Re: [V2] btrfs: drop inode reference count on error path

[Nikolay Borisov <nborisov@suse.com>]

On 18.04.19 г. 14:06 ч., Pan Bian wrote:
> The reference count of inode is incremented by ihold. It should be
> dropped if not used. However, the reference count is not dropped if
> error occurs during updating the inode or deleting orphan items. This
> patch fixes the bug.
> 
> Signed-off-by: Pan Bian <bianpan2016@163.com>
> ---
> V2: move ihold just before device_initialize to make code clearer
> ---
>  fs/btrfs/inode.c | 54 +++++++++++++++++++++++++-----------------------------
>  1 file changed, 25 insertions(+), 29 deletions(-)
> 
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 82fdda8..d6630df 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -6579,7 +6579,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
>  	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
>  	u64 index;
>  	int err;
> -	int drop_inode = 0;
> +	int log_mode;
>  
>  	/* do not allow sys_link's with other subvols of the same device */
>  	if (root->root_key.objectid != BTRFS_I(inode)->root->root_key.objectid)
> @@ -6616,41 +6616,37 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
>  	err = btrfs_add_nondir(trans, BTRFS_I(dir), dentry, BTRFS_I(inode),
>  			1, index);
>  
> -	if (err) {
> -		drop_inode = 1;
> -	} else {
> -		struct dentry *parent = dentry->d_parent;
> -		int ret;
> +	if (err)
> +		goto err_link;
>  
> -		err = btrfs_update_inode(trans, root, inode);
> +	err = btrfs_update_inode(trans, root, inode);
> +	if (err)
> +		goto err_link;
> +	if (inode->i_nlink == 1) {
> +		/*
> +		 * If new hard link count is 1, it's a file created
> +		 * with open(2) O_TMPFILE flag.
> +		 */
> +		err = btrfs_orphan_del(trans, BTRFS_I(inode));
>  		if (err)
> -			goto fail;
> -		if (inode->i_nlink == 1) {
> -			/*
> -			 * If new hard link count is 1, it's a file created
> -			 * with open(2) O_TMPFILE flag.
> -			 */
> -			err = btrfs_orphan_del(trans, BTRFS_I(inode));
> -			if (err)
> -				goto fail;
> -		}
> -		BTRFS_I(inode)->last_link_trans = trans->transid;
> -		d_instantiate(dentry, inode);
> -		ret = btrfs_log_new_name(trans, BTRFS_I(inode), NULL, parent,
> -					 true, NULL);
> -		if (ret == BTRFS_NEED_TRANS_COMMIT) {
> -			err = btrfs_commit_transaction(trans);
> -			trans = NULL;
> -		}
> +			goto err_link;
> +	}
> +	BTRFS_I(inode)->last_link_trans = trans->transid;
> +	ihold(inode);
> +	d_instantiate(dentry, inode);
> +	log_mode = btrfs_log_new_name(trans, BTRFS_I(inode), NULL,
> +			dentry->d_parent, true, NULL);
> +	if (log_mode == BTRFS_NEED_TRANS_COMMIT) {
> +		err = btrfs_commit_transaction(trans);
> +		trans = NULL;
>  	}
>  
> +err_link:
> +	if (err)
> +		inode_dec_link_count(inode);

Any particular reason why you moved this before ending the transaction?
It potentially has an effect during tyransaction commit since doing
inode_dec_link_count does mark_inode_dirty which moves the inode on the
dirty list. Have you explicitly thought about that ? Note, I'm not
saying it's wrong but I want the rationale for the code move.

>  fail:
>  	if (trans)
>  		btrfs_end_transaction(trans);
> -	if (drop_inode) {
> -		inode_dec_link_count(inode);
> -		iput(inode);
> -	}
>  	btrfs_btree_balance_dirty(fs_info);
>  	return err;
>  }
>