1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2012 6WIND S.A.
3 Copyright 2015 Mellanox Technologies, Ltd
5 MLX4 poll mode driver library
6 =============================
8 The MLX4 poll mode driver library (**librte_pmd_mlx4**) implements support
9 for **Mellanox ConnectX-3** and **Mellanox ConnectX-3 Pro** 10/40 Gbps adapters
10 as well as their virtual functions (VF) in SR-IOV context.
12 Information and documentation about this family of adapters can be found on
13 the `Mellanox website <http://www.mellanox.com>`_. Help is also provided by
14 the `Mellanox community <http://community.mellanox.com/welcome>`_.
16 There is also a `section dedicated to this poll mode driver
17 <http://www.mellanox.com/page/products_dyn?product_family=209&mtag=pmd_for_dpdk>`_.
21 Due to external dependencies, this driver is disabled by default. It must
22 be enabled manually by setting ``CONFIG_RTE_LIBRTE_MLX4_PMD=y`` and
25 Implementation details
26 ----------------------
28 Most Mellanox ConnectX-3 devices provide two ports but expose a single PCI
29 bus address, thus unlike most drivers, librte_pmd_mlx4 registers itself as a
30 PCI driver that allocates one Ethernet device per detected port.
32 For this reason, one cannot white/blacklist a single port without also
33 white/blacklisting the others on the same device.
35 Besides its dependency on libibverbs (that implies libmlx4 and associated
36 kernel support), librte_pmd_mlx4 relies heavily on system calls for control
37 operations such as querying/updating the MTU and flow control parameters.
39 For security reasons and robustness, this driver only deals with virtual
40 memory addresses. The way resources allocations are handled by the kernel
41 combined with hardware specifications that allow it to handle virtual memory
42 addresses directly ensure that DPDK applications cannot access random
43 physical memory (or memory that does not belong to the current process).
45 This capability allows the PMD to coexist with kernel network interfaces
46 which remain functional, although they stop receiving unicast packets as
47 long as they share the same MAC address.
49 The :ref:`flow_isolated_mode` is supported.
51 Compiling librte_pmd_mlx4 causes DPDK to be linked against libibverbs.
59 These options can be modified in the ``.config`` file.
61 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
63 Toggle compilation of librte_pmd_mlx4 itself.
65 - ``CONFIG_RTE_IBVERBS_LINK_DLOPEN`` (default **n**)
67 Build PMD with additional code to make it loadable without hard
68 dependencies on **libibverbs** nor **libmlx4**, which may not be installed
71 In this mode, their presence is still required for it to run properly,
72 however their absence won't prevent a DPDK application from starting (with
73 ``CONFIG_RTE_BUILD_SHARED_LIB`` disabled) and they won't show up as
74 missing with ``ldd(1)``.
76 It works by moving these dependencies to a purpose-built rdma-core "glue"
77 plug-in which must either be installed in a directory whose name is based
78 on ``CONFIG_RTE_EAL_PMD_PATH`` suffixed with ``-glue`` if set, or in a
79 standard location for the dynamic linker (e.g. ``/lib``) if left to the
80 default empty string (``""``).
82 This option has no performance impact.
84 - ``CONFIG_RTE_IBVERBS_LINK_STATIC`` (default **n**)
86 Embed static flavour of the dependencies **libibverbs** and **libmlx4**
87 in the PMD shared library or the executable static binary.
89 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
91 Toggle debugging code and stricter compilation flags. Enabling this option
92 adds additional run-time checks and debugging messages at the cost of
100 A list of directories in which to search for the rdma-core "glue" plug-in,
101 separated by colons or semi-colons.
103 Only matters when compiled with ``CONFIG_RTE_IBVERBS_LINK_DLOPEN``
104 enabled and most useful when ``CONFIG_RTE_EAL_PMD_PATH`` is also set,
105 since ``LD_LIBRARY_PATH`` has no effect in this case.
107 Run-time configuration
108 ~~~~~~~~~~~~~~~~~~~~~~
110 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
111 because it is affected by their state. Forcing them down prevents packets
114 - **ethtool** operations on related kernel interfaces also affect the PMD.
116 - ``port`` parameter [int]
118 This parameter provides a physical port to probe and can be specified multiple
119 times for additional ports. All ports are probed by default if left
122 Kernel module parameters
123 ~~~~~~~~~~~~~~~~~~~~~~~~
125 The **mlx4_core** kernel module has several parameters that affect the
126 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
129 - **num_vfs** (integer or triplet, optionally prefixed by device address
132 Create the given number of VFs on the specified devices.
134 - **log_num_mgm_entry_size** (integer)
136 Device-managed flow steering (DMFS) is required by DPDK applications. It is
137 enabled by using a negative value, the last four bits of which have a
140 - **-1**: force device-managed flow steering (DMFS).
141 - **-7**: configure optimized steering mode to improve performance with the
142 following limitation: VLAN filtering is not supported with this mode.
143 This is the recommended mode in case VLAN filter is not needed.
148 - CRC stripping is supported by default and always reported as "true".
149 The ability to enable/disable CRC stripping requires OFED version
150 4.3-1.5.0.0 and above or rdma-core version v18 and above.
152 - TSO (Transmit Segmentation Offload) is supported in OFED version
158 This driver relies on external libraries and kernel drivers for resources
159 allocations and initialization. The following dependencies are not part of
160 DPDK and must be installed separately:
162 - **libibverbs** (provided by rdma-core package)
164 User space verbs framework used by librte_pmd_mlx4. This library provides
165 a generic interface between the kernel and low-level user space drivers
168 It allows slow and privileged operations (context initialization, hardware
169 resources allocations) to be managed by the kernel and fast operations to
170 never leave user space.
172 - **libmlx4** (provided by rdma-core package)
174 Low-level user space driver library for Mellanox ConnectX-3 devices,
175 it is automatically loaded by libibverbs.
177 This library basically implements send/receive calls to the hardware
182 They provide the kernel-side verbs API and low level device drivers that
183 manage actual hardware initialization and resources sharing with user
186 Unlike most other PMDs, these modules must remain loaded and bound to
189 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
190 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
191 - mlx4_ib: InifiniBand device driver.
192 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
194 - **Firmware update**
196 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
198 Because each release provides new features, these updates must be applied to
199 match the kernel modules and libraries they come with.
203 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
206 Depending on system constraints and user preferences either RDMA core library
207 with a recent enough Linux kernel release (recommended) or Mellanox OFED,
208 which provides compatibility with older releases.
210 Current RDMA core package and Linux kernel (recommended)
211 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
213 - Minimal Linux kernel version: 4.14.
214 - Minimal RDMA core version: v15 (see `RDMA core installation documentation`_).
216 - Starting with rdma-core v21, static libraries can be built::
219 CFLAGS=-fPIC cmake -DIN_PLACE=1 -DENABLE_STATIC=1 -GNinja ..
222 .. _`RDMA core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
224 If rdma-core libraries are built but not installed, DPDK makefile can link them,
225 thanks to these environment variables:
227 - ``EXTRA_CFLAGS=-I/path/to/rdma-core/build/include``
228 - ``EXTRA_LDFLAGS=-L/path/to/rdma-core/build/lib``
229 - ``PKG_CONFIG_PATH=/path/to/rdma-core/build/lib/pkgconfig``
231 .. _Mellanox_OFED_as_a_fallback:
233 Mellanox OFED as a fallback
234 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
236 - `Mellanox OFED`_ version: **4.4, 4.5**.
237 - firmware version: **2.42.5000** and above.
239 .. _`Mellanox OFED`: http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers
243 Several versions of Mellanox OFED are available. Installing the version
244 this DPDK release was developed and tested against is strongly
245 recommended. Please check the `prerequisites`_.
247 Installing Mellanox OFED
248 ^^^^^^^^^^^^^^^^^^^^^^^^
250 1. Download latest Mellanox OFED.
252 2. Install the required libraries and kernel modules either by installing
253 only the required set, or by installing the entire Mellanox OFED:
257 .. code-block:: console
259 ./mlnxofedinstall --dpdk --upstream-libs
261 For SR-IOV hypervisors use:
263 .. code-block:: console
265 ./mlnxofedinstall --dpdk --upstream-libs --enable-sriov --hypervisor
267 For SR-IOV virtual machine use:
269 .. code-block:: console
271 ./mlnxofedinstall --dpdk --upstream-libs --guest
273 3. Verify the firmware is the correct one:
275 .. code-block:: console
279 4. Set all ports links to Ethernet, follow instructions on the screen:
281 .. code-block:: console
285 5. Continue with :ref:`section 2 of the Quick Start Guide <QSG_2>`.
290 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
297 1. Set all ports links to Ethernet
299 .. code-block:: console
301 PCI=<NIC PCI address>
302 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port0"
303 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port1"
307 If using Mellanox OFED one can permanently set the port link
308 to Ethernet using connectx_port_config tool provided by it.
309 :ref:`Mellanox_OFED_as_a_fallback`:
313 2. In case of bare metal or hypervisor, configure optimized steering mode
314 by adding the following line to ``/etc/modprobe.d/mlx4_core.conf``:
316 .. code-block:: console
318 options mlx4_core log_num_mgm_entry_size=-7
322 If VLAN filtering is used, set log_num_mgm_entry_size=-1.
323 Performance degradation can occur on this case.
325 3. Restart the driver:
327 .. code-block:: console
329 /etc/init.d/openibd restart
333 .. code-block:: console
335 service openibd restart
337 4. Compile DPDK and you are ready to go. See instructions on
338 :ref:`Development Kit Build System <Development_Kit_Build_System>`
343 1. Verify the optimized steering mode is configured:
345 .. code-block:: console
347 cat /sys/module/mlx4_core/parameters/log_num_mgm_entry_size
349 2. Use the CPU near local NUMA node to which the PCIe adapter is connected,
350 for better performance. For VMs, verify that the right CPU
351 and NUMA node are pinned according to the above. Run:
353 .. code-block:: console
357 to identify the NUMA node to which the PCIe adapter is connected.
359 3. If more than one adapter is used, and root complex capabilities allow
360 to put both adapters on the same NUMA node without PCI bandwidth degradation,
361 it is recommended to locate both adapters on the same NUMA node.
362 This in order to forward packets from one to the other without
363 NUMA performance penalty.
365 4. Disable pause frames:
367 .. code-block:: console
369 ethtool -A <netdev> rx off tx off
371 5. Verify IO non-posted prefetch is disabled by default. This can be checked
372 via the BIOS configuration. Please contact you server provider for more
373 information about the settings.
377 On some machines, depends on the machine integrator, it is beneficial
378 to set the PCI max read request parameter to 1K. This can be
379 done in the following way:
381 To query the read request size use:
383 .. code-block:: console
385 setpci -s <NIC PCI address> 68.w
387 If the output is different than 3XXX, set it by:
389 .. code-block:: console
391 setpci -s <NIC PCI address> 68.w=3XXX
393 The XXX can be different on different systems. Make sure to configure
394 according to the setpci output.
396 6. To minimize overhead of searching Memory Regions:
398 - '--socket-mem' is recommended to pin memory by predictable amount.
399 - Configure per-lcore cache when creating Mempools for packet buffer.
400 - Refrain from dynamically allocating/freeing memory in run-time.
405 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
406 devices managed by librte_pmd_mlx4.
408 #. Load the kernel modules:
410 .. code-block:: console
412 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
414 Alternatively if MLNX_OFED is fully installed, the following script can
417 .. code-block:: console
419 /etc/init.d/openibd restart
423 User space I/O kernel modules (uio and igb_uio) are not used and do
424 not have to be loaded.
426 #. Make sure Ethernet interfaces are in working order and linked to kernel
427 verbs. Related sysfs entries should be present:
429 .. code-block:: console
431 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
435 .. code-block:: console
442 #. Optionally, retrieve their PCI bus addresses for whitelisting:
444 .. code-block:: console
447 for intf in eth2 eth3 eth4 eth5;
449 (cd "/sys/class/net/${intf}/device/" && pwd -P);
452 sed -n 's,.*/\(.*\),-w \1,p'
456 .. code-block:: console
465 There are only two distinct PCI bus addresses because the Mellanox
466 ConnectX-3 adapters installed on this system are dual port.
468 #. Request huge pages:
470 .. code-block:: console
472 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
474 #. Start testpmd with basic parameters:
476 .. code-block:: console
478 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
482 .. code-block:: console
485 EAL: PCI device 0000:83:00.0 on NUMA socket 1
486 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
487 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
488 PMD: librte_pmd_mlx4: 2 port(s) detected
489 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
490 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
491 EAL: PCI device 0000:84:00.0 on NUMA socket 1
492 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
493 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
494 PMD: librte_pmd_mlx4: 2 port(s) detected
495 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
496 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
497 Interactive-mode selected
498 Configuring Port 0 (socket 0)
499 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
500 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
501 Port 0: 00:02:C9:B5:B7:50
502 Configuring Port 1 (socket 0)
503 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
504 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
505 Port 1: 00:02:C9:B5:B7:51
506 Configuring Port 2 (socket 0)
507 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
508 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
509 Port 2: 00:02:C9:B5:BA:B0
510 Configuring Port 3 (socket 0)
511 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
512 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
513 Port 3: 00:02:C9:B5:BA:B1
514 Checking link statuses...
515 Port 0 Link Up - speed 10000 Mbps - full-duplex
516 Port 1 Link Up - speed 40000 Mbps - full-duplex
517 Port 2 Link Up - speed 10000 Mbps - full-duplex
518 Port 3 Link Up - speed 40000 Mbps - full-duplex