From: "Leon M. Busch-George" <leon@georgemail.eu>

nbp_switchdev_allowed_egress uses hwdom to determine whether or not a
packet has already been forwarded to a hardware domain. For
net_bridge_ports that aren't set up to use forward offloading, hwdom is
set to 0. When both ingress and egress port have no hwdom,
'cb->src_hwdom != p->hwdom' indicates that the packet is already known in
the target domain - which it isn't - and the packet is wrongly dropped.

The error was found on a bridge containing a wifi device and a VLAN
tagging device (e.g. eth0.12). With VLAN filtering, this shouldn't happen.

This patch adds a check for p->hwdom != 0 before comparing hardware
domains to restore forwarding between ports with hwdom = 0.

fwd_hwdoms are only set for ports with offloading enabled, which also
implies a valid hwdom, so the check '!test_bit(p->hwdom, &cb->fwd_hwdoms)'
doesn't fail in this way (yet - fingers crossed..) and it is left in place.

Co-developed-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Leon M. Busch-George <leon@georgemail.eu>
---
 net/bridge/br_switchdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/bridge/br_switchdev.c b/net/bridge/br_switchdev.c
index 7b41ee8740cb..29ff9af99276 100644
--- a/net/bridge/br_switchdev.c
+++ b/net/bridge/br_switchdev.c
@@ -67,7 +67,7 @@ bool nbp_switchdev_allowed_egress(const struct net_bridge_port *p,
 	struct br_input_skb_cb *cb = BR_INPUT_SKB_CB(skb);
 
 	return !test_bit(p->hwdom, &cb->fwd_hwdoms) &&
-		(!skb->offload_fwd_mark || cb->src_hwdom != p->hwdom);
+		(!skb->offload_fwd_mark || !p->hwdom || cb->src_hwdom != p->hwdom);
 }
 
 /* Flags that can be offloaded to hardware */
-- 
2.47.0