iproute2/vlan: Idea: Support untagged VLAN interfaces

[Brian Allen Vanderburg II <brianvanderburg2@aim.com>]

[-- Attachment #1: Type: text/plain, Size: 2706 bytes --]

I've been recent playing around with some of the new VLAN filtering
features.  One feature I want to be able to do is to create an interface
that handles untagged data for VLANs as a sub-interface instead. 
Currently, untagged data is only handled by the main interface.  I can
think of some reasons to have untagged data on its own VLAN interface.

* The ability to bring up and down the untagged interface without
affecting all tagged interfaces as well.
* The ability to bridge only untagged traffic for mapping/translating. 
If an untagged sub-interface exists, then VLAN mapping can be performed
by bridging the VLAN interfaces of two NICs with differing VLANs and
including the untagged traffic.  Without untagged VLAN sub-interfaces
only tagged traffic could be mapped.  The bridge VLAN filtering allows
filtering of VLANs per port, but does not seem to allow mapping or
translating VLANs.
* Misc

The bridge VLAN filtering feature seems to make it possible, but
requires a rather large amount of configuration.  This first section
will take any traffic on eth0 and pass it in to br0.  VLAN 1 is untagged
on eth0, but is treated as tagged as it passes to br0.

    ip link add dev br0 type bridge
    ip link set dev eth0 master br0
    ip link set dev eth0 promisc on
    echo 1 > /sys/class/net/br0/bridge/vlan_filtering

    bridge vlan add dev eth0 vid 1 pvid untagged
    bridge vlan add dev eth0 vid 10
    bridge vlan add dev eth0 vid 99

    bridge vlan add dev br0 vid 1 self
    bridge vlan add dev br0 vid 10 self
    bridge vlan add dev br0 vid 99 self

Because br0 sees even the original untagged traffic as tagged, I can now
add a separate interface for the untagged VLAN as desired:

    ip link add link br0 name br0_user type vlan id 1
    ip link add link br0 name br0_priv type vlan id 10
    ip link add link br0 name br0_mgmt type vlan id 99

The interfaces can then be configured as normal.  I can bring down any
interface and it will not affect the others, or do anything else.  I
have managed to confirm this seems to work fine in a VM where untagged
traffic on eth0 was received on br0_user as expected.

This is a bit of work to get a "sub-interface" for an untagged VLAN. 
One idea I had was that if the groundwork already exists, it may be
possible to add something similar directly to the interface without the
bridge and VLAN filtering setup.  I could imagine a command such as

    ip link add link eth0 name eth0_user type vlan id 1 pvid untagged

as being able to accomplish the same thing without the overhead of
creating a bridge and assigning all the VLANs to both the physical
device and bridge device.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 901 bytes --]