2 Copyright 2012-2015 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.
30 MLX4 poll mode driver library
31 =============================
33 The MLX4 poll mode driver library (**librte_pmd_mlx4**) implements support
34 for **Mellanox ConnectX-3 EN** 10/40 Gbps adapters as well as their virtual
35 functions (VF) in SR-IOV context.
37 Information and documentation about this family of adapters can be found on
38 the `Mellanox website <http://www.mellanox.com>`_. Help is also provided by
39 the `Mellanox community <http://community.mellanox.com/welcome>`_.
41 There is also a `section dedicated to this poll mode driver
42 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`_.
46 Due to external dependencies, this driver is disabled by default. It must
47 be enabled manually by setting ``CONFIG_RTE_LIBRTE_MLX4_PMD=y`` and
52 ``CONFIG_RTE_BUILD_COMBINE_LIBS`` with ``CONFIG_RTE_BUILD_SHARED_LIB``
53 is not supported and thus the compilation will fail with this configuration.
55 Implementation details
56 ----------------------
58 Most Mellanox ConnectX-3 devices provide two ports but expose a single PCI
59 bus address, thus unlike most drivers, librte_pmd_mlx4 registers itself as a
60 PCI driver that allocates one Ethernet device per detected port.
62 For this reason, one cannot white/blacklist a single port without also
63 white/blacklisting the others on the same device.
65 Besides its dependency on libibverbs (that implies libmlx4 and associated
66 kernel support), librte_pmd_mlx4 relies heavily on system calls for control
67 operations such as querying/updating the MTU and flow control parameters.
69 For security reasons and robustness, this driver only deals with virtual
70 memory addresses. The way resources allocations are handled by the kernel
71 combined with hardware specifications that allow it to handle virtual memory
72 addresses directly ensure that DPDK applications cannot access random
73 physical memory (or memory that does not belong to the current process).
75 This capability allows the PMD to coexist with kernel network interfaces
76 which remain functional, although they stop receiving unicast packets as
77 long as they share the same MAC address.
79 Compiling librte_pmd_mlx4 causes DPDK to be linked against libibverbs.
81 Features and limitations
82 ------------------------
84 - RSS, also known as RCA, is supported. In this mode the number of
85 configured RX queues must be a power of two.
86 - VLAN filtering is supported.
87 - Link state information is provided.
88 - Promiscuous mode is supported.
89 - All multicast mode is supported.
90 - Multiple MAC addresses (unicast, multicast) can be configured.
91 - Scattered packets are supported for TX and RX.
92 - Inner L3/L4 (IP, TCP and UDP) TX/RX checksum offloading and validation.
93 - Outer L3 (IP) TX/RX checksum offloading and validation for VXLAN frames.
97 - RSS hash key cannot be modified.
98 - Hardware counters are not implemented (they are software counters).
106 These options can be modified in the ``.config`` file.
108 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
110 Toggle compilation of librte_pmd_mlx4 itself.
112 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
114 Toggle debugging code and stricter compilation flags. Enabling this option
115 adds additional run-time checks and debugging messages at the cost of
118 - ``CONFIG_RTE_LIBRTE_MLX4_SGE_WR_N`` (default **4**)
120 Number of scatter/gather elements (SGEs) per work request (WR). Lowering
121 this number improves performance but also limits the ability to receive
122 scattered packets (packets that do not fit a single mbuf). The default
123 value is a safe tradeoff.
125 - ``CONFIG_RTE_LIBRTE_MLX4_MAX_INLINE`` (default **0**)
127 Amount of data to be inlined during TX operations. Improves latency but
130 - ``CONFIG_RTE_LIBRTE_MLX4_TX_MP_CACHE`` (default **8**)
132 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
133 which buffers are to be transmitted must be associated to memory regions
134 (MRs). This is a slow operation that must be cached.
136 This value is always 1 for RX queues since they use a single MP.
138 - ``CONFIG_RTE_LIBRTE_MLX4_SOFT_COUNTERS`` (default **1**)
140 Toggle software counters. No counters are available if this option is
141 disabled since hardware counters are not supported.
143 Environment variables
144 ~~~~~~~~~~~~~~~~~~~~~
146 - ``MLX4_INLINE_RECV_SIZE``
148 A nonzero value enables inline receive for packets up to that size. May
149 significantly improve performance in some cases but lower it in
150 others. Requires careful testing.
152 Run-time configuration
153 ~~~~~~~~~~~~~~~~~~~~~~
155 - The only constraint when RSS mode is requested is to make sure the number
156 of RX queues is a power of two. This is a hardware requirement.
158 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
159 because it is affected by their state. Forcing them down prevents packets
162 - **ethtool** operations on related kernel interfaces also affect the PMD.
164 Kernel module parameters
165 ~~~~~~~~~~~~~~~~~~~~~~~~
167 The **mlx4_core** kernel module has several parameters that affect the
168 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
171 - **num_vfs** (integer or triplet, optionally prefixed by device address
174 Create the given number of VFs on the specified devices.
176 - **log_num_mgm_entry_size** (integer)
178 Device-managed flow steering (DMFS) is required by DPDK applications. It is
179 enabled by using a negative value, the last four bits of which have a
182 - **-1**: force device-managed flow steering (DMFS).
183 - **-7**: configure optimized steering mode to improve performance with the
184 following limitation: VLAN filtering is not supported with this mode.
185 This is the recommended mode in case VLAN filter is not needed.
190 This driver relies on external libraries and kernel drivers for resources
191 allocations and initialization. The following dependencies are not part of
192 DPDK and must be installed separately:
196 User space verbs framework used by librte_pmd_mlx4. This library provides
197 a generic interface between the kernel and low-level user space drivers
200 It allows slow and privileged operations (context initialization, hardware
201 resources allocations) to be managed by the kernel and fast operations to
202 never leave user space.
206 Low-level user space driver library for Mellanox ConnectX-3 devices,
207 it is automatically loaded by libibverbs.
209 This library basically implements send/receive calls to the hardware
212 - **Kernel modules** (mlnx-ofed-kernel)
214 They provide the kernel-side verbs API and low level device drivers that
215 manage actual hardware initialization and resources sharing with user
218 Unlike most other PMDs, these modules must remain loaded and bound to
221 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
222 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
223 - mlx4_ib: InifiniBand device driver.
224 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
226 - **Firmware update**
228 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
230 Because each release provides new features, these updates must be applied to
231 match the kernel modules and libraries they come with.
235 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
238 Currently supported by DPDK:
240 - Mellanox OFED **3.0**.
241 - Firmware version **2.34.5000** and higher.
243 Getting Mellanox OFED
244 ~~~~~~~~~~~~~~~~~~~~~
246 While these libraries and kernel modules are available on OpenFabrics
247 Alliance's `website <https://www.openfabrics.org/>`_ and provided by package
248 managers on most distributions, this PMD requires Ethernet extensions that
249 may not be supported at the moment (this is a work in progress).
252 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers>`_
253 includes the necessary support and should be used in the meantime. For DPDK,
254 only libibverbs, libmlx4, mlnx-ofed-kernel packages and firmware updates are
255 required from that distribution.
259 Several versions of Mellanox OFED are available. Installing the version
260 this DPDK release was developed and tested against is strongly
261 recommended. Please check the `prerequisites`_.
266 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
267 devices managed by librte_pmd_mlx4.
269 #. Load the kernel modules:
271 .. code-block:: console
273 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
277 User space I/O kernel modules (uio and igb_uio) are not used and do
278 not have to be loaded.
280 #. Make sure Ethernet interfaces are in working order and linked to kernel
281 verbs. Related sysfs entries should be present:
283 .. code-block:: console
285 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
289 .. code-block:: console
296 #. Optionally, retrieve their PCI bus addresses for whitelisting:
298 .. code-block:: console
301 for intf in eth2 eth3 eth4 eth5;
303 (cd "/sys/class/net/${intf}/device/" && pwd -P);
306 sed -n 's,.*/\(.*\),-w \1,p'
310 .. code-block:: console
319 There are only two distinct PCI bus addresses because the Mellanox
320 ConnectX-3 adapters installed on this system are dual port.
322 #. Request huge pages:
324 .. code-block:: console
326 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
328 #. Start testpmd with basic parameters:
330 .. code-block:: console
332 testpmd -c 0xff00 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
336 .. code-block:: console
339 EAL: PCI device 0000:83:00.0 on NUMA socket 1
340 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
341 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
342 PMD: librte_pmd_mlx4: 2 port(s) detected
343 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
344 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
345 EAL: PCI device 0000:84:00.0 on NUMA socket 1
346 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
347 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
348 PMD: librte_pmd_mlx4: 2 port(s) detected
349 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
350 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
351 Interactive-mode selected
352 Configuring Port 0 (socket 0)
353 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
354 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
355 Port 0: 00:02:C9:B5:B7:50
356 Configuring Port 1 (socket 0)
357 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
358 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
359 Port 1: 00:02:C9:B5:B7:51
360 Configuring Port 2 (socket 0)
361 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
362 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
363 Port 2: 00:02:C9:B5:BA:B0
364 Configuring Port 3 (socket 0)
365 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
366 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
367 Port 3: 00:02:C9:B5:BA:B1
368 Checking link statuses...
369 Port 0 Link Up - speed 10000 Mbps - full-duplex
370 Port 1 Link Up - speed 40000 Mbps - full-duplex
371 Port 2 Link Up - speed 10000 Mbps - full-duplex
372 Port 3 Link Up - speed 40000 Mbps - full-duplex