2 Copyright(c) 2016 Intel Corporation. All rights reserved.
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions
9 * Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
11 * Redistributions in binary form must reproduce the above copyright
12 notice, this list of conditions and the following disclaimer in
13 the documentation and/or other materials provided with the
15 * Neither the name of Intel Corporation nor the names of its
16 contributors may be used to endorse or promote products derived
17 from this software without specific prior written permission.
19 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 ======================
34 The I40E PMD (librte_pmd_i40e) provides poll mode driver support
35 for the Intel X710/XL710/X722 10/40 Gbps family of adapters.
41 Features of the I40E PMD are:
43 - Multiple queues for TX and RX
44 - Receiver Side Scaling (RSS)
46 - Packet type information
50 - VLAN/QinQ stripping and inserting
54 - Port hardware statistics
56 - Link state information
58 - Mirror on port, VLAN and VSI
59 - Interrupt mode for RX
60 - Scattered and gather for TX and RX
61 - Vector Poll mode driver
66 - IEEE1588/802.1AS timestamping
67 - VF Daemon (VFD) - EXPERIMENTAL
73 - Identifying your adapter using `Intel Support
74 <http://www.intel.com/support>`_ and get the latest NVM/FW images.
76 - Follow the DPDK :ref:`Getting Started Guide for Linux <linux_gsg>` to setup the basic DPDK environment.
78 - To get better performance on Intel platforms, please follow the "How to get best performance with NICs on Intel platforms"
79 section of the :ref:`Getting Started Guide for Linux <linux_gsg>`.
81 - Upgrade the NVM/FW version following the `IntelĀ® Ethernet NVM Update Tool Quick Usage Guide for Linux
82 <https://www-ssl.intel.com/content/www/us/en/embedded/products/networking/nvm-update-tool-quick-linux-usage-guide.html>`_ if needed.
84 Pre-Installation Configuration
85 ------------------------------
90 The following options can be modified in the ``config`` file.
91 Please note that enabling debugging options may affect system performance.
93 - ``CONFIG_RTE_LIBRTE_I40E_PMD`` (default ``y``)
95 Toggle compilation of the ``librte_pmd_i40e`` driver.
97 - ``CONFIG_RTE_LIBRTE_I40E_DEBUG_*`` (default ``n``)
99 Toggle display of generic debugging messages.
101 - ``CONFIG_RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC`` (default ``y``)
103 Toggle bulk allocation for RX.
105 - ``CONFIG_RTE_LIBRTE_I40E_INC_VECTOR`` (default ``n``)
107 Toggle the use of Vector PMD instead of normal RX/TX path.
108 To enable vPMD for RX, bulk allocation for Rx must be allowed.
110 - ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` (default ``n``)
112 Toggle to use a 16-byte RX descriptor, by default the RX descriptor is 32 byte.
114 - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_PF`` (default ``64``)
116 Number of queues reserved for PF.
118 - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM`` (default ``4``)
120 Number of queues reserved for each VMDQ Pool.
122 - ``CONFIG_RTE_LIBRTE_I40E_ITR_INTERVAL`` (default ``-1``)
124 Interrupt Throttling interval.
127 Runtime Config Options
128 ~~~~~~~~~~~~~~~~~~~~~~
130 - ``Number of Queues per VF`` (default ``4``)
132 The number of queue per VF is determined by its host PF. If the PCI address
133 of an i40e PF is aaaa:bb.cc, the number of queues per VF can be configured
134 with EAL parameter like -w aaaa:bb.cc,queue-num-per-vf=n. The value n can be
135 1, 2, 4, 8 or 16. If no such parameter is configured, the number of queues
136 per VF is 4 by default.
139 Driver compilation and testing
140 ------------------------------
142 Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
146 SR-IOV: Prerequisites and sample Application Notes
147 --------------------------------------------------
149 #. Load the kernel module:
151 .. code-block:: console
155 Check the output in dmesg:
157 .. code-block:: console
159 i40e 0000:83:00.1 ens802f0: renamed from eth0
161 #. Bring up the PF ports:
163 .. code-block:: console
167 #. Create VF device(s):
169 Echo the number of VFs to be created into the ``sriov_numvfs`` sysfs entry
174 .. code-block:: console
176 echo 2 > /sys/devices/pci0000:00/0000:00:03.0/0000:81:00.0/sriov_numvfs
179 #. Assign VF MAC address:
181 Assign MAC address to the VF using iproute2 utility. The syntax is:
183 .. code-block:: console
185 ip link set <PF netdev id> vf <VF id> mac <macaddr>
189 .. code-block:: console
191 ip link set ens802f0 vf 0 mac a0:b0:c0:d0:e0:f0
193 #. Assign VF to VM, and bring up the VM.
194 Please see the documentation for the *I40E/IXGBE/IGB Virtual Function Driver*.
198 Follow instructions available in the document
199 :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
204 .. code-block:: console
207 EAL: PCI device 0000:83:00.0 on NUMA socket 1
208 EAL: probe driver: 8086:1572 rte_i40e_pmd
209 EAL: PCI memory mapped at 0x7f7f80000000
210 EAL: PCI memory mapped at 0x7f7f80800000
211 PMD: eth_i40e_dev_init(): FW 5.0 API 1.5 NVM 05.00.02 eetrack 8000208a
212 Interactive-mode selected
213 Configuring Port 0 (socket 0)
216 PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are
217 satisfied.Rx Burst Bulk Alloc function will be used on port=0, queue=0.
220 Port 0: 68:05:CA:26:85:84
221 Checking link statuses...
222 Port 0 Link Up - speed 10000 Mbps - full-duplex
228 Sample Application Notes
229 ------------------------
234 Vlan filter only works when Promiscuous mode is off.
236 To start ``testpmd``, and add vlan 10 to port 0:
238 .. code-block:: console
240 ./app/testpmd -l 0-15 -n 4 -- -i --forward-mode=mac
243 testpmd> set promisc 0 off
244 testpmd> rx_vlan add 10 0
250 The Flow Director works in receive mode to identify specific flows or sets of flows and route them to specific queues.
251 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.
253 The default input set of each flow type is::
255 ipv4-other : src_ip_address, dst_ip_address
256 ipv4-frag : src_ip_address, dst_ip_address
257 ipv4-tcp : src_ip_address, dst_ip_address, src_port, dst_port
258 ipv4-udp : src_ip_address, dst_ip_address, src_port, dst_port
259 ipv4-sctp : src_ip_address, dst_ip_address, src_port, dst_port,
261 ipv6-other : src_ip_address, dst_ip_address
262 ipv6-frag : src_ip_address, dst_ip_address
263 ipv6-tcp : src_ip_address, dst_ip_address, src_port, dst_port
264 ipv6-udp : src_ip_address, dst_ip_address, src_port, dst_port
265 ipv6-sctp : src_ip_address, dst_ip_address, src_port, dst_port,
267 l2_payload : ether_type
269 The flex payload is selected from offset 0 to 15 of packet's payload by default, while it is masked out from matching.
271 Start ``testpmd`` with ``--disable-rss`` and ``--pkt-filter-mode=perfect``:
273 .. code-block:: console
275 ./app/testpmd -l 0-15 -n 4 -- -i --disable-rss --pkt-filter-mode=perfect \
276 --rxq=8 --txq=8 --nb-cores=8 --nb-ports=1
278 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:
280 .. code-block:: console
282 testpmd> flow_director_filter 0 mode IP add flow ipv4-udp \
283 src 2.2.2.3 32 dst 2.2.2.5 32 vlan 0 flexbytes () \
284 fwd pf queue 1 fd_id 1
286 Check the flow director status:
288 .. code-block:: console
290 testpmd> show port fdir 0
292 ######################## FDIR infos for port 0 ####################
294 SUPPORTED FLOW TYPE: ipv4-frag ipv4-tcp ipv4-udp ipv4-sctp ipv4-other
295 ipv6-frag ipv6-tcp ipv6-udp ipv6-sctp ipv6-other
298 max_len: 16 payload_limit: 480
299 payload_unit: 2 payload_seg: 3
300 bitmask_unit: 2 bitmask_num: 2
303 src_ipv4: 0x00000000,
304 dst_ipv4: 0x00000000,
307 src_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000,
308 dst_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000
309 FLEX PAYLOAD SRC OFFSET:
310 L2_PAYLOAD: 0 1 2 3 4 5 6 ...
311 L3_PAYLOAD: 0 1 2 3 4 5 6 ...
312 L4_PAYLOAD: 0 1 2 3 4 5 6 ...
314 ipv4-udp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
315 ipv4-tcp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
316 ipv4-sctp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
317 ipv4-other: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
318 ipv4-frag: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
319 ipv6-udp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
320 ipv6-tcp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
321 ipv6-sctp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
322 ipv6-other: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
323 ipv6-frag: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
324 l2_payload: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
325 guarant_count: 1 best_count: 0
326 guarant_space: 512 best_space: 7168
333 Delete all flow director rules on a port:
335 .. code-block:: console
337 testpmd> flush_flow_director 0
342 The IntelĀ® Ethernet Controller X710 and XL710 Family support a feature called
345 A Virtual Ethernet Bridge (VEB) is an IEEE Edge Virtual Bridging (EVB) term
346 for functionality that allows local switching between virtual endpoints within
347 a physical endpoint and also with an external bridge/network.
349 A "Floating" VEB doesn't have an uplink connection to the outside world so all
350 switching is done internally and remains within the host. As such, this
351 feature provides security benefits.
353 In addition, a Floating VEB overcomes a limitation of normal VEBs where they
354 cannot forward packets when the physical link is down. Floating VEBs don't need
355 to connect to the NIC port so they can still forward traffic from VF to VF
356 even when the physical link is down.
358 Therefore, with this feature enabled VFs can be limited to communicating with
359 each other but not an outside network, and they can do so even when there is
360 no physical uplink on the associated NIC port.
362 To enable this feature, the user should pass a ``devargs`` parameter to the
365 -w 84:00.0,enable_floating_veb=1
367 In this configuration the PMD will use the floating VEB feature for all the
368 VFs created by this PF device.
370 Alternatively, the user can specify which VFs need to connect to this floating
371 VEB using the ``floating_veb_list`` argument::
373 -w 84:00.0,enable_floating_veb=1,floating_veb_list=1;3-4
375 In this example ``VF1``, ``VF3`` and ``VF4`` connect to the floating VEB,
376 while other VFs connect to the normal VEB.
378 The current implementation only supports one floating VEB and one regular
379 VEB. VFs can connect to a floating VEB or a regular VEB according to the
380 configuration passed on the EAL command line.
382 The floating VEB functionality requires a NIC firmware version of 5.0
386 Limitations or Known issues
387 ---------------------------
389 MPLS packet classification on X710/XL710
390 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
392 For firmware versions prior to 5.0, MPLS packets are not recognized by the NIC.
393 The L2 Payload flow type in flow director can be used to classify MPLS packet
394 by using a command in testpmd like:
396 testpmd> flow_director_filter 0 mode IP add flow l2_payload ether \
397 0x8847 flexbytes () fwd pf queue <N> fd_id <M>
399 With the NIC firmware version 5.0 or greater, some limited MPLS support
400 is added: Native MPLS (MPLS in Ethernet) skip is implemented, while no
401 new packet type, no classification or offload are possible. With this change,
402 L2 Payload flow type in flow director cannot be used to classify MPLS packet
403 as with previous firmware versions. Meanwhile, the Ethertype filter can be
404 used to classify MPLS packet by using a command in testpmd like:
406 testpmd> ethertype_filter 0 add mac_ignr 00:00:00:00:00:00 ethertype \
409 16 Byte RX Descriptor setting on DPDK VF
410 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
412 Currently the VF's RX descriptor mode is decided by PF. There's no PF-VF
413 interface for VF to request the RX descriptor mode, also no interface to notify
414 VF its own RX descriptor mode.
415 For all available versions of the i40e driver, these drivers don't support 16
416 byte RX descriptor. If the Linux i40e kernel driver is used as host driver,
417 while DPDK i40e PMD is used as the VF driver, DPDK cannot choose 16 byte receive
418 descriptor. The reason is that the RX descriptor is already set to 32 byte by
419 the i40e kernel driver. That is to say, user should keep
420 ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC=n`` in config file.
421 In the future, if the Linux i40e driver supports 16 byte RX descriptor, user
422 should make sure the DPDK VF uses the same RX descriptor mode, 16 byte or 32
423 byte, as the PF driver.
425 The same rule for DPDK PF + DPDK VF. The PF and VF should use the same RX
426 descriptor mode. Or the VF RX will not work.
428 Receive packets with Ethertype 0x88A8
429 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
431 Due to the FW limitation, PF can receive packets with Ethertype 0x88A8
432 only when floating VEB is disabled.
434 Incorrect Rx statistics when packet is oversize
435 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
437 When a packet is over maximum frame size, the packet is dropped.
438 However the Rx statistics, when calling `rte_eth_stats_get` incorrectly
439 shows it as received.
441 VF & TC max bandwidth setting
442 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
444 The per VF max bandwidth and per TC max bandwidth cannot be enabled in parallel.
445 The dehavior is different when handling per VF and per TC max bandwidth setting.
446 When enabling per VF max bandwidth, SW will check if per TC max bandwidth is
447 enabled. If so, return failure.
448 When enabling per TC max bandwidth, SW will check if per VF max bandwidth
449 is enabled. If so, disable per VF max bandwidth and continue with per TC max
452 TC TX scheduling mode setting
453 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
455 There're 2 TX scheduling modes for TCs, round robin and strict priority mode.
456 If a TC is set to strict priority mode, it can consume unlimited bandwidth.
457 It means if APP has set the max bandwidth for that TC, it comes to no
459 It's suggested to set the strict priority mode for a TC that is latency
460 sensitive but no consuming much bandwidth.
462 VF performance is impacted by PCI extended tag setting
463 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
465 To reach maximum NIC performance in the VF the PCI extended tag must be
466 enabled. The DPDK I40E PF driver will set this feature during initialization,
467 but the kernel PF driver does not. So when running traffic on a VF which is
468 managed by the kernel PF driver, a significant NIC performance downgrade has
469 been observed (for 64 byte packets, there is about 25% linerate downgrade for
470 a 25G device and about 35% for a 40G device).
472 For kernel version >= 4.11, the kernel's PCI driver will enable the extended
473 tag if it detects that the device supports it. So by default, this is not an
474 issue. For kernels <= 4.11 or when the PCI extended tag is disabled it can be
475 enabled using the steps below.
477 #. Get the current value of the PCI configure register::
479 setpci -s <XX:XX.X> a8.w
483 value = value | 0x100
485 #. Set the PCI configure register with new value::
487 setpci -s <XX:XX.X> a8.w=<value>
492 The VF vlan strip function is only supported in the i40e kernel driver >= 2.1.26.
497 DCB works only when RSS is enabled.
499 High Performance of Small Packets on 40G NIC
500 --------------------------------------------
502 As there might be firmware fixes for performance enhancement in latest version
503 of firmware image, the firmware update might be needed for getting high performance.
504 Check with the local Intel's Network Division application engineers for firmware updates.
505 Users should consult the release notes specific to a DPDK release to identify
506 the validated firmware version for a NIC using the i40e driver.
508 Use 16 Bytes RX Descriptor Size
509 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
511 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.
512 Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors.
514 High Performance and per Packet Latency Tradeoff
515 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
517 Due to the hardware design, the interrupt signal inside NIC is needed for per
518 packet descriptor write-back. The minimum interval of interrupts could be set
519 at compile time by ``CONFIG_RTE_LIBRTE_I40E_ITR_INTERVAL`` in configuration files.
520 Though there is a default configuration, the interval could be tuned by the
521 users with that configuration item depends on what the user cares about more,
522 performance or per packet latency.
524 Example of getting best performance with l3fwd example
525 ------------------------------------------------------
527 The following is an example of running the DPDK ``l3fwd`` sample application to get high performance with an
528 Intel server platform and Intel XL710 NICs.
530 The example scenario is to get best performance with two Intel XL710 40GbE ports.
531 See :numref:`figure_intel_perf_test_setup` for the performance test setup.
533 .. _figure_intel_perf_test_setup:
535 .. figure:: img/intel_perf_test_setup.*
537 Performance Test Setup
540 1. Add two Intel XL710 NICs to the platform, and use one port per card to get best performance.
541 The reason for using two NICs is to overcome a PCIe Gen3's limitation since it cannot provide 80G bandwidth
542 for two 40G ports, but two different PCIe Gen3 x8 slot can.
543 Refer to the sample NICs output above, then we can select ``82:00.0`` and ``85:00.0`` as test ports::
545 82:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583]
546 85:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583]
548 2. Connect the ports to the traffic generator. For high speed testing, it's best to use a hardware traffic generator.
550 3. Check the PCI devices numa node (socket id) and get the cores number on the exact socket id.
551 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
553 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
554 cores from different cores (e.g core18 and core19).
556 4. Bind these two ports to igb_uio.
558 5. As to XL710 40G port, we need at least two queue pairs to achieve best performance, then two queues per port
559 will be required, and each queue pair will need a dedicated CPU core for receiving/transmitting packets.
561 6. The DPDK sample application ``l3fwd`` will be used for performance testing, with using two ports for bi-directional forwarding.
562 Compile the ``l3fwd sample`` with the default lpm mode.
564 7. The command line of running l3fwd would be something like the following::
566 ./l3fwd -l 18-21 -n 4 -w 82:00.0 -w 85:00.0 \
567 -- -p 0x3 --config '(0,0,18),(0,1,19),(1,0,20),(1,1,21)'
569 This means that the application uses core 18 for port 0, queue pair 0 forwarding, core 19 for port 0, queue pair 1 forwarding,
570 core 20 for port 1, queue pair 0 forwarding, and core 21 for port 1, queue pair 1 forwarding.
572 8. Configure the traffic at a traffic generator.
574 * Start creating a stream on packet generator.
576 * Set the Ethernet II type to 0x0800.