On 25.11.2025 15:58, Paolo Abeni wrote:

On 11/20/25 6:49 PM, Dmitry Skorodumov wrote:

quoted

+static inline void ipvlan_mark_skb(struct sk_buff *skb, struct net_device *dev)
+{
+	IPVL_SKB_CB(skb)->mark = dev;

+	if (ipvlan_is_skb_marked(skb, port->dev))
... and this is the receiver path. Nothing guaratees the CB is preserved
in between and that random data in there will not confuse the above
check. Also I'm not sure what you are trying to filter out here.

I need to recheck.. unfortunately I didn't wrote enough comments here and now unsure about some details..

quoted

@@ -597,6 +690,9 @@ int ipvlan_link_new(struct net_device *dev, struct rtnl_newlink_params *params,
 	port = ipvlan_port_get_rtnl(phy_dev);
 	ipvlan->port = port;
 
+	if (data && data[IFLA_IPVLAN_FLAGS])
+		port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);

This looks like a change of behavior that could potentially break the
user-space.

quoted

+
 	/* If the port-id base is at the MAX value, then wrap it around and
 	 * begin from 0x1 again. This may be due to a busy system where lots
 	 * of slaves are getting created and deleted.

@@ -625,19 +721,13 @@ int ipvlan_link_new(struct net_device *dev, struct rtnl_newlink_params *params,
 	if (err)
 		goto remove_ida;
 
-	/* Flags are per port and latest update overrides. User has
-	 * to be consistent in setting it just like the mode attribute.
-	 */
-	if (data && data[IFLA_IPVLAN_FLAGS])
-		port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]);
-
-	if (data && data[IFLA_IPVLAN_MODE])
-		mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
-
 	err = ipvlan_set_port_mode(port, mode, extack);
 	if (err)
 		goto unlink_netdev;
 

Hm... What am I missing? The intention was to know "mode" a bit earlier and generate MAC as random for macnat-mode.. it's supposed to be just a simple line move a bit upper in the code

PS: I agree with other issues and will work on it

Dmitry