Re: [RFC net-next PATCH V1 9/9] net: frag remove readers-writer lock (hack)

[Florian Westphal <fw@strlen.de>]

Jesper Dangaard Brouer <brouer@redhat.com> wrote:
> After all the other patches, the rw_lock is now the contention point.
> 
> This is a quick hack, that remove the readers-writer lock, by
> disabling/breaking hash rebuilding.  Just to see how big the
> performance gain would be.
> 
>   2x10G size(4416) result: 6481+6764 = 13245 Mbit/s (gen: 7652+8077 Mbit/s)
> 
>   4x10G size(4416) result:(5610+6283+5735+5238)=22866 Mbit/s
>                      (gen: 6530+7860+5967+5238 =25595 Mbit/s)
> 
> And the results show, that its a big win. With 4x10G size(4416)
> before: 17923 Mbit/s -> now: 22866 Mbit/s increase 4943 Mbit/s.
> With 2x10G size(4416) before 10689 Mbit/s -> 13245 Mbit/s
> increase 2556 Mbit/s.
> 
> I'll work on a real solution for removing the rw_lock while still
> supporting hash rebuilding.  Suggestions and ideas are welcome.

<devils advocate>
Why not kill it altogether, and just set new secret_interval
without moving frag queues to new location?

The only consequence is that all fragments queued at the time of
changing secret_interval will be lost, and free'd by evictor/timer.

Default secret rebuild interval is 10 minutes, should we care about
small packet loss every 10 minutes?
</devils advocate>