From mboxrd@z Thu Jan  1 00:00:00 1970
From: Song Liu <songliubraving@fb.com>
Subject: Re: pegged softirq and NAPI race (?)
Date: Tue, 18 Sep 2018 21:21:54 +0000
Message-ID: <117A22DB-A4CF-4FFB-AD08-A187E9F6EEB3@fb.com>
References: <A782704A-DF97-4E85-B10A-D2268A67DFD7@fb.com>
 <CANn89iLAPvPn2C9J+sSHsQ5EeF0hq47iMiiAfwoQkdZurQiyzA@mail.gmail.com>
 <a946c4d2-5e29-b1d4-3af6-8657caeebc11@fb.com>
 <CANn89iJTEMjYepKbr-8pmk0i03d9D+CfDFLPx+J=fqZivDJ9zQ@mail.gmail.com>
 <0FD562CC-CDE9-43C8-9623-B42AC7A208C8@fb.com>
 <CANn89i+kg=tE9aRYoFb+9ws35CCJJQz9Yc9i=9w=vDj=Epvj=A@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Cc: Alexei Starovoitov <ast@fb.com>, netdev <netdev@vger.kernel.org>,
        "Jeff Kirsher" <jeffrey.t.kirsher@intel.com>,
        Alexander Duyck <alexander.h.duyck@intel.com>,
        "michael.chan@broadcom.com" <michael.chan@broadcom.com>,
        Kernel Team <Kernel-team@fb.com>
To: Eric Dumazet <edumazet@google.com>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mx0a-00082601.pphosted.com ([67.231.145.42]:48376 "EHLO
        mx0a-00082601.pphosted.com" rhost-flags-OK-OK-OK-OK)
        by vger.kernel.org with ESMTP id S1726821AbeISC4c (ORCPT
        <rfc822;netdev@vger.kernel.org>); Tue, 18 Sep 2018 22:56:32 -0400
In-Reply-To: <CANn89i+kg=tE9aRYoFb+9ws35CCJJQz9Yc9i=9w=vDj=Epvj=A@mail.gmail.com>
Content-Language: en-US
Content-ID: <50937DF3F02752429AAA1E2F8CE10FDF@namprd15.prod.outlook.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>



> On Sep 18, 2018, at 2:13 PM, Eric Dumazet <edumazet@google.com> wrote:
>=20
> On Tue, Sep 18, 2018 at 1:37 PM Song Liu <songliubraving@fb.com> wrote:
>>=20
>=20
>> Looks like a patch like the following fixes the issue for ixgbe. But I
>> cannot explain it yet.
>>=20
>> Does this ring a bell?
>=20
> I dunno, it looks like the NIC is  generating an interrupt while it shoul=
d not,
> and constantly sets NAPI_STATE_MISSED.
>=20
> Or maybe we need to properly check napi_complete_done() return value.
>=20
> diff --git a/drivers/net/ethernet/intel/ixgb/ixgb_main.c
> b/drivers/net/ethernet/intel/ixgb/ixgb_main.c
> index d3e72d0f66ef428b08e4bd88508e05b734bc43a4..c4c565c982a98a5891603cedc=
dcb72dc1c401813
> 100644
> --- a/drivers/net/ethernet/intel/ixgb/ixgb_main.c
> +++ b/drivers/net/ethernet/intel/ixgb/ixgb_main.c
> @@ -1773,8 +1773,8 @@ ixgb_clean(struct napi_struct *napi, int budget)
>        ixgb_clean_rx_irq(adapter, &work_done, budget);
>=20
>        /* If budget not fully consumed, exit the polling mode */
> -       if (work_done < budget) {
> -               napi_complete_done(napi, work_done);
> +       if (work_done < budget &&
> +           napi_complete_done(napi, work_done)) {
>                if (!test_bit(__IXGB_DOWN, &adapter->flags))
>                        ixgb_irq_enable(adapter);
>        }

Thanks Eric!

I was looking at exactly this part. :) And it seems working!

I will run a bigger test and update shortly.=20

Best,
Song=