2014-03-26 10:57 GMT-07:00 Roopa Prabhu [off-list ref]:

On 3/26/14, 10:29 AM, Florian Fainelli wrote:

quoted

2014-03-26 9:59 GMT-07:00 Jiri Pirko [off-list ref]:

quoted

Wed, Mar 26, 2014 at 05:54:17PM CET, roopa@cumulusnetworks.com wrote:

quoted

On 3/26/14, 3:54 AM, Jamal Hadi Salim wrote:

quoted

On 03/26/14 01:37, Roopa Prabhu wrote:

quoted

On 3/25/14, 1:11 PM, Florian Fainelli wrote:

quoted

2014-03-25 12:35 GMT-07:00 Neil Horman [off-list ref]:

Sorry about getting on this thread late and possibly in the middle.
Agree on the idea of keeping the ports linked to the master switch dev
(or the 'conduit' to the switch chip) via private list instead of the
master-slave relationship proposed earlier.
By private i mean the netdev->priv linkage to the master switch dev
and
not really keeping the ports from being exposed to the user.

We think its better to keep the switch ports exposed as any other
netdev
on linux.
  This approach will make the switch ports look exactly like a nic
port
and all tools will continue to work seamlessly. The switch port
operations could internally be forwarded to the switch netdev (sw1 in
the above case).

example:
$ip link set dev sw1p0 up
$ethtool -S sw1p0

I like the approach. I know the above is a simple version, but i am
assuming you also mean i can do things like
ip route add ...
bridge fdb add ... (and if you like your brctl go ahead)
bonding ...

yes, exactly.  We support this model on our boxes today.
User can bond switch ports on our box in the exact same way as he/she
would bond two nic ports.
Our 'conduit to switch chip' reflects the corresponding lag
configuration in the switch chip.
Same goes for bridging, routing, acls.

So you implement bonding netlink api? Or you hook into bonding driver
itselt? Can you show us the code?

Before we start talking about bonding, maybe we should make sure that
we cover some basic hardware switches uses which are to make some
ports belong to certain VLANs, tagged or untagged?

It seems to me like this would become something like this, assuming P0
and P1 are two switch ports and 'eth0' is the CPU port, where P0 and
P1 belong to VLAN1 and CPU belongs to VLAN2:

ip link set dev sw1p0 up
ip link set dev sw1p1 up
ip link set dev eth0 up

ip link add link eth0 name eth0.2 type vlan id 2

ip link add link sw1p0 name sw1p0.1 type vlan id 1
ip link add link sw1p1 name sw1p1.1 type vlan id 1

ip link add sw1.1 type bridge
ip link set sw1p0.1 master sw1.1
ip link set sw1p1.1 master sw1.1

Does that fit the model correctly?

Not entirely, but close.
In our current model, there is no netdev for cpu port (or master switch
netdev):

You mean there is no netdev for the switch-side CPU-port facing the
CPU Ethernet MAC, right? There is still a netdev for the CPU Ethernet
MAC to receive packets destined to it presumably.

I do not think it hurts nor changes anything to introduce a CPU-port
netdev, this just gives greater flexibility and this should allow for
more complex setups where multiple CPU-ports exist (there are some
real devices using this...).

ip link set dev sw1p0 up
ip link set dev sw1p1 up

ip link add link sw1p0 name sw1p0.1 type vlan id 1
ip link add link sw1p1 name sw1p1.1 type vlan id 1

ip link add brvlan1 type bridge
ip link set swp1p0.1 master brvlan1
ip link set swp1p1.1 master brvlan1

switch driver programs the brvlan1 vlan in the switch asic.

bonding works in the same way.

Thanks,
Roopa

-- 
Florian