Re: Multiqueue pktgen and ingress path (Was: [PATCH v5 2/2] pktgen: introduce xmit_mode '<start_xmit|netif_receive>')

[Alexander Duyck <alexander.duyck@gmail.com>]

On 05/08/2015 10:00 AM, Alexei Starovoitov wrote:
> On 5/8/15 8:49 AM, Jesper Dangaard Brouer wrote:
>>>
>>> I've added a more plain version of a script, based on yours, below my
>>> signature.
>>
>> Now attached.
>
> thanks for the script! Really useful.
> Could you add it to samples/pktgen/ and remove useless and confusing
> pktgen.conf-2-1 ?
>
>>> The funny thing now is that scaling does not "happen" as we stall on:
>>>     atomic_long_inc(&skb->dev->rx_dropped);
>>
>> More interesting observations with the mentioned script (now attached).
>>
>> On my system the scaling stopped a 24Mpps, when I increased the number
>> of threads the collective scaling was stuck at 24Mpps.
>
> what was your config to start hitting that drop counter?
> We can convert it to per_cpu, but I'm not sure it's worth doing.
> If I send normal ip packets they don't go this path. Only
> unknown protocol packets suppose to hit it ?

I'm assuming it just has to be a packet that isn't claimed by any 
sockets or interfaces registered on top of the device.  After all it 
isn't as if pktgen sends an unknown protocol so I would assume just 
enabling promiscuous mode on an interface without a bridge or raw socket 
would probably be enough to trigger this.  The overhead itself would 
show up in __netif_receive_skb_core in perf since the atomic_inc would 
be inlined.

I would think a common case where something like this might be seen 
would be if you registered enough macvlan interfaces to force a device 
to switch into promiscuous mode and then received traffic that wasn't 
meant for any of the other interfaces.

- Alex