From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id 3E2EB71 for ; Thu, 2 Oct 2014 10:57:50 +0200 (CEST) Received: from fmsmga001.fm.intel.com ([10.253.24.23]) by fmsmga101.fm.intel.com with ESMTP; 02 Oct 2014 02:04:38 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.04,638,1406617200"; d="scan'208";a="599715179" Received: from irsmsx103.ger.corp.intel.com ([163.33.3.157]) by fmsmga001.fm.intel.com with ESMTP; 02 Oct 2014 02:04:33 -0700 Received: from irsmsx107.ger.corp.intel.com (163.33.3.99) by IRSMSX103.ger.corp.intel.com (163.33.3.157) with Microsoft SMTP Server (TLS) id 14.3.195.1; Thu, 2 Oct 2014 10:04:28 +0100 Received: from irsmsx105.ger.corp.intel.com ([169.254.7.174]) by IRSMSX107.ger.corp.intel.com ([169.254.10.68]) with mapi id 14.03.0195.001; Thu, 2 Oct 2014 10:04:28 +0100 From: "Ananyev, Konstantin" To: "Richardson, Bruce" , Neil Horman Thread-Topic: [dpdk-dev] [PATCH v3] distributor_app: new sample app Thread-Index: AQHP3Jr3nDHv7Xjdv0yMyvh7Wjw+5JwZexeAgAAMNwCAABbFAIABpRAAgAACmwCAAAuBAIABLxxg Date: Thu, 2 Oct 2014 09:04:27 +0000 Message-ID: <2601191342CEEE43887BDE71AB977258213904F6@IRSMSX105.ger.corp.intel.com> References: <1411568210-2555-1-git-send-email-reshma.pattan@intel.com> <1412073577-12248-1-git-send-email-reshma.pattan@intel.com> <20140930113445.GB2193@hmsreliant.think-freely.org> <20140930121828.GA9312@BRICHA3-MOBL> <20140930133958.GG2193@hmsreliant.think-freely.org> <3AEA2BF9852C6F48A459DA490692831FE20FB6@IRSMSX109.ger.corp.intel.com> <20141001145620.GB24028@localhost.localdomain> <20141001153730.GA9292@BRICHA3-MOBL> In-Reply-To: <20141001153730.GA9292@BRICHA3-MOBL> Accept-Language: en-IE, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [163.33.239.180] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Cc: "dev@dpdk.org" Subject: Re: [dpdk-dev] [PATCH v3] distributor_app: new sample app X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 02 Oct 2014 08:57:51 -0000 > -----Original Message----- > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Bruce Richardson > Sent: Wednesday, October 01, 2014 4:38 PM > To: Neil Horman > Cc: dev@dpdk.org > Subject: Re: [dpdk-dev] [PATCH v3] distributor_app: new sample app >=20 > On Wed, Oct 01, 2014 at 10:56:20AM -0400, Neil Horman wrote: > > On Wed, Oct 01, 2014 at 02:47:00PM +0000, Pattan, Reshma wrote: > > > > > > > > > > -----Original Message----- > > > > From: Neil Horman [mailto:nhorman@tuxdriver.com] > > > > Sent: Tuesday, September 30, 2014 2:40 PM > > > > To: Richardson, Bruce > > > > Cc: Pattan, Reshma; dev@dpdk.org > > > > Subject: Re: [dpdk-dev] [PATCH v3] distributor_app: new sample app > > > > > > > > On Tue, Sep 30, 2014 at 01:18:28PM +0100, Bruce Richardson wrote: > > > > > On Tue, Sep 30, 2014 at 07:34:45AM -0400, Neil Horman wrote: > > > > > > On Tue, Sep 30, 2014 at 11:39:37AM +0100, reshmapa wrote: > > > > > > > From: Reshma Pattan > > > > > > > > > > > > > > A new sample app that shows the usage of the distributor libr= ary. > > > > > > > This app works as follows: > > > > > > > > > > > > > > * An RX thread runs which pulls packets from each ethernet po= rt in turn > > > > > > > and passes those packets to worker using a distributor comp= onent. > > > > > > > * The workers take the packets in turn, and determine the out= put port > > > > > > > for those packets using basic l2forwarding doing an xor on = the source > > > > > > > port id. > > > > > > > * The RX thread takes the returned packets from the workers a= nd enqueue > > > > > > > those packets into an rte_ring structure. > > > > > > > * A TX thread pulls the packets off the rte_ring structure an= d then > > > > > > > sends each packet out the output port specified previously = by > > > > > > > the worker > > > > > > > * Command-line option support provided only for portmask. > > > > > > > > > > > > > > Signed-off-by: Bruce Richardson > > > > > > > Signed-off-by: Reshma Pattan > > > > > > > --- > > > > > > > examples/Makefile | 1 + > > > > > > > examples/distributor_app/Makefile | 57 ++++ > > > > > > > examples/distributor_app/main.c | 600 > > > > +++++++++++++++++++++++++++++++++++++ > > > > > > > examples/distributor_app/main.h | 46 +++ > > > > > > > 4 files changed, 704 insertions(+), 0 deletions(-) create m= ode > > > > > > > 100644 examples/distributor_app/Makefile create mode 100644 > > > > > > > examples/distributor_app/main.c create mode 100644 > > > > > > > examples/distributor_app/main.h > > > > > > > > > > > > > > diff --git a/examples/Makefile b/examples/Makefile index > > > > > > > 6245f83..2ba82b0 100644 > > > > > > > --- a/examples/Makefile > > > > > > > +++ b/examples/Makefile > > > > > > > @@ -66,5 +66,6 @@ DIRS-y +=3D vhost > > > > > > > DIRS-$(CONFIG_RTE_LIBRTE_XEN_DOM0) +=3D vhost_xen DIRS-y += =3D vmdq > > > > > > > DIRS-y +=3D vmdq_dcb > > > > > > > +DIRS-$(CONFIG_RTE_LIBRTE_DISTRIBUTOR) +=3D distributor_app > > > > > > > > > > > > > > include $(RTE_SDK)/mk/rte.extsubdir.mk diff --git > > > > > > > a/examples/distributor_app/Makefile > > > > > > > b/examples/distributor_app/Makefile > > > > > > > new file mode 100644 > > > > > > > index 0000000..6a5bada > > > > > > > --- /dev/null > > > > > > > +++ b/examples/distributor_app/Makefile > > > > > > > @@ -0,0 +1,57 @@ > > > > > > > +# BSD LICENSE > > > > > > > +# > > > > > > > +# Copyright(c) 2010-2014 Intel Corporation. All rights res= erved. > > > > > > > +# All rights reserved. > > > > > > > +# > > > > > > > +# Redistribution and use in source and binary forms, with = or without > > > > > > > +# modification, are permitted provided that the following = conditions > > > > > > > +# are met: > > > > > > > +# > > > > > > > +# * Redistributions of source code must retain the above= copyright > > > > > > > +# notice, this list of conditions and the following di= sclaimer. > > > > > > > +# * Redistributions in binary form must reproduce the ab= ove copyright > > > > > > > +# notice, this list of conditions and the following di= sclaimer in > > > > > > > +# the documentation and/or other materials provided wi= th the > > > > > > > +# distribution. > > > > > > > +# * Neither the name of Intel Corporation nor the names = of its > > > > > > > +# contributors may be used to endorse or promote produ= cts derived > > > > > > > +# from this software without specific prior written pe= rmission. > > > > > > > +# > > > > > > > +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND > > > > CONTRIBUTORS > > > > > > > +# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING= , BUT > > > > NOT > > > > > > > +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AN= D > > > > FITNESS FOR > > > > > > > +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL T= HE > > > > COPYRIGHT > > > > > > > +# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT= , > > > > INCIDENTAL, > > > > > > > +# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, > > > > BUT NOT > > > > > > > +# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; > > > > LOSS OF USE, > > > > > > > +# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUS= ED > > > > AND ON ANY > > > > > > > +# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILI= TY, OR > > > > TORT > > > > > > > +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY O= UT > > > > OF THE USE > > > > > > > +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF = SUCH > > > > DAMAGE. > > > > > > > + > > > > > > > +ifeq ($(RTE_SDK),) > > > > > > > +$(error "Please define RTE_SDK environment variable") endif > > > > > > > + > > > > > > > +# Default target, can be overriden by command line or enviro= nment > > > > > > > +RTE_TARGET ?=3D x86_64-native-linuxapp-gcc > > > > > > > + > > > > > > > +include $(RTE_SDK)/mk/rte.vars.mk > > > > > > > + > > > > > > > +# binary name > > > > > > > +APP =3D distributor_app > > > > > > > + > > > > > > > +# all source are stored in SRCS-y SRCS-y :=3D main.c > > > > > > > + > > > > > > > +CFLAGS +=3D $(WERROR_FLAGS) > > > > > > > + > > > > > > > +# workaround for a gcc bug with noreturn attribute # > > > > > > > +http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D12603 > > > > > > > +ifeq ($(CONFIG_RTE_TOOLCHAIN_GCC),y) CFLAGS_main.o +=3D > > > > > > > +-Wno-return-type endif > > > > > > > + > > > > > > > +EXTRA_CFLAGS +=3D -O3 -Wfatal-errors > > > > > > > + > > > > > > > +include $(RTE_SDK)/mk/rte.extapp.mk > > > > > > > diff --git a/examples/distributor_app/main.c > > > > > > > b/examples/distributor_app/main.c new file mode 100644 index > > > > > > > 0000000..f555d93 > > > > > > > --- /dev/null > > > > > > > +++ b/examples/distributor_app/main.c > > > > > > > @@ -0,0 +1,600 @@ > > > > > > > +/*- > > > > > > > + * BSD LICENSE > > > > > > > + * > > > > > > > + * Copyright(c) 2010-2014 Intel Corporation. All rights re= served. > > > > > > > + * All rights reserved. > > > > > > > + * > > > > > > > + * Redistribution and use in source and binary forms, with= or without > > > > > > > + * modification, are permitted provided that the following= conditions > > > > > > > + * are met: > > > > > > > + * > > > > > > > + * * Redistributions of source code must retain the abov= e copyright > > > > > > > + * notice, this list of conditions and the following d= isclaimer. > > > > > > > + * * Redistributions in binary form must reproduce the a= bove copyright > > > > > > > + * notice, this list of conditions and the following d= isclaimer in > > > > > > > + * the documentation and/or other materials provided w= ith the > > > > > > > + * distribution. > > > > > > > + * * Neither the name of Intel Corporation nor the names= of its > > > > > > > + * contributors may be used to endorse or promote prod= ucts derived > > > > > > > + * from this software without specific prior written p= ermission. > > > > > > > + * > > > > > > > + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND > > > > CONTRIBUTORS > > > > > > > + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDIN= G, > > > > BUT NOT > > > > > > > + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY A= ND > > > > FITNESS FOR > > > > > > > + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL = THE > > > > COPYRIGHT > > > > > > > + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIREC= T, > > > > INCIDENTAL, > > > > > > > + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING= , > > > > BUT NOT > > > > > > > + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES= ; > > > > LOSS OF USE, > > > > > > > + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAU= SED > > > > AND ON ANY > > > > > > > + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABIL= ITY, OR > > > > TORT > > > > > > > + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY = OUT > > > > OF THE USE > > > > > > > + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF= SUCH > > > > DAMAGE. > > > > > > > + */ > > > > > > > + > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > + > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > +#include > > > > > > > + > > > > > > > +#include "main.h" > > > > > > > + > > > > > > > +#define RX_RING_SIZE 256 > > > > > > > +#define RX_FREE_THRESH 32 > > > > > > > +#define RX_PTHRESH 8 > > > > > > > +#define RX_HTHRESH 8 > > > > > > > +#define RX_WTHRESH 0 > > > > > > > + > > > > > > > +#define TX_RING_SIZE 512 > > > > > > > +#define TX_FREE_THRESH 32 > > > > > > > +#define TX_PTHRESH 32 > > > > > > > +#define TX_HTHRESH 0 > > > > > > > +#define TX_WTHRESH 0 > > > > > > > +#define TX_RSBIT_THRESH 32 > > > > > > > +#define TX_Q_FLAGS (ETH_TXQ_FLAGS_NOMULTSEGS | > > > > ETH_TXQ_FLAGS_NOVLANOFFL |\ > > > > > > > + ETH_TXQ_FLAGS_NOXSUMSCTP | ETH_TXQ_FLAGS_NOXSUMUDP | \ > > > > > > > + ETH_TXQ_FLAGS_NOXSUMTCP) > > > > > > > + > > > > > > > +#define NUM_MBUFS ((64*1024)-1) > > > > > > > +#define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + > > > > > > > +RTE_PKTMBUF_HEADROOM) #define MBUF_CACHE_SIZE 250 #define > > > > > > > +BURST_SIZE 32 #define RTE_RING_SZ 1024 > > > > > > > + > > > > > > > +/* uncommnet below line to enable debug logs */ > > > > > > > +/* #define DEBUG */ > > > > > > > + > > > > > > > +#ifdef DEBUG > > > > > > > +#define LOG_LEVEL RTE_LOG_DEBUG > > > > > > > +#define LOG_DEBUG(log_type, fmt, args...) do { \ > > > > > > > + RTE_LOG(DEBUG, log_type, fmt, ##args) \ > > > > > > > +} while (0) > > > > > > > +#else > > > > > > > +#define LOG_LEVEL RTE_LOG_INFO > > > > > > > +#define LOG_DEBUG(log_type, fmt, args...) do {} while (0) #e= ndif > > > > > > > + > > > > > > > +#define RTE_LOGTYPE_DISTRAPP RTE_LOGTYPE_USER1 > > > > > > > + > > > > > > > +/* mask of enabled ports */ > > > > > > > +static uint32_t enabled_port_mask =3D 0; > > > > > > > + > > > > > > > +static volatile struct app_stats { > > > > > > > + struct { > > > > > > > + uint64_t rx_pkts; > > > > > > > + uint64_t returned_pkts; > > > > > > > + uint64_t enqueued_pkts; > > > > > > > + } rx __rte_cache_aligned; > > > > > > > + > > > > > > > + struct { > > > > > > > + uint64_t dequeue_pkts; > > > > > > > + uint64_t tx_pkts; > > > > > > > + } tx __rte_cache_aligned; > > > > > > > +} app_stats; > > > > > > > + > > > > > > > +static const struct rte_eth_conf port_conf_default =3D { > > > > > > > + .rxmode =3D { > > > > > > > + .mq_mode =3D ETH_MQ_RX_RSS, > > > > > > > + .max_rx_pkt_len =3D ETHER_MAX_LEN, > > > > > > > + .split_hdr_size =3D 0, > > > > > > > + .header_split =3D 0, /**< Header Split disabled */ > > > > > > > + .hw_ip_checksum =3D 0, /**< IP checksum offload enabled */ > > > > > > > + .hw_vlan_filter =3D 0, /**< VLAN filtering disabled */ > > > > > > > + .jumbo_frame =3D 0, /**< Jumbo Frame Support disabled *= / > > > > > > > + .hw_strip_crc =3D 0, /**< CRC stripped by hardware */ > > > > > > > + }, > > > > > > > + .txmode =3D { > > > > > > > + .mq_mode =3D ETH_MQ_TX_NONE, > > > > > > > + }, > > > > > > > + .lpbk_mode =3D 0, > > > > > > > + .rx_adv_conf =3D { > > > > > > > + .rss_conf =3D { > > > > > > > + .rss_hf =3D ETH_RSS_IPV4 | ETH_RSS_IPV6 | > > > > > > > + ETH_RSS_IPV4_TCP | > > > > ETH_RSS_IPV4_UDP | > > > > > > > + ETH_RSS_IPV6_TCP | > > > > ETH_RSS_IPV6_UDP, > > > > > > > + } > > > > > > > + }, > > > > > > > +}; > > > > > > > + > > > > > > > +static const struct rte_eth_rxconf rx_conf_default =3D { > > > > > > > + .rx_thresh =3D { > > > > > > > + .pthresh =3D RX_PTHRESH, > > > > > > > + .hthresh =3D RX_HTHRESH, > > > > > > > + .wthresh =3D RX_WTHRESH, > > > > > > > + }, > > > > > > > + .rx_free_thresh =3D RX_FREE_THRESH, > > > > > > > + .rx_drop_en =3D 0, > > > > > > > +}; > > > > > > > + > > > > > > > +static const struct rte_eth_txconf tx_conf_default =3D { > > > > > > > + .tx_thresh =3D { > > > > > > > + .pthresh =3D TX_PTHRESH, > > > > > > > + .hthresh =3D TX_HTHRESH, > > > > > > > + .wthresh =3D TX_WTHRESH, > > > > > > > + }, > > > > > > > + .tx_free_thresh =3D TX_FREE_THRESH, > > > > > > > + .tx_rs_thresh =3D TX_RSBIT_THRESH, > > > > > > > + .txq_flags =3D TX_Q_FLAGS > > > > > > > + > > > > > > > +}; > > > > > > > + > > > > > > > +struct output_buffer { > > > > > > > + unsigned count; > > > > > > > + struct rte_mbuf *mbufs[BURST_SIZE]; }; > > > > > > > + > > > > > > > +/* > > > > > > > + * Initialises a given port using global settings and with t= he rx > > > > > > > +buffers > > > > > > > + * coming from the mbuf_pool passed as parameter */ static > > > > > > > +inline int port_init(uint8_t port, struct rte_mempool *mbuf_= pool) > > > > > > > +{ > > > > > > > + struct rte_eth_conf port_conf =3D port_conf_default; > > > > > > > + const uint16_t rxRings =3D 1, txRings =3D rte_lcore_count()= - 1; > > > > > > > + int retval; > > > > > > > + uint16_t q; > > > > > > > + > > > > > > > + if (port >=3D rte_eth_dev_count()) > > > > > > > + return -1; > > > > > > > + > > > > > > > + retval =3D rte_eth_dev_configure(port, rxRings, txRings, &p= ort_conf); > > > > > > > + if (retval !=3D 0) > > > > > > > + return retval; > > > > > > > + > > > > > > > + for (q =3D 0; q < rxRings; q++) { > > > > > > > + retval =3D rte_eth_rx_queue_setup(port, q, RX_RING_SIZE, > > > > > > > + rte_eth_dev_socket_id(port), > > > > > > > + &rx_conf_default, mbuf_pool); > > > > > > > + if (retval < 0) > > > > > > > + return retval; > > > > > > > + } > > > > > > > + > > > > > > > + for (q =3D 0; q < txRings; q++) { > > > > > > > + retval =3D rte_eth_tx_queue_setup(port, q, TX_RING_SIZE, > > > > > > > + rte_eth_dev_socket_id(port), > > > > > > > + &tx_conf_default); > > > > > > > + if (retval < 0) > > > > > > > + return retval; > > > > > > > + } > > > > > > > + > > > > > > > + retval =3D rte_eth_dev_start(port); > > > > > > > + if (retval < 0) > > > > > > > + return retval; > > > > > > > + > > > > > > > + struct rte_eth_link link; > > > > > > > + rte_eth_link_get_nowait(port, &link); > > > > > > > + if (!link.link_status) { > > > > > > > + sleep(1); > > > > > > > + rte_eth_link_get_nowait(port, &link); > > > > > > > + } > > > > > > > + > > > > > > > + if (!link.link_status) { > > > > > > > + printf("Link down on port %"PRIu8"\n", port); > > > > > > > + return 0; > > > > > > > + } > > > > > > > + > > > > > > > + struct ether_addr addr; > > > > > > > + rte_eth_macaddr_get(port, &addr); > > > > > > > + printf("Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8 > > > > > > > + " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n", > > > > > > > + (unsigned)port, > > > > > > > + addr.addr_bytes[0], addr.addr_bytes[1], > > > > > > > + addr.addr_bytes[2], addr.addr_bytes[3], > > > > > > > + addr.addr_bytes[4], addr.addr_bytes[5]); > > > > > > > + > > > > > > > + rte_eth_promiscuous_enable(port); > > > > > > > + > > > > > > > + return 0; > > > > > > > +} > > > > > > > + > > > > > > > +struct lcore_params { > > > > > > > + unsigned worker_id; > > > > > > > + struct rte_distributor *d; > > > > > > > + struct rte_ring *r; > > > > > > > +}; > > > > > > > + > > > > > > > +static __attribute__((noreturn)) void lcore_rx(struct > > > > > > > +lcore_params *p) { > > > > > > > + struct rte_distributor *d =3D p->d; > > > > > > > + struct rte_ring *r =3D p->r; > > > > > > > + const uint8_t nb_ports =3D rte_eth_dev_count(); > > > > > > > + const int socket_id =3D rte_socket_id(); > > > > > > > + uint8_t port; > > > > > > > + > > > > > > > + for (port =3D 0; port < nb_ports; port++) { > > > > > > > + /* skip ports that are not enabled */ > > > > > > > + if ((enabled_port_mask & (1 << port)) =3D=3D 0) > > > > > > > + continue; > > > > > > > + > > > > > > > + if (rte_eth_dev_socket_id(port) > 0 && > > > > > > > + rte_eth_dev_socket_id(port) !=3D socket_id) > > > > > > > + printf("WARNING, port %u is on remote NUMA node to > > > > " > > > > > > > + "RX thread.\n\tPerformance will not " > > > > > > > + "be optimal.\n", port); > > > > > > > + } > > > > > > > + > > > > > > > + printf("\nCore %u doing packet RX.\n", rte_lcore_id()); > > > > > > > + port =3D 0; > > > > > > > + for (;;) { > > > > > > > + /* skip ports that are not enabled */ > > > > > > > + if ((enabled_port_mask & (1 << port)) =3D=3D 0) { > > > > > > > + if (++port =3D=3D nb_ports) > > > > > > > + port =3D 0; > > > > > > > + continue; > > > > > > > + } > > > > > > > + struct rte_mbuf *bufs[BURST_SIZE*2]; > > > > > > > + const uint16_t nb_rx =3D rte_eth_rx_burst(port, 0, bufs, > > > > > > > + BURST_SIZE); > > > > > > > + app_stats.rx.rx_pkts +=3D nb_rx; > > > > > > > + > > > > > > > + rte_distributor_process(d, bufs, nb_rx); > > > > > > > + const uint16_t nb_ret =3D rte_distributor_returned_pkts(d, > > > > > > > + bufs, BURST_SIZE*2); > > > > > > > + app_stats.rx.returned_pkts +=3D nb_ret; > > > > > > > + if (unlikely(nb_ret =3D=3D 0)) > > > > > > > + continue; > > > > > > > + > > > > > > > + uint16_t sent =3D rte_ring_enqueue_burst(r, (void *)bufs, = nb_ret); > > > > > > > + app_stats.rx.enqueued_pkts +=3D sent; > > > > > > > + if (unlikely(sent < nb_ret)) { > > > > > > > + LOG_DEBUG(DISTRAPP, "%s:Packet loss due to full > > > > ring\n", __func__); > > > > > > > + while (sent < nb_ret) > > > > > > > + rte_pktmbuf_free(bufs[sent++]); > > > > > > > + } > > > > > > > + if (++port =3D=3D nb_ports) > > > > > > > + port =3D 0; > > > > > > > + } > > > > > > > +} > > > > > > > + > > > > > > > +static inline void > > > > > > > +flush_one_port(struct output_buffer *outbuf, uint8_t outp) { > > > > > > > + unsigned nb_tx =3D rte_eth_tx_burst(outp, 0, outbuf->mbufs, > > > > > > > + outbuf->count); > > > > > > > + app_stats.tx.tx_pkts +=3D nb_tx; > > > > > > > + > > > > > > > + if (unlikely(nb_tx < outbuf->count)) { > > > > > > > + LOG_DEBUG(DISTRAPP, "%s:Packet loss with tx_burst\n", > > > > __func__); > > > > > > > + do { > > > > > > > + rte_pktmbuf_free(outbuf->mbufs[nb_tx]); > > > > > > > + } while (++nb_tx < outbuf->count); > > > > > > > + } > > > > > > > + outbuf->count =3D 0; > > > > > > > +} > > > > > > > + > > > > > > > +static inline void > > > > > > > +flush_all_ports(struct output_buffer *tx_buffers, uint8_t > > > > > > > +nb_ports) { > > > > > > > + uint8_t outp; > > > > > > > + for (outp =3D 0; outp < nb_ports; outp++) { > > > > > > > + /* skip ports that are not enabled */ > > > > > > > + if ((enabled_port_mask & (1 << outp)) =3D=3D 0) > > > > > > > + continue; > > > > > > > + > > > > > > > + if (tx_buffers[outp].count =3D=3D 0) > > > > > > > + continue; > > > > > > > + > > > > > > > + flush_one_port(&tx_buffers[outp], outp); > > > > > > > + } > > > > > > > +} > > > > > > > + > > > > > > > +static __attribute__((noreturn)) void lcore_tx(struct rte_ri= ng > > > > > > > +*in_r) { > > > > > > > + static struct output_buffer tx_buffers[RTE_MAX_ETHPORTS]; > > > > > > > + const uint8_t nb_ports =3D rte_eth_dev_count(); > > > > > > > + const int socket_id =3D rte_socket_id(); > > > > > > > + uint8_t port; > > > > > > > + > > > > > > > + for (port =3D 0; port < nb_ports; port++) { > > > > > > > + /* skip ports that are not enabled */ > > > > > > > + if ((enabled_port_mask & (1 << port)) =3D=3D 0) > > > > > > > + continue; > > > > > > > + > > > > > > > + if (rte_eth_dev_socket_id(port) > 0 && > > > > > > > + rte_eth_dev_socket_id(port) !=3D socket_id) > > > > > > > + printf("WARNING, port %u is on remote NUMA node to > > > > " > > > > > > > + "TX thread.\n\tPerformance will not " > > > > > > > + "be optimal.\n", port); > > > > > > > + } > > > > > > > + > > > > > > > + printf("\nCore %u doing packet TX.\n", rte_lcore_id()); > > > > > > > + for (;;) { > > > > > > > + for (port =3D 0; port < nb_ports; port++) { > > > > > > > + /* skip ports that are not enabled */ > > > > > > > + if ((enabled_port_mask & (1 << port)) =3D=3D 0) > > > > > > > + continue; > > > > > > > + > > > > > > > + struct rte_mbuf *bufs[BURST_SIZE]; > > > > > > > + const uint16_t nb_rx =3D rte_ring_dequeue_burst(in_r, > > > > > > > + (void *)bufs, BURST_SIZE); > > > > > > > + app_stats.tx.dequeue_pkts +=3D nb_rx; > > > > > > > + > > > > > > > + /* if we get no traffic, flush anything we have */ > > > > > > > + if (unlikely(nb_rx =3D=3D 0)) { > > > > > > > + flush_all_ports(tx_buffers, nb_ports); > > > > > > > + continue; > > > > > > > + } > > > > > > > + > > > > > > > + /* for traffic we receive, queue it up for transmit */ > > > > > > > + uint16_t i; > > > > > > > + _mm_prefetch(bufs[0], 0); > > > > > > > + _mm_prefetch(bufs[1], 0); > > > > > > > + _mm_prefetch(bufs[2], 0); > > > > > > > + for (i =3D 0; i < nb_rx; i++) { > > > > > > > + struct output_buffer *outbuf; > > > > > > > + uint8_t outp; > > > > > > > + _mm_prefetch(bufs[i + 3], 0); > > > > > > > + /* workers should update in_port to hold the > > > > > > > + * output port value */ > > > > > > > + outp =3D bufs[i]->port; > > > > > > > + /* skip ports that are not enabled */ > > > > > > > + if ((enabled_port_mask & (1 << outp)) =3D=3D 0) > > > > > > > + continue; > > > > > > > + > > > > > > > + outbuf =3D &tx_buffers[outp]; > > > > > > > + outbuf->mbufs[outbuf->count++] =3D bufs[i]; > > > > > > > + if (outbuf->count =3D=3D BURST_SIZE) > > > > > > > + flush_one_port(outbuf, outp); > > > > > > > + } > > > > > > > + } > > > > > > > + } > > > > > > > +} > > > > > > > + > > > > > > > + > > > > > > > +static __attribute__((noreturn)) void lcore_worker(struct > > > > > > > +lcore_params *p) { > > > > > > > + struct rte_distributor *d =3D p->d; > > > > > > > + const unsigned id =3D p->worker_id; > > > > > > > + /* for single port, xor_val will be zero so we won't modify= the output > > > > > > > + * port, otherwise we send traffic from 0 to 1, 2 to 3, and= vice versa > > > > > > > + */ > > > > > > > + const unsigned xor_val =3D (rte_eth_dev_count() > 1); > > > > > > > + struct rte_mbuf *buf =3D NULL; > > > > > > > + > > > > > > > + printf("\nCore %u acting as worker core.\n", rte_lcore_id()= ); > > > > > > > + for (;;) { > > > > > > > + buf =3D rte_distributor_get_pkt(d, id, buf); > > > > > > > + buf->port ^=3D xor_val; > > > > > > > + } > > > > > > > +} > > > > > > > + > > > > > > > +static void > > > > > > > +int_handler(int sig_num) > > > > > > > +{ > > > > > > > + struct rte_eth_stats eth_stats; > > > > > > > + unsigned i; > > > > > > > + > > > > > > > + printf("Exiting on signal %d\n", sig_num); > > > > > > > + > > > > > > > + printf("\nRX thread stats:\n"); > > > > > > > + printf(" - Received: %"PRIu64"\n", app_stats.rx.rx_pkts)= ; > > > > > > > + printf(" - Processed: %"PRIu64"\n", app_stats.rx.returned= _pkts); > > > > > > > + printf(" - Enqueued: %"PRIu64"\n", app_stats.rx.enqueued= _pkts); > > > > > > > + > > > > > > > + printf("\nTX thread stats:\n"); > > > > > > > + printf(" - Dequeued: %"PRIu64"\n", app_stats.tx.dequeue_= pkts); > > > > > > > + printf(" - Transmitted: %"PRIu64"\n", app_stats.tx.tx_pkts)= ; > > > > > > > + > > > > > > > + for (i =3D 0; i < rte_eth_dev_count(); i++) { > > > > > > > + rte_eth_stats_get(i, ð_stats); > > > > > > > + printf("\nPort %u stats:\n", i); > > > > > > > + printf(" - Pkts in: %"PRIu64"\n", eth_stats.ipackets); > > > > > > > + printf(" - Pkts out: %"PRIu64"\n", eth_stats.opackets); > > > > > > > + printf(" - In Errs: %"PRIu64"\n", eth_stats.ierrors); > > > > > > > + printf(" - Out Errs: %"PRIu64"\n", eth_stats.oerrors); > > > > > > > + printf(" - Mbuf Errs: %"PRIu64"\n", eth_stats.rx_nombuf); > > > > > > > + } > > > > > > > + exit(0); > > > > > > rte_exit here? Also, this is a pretty ungraceful exit strategy= as > > > > > > all the threads you've created and memory you've allocated are = just > > > > forgotten here. > > > > > > Given that dpdk mempools are shared, this has the potential to = leak > > > > > > lots of memory if other apps are using the dpdk at the same tim= e > > > > > > that you run this. You probably want to use the sigint handler= to > > > > > > raise a flag to the tx/rx threads to shutdown gracefully, and t= hen free your > > > > allocated memory and mempool. > > > > > > > > > > > > Neil > > > > > > > > > > > > > > > > Unless the different processes are explicitly cooperating as > > > > > primary/secondary, the mempools are not shared. I just don't see = the > > > > > need for this app to do more cleanup on ctrl-c signal, as it's no= t > > > > > intended to be a multiprocess app, and there is little that any > > > > > secondary process could do to work with this app, except possibly= some > > > > > resource monitoring, which would be completely unaffected by it e= xiting the > > > > way it does. > > > > > > > > > Ah, ok, so we don't use a common shared pool between isolated proce= sses > > > > then, thats good. Still though, this is a sample application, I th= ink its lazy > > > > programming practice to illustrate to application developers that i= ts generally ok > > > > to exit programs without freeing your resources. Its about 20 line= s of additional > > > > code to change the sigint handler to flag an exit condition, and ha= ve all the > > > > other threads join on it. > > > > > > > > Neil > > > > > > 1)I had sent v5 patch which handles graceful shutdown of rx and tx th= reads upon SIGINT > > I see it and will take a look shortly, thanks. > > > > > 2)Worker thread graceful shutdown was not handled as of now as it nee= ds some change in lcore_worker logic , which will be done > in future enhancements. > > Not sure I understand what you mean here. Can you elaborate? > > > > > 3)Freeing of mempool is also not handled , as the framework support i= s not available. > > Ew, I hadn't noticed that, freeing of mempools seems like something we = should > > implement. > > > > > 4)Cleaning of rx/tx queues not done, as it needs some extensive logic= which we haven't planned as of now. Will check the > possibility of doing it in future enhancements i.e in next version of= sample application. > > We can't just flush the queues after we shutdown the workers? I presum= e a queue > > flush operation exists, yes? > > Neil >=20 > Other than code hygiene, which does have some value in itself, I can't > really see what the practical point of such cleanup would be. >=20 > If traffic is going through the system, and the process is killed packets > will be dropped, whatever we do, as packet reception will stop. If traffi= c > is not going through the system, then there are no packets in flight and > therefore no relevant cleanup to be done. [And if the traffic is stopped > just before shutting down the app, at a throughput rate of a couple of > million packets per second, the app should be flushed of packets within t= iny > fractions of a second]. I think that in theory not resetting HW at process termination can cause a = problem. Something like that: - DPDK app has HW RX/TX queues active with armed RXDs, but the link is idle= (no packets are flying).=20 - DPDK app terminates abnormally. - user deletes DPDK hugepages files. - Hugepage memory that was used by DPDK for RXDs/data buffers are given to = other app (or kernel). - Packet arrives -a s HW is still active - it will do a write to the RXD an= d data-buffer.=20 - Silent memory corruption. Saying that, I don't think we can completely eliminate that problem from us= er-space code - as not process signals can be handled.=20 >>From other side - our whole concept is to move away from custom kernel modu= les... Though we probably can make it less possible to happen - create a terminati= on handler, that would try to reset all active HW. And make sure it is called atexit and all catchable signals that cause proc= ess termination. But I don' t think it is a good idea to duplicate such code in each and eve= ry sample app. I think it should be in the librte_eal, and yes - it is a subject of a sepa= rate patch/discussion :) Konstantin =20 > So overall, I just don't see complicated jumping through hoops for flushi= ng > and cleaning up things being worth the effort. These apps are designed to > run in a forever loop, and, in the exception case, when killed the cleanu= p > done by kernel is sufficient. >=20 > regards, > /Bruce