1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2015 6WIND S.A.
3 Copyright 2015 Mellanox Technologies, Ltd
8 The MLX5 poll mode driver library (**librte_pmd_mlx5**) provides support
9 for **Mellanox ConnectX-4**, **Mellanox ConnectX-4 Lx** , **Mellanox
10 ConnectX-5** and **Mellanox Bluefield** families of 10/25/40/50/100 Gb/s
11 adapters as well as their virtual functions (VF) in SR-IOV context.
13 Information and documentation about these adapters can be found on the
14 `Mellanox website <http://www.mellanox.com>`__. Help is also provided by the
15 `Mellanox community <http://community.mellanox.com/welcome>`__.
17 There is also a `section dedicated to this poll mode driver
18 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`__.
22 Due to external dependencies, this driver is disabled by default. It must
23 be enabled manually by setting ``CONFIG_RTE_LIBRTE_MLX5_PMD=y`` and
26 Implementation details
27 ----------------------
29 Besides its dependency on libibverbs (that implies libmlx5 and associated
30 kernel support), librte_pmd_mlx5 relies heavily on system calls for control
31 operations such as querying/updating the MTU and flow control parameters.
33 For security reasons and robustness, this driver only deals with virtual
34 memory addresses. The way resources allocations are handled by the kernel
35 combined with hardware specifications that allow it to handle virtual memory
36 addresses directly ensure that DPDK applications cannot access random
37 physical memory (or memory that does not belong to the current process).
39 This capability allows the PMD to coexist with kernel network interfaces
40 which remain functional, although they stop receiving unicast packets as
41 long as they share the same MAC address.
42 This means legacy linux control tools (for example: ethtool, ifconfig and
43 more) can operate on the same network interfaces that owned by the DPDK
46 Enabling librte_pmd_mlx5 causes DPDK applications to be linked against
52 - Multi arch support: x86_64, POWER8, ARMv8, i686.
53 - Multiple TX and RX queues.
54 - Support for scattered TX and RX frames.
55 - IPv4, IPv6, TCPv4, TCPv6, UDPv4 and UDPv6 RSS on any number of queues.
56 - Several RSS hash keys, one for each flow type.
57 - Configurable RETA table.
58 - Support for multiple MAC addresses.
62 - RX CRC stripping configuration.
64 - Multicast promiscuous mode.
65 - Hardware checksum offloads.
66 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
70 - KVM and VMware ESX SR-IOV modes are supported.
71 - RSS hash result is supported.
72 - Hardware TSO for generic IP or UDP tunnel, including VXLAN and GRE.
73 - Hardware checksum Tx offload for generic IP or UDP tunnel, including VXLAN and GRE.
75 - Statistics query including Basic, Extended and per queue.
77 - Tunnel types: VXLAN, L3 VXLAN, VXLAN-GPE, GRE, MPLSoGRE, MPLSoUDP.
78 - Tunnel HW offloads: packet type, inner/outer RSS, IP and UDP checksum verification.
83 - For secondary process:
85 - Forked secondary process not supported.
86 - All mempools must be initialized before rte_eth_dev_start().
88 - Flow pattern without any specific vlan will match for vlan packets as well:
90 When VLAN spec is not specified in the pattern, the matching rule will be created with VLAN as a wild card.
91 Meaning, the flow rule::
93 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
95 Will only match vlan packets with vid=3. and the flow rules::
97 flow create 0 ingress pattern eth / ipv4 / end ...
101 flow create 0 ingress pattern eth / vlan / ipv4 / end ...
103 Will match any ipv4 packet (VLAN included).
105 - A multi segment packet must have less than 6 segments in case the Tx burst function
106 is set to multi-packet send or Enhanced multi-packet send. Otherwise it must have
107 less than 50 segments.
109 - Count action for RTE flow is **only supported in Mellanox OFED**.
111 - Flows with a VXLAN Network Identifier equal (or ends to be equal)
112 to 0 are not supported.
114 - VXLAN TSO and checksum offloads are not supported on VM.
116 - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with MPLSoGRE and MPLSoUDP.
118 - VF: flow rules created on VF devices can only match traffic targeted at the
119 configured MAC addresses (see ``rte_eth_dev_mac_addr_add()``).
123 MAC addresses not already present in the bridge table of the associated
124 kernel network device will be added and cleaned up by the PMD when closing
125 the device. In case of ungraceful program termination, some entries may
126 remain present and should be removed manually by other means.
128 - When Multi-Packet Rx queue is configured (``mprq_en``), a Rx packet can be
129 externally attached to a user-provided mbuf with having EXT_ATTACHED_MBUF in
130 ol_flags. As the mempool for the external buffer is managed by PMD, all the
131 Rx mbufs must be freed before the device is closed. Otherwise, the mempool of
132 the external buffers will be freed by PMD and the application which still
133 holds the external buffers may be corrupted.
135 - If Multi-Packet Rx queue is configured (``mprq_en``) and Rx CQE compression is
136 enabled (``rxq_cqe_comp_en``) at the same time, RSS hash result is not fully
137 supported. Some Rx packets may not have PKT_RX_RSS_HASH.
142 MLX5 supports various of methods to report statistics:
144 Port statistics can be queried using ``rte_eth_stats_get()``. The port statistics are through SW only and counts the number of packets received or sent successfully by the PMD.
146 Extended statistics can be queried using ``rte_eth_xstats_get()``. The extended statistics expose a wider set of counters counted by the device. The extended port statistics counts the number of packets received or sent successfully by the port. As Mellanox NICs are using the :ref:`Bifurcated Linux Driver <linux_gsg_linux_drivers>` those counters counts also packet received or sent by the Linux kernel. The counters with ``_phy`` suffix counts the total events on the physical port, therefore not valid for VF.
148 Finally per-flow statistics can by queried using ``rte_flow_query`` when attaching a count action for specific flow. The flow counter counts the number of packets received successfully by the port and match the specific flow.
156 These options can be modified in the ``.config`` file.
158 - ``CONFIG_RTE_LIBRTE_MLX5_PMD`` (default **n**)
160 Toggle compilation of librte_pmd_mlx5 itself.
162 - ``CONFIG_RTE_LIBRTE_MLX5_DLOPEN_DEPS`` (default **n**)
164 Build PMD with additional code to make it loadable without hard
165 dependencies on **libibverbs** nor **libmlx5**, which may not be installed
166 on the target system.
168 In this mode, their presence is still required for it to run properly,
169 however their absence won't prevent a DPDK application from starting (with
170 ``CONFIG_RTE_BUILD_SHARED_LIB`` disabled) and they won't show up as
171 missing with ``ldd(1)``.
173 It works by moving these dependencies to a purpose-built rdma-core "glue"
174 plug-in which must either be installed in a directory whose name is based
175 on ``CONFIG_RTE_EAL_PMD_PATH`` suffixed with ``-glue`` if set, or in a
176 standard location for the dynamic linker (e.g. ``/lib``) if left to the
177 default empty string (``""``).
179 This option has no performance impact.
181 - ``CONFIG_RTE_LIBRTE_MLX5_DEBUG`` (default **n**)
183 Toggle debugging code and stricter compilation flags. Enabling this option
184 adds additional run-time checks and debugging messages at the cost of
187 Environment variables
188 ~~~~~~~~~~~~~~~~~~~~~
192 A list of directories in which to search for the rdma-core "glue" plug-in,
193 separated by colons or semi-colons.
195 Only matters when compiled with ``CONFIG_RTE_LIBRTE_MLX5_DLOPEN_DEPS``
196 enabled and most useful when ``CONFIG_RTE_EAL_PMD_PATH`` is also set,
197 since ``LD_LIBRARY_PATH`` has no effect in this case.
199 - ``MLX5_PMD_ENABLE_PADDING``
201 Enables HW packet padding in PCI bus transactions.
203 When packet size is cache aligned and CRC stripping is enabled, 4 fewer
204 bytes are written to the PCI bus. Enabling padding makes such packets
207 In cases where PCI bandwidth is the bottleneck, padding can improve
210 This is disabled by default since this can also decrease performance for
211 unaligned packet sizes.
213 - ``MLX5_SHUT_UP_BF``
215 Configures HW Tx doorbell register as IO-mapped.
217 By default, the HW Tx doorbell is configured as a write-combining register.
218 The register would be flushed to HW usually when the write-combining buffer
219 becomes full, but it depends on CPU design.
221 Except for vectorized Tx burst routines, a write memory barrier is enforced
222 after updating the register so that the update can be immediately visible to
225 When vectorized Tx burst is called, the barrier is set only if the burst size
226 is not aligned to MLX5_VPMD_TX_MAX_BURST. However, setting this environmental
227 variable will bring better latency even though the maximum throughput can
230 Run-time configuration
231 ~~~~~~~~~~~~~~~~~~~~~~
233 - librte_pmd_mlx5 brings kernel network interfaces up during initialization
234 because it is affected by their state. Forcing them down prevents packets
237 - **ethtool** operations on related kernel interfaces also affect the PMD.
239 - ``rxq_cqe_comp_en`` parameter [int]
241 A nonzero value enables the compression of CQE on RX side. This feature
242 allows to save PCI bandwidth and improve performance. Enabled by default.
246 - x86_64 with ConnectX-4, ConnectX-4 LX, ConnectX-5 and Bluefield.
247 - POWER8 and ARMv8 with ConnectX-4 LX, ConnectX-5 and Bluefield.
249 - ``mprq_en`` parameter [int]
251 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
252 configured as Multi-Packet RQ if the total number of Rx queues is
253 ``rxqs_min_mprq`` or more and Rx scatter isn't configured. Disabled by
256 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
257 by posting a single large buffer for multiple packets. Instead of posting a
258 buffers per a packet, one large buffer is posted in order to receive multiple
259 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
260 and each stride receives one packet. MPRQ can improve throughput for
261 small-packet tarffic.
263 When MPRQ is enabled, max_rx_pkt_len can be larger than the size of
264 user-provided mbuf even if DEV_RX_OFFLOAD_SCATTER isn't enabled. PMD will
265 configure large stride size enough to accommodate max_rx_pkt_len as long as
266 device allows. Note that this can waste system memory compared to enabling Rx
267 scatter and multi-segment packet.
269 - ``mprq_log_stride_num`` parameter [int]
271 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
272 strides can reduce PCIe tarffic further. If configured value is not in the
273 range of device capability, the default value will be set with a warning
274 message. The default value is 4 which is 16 strides per a buffer, valid only
275 if ``mprq_en`` is set.
277 The size of Rx queue should be bigger than the number of strides.
279 - ``mprq_max_memcpy_len`` parameter [int]
281 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
282 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
283 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
284 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
285 A mempool for external buffers will be allocated and managed by PMD. If Rx
286 packet is externally attached, ol_flags field of the mbuf will have
287 EXT_ATTACHED_MBUF and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
288 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
290 - ``rxqs_min_mprq`` parameter [int]
292 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
293 greater or equal to this value. The default value is 12, valid only if
296 - ``txq_inline`` parameter [int]
298 Amount of data to be inlined during TX operations. Improves latency.
299 Can improve PPS performance when PCI back pressure is detected and may be
300 useful for scenarios involving heavy traffic on many queues.
302 Because additional software logic is necessary to handle this mode, this
303 option should be used with care, as it can lower performance when back
304 pressure is not expected.
306 - ``txqs_min_inline`` parameter [int]
308 Enable inline send only when the number of TX queues is greater or equal
311 This option should be used in combination with ``txq_inline`` above.
313 On ConnectX-4, ConnectX-4 LX, ConnectX-5 and Bluefield without
316 - Disabled by default.
317 - In case ``txq_inline`` is set recommendation is 4.
319 On ConnectX-5 and Bluefield with Enhanced MPW:
321 - Set to 8 by default.
323 - ``txq_mpw_en`` parameter [int]
325 A nonzero value enables multi-packet send (MPS) for ConnectX-4 Lx and
326 enhanced multi-packet send (Enhanced MPS) for ConnectX-5 and Bluefield.
327 MPS allows the TX burst function to pack up multiple packets in a
328 single descriptor session in order to save PCI bandwidth and improve
329 performance at the cost of a slightly higher CPU usage. When
330 ``txq_inline`` is set along with ``txq_mpw_en``, TX burst function tries
331 to copy entire packet data on to TX descriptor instead of including
332 pointer of packet only if there is enough room remained in the
333 descriptor. ``txq_inline`` sets per-descriptor space for either pointers
334 or inlined packets. In addition, Enhanced MPS supports hybrid mode -
335 mixing inlined packets and pointers in the same descriptor.
337 This option cannot be used with certain offloads such as ``DEV_TX_OFFLOAD_TCP_TSO,
338 DEV_TX_OFFLOAD_VXLAN_TNL_TSO, DEV_TX_OFFLOAD_GRE_TNL_TSO, DEV_TX_OFFLOAD_VLAN_INSERT``.
339 When those offloads are requested the MPS send function will not be used.
341 It is currently only supported on the ConnectX-4 Lx, ConnectX-5 and Bluefield
342 families of adapters. Enabled by default.
344 - ``txq_mpw_hdr_dseg_en`` parameter [int]
346 A nonzero value enables including two pointers in the first block of TX
347 descriptor. This can be used to lessen CPU load for memory copy.
349 Effective only when Enhanced MPS is supported. Disabled by default.
351 - ``txq_max_inline_len`` parameter [int]
353 Maximum size of packet to be inlined. This limits the size of packet to
354 be inlined. If the size of a packet is larger than configured value, the
355 packet isn't inlined even though there's enough space remained in the
356 descriptor. Instead, the packet is included with pointer.
358 Effective only when Enhanced MPS is supported. The default value is 256.
360 - ``tx_vec_en`` parameter [int]
362 A nonzero value enables Tx vector on ConnectX-5 and Bluefield NICs if the number of
363 global Tx queues on the port is lesser than MLX5_VPMD_MIN_TXQS.
365 This option cannot be used with certain offloads such as ``DEV_TX_OFFLOAD_TCP_TSO,
366 DEV_TX_OFFLOAD_VXLAN_TNL_TSO, DEV_TX_OFFLOAD_GRE_TNL_TSO, DEV_TX_OFFLOAD_VLAN_INSERT``.
367 When those offloads are requested the MPS send function will not be used.
369 Enabled by default on ConnectX-5 and Bluefield.
371 - ``rx_vec_en`` parameter [int]
373 A nonzero value enables Rx vector if the port is not configured in
374 multi-segment otherwise this parameter is ignored.
378 - ``vf_nl_en`` parameter [int]
380 A nonzero value enables Netlink requests from the VF to add/remove MAC
381 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
382 Otherwise the relevant configuration must be run with Linux iproute2 tools.
383 This is a prerequisite to receive this kind of traffic.
385 Enabled by default, valid only on VF devices ignored otherwise.
387 - ``l3_vxlan_en`` parameter [int]
389 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
390 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
391 parameter. This is a prerequisite to receive this kind of traffic.
395 - ``representor`` parameter [list]
397 This parameter can be used to instantiate DPDK Ethernet devices from
398 existing port (or VF) representors configured on the device.
400 It is a standard parameter whose format is described in
401 :ref:`ethernet_device_standard_device_arguments`.
403 For instance, to probe port representors 0 through 2::
407 Firmware configuration
408 ~~~~~~~~~~~~~~~~~~~~~~
410 - L3 VXLAN and VXLAN-GPE destination UDP port
412 .. code-block:: console
414 mlxconfig -d <mst device> set IP_OVER_VXLAN_EN=1
415 mlxconfig -d <mst device> set IP_OVER_VXLAN_PORT=<udp dport>
417 Verify configurations are set:
419 .. code-block:: console
421 mlxconfig -d <mst device> query | grep IP_OVER_VXLAN
422 IP_OVER_VXLAN_EN True(1)
423 IP_OVER_VXLAN_PORT <udp dport>
428 This driver relies on external libraries and kernel drivers for resources
429 allocations and initialization. The following dependencies are not part of
430 DPDK and must be installed separately:
434 User space Verbs framework used by librte_pmd_mlx5. This library provides
435 a generic interface between the kernel and low-level user space drivers
438 It allows slow and privileged operations (context initialization, hardware
439 resources allocations) to be managed by the kernel and fast operations to
440 never leave user space.
444 Low-level user space driver library for Mellanox
445 ConnectX-4/ConnectX-5/Bluefield devices, it is automatically loaded
448 This library basically implements send/receive calls to the hardware
453 They provide the kernel-side Verbs API and low level device drivers that
454 manage actual hardware initialization and resources sharing with user
457 Unlike most other PMDs, these modules must remain loaded and bound to
460 - mlx5_core: hardware driver managing Mellanox
461 ConnectX-4/ConnectX-5/Bluefield devices and related Ethernet kernel
463 - mlx5_ib: InifiniBand device driver.
464 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
466 - **Firmware update**
468 Mellanox OFED releases include firmware updates for
469 ConnectX-4/ConnectX-5/Bluefield adapters.
471 Because each release provides new features, these updates must be applied to
472 match the kernel modules and libraries they come with.
476 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
482 Either RDMA Core library with a recent enough Linux kernel release
483 (recommended) or Mellanox OFED, which provides compatibility with older
486 RMDA Core with Linux Kernel
487 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
489 - Minimal kernel version : v4.14 or the most recent 4.14-rc (see `Linux installation documentation`_)
490 - Minimal rdma-core version: v15+ commit 0c5f5765213a ("Merge pull request #227 from yishaih/tm")
491 (see `RDMA Core installation documentation`_)
492 - When building for i686 use:
494 - rdma-core version 18.0 or above built with 32bit support.
495 - Kernel version 4.14.41 or above.
497 .. _`Linux installation documentation`: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/plain/Documentation/admin-guide/README.rst
498 .. _`RDMA Core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
503 - Mellanox OFED version: **4.2, 4.3**.
506 - ConnectX-4: **12.21.1000** and above.
507 - ConnectX-4 Lx: **14.21.1000** and above.
508 - ConnectX-5: **16.21.1000** and above.
509 - ConnectX-5 Ex: **16.21.1000** and above.
510 - Bluefield: **18.99.3950** and above.
512 While these libraries and kernel modules are available on OpenFabrics
513 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
514 managers on most distributions, this PMD requires Ethernet extensions that
515 may not be supported at the moment (this is a work in progress).
518 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__
519 includes the necessary support and should be used in the meantime. For DPDK,
520 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
521 required from that distribution.
525 Several versions of Mellanox OFED are available. Installing the version
526 this DPDK release was developed and tested against is strongly
527 recommended. Please check the `prerequisites`_.
532 * Mellanox(R) ConnectX(R)-4 10G MCX4111A-XCAT (1x10G)
533 * Mellanox(R) ConnectX(R)-4 10G MCX4121A-XCAT (2x10G)
534 * Mellanox(R) ConnectX(R)-4 25G MCX4111A-ACAT (1x25G)
535 * Mellanox(R) ConnectX(R)-4 25G MCX4121A-ACAT (2x25G)
536 * Mellanox(R) ConnectX(R)-4 40G MCX4131A-BCAT (1x40G)
537 * Mellanox(R) ConnectX(R)-4 40G MCX413A-BCAT (1x40G)
538 * Mellanox(R) ConnectX(R)-4 40G MCX415A-BCAT (1x40G)
539 * Mellanox(R) ConnectX(R)-4 50G MCX4131A-GCAT (1x50G)
540 * Mellanox(R) ConnectX(R)-4 50G MCX413A-GCAT (1x50G)
541 * Mellanox(R) ConnectX(R)-4 50G MCX414A-BCAT (2x50G)
542 * Mellanox(R) ConnectX(R)-4 50G MCX415A-GCAT (2x50G)
543 * Mellanox(R) ConnectX(R)-4 50G MCX416A-BCAT (2x50G)
544 * Mellanox(R) ConnectX(R)-4 50G MCX416A-GCAT (2x50G)
545 * Mellanox(R) ConnectX(R)-4 50G MCX415A-CCAT (1x100G)
546 * Mellanox(R) ConnectX(R)-4 100G MCX416A-CCAT (2x100G)
547 * Mellanox(R) ConnectX(R)-4 Lx 10G MCX4121A-XCAT (2x10G)
548 * Mellanox(R) ConnectX(R)-4 Lx 25G MCX4121A-ACAT (2x25G)
549 * Mellanox(R) ConnectX(R)-5 100G MCX556A-ECAT (2x100G)
550 * Mellanox(R) ConnectX(R)-5 Ex EN 100G MCX516A-CDAT (2x100G)
552 Quick Start Guide on OFED
553 -------------------------
555 1. Download latest Mellanox OFED. For more info check the `prerequisites`_.
558 2. Install the required libraries and kernel modules either by installing
559 only the required set, or by installing the entire Mellanox OFED:
561 .. code-block:: console
563 ./mlnxofedinstall --upstream-libs --dpdk
565 3. Verify the firmware is the correct one:
567 .. code-block:: console
571 4. Verify all ports links are set to Ethernet:
573 .. code-block:: console
575 mlxconfig -d <mst device> query | grep LINK_TYPE
579 Link types may have to be configured to Ethernet:
581 .. code-block:: console
583 mlxconfig -d <mst device> set LINK_TYPE_P1/2=1/2/3
585 * LINK_TYPE_P1=<1|2|3> , 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
587 For hypervisors verify SR-IOV is enabled on the NIC:
589 .. code-block:: console
591 mlxconfig -d <mst device> query | grep SRIOV_EN
594 If needed, set enable the set the relevant fields:
596 .. code-block:: console
598 mlxconfig -d <mst device> set SRIOV_EN=1 NUM_OF_VFS=16
599 mlxfwreset -d <mst device> reset
601 5. Restart the driver:
603 .. code-block:: console
605 /etc/init.d/openibd restart
609 .. code-block:: console
611 service openibd restart
613 If link type was changed, firmware must be reset as well:
615 .. code-block:: console
617 mlxfwreset -d <mst device> reset
619 For hypervisors, after reset write the sysfs number of virtual functions
622 To dynamically instantiate a given number of virtual functions (VFs):
624 .. code-block:: console
626 echo [num_vfs] > /sys/class/infiniband/mlx5_0/device/sriov_numvfs
628 6. Compile DPDK and you are ready to go. See instructions on
629 :ref:`Development Kit Build System <Development_Kit_Build_System>`
634 1. Configure aggressive CQE Zipping for maximum performance:
636 .. code-block:: console
638 mlxconfig -d <mst device> s CQE_COMPRESSION=1
640 To set it back to the default CQE Zipping mode use:
642 .. code-block:: console
644 mlxconfig -d <mst device> s CQE_COMPRESSION=0
646 2. In case of virtualization:
648 - Make sure that hypervisor kernel is 3.16 or newer.
649 - Configure boot with ``iommu=pt``.
651 - Make sure to allocate a VM on huge pages.
652 - Make sure to set CPU pinning.
654 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
655 for better performance. For VMs, verify that the right CPU
656 and NUMA node are pinned according to the above. Run:
658 .. code-block:: console
662 to identify the NUMA node to which the PCIe adapter is connected.
664 4. If more than one adapter is used, and root complex capabilities allow
665 to put both adapters on the same NUMA node without PCI bandwidth degradation,
666 it is recommended to locate both adapters on the same NUMA node.
667 This in order to forward packets from one to the other without
668 NUMA performance penalty.
670 5. Disable pause frames:
672 .. code-block:: console
674 ethtool -A <netdev> rx off tx off
676 6. Verify IO non-posted prefetch is disabled by default. This can be checked
677 via the BIOS configuration. Please contact you server provider for more
678 information about the settings.
682 On some machines, depends on the machine integrator, it is beneficial
683 to set the PCI max read request parameter to 1K. This can be
684 done in the following way:
686 To query the read request size use:
688 .. code-block:: console
690 setpci -s <NIC PCI address> 68.w
692 If the output is different than 3XXX, set it by:
694 .. code-block:: console
696 setpci -s <NIC PCI address> 68.w=3XXX
698 The XXX can be different on different systems. Make sure to configure
699 according to the setpci output.
701 7. To minimize overhead of searching Memory Regions:
703 - '--socket-mem' is recommended to pin memory by predictable amount.
704 - Configure per-lcore cache when creating Mempools for packet buffer.
705 - Refrain from dynamically allocating/freeing memory in run-time.
710 Compared to librte_pmd_mlx4 that implements a single RSS configuration per
711 port, librte_pmd_mlx5 supports per-protocol RSS configuration.
713 Since ``testpmd`` defaults to IP RSS mode and there is currently no
714 command-line parameter to enable additional protocols (UDP and TCP as well
715 as IP), the following commands must be entered from its CLI to get the same
716 behavior as librte_pmd_mlx4:
718 .. code-block:: console
721 > port config all rss all
727 This section demonstrates how to launch **testpmd** with Mellanox
728 ConnectX-4/ConnectX-5/Bluefield devices managed by librte_pmd_mlx5.
730 #. Load the kernel modules:
732 .. code-block:: console
734 modprobe -a ib_uverbs mlx5_core mlx5_ib
736 Alternatively if MLNX_OFED is fully installed, the following script can
739 .. code-block:: console
741 /etc/init.d/openibd restart
745 User space I/O kernel modules (uio and igb_uio) are not used and do
746 not have to be loaded.
748 #. Make sure Ethernet interfaces are in working order and linked to kernel
749 verbs. Related sysfs entries should be present:
751 .. code-block:: console
753 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
757 .. code-block:: console
764 #. Optionally, retrieve their PCI bus addresses for whitelisting:
766 .. code-block:: console
769 for intf in eth2 eth3 eth4 eth5;
771 (cd "/sys/class/net/${intf}/device/" && pwd -P);
774 sed -n 's,.*/\(.*\),-w \1,p'
778 .. code-block:: console
785 #. Request huge pages:
787 .. code-block:: console
789 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
791 #. Start testpmd with basic parameters:
793 .. code-block:: console
795 testpmd -l 8-15 -n 4 -w 05:00.0 -w 05:00.1 -w 06:00.0 -w 06:00.1 -- --rxq=2 --txq=2 -i
799 .. code-block:: console
802 EAL: PCI device 0000:05:00.0 on NUMA socket 0
803 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
804 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
805 PMD: librte_pmd_mlx5: 1 port(s) detected
806 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
807 EAL: PCI device 0000:05:00.1 on NUMA socket 0
808 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
809 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
810 PMD: librte_pmd_mlx5: 1 port(s) detected
811 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
812 EAL: PCI device 0000:06:00.0 on NUMA socket 0
813 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
814 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
815 PMD: librte_pmd_mlx5: 1 port(s) detected
816 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
817 EAL: PCI device 0000:06:00.1 on NUMA socket 0
818 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
819 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
820 PMD: librte_pmd_mlx5: 1 port(s) detected
821 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
822 Interactive-mode selected
823 Configuring Port 0 (socket 0)
824 PMD: librte_pmd_mlx5: 0x8cba80: TX queues number update: 0 -> 2
825 PMD: librte_pmd_mlx5: 0x8cba80: RX queues number update: 0 -> 2
826 Port 0: E4:1D:2D:E7:0C:FE
827 Configuring Port 1 (socket 0)
828 PMD: librte_pmd_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
829 PMD: librte_pmd_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
830 Port 1: E4:1D:2D:E7:0C:FF
831 Configuring Port 2 (socket 0)
832 PMD: librte_pmd_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
833 PMD: librte_pmd_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
834 Port 2: E4:1D:2D:E7:0C:FA
835 Configuring Port 3 (socket 0)
836 PMD: librte_pmd_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
837 PMD: librte_pmd_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
838 Port 3: E4:1D:2D:E7:0C:FB
839 Checking link statuses...
840 Port 0 Link Up - speed 40000 Mbps - full-duplex
841 Port 1 Link Up - speed 40000 Mbps - full-duplex
842 Port 2 Link Up - speed 10000 Mbps - full-duplex
843 Port 3 Link Up - speed 10000 Mbps - full-duplex