog(0, "getting ready to insert namelen %d into dir %llu\n",
	     namelen, (unsigned long long)OCFS2_I(dir)->ip_blkno);

	*ret_de_bh = NULL;

	if (!namelen) {
		ret = -EINVAL;
		mlog_errno(ret);
		goto out;
	}

	if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
		ret = ocfs2_find_dir_space_id(dir, parent_fe_bh, name,
					      namelen, &bh, &blocks_wanted);
	} else
		ret = ocfs2_find_dir_space_el(dir, name, namelen, &bh);

	if (ret && ret != -ENOSPC) {
		mlog_errno(ret);
		goto out;
	}

	if (ret == -ENOSPC) {
		/*
		 * We have to expand the directory to add this name.
		 */
		BUG_ON(bh);

		ret = ocfs2_extend_dir(osb, dir, parent_fe_bh, blocks_wanted,
				       &bh);
		if (ret) {
			if (ret != -ENOSPC)
				mlog_errno(ret);
			goto out;
		}

		BUG_ON(!bh);
	}

	*ret_de_bh = bh;
	bh = NULL;
out:
	if (bh)
		brelse(bh);
	return ret;
}