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.
82 - Multicast promiscuous mode.
83 - Hardware checksum offloads.
88 - KVM and VMware ESX SR-IOV modes are not supported yet.
89 - Inner RSS for VXLAN frames is not supported yet.
90 - Port statistics through software counters only.
91 - Hardware checksum offloads for VXLAN inner header are not supported yet.
92 - Secondary processes are not supported yet.
100 These options can be modified in the ``.config`` file.
102 - ``CONFIG_RTE_LIBRTE_MLX5_PMD`` (default **n**)
104 Toggle compilation of librte_pmd_mlx5 itself.
106 - ``CONFIG_RTE_LIBRTE_MLX5_DEBUG`` (default **n**)
108 Toggle debugging code and stricter compilation flags. Enabling this option
109 adds additional run-time checks and debugging messages at the cost of
112 - ``CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N`` (default **4**)
114 Number of scatter/gather elements (SGEs) per work request (WR). Lowering
115 this number improves performance but also limits the ability to receive
116 scattered packets (packets that do not fit a single mbuf). The default
117 value is a safe tradeoff.
119 - ``CONFIG_RTE_LIBRTE_MLX5_MAX_INLINE`` (default **0**)
121 Amount of data to be inlined during TX operations. Improves latency.
122 Can improve PPS performance when PCI backpressure is detected and may be
123 useful for scenarios involving heavy traffic on many queues.
125 Since the additional software logic necessary to handle this mode can
126 lower performance when there is no backpressure, it is not enabled by
129 - ``CONFIG_RTE_LIBRTE_MLX5_TX_MP_CACHE`` (default **8**)
131 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
132 which buffers are to be transmitted must be associated to memory regions
133 (MRs). This is a slow operation that must be cached.
135 This value is always 1 for RX queues since they use a single MP.
137 Environment variables
138 ~~~~~~~~~~~~~~~~~~~~~
140 - ``MLX5_ENABLE_CQE_COMPRESSION``
142 A nonzero value lets ConnectX-4 return smaller completion entries to
143 improve performance when PCI backpressure is detected. It is most useful
144 for scenarios involving heavy traffic on many queues.
146 Since the additional software logic necessary to handle this mode can
147 lower performance when there is no backpressure, it is not enabled by
150 Run-time configuration
151 ~~~~~~~~~~~~~~~~~~~~~~
153 - librte_pmd_mlx5 brings kernel network interfaces up during initialization
154 because it is affected by their state. Forcing them down prevents packets
157 - **ethtool** operations on related kernel interfaces also affect the PMD.
162 This driver relies on external libraries and kernel drivers for resources
163 allocations and initialization. The following dependencies are not part of
164 DPDK and must be installed separately:
168 User space Verbs framework used by librte_pmd_mlx5. This library provides
169 a generic interface between the kernel and low-level user space drivers
172 It allows slow and privileged operations (context initialization, hardware
173 resources allocations) to be managed by the kernel and fast operations to
174 never leave user space.
178 Low-level user space driver library for Mellanox ConnectX-4 devices,
179 it is automatically loaded by libibverbs.
181 This library basically implements send/receive calls to the hardware
184 - **Kernel modules** (mlnx-ofed-kernel)
186 They provide the kernel-side Verbs API and low level device drivers that
187 manage actual hardware initialization and resources sharing with user
190 Unlike most other PMDs, these modules must remain loaded and bound to
193 - mlx5_core: hardware driver managing Mellanox ConnectX-4 devices and
194 related Ethernet kernel network devices.
195 - mlx5_ib: InifiniBand device driver.
196 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
198 - **Firmware update**
200 Mellanox OFED releases include firmware updates for ConnectX-4 adapters.
202 Because each release provides new features, these updates must be applied to
203 match the kernel modules and libraries they come with.
207 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
210 Currently supported by DPDK:
212 - Mellanox OFED **3.1-1.0.3** or **3.1-1.5.7.1** depending on usage.
214 The following features are supported with version **3.1-1.5.7.1** and
217 - IPv6, UPDv6, TCPv6 RSS.
218 - RX checksum offloads.
221 - Minimum firmware version:
223 With MLNX_OFED **3.1-1.0.3**:
225 - ConnectX-4: **12.12.1240**
226 - ConnectX-4 Lx: **14.12.1100**
228 With MLNX_OFED **3.1-1.5.7.1**:
230 - ConnectX-4: **12.13.0144**
231 - ConnectX-4 Lx: **14.13.0144**
233 Getting Mellanox OFED
234 ~~~~~~~~~~~~~~~~~~~~~
236 While these libraries and kernel modules are available on OpenFabrics
237 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
238 managers on most distributions, this PMD requires Ethernet extensions that
239 may not be supported at the moment (this is a work in progress).
242 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__
243 includes the necessary support and should be used in the meantime. For DPDK,
244 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
245 required from that distribution.
249 Several versions of Mellanox OFED are available. Installing the version
250 this DPDK release was developed and tested against is strongly
251 recommended. Please check the `prerequisites`_.
256 Compared to librte_pmd_mlx4 that implements a single RSS configuration per
257 port, librte_pmd_mlx5 supports per-protocol RSS configuration.
259 Since ``testpmd`` defaults to IP RSS mode and there is currently no
260 command-line parameter to enable additional protocols (UDP and TCP as well
261 as IP), the following commands must be entered from its CLI to get the same
262 behavior as librte_pmd_mlx4:
264 .. code-block:: console
267 > port config all rss all
273 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-4
274 devices managed by librte_pmd_mlx5.
276 #. Load the kernel modules:
278 .. code-block:: console
280 modprobe -a ib_uverbs mlx5_core mlx5_ib
282 Alternatively if MLNX_OFED is fully installed, the following script can
285 .. code-block:: console
287 /etc/init.d/openibd restart
291 User space I/O kernel modules (uio and igb_uio) are not used and do
292 not have to be loaded.
294 #. Make sure Ethernet interfaces are in working order and linked to kernel
295 verbs. Related sysfs entries should be present:
297 .. code-block:: console
299 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
303 .. code-block:: console
310 #. Optionally, retrieve their PCI bus addresses for whitelisting:
312 .. code-block:: console
315 for intf in eth2 eth3 eth4 eth5;
317 (cd "/sys/class/net/${intf}/device/" && pwd -P);
320 sed -n 's,.*/\(.*\),-w \1,p'
324 .. code-block:: console
331 #. Request huge pages:
333 .. code-block:: console
335 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
337 #. Start testpmd with basic parameters:
339 .. code-block:: console
341 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
345 .. code-block:: console
348 EAL: PCI device 0000:05:00.0 on NUMA socket 0
349 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
350 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
351 PMD: librte_pmd_mlx5: 1 port(s) detected
352 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
353 EAL: PCI device 0000:05:00.1 on NUMA socket 0
354 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
355 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
356 PMD: librte_pmd_mlx5: 1 port(s) detected
357 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
358 EAL: PCI device 0000:06:00.0 on NUMA socket 0
359 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
360 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
361 PMD: librte_pmd_mlx5: 1 port(s) detected
362 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
363 EAL: PCI device 0000:06:00.1 on NUMA socket 0
364 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
365 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
366 PMD: librte_pmd_mlx5: 1 port(s) detected
367 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
368 Interactive-mode selected
369 Configuring Port 0 (socket 0)
370 PMD: librte_pmd_mlx5: 0x8cba80: TX queues number update: 0 -> 2
371 PMD: librte_pmd_mlx5: 0x8cba80: RX queues number update: 0 -> 2
372 Port 0: E4:1D:2D:E7:0C:FE
373 Configuring Port 1 (socket 0)
374 PMD: librte_pmd_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
375 PMD: librte_pmd_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
376 Port 1: E4:1D:2D:E7:0C:FF
377 Configuring Port 2 (socket 0)
378 PMD: librte_pmd_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
379 PMD: librte_pmd_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
380 Port 2: E4:1D:2D:E7:0C:FA
381 Configuring Port 3 (socket 0)
382 PMD: librte_pmd_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
383 PMD: librte_pmd_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
384 Port 3: E4:1D:2D:E7:0C:FB
385 Checking link statuses...
386 Port 0 Link Up - speed 40000 Mbps - full-duplex
387 Port 1 Link Up - speed 40000 Mbps - full-duplex
388 Port 2 Link Up - speed 10000 Mbps - full-duplex
389 Port 3 Link Up - speed 10000 Mbps - full-duplex