From mboxrd@z Thu Jan  1 00:00:00 1970
From: David Miller <davem@davemloft.net>
Subject: Re: [PATCH net 00/15] netpoll: avoid capture effects for NAPI
 drivers
Date: Sun, 23 Sep 2018 12:29:15 -0700 (PDT)
Message-ID: <20180923.122915.2174284095784692286.davem@davemloft.net>
References: <20180921222752.101307-1-edumazet@google.com>
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Cc: netdev@vger.kernel.org, michael.chan@broadcom.com,
        ariel.elior@cavium.com, eric.dumazet@gmail.com,
        tariqt@mellanox.com, saeedm@mellanox.com,
        jeffrey.t.kirsher@intel.com, jakub.kicinski@netronome.com,
        songliubraving@fb.com, j.vosburgh@gmail.com, vfalico@gmail.com,
        andy@greyhouse.net
To: edumazet@google.com
Return-path: <netdev-owner@vger.kernel.org>
Received: from shards.monkeyblade.net ([23.128.96.9]:34612 "EHLO
        shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1726437AbeIXB1u (ORCPT
        <rfc822;netdev@vger.kernel.org>); Sun, 23 Sep 2018 21:27:50 -0400
In-Reply-To: <20180921222752.101307-1-edumazet@google.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

From: Eric Dumazet <edumazet@google.com>
Date: Fri, 21 Sep 2018 15:27:37 -0700

> As diagnosed by Song Liu, ndo_poll_controller() can
> be very dangerous on loaded hosts, since the cpu
> calling ndo_poll_controller() might steal all NAPI
> contexts (for all RX/TX queues of the NIC).
> 
> This capture, showing one ksoftirqd eating all cycles
> can last for unlimited amount of time, since one
> cpu is generally not able to drain all the queues under load.
> 
> It seems that all networking drivers that do use NAPI
> for their TX completions, should not provide a ndo_poll_controller() :
> 
> Most NAPI drivers have netpoll support already handled
> in core networking stack, since netpoll_poll_dev()
> uses poll_napi(dev) to iterate through registered
> NAPI contexts for a device.

I'm having trouble understanding the difference.

If the drivers are processing all of the RX/TX queue draining by hand
in their ndo_poll_controller() method, how is that different from the
generic code walking all of the registererd NAPI instances one by one?