Re: [PATCH net-next] net: introduce SO_INCOMING_CPU

[David Miller <davem@davemloft.net>]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Mon, 10 Nov 2014 15:51:21 -0800

> On Mon, 2014-11-10 at 15:08 -0500, David Miller wrote:
>> From: Eric Dumazet <eric.dumazet@gmail.com>
>> Date: Fri, 07 Nov 2014 12:51:12 -0800
>> 
>> > @@ -1455,6 +1455,7 @@ process:
>> >  		goto discard_and_relse;
>> >  
>> >  	sk_mark_napi_id(sk, skb);
>> > +	sk_incoming_cpu_update(sk);
>> 
>> Just make sk_mark_napi_id() call sk_incoming_cpu_update().
>> 
>> You've matched up the calls precisely in this patch, and I can't think
>> of any situation where we'd add a sk_mark_napi_call() not not want to
>> do an sk_incoming_cpu_update().
> 
> I believe this was a coincidence.
> 
> In fact some sk_mark_napi_id() calls are not at the right place.
> It makes little sense to change sk->sk_napi_id for a listener socket.
> 
> sk_mark_napi_id() should better be done [1] at the same time we call
> sock_rps_save_rxhash
> 
> But we need to store cpu before prequeue or backlog (as I did in my
> patch)

Ok, so why don't you submit two patches.  First, fix the locations
where we do sk_mark_napi_id(), then your SO_INCOMING_CPU patch
(unchanged)?

Thanks.