From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: [net-next, RFC, 4/8] net: core: add recycle capabilities on skbs
 via page_pool API
Date: Sat, 8 Dec 2018 04:34:15 -0800
Message-ID: <50266e0e-e2c6-efee-89d2-7518ee2761d0@gmail.com>
References: <154413874729.21735.10644578158550468689.stgit@firesoul>
 <20181208095758.GA32028@strlen.de>
 <72f33f12-9222-cbe7-6ff2-e4b4f86fb17c@gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Cc: netdev@vger.kernel.org, "David S. Miller" <davem@davemloft.net>,
        =?UTF-8?Q?Toke_H=c3=b8iland-J=c3=b8rgensen?= <toke@toke.dk>,
        ard.biesheuvel@linaro.org, Jason Wang <jasowang@redhat.com>,
        ilias.apalodimas@linaro.org,
        =?UTF-8?B?QmrDtnJuVMO2cGVs?= <bjorn.topel@intel.com>, w@1wt.eu,
        Saeed Mahameed <saeedm@mellanox.com>,
        mykyta.iziumtsev@gmail.com,
        Daniel Borkmann <borkmann@iogearbox.net>,
        Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        Tariq Toukan <tariqt@mellanox.com>
To: Florian Westphal <fw@strlen.de>,
        Jesper Dangaard Brouer <brouer@redhat.com>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mail-pf1-f170.google.com ([209.85.210.170]:33454 "EHLO
        mail-pf1-f170.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1726121AbeLHMeS (ORCPT
        <rfc822;netdev@vger.kernel.org>); Sat, 8 Dec 2018 07:34:18 -0500
Received: by mail-pf1-f170.google.com with SMTP id c123so3229138pfb.0
        for <netdev@vger.kernel.org>; Sat, 08 Dec 2018 04:34:17 -0800 (PST)
In-Reply-To: <72f33f12-9222-cbe7-6ff2-e4b4f86fb17c@gmail.com>
Content-Language: en-US
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>



On 12/08/2018 04:29 AM, Eric Dumazet wrote:
> 

> But I do not get why the patch is needed.
> 
> Adding extra cost for each skb destruction is costly.
> 
> I though XDP was all about _not_ having skbs.
> 
> Please let's do not slow down the non XDP stack only to make XDP more appealing.
> 

I have a similar concern with napi_consume_skb() :

This added a cost for locally generated TCP traffic, since most TX packets are fast clones.