Re: [PATCH] ice: Fix improper handling of refcount in ice_dpll_init_rclk_pins()

[Simon Horman <horms@kernel.org>]

On Fri, Aug 09, 2024 at 01:02:15PM +0800, Gui-Dong Han wrote:
> This patch addresses a reference count handling issue in the
> ice_dpll_init_rclk_pins() function. The function calls ice_dpll_get_pins(),
> which increments the reference count of the relevant resources. However,
> if the condition WARN_ON((!vsi || !vsi->netdev)) is met, the function
> currently returns an error without properly releasing the resources
> acquired by ice_dpll_get_pins(), leading to a reference count leak.
> 
> To resolve this, the patch introduces a goto unregister_pins; statement
> when the condition is met, ensuring that the resources are correctly
> released and the reference count is decremented before returning the error.
> This change prevents potential memory leaks and ensures proper resource
>  management within the function.
> 
> This bug was identified by an experimental static analysis tool developed
> by our team. The tool specializes in analyzing reference count operations
> and detecting potential issues where resources are not properly managed.
> In this case, the tool flagged the missing release operation as a
> potential problem, which led to the development of this patch.
> 
> Fixes: d7999f5ea64b ("ice: implement dpll interface to control cgu")
> Cc: stable@vger.kernel.org
> Signed-off-by: Gui-Dong Han <hanguidong02@outlook.com>
> ---
>  drivers/net/ethernet/intel/ice/ice_dpll.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_dpll.c b/drivers/net/ethernet/intel/ice/ice_dpll.c
> index e92be6f130a3..f3f204cae093 100644
> --- a/drivers/net/ethernet/intel/ice/ice_dpll.c
> +++ b/drivers/net/ethernet/intel/ice/ice_dpll.c
> @@ -1641,8 +1641,10 @@ ice_dpll_init_rclk_pins(struct ice_pf *pf, struct ice_dpll_pin *pin,
>  		if (ret)
>  			goto unregister_pins;
>  	}
> -	if (WARN_ON((!vsi || !vsi->netdev)))
> -		return -EINVAL;
> +	if (WARN_ON((!vsi || !vsi->netdev))) {
> +		ret = -EINVAL;
> +		goto unregister_pins;
> +	}

Hi,

I wonder if it would make sense to move the check to the
top of the function. It seems to be more of a verification
of state at the time the function is run than anything else.

Doing so would avoid the need to handle unwind in this case.

>  	dpll_netdev_pin_set(vsi->netdev, pf->dplls.rclk.pin);
>  
>  	return 0;
> -- 
> 2.25.1
> 
>