[Intel-wired-lan] [PATCH] ice: Fix possible double free in error handling path

Alexander Lobakin aleksander.lobakin at intel.com
Fri Aug 30 13:49:39 UTC 2024


From: Ma Ke <make24 at iscas.ac.cn>
Date: Fri, 30 Aug 2024 21:33:24 +0800

> When auxiliary_device_add() returns error and then calls
> auxiliary_device_uninit(), callback function adev_release
> calls kfree(iadev). We shouldn't call kfree(iadev) again
> in the error handling path. Set 'iadev' to NULL.
> 
> Cc: stable at vger.kernel.org

You need to add a comment with the first kernel version which has this
bug. E.g.

Cc: stable at vger.kernel.org # 9.9+

(replace 9.9 with the actual version)

BTW usually Cc: stable goes *after* "Fixes:".

> Fixes: f9f5301e7e2d ("ice: Register auxiliary device to provide RDMA")
> Signed-off-by: Ma Ke <make24 at iscas.ac.cn>
> ---
>  drivers/net/ethernet/intel/ice/ice_idc.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
> index 145b27f2a4ce..5db05f54a80c 100644
> --- a/drivers/net/ethernet/intel/ice/ice_idc.c
> +++ b/drivers/net/ethernet/intel/ice/ice_idc.c
> @@ -330,6 +330,7 @@ int ice_plug_aux_dev(struct ice_pf *pf)
>  		return ret;
>  	}
>  
> +	iadev = NULL;
>  	ret = auxiliary_device_add(adev);
>  	if (ret) {
>  		auxiliary_device_uninit(adev);

I don't understand this patch.
I checked net-next, tnguy/net-queue, tnguy/next-queue and not in a
single branch @iadev is used after auxiliary_device_add().
IOW you just introduce a dead store which will be removed by the
compiler. The kfree() above that you mentioned is called only once and
never after auxiliary_device_uninit().
Can you draw a call ladder with the related code pieces how I can
reproduce the issue?

Thanks,
Olek


More information about the Intel-wired-lan mailing list