2 Licensed under the Apache License, Version 2.0 (the "License"); you may
3 not use this file except in compliance with the License. You may obtain
4 a copy of the License at
6 http://www.apache.org/licenses/LICENSE-2.0
8 Unless required by applicable law or agreed to in writing, software
9 distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
10 WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
11 License for the specific language governing permissions and limitations
14 Convention for heading levels in Open vSwitch documentation:
16 ======= Heading 0 (reserved for the title in a document)
22 Avoid deeper levels because they do not render well.
24 ============================
25 Using Open vSwitch with DPDK
26 ============================
28 This document describes how to use Open vSwitch with DPDK datapath.
32 Using the DPDK datapath requires building OVS with DPDK support. Refer to
33 :doc:`/intro/install/dpdk` for more information.
38 ovs-vsctl can be used to set up bridges and other Open vSwitch features.
39 Bridges should be created with a ``datapath_type=netdev``::
41 $ ovs-vsctl add-br br0 -- set bridge br0 datapath_type=netdev
43 ovs-vsctl can also be used to add DPDK devices. ovs-vswitchd should print the
44 number of dpdk devices found in the log file::
46 $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
47 options:dpdk-devargs=0000:01:00.0
48 $ ovs-vsctl add-port br0 dpdk-p1 -- set Interface dpdk-p1 type=dpdk \
49 options:dpdk-devargs=0000:01:00.1
51 After the DPDK ports get added to switch, a polling thread continuously polls
52 DPDK devices and consumes 100% of the core, as can be checked from ``top`` and
56 $ ps -eLo pid,psr,comm | grep pmd
58 Creating bonds of DPDK interfaces is slightly different to creating bonds of
59 system interfaces. For DPDK, the interface type and devargs must be explicitly
62 $ ovs-vsctl add-bond br0 dpdkbond p0 p1 \
63 -- set Interface p0 type=dpdk options:dpdk-devargs=0000:01:00.0 \
64 -- set Interface p1 type=dpdk options:dpdk-devargs=0000:01:00.1
66 To stop ovs-vswitchd & delete bridge, run::
68 $ ovs-appctl -t ovs-vswitchd exit
69 $ ovs-appctl -t ovsdb-server exit
70 $ ovs-vsctl del-br br0
75 To show current stats::
77 $ ovs-appctl dpif-netdev/pmd-stats-show
79 To clear previous stats::
81 $ ovs-appctl dpif-netdev/pmd-stats-clear
83 Port/RXQ Assigment to PMD Threads
84 ---------------------------------
86 To show port/rxq assignment::
88 $ ovs-appctl dpif-netdev/pmd-rxq-show
90 To change default rxq assignment to pmd threads, rxqs may be manually pinned to
93 $ ovs-vsctl set Interface <iface> \
94 other_config:pmd-rxq-affinity=<rxq-affinity-list>
98 - ``<rxq-affinity-list>`` is a CSV list of ``<queue-id>:<core-id>`` values
102 $ ovs-vsctl set interface dpdk-p0 options:n_rxq=4 \
103 other_config:pmd-rxq-affinity="0:3,1:7,3:8"
107 - Queue #0 pinned to core 3
108 - Queue #1 pinned to core 7
109 - Queue #2 not pinned
110 - Queue #3 pinned to core 8
112 After that PMD threads on cores where RX queues was pinned will become
113 ``isolated``. This means that this thread will poll only pinned RX queues.
116 If there are no ``non-isolated`` PMD threads, ``non-pinned`` RX queues will
117 not be polled. Also, if provided ``core_id`` is not available (ex. this
118 ``core_id`` not in ``pmd-cpu-mask``), RX queue will not be polled by any PMD
121 If pmd-rxq-affinity is not set for rxqs, they will be assigned to pmds (cores)
122 automatically. The processing cycles that have been stored for each rxq
123 will be used where known to assign rxqs to pmd based on a round robin of the
126 For example, in the case where here there are 5 rxqs and 3 cores (e.g. 3,7,8)
127 available, and the measured usage of core cycles per rxq over the last
128 interval is seen to be:
136 The rxqs will be assigned to cores 3,7,8 in the following order:
139 Core 7: Q4 (70%) | Q5 (10%)
140 core 8: Q3 (60%) | Q0 (30%)
142 Rxq to pmds assignment takes place whenever there are configuration changes.
147 Assuming you have a vhost-user port transmitting traffic consisting of packets
148 of size 64 bytes, the following command would limit the egress transmission
149 rate of the port to ~1,000,000 packets per second::
151 $ ovs-vsctl set port vhost-user0 qos=@newqos -- \
152 --id=@newqos create qos type=egress-policer other-config:cir=46000000 \
153 other-config:cbs=2048`
155 To examine the QoS configuration of the port, run::
157 $ ovs-appctl -t ovs-vswitchd qos/show vhost-user0
159 To clear the QoS configuration from the port and ovsdb, run::
161 $ ovs-vsctl destroy QoS vhost-user0 -- clear Port vhost-user0 qos
163 Refer to vswitch.xml for more details on egress-policer.
168 Here is an example on Ingress Policing usage. Assuming you have a vhost-user
169 port receiving traffic consisting of packets of size 64 bytes, the following
170 command would limit the reception rate of the port to ~1,000,000 packets per
173 $ ovs-vsctl set interface vhost-user0 ingress_policing_rate=368000 \
174 ingress_policing_burst=1000`
176 To examine the ingress policer configuration of the port::
178 $ ovs-vsctl list interface vhost-user0
180 To clear the ingress policer configuration from the port::
182 $ ovs-vsctl set interface vhost-user0 ingress_policing_rate=0
184 Refer to vswitch.xml for more details on ingress-policer.
189 Flow control can be enabled only on DPDK physical ports. To enable flow control
190 support at tx side while adding a port, run::
192 $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
193 options:dpdk-devargs=0000:01:00.0 options:tx-flow-ctrl=true
195 Similarly, to enable rx flow control, run::
197 $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
198 options:dpdk-devargs=0000:01:00.0 options:rx-flow-ctrl=true
200 To enable flow control auto-negotiation, run::
202 $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
203 options:dpdk-devargs=0000:01:00.0 options:flow-ctrl-autoneg=true
205 To turn ON the tx flow control at run time for an existing port, run::
207 $ ovs-vsctl set Interface dpdk-p0 options:tx-flow-ctrl=true
209 The flow control parameters can be turned off by setting ``false`` to the
210 respective parameter. To disable the flow control at tx side, run::
212 $ ovs-vsctl set Interface dpdk-p0 options:tx-flow-ctrl=false
217 pdump allows you to listen on DPDK ports and view the traffic that is passing
218 on them. To use this utility, one must have libpcap installed on the system.
219 Furthermore, DPDK must be built with ``CONFIG_RTE_LIBRTE_PDUMP=y`` and
220 ``CONFIG_RTE_LIBRTE_PMD_PCAP=y``.
223 A performance decrease is expected when using a monitoring application like
226 To use pdump, simply launch OVS as usual, then navigate to the ``app/pdump``
227 directory in DPDK, ``make`` the application and run like so::
229 $ sudo ./build/app/dpdk-pdump -- \
230 --pdump port=0,queue=0,rx-dev=/tmp/pkts.pcap \
231 --server-socket-path=/usr/local/var/run/openvswitch
233 The above command captures traffic received on queue 0 of port 0 and stores it
234 in ``/tmp/pkts.pcap``. Other combinations of port numbers, queues numbers and
235 pcap locations are of course also available to use. For example, to capture all
236 packets that traverse port 0 in a single pcap file::
238 $ sudo ./build/app/dpdk-pdump -- \
239 --pdump 'port=0,queue=*,rx-dev=/tmp/pkts.pcap,tx-dev=/tmp/pkts.pcap' \
240 --server-socket-path=/usr/local/var/run/openvswitch
242 ``server-socket-path`` must be set to the value of ``ovs_rundir()`` which
243 typically resolves to ``/usr/local/var/run/openvswitch``.
245 Many tools are available to view the contents of the pcap file. Once example is
246 tcpdump. Issue the following command to view the contents of ``pkts.pcap``::
248 $ tcpdump -r pkts.pcap
250 More information on the pdump app and its usage can be found in the `DPDK docs
251 <http://dpdk.org/doc/guides/tools/pdump.html>`__.
256 By default, DPDK ports are configured with standard Ethernet MTU (1500B). To
257 enable Jumbo Frames support for a DPDK port, change the Interface's
258 ``mtu_request`` attribute to a sufficiently large value. For example, to add a
259 DPDK Phy port with MTU of 9000::
261 $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
262 options:dpdk-devargs=0000:01:00.0 mtu_request=9000
264 Similarly, to change the MTU of an existing port to 6200::
266 $ ovs-vsctl set Interface dpdk-p0 mtu_request=6200
268 Some additional configuration is needed to take advantage of jumbo frames with
271 1. *mergeable buffers* must be enabled for vHost ports, as demonstrated in the
272 QEMU command line snippet below::
274 -netdev type=vhost-user,id=mynet1,chardev=char0,vhostforce \
275 -device virtio-net-pci,mac=00:00:00:00:00:01,netdev=mynet1,mrg_rxbuf=on
277 2. Where virtio devices are bound to the Linux kernel driver in a guest
278 environment (i.e. interfaces are not bound to an in-guest DPDK driver), the
279 MTU of those logical network interfaces must also be increased to a
280 sufficiently large value. This avoids segmentation of Jumbo Frames received
281 in the guest. Note that 'MTU' refers to the length of the IP packet only,
282 and not that of the entire frame.
284 To calculate the exact MTU of a standard IPv4 frame, subtract the L2 header
285 and CRC lengths (i.e. 18B) from the max supported frame size. So, to set
286 the MTU for a 9018B Jumbo Frame::
288 $ ip link set eth1 mtu 9000
290 When Jumbo Frames are enabled, the size of a DPDK port's mbuf segments are
291 increased, such that a full Jumbo Frame of a specific size may be accommodated
292 within a single mbuf segment.
294 Jumbo frame support has been validated against 9728B frames, which is the
295 largest frame size supported by Fortville NIC using the DPDK i40e driver, but
296 larger frames and other DPDK NIC drivers may be supported. These cases are
297 common for use cases involving East-West traffic only.
302 By default, DPDK physical ports are enabled with Rx checksum offload.
304 Rx checksum offload can offer performance improvement only for tunneling
305 traffic in OVS-DPDK because the checksum validation of tunnel packets is
306 offloaded to the NIC. Also enabling Rx checksum may slightly reduce the
307 performance of non-tunnel traffic, specifically for smaller size packet.
309 .. _extended-statistics:
314 DPDK Extended Statistics API allows PMD to expose unique set of statistics.
315 The Extended statistics are implemented and supported only for DPDK physical
318 To enable statistics, you have to enable OpenFlow 1.4 support for OVS.
319 Configure bridge br0 to support OpenFlow version 1.4::
321 $ ovs-vsctl set bridge br0 datapath_type=netdev \
322 protocols=OpenFlow10,OpenFlow11,OpenFlow12,OpenFlow13,OpenFlow14
324 Check the OVSDB protocols column in the bridge table if OpenFlow 1.4 support
327 $ ovsdb-client dump Bridge protocols
329 Query the port statistics by explicitly specifying -O OpenFlow14 option::
331 $ ovs-ofctl -O OpenFlow14 dump-ports br0
333 Note: vHost ports supports only partial statistics. RX packet size based
334 counter are only supported and doesn't include TX packet size counters.
341 OVS supports port hotplugging, allowing the use of ports that were not bound
342 to DPDK when vswitchd was started.
343 In order to attach a port, it has to be bound to DPDK using the
344 ``dpdk_nic_bind.py`` script::
346 $ $DPDK_DIR/tools/dpdk_nic_bind.py --bind=igb_uio 0000:01:00.0
348 Then it can be attached to OVS::
350 $ ovs-vsctl add-port br0 dpdkx -- set Interface dpdkx type=dpdk \
351 options:dpdk-devargs=0000:01:00.0
353 Detaching will be performed while processing del-port command::
355 $ ovs-vsctl del-port dpdkx
357 Sometimes, the del-port command may not detach the device.
358 Detaching can be confirmed by the appearance of an INFO log.
361 INFO|Device '0000:04:00.1' has been detached
363 If the log is not seen, then the port can be detached using::
365 $ ovs-appctl netdev-dpdk/detach 0000:01:00.0
367 Detaching can be confirmed by console output::
369 Device '0000:04:00.1' has been detached
372 Detaching should not be done if a device is known to be non-detachable, as
373 this may cause the device to behave improperly when added back with
374 add-port. The Chelsio Terminator adapters which use the cxgbe driver seem
375 to be an example of this behavior; check the driver documentation if this
378 This feature does not work with some NICs.
379 For more information please refer to the `DPDK Port Hotplug Framework
380 <http://dpdk.org/doc/guides/prog_guide/port_hotplug_framework.html#hotplug>`__.
387 DPDK provides drivers for both physical and virtual devices. Physical DPDK
388 devices are added to OVS by specifying a valid PCI address in 'dpdk-devargs'.
389 Virtual DPDK devices which do not have PCI addresses can be added using a
390 different format for 'dpdk-devargs'.
392 Typically, the format expected is 'eth_<driver_name><x>' where 'x' is a
393 unique identifier of your choice for the given port.
395 For example to add a dpdk port that uses the 'null' DPDK PMD driver::
397 $ ovs-vsctl add-port br0 null0 -- set Interface null0 type=dpdk \
398 options:dpdk-devargs=eth_null0
400 Similarly, to add a dpdk port that uses the 'af_packet' DPDK PMD driver::
402 $ ovs-vsctl add-port br0 myeth0 -- set Interface myeth0 type=dpdk \
403 options:dpdk-devargs=eth_af_packet0,iface=eth0
405 More information on the different types of virtual DPDK PMDs can be found in
406 the `DPDK documentation
407 <http://dpdk.org/doc/guides/nics/overview.html>`__.
409 Note: Not all DPDK virtual PMD drivers have been tested and verified to work.
411 EMC Insertion Probability
412 -------------------------
413 By default 1 in every 100 flows are inserted into the Exact Match Cache (EMC).
414 It is possible to change this insertion probability by setting the
415 ``emc-insert-inv-prob`` option::
417 $ ovs-vsctl --no-wait set Open_vSwitch . other_config:emc-insert-inv-prob=N
422 is a positive integer representing the inverse probability of insertion ie.
423 on average 1 in every N packets with a unique flow will generate an EMC
426 If ``N`` is set to 1, an insertion will be performed for every flow. If set to
427 0, no insertions will be performed and the EMC will effectively be disabled.
429 With default ``N`` set to 100, higher megaflow hits will occur initially
430 as observed with pmd stats::
432 $ ovs-appctl dpif-netdev/pmd-stats-show
434 For certain traffic profiles with many parallel flows, it's recommended to set
435 ``N`` to '0' to achieve higher forwarding performance.
437 For more information on the EMC refer to :doc:`/intro/install/dpdk` .
439 .. _dpdk-ovs-in-guest:
441 OVS with DPDK Inside VMs
442 ------------------------
444 Additional configuration is required if you want to run ovs-vswitchd with DPDK
445 backend inside a QEMU virtual machine. ovs-vswitchd creates separate DPDK TX
446 queues for each CPU core available. This operation fails inside QEMU virtual
447 machine because, by default, VirtIO NIC provided to the guest is configured to
448 support only single TX queue and single RX queue. To change this behavior, you
449 need to turn on ``mq`` (multiqueue) property of all ``virtio-net-pci`` devices
450 emulated by QEMU and used by DPDK. You may do it manually (by changing QEMU
451 command line) or, if you use Libvirt, by adding the following string to
452 ``<interface>`` sections of all network devices used by DPDK::
454 <driver name='vhost' queues='N'/>
459 determines how many queues can be used by the guest.
461 This requires QEMU >= 2.2.
468 Add a userspace bridge and two ``dpdk`` (PHY) ports::
470 # Add userspace bridge
471 $ ovs-vsctl add-br br0 -- set bridge br0 datapath_type=netdev
474 $ ovs-vsctl add-port br0 phy0 -- set Interface phy0 type=dpdk \
475 options:dpdk-devargs=0000:01:00.0 ofport_request=1
477 $ ovs-vsctl add-port br0 phy1 -- set Interface phy1 type=dpdk
478 options:dpdk-devargs=0000:01:00.1 ofport_request=2
480 Add test flows to forward packets betwen DPDK port 0 and port 1::
482 # Clear current flows
483 $ ovs-ofctl del-flows br0
485 # Add flows between port 1 (phy0) to port 2 (phy1)
486 $ ovs-ofctl add-flow br0 in_port=1,action=output:2
487 $ ovs-ofctl add-flow br0 in_port=2,action=output:1
489 Transmit traffic into either port. You should see it returned via the other.
491 .. _dpdk-vhost-loopback:
493 PHY-VM-PHY (vHost Loopback)
494 ---------------------------
496 Add a userspace bridge, two ``dpdk`` (PHY) ports, and two ``dpdkvhostuser``
499 # Add userspace bridge
500 $ ovs-vsctl add-br br0 -- set bridge br0 datapath_type=netdev
503 $ ovs-vsctl add-port br0 phy0 -- set Interface phy0 type=dpdk \
504 options:dpdk-devargs=0000:01:00.0 ofport_request=1
506 $ ovs-vsctl add-port br0 phy1 -- set Interface phy1 type=dpdk
507 options:dpdk-devargs=0000:01:00.1 ofport_request=2
509 # Add two dpdkvhostuser ports
510 $ ovs-vsctl add-port br0 dpdkvhostuser0 \
511 -- set Interface dpdkvhostuser0 type=dpdkvhostuser ofport_request=3
512 $ ovs-vsctl add-port br0 dpdkvhostuser1 \
513 -- set Interface dpdkvhostuser1 type=dpdkvhostuser ofport_request=4
515 Add test flows to forward packets betwen DPDK devices and VM ports::
517 # Clear current flows
518 $ ovs-ofctl del-flows br0
521 $ ovs-ofctl add-flow br0 in_port=1,action=output:3
522 $ ovs-ofctl add-flow br0 in_port=3,action=output:1
523 $ ovs-ofctl add-flow br0 in_port=4,action=output:2
524 $ ovs-ofctl add-flow br0 in_port=2,action=output:4
527 $ ovs-ofctl dump-flows br0
529 Create a VM using the following configuration:
531 +----------------------+--------+-----------------+
532 | configuration | values | comments |
533 +----------------------+--------+-----------------+
534 | qemu version | 2.2.0 | n/a |
535 | qemu thread affinity | core 5 | taskset 0x20 |
536 | memory | 4GB | n/a |
538 | Qcow2 image | CentOS7| n/a |
539 | mrg_rxbuf | off | n/a |
540 +----------------------+--------+-----------------+
542 You can do this directly with QEMU via the ``qemu-system-x86_64`` application::
544 $ export VM_NAME=vhost-vm
545 $ export GUEST_MEM=3072M
546 $ export QCOW2_IMAGE=/root/CentOS7_x86_64.qcow2
547 $ export VHOST_SOCK_DIR=/usr/local/var/run/openvswitch
549 $ taskset 0x20 qemu-system-x86_64 -name $VM_NAME -cpu host -enable-kvm \
550 -m $GUEST_MEM -drive file=$QCOW2_IMAGE --nographic -snapshot \
551 -numa node,memdev=mem -mem-prealloc -smp sockets=1,cores=2 \
552 -object memory-backend-file,id=mem,size=$GUEST_MEM,mem-path=/dev/hugepages,share=on \
553 -chardev socket,id=char0,path=$VHOST_SOCK_DIR/dpdkvhostuser0 \
554 -netdev type=vhost-user,id=mynet1,chardev=char0,vhostforce \
555 -device virtio-net-pci,mac=00:00:00:00:00:01,netdev=mynet1,mrg_rxbuf=off \
556 -chardev socket,id=char1,path=$VHOST_SOCK_DIR/dpdkvhostuser1 \
557 -netdev type=vhost-user,id=mynet2,chardev=char1,vhostforce \
558 -device virtio-net-pci,mac=00:00:00:00:00:02,netdev=mynet2,mrg_rxbuf=off
560 For a explanation of this command, along with alternative approaches such as
561 booting the VM via libvirt, refer to :doc:`/topics/dpdk/vhost-user`.
563 Once the guest is configured and booted, configure DPDK packet forwarding
564 within the guest. To accomplish this, build the ``testpmd`` application as
565 described in :ref:`dpdk-testpmd`. Once compiled, run the application::
567 $ cd $DPDK_DIR/app/test-pmd;
568 $ ./testpmd -c 0x3 -n 4 --socket-mem 1024 -- \
569 --burst=64 -i --txqflags=0xf00 --disable-hw-vlan
573 When you finish testing, bind the vNICs back to kernel::
575 $ $DPDK_DIR/usertools/dpdk-devbind.py --bind=virtio-pci 0000:00:03.0
576 $ $DPDK_DIR/usertools/dpdk-devbind.py --bind=virtio-pci 0000:00:04.0
580 Valid PCI IDs must be passed in above example. The PCI IDs can be retrieved
583 $ $DPDK_DIR/usertools/dpdk-devbind.py --status
585 More information on the dpdkvhostuser ports can be found in
586 :doc:`/topics/dpdk/vhost-user`.
588 PHY-VM-PHY (vHost Loopback) (Kernel Forwarding)
589 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
591 :ref:`dpdk-vhost-loopback` details steps for PHY-VM-PHY loopback
592 testcase and packet forwarding using DPDK testpmd application in the Guest VM.
593 For users wishing to do packet forwarding using kernel stack below, you need to
594 run the below commands on the guest::
596 $ ip addr add 1.1.1.2/24 dev eth1
597 $ ip addr add 1.1.2.2/24 dev eth2
598 $ ip link set eth1 up
599 $ ip link set eth2 up
600 $ systemctl stop firewalld.service
601 $ systemctl stop iptables.service
602 $ sysctl -w net.ipv4.ip_forward=1
603 $ sysctl -w net.ipv4.conf.all.rp_filter=0
604 $ sysctl -w net.ipv4.conf.eth1.rp_filter=0
605 $ sysctl -w net.ipv4.conf.eth2.rp_filter=0
606 $ route add -net 1.1.2.0/24 eth2
607 $ route add -net 1.1.1.0/24 eth1
608 $ arp -s 1.1.2.99 DE:AD:BE:EF:CA:FE
609 $ arp -s 1.1.1.99 DE:AD:BE:EF:CA:EE
611 PHY-VM-PHY (vHost Multiqueue)
612 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
614 vHost Multiqueue functionality can also be validated using the PHY-VM-PHY
615 configuration. To begin, follow the steps described in :ref:`dpdk-phy-phy` to
616 create and initialize the database, start ovs-vswitchd and add ``dpdk``-type
617 devices to bridge ``br0``. Once complete, follow the below steps:
619 1. Configure PMD and RXQs.
621 For example, set the number of dpdk port rx queues to at least 2 The number
622 of rx queues at vhost-user interface gets automatically configured after
623 virtio device connection and doesn't need manual configuration::
625 $ ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=0xc
626 $ ovs-vsctl set Interface phy0 options:n_rxq=2
627 $ ovs-vsctl set Interface phy1 options:n_rxq=2
629 2. Instantiate Guest VM using QEMU cmdline
631 We must configure with appropriate software versions to ensure this feature
634 .. list-table:: Recommended BIOS Settings
641 * - QEMU thread affinity
642 - 2 cores (taskset 0x30)
652 To do this, instantiate the guest as follows::
654 $ export VM_NAME=vhost-vm
655 $ export GUEST_MEM=4096M
656 $ export QCOW2_IMAGE=/root/Fedora22_x86_64.qcow2
657 $ export VHOST_SOCK_DIR=/usr/local/var/run/openvswitch
658 $ taskset 0x30 qemu-system-x86_64 -cpu host -smp 2,cores=2 -m 4096M \
659 -drive file=$QCOW2_IMAGE --enable-kvm -name $VM_NAME \
660 -nographic -numa node,memdev=mem -mem-prealloc \
661 -object memory-backend-file,id=mem,size=$GUEST_MEM,mem-path=/dev/hugepages,share=on \
662 -chardev socket,id=char1,path=$VHOST_SOCK_DIR/dpdkvhostuser0 \
663 -netdev type=vhost-user,id=mynet1,chardev=char1,vhostforce,queues=2 \
664 -device virtio-net-pci,mac=00:00:00:00:00:01,netdev=mynet1,mq=on,vectors=6 \
665 -chardev socket,id=char2,path=$VHOST_SOCK_DIR/dpdkvhostuser1 \
666 -netdev type=vhost-user,id=mynet2,chardev=char2,vhostforce,queues=2 \
667 -device virtio-net-pci,mac=00:00:00:00:00:02,netdev=mynet2,mq=on,vectors=6
670 Queue value above should match the queues configured in OVS, The vector
671 value should be set to "number of queues x 2 + 2"
673 3. Configure the guest interface
675 Assuming there are 2 interfaces in the guest named eth0, eth1 check the
676 channel configuration and set the number of combined channels to 2 for
680 $ ethtool -L eth0 combined 2
681 $ ethtool -L eth1 combined 2
683 More information can be found in vHost walkthrough section.
685 4. Configure kernel packet forwarding
687 Configure IP and enable interfaces::
689 $ ip addr add 5.5.5.1/24 dev eth0
690 $ ip addr add 90.90.90.1/24 dev eth1
691 $ ip link set eth0 up
692 $ ip link set eth1 up
694 Configure IP forwarding and add route entries::
696 $ sysctl -w net.ipv4.ip_forward=1
697 $ sysctl -w net.ipv4.conf.all.rp_filter=0
698 $ sysctl -w net.ipv4.conf.eth0.rp_filter=0
699 $ sysctl -w net.ipv4.conf.eth1.rp_filter=0
700 $ ip route add 2.1.1.0/24 dev eth1
701 $ route add default gw 2.1.1.2 eth1
702 $ route add default gw 90.90.90.90 eth1
703 $ arp -s 90.90.90.90 DE:AD:BE:EF:CA:FE
704 $ arp -s 2.1.1.2 DE:AD:BE:EF:CA:FA
706 Check traffic on multiple queues::
708 $ cat /proc/interrupts | grep virtio