1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2015 6WIND S.A.
3 Copyright 2015 Mellanox Technologies, Ltd
5 .. include:: <isonum.txt>
7 MLX5 Ethernet Poll Mode Driver
8 ==============================
10 The mlx5 Ethernet 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)
21 Besides its dependency on libibverbs (that implies libmlx5 and associated
22 kernel support), librte_net_mlx5 relies heavily on system calls for control
23 operations such as querying/updating the MTU and flow control parameters.
25 This capability allows the PMD to coexist with kernel network interfaces
26 which remain functional, although they stop receiving unicast packets as
27 long as they share the same MAC address.
28 This means legacy linux control tools (for example: ethtool, ifconfig and
29 more) can operate on the same network interfaces that owned by the DPDK
32 See :doc:`../../platform/mlx5` guide for more design details.
37 - Multi arch support: x86_64, POWER8, ARMv8, i686.
38 - Multiple TX and RX queues.
40 - Rx queue delay drop.
41 - Support for scattered TX frames.
42 - Advanced support for scattered Rx frames with tunable buffer attributes.
43 - IPv4, IPv6, TCPv4, TCPv6, UDPv4 and UDPv6 RSS on any number of queues.
44 - RSS using different combinations of fields: L3 only, L4 only or both,
45 and source only, destination only or both.
46 - Several RSS hash keys, one for each flow type.
47 - Default RSS operation with no hash key specification.
48 - Configurable RETA table.
49 - Link flow control (pause frame).
50 - Support for multiple MAC addresses.
54 - RX CRC stripping configuration.
55 - TX mbuf fast free offload.
56 - Promiscuous mode on PF and VF.
57 - Multicast promiscuous mode on PF and VF.
58 - Hardware checksum offloads.
59 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
61 - Flow API, including :ref:`flow_isolated_mode`.
63 - KVM and VMware ESX SR-IOV modes are supported.
64 - RSS hash result is supported.
65 - Hardware TSO for generic IP or UDP tunnel, including VXLAN and GRE.
66 - Hardware checksum Tx offload for generic IP or UDP tunnel, including VXLAN and GRE.
68 - Statistics query including Basic, Extended and per queue.
70 - Tunnel types: VXLAN, L3 VXLAN, VXLAN-GPE, GRE, MPLSoGRE, MPLSoUDP, IP-in-IP, Geneve, GTP.
71 - Tunnel HW offloads: packet type, inner/outer RSS, IP and UDP checksum verification.
72 - NIC HW offloads: encapsulation (vxlan, gre, mplsoudp, mplsogre), NAT, routing, TTL
73 increment/decrement, count, drop, mark. For details please see :ref:`mlx5_offloads_support`.
74 - Flow insertion rate of more then million flows per second, when using Direct Rules.
75 - Support for multiple rte_flow groups.
76 - Per packet no-inline hint flag to disable packet data copying into Tx descriptors.
79 - Multiple-thread flow insertion.
80 - Matching on IPv4 Internet Header Length (IHL).
81 - Matching on GTP extension header with raw encap/decap action.
82 - Matching on Geneve TLV option header with raw encap/decap action.
83 - RSS support in sample action.
84 - E-Switch mirroring and jump.
85 - E-Switch mirroring and modify.
86 - 21844 flow priorities for ingress or egress flow groups greater than 0 and for any transfer
88 - Flow metering, including meter policy API.
89 - Flow meter hierarchy.
90 - Flow integrity offload API.
91 - Connection tracking.
92 - Sub-Function representors.
101 On Windows, the features are limited:
103 - Promiscuous mode is not supported
104 - The following rules are supported:
106 - IPv4/UDP with CVLAN filtering
107 - Unicast MAC filtering
109 - Additional rules are supported from WinOF2 version 2.70:
111 - IPv4/TCP with CVLAN filtering
112 - L4 steering rules for port RSS of UDP, TCP and IP
114 - For secondary process:
116 - Forked secondary process not supported.
117 - MPRQ is not supported. Callback to free externally attached MPRQ buffer is set
118 in a primary process, but has a different virtual address in a secondary process.
119 Calling a function at the wrong address leads to a segmentation fault.
120 - External memory unregistered in EAL memseg list cannot be used for DMA
121 unless such memory has been registered by ``mlx5_mr_update_ext_mp()`` in
122 primary process and remapped to the same virtual address in secondary
123 process. If the external memory is registered by primary process but has
124 different virtual address in secondary process, unexpected error may happen.
128 - Counters of received packets and bytes number of devices in same share group are same.
129 - Counters of received packets and bytes number of queues in same group and queue ID are same.
131 - When using Verbs flow engine (``dv_flow_en`` = 0), flow pattern without any
132 specific VLAN will match for VLAN packets as well:
134 When VLAN spec is not specified in the pattern, the matching rule will be created with VLAN as a wild card.
135 Meaning, the flow rule::
137 flow create 0 ingress pattern eth / vlan vid is 3 / ipv4 / end ...
139 Will only match vlan packets with vid=3. and the flow rule::
141 flow create 0 ingress pattern eth / ipv4 / end ...
143 Will match any ipv4 packet (VLAN included).
145 - When using Verbs flow engine (``dv_flow_en`` = 0), multi-tagged(QinQ) match is not supported.
147 - 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.
150 flow create 0 ingress pattern eth / ipv4 / end ...
152 Will match any ipv4 packet.
155 flow create 0 ingress pattern eth / vlan / end ...
156 flow create 0 ingress pattern eth has_vlan is 1 / end ...
157 flow create 0 ingress pattern eth type is 0x8100 / end ...
159 Will match single-tagged packets only, with any VLAN ID value.
162 flow create 0 ingress pattern eth type is 0x88A8 / end ...
163 flow create 0 ingress pattern eth / vlan has_more_vlan is 1 / end ...
165 Will match multi-tagged packets only, with any VLAN ID value.
167 - A flow pattern with 2 sequential VLAN items is not supported.
169 - VLAN pop offload command:
171 - Flow rules having a VLAN pop offload command as one of their actions and
172 are lacking a match on VLAN as one of their items are not supported.
173 - The command is not supported on egress traffic in NIC mode.
175 - VLAN push offload is not supported on ingress traffic in NIC mode.
177 - VLAN set PCP offload is not supported on existing headers.
179 - A multi segment packet must have not more segments than reported by dev_infos_get()
180 in tx_desc_lim.nb_seg_max field. This value depends on maximal supported Tx descriptor
181 size and ``txq_inline_min`` settings and may be from 2 (worst case forced by maximal
182 inline settings) to 58.
184 - Match on VXLAN supports the following fields only:
187 - Last reserved 8-bits
189 Last reserved 8-bits matching is only supported When using DV flow
190 engine (``dv_flow_en`` = 1).
191 For ConnectX-5, the UDP destination port must be the standard one (4789).
192 Group zero's behavior may differ which depends on FW.
193 Matching value equals 0 (value & mask) is not supported.
195 - L3 VXLAN and VXLAN-GPE tunnels cannot be supported together with MPLSoGRE and MPLSoUDP.
197 - Match on Geneve header supports the following fields only:
204 - Match on Geneve TLV option is supported on the following fields:
211 Only one Class/Type/Length Geneve TLV option is supported per shared device.
212 Class/Type/Length fields must be specified as well as masks.
213 Class/Type/Length specified masks must be full.
214 Matching Geneve TLV option without specifying data is not supported.
215 Matching Geneve TLV option with ``data & mask == 0`` is not supported.
217 - VF: flow rules created on VF devices can only match traffic targeted at the
218 configured MAC addresses (see ``rte_eth_dev_mac_addr_add()``).
220 - Match on GTP tunnel header item supports the following fields only:
222 - v_pt_rsv_flags: E flag, S flag, PN flag
226 - Match on GTP extension header only for GTP PDU session container (next
227 extension header type = 0x85).
228 - Match on GTP extension header is not supported in group 0.
232 - Hardware support: BlueField-2.
233 - Flex item is supported on PF only.
234 - Hardware limits ``header_length_mask_width`` up to 6 bits.
235 - Firmware supports 8 global sample fields.
236 Each flex item allocates non-shared sample fields from that pool.
237 - Supported flex item can have 1 input link - ``eth`` or ``udp``
238 and up to 2 output links - ``ipv4`` or ``ipv6``.
239 - Flex item fields (``next_header``, ``next_protocol``, ``samples``)
240 do not participate in RSS hash functions.
241 - In flex item configuration, ``next_header.field_base`` value
242 must be byte aligned (multiple of 8).
244 - No Tx metadata go to the E-Switch steering domain for the Flow group 0.
245 The flows within group 0 and set metadata action are rejected by hardware.
249 MAC addresses not already present in the bridge table of the associated
250 kernel network device will be added and cleaned up by the PMD when closing
251 the device. In case of ungraceful program termination, some entries may
252 remain present and should be removed manually by other means.
254 - Buffer split offload is supported with regular Rx burst routine only,
255 no MPRQ feature or vectorized code can be engaged.
257 - When Multi-Packet Rx queue is configured (``mprq_en``), a Rx packet can be
258 externally attached to a user-provided mbuf with having RTE_MBUF_F_EXTERNAL in
259 ol_flags. As the mempool for the external buffer is managed by PMD, all the
260 Rx mbufs must be freed before the device is closed. Otherwise, the mempool of
261 the external buffers will be freed by PMD and the application which still
262 holds the external buffers may be corrupted.
264 - If Multi-Packet Rx queue is configured (``mprq_en``) and Rx CQE compression is
265 enabled (``rxq_cqe_comp_en``) at the same time, RSS hash result is not fully
266 supported. Some Rx packets may not have RTE_MBUF_F_RX_RSS_HASH.
268 - IPv6 Multicast messages are not supported on VM, while promiscuous mode
269 and allmulticast mode are both set to off.
270 To receive IPv6 Multicast messages on VM, explicitly set the relevant
271 MAC address using rte_eth_dev_mac_addr_add() API.
273 - To support a mixed traffic pattern (some buffers from local host memory, some
274 buffers from other devices) with high bandwidth, a mbuf flag is used.
276 An application hints the PMD whether or not it should try to inline the
277 given mbuf data buffer. PMD should do the best effort to act upon this request.
279 The hint flag ``RTE_PMD_MLX5_FINE_GRANULARITY_INLINE`` is dynamic,
280 registered by application with rte_mbuf_dynflag_register(). This flag is
281 purely driver-specific and declared in PMD specific header ``rte_pmd_mlx5.h``,
282 which is intended to be used by the application.
284 To query the supported specific flags in runtime,
285 the function ``rte_pmd_mlx5_get_dyn_flag_names`` returns the array of
286 currently (over present hardware and configuration) supported specific flags.
287 The "not inline hint" feature operating flow is the following one:
290 - probe the devices, ports are created
291 - query the port capabilities
292 - if port supporting the feature is found
293 - register dynamic flag ``RTE_PMD_MLX5_FINE_GRANULARITY_INLINE``
294 - application starts the ports
295 - on ``dev_start()`` PMD checks whether the feature flag is registered and
296 enables the feature support in datapath
297 - application might set the registered flag bit in ``ol_flags`` field
298 of mbuf being sent and PMD will handle ones appropriately.
300 - The amount of descriptors in Tx queue may be limited by data inline settings.
301 Inline data require the more descriptor building blocks and overall block
302 amount may exceed the hardware supported limits. The application should
303 reduce the requested Tx size or adjust data inline settings with
304 ``txq_inline_max`` and ``txq_inline_mpw`` devargs keys.
306 - To provide the packet send scheduling on mbuf timestamps the ``tx_pp``
307 parameter should be specified.
308 When PMD sees the RTE_MBUF_DYNFLAG_TX_TIMESTAMP_NAME set on the packet
309 being sent it tries to synchronize the time of packet appearing on
310 the wire with the specified packet timestamp. It the specified one
311 is in the past it should be ignored, if one is in the distant future
312 it should be capped with some reasonable value (in range of seconds).
313 These specific cases ("too late" and "distant future") can be optionally
314 reported via device xstats to assist applications to detect the
315 time-related problems.
317 The timestamp upper "too-distant-future" limit
318 at the moment of invoking the Tx burst routine
319 can be estimated as ``tx_pp`` option (in nanoseconds) multiplied by 2^23.
320 Please note, for the testpmd txonly mode,
321 the limit is deduced from the expression::
323 (n_tx_descriptors / burst_size + 1) * inter_burst_gap
325 There is no any packet reordering according timestamps is supposed,
326 neither within packet burst, nor between packets, it is an entirely
327 application responsibility to generate packets and its timestamps
328 in desired order. The timestamps can be put only in the first packet
329 in the burst providing the entire burst scheduling.
331 - E-Switch decapsulation Flow:
333 - can be applied to PF port only.
334 - must specify VF port action (packet redirection from PF to VF).
335 - optionally may specify tunnel inner source and destination MAC addresses.
337 - E-Switch encapsulation Flow:
339 - can be applied to VF ports only.
340 - must specify PF port action (packet redirection from VF to PF).
344 - The input buffer, used as outer header, is not validated.
348 - The decapsulation is always done up to the outermost tunnel detected by the HW.
349 - The input buffer, providing the removal size, is not validated.
350 - The buffer size must match the length of the headers to be removed.
352 - ICMP(code/type/identifier/sequence number) / ICMP6(code/type) matching, IP-in-IP and MPLS flow matching are all
353 mutually exclusive features which cannot be supported together
354 (see :ref:`mlx5_firmware_config`).
358 - Requires DevX and DV flow to be enabled.
359 - KEEP_CRC offload cannot be supported with LRO.
360 - The first mbuf length, without head-room, must be big enough to include the
362 - Rx queue with LRO offload enabled, receiving a non-LRO packet, can forward
363 it with size limited to max LRO size, not to max RX packet length.
364 - LRO can be used with outer header of TCP packets of the standard format:
365 eth (with or without vlan) / ipv4 or ipv6 / tcp / payload
367 Other TCP packets (e.g. with MPLS label) received on Rx queue with LRO enabled, will be received with bad checksum.
368 - LRO packet aggregation is performed by HW only for packet size larger than
369 ``lro_min_mss_size``. This value is reported on device start, when debug
374 - ``RTE_ETH_RX_OFFLOAD_KEEP_CRC`` cannot be supported with decapsulation
375 for some NICs (such as ConnectX-6 Dx, ConnectX-6 Lx, and BlueField-2).
376 The capability bit ``scatter_fcs_w_decap_disable`` shows NIC support.
380 - fast free offload assumes the all mbufs being sent are originated from the
381 same memory pool and there is no any extra references to the mbufs (the
382 reference counter for each mbuf is equal 1 on tx_burst call). The latter
383 means there should be no any externally attached buffers in mbufs. It is
384 an application responsibility to provide the correct mbufs if the fast
385 free offload is engaged. The mlx5 PMD implicitly produces the mbufs with
386 externally attached buffers if MPRQ option is enabled, hence, the fast
387 free offload is neither supported nor advertised if there is MPRQ enabled.
391 - Supports ``RTE_FLOW_ACTION_TYPE_SAMPLE`` action only within NIC Rx and
392 E-Switch steering domain.
393 - For E-Switch Sampling flow with sample ratio > 1, additional actions are not
394 supported in the sample actions list.
395 - For ConnectX-5, the ``RTE_FLOW_ACTION_TYPE_SAMPLE`` is typically used as
396 first action in the E-Switch egress flow if with header modify or
397 encapsulation actions.
398 - For NIC Rx flow, supports ``MARK``, ``COUNT``, ``QUEUE``, ``RSS`` in the
400 - For E-Switch mirroring flow, supports ``RAW ENCAP``, ``Port ID``,
401 ``VXLAN ENCAP``, ``NVGRE ENCAP`` in the sample actions list.
405 - Supports the 'set' operation only for ``RTE_FLOW_ACTION_TYPE_MODIFY_FIELD`` action.
406 - Modification of an arbitrary place in a packet via the special ``RTE_FLOW_FIELD_START`` Field ID is not supported.
407 - Modification of the 802.1Q Tag, VXLAN Network or GENEVE Network ID's is not supported.
408 - Encapsulation levels are not supported, can modify outermost header fields only.
409 - Offsets must be 32-bits aligned, cannot skip past the boundary of a field.
411 - IPv6 header item 'proto' field, indicating the next header protocol, should
412 not be set as extension header.
413 In case the next header is an extension header, it should not be specified in
414 IPv6 header item 'proto' field.
415 The last extension header item 'next header' field can specify the following
416 header protocol type.
420 - 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.
421 - Hairpin in switchdev SR-IOV mode is not supported till now.
425 - All the meter colors with drop action will be counted only by the global drop statistics.
426 - Yellow detection is only supported with ASO metering.
427 - Red color must be with drop action.
428 - Meter statistics are supported only for drop case.
429 - A meter action created with pre-defined policy must be the last action in the flow except single case where the policy actions are:
430 - green: NULL or END.
431 - yellow: NULL or END.
433 - The only supported meter policy actions:
434 - green: QUEUE, RSS, PORT_ID, REPRESENTED_PORT, JUMP, DROP, MARK and SET_TAG.
435 - yellow: QUEUE, RSS, PORT_ID, REPRESENTED_PORT, JUMP, DROP, MARK and SET_TAG.
437 - Policy actions of RSS for green and yellow should have the same configuration except queues.
438 - Policy with RSS/queue action is not supported when ``dv_xmeta_en`` enabled.
439 - meter profile packet mode is supported.
440 - meter profiles of RFC2697, RFC2698 and RFC4115 are supported.
444 - Integrity offload is enabled for **ConnectX-6** family.
445 - Verification bits provided by the hardware are ``l3_ok``, ``ipv4_csum_ok``, ``l4_ok``, ``l4_csum_ok``.
446 - ``level`` value 0 references outer headers.
447 - Multiple integrity items not supported in a single flow rule.
448 - Flow rule items supplied by application must explicitly specify network headers referred by integrity item.
449 For example, if integrity item mask sets ``l4_ok`` or ``l4_csum_ok`` bits, reference to L4 network header,
450 TCP or UDP, must be in the rule pattern as well::
452 flow create 0 ingress pattern integrity level is 0 value mask l3_ok value spec l3_ok / eth / ipv6 / end …
454 flow create 0 ingress pattern integrity level is 0 value mask l4_ok value spec 0 / eth / ipv4 proto is udp / end …
456 - Connection tracking:
458 - Cannot co-exist with ASO meter, ASO age action in a single flow rule.
459 - Flow rules insertion rate and memory consumption need more optimization.
461 - 4M connections maximum.
463 - Multi-thread flow insertion:
465 - In order to achieve best insertion rate, application should manage the flows per lcore.
466 - Better to disable memory reclaim by setting ``reclaim_mem_mode`` to 0 to accelerate the flow object allocation and release with cache.
470 - TXQ affinity subjects to HW hash once enabled.
472 - Bonding under socket direct mode
478 - CQE timestamp field width is limited by hardware to 63 bits, MSB is zero.
479 - In the free-running mode the timestamp counter is reset on power on
480 and 63-bit value provides over 1800 years of uptime till overflow.
481 - In the real-time mode
482 (configurable with ``REAL_TIME_CLOCK_ENABLE`` firmware settings),
483 the timestamp presents the nanoseconds elapsed since 01-Jan-1970,
484 hardware timestamp overflow will happen on 19-Jan-2038
485 (0x80000000 seconds since 01-Jan-1970).
486 - The send scheduling is based on timestamps
487 from the reference "Clock Queue" completions,
488 the scheduled send timestamps should not be specified with non-zero MSB.
493 MLX5 supports various methods to report statistics:
495 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.
497 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.
499 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.
505 See :ref:`mlx5 common compilation <mlx5_common_compilation>`.
511 Environment Configuration
512 ~~~~~~~~~~~~~~~~~~~~~~~~~
514 See :ref:`mlx5 common configuration <mlx5_common_env>`.
516 Firmware configuration
517 ~~~~~~~~~~~~~~~~~~~~~~
519 See :ref:`mlx5_firmware_config` guide.
524 Please refer to :ref:`mlx5 common options <mlx5_common_driver_options>`
525 for an additional list of options shared with other mlx5 drivers.
527 - ``rxq_cqe_comp_en`` parameter [int]
529 A nonzero value enables the compression of CQE on RX side. This feature
530 allows to save PCI bandwidth and improve performance. Enabled by default.
531 Different compression formats are supported in order to achieve the best
532 performance for different traffic patterns. Default format depends on
533 Multi-Packet Rx queue configuration: Hash RSS format is used in case
534 MPRQ is disabled, Checksum format is used in case MPRQ is enabled.
536 Specifying 2 as a ``rxq_cqe_comp_en`` value selects Flow Tag format for
537 better compression rate in case of RTE Flow Mark traffic.
538 Specifying 3 as a ``rxq_cqe_comp_en`` value selects Checksum format.
539 Specifying 4 as a ``rxq_cqe_comp_en`` value selects L3/L4 Header format for
540 better compression rate in case of mixed TCP/UDP and IPv4/IPv6 traffic.
541 CQE compression format selection requires DevX to be enabled. If there is
542 no DevX enabled/supported the value is reset to 1 by default.
546 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
547 ConnectX-6 Lx, BlueField and BlueField-2.
548 - POWER9 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
549 ConnectX-6 Lx, BlueField and BlueField-2.
551 - ``rxq_pkt_pad_en`` parameter [int]
553 A nonzero value enables padding Rx packet to the size of cacheline on PCI
554 transaction. This feature would waste PCI bandwidth but could improve
555 performance by avoiding partial cacheline write which may cause costly
556 read-modify-copy in memory transaction on some architectures. Disabled by
561 - x86_64 with ConnectX-4, ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
562 ConnectX-6 Lx, BlueField and BlueField-2.
563 - POWER8 and ARMv8 with ConnectX-4 Lx, ConnectX-5, ConnectX-6, ConnectX-6 Dx,
564 ConnectX-6 Lx, BlueField and BlueField-2.
566 - ``delay_drop`` parameter [int]
568 Bitmask value for the Rx queue delay drop attribute. Bit 0 is used for the
569 standard Rx queue and bit 1 is used for the hairpin Rx queue. By default, the
570 delay drop is disabled for all Rx queues. It will be ignored if the port does
571 not support the attribute even if it is enabled explicitly.
573 The packets being received will not be dropped immediately when the WQEs are
574 exhausted in a Rx queue with delay drop enabled.
576 A timeout value is set in the driver to control the waiting time before
577 dropping a packet. Once the timer is expired, the delay drop will be
578 deactivated for all the Rx queues with this feature enable. To re-activate
579 it, a rearming is needed and it is part of the kernel driver starting from
582 To enable / disable the delay drop rearming, the private flag ``dropless_rq``
583 can be set and queried via ethtool:
585 - ethtool --set-priv-flags <netdev> dropless_rq on (/ off)
586 - ethtool --show-priv-flags <netdev>
588 The configuration flag is global per PF and can only be set on the PF, once
589 it is on, all the VFs', SFs' and representors' Rx queues will share the timer
592 - ``mprq_en`` parameter [int]
594 A nonzero value enables configuring Multi-Packet Rx queues. Rx queue is
595 configured as Multi-Packet RQ if the total number of Rx queues is
596 ``rxqs_min_mprq`` or more. Disabled by default.
598 Multi-Packet Rx Queue (MPRQ a.k.a Striding RQ) can further save PCIe bandwidth
599 by posting a single large buffer for multiple packets. Instead of posting a
600 buffers per a packet, one large buffer is posted in order to receive multiple
601 packets on the buffer. A MPRQ buffer consists of multiple fixed-size strides
602 and each stride receives one packet. MPRQ can improve throughput for
603 small-packet traffic.
605 When MPRQ is enabled, MTU can be larger than the size of
606 user-provided mbuf even if RTE_ETH_RX_OFFLOAD_SCATTER isn't enabled. PMD will
607 configure large stride size enough to accommodate MTU as long as
608 device allows. Note that this can waste system memory compared to enabling Rx
609 scatter and multi-segment packet.
611 - ``mprq_log_stride_num`` parameter [int]
613 Log 2 of the number of strides for Multi-Packet Rx queue. Configuring more
614 strides can reduce PCIe traffic further. If configured value is not in the
615 range of device capability, the default value will be set with a warning
616 message. The default value is 4 which is 16 strides per a buffer, valid only
617 if ``mprq_en`` is set.
619 The size of Rx queue should be bigger than the number of strides.
621 - ``mprq_log_stride_size`` parameter [int]
623 Log 2 of the size of a stride for Multi-Packet Rx queue. Configuring a smaller
624 stride size can save some memory and reduce probability of a depletion of all
625 available strides due to unreleased packets by an application. If configured
626 value is not in the range of device capability, the default value will be set
627 with a warning message. The default value is 11 which is 2048 bytes per a
628 stride, valid only if ``mprq_en`` is set. With ``mprq_log_stride_size`` set
629 it is possible for a packet to span across multiple strides. This mode allows
630 support of jumbo frames (9K) with MPRQ. The memcopy of some packets (or part
631 of a packet if Rx scatter is configured) may be required in case there is no
632 space left for a head room at the end of a stride which incurs some
635 - ``mprq_max_memcpy_len`` parameter [int]
637 The maximum length of packet to memcpy in case of Multi-Packet Rx queue. Rx
638 packet is mem-copied to a user-provided mbuf if the size of Rx packet is less
639 than or equal to this parameter. Otherwise, PMD will attach the Rx packet to
640 the mbuf by external buffer attachment - ``rte_pktmbuf_attach_extbuf()``.
641 A mempool for external buffers will be allocated and managed by PMD. If Rx
642 packet is externally attached, ol_flags field of the mbuf will have
643 RTE_MBUF_F_EXTERNAL and this flag must be preserved. ``RTE_MBUF_HAS_EXTBUF()``
644 checks the flag. The default value is 128, valid only if ``mprq_en`` is set.
646 - ``rxqs_min_mprq`` parameter [int]
648 Configure Rx queues as Multi-Packet RQ if the total number of Rx queues is
649 greater or equal to this value. The default value is 12, valid only if
652 - ``txq_inline`` parameter [int]
654 Amount of data to be inlined during TX operations. This parameter is
655 deprecated and converted to the new parameter ``txq_inline_max`` providing
656 partial compatibility.
658 - ``txqs_min_inline`` parameter [int]
660 Enable inline data send only when the number of TX queues is greater or equal
663 This option should be used in combination with ``txq_inline_max`` and
664 ``txq_inline_mpw`` below and does not affect ``txq_inline_min`` settings above.
666 If this option is not specified the default value 16 is used for BlueField
667 and 8 for other platforms
669 The data inlining consumes the CPU cycles, so this option is intended to
670 auto enable inline data if we have enough Tx queues, which means we have
671 enough CPU cores and PCI bandwidth is getting more critical and CPU
672 is not supposed to be bottleneck anymore.
674 The copying data into WQE improves latency and can improve PPS performance
675 when PCI back pressure is detected and may be useful for scenarios involving
676 heavy traffic on many queues.
678 Because additional software logic is necessary to handle this mode, this
679 option should be used with care, as it may lower performance when back
680 pressure is not expected.
682 If inline data are enabled it may affect the maximal size of Tx queue in
683 descriptors because the inline data increase the descriptor size and
684 queue size limits supported by hardware may be exceeded.
686 - ``txq_inline_min`` parameter [int]
688 Minimal amount of data to be inlined into WQE during Tx operations. NICs
689 may require this minimal data amount to operate correctly. The exact value
690 may depend on NIC operation mode, requested offloads, etc. It is strongly
691 recommended to omit this parameter and use the default values. Anyway,
692 applications using this parameter should take into consideration that
693 specifying an inconsistent value may prevent the NIC from sending packets.
695 If ``txq_inline_min`` key is present the specified value (may be aligned
696 by the driver in order not to exceed the limits and provide better descriptor
697 space utilization) will be used by the driver and it is guaranteed that
698 requested amount of data bytes are inlined into the WQE beside other inline
699 settings. This key also may update ``txq_inline_max`` value (default
700 or specified explicitly in devargs) to reserve the space for inline data.
702 If ``txq_inline_min`` key is not present, the value may be queried by the
703 driver from the NIC via DevX if this feature is available. If there is no DevX
704 enabled/supported the value 18 (supposing L2 header including VLAN) is set
705 for ConnectX-4 and ConnectX-4 Lx, and 0 is set by default for ConnectX-5
706 and newer NICs. If packet is shorter the ``txq_inline_min`` value, the entire
709 For ConnectX-4 NIC, driver does not allow specifying value below 18
710 (minimal L2 header, including VLAN), error will be raised.
712 For ConnectX-4 Lx NIC, it is allowed to specify values below 18, but
713 it is not recommended and may prevent NIC from sending packets over
716 For ConnectX-4 and ConnectX-4 Lx NICs, automatically configured value
717 is insufficient for some traffic, because they require at least all L2 headers
718 to be inlined. For example, Q-in-Q adds 4 bytes to default 18 bytes
719 of Ethernet and VLAN, thus ``txq_inline_min`` must be set to 22.
720 MPLS would add 4 bytes per label. Final value must account for all possible
721 L2 encapsulation headers used in particular environment.
723 Please, note, this minimal data inlining disengages eMPW feature (Enhanced
724 Multi-Packet Write), because last one does not support partial packet inlining.
725 This is not very critical due to minimal data inlining is mostly required
726 by ConnectX-4 and ConnectX-4 Lx, these NICs do not support eMPW feature.
728 - ``txq_inline_max`` parameter [int]
730 Specifies the maximal packet length to be completely inlined into WQE
731 Ethernet Segment for ordinary SEND method. If packet is larger than specified
732 value, the packet data won't be copied by the driver at all, data buffer
733 is addressed with a pointer. If packet length is less or equal all packet
734 data will be copied into WQE. This may improve PCI bandwidth utilization for
735 short packets significantly but requires the extra CPU cycles.
737 The data inline feature is controlled by number of Tx queues, if number of Tx
738 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
739 is engaged, if there are not enough Tx queues (which means not enough CPU cores
740 and CPU resources are scarce), data inline is not performed by the driver.
741 Assigning ``txqs_min_inline`` with zero always enables the data inline.
743 The default ``txq_inline_max`` value is 290. The specified value may be adjusted
744 by the driver in order not to exceed the limit (930 bytes) and to provide better
745 WQE space filling without gaps, the adjustment is reflected in the debug log.
746 Also, the default value (290) may be decreased in run-time if the large transmit
747 queue size is requested and hardware does not support enough descriptor
748 amount, in this case warning is emitted. If ``txq_inline_max`` key is
749 specified and requested inline settings can not be satisfied then error
752 - ``txq_inline_mpw`` parameter [int]
754 Specifies the maximal packet length to be completely inlined into WQE for
755 Enhanced MPW method. If packet is large the specified value, the packet data
756 won't be copied, and data buffer is addressed with pointer. If packet length
757 is less or equal, all packet data will be copied into WQE. This may improve PCI
758 bandwidth utilization for short packets significantly but requires the extra
761 The data inline feature is controlled by number of TX queues, if number of Tx
762 queues is larger than ``txqs_min_inline`` key parameter, the inline feature
763 is engaged, if there are not enough Tx queues (which means not enough CPU cores
764 and CPU resources are scarce), data inline is not performed by the driver.
765 Assigning ``txqs_min_inline`` with zero always enables the data inline.
767 The default ``txq_inline_mpw`` value is 268. The specified value may be adjusted
768 by the driver in order not to exceed the limit (930 bytes) and to provide better
769 WQE space filling without gaps, the adjustment is reflected in the debug log.
770 Due to multiple packets may be included to the same WQE with Enhanced Multi
771 Packet Write Method and overall WQE size is limited it is not recommended to
772 specify large values for the ``txq_inline_mpw``. Also, the default value (268)
773 may be decreased in run-time if the large transmit queue size is requested
774 and hardware does not support enough descriptor amount, in this case warning
775 is emitted. If ``txq_inline_mpw`` key is specified and requested inline
776 settings can not be satisfied then error will be raised.
778 - ``txqs_max_vec`` parameter [int]
780 Enable vectorized Tx only when the number of TX queues is less than or
781 equal to this value. This parameter is deprecated and ignored, kept
782 for compatibility issue to not prevent driver from probing.
784 - ``txq_mpw_hdr_dseg_en`` parameter [int]
786 A nonzero value enables including two pointers in the first block of TX
787 descriptor. The parameter is deprecated and ignored, kept for compatibility
790 - ``txq_max_inline_len`` parameter [int]
792 Maximum size of packet to be inlined. This limits the size of packet to
793 be inlined. If the size of a packet is larger than configured value, the
794 packet isn't inlined even though there's enough space remained in the
795 descriptor. Instead, the packet is included with pointer. This parameter
796 is deprecated and converted directly to ``txq_inline_mpw`` providing full
797 compatibility. Valid only if eMPW feature is engaged.
799 - ``txq_mpw_en`` parameter [int]
801 A nonzero value enables Enhanced Multi-Packet Write (eMPW) for ConnectX-5,
802 ConnectX-6, ConnectX-6 Dx, ConnectX-6 Lx, BlueField, BlueField-2.
803 eMPW allows the Tx burst function to pack up multiple packets
804 in a single descriptor session in order to save PCI bandwidth
805 and improve performance at the cost of a slightly higher CPU usage.
806 When ``txq_inline_mpw`` is set along with ``txq_mpw_en``,
807 Tx burst function copies entire packet data on to Tx descriptor
808 instead of including pointer of packet.
810 The Enhanced Multi-Packet Write feature is enabled by default if NIC supports
811 it, can be disabled by explicit specifying 0 value for ``txq_mpw_en`` option.
812 Also, if minimal data inlining is requested by non-zero ``txq_inline_min``
813 option or reported by the NIC, the eMPW feature is disengaged.
815 - ``tx_db_nc`` parameter [int]
817 This parameter name is deprecated and ignored.
818 The new name for this parameter is ``sq_db_nc``.
819 See :ref:`common driver options <mlx5_common_driver_options>`.
821 - ``tx_pp`` parameter [int]
823 If a nonzero value is specified the driver creates all necessary internal
824 objects to provide accurate packet send scheduling on mbuf timestamps.
825 The positive value specifies the scheduling granularity in nanoseconds,
826 the packet send will be accurate up to specified digits. The allowed range is
827 from 500 to 1 million of nanoseconds. The negative value specifies the module
828 of granularity and engages the special test mode the check the schedule rate.
829 By default (if the ``tx_pp`` is not specified) send scheduling on timestamps
832 Starting with ConnectX-7 the capability to schedule traffic directly
833 on timestamp specified in descriptor is provided,
834 no extra objects are needed anymore and scheduling capability
835 is advertised and handled regardless ``tx_pp`` parameter presence.
837 - ``tx_skew`` parameter [int]
839 The parameter adjusts the send packet scheduling on timestamps and represents
840 the average delay between beginning of the transmitting descriptor processing
841 by the hardware and appearance of actual packet data on the wire. The value
842 should be provided in nanoseconds and is valid only if ``tx_pp`` parameter is
843 specified. The default value is zero.
845 - ``tx_vec_en`` parameter [int]
847 A nonzero value enables Tx vector on ConnectX-5, ConnectX-6, ConnectX-6 Dx,
848 ConnectX-6 Lx, BlueField and BlueField-2 NICs
849 if the number of global Tx queues on the port is less than ``txqs_max_vec``.
850 The parameter is deprecated and ignored.
852 - ``rx_vec_en`` parameter [int]
854 A nonzero value enables Rx vector if the port is not configured in
855 multi-segment otherwise this parameter is ignored.
859 - ``vf_nl_en`` parameter [int]
861 A nonzero value enables Netlink requests from the VF to add/remove MAC
862 addresses or/and enable/disable promiscuous/all multicast on the Netdevice.
863 Otherwise the relevant configuration must be run with Linux iproute2 tools.
864 This is a prerequisite to receive this kind of traffic.
866 Enabled by default, valid only on VF devices ignored otherwise.
868 - ``l3_vxlan_en`` parameter [int]
870 A nonzero value allows L3 VXLAN and VXLAN-GPE flow creation. To enable
871 L3 VXLAN or VXLAN-GPE, users has to configure firmware and enable this
872 parameter. This is a prerequisite to receive this kind of traffic.
876 - ``dv_xmeta_en`` parameter [int]
878 A nonzero value enables extensive flow metadata support if device is
879 capable and driver supports it. This can enable extensive support of
880 ``MARK`` and ``META`` item of ``rte_flow``. The newly introduced
881 ``SET_TAG`` and ``SET_META`` actions do not depend on ``dv_xmeta_en``.
883 There are some possible configurations, depending on parameter value:
885 - 0, this is default value, defines the legacy mode, the ``MARK`` and
886 ``META`` related actions and items operate only within NIC Tx and
887 NIC Rx steering domains, no ``MARK`` and ``META`` information crosses
888 the domain boundaries. The ``MARK`` item is 24 bits wide, the ``META``
889 item is 32 bits wide and match supported on egress only.
891 - 1, 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 ``MARK`` item is 24 bits wide, the ``META`` item width
895 depends on kernel and firmware configurations and might be 0, 16 or
896 32 bits. Within NIC Tx domain ``META`` data width is 32 bits for
897 compatibility, the actual width of data transferred to the FDB domain
898 depends on kernel configuration and may be vary. The actual supported
899 width can be retrieved in runtime by series of rte_flow_validate()
902 - 2, this engages extensive metadata mode, the ``MARK`` and ``META``
903 related actions and items operate within all supported steering domains,
904 including FDB, ``MARK`` and ``META`` information may cross the domain
905 boundaries. The ``META`` item is 32 bits wide, the ``MARK`` item width
906 depends on kernel and firmware configurations and might be 0, 16 or
907 24 bits. The actual supported width can be retrieved in runtime by
908 series of rte_flow_validate() trials.
910 - 3, this engages tunnel offload mode. In E-Switch configuration, that
911 mode implicitly activates ``dv_xmeta_en=1``.
913 +------+-----------+-----------+-------------+-------------+
914 | Mode | ``MARK`` | ``META`` | ``META`` Tx | FDB/Through |
915 +======+===========+===========+=============+=============+
916 | 0 | 24 bits | 32 bits | 32 bits | no |
917 +------+-----------+-----------+-------------+-------------+
918 | 1 | 24 bits | vary 0-32 | 32 bits | yes |
919 +------+-----------+-----------+-------------+-------------+
920 | 2 | vary 0-24 | 32 bits | 32 bits | yes |
921 +------+-----------+-----------+-------------+-------------+
923 If there is no E-Switch configuration the ``dv_xmeta_en`` parameter is
924 ignored and the device is configured to operate in legacy mode (0).
926 Disabled by default (set to 0).
928 The Direct Verbs/Rules (engaged with ``dv_flow_en`` = 1) supports all
929 of the extensive metadata features. The legacy Verbs supports FLAG and
930 MARK metadata actions over NIC Rx steering domain only.
932 Setting META value to zero in flow action means there is no item provided
933 and receiving datapath will not report in mbufs the metadata are present.
934 Setting MARK value to zero in flow action means the zero FDIR ID value
935 will be reported on packet receiving.
937 For the MARK action the last 16 values in the full range are reserved for
938 internal PMD purposes (to emulate FLAG action). The valid range for the
939 MARK action values is 0-0xFFEF for the 16-bit mode and 0-0xFFFFEF
940 for the 24-bit mode, the flows with the MARK action value outside
941 the specified range will be rejected.
943 - ``dv_flow_en`` parameter [int]
945 A nonzero value enables the DV flow steering assuming it is supported
946 by the driver (RDMA Core library version is rdma-core-24.0 or higher).
948 Enabled by default if supported.
950 - ``dv_esw_en`` parameter [int]
952 A nonzero value enables E-Switch using Direct Rules.
954 Enabled by default if supported.
956 - ``lacp_by_user`` parameter [int]
958 A nonzero value enables the control of LACP traffic by the user application.
959 When a bond exists in the driver, by default it should be managed by the
960 kernel and therefore LACP traffic should be steered to the kernel.
961 If this devarg is set to 1 it will allow the user to manage the bond by
962 itself and not steer LACP traffic to the kernel.
964 Disabled by default (set to 0).
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 - ``decap_en`` parameter [int]
1038 Some devices do not support FCS (frame checksum) scattering for
1039 tunnel-decapsulated packets.
1040 If set to 0, this option forces the FCS feature and rejects tunnel
1041 decapsulation in the flow engine for such devices.
1043 By default, the PMD will set this value to 1.
1045 - ``allow_duplicate_pattern`` parameter [int]
1047 There are two options to choose:
1049 - 0. Prevent insertion of rules with the same pattern items on non-root table.
1050 In this case, only the first rule is inserted and the following rules are
1051 rejected and error code EEXIST is returned.
1053 - 1. Allow insertion of rules with the same pattern items.
1054 In this case, all rules are inserted but only the first rule takes effect,
1055 the next rule takes effect only if the previous rules are deleted.
1057 By default, the PMD will set this value to 1.
1063 The following Mellanox device families are supported by the same mlx5 driver:
1075 Below are detailed device names:
1077 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX4111A-XCAT (1x10G)
1078 * Mellanox\ |reg| ConnectX\ |reg|-4 10G MCX412A-XCAT (2x10G)
1079 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX4111A-ACAT (1x25G)
1080 * Mellanox\ |reg| ConnectX\ |reg|-4 25G MCX412A-ACAT (2x25G)
1081 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX413A-BCAT (1x40G)
1082 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX4131A-BCAT (1x40G)
1083 * Mellanox\ |reg| ConnectX\ |reg|-4 40G MCX415A-BCAT (1x40G)
1084 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX413A-GCAT (1x50G)
1085 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX4131A-GCAT (1x50G)
1086 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX414A-BCAT (2x50G)
1087 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-GCAT (1x50G)
1088 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-BCAT (2x50G)
1089 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX416A-GCAT (2x50G)
1090 * Mellanox\ |reg| ConnectX\ |reg|-4 50G MCX415A-CCAT (1x100G)
1091 * Mellanox\ |reg| ConnectX\ |reg|-4 100G MCX416A-CCAT (2x100G)
1092 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4111A-XCAT (1x10G)
1093 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 10G MCX4121A-XCAT (2x10G)
1094 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4111A-ACAT (1x25G)
1095 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 25G MCX4121A-ACAT (2x25G)
1096 * Mellanox\ |reg| ConnectX\ |reg|-4 Lx 40G MCX4131A-BCAT (1x40G)
1097 * Mellanox\ |reg| ConnectX\ |reg|-5 100G MCX556A-ECAT (2x100G)
1098 * Mellanox\ |reg| ConnectX\ |reg|-5 Ex EN 100G MCX516A-CDAT (2x100G)
1099 * Mellanox\ |reg| ConnectX\ |reg|-6 200G MCX654106A-HCAT (2x200G)
1100 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 100G MCX623106AN-CDAT (2x100G)
1101 * Mellanox\ |reg| ConnectX\ |reg|-6 Dx EN 200G MCX623105AN-VDAT (1x200G)
1102 * Mellanox\ |reg| ConnectX\ |reg|-6 Lx EN 25G MCX631102AN-ADAT (2x25G)
1108 See :ref:`mlx5_sub_function`.
1110 Sub-Function representor support
1111 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1113 A SF netdev supports E-Switch representation offload
1114 similar to PF and VF representors.
1115 Use <sfnum> to probe SF representor::
1117 testpmd> port attach <PCI_BDF>,representor=sf<sfnum>,dv_flow_en=1
1123 1. Configure aggressive CQE Zipping for maximum performance::
1125 mlxconfig -d <mst device> s CQE_COMPRESSION=1
1127 To set it back to the default CQE Zipping mode use::
1129 mlxconfig -d <mst device> s CQE_COMPRESSION=0
1131 2. In case of virtualization:
1133 - Make sure that hypervisor kernel is 3.16 or newer.
1134 - Configure boot with ``iommu=pt``.
1135 - Use 1G huge pages.
1136 - Make sure to allocate a VM on huge pages.
1137 - Make sure to set CPU pinning.
1139 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
1140 for better performance. For VMs, verify that the right CPU
1141 and NUMA node are pinned according to the above. Run::
1143 lstopo-no-graphics --merge
1145 to identify the NUMA node to which the PCIe adapter is connected.
1147 4. If more than one adapter is used, and root complex capabilities allow
1148 to put both adapters on the same NUMA node without PCI bandwidth degradation,
1149 it is recommended to locate both adapters on the same NUMA node.
1150 This in order to forward packets from one to the other without
1151 NUMA performance penalty.
1153 5. Disable pause frames::
1155 ethtool -A <netdev> rx off tx off
1157 6. Verify IO non-posted prefetch is disabled by default. This can be checked
1158 via the BIOS configuration. Please contact you server provider for more
1159 information about the settings.
1163 On some machines, depends on the machine integrator, it is beneficial
1164 to set the PCI max read request parameter to 1K. This can be
1165 done in the following way:
1167 To query the read request size use::
1169 setpci -s <NIC PCI address> 68.w
1171 If the output is different than 3XXX, set it by::
1173 setpci -s <NIC PCI address> 68.w=3XXX
1175 The XXX can be different on different systems. Make sure to configure
1176 according to the setpci output.
1178 7. To minimize overhead of searching Memory Regions:
1180 - '--socket-mem' is recommended to pin memory by predictable amount.
1181 - Configure per-lcore cache when creating Mempools for packet buffer.
1182 - Refrain from dynamically allocating/freeing memory in run-time.
1187 There are multiple Rx burst functions with different advantages and limitations.
1189 .. table:: Rx burst functions
1191 +-------------------+------------------------+---------+-----------------+------+-------+
1192 || Function Name || Enabler || Scatter|| Error Recovery || CQE || Large|
1193 | | | | || comp|| MTU |
1194 +===================+========================+=========+=================+======+=======+
1195 | rx_burst | rx_vec_en=0 | Yes | Yes | Yes | Yes |
1196 +-------------------+------------------------+---------+-----------------+------+-------+
1197 | rx_burst_vec | rx_vec_en=1 (default) | No | if CQE comp off | Yes | No |
1198 +-------------------+------------------------+---------+-----------------+------+-------+
1199 | rx_burst_mprq || mprq_en=1 | No | Yes | Yes | Yes |
1200 | || RxQs >= rxqs_min_mprq | | | | |
1201 +-------------------+------------------------+---------+-----------------+------+-------+
1202 | rx_burst_mprq_vec || rx_vec_en=1 (default) | No | if CQE comp off | Yes | Yes |
1203 | || mprq_en=1 | | | | |
1204 | || RxQs >= rxqs_min_mprq | | | | |
1205 +-------------------+------------------------+---------+-----------------+------+-------+
1207 .. _mlx5_offloads_support:
1209 Supported hardware offloads
1210 ---------------------------
1212 .. table:: Minimal SW/HW versions for queue offloads
1214 ============== ===== ===== ========= ===== ========== =============
1215 Offload DPDK Linux rdma-core OFED firmware hardware
1216 ============== ===== ===== ========= ===== ========== =============
1217 common base 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1218 checksums 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1219 Rx timestamp 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1220 TSO 17.11 4.14 16 4.2-1 12.21.1000 ConnectX-4
1221 LRO 19.08 N/A N/A 4.6-4 16.25.6406 ConnectX-5
1222 Tx scheduling 20.08 N/A N/A 5.1-2 22.28.2006 ConnectX-6 Dx
1223 Buffer Split 20.11 N/A N/A 5.1-2 16.28.2006 ConnectX-5
1224 ============== ===== ===== ========= ===== ========== =============
1226 .. table:: Minimal SW/HW versions for rte_flow offloads
1228 +-----------------------+-----------------+-----------------+
1229 | Offload | with E-Switch | with NIC |
1230 +=======================+=================+=================+
1231 | Count | | DPDK 19.05 | | DPDK 19.02 |
1232 | | | OFED 4.6 | | OFED 4.6 |
1233 | | | rdma-core 24 | | rdma-core 23 |
1234 | | | ConnectX-5 | | ConnectX-5 |
1235 +-----------------------+-----------------+-----------------+
1236 | Drop | | DPDK 19.05 | | DPDK 18.11 |
1237 | | | OFED 4.6 | | OFED 4.5 |
1238 | | | rdma-core 24 | | rdma-core 23 |
1239 | | | ConnectX-5 | | ConnectX-4 |
1240 +-----------------------+-----------------+-----------------+
1241 | Queue / RSS | | | | DPDK 18.11 |
1242 | | | N/A | | OFED 4.5 |
1243 | | | | | rdma-core 23 |
1244 | | | | | ConnectX-4 |
1245 +-----------------------+-----------------+-----------------+
1246 | Shared action | | | | |
1247 | | | :numref:`sact`| | :numref:`sact`|
1250 +-----------------------+-----------------+-----------------+
1251 | | VLAN | | DPDK 19.11 | | DPDK 19.11 |
1252 | | (of_pop_vlan / | | OFED 4.7-1 | | OFED 4.7-1 |
1253 | | of_push_vlan / | | ConnectX-5 | | ConnectX-5 |
1254 | | of_set_vlan_pcp / | | | | |
1255 | | of_set_vlan_vid) | | | | |
1256 +-----------------------+-----------------+-----------------+
1257 | | VLAN | | DPDK 21.05 | | |
1258 | | ingress and / | | OFED 5.3 | | N/A |
1259 | | of_push_vlan / | | ConnectX-6 Dx | | |
1260 +-----------------------+-----------------+-----------------+
1261 | | VLAN | | DPDK 21.05 | | |
1262 | | egress and / | | OFED 5.3 | | N/A |
1263 | | of_pop_vlan / | | ConnectX-6 Dx | | |
1264 +-----------------------+-----------------+-----------------+
1265 | Encapsulation | | DPDK 19.05 | | DPDK 19.02 |
1266 | (VXLAN / NVGRE / RAW) | | OFED 4.7-1 | | OFED 4.6 |
1267 | | | rdma-core 24 | | rdma-core 23 |
1268 | | | ConnectX-5 | | ConnectX-5 |
1269 +-----------------------+-----------------+-----------------+
1270 | Encapsulation | | DPDK 19.11 | | DPDK 19.11 |
1271 | GENEVE | | OFED 4.7-3 | | OFED 4.7-3 |
1272 | | | rdma-core 27 | | rdma-core 27 |
1273 | | | ConnectX-5 | | ConnectX-5 |
1274 +-----------------------+-----------------+-----------------+
1275 | Tunnel Offload | | DPDK 20.11 | | DPDK 20.11 |
1276 | | | OFED 5.1-2 | | OFED 5.1-2 |
1277 | | | rdma-core 32 | | N/A |
1278 | | | ConnectX-5 | | ConnectX-5 |
1279 +-----------------------+-----------------+-----------------+
1280 | | Header rewrite | | DPDK 19.05 | | DPDK 19.02 |
1281 | | (set_ipv4_src / | | OFED 4.7-1 | | OFED 4.7-1 |
1282 | | set_ipv4_dst / | | rdma-core 24 | | rdma-core 24 |
1283 | | set_ipv6_src / | | ConnectX-5 | | ConnectX-5 |
1284 | | set_ipv6_dst / | | | | |
1285 | | set_tp_src / | | | | |
1286 | | set_tp_dst / | | | | |
1287 | | dec_ttl / | | | | |
1288 | | set_ttl / | | | | |
1289 | | set_mac_src / | | | | |
1290 | | set_mac_dst) | | | | |
1291 +-----------------------+-----------------+-----------------+
1292 | | Header rewrite | | DPDK 20.02 | | DPDK 20.02 |
1293 | | (set_dscp) | | OFED 5.0 | | OFED 5.0 |
1294 | | | | rdma-core 24 | | rdma-core 24 |
1295 | | | | ConnectX-5 | | ConnectX-5 |
1296 +-----------------------+-----------------+-----------------+
1297 | Jump | | DPDK 19.05 | | DPDK 19.02 |
1298 | | | OFED 4.7-1 | | OFED 4.7-1 |
1299 | | | rdma-core 24 | | N/A |
1300 | | | ConnectX-5 | | ConnectX-5 |
1301 +-----------------------+-----------------+-----------------+
1302 | Mark / Flag | | DPDK 19.05 | | DPDK 18.11 |
1303 | | | OFED 4.6 | | OFED 4.5 |
1304 | | | rdma-core 24 | | rdma-core 23 |
1305 | | | ConnectX-5 | | ConnectX-4 |
1306 +-----------------------+-----------------+-----------------+
1307 | Meta data | | DPDK 19.11 | | DPDK 19.11 |
1308 | | | OFED 4.7-3 | | OFED 4.7-3 |
1309 | | | rdma-core 26 | | rdma-core 26 |
1310 | | | ConnectX-5 | | ConnectX-5 |
1311 +-----------------------+-----------------+-----------------+
1312 | Port ID | | DPDK 19.05 | | N/A |
1313 | | | OFED 4.7-1 | | N/A |
1314 | | | rdma-core 24 | | N/A |
1315 | | | ConnectX-5 | | N/A |
1316 +-----------------------+-----------------+-----------------+
1317 | Hairpin | | | | DPDK 19.11 |
1318 | | | N/A | | OFED 4.7-3 |
1319 | | | | | rdma-core 26 |
1320 | | | | | ConnectX-5 |
1321 +-----------------------+-----------------+-----------------+
1322 | 2-port Hairpin | | | | DPDK 20.11 |
1323 | | | N/A | | OFED 5.1-2 |
1325 | | | | | ConnectX-5 |
1326 +-----------------------+-----------------+-----------------+
1327 | Metering | | DPDK 19.11 | | DPDK 19.11 |
1328 | | | OFED 4.7-3 | | OFED 4.7-3 |
1329 | | | rdma-core 26 | | rdma-core 26 |
1330 | | | ConnectX-5 | | ConnectX-5 |
1331 +-----------------------+-----------------+-----------------+
1332 | ASO Metering | | DPDK 21.05 | | DPDK 21.05 |
1333 | | | OFED 5.3 | | OFED 5.3 |
1334 | | | rdma-core 33 | | rdma-core 33 |
1335 | | | ConnectX-6 Dx| | ConnectX-6 Dx |
1336 +-----------------------+-----------------+-----------------+
1337 | Metering Hierarchy | | DPDK 21.08 | | DPDK 21.08 |
1338 | | | OFED 5.3 | | OFED 5.3 |
1340 | | | ConnectX-6 Dx| | ConnectX-6 Dx |
1341 +-----------------------+-----------------+-----------------+
1342 | Sampling | | DPDK 20.11 | | DPDK 20.11 |
1343 | | | OFED 5.1-2 | | OFED 5.1-2 |
1344 | | | rdma-core 32 | | N/A |
1345 | | | ConnectX-5 | | ConnectX-5 |
1346 +-----------------------+-----------------+-----------------+
1347 | Encapsulation | | DPDK 21.02 | | DPDK 21.02 |
1348 | GTP PSC | | OFED 5.2 | | OFED 5.2 |
1349 | | | rdma-core 35 | | rdma-core 35 |
1350 | | | ConnectX-6 Dx| | ConnectX-6 Dx |
1351 +-----------------------+-----------------+-----------------+
1352 | Encapsulation | | DPDK 21.02 | | DPDK 21.02 |
1353 | GENEVE TLV option | | OFED 5.2 | | OFED 5.2 |
1354 | | | rdma-core 34 | | rdma-core 34 |
1355 | | | ConnectX-6 Dx | | ConnectX-6 Dx |
1356 +-----------------------+-----------------+-----------------+
1357 | Modify Field | | DPDK 21.02 | | DPDK 21.02 |
1358 | | | OFED 5.2 | | OFED 5.2 |
1359 | | | rdma-core 35 | | rdma-core 35 |
1360 | | | ConnectX-5 | | ConnectX-5 |
1361 +-----------------------+-----------------+-----------------+
1362 | Connection tracking | | | | DPDK 21.05 |
1363 | | | N/A | | OFED 5.3 |
1364 | | | | | rdma-core 35 |
1365 | | | | | ConnectX-6 Dx |
1366 +-----------------------+-----------------+-----------------+
1368 .. table:: Minimal SW/HW versions for shared action offload
1371 +-----------------------+-----------------+-----------------+
1372 | Shared Action | with E-Switch | with NIC |
1373 +=======================+=================+=================+
1374 | RSS | | | | DPDK 20.11 |
1375 | | | N/A | | OFED 5.2 |
1376 | | | | | rdma-core 33 |
1377 | | | | | ConnectX-5 |
1378 +-----------------------+-----------------+-----------------+
1379 | Age | | DPDK 20.11 | | DPDK 20.11 |
1380 | | | OFED 5.2 | | OFED 5.2 |
1381 | | | rdma-core 32 | | rdma-core 32 |
1382 | | | ConnectX-6 Dx | | ConnectX-6 Dx |
1383 +-----------------------+-----------------+-----------------+
1384 | Count | | DPDK 21.05 | | DPDK 21.05 |
1385 | | | OFED 4.6 | | OFED 4.6 |
1386 | | | rdma-core 24 | | rdma-core 23 |
1387 | | | ConnectX-5 | | ConnectX-5 |
1388 +-----------------------+-----------------+-----------------+
1393 MARK and META items are interrelated with datapath - they might move from/to
1394 the applications in mbuf fields. Hence, zero value for these items has the
1395 special meaning - it means "no metadata are provided", not zero values are
1396 treated by applications and PMD as valid ones.
1398 Moreover in the flow engine domain the value zero is acceptable to match and
1399 set, and we should allow to specify zero values as rte_flow parameters for the
1400 META and MARK items and actions. In the same time zero mask has no meaning and
1401 should be rejected on validation stage.
1406 Flows are not cached in the driver.
1407 When stopping a device port, all the flows created on this port from the
1408 application will be flushed automatically in the background.
1409 After stopping the device port, all flows on this port become invalid and
1410 not represented in the system.
1411 All references to these flows held by the application should be discarded
1412 directly but neither destroyed nor flushed.
1414 The application should re-create the flows as required after the port restart.
1419 Compared to librte_net_mlx4 that implements a single RSS configuration per
1420 port, librte_net_mlx5 supports per-protocol RSS configuration.
1422 Since ``testpmd`` defaults to IP RSS mode and there is currently no
1423 command-line parameter to enable additional protocols (UDP and TCP as well
1424 as IP), the following commands must be entered from its CLI to get the same
1425 behavior as librte_net_mlx4::
1428 > port config all rss all
1434 This section demonstrates how to launch **testpmd** with Mellanox
1435 ConnectX-4/ConnectX-5/ConnectX-6/BlueField devices managed by librte_net_mlx5.
1437 #. Load the kernel modules::
1439 modprobe -a ib_uverbs mlx5_core mlx5_ib
1441 Alternatively if MLNX_OFED/MLNX_EN is fully installed, the following script
1444 /etc/init.d/openibd restart
1448 User space I/O kernel modules (uio and igb_uio) are not used and do
1449 not have to be loaded.
1451 #. Make sure Ethernet interfaces are in working order and linked to kernel
1452 verbs. Related sysfs entries should be present::
1454 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
1463 #. Optionally, retrieve their PCI bus addresses for to be used with the allow list::
1466 for intf in eth2 eth3 eth4 eth5;
1468 (cd "/sys/class/net/${intf}/device/" && pwd -P);
1471 sed -n 's,.*/\(.*\),-a \1,p'
1480 #. Request huge pages::
1482 dpdk-hugepages.py --setup 2G
1484 #. Start testpmd with basic parameters::
1486 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
1491 EAL: PCI device 0000:05:00.0 on NUMA socket 0
1492 EAL: probe driver: 15b3:1013 librte_net_mlx5
1493 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
1494 PMD: librte_net_mlx5: 1 port(s) detected
1495 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
1496 EAL: PCI device 0000:05:00.1 on NUMA socket 0
1497 EAL: probe driver: 15b3:1013 librte_net_mlx5
1498 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
1499 PMD: librte_net_mlx5: 1 port(s) detected
1500 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
1501 EAL: PCI device 0000:06:00.0 on NUMA socket 0
1502 EAL: probe driver: 15b3:1013 librte_net_mlx5
1503 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
1504 PMD: librte_net_mlx5: 1 port(s) detected
1505 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
1506 EAL: PCI device 0000:06:00.1 on NUMA socket 0
1507 EAL: probe driver: 15b3:1013 librte_net_mlx5
1508 PMD: librte_net_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
1509 PMD: librte_net_mlx5: 1 port(s) detected
1510 PMD: librte_net_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
1511 Interactive-mode selected
1512 Configuring Port 0 (socket 0)
1513 PMD: librte_net_mlx5: 0x8cba80: TX queues number update: 0 -> 2
1514 PMD: librte_net_mlx5: 0x8cba80: RX queues number update: 0 -> 2
1515 Port 0: E4:1D:2D:E7:0C:FE
1516 Configuring Port 1 (socket 0)
1517 PMD: librte_net_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
1518 PMD: librte_net_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
1519 Port 1: E4:1D:2D:E7:0C:FF
1520 Configuring Port 2 (socket 0)
1521 PMD: librte_net_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
1522 PMD: librte_net_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
1523 Port 2: E4:1D:2D:E7:0C:FA
1524 Configuring Port 3 (socket 0)
1525 PMD: librte_net_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
1526 PMD: librte_net_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
1527 Port 3: E4:1D:2D:E7:0C:FB
1528 Checking link statuses...
1529 Port 0 Link Up - speed 40000 Mbps - full-duplex
1530 Port 1 Link Up - speed 40000 Mbps - full-duplex
1531 Port 2 Link Up - speed 10000 Mbps - full-duplex
1532 Port 3 Link Up - speed 10000 Mbps - full-duplex
1539 This section demonstrates how to dump flows. Currently, it's possible to dump
1540 all flows with assistance of external tools.
1542 #. 2 ways to get flow raw file:
1544 - Using testpmd CLI:
1546 .. code-block:: console
1549 testpmd> flow dump <port> all <output_file>
1551 testpmd> flow dump <port> rule <rule_id> <output_file>
1553 - call rte_flow_dev_dump api:
1555 .. code-block:: console
1557 rte_flow_dev_dump(port, flow, file, NULL);
1559 #. Dump human-readable flows from raw file:
1561 Get flow parsing tool from: https://github.com/Mellanox/mlx_steering_dump
1563 .. code-block:: console
1565 mlx_steering_dump.py -f <output_file> -flowptr <flow_ptr>
1567 How to share a meter between ports in the same switch domain
1568 ------------------------------------------------------------
1570 This section demonstrates how to use the shared meter. A meter M can be created
1571 on port X and to be shared with a port Y on the same switch domain by the next way:
1573 .. code-block:: console
1575 flow create X ingress transfer pattern eth / port_id id is Y / end actions meter mtr_id M / end
1577 How to use meter hierarchy
1578 --------------------------
1580 This section demonstrates how to create and use a meter hierarchy.
1581 A termination meter M can be the policy green action of another termination meter N.
1582 The two meters are chained together as a chain. Using meter N in a flow will apply
1583 both the meters in hierarchy on that flow.
1585 .. code-block:: console
1587 add port meter policy 0 1 g_actions queue index 0 / end y_actions end r_actions drop / end
1588 create port meter 0 M 1 1 yes 0xffff 1 0
1589 add port meter policy 0 2 g_actions meter mtr_id M / end y_actions end r_actions drop / end
1590 create port meter 0 N 2 2 yes 0xffff 1 0
1591 flow create 0 ingress group 1 pattern eth / end actions meter mtr_id N / end