2 Copyright 2015 6WIND S.A.
4 Redistribution and use in source and binary forms, with or without
5 modification, are permitted provided that the following conditions
8 * Redistributions of source code must retain the above copyright
9 notice, this list of conditions and the following disclaimer.
10 * Redistributions in binary form must reproduce the above copyright
11 notice, this list of conditions and the following disclaimer in
12 the documentation and/or other materials provided with the
14 * Neither the name of 6WIND S.A. nor the names of its
15 contributors may be used to endorse or promote products derived
16 from this software without specific prior written permission.
18 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 The MLX5 poll mode driver library (**librte_pmd_mlx5**) provides support for
34 **Mellanox ConnectX-4** and **Mellanox ConnectX-4 Lx** families of
35 10/25/40/50/100 Gb/s adapters as well as their virtual functions (VF) in
38 Information and documentation about these adapters can be found on the
39 `Mellanox website <http://www.mellanox.com>`__. Help is also provided by the
40 `Mellanox community <http://community.mellanox.com/welcome>`__.
42 There is also a `section dedicated to this poll mode driver
43 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`__.
47 Due to external dependencies, this driver is disabled by default. It must
48 be enabled manually by setting ``CONFIG_RTE_LIBRTE_MLX5_PMD=y`` and
51 Implementation details
52 ----------------------
54 Besides its dependency on libibverbs (that implies libmlx5 and associated
55 kernel support), librte_pmd_mlx5 relies heavily on system calls for control
56 operations such as querying/updating the MTU and flow control parameters.
58 For security reasons and robustness, this driver only deals with virtual
59 memory addresses. The way resources allocations are handled by the kernel
60 combined with hardware specifications that allow it to handle virtual memory
61 addresses directly ensure that DPDK applications cannot access random
62 physical memory (or memory that does not belong to the current process).
64 This capability allows the PMD to coexist with kernel network interfaces
65 which remain functional, although they stop receiving unicast packets as
66 long as they share the same MAC address.
68 Enabling librte_pmd_mlx5 causes DPDK applications to be linked against
74 - Multiple TX and RX queues.
75 - Support for scattered TX and RX frames.
76 - IPv4, IPv6, TCPv4, TCPv6, UDPv4 and UDPv6 RSS on any number of queues.
77 - Several RSS hash keys, one for each flow type.
78 - Configurable RETA table.
79 - Support for multiple MAC addresses.
83 - RX CRC stripping configuration.
85 - Multicast promiscuous mode.
86 - Hardware checksum offloads.
87 - Flow director (RTE_FDIR_MODE_PERFECT, RTE_FDIR_MODE_PERFECT_MAC_VLAN and
89 - Secondary process TX is supported.
90 - KVM and VMware ESX SR-IOV modes are supported.
91 - RSS hash result is supported.
96 - Inner RSS for VXLAN frames is not supported yet.
97 - Port statistics through software counters only.
98 - Hardware checksum offloads for VXLAN inner header are not supported yet.
99 - Secondary process RX is not supported.
107 These options can be modified in the ``.config`` file.
109 - ``CONFIG_RTE_LIBRTE_MLX5_PMD`` (default **n**)
111 Toggle compilation of librte_pmd_mlx5 itself.
113 - ``CONFIG_RTE_LIBRTE_MLX5_DEBUG`` (default **n**)
115 Toggle debugging code and stricter compilation flags. Enabling this option
116 adds additional run-time checks and debugging messages at the cost of
119 - ``CONFIG_RTE_LIBRTE_MLX5_TX_MP_CACHE`` (default **8**)
121 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
122 which buffers are to be transmitted must be associated to memory regions
123 (MRs). This is a slow operation that must be cached.
125 This value is always 1 for RX queues since they use a single MP.
127 Environment variables
128 ~~~~~~~~~~~~~~~~~~~~~
130 - ``MLX5_PMD_ENABLE_PADDING``
132 Enables HW packet padding in PCI bus transactions.
134 When packet size is cache aligned and CRC stripping is enabled, 4 fewer
135 bytes are written to the PCI bus. Enabling padding makes such packets
138 In cases where PCI bandwidth is the bottleneck, padding can improve
141 This is disabled by default since this can also decrease performance for
142 unaligned packet sizes.
144 Run-time configuration
145 ~~~~~~~~~~~~~~~~~~~~~~
147 - librte_pmd_mlx5 brings kernel network interfaces up during initialization
148 because it is affected by their state. Forcing them down prevents packets
151 - **ethtool** operations on related kernel interfaces also affect the PMD.
153 - ``rxq_cqe_comp_en`` parameter [int]
155 A nonzero value enables the compression of CQE on RX side. This feature
156 allows to save PCI bandwidth and improve performance at the cost of a
157 slightly higher CPU usage. Enabled by default.
161 - x86_64 with ConnectX4 and ConnectX4 LX
162 - Power8 with ConnectX4 LX
164 - ``txq_inline`` parameter [int]
166 Amount of data to be inlined during TX operations. Improves latency.
167 Can improve PPS performance when PCI back pressure is detected and may be
168 useful for scenarios involving heavy traffic on many queues.
170 It is not enabled by default (set to 0) since the additional software
171 logic necessary to handle this mode can lower performance when back
172 pressure is not expected.
174 - ``txqs_min_inline`` parameter [int]
176 Enable inline send only when the number of TX queues is greater or equal
179 This option should be used in combination with ``txq_inline`` above.
181 - ``txq_mpw_en`` parameter [int]
183 A nonzero value enables multi-packet send. This feature allows the TX
184 burst function to pack up to five packets in two descriptors in order to
185 save PCI bandwidth and improve performance at the cost of a slightly
188 It is currently only supported on the ConnectX-4 Lx family of adapters.
194 This driver relies on external libraries and kernel drivers for resources
195 allocations and initialization. The following dependencies are not part of
196 DPDK and must be installed separately:
200 User space Verbs framework used by librte_pmd_mlx5. This library provides
201 a generic interface between the kernel and low-level user space drivers
204 It allows slow and privileged operations (context initialization, hardware
205 resources allocations) to be managed by the kernel and fast operations to
206 never leave user space.
210 Low-level user space driver library for Mellanox ConnectX-4 devices,
211 it is automatically loaded by libibverbs.
213 This library basically implements send/receive calls to the hardware
216 - **Kernel modules** (mlnx-ofed-kernel)
218 They provide the kernel-side Verbs API and low level device drivers that
219 manage actual hardware initialization and resources sharing with user
222 Unlike most other PMDs, these modules must remain loaded and bound to
225 - mlx5_core: hardware driver managing Mellanox ConnectX-4 devices and
226 related Ethernet kernel network devices.
227 - mlx5_ib: InifiniBand device driver.
228 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
230 - **Firmware update**
232 Mellanox OFED releases include firmware updates for ConnectX-4 adapters.
234 Because each release provides new features, these updates must be applied to
235 match the kernel modules and libraries they come with.
239 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
242 Currently supported by DPDK:
244 - Mellanox OFED **3.4-1.0.0.0**.
248 - ConnectX-4: **12.17.1010**
249 - ConnectX-4 Lx: **14.17.1010**
251 Getting Mellanox OFED
252 ~~~~~~~~~~~~~~~~~~~~~
254 While these libraries and kernel modules are available on OpenFabrics
255 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
256 managers on most distributions, this PMD requires Ethernet extensions that
257 may not be supported at the moment (this is a work in progress).
260 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__
261 includes the necessary support and should be used in the meantime. For DPDK,
262 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
263 required from that distribution.
267 Several versions of Mellanox OFED are available. Installing the version
268 this DPDK release was developed and tested against is strongly
269 recommended. Please check the `prerequisites`_.
274 Compared to librte_pmd_mlx4 that implements a single RSS configuration per
275 port, librte_pmd_mlx5 supports per-protocol RSS configuration.
277 Since ``testpmd`` defaults to IP RSS mode and there is currently no
278 command-line parameter to enable additional protocols (UDP and TCP as well
279 as IP), the following commands must be entered from its CLI to get the same
280 behavior as librte_pmd_mlx4:
282 .. code-block:: console
285 > port config all rss all
291 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-4
292 devices managed by librte_pmd_mlx5.
294 #. Load the kernel modules:
296 .. code-block:: console
298 modprobe -a ib_uverbs mlx5_core mlx5_ib
300 Alternatively if MLNX_OFED is fully installed, the following script can
303 .. code-block:: console
305 /etc/init.d/openibd restart
309 User space I/O kernel modules (uio and igb_uio) are not used and do
310 not have to be loaded.
312 #. Make sure Ethernet interfaces are in working order and linked to kernel
313 verbs. Related sysfs entries should be present:
315 .. code-block:: console
317 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
321 .. code-block:: console
328 #. Optionally, retrieve their PCI bus addresses for whitelisting:
330 .. code-block:: console
333 for intf in eth2 eth3 eth4 eth5;
335 (cd "/sys/class/net/${intf}/device/" && pwd -P);
338 sed -n 's,.*/\(.*\),-w \1,p'
342 .. code-block:: console
349 #. Request huge pages:
351 .. code-block:: console
353 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
355 #. Start testpmd with basic parameters:
357 .. code-block:: console
359 testpmd -c 0xff00 -n 4 -w 05:00.0 -w 05:00.1 -w 06:00.0 -w 06:00.1 -- --rxq=2 --txq=2 -i
363 .. code-block:: console
366 EAL: PCI device 0000:05:00.0 on NUMA socket 0
367 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
368 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
369 PMD: librte_pmd_mlx5: 1 port(s) detected
370 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
371 EAL: PCI device 0000:05:00.1 on NUMA socket 0
372 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
373 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
374 PMD: librte_pmd_mlx5: 1 port(s) detected
375 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
376 EAL: PCI device 0000:06:00.0 on NUMA socket 0
377 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
378 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
379 PMD: librte_pmd_mlx5: 1 port(s) detected
380 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
381 EAL: PCI device 0000:06:00.1 on NUMA socket 0
382 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
383 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
384 PMD: librte_pmd_mlx5: 1 port(s) detected
385 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
386 Interactive-mode selected
387 Configuring Port 0 (socket 0)
388 PMD: librte_pmd_mlx5: 0x8cba80: TX queues number update: 0 -> 2
389 PMD: librte_pmd_mlx5: 0x8cba80: RX queues number update: 0 -> 2
390 Port 0: E4:1D:2D:E7:0C:FE
391 Configuring Port 1 (socket 0)
392 PMD: librte_pmd_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
393 PMD: librte_pmd_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
394 Port 1: E4:1D:2D:E7:0C:FF
395 Configuring Port 2 (socket 0)
396 PMD: librte_pmd_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
397 PMD: librte_pmd_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
398 Port 2: E4:1D:2D:E7:0C:FA
399 Configuring Port 3 (socket 0)
400 PMD: librte_pmd_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
401 PMD: librte_pmd_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
402 Port 3: E4:1D:2D:E7:0C:FB
403 Checking link statuses...
404 Port 0 Link Up - speed 40000 Mbps - full-duplex
405 Port 1 Link Up - speed 40000 Mbps - full-duplex
406 Port 2 Link Up - speed 10000 Mbps - full-duplex
407 Port 3 Link Up - speed 10000 Mbps - full-duplex