]> git.baikalelectronics.ru Git - kernel.git/commit
net: dsa: sja1105: fix address learning getting disabled on the CPU port
authorVladimir Oltean <vladimir.oltean@nxp.com>
Tue, 13 Jul 2021 09:37:19 +0000 (12:37 +0300)
committerDavid S. Miller <davem@davemloft.net>
Tue, 13 Jul 2021 16:32:41 +0000 (09:32 -0700)
commit116428662b82fd33547e78b7305186d266e0c6ee
treede3dbb8667d09b9e30d0bfe93a7a4a4dd39e10f1
parent86d3f37e2e56ac23f18488a3809413f2901d3a81
net: dsa: sja1105: fix address learning getting disabled on the CPU port

In May 2019 when commit 30b690660842 ("net: dsa: sja1105: Add support
for Spanning Tree Protocol") was introduced, the comment that "STP does
not get called for the CPU port" was true. This changed after commit
404668d1f07e ("net: dsa: enable and disable all ports") in August 2019
and went largely unnoticed, because the sja1105_bridge_stp_state_set()
method did nothing different compared to the static setup done by
sja1105_init_mac_settings().

With the ability to turn address learning off introduced by the blamed
commit, there is a new priv->learn_ena port mask in the driver. When
sja1105_bridge_stp_state_set() gets called and we are in
BR_STATE_LEARNING or later, address learning is enabled or not depending
on priv->learn_ena & BIT(port).

So what happens is that priv->learn_ena is not being set from anywhere
for the CPU port, and the static configuration done by
sja1105_init_mac_settings() is being overwritten.

To solve this, acknowledge that the static configuration of STP state is
no longer necessary because the STP state is being set by the DSA core
now, but what is necessary is to set priv->learn_ena for the CPU port.

Fixes: 30abb4062fc1 ("net: dsa: sja1105: offload bridge port flags to device")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/dsa/sja1105/sja1105_main.c