Re: unexpected GRO/veth behavior

[Eric Dumazet]

On 09/10/2018 08:22 AM, Paolo Abeni wrote:
 in this already heavy cost engine.
> 
> Yup, even if I do not see any measurable cost added by the posted code.

Sure, micro bench marks wont show anything.

Now, if GRO receives one packet every 100 usec, as many hosts in the wild do,
there is an additional cost because of icache being wasted.