On Wed, Dec 16, 2020 at 18:27, Vladimir Oltean [off-list ref] wrote:

On Wed, Dec 16, 2020 at 05:00:53PM +0100, Tobias Waldekranz wrote:

quoted

In a situation where a standalone port is indirectly attached to a
bridge (e.g. via a LAG) which is not offloaded, do not offload any
port attributes either. The port should behave as a standard NIC.

Previously, on mv88e6xxx, this meant that in the following setup:

     br0
     /
  team0
   / \
swp0 swp1

If vlan filtering was enabled on br0, swp0's and swp1's QMode was set
to "secure". This caused all untagged packets to be dropped, as their
default VID (0) was not loaded into the VTU.

Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
---
 net/dsa/slave.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/net/dsa/slave.c b/net/dsa/slave.c
index 4a0498bf6c65..faae8dcc0849 100644
--- a/net/dsa/slave.c
+++ b/net/dsa/slave.c

@@ -274,6 +274,9 @@ static int dsa_slave_port_attr_set(struct net_device *dev,
 	struct dsa_port *dp = dsa_slave_to_port(dev);
 	int ret;
 
+	if (attr->orig_dev != dev)
+		return -EOPNOTSUPP;
+

Should this not be:

	if (!dsa_port_offloads_netdev(dp, attr->orig_dev))
		return -EOPNOTSUPP;

?

That function is born in the following patch. So here I just align the
filtering with how switchdev objects were handled before this
series. Then in the next patch, all instances are converted to the exact
statement you suggest.

quoted

 	switch (attr->id) {
 	case SWITCHDEV_ATTR_ID_PORT_STP_STATE:
 		ret = dsa_port_set_state(dp, attr->u.stp_state, trans);
-- 
2.17.1