From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?iso-8859-1?Q?N=E9lio?= Laranjeiro Subject: Re: [PATCH v3 4/5] testpmd: handle all rxqs in rss setup Date: Tue, 28 Jun 2016 10:34:19 +0200 Message-ID: <20160628083419.GT14221@autoinstall.dev.6wind.com> References: <1462488421-118990-1-git-send-email-zhihong.wang@intel.com> <1465945686-142094-1-git-send-email-zhihong.wang@intel.com> <1465945686-142094-5-git-send-email-zhihong.wang@intel.com> <20160627142340.GO14221@autoinstall.dev.6wind.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable Cc: "Wang, Zhihong" , "dev@dpdk.org" , "Ananyev, Konstantin" , "Richardson, Bruce" , "thomas.monjalon@6wind.com" To: "De Lara Guarch, Pablo" Return-path: Received: from mail-wm0-f41.google.com (mail-wm0-f41.google.com [74.125.82.41]) by dpdk.org (Postfix) with ESMTP id BC35E2C34 for ; Tue, 28 Jun 2016 10:34:21 +0200 (CEST) Received: by mail-wm0-f41.google.com with SMTP id a66so16176496wme.0 for ; Tue, 28 Jun 2016 01:34:21 -0700 (PDT) Content-Disposition: inline In-Reply-To: List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Hi Pablo, On Mon, Jun 27, 2016 at 10:36:38PM +0000, De Lara Guarch, Pablo wrote: > Hi Nelio, >=20 > > -----Original Message----- > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of N=E9lio Laranjei= ro > > Sent: Monday, June 27, 2016 3:24 PM > > To: Wang, Zhihong > > Cc: dev@dpdk.org; Ananyev, Konstantin; Richardson, Bruce; De Lara Gua= rch, > > Pablo; thomas.monjalon@6wind.com > > Subject: Re: [dpdk-dev] [PATCH v3 4/5] testpmd: handle all rxqs in rs= s setup > >=20 > > On Tue, Jun 14, 2016 at 07:08:05PM -0400, Zhihong Wang wrote: > > > This patch removes constraints in rxq handling when multiqueue is e= nabled > > > to handle all the rxqs. > > > > > > Current testpmd forces a dedicated core for each rxq, some rxqs may= be > > > ignored when core number is less than rxq number, and that causes > > confusion > > > and inconvenience. > > > > > > One example: One Red Hat engineer was doing multiqueue test, there'= re 2 > > > ports in guest each with 4 queues, and testpmd was used as the forw= arding > > > engine in guest, as usual he used 1 core for forwarding, as a resul= ts he > > > only saw traffic from port 0 queue 0 to port 1 queue 0, then a lot = of > > > emails and quite some time are spent to root cause it, and of cours= e it's > > > caused by this unreasonable testpmd behavior. > > > > > > Moreover, even if we understand this behavior, if we want to test t= he > > > above case, we still need 8 cores for a single guest to poll all th= e > > > rxqs, obviously this is too expensive. > > > > > > We met quite a lot cases like this, one recent example: > > > http://openvswitch.org/pipermail/dev/2016-June/072110.html > > > > > > > > > Signed-off-by: Zhihong Wang > > > --- > > > app/test-pmd/config.c | 8 +------- > > > 1 file changed, 1 insertion(+), 7 deletions(-) > > > > > > diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c > > > index ede7c78..4719a08 100644 > > > --- a/app/test-pmd/config.c > > > +++ b/app/test-pmd/config.c > > > @@ -1199,19 +1199,13 @@ rss_fwd_config_setup(void) > > > cur_fwd_config.nb_fwd_ports =3D nb_fwd_ports; > > > cur_fwd_config.nb_fwd_streams =3D > > > (streamid_t) (nb_q * cur_fwd_config.nb_fwd_ports); > > > - if (cur_fwd_config.nb_fwd_streams > cur_fwd_config.nb_fwd_lcores) > > > - cur_fwd_config.nb_fwd_streams =3D > > > - (streamid_t)cur_fwd_config.nb_fwd_lcores; > > > - else > > > - cur_fwd_config.nb_fwd_lcores =3D > > > - (lcoreid_t)cur_fwd_config.nb_fwd_streams; > > > > > > /* reinitialize forwarding streams */ > > > init_fwd_streams(); > > > > > > setup_fwd_config_of_each_lcore(&cur_fwd_config); > > > rxp =3D 0; rxq =3D 0; > > > - for (lc_id =3D 0; lc_id < cur_fwd_config.nb_fwd_lcores; lc_id++) = { > > > + for (lc_id =3D 0; lc_id < cur_fwd_config.nb_fwd_streams; lc_id++)= { > > > struct fwd_stream *fs; > > > > > > fs =3D fwd_streams[lc_id]; > > > -- > > > 2.5.0 > >=20 > > Hi Zhihong, > >=20 > > It seems this commits introduce a bug in pkt_burst_transmit(), this o= nly > > occurs when the number of cores present in the coremask is greater th= an > > the number of queues i.e. coremask=3D0xffe --txq=3D4 --rxq=3D4. > >=20 > > Port 0 Link Up - speed 40000 Mbps - full-duplex > > Port 1 Link Up - speed 40000 Mbps - full-duplex > > Done > > testpmd> start tx_first > > io packet forwarding - CRC stripping disabled - packets/burst=3D6= 4 > > nb forwarding cores=3D10 - nb forwarding ports=3D2 > > RX queues=3D4 - RX desc=3D256 - RX free threshold=3D0 > > RX threshold registers: pthresh=3D0 hthresh=3D0 wthresh=3D0 > > TX queues=3D4 - TX desc=3D256 - TX free threshold=3D0 > > TX threshold registers: pthresh=3D0 hthresh=3D0 wthresh=3D0 > > TX RS bit threshold=3D0 - TXQ flags=3D0x0 > > Segmentation fault (core dumped) > >=20 > >=20 > > If I start testpmd with a coremask with at most as many cores as queu= es, > > everything works well (i.e. coremask=3D0xff0, or 0xf00). > >=20 > > Are you able to reproduce the same issue? > > Note: It only occurs on dpdk/master branch (commit f2bb7ae1d204). >=20 > Thanks for reporting this. I was able to reproduce this issue and > sent a patch that should fix it. Could you verify it? > http://dpdk.org/dev/patchwork/patch/14430/ I have tested it, it works, I will add a test report on the corresponding email. Thanks >=20 >=20 > Thanks > Pablo > >=20 > > Regards, --=20 N=E9lio Laranjeiro 6WIND