2 Copyright 2012 6WIND S.A.
3 Copyright 2015 Mellanox
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions
9 * Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
11 * Redistributions in binary form must reproduce the above copyright
12 notice, this list of conditions and the following disclaimer in
13 the documentation and/or other materials provided with the
15 * Neither the name of 6WIND S.A. nor the names of its
16 contributors may be used to endorse or promote products derived
17 from this software without specific prior written permission.
19 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 MLX4 poll mode driver library
32 =============================
34 The MLX4 poll mode driver library (**librte_pmd_mlx4**) implements support
35 for **Mellanox ConnectX-3** and **Mellanox ConnectX-3 Pro** 10/40 Gbps adapters
36 as well as their virtual functions (VF) in SR-IOV context.
38 Information and documentation about this family of adapters can be found on
39 the `Mellanox website <http://www.mellanox.com>`_. Help is also provided by
40 the `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_MLX4_PMD=y`` and
51 Implementation details
52 ----------------------
54 Most Mellanox ConnectX-3 devices provide two ports but expose a single PCI
55 bus address, thus unlike most drivers, librte_pmd_mlx4 registers itself as a
56 PCI driver that allocates one Ethernet device per detected port.
58 For this reason, one cannot white/blacklist a single port without also
59 white/blacklisting the others on the same device.
61 Besides its dependency on libibverbs (that implies libmlx4 and associated
62 kernel support), librte_pmd_mlx4 relies heavily on system calls for control
63 operations such as querying/updating the MTU and flow control parameters.
65 For security reasons and robustness, this driver only deals with virtual
66 memory addresses. The way resources allocations are handled by the kernel
67 combined with hardware specifications that allow it to handle virtual memory
68 addresses directly ensure that DPDK applications cannot access random
69 physical memory (or memory that does not belong to the current process).
71 This capability allows the PMD to coexist with kernel network interfaces
72 which remain functional, although they stop receiving unicast packets as
73 long as they share the same MAC address.
75 Compiling librte_pmd_mlx4 causes DPDK to be linked against libibverbs.
83 These options can be modified in the ``.config`` file.
85 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
87 Toggle compilation of librte_pmd_mlx4 itself.
89 - ``CONFIG_RTE_LIBRTE_MLX4_DLOPEN_DEPS`` (default **n**)
91 Build PMD with additional code to make it loadable without hard
92 dependencies on **libibverbs** nor **libmlx4**, which may not be installed
95 In this mode, their presence is still required for it to run properly,
96 however their absence won't prevent a DPDK application from starting (with
97 ``CONFIG_RTE_BUILD_SHARED_LIB`` disabled) and they won't show up as
98 missing with ``ldd(1)``.
100 This option has no performance impact.
102 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
104 Toggle debugging code and stricter compilation flags. Enabling this option
105 adds additional run-time checks and debugging messages at the cost of
108 - ``CONFIG_RTE_LIBRTE_MLX4_TX_MP_CACHE`` (default **8**)
110 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
111 which buffers are to be transmitted must be associated to memory regions
112 (MRs). This is a slow operation that must be cached.
114 This value is always 1 for RX queues since they use a single MP.
116 Run-time configuration
117 ~~~~~~~~~~~~~~~~~~~~~~
119 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
120 because it is affected by their state. Forcing them down prevents packets
123 - **ethtool** operations on related kernel interfaces also affect the PMD.
125 - ``port`` parameter [int]
127 This parameter provides a physical port to probe and can be specified multiple
128 times for additional ports. All ports are probed by default if left
131 Kernel module parameters
132 ~~~~~~~~~~~~~~~~~~~~~~~~
134 The **mlx4_core** kernel module has several parameters that affect the
135 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
138 - **num_vfs** (integer or triplet, optionally prefixed by device address
141 Create the given number of VFs on the specified devices.
143 - **log_num_mgm_entry_size** (integer)
145 Device-managed flow steering (DMFS) is required by DPDK applications. It is
146 enabled by using a negative value, the last four bits of which have a
149 - **-1**: force device-managed flow steering (DMFS).
150 - **-7**: configure optimized steering mode to improve performance with the
151 following limitation: VLAN filtering is not supported with this mode.
152 This is the recommended mode in case VLAN filter is not needed.
157 This driver relies on external libraries and kernel drivers for resources
158 allocations and initialization. The following dependencies are not part of
159 DPDK and must be installed separately:
161 - **libibverbs** (provided by rdma-core package)
163 User space verbs framework used by librte_pmd_mlx4. This library provides
164 a generic interface between the kernel and low-level user space drivers
167 It allows slow and privileged operations (context initialization, hardware
168 resources allocations) to be managed by the kernel and fast operations to
169 never leave user space.
171 - **libmlx4** (provided by rdma-core package)
173 Low-level user space driver library for Mellanox ConnectX-3 devices,
174 it is automatically loaded by libibverbs.
176 This library basically implements send/receive calls to the hardware
181 They provide the kernel-side verbs API and low level device drivers that
182 manage actual hardware initialization and resources sharing with user
185 Unlike most other PMDs, these modules must remain loaded and bound to
188 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
189 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
190 - mlx4_ib: InifiniBand device driver.
191 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
193 - **Firmware update**
195 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
197 Because each release provides new features, these updates must be applied to
198 match the kernel modules and libraries they come with.
202 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
205 Depending on system constraints and user preferences either RDMA core library
206 with a recent enough Linux kernel release (recommended) or Mellanox OFED,
207 which provides compatibility with older releases.
209 Current RDMA core package and Linux kernel (recommended)
210 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
212 - Minimal Linux kernel version: 4.14.
213 - Minimal RDMA core version: v15 (see `RDMA core installation documentation`_).
215 .. _`RDMA core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
217 .. _Mellanox_OFED_as_a_fallback:
219 Mellanox OFED as a fallback
220 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
222 - `Mellanox OFED`_ version: **4.2, 4.3**.
223 - firmware version: **2.42.5000** and above.
225 .. _`Mellanox OFED`: http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers
229 Several versions of Mellanox OFED are available. Installing the version
230 this DPDK release was developed and tested against is strongly
231 recommended. Please check the `prerequisites`_.
233 Installing Mellanox OFED
234 ^^^^^^^^^^^^^^^^^^^^^^^^
236 1. Download latest Mellanox OFED.
238 2. Install the required libraries and kernel modules either by installing
239 only the required set, or by installing the entire Mellanox OFED:
243 .. code-block:: console
245 ./mlnxofedinstall --dpdk --upstream-libs
247 For SR-IOV hypervisors use:
249 .. code-block:: console
251 ./mlnxofedinstall --dpdk --upstream-libs --enable-sriov --hypervisor
253 For SR-IOV virtual machine use:
255 .. code-block:: console
257 ./mlnxofedinstall --dpdk --upstream-libs --guest
259 3. Verify the firmware is the correct one:
261 .. code-block:: console
265 4. Set all ports links to Ethernet, follow instructions on the screen:
267 .. code-block:: console
271 5. Continue with :ref:`section 2 of the Quick Start Guide <QSG_2>`.
276 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
283 1. Set all ports links to Ethernet
285 .. code-block:: console
287 PCI=<NIC PCI address>
288 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port0"
289 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port1"
293 If using Mellanox OFED one can permanently set the port link
294 to Ethernet using connectx_port_config tool provided by it.
295 :ref:`Mellanox_OFED_as_a_fallback`:
299 2. In case of bare metal or hypervisor, configure optimized steering mode
300 by adding the following line to ``/etc/modprobe.d/mlx4_core.conf``:
302 .. code-block:: console
304 options mlx4_core log_num_mgm_entry_size=-7
308 If VLAN filtering is used, set log_num_mgm_entry_size=-1.
309 Performance degradation can occur on this case.
311 3. Restart the driver:
313 .. code-block:: console
315 /etc/init.d/openibd restart
319 .. code-block:: console
321 service openibd restart
323 4. Compile DPDK and you are ready to go. See instructions on
324 :ref:`Development Kit Build System <Development_Kit_Build_System>`
329 1. Verify the optimized steering mode is configured:
331 .. code-block:: console
333 cat /sys/module/mlx4_core/parameters/log_num_mgm_entry_size
335 2. Use the CPU near local NUMA node to which the PCIe adapter is connected,
336 for better performance. For VMs, verify that the right CPU
337 and NUMA node are pinned according to the above. Run:
339 .. code-block:: console
343 to identify the NUMA node to which the PCIe adapter is connected.
345 3. If more than one adapter is used, and root complex capabilities allow
346 to put both adapters on the same NUMA node without PCI bandwidth degradation,
347 it is recommended to locate both adapters on the same NUMA node.
348 This in order to forward packets from one to the other without
349 NUMA performance penalty.
351 4. Disable pause frames:
353 .. code-block:: console
355 ethtool -A <netdev> rx off tx off
357 5. Verify IO non-posted prefetch is disabled by default. This can be checked
358 via the BIOS configuration. Please contact you server provider for more
359 information about the settings.
363 On some machines, depends on the machine integrator, it is beneficial
364 to set the PCI max read request parameter to 1K. This can be
365 done in the following way:
367 To query the read request size use:
369 .. code-block:: console
371 setpci -s <NIC PCI address> 68.w
373 If the output is different than 3XXX, set it by:
375 .. code-block:: console
377 setpci -s <NIC PCI address> 68.w=3XXX
379 The XXX can be different on different systems. Make sure to configure
380 according to the setpci output.
385 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
386 devices managed by librte_pmd_mlx4.
388 #. Load the kernel modules:
390 .. code-block:: console
392 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
394 Alternatively if MLNX_OFED is fully installed, the following script can
397 .. code-block:: console
399 /etc/init.d/openibd restart
403 User space I/O kernel modules (uio and igb_uio) are not used and do
404 not have to be loaded.
406 #. Make sure Ethernet interfaces are in working order and linked to kernel
407 verbs. Related sysfs entries should be present:
409 .. code-block:: console
411 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
415 .. code-block:: console
422 #. Optionally, retrieve their PCI bus addresses for whitelisting:
424 .. code-block:: console
427 for intf in eth2 eth3 eth4 eth5;
429 (cd "/sys/class/net/${intf}/device/" && pwd -P);
432 sed -n 's,.*/\(.*\),-w \1,p'
436 .. code-block:: console
445 There are only two distinct PCI bus addresses because the Mellanox
446 ConnectX-3 adapters installed on this system are dual port.
448 #. Request huge pages:
450 .. code-block:: console
452 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
454 #. Start testpmd with basic parameters:
456 .. code-block:: console
458 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
462 .. code-block:: console
465 EAL: PCI device 0000:83:00.0 on NUMA socket 1
466 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
467 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
468 PMD: librte_pmd_mlx4: 2 port(s) detected
469 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
470 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
471 EAL: PCI device 0000:84:00.0 on NUMA socket 1
472 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
473 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
474 PMD: librte_pmd_mlx4: 2 port(s) detected
475 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
476 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
477 Interactive-mode selected
478 Configuring Port 0 (socket 0)
479 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
480 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
481 Port 0: 00:02:C9:B5:B7:50
482 Configuring Port 1 (socket 0)
483 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
484 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
485 Port 1: 00:02:C9:B5:B7:51
486 Configuring Port 2 (socket 0)
487 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
488 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
489 Port 2: 00:02:C9:B5:BA:B0
490 Configuring Port 3 (socket 0)
491 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
492 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
493 Port 3: 00:02:C9:B5:BA:B1
494 Checking link statuses...
495 Port 0 Link Up - speed 10000 Mbps - full-duplex
496 Port 1 Link Up - speed 40000 Mbps - full-duplex
497 Port 2 Link Up - speed 10000 Mbps - full-duplex
498 Port 3 Link Up - speed 40000 Mbps - full-duplex