Re: [PATCH] evl/net: rx filter precedence over loopback

Brandon Ho <[email protected]> writes:

> Signed-off-by: Brandon Ho <[email protected]>
> ---
>  include/evl/net/device.h | 17 +++++++++--------
>  1 file changed, 9 insertions(+), 8 deletions(-)
>
> diff --git a/include/evl/net/device.h b/include/evl/net/device.h
> index d0085b3c8..b8ed44377 100644
> --- a/include/evl/net/device.h
> +++ b/include/evl/net/device.h
> @@ -53,18 +53,19 @@ evl_net_filter_rx(struct net_device *dev, struct sk_buff *skb)
>  	struct evl_netdev_state *est = dev->oob_state.estate;
>  
>  	/*
> -	 * Unconditionally accept all traffic sent to oob-enabled
> -	 * loopback devices from the oob stage. Those devices are very
> -	 * unlikely to support VLANs, but we could still use a filter
> -	 * to accept in-band traffic, so keep on checking for the
> -	 * latter.
> +	 * RX filter takes precedence over loopback acceptance, because
> +	 * it may be used to route packets to in-band consumers.
> +	*/
> +	if (test_bit(EVL_NETDEV_RX_FILTER_BIT, &est->flags))
> +		return __evl_net_filter_rx(est, skb);
> +
> +	/*
> +	 * Accept all traffic sent to oob-enabled loopback devices 
> +	 * from the oob stage.
>  	 */
>  	if (dev->flags & IFF_LOOPBACK && running_oob())
>  		return EVL_RX_ACCEPT;
>  
> -	if (test_bit(EVL_NETDEV_RX_FILTER_BIT, &est->flags))
> -		return __evl_net_filter_rx(est, skb);
> -
>  	return EVL_RX_VLAN;
>  }

Thanks for contributing, appreciated. The change above is superseded by
the following (recent) commit:

https://gitlab.com/Xenomai/xenomai4/linux-evl/-/commit/0bf66e53dcf91da8b3a8c3283cb78ff17712c670

-- 
Philippe.