1 Using Open vSwitch with DPDK
2 ============================
4 Open vSwitch can use Intel(R) DPDK lib to operate entirely in
5 userspace. This file explains how to install and use Open vSwitch in
8 The DPDK support of Open vSwitch is considered experimental.
9 It has not been thoroughly tested.
11 This version of Open vSwitch should be built manually with `configure`
14 OVS needs a system with 1GB hugepages support.
16 Building and Installing:
17 ------------------------
20 Optional (if building with vhost-cuse): `fuse`, `fuse-devel` (`libfuse-dev`
23 1. Configure build & install DPDK:
27 export DPDK_DIR=/usr/src/dpdk-16.04
31 2. Then run `make install` to build and install the library.
32 For default install without IVSHMEM:
34 `make install T=x86_64-native-linuxapp-gcc DESTDIR=install`
36 To include IVSHMEM (shared memory):
38 `make install T=x86_64-ivshmem-linuxapp-gcc DESTDIR=install`
40 For further details refer to http://dpdk.org/
42 2. Configure & build the Linux kernel:
44 Refer to intel-dpdk-getting-started-guide.pdf for understanding
45 DPDK kernel requirement.
47 3. Configure & build OVS:
51 `export DPDK_BUILD=$DPDK_DIR/x86_64-native-linuxapp-gcc/`
55 `export DPDK_BUILD=$DPDK_DIR/x86_64-ivshmem-linuxapp-gcc/`
60 ./configure --with-dpdk=$DPDK_BUILD [CFLAGS="-g -O2 -Wno-cast-align"]
64 Note: 'clang' users may specify the '-Wno-cast-align' flag to suppress DPDK cast-align warnings.
66 To have better performance one can enable aggressive compiler optimizations and
67 use the special instructions(popcnt, crc32) that may not be available on all
68 machines. Instead of typing `make`, type:
70 `make CFLAGS='-O3 -march=native'`
72 Refer to [INSTALL.userspace.md] for general requirements of building userspace OVS.
74 Using the DPDK with ovs-vswitchd:
75 ---------------------------------
78 Add the following options to the kernel bootline:
80 `default_hugepagesz=1GB hugepagesz=1G hugepages=1`
82 2. Setup DPDK devices:
84 DPDK devices can be setup using either the VFIO (for DPDK 1.7+) or UIO
85 modules. UIO requires inserting an out of tree driver igb_uio.ko that is
86 available in DPDK. Setup for both methods are described below.
89 1. insert uio.ko: `modprobe uio`
90 2. insert igb_uio.ko: `insmod $DPDK_BUILD/kmod/igb_uio.ko`
91 3. Bind network device to igb_uio:
92 `$DPDK_DIR/tools/dpdk_nic_bind.py --bind=igb_uio eth1`
96 VFIO needs to be supported in the kernel and the BIOS. More information
97 can be found in the [DPDK Linux GSG].
99 1. Insert vfio-pci.ko: `modprobe vfio-pci`
100 2. Set correct permissions on vfio device: `sudo /usr/bin/chmod a+x /dev/vfio`
101 and: `sudo /usr/bin/chmod 0666 /dev/vfio/*`
102 3. Bind network device to vfio-pci:
103 `$DPDK_DIR/tools/dpdk_nic_bind.py --bind=vfio-pci eth1`
105 3. Mount the hugetable filesystem
107 `mount -t hugetlbfs -o pagesize=1G none /dev/hugepages`
109 Ref to http://www.dpdk.org/doc/quick-start for verifying DPDK setup.
111 4. Follow the instructions in [INSTALL.md] to install only the
112 userspace daemons and utilities (via 'make install').
113 1. First time only db creation (or clearing):
116 mkdir -p /usr/local/etc/openvswitch
117 mkdir -p /usr/local/var/run/openvswitch
118 rm /usr/local/etc/openvswitch/conf.db
119 ovsdb-tool create /usr/local/etc/openvswitch/conf.db \
120 /usr/local/share/openvswitch/vswitch.ovsschema
123 2. Start ovsdb-server
126 ovsdb-server --remote=punix:/usr/local/var/run/openvswitch/db.sock \
127 --remote=db:Open_vSwitch,Open_vSwitch,manager_options \
128 --private-key=db:Open_vSwitch,SSL,private_key \
129 --certificate=Open_vSwitch,SSL,certificate \
130 --bootstrap-ca-cert=db:Open_vSwitch,SSL,ca_cert --pidfile --detach
133 3. First time after db creation, initialize:
136 ovs-vsctl --no-wait init
141 DPDK configuration arguments can be passed to vswitchd via Open_vSwitch
142 other_config column. The recognized configuration options are listed.
143 Defaults will be provided for all values not explicitly set.
146 Specifies whether OVS should initialize and support DPDK ports. This is
147 a boolean, and defaults to false.
150 Specifies the CPU cores on which dpdk lcore threads should be spawned.
151 The DPDK lcore threads are used for DPDK library tasks, such as
152 library internal message processing, logging, etc. Value should be in
153 the form of a hex string (so '0x123') similar to the 'taskset' mask
155 If not specified, the value will be determined by choosing the lowest
156 CPU core from initial cpu affinity list. Otherwise, the value will be
157 passed directly to the DPDK library.
158 For performance reasons, it is best to set this to a single core on
159 the system, rather than allow lcore threads to float.
162 This sets the total memory to preallocate from hugepages regardless of
163 processor socket. It is recommended to use dpdk-socket-mem instead.
166 Comma separated list of memory to pre-allocate from hugepages on specific
170 Directory where hugetlbfs is mounted
173 Extra arguments to provide to DPDK EAL, as previously specified on the
174 command line. Do not pass '--no-huge' to the system in this way. Support
175 for running the system without hugepages is nonexistent.
178 Option to set the vhost_cuse character device name.
181 Option to set the path to the vhost_user unix socket files.
183 NOTE: Changing any of these options requires restarting the ovs-vswitchd
186 Open vSwitch can be started as normal. DPDK will be initialized as long
187 as the dpdk-init option has been set to 'true'.
191 export DB_SOCK=/usr/local/var/run/openvswitch/db.sock
192 ovs-vsctl --no-wait set Open_vSwitch . other_config:dpdk-init=true
193 ovs-vswitchd unix:$DB_SOCK --pidfile --detach
196 If allocated more than one GB hugepage (as for IVSHMEM), set amount and
197 use NUMA node 0 memory:
200 ovs-vsctl --no-wait set Open_vSwitch . other_config:dpdk-socket-mem="1024,0"
201 ovs-vswitchd unix:$DB_SOCK --pidfile --detach
204 6. Add bridge & ports
206 To use ovs-vswitchd with DPDK, create a bridge with datapath_type
207 "netdev" in the configuration database. For example:
209 `ovs-vsctl add-br br0 -- set bridge br0 datapath_type=netdev`
211 Now you can add dpdk devices. OVS expects DPDK device names to start with
212 "dpdk" and end with a portid. vswitchd should print (in the log file) the
213 number of dpdk devices found.
216 ovs-vsctl add-port br0 dpdk0 -- set Interface dpdk0 type=dpdk
217 ovs-vsctl add-port br0 dpdk1 -- set Interface dpdk1 type=dpdk
220 Once first DPDK port is added to vswitchd, it creates a Polling thread and
221 polls dpdk device in continuous loop. Therefore CPU utilization
222 for that thread is always 100%.
224 Note: creating bonds of DPDK interfaces is slightly different to creating
225 bonds of system interfaces. For DPDK, the interface type must be explicitly
229 ovs-vsctl add-bond br0 dpdkbond dpdk0 dpdk1 -- set Interface dpdk0 type=dpdk -- set Interface dpdk1 type=dpdk
234 Test flow script across NICs (assuming ovs in /usr/src/ovs):
239 # Move to command directory
240 cd /usr/src/ovs/utilities/
242 # Clear current flows
243 ./ovs-ofctl del-flows br0
245 # Add flows between port 1 (dpdk0) to port 2 (dpdk1)
246 ./ovs-ofctl add-flow br0 in_port=1,action=output:2
247 ./ovs-ofctl add-flow br0 in_port=2,action=output:1
252 Assuming you have a vhost-user port transmitting traffic consisting of
253 packets of size 64 bytes, the following command would limit the egress
254 transmission rate of the port to ~1,000,000 packets per second:
256 `ovs-vsctl set port vhost-user0 qos=@newqos -- --id=@newqos create qos
257 type=egress-policer other-config:cir=46000000 other-config:cbs=2048`
259 To examine the QoS configuration of the port:
261 `ovs-appctl -t ovs-vswitchd qos/show vhost-user0`
263 To clear the QoS configuration from the port and ovsdb use the following:
265 `ovs-vsctl destroy QoS vhost-user0 -- clear Port vhost-user0 qos`
267 For more details regarding egress-policer parameters please refer to the
273 1. PMD affinitization
275 A poll mode driver (pmd) thread handles the I/O of all DPDK
276 interfaces assigned to it. A pmd thread will busy loop through
277 the assigned port/rxq's polling for packets, switch the packets
278 and send to a tx port if required. Typically, it is found that
279 a pmd thread is CPU bound, meaning that the greater the CPU
280 occupancy the pmd thread can get, the better the performance. To
281 that end, it is good practice to ensure that a pmd thread has as
282 many cycles on a core available to it as possible. This can be
283 achieved by affinitizing the pmd thread with a core that has no
284 other workload. See section 7 below for a description of how to
285 isolate cores for this purpose also.
287 The following command can be used to specify the affinity of the
290 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=<hex string>`
292 By setting a bit in the mask, a pmd thread is created and pinned
293 to the corresponding CPU core. e.g. to run a pmd thread on core 1
295 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=2`
297 For more information, please refer to the Open_vSwitch TABLE section in
299 `man ovs-vswitchd.conf.db`
301 Note, that a pmd thread on a NUMA node is only created if there is
302 at least one DPDK interface from that NUMA node added to OVS.
304 2. Multiple poll mode driver threads
306 With pmd multi-threading support, OVS creates one pmd thread
307 for each NUMA node by default. However, it can be seen that in cases
308 where there are multiple ports/rxq's producing traffic, performance
309 can be improved by creating multiple pmd threads running on separate
310 cores. These pmd threads can then share the workload by each being
311 responsible for different ports/rxq's. Assignment of ports/rxq's to
312 pmd threads is done automatically.
314 The following command can be used to specify the affinity of the
317 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=<hex string>`
319 A set bit in the mask means a pmd thread is created and pinned
320 to the corresponding CPU core. e.g. to run pmd threads on core 1 and 2
322 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=6`
324 For more information, please refer to the Open_vSwitch TABLE section in
326 `man ovs-vswitchd.conf.db`
328 For example, when using dpdk and dpdkvhostuser ports in a bi-directional
329 VM loopback as shown below, spreading the workload over 2 or 4 pmd
330 threads shows significant improvements as there will be more total CPU
333 NIC port0 <-> OVS <-> VM <-> OVS <-> NIC port 1
335 The following command can be used to confirm that the port/rxq assignment
336 to pmd threads is as required:
338 `ovs-appctl dpif-netdev/pmd-rxq-show`
340 This can also be checked with:
344 taskset -p <pid_of_pmd>
347 To understand where most of the pmd thread time is spent and whether the
348 caches are being utilized, these commands can be used:
351 # Clear previous stats
352 ovs-appctl dpif-netdev/pmd-stats-clear
354 # Check current stats
355 ovs-appctl dpif-netdev/pmd-stats-show
358 3. DPDK port Rx Queues
360 `ovs-vsctl set Interface <DPDK interface> options:n_rxq=<integer>`
362 The command above sets the number of rx queues for DPDK interface.
363 The rx queues are assigned to pmd threads on the same NUMA node in a
364 round-robin fashion. For more information, please refer to the
365 Open_vSwitch TABLE section in
367 `man ovs-vswitchd.conf.db`
371 Each pmd thread contains one EMC. After initial flow setup in the
372 datapath, the EMC contains a single table and provides the lowest level
373 (fastest) switching for DPDK ports. If there is a miss in the EMC then
374 the next level where switching will occur is the datapath classifier.
375 Missing in the EMC and looking up in the datapath classifier incurs a
376 significant performance penalty. If lookup misses occur in the EMC
377 because it is too small to handle the number of flows, its size can
378 be increased. The EMC size can be modified by editing the define
379 EM_FLOW_HASH_SHIFT in lib/dpif-netdev.c.
381 As mentioned above an EMC is per pmd thread. So an alternative way of
382 increasing the aggregate amount of possible flow entries in EMC and
383 avoiding datapath classifier lookups is to have multiple pmd threads
384 running. This can be done as described in section 2.
388 The default compiler optimization level is '-O2'. Changing this to
389 more aggressive compiler optimizations such as '-O3' or
390 '-Ofast -march=native' with gcc can produce performance gains.
392 6. Simultaneous Multithreading (SMT)
394 With SMT enabled, one physical core appears as two logical cores
395 which can improve performance.
397 SMT can be utilized to add additional pmd threads without consuming
398 additional physical cores. Additional pmd threads may be added in the
399 same manner as described in section 2. If trying to minimize the use
400 of physical cores for pmd threads, care must be taken to set the
401 correct bits in the pmd-cpu-mask to ensure that the pmd threads are
402 pinned to SMT siblings.
404 For example, when using 2x 10 core processors in a dual socket system
405 with HT enabled, /proc/cpuinfo will report 40 logical cores. To use
406 two logical cores which share the same physical core for pmd threads,
407 the following command can be used to identify a pair of logical cores.
409 `cat /sys/devices/system/cpu/cpuN/topology/thread_siblings_list`
411 where N is the logical core number. In this example, it would show that
412 cores 1 and 21 share the same physical core. The pmd-cpu-mask to enable
413 two pmd threads running on these two logical cores (one physical core)
416 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=100002`
418 Note that SMT is enabled by the Hyper-Threading section in the
419 BIOS, and as such will apply to the whole system. So the impact of
420 enabling/disabling it for the whole system should be considered
421 e.g. If workloads on the system can scale across multiple cores,
422 SMT may very beneficial. However, if they do not and perform best
423 on a single physical core, SMT may not be beneficial.
425 7. The isolcpus kernel boot parameter
427 isolcpus can be used on the kernel bootline to isolate cores from the
428 kernel scheduler and hence dedicate them to OVS or other packet
429 forwarding related workloads. For example a Linux kernel boot-line
432 'GRUB_CMDLINE_LINUX_DEFAULT="quiet hugepagesz=1G hugepages=4 default_hugepagesz=1G 'intel_iommu=off' isolcpus=1-19"'
434 8. NUMA/Cluster On Die
436 Ideally inter NUMA datapaths should be avoided where possible as packets
437 will go across QPI and there may be a slight performance penalty when
438 compared with intra NUMA datapaths. On Intel Xeon Processor E5 v3,
439 Cluster On Die is introduced on models that have 10 cores or more.
440 This makes it possible to logically split a socket into two NUMA regions
441 and again it is preferred where possible to keep critical datapaths
442 within the one cluster.
444 It is good practice to ensure that threads that are in the datapath are
445 pinned to cores in the same NUMA area. e.g. pmd threads and QEMU vCPUs
446 responsible for forwarding.
448 9. Rx Mergeable buffers
450 Rx Mergeable buffers is a virtio feature that allows chaining of multiple
451 virtio descriptors to handle large packet sizes. As such, large packets
452 are handled by reserving and chaining multiple free descriptors
453 together. Mergeable buffer support is negotiated between the virtio
454 driver and virtio device and is supported by the DPDK vhost library.
455 This behavior is typically supported and enabled by default, however
456 in the case where the user knows that rx mergeable buffers are not needed
457 i.e. jumbo frames are not needed, it can be forced off by adding
458 mrg_rxbuf=off to the QEMU command line options. By not reserving multiple
459 chains of descriptors it will make more individual virtio descriptors
460 available for rx to the guest using dpdkvhost ports and this can improve
463 10. Packet processing in the guest
465 It is good practice whether simply forwarding packets from one
466 interface to another or more complex packet processing in the guest,
467 to ensure that the thread performing this work has as much CPU
468 occupancy as possible. For example when the DPDK sample application
469 `testpmd` is used to forward packets in the guest, multiple QEMU vCPU
470 threads can be created. Taskset can then be used to affinitize the
471 vCPU thread responsible for forwarding to a dedicated core not used
472 for other general processing on the host system.
474 11. DPDK virtio pmd in the guest
476 dpdkvhostcuse or dpdkvhostuser ports can be used to accelerate the path
477 to the guest using the DPDK vhost library. This library is compatible with
478 virtio-net drivers in the guest but significantly better performance can
479 be observed when using the DPDK virtio pmd driver in the guest. The DPDK
480 `testpmd` application can be used in the guest as an example application
481 that forwards packet from one DPDK vhost port to another. An example of
482 running `testpmd` in the guest can be seen here.
484 `./testpmd -c 0x3 -n 4 --socket-mem 512 -- --burst=64 -i --txqflags=0xf00 --disable-hw-vlan --forward-mode=io --auto-start`
486 See below information on dpdkvhostcuse and dpdkvhostuser ports.
487 See [DPDK Docs] for more information on `testpmd`.
494 Following the steps above to create a bridge, you can now add dpdk rings
495 as a port to the vswitch. OVS will expect the DPDK ring device name to
496 start with dpdkr and end with a portid.
498 `ovs-vsctl add-port br0 dpdkr0 -- set Interface dpdkr0 type=dpdkr`
500 DPDK rings client test application
502 Included in the test directory is a sample DPDK application for testing
503 the rings. This is from the base dpdk directory and modified to work
504 with the ring naming used within ovs.
506 location tests/ovs_client
511 cd /usr/src/ovs/tests/
512 ovsclient -c 1 -n 4 --proc-type=secondary -- -n "port id you gave dpdkr"
515 In the case of the dpdkr example above the "port id you gave dpdkr" is 0.
517 It is essential to have --proc-type=secondary
519 The application simply receives an mbuf on the receive queue of the
520 ethernet ring and then places that same mbuf on the transmit ring of
521 the ethernet ring. It is a trivial loopback application.
523 DPDK rings in VM (IVSHMEM shared memory communications)
524 -------------------------------------------------------
526 In addition to executing the client in the host, you can execute it within
527 a guest VM. To do so you will need a patched qemu. You can download the
528 patch and getting started guide at :
530 https://01.org/packet-processing/downloads
532 A general rule of thumb for better performance is that the client
533 application should not be assigned the same dpdk core mask "-c" as
539 DPDK 16.04 supports two types of vhost:
544 Whatever type of vhost is enabled in the DPDK build specified, is the type
545 that will be enabled in OVS. By default, vhost-user is enabled in DPDK.
546 Therefore, unless vhost-cuse has been enabled in DPDK, vhost-user ports
547 will be enabled in OVS.
548 Please note that support for vhost-cuse is intended to be deprecated in OVS
554 The following sections describe the use of vhost-user 'dpdkvhostuser' ports
557 DPDK vhost-user Prerequisites:
558 -------------------------
560 1. DPDK 16.04 with vhost support enabled as documented in the "Building and
563 2. QEMU version v2.1.0+
565 QEMU v2.1.0 will suffice, but it is recommended to use v2.2.0 if providing
566 your VM with memory greater than 1GB due to potential issues with memory
567 mapping larger areas.
569 Adding DPDK vhost-user ports to the Switch:
570 --------------------------------------
572 Following the steps above to create a bridge, you can now add DPDK vhost-user
573 as a port to the vswitch. Unlike DPDK ring ports, DPDK vhost-user ports can
574 have arbitrary names, except that forward and backward slashes are prohibited
577 - For vhost-user, the name of the port type is `dpdkvhostuser`
580 ovs-vsctl add-port br0 vhost-user-1 -- set Interface vhost-user-1
584 This action creates a socket located at
585 `/usr/local/var/run/openvswitch/vhost-user-1`, which you must provide
586 to your VM on the QEMU command line. More instructions on this can be
587 found in the next section "DPDK vhost-user VM configuration"
588 - If you wish for the vhost-user sockets to be created in a sub-directory of
589 `/usr/local/var/run/openvswitch`, you may specify this directory in the
592 `./utilities/ovs-vsctl --no-wait \
593 set Open_vSwitch . other_config:vhost-sock-dir=subdir`
595 DPDK vhost-user VM configuration:
596 ---------------------------------
597 Follow the steps below to attach vhost-user port(s) to a VM.
599 1. Configure sockets.
600 Pass the following parameters to QEMU to attach a vhost-user device:
603 -chardev socket,id=char1,path=/usr/local/var/run/openvswitch/vhost-user-1
604 -netdev type=vhost-user,id=mynet1,chardev=char1,vhostforce
605 -device virtio-net-pci,mac=00:00:00:00:00:01,netdev=mynet1
608 ...where vhost-user-1 is the name of the vhost-user port added
610 Repeat the above parameters for multiple devices, changing the
611 chardev path and id as necessary. Note that a separate and different
612 chardev path needs to be specified for each vhost-user device. For
613 example you have a second vhost-user port named 'vhost-user-2', you
614 append your QEMU command line with an additional set of parameters:
617 -chardev socket,id=char2,path=/usr/local/var/run/openvswitch/vhost-user-2
618 -netdev type=vhost-user,id=mynet2,chardev=char2,vhostforce
619 -device virtio-net-pci,mac=00:00:00:00:00:02,netdev=mynet2
622 2. Configure huge pages.
623 QEMU must allocate the VM's memory on hugetlbfs. vhost-user ports access
624 a virtio-net device's virtual rings and packet buffers mapping the VM's
625 physical memory on hugetlbfs. To enable vhost-user ports to map the VM's
626 memory into their process address space, pass the following paramters
630 -object memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,
632 -numa node,memdev=mem -mem-prealloc
635 3. Optional: Enable multiqueue support
636 The vhost-user interface must be configured in Open vSwitch with the
637 desired amount of queues with:
640 ovs-vsctl set Interface vhost-user-2 options:n_rxq=<requested queues>
643 QEMU needs to be configured as well.
644 The $q below should match the queues requested in OVS (if $q is more,
645 packets will not be received).
646 The $v is the number of vectors, which is '$q x 2 + 2'.
649 -chardev socket,id=char2,path=/usr/local/var/run/openvswitch/vhost-user-2
650 -netdev type=vhost-user,id=mynet2,chardev=char2,vhostforce,queues=$q
651 -device virtio-net-pci,mac=00:00:00:00:00:02,netdev=mynet2,mq=on,vectors=$v
654 If one wishes to use multiple queues for an interface in the guest, the
655 driver in the guest operating system must be configured to do so. It is
656 recommended that the number of queues configured be equal to '$q'.
658 For example, this can be done for the Linux kernel virtio-net driver with:
661 ethtool -L <DEV> combined <$q>
664 A note on the command above:
666 `-L`: Changes the numbers of channels of the specified network device
668 `combined`: Changes the number of multi-purpose channels.
673 The following sections describe the use of vhost-cuse 'dpdkvhostcuse' ports
676 DPDK vhost-cuse Prerequisites:
677 -------------------------
679 1. DPDK 16.04 with vhost support enabled as documented in the "Building and
681 As an additional step, you must enable vhost-cuse in DPDK by setting the
682 following additional flag in `config/common_base`:
684 `CONFIG_RTE_LIBRTE_VHOST_USER=n`
686 Following this, rebuild DPDK as per the instructions in the "Building and
687 Installing" section. Finally, rebuild OVS as per step 3 in the "Building
688 and Installing" section - OVS will detect that DPDK has vhost-cuse libraries
689 compiled and in turn will enable support for it in the switch and disable
692 2. Insert the Cuse module:
696 3. Build and insert the `eventfd_link` module:
699 cd $DPDK_DIR/lib/librte_vhost/eventfd_link/
701 insmod $DPDK_DIR/lib/librte_vhost/eventfd_link.ko
704 4. QEMU version v2.1.0+
706 vhost-cuse will work with QEMU v2.1.0 and above, however it is recommended to
707 use v2.2.0 if providing your VM with memory greater than 1GB due to potential
708 issues with memory mapping larger areas.
709 Note: QEMU v1.6.2 will also work, with slightly different command line parameters,
710 which are specified later in this document.
712 Adding DPDK vhost-cuse ports to the Switch:
713 --------------------------------------
715 Following the steps above to create a bridge, you can now add DPDK vhost-cuse
716 as a port to the vswitch. Unlike DPDK ring ports, DPDK vhost-cuse ports can have
719 - For vhost-cuse, the name of the port type is `dpdkvhostcuse`
722 ovs-vsctl add-port br0 vhost-cuse-1 -- set Interface vhost-cuse-1
726 When attaching vhost-cuse ports to QEMU, the name provided during the
727 add-port operation must match the ifname parameter on the QEMU command
728 line. More instructions on this can be found in the next section.
730 DPDK vhost-cuse VM configuration:
731 ---------------------------------
733 vhost-cuse ports use a Linux* character device to communicate with QEMU.
734 By default it is set to `/dev/vhost-net`. It is possible to reuse this
735 standard device for DPDK vhost, which makes setup a little simpler but it
736 is better practice to specify an alternative character device in order to
737 avoid any conflicts if kernel vhost is to be used in parallel.
739 1. This step is only needed if using an alternative character device.
741 The new character device filename must be specified in the ovsdb:
743 `./utilities/ovs-vsctl --no-wait set Open_vSwitch . \
744 other_config:cuse-dev-name=my-vhost-net`
746 In the example above, the character device to be used will be
749 2. This step is only needed if reusing the standard character device. It will
750 conflict with the kernel vhost character device so the user must first
753 `rm -rf /dev/vhost-net`
755 3a. Configure virtio-net adaptors:
756 The following parameters must be passed to the QEMU binary:
759 -netdev tap,id=<id>,script=no,downscript=no,ifname=<name>,vhost=on
760 -device virtio-net-pci,netdev=net1,mac=<mac>
763 Repeat the above parameters for multiple devices.
765 The DPDK vhost library will negiotiate its own features, so they
766 need not be passed in as command line params. Note that as offloads are
767 disabled this is the equivalent of setting:
769 `csum=off,gso=off,guest_tso4=off,guest_tso6=off,guest_ecn=off`
771 3b. If using an alternative character device. It must be also explicitly
772 passed to QEMU using the `vhostfd` argument:
775 -netdev tap,id=<id>,script=no,downscript=no,ifname=<name>,vhost=on,
777 -device virtio-net-pci,netdev=net1,mac=<mac>
780 The open file descriptor must be passed to QEMU running as a child
781 process. This could be done with a simple python script.
785 fd = os.open("/dev/usvhost", os.O_RDWR)
786 subprocess.call("qemu-system-x86_64 .... -netdev tap,id=vhostnet0,\
787 vhost=on,vhostfd=" + fd +"...", shell=True)
789 Alternatively the `qemu-wrap.py` script can be used to automate the
790 requirements specified above and can be used in conjunction with libvirt if
791 desired. See the "DPDK vhost VM configuration with QEMU wrapper" section
794 4. Configure huge pages:
795 QEMU must allocate the VM's memory on hugetlbfs. Vhost ports access a
796 virtio-net device's virtual rings and packet buffers mapping the VM's
797 physical memory on hugetlbfs. To enable vhost-ports to map the VM's
798 memory into their process address space, pass the following parameters
801 `-object memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,
802 share=on -numa node,memdev=mem -mem-prealloc`
804 Note: For use with an earlier QEMU version such as v1.6.2, use the
805 following to configure hugepages instead:
807 `-mem-path /dev/hugepages -mem-prealloc`
809 DPDK vhost-cuse VM configuration with QEMU wrapper:
810 ---------------------------------------------------
811 The QEMU wrapper script automatically detects and calls QEMU with the
812 necessary parameters. It performs the following actions:
814 * Automatically detects the location of the hugetlbfs and inserts this
815 into the command line parameters.
816 * Automatically open file descriptors for each virtio-net device and
817 inserts this into the command line parameters.
818 * Calls QEMU passing both the command line parameters passed to the
819 script itself and those it has auto-detected.
821 Before use, you **must** edit the configuration parameters section of the
822 script to point to the correct emulator location and set additional
823 settings. Of these settings, `emul_path` and `us_vhost_path` **must** be
824 set. All other settings are optional.
826 To use directly from the command line simply pass the wrapper some of the
827 QEMU parameters: it will configure the rest. For example:
830 qemu-wrap.py -cpu host -boot c -hda <disk image> -m 4096 -smp 4
831 --enable-kvm -nographic -vnc none -net none -netdev tap,id=net1,
832 script=no,downscript=no,ifname=if1,vhost=on -device virtio-net-pci,
833 netdev=net1,mac=00:00:00:00:00:01
836 DPDK vhost-cuse VM configuration with libvirt:
837 ----------------------------------------------
839 If you are using libvirt, you must enable libvirt to access the character
840 device by adding it to controllers cgroup for libvirtd using the following
843 1. In `/etc/libvirt/qemu.conf` add/edit the following lines:
846 1) clear_emulator_capabilities = 0
849 4) cgroup_device_acl = [
850 "/dev/null", "/dev/full", "/dev/zero",
851 "/dev/random", "/dev/urandom",
852 "/dev/ptmx", "/dev/kvm", "/dev/kqemu",
853 "/dev/rtc", "/dev/hpet", "/dev/net/tun",
854 "/dev/<my-vhost-device>",
858 <my-vhost-device> refers to "vhost-net" if using the `/dev/vhost-net`
859 device. If you have specificed a different name in the database
860 using the "other_config:cuse-dev-name" parameter, please specify that
863 2. Disable SELinux or set to permissive mode
865 3. Restart the libvirtd process
866 For example, on Fedora:
868 `systemctl restart libvirtd.service`
870 After successfully editing the configuration, you may launch your
871 vhost-enabled VM. The XML describing the VM can be configured like so
872 within the <qemu:commandline> section:
874 1. Set up shared hugepages:
877 <qemu:arg value='-object'/>
878 <qemu:arg value='memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,share=on'/>
879 <qemu:arg value='-numa'/>
880 <qemu:arg value='node,memdev=mem'/>
881 <qemu:arg value='-mem-prealloc'/>
884 2. Set up your tap devices:
887 <qemu:arg value='-netdev'/>
888 <qemu:arg value='type=tap,id=net1,script=no,downscript=no,ifname=vhost0,vhost=on'/>
889 <qemu:arg value='-device'/>
890 <qemu:arg value='virtio-net-pci,netdev=net1,mac=00:00:00:00:00:01'/>
893 Repeat for as many devices as are desired, modifying the id, ifname
894 and mac as necessary.
896 Again, if you are using an alternative character device (other than
897 `/dev/vhost-net`), please specify the file descriptor like so:
899 `<qemu:arg value='type=tap,id=net3,script=no,downscript=no,ifname=vhost0,vhost=on,vhostfd=<open_fd>'/>`
901 Where <open_fd> refers to the open file descriptor of the character device.
902 Instructions of how to retrieve the file descriptor can be found in the
903 "DPDK vhost VM configuration" section.
904 Alternatively, the process is automated with the qemu-wrap.py script,
905 detailed in the next section.
907 Now you may launch your VM using virt-manager, or like so:
909 `virsh create my_vhost_vm.xml`
911 DPDK vhost-cuse VM configuration with libvirt and QEMU wrapper:
912 ----------------------------------------------------------
914 To use the qemu-wrapper script in conjuntion with libvirt, follow the
915 steps in the previous section before proceeding with the following steps:
917 1. Place `qemu-wrap.py` in libvirtd's binary search PATH ($PATH)
918 Ideally in the same directory that the QEMU binary is located.
920 2. Ensure that the script has the same owner/group and file permissions
923 3. Update the VM xml file using "virsh edit VM.xml"
925 1. Set the VM to use the launch script.
926 Set the emulator path contained in the `<emulator><emulator/>` tags.
927 For example, replace:
929 `<emulator>/usr/bin/qemu-kvm<emulator/>`
933 `<emulator>/usr/bin/qemu-wrap.py<emulator/>`
935 4. Edit the Configuration Parameters section of the script to point to
936 the correct emulator location and set any additional options. If you are
937 using a alternative character device name, please set "us_vhost_path" to the
938 location of that device. The script will automatically detect and insert
939 the correct "vhostfd" value in the QEMU command line arguments.
941 5. Use virt-manager to launch the VM
943 Running ovs-vswitchd with DPDK backend inside a VM
944 --------------------------------------------------
946 Please note that additional configuration is required if you want to run
947 ovs-vswitchd with DPDK backend inside a QEMU virtual machine. Ovs-vswitchd
948 creates separate DPDK TX queues for each CPU core available. This operation
949 fails inside QEMU virtual machine because, by default, VirtIO NIC provided
950 to the guest is configured to support only single TX queue and single RX
951 queue. To change this behavior, you need to turn on 'mq' (multiqueue)
952 property of all virtio-net-pci devices emulated by QEMU and used by DPDK.
953 You may do it manually (by changing QEMU command line) or, if you use Libvirt,
954 by adding the following string:
956 `<driver name='vhost' queues='N'/>`
958 to <interface> sections of all network devices used by DPDK. Parameter 'N'
959 determines how many queues can be used by the guest.
964 - Work with 1500 MTU, needs few changes in DPDK lib to fix this issue.
965 - Currently DPDK port does not make use any offload functionality.
966 - DPDK-vHost support works with 1G huge pages.
969 - If you run Open vSwitch with smaller page sizes (e.g. 2MB), you may be
970 unable to share any rings or mempools with a virtual machine.
971 This is because the current implementation of ivshmem works by sharing
972 a single 1GB huge page from the host operating system to any guest
973 operating system through the Qemu ivshmem device. When using smaller
974 page sizes, multiple pages may be required to hold the ring descriptors
975 and buffer pools. The Qemu ivshmem device does not allow you to share
976 multiple file descriptors to the guest operating system. However, if you
977 want to share dpdkr rings with other processes on the host, you can do
978 this with smaller page sizes.
980 Platform and Network Interface:
981 - By default with DPDK 16.04, a maximum of 64 TX queues can be used with an
982 Intel XL710 Network Interface on a platform with more than 64 logical
983 cores. If a user attempts to add an XL710 interface as a DPDK port type to
984 a system as described above, an error will be reported that initialization
985 failed for the 65th queue. OVS will then roll back to the previous
986 successful queue initialization and use that value as the total number of
987 TX queues available with queue locking. If a user wishes to use more than
988 64 queues and avoid locking, then the
989 `CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_PF` config parameter in DPDK must be
990 increased to the desired number of queues. Both DPDK and OVS must be
991 recompiled for this change to take effect.
996 Please report problems to bugs@openvswitch.org.
998 [INSTALL.userspace.md]:INSTALL.userspace.md
999 [INSTALL.md]:INSTALL.md
1000 [DPDK Linux GSG]: http://www.dpdk.org/doc/guides/linux_gsg/build_dpdk.html#binding-and-unbinding-network-ports-to-from-the-igb-uioor-vfio-modules
1001 [DPDK Docs]: http://dpdk.org/doc