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 - Default RSS operation with no hash key specification.
58 - Configurable RETA table.
59 - Support for multiple MAC addresses.
63 - RX CRC stripping configuration.
65 - Multicast promiscuous mode.
66 - Hardware checksum offloads.
67 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
71 - KVM and VMware ESX SR-IOV modes are supported.
72 - RSS hash result is supported.
73 - Hardware TSO for generic IP or UDP tunnel, including VXLAN and GRE.
74 - Hardware checksum Tx offload for generic IP or UDP tunnel, including VXLAN and GRE.
76 - Statistics query including Basic, Extended and per queue.
78 - Tunnel types: VXLAN, L3 VXLAN, VXLAN-GPE, GRE, MPLSoGRE, MPLSoUDP.
79 - Tunnel HW offloads: packet type, inner/outer RSS, IP and UDP checksum verification.
84 - For secondary process:
86 - Forked secondary process not supported.
87 - All mempools must be initialized before rte_eth_dev_start().
89 - Flow pattern without any specific vlan will match for vlan packets as well:
91 When VLAN spec is not specified in the pattern, the matching rule will be created with VLAN as a wild card.
92 Meaning, the flow rule::
94 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
96 Will only match vlan packets with vid=3. and the flow rules::
98 flow create 0 ingress pattern eth / ipv4 / end ...
102 flow create 0 ingress pattern eth / vlan / ipv4 / end ...
104 Will match any ipv4 packet (VLAN included).
106 - A multi segment packet must have less than 6 segments in case the Tx burst function
107 is set to multi-packet send or Enhanced multi-packet send. Otherwise it must have
108 less than 50 segments.
110 - Count action for RTE flow is **only supported in Mellanox OFED**.
112 - Flows with a VXLAN Network Identifier equal (or ends to be equal)
113 to 0 are not supported.
115 - VXLAN TSO and checksum offloads are not supported on VM.
117 - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with MPLSoGRE and MPLSoUDP.
119 - VF: flow rules created on VF devices can only match traffic targeted at the
120 configured MAC addresses (see ``rte_eth_dev_mac_addr_add()``).
124 MAC addresses not already present in the bridge table of the associated
125 kernel network device will be added and cleaned up by the PMD when closing
126 the device. In case of ungraceful program termination, some entries may
127 remain present and should be removed manually by other means.
129 - When Multi-Packet Rx queue is configured (``mprq_en``), a Rx packet can be
130 externally attached to a user-provided mbuf with having EXT_ATTACHED_MBUF in
131 ol_flags. As the mempool for the external buffer is managed by PMD, all the
132 Rx mbufs must be freed before the device is closed. Otherwise, the mempool of
133 the external buffers will be freed by PMD and the application which still
134 holds the external buffers may be corrupted.
136 - If Multi-Packet Rx queue is configured (``mprq_en``) and Rx CQE compression is
137 enabled (``rxq_cqe_comp_en``) at the same time, RSS hash result is not fully
138 supported. Some Rx packets may not have PKT_RX_RSS_HASH.
140 - IPv6 Multicast messages are not supported on VM, while promiscuous mode
141 and allmulticast mode are both set to off.
142 To receive IPv6 Multicast messages on VM, explicitly set the relevant
143 MAC address using rte_eth_dev_mac_addr_add() API.
145 - E-Switch VXLAN tunnel is not supported together with outer VLAN.
147 - E-Switch Flows with VNI pattern must include the VXLAN decapsulation action.
149 - E-Switch VXLAN decapsulation Flow:
151 - can be appiled to PF port only.
152 - must specify VF port action (packet redirection from PF to VF).
153 - must specify tunnel outer UDP local (destination) port, wildcards not allowed.
154 - must specify tunnel outer VNI, wildcards not allowed.
155 - must specify tunnel outer local (destination) IPv4 or IPv6 address, wildcards not allowed.
156 - optionally may specify tunnel outer remote (source) IPv4 or IPv6, wildcards or group IPs allowed.
157 - optionally may specify tunnel inner source and destination MAC addresses.
159 - E-Switch VXLAN encapsulation Flow:
161 - can be applied to VF ports only.
162 - must specify PF port action (packet redirection from VF to PF).
163 - must specify the VXLAN item with tunnel outer parameters.
164 - must specify the tunnel outer VNI in the VXLAN item.
165 - must specify the tunnel outer remote (destination) UDP port in the VXLAN item.
166 - must specify the tunnel outer local (source) IPv4 or IPv6 in the , this address will locally (with scope link) assigned to the outer network interace, wildcards not allowed.
167 - must specify the tunnel outer remote (destination) IPv4 or IPv6 in the VXLAN item, group IPs allowed.
168 - must specify the tunnel outer destination MAC address in the VXLAN item, this address will be used to create neigh rule.
173 MLX5 supports various of methods to report statistics:
175 Port statistics can be queried using ``rte_eth_stats_get()``. The received and sent statistics are through SW only and counts the number of packets received or sent successfully by the PMD. The imissed counter is the amount of packets that could not be delivered to SW because a queue was full. Packets not received due to congestion in the bus or on the NIC can be queried via the rx_discards_phy xstats counter.
177 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.
179 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.
187 These options can be modified in the ``.config`` file.
189 - ``CONFIG_RTE_LIBRTE_MLX5_PMD`` (default **n**)
191 Toggle compilation of librte_pmd_mlx5 itself.
193 - ``CONFIG_RTE_LIBRTE_MLX5_DLOPEN_DEPS`` (default **n**)
195 Build PMD with additional code to make it loadable without hard
196 dependencies on **libibverbs** nor **libmlx5**, which may not be installed
197 on the target system.
199 In this mode, their presence is still required for it to run properly,
200 however their absence won't prevent a DPDK application from starting (with
201 ``CONFIG_RTE_BUILD_SHARED_LIB`` disabled) and they won't show up as
202 missing with ``ldd(1)``.
204 It works by moving these dependencies to a purpose-built rdma-core "glue"
205 plug-in which must either be installed in a directory whose name is based
206 on ``CONFIG_RTE_EAL_PMD_PATH`` suffixed with ``-glue`` if set, or in a
207 standard location for the dynamic linker (e.g. ``/lib``) if left to the
208 default empty string (``""``).
210 This option has no performance impact.
212 - ``CONFIG_RTE_LIBRTE_MLX5_DEBUG`` (default **n**)
214 Toggle debugging code and stricter compilation flags. Enabling this option
215 adds additional run-time checks and debugging messages at the cost of
218 Environment variables
219 ~~~~~~~~~~~~~~~~~~~~~
223 A list of directories in which to search for the rdma-core "glue" plug-in,
224 separated by colons or semi-colons.
226 Only matters when compiled with ``CONFIG_RTE_LIBRTE_MLX5_DLOPEN_DEPS``
227 enabled and most useful when ``CONFIG_RTE_EAL_PMD_PATH`` is also set,
228 since ``LD_LIBRARY_PATH`` has no effect in this case.
230 - ``MLX5_PMD_ENABLE_PADDING``
232 Enables HW packet padding in PCI bus transactions.
234 When packet size is cache aligned and CRC stripping is enabled, 4 fewer
235 bytes are written to the PCI bus. Enabling padding makes such packets
238 In cases where PCI bandwidth is the bottleneck, padding can improve
241 This is disabled by default since this can also decrease performance for
242 unaligned packet sizes.
244 - ``MLX5_SHUT_UP_BF``
246 Configures HW Tx doorbell register as IO-mapped.
248 By default, the HW Tx doorbell is configured as a write-combining register.
249 The register would be flushed to HW usually when the write-combining buffer
250 becomes full, but it depends on CPU design.
252 Except for vectorized Tx burst routines, a write memory barrier is enforced
253 after updating the register so that the update can be immediately visible to
256 When vectorized Tx burst is called, the barrier is set only if the burst size
257 is not aligned to MLX5_VPMD_TX_MAX_BURST. However, setting this environmental
258 variable will bring better latency even though the maximum throughput can
261 Run-time configuration
262 ~~~~~~~~~~~~~~~~~~~~~~
264 - librte_pmd_mlx5 brings kernel network interfaces up during initialization
265 because it is affected by their state. Forcing them down prevents packets
268 - **ethtool** operations on related kernel interfaces also affect the PMD.
270 - ``rxq_cqe_comp_en`` parameter [int]
272 A nonzero value enables the compression of CQE on RX side. This feature
273 allows to save PCI bandwidth and improve performance. Enabled by default.
277 - x86_64 with ConnectX-4, ConnectX-4 LX, ConnectX-5 and Bluefield.
278 - POWER8 and ARMv8 with ConnectX-4 LX, ConnectX-5 and Bluefield.
280 - ``rxq_cqe_pad_en`` parameter [int]
282 A nonzero value enables 128B padding of CQE on RX side. The size of CQE
283 is aligned with the size of a cacheline of the core. If cacheline size is
284 128B, the CQE size is configured to be 128B even though the device writes
285 only 64B data on the cacheline. This is to avoid unnecessary cache
286 invalidation by device's two consecutive writes on to one cacheline.
287 However in some architecture, it is more beneficial to update entire
288 cacheline with padding the rest 64B rather than striding because
289 read-modify-write could drop performance a lot. On the other hand,
290 writing extra data will consume more PCIe bandwidth and could also drop
291 the maximum throughput. It is recommended to empirically set this
292 parameter. Disabled by default.
296 - CPU having 128B cacheline with ConnectX-5 and Bluefield.
298 - ``mprq_en`` parameter [int]
300 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
301 configured as Multi-Packet RQ if the total number of Rx queues is
302 ``rxqs_min_mprq`` or more and Rx scatter isn't configured. Disabled by
305 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
306 by posting a single large buffer for multiple packets. Instead of posting a
307 buffers per a packet, one large buffer is posted in order to receive multiple
308 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
309 and each stride receives one packet. MPRQ can improve throughput for
310 small-packet tarffic.
312 When MPRQ is enabled, max_rx_pkt_len can be larger than the size of
313 user-provided mbuf even if DEV_RX_OFFLOAD_SCATTER isn't enabled. PMD will
314 configure large stride size enough to accommodate max_rx_pkt_len as long as
315 device allows. Note that this can waste system memory compared to enabling Rx
316 scatter and multi-segment packet.
318 - ``mprq_log_stride_num`` parameter [int]
320 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
321 strides can reduce PCIe tarffic further. If configured value is not in the
322 range of device capability, the default value will be set with a warning
323 message. The default value is 4 which is 16 strides per a buffer, valid only
324 if ``mprq_en`` is set.
326 The size of Rx queue should be bigger than the number of strides.
328 - ``mprq_max_memcpy_len`` parameter [int]
330 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
331 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
332 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
333 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
334 A mempool for external buffers will be allocated and managed by PMD. If Rx
335 packet is externally attached, ol_flags field of the mbuf will have
336 EXT_ATTACHED_MBUF and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
337 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
339 - ``rxqs_min_mprq`` parameter [int]
341 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
342 greater or equal to this value. The default value is 12, valid only if
345 - ``txq_inline`` parameter [int]
347 Amount of data to be inlined during TX operations. Improves latency.
348 Can improve PPS performance when PCI back pressure is detected and may be
349 useful for scenarios involving heavy traffic on many queues.
351 Because additional software logic is necessary to handle this mode, this
352 option should be used with care, as it can lower performance when back
353 pressure is not expected.
355 - ``txqs_min_inline`` parameter [int]
357 Enable inline send only when the number of TX queues is greater or equal
360 This option should be used in combination with ``txq_inline`` above.
362 On ConnectX-4, ConnectX-4 LX, ConnectX-5 and Bluefield without
365 - Disabled by default.
366 - In case ``txq_inline`` is set recommendation is 4.
368 On ConnectX-5 and Bluefield with Enhanced MPW:
370 - Set to 8 by default.
372 - ``txqs_max_vec`` parameter [int]
374 Enable vectorized Tx only when the number of TX queues is less than or
375 equal to this value. Effective only when ``tx_vec_en`` is enabled.
379 - Set to 8 by default on ARMv8.
380 - Set to 4 by default otherwise.
384 - Set to 16 by default.
386 - ``txq_mpw_en`` parameter [int]
388 A nonzero value enables multi-packet send (MPS) for ConnectX-4 Lx and
389 enhanced multi-packet send (Enhanced MPS) for ConnectX-5 and Bluefield.
390 MPS allows the TX burst function to pack up multiple packets in a
391 single descriptor session in order to save PCI bandwidth and improve
392 performance at the cost of a slightly higher CPU usage. When
393 ``txq_inline`` is set along with ``txq_mpw_en``, TX burst function tries
394 to copy entire packet data on to TX descriptor instead of including
395 pointer of packet only if there is enough room remained in the
396 descriptor. ``txq_inline`` sets per-descriptor space for either pointers
397 or inlined packets. In addition, Enhanced MPS supports hybrid mode -
398 mixing inlined packets and pointers in the same descriptor.
400 This option cannot be used with certain offloads such as ``DEV_TX_OFFLOAD_TCP_TSO,
401 DEV_TX_OFFLOAD_VXLAN_TNL_TSO, DEV_TX_OFFLOAD_GRE_TNL_TSO, DEV_TX_OFFLOAD_VLAN_INSERT``.
402 When those offloads are requested the MPS send function will not be used.
404 It is currently only supported on the ConnectX-4 Lx, ConnectX-5 and Bluefield
405 families of adapters.
406 On ConnectX-4 Lx the MPW is considered un-secure hence disabled by default.
407 Users which enable the MPW should be aware that application which provides incorrect
408 mbuf descriptors in the Tx burst can lead to serious errors in the host including, on some cases,
410 On ConnectX-5 and Bluefield the MPW is secure and enabled by default.
412 - ``txq_mpw_hdr_dseg_en`` parameter [int]
414 A nonzero value enables including two pointers in the first block of TX
415 descriptor. This can be used to lessen CPU load for memory copy.
417 Effective only when Enhanced MPS is supported. Disabled by default.
419 - ``txq_max_inline_len`` parameter [int]
421 Maximum size of packet to be inlined. This limits the size of packet to
422 be inlined. If the size of a packet is larger than configured value, the
423 packet isn't inlined even though there's enough space remained in the
424 descriptor. Instead, the packet is included with pointer.
426 Effective only when Enhanced MPS is supported. The default value is 256.
428 - ``tx_vec_en`` parameter [int]
430 A nonzero value enables Tx vector on ConnectX-5 and Bluefield NICs if the number of
431 global Tx queues on the port is less than ``txqs_max_vec``.
433 This option cannot be used with certain offloads such as ``DEV_TX_OFFLOAD_TCP_TSO,
434 DEV_TX_OFFLOAD_VXLAN_TNL_TSO, DEV_TX_OFFLOAD_GRE_TNL_TSO, DEV_TX_OFFLOAD_VLAN_INSERT``.
435 When those offloads are requested the MPS send function will not be used.
437 Enabled by default on ConnectX-5 and Bluefield.
439 - ``rx_vec_en`` parameter [int]
441 A nonzero value enables Rx vector if the port is not configured in
442 multi-segment otherwise this parameter is ignored.
446 - ``vf_nl_en`` parameter [int]
448 A nonzero value enables Netlink requests from the VF to add/remove MAC
449 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
450 Otherwise the relevant configuration must be run with Linux iproute2 tools.
451 This is a prerequisite to receive this kind of traffic.
453 Enabled by default, valid only on VF devices ignored otherwise.
455 - ``l3_vxlan_en`` parameter [int]
457 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
458 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
459 parameter. This is a prerequisite to receive this kind of traffic.
463 - ``dv_flow_en`` parameter [int]
465 A nonzero value enables the DV flow steering assuming it is supported
467 The DV flow steering is not supported on switchdev mode.
471 - ``representor`` parameter [list]
473 This parameter can be used to instantiate DPDK Ethernet devices from
474 existing port (or VF) representors configured on the device.
476 It is a standard parameter whose format is described in
477 :ref:`ethernet_device_standard_device_arguments`.
479 For instance, to probe port representors 0 through 2::
483 Firmware configuration
484 ~~~~~~~~~~~~~~~~~~~~~~
486 - L3 VXLAN and VXLAN-GPE destination UDP port
488 .. code-block:: console
490 mlxconfig -d <mst device> set IP_OVER_VXLAN_EN=1
491 mlxconfig -d <mst device> set IP_OVER_VXLAN_PORT=<udp dport>
493 Verify configurations are set:
495 .. code-block:: console
497 mlxconfig -d <mst device> query | grep IP_OVER_VXLAN
498 IP_OVER_VXLAN_EN True(1)
499 IP_OVER_VXLAN_PORT <udp dport>
504 This driver relies on external libraries and kernel drivers for resources
505 allocations and initialization. The following dependencies are not part of
506 DPDK and must be installed separately:
510 User space Verbs framework used by librte_pmd_mlx5. This library provides
511 a generic interface between the kernel and low-level user space drivers
514 It allows slow and privileged operations (context initialization, hardware
515 resources allocations) to be managed by the kernel and fast operations to
516 never leave user space.
520 Low-level user space driver library for Mellanox
521 ConnectX-4/ConnectX-5/Bluefield devices, it is automatically loaded
524 This library basically implements send/receive calls to the hardware
529 Minimalistic Netlink library mainly relied on to manage E-Switch flow
530 rules (i.e. those with the "transfer" attribute and typically involving
535 They provide the kernel-side Verbs API and low level device drivers that
536 manage actual hardware initialization and resources sharing with user
539 Unlike most other PMDs, these modules must remain loaded and bound to
542 - mlx5_core: hardware driver managing Mellanox
543 ConnectX-4/ConnectX-5/Bluefield devices and related Ethernet kernel
545 - mlx5_ib: InifiniBand device driver.
546 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
548 - **Firmware update**
550 Mellanox OFED releases include firmware updates for
551 ConnectX-4/ConnectX-5/Bluefield adapters.
553 Because each release provides new features, these updates must be applied to
554 match the kernel modules and libraries they come with.
558 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
564 Either RDMA Core library with a recent enough Linux kernel release
565 (recommended) or Mellanox OFED, which provides compatibility with older
568 RMDA Core with Linux Kernel
569 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
571 - Minimal kernel version : v4.14 or the most recent 4.14-rc (see `Linux installation documentation`_)
572 - Minimal rdma-core version: v15+ commit 0c5f5765213a ("Merge pull request #227 from yishaih/tm")
573 (see `RDMA Core installation documentation`_)
574 - When building for i686 use:
576 - rdma-core version 18.0 or above built with 32bit support.
577 - Kernel version 4.14.41 or above.
579 .. _`Linux installation documentation`: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/plain/Documentation/admin-guide/README.rst
580 .. _`RDMA Core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
585 - Mellanox OFED version: **4.4, 4.5**.
588 - ConnectX-4: **12.21.1000** and above.
589 - ConnectX-4 Lx: **14.21.1000** and above.
590 - ConnectX-5: **16.21.1000** and above.
591 - ConnectX-5 Ex: **16.21.1000** and above.
592 - Bluefield: **18.99.3950** and above.
594 While these libraries and kernel modules are available on OpenFabrics
595 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
596 managers on most distributions, this PMD requires Ethernet extensions that
597 may not be supported at the moment (this is a work in progress).
600 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__
601 includes the necessary support and should be used in the meantime. For DPDK,
602 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
603 required from that distribution.
607 Several versions of Mellanox OFED are available. Installing the version
608 this DPDK release was developed and tested against is strongly
609 recommended. Please check the `prerequisites`_.
614 Minimal version for libmnl is **1.0.3**.
616 As a dependency of the **iproute2** suite, this library is often installed
617 by default. It is otherwise readily available through standard system
620 Its development headers must be installed in order to compile this PMD.
621 These packages are usually named **libmnl-dev** or **libmnl-devel**
622 depending on the Linux distribution.
627 * Mellanox(R) ConnectX(R)-4 10G MCX4111A-XCAT (1x10G)
628 * Mellanox(R) ConnectX(R)-4 10G MCX4121A-XCAT (2x10G)
629 * Mellanox(R) ConnectX(R)-4 25G MCX4111A-ACAT (1x25G)
630 * Mellanox(R) ConnectX(R)-4 25G MCX4121A-ACAT (2x25G)
631 * Mellanox(R) ConnectX(R)-4 40G MCX4131A-BCAT (1x40G)
632 * Mellanox(R) ConnectX(R)-4 40G MCX413A-BCAT (1x40G)
633 * Mellanox(R) ConnectX(R)-4 40G MCX415A-BCAT (1x40G)
634 * Mellanox(R) ConnectX(R)-4 50G MCX4131A-GCAT (1x50G)
635 * Mellanox(R) ConnectX(R)-4 50G MCX413A-GCAT (1x50G)
636 * Mellanox(R) ConnectX(R)-4 50G MCX414A-BCAT (2x50G)
637 * Mellanox(R) ConnectX(R)-4 50G MCX415A-GCAT (2x50G)
638 * Mellanox(R) ConnectX(R)-4 50G MCX416A-BCAT (2x50G)
639 * Mellanox(R) ConnectX(R)-4 50G MCX416A-GCAT (2x50G)
640 * Mellanox(R) ConnectX(R)-4 50G MCX415A-CCAT (1x100G)
641 * Mellanox(R) ConnectX(R)-4 100G MCX416A-CCAT (2x100G)
642 * Mellanox(R) ConnectX(R)-4 Lx 10G MCX4121A-XCAT (2x10G)
643 * Mellanox(R) ConnectX(R)-4 Lx 25G MCX4121A-ACAT (2x25G)
644 * Mellanox(R) ConnectX(R)-5 100G MCX556A-ECAT (2x100G)
645 * Mellanox(R) ConnectX(R)-5 Ex EN 100G MCX516A-CDAT (2x100G)
647 Quick Start Guide on OFED
648 -------------------------
650 1. Download latest Mellanox OFED. For more info check the `prerequisites`_.
653 2. Install the required libraries and kernel modules either by installing
654 only the required set, or by installing the entire Mellanox OFED:
656 .. code-block:: console
658 ./mlnxofedinstall --upstream-libs --dpdk
660 3. Verify the firmware is the correct one:
662 .. code-block:: console
666 4. Verify all ports links are set to Ethernet:
668 .. code-block:: console
670 mlxconfig -d <mst device> query | grep LINK_TYPE
674 Link types may have to be configured to Ethernet:
676 .. code-block:: console
678 mlxconfig -d <mst device> set LINK_TYPE_P1/2=1/2/3
680 * LINK_TYPE_P1=<1|2|3> , 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
682 For hypervisors verify SR-IOV is enabled on the NIC:
684 .. code-block:: console
686 mlxconfig -d <mst device> query | grep SRIOV_EN
689 If needed, set enable the set the relevant fields:
691 .. code-block:: console
693 mlxconfig -d <mst device> set SRIOV_EN=1 NUM_OF_VFS=16
694 mlxfwreset -d <mst device> reset
696 5. Restart the driver:
698 .. code-block:: console
700 /etc/init.d/openibd restart
704 .. code-block:: console
706 service openibd restart
708 If link type was changed, firmware must be reset as well:
710 .. code-block:: console
712 mlxfwreset -d <mst device> reset
714 For hypervisors, after reset write the sysfs number of virtual functions
717 To dynamically instantiate a given number of virtual functions (VFs):
719 .. code-block:: console
721 echo [num_vfs] > /sys/class/infiniband/mlx5_0/device/sriov_numvfs
723 6. Compile DPDK and you are ready to go. See instructions on
724 :ref:`Development Kit Build System <Development_Kit_Build_System>`
729 1. Configure aggressive CQE Zipping for maximum performance:
731 .. code-block:: console
733 mlxconfig -d <mst device> s CQE_COMPRESSION=1
735 To set it back to the default CQE Zipping mode use:
737 .. code-block:: console
739 mlxconfig -d <mst device> s CQE_COMPRESSION=0
741 2. In case of virtualization:
743 - Make sure that hypervisor kernel is 3.16 or newer.
744 - Configure boot with ``iommu=pt``.
746 - Make sure to allocate a VM on huge pages.
747 - Make sure to set CPU pinning.
749 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
750 for better performance. For VMs, verify that the right CPU
751 and NUMA node are pinned according to the above. Run:
753 .. code-block:: console
757 to identify the NUMA node to which the PCIe adapter is connected.
759 4. If more than one adapter is used, and root complex capabilities allow
760 to put both adapters on the same NUMA node without PCI bandwidth degradation,
761 it is recommended to locate both adapters on the same NUMA node.
762 This in order to forward packets from one to the other without
763 NUMA performance penalty.
765 5. Disable pause frames:
767 .. code-block:: console
769 ethtool -A <netdev> rx off tx off
771 6. Verify IO non-posted prefetch is disabled by default. This can be checked
772 via the BIOS configuration. Please contact you server provider for more
773 information about the settings.
777 On some machines, depends on the machine integrator, it is beneficial
778 to set the PCI max read request parameter to 1K. This can be
779 done in the following way:
781 To query the read request size use:
783 .. code-block:: console
785 setpci -s <NIC PCI address> 68.w
787 If the output is different than 3XXX, set it by:
789 .. code-block:: console
791 setpci -s <NIC PCI address> 68.w=3XXX
793 The XXX can be different on different systems. Make sure to configure
794 according to the setpci output.
796 7. To minimize overhead of searching Memory Regions:
798 - '--socket-mem' is recommended to pin memory by predictable amount.
799 - Configure per-lcore cache when creating Mempools for packet buffer.
800 - Refrain from dynamically allocating/freeing memory in run-time.
805 Compared to librte_pmd_mlx4 that implements a single RSS configuration per
806 port, librte_pmd_mlx5 supports per-protocol RSS configuration.
808 Since ``testpmd`` defaults to IP RSS mode and there is currently no
809 command-line parameter to enable additional protocols (UDP and TCP as well
810 as IP), the following commands must be entered from its CLI to get the same
811 behavior as librte_pmd_mlx4:
813 .. code-block:: console
816 > port config all rss all
822 This section demonstrates how to launch **testpmd** with Mellanox
823 ConnectX-4/ConnectX-5/Bluefield devices managed by librte_pmd_mlx5.
825 #. Load the kernel modules:
827 .. code-block:: console
829 modprobe -a ib_uverbs mlx5_core mlx5_ib
831 Alternatively if MLNX_OFED is fully installed, the following script can
834 .. code-block:: console
836 /etc/init.d/openibd restart
840 User space I/O kernel modules (uio and igb_uio) are not used and do
841 not have to be loaded.
843 #. Make sure Ethernet interfaces are in working order and linked to kernel
844 verbs. Related sysfs entries should be present:
846 .. code-block:: console
848 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
852 .. code-block:: console
859 #. Optionally, retrieve their PCI bus addresses for whitelisting:
861 .. code-block:: console
864 for intf in eth2 eth3 eth4 eth5;
866 (cd "/sys/class/net/${intf}/device/" && pwd -P);
869 sed -n 's,.*/\(.*\),-w \1,p'
873 .. code-block:: console
880 #. Request huge pages:
882 .. code-block:: console
884 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
886 #. Start testpmd with basic parameters:
888 .. code-block:: console
890 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
894 .. code-block:: console
897 EAL: PCI device 0000:05:00.0 on NUMA socket 0
898 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
899 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
900 PMD: librte_pmd_mlx5: 1 port(s) detected
901 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
902 EAL: PCI device 0000:05:00.1 on NUMA socket 0
903 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
904 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
905 PMD: librte_pmd_mlx5: 1 port(s) detected
906 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
907 EAL: PCI device 0000:06:00.0 on NUMA socket 0
908 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
909 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
910 PMD: librte_pmd_mlx5: 1 port(s) detected
911 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
912 EAL: PCI device 0000:06:00.1 on NUMA socket 0
913 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
914 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
915 PMD: librte_pmd_mlx5: 1 port(s) detected
916 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
917 Interactive-mode selected
918 Configuring Port 0 (socket 0)
919 PMD: librte_pmd_mlx5: 0x8cba80: TX queues number update: 0 -> 2
920 PMD: librte_pmd_mlx5: 0x8cba80: RX queues number update: 0 -> 2
921 Port 0: E4:1D:2D:E7:0C:FE
922 Configuring Port 1 (socket 0)
923 PMD: librte_pmd_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
924 PMD: librte_pmd_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
925 Port 1: E4:1D:2D:E7:0C:FF
926 Configuring Port 2 (socket 0)
927 PMD: librte_pmd_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
928 PMD: librte_pmd_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
929 Port 2: E4:1D:2D:E7:0C:FA
930 Configuring Port 3 (socket 0)
931 PMD: librte_pmd_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
932 PMD: librte_pmd_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
933 Port 3: E4:1D:2D:E7:0C:FB
934 Checking link statuses...
935 Port 0 Link Up - speed 40000 Mbps - full-duplex
936 Port 1 Link Up - speed 40000 Mbps - full-duplex
937 Port 2 Link Up - speed 10000 Mbps - full-duplex
938 Port 3 Link Up - speed 10000 Mbps - full-duplex