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.
317 MLX5 supports various methods to report statistics:
319 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.
321 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.
323 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.
331 The ibverbs libraries can be linked with this PMD in a number of ways,
332 configured by the ``ibverbs_link`` build option:
334 - ``shared`` (default): the PMD depends on some .so files.
336 - ``dlopen``: Split the dependencies glue in a separate library
337 loaded when needed by dlopen.
338 It make dependencies on libibverbs and libmlx4 optional,
339 and has no performance impact.
341 - ``static``: Embed static flavor of the dependencies libibverbs and libmlx4
342 in the PMD shared library or the executable static binary.
344 Environment variables
345 ~~~~~~~~~~~~~~~~~~~~~
349 A list of directories in which to search for the rdma-core "glue" plug-in,
350 separated by colons or semi-colons.
352 - ``MLX5_SHUT_UP_BF``
354 Configures HW Tx doorbell register as IO-mapped.
356 By default, the HW Tx doorbell is configured as a write-combining register.
357 The register would be flushed to HW usually when the write-combining buffer
358 becomes full, but it depends on CPU design.
360 Except for vectorized Tx burst routines, a write memory barrier is enforced
361 after updating the register so that the update can be immediately visible to
364 When vectorized Tx burst is called, the barrier is set only if the burst size
365 is not aligned to MLX5_VPMD_TX_MAX_BURST. However, setting this environmental
366 variable will bring better latency even though the maximum throughput can
369 Run-time configuration
370 ~~~~~~~~~~~~~~~~~~~~~~
372 - librte_net_mlx5 brings kernel network interfaces up during initialization
373 because it is affected by their state. Forcing them down prevents packets
376 - **ethtool** operations on related kernel interfaces also affect the PMD.
381 In order to run as a non-root user,
382 some capabilities must be granted to the application::
384 setcap cap_sys_admin,cap_net_admin,cap_net_raw,cap_ipc_lock+ep <dpdk-app>
386 Below are the reasons of the need for each capability:
389 When using physical addresses (PA mode), with Linux >= 4.0,
390 for access to ``/proc/self/pagemap``.
393 For device configuration.
396 For raw ethernet queue allocation through kernel driver.
399 For DMA memory pinning.
404 - ``rxq_cqe_comp_en`` parameter [int]
406 A nonzero value enables the compression of CQE on RX side. This feature
407 allows to save PCI bandwidth and improve performance. Enabled by default.
411 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
413 - POWER9 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
416 - ``rxq_cqe_pad_en`` parameter [int]
418 A nonzero value enables 128B padding of CQE on RX side. The size of CQE
419 is aligned with the size of a cacheline of the core. If cacheline size is
420 128B, the CQE size is configured to be 128B even though the device writes
421 only 64B data on the cacheline. This is to avoid unnecessary cache
422 invalidation by device's two consecutive writes on to one cacheline.
423 However in some architecture, it is more beneficial to update entire
424 cacheline with padding the rest 64B rather than striding because
425 read-modify-write could drop performance a lot. On the other hand,
426 writing extra data will consume more PCIe bandwidth and could also drop
427 the maximum throughput. It is recommended to empirically set this
428 parameter. Disabled by default.
432 - CPU having 128B cacheline with ConnectX-5 and BlueField.
434 - ``rxq_pkt_pad_en`` parameter [int]
436 A nonzero value enables padding Rx packet to the size of cacheline on PCI
437 transaction. This feature would waste PCI bandwidth but could improve
438 performance by avoiding partial cacheline write which may cause costly
439 read-modify-copy in memory transaction on some architectures. Disabled by
444 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
446 - POWER8 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx
449 - ``mprq_en`` parameter [int]
451 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
452 configured as Multi-Packet RQ if the total number of Rx queues is
453 ``rxqs_min_mprq`` or more. Disabled by default.
455 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
456 by posting a single large buffer for multiple packets. Instead of posting a
457 buffers per a packet, one large buffer is posted in order to receive multiple
458 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
459 and each stride receives one packet. MPRQ can improve throughput for
460 small-packet traffic.
462 When MPRQ is enabled, max_rx_pkt_len can be larger than the size of
463 user-provided mbuf even if DEV_RX_OFFLOAD_SCATTER isn't enabled. PMD will
464 configure large stride size enough to accommodate max_rx_pkt_len as long as
465 device allows. Note that this can waste system memory compared to enabling Rx
466 scatter and multi-segment packet.
468 - ``mprq_log_stride_num`` parameter [int]
470 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
471 strides can reduce PCIe traffic further. If configured value is not in the
472 range of device capability, the default value will be set with a warning
473 message. The default value is 4 which is 16 strides per a buffer, valid only
474 if ``mprq_en`` is set.
476 The size of Rx queue should be bigger than the number of strides.
478 - ``mprq_log_stride_size`` parameter [int]
480 Log 2 of the size of a stride for Multi-Packet Rx queue. Configuring a smaller
481 stride size can save some memory and reduce probability of a depletion of all
482 available strides due to unreleased packets by an application. If configured
483 value is not in the range of device capability, the default value will be set
484 with a warning message. The default value is 11 which is 2048 bytes per a
485 stride, valid only if ``mprq_en`` is set. With ``mprq_log_stride_size`` set
486 it is possible for a packet to span across multiple strides. This mode allows
487 support of jumbo frames (9K) with MPRQ. The memcopy of some packets (or part
488 of a packet if Rx scatter is configured) may be required in case there is no
489 space left for a head room at the end of a stride which incurs some
492 - ``mprq_max_memcpy_len`` parameter [int]
494 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
495 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
496 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
497 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
498 A mempool for external buffers will be allocated and managed by PMD. If Rx
499 packet is externally attached, ol_flags field of the mbuf will have
500 EXT_ATTACHED_MBUF and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
501 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
503 - ``rxqs_min_mprq`` parameter [int]
505 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
506 greater or equal to this value. The default value is 12, valid only if
509 - ``txq_inline`` parameter [int]
511 Amount of data to be inlined during TX operations. This parameter is
512 deprecated and converted to the new parameter ``txq_inline_max`` providing
513 partial compatibility.
515 - ``txqs_min_inline`` parameter [int]
517 Enable inline data send only when the number of TX queues is greater or equal
520 This option should be used in combination with ``txq_inline_max`` and
521 ``txq_inline_mpw`` below and does not affect ``txq_inline_min`` settings above.
523 If this option is not specified the default value 16 is used for BlueField
524 and 8 for other platforms
526 The data inlining consumes the CPU cycles, so this option is intended to
527 auto enable inline data if we have enough Tx queues, which means we have
528 enough CPU cores and PCI bandwidth is getting more critical and CPU
529 is not supposed to be bottleneck anymore.
531 The copying data into WQE improves latency and can improve PPS performance
532 when PCI back pressure is detected and may be useful for scenarios involving
533 heavy traffic on many queues.
535 Because additional software logic is necessary to handle this mode, this
536 option should be used with care, as it may lower performance when back
537 pressure is not expected.
539 If inline data are enabled it may affect the maximal size of Tx queue in
540 descriptors because the inline data increase the descriptor size and
541 queue size limits supported by hardware may be exceeded.
543 - ``txq_inline_min`` parameter [int]
545 Minimal amount of data to be inlined into WQE during Tx operations. NICs
546 may require this minimal data amount to operate correctly. The exact value
547 may depend on NIC operation mode, requested offloads, etc. It is strongly
548 recommended to omit this parameter and use the default values. Anyway,
549 applications using this parameter should take into consideration that
550 specifying an inconsistent value may prevent the NIC from sending packets.
552 If ``txq_inline_min`` key is present the specified value (may be aligned
553 by the driver in order not to exceed the limits and provide better descriptor
554 space utilization) will be used by the driver and it is guaranteed that
555 requested amount of data bytes are inlined into the WQE beside other inline
556 settings. This key also may update ``txq_inline_max`` value (default
557 or specified explicitly in devargs) to reserve the space for inline data.
559 If ``txq_inline_min`` key is not present, the value may be queried by the
560 driver from the NIC via DevX if this feature is available. If there is no DevX
561 enabled/supported the value 18 (supposing L2 header including VLAN) is set
562 for ConnectX-4 and ConnectX-4 Lx, and 0 is set by default for ConnectX-5
563 and newer NICs. If packet is shorter the ``txq_inline_min`` value, the entire
566 For ConnectX-4 NIC, driver does not allow specifying value below 18
567 (minimal L2 header, including VLAN), error will be raised.
569 For ConnectX-4 Lx NIC, it is allowed to specify values below 18, but
570 it is not recommended and may prevent NIC from sending packets over
573 Please, note, this minimal data inlining disengages eMPW feature (Enhanced
574 Multi-Packet Write), because last one does not support partial packet inlining.
575 This is not very critical due to minimal data inlining is mostly required
576 by ConnectX-4 and ConnectX-4 Lx, these NICs do not support eMPW feature.
578 - ``txq_inline_max`` parameter [int]
580 Specifies the maximal packet length to be completely inlined into WQE
581 Ethernet Segment for ordinary SEND method. If packet is larger than specified
582 value, the packet data won't be copied by the driver at all, data buffer
583 is addressed with a pointer. If packet length is less or equal all packet
584 data will be copied into WQE. This may improve PCI bandwidth utilization for
585 short packets significantly but requires the extra CPU cycles.
587 The data inline feature is controlled by number of Tx queues, if number of Tx
588 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
589 is engaged, if there are not enough Tx queues (which means not enough CPU cores
590 and CPU resources are scarce), data inline is not performed by the driver.
591 Assigning ``txqs_min_inline`` with zero always enables the data inline.
593 The default ``txq_inline_max`` value is 290. The specified value may be adjusted
594 by the driver in order not to exceed the limit (930 bytes) and to provide better
595 WQE space filling without gaps, the adjustment is reflected in the debug log.
596 Also, the default value (290) may be decreased in run-time if the large transmit
597 queue size is requested and hardware does not support enough descriptor
598 amount, in this case warning is emitted. If ``txq_inline_max`` key is
599 specified and requested inline settings can not be satisfied then error
602 - ``txq_inline_mpw`` parameter [int]
604 Specifies the maximal packet length to be completely inlined into WQE for
605 Enhanced MPW method. If packet is large the specified value, the packet data
606 won't be copied, and data buffer is addressed with pointer. If packet length
607 is less or equal, all packet data will be copied into WQE. This may improve PCI
608 bandwidth utilization for short packets significantly but requires the extra
611 The data inline feature is controlled by number of TX queues, if number of Tx
612 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
613 is engaged, if there are not enough Tx queues (which means not enough CPU cores
614 and CPU resources are scarce), data inline is not performed by the driver.
615 Assigning ``txqs_min_inline`` with zero always enables the data inline.
617 The default ``txq_inline_mpw`` value is 268. The specified value may be adjusted
618 by the driver in order not to exceed the limit (930 bytes) and to provide better
619 WQE space filling without gaps, the adjustment is reflected in the debug log.
620 Due to multiple packets may be included to the same WQE with Enhanced Multi
621 Packet Write Method and overall WQE size is limited it is not recommended to
622 specify large values for the ``txq_inline_mpw``. Also, the default value (268)
623 may be decreased in run-time if the large transmit queue size is requested
624 and hardware does not support enough descriptor amount, in this case warning
625 is emitted. If ``txq_inline_mpw`` key is specified and requested inline
626 settings can not be satisfied then error will be raised.
628 - ``txqs_max_vec`` parameter [int]
630 Enable vectorized Tx only when the number of TX queues is less than or
631 equal to this value. This parameter is deprecated and ignored, kept
632 for compatibility issue to not prevent driver from probing.
634 - ``txq_mpw_hdr_dseg_en`` parameter [int]
636 A nonzero value enables including two pointers in the first block of TX
637 descriptor. The parameter is deprecated and ignored, kept for compatibility
640 - ``txq_max_inline_len`` parameter [int]
642 Maximum size of packet to be inlined. This limits the size of packet to
643 be inlined. If the size of a packet is larger than configured value, the
644 packet isn't inlined even though there's enough space remained in the
645 descriptor. Instead, the packet is included with pointer. This parameter
646 is deprecated and converted directly to ``txq_inline_mpw`` providing full
647 compatibility. Valid only if eMPW feature is engaged.
649 - ``txq_mpw_en`` parameter [int]
651 A nonzero value enables Enhanced Multi-Packet Write (eMPW) for ConnectX-5,
652 ConnectX-6, ConnectX-6 Dx and BlueField. eMPW allows the TX burst function to pack
653 up multiple packets in a single descriptor session in order to save PCI bandwidth
654 and improve performance at the cost of a slightly higher CPU usage. When
655 ``txq_inline_mpw`` is set along with ``txq_mpw_en``, TX burst function copies
656 entire packet data on to TX descriptor instead of including pointer of packet.
658 The Enhanced Multi-Packet Write feature is enabled by default if NIC supports
659 it, can be disabled by explicit specifying 0 value for ``txq_mpw_en`` option.
660 Also, if minimal data inlining is requested by non-zero ``txq_inline_min``
661 option or reported by the NIC, the eMPW feature is disengaged.
663 - ``tx_db_nc`` parameter [int]
665 The rdma core library can map doorbell register in two ways, depending on the
666 environment variable "MLX5_SHUT_UP_BF":
668 - As regular cached memory (usually with write combining attribute), if the
669 variable is either missing or set to zero.
670 - As non-cached memory, if the variable is present and set to not "0" value.
672 The type of mapping may slightly affect the Tx performance, the optimal choice
673 is strongly relied on the host architecture and should be deduced practically.
675 If ``tx_db_nc`` is set to zero, the doorbell is forced to be mapped to regular
676 memory (with write combining), the PMD will perform the extra write memory barrier
677 after writing to doorbell, it might increase the needed CPU clocks per packet
678 to send, but latency might be improved.
680 If ``tx_db_nc`` is set to one, the doorbell is forced to be mapped to non
681 cached memory, the PMD will not perform the extra write memory barrier
682 after writing to doorbell, on some architectures it might improve the
685 If ``tx_db_nc`` is set to two, the doorbell is forced to be mapped to regular
686 memory, the PMD will use heuristics to decide whether write memory barrier
687 should be performed. For bursts with size multiple of recommended one (64 pkts)
688 it is supposed the next burst is coming and no need to issue the extra memory
689 barrier (it is supposed to be issued in the next coming burst, at least after
690 descriptor writing). It might increase latency (on some hosts till next
691 packets transmit) and should be used with care.
693 If ``tx_db_nc`` is omitted or set to zero, the preset (if any) environment
694 variable "MLX5_SHUT_UP_BF" value is used. If there is no "MLX5_SHUT_UP_BF",
695 the default ``tx_db_nc`` value is zero for ARM64 hosts and one for others.
697 - ``tx_pp`` parameter [int]
699 If a nonzero value is specified the driver creates all necessary internal
700 objects to provide accurate packet send scheduling on mbuf timestamps.
701 The positive value specifies the scheduling granularity in nanoseconds,
702 the packet send will be accurate up to specified digits. The allowed range is
703 from 500 to 1 million of nanoseconds. The negative value specifies the module
704 of granularity and engages the special test mode the check the schedule rate.
705 By default (if the ``tx_pp`` is not specified) send scheduling on timestamps
708 - ``tx_skew`` parameter [int]
710 The parameter adjusts the send packet scheduling on timestamps and represents
711 the average delay between beginning of the transmitting descriptor processing
712 by the hardware and appearance of actual packet data on the wire. The value
713 should be provided in nanoseconds and is valid only if ``tx_pp`` parameter is
714 specified. The default value is zero.
716 - ``tx_vec_en`` parameter [int]
718 A nonzero value enables Tx vector on ConnectX-5, ConnectX-6, ConnectX-6 Dx
719 and BlueField NICs if the number of global Tx queues on the port is less than
720 ``txqs_max_vec``. The parameter is deprecated and ignored.
722 - ``rx_vec_en`` parameter [int]
724 A nonzero value enables Rx vector if the port is not configured in
725 multi-segment otherwise this parameter is ignored.
729 - ``vf_nl_en`` parameter [int]
731 A nonzero value enables Netlink requests from the VF to add/remove MAC
732 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
733 Otherwise the relevant configuration must be run with Linux iproute2 tools.
734 This is a prerequisite to receive this kind of traffic.
736 Enabled by default, valid only on VF devices ignored otherwise.
738 - ``l3_vxlan_en`` parameter [int]
740 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
741 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
742 parameter. This is a prerequisite to receive this kind of traffic.
746 - ``dv_xmeta_en`` parameter [int]
748 A nonzero value enables extensive flow metadata support if device is
749 capable and driver supports it. This can enable extensive support of
750 ``MARK`` and ``META`` item of ``rte_flow``. The newly introduced
751 ``SET_TAG`` and ``SET_META`` actions do not depend on ``dv_xmeta_en``.
753 There are some possible configurations, depending on parameter value:
755 - 0, this is default value, defines the legacy mode, the ``MARK`` and
756 ``META`` related actions and items operate only within NIC Tx and
757 NIC Rx steering domains, no ``MARK`` and ``META`` information crosses
758 the domain boundaries. The ``MARK`` item is 24 bits wide, the ``META``
759 item is 32 bits wide and match supported on egress only.
761 - 1, this engages extensive metadata mode, the ``MARK`` and ``META``
762 related actions and items operate within all supported steering domains,
763 including FDB, ``MARK`` and ``META`` information may cross the domain
764 boundaries. The ``MARK`` item is 24 bits wide, the ``META`` item width
765 depends on kernel and firmware configurations and might be 0, 16 or
766 32 bits. Within NIC Tx domain ``META`` data width is 32 bits for
767 compatibility, the actual width of data transferred to the FDB domain
768 depends on kernel configuration and may be vary. The actual supported
769 width can be retrieved in runtime by series of rte_flow_validate()
772 - 2, this engages extensive metadata mode, the ``MARK`` and ``META``
773 related actions and items operate within all supported steering domains,
774 including FDB, ``MARK`` and ``META`` information may cross the domain
775 boundaries. The ``META`` item is 32 bits wide, the ``MARK`` item width
776 depends on kernel and firmware configurations and might be 0, 16 or
777 24 bits. The actual supported width can be retrieved in runtime by
778 series of rte_flow_validate() trials.
780 +------+-----------+-----------+-------------+-------------+
781 | Mode | ``MARK`` | ``META`` | ``META`` Tx | FDB/Through |
782 +======+===========+===========+=============+=============+
783 | 0 | 24 bits | 32 bits | 32 bits | no |
784 +------+-----------+-----------+-------------+-------------+
785 | 1 | 24 bits | vary 0-32 | 32 bits | yes |
786 +------+-----------+-----------+-------------+-------------+
787 | 2 | vary 0-32 | 32 bits | 32 bits | yes |
788 +------+-----------+-----------+-------------+-------------+
790 If there is no E-Switch configuration the ``dv_xmeta_en`` parameter is
791 ignored and the device is configured to operate in legacy mode (0).
793 Disabled by default (set to 0).
795 The Direct Verbs/Rules (engaged with ``dv_flow_en`` = 1) supports all
796 of the extensive metadata features. The legacy Verbs supports FLAG and
797 MARK metadata actions over NIC Rx steering domain only.
799 - ``dv_flow_en`` parameter [int]
801 A nonzero value enables the DV flow steering assuming it is supported
802 by the driver (RDMA Core library version is rdma-core-24.0 or higher).
804 Enabled by default if supported.
806 - ``dv_esw_en`` parameter [int]
808 A nonzero value enables E-Switch using Direct Rules.
810 Enabled by default if supported.
812 - ``lacp_by_user`` parameter [int]
814 A nonzero value enables the control of LACP traffic by the user application.
815 When a bond exists in the driver, by default it should be managed by the
816 kernel and therefore LACP traffic should be steered to the kernel.
817 If this devarg is set to 1 it will allow the user to manage the bond by
818 itself and not steer LACP traffic to the kernel.
820 Disabled by default (set to 0).
822 - ``mr_ext_memseg_en`` parameter [int]
824 A nonzero value enables extending memseg when registering DMA memory. If
825 enabled, the number of entries in MR (Memory Region) lookup table on datapath
826 is minimized and it benefits performance. On the other hand, it worsens memory
827 utilization because registered memory is pinned by kernel driver. Even if a
828 page in the extended chunk is freed, that doesn't become reusable until the
829 entire memory is freed.
833 - ``representor`` parameter [list]
835 This parameter can be used to instantiate DPDK Ethernet devices from
836 existing port (or VF) representors configured on the device.
838 It is a standard parameter whose format is described in
839 :ref:`ethernet_device_standard_device_arguments`.
841 For instance, to probe port representors 0 through 2::
845 - ``max_dump_files_num`` parameter [int]
847 The maximum number of files per PMD entity that may be created for debug information.
848 The files will be created in /var/log directory or in current directory.
850 set to 128 by default.
852 - ``lro_timeout_usec`` parameter [int]
854 The maximum allowed duration of an LRO session, in micro-seconds.
855 PMD will set the nearest value supported by HW, which is not bigger than
856 the input ``lro_timeout_usec`` value.
857 If this parameter is not specified, by default PMD will set
858 the smallest value supported by HW.
860 - ``hp_buf_log_sz`` parameter [int]
862 The total data buffer size of a hairpin queue (logarithmic form), in bytes.
863 PMD will set the data buffer size to 2 ** ``hp_buf_log_sz``, both for RX & TX.
864 The capacity of the value is specified by the firmware and the initialization
865 will get a failure if it is out of scope.
866 The range of the value is from 11 to 19 right now, and the supported frame
867 size of a single packet for hairpin is from 512B to 128KB. It might change if
868 different firmware release is being used. By using a small value, it could
869 reduce memory consumption but not work with a large frame. If the value is
870 too large, the memory consumption will be high and some potential performance
871 degradation will be introduced.
872 By default, the PMD will set this value to 16, which means that 9KB jumbo
873 frames will be supported.
875 - ``reclaim_mem_mode`` parameter [int]
877 Cache some resources in flow destroy will help flow recreation more efficient.
878 While some systems may require the all the resources can be reclaimed after
880 The parameter ``reclaim_mem_mode`` provides the option for user to configure
881 if the resource cache is needed or not.
883 There are three options to choose:
885 - 0. It means the flow resources will be cached as usual. The resources will
886 be cached, helpful with flow insertion rate.
888 - 1. It will only enable the DPDK PMD level resources reclaim.
890 - 2. Both DPDK PMD level and rdma-core low level will be configured as
893 By default, the PMD will set this value to 0.
895 - ``sys_mem_en`` parameter [int]
897 A non-zero value enables the PMD memory management allocating memory
898 from system by default, without explicit rte memory flag.
900 By default, the PMD will set this value to 0.
902 - ``decap_en`` parameter [int]
904 Some devices do not support FCS (frame checksum) scattering for
905 tunnel-decapsulated packets.
906 If set to 0, this option forces the FCS feature and rejects tunnel
907 decapsulation in the flow engine for such devices.
909 By default, the PMD will set this value to 1.
911 .. _mlx5_firmware_config:
913 Firmware configuration
914 ~~~~~~~~~~~~~~~~~~~~~~
916 Firmware features can be configured as key/value pairs.
918 The command to set a value is::
920 mlxconfig -d <device> set <key>=<value>
922 The command to query a value is::
924 mlxconfig -d <device> query | grep <key>
926 The device name for the command ``mlxconfig`` can be either the PCI address,
927 or the mst device name found with::
931 Below are some firmware configurations listed.
937 value: 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
943 - maximum number of SR-IOV virtual functions::
947 - enable DevX (required by Direct Rules and other features)::
951 - aggressive CQE zipping::
955 - L3 VXLAN and VXLAN-GPE destination UDP port::
958 IP_OVER_VXLAN_PORT=<udp dport>
960 - enable VXLAN-GPE tunnel flow matching::
962 FLEX_PARSER_PROFILE_ENABLE=0
964 FLEX_PARSER_PROFILE_ENABLE=2
966 - enable IP-in-IP tunnel flow matching::
968 FLEX_PARSER_PROFILE_ENABLE=0
970 - enable MPLS flow matching::
972 FLEX_PARSER_PROFILE_ENABLE=1
974 - enable ICMP(code/type/identifier/sequence number) / ICMP6(code/type) fields matching::
976 FLEX_PARSER_PROFILE_ENABLE=2
978 - enable Geneve flow matching::
980 FLEX_PARSER_PROFILE_ENABLE=0
982 FLEX_PARSER_PROFILE_ENABLE=1
984 - enable GTP flow matching::
986 FLEX_PARSER_PROFILE_ENABLE=3
988 - enable eCPRI flow matching::
990 FLEX_PARSER_PROFILE_ENABLE=4
996 This driver relies on external libraries and kernel drivers for resources
997 allocations and initialization. The following dependencies are not part of
998 DPDK and must be installed separately:
1002 User space Verbs framework used by librte_net_mlx5. This library provides
1003 a generic interface between the kernel and low-level user space drivers
1006 It allows slow and privileged operations (context initialization, hardware
1007 resources allocations) to be managed by the kernel and fast operations to
1008 never leave user space.
1012 Low-level user space driver library for Mellanox
1013 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices, it is automatically loaded
1016 This library basically implements send/receive calls to the hardware
1019 - **Kernel modules**
1021 They provide the kernel-side Verbs API and low level device drivers that
1022 manage actual hardware initialization and resources sharing with user
1025 Unlike most other PMDs, these modules must remain loaded and bound to
1028 - mlx5_core: hardware driver managing Mellanox
1029 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices and related Ethernet kernel
1031 - mlx5_ib: InifiniBand device driver.
1032 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
1034 - **Firmware update**
1036 Mellanox OFED/EN releases include firmware updates for
1037 ConnectX-4/ConnectX-5/ConnectX-6/BlueField adapters.
1039 Because each release provides new features, these updates must be applied to
1040 match the kernel modules and libraries they come with.
1044 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
1050 Either RDMA Core library with a recent enough Linux kernel release
1051 (recommended) or Mellanox OFED/EN, which provides compatibility with older
1054 RDMA Core with Linux Kernel
1055 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1057 - Minimal kernel version : v4.14 or the most recent 4.14-rc (see `Linux installation documentation`_)
1058 - Minimal rdma-core version: v15+ commit 0c5f5765213a ("Merge pull request #227 from yishaih/tm")
1059 (see `RDMA Core installation documentation`_)
1060 - When building for i686 use:
1062 - rdma-core version 18.0 or above built with 32bit support.
1063 - Kernel version 4.14.41 or above.
1065 - Starting with rdma-core v21, static libraries can be built::
1068 CFLAGS=-fPIC cmake -DIN_PLACE=1 -DENABLE_STATIC=1 -GNinja ..
1071 .. _`Linux installation documentation`: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/plain/Documentation/admin-guide/README.rst
1072 .. _`RDMA Core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
1078 - Mellanox OFED version: **4.5** and above /
1079 Mellanox EN version: **4.5** and above
1082 - ConnectX-4: **12.21.1000** and above.
1083 - ConnectX-4 Lx: **14.21.1000** and above.
1084 - ConnectX-5: **16.21.1000** and above.
1085 - ConnectX-5 Ex: **16.21.1000** and above.
1086 - ConnectX-6: **20.27.0090** and above.
1087 - ConnectX-6 Dx: **22.27.0090** and above.
1088 - BlueField: **18.25.1010** and above.
1090 While these libraries and kernel modules are available on OpenFabrics
1091 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
1092 managers on most distributions, this PMD requires Ethernet extensions that
1093 may not be supported at the moment (this is a work in progress).
1096 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__ and
1098 <http://www.mellanox.com/page/products_dyn?product_family=27&mtag=linux>`__
1099 include the necessary support and should be used in the meantime. For DPDK,
1100 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
1101 required from that distribution.
1105 Several versions of Mellanox OFED/EN are available. Installing the version
1106 this DPDK release was developed and tested against is strongly
1107 recommended. Please check the `prerequisites`_.
1112 The following Mellanox device families are supported by the same mlx5 driver:
1122 Below are detailed device names:
1124 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX4111A-XCAT (1x10G)
1125 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX412A-XCAT (2x10G)
1126 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX4111A-ACAT (1x25G)
1127 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX412A-ACAT (2x25G)
1128 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX413A-BCAT (1x40G)
1129 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX4131A-BCAT (1x40G)
1130 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX415A-BCAT (1x40G)
1131 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX413A-GCAT (1x50G)
1132 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX4131A-GCAT (1x50G)
1133 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX414A-BCAT (2x50G)
1134 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-GCAT (1x50G)
1135 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-BCAT (2x50G)
1136 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-GCAT (2x50G)
1137 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-CCAT (1x100G)
1138 * Mellanox\ |reg| ConnectX\ |reg|-4 100G MCX416A-CCAT (2x100G)
1139 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4111A-XCAT (1x10G)
1140 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4121A-XCAT (2x10G)
1141 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4111A-ACAT (1x25G)
1142 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4121A-ACAT (2x25G)
1143 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 40G MCX4131A-BCAT (1x40G)
1144 * Mellanox\ |reg| ConnectX\ |reg|-5 100G MCX556A-ECAT (2x100G)
1145 * Mellanox\ |reg| ConnectX\ |reg|-5 Ex EN 100G MCX516A-CDAT (2x100G)
1146 * Mellanox\ |reg| ConnectX\ |reg|-6 200G MCX654106A-HCAT (2x200G)
1147 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 100G MCX623106AN-CDAT (2x100G)
1148 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 200G MCX623105AN-VDAT (1x200G)
1150 Quick Start Guide on OFED/EN
1151 ----------------------------
1153 1. Download latest Mellanox OFED/EN. For more info check the `prerequisites`_.
1156 2. Install the required libraries and kernel modules either by installing
1157 only the required set, or by installing the entire Mellanox OFED/EN::
1159 ./mlnxofedinstall --upstream-libs --dpdk
1161 3. Verify the firmware is the correct one::
1165 4. Verify all ports links are set to Ethernet::
1167 mlxconfig -d <mst device> query | grep LINK_TYPE
1171 Link types may have to be configured to Ethernet::
1173 mlxconfig -d <mst device> set LINK_TYPE_P1/2=1/2/3
1175 * LINK_TYPE_P1=<1|2|3> , 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
1177 For hypervisors, verify SR-IOV is enabled on the NIC::
1179 mlxconfig -d <mst device> query | grep SRIOV_EN
1182 If needed, configure SR-IOV::
1184 mlxconfig -d <mst device> set SRIOV_EN=1 NUM_OF_VFS=16
1185 mlxfwreset -d <mst device> reset
1187 5. Restart the driver::
1189 /etc/init.d/openibd restart
1193 service openibd restart
1195 If link type was changed, firmware must be reset as well::
1197 mlxfwreset -d <mst device> reset
1199 For hypervisors, after reset write the sysfs number of virtual functions
1202 To dynamically instantiate a given number of virtual functions (VFs)::
1204 echo [num_vfs] > /sys/class/infiniband/mlx5_0/device/sriov_numvfs
1206 6. Install DPDK and you are ready to go.
1207 See :doc:`compilation instructions <../linux_gsg/build_dpdk>`.
1209 Enable switchdev mode
1210 ---------------------
1212 Switchdev mode is a mode in E-Switch, that binds between representor and VF.
1213 Representor is a port in DPDK that is connected to a VF in such a way
1214 that assuming there are no offload flows, each packet that is sent from the VF
1215 will be received by the corresponding representor. While each packet that is
1216 sent to a representor will be received by the VF.
1217 This is very useful in case of SRIOV mode, where the first packet that is sent
1218 by the VF will be received by the DPDK application which will decide if this
1219 flow should be offloaded to the E-Switch. After offloading the flow packet
1220 that the VF that are matching the flow will not be received any more by
1221 the DPDK application.
1223 1. Enable SRIOV mode::
1225 mlxconfig -d <mst device> set SRIOV_EN=true
1227 2. Configure the max number of VFs::
1229 mlxconfig -d <mst device> set NUM_OF_VFS=<num of vfs>
1233 mlxfwreset -d <mst device> reset
1235 3. Configure the actual number of VFs::
1237 echo <num of vfs > /sys/class/net/<net device>/device/sriov_numvfs
1239 4. Unbind the device (can be rebind after the switchdev mode)::
1241 echo -n "<device pci address" > /sys/bus/pci/drivers/mlx5_core/unbind
1243 5. Enbale switchdev mode::
1245 echo switchdev > /sys/class/net/<net device>/compat/devlink/mode
1250 1. Configure aggressive CQE Zipping for maximum performance::
1252 mlxconfig -d <mst device> s CQE_COMPRESSION=1
1254 To set it back to the default CQE Zipping mode use::
1256 mlxconfig -d <mst device> s CQE_COMPRESSION=0
1258 2. In case of virtualization:
1260 - Make sure that hypervisor kernel is 3.16 or newer.
1261 - Configure boot with ``iommu=pt``.
1262 - Use 1G huge pages.
1263 - Make sure to allocate a VM on huge pages.
1264 - Make sure to set CPU pinning.
1266 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
1267 for better performance. For VMs, verify that the right CPU
1268 and NUMA node are pinned according to the above. Run::
1272 to identify the NUMA node to which the PCIe adapter is connected.
1274 4. If more than one adapter is used, and root complex capabilities allow
1275 to put both adapters on the same NUMA node without PCI bandwidth degradation,
1276 it is recommended to locate both adapters on the same NUMA node.
1277 This in order to forward packets from one to the other without
1278 NUMA performance penalty.
1280 5. Disable pause frames::
1282 ethtool -A <netdev> rx off tx off
1284 6. Verify IO non-posted prefetch is disabled by default. This can be checked
1285 via the BIOS configuration. Please contact you server provider for more
1286 information about the settings.
1290 On some machines, depends on the machine integrator, it is beneficial
1291 to set the PCI max read request parameter to 1K. This can be
1292 done in the following way:
1294 To query the read request size use::
1296 setpci -s <NIC PCI address> 68.w
1298 If the output is different than 3XXX, set it by::
1300 setpci -s <NIC PCI address> 68.w=3XXX
1302 The XXX can be different on different systems. Make sure to configure
1303 according to the setpci output.
1305 7. To minimize overhead of searching Memory Regions:
1307 - '--socket-mem' is recommended to pin memory by predictable amount.
1308 - Configure per-lcore cache when creating Mempools for packet buffer.
1309 - Refrain from dynamically allocating/freeing memory in run-time.
1311 .. _mlx5_offloads_support:
1313 Supported hardware offloads
1314 ---------------------------
1316 .. table:: Minimal SW/HW versions for queue offloads
1318 ============== ===== ===== ========= ===== ========== ==========
1319 Offload DPDK Linux rdma-core OFED firmware hardware
1320 ============== ===== ===== ========= ===== ========== ==========
1321 common base 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1322 checksums 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1323 Rx timestamp 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1324 TSO 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1325 LRO 19.08 N/A N/A 4.6-4 16.25.6406 ConnectX-5
1326 ============== ===== ===== ========= ===== ========== ==========
1328 .. table:: Minimal SW/HW versions for rte_flow offloads
1330 +-----------------------+-----------------+-----------------+
1331 | Offload | with E-Switch | with NIC |
1332 +=======================+=================+=================+
1333 | Count | | DPDK 19.05 | | DPDK 19.02 |
1334 | | | OFED 4.6 | | OFED 4.6 |
1335 | | | rdma-core 24 | | rdma-core 23 |
1336 | | | ConnectX-5 | | ConnectX-5 |
1337 +-----------------------+-----------------+-----------------+
1338 | Drop | | DPDK 19.05 | | DPDK 18.11 |
1339 | | | OFED 4.6 | | OFED 4.5 |
1340 | | | rdma-core 24 | | rdma-core 23 |
1341 | | | ConnectX-5 | | ConnectX-4 |
1342 +-----------------------+-----------------+-----------------+
1343 | Queue / RSS | | | | DPDK 18.11 |
1344 | | | N/A | | OFED 4.5 |
1345 | | | | | rdma-core 23 |
1346 | | | | | ConnectX-4 |
1347 +-----------------------+-----------------+-----------------+
1348 | Encapsulation | | DPDK 19.05 | | DPDK 19.02 |
1349 | (VXLAN / NVGRE / RAW) | | OFED 4.7-1 | | OFED 4.6 |
1350 | | | rdma-core 24 | | rdma-core 23 |
1351 | | | ConnectX-5 | | ConnectX-5 |
1352 +-----------------------+-----------------+-----------------+
1353 | Encapsulation | | DPDK 19.11 | | DPDK 19.11 |
1354 | GENEVE | | OFED 4.7-3 | | OFED 4.7-3 |
1355 | | | rdma-core 27 | | rdma-core 27 |
1356 | | | ConnectX-5 | | ConnectX-5 |
1357 +-----------------------+-----------------+-----------------+
1358 | | Header rewrite | | DPDK 19.05 | | DPDK 19.02 |
1359 | | (set_ipv4_src / | | OFED 4.7-1 | | OFED 4.7-1 |
1360 | | set_ipv4_dst / | | rdma-core 24 | | rdma-core 24 |
1361 | | set_ipv6_src / | | ConnectX-5 | | ConnectX-5 |
1362 | | set_ipv6_dst / | | | | |
1363 | | set_tp_src / | | | | |
1364 | | set_tp_dst / | | | | |
1365 | | dec_ttl / | | | | |
1366 | | set_ttl / | | | | |
1367 | | set_mac_src / | | | | |
1368 | | set_mac_dst) | | | | |
1369 +-----------------------+-----------------+-----------------+
1370 | | Header rewrite | | DPDK 20.02 | | DPDK 20.02 |
1371 | | (set_dscp) | | OFED 5.0 | | OFED 5.0 |
1372 | | | | rdma-core 24 | | rdma-core 24 |
1373 | | | | ConnectX-5 | | ConnectX-5 |
1374 +-----------------------+-----------------+-----------------+
1375 | Jump | | DPDK 19.05 | | DPDK 19.02 |
1376 | | | OFED 4.7-1 | | OFED 4.7-1 |
1377 | | | rdma-core 24 | | N/A |
1378 | | | ConnectX-5 | | ConnectX-5 |
1379 +-----------------------+-----------------+-----------------+
1380 | Mark / Flag | | DPDK 19.05 | | DPDK 18.11 |
1381 | | | OFED 4.6 | | OFED 4.5 |
1382 | | | rdma-core 24 | | rdma-core 23 |
1383 | | | ConnectX-5 | | ConnectX-4 |
1384 +-----------------------+-----------------+-----------------+
1385 | Port ID | | DPDK 19.05 | | N/A |
1386 | | | OFED 4.7-1 | | N/A |
1387 | | | rdma-core 24 | | N/A |
1388 | | | ConnectX-5 | | N/A |
1389 +-----------------------+-----------------+-----------------+
1390 | | VLAN | | DPDK 19.11 | | DPDK 19.11 |
1391 | | (of_pop_vlan / | | OFED 4.7-1 | | OFED 4.7-1 |
1392 | | of_push_vlan / | | ConnectX-5 | | ConnectX-5 |
1393 | | of_set_vlan_pcp / | | | | |
1394 | | of_set_vlan_vid) | | | | |
1395 +-----------------------+-----------------+-----------------+
1396 | Hairpin | | | | DPDK 19.11 |
1397 | | | N/A | | OFED 4.7-3 |
1398 | | | | | rdma-core 26 |
1399 | | | | | ConnectX-5 |
1400 +-----------------------+-----------------+-----------------+
1401 | Meta data | | DPDK 19.11 | | DPDK 19.11 |
1402 | | | OFED 4.7-3 | | OFED 4.7-3 |
1403 | | | rdma-core 26 | | rdma-core 26 |
1404 | | | ConnectX-5 | | ConnectX-5 |
1405 +-----------------------+-----------------+-----------------+
1406 | Metering | | DPDK 19.11 | | DPDK 19.11 |
1407 | | | OFED 4.7-3 | | OFED 4.7-3 |
1408 | | | rdma-core 26 | | rdma-core 26 |
1409 | | | ConnectX-5 | | ConnectX-5 |
1410 +-----------------------+-----------------+-----------------+
1411 | Sampling | | DPDK 20.11 | | DPDK 20.11 |
1412 | | | OFED 5.2 | | OFED 5.2 |
1413 | | | rdma-core 32 | | rdma-core 32 |
1414 | | | ConnectX-5 | | ConnectX-5 |
1415 +-----------------------+-----------------+-----------------+
1420 MARK and META items are interrelated with datapath - they might move from/to
1421 the applications in mbuf fields. Hence, zero value for these items has the
1422 special meaning - it means "no metadata are provided", not zero values are
1423 treated by applications and PMD as valid ones.
1425 Moreover in the flow engine domain the value zero is acceptable to match and
1426 set, and we should allow to specify zero values as rte_flow parameters for the
1427 META and MARK items and actions. In the same time zero mask has no meaning and
1428 should be rejected on validation stage.
1433 Flows are not cached in the driver.
1434 When stopping a device port, all the flows created on this port from the
1435 application will be flushed automatically in the background.
1436 After stopping the device port, all flows on this port become invalid and
1437 not represented in the system.
1438 All references to these flows held by the application should be discarded
1439 directly but neither destroyed nor flushed.
1441 The application should re-create the flows as required after the port restart.
1446 Compared to librte_net_mlx4 that implements a single RSS configuration per
1447 port, librte_net_mlx5 supports per-protocol RSS configuration.
1449 Since ``testpmd`` defaults to IP RSS mode and there is currently no
1450 command-line parameter to enable additional protocols (UDP and TCP as well
1451 as IP), the following commands must be entered from its CLI to get the same
1452 behavior as librte_net_mlx4::
1455 > port config all rss all
1461 This section demonstrates how to launch **testpmd** with Mellanox
1462 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices managed by librte_net_mlx5.
1464 #. Load the kernel modules::
1466 modprobe -a ib_uverbs mlx5_core mlx5_ib
1468 Alternatively if MLNX_OFED/MLNX_EN is fully installed, the following script
1471 /etc/init.d/openibd restart
1475 User space I/O kernel modules (uio and igb_uio) are not used and do
1476 not have to be loaded.
1478 #. Make sure Ethernet interfaces are in working order and linked to kernel
1479 verbs. Related sysfs entries should be present::
1481 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
1490 #. Optionally, retrieve their PCI bus addresses for whitelisting::
1493 for intf in eth2 eth3 eth4 eth5;
1495 (cd "/sys/class/net/${intf}/device/" && pwd -P);
1498 sed -n 's,.*/\(.*\),-w \1,p'
1507 #. Request huge pages::
1509 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
1511 #. Start testpmd with basic parameters::
1513 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
1518 EAL: PCI device 0000:05:00.0 on NUMA socket 0
1519 EAL: probe driver: 15b3:1013 librte_net_mlx5
1520 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
1521 PMD: librte_net_mlx5: 1 port(s) detected
1522 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
1523 EAL: PCI device 0000:05:00.1 on NUMA socket 0
1524 EAL: probe driver: 15b3:1013 librte_net_mlx5
1525 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
1526 PMD: librte_net_mlx5: 1 port(s) detected
1527 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
1528 EAL: PCI device 0000:06:00.0 on NUMA socket 0
1529 EAL: probe driver: 15b3:1013 librte_net_mlx5
1530 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
1531 PMD: librte_net_mlx5: 1 port(s) detected
1532 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
1533 EAL: PCI device 0000:06:00.1 on NUMA socket 0
1534 EAL: probe driver: 15b3:1013 librte_net_mlx5
1535 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
1536 PMD: librte_net_mlx5: 1 port(s) detected
1537 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
1538 Interactive-mode selected
1539 Configuring Port 0 (socket 0)
1540 PMD: librte_net_mlx5: 0x8cba80: TX queues number update: 0 -> 2
1541 PMD: librte_net_mlx5: 0x8cba80: RX queues number update: 0 -> 2
1542 Port 0: E4:1D:2D:E7:0C:FE
1543 Configuring Port 1 (socket 0)
1544 PMD: librte_net_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
1545 PMD: librte_net_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
1546 Port 1: E4:1D:2D:E7:0C:FF
1547 Configuring Port 2 (socket 0)
1548 PMD: librte_net_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
1549 PMD: librte_net_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
1550 Port 2: E4:1D:2D:E7:0C:FA
1551 Configuring Port 3 (socket 0)
1552 PMD: librte_net_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
1553 PMD: librte_net_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
1554 Port 3: E4:1D:2D:E7:0C:FB
1555 Checking link statuses...
1556 Port 0 Link Up - speed 40000 Mbps - full-duplex
1557 Port 1 Link Up - speed 40000 Mbps - full-duplex
1558 Port 2 Link Up - speed 10000 Mbps - full-duplex
1559 Port 3 Link Up - speed 10000 Mbps - full-duplex
1566 This section demonstrates how to dump flows. Currently, it's possible to dump
1567 all flows with assistance of external tools.
1569 #. 2 ways to get flow raw file:
1571 - Using testpmd CLI:
1573 .. code-block:: console
1575 testpmd> flow dump <port> <output_file>
1577 - call rte_flow_dev_dump api:
1579 .. code-block:: console
1581 rte_flow_dev_dump(port, file, NULL);
1583 #. Dump human-readable flows from raw file:
1585 Get flow parsing tool from: https://github.com/Mellanox/mlx_steering_dump
1587 .. code-block:: console
1589 mlx_steering_dump.py -f <output_file>