Re: [PATCH] mlxsw_spectrum_router: add check for return value of 'mlxsw_sp_rif_find_by_dev'

[Petr Machata <petrm@nvidia.com>]

Natalia Petrova <n.petrova@fintech.ru> writes:

> Pointer 'rif' that contains the return value of 'mlxsw_sp_rif_find_by_dev'
> is checked for NULL to avoid possible undefined behavior below caused by
> dereference in 'mlxsw_sp_rif_destroy'.
>
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Fixes: e4f3c1c17b6d ("mlxsw: spectrum_router: Implement common RIF core")
> Signed-off-by: Natalia Petrova <n.petrova@fintech.ru>
> ---
>  drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
> index 2c4443c6b964..4f41b83d7c9e 100644
> --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
> +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
> @@ -8658,7 +8658,8 @@ static int mlxsw_sp_inetaddr_bridge_event(struct mlxsw_sp *mlxsw_sp,
>  		break;
>  	case NETDEV_DOWN:
>  		rif = mlxsw_sp_rif_find_by_dev(mlxsw_sp, l3_dev);
> -		mlxsw_sp_rif_destroy(rif);
> +		if (rif)
> +			mlxsw_sp_rif_destroy(rif);
>  		break;
>  	}

I don't think this can happen. The corresponding NETDEV_UP is invoked
through address validator chain, so failures to create a RIF would be
vetoed. Furthermore the DOWN event itself is invoked by
mlxsw_sp_inetaddr_event(), where the invocation is contingent on
mlxsw_sp_rif_should_config(). That would be false if there's no RIF.