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 - RX CRC stripping configuration.
84 - Multicast promiscuous mode.
85 - Hardware checksum offloads.
86 - Flow director (RTE_FDIR_MODE_PERFECT and RTE_FDIR_MODE_PERFECT_MAC_VLAN).
87 - Secondary process TX is supported.
92 - KVM and VMware ESX SR-IOV modes are not supported yet.
93 - Inner RSS for VXLAN frames is not supported yet.
94 - Port statistics through software counters only.
95 - Hardware checksum offloads for VXLAN inner header are not supported yet.
96 - Secondary process RX is not supported.
104 These options can be modified in the ``.config`` file.
106 - ``CONFIG_RTE_LIBRTE_MLX5_PMD`` (default **n**)
108 Toggle compilation of librte_pmd_mlx5 itself.
110 - ``CONFIG_RTE_LIBRTE_MLX5_DEBUG`` (default **n**)
112 Toggle debugging code and stricter compilation flags. Enabling this option
113 adds additional run-time checks and debugging messages at the cost of
116 - ``CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N`` (default **4**)
118 Number of scatter/gather elements (SGEs) per work request (WR). Lowering
119 this number improves performance but also limits the ability to receive
120 scattered packets (packets that do not fit a single mbuf). The default
121 value is a safe tradeoff.
123 - ``CONFIG_RTE_LIBRTE_MLX5_MAX_INLINE`` (default **0**)
125 Amount of data to be inlined during TX operations. Improves latency.
126 Can improve PPS performance when PCI backpressure is detected and may be
127 useful for scenarios involving heavy traffic on many queues.
129 Since the additional software logic necessary to handle this mode can
130 lower performance when there is no backpressure, it is not enabled by
133 - ``CONFIG_RTE_LIBRTE_MLX5_TX_MP_CACHE`` (default **8**)
135 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
136 which buffers are to be transmitted must be associated to memory regions
137 (MRs). This is a slow operation that must be cached.
139 This value is always 1 for RX queues since they use a single MP.
141 Environment variables
142 ~~~~~~~~~~~~~~~~~~~~~
144 - ``MLX5_ENABLE_CQE_COMPRESSION``
146 A nonzero value lets ConnectX-4 return smaller completion entries to
147 improve performance when PCI backpressure is detected. It is most useful
148 for scenarios involving heavy traffic on many queues.
150 Since the additional software logic necessary to handle this mode can
151 lower performance when there is no backpressure, it is not enabled by
154 - ``MLX5_PMD_ENABLE_PADDING``
156 Enables HW packet padding in PCI bus transactions.
158 When packet size is cache aligned and CRC stripping is enabled, 4 fewer
159 bytes are written to the PCI bus. Enabling padding makes such packets
162 In cases where PCI bandwidth is the bottleneck, padding can improve
165 This is disabled by default since this can also decrease performance for
166 unaligned packet sizes.
168 Run-time configuration
169 ~~~~~~~~~~~~~~~~~~~~~~
171 - librte_pmd_mlx5 brings kernel network interfaces up during initialization
172 because it is affected by their state. Forcing them down prevents packets
175 - **ethtool** operations on related kernel interfaces also affect the PMD.
180 This driver relies on external libraries and kernel drivers for resources
181 allocations and initialization. The following dependencies are not part of
182 DPDK and must be installed separately:
186 User space Verbs framework used by librte_pmd_mlx5. This library provides
187 a generic interface between the kernel and low-level user space drivers
190 It allows slow and privileged operations (context initialization, hardware
191 resources allocations) to be managed by the kernel and fast operations to
192 never leave user space.
196 Low-level user space driver library for Mellanox ConnectX-4 devices,
197 it is automatically loaded by libibverbs.
199 This library basically implements send/receive calls to the hardware
202 - **Kernel modules** (mlnx-ofed-kernel)
204 They provide the kernel-side Verbs API and low level device drivers that
205 manage actual hardware initialization and resources sharing with user
208 Unlike most other PMDs, these modules must remain loaded and bound to
211 - mlx5_core: hardware driver managing Mellanox ConnectX-4 devices and
212 related Ethernet kernel network devices.
213 - mlx5_ib: InifiniBand device driver.
214 - ib_uverbs: user space driver for Verbs (entry point for libibverbs).
216 - **Firmware update**
218 Mellanox OFED releases include firmware updates for ConnectX-4 adapters.
220 Because each release provides new features, these updates must be applied to
221 match the kernel modules and libraries they come with.
225 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
228 Currently supported by DPDK:
230 - Mellanox OFED **3.1-1.0.3**, **3.1-1.5.7.1** or **3.2-2.0.0.0** depending
233 The following features are supported with version **3.1-1.5.7.1** and
236 - IPv6, UPDv6, TCPv6 RSS.
237 - RX checksum offloads.
240 The following features are supported with version **3.2-2.0.0.0** and
245 - RX CRC stripping configuration.
247 - Minimum firmware version:
249 With MLNX_OFED **3.1-1.0.3**:
251 - ConnectX-4: **12.12.1240**
252 - ConnectX-4 Lx: **14.12.1100**
254 With MLNX_OFED **3.1-1.5.7.1**:
256 - ConnectX-4: **12.13.0144**
257 - ConnectX-4 Lx: **14.13.0144**
259 With MLNX_OFED **3.2-2.0.0.0**:
261 - ConnectX-4: **12.14.2036**
262 - ConnectX-4 Lx: **14.14.2036**
264 Getting Mellanox OFED
265 ~~~~~~~~~~~~~~~~~~~~~
267 While these libraries and kernel modules are available on OpenFabrics
268 Alliance's `website <https://www.openfabrics.org/>`__ and provided by package
269 managers on most distributions, this PMD requires Ethernet extensions that
270 may not be supported at the moment (this is a work in progress).
273 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux>`__
274 includes the necessary support and should be used in the meantime. For DPDK,
275 only libibverbs, libmlx5, mlnx-ofed-kernel packages and firmware updates are
276 required from that distribution.
280 Several versions of Mellanox OFED are available. Installing the version
281 this DPDK release was developed and tested against is strongly
282 recommended. Please check the `prerequisites`_.
287 Compared to librte_pmd_mlx4 that implements a single RSS configuration per
288 port, librte_pmd_mlx5 supports per-protocol RSS configuration.
290 Since ``testpmd`` defaults to IP RSS mode and there is currently no
291 command-line parameter to enable additional protocols (UDP and TCP as well
292 as IP), the following commands must be entered from its CLI to get the same
293 behavior as librte_pmd_mlx4:
295 .. code-block:: console
298 > port config all rss all
304 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-4
305 devices managed by librte_pmd_mlx5.
307 #. Load the kernel modules:
309 .. code-block:: console
311 modprobe -a ib_uverbs mlx5_core mlx5_ib
313 Alternatively if MLNX_OFED is fully installed, the following script can
316 .. code-block:: console
318 /etc/init.d/openibd restart
322 User space I/O kernel modules (uio and igb_uio) are not used and do
323 not have to be loaded.
325 #. Make sure Ethernet interfaces are in working order and linked to kernel
326 verbs. Related sysfs entries should be present:
328 .. code-block:: console
330 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
334 .. code-block:: console
341 #. Optionally, retrieve their PCI bus addresses for whitelisting:
343 .. code-block:: console
346 for intf in eth2 eth3 eth4 eth5;
348 (cd "/sys/class/net/${intf}/device/" && pwd -P);
351 sed -n 's,.*/\(.*\),-w \1,p'
355 .. code-block:: console
362 #. Request huge pages:
364 .. code-block:: console
366 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
368 #. Start testpmd with basic parameters:
370 .. code-block:: console
372 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
376 .. code-block:: console
379 EAL: PCI device 0000:05:00.0 on NUMA socket 0
380 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
381 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_0" (VF: false)
382 PMD: librte_pmd_mlx5: 1 port(s) detected
383 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fe
384 EAL: PCI device 0000:05:00.1 on NUMA socket 0
385 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
386 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_1" (VF: false)
387 PMD: librte_pmd_mlx5: 1 port(s) detected
388 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:ff
389 EAL: PCI device 0000:06:00.0 on NUMA socket 0
390 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
391 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_2" (VF: false)
392 PMD: librte_pmd_mlx5: 1 port(s) detected
393 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fa
394 EAL: PCI device 0000:06:00.1 on NUMA socket 0
395 EAL: probe driver: 15b3:1013 librte_pmd_mlx5
396 PMD: librte_pmd_mlx5: PCI information matches, using device "mlx5_3" (VF: false)
397 PMD: librte_pmd_mlx5: 1 port(s) detected
398 PMD: librte_pmd_mlx5: port 1 MAC address is e4:1d:2d:e7:0c:fb
399 Interactive-mode selected
400 Configuring Port 0 (socket 0)
401 PMD: librte_pmd_mlx5: 0x8cba80: TX queues number update: 0 -> 2
402 PMD: librte_pmd_mlx5: 0x8cba80: RX queues number update: 0 -> 2
403 Port 0: E4:1D:2D:E7:0C:FE
404 Configuring Port 1 (socket 0)
405 PMD: librte_pmd_mlx5: 0x8ccac8: TX queues number update: 0 -> 2
406 PMD: librte_pmd_mlx5: 0x8ccac8: RX queues number update: 0 -> 2
407 Port 1: E4:1D:2D:E7:0C:FF
408 Configuring Port 2 (socket 0)
409 PMD: librte_pmd_mlx5: 0x8cdb10: TX queues number update: 0 -> 2
410 PMD: librte_pmd_mlx5: 0x8cdb10: RX queues number update: 0 -> 2
411 Port 2: E4:1D:2D:E7:0C:FA
412 Configuring Port 3 (socket 0)
413 PMD: librte_pmd_mlx5: 0x8ceb58: TX queues number update: 0 -> 2
414 PMD: librte_pmd_mlx5: 0x8ceb58: RX queues number update: 0 -> 2
415 Port 3: E4:1D:2D:E7:0C:FB
416 Checking link statuses...
417 Port 0 Link Up - speed 40000 Mbps - full-duplex
418 Port 1 Link Up - speed 40000 Mbps - full-duplex
419 Port 2 Link Up - speed 10000 Mbps - full-duplex
420 Port 3 Link Up - speed 10000 Mbps - full-duplex