xref: /dpdk/doc/guides/nics/i40e.rst (revision 089e5ed727a15da2729cfee9b63533dd120bd04c)
1..  SPDX-License-Identifier: BSD-3-Clause
2    Copyright(c) 2016 Intel Corporation.
3
4I40E Poll Mode Driver
5======================
6
7The i40e PMD (librte_pmd_i40e) provides poll mode driver support for
810/25/40 Gbps Intel® Ethernet 700 Series Network Adapters based on
9the Intel Ethernet Controller X710/XL710/XXV710 and Intel Ethernet
10Connection X722 (only support part of features).
11
12
13Features
14--------
15
16Features of the i40e PMD are:
17
18- Multiple queues for TX and RX
19- Receiver Side Scaling (RSS)
20- MAC/VLAN filtering
21- Packet type information
22- Flow director
23- Cloud filter
24- Checksum offload
25- VLAN/QinQ stripping and inserting
26- TSO offload
27- Promiscuous mode
28- Multicast mode
29- Port hardware statistics
30- Jumbo frames
31- Link state information
32- Link flow control
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
37- DCB
38- VMDQ
39- SR-IOV VF
40- Hot plug
41- IEEE1588/802.1AS timestamping
42- VF Daemon (VFD) - EXPERIMENTAL
43- Dynamic Device Personalization (DDP)
44- Queue region configuration
45- Virtual Function Port Representors
46
47Prerequisites
48-------------
49
50- Identifying your adapter using `Intel Support
51  <http://www.intel.com/support>`_ and get the latest NVM/FW images.
52
53- Follow the DPDK :ref:`Getting Started Guide for Linux <linux_gsg>` to setup the basic DPDK environment.
54
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>`.
57
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.
60
61Recommended Matching List
62-------------------------
63
64It is highly recommended to upgrade the i40e kernel driver and firmware to
65avoid the compatibility issues with i40e PMD. Here is the suggested matching
66list which has been tested and verified. The detailed information can refer
67to chapter Tested Platforms/Tested NICs in release notes.
68
69   +--------------+-----------------------+------------------+
70   | DPDK version | Kernel driver version | Firmware version |
71   +==============+=======================+==================+
72   |    19.08     |         2.9.21        |       7.00       |
73   +--------------+-----------------------+------------------+
74   |    19.05     |         2.7.29        |       6.80       |
75   +--------------+-----------------------+------------------+
76   |    19.02     |         2.7.26        |       6.80       |
77   +--------------+-----------------------+------------------+
78   |    18.11     |         2.4.6         |       6.01       |
79   +--------------+-----------------------+------------------+
80   |    18.08     |         2.4.6         |       6.01       |
81   +--------------+-----------------------+------------------+
82   |    18.05     |         2.4.6         |       6.01       |
83   +--------------+-----------------------+------------------+
84   |    18.02     |         2.4.3         |       6.01       |
85   +--------------+-----------------------+------------------+
86   |    17.11     |         2.1.26        |       6.01       |
87   +--------------+-----------------------+------------------+
88   |    17.08     |         2.0.19        |       6.01       |
89   +--------------+-----------------------+------------------+
90   |    17.05     |         1.5.23        |       5.05       |
91   +--------------+-----------------------+------------------+
92   |    17.02     |         1.5.23        |       5.05       |
93   +--------------+-----------------------+------------------+
94   |    16.11     |         1.5.23        |       5.05       |
95   +--------------+-----------------------+------------------+
96   |    16.07     |         1.4.25        |       5.04       |
97   +--------------+-----------------------+------------------+
98   |    16.04     |         1.4.25        |       5.02       |
99   +--------------+-----------------------+------------------+
100
101Pre-Installation Configuration
102------------------------------
103
104Config File Options
105~~~~~~~~~~~~~~~~~~~
106
107The following options can be modified in the ``config`` file.
108Please note that enabling debugging options may affect system performance.
109
110- ``CONFIG_RTE_LIBRTE_I40E_PMD`` (default ``y``)
111
112  Toggle compilation of the ``librte_pmd_i40e`` driver.
113
114- ``CONFIG_RTE_LIBRTE_I40E_DEBUG_*`` (default ``n``)
115
116  Toggle display of generic debugging messages.
117
118- ``CONFIG_RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC`` (default ``y``)
119
120  Toggle bulk allocation for RX.
121
122- ``CONFIG_RTE_LIBRTE_I40E_INC_VECTOR`` (default ``n``)
123
124  Toggle the use of Vector PMD instead of normal RX/TX path.
125  To enable vPMD for RX, bulk allocation for Rx must be allowed.
126
127- ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` (default ``n``)
128
129  Toggle to use a 16-byte RX descriptor, by default the RX descriptor is 32 byte.
130
131- ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_PF`` (default ``64``)
132
133  Number of queues reserved for PF.
134
135- ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM`` (default ``4``)
136
137  Number of queues reserved for each VMDQ Pool.
138
139Runtime Config Options
140~~~~~~~~~~~~~~~~~~~~~~
141
142- ``Reserved number of Queues per VF`` (default ``4``)
143
144  The number of reserved queue per VF is determined by its host PF. If the
145  PCI address of an i40e PF is aaaa:bb.cc, the number of reserved queues per
146  VF can be configured with EAL parameter like -w aaaa:bb.cc,queue-num-per-vf=n.
147  The value n can be 1, 2, 4, 8 or 16. If no such parameter is configured, the
148  number of reserved queues per VF is 4 by default. If VF request more than
149  reserved queues per VF, PF will able to allocate max to 16 queues after a VF
150  reset.
151
152
153- ``Support multiple driver`` (default ``disable``)
154
155  There was a multiple driver support issue during use of 700 series Ethernet
156  Adapter with both Linux kernel and DPDK PMD. To fix this issue, ``devargs``
157  parameter ``support-multi-driver`` is introduced, for example::
158
159    -w 84:00.0,support-multi-driver=1
160
161  With the above configuration, DPDK PMD will not change global registers, and
162  will switch PF interrupt from IntN to Int0 to avoid interrupt conflict between
163  DPDK and Linux Kernel.
164
165- ``Support VF Port Representor`` (default ``not enabled``)
166
167  The i40e PF PMD supports the creation of VF port representors for the control
168  and monitoring of i40e virtual function devices. Each port representor
169  corresponds to a single virtual function of that device. Using the ``devargs``
170  option ``representor`` the user can specify which virtual functions to create
171  port representors for on initialization of the PF PMD by passing the VF IDs of
172  the VFs which are required.::
173
174  -w DBDF,representor=[0,1,4]
175
176  Currently hot-plugging of representor ports is not supported so all required
177  representors must be specified on the creation of the PF.
178
179- ``Use latest supported vector`` (default ``disable``)
180
181  Latest supported vector path may not always get the best perf so vector path was
182  recommended to use only on later platform. But users may want the latest vector path
183  since it can get better perf in some real work loading cases. So ``devargs`` param
184  ``use-latest-supported-vec`` is introduced, for example::
185
186  -w 84:00.0,use-latest-supported-vec=1
187
188Vector RX Pre-conditions
189~~~~~~~~~~~~~~~~~~~~~~~~
190For Vector RX it is assumed that the number of descriptor rings will be a power
191of 2. With this pre-condition, the ring pointer can easily scroll back to the
192head after hitting the tail without a conditional check. In addition Vector RX
193can use this assumption to do a bit mask using ``ring_size - 1``.
194
195Driver compilation and testing
196------------------------------
197
198Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
199for details.
200
201
202SR-IOV: Prerequisites and sample Application Notes
203--------------------------------------------------
204
205#. Load the kernel module:
206
207   .. code-block:: console
208
209      modprobe i40e
210
211   Check the output in dmesg:
212
213   .. code-block:: console
214
215      i40e 0000:83:00.1 ens802f0: renamed from eth0
216
217#. Bring up the PF ports:
218
219   .. code-block:: console
220
221      ifconfig ens802f0 up
222
223#. Create VF device(s):
224
225   Echo the number of VFs to be created into the ``sriov_numvfs`` sysfs entry
226   of the parent PF.
227
228   Example:
229
230   .. code-block:: console
231
232      echo 2 > /sys/devices/pci0000:00/0000:00:03.0/0000:81:00.0/sriov_numvfs
233
234
235#. Assign VF MAC address:
236
237   Assign MAC address to the VF using iproute2 utility. The syntax is:
238
239   .. code-block:: console
240
241      ip link set <PF netdev id> vf <VF id> mac <macaddr>
242
243   Example:
244
245   .. code-block:: console
246
247      ip link set ens802f0 vf 0 mac a0:b0:c0:d0:e0:f0
248
249#. Assign VF to VM, and bring up the VM.
250   Please see the documentation for the *I40E/IXGBE/IGB Virtual Function Driver*.
251
252#. Running testpmd:
253
254   Follow instructions available in the document
255   :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
256   to run testpmd.
257
258   Example output:
259
260   .. code-block:: console
261
262      ...
263      EAL: PCI device 0000:83:00.0 on NUMA socket 1
264      EAL: probe driver: 8086:1572 rte_i40e_pmd
265      EAL: PCI memory mapped at 0x7f7f80000000
266      EAL: PCI memory mapped at 0x7f7f80800000
267      PMD: eth_i40e_dev_init(): FW 5.0 API 1.5 NVM 05.00.02 eetrack 8000208a
268      Interactive-mode selected
269      Configuring Port 0 (socket 0)
270      ...
271
272      PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are
273      satisfied.Rx Burst Bulk Alloc function will be used on port=0, queue=0.
274
275      ...
276      Port 0: 68:05:CA:26:85:84
277      Checking link statuses...
278      Port 0 Link Up - speed 10000 Mbps - full-duplex
279      Done
280
281      testpmd>
282
283
284Sample Application Notes
285------------------------
286
287Vlan filter
288~~~~~~~~~~~
289
290Vlan filter only works when Promiscuous mode is off.
291
292To start ``testpmd``, and add vlan 10 to port 0:
293
294.. code-block:: console
295
296    ./app/testpmd -l 0-15 -n 4 -- -i --forward-mode=mac
297    ...
298
299    testpmd> set promisc 0 off
300    testpmd> rx_vlan add 10 0
301
302
303Flow Director
304~~~~~~~~~~~~~
305
306The Flow Director works in receive mode to identify specific flows or sets of flows and route them to specific queues.
307The 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.
308
309The default input set of each flow type is::
310
311   ipv4-other : src_ip_address, dst_ip_address
312   ipv4-frag  : src_ip_address, dst_ip_address
313   ipv4-tcp   : src_ip_address, dst_ip_address, src_port, dst_port
314   ipv4-udp   : src_ip_address, dst_ip_address, src_port, dst_port
315   ipv4-sctp  : src_ip_address, dst_ip_address, src_port, dst_port,
316                verification_tag
317   ipv6-other : src_ip_address, dst_ip_address
318   ipv6-frag  : src_ip_address, dst_ip_address
319   ipv6-tcp   : src_ip_address, dst_ip_address, src_port, dst_port
320   ipv6-udp   : src_ip_address, dst_ip_address, src_port, dst_port
321   ipv6-sctp  : src_ip_address, dst_ip_address, src_port, dst_port,
322                verification_tag
323   l2_payload : ether_type
324
325The flex payload is selected from offset 0 to 15 of packet's payload by default, while it is masked out from matching.
326
327Start ``testpmd`` with ``--disable-rss`` and ``--pkt-filter-mode=perfect``:
328
329.. code-block:: console
330
331   ./app/testpmd -l 0-15 -n 4 -- -i --disable-rss --pkt-filter-mode=perfect \
332                 --rxq=8 --txq=8 --nb-cores=8 --nb-ports=1
333
334Add 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:
335
336.. code-block:: console
337
338   testpmd> flow_director_filter 0 mode IP add flow ipv4-udp  \
339            src 2.2.2.3 32 dst 2.2.2.5 32 vlan 0 flexbytes () \
340	    fwd pf queue 1 fd_id 1
341
342Check the flow director status:
343
344.. code-block:: console
345
346   testpmd> show port fdir 0
347
348   ######################## FDIR infos for port 0      ####################
349     MODE:   PERFECT
350     SUPPORTED FLOW TYPE:  ipv4-frag ipv4-tcp ipv4-udp ipv4-sctp ipv4-other
351                           ipv6-frag ipv6-tcp ipv6-udp ipv6-sctp ipv6-other
352			   l2_payload
353     FLEX PAYLOAD INFO:
354     max_len:	    16	        payload_limit: 480
355     payload_unit:  2	        payload_seg:   3
356     bitmask_unit:  2	        bitmask_num:   2
357     MASK:
358       vlan_tci: 0x0000,
359       src_ipv4: 0x00000000,
360       dst_ipv4: 0x00000000,
361       src_port: 0x0000,
362       dst_port: 0x0000
363       src_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000,
364       dst_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000
365     FLEX PAYLOAD SRC OFFSET:
366       L2_PAYLOAD:    0      1	    2	   3	  4	 5	6  ...
367       L3_PAYLOAD:    0      1	    2	   3	  4	 5	6  ...
368       L4_PAYLOAD:    0      1	    2	   3	  4	 5	6  ...
369     FLEX MASK CFG:
370       ipv4-udp:    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
371       ipv4-tcp:    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
372       ipv4-sctp:   00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
373       ipv4-other:  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
374       ipv4-frag:   00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
375       ipv6-udp:    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
376       ipv6-tcp:    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
377       ipv6-sctp:   00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
378       ipv6-other:  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
379       ipv6-frag:   00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
380       l2_payload:  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
381     guarant_count: 1	        best_count:    0
382     guarant_space: 512         best_space:    7168
383     collision:     0	        free:	       0
384     maxhash:	    0	        maxlen:        0
385     add:	    0	        remove:        0
386     f_add:	    0	        f_remove:      0
387
388
389Delete all flow director rules on a port:
390
391.. code-block:: console
392
393   testpmd> flush_flow_director 0
394
395Floating VEB
396~~~~~~~~~~~~~
397
398The Intel® Ethernet 700 Series support a feature called
399"Floating VEB".
400
401A Virtual Ethernet Bridge (VEB) is an IEEE Edge Virtual Bridging (EVB) term
402for functionality that allows local switching between virtual endpoints within
403a physical endpoint and also with an external bridge/network.
404
405A "Floating" VEB doesn't have an uplink connection to the outside world so all
406switching is done internally and remains within the host. As such, this
407feature provides security benefits.
408
409In addition, a Floating VEB overcomes a limitation of normal VEBs where they
410cannot forward packets when the physical link is down. Floating VEBs don't need
411to connect to the NIC port so they can still forward traffic from VF to VF
412even when the physical link is down.
413
414Therefore, with this feature enabled VFs can be limited to communicating with
415each other but not an outside network, and they can do so even when there is
416no physical uplink on the associated NIC port.
417
418To enable this feature, the user should pass a ``devargs`` parameter to the
419EAL, for example::
420
421    -w 84:00.0,enable_floating_veb=1
422
423In this configuration the PMD will use the floating VEB feature for all the
424VFs created by this PF device.
425
426Alternatively, the user can specify which VFs need to connect to this floating
427VEB using the ``floating_veb_list`` argument::
428
429    -w 84:00.0,enable_floating_veb=1,floating_veb_list=1;3-4
430
431In this example ``VF1``, ``VF3`` and ``VF4`` connect to the floating VEB,
432while other VFs connect to the normal VEB.
433
434The current implementation only supports one floating VEB and one regular
435VEB. VFs can connect to a floating VEB or a regular VEB according to the
436configuration passed on the EAL command line.
437
438The floating VEB functionality requires a NIC firmware version of 5.0
439or greater.
440
441Dynamic Device Personalization (DDP)
442~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
443
444The Intel® Ethernet 700 Series except for the Intel Ethernet Connection
445X722 support a feature called "Dynamic Device Personalization (DDP)",
446which is used to configure hardware by downloading a profile to support
447protocols/filters which are not supported by default. The DDP
448functionality requires a NIC firmware version of 6.0 or greater.
449
450Current implementation supports GTP-C/GTP-U/PPPoE/PPPoL2TP,
451steering can be used with rte_flow API.
452
453GTPv1 package is released, and it can be downloaded from
454https://downloadcenter.intel.com/download/27587.
455
456PPPoE package is released, and it can be downloaded from
457https://downloadcenter.intel.com/download/28040.
458
459Load a profile which supports GTP and store backup profile:
460
461.. code-block:: console
462
463   testpmd> ddp add 0 ./gtp.pkgo,./backup.pkgo
464
465Delete a GTP profile and restore backup profile:
466
467.. code-block:: console
468
469   testpmd> ddp del 0 ./backup.pkgo
470
471Get loaded DDP package info list:
472
473.. code-block:: console
474
475   testpmd> ddp get list 0
476
477Display information about a GTP profile:
478
479.. code-block:: console
480
481   testpmd> ddp get info ./gtp.pkgo
482
483Input set configuration
484~~~~~~~~~~~~~~~~~~~~~~~
485Input set for any PCTYPE can be configured with user defined configuration,
486For example, to use only 48bit prefix for IPv6 src address for IPv6 TCP RSS:
487
488.. code-block:: console
489
490   testpmd> port config 0 pctype 43 hash_inset clear all
491   testpmd> port config 0 pctype 43 hash_inset set field 13
492   testpmd> port config 0 pctype 43 hash_inset set field 14
493   testpmd> port config 0 pctype 43 hash_inset set field 15
494
495Queue region configuration
496~~~~~~~~~~~~~~~~~~~~~~~~~~~
497The Intel® Ethernet 700 Series supports a feature of queue regions
498configuration for RSS in the PF, so that different traffic classes or
499different packet classification types can be separated to different
500queues in different queue regions. There is an API for configuration
501of queue regions in RSS with a command line. It can parse the parameters
502of the region index, queue number, queue start index, user priority, traffic
503classes and so on. Depending on commands from the command line, it will call
504i40e private APIs and start the process of setting or flushing the queue
505region configuration. As this feature is specific for i40e only private
506APIs are used. These new ``test_pmd`` commands are as shown below. For
507details please refer to :doc:`../testpmd_app_ug/index`.
508
509.. code-block:: console
510
511   testpmd> set port (port_id) queue-region region_id (value) \
512		queue_start_index (value) queue_num (value)
513   testpmd> set port (port_id) queue-region region_id (value) flowtype (value)
514   testpmd> set port (port_id) queue-region UP (value) region_id (value)
515   testpmd> set port (port_id) queue-region flush (on|off)
516   testpmd> show port (port_id) queue-region
517
518Limitations or Known issues
519---------------------------
520
521MPLS packet classification
522~~~~~~~~~~~~~~~~~~~~~~~~~~
523
524For firmware versions prior to 5.0, MPLS packets are not recognized by the NIC.
525The L2 Payload flow type in flow director can be used to classify MPLS packet
526by using a command in testpmd like:
527
528   testpmd> flow_director_filter 0 mode IP add flow l2_payload ether \
529            0x8847 flexbytes () fwd pf queue <N> fd_id <M>
530
531With the NIC firmware version 5.0 or greater, some limited MPLS support
532is added: Native MPLS (MPLS in Ethernet) skip is implemented, while no
533new packet type, no classification or offload are possible. With this change,
534L2 Payload flow type in flow director cannot be used to classify MPLS packet
535as with previous firmware versions. Meanwhile, the Ethertype filter can be
536used to classify MPLS packet by using a command in testpmd like:
537
538   testpmd> ethertype_filter 0 add mac_ignr 00:00:00:00:00:00 ethertype \
539            0x8847 fwd queue <M>
540
54116 Byte RX Descriptor setting on DPDK VF
542~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
543
544Currently the VF's RX descriptor mode is decided by PF. There's no PF-VF
545interface for VF to request the RX descriptor mode, also no interface to notify
546VF its own RX descriptor mode.
547For all available versions of the i40e driver, these drivers don't support 16
548byte RX descriptor. If the Linux i40e kernel driver is used as host driver,
549while DPDK i40e PMD is used as the VF driver, DPDK cannot choose 16 byte receive
550descriptor. The reason is that the RX descriptor is already set to 32 byte by
551the i40e kernel driver. That is to say, user should keep
552``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC=n`` in config file.
553In the future, if the Linux i40e driver supports 16 byte RX descriptor, user
554should make sure the DPDK VF uses the same RX descriptor mode, 16 byte or 32
555byte, as the PF driver.
556
557The same rule for DPDK PF + DPDK VF. The PF and VF should use the same RX
558descriptor mode. Or the VF RX will not work.
559
560Receive packets with Ethertype 0x88A8
561~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
562
563Due to the FW limitation, PF can receive packets with Ethertype 0x88A8
564only when floating VEB is disabled.
565
566Incorrect Rx statistics when packet is oversize
567~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
568
569When a packet is over maximum frame size, the packet is dropped.
570However, the Rx statistics, when calling `rte_eth_stats_get` incorrectly
571shows it as received.
572
573VF & TC max bandwidth setting
574~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
575
576The per VF max bandwidth and per TC max bandwidth cannot be enabled in parallel.
577The behavior is different when handling per VF and per TC max bandwidth setting.
578When enabling per VF max bandwidth, SW will check if per TC max bandwidth is
579enabled. If so, return failure.
580When enabling per TC max bandwidth, SW will check if per VF max bandwidth
581is enabled. If so, disable per VF max bandwidth and continue with per TC max
582bandwidth setting.
583
584TC TX scheduling mode setting
585~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
586
587There are 2 TX scheduling modes for TCs, round robin and strict priority mode.
588If a TC is set to strict priority mode, it can consume unlimited bandwidth.
589It means if APP has set the max bandwidth for that TC, it comes to no
590effect.
591It's suggested to set the strict priority mode for a TC that is latency
592sensitive but no consuming much bandwidth.
593
594VF performance is impacted by PCI extended tag setting
595~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
596
597To reach maximum NIC performance in the VF the PCI extended tag must be
598enabled. The DPDK i40e PF driver will set this feature during initialization,
599but the kernel PF driver does not. So when running traffic on a VF which is
600managed by the kernel PF driver, a significant NIC performance downgrade has
601been observed (for 64 byte packets, there is about 25% line-rate downgrade for
602a 25GbE device and about 35% for a 40GbE device).
603
604For kernel version >= 4.11, the kernel's PCI driver will enable the extended
605tag if it detects that the device supports it. So by default, this is not an
606issue. For kernels <= 4.11 or when the PCI extended tag is disabled it can be
607enabled using the steps below.
608
609#. Get the current value of the PCI configure register::
610
611      setpci -s <XX:XX.X> a8.w
612
613#. Set bit 8::
614
615      value = value | 0x100
616
617#. Set the PCI configure register with new value::
618
619      setpci -s <XX:XX.X> a8.w=<value>
620
621Vlan strip of VF
622~~~~~~~~~~~~~~~~
623
624The VF vlan strip function is only supported in the i40e kernel driver >= 2.1.26.
625
626DCB function
627~~~~~~~~~~~~
628
629DCB works only when RSS is enabled.
630
631Global configuration warning
632~~~~~~~~~~~~~~~~~~~~~~~~~~~~
633
634I40E PMD will set some global registers to enable some function or set some
635configure. Then when using different ports of the same NIC with Linux kernel
636and DPDK, the port with Linux kernel will be impacted by the port with DPDK.
637For example, register I40E_GL_SWT_L2TAGCTRL is used to control L2 tag, i40e
638PMD uses I40E_GL_SWT_L2TAGCTRL to set vlan TPID. If setting TPID in port A
639with DPDK, then the configuration will also impact port B in the NIC with
640kernel driver, which don't want to use the TPID.
641So PMD reports warning to clarify what is changed by writing global register.
642
643High Performance of Small Packets on 40GbE NIC
644----------------------------------------------
645
646As there might be firmware fixes for performance enhancement in latest version
647of firmware image, the firmware update might be needed for getting high performance.
648Check the Intel support website for the latest firmware updates.
649Users should consult the release notes specific to a DPDK release to identify
650the validated firmware version for a NIC using the i40e driver.
651
652Use 16 Bytes RX Descriptor Size
653~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
654
655As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets.
656Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors.
657
658Example of getting best performance with l3fwd example
659------------------------------------------------------
660
661The following is an example of running the DPDK ``l3fwd`` sample application to get high performance with a
662server with Intel Xeon processors and Intel Ethernet CNA XL710.
663
664The example scenario is to get best performance with two Intel Ethernet CNA XL710 40GbE ports.
665See :numref:`figure_intel_perf_test_setup` for the performance test setup.
666
667.. _figure_intel_perf_test_setup:
668
669.. figure:: img/intel_perf_test_setup.*
670
671   Performance Test Setup
672
673
6741. Add two Intel Ethernet CNA XL710 to the platform, and use one port per card to get best performance.
675   The reason for using two NICs is to overcome a PCIe v3.0 limitation since it cannot provide 80GbE bandwidth
676   for two 40GbE ports, but two different PCIe v3.0 x8 slot can.
677   Refer to the sample NICs output above, then we can select ``82:00.0`` and ``85:00.0`` as test ports::
678
679      82:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583]
680      85:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583]
681
6822. Connect the ports to the traffic generator. For high speed testing, it's best to use a hardware traffic generator.
683
6843. Check the PCI devices numa node (socket id) and get the cores number on the exact socket id.
685   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
686   are 18-35 and 54-71.
687   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
688   cores from different cores (e.g core18 and core19).
689
6904. Bind these two ports to igb_uio.
691
6925. As to Intel Ethernet CNA XL710 40GbE port, we need at least two queue pairs to achieve best performance, then two queues per port
693   will be required, and each queue pair will need a dedicated CPU core for receiving/transmitting packets.
694
6956. The DPDK sample application ``l3fwd`` will be used for performance testing, with using two ports for bi-directional forwarding.
696   Compile the ``l3fwd sample`` with the default lpm mode.
697
6987. The command line of running l3fwd would be something like the following::
699
700      ./l3fwd -l 18-21 -n 4 -w 82:00.0 -w 85:00.0 \
701              -- -p 0x3 --config '(0,0,18),(0,1,19),(1,0,20),(1,1,21)'
702
703   This means that the application uses core 18 for port 0, queue pair 0 forwarding, core 19 for port 0, queue pair 1 forwarding,
704   core 20 for port 1, queue pair 0 forwarding, and core 21 for port 1, queue pair 1 forwarding.
705
7068. Configure the traffic at a traffic generator.
707
708   * Start creating a stream on packet generator.
709
710   * Set the Ethernet II type to 0x0800.
711
712Tx bytes affected by the link status change
713~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
714
715For firmware versions prior to 6.01 for X710 series and 3.33 for X722 series, the tx_bytes statistics data is affected by
716the link down event. Each time the link status changes to down, the tx_bytes decreases 110 bytes.
717