1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2016 Intel Corporation.
7 The i40e PMD (librte_pmd_i40e) provides poll mode driver support for
8 10/25/40 Gbps Intel® Ethernet 700 Series Network Adapters based on
9 the Intel Ethernet Controller X710/XL710/XXV710 and Intel Ethernet
10 Connection X722 (only support part of features).
16 Features of the i40e PMD are:
18 - Multiple queues for TX and RX
19 - Receiver Side Scaling (RSS)
21 - Packet type information
25 - VLAN/QinQ stripping and inserting
29 - Port hardware statistics
31 - Link state information
33 - Mirror on port, VLAN and VSI
34 - Interrupt mode for RX
35 - Scattered and gather for TX and RX
36 - Vector Poll mode driver
41 - IEEE1588/802.1AS timestamping
42 - VF Daemon (VFD) - EXPERIMENTAL
43 - Dynamic Device Personalization (DDP)
44 - Queue region configuration
45 - Virtual Function Port Representors
50 - Identifying your adapter using `Intel Support
51 <http://www.intel.com/support>`_ and get the latest NVM/FW images.
53 - Follow the DPDK :ref:`Getting Started Guide for Linux <linux_gsg>` to setup the basic DPDK environment.
55 - To get better performance on Intel platforms, please follow the "How to get best performance with NICs on Intel platforms"
56 section of the :ref:`Getting Started Guide for Linux <linux_gsg>`.
58 - Upgrade the NVM/FW version following the `Intel® Ethernet NVM Update Tool Quick Usage Guide for Linux
59 <https://www-ssl.intel.com/content/www/us/en/embedded/products/networking/nvm-update-tool-quick-linux-usage-guide.html>`_ and `Intel® Ethernet NVM Update Tool: Quick Usage Guide for EFI <https://www.intel.com/content/www/us/en/embedded/products/networking/nvm-update-tool-quick-efi-usage-guide.html>`_ if needed.
61 Recommended Matching List
62 -------------------------
64 It is highly recommended to upgrade the i40e kernel driver and firmware to
65 avoid the compatibility issues with i40e PMD. Here is the suggested matching
66 list which has been tested and verified. The detailed information can refer
67 to chapter Tested Platforms/Tested NICs in release notes.
69 +--------------+-----------------------+------------------+
70 | DPDK version | Kernel driver version | Firmware version |
71 +==============+=======================+==================+
72 | 18.05 | 2.4.6 | 6.01 |
73 +--------------+-----------------------+------------------+
74 | 18.02 | 2.4.3 | 6.01 |
75 +--------------+-----------------------+------------------+
76 | 17.11 | 2.1.26 | 6.01 |
77 +--------------+-----------------------+------------------+
78 | 17.08 | 2.0.19 | 6.01 |
79 +--------------+-----------------------+------------------+
80 | 17.05 | 1.5.23 | 5.05 |
81 +--------------+-----------------------+------------------+
82 | 17.02 | 1.5.23 | 5.05 |
83 +--------------+-----------------------+------------------+
84 | 16.11 | 1.5.23 | 5.05 |
85 +--------------+-----------------------+------------------+
86 | 16.07 | 1.4.25 | 5.04 |
87 +--------------+-----------------------+------------------+
88 | 16.04 | 1.4.25 | 5.02 |
89 +--------------+-----------------------+------------------+
91 Pre-Installation Configuration
92 ------------------------------
97 The following options can be modified in the ``config`` file.
98 Please note that enabling debugging options may affect system performance.
100 - ``CONFIG_RTE_LIBRTE_I40E_PMD`` (default ``y``)
102 Toggle compilation of the ``librte_pmd_i40e`` driver.
104 - ``CONFIG_RTE_LIBRTE_I40E_DEBUG_*`` (default ``n``)
106 Toggle display of generic debugging messages.
108 - ``CONFIG_RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC`` (default ``y``)
110 Toggle bulk allocation for RX.
112 - ``CONFIG_RTE_LIBRTE_I40E_INC_VECTOR`` (default ``n``)
114 Toggle the use of Vector PMD instead of normal RX/TX path.
115 To enable vPMD for RX, bulk allocation for Rx must be allowed.
117 - ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` (default ``n``)
119 Toggle to use a 16-byte RX descriptor, by default the RX descriptor is 32 byte.
121 - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_PF`` (default ``64``)
123 Number of queues reserved for PF.
125 - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM`` (default ``4``)
127 Number of queues reserved for each VMDQ Pool.
129 Runtime Config Options
130 ~~~~~~~~~~~~~~~~~~~~~~
132 - ``Number of Queues per VF`` (default ``4``)
134 The number of queue per VF is determined by its host PF. If the PCI address
135 of an i40e PF is aaaa:bb.cc, the number of queues per VF can be configured
136 with EAL parameter like -w aaaa:bb.cc,queue-num-per-vf=n. The value n can be
137 1, 2, 4, 8 or 16. If no such parameter is configured, the number of queues
138 per VF is 4 by default.
140 - ``Support multiple driver`` (default ``disable``)
142 There was a multiple driver support issue during use of 700 series Ethernet
143 Adapter with both Linux kernel and DPDK PMD. To fix this issue, ``devargs``
144 parameter ``support-multi-driver`` is introduced, for example::
146 -w 84:00.0,support-multi-driver=1
148 With the above configuration, DPDK PMD will not change global registers, and
149 will switch PF interrupt from IntN to Int0 to avoid interrupt conflict between
150 DPDK and Linux Kernel.
152 - ``Support VF Port Representor`` (default ``not enabled``)
154 The i40e PF PMD supports the creation of VF port representors for the control
155 and monitoring of i40e virtual function devices. Each port representor
156 corresponds to a single virtual function of that device. Using the ``devargs``
157 option ``representor`` the user can specify which virtual functions to create
158 port representors for on initialization of the PF PMD by passing the VF IDs of
159 the VFs which are required.::
161 -w DBDF,representor=[0,1,4]
163 Currently hot-plugging of representor ports is not supported so all required
164 representors must be specified on the creation of the PF.
166 - ``Use latest supported vector`` (default ``disable``)
168 Latest supported vector path may not always get the best perf so vector path was
169 recommended to use only on later platform. But users may want the latest vector path
170 since it can get better perf in some real work loading cases. So ``devargs`` param
171 ``use-latest-supported-vec`` is introduced, for example::
173 -w 84:00.0,use-latest-supported-vec=1
175 Driver compilation and testing
176 ------------------------------
178 Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
182 SR-IOV: Prerequisites and sample Application Notes
183 --------------------------------------------------
185 #. Load the kernel module:
187 .. code-block:: console
191 Check the output in dmesg:
193 .. code-block:: console
195 i40e 0000:83:00.1 ens802f0: renamed from eth0
197 #. Bring up the PF ports:
199 .. code-block:: console
203 #. Create VF device(s):
205 Echo the number of VFs to be created into the ``sriov_numvfs`` sysfs entry
210 .. code-block:: console
212 echo 2 > /sys/devices/pci0000:00/0000:00:03.0/0000:81:00.0/sriov_numvfs
215 #. Assign VF MAC address:
217 Assign MAC address to the VF using iproute2 utility. The syntax is:
219 .. code-block:: console
221 ip link set <PF netdev id> vf <VF id> mac <macaddr>
225 .. code-block:: console
227 ip link set ens802f0 vf 0 mac a0:b0:c0:d0:e0:f0
229 #. Assign VF to VM, and bring up the VM.
230 Please see the documentation for the *I40E/IXGBE/IGB Virtual Function Driver*.
234 Follow instructions available in the document
235 :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
240 .. code-block:: console
243 EAL: PCI device 0000:83:00.0 on NUMA socket 1
244 EAL: probe driver: 8086:1572 rte_i40e_pmd
245 EAL: PCI memory mapped at 0x7f7f80000000
246 EAL: PCI memory mapped at 0x7f7f80800000
247 PMD: eth_i40e_dev_init(): FW 5.0 API 1.5 NVM 05.00.02 eetrack 8000208a
248 Interactive-mode selected
249 Configuring Port 0 (socket 0)
252 PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are
253 satisfied.Rx Burst Bulk Alloc function will be used on port=0, queue=0.
256 Port 0: 68:05:CA:26:85:84
257 Checking link statuses...
258 Port 0 Link Up - speed 10000 Mbps - full-duplex
264 Sample Application Notes
265 ------------------------
270 Vlan filter only works when Promiscuous mode is off.
272 To start ``testpmd``, and add vlan 10 to port 0:
274 .. code-block:: console
276 ./app/testpmd -l 0-15 -n 4 -- -i --forward-mode=mac
279 testpmd> set promisc 0 off
280 testpmd> rx_vlan add 10 0
286 The Flow Director works in receive mode to identify specific flows or sets of flows and route them to specific queues.
287 The Flow Director filters can match the different fields for different type of packet: flow type, specific input set per flow type and the flexible payload.
289 The default input set of each flow type is::
291 ipv4-other : src_ip_address, dst_ip_address
292 ipv4-frag : src_ip_address, dst_ip_address
293 ipv4-tcp : src_ip_address, dst_ip_address, src_port, dst_port
294 ipv4-udp : src_ip_address, dst_ip_address, src_port, dst_port
295 ipv4-sctp : src_ip_address, dst_ip_address, src_port, dst_port,
297 ipv6-other : src_ip_address, dst_ip_address
298 ipv6-frag : src_ip_address, dst_ip_address
299 ipv6-tcp : src_ip_address, dst_ip_address, src_port, dst_port
300 ipv6-udp : src_ip_address, dst_ip_address, src_port, dst_port
301 ipv6-sctp : src_ip_address, dst_ip_address, src_port, dst_port,
303 l2_payload : ether_type
305 The flex payload is selected from offset 0 to 15 of packet's payload by default, while it is masked out from matching.
307 Start ``testpmd`` with ``--disable-rss`` and ``--pkt-filter-mode=perfect``:
309 .. code-block:: console
311 ./app/testpmd -l 0-15 -n 4 -- -i --disable-rss --pkt-filter-mode=perfect \
312 --rxq=8 --txq=8 --nb-cores=8 --nb-ports=1
314 Add a rule to direct ``ipv4-udp`` packet whose ``dst_ip=2.2.2.5, src_ip=2.2.2.3, src_port=32, dst_port=32`` to queue 1:
316 .. code-block:: console
318 testpmd> flow_director_filter 0 mode IP add flow ipv4-udp \
319 src 2.2.2.3 32 dst 2.2.2.5 32 vlan 0 flexbytes () \
320 fwd pf queue 1 fd_id 1
322 Check the flow director status:
324 .. code-block:: console
326 testpmd> show port fdir 0
328 ######################## FDIR infos for port 0 ####################
330 SUPPORTED FLOW TYPE: ipv4-frag ipv4-tcp ipv4-udp ipv4-sctp ipv4-other
331 ipv6-frag ipv6-tcp ipv6-udp ipv6-sctp ipv6-other
334 max_len: 16 payload_limit: 480
335 payload_unit: 2 payload_seg: 3
336 bitmask_unit: 2 bitmask_num: 2
339 src_ipv4: 0x00000000,
340 dst_ipv4: 0x00000000,
343 src_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000,
344 dst_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000
345 FLEX PAYLOAD SRC OFFSET:
346 L2_PAYLOAD: 0 1 2 3 4 5 6 ...
347 L3_PAYLOAD: 0 1 2 3 4 5 6 ...
348 L4_PAYLOAD: 0 1 2 3 4 5 6 ...
350 ipv4-udp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
351 ipv4-tcp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
352 ipv4-sctp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
353 ipv4-other: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
354 ipv4-frag: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
355 ipv6-udp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
356 ipv6-tcp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
357 ipv6-sctp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
358 ipv6-other: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
359 ipv6-frag: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
360 l2_payload: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
361 guarant_count: 1 best_count: 0
362 guarant_space: 512 best_space: 7168
369 Delete all flow director rules on a port:
371 .. code-block:: console
373 testpmd> flush_flow_director 0
378 The Intel® Ethernet 700 Series support a feature called
381 A Virtual Ethernet Bridge (VEB) is an IEEE Edge Virtual Bridging (EVB) term
382 for functionality that allows local switching between virtual endpoints within
383 a physical endpoint and also with an external bridge/network.
385 A "Floating" VEB doesn't have an uplink connection to the outside world so all
386 switching is done internally and remains within the host. As such, this
387 feature provides security benefits.
389 In addition, a Floating VEB overcomes a limitation of normal VEBs where they
390 cannot forward packets when the physical link is down. Floating VEBs don't need
391 to connect to the NIC port so they can still forward traffic from VF to VF
392 even when the physical link is down.
394 Therefore, with this feature enabled VFs can be limited to communicating with
395 each other but not an outside network, and they can do so even when there is
396 no physical uplink on the associated NIC port.
398 To enable this feature, the user should pass a ``devargs`` parameter to the
401 -w 84:00.0,enable_floating_veb=1
403 In this configuration the PMD will use the floating VEB feature for all the
404 VFs created by this PF device.
406 Alternatively, the user can specify which VFs need to connect to this floating
407 VEB using the ``floating_veb_list`` argument::
409 -w 84:00.0,enable_floating_veb=1,floating_veb_list=1;3-4
411 In this example ``VF1``, ``VF3`` and ``VF4`` connect to the floating VEB,
412 while other VFs connect to the normal VEB.
414 The current implementation only supports one floating VEB and one regular
415 VEB. VFs can connect to a floating VEB or a regular VEB according to the
416 configuration passed on the EAL command line.
418 The floating VEB functionality requires a NIC firmware version of 5.0
421 Dynamic Device Personalization (DDP)
422 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
424 The Intel® Ethernet 700 Series except for the Intel Ethernet Connection
425 X722 support a feature called "Dynamic Device Personalization (DDP)",
426 which is used to configure hardware by downloading a profile to support
427 protocols/filters which are not supported by default. The DDP
428 functionality requires a NIC firmware version of 6.0 or greater.
430 Current implementation supports GTP-C/GTP-U/PPPoE/PPPoL2TP,
431 steering can be used with rte_flow API.
433 GTPv1 package is released, and it can be downloaded from
434 https://downloadcenter.intel.com/download/27587.
436 PPPoE package is released, and it can be downloaded from
437 https://downloadcenter.intel.com/download/28040.
439 Load a profile which supports GTP and store backup profile:
441 .. code-block:: console
443 testpmd> ddp add 0 ./gtp.pkgo,./backup.pkgo
445 Delete a GTP profile and restore backup profile:
447 .. code-block:: console
449 testpmd> ddp del 0 ./backup.pkgo
451 Get loaded DDP package info list:
453 .. code-block:: console
455 testpmd> ddp get list 0
457 Display information about a GTP profile:
459 .. code-block:: console
461 testpmd> ddp get info ./gtp.pkgo
463 Input set configuration
464 ~~~~~~~~~~~~~~~~~~~~~~~
465 Input set for any PCTYPE can be configured with user defined configuration,
466 For example, to use only 48bit prefix for IPv6 src address for IPv6 TCP RSS:
468 .. code-block:: console
470 testpmd> port config 0 pctype 43 hash_inset clear all
471 testpmd> port config 0 pctype 43 hash_inset set field 13
472 testpmd> port config 0 pctype 43 hash_inset set field 14
473 testpmd> port config 0 pctype 43 hash_inset set field 15
475 Queue region configuration
476 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
477 The Intel® Ethernet 700 Series supports a feature of queue regions
478 configuration for RSS in the PF, so that different traffic classes or
479 different packet classification types can be separated to different
480 queues in different queue regions. There is an API for configuration
481 of queue regions in RSS with a command line. It can parse the parameters
482 of the region index, queue number, queue start index, user priority, traffic
483 classes and so on. Depending on commands from the command line, it will call
484 i40e private APIs and start the process of setting or flushing the queue
485 region configuration. As this feature is specific for i40e only private
486 APIs are used. These new ``test_pmd`` commands are as shown below. For
487 details please refer to :doc:`../testpmd_app_ug/index`.
489 .. code-block:: console
491 testpmd> set port (port_id) queue-region region_id (value) \
492 queue_start_index (value) queue_num (value)
493 testpmd> set port (port_id) queue-region region_id (value) flowtype (value)
494 testpmd> set port (port_id) queue-region UP (value) region_id (value)
495 testpmd> set port (port_id) queue-region flush (on|off)
496 testpmd> show port (port_id) queue-region
498 Limitations or Known issues
499 ---------------------------
501 MPLS packet classification
502 ~~~~~~~~~~~~~~~~~~~~~~~~~~
504 For firmware versions prior to 5.0, MPLS packets are not recognized by the NIC.
505 The L2 Payload flow type in flow director can be used to classify MPLS packet
506 by using a command in testpmd like:
508 testpmd> flow_director_filter 0 mode IP add flow l2_payload ether \
509 0x8847 flexbytes () fwd pf queue <N> fd_id <M>
511 With the NIC firmware version 5.0 or greater, some limited MPLS support
512 is added: Native MPLS (MPLS in Ethernet) skip is implemented, while no
513 new packet type, no classification or offload are possible. With this change,
514 L2 Payload flow type in flow director cannot be used to classify MPLS packet
515 as with previous firmware versions. Meanwhile, the Ethertype filter can be
516 used to classify MPLS packet by using a command in testpmd like:
518 testpmd> ethertype_filter 0 add mac_ignr 00:00:00:00:00:00 ethertype \
521 16 Byte RX Descriptor setting on DPDK VF
522 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
524 Currently the VF's RX descriptor mode is decided by PF. There's no PF-VF
525 interface for VF to request the RX descriptor mode, also no interface to notify
526 VF its own RX descriptor mode.
527 For all available versions of the i40e driver, these drivers don't support 16
528 byte RX descriptor. If the Linux i40e kernel driver is used as host driver,
529 while DPDK i40e PMD is used as the VF driver, DPDK cannot choose 16 byte receive
530 descriptor. The reason is that the RX descriptor is already set to 32 byte by
531 the i40e kernel driver. That is to say, user should keep
532 ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC=n`` in config file.
533 In the future, if the Linux i40e driver supports 16 byte RX descriptor, user
534 should make sure the DPDK VF uses the same RX descriptor mode, 16 byte or 32
535 byte, as the PF driver.
537 The same rule for DPDK PF + DPDK VF. The PF and VF should use the same RX
538 descriptor mode. Or the VF RX will not work.
540 Receive packets with Ethertype 0x88A8
541 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
543 Due to the FW limitation, PF can receive packets with Ethertype 0x88A8
544 only when floating VEB is disabled.
546 Incorrect Rx statistics when packet is oversize
547 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
549 When a packet is over maximum frame size, the packet is dropped.
550 However, the Rx statistics, when calling `rte_eth_stats_get` incorrectly
551 shows it as received.
553 VF & TC max bandwidth setting
554 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
556 The per VF max bandwidth and per TC max bandwidth cannot be enabled in parallel.
557 The behavior is different when handling per VF and per TC max bandwidth setting.
558 When enabling per VF max bandwidth, SW will check if per TC max bandwidth is
559 enabled. If so, return failure.
560 When enabling per TC max bandwidth, SW will check if per VF max bandwidth
561 is enabled. If so, disable per VF max bandwidth and continue with per TC max
564 TC TX scheduling mode setting
565 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
567 There're 2 TX scheduling modes for TCs, round robin and strict priority mode.
568 If a TC is set to strict priority mode, it can consume unlimited bandwidth.
569 It means if APP has set the max bandwidth for that TC, it comes to no
571 It's suggested to set the strict priority mode for a TC that is latency
572 sensitive but no consuming much bandwidth.
574 VF performance is impacted by PCI extended tag setting
575 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
577 To reach maximum NIC performance in the VF the PCI extended tag must be
578 enabled. The DPDK i40e PF driver will set this feature during initialization,
579 but the kernel PF driver does not. So when running traffic on a VF which is
580 managed by the kernel PF driver, a significant NIC performance downgrade has
581 been observed (for 64 byte packets, there is about 25% line-rate downgrade for
582 a 25GbE device and about 35% for a 40GbE device).
584 For kernel version >= 4.11, the kernel's PCI driver will enable the extended
585 tag if it detects that the device supports it. So by default, this is not an
586 issue. For kernels <= 4.11 or when the PCI extended tag is disabled it can be
587 enabled using the steps below.
589 #. Get the current value of the PCI configure register::
591 setpci -s <XX:XX.X> a8.w
595 value = value | 0x100
597 #. Set the PCI configure register with new value::
599 setpci -s <XX:XX.X> a8.w=<value>
604 The VF vlan strip function is only supported in the i40e kernel driver >= 2.1.26.
609 DCB works only when RSS is enabled.
611 Global configuration warning
612 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
614 I40E PMD will set some global registers to enable some function or set some
615 configure. Then when using different ports of the same NIC with Linux kernel
616 and DPDK, the port with Linux kernel will be impacted by the port with DPDK.
617 For example, register I40E_GL_SWT_L2TAGCTRL is used to control L2 tag, i40e
618 PMD uses I40E_GL_SWT_L2TAGCTRL to set vlan TPID. If setting TPID in port A
619 with DPDK, then the configuration will also impact port B in the NIC with
620 kernel driver, which don't want to use the TPID.
621 So PMD reports warning to clarify what is changed by writing global register.
623 High Performance of Small Packets on 40GbE NIC
624 ----------------------------------------------
626 As there might be firmware fixes for performance enhancement in latest version
627 of firmware image, the firmware update might be needed for getting high performance.
628 Check the Intel support website for the latest firmware updates.
629 Users should consult the release notes specific to a DPDK release to identify
630 the validated firmware version for a NIC using the i40e driver.
632 Use 16 Bytes RX Descriptor Size
633 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
635 As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets.
636 Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors.
638 Example of getting best performance with l3fwd example
639 ------------------------------------------------------
641 The following is an example of running the DPDK ``l3fwd`` sample application to get high performance with a
642 server with Intel Xeon processors and Intel Ethernet CNA XL710.
644 The example scenario is to get best performance with two Intel Ethernet CNA XL710 40GbE ports.
645 See :numref:`figure_intel_perf_test_setup` for the performance test setup.
647 .. _figure_intel_perf_test_setup:
649 .. figure:: img/intel_perf_test_setup.*
651 Performance Test Setup
654 1. Add two Intel Ethernet CNA XL710 to the platform, and use one port per card to get best performance.
655 The reason for using two NICs is to overcome a PCIe v3.0 limitation since it cannot provide 80GbE bandwidth
656 for two 40GbE ports, but two different PCIe v3.0 x8 slot can.
657 Refer to the sample NICs output above, then we can select ``82:00.0`` and ``85:00.0`` as test ports::
659 82:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583]
660 85:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583]
662 2. Connect the ports to the traffic generator. For high speed testing, it's best to use a hardware traffic generator.
664 3. Check the PCI devices numa node (socket id) and get the cores number on the exact socket id.
665 In this case, ``82:00.0`` and ``85:00.0`` are both in socket 1, and the cores on socket 1 in the referenced platform
667 Note: Don't use 2 logical cores on the same core (e.g core18 has 2 logical cores, core18 and core54), instead, use 2 logical
668 cores from different cores (e.g core18 and core19).
670 4. Bind these two ports to igb_uio.
672 5. As to Intel Ethernet CNA XL710 40GbE port, we need at least two queue pairs to achieve best performance, then two queues per port
673 will be required, and each queue pair will need a dedicated CPU core for receiving/transmitting packets.
675 6. The DPDK sample application ``l3fwd`` will be used for performance testing, with using two ports for bi-directional forwarding.
676 Compile the ``l3fwd sample`` with the default lpm mode.
678 7. The command line of running l3fwd would be something like the following::
680 ./l3fwd -l 18-21 -n 4 -w 82:00.0 -w 85:00.0 \
681 -- -p 0x3 --config '(0,0,18),(0,1,19),(1,0,20),(1,1,21)'
683 This means that the application uses core 18 for port 0, queue pair 0 forwarding, core 19 for port 0, queue pair 1 forwarding,
684 core 20 for port 1, queue pair 0 forwarding, and core 21 for port 1, queue pair 1 forwarding.
686 8. Configure the traffic at a traffic generator.
688 * Start creating a stream on packet generator.
690 * Set the Ethernet II type to 0x0800.