Add support to use vswitch to handle V2V in PVVP SRIOV scenario

1. Add support to use vswitch to handle V2V in PVVP SRIOV scenario
2. Update nfvbenchvm to 0.5:
    (1) Update VPP to 17.10;
    (2) Update DPDK testpmd to 17.08;
    (3) Change kernel to based on longterm lineup;

Change-Id: I944489579a4cd92d17075e80870bbdb32512a150
Signed-off-by: Yichen Wang <yicwang@cisco.com>

13 files changed, 99 insertions, 41 deletions

diff --git a/nfvbench/cfg.default.yaml b/nfvbench/cfg.default.yaml
index 83dd5ac..a8bdc2b 100644
--- a/nfvbench/cfg.default.yaml
+++ b/nfvbench/cfg.default.yaml
@@ -253,6 +253,11 @@ internal_networks:
         segmentation_id:
         physical_network:
 
+# In the scenario of PVVP + SRIOV, there is choice of how the traffic will be
+# handled in the middle network. The default (false) will use vswitch, while
+# SRIOV can be used by toggling below setting.
+use_sriov_middle_net: false
+
 # EXT chain only. Names of edge networks which will be used to send traffic via traffic generator.
 external_networks:
     left: 'nfvbench-net0'
@@ -400,4 +405,4 @@ factory_class: 'BasicFactory'
 
 # Custom label added for every perf record generated during this run.
 # Can be overriden by --user-label
-user_label:
\ No newline at end of file
+user_label:
diff --git a/nfvbench/chain_clients.py b/nfvbench/chain_clients.py
index d9a39af..57b15ee 100644
--- a/nfvbench/chain_clients.py
+++ b/nfvbench/chain_clients.py
@@ -145,11 +145,11 @@ class BasicStageClient(object):
         LOG.info('Created network: %s.', name)
         return network
 
-    def _create_port(self, net):
+    def _create_port(self, net, vnic_type='normal'):
         body = {
             "port": {
                 'network_id': net['id'],
-                'binding:vnic_type': 'direct' if self.config.sriov else 'normal'
+                'binding:vnic_type': vnic_type
             }
         }
         port = self.neutron.create_port(body)
@@ -305,7 +305,7 @@ class BasicStageClient(object):
         else:
             LOG.error('Unable to delete flavor: %s', self.config.flavor_type)
 
-    def get_config_file(self, chain_index, src_mac, dst_mac):
+    def get_config_file(self, chain_index, src_mac, dst_mac, intf_mac1, intf_mac2):
         boot_script_file = os.path.join(os.path.dirname(os.path.abspath(__file__)),
                                         'nfvbenchvm/', self.nfvbenchvm_config_name)
 
@@ -317,6 +317,8 @@ class BasicStageClient(object):
 
         vm_config = {
             'forwarder': self.config.vm_forwarder,
+            'intf_mac1': intf_mac1,
+            'intf_mac2': intf_mac2,
             'tg_gateway1_ip': self.config.traffic_generator.tg_gateway_ip_addrs[0],
             'tg_gateway2_ip': self.config.traffic_generator.tg_gateway_ip_addrs[1],
             'tg_net1': self.config.traffic_generator.ip_addrs[0],
@@ -479,11 +481,13 @@ class PVPStageClient(BasicStageClient):
             if reusable_vm:
                 self.vms.append(reusable_vm)
             else:
+                vnic_type = 'direct' if self.config.sriov else 'normal'
+                ports = [self._create_port(net, vnic_type) for net in self.nets]
                 config_file = self.get_config_file(chain_index,
                                                    self.config.generator_config.src_device.mac,
-                                                   self.config.generator_config.dst_device.mac)
-
-                ports = [self._create_port(net) for net in self.nets]
+                                                   self.config.generator_config.dst_device.mac,
+                                                   ports[0]['mac_address'],
+                                                   ports[1]['mac_address'])
                 self.created_ports.extend(ports)
                 self.vms.append(self._create_server(name, ports, az, config_file))
         self._ensure_vms_active()
@@ -542,11 +546,15 @@ class PVVPStageClient(BasicStageClient):
             if reusable_vm0 and reusable_vm1:
                 self.vms.extend([reusable_vm0, reusable_vm1])
             else:
-                vm0_port_net0 = self._create_port(vm0_nets[0])
-                vm0_port_net2 = self._create_port(vm0_nets[1])
+                edge_vnic_type = 'direct' if self.config.sriov else 'normal'
+                middle_vnic_type = 'direct' \
+                    if self.config.sriov and self.config.use_sriov_middle_net \
+                    else 'normal'
+                vm0_port_net0 = self._create_port(vm0_nets[0], edge_vnic_type)
+                vm0_port_net2 = self._create_port(vm0_nets[1], middle_vnic_type)
 
-                vm1_port_net2 = self._create_port(vm1_nets[1])
-                vm1_port_net1 = self._create_port(vm1_nets[0])
+                vm1_port_net2 = self._create_port(vm1_nets[1], middle_vnic_type)
+                vm1_port_net1 = self._create_port(vm1_nets[0], edge_vnic_type)
 
                 self.created_ports.extend([vm0_port_net0,
                                            vm0_port_net2,
@@ -558,10 +566,14 @@ class PVVPStageClient(BasicStageClient):
                 # TG0 (net0) -> VM0 (net2) -> VM1 (net2) -> TG1 (net1)
                 config_file0 = self.get_config_file(chain_index,
                                                     self.config.generator_config.src_device.mac,
-                                                    vm1_port_net2['mac_address'])
+                                                    vm1_port_net2['mac_address'],
+                                                    vm0_port_net0['mac_address'],
+                                                    vm0_port_net2['mac_address'])
                 config_file1 = self.get_config_file(chain_index,
                                                     vm0_port_net2['mac_address'],
-                                                    self.config.generator_config.dst_device.mac)
+                                                    self.config.generator_config.dst_device.mac,
+                                                    vm1_port_net2['mac_address'],
+                                                    vm1_port_net1['mac_address'])
 
                 self.vms.append(self._create_server(name0,
                                                     [vm0_port_net0, vm0_port_net2],
diff --git a/nfvbench/nfvbench.py b/nfvbench/nfvbench.py
index d1bd0d9..4c9f56c 100644
--- a/nfvbench/nfvbench.py
+++ b/nfvbench/nfvbench.py
@@ -284,6 +284,12 @@ def parse_opts_from_cli():
                         action='store_true',
                         help='Use SRIOV (no vswitch - requires SRIOV support in compute nodes)')
 
+    parser.add_argument('--use-sriov-middle-net', dest='use_sriov_middle_net',
+                        default=None,
+                        action='store_true',
+                        help='Use SRIOV to handle the middle network traffic '
+                             '(PVVP with SRIOV only)')
+
     parser.add_argument('-d', '--debug', dest='debug',
                         action='store_true',
                         default=None,
@@ -491,20 +497,29 @@ def main():
             config.sriov = True
         if opts.log_file:
             config.log_file = opts.log_file
+        if opts.service_chain:
+            config.service_chain = opts.service_chain
+        if opts.service_chain_count:
+            config.service_chain_count = opts.service_chain_count
 
-        # show running config in json format
-        if opts.show_config:
-            print json.dumps(config, sort_keys=True, indent=4)
-            sys.exit(0)
+        if opts.use_sriov_middle_net:
+            if (not config.sriov) or (not config.service_chain == ChainType.PVVP):
+                raise Exception("--use-sriov-middle-net is only valid for PVVP with SRIOV")
+            config.use_sriov_middle_net = True
 
         if config.sriov and config.service_chain != ChainType.EXT:
             # if sriov is requested (does not apply to ext chains)
             # make sure the physnet names are specified
             check_physnet("left", config.internal_networks.left)
             check_physnet("right", config.internal_networks.right)
-            if config.service_chain == ChainType.PVVP:
+            if config.service_chain == ChainType.PVVP and config.use_sriov_middle_net:
                 check_physnet("middle", config.internal_networks.middle)
 
+        # show running config in json format
+        if opts.show_config:
+            print json.dumps(config, sort_keys=True, indent=4)
+            sys.exit(0)
+
         # update the config in the config plugin as it might have changed
         # in a copy of the dict (config plugin still holds the original dict)
         config_plugin.set_config(config)
diff --git a/nfvbench/nfvbenchvm/nfvbenchvm.conf b/nfvbench/nfvbenchvm/nfvbenchvm.conf
index 0b76244..3bc6ace 100644
--- a/nfvbench/nfvbenchvm/nfvbenchvm.conf
+++ b/nfvbench/nfvbenchvm/nfvbenchvm.conf
@@ -1,4 +1,6 @@
 FORWARDER={forwarder}
+INTF_MAC1={intf_mac1}
+INTF_MAC2={intf_mac2}
 TG_MAC1={tg_mac1}
 TG_MAC2={tg_mac2}
 VNF_GATEWAY1_CIDR={vnf_gateway1_cidr}
diff --git a/nfvbenchvm/README.rst b/nfvbenchvm/README.rst
index baa6996..1bf0bbf 100644
--- a/nfvbenchvm/README.rst
+++ b/nfvbenchvm/README.rst
@@ -1,4 +1,4 @@
-NFVBENCH VM IMAGE FOR OPENSTACK 
+NFVBENCH VM IMAGE FOR OPENSTACK
 +++++++++++++++++++++++++++++++
 
 This repo will build a centos 7 image with testpmd and VPP installed.
@@ -18,7 +18,7 @@ Pre-requisites
 Build the image
 ---------------
 - cd dib
-- update the version number for the image (if needed) by modifying __version__ in build-image.sh 
+- update the version number for the image (if needed) by modifying __version__ in build-image.sh
 - setup your http_proxy if needed
 - bash build-image.sh
 
@@ -48,15 +48,17 @@ nfvbenchvm config file is located at ``/etc/nfvbenchvm.conf``.
 
 .. code-block:: bash
 
-    FORWARDER=VPP
-    TG_MAC0=00:10:94:00:0A:00
-    TG_MAC1=00:11:94:00:0A:00
-    VNF1_GATEWAY_CIDR=1.1.0.2/8
-    VNF2_GATEWAY_CIDR=2.2.0.2/8
-    TG1_NET=10.0.0.0/8
-    TG2_NET=20.0.0.0/8
-    TG1_GATEWAY_IP=1.1.0.100
-    TG1_GATEWAY_IP=2.2.0.100
+    FORWARDER=testpmd
+    INTF_MAC1=FA:16:3E:A2:30:41
+    INTF_MAC2=FA:16:3E:10:DA:10
+    TG_MAC1=00:10:94:00:0A:00
+    TG_MAC2=00:11:94:00:0A:00
+    VNF_GATEWAY1_CIDR=1.1.0.2/8
+    VNF_GATEWAY2_CIDR=2.2.0.2/8
+    TG_NET1=10.0.0.0/8
+    TG_NET2=20.0.0.0/8
+    TG_GATEWAY1_IP=1.1.0.100
+    TG_GATEWAY2_IP=2.2.0.100
 
 
 Launching nfvbenchvm VM
@@ -79,6 +81,6 @@ To check if VPP is running, you can run this command in VNC console:
 
 Hardcoded Username and Password
 --------------------------------
-- Username: nfvbench 
-- Password: nfvbench 
+- Username: nfvbench
+- Password: nfvbench
 
diff --git a/nfvbenchvm/dib/build-image.sh b/nfvbenchvm/dib/build-image.sh
index 9326762..605db54 100755
--- a/nfvbenchvm/dib/build-image.sh
+++ b/nfvbenchvm/dib/build-image.sh
@@ -11,7 +11,7 @@ set -e
 gs_url=artifacts.opnfv.org/nfvbench/images
 
 # image version number
-__version__=0.4
+__version__=0.5
 image_name=nfvbenchvm_centos-$__version__
 
 # if image exists skip building
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/fdio-release.repo b/nfvbenchvm/dib/elements/nfvbenchvm/fdio-release.repo
index a26aa42..3ad12fb 100644
--- a/nfvbenchvm/dib/elements/nfvbenchvm/fdio-release.repo
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/fdio-release.repo
@@ -1,5 +1,5 @@
 [fdio-release]
 name=fd.io release branch latest merge
-baseurl=https://nexus.fd.io/content/repositories/fd.io.stable.1707.centos7/
+baseurl=https://nexus.fd.io/content/repositories/fd.io.stable.1710.centos7/
 enabled=1
 gpgcheck=0
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/package-installs.yaml b/nfvbenchvm/dib/elements/nfvbenchvm/package-installs.yaml
index 36e0196..e3184c7 100644
--- a/nfvbenchvm/dib/elements/nfvbenchvm/package-installs.yaml
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/package-installs.yaml
@@ -6,6 +6,8 @@ screen:
 telnet:
 python-devel:
 libyaml-devel:
+numactl-libs:
+numactl-devel:
 vpp:
 vpp-plugins:
 kernel-firmware:
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/01-update-kernel b/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/01-update-kernel
index d884f79..8094006 100755
--- a/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/01-update-kernel
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/01-update-kernel
@@ -8,7 +8,7 @@ fi
 rpm --import https://www.elrepo.org/RPM-GPG-KEY-elrepo.org
 rpm -Uvh http://www.elrepo.org/elrepo-release-7.0-2.el7.elrepo.noarch.rpm
 yum remove -y kernel-firmware kernel-headers kernel-devel
-yum install -y --enablerepo=elrepo-kernel kernel-ml kernel-ml-headers kernel-ml-devel
+yum install -y --enablerepo=elrepo-kernel kernel-lt kernel-lt-headers kernel-lt-devel
 
 # gcc will be removed with old kernel as dependency, so reinstalling it back
 yum install -y gcc
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/02-testpmd-script b/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/02-testpmd-script
index acdc6a3..2136c3a 100755
--- a/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/02-testpmd-script
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/02-testpmd-script
@@ -1,7 +1,7 @@
 #!/bin/bash
 
-DPDK=dpdk-17.05.2
-DPDK_UNTAR=dpdk-stable-17.05.2
+DPDK=dpdk-17.08
+DPDK_UNTAR=dpdk-17.08
 
 # pick up the kernel version for the target image
 kernel_version=`ls -t /lib/modules | awk 'NR==1 {print}'`
@@ -19,6 +19,7 @@ cp usertools/dpdk-devbind.py ../dpdk
 # cp tools/dpdk_nic_bind.py ../dpdk/dpdk-devbind.py
 cp x86_64-native-linuxapp-gcc/app/testpmd ../dpdk
 cp x86_64-native-linuxapp-gcc/kmod/igb_uio.ko ../dpdk
+echo "set promisc all off" > /dpdk/testpmd_cmd.txt
 
 cd ..
 rm -f $DPDK.tar.xz
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/99-cleanup b/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/99-cleanup
index 83d4fc5..14e9f27 100755
--- a/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/99-cleanup
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/post-install.d/99-cleanup
@@ -1,3 +1,3 @@
 #!/bin/bash
 
-yum erase -y python-devel libyaml-devel kernel-devel kernel-headers kernel-ml-headers kernel-ml-devel gcc
+yum erase -y python-devel libyaml-devel numactl-devel kernel-devel kernel-headers kernel-lt-headers kernel-lt-devel gcc
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/static/etc/rc.d/rc.local b/nfvbenchvm/dib/elements/nfvbenchvm/static/etc/rc.d/rc.local
index 596cfdb..8aab1ae 100644
--- a/nfvbenchvm/dib/elements/nfvbenchvm/static/etc/rc.d/rc.local
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/static/etc/rc.d/rc.local
@@ -34,6 +34,25 @@ for irq in `ls /proc/irq/`; do
 done
 tuna -c $(seq -s, 1 1 $WORKER_CORES) --isolate
 
+# Sometimes the interfaces on the loopback VM will use different drivers, e.g.
+# one from vswitch which is virtio based, one is from SRIOV VF. In this case,
+# we have to make sure the forwarder uses them in the right order, which is
+# especially important if the VM is in a PVVP chain.
+if [ -z $INTF_MAC1 ] && [ -z $INTF_MAC2 ]; then
+    NET_PATH=/sys/class/net
+    EXP_INTF_1=$(for f in $(ls $NET_PATH/); do if [ ! $(grep -o "$INTF_MAC1" $NET_PATH/$f/address) ]; then break; fi; done)
+    EXP_PCI_ADDRESS_1=$(basename $(readlink $NET_PATH/$EXP_INTF_1/device))
+    EXP_INTF_2=$(for f in $(ls $NET_PATH/); do if [ ! $(grep -o "$INTF_MAC2" $NET_PATH/$f/address) ]; then break; fi; done)
+    EXP_PCI_ADDRESS_2=$(basename $(readlink $NET_PATH/$EXP_INTF_2/device))
+    if [ "$PCI_ADDRESS_1" == "$EXP_PCI_ADDRESS_2" ] && [ "$PCI_ADDRESS_2" == "$EXP_PCI_ADDRESS_1" ]; then
+        # Interfaces are not coming in the expected order:
+        #     (1) Swap the MAC in the case of testpmd;
+        #     (2) Swap the interface order in the case of VPP;
+        TEMP=$PCI_ADDRESS_1; PCI_ADDRESS_1=$PCI_ADDRESS_2; PCI_ADDRESS_2=$TEMP
+        TEMP=$TG_MAC1; TG_MAC1=$TG_MAC2; TG_MAC2=$TEMP
+    fi
+fi
+
 # Configure the forwarder
 if [ "$FORWARDER" == "testpmd" ]; then
     echo "Configuring testpmd..."
@@ -56,8 +75,9 @@ if [ "$FORWARDER" == "testpmd" ]; then
                              --eth-peer=0,$TG_MAC1 \
                              --eth-peer=1,$TG_MAC2 \
                              --forward-mode=mac \
-                             --nb-cores=$WORKER_CORES\
-                             --max-pkt-len 9000
+                             --nb-cores=$WORKER_CORES \
+                             --max-pkt-len=9000 \
+                             --cmdline-file=/dpdk/testpmd_cmd.txt
 else
     echo "Configuring vpp..."
     cp /vpp/startup.conf /etc/vpp/startup.conf
@@ -81,6 +101,5 @@ else
     sed -i "s/{{TG_NET2}}/${TG_NET2//\//\/}/g" /etc/vpp/vm.conf
     sed -i "s/{{TG_GATEWAY1_IP}}/${TG_GATEWAY1_IP}/g" /etc/vpp/vm.conf
     sed -i "s/{{TG_GATEWAY2_IP}}/${TG_GATEWAY2_IP}/g" /etc/vpp/vm.conf
-
     service vpp restart
 fi
diff --git a/nfvbenchvm/dib/elements/nfvbenchvm/static/vpp/startup.conf b/nfvbenchvm/dib/elements/nfvbenchvm/static/vpp/startup.conf
index 811eee1..4306fe9 100644
--- a/nfvbenchvm/dib/elements/nfvbenchvm/static/vpp/startup.conf
+++ b/nfvbenchvm/dib/elements/nfvbenchvm/static/vpp/startup.conf
@@ -3,7 +3,7 @@ unix {
   log /tmp/vpp.log
   full-coredump
   startup-config /etc/vpp/vm.conf
-  cli-listen localhost:5002
+  cli-listen /run/vpp/cli.sock
 }
 
 api-trace {