Linux Ethernet Bridging - May 2021 - [Bridge] [RFC net-next 4/9] net: bridge: switchdev: Forward offloading

On Mon, Apr 26, 2021 at 07:04:06PM +0200, Tobias Waldekranz
wrote:
> +static void nbp_switchdev_fwd_offload_add(struct net_bridge_port *p)
> +{
> +	void *priv;
> +
> +	if (!(p->dev->features & NETIF_F_HW_L2FW_DOFFLOAD))
> +		return;
> +
> +	priv = p->dev->netdev_ops->ndo_dfwd_add_station(p->dev,
p->br->dev);
Some changes to team/bond/8021q will be needed in order to get this
optimization to work when they are enslaved to the bridge instead of the
front panel port itself?
> +	if (!IS_ERR_OR_NULL(priv))
> +		p->accel_priv = priv;
> +}
> +
> +static void nbp_switchdev_fwd_offload_del(struct net_bridge_port *p)
> +{
> +	if (!p->accel_priv)
> +		return;
> +
> +	p->dev->netdev_ops->ndo_dfwd_del_station(p->dev,
p->accel_priv);
> +	p->accel_priv = NULL;
> +}

On Sun, May 02, 2021 at 18:04, Ido Schimmel <idosch at idosch.org>
wrote:
> On Mon, Apr 26, 2021 at 07:04:06PM +0200, Tobias Waldekranz wrote:
>> +static void nbp_switchdev_fwd_offload_add(struct net_bridge_port *p)
>> +{
>> +	void *priv;
>> +
>> +	if (!(p->dev->features & NETIF_F_HW_L2FW_DOFFLOAD))
>> +		return;
>> +
>> +	priv = p->dev->netdev_ops->ndo_dfwd_add_station(p->dev,
p->br->dev);
>
> Some changes to team/bond/8021q will be needed in order to get this
> optimization to work when they are enslaved to the bridge instead of the
> front panel port itself?
Right you are. We should probably do something similar to the
switchdev_handle_port_* family of helpers that could be reused in
stacked devices. I will look at it for v1.
>> +	if (!IS_ERR_OR_NULL(priv))
>> +		p->accel_priv = priv;
>> +}
>> +
>> +static void nbp_switchdev_fwd_offload_del(struct net_bridge_port *p)
>> +{
>> +	if (!p->accel_priv)
>> +		return;
>> +
>> +	p->dev->netdev_ops->ndo_dfwd_del_station(p->dev,
p->accel_priv);
>> +	p->accel_priv = NULL;
>> +}