Re: [PATCH RFC 2/2] net: dsa: bcm_sf2: implement HW bridging operations

[Florian Fainelli <f.fainelli@gmail.com>]

On 19/02/15 16:09, Guenter Roeck wrote:
> On Thu, Feb 19, 2015 at 03:50:53PM -0800, Florian Fainelli wrote:
>> On 19/02/15 09:46, Guenter Roeck wrote:
>>> On Thu, Feb 19, 2015 at 09:27:23AM -0800, Florian Fainelli wrote:
>>>> On 18/02/15 21:59, Guenter Roeck wrote:
>>>>> On Wed, Feb 18, 2015 at 06:48:19PM -0800, Florian Fainelli wrote:
>>>>>> On 17/02/15 11:26, Florian Fainelli wrote:
>>>>>>> Update the Broadcom Starfighter 2 switch driver to implement the
>>>>>>> join/leave/stp_update callbacks required for basic hardware bridging
>>>>>>> support.
>>>>>>>
>>>>>>> There is not much to be done at the driver level but translating the
>>>>>>> STP state from Linux to their HW values.
>>>>>>>
>>>>>>> Joining a bridge means that the joining port and the other port members
>>>>>>> need to be in the same VLAN membership as the CPU, while leaving the
>>>>>>> bridge puts the port back into a separate VLAN membership with only the
>>>>>>> CPU.
>>>>>>
>>>>>> I found a couple additional issues while testing:
>>>>>>
>>>>>> - manipulating UP/DOWN state of interfaces that are part of a bridge
>>>>>> would not restore their bridge membership
>>>>>>
>>>>>> - removing an interface from a bridge and bringing it back up would
>>>>>> leave it in blocked state
>>>>>>
>>>>> Is this a problem with your implementation for sf2 or a generic problem
>>>>> with the first patch, such as some missing state transitions ?
>>>>
>>>> This is more of a side effect of having HW (an Ethernet switch) that can
>>>> really block a given port, based on last night's discussion with Roopa,
>>>> we can either fix this at the DSA level (in our case) or better fix this
>>>> at the bridge layer, I will propose a fix for this shortly.
>>>>
>>>>>
>>>>> For sf2, you might have to set the port state as well as the bridge
>>>>> association in the port_setup function. That is of course just a
>>>>> wild guess.
>>>>
>>>> Right, that's what I ended up doing. Thanks!
>>>
>>> Great.
>>>
>>> Another question: How do you handle flushing the forwarding database ?
>>>
>>> My current code for mv88e6352 flushes the forwarding database for a bridge
>>> group if the port association for that group changes (whenever a port joins
>>> or leaves a group, or whenever the state of a port in a group changes).
>>> There is, however, another situation where the forwarding database may
>>> have to be flushed - essentially on each topology change.
>>>
>>> How do you handle this situation ? Is it a real problem or do I just
>>> imagine that it is ?
>>
>> This is a real problem, for once I was working under the assumption that
>> the SF2 hardware was doing an automatic FDB flushing, but after
>> re-reading the documentation, this is not the case. My lab network does
>> not have many stations and I certainly did not catch that case.
> 
> The rocker code implements the fdb flush operation pretty much the same
> way I do, so I seem to be doing something right.
> 
> Not sure yet what to do about setting the fdb aging time. I don't see a
> mechanism to do that. No idea how important that is.
> 
> I think we'll also need support for ndo_vlan_rx_add_vid, ndo_vlan_rx_kill_vid,
> ndo_fdb_add, ndo_fdb_del, and ndo_fdb_dump. Have you thought about those ?

Yes, we will need those as well, although for now, I think we might just
be able to get basic HW bridging to work with the current patches +
driver specific FDB flushing, what do you think?

The only part thing that I cannot really map to HW switches today is
that the ndo_vlan_rx_add_vid() does not specify whether the VLAN id you
are adding is tagged or untagged (aka native). You can configure that
using the "bridge" sub-command from iproute2, but I am not sure how this
translates in terms of programming the hardware with the tagged/untagged
bit yet.

Maybe we should update ndo_vlan_rx_add_vid() to allow for a tag bit to
specified?
-- 
Florian