Re: [PATCH 3/7] fs, xfs: convert xfs_buf_log_item.bli_refcount from atomic_t to refcount_t

[Peter Zijlstra]

On Tue, Feb 21, 2017 at 05:49:03PM +0200, Elena Reshetova wrote:
> refcount_t type and corresponding API should be
> used instead of atomic_t when the variable is used as
> a reference counter. This allows to avoid accidental
> refcounter overflows that might lead to use-after-free
> situations.

Changelog forgets to mention if this was runtime tested..


> @@ -371,7 +371,7 @@ xfs_trans_brelse(xfs_trans_t	*tp,
>  	ASSERT(bip->bli_item.li_type == XFS_LI_BUF);
>  	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
>  	ASSERT(!(bip->__bli_format.blf_flags & XFS_BLF_CANCEL));
> -	ASSERT(atomic_read(&bip->bli_refcount) > 0);
> +	ASSERT(refcount_read(&bip->bli_refcount) > 0);
>  
>  	trace_xfs_trans_brelse(bip);
>  
> @@ -419,7 +419,7 @@ xfs_trans_brelse(xfs_trans_t	*tp,
>  	/*
>  	 * Drop our reference to the buf log item.
>  	 */
> -	atomic_dec(&bip->bli_refcount);
> +	refcount_dec(&bip->bli_refcount);
>  
>  	/*
>  	 * If the buf item is not tracking data in the log, then
> @@ -432,7 +432,7 @@ xfs_trans_brelse(xfs_trans_t	*tp,
>  /***
>  		ASSERT(bp->b_pincount == 0);
>  ***/
> -		ASSERT(atomic_read(&bip->bli_refcount) == 0);
> +		ASSERT(refcount_read(&bip->bli_refcount) == 0);
>  		ASSERT(!(bip->bli_item.li_flags & XFS_LI_IN_AIL));
>  		ASSERT(!(bip->bli_flags & XFS_BLI_INODE_ALLOC_BUF));
>  		xfs_buf_item_relse(bp);


This for example looks dodgy.

That seems to suggest the atomic_dec() there can actually hit 0, which
_will_ generate a WARN.