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**, **Mellanox ConnectX-6** and **Mellanox BlueField** families
11 of 10/25/40/50/100/200 Gb/s adapters as well as their virtual functions (VF)
14 Information and documentation about these adapters can be found on the
15 `Mellanox website <http://www.mellanox.com>`__. Help is also provided by the
16 `Mellanox community <http://community.mellanox.com/welcome>`__.
18 There is also a `section dedicated to this poll mode driver
19 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`__.
23 Due to external dependencies, this driver is disabled by default. It must
24 be enabled manually by setting ``CONFIG_RTE_LIBRTE_MLX5_PMD=y`` and
27 Implementation details
28 ----------------------
30 Besides its dependency on libibverbs (that implies libmlx5 and associated
31 kernel support), librte_pmd_mlx5 relies heavily on system calls for control
32 operations such as querying/updating the MTU and flow control parameters.
34 For security reasons and robustness, this driver only deals with virtual
35 memory addresses. The way resources allocations are handled by the kernel
36 combined with hardware specifications that allow it to handle virtual memory
37 addresses directly ensure that DPDK applications cannot access random
38 physical memory (or memory that does not belong to the current process).
40 This capability allows the PMD to coexist with kernel network interfaces
41 which remain functional, although they stop receiving unicast packets as
42 long as they share the same MAC address.
43 This means legacy linux control tools (for example: ethtool, ifconfig and
44 more) can operate on the same network interfaces that owned by the DPDK
47 Enabling librte_pmd_mlx5 causes DPDK applications to be linked against
53 - Multi arch support: x86_64, POWER8, ARMv8, i686.
54 - Multiple TX and RX queues.
55 - Support for scattered TX and RX frames.
56 - IPv4, IPv6, TCPv4, TCPv6, UDPv4 and UDPv6 RSS on any number of queues.
57 - Several RSS hash keys, one for each flow type.
58 - Default RSS operation with no hash key specification.
59 - Configurable RETA table.
60 - Support for multiple MAC addresses.
64 - RX CRC stripping configuration.
65 - Promiscuous mode on PF and VF.
66 - Multicast promiscuous mode on PF and VF.
67 - Hardware checksum offloads.
68 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
70 - Flow API, including :ref:`flow_isolated_mode`.
72 - KVM and VMware ESX SR-IOV modes are supported.
73 - RSS hash result is supported.
74 - Hardware TSO for generic IP or UDP tunnel, including VXLAN and GRE.
75 - Hardware checksum Tx offload for generic IP or UDP tunnel, including VXLAN and GRE.
77 - Statistics query including Basic, Extended and per queue.
79 - Tunnel types: VXLAN, L3 VXLAN, VXLAN-GPE, GRE, MPLSoGRE, MPLSoUDP, IP-in-IP.
80 - Tunnel HW offloads: packet type, inner/outer RSS, IP and UDP checksum verification.
81 - NIC HW offloads: encapsulation (vxlan, gre, mplsoudp, mplsogre), NAT, routing, TTL
82 increment/decrement, count, drop, mark. For details please see :ref:`Supported hardware offloads using rte_flow API`.
83 - Flow insertion rate of more then million flows per second, when using Direct Rules.
84 - Support for multiple rte_flow groups.
90 - For secondary process:
92 - Forked secondary process not supported.
93 - External memory unregistered in EAL memseg list cannot be used for DMA
94 unless such memory has been registered by ``mlx5_mr_update_ext_mp()`` in
95 primary process and remapped to the same virtual address in secondary
96 process. If the external memory is registered by primary process but has
97 different virtual address in secondary process, unexpected error may happen.
99 - Flow pattern without any specific vlan will match for vlan packets as well:
101 When VLAN spec is not specified in the pattern, the matching rule will be created with VLAN as a wild card.
102 Meaning, the flow rule::
104 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
106 Will only match vlan packets with vid=3. and the flow rules::
108 flow create 0 ingress pattern eth / ipv4 / end ...
112 flow create 0 ingress pattern eth / vlan / ipv4 / end ...
114 Will match any ipv4 packet (VLAN included).
116 - A multi segment packet must have less than 6 segments in case the Tx burst function
117 is set to multi-packet send or Enhanced multi-packet send. Otherwise it must have
118 less than 50 segments.
120 - Flows with a VXLAN Network Identifier equal (or ends to be equal)
121 to 0 are not supported.
123 - VXLAN TSO and checksum offloads are not supported on VM.
125 - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with MPLSoGRE and MPLSoUDP.
127 - VF: flow rules created on VF devices can only match traffic targeted at the
128 configured MAC addresses (see ``rte_eth_dev_mac_addr_add()``).
132 MAC addresses not already present in the bridge table of the associated
133 kernel network device will be added and cleaned up by the PMD when closing
134 the device. In case of ungraceful program termination, some entries may
135 remain present and should be removed manually by other means.
137 - When Multi-Packet Rx queue is configured (``mprq_en``), a Rx packet can be
138 externally attached to a user-provided mbuf with having EXT_ATTACHED_MBUF in
139 ol_flags. As the mempool for the external buffer is managed by PMD, all the
140 Rx mbufs must be freed before the device is closed. Otherwise, the mempool of
141 the external buffers will be freed by PMD and the application which still
142 holds the external buffers may be corrupted.
144 - If Multi-Packet Rx queue is configured (``mprq_en``) and Rx CQE compression is
145 enabled (``rxq_cqe_comp_en``) at the same time, RSS hash result is not fully
146 supported. Some Rx packets may not have PKT_RX_RSS_HASH.
148 - IPv6 Multicast messages are not supported on VM, while promiscuous mode
149 and allmulticast mode are both set to off.
150 To receive IPv6 Multicast messages on VM, explicitly set the relevant
151 MAC address using rte_eth_dev_mac_addr_add() API.
153 - E-Switch decapsulation Flow:
155 - can be applied to PF port only.
156 - must specify VF port action (packet redirection from PF to VF).
157 - optionally may specify tunnel inner source and destination MAC addresses.
159 - E-Switch encapsulation Flow:
161 - can be applied to VF ports only.
162 - must specify PF port action (packet redirection from VF to PF).
164 - ICMP/ICMP6 code/type matching cannot be supported togeter with IP-in-IP tunnel.
168 - ``scatter_fcs`` is disabled when LRO is configured.
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_IBVERBS_LINK_DLOPEN`` (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_IBVERBS_LINK_STATIC`` (default **n**)
214 Embed static flavor of the dependencies **libibverbs** and **libmlx5**
215 in the PMD shared library or the executable static binary.
217 - ``CONFIG_RTE_LIBRTE_MLX5_DEBUG`` (default **n**)
219 Toggle debugging code and stricter compilation flags. Enabling this option
220 adds additional run-time checks and debugging messages at the cost of
225 For BlueField, target should be set to ``arm64-bluefield-linux-gcc``. This
226 will enable ``CONFIG_RTE_LIBRTE_MLX5_PMD`` and set ``RTE_CACHE_LINE_SIZE`` to
227 64. Default armv8a configuration of make build and meson build set it to 128
228 then brings performance degradation.
230 Environment variables
231 ~~~~~~~~~~~~~~~~~~~~~
235 A list of directories in which to search for the rdma-core "glue" plug-in,
236 separated by colons or semi-colons.
238 Only matters when compiled with ``CONFIG_RTE_IBVERBS_LINK_DLOPEN``
239 enabled and most useful when ``CONFIG_RTE_EAL_PMD_PATH`` is also set,
240 since ``LD_LIBRARY_PATH`` has no effect in this case.
242 - ``MLX5_SHUT_UP_BF``
244 Configures HW Tx doorbell register as IO-mapped.
246 By default, the HW Tx doorbell is configured as a write-combining register.
247 The register would be flushed to HW usually when the write-combining buffer
248 becomes full, but it depends on CPU design.
250 Except for vectorized Tx burst routines, a write memory barrier is enforced
251 after updating the register so that the update can be immediately visible to
254 When vectorized Tx burst is called, the barrier is set only if the burst size
255 is not aligned to MLX5_VPMD_TX_MAX_BURST. However, setting this environmental
256 variable will bring better latency even though the maximum throughput can
259 Run-time configuration
260 ~~~~~~~~~~~~~~~~~~~~~~
262 - librte_pmd_mlx5 brings kernel network interfaces up during initialization
263 because it is affected by their state. Forcing them down prevents packets
266 - **ethtool** operations on related kernel interfaces also affect the PMD.
268 - ``rxq_cqe_comp_en`` parameter [int]
270 A nonzero value enables the compression of CQE on RX side. This feature
271 allows to save PCI bandwidth and improve performance. Enabled by default.
275 - x86_64 with ConnectX-4, ConnectX-4 LX, ConnectX-5, ConnectX-6 and BlueField.
276 - POWER9 and ARMv8 with ConnectX-4 LX, ConnectX-5, ConnectX-6 and BlueField.
278 - ``rxq_cqe_pad_en`` parameter [int]
280 A nonzero value enables 128B padding of CQE on RX side. The size of CQE
281 is aligned with the size of a cacheline of the core. If cacheline size is
282 128B, the CQE size is configured to be 128B even though the device writes
283 only 64B data on the cacheline. This is to avoid unnecessary cache
284 invalidation by device's two consecutive writes on to one cacheline.
285 However in some architecture, it is more beneficial to update entire
286 cacheline with padding the rest 64B rather than striding because
287 read-modify-write could drop performance a lot. On the other hand,
288 writing extra data will consume more PCIe bandwidth and could also drop
289 the maximum throughput. It is recommended to empirically set this
290 parameter. Disabled by default.
294 - CPU having 128B cacheline with ConnectX-5 and BlueField.
296 - ``rxq_pkt_pad_en`` parameter [int]
298 A nonzero value enables padding Rx packet to the size of cacheline on PCI
299 transaction. This feature would waste PCI bandwidth but could improve
300 performance by avoiding partial cacheline write which may cause costly
301 read-modify-copy in memory transaction on some architectures. Disabled by
306 - x86_64 with ConnectX-4, ConnectX-4 LX, ConnectX-5, ConnectX-6 and BlueField.
307 - POWER8 and ARMv8 with ConnectX-4 LX, ConnectX-5, ConnectX-6 and BlueField.
309 - ``mprq_en`` parameter [int]
311 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
312 configured as Multi-Packet RQ if the total number of Rx queues is
313 ``rxqs_min_mprq`` or more and Rx scatter isn't configured. Disabled by
316 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
317 by posting a single large buffer for multiple packets. Instead of posting a
318 buffers per a packet, one large buffer is posted in order to receive multiple
319 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
320 and each stride receives one packet. MPRQ can improve throughput for
321 small-packet traffic.
323 When MPRQ is enabled, max_rx_pkt_len can be larger than the size of
324 user-provided mbuf even if DEV_RX_OFFLOAD_SCATTER isn't enabled. PMD will
325 configure large stride size enough to accommodate max_rx_pkt_len as long as
326 device allows. Note that this can waste system memory compared to enabling Rx
327 scatter and multi-segment packet.
329 - ``mprq_log_stride_num`` parameter [int]
331 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
332 strides can reduce PCIe traffic further. If configured value is not in the
333 range of device capability, the default value will be set with a warning
334 message. The default value is 4 which is 16 strides per a buffer, valid only
335 if ``mprq_en`` is set.
337 The size of Rx queue should be bigger than the number of strides.
339 - ``mprq_max_memcpy_len`` parameter [int]
341 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
342 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
343 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
344 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
345 A mempool for external buffers will be allocated and managed by PMD. If Rx
346 packet is externally attached, ol_flags field of the mbuf will have
347 EXT_ATTACHED_MBUF and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
348 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
350 - ``rxqs_min_mprq`` parameter [int]
352 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
353 greater or equal to this value. The default value is 12, valid only if
356 - ``txq_inline`` parameter [int]
358 Amount of data to be inlined during TX operations. This parameter is
359 deprecated and converted to the new parameter ``txq_inline_max`` providing
360 partial compatibility.
362 - ``txqs_min_inline`` parameter [int]
364 Enable inline data send only when the number of TX queues is greater or equal
367 This option should be used in combination with ``txq_inline_max`` and
368 ``txq_inline_mpw`` below and does not affect ``txq_inline_min`` settings above.
370 If this option is not specified the default value 16 is used for BlueField
371 and 8 for other platforms
373 The data inlining consumes the CPU cycles, so this option is intended to
374 auto enable inline data if we have enough Tx queues, which means we have
375 enough CPU cores and PCI bandwidth is getting more critical and CPU
376 is not supposed to be bottleneck anymore.
378 The copying data into WQE improves latency and can improve PPS performance
379 when PCI back pressure is detected and may be useful for scenarios involving
380 heavy traffic on many queues.
382 Because additional software logic is necessary to handle this mode, this
383 option should be used with care, as it may lower performance when back
384 pressure is not expected.
386 - ``txq_inline_min`` parameter [int]
388 Minimal amount of data to be inlined into WQE during Tx operations. NICs
389 may require this minimal data amount to operate correctly. The exact value
390 may depend on NIC operation mode, requested offloads, etc.
392 If ``txq_inline_min`` key is present the specified value (may be aligned
393 by the driver in order not to exceed the limits and provide better descriptor
394 space utilization) will be used by the driver and it is guaranteed the
395 requested data bytes are inlined into the WQE beside other inline settings.
396 This keys also may update ``txq_inline_max`` value (default of specified
397 explicitly in devargs) to reserve the space for inline data.
399 If ``txq_inline_min`` key is not present, the value may be queried by the
400 driver from the NIC via DevX if this feature is available. If there is no DevX
401 enabled/supported the value 18 (supposing L2 header including VLAN) is set
402 for ConnectX-4, value 58 (supposing L2-L4 headers, required by configurations
403 over E-Switch) is set for ConnectX-4 Lx, and 0 is set by default for ConnectX-5
404 and newer NICs. If packet is shorter the ``txq_inline_min`` value, the entire
407 For the ConnectX-4 and ConnectX-4 Lx NICs driver does not allow to set
408 this value below 18 (minimal L2 header, including VLAN).
410 Please, note, this minimal data inlining disengages eMPW feature (Enhanced
411 Multi-Packet Write), because last one does not support partial packet inlining.
412 This is not very critical due to minimal data inlining is mostly required
413 by ConnectX-4 and ConnectX-4 Lx, these NICs do not support eMPW feature.
415 - ``txq_inline_max`` parameter [int]
417 Specifies the maximal packet length to be completely inlined into WQE
418 Ethernet Segment for ordinary SEND method. If packet is larger than specified
419 value, the packet data won't be copied by the driver at all, data buffer
420 is addressed with a pointer. If packet length is less or equal all packet
421 data will be copied into WQE. This may improve PCI bandwidth utilization for
422 short packets significantly but requires the extra CPU cycles.
424 The data inline feature is controlled by number of Tx queues, if number of Tx
425 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
426 is engaged, if there are not enough Tx queues (which means not enough CPU cores
427 and CPU resources are scarce), data inline is not performed by the driver.
428 Assigning ``txqs_min_inline`` with zero always enables the data inline.
430 The default ``txq_inline_max`` value is 290. The specified value may be adjusted
431 by the driver in order not to exceed the limit (930 bytes) and to provide better
432 WQE space filling without gaps, the adjustment is reflected in the debug log.
434 - ``txq_inline_mpw`` parameter [int]
436 Specifies the maximal packet length to be completely inlined into WQE for
437 Enhanced MPW method. If packet is large the specified value, the packet data
438 won't be copied, and data buffer is addressed with pointer. If packet length
439 is less or equal, all packet data will be copied into WQE. This may improve PCI
440 bandwidth utilization for short packets significantly but requires the extra
443 The data inline feature is controlled by number of TX queues, if number of Tx
444 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
445 is engaged, if there are not enough Tx queues (which means not enough CPU cores
446 and CPU resources are scarce), data inline is not performed by the driver.
447 Assigning ``txqs_min_inline`` with zero always enables the data inline.
449 The default ``txq_inline_mpw`` value is 188. The specified value may be adjusted
450 by the driver in order not to exceed the limit (930 bytes) and to provide better
451 WQE space filling without gaps, the adjustment is reflected in the debug log.
452 Due to multiple packets may be included to the same WQE with Enhanced Multi
453 Packet Write Method and overall WQE size is limited it is not recommended to
454 specify large values for the ``txq_inline_mpw``.
456 - ``txqs_max_vec`` parameter [int]
458 Enable vectorized Tx only when the number of TX queues is less than or
459 equal to this value. This parameter is deprecated and ignored, kept
460 for compatibility issue to not prevent driver from probing.
462 - ``txq_mpw_hdr_dseg_en`` parameter [int]
464 A nonzero value enables including two pointers in the first block of TX
465 descriptor. The parameter is deprecated and ignored, kept for compatibility
468 - ``txq_max_inline_len`` parameter [int]
470 Maximum size of packet to be inlined. This limits the size of packet to
471 be inlined. If the size of a packet is larger than configured value, the
472 packet isn't inlined even though there's enough space remained in the
473 descriptor. Instead, the packet is included with pointer. This parameter
474 is deprecated and converted directly to ``txq_inline_mpw`` providing full
475 compatibility. Valid only if eMPW feature is engaged.
477 - ``txq_mpw_en`` parameter [int]
479 A nonzero value enables Enhanced Multi-Packet Write (eMPW) for ConnectX-5,
480 ConnectX-6 and BlueField. eMPW allows the TX burst function to pack up multiple
481 packets in a single descriptor session in order to save PCI bandwidth and improve
482 performance at the cost of a slightly higher CPU usage. When ``txq_inline_mpw``
483 is set along with ``txq_mpw_en``, TX burst function copies entire packet
484 data on to TX descriptor instead of including pointer of packet.
486 The Enhanced Multi-Packet Write feature is enabled by default if NIC supports
487 it, can be disabled by explicit specifying 0 value for ``txq_mpw_en`` option.
488 Also, if minimal data inlining is requested by non-zero ``txq_inline_min``
489 option or reported by the NIC, the eMPW feature is disengaged.
491 - ``tx_vec_en`` parameter [int]
493 A nonzero value enables Tx vector on ConnectX-5, ConnectX-6 and BlueField
494 NICs if the number of global Tx queues on the port is less than
495 ``txqs_max_vec``. The parameter is deprecated and ignored.
497 - ``rx_vec_en`` parameter [int]
499 A nonzero value enables Rx vector if the port is not configured in
500 multi-segment otherwise this parameter is ignored.
504 - ``vf_nl_en`` parameter [int]
506 A nonzero value enables Netlink requests from the VF to add/remove MAC
507 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
508 Otherwise the relevant configuration must be run with Linux iproute2 tools.
509 This is a prerequisite to receive this kind of traffic.
511 Enabled by default, valid only on VF devices ignored otherwise.
513 - ``l3_vxlan_en`` parameter [int]
515 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
516 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
517 parameter. This is a prerequisite to receive this kind of traffic.
521 - ``dv_flow_en`` parameter [int]
523 A nonzero value enables the DV flow steering assuming it is supported
528 - ``dv_esw_en`` parameter [int]
530 A nonzero value enables E-Switch using Direct Rules.
532 Enabled by default if supported.
534 - ``mr_ext_memseg_en`` parameter [int]
536 A nonzero value enables extending memseg when registering DMA memory. If
537 enabled, the number of entries in MR (Memory Region) lookup table on datapath
538 is minimized and it benefits performance. On the other hand, it worsens memory
539 utilization because registered memory is pinned by kernel driver. Even if a
540 page in the extended chunk is freed, that doesn't become reusable until the
541 entire memory is freed.
545 - ``representor`` parameter [list]
547 This parameter can be used to instantiate DPDK Ethernet devices from
548 existing port (or VF) representors configured on the device.
550 It is a standard parameter whose format is described in
551 :ref:`ethernet_device_standard_device_arguments`.
553 For instance, to probe port representors 0 through 2::
557 - ``max_dump_files_num`` parameter [int]
559 The maximum number of files per PMD entity that may be created for debug information.
560 The files will be created in /var/log directory or in current directory.
562 set to 128 by default.
564 - ``lro_timeout_usec`` parameter [int]
566 The maximum allowed duration of an LRO session, in micro-seconds.
567 PMD will set the nearest value supported by HW, which is not bigger than
568 the input ``lro_timeout_usec`` value.
569 If this parameter is not specified, by default PMD will set
570 the smallest value supported by HW.
572 Firmware configuration
573 ~~~~~~~~~~~~~~~~~~~~~~
575 - L3 VXLAN and VXLAN-GPE destination UDP port
577 .. code-block:: console
579 mlxconfig -d <mst device> set IP_OVER_VXLAN_EN=1
580 mlxconfig -d <mst device> set IP_OVER_VXLAN_PORT=<udp dport>
582 Verify configurations are set:
584 .. code-block:: console
586 mlxconfig -d <mst device> query | grep IP_OVER_VXLAN
587 IP_OVER_VXLAN_EN True(1)
588 IP_OVER_VXLAN_PORT <udp dport>
590 - enable ICMP/ICMP6's code/type field matching
592 .. code-block:: console
594 mlxconfig -d <mst device> set FLEX_PARSER_PROFILE_ENABLE=2
596 Verify configurations are set:
598 .. code-block:: console
600 mlxconfig -d <mst device> query | grep FLEX_PARSER_PROFILE_ENABLE
601 FLEX_PARSER_PROFILE_ENABLE 2
603 - IP-in-IP tunnel enable
605 .. code-block:: console
607 mlxconfig -d <mst device> set FLEX_PARSER_PROFILE_ENABLE=0
609 Verify configurations are set:
611 .. code-block:: console
613 mlxconfig -d <mst device> query | grep FLEX_PARSER_PROFILE_ENABLE
614 FLEX_PARSER_PROFILE_ENABLE 0
619 This driver relies on external libraries and kernel drivers for resources
620 allocations and initialization. The following dependencies are not part of
621 DPDK and must be installed separately:
625 User space Verbs framework used by librte_pmd_mlx5. This library provides
626 a generic interface between the kernel and low-level user space drivers
629 It allows slow and privileged operations (context initialization, hardware
630 resources allocations) to be managed by the kernel and fast operations to
631 never leave user space.
635 Low-level user space driver library for Mellanox
636 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices, it is automatically loaded
639 This library basically implements send/receive calls to the hardware
644 They provide the kernel-side Verbs API and low level device drivers that
645 manage actual hardware initialization and resources sharing with user
648 Unlike most other PMDs, these modules must remain loaded and bound to
651 - mlx5_core: hardware driver managing Mellanox
652 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices and related Ethernet kernel
654 - mlx5_ib: InifiniBand device driver.
655 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
657 - **Firmware update**
659 Mellanox OFED/EN releases include firmware updates for
660 ConnectX-4/ConnectX-5/ConnectX-6/BlueField adapters.
662 Because each release provides new features, these updates must be applied to
663 match the kernel modules and libraries they come with.
667 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
673 Either RDMA Core library with a recent enough Linux kernel release
674 (recommended) or Mellanox OFED/EN, which provides compatibility with older
677 RDMA Core with Linux Kernel
678 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
680 - Minimal kernel version : v4.14 or the most recent 4.14-rc (see `Linux installation documentation`_)
681 - Minimal rdma-core version: v15+ commit 0c5f5765213a ("Merge pull request #227 from yishaih/tm")
682 (see `RDMA Core installation documentation`_)
683 - When building for i686 use:
685 - rdma-core version 18.0 or above built with 32bit support.
686 - Kernel version 4.14.41 or above.
688 - Starting with rdma-core v21, static libraries can be built::
691 CFLAGS=-fPIC cmake -DIN_PLACE=1 -DENABLE_STATIC=1 -GNinja ..
694 .. _`Linux installation documentation`: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/plain/Documentation/admin-guide/README.rst
695 .. _`RDMA Core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
697 If rdma-core libraries are built but not installed, DPDK makefile can link them,
698 thanks to these environment variables:
700 - ``EXTRA_CFLAGS=-I/path/to/rdma-core/build/include``
701 - ``EXTRA_LDFLAGS=-L/path/to/rdma-core/build/lib``
702 - ``PKG_CONFIG_PATH=/path/to/rdma-core/build/lib/pkgconfig``
707 - Mellanox OFED version: ** 4.5, 4.6** /
708 Mellanox EN version: **4.5, 4.6**
711 - ConnectX-4: **12.21.1000** and above.
712 - ConnectX-4 Lx: **14.21.1000** and above.
713 - ConnectX-5: **16.21.1000** and above.
714 - ConnectX-5 Ex: **16.21.1000** and above.
715 - ConnectX-6: **20.99.5374** and above.
716 - BlueField: **18.25.1010** and above.
718 While these libraries and kernel modules are available on OpenFabrics
719 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
720 managers on most distributions, this PMD requires Ethernet extensions that
721 may not be supported at the moment (this is a work in progress).
724 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__ and
726 <http://www.mellanox.com/page/products_dyn?product_family=27&mtag=linux>`__
727 include the necessary support and should be used in the meantime. For DPDK,
728 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
729 required from that distribution.
733 Several versions of Mellanox OFED/EN are available. Installing the version
734 this DPDK release was developed and tested against is strongly
735 recommended. Please check the `prerequisites`_.
740 * Mellanox(R) ConnectX(R)-4 10G MCX4111A-XCAT (1x10G)
741 * Mellanox(R) ConnectX(R)-4 10G MCX4121A-XCAT (2x10G)
742 * Mellanox(R) ConnectX(R)-4 25G MCX4111A-ACAT (1x25G)
743 * Mellanox(R) ConnectX(R)-4 25G MCX4121A-ACAT (2x25G)
744 * Mellanox(R) ConnectX(R)-4 40G MCX4131A-BCAT (1x40G)
745 * Mellanox(R) ConnectX(R)-4 40G MCX413A-BCAT (1x40G)
746 * Mellanox(R) ConnectX(R)-4 40G MCX415A-BCAT (1x40G)
747 * Mellanox(R) ConnectX(R)-4 50G MCX4131A-GCAT (1x50G)
748 * Mellanox(R) ConnectX(R)-4 50G MCX413A-GCAT (1x50G)
749 * Mellanox(R) ConnectX(R)-4 50G MCX414A-BCAT (2x50G)
750 * Mellanox(R) ConnectX(R)-4 50G MCX415A-GCAT (2x50G)
751 * Mellanox(R) ConnectX(R)-4 50G MCX416A-BCAT (2x50G)
752 * Mellanox(R) ConnectX(R)-4 50G MCX416A-GCAT (2x50G)
753 * Mellanox(R) ConnectX(R)-4 50G MCX415A-CCAT (1x100G)
754 * Mellanox(R) ConnectX(R)-4 100G MCX416A-CCAT (2x100G)
755 * Mellanox(R) ConnectX(R)-4 Lx 10G MCX4121A-XCAT (2x10G)
756 * Mellanox(R) ConnectX(R)-4 Lx 25G MCX4121A-ACAT (2x25G)
757 * Mellanox(R) ConnectX(R)-5 100G MCX556A-ECAT (2x100G)
758 * Mellanox(R) ConnectX(R)-5 Ex EN 100G MCX516A-CDAT (2x100G)
760 Quick Start Guide on OFED/EN
761 ----------------------------
763 1. Download latest Mellanox OFED/EN. For more info check the `prerequisites`_.
766 2. Install the required libraries and kernel modules either by installing
767 only the required set, or by installing the entire Mellanox OFED/EN:
769 .. code-block:: console
771 ./mlnxofedinstall --upstream-libs --dpdk
773 3. Verify the firmware is the correct one:
775 .. code-block:: console
779 4. Verify all ports links are set to Ethernet:
781 .. code-block:: console
783 mlxconfig -d <mst device> query | grep LINK_TYPE
787 Link types may have to be configured to Ethernet:
789 .. code-block:: console
791 mlxconfig -d <mst device> set LINK_TYPE_P1/2=1/2/3
793 * LINK_TYPE_P1=<1|2|3> , 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
795 For hypervisors verify SR-IOV is enabled on the NIC:
797 .. code-block:: console
799 mlxconfig -d <mst device> query | grep SRIOV_EN
802 If needed, set enable the set the relevant fields:
804 .. code-block:: console
806 mlxconfig -d <mst device> set SRIOV_EN=1 NUM_OF_VFS=16
807 mlxfwreset -d <mst device> reset
809 5. Restart the driver:
811 .. code-block:: console
813 /etc/init.d/openibd restart
817 .. code-block:: console
819 service openibd restart
821 If link type was changed, firmware must be reset as well:
823 .. code-block:: console
825 mlxfwreset -d <mst device> reset
827 For hypervisors, after reset write the sysfs number of virtual functions
830 To dynamically instantiate a given number of virtual functions (VFs):
832 .. code-block:: console
834 echo [num_vfs] > /sys/class/infiniband/mlx5_0/device/sriov_numvfs
836 6. Compile DPDK and you are ready to go. See instructions on
837 :ref:`Development Kit Build System <Development_Kit_Build_System>`
839 Enable switchdev mode
840 ---------------------
842 Switchdev mode is a mode in E-Switch, that binds between representor and VF.
843 Representor is a port in DPDK that is connected to a VF in such a way
844 that assuming there are no offload flows, each packet that is sent from the VF
845 will be received by the corresponding representor. While each packet that is
846 sent to a representor will be received by the VF.
847 This is very useful in case of SRIOV mode, where the first packet that is sent
848 by the VF will be received by the DPDK application which will decide if this
849 flow should be offloaded to the E-Switch. After offloading the flow packet
850 that the VF that are matching the flow will not be received any more by
851 the DPDK application.
853 1. Enable SRIOV mode:
855 .. code-block:: console
857 mlxconfig -d <mst device> set SRIOV_EN=true
859 2. Configure the max number of VFs:
861 .. code-block:: console
863 mlxconfig -d <mst device> set NUM_OF_VFS=<num of vfs>
867 .. code-block:: console
869 mlxfwreset -d <mst device> reset
871 3. Configure the actual number of VFs:
873 .. code-block:: console
875 echo <num of vfs > /sys/class/net/<net device>/device/sriov_numvfs
877 4. Unbind the device (can be rebind after the switchdev mode):
879 .. code-block:: console
881 echo -n "<device pci address" > /sys/bus/pci/drivers/mlx5_core/unbind
883 5. Enbale switchdev mode:
885 .. code-block:: console
887 echo switchdev > /sys/class/net/<net device>/compat/devlink/mode
892 1. Configure aggressive CQE Zipping for maximum performance:
894 .. code-block:: console
896 mlxconfig -d <mst device> s CQE_COMPRESSION=1
898 To set it back to the default CQE Zipping mode use:
900 .. code-block:: console
902 mlxconfig -d <mst device> s CQE_COMPRESSION=0
904 2. In case of virtualization:
906 - Make sure that hypervisor kernel is 3.16 or newer.
907 - Configure boot with ``iommu=pt``.
909 - Make sure to allocate a VM on huge pages.
910 - Make sure to set CPU pinning.
912 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
913 for better performance. For VMs, verify that the right CPU
914 and NUMA node are pinned according to the above. Run:
916 .. code-block:: console
920 to identify the NUMA node to which the PCIe adapter is connected.
922 4. If more than one adapter is used, and root complex capabilities allow
923 to put both adapters on the same NUMA node without PCI bandwidth degradation,
924 it is recommended to locate both adapters on the same NUMA node.
925 This in order to forward packets from one to the other without
926 NUMA performance penalty.
928 5. Disable pause frames:
930 .. code-block:: console
932 ethtool -A <netdev> rx off tx off
934 6. Verify IO non-posted prefetch is disabled by default. This can be checked
935 via the BIOS configuration. Please contact you server provider for more
936 information about the settings.
940 On some machines, depends on the machine integrator, it is beneficial
941 to set the PCI max read request parameter to 1K. This can be
942 done in the following way:
944 To query the read request size use:
946 .. code-block:: console
948 setpci -s <NIC PCI address> 68.w
950 If the output is different than 3XXX, set it by:
952 .. code-block:: console
954 setpci -s <NIC PCI address> 68.w=3XXX
956 The XXX can be different on different systems. Make sure to configure
957 according to the setpci output.
959 7. To minimize overhead of searching Memory Regions:
961 - '--socket-mem' is recommended to pin memory by predictable amount.
962 - Configure per-lcore cache when creating Mempools for packet buffer.
963 - Refrain from dynamically allocating/freeing memory in run-time.
965 Supported hardware offloads using rte_flow API
966 ----------------------------------------------
968 .. _Supported hardware offloads using rte_flow API:
970 .. table:: Supported hardware offloads using rte_flow API
972 +-----------------------+-----------------+-----------------+
973 | Offload | E-Switch | NIC |
975 +=======================+=================+=================+
976 | Count | | DPDK 19.05 | | DPDK 19.02 |
977 | | | OFED 4.6 | | OFED 4.6 |
978 | | | RDMA-CORE V24 | | RDMA-CORE V23 |
979 | | | ConnectX-5 | | ConnectX-5 |
980 +-----------------------+-----------------+-----------------+
981 | Drop / Queue / RSS | | DPDK 19.05 | | DPDK 18.11 |
982 | | | OFED 4.6 | | OFED 4.5 |
983 | | | RDMA-CORE V24 | | RDMA-CORE V23 |
984 | | | ConnectX-5 | | ConnectX-4 |
985 +-----------------------+-----------------+-----------------+
986 | Encapsulation | | DPDK 19.05 | | DPDK 19.02 |
987 | (VXLAN / NVGRE / RAW) | | OFED 4.6.2 | | OFED 4.6 |
988 | | | RDMA-CORE V24 | | RDMA-CORE V23 |
989 | | | ConnectX-5 | | ConnectX-5 |
990 +-----------------------+-----------------+-----------------+
991 | Header rewrite | | DPDK 19.05 | | DPDK 19.02 |
992 | (set_ipv4_src / | | OFED 4.6.2 | | OFED 4.6.2 |
993 | set_ipv4_dst / | | RDMA-CORE V24 | | RDMA-CORE V23 |
994 | set_ipv6_src / | | ConnectX-5 | | ConnectX-5 |
995 | set_ipv6_dst / | | |
1000 | set_mac_src / | | |
1001 | set_mac_dst) | | |
1002 +-----------------------+-----------------+-----------------+
1003 | Jump | | DPDK 19.05 | | DPDK 19.02 |
1004 | | | OFED 4.6.2 | | OFED 4.6.2 |
1005 | | | RDMA-CORE V24 | | N/A |
1006 | | | ConnectX-5 | | ConnectX-5 |
1007 +-----------------------+-----------------+-----------------+
1008 | Mark / Flag | | DPDK 19.05 | | DPDK 18.11 |
1009 | | | OFED 4.6 | | OFED 4.5 |
1010 | | | RDMA-CORE V24 | | RDMA-CORE V23 |
1011 | | | ConnectX-5 | | ConnectX-4 |
1012 +-----------------------+-----------------+-----------------+
1013 | Port ID | | DPDK 19.05 | | N/A |
1014 | | | OFED 4.6 | | N/A |
1015 | | | RDMA-CORE V24 | | N/A |
1016 | | | ConnectX-5 | | N/A |
1017 +-----------------------+-----------------+-----------------+
1019 * Minimum version for each component and nic.
1024 Compared to librte_pmd_mlx4 that implements a single RSS configuration per
1025 port, librte_pmd_mlx5 supports per-protocol RSS configuration.
1027 Since ``testpmd`` defaults to IP RSS mode and there is currently no
1028 command-line parameter to enable additional protocols (UDP and TCP as well
1029 as IP), the following commands must be entered from its CLI to get the same
1030 behavior as librte_pmd_mlx4:
1032 .. code-block:: console
1035 > port config all rss all
1041 This section demonstrates how to launch **testpmd** with Mellanox
1042 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices managed by librte_pmd_mlx5.
1044 #. Load the kernel modules:
1046 .. code-block:: console
1048 modprobe -a ib_uverbs mlx5_core mlx5_ib
1050 Alternatively if MLNX_OFED/MLNX_EN is fully installed, the following script
1053 .. code-block:: console
1055 /etc/init.d/openibd restart
1059 User space I/O kernel modules (uio and igb_uio) are not used and do
1060 not have to be loaded.
1062 #. Make sure Ethernet interfaces are in working order and linked to kernel
1063 verbs. Related sysfs entries should be present:
1065 .. code-block:: console
1067 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
1071 .. code-block:: console
1078 #. Optionally, retrieve their PCI bus addresses for whitelisting:
1080 .. code-block:: console
1083 for intf in eth2 eth3 eth4 eth5;
1085 (cd "/sys/class/net/${intf}/device/" && pwd -P);
1088 sed -n 's,.*/\(.*\),-w \1,p'
1092 .. code-block:: console
1099 #. Request huge pages:
1101 .. code-block:: console
1103 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
1105 #. Start testpmd with basic parameters:
1107 .. code-block:: console
1109 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
1113 .. code-block:: console
1116 EAL: PCI device 0000:05:00.0 on NUMA socket 0
1117 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
1118 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
1119 PMD: librte_pmd_mlx5: 1 port(s) detected
1120 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
1121 EAL: PCI device 0000:05:00.1 on NUMA socket 0
1122 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
1123 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
1124 PMD: librte_pmd_mlx5: 1 port(s) detected
1125 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
1126 EAL: PCI device 0000:06:00.0 on NUMA socket 0
1127 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
1128 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
1129 PMD: librte_pmd_mlx5: 1 port(s) detected
1130 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
1131 EAL: PCI device 0000:06:00.1 on NUMA socket 0
1132 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
1133 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
1134 PMD: librte_pmd_mlx5: 1 port(s) detected
1135 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
1136 Interactive-mode selected
1137 Configuring Port 0 (socket 0)
1138 PMD: librte_pmd_mlx5: 0x8cba80: TX queues number update: 0 -> 2
1139 PMD: librte_pmd_mlx5: 0x8cba80: RX queues number update: 0 -> 2
1140 Port 0: E4:1D:2D:E7:0C:FE
1141 Configuring Port 1 (socket 0)
1142 PMD: librte_pmd_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
1143 PMD: librte_pmd_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
1144 Port 1: E4:1D:2D:E7:0C:FF
1145 Configuring Port 2 (socket 0)
1146 PMD: librte_pmd_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
1147 PMD: librte_pmd_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
1148 Port 2: E4:1D:2D:E7:0C:FA
1149 Configuring Port 3 (socket 0)
1150 PMD: librte_pmd_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
1151 PMD: librte_pmd_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
1152 Port 3: E4:1D:2D:E7:0C:FB
1153 Checking link statuses...
1154 Port 0 Link Up - speed 40000 Mbps - full-duplex
1155 Port 1 Link Up - speed 40000 Mbps - full-duplex
1156 Port 2 Link Up - speed 10000 Mbps - full-duplex
1157 Port 3 Link Up - speed 10000 Mbps - full-duplex