[PATCH net-next v5 2/3] net: ti: icssm-prueth: Adds switchdev support for icssm_prueth driver

Simon Horman horms at kernel.org
Mon Nov 17 13:08:29 PST 2025


On Thu, Nov 13, 2025 at 03:40:22PM +0530, Parvathi Pudi wrote:

...

> @@ -222,12 +229,14 @@ struct prueth_emac {
>  	const char *phy_id;
>  	u32 msg_enable;
>  	u8 mac_addr[6];
> +	unsigned char mc_filter_mask[ETH_ALEN]; /* for multicast filtering */
>  	phy_interface_t phy_if;
>  
>  	/* spin lock used to protect
>  	 * during link configuration
>  	 */
>  	spinlock_t lock;
> +	spinlock_t addr_lock;   /* serialize access to VLAN/MC filter table */

addr_lock does not appear to be initialised anywhere.

...

> +static int icssm_prueth_switchdev_obj_del(struct net_device *ndev,
> +					  const void *ctx,
> +					  const struct switchdev_obj *obj)
> +{
> +	struct switchdev_obj_port_mdb *mdb = SWITCHDEV_OBJ_PORT_MDB(obj);
> +	struct prueth_emac *emac = netdev_priv(ndev);
> +	struct prueth *prueth = emac->prueth;
> +	struct netdev_hw_addr *ha;
> +	u8 hash, tmp_hash;
> +	int ret = 0;
> +
> +	switch (obj->id) {
> +	case SWITCHDEV_OBJ_ID_HOST_MDB:
> +		dev_dbg(prueth->dev, "MDB del: %s: vid %u:%pM  port: %x\n",
> +			ndev->name, mdb->vid, mdb->addr, emac->port_id);
> +		hash = icssm_emac_get_mc_hash(mdb->addr, emac->mc_filter_mask);
> +		netdev_for_each_mc_addr(ha, prueth->hw_bridge_dev) {

Is there anything stopping this event from occurring when
the port is not the lower device of a bridge - before being added
or after being removed?

If not, then passing prueth->hw_bridge_dev to netdev_for_each_mc_addr()
will result in a null pointer dereference.

...



More information about the linux-arm-kernel mailing list