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** 10/40 Gbps adapters (EN 40, EN 10, Pro EN 40) as
35 well as their virtual functions (VF) in SR-IOV context.
39 Due to external dependencies, this driver is disabled by default. It must
40 be enabled manually by setting ``CONFIG_RTE_LIBRTE_MLX4_PMD=y`` and
43 Implementation details
44 ----------------------
46 Most Mellanox ConnectX-3 devices provide two ports but expose a single PCI
47 bus address, thus unlike most drivers, librte_pmd_mlx4 registers itself as a
48 PCI driver that allocates one Ethernet device per detected port.
50 For this reason, one cannot white/blacklist a single port without also
51 white/blacklisting the others on the same device.
53 Besides its dependency on libibverbs (that implies libmlx4 and associated
54 kernel support), librte_pmd_mlx4 relies heavily on system calls for control
55 operations such as querying/updating the MTU and flow control parameters.
57 For security reasons and robustness, this driver only deals with virtual
58 memory addresses. The way resources allocations are handled by the kernel
59 combined with hardware specifications that allow it to handle virtual memory
60 addresses directly ensure that DPDK applications cannot access random
61 physical memory (or memory that does not belong to the current process).
63 This capability allows the PMD to coexist with kernel network interfaces
64 which remain functional, although they stop receiving unicast packets as
65 long as they share the same MAC address.
67 Compiling librte_pmd_mlx4 causes DPDK to be linked against libibverbs.
69 Features and limitations
70 ------------------------
72 - RSS, also known as RCA, is supported. In this mode the number of
73 configured RX queues must be a power of two.
74 - VLAN filtering is supported.
75 - Link state information is provided.
76 - Promiscuous mode is supported.
77 - All multicast mode is supported.
78 - Multiple MAC addresses (unicast, multicast) can be configured.
79 - Scattered packets are supported for TX and RX.
83 - RSS hash key cannot be modified.
84 - Hardware counters are not implemented (they are software counters).
85 - Checksum offloads are not supported yet.
93 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
95 Toggle compilation of librte_pmd_mlx4 itself.
97 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
99 Toggle debugging code and stricter compilation flags. Enabling this option
100 adds additional run-time checks and debugging messages at the cost of
103 - ``CONFIG_RTE_LIBRTE_MLX4_SGE_WR_N`` (default **4**)
105 Number of scatter/gather elements (SGEs) per work request (WR). Lowering
106 this number improves performance but also limits the ability to receive
107 scattered packets (packets that do not fit a single mbuf). The default
108 value is a safe tradeoff.
110 - ``CONFIG_RTE_LIBRTE_MLX4_MAX_INLINE`` (default **0**)
112 Amount of data to be inlined during TX operations. Improves latency but
115 - ``CONFIG_RTE_LIBRTE_MLX4_TX_MP_CACHE`` (default **8**)
117 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
118 which buffers are to be transmitted must be associated to memory regions
119 (MRs). This is a slow operation that must be cached.
121 This value is always 1 for RX queues since they use a single MP.
123 - ``CONFIG_RTE_LIBRTE_MLX4_SOFT_COUNTERS`` (default **1**)
125 Toggle software counters. No counters are available if this option is
126 disabled since hardware counters are not supported.
128 Environment variables
129 ~~~~~~~~~~~~~~~~~~~~~
131 - ``MLX4_INLINE_RECV_SIZE``
133 A nonzero value enables inline receive for packets up to that size. May
134 significantly improve performance in some cases but lower it in
135 others. Requires careful testing.
137 Run-time configuration
138 ~~~~~~~~~~~~~~~~~~~~~~
140 - The only constraint when RSS mode is requested is to make sure the number
141 of RX queues is a power of two. This is a hardware requirement.
143 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
144 because it is affected by their state. Forcing them down prevents packets
147 - **ethtool** operations on related kernel interfaces also affect the PMD.
152 This driver relies on external libraries and kernel drivers for resources
153 allocations and initialization. The following dependencies are not part of
154 DPDK and must be installed separately:
158 User space verbs framework used by librte_pmd_mlx4. This library provides
159 a generic interface between the kernel and low-level user space drivers
162 It allows slow and privileged operations (context initialization, hardware
163 resources allocations) to be managed by the kernel and fast operations to
164 never leave user space.
168 Low-level user space driver library for Mellanox ConnectX-3 devices,
169 it is automatically loaded by libibverbs.
171 This library basically implements send/receive calls to the hardware
174 - **Kernel modules** (mlnx-ofed-kernel)
176 They provide the kernel-side verbs API and low level device drivers that
177 manage actual hardware initialization and resources sharing with user
180 Unlike most other PMDs, these modules must remain loaded and bound to
183 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
184 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
185 - mlx4_ib: InifiniBand device driver.
186 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
188 While these libraries and kernel modules are available on OpenFabrics
189 Aliance's `website <https://www.openfabrics.org/>`_ and provided by package
190 managers on most distributions, this PMD requires Ethernet extensions that
191 may not be supported at the moment (this is a work in progress).
194 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers>`_
195 includes the necessary support and should be used in the meantime. For DPDK,
196 only libibverbs, libmlx4 and mlnx-ofed-kernel packages are required from
201 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
207 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
208 devices managed by librte_pmd_mlx4.
210 #. Load the kernel modules:
212 .. code-block:: console
214 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
218 User space I/O kernel modules (uio and igb_uio) are not used and do
219 not have to be loaded.
221 #. Make sure Ethernet interfaces are in working order and linked to kernel
222 verbs. Related sysfs entries should be present:
224 .. code-block:: console
226 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
230 .. code-block:: console
237 #. Optionally, retrieve their PCI bus addresses for whitelisting:
239 .. code-block:: console
242 for intf in eth2 eth3 eth4 eth5;
244 (cd "/sys/class/net/${intf}/device/" && pwd -P);
247 sed -n 's,.*/\(.*\),-w \1,p'
251 .. code-block:: console
260 There are only two distinct PCI bus addresses because the Mellanox
261 ConnectX-3 adapters installed on this system are dual port.
263 #. Request huge pages:
265 .. code-block:: console
267 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
269 #. Start testpmd with basic parameters:
271 .. code-block:: console
273 testpmd -c 0xff00 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
277 .. code-block:: console
280 EAL: PCI device 0000:83:00.0 on NUMA socket 1
281 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
282 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
283 PMD: librte_pmd_mlx4: 2 port(s) detected
284 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
285 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
286 EAL: PCI device 0000:84:00.0 on NUMA socket 1
287 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
288 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
289 PMD: librte_pmd_mlx4: 2 port(s) detected
290 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
291 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
292 Interactive-mode selected
293 Configuring Port 0 (socket 0)
294 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
295 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
296 Port 0: 00:02:C9:B5:B7:50
297 Configuring Port 1 (socket 0)
298 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
299 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
300 Port 1: 00:02:C9:B5:B7:51
301 Configuring Port 2 (socket 0)
302 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
303 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
304 Port 2: 00:02:C9:B5:BA:B0
305 Configuring Port 3 (socket 0)
306 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
307 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
308 Port 3: 00:02:C9:B5:BA:B1
309 Checking link statuses...
310 Port 0 Link Up - speed 10000 Mbps - full-duplex
311 Port 1 Link Up - speed 40000 Mbps - full-duplex
312 Port 2 Link Up - speed 10000 Mbps - full-duplex
313 Port 3 Link Up - speed 40000 Mbps - full-duplex