1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2015 6WIND S.A.
3 Copyright 2015 Mellanox Technologies, Ltd
5 .. include:: <isonum.txt>
10 The MLX5 poll mode driver library (**librte_net_mlx5**) provides support
11 for **Mellanox ConnectX-4**, **Mellanox ConnectX-4 Lx** , **Mellanox
12 ConnectX-5**, **Mellanox ConnectX-6**, **Mellanox ConnectX-6 Dx** and
13 **Mellanox BlueField** families of 10/25/40/50/100/200 Gb/s adapters
14 as well as their virtual functions (VF) in SR-IOV context.
16 Information and documentation about these adapters can be found on the
17 `Mellanox website <http://www.mellanox.com>`__. Help is also provided by the
18 `Mellanox community <http://community.mellanox.com/welcome>`__.
20 There is also a `section dedicated to this poll mode driver
21 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`__.
27 Besides its dependency on libibverbs (that implies libmlx5 and associated
28 kernel support), librte_net_mlx5 relies heavily on system calls for control
29 operations such as querying/updating the MTU and flow control parameters.
31 For security reasons and robustness, this driver only deals with virtual
32 memory addresses. The way resources allocations are handled by the kernel,
33 combined with hardware specifications that allow to handle virtual memory
34 addresses directly, ensure that DPDK applications cannot access random
35 physical memory (or memory that does not belong to the current process).
37 This capability allows the PMD to coexist with kernel network interfaces
38 which remain functional, although they stop receiving unicast packets as
39 long as they share the same MAC address.
40 This means legacy linux control tools (for example: ethtool, ifconfig and
41 more) can operate on the same network interfaces that owned by the DPDK
44 The PMD can use libibverbs and libmlx5 to access the device firmware
45 or directly the hardware components.
46 There are different levels of objects and bypassing abilities
47 to get the best performances:
49 - Verbs is a complete high-level generic API
50 - Direct Verbs is a device-specific API
51 - DevX allows to access firmware objects
52 - Direct Rules manages flow steering at low-level hardware layer
54 Enabling librte_net_mlx5 causes DPDK applications to be linked against
60 - Multi arch support: x86_64, POWER8, ARMv8, i686.
61 - Multiple TX and RX queues.
62 - Support for scattered TX and RX frames.
63 - IPv4, IPv6, TCPv4, TCPv6, UDPv4 and UDPv6 RSS on any number of queues.
64 - RSS using different combinations of fields: L3 only, L4 only or both,
65 and source only, destination only or both.
66 - Several RSS hash keys, one for each flow type.
67 - Default RSS operation with no hash key specification.
68 - Configurable RETA table.
69 - Link flow control (pause frame).
70 - Support for multiple MAC addresses.
74 - RX CRC stripping configuration.
75 - Promiscuous mode on PF and VF.
76 - Multicast promiscuous mode on PF and VF.
77 - Hardware checksum offloads.
78 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
80 - Flow API, including :ref:`flow_isolated_mode`.
82 - KVM and VMware ESX SR-IOV modes are supported.
83 - RSS hash result is supported.
84 - Hardware TSO for generic IP or UDP tunnel, including VXLAN and GRE.
85 - Hardware checksum Tx offload for generic IP or UDP tunnel, including VXLAN and GRE.
87 - Statistics query including Basic, Extended and per queue.
89 - Tunnel types: VXLAN, L3 VXLAN, VXLAN-GPE, GRE, MPLSoGRE, MPLSoUDP, IP-in-IP, Geneve, GTP.
90 - Tunnel HW offloads: packet type, inner/outer RSS, IP and UDP checksum verification.
91 - NIC HW offloads: encapsulation (vxlan, gre, mplsoudp, mplsogre), NAT, routing, TTL
92 increment/decrement, count, drop, mark. For details please see :ref:`mlx5_offloads_support`.
93 - Flow insertion rate of more then million flows per second, when using Direct Rules.
94 - Support for multiple rte_flow groups.
95 - Per packet no-inline hint flag to disable packet data copying into Tx descriptors.
102 - For secondary process:
104 - Forked secondary process not supported.
105 - External memory unregistered in EAL memseg list cannot be used for DMA
106 unless such memory has been registered by ``mlx5_mr_update_ext_mp()`` in
107 primary process and remapped to the same virtual address in secondary
108 process. If the external memory is registered by primary process but has
109 different virtual address in secondary process, unexpected error may happen.
111 - When using Verbs flow engine (``dv_flow_en`` = 0), flow pattern without any
112 specific VLAN will match for VLAN packets as well:
114 When VLAN spec is not specified in the pattern, the matching rule will be created with VLAN as a wild card.
115 Meaning, the flow rule::
117 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
119 Will only match vlan packets with vid=3. and the flow rule::
121 flow create 0 ingress pattern eth / ipv4 / end ...
123 Will match any ipv4 packet (VLAN included).
125 - When using DV flow engine (``dv_flow_en`` = 1), flow pattern without VLAN item
126 will match untagged packets only.
129 flow create 0 ingress pattern eth / ipv4 / end ...
131 Will match untagged packets only.
134 flow create 0 ingress pattern eth / vlan / ipv4 / end ...
136 Will match tagged packets only, with any VLAN ID value.
139 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
141 Will only match tagged packets with VLAN ID 3.
143 - VLAN pop offload command:
145 - Flow rules having a VLAN pop offload command as one of their actions and
146 are lacking a match on VLAN as one of their items are not supported.
147 - The command is not supported on egress traffic.
149 - VLAN push offload is not supported on ingress traffic.
151 - VLAN set PCP offload is not supported on existing headers.
153 - A multi segment packet must have not more segments than reported by dev_infos_get()
154 in tx_desc_lim.nb_seg_max field. This value depends on maximal supported Tx descriptor
155 size and ``txq_inline_min`` settings and may be from 2 (worst case forced by maximal
156 inline settings) to 58.
158 - Flows with a VXLAN Network Identifier equal (or ends to be equal)
159 to 0 are not supported.
161 - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with MPLSoGRE and MPLSoUDP.
163 - Match on Geneve header supports the following fields only:
169 Currently, the only supported options length value is 0.
171 - VF: flow rules created on VF devices can only match traffic targeted at the
172 configured MAC addresses (see ``rte_eth_dev_mac_addr_add()``).
174 - Match on GTP tunnel header item supports the following fields only:
176 - v_pt_rsv_flags: E flag, S flag, PN flag
180 - No Tx metadata go to the E-Switch steering domain for the Flow group 0.
181 The flows within group 0 and set metadata action are rejected by hardware.
185 MAC addresses not already present in the bridge table of the associated
186 kernel network device will be added and cleaned up by the PMD when closing
187 the device. In case of ungraceful program termination, some entries may
188 remain present and should be removed manually by other means.
190 - When Multi-Packet Rx queue is configured (``mprq_en``), a Rx packet can be
191 externally attached to a user-provided mbuf with having EXT_ATTACHED_MBUF in
192 ol_flags. As the mempool for the external buffer is managed by PMD, all the
193 Rx mbufs must be freed before the device is closed. Otherwise, the mempool of
194 the external buffers will be freed by PMD and the application which still
195 holds the external buffers may be corrupted.
197 - If Multi-Packet Rx queue is configured (``mprq_en``) and Rx CQE compression is
198 enabled (``rxq_cqe_comp_en``) at the same time, RSS hash result is not fully
199 supported. Some Rx packets may not have PKT_RX_RSS_HASH.
201 - IPv6 Multicast messages are not supported on VM, while promiscuous mode
202 and allmulticast mode are both set to off.
203 To receive IPv6 Multicast messages on VM, explicitly set the relevant
204 MAC address using rte_eth_dev_mac_addr_add() API.
206 - To support a mixed traffic pattern (some buffers from local host memory, some
207 buffers from other devices) with high bandwidth, a mbuf flag is used.
209 An application hints the PMD whether or not it should try to inline the
210 given mbuf data buffer. PMD should do the best effort to act upon this request.
212 The hint flag ``RTE_PMD_MLX5_FINE_GRANULARITY_INLINE`` is dynamic,
213 registered by application with rte_mbuf_dynflag_register(). This flag is
214 purely driver-specific and declared in PMD specific header ``rte_pmd_mlx5.h``,
215 which is intended to be used by the application.
217 To query the supported specific flags in runtime,
218 the function ``rte_pmd_mlx5_get_dyn_flag_names`` returns the array of
219 currently (over present hardware and configuration) supported specific flags.
220 The "not inline hint" feature operating flow is the following one:
223 - probe the devices, ports are created
224 - query the port capabilities
225 - if port supporting the feature is found
226 - register dynamic flag ``RTE_PMD_MLX5_FINE_GRANULARITY_INLINE``
227 - application starts the ports
228 - on ``dev_start()`` PMD checks whether the feature flag is registered and
229 enables the feature support in datapath
230 - application might set the registered flag bit in ``ol_flags`` field
231 of mbuf being sent and PMD will handle ones appropriately.
233 - The amount of descriptors in Tx queue may be limited by data inline settings.
234 Inline data require the more descriptor building blocks and overall block
235 amount may exceed the hardware supported limits. The application should
236 reduce the requested Tx size or adjust data inline settings with
237 ``txq_inline_max`` and ``txq_inline_mpw`` devargs keys.
239 - To provide the packet send scheduling on mbuf timestamps the ``tx_pp``
240 parameter should be specified.
241 When PMD sees the RTE_MBUF_DYNFLAG_TX_TIMESTAMP_NAME set on the packet
242 being sent it tries to synchronize the time of packet appearing on
243 the wire with the specified packet timestamp. It the specified one
244 is in the past it should be ignored, if one is in the distant future
245 it should be capped with some reasonable value (in range of seconds).
246 These specific cases ("too late" and "distant future") can be optionally
247 reported via device xstats to assist applications to detect the
248 time-related problems.
250 The timestamp upper "too-distant-future" limit
251 at the moment of invoking the Tx burst routine
252 can be estimated as ``tx_pp`` option (in nanoseconds) multiplied by 2^23.
253 Please note, for the testpmd txonly mode,
254 the limit is deduced from the expression::
256 (n_tx_descriptors / burst_size + 1) * inter_burst_gap
258 There is no any packet reordering according timestamps is supposed,
259 neither within packet burst, nor between packets, it is an entirely
260 application responsibility to generate packets and its timestamps
261 in desired order. The timestamps can be put only in the first packet
262 in the burst providing the entire burst scheduling.
264 - E-Switch decapsulation Flow:
266 - can be applied to PF port only.
267 - must specify VF port action (packet redirection from PF to VF).
268 - optionally may specify tunnel inner source and destination MAC addresses.
270 - E-Switch encapsulation Flow:
272 - can be applied to VF ports only.
273 - must specify PF port action (packet redirection from VF to PF).
277 - The input buffer, used as outer header, is not validated.
281 - The decapsulation is always done up to the outermost tunnel detected by the HW.
282 - The input buffer, providing the removal size, is not validated.
283 - The buffer size must match the length of the headers to be removed.
285 - ICMP(code/type/identifier/sequence number) / ICMP6(code/type) matching, IP-in-IP and MPLS flow matching are all
286 mutually exclusive features which cannot be supported together
287 (see :ref:`mlx5_firmware_config`).
291 - Requires DevX and DV flow to be enabled.
292 - KEEP_CRC offload cannot be supported with LRO.
293 - The first mbuf length, without head-room, must be big enough to include the
295 - Rx queue with LRO offload enabled, receiving a non-LRO packet, can forward
296 it with size limited to max LRO size, not to max RX packet length.
297 - LRO can be used with outer header of TCP packets of the standard format:
298 eth (with or without vlan) / ipv4 or ipv6 / tcp / payload
300 Other TCP packets (e.g. with MPLS label) received on Rx queue with LRO enabled, will be received with bad checksum.
304 - ``DEV_RX_OFFLOAD_KEEP_CRC`` cannot be supported with decapsulation
305 for some NICs (such as ConnectX-6 Dx and BlueField 2).
306 The capability bit ``scatter_fcs_w_decap_disable`` shows NIC support.
310 - Supports ``RTE_FLOW_ACTION_TYPE_SAMPLE`` action only within NIC Rx and E-Switch steering domain.
311 - The E-Switch Sample flow must have the eswitch_manager VPORT destination (PF or ECPF) and no additional actions.
312 - For ConnectX-5, the ``RTE_FLOW_ACTION_TYPE_SAMPLE`` is typically used as first action in the E-Switch egress flow if with header modify or encapsulation actions.
314 - IPv6 header item 'proto' field, indicating the next header protocol, should
315 not be set as extension header.
316 In case the next header is an extension header, it should not be specified in
317 IPv6 header item 'proto' field.
318 The last extension header item 'next header' field can specify the following
319 header protocol type.
324 MLX5 supports various methods to report statistics:
326 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.
328 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.
330 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.
338 The ibverbs libraries can be linked with this PMD in a number of ways,
339 configured by the ``ibverbs_link`` build option:
341 - ``shared`` (default): the PMD depends on some .so files.
343 - ``dlopen``: Split the dependencies glue in a separate library
344 loaded when needed by dlopen.
345 It make dependencies on libibverbs and libmlx4 optional,
346 and has no performance impact.
348 - ``static``: Embed static flavor of the dependencies libibverbs and libmlx4
349 in the PMD shared library or the executable static binary.
351 Environment variables
352 ~~~~~~~~~~~~~~~~~~~~~
356 A list of directories in which to search for the rdma-core "glue" plug-in,
357 separated by colons or semi-colons.
359 - ``MLX5_SHUT_UP_BF``
361 Configures HW Tx doorbell register as IO-mapped.
363 By default, the HW Tx doorbell is configured as a write-combining register.
364 The register would be flushed to HW usually when the write-combining buffer
365 becomes full, but it depends on CPU design.
367 Except for vectorized Tx burst routines, a write memory barrier is enforced
368 after updating the register so that the update can be immediately visible to
371 When vectorized Tx burst is called, the barrier is set only if the burst size
372 is not aligned to MLX5_VPMD_TX_MAX_BURST. However, setting this environmental
373 variable will bring better latency even though the maximum throughput can
376 Run-time configuration
377 ~~~~~~~~~~~~~~~~~~~~~~
379 - librte_net_mlx5 brings kernel network interfaces up during initialization
380 because it is affected by their state. Forcing them down prevents packets
383 - **ethtool** operations on related kernel interfaces also affect the PMD.
388 In order to run as a non-root user,
389 some capabilities must be granted to the application::
391 setcap cap_sys_admin,cap_net_admin,cap_net_raw,cap_ipc_lock+ep <dpdk-app>
393 Below are the reasons of the need for each capability:
396 When using physical addresses (PA mode), with Linux >= 4.0,
397 for access to ``/proc/self/pagemap``.
400 For device configuration.
403 For raw ethernet queue allocation through kernel driver.
406 For DMA memory pinning.
411 - ``rxq_cqe_comp_en`` parameter [int]
413 A nonzero value enables the compression of CQE on RX side. This feature
414 allows to save PCI bandwidth and improve performance. Enabled by default.
418 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
420 - POWER9 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
423 - ``rxq_cqe_pad_en`` parameter [int]
425 A nonzero value enables 128B padding of CQE on RX side. The size of CQE
426 is aligned with the size of a cacheline of the core. If cacheline size is
427 128B, the CQE size is configured to be 128B even though the device writes
428 only 64B data on the cacheline. This is to avoid unnecessary cache
429 invalidation by device's two consecutive writes on to one cacheline.
430 However in some architecture, it is more beneficial to update entire
431 cacheline with padding the rest 64B rather than striding because
432 read-modify-write could drop performance a lot. On the other hand,
433 writing extra data will consume more PCIe bandwidth and could also drop
434 the maximum throughput. It is recommended to empirically set this
435 parameter. Disabled by default.
439 - CPU having 128B cacheline with ConnectX-5 and BlueField.
441 - ``rxq_pkt_pad_en`` parameter [int]
443 A nonzero value enables padding Rx packet to the size of cacheline on PCI
444 transaction. This feature would waste PCI bandwidth but could improve
445 performance by avoiding partial cacheline write which may cause costly
446 read-modify-copy in memory transaction on some architectures. Disabled by
451 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
453 - POWER8 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
456 - ``mprq_en`` parameter [int]
458 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
459 configured as Multi-Packet RQ if the total number of Rx queues is
460 ``rxqs_min_mprq`` or more. Disabled by default.
462 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
463 by posting a single large buffer for multiple packets. Instead of posting a
464 buffers per a packet, one large buffer is posted in order to receive multiple
465 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
466 and each stride receives one packet. MPRQ can improve throughput for
467 small-packet traffic.
469 When MPRQ is enabled, max_rx_pkt_len can be larger than the size of
470 user-provided mbuf even if DEV_RX_OFFLOAD_SCATTER isn't enabled. PMD will
471 configure large stride size enough to accommodate max_rx_pkt_len as long as
472 device allows. Note that this can waste system memory compared to enabling Rx
473 scatter and multi-segment packet.
475 - ``mprq_log_stride_num`` parameter [int]
477 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
478 strides can reduce PCIe traffic further. If configured value is not in the
479 range of device capability, the default value will be set with a warning
480 message. The default value is 4 which is 16 strides per a buffer, valid only
481 if ``mprq_en`` is set.
483 The size of Rx queue should be bigger than the number of strides.
485 - ``mprq_log_stride_size`` parameter [int]
487 Log 2 of the size of a stride for Multi-Packet Rx queue. Configuring a smaller
488 stride size can save some memory and reduce probability of a depletion of all
489 available strides due to unreleased packets by an application. If configured
490 value is not in the range of device capability, the default value will be set
491 with a warning message. The default value is 11 which is 2048 bytes per a
492 stride, valid only if ``mprq_en`` is set. With ``mprq_log_stride_size`` set
493 it is possible for a packet to span across multiple strides. This mode allows
494 support of jumbo frames (9K) with MPRQ. The memcopy of some packets (or part
495 of a packet if Rx scatter is configured) may be required in case there is no
496 space left for a head room at the end of a stride which incurs some
499 - ``mprq_max_memcpy_len`` parameter [int]
501 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
502 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
503 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
504 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
505 A mempool for external buffers will be allocated and managed by PMD. If Rx
506 packet is externally attached, ol_flags field of the mbuf will have
507 EXT_ATTACHED_MBUF and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
508 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
510 - ``rxqs_min_mprq`` parameter [int]
512 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
513 greater or equal to this value. The default value is 12, valid only if
516 - ``txq_inline`` parameter [int]
518 Amount of data to be inlined during TX operations. This parameter is
519 deprecated and converted to the new parameter ``txq_inline_max`` providing
520 partial compatibility.
522 - ``txqs_min_inline`` parameter [int]
524 Enable inline data send only when the number of TX queues is greater or equal
527 This option should be used in combination with ``txq_inline_max`` and
528 ``txq_inline_mpw`` below and does not affect ``txq_inline_min`` settings above.
530 If this option is not specified the default value 16 is used for BlueField
531 and 8 for other platforms
533 The data inlining consumes the CPU cycles, so this option is intended to
534 auto enable inline data if we have enough Tx queues, which means we have
535 enough CPU cores and PCI bandwidth is getting more critical and CPU
536 is not supposed to be bottleneck anymore.
538 The copying data into WQE improves latency and can improve PPS performance
539 when PCI back pressure is detected and may be useful for scenarios involving
540 heavy traffic on many queues.
542 Because additional software logic is necessary to handle this mode, this
543 option should be used with care, as it may lower performance when back
544 pressure is not expected.
546 If inline data are enabled it may affect the maximal size of Tx queue in
547 descriptors because the inline data increase the descriptor size and
548 queue size limits supported by hardware may be exceeded.
550 - ``txq_inline_min`` parameter [int]
552 Minimal amount of data to be inlined into WQE during Tx operations. NICs
553 may require this minimal data amount to operate correctly. The exact value
554 may depend on NIC operation mode, requested offloads, etc. It is strongly
555 recommended to omit this parameter and use the default values. Anyway,
556 applications using this parameter should take into consideration that
557 specifying an inconsistent value may prevent the NIC from sending packets.
559 If ``txq_inline_min`` key is present the specified value (may be aligned
560 by the driver in order not to exceed the limits and provide better descriptor
561 space utilization) will be used by the driver and it is guaranteed that
562 requested amount of data bytes are inlined into the WQE beside other inline
563 settings. This key also may update ``txq_inline_max`` value (default
564 or specified explicitly in devargs) to reserve the space for inline data.
566 If ``txq_inline_min`` key is not present, the value may be queried by the
567 driver from the NIC via DevX if this feature is available. If there is no DevX
568 enabled/supported the value 18 (supposing L2 header including VLAN) is set
569 for ConnectX-4 and ConnectX-4 Lx, and 0 is set by default for ConnectX-5
570 and newer NICs. If packet is shorter the ``txq_inline_min`` value, the entire
573 For ConnectX-4 NIC, driver does not allow specifying value below 18
574 (minimal L2 header, including VLAN), error will be raised.
576 For ConnectX-4 Lx NIC, it is allowed to specify values below 18, but
577 it is not recommended and may prevent NIC from sending packets over
580 Please, note, this minimal data inlining disengages eMPW feature (Enhanced
581 Multi-Packet Write), because last one does not support partial packet inlining.
582 This is not very critical due to minimal data inlining is mostly required
583 by ConnectX-4 and ConnectX-4 Lx, these NICs do not support eMPW feature.
585 - ``txq_inline_max`` parameter [int]
587 Specifies the maximal packet length to be completely inlined into WQE
588 Ethernet Segment for ordinary SEND method. If packet is larger than specified
589 value, the packet data won't be copied by the driver at all, data buffer
590 is addressed with a pointer. If packet length is less or equal all packet
591 data will be copied into WQE. This may improve PCI bandwidth utilization for
592 short packets significantly but requires the extra CPU cycles.
594 The data inline feature is controlled by number of Tx queues, if number of Tx
595 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
596 is engaged, if there are not enough Tx queues (which means not enough CPU cores
597 and CPU resources are scarce), data inline is not performed by the driver.
598 Assigning ``txqs_min_inline`` with zero always enables the data inline.
600 The default ``txq_inline_max`` value is 290. The specified value may be adjusted
601 by the driver in order not to exceed the limit (930 bytes) and to provide better
602 WQE space filling without gaps, the adjustment is reflected in the debug log.
603 Also, the default value (290) may be decreased in run-time if the large transmit
604 queue size is requested and hardware does not support enough descriptor
605 amount, in this case warning is emitted. If ``txq_inline_max`` key is
606 specified and requested inline settings can not be satisfied then error
609 - ``txq_inline_mpw`` parameter [int]
611 Specifies the maximal packet length to be completely inlined into WQE for
612 Enhanced MPW method. If packet is large the specified value, the packet data
613 won't be copied, and data buffer is addressed with pointer. If packet length
614 is less or equal, all packet data will be copied into WQE. This may improve PCI
615 bandwidth utilization for short packets significantly but requires the extra
618 The data inline feature is controlled by number of TX queues, if number of Tx
619 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
620 is engaged, if there are not enough Tx queues (which means not enough CPU cores
621 and CPU resources are scarce), data inline is not performed by the driver.
622 Assigning ``txqs_min_inline`` with zero always enables the data inline.
624 The default ``txq_inline_mpw`` value is 268. The specified value may be adjusted
625 by the driver in order not to exceed the limit (930 bytes) and to provide better
626 WQE space filling without gaps, the adjustment is reflected in the debug log.
627 Due to multiple packets may be included to the same WQE with Enhanced Multi
628 Packet Write Method and overall WQE size is limited it is not recommended to
629 specify large values for the ``txq_inline_mpw``. Also, the default value (268)
630 may be decreased in run-time if the large transmit queue size is requested
631 and hardware does not support enough descriptor amount, in this case warning
632 is emitted. If ``txq_inline_mpw`` key is specified and requested inline
633 settings can not be satisfied then error will be raised.
635 - ``txqs_max_vec`` parameter [int]
637 Enable vectorized Tx only when the number of TX queues is less than or
638 equal to this value. This parameter is deprecated and ignored, kept
639 for compatibility issue to not prevent driver from probing.
641 - ``txq_mpw_hdr_dseg_en`` parameter [int]
643 A nonzero value enables including two pointers in the first block of TX
644 descriptor. The parameter is deprecated and ignored, kept for compatibility
647 - ``txq_max_inline_len`` parameter [int]
649 Maximum size of packet to be inlined. This limits the size of packet to
650 be inlined. If the size of a packet is larger than configured value, the
651 packet isn't inlined even though there's enough space remained in the
652 descriptor. Instead, the packet is included with pointer. This parameter
653 is deprecated and converted directly to ``txq_inline_mpw`` providing full
654 compatibility. Valid only if eMPW feature is engaged.
656 - ``txq_mpw_en`` parameter [int]
658 A nonzero value enables Enhanced Multi-Packet Write (eMPW) for ConnectX-5,
659 ConnectX-6, ConnectX-6 Dx and BlueField. eMPW allows the TX burst function to pack
660 up multiple packets in a single descriptor session in order to save PCI bandwidth
661 and improve performance at the cost of a slightly higher CPU usage. When
662 ``txq_inline_mpw`` is set along with ``txq_mpw_en``, TX burst function copies
663 entire packet data on to TX descriptor instead of including pointer of packet.
665 The Enhanced Multi-Packet Write feature is enabled by default if NIC supports
666 it, can be disabled by explicit specifying 0 value for ``txq_mpw_en`` option.
667 Also, if minimal data inlining is requested by non-zero ``txq_inline_min``
668 option or reported by the NIC, the eMPW feature is disengaged.
670 - ``tx_db_nc`` parameter [int]
672 The rdma core library can map doorbell register in two ways, depending on the
673 environment variable "MLX5_SHUT_UP_BF":
675 - As regular cached memory (usually with write combining attribute), if the
676 variable is either missing or set to zero.
677 - As non-cached memory, if the variable is present and set to not "0" value.
679 The type of mapping may slightly affect the Tx performance, the optimal choice
680 is strongly relied on the host architecture and should be deduced practically.
682 If ``tx_db_nc`` is set to zero, the doorbell is forced to be mapped to regular
683 memory (with write combining), the PMD will perform the extra write memory barrier
684 after writing to doorbell, it might increase the needed CPU clocks per packet
685 to send, but latency might be improved.
687 If ``tx_db_nc`` is set to one, the doorbell is forced to be mapped to non
688 cached memory, the PMD will not perform the extra write memory barrier
689 after writing to doorbell, on some architectures it might improve the
692 If ``tx_db_nc`` is set to two, the doorbell is forced to be mapped to regular
693 memory, the PMD will use heuristics to decide whether write memory barrier
694 should be performed. For bursts with size multiple of recommended one (64 pkts)
695 it is supposed the next burst is coming and no need to issue the extra memory
696 barrier (it is supposed to be issued in the next coming burst, at least after
697 descriptor writing). It might increase latency (on some hosts till next
698 packets transmit) and should be used with care.
700 If ``tx_db_nc`` is omitted or set to zero, the preset (if any) environment
701 variable "MLX5_SHUT_UP_BF" value is used. If there is no "MLX5_SHUT_UP_BF",
702 the default ``tx_db_nc`` value is zero for ARM64 hosts and one for others.
704 - ``tx_pp`` parameter [int]
706 If a nonzero value is specified the driver creates all necessary internal
707 objects to provide accurate packet send scheduling on mbuf timestamps.
708 The positive value specifies the scheduling granularity in nanoseconds,
709 the packet send will be accurate up to specified digits. The allowed range is
710 from 500 to 1 million of nanoseconds. The negative value specifies the module
711 of granularity and engages the special test mode the check the schedule rate.
712 By default (if the ``tx_pp`` is not specified) send scheduling on timestamps
715 - ``tx_skew`` parameter [int]
717 The parameter adjusts the send packet scheduling on timestamps and represents
718 the average delay between beginning of the transmitting descriptor processing
719 by the hardware and appearance of actual packet data on the wire. The value
720 should be provided in nanoseconds and is valid only if ``tx_pp`` parameter is
721 specified. The default value is zero.
723 - ``tx_vec_en`` parameter [int]
725 A nonzero value enables Tx vector on ConnectX-5, ConnectX-6, ConnectX-6 Dx
726 and BlueField NICs if the number of global Tx queues on the port is less than
727 ``txqs_max_vec``. The parameter is deprecated and ignored.
729 - ``rx_vec_en`` parameter [int]
731 A nonzero value enables Rx vector if the port is not configured in
732 multi-segment otherwise this parameter is ignored.
736 - ``vf_nl_en`` parameter [int]
738 A nonzero value enables Netlink requests from the VF to add/remove MAC
739 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
740 Otherwise the relevant configuration must be run with Linux iproute2 tools.
741 This is a prerequisite to receive this kind of traffic.
743 Enabled by default, valid only on VF devices ignored otherwise.
745 - ``l3_vxlan_en`` parameter [int]
747 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
748 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
749 parameter. This is a prerequisite to receive this kind of traffic.
753 - ``dv_xmeta_en`` parameter [int]
755 A nonzero value enables extensive flow metadata support if device is
756 capable and driver supports it. This can enable extensive support of
757 ``MARK`` and ``META`` item of ``rte_flow``. The newly introduced
758 ``SET_TAG`` and ``SET_META`` actions do not depend on ``dv_xmeta_en``.
760 There are some possible configurations, depending on parameter value:
762 - 0, this is default value, defines the legacy mode, the ``MARK`` and
763 ``META`` related actions and items operate only within NIC Tx and
764 NIC Rx steering domains, no ``MARK`` and ``META`` information crosses
765 the domain boundaries. The ``MARK`` item is 24 bits wide, the ``META``
766 item is 32 bits wide and match supported on egress only.
768 - 1, this engages extensive metadata mode, the ``MARK`` and ``META``
769 related actions and items operate within all supported steering domains,
770 including FDB, ``MARK`` and ``META`` information may cross the domain
771 boundaries. The ``MARK`` item is 24 bits wide, the ``META`` item width
772 depends on kernel and firmware configurations and might be 0, 16 or
773 32 bits. Within NIC Tx domain ``META`` data width is 32 bits for
774 compatibility, the actual width of data transferred to the FDB domain
775 depends on kernel configuration and may be vary. The actual supported
776 width can be retrieved in runtime by series of rte_flow_validate()
779 - 2, this engages extensive metadata mode, the ``MARK`` and ``META``
780 related actions and items operate within all supported steering domains,
781 including FDB, ``MARK`` and ``META`` information may cross the domain
782 boundaries. The ``META`` item is 32 bits wide, the ``MARK`` item width
783 depends on kernel and firmware configurations and might be 0, 16 or
784 24 bits. The actual supported width can be retrieved in runtime by
785 series of rte_flow_validate() trials.
787 +------+-----------+-----------+-------------+-------------+
788 | Mode | ``MARK`` | ``META`` | ``META`` Tx | FDB/Through |
789 +======+===========+===========+=============+=============+
790 | 0 | 24 bits | 32 bits | 32 bits | no |
791 +------+-----------+-----------+-------------+-------------+
792 | 1 | 24 bits | vary 0-32 | 32 bits | yes |
793 +------+-----------+-----------+-------------+-------------+
794 | 2 | vary 0-32 | 32 bits | 32 bits | yes |
795 +------+-----------+-----------+-------------+-------------+
797 If there is no E-Switch configuration the ``dv_xmeta_en`` parameter is
798 ignored and the device is configured to operate in legacy mode (0).
800 Disabled by default (set to 0).
802 The Direct Verbs/Rules (engaged with ``dv_flow_en`` = 1) supports all
803 of the extensive metadata features. The legacy Verbs supports FLAG and
804 MARK metadata actions over NIC Rx steering domain only.
806 - ``dv_flow_en`` parameter [int]
808 A nonzero value enables the DV flow steering assuming it is supported
809 by the driver (RDMA Core library version is rdma-core-24.0 or higher).
811 Enabled by default if supported.
813 - ``dv_esw_en`` parameter [int]
815 A nonzero value enables E-Switch using Direct Rules.
817 Enabled by default if supported.
819 - ``lacp_by_user`` parameter [int]
821 A nonzero value enables the control of LACP traffic by the user application.
822 When a bond exists in the driver, by default it should be managed by the
823 kernel and therefore LACP traffic should be steered to the kernel.
824 If this devarg is set to 1 it will allow the user to manage the bond by
825 itself and not steer LACP traffic to the kernel.
827 Disabled by default (set to 0).
829 - ``mr_ext_memseg_en`` parameter [int]
831 A nonzero value enables extending memseg when registering DMA memory. If
832 enabled, the number of entries in MR (Memory Region) lookup table on datapath
833 is minimized and it benefits performance. On the other hand, it worsens memory
834 utilization because registered memory is pinned by kernel driver. Even if a
835 page in the extended chunk is freed, that doesn't become reusable until the
836 entire memory is freed.
840 - ``representor`` parameter [list]
842 This parameter can be used to instantiate DPDK Ethernet devices from
843 existing port (or VF) representors configured on the device.
845 It is a standard parameter whose format is described in
846 :ref:`ethernet_device_standard_device_arguments`.
848 For instance, to probe port representors 0 through 2::
852 - ``max_dump_files_num`` parameter [int]
854 The maximum number of files per PMD entity that may be created for debug information.
855 The files will be created in /var/log directory or in current directory.
857 set to 128 by default.
859 - ``lro_timeout_usec`` parameter [int]
861 The maximum allowed duration of an LRO session, in micro-seconds.
862 PMD will set the nearest value supported by HW, which is not bigger than
863 the input ``lro_timeout_usec`` value.
864 If this parameter is not specified, by default PMD will set
865 the smallest value supported by HW.
867 - ``hp_buf_log_sz`` parameter [int]
869 The total data buffer size of a hairpin queue (logarithmic form), in bytes.
870 PMD will set the data buffer size to 2 ** ``hp_buf_log_sz``, both for RX & TX.
871 The capacity of the value is specified by the firmware and the initialization
872 will get a failure if it is out of scope.
873 The range of the value is from 11 to 19 right now, and the supported frame
874 size of a single packet for hairpin is from 512B to 128KB. It might change if
875 different firmware release is being used. By using a small value, it could
876 reduce memory consumption but not work with a large frame. If the value is
877 too large, the memory consumption will be high and some potential performance
878 degradation will be introduced.
879 By default, the PMD will set this value to 16, which means that 9KB jumbo
880 frames will be supported.
882 - ``reclaim_mem_mode`` parameter [int]
884 Cache some resources in flow destroy will help flow recreation more efficient.
885 While some systems may require the all the resources can be reclaimed after
887 The parameter ``reclaim_mem_mode`` provides the option for user to configure
888 if the resource cache is needed or not.
890 There are three options to choose:
892 - 0. It means the flow resources will be cached as usual. The resources will
893 be cached, helpful with flow insertion rate.
895 - 1. It will only enable the DPDK PMD level resources reclaim.
897 - 2. Both DPDK PMD level and rdma-core low level will be configured as
900 By default, the PMD will set this value to 0.
902 - ``sys_mem_en`` parameter [int]
904 A non-zero value enables the PMD memory management allocating memory
905 from system by default, without explicit rte memory flag.
907 By default, the PMD will set this value to 0.
909 - ``decap_en`` parameter [int]
911 Some devices do not support FCS (frame checksum) scattering for
912 tunnel-decapsulated packets.
913 If set to 0, this option forces the FCS feature and rejects tunnel
914 decapsulation in the flow engine for such devices.
916 By default, the PMD will set this value to 1.
918 .. _mlx5_firmware_config:
920 Firmware configuration
921 ~~~~~~~~~~~~~~~~~~~~~~
923 Firmware features can be configured as key/value pairs.
925 The command to set a value is::
927 mlxconfig -d <device> set <key>=<value>
929 The command to query a value is::
931 mlxconfig -d <device> query | grep <key>
933 The device name for the command ``mlxconfig`` can be either the PCI address,
934 or the mst device name found with::
938 Below are some firmware configurations listed.
944 value: 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
950 - maximum number of SR-IOV virtual functions::
954 - enable DevX (required by Direct Rules and other features)::
958 - aggressive CQE zipping::
962 - L3 VXLAN and VXLAN-GPE destination UDP port::
965 IP_OVER_VXLAN_PORT=<udp dport>
967 - enable VXLAN-GPE tunnel flow matching::
969 FLEX_PARSER_PROFILE_ENABLE=0
971 FLEX_PARSER_PROFILE_ENABLE=2
973 - enable IP-in-IP tunnel flow matching::
975 FLEX_PARSER_PROFILE_ENABLE=0
977 - enable MPLS flow matching::
979 FLEX_PARSER_PROFILE_ENABLE=1
981 - enable ICMP(code/type/identifier/sequence number) / ICMP6(code/type) fields matching::
983 FLEX_PARSER_PROFILE_ENABLE=2
985 - enable Geneve flow matching::
987 FLEX_PARSER_PROFILE_ENABLE=0
989 FLEX_PARSER_PROFILE_ENABLE=1
991 - enable GTP flow matching::
993 FLEX_PARSER_PROFILE_ENABLE=3
995 - enable eCPRI flow matching::
997 FLEX_PARSER_PROFILE_ENABLE=4
1003 This driver relies on external libraries and kernel drivers for resources
1004 allocations and initialization. The following dependencies are not part of
1005 DPDK and must be installed separately:
1009 User space Verbs framework used by librte_net_mlx5. This library provides
1010 a generic interface between the kernel and low-level user space drivers
1013 It allows slow and privileged operations (context initialization, hardware
1014 resources allocations) to be managed by the kernel and fast operations to
1015 never leave user space.
1019 Low-level user space driver library for Mellanox
1020 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices, it is automatically loaded
1023 This library basically implements send/receive calls to the hardware
1026 - **Kernel modules**
1028 They provide the kernel-side Verbs API and low level device drivers that
1029 manage actual hardware initialization and resources sharing with user
1032 Unlike most other PMDs, these modules must remain loaded and bound to
1035 - mlx5_core: hardware driver managing Mellanox
1036 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices and related Ethernet kernel
1038 - mlx5_ib: InifiniBand device driver.
1039 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
1041 - **Firmware update**
1043 Mellanox OFED/EN releases include firmware updates for
1044 ConnectX-4/ConnectX-5/ConnectX-6/BlueField adapters.
1046 Because each release provides new features, these updates must be applied to
1047 match the kernel modules and libraries they come with.
1051 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
1057 Either RDMA Core library with a recent enough Linux kernel release
1058 (recommended) or Mellanox OFED/EN, which provides compatibility with older
1061 RDMA Core with Linux Kernel
1062 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1064 - Minimal kernel version : v4.14 or the most recent 4.14-rc (see `Linux installation documentation`_)
1065 - Minimal rdma-core version: v15+ commit 0c5f5765213a ("Merge pull request #227 from yishaih/tm")
1066 (see `RDMA Core installation documentation`_)
1067 - When building for i686 use:
1069 - rdma-core version 18.0 or above built with 32bit support.
1070 - Kernel version 4.14.41 or above.
1072 - Starting with rdma-core v21, static libraries can be built::
1075 CFLAGS=-fPIC cmake -DIN_PLACE=1 -DENABLE_STATIC=1 -GNinja ..
1078 .. _`Linux installation documentation`: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/plain/Documentation/admin-guide/README.rst
1079 .. _`RDMA Core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
1085 - Mellanox OFED version: **4.5** and above /
1086 Mellanox EN version: **4.5** and above
1089 - ConnectX-4: **12.21.1000** and above.
1090 - ConnectX-4 Lx: **14.21.1000** and above.
1091 - ConnectX-5: **16.21.1000** and above.
1092 - ConnectX-5 Ex: **16.21.1000** and above.
1093 - ConnectX-6: **20.27.0090** and above.
1094 - ConnectX-6 Dx: **22.27.0090** and above.
1095 - BlueField: **18.25.1010** and above.
1097 While these libraries and kernel modules are available on OpenFabrics
1098 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
1099 managers on most distributions, this PMD requires Ethernet extensions that
1100 may not be supported at the moment (this is a work in progress).
1103 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__ and
1105 <http://www.mellanox.com/page/products_dyn?product_family=27&mtag=linux>`__
1106 include the necessary support and should be used in the meantime. For DPDK,
1107 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
1108 required from that distribution.
1112 Several versions of Mellanox OFED/EN are available. Installing the version
1113 this DPDK release was developed and tested against is strongly
1114 recommended. Please check the `prerequisites`_.
1119 The following Mellanox device families are supported by the same mlx5 driver:
1129 Below are detailed device names:
1131 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX4111A-XCAT (1x10G)
1132 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX412A-XCAT (2x10G)
1133 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX4111A-ACAT (1x25G)
1134 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX412A-ACAT (2x25G)
1135 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX413A-BCAT (1x40G)
1136 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX4131A-BCAT (1x40G)
1137 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX415A-BCAT (1x40G)
1138 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX413A-GCAT (1x50G)
1139 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX4131A-GCAT (1x50G)
1140 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX414A-BCAT (2x50G)
1141 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-GCAT (1x50G)
1142 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-BCAT (2x50G)
1143 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-GCAT (2x50G)
1144 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-CCAT (1x100G)
1145 * Mellanox\ |reg| ConnectX\ |reg|-4 100G MCX416A-CCAT (2x100G)
1146 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4111A-XCAT (1x10G)
1147 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4121A-XCAT (2x10G)
1148 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4111A-ACAT (1x25G)
1149 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4121A-ACAT (2x25G)
1150 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 40G MCX4131A-BCAT (1x40G)
1151 * Mellanox\ |reg| ConnectX\ |reg|-5 100G MCX556A-ECAT (2x100G)
1152 * Mellanox\ |reg| ConnectX\ |reg|-5 Ex EN 100G MCX516A-CDAT (2x100G)
1153 * Mellanox\ |reg| ConnectX\ |reg|-6 200G MCX654106A-HCAT (2x200G)
1154 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 100G MCX623106AN-CDAT (2x100G)
1155 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 200G MCX623105AN-VDAT (1x200G)
1157 Quick Start Guide on OFED/EN
1158 ----------------------------
1160 1. Download latest Mellanox OFED/EN. For more info check the `prerequisites`_.
1163 2. Install the required libraries and kernel modules either by installing
1164 only the required set, or by installing the entire Mellanox OFED/EN::
1166 ./mlnxofedinstall --upstream-libs --dpdk
1168 3. Verify the firmware is the correct one::
1172 4. Verify all ports links are set to Ethernet::
1174 mlxconfig -d <mst device> query | grep LINK_TYPE
1178 Link types may have to be configured to Ethernet::
1180 mlxconfig -d <mst device> set LINK_TYPE_P1/2=1/2/3
1182 * LINK_TYPE_P1=<1|2|3> , 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
1184 For hypervisors, verify SR-IOV is enabled on the NIC::
1186 mlxconfig -d <mst device> query | grep SRIOV_EN
1189 If needed, configure SR-IOV::
1191 mlxconfig -d <mst device> set SRIOV_EN=1 NUM_OF_VFS=16
1192 mlxfwreset -d <mst device> reset
1194 5. Restart the driver::
1196 /etc/init.d/openibd restart
1200 service openibd restart
1202 If link type was changed, firmware must be reset as well::
1204 mlxfwreset -d <mst device> reset
1206 For hypervisors, after reset write the sysfs number of virtual functions
1209 To dynamically instantiate a given number of virtual functions (VFs)::
1211 echo [num_vfs] > /sys/class/infiniband/mlx5_0/device/sriov_numvfs
1213 6. Install DPDK and you are ready to go.
1214 See :doc:`compilation instructions <../linux_gsg/build_dpdk>`.
1216 Enable switchdev mode
1217 ---------------------
1219 Switchdev mode is a mode in E-Switch, that binds between representor and VF.
1220 Representor is a port in DPDK that is connected to a VF in such a way
1221 that assuming there are no offload flows, each packet that is sent from the VF
1222 will be received by the corresponding representor. While each packet that is
1223 sent to a representor will be received by the VF.
1224 This is very useful in case of SRIOV mode, where the first packet that is sent
1225 by the VF will be received by the DPDK application which will decide if this
1226 flow should be offloaded to the E-Switch. After offloading the flow packet
1227 that the VF that are matching the flow will not be received any more by
1228 the DPDK application.
1230 1. Enable SRIOV mode::
1232 mlxconfig -d <mst device> set SRIOV_EN=true
1234 2. Configure the max number of VFs::
1236 mlxconfig -d <mst device> set NUM_OF_VFS=<num of vfs>
1240 mlxfwreset -d <mst device> reset
1242 3. Configure the actual number of VFs::
1244 echo <num of vfs > /sys/class/net/<net device>/device/sriov_numvfs
1246 4. Unbind the device (can be rebind after the switchdev mode)::
1248 echo -n "<device pci address" > /sys/bus/pci/drivers/mlx5_core/unbind
1250 5. Enbale switchdev mode::
1252 echo switchdev > /sys/class/net/<net device>/compat/devlink/mode
1257 1. Configure aggressive CQE Zipping for maximum performance::
1259 mlxconfig -d <mst device> s CQE_COMPRESSION=1
1261 To set it back to the default CQE Zipping mode use::
1263 mlxconfig -d <mst device> s CQE_COMPRESSION=0
1265 2. In case of virtualization:
1267 - Make sure that hypervisor kernel is 3.16 or newer.
1268 - Configure boot with ``iommu=pt``.
1269 - Use 1G huge pages.
1270 - Make sure to allocate a VM on huge pages.
1271 - Make sure to set CPU pinning.
1273 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
1274 for better performance. For VMs, verify that the right CPU
1275 and NUMA node are pinned according to the above. Run::
1279 to identify the NUMA node to which the PCIe adapter is connected.
1281 4. If more than one adapter is used, and root complex capabilities allow
1282 to put both adapters on the same NUMA node without PCI bandwidth degradation,
1283 it is recommended to locate both adapters on the same NUMA node.
1284 This in order to forward packets from one to the other without
1285 NUMA performance penalty.
1287 5. Disable pause frames::
1289 ethtool -A <netdev> rx off tx off
1291 6. Verify IO non-posted prefetch is disabled by default. This can be checked
1292 via the BIOS configuration. Please contact you server provider for more
1293 information about the settings.
1297 On some machines, depends on the machine integrator, it is beneficial
1298 to set the PCI max read request parameter to 1K. This can be
1299 done in the following way:
1301 To query the read request size use::
1303 setpci -s <NIC PCI address> 68.w
1305 If the output is different than 3XXX, set it by::
1307 setpci -s <NIC PCI address> 68.w=3XXX
1309 The XXX can be different on different systems. Make sure to configure
1310 according to the setpci output.
1312 7. To minimize overhead of searching Memory Regions:
1314 - '--socket-mem' is recommended to pin memory by predictable amount.
1315 - Configure per-lcore cache when creating Mempools for packet buffer.
1316 - Refrain from dynamically allocating/freeing memory in run-time.
1318 .. _mlx5_offloads_support:
1320 Supported hardware offloads
1321 ---------------------------
1323 .. table:: Minimal SW/HW versions for queue offloads
1325 ============== ===== ===== ========= ===== ========== ==========
1326 Offload DPDK Linux rdma-core OFED firmware hardware
1327 ============== ===== ===== ========= ===== ========== ==========
1328 common base 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1329 checksums 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1330 Rx timestamp 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1331 TSO 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1332 LRO 19.08 N/A N/A 4.6-4 16.25.6406 ConnectX-5
1333 ============== ===== ===== ========= ===== ========== ==========
1335 .. table:: Minimal SW/HW versions for rte_flow offloads
1337 +-----------------------+-----------------+-----------------+
1338 | Offload | with E-Switch | with NIC |
1339 +=======================+=================+=================+
1340 | Count | | DPDK 19.05 | | DPDK 19.02 |
1341 | | | OFED 4.6 | | OFED 4.6 |
1342 | | | rdma-core 24 | | rdma-core 23 |
1343 | | | ConnectX-5 | | ConnectX-5 |
1344 +-----------------------+-----------------+-----------------+
1345 | Drop | | DPDK 19.05 | | DPDK 18.11 |
1346 | | | OFED 4.6 | | OFED 4.5 |
1347 | | | rdma-core 24 | | rdma-core 23 |
1348 | | | ConnectX-5 | | ConnectX-4 |
1349 +-----------------------+-----------------+-----------------+
1350 | Queue / RSS | | | | DPDK 18.11 |
1351 | | | N/A | | OFED 4.5 |
1352 | | | | | rdma-core 23 |
1353 | | | | | ConnectX-4 |
1354 +-----------------------+-----------------+-----------------+
1355 | Encapsulation | | DPDK 19.05 | | DPDK 19.02 |
1356 | (VXLAN / NVGRE / RAW) | | OFED 4.7-1 | | OFED 4.6 |
1357 | | | rdma-core 24 | | rdma-core 23 |
1358 | | | ConnectX-5 | | ConnectX-5 |
1359 +-----------------------+-----------------+-----------------+
1360 | Encapsulation | | DPDK 19.11 | | DPDK 19.11 |
1361 | GENEVE | | OFED 4.7-3 | | OFED 4.7-3 |
1362 | | | rdma-core 27 | | rdma-core 27 |
1363 | | | ConnectX-5 | | ConnectX-5 |
1364 +-----------------------+-----------------+-----------------+
1365 | | Header rewrite | | DPDK 19.05 | | DPDK 19.02 |
1366 | | (set_ipv4_src / | | OFED 4.7-1 | | OFED 4.7-1 |
1367 | | set_ipv4_dst / | | rdma-core 24 | | rdma-core 24 |
1368 | | set_ipv6_src / | | ConnectX-5 | | ConnectX-5 |
1369 | | set_ipv6_dst / | | | | |
1370 | | set_tp_src / | | | | |
1371 | | set_tp_dst / | | | | |
1372 | | dec_ttl / | | | | |
1373 | | set_ttl / | | | | |
1374 | | set_mac_src / | | | | |
1375 | | set_mac_dst) | | | | |
1376 +-----------------------+-----------------+-----------------+
1377 | | Header rewrite | | DPDK 20.02 | | DPDK 20.02 |
1378 | | (set_dscp) | | OFED 5.0 | | OFED 5.0 |
1379 | | | | rdma-core 24 | | rdma-core 24 |
1380 | | | | ConnectX-5 | | ConnectX-5 |
1381 +-----------------------+-----------------+-----------------+
1382 | Jump | | DPDK 19.05 | | DPDK 19.02 |
1383 | | | OFED 4.7-1 | | OFED 4.7-1 |
1384 | | | rdma-core 24 | | N/A |
1385 | | | ConnectX-5 | | ConnectX-5 |
1386 +-----------------------+-----------------+-----------------+
1387 | Mark / Flag | | DPDK 19.05 | | DPDK 18.11 |
1388 | | | OFED 4.6 | | OFED 4.5 |
1389 | | | rdma-core 24 | | rdma-core 23 |
1390 | | | ConnectX-5 | | ConnectX-4 |
1391 +-----------------------+-----------------+-----------------+
1392 | Port ID | | DPDK 19.05 | | N/A |
1393 | | | OFED 4.7-1 | | N/A |
1394 | | | rdma-core 24 | | N/A |
1395 | | | ConnectX-5 | | N/A |
1396 +-----------------------+-----------------+-----------------+
1397 | | VLAN | | DPDK 19.11 | | DPDK 19.11 |
1398 | | (of_pop_vlan / | | OFED 4.7-1 | | OFED 4.7-1 |
1399 | | of_push_vlan / | | ConnectX-5 | | ConnectX-5 |
1400 | | of_set_vlan_pcp / | | | | |
1401 | | of_set_vlan_vid) | | | | |
1402 +-----------------------+-----------------+-----------------+
1403 | Hairpin | | | | DPDK 19.11 |
1404 | | | N/A | | OFED 4.7-3 |
1405 | | | | | rdma-core 26 |
1406 | | | | | ConnectX-5 |
1407 +-----------------------+-----------------+-----------------+
1408 | Meta data | | DPDK 19.11 | | DPDK 19.11 |
1409 | | | OFED 4.7-3 | | OFED 4.7-3 |
1410 | | | rdma-core 26 | | rdma-core 26 |
1411 | | | ConnectX-5 | | ConnectX-5 |
1412 +-----------------------+-----------------+-----------------+
1413 | Metering | | DPDK 19.11 | | DPDK 19.11 |
1414 | | | OFED 4.7-3 | | OFED 4.7-3 |
1415 | | | rdma-core 26 | | rdma-core 26 |
1416 | | | ConnectX-5 | | ConnectX-5 |
1417 +-----------------------+-----------------+-----------------+
1418 | Sampling | | DPDK 20.11 | | DPDK 20.11 |
1419 | | | OFED 5.2 | | OFED 5.2 |
1420 | | | rdma-core 32 | | rdma-core 32 |
1421 | | | ConnectX-5 | | ConnectX-5 |
1422 +-----------------------+-----------------+-----------------+
1427 MARK and META items are interrelated with datapath - they might move from/to
1428 the applications in mbuf fields. Hence, zero value for these items has the
1429 special meaning - it means "no metadata are provided", not zero values are
1430 treated by applications and PMD as valid ones.
1432 Moreover in the flow engine domain the value zero is acceptable to match and
1433 set, and we should allow to specify zero values as rte_flow parameters for the
1434 META and MARK items and actions. In the same time zero mask has no meaning and
1435 should be rejected on validation stage.
1440 Flows are not cached in the driver.
1441 When stopping a device port, all the flows created on this port from the
1442 application will be flushed automatically in the background.
1443 After stopping the device port, all flows on this port become invalid and
1444 not represented in the system.
1445 All references to these flows held by the application should be discarded
1446 directly but neither destroyed nor flushed.
1448 The application should re-create the flows as required after the port restart.
1453 Compared to librte_net_mlx4 that implements a single RSS configuration per
1454 port, librte_net_mlx5 supports per-protocol RSS configuration.
1456 Since ``testpmd`` defaults to IP RSS mode and there is currently no
1457 command-line parameter to enable additional protocols (UDP and TCP as well
1458 as IP), the following commands must be entered from its CLI to get the same
1459 behavior as librte_net_mlx4::
1462 > port config all rss all
1468 This section demonstrates how to launch **testpmd** with Mellanox
1469 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices managed by librte_net_mlx5.
1471 #. Load the kernel modules::
1473 modprobe -a ib_uverbs mlx5_core mlx5_ib
1475 Alternatively if MLNX_OFED/MLNX_EN is fully installed, the following script
1478 /etc/init.d/openibd restart
1482 User space I/O kernel modules (uio and igb_uio) are not used and do
1483 not have to be loaded.
1485 #. Make sure Ethernet interfaces are in working order and linked to kernel
1486 verbs. Related sysfs entries should be present::
1488 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
1497 #. Optionally, retrieve their PCI bus addresses for whitelisting::
1500 for intf in eth2 eth3 eth4 eth5;
1502 (cd "/sys/class/net/${intf}/device/" && pwd -P);
1505 sed -n 's,.*/\(.*\),-w \1,p'
1514 #. Request huge pages::
1516 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
1518 #. Start testpmd with basic parameters::
1520 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
1525 EAL: PCI device 0000:05:00.0 on NUMA socket 0
1526 EAL: probe driver: 15b3:1013 librte_net_mlx5
1527 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
1528 PMD: librte_net_mlx5: 1 port(s) detected
1529 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
1530 EAL: PCI device 0000:05:00.1 on NUMA socket 0
1531 EAL: probe driver: 15b3:1013 librte_net_mlx5
1532 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
1533 PMD: librte_net_mlx5: 1 port(s) detected
1534 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
1535 EAL: PCI device 0000:06:00.0 on NUMA socket 0
1536 EAL: probe driver: 15b3:1013 librte_net_mlx5
1537 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
1538 PMD: librte_net_mlx5: 1 port(s) detected
1539 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
1540 EAL: PCI device 0000:06:00.1 on NUMA socket 0
1541 EAL: probe driver: 15b3:1013 librte_net_mlx5
1542 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
1543 PMD: librte_net_mlx5: 1 port(s) detected
1544 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
1545 Interactive-mode selected
1546 Configuring Port 0 (socket 0)
1547 PMD: librte_net_mlx5: 0x8cba80: TX queues number update: 0 -> 2
1548 PMD: librte_net_mlx5: 0x8cba80: RX queues number update: 0 -> 2
1549 Port 0: E4:1D:2D:E7:0C:FE
1550 Configuring Port 1 (socket 0)
1551 PMD: librte_net_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
1552 PMD: librte_net_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
1553 Port 1: E4:1D:2D:E7:0C:FF
1554 Configuring Port 2 (socket 0)
1555 PMD: librte_net_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
1556 PMD: librte_net_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
1557 Port 2: E4:1D:2D:E7:0C:FA
1558 Configuring Port 3 (socket 0)
1559 PMD: librte_net_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
1560 PMD: librte_net_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
1561 Port 3: E4:1D:2D:E7:0C:FB
1562 Checking link statuses...
1563 Port 0 Link Up - speed 40000 Mbps - full-duplex
1564 Port 1 Link Up - speed 40000 Mbps - full-duplex
1565 Port 2 Link Up - speed 10000 Mbps - full-duplex
1566 Port 3 Link Up - speed 10000 Mbps - full-duplex
1573 This section demonstrates how to dump flows. Currently, it's possible to dump
1574 all flows with assistance of external tools.
1576 #. 2 ways to get flow raw file:
1578 - Using testpmd CLI:
1580 .. code-block:: console
1582 testpmd> flow dump <port> <output_file>
1584 - call rte_flow_dev_dump api:
1586 .. code-block:: console
1588 rte_flow_dev_dump(port, file, NULL);
1590 #. Dump human-readable flows from raw file:
1592 Get flow parsing tool from: https://github.com/Mellanox/mlx_steering_dump
1594 .. code-block:: console
1596 mlx_steering_dump.py -f <output_file>