Re: [PATCH] veth: replace iflink by a dedicated symlink in sysfs

[David Miller <davem@davemloft.net>]

From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Date: Thu, 20 Aug 2015 16:31:11 +0200

> Le 20/08/2015 13:53, Jiri Benc a écrit :
>> On Wed, 19 Aug 2015 18:33:14 +0200, Nicolas Dichtel wrote:
>>> Probably better to introduce veth netlink attribute then, something
>>> like
>>> IFLA_VETH_PEER and keeps IFLA_LINK_NETNSID.
>>
>> I'd prefer IFLA_PEER. More generic attribute will be helpful should we
>> introduce an interface similar to veth in the future.s
> Ok.
> 
>>
>> Also, I'd not combine IFLA_LINK_NETNSID with IFLA_PEER. There might
>> very well be an interface in the future that will need both IFLA_LINK
>> and
>> IFLA_PEER and this would just create a confusion. It may be unlikely
>> but the attributes are cheap and it doesn't make sense to design uAPI
>> in a way that might bring problems in the future.
> Ok, but then this IFLA_PEER can include the ifindex and the nsid. No
> need
> to have two new attributes.
> 
>>
>>> I also don't know what is the best way to handle this. veth advertises
>>> its peer via IFLA_LINK since 4.1, so it's too late to change it for
>>> this
>>> release.
>>
>> Apparently we need to pick our poison. Either way, we break something.
> Sure. I would prefer to have the same mechanism in all version, but I
> can
> live with the other solution.
> 
> David, any thoughts about this?

You can't break the 4.1 semantics, it's in a released kernel and people
_ARE_ using it.