Re: [PATCH net-next] net, sched: add clsact qdisc

[Hannes Frederic Sowa <hannes@stressinduktion.org>]

On 07.01.2016 12:58, Daniel Borkmann wrote:
> On 01/07/2016 11:09 AM, Hannes Frederic Sowa wrote:
>> Hi Daniel and Alexei,
>>
>> On 07.01.2016 04:53, Alexei Starovoitov wrote:
>>> On Wed, Jan 06, 2016 at 02:00:56AM +0100, Daniel Borkmann wrote:
>>>>
>>>> I decided to extend the sch_ingress module with clsact functionality so
>>>> that commonly used code can be reused, the module is being aliased with
>>>> sch_clsact so that it can be auto-loaded properly. Alternative would
>>>> have been
>>>> to add a flag when initializing ingress to alter its behaviour plus
>>>> aliasing
>>>> to a different name (as it's more than just ingress). However, the
>>>> first would
>>>> end up, based on the flag, choosing the new/old behaviour by calling
>>>> different
>>>> function implementations to handle each anyway, the latter would
>>>> require to
>>>> register ingress qdisc once again under different alias. So, this
>>>> really begs
>>>> to provide a minimal, cleaner approach to have Qdisc_ops and
>>>> Qdisc_class_ops
>>>> by its own that share callbacks used by both.
>>> ...
>>>> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
>>>
>>> we've been going back and forth on the design and this final approach
>>> presented seems to be the best, since pros outweigh the cons.
>>>
>>> Acked-by: Alexei Starovoitov <ast@kernel.org>
>>
>> One question:
>>
>> With the advance in lockless qdiscs by John Fastabend, is it possible
>> to push the handle_egress hook further down into sched layer?
>
> Idea was that this is done before we pick txq as stated. F.e., could also
> be that we end up not having enqueue handler, thus moving this further down
> (not sure if there's a good place?), might make it all more scattered resp.
> complex to cover all parts.

My understanding is that using redirects and something like 
ifb/imq-alike could already solve some of those problems while not 
adding yet another handler to the stack?

Bye,
Hannes