From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from dpdk.org (dpdk.org [92.243.14.124]) by inbox.dpdk.org (Postfix) with ESMTP id E2A5FA046B for ; Tue, 23 Jul 2019 07:47:11 +0200 (CEST) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id B55371BF7E; Tue, 23 Jul 2019 07:47:11 +0200 (CEST) Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by dpdk.org (Postfix) with ESMTP id BFE171BF7C for ; Tue, 23 Jul 2019 07:47:08 +0200 (CEST) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga001.jf.intel.com ([10.7.209.18]) by fmsmga104.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 22 Jul 2019 22:47:07 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.64,297,1559545200"; d="scan'208";a="253128012" Received: from unknown (HELO localhost.localdomain.sh.intel.com) ([10.240.176.135]) by orsmga001.jf.intel.com with ESMTP; 22 Jul 2019 22:47:06 -0700 From: Peng Yuan To: dts@dpdk.org Cc: Peng Yuan Date: Tue, 23 Jul 2019 13:51:45 +0800 Message-Id: <1563861105-40707-1-git-send-email-yuan.peng@intel.com> X-Mailer: git-send-email 1.9.3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Subject: [dts] [PATCH]test_plans: add new cases to af_xdp_test_plan X-BeenThere: dts@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: test suite reviews and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dts-bounces@dpdk.org Sender: "dts" Add five cases to af_xdp_test_plan.rst covering zero copy, multiqueue and need_wakeup Signed-off-by: Peng Yuan diff --git a/test_plans/af_xdp_test_plan.rst b/test_plans/af_xdp_test_plan.rst index 4576fef..d12ef0c 100644 --- a/test_plans/af_xdp_test_plan.rst +++ b/test_plans/af_xdp_test_plan.rst @@ -51,7 +51,7 @@ Prerequisites 2. The NIC is located on the socket 1, so we define the cores of socket 1. -3. Take the kernel >= v5.1-rc1, build kernel and replace your host +3. Take the kernel >= v5.2-rc2, build kernel and replace your host kernel with it. Update compiler to the proper version. Make sure you turn on XDP sockets when compiling:: @@ -60,59 +60,239 @@ Prerequisites Networking options --> [ * ] XDP sockets + Then compile the kernel:: + + make -j16 + make modules_install install + 4. Build libbpf in tools/lib/bpf:: cd tools/lib/bpf - make + make install_lib prefix=/usr + make install_headers prefix=/usr + +5. Explicitly enable AF_XDP pmd by adding below line to + config/common_linux:: + + CONFIG_RTE_LIBRTE_PMD_AF_XDP=y - Then copy the libbpf.a and libbpf.so to /usr/lib64 + Then build DPDK. -5. Set DUT port only has one queue:: +6. Set DUT port only has one queue:: ethtool -L enp216s0f0 combined 1 ethtool -L enp216s0f1 combined 1 -6. Explicitly enable AF_XDP pmd by adding below line to - config/common_linux:: +Test case 1: single port +======================== - CONFIG_RTE_LIBRTE_PMD_AF_XDP=y +1. Start the testpmd:: + + ./testpmd -l 29,30 -n 6 --no-pci --vdev net_af_xdp0,iface=enp216s0f0 \ + -- -i --nb-cores=1 --rxq=1 --txq=1 --port-topology=loop + +2. Assign the kernel core:: + + ./set_irq_affinity 34 enp216s0f0 + +3. Send packets by packet generator with different packet size, + from 64 bytes to 1518 bytes, check the throughput. -Test case 1: 1 port, 1 kernel core, 1 user core -=============================================== +Test case 2: two ports +====================== 1. Start the testpmd:: - ./x86_64-native-linuxapp-gcc/app/testpmd -l 30,31 --no-pci -n 6 \ - --vdev net_af_xdp0,iface=enp216s0f0,queue=0 --file-prefix=port0 \ - --socket-mem 1024,1024 -- -a --nb-cores=1 -i --rxq=1 --txq=1 \ - --port-topology=loop + ./x86_64-native-linuxapp-gcc/app/testpmd -l 29,30-31 --no-pci -n 6 \ + --vdev net_af_xdp0,iface=enp216s0f0 --vdev net_af_xdp1,iface=enp216s0f1 \ + -- -i --nb-cores=2 --rxq=1 --txq=1 2. Assign the kernel core:: ./set_irq_affinity 33 enp216s0f0 + ./set_irq_affinity 34 enp216s0f1 -3. Send packet with packet generator with different packet size, - from 64 bytes to 1518 bytes, check the throughput. +3. Send packets by packet generator port0 with different packet size, + from 64 bytes to 1518 bytes, check the throughput at port1. -Test case 2: 2 ports, 2 kernel cores, 2 user cores -================================================== +4. Send packets by packet generator port0 and port1 with different packet size, + from 64 bytes to 1518 bytes, check the throughput at port0 and port1. -1. Start the testpmd:: +Test case 3: zero copy +====================== - ./x86_64-native-linuxapp-gcc/app/testpmd -l 30,32 --no-pci -n 6 \ - --vdev net_af_xdp0,iface=enp216s0f0,queue=0 --file-prefix=port0 \ - --socket-mem 1024,1024 -- -a --nb-cores=1 -i --rxq=1 --txq=1 \ - --port-topology=loop +1. Start the testpmd:: - ./x86_64-native-linuxapp-gcc/app/testpmd -l 31,33 --no-pci -n 6 \ - --vdev net_af_xdp1,iface=enp216s0f1,queue=0 --file-prefix=port1 \ - --socket-mem 1024,1024 -- -a --nb-cores=1 -i --rxq=1 --txq=1 \ - --port-topology=loop + ./testpmd -l 29,30 -n 6 --no-pci \ + --vdev net_af_xdp0,iface=enp216s0f0,pmd_zero_copy=1 \ + -- -i --nb-cores=1 --rxq=1 --txq=1 --port-topology=loop 2. Assign the kernel core:: - ./set_irq_affinity 35 enp216s0f0 - ./set_irq_affinity 36 enp216s0f1 + ./set_irq_affinity 34 enp216s0f0 -3. Send packet with packet generator with different packet size, +3. Send packets by packet generator with different packet size, from 64 bytes to 1518 bytes, check the throughput. + +Test case 4: multiqueue +======================= + +1. One queue. + + 1) Start the testpmd with one queue:: + + ./testpmd -l 29,30 -n 6 --no-pci \ + --vdev net_af_xdp0,iface=enp216s0f0,start_queue=0,queue_count=1 \ + -- -i --nb-cores=1 --rxq=1 --txq=1 --port-topology=loop + + 2) Assign the kernel core:: + + ./set_irq_affinity 34 enp216s0f0 + + 3) Send packets with different dst IP address by packet generator + with different packet size from 64 bytes to 1518 bytes, check the throughput. + +2. Four queues. + + 1) Set hardware queue:: + + ethtool -L enp216s0f0 combined 4 + + 2）Start the testpmd with four queues:: + + ./testpmd -l 29,30-33 -n 6 --no-pci \ + --vdev net_af_xdp0,iface=enp216s0f0,start_queue=0,queue_count=4 \ + -- -i --nb-cores=4 --rxq=4 --txq=4 --port-topology=loop + + 3）Assign the kernel core:: + + ./set_irq_affinity 34-37 enp216s0f0 + + 4）Send packets with different dst IP address by packet generator + with different packet size from 64 bytes to 1518 bytes, check the throughput. + The packets were distributed to the four queues. + +Test case 5: multiqueue and zero copy +===================================== + +1. One queue and zero copy. + + 1) Set hardware queue:: + + ethtool -L enp216s0f0 combined 1 + + 2) Start the testpmd with one queue:: + + ./testpmd -l 29,30 -n 6 --no-pci \ + --vdev net_af_xdp0,iface=enp216s0f0,start_queue=0,queue_count=1,pmd_zero_copy=1 \ + -- -i --nb-cores=1 --rxq=1 --txq=1 --port-topology=loop + + 3) Assign the kernel core:: + + ./set_irq_affinity 34 enp216s0f0 + + 4) Send packets with different dst IP address by packet generator + with different packet size from 64 bytes to 1518 bytes, check the throughput. + Expect the performance is better than non-zero-copy. + +2. Four queues and zero copy. + + 1) Set hardware queue:: + + ethtool -L enp216s0f0 combined 4 + + 2) Start the testpmd with four queues:: + + ./testpmd -l 29,30-33 -n 6 --no-pci \ + --vdev net_af_xdp0,iface=enp216s0f0,start_queue=0,queue_count=4,pmd_zero_copy=1 \ + -- -i --nb-cores=4 --rxq=4 --txq=4 --port-topology=loop + + 3) Assign the kernel core:: + + ./set_irq_affinity 34-37 enp216s0f0 + + 4) Send packets with different dst IP address by packet generator + with different packet size from 64 bytes to 1518 bytes, check the throughput. + The packets were distributed to the four queues. + Expect the performance of four queues is better than one queue. + Expect the performance is better than non-zero-copy. + +Test case 6: need_wakeup +======================== + +1. Set hardware queue:: + + ethtool -L enp216s0f0 combined 1 + +2. Start the testpmd with one queue:: + + ./testpmd -l 29,30 -n 6 --no-pci --vdev net_af_xdp0,iface=enp216s0f0 \ + -- -i --nb-cores=1 --rxq=1 --txq=1 --port-topology=loop + +3. Assign the same core:: + + ./set_irq_affinity 30 enp216s0f0 + +4. Send packets by packet generator with different packet size from 64 bytes + to 1518 bytes, check the throughput. + Expect the performance is better than no need_wakeup. + +Test case 7: xdpsock sample performance +======================================= + +1. One queue. + + 1) Set hardware queue:: + + ethtool -L enp216s0f0 combined 1 + + 2) Start the xdp socket with one queue:: + + #taskset -c 30 ./xdpsock -l -i enp216s0f0 + + 3) Assign the kernel core:: + + ./set_irq_affinity 34 enp216s0f0 + + 4) Send packets with different dst IP address by packet generator + with different packet size from 64 bytes to 1518 bytes, check the throughput. + +2. Four queues. + + 1) Set hardware queue:: + + ethtool -L enp216s0f0 combined 4 + + 2) Start the xdp socket with four queues:: + + #taskset -c 30 ./xdpsock -l -i enp216s0f0 -q 0 + #taskset -c 31 ./xdpsock -l -i enp216s0f0 -q 1 + #taskset -c 32 ./xdpsock -l -i enp216s0f0 -q 2 + #taskset -c 33 ./xdpsock -l -i enp216s0f0 -q 3 + + 3）Assign the kernel core:: + + ./set_irq_affinity 34-37 enp216s0f0 + + 4）Send packets with different dst IP address by packet generator + with different packet size from 64 bytes to 1518 bytes, check the throughput. + The packets were distributed to the four queues. + Expect the performance of four queues is better than one queue. + +3. Need_wakeup. + + 1) Set hardware queue:: + + ethtool -L enp216s0f0 combined 1 + + 2) Start the xdp socket with four queues:: + + #taskset -c 30 ./xdpsock -l -i enp216s0f0 + + 3）Assign the kernel core:: + + ./set_irq_affinity 30 enp216s0f0 + + 4) Send packets by packet generator with different packet size from 64 bytes + to 1518 bytes, check the throughput. + Expect the performance is better than no need_wakeup. -- 2.14.3