Re: [PATCH 5/5 v4] net: add old_queue_mapping into skb->cb

[jamal <hadi@cyberus.ca>]

On Tue, 2010-12-21 at 22:03 +0800, Changli Gao wrote:

> When I tested it, my OS got frozen.

I will look into it the next opportunity i get. The example i showed
is on egress btw. A ping from outside that matches the filter
will be a good test.

> Currently, you can only change the rx queue mapping, because for tx,
> dev_pick_tx() doesn't use skb->queue_mapping to choose tx queue.

If skbedit is on egress, it will happen after (and override whatever
dev_pick_tx() chose), no? Thats the whole point for skbedits queuemap
editing.

> However, I don't think change the rx queue mapping is a good idea.

I agree for that as a default policy. But it is
policy that skbedit can and should be able to override.

> When the skbs returned from ifb enter netif_receive_skb() again,
> get_rps_cpu() may warn about the wrong rx queue, and my this patch is
> used to solve this problem. Even though the rx queue is legal, a
> different rps_cpus settings will be used, and the skbs may be
> redirected to different CPUs. Is it expected?

I am not sure without analyzing what performance impact would be, i.e i
think that the only reason i wouldnt do it is because it may have crazy
effect on performance but:
If i wanted to override the choice made by rps through some policy, why
shouldnt i be able to do it? Same thing if i wanted to bypass rps. tc
level seems appropriate.
I may be misreading the code: Quick glance at the code indicates users
have no choice on ingress: rps happens first then we can do tc level -
so it doesnt matter what changes we make to the queue map it will not
take effect in any case. Am i mistaken?

cheers,
jamal