From mboxrd@z Thu Jan 1 00:00:00 1970 From: Song Liu Subject: Re: [PATCH net] ixgbe: check return value of napi_complete_done() Date: Fri, 21 Sep 2018 07:17:03 +0000 Message-ID: References: <20180920190113.490005-1-songliubraving@fb.com> <2ca0e823642d232092017f66e8151652e22e74a1.camel@intel.com> <0DAF1AF9-E98D-4D0F-BD68-F5936A0312C6@fb.com> <028b4cea-0e3f-fab5-7a74-cf003bbd1134@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Cc: Jeff Kirsher , netdev , "intel-wired-lan@lists.osuosl.org" , Kernel Team , "stable@vger.kernel.org" , Alexei Starovoitov To: Eric Dumazet Return-path: In-Reply-To: <028b4cea-0e3f-fab5-7a74-cf003bbd1134@gmail.com> Content-Language: en-US Content-ID: Sender: stable-owner@vger.kernel.org List-Id: netdev.vger.kernel.org > On Sep 20, 2018, at 4:49 PM, Eric Dumazet wrote: >=20 >=20 >=20 > On 09/20/2018 04:43 PM, Song Liu wrote: >>=20 >=20 >> I tried to totally skip ndo_poll_controller() here. It did avoid hitting >> the issue. However, netpoll will drop (fail to send) more packets.=20 >>=20 >=20 > Why is it failing ? >=20 > If you are under high memory pressure, then maybe if you absolutely want = memory to send > netpoll packets, you want to grab all NAPI contexts as a way to prevent o= ther cpus=20 > from feeding incoming packets to the host and add more memory pressure ;) >=20 I did the test with Eric's latest patch (and disable ndo_poll_controller=20 in driver). The result didn't show significant increase in drop packets.=20 I guess packet drops in my earlier test was caused by some other changes I mixed there.=20 So I think this patch does fix the issue. Thanks Eric! For ixgbe, I think we need to check napi_complete_done() return value anyway. Otherwise, the driver will enable IRQ in polling mode. =20 Song=