Re: [PATCH 2/2] xfs_refcount: Preventing integer overflow

[Darrick J. Wong]

On Sat, Mar 23, 2024 at 09:26:03AM +0300, Andrey Shumilin wrote:
> Multiplying variables can overflow the "overhead" variable.
> To fix this, the variable type has been increased.
> Next, a subtraction operation occurs with it,
> but before that it is checked.

Under what circumstances will pre-multiplication @overhead have a large
enough value to overflow?  The blocksize cannot be larger than 2^16, and
full splits of three btrees should never require anywhere close to 2^16
blocks, right?  Did your analysis tool find a scenario where this
actually happens?

--D

> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> Signed-off-by: Andrey Shumilin <shum.sdl@xxxxxxxx>
> ---
>  fs/xfs/libxfs/xfs_refcount.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_refcount.c b/fs/xfs/libxfs/xfs_refcount.c
> index 511c912d515c..cbf07552eaff 100644
> --- a/fs/xfs/libxfs/xfs_refcount.c
> +++ b/fs/xfs/libxfs/xfs_refcount.c
> @@ -1070,7 +1070,7 @@ static bool
>  xfs_refcount_still_have_space(
>  	struct xfs_btree_cur		*cur)
>  {
> -	unsigned long			overhead;
> +	unsigned long long		overhead;
>  
>  	/*
>  	 * Worst case estimate: full splits of the free space and rmap btrees
> -- 
> 2.30.2
> 
>