Hi Pablo, Please do not apply v8. Sashiko found an out-of-bounds read in it: with a vlan device of one bridge enslaved to another bridge, port_vlans counts the vlans below both bridges, so the gc looks up the bridge port from the wrong device and reads past encap[]. Resetting port_vlans at each bridge fixes it, and a selftest case covers it.
v9 has that fix, credits your patch as its author with your sign-off and a note of my changes instead of Co-developed-by, and is rebased onto nf-next. I will post it once nf-next reopens. pw-bot: changes-requested Julius
