.. BSD LICENSE Copyright(c) 2016 Intel Corporation. All rights reserved. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of Intel Corporation nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. I40E Poll Mode Driver ====================== The I40E PMD (librte_pmd_i40e) provides poll mode driver support for the Intel X710/XL710/X722 10/40 Gbps family of adapters. Features -------- Features of the I40E PMD are: - Multiple queues for TX and RX - Receiver Side Scaling (RSS) - MAC/VLAN filtering - Packet type information - Flow director - Cloud filter - Checksum offload - VLAN/QinQ stripping and inserting - TSO offload - Promiscuous mode - Multicast mode - Port hardware statistics - Jumbo frames - Link state information - Link flow control - Mirror on port, VLAN and VSI - Interrupt mode for RX - Scattered and gather for TX and RX - Vector Poll mode driver - DCB - VMDQ - SR-IOV VF - Hot plug - IEEE1588/802.1AS timestamping - VF Daemon (VFD) - EXPERIMENTAL Prerequisites ------------- - Identifying your adapter using `Intel Support `_ and get the latest NVM/FW images. - Follow the DPDK :ref:`Getting Started Guide for Linux ` to setup the basic DPDK environment. - To get better performance on Intel platforms, please follow the "How to get best performance with NICs on Intel platforms" section of the :ref:`Getting Started Guide for Linux `. - Upgrade the NVM/FW version following the `Intel® Ethernet NVM Update Tool Quick Usage Guide for Linux `_ if needed. Pre-Installation Configuration ------------------------------ Config File Options ~~~~~~~~~~~~~~~~~~~ The following options can be modified in the ``config`` file. Please note that enabling debugging options may affect system performance. - ``CONFIG_RTE_LIBRTE_I40E_PMD`` (default ``y``) Toggle compilation of the ``librte_pmd_i40e`` driver. - ``CONFIG_RTE_LIBRTE_I40E_DEBUG_*`` (default ``n``) Toggle display of generic debugging messages. - ``CONFIG_RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC`` (default ``y``) Toggle bulk allocation for RX. - ``CONFIG_RTE_LIBRTE_I40E_INC_VECTOR`` (default ``n``) Toggle the use of Vector PMD instead of normal RX/TX path. To enable vPMD for RX, bulk allocation for Rx must be allowed. - ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` (default ``n``) Toggle to use a 16-byte RX descriptor, by default the RX descriptor is 32 byte. - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_PF`` (default ``64``) Number of queues reserved for PF. - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VF`` (default ``4``) Number of queues reserved for each SR-IOV VF. - ``CONFIG_RTE_LIBRTE_I40E_QUEUE_NUM_PER_VM`` (default ``4``) Number of queues reserved for each VMDQ Pool. - ``CONFIG_RTE_LIBRTE_I40E_ITR_INTERVAL`` (default ``-1``) Interrupt Throttling interval. Driver compilation and testing ------------------------------ Refer to the document :ref:`compiling and testing a PMD for a NIC ` for details. SR-IOV: Prerequisites and sample Application Notes -------------------------------------------------- #. Load the kernel module: .. code-block:: console modprobe i40e Check the output in dmesg: .. code-block:: console i40e 0000:83:00.1 ens802f0: renamed from eth0 #. Bring up the PF ports: .. code-block:: console ifconfig ens802f0 up #. Create VF device(s): Echo the number of VFs to be created into the ``sriov_numvfs`` sysfs entry of the parent PF. Example: .. code-block:: console echo 2 > /sys/devices/pci0000:00/0000:00:03.0/0000:81:00.0/sriov_numvfs #. Assign VF MAC address: Assign MAC address to the VF using iproute2 utility. The syntax is: .. code-block:: console ip link set vf mac Example: .. code-block:: console ip link set ens802f0 vf 0 mac a0:b0:c0:d0:e0:f0 #. Assign VF to VM, and bring up the VM. Please see the documentation for the *I40E/IXGBE/IGB Virtual Function Driver*. #. Running testpmd: Follow instructions available in the document :ref:`compiling and testing a PMD for a NIC ` to run testpmd. Example output: .. code-block:: console ... EAL: PCI device 0000:83:00.0 on NUMA socket 1 EAL: probe driver: 8086:1572 rte_i40e_pmd EAL: PCI memory mapped at 0x7f7f80000000 EAL: PCI memory mapped at 0x7f7f80800000 PMD: eth_i40e_dev_init(): FW 5.0 API 1.5 NVM 05.00.02 eetrack 8000208a Interactive-mode selected Configuring Port 0 (socket 0) ... PMD: i40e_dev_rx_queue_setup(): Rx Burst Bulk Alloc Preconditions are satisfied.Rx Burst Bulk Alloc function will be used on port=0, queue=0. ... Port 0: 68:05:CA:26:85:84 Checking link statuses... Port 0 Link Up - speed 10000 Mbps - full-duplex Done testpmd> Sample Application Notes ------------------------ Vlan filter ~~~~~~~~~~~ Vlan filter only works when Promiscuous mode is off. To start ``testpmd``, and add vlan 10 to port 0: .. code-block:: console ./app/testpmd -l 0-15 -n 4 -- -i --forward-mode=mac ... testpmd> set promisc 0 off testpmd> rx_vlan add 10 0 Flow Director ~~~~~~~~~~~~~ The Flow Director works in receive mode to identify specific flows or sets of flows and route them to specific queues. 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. The default input set of each flow type is:: ipv4-other : src_ip_address, dst_ip_address ipv4-frag : src_ip_address, dst_ip_address ipv4-tcp : src_ip_address, dst_ip_address, src_port, dst_port ipv4-udp : src_ip_address, dst_ip_address, src_port, dst_port ipv4-sctp : src_ip_address, dst_ip_address, src_port, dst_port, verification_tag ipv6-other : src_ip_address, dst_ip_address ipv6-frag : src_ip_address, dst_ip_address ipv6-tcp : src_ip_address, dst_ip_address, src_port, dst_port ipv6-udp : src_ip_address, dst_ip_address, src_port, dst_port ipv6-sctp : src_ip_address, dst_ip_address, src_port, dst_port, verification_tag l2_payload : ether_type The flex payload is selected from offset 0 to 15 of packet's payload by default, while it is masked out from matching. Start ``testpmd`` with ``--disable-rss`` and ``--pkt-filter-mode=perfect``: .. code-block:: console ./app/testpmd -l 0-15 -n 4 -- -i --disable-rss --pkt-filter-mode=perfect \ --rxq=8 --txq=8 --nb-cores=8 --nb-ports=1 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: .. code-block:: console testpmd> flow_director_filter 0 mode IP add flow ipv4-udp \ src 2.2.2.3 32 dst 2.2.2.5 32 vlan 0 flexbytes () \ fwd pf queue 1 fd_id 1 Check the flow director status: .. code-block:: console testpmd> show port fdir 0 ######################## FDIR infos for port 0 #################### MODE: PERFECT SUPPORTED FLOW TYPE: ipv4-frag ipv4-tcp ipv4-udp ipv4-sctp ipv4-other ipv6-frag ipv6-tcp ipv6-udp ipv6-sctp ipv6-other l2_payload FLEX PAYLOAD INFO: max_len: 16 payload_limit: 480 payload_unit: 2 payload_seg: 3 bitmask_unit: 2 bitmask_num: 2 MASK: vlan_tci: 0x0000, src_ipv4: 0x00000000, dst_ipv4: 0x00000000, src_port: 0x0000, dst_port: 0x0000 src_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000, dst_ipv6: 0x00000000,0x00000000,0x00000000,0x00000000 FLEX PAYLOAD SRC OFFSET: L2_PAYLOAD: 0 1 2 3 4 5 6 ... L3_PAYLOAD: 0 1 2 3 4 5 6 ... L4_PAYLOAD: 0 1 2 3 4 5 6 ... FLEX MASK CFG: ipv4-udp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv4-tcp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv4-sctp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv4-other: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv4-frag: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv6-udp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv6-tcp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv6-sctp: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv6-other: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ipv6-frag: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 l2_payload: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 guarant_count: 1 best_count: 0 guarant_space: 512 best_space: 7168 collision: 0 free: 0 maxhash: 0 maxlen: 0 add: 0 remove: 0 f_add: 0 f_remove: 0 Delete all flow director rules on a port: .. code-block:: console testpmd> flush_flow_director 0 Floating VEB ~~~~~~~~~~~~~ The Intel® Ethernet Controller X710 and XL710 Family support a feature called "Floating VEB". A Virtual Ethernet Bridge (VEB) is an IEEE Edge Virtual Bridging (EVB) term for functionality that allows local switching between virtual endpoints within a physical endpoint and also with an external bridge/network. A "Floating" VEB doesn't have an uplink connection to the outside world so all switching is done internally and remains within the host. As such, this feature provides security benefits. In addition, a Floating VEB overcomes a limitation of normal VEBs where they cannot forward packets when the physical link is down. Floating VEBs don't need to connect to the NIC port so they can still forward traffic from VF to VF even when the physical link is down. Therefore, with this feature enabled VFs can be limited to communicating with each other but not an outside network, and they can do so even when there is no physical uplink on the associated NIC port. To enable this feature, the user should pass a ``devargs`` parameter to the EAL, for example:: -w 84:00.0,enable_floating_veb=1 In this configuration the PMD will use the floating VEB feature for all the VFs created by this PF device. Alternatively, the user can specify which VFs need to connect to this floating VEB using the ``floating_veb_list`` argument:: -w 84:00.0,enable_floating_veb=1,floating_veb_list=1;3-4 In this example ``VF1``, ``VF3`` and ``VF4`` connect to the floating VEB, while other VFs connect to the normal VEB. The current implementation only supports one floating VEB and one regular VEB. VFs can connect to a floating VEB or a regular VEB according to the configuration passed on the EAL command line. The floating VEB functionality requires a NIC firmware version of 5.0 or greater. Limitations or Known issues --------------------------- MPLS packet classification on X710/XL710 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For firmware versions prior to 5.0, MPLS packets are not recognized by the NIC. The L2 Payload flow type in flow director can be used to classify MPLS packet by using a command in testpmd like: testpmd> flow_director_filter 0 mode IP add flow l2_payload ether \ 0x8847 flexbytes () fwd pf queue fd_id With the NIC firmware version 5.0 or greater, some limited MPLS support is added: Native MPLS (MPLS in Ethernet) skip is implemented, while no new packet type, no classification or offload are possible. With this change, L2 Payload flow type in flow director cannot be used to classify MPLS packet as with previous firmware versions. Meanwhile, the Ethertype filter can be used to classify MPLS packet by using a command in testpmd like: testpmd> ethertype_filter 0 add mac_ignr 00:00:00:00:00:00 ethertype \ 0x8847 fwd queue 16 Byte RX Descriptor setting on DPDK VF ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Currently the VF's RX descriptor mode is decided by PF. There's no PF-VF interface for VF to request the RX descriptor mode, also no interface to notify VF its own RX descriptor mode. For all available versions of the i40e driver, these drivers don't support 16 byte RX descriptor. If the Linux i40e kernel driver is used as host driver, while DPDK i40e PMD is used as the VF driver, DPDK cannot choose 16 byte receive descriptor. The reason is that the RX descriptor is already set to 32 byte by the i40e kernel driver. That is to say, user should keep ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC=n`` in config file. In the future, if the Linux i40e driver supports 16 byte RX descriptor, user should make sure the DPDK VF uses the same RX descriptor mode, 16 byte or 32 byte, as the PF driver. The same rule for DPDK PF + DPDK VF. The PF and VF should use the same RX descriptor mode. Or the VF RX will not work. Receive packets with Ethertype 0x88A8 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Due to the FW limitation, PF can receive packets with Ethertype 0x88A8 only when floating VEB is disabled. Incorrect Rx statistics when packet is oversize ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ When a packet is over maximum frame size, the packet is dropped. However the Rx statistics, when calling `rte_eth_stats_get` incorrectly shows it as received. VF & TC max bandwidth setting ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The per VF max bandwidth and per TC max bandwidth cannot be enabled in parallel. The dehavior is different when handling per VF and per TC max bandwidth setting. When enabling per VF max bandwidth, SW will check if per TC max bandwidth is enabled. If so, return failure. When enabling per TC max bandwidth, SW will check if per VF max bandwidth is enabled. If so, disable per VF max bandwidth and continue with per TC max bandwidth setting. TC TX scheduling mode setting ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ There're 2 TX scheduling modes for TCs, round robin and strict priority mode. If a TC is set to strict priority mode, it can consume unlimited bandwidth. It means if APP has set the max bandwidth for that TC, it comes to no effect. It's suggested to set the strict priority mode for a TC that is latency sensitive but no consuming much bandwidth. VF performance is impacted by PCI extended tag setting ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To reach maximum NIC performance in the VF the PCI extended tag must be enabled. The DPDK I40E PF driver will set this feature during initialization, but the kernel PF driver does not. So when running traffic on a VF which is managed by the kernel PF driver, a significant NIC performance downgrade has been observed (for 64 byte packets, there is about 25% linerate downgrade for a 25G device and about 35% for a 40G device). For kernel version >= 4.11, the kernel's PCI driver will enable the extended tag if it detects that the device supports it. So by default, this is not an issue. For kernels <= 4.11 or when the PCI extended tag is disabled it can be enabled using the steps below. #. Get the current value of the PCI configure register:: setpci -s a8.w #. Set bit 8:: value = value | 0x100 #. Set the PCI configure register with new value:: setpci -s a8.w= Vlan strip of VF ~~~~~~~~~~~~~~~~ The VF vlan strip function is only supported in the i40e kernel driver >= 2.1.26. High Performance of Small Packets on 40G NIC -------------------------------------------- As there might be firmware fixes for performance enhancement in latest version of firmware image, the firmware update might be needed for getting high performance. Check with the local Intel's Network Division application engineers for firmware updates. Users should consult the release notes specific to a DPDK release to identify the validated firmware version for a NIC using the i40e driver. Use 16 Bytes RX Descriptor Size ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 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. Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors. High Performance and per Packet Latency Tradeoff ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Due to the hardware design, the interrupt signal inside NIC is needed for per packet descriptor write-back. The minimum interval of interrupts could be set at compile time by ``CONFIG_RTE_LIBRTE_I40E_ITR_INTERVAL`` in configuration files. Though there is a default configuration, the interval could be tuned by the users with that configuration item depends on what the user cares about more, performance or per packet latency. Example of getting best performance with l3fwd example ------------------------------------------------------ The following is an example of running the DPDK ``l3fwd`` sample application to get high performance with an Intel server platform and Intel XL710 NICs. The example scenario is to get best performance with two Intel XL710 40GbE ports. See :numref:`figure_intel_perf_test_setup` for the performance test setup. .. _figure_intel_perf_test_setup: .. figure:: img/intel_perf_test_setup.* Performance Test Setup 1. Add two Intel XL710 NICs to the platform, and use one port per card to get best performance. The reason for using two NICs is to overcome a PCIe Gen3's limitation since it cannot provide 80G bandwidth for two 40G ports, but two different PCIe Gen3 x8 slot can. Refer to the sample NICs output above, then we can select ``82:00.0`` and ``85:00.0`` as test ports:: 82:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583] 85:00.0 Ethernet [0200]: Intel XL710 for 40GbE QSFP+ [8086:1583] 2. Connect the ports to the traffic generator. For high speed testing, it's best to use a hardware traffic generator. 3. Check the PCI devices numa node (socket id) and get the cores number on the exact socket id. 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 are 18-35 and 54-71. 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 cores from different cores (e.g core18 and core19). 4. Bind these two ports to igb_uio. 5. As to XL710 40G port, we need at least two queue pairs to achieve best performance, then two queues per port will be required, and each queue pair will need a dedicated CPU core for receiving/transmitting packets. 6. The DPDK sample application ``l3fwd`` will be used for performance testing, with using two ports for bi-directional forwarding. Compile the ``l3fwd sample`` with the default lpm mode. 7. The command line of running l3fwd would be something like the following:: ./l3fwd -l 18-21 -n 4 -w 82:00.0 -w 85:00.0 \ -- -p 0x3 --config '(0,0,18),(0,1,19),(1,0,20),(1,1,21)' This means that the application uses core 18 for port 0, queue pair 0 forwarding, core 19 for port 0, queue pair 1 forwarding, core 20 for port 1, queue pair 0 forwarding, and core 21 for port 1, queue pair 1 forwarding. 8. Configure the traffic at a traffic generator. * Start creating a stream on packet generator. * Set the Ethernet II type to 0x0800.