I'd like confirmation about a relationship between interfaces, vlans and bridges in the series 5 kernels and up.
Suppose we have a few similar hosts: each with a few interfaces, and each interface supports let's say two vlans, and has untagged traffic as well. So, three linux logical interfaces per physical interface. And, for simplicity sake let's say there are no L2 loops or cross connections. Each vlan interface and untagged interface has a different, unique mac address.
There are three 'simple' (not vlan aware, no PVID, egress untagged etc) bridges defined on each host: one has all the untagged interfaces as ports and no others, and the other two each collect as port members like vlan interfaces and no others. There are no L3 addresses on any of the ports (not even IPv6 fe80:xxx). The bridges may or may not have L3 ip4 / ip6 addresses. Let's ignore traffic sourced or destined for any of these hosts.
Am I correct that the bridge code will never send a copy of a packet out on other than the vlan (or untagged) it entered?
Also, if STP is enabled on a bridge, will it keep the bdpus and related on the assigned vlan/untagged that bridge was designed for and no others?
I've read some puzzling official seeming documentation that suggests 'bridge code' manages packets 'before vlan processing'.
(I am aware much code exists for one bridge be vlan aware. However, that code led to packet storms even when stp was enabled on vlans that had l2 loops.)
Suppose we have a few similar hosts: each with a few interfaces, and each interface supports let's say two vlans, and has untagged traffic as well. So, three linux logical interfaces per physical interface. And, for simplicity sake let's say there are no L2 loops or cross connections. Each vlan interface and untagged interface has a different, unique mac address.
There are three 'simple' (not vlan aware, no PVID, egress untagged etc) bridges defined on each host: one has all the untagged interfaces as ports and no others, and the other two each collect as port members like vlan interfaces and no others. There are no L3 addresses on any of the ports (not even IPv6 fe80:xxx). The bridges may or may not have L3 ip4 / ip6 addresses. Let's ignore traffic sourced or destined for any of these hosts.
Am I correct that the bridge code will never send a copy of a packet out on other than the vlan (or untagged) it entered?
Also, if STP is enabled on a bridge, will it keep the bdpus and related on the assigned vlan/untagged that bridge was designed for and no others?
I've read some puzzling official seeming documentation that suggests 'bridge code' manages packets 'before vlan processing'.
(I am aware much code exists for one bridge be vlan aware. However, that code led to packet storms even when stp was enabled on vlans that had l2 loops.)
Last edited:
