* [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect
@ 2019-07-16 1:11 lihong
2019-07-17 6:18 ` Wang, Yinan
2019-08-07 3:44 ` Tu, Lijuan
0 siblings, 2 replies; 4+ messages in thread
From: lihong @ 2019-07-16 1:11 UTC (permalink / raw)
To: dts; +Cc: yinan.wang, lihong
Signed-off-by: lihong <lihongx.ma@intel.com>
---
tests/TestSuite_pvp_vhost_user_reconnect.py | 407 ++++++++++++++++++++++++++++
1 file changed, 407 insertions(+)
create mode 100644 tests/TestSuite_pvp_vhost_user_reconnect.py
diff --git a/tests/TestSuite_pvp_vhost_user_reconnect.py b/tests/TestSuite_pvp_vhost_user_reconnect.py
new file mode 100644
index 0000000..5568376
--- /dev/null
+++ b/tests/TestSuite_pvp_vhost_user_reconnect.py
@@ -0,0 +1,407 @@
+# BSD LICENSE
+#
+# Copyright(c) <2019> Intel Corporation.
+# 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 disclaimer.
+# * Redistributions in binary form must reproduce the above copyright
+# notice, this list of conditions and the following disclaimer in
+# the documentation and/or other materials provided with the
+# distribution.
+# * Neither the name of Intel Corporation nor the names of its
+# contributors may be used to endorse or promote products derived
+# from this software without specific prior written permission.
+#
+# 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 AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 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 CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 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.
+
+"""
+DPDK Test suite.
+
+Vhost reconnect two VM test suite.
+Becase this suite will use the reconnet feature, the VM will start as
+server mode, so the qemu version should greater than 2.7
+"""
+import re
+import utils
+import time
+from test_case import TestCase
+from settings import HEADER_SIZE
+from virt_common import VM
+from pktgen import PacketGeneratorHelper
+
+
+class TestPVPVhostUserReconnect(TestCase):
+
+ def set_up_all(self):
+
+ # Get and verify the ports
+ self.dut_ports = self.dut.get_ports()
+ self.verify(len(self.dut_ports) >= 1, "Insufficient ports for testing")
+
+ # Get the port's socket
+ self.pf = self.dut_ports[0]
+ netdev = self.dut.ports_info[self.pf]['port']
+ self.socket = netdev.get_nic_socket()
+ self.cores = self.dut.get_core_list("1S/2C/1T", socket=self.socket)
+ self.coremask = utils.create_mask(self.cores)
+ self.memory_channel = self.dut.get_memory_channels()
+ self.dst_mac = self.dut.get_mac_address(self.dut_ports[0])
+ # set diff arg about mem_socket base on socket number
+ if len(set([int(core['socket']) for core in self.dut.cores])) == 1:
+ self.socket_mem = '1024'
+ else:
+ self.socket_mem = '1024,1024'
+
+ self.reconnect_times = 5
+ self.vm_num = 1
+ self.frame_sizes = [64, 1518]
+ self.virtio_ip = ["1.1.1.2", "1.1.1.3"]
+ self.virtio_mac = ["52:54:00:00:00:01",
+ "52:54:00:00:00:02"]
+ self.src1 = "192.168.4.1"
+ self.dst1 = "192.168.3.1"
+ self.checked_vm = False
+ self.header_size = HEADER_SIZE['eth'] + HEADER_SIZE['ip'] + HEADER_SIZE['udp']
+
+ # create an instance to set stream field setting
+ self.pktgen_helper = PacketGeneratorHelper()
+
+ def set_up(self):
+ """
+ run before each test case.
+ clear the execution ENV
+ """
+ self.dut.send_expect("killall -s INT testpmd", "# ")
+ self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
+ self.dut.send_expect("rm -rf ./vhost-net*", "# ")
+ self.vhost_user = self.dut.new_session(suite="vhost-user")
+
+ def launch_testpmd_as_vhost_user(self):
+ """
+ launch the testpmd as vhost user
+ """
+ vdev_info = ""
+ for i in range(self.vm_num):
+ vdev_info += "--vdev 'net_vhost%d,iface=vhost-net%d,client=1,queues=1' " % (i, i)
+ self.vhostapp_testcmd = self.dut.base_dir + \
+ "/%s/app/testpmd -c %s -n %d --socket-mem %s --legacy-mem" + \
+ " --file-prefix=vhost %s" + \
+ " -- -i --port-topology=chained --nb-cores=1" + \
+ " --txd=1024 --rxd=1024"
+ self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
+ self.coremask, self.memory_channel,
+ self.socket_mem, vdev_info)
+ self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
+ self.vhost_user.send_expect("set fwd mac", "testpmd> ", 40)
+ self.vhost_user.send_expect("start", "testpmd> ", 40)
+
+ def launch_testpmd_as_vhost_user_with_no_pci(self):
+ """
+ launch the testpmd as vhost user
+ """
+ vdev_info = ""
+ for i in range(self.vm_num):
+ vdev_info += "--vdev 'net_vhost%d,iface=vhost-net%d,client=1,queues=1' " % (i, i)
+ self.vhostapp_testcmd = self.dut.base_dir + \
+ "/%s/app/testpmd -c %s -n %d --socket-mem %s --legacy-mem" + \
+ " --no-pci --file-prefix=vhost %s" + \
+ " -- -i --nb-cores=1 --txd=1024 --rxd=1024"
+ self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
+ self.coremask, self.memory_channel,
+ self.socket_mem, vdev_info)
+ self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
+ self.vhost_user.send_expect("start", "testpmd> ", 40)
+
+ def check_link_status_after_testpmd_start(self, dut_info):
+ """
+ check the link status is up after testpmd start
+ """
+ loop = 1
+ while(loop <= 5):
+ out = dut_info.send_expect("show port info all", "testpmd> ", 120)
+ port_status = re.findall("Link\s*status:\s*([a-z]*)", out)
+ if("down" not in port_status):
+ break
+ time.sleep(3)
+ loop = loop + 1
+
+ self.verify("down" not in port_status, "port can not up after restart")
+
+ def check_qemu_version(self, vm_config):
+ """
+ in this suite, the qemu version should greater 2.7
+ """
+ if self.checked_vm:
+ return
+
+ self.vm_qemu_version = vm_config.qemu_emulator
+ params_number = len(vm_config.params)
+ for i in range(params_number):
+ if vm_config.params[i].keys()[0] == 'qemu':
+ self.vm_qemu_version = vm_config.params[i]['qemu'][0]['path']
+
+ out = self.dut.send_expect("%s --version" % self.vm_qemu_version, "#")
+ result = re.search("QEMU\s*emulator\s*version\s*(\d*.\d*)", out)
+ self.verify(result is not None,
+ 'the qemu path may be not right: %s' % self.vm_qemu_version)
+ version = result.group(1)
+ index = version.find('.')
+ self.verify(int(version[:index]) > 2 or
+ (int(version[:index]) == 2 and int(version[index+1:]) >= 7),
+ 'This qemu version should greater than 2.7 ' + \
+ 'in this suite, please config it in vhost_sample.cfg file')
+ self.checked_vm = True
+
+ def start_vms(self):
+ """
+ start two VM
+ """
+ self.vm_dut = []
+ self.vm = []
+ for i in range(self.vm_num):
+ vm_info = VM(self.dut, 'vm%d' % i, 'vhost_sample')
+ vm_params = {}
+ vm_params['driver'] = 'vhost-user'
+ vm_params['opt_path'] = './vhost-net%d' % (i)
+ vm_params['opt_mac'] = '52:54:00:00:00:0%d' % (i+1)
+ vm_params['opt_server'] = 'server'
+ vm_params['opt_settings'] = 'mrg_rxbuf=on,rx_queue_size=1024,tx_queue_size=1024'
+ vm_info.set_vm_device(**vm_params)
+ self.check_qemu_version(vm_info)
+
+ try:
+ vm_dut = None
+ vm_dut = vm_info.start()
+ if vm_dut is None:
+ raise Exception("Set up VM ENV failed")
+ except Exception as e:
+ print utils.RED("Failure for %s" % str(e))
+ self.verify(vm_dut is not None, "start vm failed")
+ self.vm_dut.append(vm_dut)
+ self.vm.append(vm_info)
+
+ def vm_testpmd_start(self):
+ """
+ start testpmd in vm
+ """
+ vm_testpmd = self.dut.target + "/app/testpmd -c 0x3 -n 4 " + \
+ "-- -i --port-topology=chained --txd=1024 --rxd=1024 "
+ for i in range(len(self.vm_dut)):
+ self.vm_dut[i].send_expect(vm_testpmd, "testpmd> ", 20)
+ self.vm_dut[i].send_expect("set fwd mac", "testpmd> ")
+ self.vm_dut[i].send_expect("start", "testpmd> ")
+
+ self.check_link_status_after_testpmd_start(self.vhost_user)
+
+ def stop_all_apps(self):
+ """
+ quit the testpmd in vm and stop all apps
+ """
+ for i in range(len(self.vm_dut)):
+ self.vm_dut[i].send_expect("stop", "testpmd> ", 20)
+ self.vm_dut[i].send_expect("quit", "# ", 20)
+ self.vm[i].stop()
+ self.vhost_user.send_expect("quit", "# ", 20)
+
+ def config_vm_intf(self):
+ """
+ restore vm interfaces and config intf arp
+ """
+ for i in range(len(self.vm_dut)):
+ self.vm_dut[i].restore_interfaces()
+ time.sleep(5)
+ vm_intf = self.vm_dut[i].ports_info[0]['intf']
+ self.vm_dut[i].send_expect("ifconfig %s %s" %
+ (vm_intf, self.virtio_ip[i]), "#", 10)
+ self.vm_dut[i].send_expect("ifconfig %s up" % vm_intf, "#", 10)
+
+ self.vm_dut[0].send_expect('arp -s %s %s' %
+ (self.virtio_ip[1], self.virtio_mac[1]), '#', 10)
+ self.vm_dut[1].send_expect('arp -s %s %s' %
+ (self.virtio_ip[0], self.virtio_mac[0]), '#', 10)
+
+ def start_iperf(self):
+ """
+ start iperf
+ """
+ self.vm_dut[0].send_expect(
+ 'iperf -s -p 12345 -i 1 > iperf_server.log &', '', 10)
+ self.vm_dut[1].send_expect(
+ 'iperf -c %s -p 12345 -i 1 -t 5 > iperf_client.log &' %
+ self.virtio_ip[0], '', 60)
+ time.sleep(20)
+
+ def iperf_result_verify(self, cycle, tinfo):
+ """
+ verify the Iperf test result
+ """
+ # copy iperf_client file from vm1
+ self.vm_dut[1].session.copy_file_from("%s/iperf_client.log" %
+ self.dut.base_dir)
+ fp = open("./iperf_client.log")
+ fmsg = fp.read()
+ fp.close()
+ iperfdata = re.compile('\S*\s*[M|G]bits/sec').findall(fmsg)
+ self.verify(len(iperfdata) != 0, "The iperf data between to vms is 0")
+ if cycle == 0:
+ cinfo = "Before reconnet"
+ else:
+ cinfo = tinfo
+ self.result_table_add(["vm2vm iperf", iperfdata[-1], cinfo])
+
+ def send_and_verify(self, cycle=0, tinfo=""):
+ for frame_size in self.frame_sizes:
+ payload = frame_size - self.header_size
+ flow = '[Ether(dst="%s")/IP(src="%s",dst="%s")/UDP()/("X"*%d)]' % (
+ self.dst_mac, self.src1, self.dst1, payload)
+ self.tester.scapy_append('wrpcap("reconnect.pcap", %s)' % flow)
+ self.tester.scapy_execute()
+
+ tgenInput = []
+ port = self.tester.get_local_port(self.pf)
+ tgenInput.append((port, port, "reconnect.pcap"))
+
+ self.tester.pktgen.clear_streams()
+ streams = self.pktgen_helper.prepare_stream_from_tginput(tgenInput, 100,
+ None, self.tester.pktgen)
+ traffic_opt = {'delay': 30, }
+ _, pps = self.tester.pktgen.measure_throughput(stream_ids=streams, options=traffic_opt)
+ Mpps = pps / 1000000.0
+ self.verify(Mpps > 0, "can not receive packets of frame size %d" % (frame_size))
+ pct = Mpps * 100 / \
+ float(self.wirespeed(self.nic, frame_size, 1))
+ if cycle == 0:
+ data_row = [tinfo, frame_size, str(Mpps), str(pct),
+ "Before relaunch", "1"]
+ elif cycle == 1:
+ data_row = [tinfo, frame_size, str(Mpps), str(pct),
+ "After relaunch", "1"]
+ self.result_table_add(data_row)
+
+ def test_perf_vhost_user_reconnet_one_vm(self):
+ """
+ test reconnect stability test of one vm
+ """
+ self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
+ "LineRate(%)", "Cycle", "Queue Number"]
+ self.result_table_create(self.header_row)
+ vm_cycle = 0
+ self.vm_num = 1
+ self.launch_testpmd_as_vhost_user()
+ self.start_vms()
+ self.vm_testpmd_start()
+ self.send_and_verify(vm_cycle, "reconnet one vm")
+
+ vm_cycle = 1
+ # reconnet from vhost
+ self.logger.info('now reconnect from vhost')
+ for i in range(self.reconnect_times):
+ self.dut.send_expect("killall -s INT testpmd", "# ")
+ self.launch_testpmd_as_vhost_user()
+ self.send_and_verify(vm_cycle, "reconnet from vhost")
+
+ # reconnet from qemu
+ self.logger.info('now reconnect from vm')
+ for i in range(self.reconnect_times):
+ self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
+ self.start_vms()
+ self.vm_testpmd_start()
+ self.send_and_verify(vm_cycle, "reconnet from VM")
+ self.result_table_print()
+ self.stop_all_apps()
+
+ def test_perf_vhost_user_reconnet_two_vms(self):
+ """
+ test reconnect stability test of two vms
+ """
+ self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
+ "LineRate(%)", "Cycle", "Queue Number"]
+ self.result_table_create(self.header_row)
+ vm_cycle = 0
+ self.vm_num = 2
+ self.launch_testpmd_as_vhost_user()
+ self.start_vms()
+ self.vm_testpmd_start()
+ self.send_and_verify(vm_cycle, "reconnet two vm")
+
+ vm_cycle = 1
+ # reconnet from vhost
+ self.logger.info('now reconnect from vhost')
+ for i in range(self.reconnect_times):
+ self.dut.send_expect("killall -s INT testpmd", "# ")
+ self.launch_testpmd_as_vhost_user()
+ self.send_and_verify(vm_cycle, "reconnet from vhost")
+
+ # reconnet from qemu
+ self.logger.info('now reconnect from vm')
+ for i in range(self.reconnect_times):
+ self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
+ self.start_vms()
+ self.vm_testpmd_start()
+ self.send_and_verify(vm_cycle, "reconnet from VM")
+ self.result_table_print()
+ self.stop_all_apps()
+
+ def test_perf_vhost_vm2vm_virtio_net_reconnet_two_vms(self):
+ """
+ test the iperf traffice can resume after reconnet
+ """
+ self.header_row = ["Mode", "[M|G]bits/sec", "Cycle"]
+ self.result_table_create(self.header_row)
+ self.vm_num = 2
+ vm_cycle = 0
+ self.launch_testpmd_as_vhost_user_with_no_pci()
+ self.start_vms()
+ self.config_vm_intf()
+ self.start_iperf()
+ self.iperf_result_verify(vm_cycle, 'before reconnet')
+
+ vm_cycle = 1
+ # reconnet from vhost
+ self.logger.info('now reconnect from vhost')
+ for i in range(self.reconnect_times):
+ self.dut.send_expect("killall -s INT testpmd", "# ")
+ self.launch_testpmd_as_vhost_user_with_no_pci()
+ self.start_iperf()
+ self.iperf_result_verify(vm_cycle, 'reconnet from vhost')
+
+ # reconnet from VM
+ self.logger.info('now reconnect from vm')
+ for i in range(self.reconnect_times):
+ self.vm_dut[0].send_expect('rm iperf_server.log', '# ', 10)
+ self.vm_dut[1].send_expect('rm iperf_client.log', '# ', 10)
+ self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
+ self.start_vms()
+ self.config_vm_intf()
+ self.start_iperf()
+ self.iperf_result_verify(vm_cycle, 'reconnet from vm')
+ self.result_table_print()
+
+ def tear_down(self):
+ #
+ # Run after each test case.
+ #
+ self.dut.send_expect("killall -s INT testpmd", "# ")
+ self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
+ time.sleep(2)
+
+ def tear_down_all(self):
+ """
+ Run after each test suite.
+ """
+ pass
--
2.7.4
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect
2019-07-16 1:11 [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect lihong
@ 2019-07-17 6:18 ` Wang, Yinan
2019-07-17 6:22 ` Ma, LihongX
2019-08-07 3:44 ` Tu, Lijuan
1 sibling, 1 reply; 4+ messages in thread
From: Wang, Yinan @ 2019-07-17 6:18 UTC (permalink / raw)
To: Ma, LihongX, dts
Acked-by: Wang, Yinan <yinan.wang@intel.com>
> -----Original Message-----
> From: Ma, LihongX
> Sent: 2019年7月16日 9:11
> To: dts@dpdk.org
> Cc: Wang, Yinan <yinan.wang@intel.com>; Ma, LihongX <lihongx.ma@intel.com>
> Subject: [dts][PATCH V1] tests: add testsuite pvp vhost user reconnect
>
> Signed-off-by: lihong <lihongx.ma@intel.com>
> ---
> tests/TestSuite_pvp_vhost_user_reconnect.py | 407
> ++++++++++++++++++++++++++++
> 1 file changed, 407 insertions(+)
> create mode 100644 tests/TestSuite_pvp_vhost_user_reconnect.py
>
> diff --git a/tests/TestSuite_pvp_vhost_user_reconnect.py
> b/tests/TestSuite_pvp_vhost_user_reconnect.py
> new file mode 100644
> index 0000000..5568376
> --- /dev/null
> +++ b/tests/TestSuite_pvp_vhost_user_reconnect.py
> @@ -0,0 +1,407 @@
> +# BSD LICENSE
> +#
> +# Copyright(c) <2019> Intel Corporation.
> +# 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 disclaimer.
> +# * Redistributions in binary form must reproduce the above copyright
> +# notice, this list of conditions and the following disclaimer in
> +# the documentation and/or other materials provided with the
> +# distribution.
> +# * Neither the name of Intel Corporation nor the names of its
> +# contributors may be used to endorse or promote products derived
> +# from this software without specific prior written permission.
> +#
> +# 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 AND FITNESS FOR
> #
> +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 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 CAUSED AND ON
> ANY #
> +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 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.
> +
> +"""
> +DPDK Test suite.
> +
> +Vhost reconnect two VM test suite.
> +Becase this suite will use the reconnet feature, the VM will start as
> +server mode, so the qemu version should greater than 2.7 """
> +import re
> +import utils
> +import time
> +from test_case import TestCase
> +from settings import HEADER_SIZE
> +from virt_common import VM
> +from pktgen import PacketGeneratorHelper
> +
> +
> +class TestPVPVhostUserReconnect(TestCase):
> +
> + def set_up_all(self):
> +
> + # Get and verify the ports
> + self.dut_ports = self.dut.get_ports()
> + self.verify(len(self.dut_ports) >= 1, "Insufficient ports for
> + testing")
> +
> + # Get the port's socket
> + self.pf = self.dut_ports[0]
> + netdev = self.dut.ports_info[self.pf]['port']
> + self.socket = netdev.get_nic_socket()
> + self.cores = self.dut.get_core_list("1S/2C/1T", socket=self.socket)
> + self.coremask = utils.create_mask(self.cores)
> + self.memory_channel = self.dut.get_memory_channels()
> + self.dst_mac = self.dut.get_mac_address(self.dut_ports[0])
> + # set diff arg about mem_socket base on socket number
> + if len(set([int(core['socket']) for core in self.dut.cores])) == 1:
> + self.socket_mem = '1024'
> + else:
> + self.socket_mem = '1024,1024'
> +
> + self.reconnect_times = 5
> + self.vm_num = 1
> + self.frame_sizes = [64, 1518]
> + self.virtio_ip = ["1.1.1.2", "1.1.1.3"]
> + self.virtio_mac = ["52:54:00:00:00:01",
> + "52:54:00:00:00:02"]
> + self.src1 = "192.168.4.1"
> + self.dst1 = "192.168.3.1"
> + self.checked_vm = False
> + self.header_size = HEADER_SIZE['eth'] + HEADER_SIZE['ip'] +
> + HEADER_SIZE['udp']
> +
> + # create an instance to set stream field setting
> + self.pktgen_helper = PacketGeneratorHelper()
> +
> + def set_up(self):
> + """
> + run before each test case.
> + clear the execution ENV
> + """
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.dut.send_expect("rm -rf ./vhost-net*", "# ")
> + self.vhost_user = self.dut.new_session(suite="vhost-user")
> +
> + def launch_testpmd_as_vhost_user(self):
> + """
> + launch the testpmd as vhost user
> + """
> + vdev_info = ""
> + for i in range(self.vm_num):
> + vdev_info += "--vdev
> 'net_vhost%d,iface=vhost-net%d,client=1,queues=1' " % (i, i)
> + self.vhostapp_testcmd = self.dut.base_dir + \
> + "/%s/app/testpmd -c %s -n %d --socket-mem %s
> --legacy-mem" + \
> + " --file-prefix=vhost %s" + \
> + " -- -i --port-topology=chained --nb-cores=1" + \
> + " --txd=1024 --rxd=1024"
> + self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
> + self.coremask, self.memory_channel,
> + self.socket_mem, vdev_info)
> + self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
> + self.vhost_user.send_expect("set fwd mac", "testpmd> ", 40)
> + self.vhost_user.send_expect("start", "testpmd> ", 40)
> +
> + def launch_testpmd_as_vhost_user_with_no_pci(self):
> + """
> + launch the testpmd as vhost user
> + """
> + vdev_info = ""
> + for i in range(self.vm_num):
> + vdev_info += "--vdev
> 'net_vhost%d,iface=vhost-net%d,client=1,queues=1' " % (i, i)
> + self.vhostapp_testcmd = self.dut.base_dir + \
> + "/%s/app/testpmd -c %s -n %d --socket-mem %s
> --legacy-mem" + \
> + " --no-pci --file-prefix=vhost %s" + \
> + " -- -i --nb-cores=1 --txd=1024 --rxd=1024"
> + self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
> + self.coremask, self.memory_channel,
> + self.socket_mem, vdev_info)
> + self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
> + self.vhost_user.send_expect("start", "testpmd> ", 40)
> +
> + def check_link_status_after_testpmd_start(self, dut_info):
> + """
> + check the link status is up after testpmd start
> + """
> + loop = 1
> + while(loop <= 5):
> + out = dut_info.send_expect("show port info all", "testpmd> ",
> 120)
> + port_status = re.findall("Link\s*status:\s*([a-z]*)", out)
> + if("down" not in port_status):
> + break
> + time.sleep(3)
> + loop = loop + 1
> +
> + self.verify("down" not in port_status, "port can not up after
> + restart")
> +
> + def check_qemu_version(self, vm_config):
> + """
> + in this suite, the qemu version should greater 2.7
> + """
> + if self.checked_vm:
> + return
> +
> + self.vm_qemu_version = vm_config.qemu_emulator
> + params_number = len(vm_config.params)
> + for i in range(params_number):
> + if vm_config.params[i].keys()[0] == 'qemu':
> + self.vm_qemu_version =
> + vm_config.params[i]['qemu'][0]['path']
> +
> + out = self.dut.send_expect("%s --version" % self.vm_qemu_version,
> "#")
> + result = re.search("QEMU\s*emulator\s*version\s*(\d*.\d*)", out)
> + self.verify(result is not None,
> + 'the qemu path may be not right: %s' %
> self.vm_qemu_version)
> + version = result.group(1)
> + index = version.find('.')
> + self.verify(int(version[:index]) > 2 or
> + (int(version[:index]) == 2 and int(version[index+1:]) >=
> 7),
> + 'This qemu version should greater than 2.7 ' + \
> + 'in this suite, please config it in vhost_sample.cfg file')
> + self.checked_vm = True
> +
> + def start_vms(self):
> + """
> + start two VM
> + """
> + self.vm_dut = []
> + self.vm = []
> + for i in range(self.vm_num):
> + vm_info = VM(self.dut, 'vm%d' % i, 'vhost_sample')
> + vm_params = {}
> + vm_params['driver'] = 'vhost-user'
> + vm_params['opt_path'] = './vhost-net%d' % (i)
> + vm_params['opt_mac'] = '52:54:00:00:00:0%d' % (i+1)
> + vm_params['opt_server'] = 'server'
> + vm_params['opt_settings'] =
> 'mrg_rxbuf=on,rx_queue_size=1024,tx_queue_size=1024'
> + vm_info.set_vm_device(**vm_params)
> + self.check_qemu_version(vm_info)
> +
> + try:
> + vm_dut = None
> + vm_dut = vm_info.start()
> + if vm_dut is None:
> + raise Exception("Set up VM ENV failed")
> + except Exception as e:
> + print utils.RED("Failure for %s" % str(e))
> + self.verify(vm_dut is not None, "start vm failed")
> + self.vm_dut.append(vm_dut)
> + self.vm.append(vm_info)
> +
> + def vm_testpmd_start(self):
> + """
> + start testpmd in vm
> + """
> + vm_testpmd = self.dut.target + "/app/testpmd -c 0x3 -n 4 " + \
> + "-- -i --port-topology=chained --txd=1024 --rxd=1024
> "
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].send_expect(vm_testpmd, "testpmd> ", 20)
> + self.vm_dut[i].send_expect("set fwd mac", "testpmd> ")
> + self.vm_dut[i].send_expect("start", "testpmd> ")
> +
> + self.check_link_status_after_testpmd_start(self.vhost_user)
> +
> + def stop_all_apps(self):
> + """
> + quit the testpmd in vm and stop all apps
> + """
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].send_expect("stop", "testpmd> ", 20)
> + self.vm_dut[i].send_expect("quit", "# ", 20)
> + self.vm[i].stop()
> + self.vhost_user.send_expect("quit", "# ", 20)
> +
> + def config_vm_intf(self):
> + """
> + restore vm interfaces and config intf arp
> + """
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].restore_interfaces()
> + time.sleep(5)
> + vm_intf = self.vm_dut[i].ports_info[0]['intf']
> + self.vm_dut[i].send_expect("ifconfig %s %s" %
> + (vm_intf, self.virtio_ip[i]), "#", 10)
> + self.vm_dut[i].send_expect("ifconfig %s up" % vm_intf, "#",
> + 10)
> +
> + self.vm_dut[0].send_expect('arp -s %s %s' %
> + (self.virtio_ip[1], self.virtio_mac[1]), '#',
> 10)
> + self.vm_dut[1].send_expect('arp -s %s %s' %
> + (self.virtio_ip[0],
> + self.virtio_mac[0]), '#', 10)
> +
> + def start_iperf(self):
> + """
> + start iperf
> + """
> + self.vm_dut[0].send_expect(
> + 'iperf -s -p 12345 -i 1 > iperf_server.log &', '', 10)
> + self.vm_dut[1].send_expect(
> + 'iperf -c %s -p 12345 -i 1 -t 5 > iperf_client.log &' %
> + self.virtio_ip[0], '', 60)
> + time.sleep(20)
> +
> + def iperf_result_verify(self, cycle, tinfo):
> + """
> + verify the Iperf test result
> + """
> + # copy iperf_client file from vm1
> + self.vm_dut[1].session.copy_file_from("%s/iperf_client.log" %
> + self.dut.base_dir)
> + fp = open("./iperf_client.log")
> + fmsg = fp.read()
> + fp.close()
> + iperfdata = re.compile('\S*\s*[M|G]bits/sec').findall(fmsg)
> + self.verify(len(iperfdata) != 0, "The iperf data between to vms is 0")
> + if cycle == 0:
> + cinfo = "Before reconnet"
> + else:
> + cinfo = tinfo
> + self.result_table_add(["vm2vm iperf", iperfdata[-1], cinfo])
> +
> + def send_and_verify(self, cycle=0, tinfo=""):
> + for frame_size in self.frame_sizes:
> + payload = frame_size - self.header_size
> + flow =
> '[Ether(dst="%s")/IP(src="%s",dst="%s")/UDP()/("X"*%d)]' % (
> + self.dst_mac, self.src1, self.dst1, payload)
> + self.tester.scapy_append('wrpcap("reconnect.pcap", %s)' % flow)
> + self.tester.scapy_execute()
> +
> + tgenInput = []
> + port = self.tester.get_local_port(self.pf)
> + tgenInput.append((port, port, "reconnect.pcap"))
> +
> + self.tester.pktgen.clear_streams()
> + streams =
> self.pktgen_helper.prepare_stream_from_tginput(tgenInput, 100,
> + None, self.tester.pktgen)
> + traffic_opt = {'delay': 30, }
> + _, pps =
> self.tester.pktgen.measure_throughput(stream_ids=streams,
> options=traffic_opt)
> + Mpps = pps / 1000000.0
> + self.verify(Mpps > 0, "can not receive packets of frame size %d" %
> (frame_size))
> + pct = Mpps * 100 / \
> + float(self.wirespeed(self.nic, frame_size, 1))
> + if cycle == 0:
> + data_row = [tinfo, frame_size, str(Mpps), str(pct),
> + "Before relaunch", "1"]
> + elif cycle == 1:
> + data_row = [tinfo, frame_size, str(Mpps), str(pct),
> + "After relaunch", "1"]
> + self.result_table_add(data_row)
> +
> + def test_perf_vhost_user_reconnet_one_vm(self):
> + """
> + test reconnect stability test of one vm
> + """
> + self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
> + "LineRate(%)", "Cycle", "Queue Number"]
> + self.result_table_create(self.header_row)
> + vm_cycle = 0
> + self.vm_num = 1
> + self.launch_testpmd_as_vhost_user()
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet one vm")
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user()
> + self.send_and_verify(vm_cycle, "reconnet from vhost")
> +
> + # reconnet from qemu
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet from VM")
> + self.result_table_print()
> + self.stop_all_apps()
> +
> + def test_perf_vhost_user_reconnet_two_vms(self):
> + """
> + test reconnect stability test of two vms
> + """
> + self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
> + "LineRate(%)", "Cycle", "Queue Number"]
> + self.result_table_create(self.header_row)
> + vm_cycle = 0
> + self.vm_num = 2
> + self.launch_testpmd_as_vhost_user()
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet two vm")
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user()
> + self.send_and_verify(vm_cycle, "reconnet from vhost")
> +
> + # reconnet from qemu
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet from VM")
> + self.result_table_print()
> + self.stop_all_apps()
> +
> + def test_perf_vhost_vm2vm_virtio_net_reconnet_two_vms(self):
> + """
> + test the iperf traffice can resume after reconnet
> + """
> + self.header_row = ["Mode", "[M|G]bits/sec", "Cycle"]
> + self.result_table_create(self.header_row)
> + self.vm_num = 2
> + vm_cycle = 0
> + self.launch_testpmd_as_vhost_user_with_no_pci()
> + self.start_vms()
> + self.config_vm_intf()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'before reconnet')
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user_with_no_pci()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'reconnet from vhost')
> +
> + # reconnet from VM
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.vm_dut[0].send_expect('rm iperf_server.log', '# ', 10)
> + self.vm_dut[1].send_expect('rm iperf_client.log', '# ', 10)
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.config_vm_intf()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'reconnet from vm')
> + self.result_table_print()
> +
> + def tear_down(self):
> + #
> + # Run after each test case.
> + #
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + time.sleep(2)
> +
> + def tear_down_all(self):
> + """
> + Run after each test suite.
> + """
> + pass
> --
> 2.7.4
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect
2019-07-17 6:18 ` Wang, Yinan
@ 2019-07-17 6:22 ` Ma, LihongX
0 siblings, 0 replies; 4+ messages in thread
From: Ma, LihongX @ 2019-07-17 6:22 UTC (permalink / raw)
To: Wang, Yinan, dts
Tested-by: lihong<lihongx.ma@intel.com>
-----Original Message-----
From: Wang, Yinan
Sent: Wednesday, July 17, 2019 2:18 PM
To: Ma, LihongX <lihongx.ma@intel.com>; dts@dpdk.org
Subject: RE: [dts][PATCH V1] tests: add testsuite pvp vhost user reconnect
Acked-by: Wang, Yinan <yinan.wang@intel.com>
> -----Original Message-----
> From: Ma, LihongX
> Sent: 2019年7月16日 9:11
> To: dts@dpdk.org
> Cc: Wang, Yinan <yinan.wang@intel.com>; Ma, LihongX
> <lihongx.ma@intel.com>
> Subject: [dts][PATCH V1] tests: add testsuite pvp vhost user reconnect
>
> Signed-off-by: lihong <lihongx.ma@intel.com>
> ---
> tests/TestSuite_pvp_vhost_user_reconnect.py | 407
> ++++++++++++++++++++++++++++
> 1 file changed, 407 insertions(+)
> create mode 100644 tests/TestSuite_pvp_vhost_user_reconnect.py
>
> diff --git a/tests/TestSuite_pvp_vhost_user_reconnect.py
> b/tests/TestSuite_pvp_vhost_user_reconnect.py
> new file mode 100644
> index 0000000..5568376
> --- /dev/null
> +++ b/tests/TestSuite_pvp_vhost_user_reconnect.py
> @@ -0,0 +1,407 @@
> +# BSD LICENSE
> +#
> +# Copyright(c) <2019> Intel Corporation.
> +# 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 disclaimer.
> +# * Redistributions in binary form must reproduce the above copyright
> +# notice, this list of conditions and the following disclaimer in
> +# the documentation and/or other materials provided with the
> +# distribution.
> +# * Neither the name of Intel Corporation nor the names of its
> +# contributors may be used to endorse or promote products derived
> +# from this software without specific prior written permission.
> +#
> +# 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 AND FITNESS FOR
> #
> +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 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 CAUSED AND ON
> ANY #
> +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 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.
> +
> +"""
> +DPDK Test suite.
> +
> +Vhost reconnect two VM test suite.
> +Becase this suite will use the reconnet feature, the VM will start as
> +server mode, so the qemu version should greater than 2.7 """
> +import re
> +import utils
> +import time
> +from test_case import TestCase
> +from settings import HEADER_SIZE
> +from virt_common import VM
> +from pktgen import PacketGeneratorHelper
> +
> +
> +class TestPVPVhostUserReconnect(TestCase):
> +
> + def set_up_all(self):
> +
> + # Get and verify the ports
> + self.dut_ports = self.dut.get_ports()
> + self.verify(len(self.dut_ports) >= 1, "Insufficient ports for
> + testing")
> +
> + # Get the port's socket
> + self.pf = self.dut_ports[0]
> + netdev = self.dut.ports_info[self.pf]['port']
> + self.socket = netdev.get_nic_socket()
> + self.cores = self.dut.get_core_list("1S/2C/1T", socket=self.socket)
> + self.coremask = utils.create_mask(self.cores)
> + self.memory_channel = self.dut.get_memory_channels()
> + self.dst_mac = self.dut.get_mac_address(self.dut_ports[0])
> + # set diff arg about mem_socket base on socket number
> + if len(set([int(core['socket']) for core in self.dut.cores])) == 1:
> + self.socket_mem = '1024'
> + else:
> + self.socket_mem = '1024,1024'
> +
> + self.reconnect_times = 5
> + self.vm_num = 1
> + self.frame_sizes = [64, 1518]
> + self.virtio_ip = ["1.1.1.2", "1.1.1.3"]
> + self.virtio_mac = ["52:54:00:00:00:01",
> + "52:54:00:00:00:02"]
> + self.src1 = "192.168.4.1"
> + self.dst1 = "192.168.3.1"
> + self.checked_vm = False
> + self.header_size = HEADER_SIZE['eth'] + HEADER_SIZE['ip'] +
> + HEADER_SIZE['udp']
> +
> + # create an instance to set stream field setting
> + self.pktgen_helper = PacketGeneratorHelper()
> +
> + def set_up(self):
> + """
> + run before each test case.
> + clear the execution ENV
> + """
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.dut.send_expect("rm -rf ./vhost-net*", "# ")
> + self.vhost_user = self.dut.new_session(suite="vhost-user")
> +
> + def launch_testpmd_as_vhost_user(self):
> + """
> + launch the testpmd as vhost user
> + """
> + vdev_info = ""
> + for i in range(self.vm_num):
> + vdev_info += "--vdev
> 'net_vhost%d,iface=vhost-net%d,client=1,queues=1' " % (i, i)
> + self.vhostapp_testcmd = self.dut.base_dir + \
> + "/%s/app/testpmd -c %s -n %d --socket-mem %s
> --legacy-mem" + \
> + " --file-prefix=vhost %s" + \
> + " -- -i --port-topology=chained --nb-cores=1" + \
> + " --txd=1024 --rxd=1024"
> + self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
> + self.coremask, self.memory_channel,
> + self.socket_mem, vdev_info)
> + self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
> + self.vhost_user.send_expect("set fwd mac", "testpmd> ", 40)
> + self.vhost_user.send_expect("start", "testpmd> ", 40)
> +
> + def launch_testpmd_as_vhost_user_with_no_pci(self):
> + """
> + launch the testpmd as vhost user
> + """
> + vdev_info = ""
> + for i in range(self.vm_num):
> + vdev_info += "--vdev
> 'net_vhost%d,iface=vhost-net%d,client=1,queues=1' " % (i, i)
> + self.vhostapp_testcmd = self.dut.base_dir + \
> + "/%s/app/testpmd -c %s -n %d --socket-mem %s
> --legacy-mem" + \
> + " --no-pci --file-prefix=vhost %s" + \
> + " -- -i --nb-cores=1 --txd=1024 --rxd=1024"
> + self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
> + self.coremask, self.memory_channel,
> + self.socket_mem, vdev_info)
> + self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
> + self.vhost_user.send_expect("start", "testpmd> ", 40)
> +
> + def check_link_status_after_testpmd_start(self, dut_info):
> + """
> + check the link status is up after testpmd start
> + """
> + loop = 1
> + while(loop <= 5):
> + out = dut_info.send_expect("show port info all",
> + "testpmd> ",
> 120)
> + port_status = re.findall("Link\s*status:\s*([a-z]*)", out)
> + if("down" not in port_status):
> + break
> + time.sleep(3)
> + loop = loop + 1
> +
> + self.verify("down" not in port_status, "port can not up after
> + restart")
> +
> + def check_qemu_version(self, vm_config):
> + """
> + in this suite, the qemu version should greater 2.7
> + """
> + if self.checked_vm:
> + return
> +
> + self.vm_qemu_version = vm_config.qemu_emulator
> + params_number = len(vm_config.params)
> + for i in range(params_number):
> + if vm_config.params[i].keys()[0] == 'qemu':
> + self.vm_qemu_version =
> + vm_config.params[i]['qemu'][0]['path']
> +
> + out = self.dut.send_expect("%s --version" %
> + self.vm_qemu_version,
> "#")
> + result = re.search("QEMU\s*emulator\s*version\s*(\d*.\d*)", out)
> + self.verify(result is not None,
> + 'the qemu path may be not right: %s' %
> self.vm_qemu_version)
> + version = result.group(1)
> + index = version.find('.')
> + self.verify(int(version[:index]) > 2 or
> + (int(version[:index]) == 2 and
> + int(version[index+1:]) >=
> 7),
> + 'This qemu version should greater than 2.7 ' + \
> + 'in this suite, please config it in vhost_sample.cfg file')
> + self.checked_vm = True
> +
> + def start_vms(self):
> + """
> + start two VM
> + """
> + self.vm_dut = []
> + self.vm = []
> + for i in range(self.vm_num):
> + vm_info = VM(self.dut, 'vm%d' % i, 'vhost_sample')
> + vm_params = {}
> + vm_params['driver'] = 'vhost-user'
> + vm_params['opt_path'] = './vhost-net%d' % (i)
> + vm_params['opt_mac'] = '52:54:00:00:00:0%d' % (i+1)
> + vm_params['opt_server'] = 'server'
> + vm_params['opt_settings'] =
> 'mrg_rxbuf=on,rx_queue_size=1024,tx_queue_size=1024'
> + vm_info.set_vm_device(**vm_params)
> + self.check_qemu_version(vm_info)
> +
> + try:
> + vm_dut = None
> + vm_dut = vm_info.start()
> + if vm_dut is None:
> + raise Exception("Set up VM ENV failed")
> + except Exception as e:
> + print utils.RED("Failure for %s" % str(e))
> + self.verify(vm_dut is not None, "start vm failed")
> + self.vm_dut.append(vm_dut)
> + self.vm.append(vm_info)
> +
> + def vm_testpmd_start(self):
> + """
> + start testpmd in vm
> + """
> + vm_testpmd = self.dut.target + "/app/testpmd -c 0x3 -n 4 " + \
> + "-- -i --port-topology=chained --txd=1024
> + --rxd=1024
> "
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].send_expect(vm_testpmd, "testpmd> ", 20)
> + self.vm_dut[i].send_expect("set fwd mac", "testpmd> ")
> + self.vm_dut[i].send_expect("start", "testpmd> ")
> +
> + self.check_link_status_after_testpmd_start(self.vhost_user)
> +
> + def stop_all_apps(self):
> + """
> + quit the testpmd in vm and stop all apps
> + """
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].send_expect("stop", "testpmd> ", 20)
> + self.vm_dut[i].send_expect("quit", "# ", 20)
> + self.vm[i].stop()
> + self.vhost_user.send_expect("quit", "# ", 20)
> +
> + def config_vm_intf(self):
> + """
> + restore vm interfaces and config intf arp
> + """
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].restore_interfaces()
> + time.sleep(5)
> + vm_intf = self.vm_dut[i].ports_info[0]['intf']
> + self.vm_dut[i].send_expect("ifconfig %s %s" %
> + (vm_intf, self.virtio_ip[i]), "#", 10)
> + self.vm_dut[i].send_expect("ifconfig %s up" % vm_intf,
> + "#",
> + 10)
> +
> + self.vm_dut[0].send_expect('arp -s %s %s' %
> + (self.virtio_ip[1],
> + self.virtio_mac[1]), '#',
> 10)
> + self.vm_dut[1].send_expect('arp -s %s %s' %
> + (self.virtio_ip[0],
> + self.virtio_mac[0]), '#', 10)
> +
> + def start_iperf(self):
> + """
> + start iperf
> + """
> + self.vm_dut[0].send_expect(
> + 'iperf -s -p 12345 -i 1 > iperf_server.log &', '', 10)
> + self.vm_dut[1].send_expect(
> + 'iperf -c %s -p 12345 -i 1 -t 5 > iperf_client.log &' %
> + self.virtio_ip[0], '', 60)
> + time.sleep(20)
> +
> + def iperf_result_verify(self, cycle, tinfo):
> + """
> + verify the Iperf test result
> + """
> + # copy iperf_client file from vm1
> + self.vm_dut[1].session.copy_file_from("%s/iperf_client.log" %
> + self.dut.base_dir)
> + fp = open("./iperf_client.log")
> + fmsg = fp.read()
> + fp.close()
> + iperfdata = re.compile('\S*\s*[M|G]bits/sec').findall(fmsg)
> + self.verify(len(iperfdata) != 0, "The iperf data between to vms is 0")
> + if cycle == 0:
> + cinfo = "Before reconnet"
> + else:
> + cinfo = tinfo
> + self.result_table_add(["vm2vm iperf", iperfdata[-1], cinfo])
> +
> + def send_and_verify(self, cycle=0, tinfo=""):
> + for frame_size in self.frame_sizes:
> + payload = frame_size - self.header_size
> + flow =
> '[Ether(dst="%s")/IP(src="%s",dst="%s")/UDP()/("X"*%d)]' % (
> + self.dst_mac, self.src1, self.dst1, payload)
> + self.tester.scapy_append('wrpcap("reconnect.pcap", %s)' % flow)
> + self.tester.scapy_execute()
> +
> + tgenInput = []
> + port = self.tester.get_local_port(self.pf)
> + tgenInput.append((port, port, "reconnect.pcap"))
> +
> + self.tester.pktgen.clear_streams()
> + streams =
> self.pktgen_helper.prepare_stream_from_tginput(tgenInput, 100,
> + None, self.tester.pktgen)
> + traffic_opt = {'delay': 30, }
> + _, pps =
> self.tester.pktgen.measure_throughput(stream_ids=streams,
> options=traffic_opt)
> + Mpps = pps / 1000000.0
> + self.verify(Mpps > 0, "can not receive packets of frame
> + size %d" %
> (frame_size))
> + pct = Mpps * 100 / \
> + float(self.wirespeed(self.nic, frame_size, 1))
> + if cycle == 0:
> + data_row = [tinfo, frame_size, str(Mpps), str(pct),
> + "Before relaunch", "1"]
> + elif cycle == 1:
> + data_row = [tinfo, frame_size, str(Mpps), str(pct),
> + "After relaunch", "1"]
> + self.result_table_add(data_row)
> +
> + def test_perf_vhost_user_reconnet_one_vm(self):
> + """
> + test reconnect stability test of one vm
> + """
> + self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
> + "LineRate(%)", "Cycle", "Queue Number"]
> + self.result_table_create(self.header_row)
> + vm_cycle = 0
> + self.vm_num = 1
> + self.launch_testpmd_as_vhost_user()
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet one vm")
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user()
> + self.send_and_verify(vm_cycle, "reconnet from vhost")
> +
> + # reconnet from qemu
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet from VM")
> + self.result_table_print()
> + self.stop_all_apps()
> +
> + def test_perf_vhost_user_reconnet_two_vms(self):
> + """
> + test reconnect stability test of two vms
> + """
> + self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
> + "LineRate(%)", "Cycle", "Queue Number"]
> + self.result_table_create(self.header_row)
> + vm_cycle = 0
> + self.vm_num = 2
> + self.launch_testpmd_as_vhost_user()
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet two vm")
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user()
> + self.send_and_verify(vm_cycle, "reconnet from vhost")
> +
> + # reconnet from qemu
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet from VM")
> + self.result_table_print()
> + self.stop_all_apps()
> +
> + def test_perf_vhost_vm2vm_virtio_net_reconnet_two_vms(self):
> + """
> + test the iperf traffice can resume after reconnet
> + """
> + self.header_row = ["Mode", "[M|G]bits/sec", "Cycle"]
> + self.result_table_create(self.header_row)
> + self.vm_num = 2
> + vm_cycle = 0
> + self.launch_testpmd_as_vhost_user_with_no_pci()
> + self.start_vms()
> + self.config_vm_intf()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'before reconnet')
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user_with_no_pci()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'reconnet from vhost')
> +
> + # reconnet from VM
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.vm_dut[0].send_expect('rm iperf_server.log', '# ', 10)
> + self.vm_dut[1].send_expect('rm iperf_client.log', '# ', 10)
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.config_vm_intf()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'reconnet from vm')
> + self.result_table_print()
> +
> + def tear_down(self):
> + #
> + # Run after each test case.
> + #
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + time.sleep(2)
> +
> + def tear_down_all(self):
> + """
> + Run after each test suite.
> + """
> + pass
> --
> 2.7.4
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect
2019-07-16 1:11 [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect lihong
2019-07-17 6:18 ` Wang, Yinan
@ 2019-08-07 3:44 ` Tu, Lijuan
1 sibling, 0 replies; 4+ messages in thread
From: Tu, Lijuan @ 2019-08-07 3:44 UTC (permalink / raw)
To: Ma, LihongX, dts; +Cc: Wang, Yinan, Ma, LihongX
Applied, thanks
> -----Original Message-----
> From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of lihong
> Sent: Tuesday, July 16, 2019 9:11 AM
> To: dts@dpdk.org
> Cc: Wang, Yinan <yinan.wang@intel.com>; Ma, LihongX
> <lihongx.ma@intel.com>
> Subject: [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect
>
> Signed-off-by: lihong <lihongx.ma@intel.com>
> ---
> tests/TestSuite_pvp_vhost_user_reconnect.py | 407
> ++++++++++++++++++++++++++++
> 1 file changed, 407 insertions(+)
> create mode 100644 tests/TestSuite_pvp_vhost_user_reconnect.py
>
> diff --git a/tests/TestSuite_pvp_vhost_user_reconnect.py
> b/tests/TestSuite_pvp_vhost_user_reconnect.py
> new file mode 100644
> index 0000000..5568376
> --- /dev/null
> +++ b/tests/TestSuite_pvp_vhost_user_reconnect.py
> @@ -0,0 +1,407 @@
> +# BSD LICENSE
> +#
> +# Copyright(c) <2019> Intel Corporation.
> +# 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 disclaimer.
> +# * Redistributions in binary form must reproduce the above copyright
> +# notice, this list of conditions and the following disclaimer in
> +# the documentation and/or other materials provided with the
> +# distribution.
> +# * Neither the name of Intel Corporation nor the names of its
> +# contributors may be used to endorse or promote products derived
> +# from this software without specific prior written permission.
> +#
> +# 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 AND FITNESS
> FOR #
> +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> 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 CAUSED AND
> ON ANY #
> +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 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.
> +
> +"""
> +DPDK Test suite.
> +
> +Vhost reconnect two VM test suite.
> +Becase this suite will use the reconnet feature, the VM will start as
> +server mode, so the qemu version should greater than 2.7 """
> +import re
> +import utils
> +import time
> +from test_case import TestCase
> +from settings import HEADER_SIZE
> +from virt_common import VM
> +from pktgen import PacketGeneratorHelper
> +
> +
> +class TestPVPVhostUserReconnect(TestCase):
> +
> + def set_up_all(self):
> +
> + # Get and verify the ports
> + self.dut_ports = self.dut.get_ports()
> + self.verify(len(self.dut_ports) >= 1, "Insufficient ports for
> + testing")
> +
> + # Get the port's socket
> + self.pf = self.dut_ports[0]
> + netdev = self.dut.ports_info[self.pf]['port']
> + self.socket = netdev.get_nic_socket()
> + self.cores = self.dut.get_core_list("1S/2C/1T", socket=self.socket)
> + self.coremask = utils.create_mask(self.cores)
> + self.memory_channel = self.dut.get_memory_channels()
> + self.dst_mac = self.dut.get_mac_address(self.dut_ports[0])
> + # set diff arg about mem_socket base on socket number
> + if len(set([int(core['socket']) for core in self.dut.cores])) == 1:
> + self.socket_mem = '1024'
> + else:
> + self.socket_mem = '1024,1024'
> +
> + self.reconnect_times = 5
> + self.vm_num = 1
> + self.frame_sizes = [64, 1518]
> + self.virtio_ip = ["1.1.1.2", "1.1.1.3"]
> + self.virtio_mac = ["52:54:00:00:00:01",
> + "52:54:00:00:00:02"]
> + self.src1 = "192.168.4.1"
> + self.dst1 = "192.168.3.1"
> + self.checked_vm = False
> + self.header_size = HEADER_SIZE['eth'] + HEADER_SIZE['ip'] +
> + HEADER_SIZE['udp']
> +
> + # create an instance to set stream field setting
> + self.pktgen_helper = PacketGeneratorHelper()
> +
> + def set_up(self):
> + """
> + run before each test case.
> + clear the execution ENV
> + """
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.dut.send_expect("rm -rf ./vhost-net*", "# ")
> + self.vhost_user = self.dut.new_session(suite="vhost-user")
> +
> + def launch_testpmd_as_vhost_user(self):
> + """
> + launch the testpmd as vhost user
> + """
> + vdev_info = ""
> + for i in range(self.vm_num):
> + vdev_info += "--vdev 'net_vhost%d,iface=vhost-
> net%d,client=1,queues=1' " % (i, i)
> + self.vhostapp_testcmd = self.dut.base_dir + \
> + "/%s/app/testpmd -c %s -n %d --socket-mem %s --legacy-mem"
> + \
> + " --file-prefix=vhost %s" + \
> + " -- -i --port-topology=chained --nb-cores=1" + \
> + " --txd=1024 --rxd=1024"
> + self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
> + self.coremask, self.memory_channel,
> + self.socket_mem, vdev_info)
> + self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
> + self.vhost_user.send_expect("set fwd mac", "testpmd> ", 40)
> + self.vhost_user.send_expect("start", "testpmd> ", 40)
> +
> + def launch_testpmd_as_vhost_user_with_no_pci(self):
> + """
> + launch the testpmd as vhost user
> + """
> + vdev_info = ""
> + for i in range(self.vm_num):
> + vdev_info += "--vdev 'net_vhost%d,iface=vhost-
> net%d,client=1,queues=1' " % (i, i)
> + self.vhostapp_testcmd = self.dut.base_dir + \
> + "/%s/app/testpmd -c %s -n %d --socket-mem %s --legacy-mem"
> + \
> + " --no-pci --file-prefix=vhost %s" + \
> + " -- -i --nb-cores=1 --txd=1024 --rxd=1024"
> + self.vhostapp_testcmd = self.vhostapp_testcmd % (self.target,
> + self.coremask, self.memory_channel,
> + self.socket_mem, vdev_info)
> + self.vhost_user.send_expect(self.vhostapp_testcmd, "testpmd> ", 40)
> + self.vhost_user.send_expect("start", "testpmd> ", 40)
> +
> + def check_link_status_after_testpmd_start(self, dut_info):
> + """
> + check the link status is up after testpmd start
> + """
> + loop = 1
> + while(loop <= 5):
> + out = dut_info.send_expect("show port info all", "testpmd> ", 120)
> + port_status = re.findall("Link\s*status:\s*([a-z]*)", out)
> + if("down" not in port_status):
> + break
> + time.sleep(3)
> + loop = loop + 1
> +
> + self.verify("down" not in port_status, "port can not up after
> + restart")
> +
> + def check_qemu_version(self, vm_config):
> + """
> + in this suite, the qemu version should greater 2.7
> + """
> + if self.checked_vm:
> + return
> +
> + self.vm_qemu_version = vm_config.qemu_emulator
> + params_number = len(vm_config.params)
> + for i in range(params_number):
> + if vm_config.params[i].keys()[0] == 'qemu':
> + self.vm_qemu_version =
> + vm_config.params[i]['qemu'][0]['path']
> +
> + out = self.dut.send_expect("%s --version" % self.vm_qemu_version, "#")
> + result = re.search("QEMU\s*emulator\s*version\s*(\d*.\d*)", out)
> + self.verify(result is not None,
> + 'the qemu path may be not right: %s' % self.vm_qemu_version)
> + version = result.group(1)
> + index = version.find('.')
> + self.verify(int(version[:index]) > 2 or
> + (int(version[:index]) == 2 and int(version[index+1:]) >= 7),
> + 'This qemu version should greater than 2.7 ' + \
> + 'in this suite, please config it in vhost_sample.cfg file')
> + self.checked_vm = True
> +
> + def start_vms(self):
> + """
> + start two VM
> + """
> + self.vm_dut = []
> + self.vm = []
> + for i in range(self.vm_num):
> + vm_info = VM(self.dut, 'vm%d' % i, 'vhost_sample')
> + vm_params = {}
> + vm_params['driver'] = 'vhost-user'
> + vm_params['opt_path'] = './vhost-net%d' % (i)
> + vm_params['opt_mac'] = '52:54:00:00:00:0%d' % (i+1)
> + vm_params['opt_server'] = 'server'
> + vm_params['opt_settings'] =
> 'mrg_rxbuf=on,rx_queue_size=1024,tx_queue_size=1024'
> + vm_info.set_vm_device(**vm_params)
> + self.check_qemu_version(vm_info)
> +
> + try:
> + vm_dut = None
> + vm_dut = vm_info.start()
> + if vm_dut is None:
> + raise Exception("Set up VM ENV failed")
> + except Exception as e:
> + print utils.RED("Failure for %s" % str(e))
> + self.verify(vm_dut is not None, "start vm failed")
> + self.vm_dut.append(vm_dut)
> + self.vm.append(vm_info)
> +
> + def vm_testpmd_start(self):
> + """
> + start testpmd in vm
> + """
> + vm_testpmd = self.dut.target + "/app/testpmd -c 0x3 -n 4 " + \
> + "-- -i --port-topology=chained --txd=1024 --rxd=1024 "
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].send_expect(vm_testpmd, "testpmd> ", 20)
> + self.vm_dut[i].send_expect("set fwd mac", "testpmd> ")
> + self.vm_dut[i].send_expect("start", "testpmd> ")
> +
> + self.check_link_status_after_testpmd_start(self.vhost_user)
> +
> + def stop_all_apps(self):
> + """
> + quit the testpmd in vm and stop all apps
> + """
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].send_expect("stop", "testpmd> ", 20)
> + self.vm_dut[i].send_expect("quit", "# ", 20)
> + self.vm[i].stop()
> + self.vhost_user.send_expect("quit", "# ", 20)
> +
> + def config_vm_intf(self):
> + """
> + restore vm interfaces and config intf arp
> + """
> + for i in range(len(self.vm_dut)):
> + self.vm_dut[i].restore_interfaces()
> + time.sleep(5)
> + vm_intf = self.vm_dut[i].ports_info[0]['intf']
> + self.vm_dut[i].send_expect("ifconfig %s %s" %
> + (vm_intf, self.virtio_ip[i]), "#", 10)
> + self.vm_dut[i].send_expect("ifconfig %s up" % vm_intf, "#",
> + 10)
> +
> + self.vm_dut[0].send_expect('arp -s %s %s' %
> + (self.virtio_ip[1], self.virtio_mac[1]), '#', 10)
> + self.vm_dut[1].send_expect('arp -s %s %s' %
> + (self.virtio_ip[0],
> + self.virtio_mac[0]), '#', 10)
> +
> + def start_iperf(self):
> + """
> + start iperf
> + """
> + self.vm_dut[0].send_expect(
> + 'iperf -s -p 12345 -i 1 > iperf_server.log &', '', 10)
> + self.vm_dut[1].send_expect(
> + 'iperf -c %s -p 12345 -i 1 -t 5 > iperf_client.log &' %
> + self.virtio_ip[0], '', 60)
> + time.sleep(20)
> +
> + def iperf_result_verify(self, cycle, tinfo):
> + """
> + verify the Iperf test result
> + """
> + # copy iperf_client file from vm1
> + self.vm_dut[1].session.copy_file_from("%s/iperf_client.log" %
> + self.dut.base_dir)
> + fp = open("./iperf_client.log")
> + fmsg = fp.read()
> + fp.close()
> + iperfdata = re.compile('\S*\s*[M|G]bits/sec').findall(fmsg)
> + self.verify(len(iperfdata) != 0, "The iperf data between to vms is 0")
> + if cycle == 0:
> + cinfo = "Before reconnet"
> + else:
> + cinfo = tinfo
> + self.result_table_add(["vm2vm iperf", iperfdata[-1], cinfo])
> +
> + def send_and_verify(self, cycle=0, tinfo=""):
> + for frame_size in self.frame_sizes:
> + payload = frame_size - self.header_size
> + flow = '[Ether(dst="%s")/IP(src="%s",dst="%s")/UDP()/("X"*%d)]' % (
> + self.dst_mac, self.src1, self.dst1, payload)
> + self.tester.scapy_append('wrpcap("reconnect.pcap", %s)' % flow)
> + self.tester.scapy_execute()
> +
> + tgenInput = []
> + port = self.tester.get_local_port(self.pf)
> + tgenInput.append((port, port, "reconnect.pcap"))
> +
> + self.tester.pktgen.clear_streams()
> + streams =
> self.pktgen_helper.prepare_stream_from_tginput(tgenInput, 100,
> + None, self.tester.pktgen)
> + traffic_opt = {'delay': 30, }
> + _, pps = self.tester.pktgen.measure_throughput(stream_ids=streams,
> options=traffic_opt)
> + Mpps = pps / 1000000.0
> + self.verify(Mpps > 0, "can not receive packets of frame size %d" %
> (frame_size))
> + pct = Mpps * 100 / \
> + float(self.wirespeed(self.nic, frame_size, 1))
> + if cycle == 0:
> + data_row = [tinfo, frame_size, str(Mpps), str(pct),
> + "Before relaunch", "1"]
> + elif cycle == 1:
> + data_row = [tinfo, frame_size, str(Mpps), str(pct),
> + "After relaunch", "1"]
> + self.result_table_add(data_row)
> +
> + def test_perf_vhost_user_reconnet_one_vm(self):
> + """
> + test reconnect stability test of one vm
> + """
> + self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
> + "LineRate(%)", "Cycle", "Queue Number"]
> + self.result_table_create(self.header_row)
> + vm_cycle = 0
> + self.vm_num = 1
> + self.launch_testpmd_as_vhost_user()
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet one vm")
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user()
> + self.send_and_verify(vm_cycle, "reconnet from vhost")
> +
> + # reconnet from qemu
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet from VM")
> + self.result_table_print()
> + self.stop_all_apps()
> +
> + def test_perf_vhost_user_reconnet_two_vms(self):
> + """
> + test reconnect stability test of two vms
> + """
> + self.header_row = ["Mode", "FrameSize(B)", "Throughput(Mpps)",
> + "LineRate(%)", "Cycle", "Queue Number"]
> + self.result_table_create(self.header_row)
> + vm_cycle = 0
> + self.vm_num = 2
> + self.launch_testpmd_as_vhost_user()
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet two vm")
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user()
> + self.send_and_verify(vm_cycle, "reconnet from vhost")
> +
> + # reconnet from qemu
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.vm_testpmd_start()
> + self.send_and_verify(vm_cycle, "reconnet from VM")
> + self.result_table_print()
> + self.stop_all_apps()
> +
> + def test_perf_vhost_vm2vm_virtio_net_reconnet_two_vms(self):
> + """
> + test the iperf traffice can resume after reconnet
> + """
> + self.header_row = ["Mode", "[M|G]bits/sec", "Cycle"]
> + self.result_table_create(self.header_row)
> + self.vm_num = 2
> + vm_cycle = 0
> + self.launch_testpmd_as_vhost_user_with_no_pci()
> + self.start_vms()
> + self.config_vm_intf()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'before reconnet')
> +
> + vm_cycle = 1
> + # reconnet from vhost
> + self.logger.info('now reconnect from vhost')
> + for i in range(self.reconnect_times):
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.launch_testpmd_as_vhost_user_with_no_pci()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'reconnet from vhost')
> +
> + # reconnet from VM
> + self.logger.info('now reconnect from vm')
> + for i in range(self.reconnect_times):
> + self.vm_dut[0].send_expect('rm iperf_server.log', '# ', 10)
> + self.vm_dut[1].send_expect('rm iperf_client.log', '# ', 10)
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + self.start_vms()
> + self.config_vm_intf()
> + self.start_iperf()
> + self.iperf_result_verify(vm_cycle, 'reconnet from vm')
> + self.result_table_print()
> +
> + def tear_down(self):
> + #
> + # Run after each test case.
> + #
> + self.dut.send_expect("killall -s INT testpmd", "# ")
> + self.dut.send_expect("killall -s INT qemu-system-x86_64", "# ")
> + time.sleep(2)
> +
> + def tear_down_all(self):
> + """
> + Run after each test suite.
> + """
> + pass
> --
> 2.7.4
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2019-08-07 3:44 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-07-16 1:11 [dts] [PATCH V1] tests: add testsuite pvp vhost user reconnect lihong
2019-07-17 6:18 ` Wang, Yinan
2019-07-17 6:22 ` Ma, LihongX
2019-08-07 3:44 ` Tu, Lijuan
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).