On 16-02-24 12:46 AM, Simon Horman wrote:
On Tue, Feb 23, 2016 at 05:12:34PM +0100, Daniel Borkmann wrote:
From my point of view the test should be weather the encapsulation might reasonably be expected to be used outside of the context of tc. If so then it makes sense to use a netdev to allow sharing of infrastructure between different kernel components. I suspect the answer to that question is no and thus IMHO a netdev would be nice to have rather than compelling. With regards to overhead of netdevs: I think that could be mitigated to some extent by using LWT or some other metadata-based approach to allow a single netdev to be use by multiple tc action instances.
We actually have a use case where we offload this thing into an embedded NIC. In any case it doesnt make much sense to use a netdev for reasons i specified. Just like it doesnt make sense when i want a policy which pushes or pops vlans or vxlans to use netdevs either. Yes it quacks like a duck(i.e has receive) and walks like a duck(has stats) but it looks like an ostrich;-> cheers, jamal
