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**, **Mellanox
13 ConnectX-6 Lx**, **Mellanox BlueField** and **Mellanox BlueField-2** families
14 of 10/25/40/50/100/200 Gb/s adapters as well as their virtual functions (VF)
17 Information and documentation about these adapters can be found on the
18 `Mellanox website <http://www.mellanox.com>`__. Help is also provided by the
19 `Mellanox community <http://community.mellanox.com/welcome>`__.
21 There is also a `section dedicated to this poll mode driver
22 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`__.
28 Besides its dependency on libibverbs (that implies libmlx5 and associated
29 kernel support), librte_net_mlx5 relies heavily on system calls for control
30 operations such as querying/updating the MTU and flow control parameters.
32 For security reasons and robustness, this driver only deals with virtual
33 memory addresses. The way resources allocations are handled by the kernel,
34 combined with hardware specifications that allow to handle virtual memory
35 addresses directly, ensure that DPDK applications cannot access random
36 physical memory (or memory that does not belong to the current process).
38 This capability allows the PMD to coexist with kernel network interfaces
39 which remain functional, although they stop receiving unicast packets as
40 long as they share the same MAC address.
41 This means legacy linux control tools (for example: ethtool, ifconfig and
42 more) can operate on the same network interfaces that owned by the DPDK
45 The PMD can use libibverbs and libmlx5 to access the device firmware
46 or directly the hardware components.
47 There are different levels of objects and bypassing abilities
48 to get the best performances:
50 - Verbs is a complete high-level generic API
51 - Direct Verbs is a device-specific API
52 - DevX allows to access firmware objects
53 - Direct Rules manages flow steering at low-level hardware layer
55 Enabling librte_net_mlx5 causes DPDK applications to be linked against
61 - Multi arch support: x86_64, POWER8, ARMv8, i686.
62 - Multiple TX and RX queues.
63 - Support for scattered TX frames.
64 - Advanced support for scattered Rx frames with tunable buffer attributes.
65 - IPv4, IPv6, TCPv4, TCPv6, UDPv4 and UDPv6 RSS on any number of queues.
66 - RSS using different combinations of fields: L3 only, L4 only or both,
67 and source only, destination only or both.
68 - Several RSS hash keys, one for each flow type.
69 - Default RSS operation with no hash key specification.
70 - Configurable RETA table.
71 - Link flow control (pause frame).
72 - Support for multiple MAC addresses.
76 - RX CRC stripping configuration.
77 - TX mbuf fast free offload.
78 - Promiscuous mode on PF and VF.
79 - Multicast promiscuous mode on PF and VF.
80 - Hardware checksum offloads.
81 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
83 - Flow API, including :ref:`flow_isolated_mode`.
85 - KVM and VMware ESX SR-IOV modes are supported.
86 - RSS hash result is supported.
87 - Hardware TSO for generic IP or UDP tunnel, including VXLAN and GRE.
88 - Hardware checksum Tx offload for generic IP or UDP tunnel, including VXLAN and GRE.
90 - Statistics query including Basic, Extended and per queue.
92 - Tunnel types: VXLAN, L3 VXLAN, VXLAN-GPE, GRE, MPLSoGRE, MPLSoUDP, IP-in-IP, Geneve, GTP.
93 - Tunnel HW offloads: packet type, inner/outer RSS, IP and UDP checksum verification.
94 - NIC HW offloads: encapsulation (vxlan, gre, mplsoudp, mplsogre), NAT, routing, TTL
95 increment/decrement, count, drop, mark. For details please see :ref:`mlx5_offloads_support`.
96 - Flow insertion rate of more then million flows per second, when using Direct Rules.
97 - Support for multiple rte_flow groups.
98 - Per packet no-inline hint flag to disable packet data copying into Tx descriptors.
101 - Multiple-thread flow insertion.
102 - Matching on GTP extension header with raw encap/decap action.
103 - Matching on Geneve TLV option header with raw encap/decap action.
104 - RSS support in sample action.
105 - E-Switch mirroring and jump.
106 - E-Switch mirroring and modify.
107 - 21844 flow priorities for ingress or egress flow groups greater than 0 and for any transfer
109 - Flow metering, including meter policy API.
110 - Flow integrity offload API.
117 On Windows, the features are limited:
119 - Promiscuous mode is not supported
120 - The following rules are supported:
122 - IPv4/UDP with CVLAN filtering
123 - Unicast MAC filtering
125 - For secondary process:
127 - Forked secondary process not supported.
128 - External memory unregistered in EAL memseg list cannot be used for DMA
129 unless such memory has been registered by ``mlx5_mr_update_ext_mp()`` in
130 primary process and remapped to the same virtual address in secondary
131 process. If the external memory is registered by primary process but has
132 different virtual address in secondary process, unexpected error may happen.
134 - When using Verbs flow engine (``dv_flow_en`` = 0), flow pattern without any
135 specific VLAN will match for VLAN packets as well:
137 When VLAN spec is not specified in the pattern, the matching rule will be created with VLAN as a wild card.
138 Meaning, the flow rule::
140 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
142 Will only match vlan packets with vid=3. and the flow rule::
144 flow create 0 ingress pattern eth / ipv4 / end ...
146 Will match any ipv4 packet (VLAN included).
148 - When using Verbs flow engine (``dv_flow_en`` = 0), multi-tagged(QinQ) match is not supported.
150 - When using DV flow engine (``dv_flow_en`` = 1), flow pattern with any VLAN specification will match only single-tagged packets unless the ETH item ``type`` field is 0x88A8 or the VLAN item ``has_more_vlan`` field is 1.
153 flow create 0 ingress pattern eth / ipv4 / end ...
155 Will match any ipv4 packet.
158 flow create 0 ingress pattern eth / vlan / end ...
159 flow create 0 ingress pattern eth has_vlan is 1 / end ...
160 flow create 0 ingress pattern eth type is 0x8100 / end ...
162 Will match single-tagged packets only, with any VLAN ID value.
165 flow create 0 ingress pattern eth type is 0x88A8 / end ...
166 flow create 0 ingress pattern eth / vlan has_more_vlan is 1 / end ...
168 Will match multi-tagged packets only, with any VLAN ID value.
170 - A flow pattern with 2 sequential VLAN items is not supported.
172 - VLAN pop offload command:
174 - Flow rules having a VLAN pop offload command as one of their actions and
175 are lacking a match on VLAN as one of their items are not supported.
176 - The command is not supported on egress traffic in NIC mode.
178 - VLAN push offload is not supported on ingress traffic in NIC mode.
180 - VLAN set PCP offload is not supported on existing headers.
182 - A multi segment packet must have not more segments than reported by dev_infos_get()
183 in tx_desc_lim.nb_seg_max field. This value depends on maximal supported Tx descriptor
184 size and ``txq_inline_min`` settings and may be from 2 (worst case forced by maximal
185 inline settings) to 58.
187 - Flows with a VXLAN Network Identifier equal (or ends to be equal)
188 to 0 are not supported.
190 - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with MPLSoGRE and MPLSoUDP.
192 - Match on Geneve header supports the following fields only:
199 - Match on Geneve TLV option is supported on the following fields:
206 Only one Class/Type/Length Geneve TLV option is supported per shared device.
207 Class/Type/Length fields must be specified as well as masks.
208 Class/Type/Length specified masks must be full.
209 Matching Geneve TLV option without specifying data is not supported.
210 Matching Geneve TLV option with ``data & mask == 0`` is not supported.
212 - VF: flow rules created on VF devices can only match traffic targeted at the
213 configured MAC addresses (see ``rte_eth_dev_mac_addr_add()``).
215 - Match on GTP tunnel header item supports the following fields only:
217 - v_pt_rsv_flags: E flag, S flag, PN flag
221 - Match on GTP extension header only for GTP PDU session container (next
222 extension header type = 0x85).
223 - Match on GTP extension header is not supported in group 0.
225 - No Tx metadata go to the E-Switch steering domain for the Flow group 0.
226 The flows within group 0 and set metadata action are rejected by hardware.
230 MAC addresses not already present in the bridge table of the associated
231 kernel network device will be added and cleaned up by the PMD when closing
232 the device. In case of ungraceful program termination, some entries may
233 remain present and should be removed manually by other means.
235 - Buffer split offload is supported with regular Rx burst routine only,
236 no MPRQ feature or vectorized code can be engaged.
238 - When Multi-Packet Rx queue is configured (``mprq_en``), a Rx packet can be
239 externally attached to a user-provided mbuf with having EXT_ATTACHED_MBUF in
240 ol_flags. As the mempool for the external buffer is managed by PMD, all the
241 Rx mbufs must be freed before the device is closed. Otherwise, the mempool of
242 the external buffers will be freed by PMD and the application which still
243 holds the external buffers may be corrupted.
245 - If Multi-Packet Rx queue is configured (``mprq_en``) and Rx CQE compression is
246 enabled (``rxq_cqe_comp_en``) at the same time, RSS hash result is not fully
247 supported. Some Rx packets may not have PKT_RX_RSS_HASH.
249 - IPv6 Multicast messages are not supported on VM, while promiscuous mode
250 and allmulticast mode are both set to off.
251 To receive IPv6 Multicast messages on VM, explicitly set the relevant
252 MAC address using rte_eth_dev_mac_addr_add() API.
254 - To support a mixed traffic pattern (some buffers from local host memory, some
255 buffers from other devices) with high bandwidth, a mbuf flag is used.
257 An application hints the PMD whether or not it should try to inline the
258 given mbuf data buffer. PMD should do the best effort to act upon this request.
260 The hint flag ``RTE_PMD_MLX5_FINE_GRANULARITY_INLINE`` is dynamic,
261 registered by application with rte_mbuf_dynflag_register(). This flag is
262 purely driver-specific and declared in PMD specific header ``rte_pmd_mlx5.h``,
263 which is intended to be used by the application.
265 To query the supported specific flags in runtime,
266 the function ``rte_pmd_mlx5_get_dyn_flag_names`` returns the array of
267 currently (over present hardware and configuration) supported specific flags.
268 The "not inline hint" feature operating flow is the following one:
271 - probe the devices, ports are created
272 - query the port capabilities
273 - if port supporting the feature is found
274 - register dynamic flag ``RTE_PMD_MLX5_FINE_GRANULARITY_INLINE``
275 - application starts the ports
276 - on ``dev_start()`` PMD checks whether the feature flag is registered and
277 enables the feature support in datapath
278 - application might set the registered flag bit in ``ol_flags`` field
279 of mbuf being sent and PMD will handle ones appropriately.
281 - The amount of descriptors in Tx queue may be limited by data inline settings.
282 Inline data require the more descriptor building blocks and overall block
283 amount may exceed the hardware supported limits. The application should
284 reduce the requested Tx size or adjust data inline settings with
285 ``txq_inline_max`` and ``txq_inline_mpw`` devargs keys.
287 - To provide the packet send scheduling on mbuf timestamps the ``tx_pp``
288 parameter should be specified.
289 When PMD sees the RTE_MBUF_DYNFLAG_TX_TIMESTAMP_NAME set on the packet
290 being sent it tries to synchronize the time of packet appearing on
291 the wire with the specified packet timestamp. It the specified one
292 is in the past it should be ignored, if one is in the distant future
293 it should be capped with some reasonable value (in range of seconds).
294 These specific cases ("too late" and "distant future") can be optionally
295 reported via device xstats to assist applications to detect the
296 time-related problems.
298 The timestamp upper "too-distant-future" limit
299 at the moment of invoking the Tx burst routine
300 can be estimated as ``tx_pp`` option (in nanoseconds) multiplied by 2^23.
301 Please note, for the testpmd txonly mode,
302 the limit is deduced from the expression::
304 (n_tx_descriptors / burst_size + 1) * inter_burst_gap
306 There is no any packet reordering according timestamps is supposed,
307 neither within packet burst, nor between packets, it is an entirely
308 application responsibility to generate packets and its timestamps
309 in desired order. The timestamps can be put only in the first packet
310 in the burst providing the entire burst scheduling.
312 - E-Switch decapsulation Flow:
314 - can be applied to PF port only.
315 - must specify VF port action (packet redirection from PF to VF).
316 - optionally may specify tunnel inner source and destination MAC addresses.
318 - E-Switch encapsulation Flow:
320 - can be applied to VF ports only.
321 - must specify PF port action (packet redirection from VF to PF).
325 - The input buffer, used as outer header, is not validated.
329 - The decapsulation is always done up to the outermost tunnel detected by the HW.
330 - The input buffer, providing the removal size, is not validated.
331 - The buffer size must match the length of the headers to be removed.
333 - ICMP(code/type/identifier/sequence number) / ICMP6(code/type) matching, IP-in-IP and MPLS flow matching are all
334 mutually exclusive features which cannot be supported together
335 (see :ref:`mlx5_firmware_config`).
339 - Requires DevX and DV flow to be enabled.
340 - KEEP_CRC offload cannot be supported with LRO.
341 - The first mbuf length, without head-room, must be big enough to include the
343 - Rx queue with LRO offload enabled, receiving a non-LRO packet, can forward
344 it with size limited to max LRO size, not to max RX packet length.
345 - LRO can be used with outer header of TCP packets of the standard format:
346 eth (with or without vlan) / ipv4 or ipv6 / tcp / payload
348 Other TCP packets (e.g. with MPLS label) received on Rx queue with LRO enabled, will be received with bad checksum.
349 - LRO packet aggregation is performed by HW only for packet size larger than
350 ``lro_min_mss_size``. This value is reported on device start, when debug
355 - ``DEV_RX_OFFLOAD_KEEP_CRC`` cannot be supported with decapsulation
356 for some NICs (such as ConnectX-6 Dx, ConnectX-6 Lx, and BlueField-2).
357 The capability bit ``scatter_fcs_w_decap_disable`` shows NIC support.
361 - fast free offload assumes the all mbufs being sent are originated from the
362 same memory pool and there is no any extra references to the mbufs (the
363 reference counter for each mbuf is equal 1 on tx_burst call). The latter
364 means there should be no any externally attached buffers in mbufs. It is
365 an application responsibility to provide the correct mbufs if the fast
366 free offload is engaged. The mlx5 PMD implicitly produces the mbufs with
367 externally attached buffers if MPRQ option is enabled, hence, the fast
368 free offload is neither supported nor advertised if there is MPRQ enabled.
372 - Supports ``RTE_FLOW_ACTION_TYPE_SAMPLE`` action only within NIC Rx and
373 E-Switch steering domain.
374 - For E-Switch Sampling flow with sample ratio > 1, additional actions are not
375 supported in the sample actions list.
376 - For ConnectX-5, the ``RTE_FLOW_ACTION_TYPE_SAMPLE`` is typically used as
377 first action in the E-Switch egress flow if with header modify or
378 encapsulation actions.
379 - For NIC Rx flow, supports ``MARK``, ``COUNT``, ``QUEUE``, ``RSS`` in the
381 - For E-Switch mirroring flow, supports ``RAW ENCAP``, ``Port ID``,
382 ``VXLAN ENCAP``, ``NVGRE ENCAP`` in the sample actions list.
386 - Supports the 'set' operation only for ``RTE_FLOW_ACTION_TYPE_MODIFY_FIELD`` action.
387 - Modification of an arbitrary place in a packet via the special ``RTE_FLOW_FIELD_START`` Field ID is not supported.
388 - Modification of the 802.1Q Tag, VXLAN Network or GENEVE Network ID's is not supported.
389 - Encapsulation levels are not supported, can modify outermost header fields only.
390 - Offsets must be 32-bits aligned, cannot skip past the boundary of a field.
392 - IPv6 header item 'proto' field, indicating the next header protocol, should
393 not be set as extension header.
394 In case the next header is an extension header, it should not be specified in
395 IPv6 header item 'proto' field.
396 The last extension header item 'next header' field can specify the following
397 header protocol type.
401 - Hairpin between two ports could only manual binding and explicit Tx flow mode. For single port hairpin, all the combinations of auto/manual binding and explicit/implicit Tx flow mode could be supported.
402 - Hairpin in switchdev SR-IOV mode is not supported till now.
406 - All the meter colors with drop action will be counted only by the global drop statistics.
407 - Green color is not supported with drop action.
408 - Yellow detection is not supported.
409 - Red color must be with drop action.
410 - Meter statistics are supported only for drop case.
411 - Meter yellow color detection is not supported.
412 - A meter action created with pre-defined policy must be the last action in the flow except single case where the policy actions are:
413 - green: NULL or END.
414 - yellow: NULL or END.
416 - The only supported meter policy actions:
417 - green: QUEUE, RSS, PORT_ID, JUMP, MARK and SET_TAG.
418 - yellow: must be empty.
420 - meter profile packet mode is supported.
424 - Integrity offload is enabled for **ConnectX-6** family.
425 - Verification bits provided by the hardware are ``l3_ok``, ``ipv4_csum_ok``, ``l4_ok``, ``l4_csum_ok``.
426 - ``level`` value 0 references outer headers.
427 - Multiple integrity items not supported in a single flow rule.
428 - Flow rule items supplied by application must explicitly specify network headers referred by integrity item.
429 For example, if integrity item mask sets ``l4_ok`` or ``l4_csum_ok`` bits, reference to L4 network header,
430 TCP or UDP, must be in the rule pattern as well::
432 flow create 0 ingress pattern integrity level is 0 value mask l3_ok value spec l3_ok / eth / ipv6 / end …
434 flow create 0 ingress pattern integrity level is 0 value mask l4_ok value spec 0 / eth / ipv4 proto is udp / end …
439 MLX5 supports various methods to report statistics:
441 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.
443 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.
445 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.
453 The ibverbs libraries can be linked with this PMD in a number of ways,
454 configured by the ``ibverbs_link`` build option:
456 - ``shared`` (default): the PMD depends on some .so files.
458 - ``dlopen``: Split the dependencies glue in a separate library
459 loaded when needed by dlopen.
460 It make dependencies on libibverbs and libmlx4 optional,
461 and has no performance impact.
463 - ``static``: Embed static flavor of the dependencies libibverbs and libmlx4
464 in the PMD shared library or the executable static binary.
466 Environment variables
467 ~~~~~~~~~~~~~~~~~~~~~
471 A list of directories in which to search for the rdma-core "glue" plug-in,
472 separated by colons or semi-colons.
474 - ``MLX5_SHUT_UP_BF``
476 Configures HW Tx doorbell register as IO-mapped.
478 By default, the HW Tx doorbell is configured as a write-combining register.
479 The register would be flushed to HW usually when the write-combining buffer
480 becomes full, but it depends on CPU design.
482 Except for vectorized Tx burst routines, a write memory barrier is enforced
483 after updating the register so that the update can be immediately visible to
486 When vectorized Tx burst is called, the barrier is set only if the burst size
487 is not aligned to MLX5_VPMD_TX_MAX_BURST. However, setting this environmental
488 variable will bring better latency even though the maximum throughput can
491 Run-time configuration
492 ~~~~~~~~~~~~~~~~~~~~~~
494 - librte_net_mlx5 brings kernel network interfaces up during initialization
495 because it is affected by their state. Forcing them down prevents packets
498 - **ethtool** operations on related kernel interfaces also affect the PMD.
503 In order to run as a non-root user,
504 some capabilities must be granted to the application::
506 setcap cap_sys_admin,cap_net_admin,cap_net_raw,cap_ipc_lock+ep <dpdk-app>
508 Below are the reasons of the need for each capability:
511 When using physical addresses (PA mode), with Linux >= 4.0,
512 for access to ``/proc/self/pagemap``.
515 For device configuration.
518 For raw ethernet queue allocation through kernel driver.
521 For DMA memory pinning.
526 - ``rxq_cqe_comp_en`` parameter [int]
528 A nonzero value enables the compression of CQE on RX side. This feature
529 allows to save PCI bandwidth and improve performance. Enabled by default.
530 Different compression formats are supported in order to achieve the best
531 performance for different traffic patterns. Default format depends on
532 Multi-Packet Rx queue configuration: Hash RSS format is used in case
533 MPRQ is disabled, Checksum format is used in case MPRQ is enabled.
535 Specifying 2 as a ``rxq_cqe_comp_en`` value selects Flow Tag format for
536 better compression rate in case of RTE Flow Mark traffic.
537 Specifying 3 as a ``rxq_cqe_comp_en`` value selects Checksum format.
538 Specifying 4 as a ``rxq_cqe_comp_en`` value selects L3/L4 Header format for
539 better compression rate in case of mixed TCP/UDP and IPv4/IPv6 traffic.
540 CQE compression format selection requires DevX to be enabled. If there is
541 no DevX enabled/supported the value is reset to 1 by default.
545 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
546 ConnectX-6 Lx, BlueField and BlueField-2.
547 - POWER9 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
548 ConnectX-6 Lx, BlueField and BlueField-2.
550 - ``rxq_pkt_pad_en`` parameter [int]
552 A nonzero value enables padding Rx packet to the size of cacheline on PCI
553 transaction. This feature would waste PCI bandwidth but could improve
554 performance by avoiding partial cacheline write which may cause costly
555 read-modify-copy in memory transaction on some architectures. Disabled by
560 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
561 ConnectX-6 Lx, BlueField and BlueField-2.
562 - POWER8 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
563 ConnectX-6 Lx, BlueField and BlueField-2.
565 - ``mprq_en`` parameter [int]
567 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
568 configured as Multi-Packet RQ if the total number of Rx queues is
569 ``rxqs_min_mprq`` or more. Disabled by default.
571 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
572 by posting a single large buffer for multiple packets. Instead of posting a
573 buffers per a packet, one large buffer is posted in order to receive multiple
574 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
575 and each stride receives one packet. MPRQ can improve throughput for
576 small-packet traffic.
578 When MPRQ is enabled, max_rx_pkt_len can be larger than the size of
579 user-provided mbuf even if DEV_RX_OFFLOAD_SCATTER isn't enabled. PMD will
580 configure large stride size enough to accommodate max_rx_pkt_len as long as
581 device allows. Note that this can waste system memory compared to enabling Rx
582 scatter and multi-segment packet.
584 - ``mprq_log_stride_num`` parameter [int]
586 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
587 strides can reduce PCIe traffic further. If configured value is not in the
588 range of device capability, the default value will be set with a warning
589 message. The default value is 4 which is 16 strides per a buffer, valid only
590 if ``mprq_en`` is set.
592 The size of Rx queue should be bigger than the number of strides.
594 - ``mprq_log_stride_size`` parameter [int]
596 Log 2 of the size of a stride for Multi-Packet Rx queue. Configuring a smaller
597 stride size can save some memory and reduce probability of a depletion of all
598 available strides due to unreleased packets by an application. If configured
599 value is not in the range of device capability, the default value will be set
600 with a warning message. The default value is 11 which is 2048 bytes per a
601 stride, valid only if ``mprq_en`` is set. With ``mprq_log_stride_size`` set
602 it is possible for a packet to span across multiple strides. This mode allows
603 support of jumbo frames (9K) with MPRQ. The memcopy of some packets (or part
604 of a packet if Rx scatter is configured) may be required in case there is no
605 space left for a head room at the end of a stride which incurs some
608 - ``mprq_max_memcpy_len`` parameter [int]
610 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
611 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
612 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
613 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
614 A mempool for external buffers will be allocated and managed by PMD. If Rx
615 packet is externally attached, ol_flags field of the mbuf will have
616 EXT_ATTACHED_MBUF and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
617 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
619 - ``rxqs_min_mprq`` parameter [int]
621 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
622 greater or equal to this value. The default value is 12, valid only if
625 - ``txq_inline`` parameter [int]
627 Amount of data to be inlined during TX operations. This parameter is
628 deprecated and converted to the new parameter ``txq_inline_max`` providing
629 partial compatibility.
631 - ``txqs_min_inline`` parameter [int]
633 Enable inline data send only when the number of TX queues is greater or equal
636 This option should be used in combination with ``txq_inline_max`` and
637 ``txq_inline_mpw`` below and does not affect ``txq_inline_min`` settings above.
639 If this option is not specified the default value 16 is used for BlueField
640 and 8 for other platforms
642 The data inlining consumes the CPU cycles, so this option is intended to
643 auto enable inline data if we have enough Tx queues, which means we have
644 enough CPU cores and PCI bandwidth is getting more critical and CPU
645 is not supposed to be bottleneck anymore.
647 The copying data into WQE improves latency and can improve PPS performance
648 when PCI back pressure is detected and may be useful for scenarios involving
649 heavy traffic on many queues.
651 Because additional software logic is necessary to handle this mode, this
652 option should be used with care, as it may lower performance when back
653 pressure is not expected.
655 If inline data are enabled it may affect the maximal size of Tx queue in
656 descriptors because the inline data increase the descriptor size and
657 queue size limits supported by hardware may be exceeded.
659 - ``txq_inline_min`` parameter [int]
661 Minimal amount of data to be inlined into WQE during Tx operations. NICs
662 may require this minimal data amount to operate correctly. The exact value
663 may depend on NIC operation mode, requested offloads, etc. It is strongly
664 recommended to omit this parameter and use the default values. Anyway,
665 applications using this parameter should take into consideration that
666 specifying an inconsistent value may prevent the NIC from sending packets.
668 If ``txq_inline_min`` key is present the specified value (may be aligned
669 by the driver in order not to exceed the limits and provide better descriptor
670 space utilization) will be used by the driver and it is guaranteed that
671 requested amount of data bytes are inlined into the WQE beside other inline
672 settings. This key also may update ``txq_inline_max`` value (default
673 or specified explicitly in devargs) to reserve the space for inline data.
675 If ``txq_inline_min`` key is not present, the value may be queried by the
676 driver from the NIC via DevX if this feature is available. If there is no DevX
677 enabled/supported the value 18 (supposing L2 header including VLAN) is set
678 for ConnectX-4 and ConnectX-4 Lx, and 0 is set by default for ConnectX-5
679 and newer NICs. If packet is shorter the ``txq_inline_min`` value, the entire
682 For ConnectX-4 NIC, driver does not allow specifying value below 18
683 (minimal L2 header, including VLAN), error will be raised.
685 For ConnectX-4 Lx NIC, it is allowed to specify values below 18, but
686 it is not recommended and may prevent NIC from sending packets over
689 Please, note, this minimal data inlining disengages eMPW feature (Enhanced
690 Multi-Packet Write), because last one does not support partial packet inlining.
691 This is not very critical due to minimal data inlining is mostly required
692 by ConnectX-4 and ConnectX-4 Lx, these NICs do not support eMPW feature.
694 - ``txq_inline_max`` parameter [int]
696 Specifies the maximal packet length to be completely inlined into WQE
697 Ethernet Segment for ordinary SEND method. If packet is larger than specified
698 value, the packet data won't be copied by the driver at all, data buffer
699 is addressed with a pointer. If packet length is less or equal all packet
700 data will be copied into WQE. This may improve PCI bandwidth utilization for
701 short packets significantly but requires the extra CPU cycles.
703 The data inline feature is controlled by number of Tx queues, if number of Tx
704 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
705 is engaged, if there are not enough Tx queues (which means not enough CPU cores
706 and CPU resources are scarce), data inline is not performed by the driver.
707 Assigning ``txqs_min_inline`` with zero always enables the data inline.
709 The default ``txq_inline_max`` value is 290. The specified value may be adjusted
710 by the driver in order not to exceed the limit (930 bytes) and to provide better
711 WQE space filling without gaps, the adjustment is reflected in the debug log.
712 Also, the default value (290) may be decreased in run-time if the large transmit
713 queue size is requested and hardware does not support enough descriptor
714 amount, in this case warning is emitted. If ``txq_inline_max`` key is
715 specified and requested inline settings can not be satisfied then error
718 - ``txq_inline_mpw`` parameter [int]
720 Specifies the maximal packet length to be completely inlined into WQE for
721 Enhanced MPW method. If packet is large the specified value, the packet data
722 won't be copied, and data buffer is addressed with pointer. If packet length
723 is less or equal, all packet data will be copied into WQE. This may improve PCI
724 bandwidth utilization for short packets significantly but requires the extra
727 The data inline feature is controlled by number of TX queues, if number of Tx
728 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
729 is engaged, if there are not enough Tx queues (which means not enough CPU cores
730 and CPU resources are scarce), data inline is not performed by the driver.
731 Assigning ``txqs_min_inline`` with zero always enables the data inline.
733 The default ``txq_inline_mpw`` value is 268. The specified value may be adjusted
734 by the driver in order not to exceed the limit (930 bytes) and to provide better
735 WQE space filling without gaps, the adjustment is reflected in the debug log.
736 Due to multiple packets may be included to the same WQE with Enhanced Multi
737 Packet Write Method and overall WQE size is limited it is not recommended to
738 specify large values for the ``txq_inline_mpw``. Also, the default value (268)
739 may be decreased in run-time if the large transmit queue size is requested
740 and hardware does not support enough descriptor amount, in this case warning
741 is emitted. If ``txq_inline_mpw`` key is specified and requested inline
742 settings can not be satisfied then error will be raised.
744 - ``txqs_max_vec`` parameter [int]
746 Enable vectorized Tx only when the number of TX queues is less than or
747 equal to this value. This parameter is deprecated and ignored, kept
748 for compatibility issue to not prevent driver from probing.
750 - ``txq_mpw_hdr_dseg_en`` parameter [int]
752 A nonzero value enables including two pointers in the first block of TX
753 descriptor. The parameter is deprecated and ignored, kept for compatibility
756 - ``txq_max_inline_len`` parameter [int]
758 Maximum size of packet to be inlined. This limits the size of packet to
759 be inlined. If the size of a packet is larger than configured value, the
760 packet isn't inlined even though there's enough space remained in the
761 descriptor. Instead, the packet is included with pointer. This parameter
762 is deprecated and converted directly to ``txq_inline_mpw`` providing full
763 compatibility. Valid only if eMPW feature is engaged.
765 - ``txq_mpw_en`` parameter [int]
767 A nonzero value enables Enhanced Multi-Packet Write (eMPW) for ConnectX-5,
768 ConnectX-6, ConnectX-6 Dx, ConnectX-6 Lx, BlueField, BlueField-2.
769 eMPW allows the Tx burst function to pack up multiple packets
770 in a single descriptor session in order to save PCI bandwidth
771 and improve performance at the cost of a slightly higher CPU usage.
772 When ``txq_inline_mpw`` is set along with ``txq_mpw_en``,
773 Tx burst function copies entire packet data on to Tx descriptor
774 instead of including pointer of packet.
776 The Enhanced Multi-Packet Write feature is enabled by default if NIC supports
777 it, can be disabled by explicit specifying 0 value for ``txq_mpw_en`` option.
778 Also, if minimal data inlining is requested by non-zero ``txq_inline_min``
779 option or reported by the NIC, the eMPW feature is disengaged.
781 - ``tx_db_nc`` parameter [int]
783 The rdma core library can map doorbell register in two ways, depending on the
784 environment variable "MLX5_SHUT_UP_BF":
786 - As regular cached memory (usually with write combining attribute), if the
787 variable is either missing or set to zero.
788 - As non-cached memory, if the variable is present and set to not "0" value.
790 The type of mapping may slightly affect the Tx performance, the optimal choice
791 is strongly relied on the host architecture and should be deduced practically.
793 If ``tx_db_nc`` is set to zero, the doorbell is forced to be mapped to regular
794 memory (with write combining), the PMD will perform the extra write memory barrier
795 after writing to doorbell, it might increase the needed CPU clocks per packet
796 to send, but latency might be improved.
798 If ``tx_db_nc`` is set to one, the doorbell is forced to be mapped to non
799 cached memory, the PMD will not perform the extra write memory barrier
800 after writing to doorbell, on some architectures it might improve the
803 If ``tx_db_nc`` is set to two, the doorbell is forced to be mapped to regular
804 memory, the PMD will use heuristics to decide whether write memory barrier
805 should be performed. For bursts with size multiple of recommended one (64 pkts)
806 it is supposed the next burst is coming and no need to issue the extra memory
807 barrier (it is supposed to be issued in the next coming burst, at least after
808 descriptor writing). It might increase latency (on some hosts till next
809 packets transmit) and should be used with care.
811 If ``tx_db_nc`` is omitted or set to zero, the preset (if any) environment
812 variable "MLX5_SHUT_UP_BF" value is used. If there is no "MLX5_SHUT_UP_BF",
813 the default ``tx_db_nc`` value is zero for ARM64 hosts and one for others.
815 - ``tx_pp`` parameter [int]
817 If a nonzero value is specified the driver creates all necessary internal
818 objects to provide accurate packet send scheduling on mbuf timestamps.
819 The positive value specifies the scheduling granularity in nanoseconds,
820 the packet send will be accurate up to specified digits. The allowed range is
821 from 500 to 1 million of nanoseconds. The negative value specifies the module
822 of granularity and engages the special test mode the check the schedule rate.
823 By default (if the ``tx_pp`` is not specified) send scheduling on timestamps
826 - ``tx_skew`` parameter [int]
828 The parameter adjusts the send packet scheduling on timestamps and represents
829 the average delay between beginning of the transmitting descriptor processing
830 by the hardware and appearance of actual packet data on the wire. The value
831 should be provided in nanoseconds and is valid only if ``tx_pp`` parameter is
832 specified. The default value is zero.
834 - ``tx_vec_en`` parameter [int]
836 A nonzero value enables Tx vector on ConnectX-5, ConnectX-6, ConnectX-6 Dx,
837 ConnectX-6 Lx, BlueField and BlueField-2 NICs
838 if the number of global Tx queues on the port is less than ``txqs_max_vec``.
839 The parameter is deprecated and ignored.
841 - ``rx_vec_en`` parameter [int]
843 A nonzero value enables Rx vector if the port is not configured in
844 multi-segment otherwise this parameter is ignored.
848 - ``vf_nl_en`` parameter [int]
850 A nonzero value enables Netlink requests from the VF to add/remove MAC
851 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
852 Otherwise the relevant configuration must be run with Linux iproute2 tools.
853 This is a prerequisite to receive this kind of traffic.
855 Enabled by default, valid only on VF devices ignored otherwise.
857 - ``l3_vxlan_en`` parameter [int]
859 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
860 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
861 parameter. This is a prerequisite to receive this kind of traffic.
865 - ``dv_xmeta_en`` parameter [int]
867 A nonzero value enables extensive flow metadata support if device is
868 capable and driver supports it. This can enable extensive support of
869 ``MARK`` and ``META`` item of ``rte_flow``. The newly introduced
870 ``SET_TAG`` and ``SET_META`` actions do not depend on ``dv_xmeta_en``.
872 There are some possible configurations, depending on parameter value:
874 - 0, this is default value, defines the legacy mode, the ``MARK`` and
875 ``META`` related actions and items operate only within NIC Tx and
876 NIC Rx steering domains, no ``MARK`` and ``META`` information crosses
877 the domain boundaries. The ``MARK`` item is 24 bits wide, the ``META``
878 item is 32 bits wide and match supported on egress only.
880 - 1, this engages extensive metadata mode, the ``MARK`` and ``META``
881 related actions and items operate within all supported steering domains,
882 including FDB, ``MARK`` and ``META`` information may cross the domain
883 boundaries. The ``MARK`` item is 24 bits wide, the ``META`` item width
884 depends on kernel and firmware configurations and might be 0, 16 or
885 32 bits. Within NIC Tx domain ``META`` data width is 32 bits for
886 compatibility, the actual width of data transferred to the FDB domain
887 depends on kernel configuration and may be vary. The actual supported
888 width can be retrieved in runtime by series of rte_flow_validate()
891 - 2, this engages extensive metadata mode, the ``MARK`` and ``META``
892 related actions and items operate within all supported steering domains,
893 including FDB, ``MARK`` and ``META`` information may cross the domain
894 boundaries. The ``META`` item is 32 bits wide, the ``MARK`` item width
895 depends on kernel and firmware configurations and might be 0, 16 or
896 24 bits. The actual supported width can be retrieved in runtime by
897 series of rte_flow_validate() trials.
899 - 3, this engages tunnel offload mode. In E-Switch configuration, that
900 mode implicitly activates ``dv_xmeta_en=1``.
902 +------+-----------+-----------+-------------+-------------+
903 | Mode | ``MARK`` | ``META`` | ``META`` Tx | FDB/Through |
904 +======+===========+===========+=============+=============+
905 | 0 | 24 bits | 32 bits | 32 bits | no |
906 +------+-----------+-----------+-------------+-------------+
907 | 1 | 24 bits | vary 0-32 | 32 bits | yes |
908 +------+-----------+-----------+-------------+-------------+
909 | 2 | vary 0-24 | 32 bits | 32 bits | yes |
910 +------+-----------+-----------+-------------+-------------+
912 If there is no E-Switch configuration the ``dv_xmeta_en`` parameter is
913 ignored and the device is configured to operate in legacy mode (0).
915 Disabled by default (set to 0).
917 The Direct Verbs/Rules (engaged with ``dv_flow_en`` = 1) supports all
918 of the extensive metadata features. The legacy Verbs supports FLAG and
919 MARK metadata actions over NIC Rx steering domain only.
921 Setting META value to zero in flow action means there is no item provided
922 and receiving datapath will not report in mbufs the metadata are present.
923 Setting MARK value to zero in flow action means the zero FDIR ID value
924 will be reported on packet receiving.
926 For the MARK action the last 16 values in the full range are reserved for
927 internal PMD purposes (to emulate FLAG action). The valid range for the
928 MARK action values is 0-0xFFEF for the 16-bit mode and 0-xFFFFEF
929 for the 24-bit mode, the flows with the MARK action value outside
930 the specified range will be rejected.
932 - ``dv_flow_en`` parameter [int]
934 A nonzero value enables the DV flow steering assuming it is supported
935 by the driver (RDMA Core library version is rdma-core-24.0 or higher).
937 Enabled by default if supported.
939 - ``dv_esw_en`` parameter [int]
941 A nonzero value enables E-Switch using Direct Rules.
943 Enabled by default if supported.
945 - ``lacp_by_user`` parameter [int]
947 A nonzero value enables the control of LACP traffic by the user application.
948 When a bond exists in the driver, by default it should be managed by the
949 kernel and therefore LACP traffic should be steered to the kernel.
950 If this devarg is set to 1 it will allow the user to manage the bond by
951 itself and not steer LACP traffic to the kernel.
953 Disabled by default (set to 0).
955 - ``mr_ext_memseg_en`` parameter [int]
957 A nonzero value enables extending memseg when registering DMA memory. If
958 enabled, the number of entries in MR (Memory Region) lookup table on datapath
959 is minimized and it benefits performance. On the other hand, it worsens memory
960 utilization because registered memory is pinned by kernel driver. Even if a
961 page in the extended chunk is freed, that doesn't become reusable until the
962 entire memory is freed.
966 - ``representor`` parameter [list]
968 This parameter can be used to instantiate DPDK Ethernet devices from
969 existing port (PF, VF or SF) representors configured on the device.
971 It is a standard parameter whose format is described in
972 :ref:`ethernet_device_standard_device_arguments`.
974 For instance, to probe VF port representors 0 through 2::
976 <PCI_BDF>,representor=vf[0-2]
978 To probe SF port representors 0 through 2::
980 <PCI_BDF>,representor=sf[0-2]
982 To probe VF port representors 0 through 2 on both PFs of bonding device::
984 <Primary_PCI_BDF>,representor=pf[0,1]vf[0-2]
986 - ``max_dump_files_num`` parameter [int]
988 The maximum number of files per PMD entity that may be created for debug information.
989 The files will be created in /var/log directory or in current directory.
991 set to 128 by default.
993 - ``lro_timeout_usec`` parameter [int]
995 The maximum allowed duration of an LRO session, in micro-seconds.
996 PMD will set the nearest value supported by HW, which is not bigger than
997 the input ``lro_timeout_usec`` value.
998 If this parameter is not specified, by default PMD will set
999 the smallest value supported by HW.
1001 - ``hp_buf_log_sz`` parameter [int]
1003 The total data buffer size of a hairpin queue (logarithmic form), in bytes.
1004 PMD will set the data buffer size to 2 ** ``hp_buf_log_sz``, both for RX & TX.
1005 The capacity of the value is specified by the firmware and the initialization
1006 will get a failure if it is out of scope.
1007 The range of the value is from 11 to 19 right now, and the supported frame
1008 size of a single packet for hairpin is from 512B to 128KB. It might change if
1009 different firmware release is being used. By using a small value, it could
1010 reduce memory consumption but not work with a large frame. If the value is
1011 too large, the memory consumption will be high and some potential performance
1012 degradation will be introduced.
1013 By default, the PMD will set this value to 16, which means that 9KB jumbo
1014 frames will be supported.
1016 - ``reclaim_mem_mode`` parameter [int]
1018 Cache some resources in flow destroy will help flow recreation more efficient.
1019 While some systems may require the all the resources can be reclaimed after
1021 The parameter ``reclaim_mem_mode`` provides the option for user to configure
1022 if the resource cache is needed or not.
1024 There are three options to choose:
1026 - 0. It means the flow resources will be cached as usual. The resources will
1027 be cached, helpful with flow insertion rate.
1029 - 1. It will only enable the DPDK PMD level resources reclaim.
1031 - 2. Both DPDK PMD level and rdma-core low level will be configured as
1034 By default, the PMD will set this value to 0.
1036 - ``sys_mem_en`` parameter [int]
1038 A non-zero value enables the PMD memory management allocating memory
1039 from system by default, without explicit rte memory flag.
1041 By default, the PMD will set this value to 0.
1043 - ``decap_en`` parameter [int]
1045 Some devices do not support FCS (frame checksum) scattering for
1046 tunnel-decapsulated packets.
1047 If set to 0, this option forces the FCS feature and rejects tunnel
1048 decapsulation in the flow engine for such devices.
1050 By default, the PMD will set this value to 1.
1052 .. _mlx5_firmware_config:
1054 Firmware configuration
1055 ~~~~~~~~~~~~~~~~~~~~~~
1057 Firmware features can be configured as key/value pairs.
1059 The command to set a value is::
1061 mlxconfig -d <device> set <key>=<value>
1063 The command to query a value is::
1065 mlxconfig -d <device> query | grep <key>
1067 The device name for the command ``mlxconfig`` can be either the PCI address,
1068 or the mst device name found with::
1072 Below are some firmware configurations listed.
1078 value: 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
1084 - maximum number of SR-IOV virtual functions::
1088 - enable DevX (required by Direct Rules and other features)::
1092 - aggressive CQE zipping::
1096 - L3 VXLAN and VXLAN-GPE destination UDP port::
1099 IP_OVER_VXLAN_PORT=<udp dport>
1101 - enable VXLAN-GPE tunnel flow matching::
1103 FLEX_PARSER_PROFILE_ENABLE=0
1105 FLEX_PARSER_PROFILE_ENABLE=2
1107 - enable IP-in-IP tunnel flow matching::
1109 FLEX_PARSER_PROFILE_ENABLE=0
1111 - enable MPLS flow matching::
1113 FLEX_PARSER_PROFILE_ENABLE=1
1115 - enable ICMP(code/type/identifier/sequence number) / ICMP6(code/type) fields matching::
1117 FLEX_PARSER_PROFILE_ENABLE=2
1119 - enable Geneve flow matching::
1121 FLEX_PARSER_PROFILE_ENABLE=0
1123 FLEX_PARSER_PROFILE_ENABLE=1
1125 - enable Geneve TLV option flow matching::
1127 FLEX_PARSER_PROFILE_ENABLE=0
1129 - enable GTP flow matching::
1131 FLEX_PARSER_PROFILE_ENABLE=3
1133 - enable eCPRI flow matching::
1135 FLEX_PARSER_PROFILE_ENABLE=4
1141 This driver relies on external libraries and kernel drivers for resources
1142 allocations and initialization. The following dependencies are not part of
1143 DPDK and must be installed separately:
1147 User space Verbs framework used by librte_net_mlx5. This library provides
1148 a generic interface between the kernel and low-level user space drivers
1151 It allows slow and privileged operations (context initialization, hardware
1152 resources allocations) to be managed by the kernel and fast operations to
1153 never leave user space.
1157 Low-level user space driver library for Mellanox
1158 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices, it is automatically loaded
1161 This library basically implements send/receive calls to the hardware
1164 - **Kernel modules**
1166 They provide the kernel-side Verbs API and low level device drivers that
1167 manage actual hardware initialization and resources sharing with user
1170 Unlike most other PMDs, these modules must remain loaded and bound to
1173 - mlx5_core: hardware driver managing Mellanox
1174 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices and related Ethernet kernel
1176 - mlx5_ib: InifiniBand device driver.
1177 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
1179 - **Firmware update**
1181 Mellanox OFED/EN releases include firmware updates for
1182 ConnectX-4/ConnectX-5/ConnectX-6/BlueField adapters.
1184 Because each release provides new features, these updates must be applied to
1185 match the kernel modules and libraries they come with.
1189 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
1195 Either RDMA Core library with a recent enough Linux kernel release
1196 (recommended) or Mellanox OFED/EN, which provides compatibility with older
1199 RDMA Core with Linux Kernel
1200 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1202 - Minimal kernel version : v4.14 or the most recent 4.14-rc (see `Linux installation documentation`_)
1203 - Minimal rdma-core version: v15+ commit 0c5f5765213a ("Merge pull request #227 from yishaih/tm")
1204 (see `RDMA Core installation documentation`_)
1205 - When building for i686 use:
1207 - rdma-core version 18.0 or above built with 32bit support.
1208 - Kernel version 4.14.41 or above.
1210 - Starting with rdma-core v21, static libraries can be built::
1213 CFLAGS=-fPIC cmake -DIN_PLACE=1 -DENABLE_STATIC=1 -GNinja ..
1216 .. _`Linux installation documentation`: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/plain/Documentation/admin-guide/README.rst
1217 .. _`RDMA Core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
1223 - Mellanox OFED version: **4.5** and above /
1224 Mellanox EN version: **4.5** and above
1227 - ConnectX-4: **12.21.1000** and above.
1228 - ConnectX-4 Lx: **14.21.1000** and above.
1229 - ConnectX-5: **16.21.1000** and above.
1230 - ConnectX-5 Ex: **16.21.1000** and above.
1231 - ConnectX-6: **20.27.0090** and above.
1232 - ConnectX-6 Dx: **22.27.0090** and above.
1233 - BlueField: **18.25.1010** and above.
1235 While these libraries and kernel modules are available on OpenFabrics
1236 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
1237 managers on most distributions, this PMD requires Ethernet extensions that
1238 may not be supported at the moment (this is a work in progress).
1241 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__ and
1243 <http://www.mellanox.com/page/products_dyn?product_family=27&mtag=linux>`__
1244 include the necessary support and should be used in the meantime. For DPDK,
1245 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
1246 required from that distribution.
1250 Several versions of Mellanox OFED/EN are available. Installing the version
1251 this DPDK release was developed and tested against is strongly
1252 recommended. Please check the `linux prerequisites`_.
1254 Windows Prerequisites
1255 ---------------------
1257 This driver relies on external libraries and kernel drivers for resources
1258 allocations and initialization. The dependencies in the following sub-sections
1259 are not part of DPDK, and must be installed separately.
1261 Compilation Prerequisites
1262 ~~~~~~~~~~~~~~~~~~~~~~~~~
1264 DevX SDK installation
1265 ^^^^^^^^^^^^^^^^^^^^^
1267 The DevX SDK must be installed on the machine building the Windows PMD.
1268 Additional information can be found at
1269 `How to Integrate Windows DevX in Your Development Environment
1270 <https://docs.mellanox.com/display/winof2v250/RShim+Drivers+and+Usage#RShimDriversandUsage-DevXInterface>`__.
1272 Runtime Prerequisites
1273 ~~~~~~~~~~~~~~~~~~~~~
1275 WinOF2 version 2.60 or higher must be installed on the machine.
1280 The driver can be downloaded from the following site:
1282 <https://www.mellanox.com/products/adapter-software/ethernet/windows/winof-2>`__
1287 DevX for Windows must be enabled in the Windows registry.
1288 The keys ``DevxEnabled`` and ``DevxFsRules`` must be set.
1289 Additional information can be found in the WinOF2 user manual.
1294 The following Mellanox device families are supported by the same mlx5 driver:
1306 Below are detailed device names:
1308 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX4111A-XCAT (1x10G)
1309 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX412A-XCAT (2x10G)
1310 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX4111A-ACAT (1x25G)
1311 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX412A-ACAT (2x25G)
1312 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX413A-BCAT (1x40G)
1313 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX4131A-BCAT (1x40G)
1314 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX415A-BCAT (1x40G)
1315 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX413A-GCAT (1x50G)
1316 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX4131A-GCAT (1x50G)
1317 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX414A-BCAT (2x50G)
1318 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-GCAT (1x50G)
1319 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-BCAT (2x50G)
1320 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-GCAT (2x50G)
1321 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-CCAT (1x100G)
1322 * Mellanox\ |reg| ConnectX\ |reg|-4 100G MCX416A-CCAT (2x100G)
1323 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4111A-XCAT (1x10G)
1324 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4121A-XCAT (2x10G)
1325 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4111A-ACAT (1x25G)
1326 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4121A-ACAT (2x25G)
1327 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 40G MCX4131A-BCAT (1x40G)
1328 * Mellanox\ |reg| ConnectX\ |reg|-5 100G MCX556A-ECAT (2x100G)
1329 * Mellanox\ |reg| ConnectX\ |reg|-5 Ex EN 100G MCX516A-CDAT (2x100G)
1330 * Mellanox\ |reg| ConnectX\ |reg|-6 200G MCX654106A-HCAT (2x200G)
1331 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 100G MCX623106AN-CDAT (2x100G)
1332 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 200G MCX623105AN-VDAT (1x200G)
1333 * Mellanox\ |reg| ConnectX\ |reg|-6 Lx EN 25G MCX631102AN-ADAT (2x25G)
1335 Quick Start Guide on OFED/EN
1336 ----------------------------
1338 1. Download latest Mellanox OFED/EN. For more info check the `linux prerequisites`_.
1341 2. Install the required libraries and kernel modules either by installing
1342 only the required set, or by installing the entire Mellanox OFED/EN::
1344 ./mlnxofedinstall --upstream-libs --dpdk
1346 3. Verify the firmware is the correct one::
1350 4. Verify all ports links are set to Ethernet::
1352 mlxconfig -d <mst device> query | grep LINK_TYPE
1356 Link types may have to be configured to Ethernet::
1358 mlxconfig -d <mst device> set LINK_TYPE_P1/2=1/2/3
1360 * LINK_TYPE_P1=<1|2|3> , 1=Infiniband 2=Ethernet 3=VPI(auto-sense)
1362 For hypervisors, verify SR-IOV is enabled on the NIC::
1364 mlxconfig -d <mst device> query | grep SRIOV_EN
1367 If needed, configure SR-IOV::
1369 mlxconfig -d <mst device> set SRIOV_EN=1 NUM_OF_VFS=16
1370 mlxfwreset -d <mst device> reset
1372 5. Restart the driver::
1374 /etc/init.d/openibd restart
1378 service openibd restart
1380 If link type was changed, firmware must be reset as well::
1382 mlxfwreset -d <mst device> reset
1384 For hypervisors, after reset write the sysfs number of virtual functions
1387 To dynamically instantiate a given number of virtual functions (VFs)::
1389 echo [num_vfs] > /sys/class/infiniband/mlx5_0/device/sriov_numvfs
1391 6. Install DPDK and you are ready to go.
1392 See :doc:`compilation instructions <../linux_gsg/build_dpdk>`.
1394 Enable switchdev mode
1395 ---------------------
1397 Switchdev mode is a mode in E-Switch, that binds between representor and VF or SF.
1398 Representor is a port in DPDK that is connected to a VF or SF in such a way
1399 that assuming there are no offload flows, each packet that is sent from the VF or SF
1400 will be received by the corresponding representor. While each packet that is or SF
1401 sent to a representor will be received by the VF or SF.
1402 This is very useful in case of SRIOV mode, where the first packet that is sent
1403 by the VF or SF will be received by the DPDK application which will decide if this
1404 flow should be offloaded to the E-Switch. After offloading the flow packet
1405 that the VF or SF that are matching the flow will not be received any more by
1406 the DPDK application.
1408 1. Enable SRIOV mode::
1410 mlxconfig -d <mst device> set SRIOV_EN=true
1412 2. Configure the max number of VFs::
1414 mlxconfig -d <mst device> set NUM_OF_VFS=<num of vfs>
1418 mlxfwreset -d <mst device> reset
1420 3. Configure the actual number of VFs::
1422 echo <num of vfs > /sys/class/net/<net device>/device/sriov_numvfs
1424 4. Unbind the device (can be rebind after the switchdev mode)::
1426 echo -n "<device pci address" > /sys/bus/pci/drivers/mlx5_core/unbind
1428 5. Enbale switchdev mode::
1430 echo switchdev > /sys/class/net/<net device>/compat/devlink/mode
1432 SubFunction representor support
1433 -------------------------------
1434 SubFunction is a portion of the PCI device, a SF netdev has its own
1435 dedicated queues(txq, rxq). A SF netdev supports E-Switch representation
1436 offload similar to existing PF and VF representors. A SF shares PCI
1437 level resources with other SFs and/or with its parent PCI function.
1439 1. Configure SF feature::
1441 mlxconfig -d <mst device> set PF_BAR2_SIZE=<0/1/2/3> PF_BAR2_ENABLE=1
1443 Value of PF_BAR2_SIZE:
1452 mlxfwreset -d <mst device> reset
1454 3. Enable switchdev mode::
1456 echo switchdev > /sys/class/net/<net device>/compat/devlink/mode
1460 mlnx-sf -d <PCI_BDF> -a create
1462 5. Probe SF representor::
1464 testpmd> port attach <PCI_BDF>,representor=sf0,dv_flow_en=1
1469 1. Configure aggressive CQE Zipping for maximum performance::
1471 mlxconfig -d <mst device> s CQE_COMPRESSION=1
1473 To set it back to the default CQE Zipping mode use::
1475 mlxconfig -d <mst device> s CQE_COMPRESSION=0
1477 2. In case of virtualization:
1479 - Make sure that hypervisor kernel is 3.16 or newer.
1480 - Configure boot with ``iommu=pt``.
1481 - Use 1G huge pages.
1482 - Make sure to allocate a VM on huge pages.
1483 - Make sure to set CPU pinning.
1485 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
1486 for better performance. For VMs, verify that the right CPU
1487 and NUMA node are pinned according to the above. Run::
1491 to identify the NUMA node to which the PCIe adapter is connected.
1493 4. If more than one adapter is used, and root complex capabilities allow
1494 to put both adapters on the same NUMA node without PCI bandwidth degradation,
1495 it is recommended to locate both adapters on the same NUMA node.
1496 This in order to forward packets from one to the other without
1497 NUMA performance penalty.
1499 5. Disable pause frames::
1501 ethtool -A <netdev> rx off tx off
1503 6. Verify IO non-posted prefetch is disabled by default. This can be checked
1504 via the BIOS configuration. Please contact you server provider for more
1505 information about the settings.
1509 On some machines, depends on the machine integrator, it is beneficial
1510 to set the PCI max read request parameter to 1K. This can be
1511 done in the following way:
1513 To query the read request size use::
1515 setpci -s <NIC PCI address> 68.w
1517 If the output is different than 3XXX, set it by::
1519 setpci -s <NIC PCI address> 68.w=3XXX
1521 The XXX can be different on different systems. Make sure to configure
1522 according to the setpci output.
1524 7. To minimize overhead of searching Memory Regions:
1526 - '--socket-mem' is recommended to pin memory by predictable amount.
1527 - Configure per-lcore cache when creating Mempools for packet buffer.
1528 - Refrain from dynamically allocating/freeing memory in run-time.
1533 There are multiple Rx burst functions with different advantages and limitations.
1535 .. table:: Rx burst functions
1537 +-------------------+------------------------+---------+-----------------+------+-------+
1538 || Function Name || Enabler || Scatter|| Error Recovery || CQE || Large|
1539 | | | | || comp|| MTU |
1540 +===================+========================+=========+=================+======+=======+
1541 | rx_burst | rx_vec_en=0 | Yes | Yes | Yes | Yes |
1542 +-------------------+------------------------+---------+-----------------+------+-------+
1543 | rx_burst_vec | rx_vec_en=1 (default) | No | if CQE comp off | Yes | No |
1544 +-------------------+------------------------+---------+-----------------+------+-------+
1545 | rx_burst_mprq || mprq_en=1 | No | Yes | Yes | Yes |
1546 | || RxQs >= rxqs_min_mprq | | | | |
1547 +-------------------+------------------------+---------+-----------------+------+-------+
1548 | rx_burst_mprq_vec || rx_vec_en=1 (default) | No | if CQE comp off | Yes | Yes |
1549 | || mprq_en=1 | | | | |
1550 | || RxQs >= rxqs_min_mprq | | | | |
1551 +-------------------+------------------------+---------+-----------------+------+-------+
1553 .. _mlx5_offloads_support:
1555 Supported hardware offloads
1556 ---------------------------
1558 .. table:: Minimal SW/HW versions for queue offloads
1560 ============== ===== ===== ========= ===== ========== =============
1561 Offload DPDK Linux rdma-core OFED firmware hardware
1562 ============== ===== ===== ========= ===== ========== =============
1563 common base 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1564 checksums 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1565 Rx timestamp 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1566 TSO 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1567 LRO 19.08 N/A N/A 4.6-4 16.25.6406 ConnectX-5
1568 Tx scheduling 20.08 N/A N/A 5.1-2 22.28.2006 ConnectX-6 Dx
1569 Buffer Split 20.11 N/A N/A 5.1-2 16.28.2006 ConnectX-5
1570 ============== ===== ===== ========= ===== ========== =============
1572 .. table:: Minimal SW/HW versions for rte_flow offloads
1574 +-----------------------+-----------------+-----------------+
1575 | Offload | with E-Switch | with NIC |
1576 +=======================+=================+=================+
1577 | Count | | DPDK 19.05 | | DPDK 19.02 |
1578 | | | OFED 4.6 | | OFED 4.6 |
1579 | | | rdma-core 24 | | rdma-core 23 |
1580 | | | ConnectX-5 | | ConnectX-5 |
1581 +-----------------------+-----------------+-----------------+
1582 | Drop | | DPDK 19.05 | | DPDK 18.11 |
1583 | | | OFED 4.6 | | OFED 4.5 |
1584 | | | rdma-core 24 | | rdma-core 23 |
1585 | | | ConnectX-5 | | ConnectX-4 |
1586 +-----------------------+-----------------+-----------------+
1587 | Queue / RSS | | | | DPDK 18.11 |
1588 | | | N/A | | OFED 4.5 |
1589 | | | | | rdma-core 23 |
1590 | | | | | ConnectX-4 |
1591 +-----------------------+-----------------+-----------------+
1592 | Shared action | | | | |
1593 | | | :numref:`sact`| | :numref:`sact`|
1596 +-----------------------+-----------------+-----------------+
1597 | | VLAN | | DPDK 19.11 | | DPDK 19.11 |
1598 | | (of_pop_vlan / | | OFED 4.7-1 | | OFED 4.7-1 |
1599 | | of_push_vlan / | | ConnectX-5 | | ConnectX-5 |
1600 | | of_set_vlan_pcp / | | | | |
1601 | | of_set_vlan_vid) | | | | |
1602 +-----------------------+-----------------+-----------------+
1603 | | VLAN | | DPDK 21.05 | | |
1604 | | ingress and / | | OFED 5.3 | | N/A |
1605 | | of_push_vlan / | | ConnectX-6 Dx | | |
1606 +-----------------------+-----------------+-----------------+
1607 | | VLAN | | DPDK 21.05 | | |
1608 | | egress and / | | OFED 5.3 | | N/A |
1609 | | of_pop_vlan / | | ConnectX-6 Dx | | |
1610 +-----------------------+-----------------+-----------------+
1611 | Encapsulation | | DPDK 19.05 | | DPDK 19.02 |
1612 | (VXLAN / NVGRE / RAW) | | OFED 4.7-1 | | OFED 4.6 |
1613 | | | rdma-core 24 | | rdma-core 23 |
1614 | | | ConnectX-5 | | ConnectX-5 |
1615 +-----------------------+-----------------+-----------------+
1616 | Encapsulation | | DPDK 19.11 | | DPDK 19.11 |
1617 | GENEVE | | OFED 4.7-3 | | OFED 4.7-3 |
1618 | | | rdma-core 27 | | rdma-core 27 |
1619 | | | ConnectX-5 | | ConnectX-5 |
1620 +-----------------------+-----------------+-----------------+
1621 | Tunnel Offload | | DPDK 20.11 | | DPDK 20.11 |
1622 | | | OFED 5.1-2 | | OFED 5.1-2 |
1623 | | | rdma-core 32 | | N/A |
1624 | | | ConnectX-5 | | ConnectX-5 |
1625 +-----------------------+-----------------+-----------------+
1626 | | Header rewrite | | DPDK 19.05 | | DPDK 19.02 |
1627 | | (set_ipv4_src / | | OFED 4.7-1 | | OFED 4.7-1 |
1628 | | set_ipv4_dst / | | rdma-core 24 | | rdma-core 24 |
1629 | | set_ipv6_src / | | ConnectX-5 | | ConnectX-5 |
1630 | | set_ipv6_dst / | | | | |
1631 | | set_tp_src / | | | | |
1632 | | set_tp_dst / | | | | |
1633 | | dec_ttl / | | | | |
1634 | | set_ttl / | | | | |
1635 | | set_mac_src / | | | | |
1636 | | set_mac_dst) | | | | |
1637 +-----------------------+-----------------+-----------------+
1638 | | Header rewrite | | DPDK 20.02 | | DPDK 20.02 |
1639 | | (set_dscp) | | OFED 5.0 | | OFED 5.0 |
1640 | | | | rdma-core 24 | | rdma-core 24 |
1641 | | | | ConnectX-5 | | ConnectX-5 |
1642 +-----------------------+-----------------+-----------------+
1643 | Jump | | DPDK 19.05 | | DPDK 19.02 |
1644 | | | OFED 4.7-1 | | OFED 4.7-1 |
1645 | | | rdma-core 24 | | N/A |
1646 | | | ConnectX-5 | | ConnectX-5 |
1647 +-----------------------+-----------------+-----------------+
1648 | Mark / Flag | | DPDK 19.05 | | DPDK 18.11 |
1649 | | | OFED 4.6 | | OFED 4.5 |
1650 | | | rdma-core 24 | | rdma-core 23 |
1651 | | | ConnectX-5 | | ConnectX-4 |
1652 +-----------------------+-----------------+-----------------+
1653 | Meta data | | DPDK 19.11 | | DPDK 19.11 |
1654 | | | OFED 4.7-3 | | OFED 4.7-3 |
1655 | | | rdma-core 26 | | rdma-core 26 |
1656 | | | ConnectX-5 | | ConnectX-5 |
1657 +-----------------------+-----------------+-----------------+
1658 | Port ID | | DPDK 19.05 | | N/A |
1659 | | | OFED 4.7-1 | | N/A |
1660 | | | rdma-core 24 | | N/A |
1661 | | | ConnectX-5 | | N/A |
1662 +-----------------------+-----------------+-----------------+
1663 | Hairpin | | | | DPDK 19.11 |
1664 | | | N/A | | OFED 4.7-3 |
1665 | | | | | rdma-core 26 |
1666 | | | | | ConnectX-5 |
1667 +-----------------------+-----------------+-----------------+
1668 | 2-port Hairpin | | | | DPDK 20.11 |
1669 | | | N/A | | OFED 5.1-2 |
1671 | | | | | ConnectX-5 |
1672 +-----------------------+-----------------+-----------------+
1673 | Metering | | DPDK 19.11 | | DPDK 19.11 |
1674 | | | OFED 4.7-3 | | OFED 4.7-3 |
1675 | | | rdma-core 26 | | rdma-core 26 |
1676 | | | ConnectX-5 | | ConnectX-5 |
1677 +-----------------------+-----------------+-----------------+
1678 | Sampling | | DPDK 20.11 | | DPDK 20.11 |
1679 | | | OFED 5.1-2 | | OFED 5.1-2 |
1680 | | | rdma-core 32 | | N/A |
1681 | | | ConnectX-5 | | ConnectX-5 |
1682 +-----------------------+-----------------+-----------------+
1683 | Encapsulation | | DPDK 21.02 | | DPDK 21.02 |
1684 | GTP PSC | | OFED 5.2 | | OFED 5.2 |
1685 | | | rdma-core 35 | | rdma-core 35 |
1686 | | | ConnectX-6 Dx| | ConnectX-6 Dx |
1687 +-----------------------+-----------------+-----------------+
1688 | Encapsulation | | DPDK 21.02 | | DPDK 21.02 |
1689 | GENEVE TLV option | | OFED 5.2 | | OFED 5.2 |
1690 | | | rdma-core 34 | | rdma-core 34 |
1691 | | | ConnectX-6 Dx | | ConnectX-6 Dx |
1692 +-----------------------+-----------------+-----------------+
1693 | Modify Field | | DPDK 21.02 | | DPDK 21.02 |
1694 | | | OFED 5.2 | | OFED 5.2 |
1695 | | | rdma-core 35 | | rdma-core 35 |
1696 | | | ConnectX-5 | | ConnectX-5 |
1697 +-----------------------+-----------------+-----------------+
1699 .. table:: Minimal SW/HW versions for shared action offload
1702 +-----------------------+-----------------+-----------------+
1703 | Shared Action | with E-Switch | with NIC |
1704 +=======================+=================+=================+
1705 | RSS | | | | DPDK 20.11 |
1706 | | | N/A | | OFED 5.2 |
1707 | | | | | rdma-core 33 |
1708 | | | | | ConnectX-5 |
1709 +-----------------------+-----------------+-----------------+
1710 | Age | | DPDK 20.11 | | DPDK 20.11 |
1711 | | | OFED 5.2 | | OFED 5.2 |
1712 | | | rdma-core 32 | | rdma-core 32 |
1713 | | | ConnectX-6 Dx | | ConnectX-6 Dx |
1714 +-----------------------+-----------------+-----------------+
1715 | Count | | DPDK 21.05 | | DPDK 21.05 |
1716 | | | OFED 4.6 | | OFED 4.6 |
1717 | | | rdma-core 24 | | rdma-core 23 |
1718 | | | ConnectX-5 | | ConnectX-5 |
1719 +-----------------------+-----------------+-----------------+
1724 MARK and META items are interrelated with datapath - they might move from/to
1725 the applications in mbuf fields. Hence, zero value for these items has the
1726 special meaning - it means "no metadata are provided", not zero values are
1727 treated by applications and PMD as valid ones.
1729 Moreover in the flow engine domain the value zero is acceptable to match and
1730 set, and we should allow to specify zero values as rte_flow parameters for the
1731 META and MARK items and actions. In the same time zero mask has no meaning and
1732 should be rejected on validation stage.
1737 Flows are not cached in the driver.
1738 When stopping a device port, all the flows created on this port from the
1739 application will be flushed automatically in the background.
1740 After stopping the device port, all flows on this port become invalid and
1741 not represented in the system.
1742 All references to these flows held by the application should be discarded
1743 directly but neither destroyed nor flushed.
1745 The application should re-create the flows as required after the port restart.
1750 Compared to librte_net_mlx4 that implements a single RSS configuration per
1751 port, librte_net_mlx5 supports per-protocol RSS configuration.
1753 Since ``testpmd`` defaults to IP RSS mode and there is currently no
1754 command-line parameter to enable additional protocols (UDP and TCP as well
1755 as IP), the following commands must be entered from its CLI to get the same
1756 behavior as librte_net_mlx4::
1759 > port config all rss all
1765 This section demonstrates how to launch **testpmd** with Mellanox
1766 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices managed by librte_net_mlx5.
1768 #. Load the kernel modules::
1770 modprobe -a ib_uverbs mlx5_core mlx5_ib
1772 Alternatively if MLNX_OFED/MLNX_EN is fully installed, the following script
1775 /etc/init.d/openibd restart
1779 User space I/O kernel modules (uio and igb_uio) are not used and do
1780 not have to be loaded.
1782 #. Make sure Ethernet interfaces are in working order and linked to kernel
1783 verbs. Related sysfs entries should be present::
1785 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
1794 #. Optionally, retrieve their PCI bus addresses for to be used with the allow list::
1797 for intf in eth2 eth3 eth4 eth5;
1799 (cd "/sys/class/net/${intf}/device/" && pwd -P);
1802 sed -n 's,.*/\(.*\),-a \1,p'
1811 #. Request huge pages::
1813 dpdk-hugepages.py --setup 2G
1815 #. Start testpmd with basic parameters::
1817 dpdk-testpmd -l 8-15 -n 4 -a 05:00.0 -a 05:00.1 -a 06:00.0 -a 06:00.1 -- --rxq=2 --txq=2 -i
1822 EAL: PCI device 0000:05:00.0 on NUMA socket 0
1823 EAL: probe driver: 15b3:1013 librte_net_mlx5
1824 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
1825 PMD: librte_net_mlx5: 1 port(s) detected
1826 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
1827 EAL: PCI device 0000:05:00.1 on NUMA socket 0
1828 EAL: probe driver: 15b3:1013 librte_net_mlx5
1829 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
1830 PMD: librte_net_mlx5: 1 port(s) detected
1831 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
1832 EAL: PCI device 0000:06:00.0 on NUMA socket 0
1833 EAL: probe driver: 15b3:1013 librte_net_mlx5
1834 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
1835 PMD: librte_net_mlx5: 1 port(s) detected
1836 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
1837 EAL: PCI device 0000:06:00.1 on NUMA socket 0
1838 EAL: probe driver: 15b3:1013 librte_net_mlx5
1839 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
1840 PMD: librte_net_mlx5: 1 port(s) detected
1841 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
1842 Interactive-mode selected
1843 Configuring Port 0 (socket 0)
1844 PMD: librte_net_mlx5: 0x8cba80: TX queues number update: 0 -> 2
1845 PMD: librte_net_mlx5: 0x8cba80: RX queues number update: 0 -> 2
1846 Port 0: E4:1D:2D:E7:0C:FE
1847 Configuring Port 1 (socket 0)
1848 PMD: librte_net_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
1849 PMD: librte_net_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
1850 Port 1: E4:1D:2D:E7:0C:FF
1851 Configuring Port 2 (socket 0)
1852 PMD: librte_net_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
1853 PMD: librte_net_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
1854 Port 2: E4:1D:2D:E7:0C:FA
1855 Configuring Port 3 (socket 0)
1856 PMD: librte_net_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
1857 PMD: librte_net_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
1858 Port 3: E4:1D:2D:E7:0C:FB
1859 Checking link statuses...
1860 Port 0 Link Up - speed 40000 Mbps - full-duplex
1861 Port 1 Link Up - speed 40000 Mbps - full-duplex
1862 Port 2 Link Up - speed 10000 Mbps - full-duplex
1863 Port 3 Link Up - speed 10000 Mbps - full-duplex
1870 This section demonstrates how to dump flows. Currently, it's possible to dump
1871 all flows with assistance of external tools.
1873 #. 2 ways to get flow raw file:
1875 - Using testpmd CLI:
1877 .. code-block:: console
1880 testpmd> flow dump <port> all <output_file>
1882 testpmd> flow dump <port> rule <rule_id> <output_file>
1884 - call rte_flow_dev_dump api:
1886 .. code-block:: console
1888 rte_flow_dev_dump(port, flow, file, NULL);
1890 #. Dump human-readable flows from raw file:
1892 Get flow parsing tool from: https://github.com/Mellanox/mlx_steering_dump
1894 .. code-block:: console
1896 mlx_steering_dump.py -f <output_file> -flowptr <flow_ptr>