Re: [patch net-next v2 03/10] rtnl: expose physical switch id for particular device

[John Fastabend <john.fastabend@gmail.com>]

On 11/10/2014 02:14 PM, Jiri Pirko wrote:
> Mon, Nov 10, 2014 at 06:58:08PM CET, roopa@cumulusnetworks.com wrote:
>> On 11/9/14, 2:51 AM, Jiri Pirko wrote:
>>> The netdevice represents a port in a switch, it will expose
>>> IFLA_PHYS_SWITCH_ID value via rtnl. Two netdevices with the same value
>>> belong to one physical switch.
>>>
>>> Signed-off-by: Jiri Pirko <jiri@resnulli.us>
>>> ---
>>>   include/uapi/linux/if_link.h |  1 +
>>>   net/core/rtnetlink.c         | 26 +++++++++++++++++++++++++-
>>>   2 files changed, 26 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
>>> index 7072d83..4163753 100644
>>> --- a/include/uapi/linux/if_link.h
>>> +++ b/include/uapi/linux/if_link.h
>>> @@ -145,6 +145,7 @@ enum {
>>>   	IFLA_CARRIER,
>>>   	IFLA_PHYS_PORT_ID,
>>>   	IFLA_CARRIER_CHANGES,
>>> +	IFLA_PHYS_SWITCH_ID,
>>
>> Jiri, since we have not really converged on the switchdev class or having a
>> separate switchdev instance,
>> am thinking it is better if we dont expose any such switch_id to userspace
>> yet until absolutely needed. Do you need it today ?
>> There is no real in kernel hw switch driver that will use it today. And quite
>> likely this will need to change when we introduce real hw switch drivers.
>
> When and if the switchdev class is introduced, switch id can happily
> live on. It is nothing against it. Userspace should use this id to group
> the ports of physical switch.
>
>

+1 I think we need this otherwise how will userspace "know" how the
ports are related? If I have two switch silicon blocks in a single
platform or perhaps have a switch silicon with traditional host
nics on the same platform.


-- 
John Fastabend         Intel Corporation