Re: [PATCH RFC 1/2] net: dsa: integrate with SWITCHDEV for HW bridging

[Guenter Roeck <linux@roeck-us.net>]

On 03/02/2015 07:27 AM, Andrey Volkov wrote:
> On 02/03/2015 15:49, Guenter Roeck wrote:
>> On 03/02/2015 06:38 AM, Andrey Volkov wrote:
>>>
>>> On 28/02/2015 08:53, Guenter Roeck wrote:
>>>> On 02/27/2015 09:09 AM, Andrey Volkov wrote:
>>>>> Gunter,
>>>>>
>>>>> Sorry with response delay, I very was busy yesterday
>>>>>
>>>>> Le 25/02/2015 15:25, Guenter Roeck a écrit :
>>>>>> Andrey,
>>>>>
>>>>> ------- snip -------
>>>>>
>>>>>>>>
>>>>>>> I simply modify port's fid to the new one in the leave routine and set to common bridge FID in enter
>>>>>>> (I'm using Marvell's chips). So the port's database will cleaned up automatically for the leave and will
>>>>>>> contain something useful at the enter time. Also I've look through yours patches and I haven't
>>>>>>
>>>>>> Does removing a port from a fid clean up the entries associated with it
>>>>>> in the database ?
>>>>>
>>>>> I've checked what happened when port changed its FID: switch logic block traffic to it
>>>>> immediately, as far as I can see, meanwhile record still exists in the bridge database,
>>>>> it was checked on 88e6185, 88e6097 and 88e6352 chips. And yet another 5c: changing of group membership is
>>>>> not atomic operation in the Marvell's chips known for me, so the port must be in the disabled state when it
>>>>> will happened.
>>>>>
>>>> Hmm - interesting. I assume you mean updating port registers 5 and 6 ?
>>> Yes sure, it's reason why we must disable the port before changing the FID.
>>>
>> Yes, I think we'll need to do that once we use the bits in register 5.
>>
>>>>
>>>> Different question: For 6185, did you write a new driver or extend an existing one ?
>>>> I found that it is quite similar to 6131, and that adding support for it to the 6131
>>>> driver should be straightforward.
>>> Yes again :), and 88E6097 have same core as 6123_61_65. Difference in both cases only in the number
>>> of supported ports, and it was main reason why hardcoded port's number was unacceptable for me, difference is
>>> large enough: for ex. 88e6123 have only 3 ports, but 88E6097 - 11.
>>>
>>
>> I have a patch set to change the number of ports to a variable in the 6131 driver.
>> Want me to submit it now ? Though I guess you must have pretty much the same,
>> so we can also use your approach. Let me know.
>
> I think that better to start from my patches: they are more generic and have support of sysfs
> (should be useful for "MII over ethernet"). Also IMO it will be better if we continue exchange/prereview

Sure, no problem. Only concern I have is that your patches don't seem to be available
in public, or maybe I missed the reference to it.

> our patches in more narrow mail list, since I do not want to pollute netdev by useless discussions about drafts.
> Objections/suggestions?
>
Sure, no problem, though personally I have no issues with the discussions
or with submitting draft patches, and I did not have the impression that
they are useless.

My patches are all in my repository at kernel.org; it is fine with me to keep
them (only) there if submitting drafts to netdev is considered pollution.

Thanks,
Guenter