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 - For secondary process:
150 - Forked secondary process not supported.
151 - All mempools must be initialized before rte_eth_dev_start().
152 - External memory unregistered in EAL memseg list cannot be used for DMA
153 unless such memory has been registered by ``mlx4_mr_update_ext_mp()`` in
154 primary process and remapped to the same virtual address in secondary
155 process. If the external memory is registered by primary process but has
156 different virtual address in secondary process, unexpected error may happen.
158 - CRC stripping is supported by default and always reported as "true".
159 The ability to enable/disable CRC stripping requires OFED version
160 4.3-1.5.0.0 and above or rdma-core version v18 and above.
162 - TSO (Transmit Segmentation Offload) is supported in OFED version
168 This driver relies on external libraries and kernel drivers for resources
169 allocations and initialization. The following dependencies are not part of
170 DPDK and must be installed separately:
172 - **libibverbs** (provided by rdma-core package)
174 User space verbs framework used by librte_pmd_mlx4. This library provides
175 a generic interface between the kernel and low-level user space drivers
178 It allows slow and privileged operations (context initialization, hardware
179 resources allocations) to be managed by the kernel and fast operations to
180 never leave user space.
182 - **libmlx4** (provided by rdma-core package)
184 Low-level user space driver library for Mellanox ConnectX-3 devices,
185 it is automatically loaded by libibverbs.
187 This library basically implements send/receive calls to the hardware
192 They provide the kernel-side verbs API and low level device drivers that
193 manage actual hardware initialization and resources sharing with user
196 Unlike most other PMDs, these modules must remain loaded and bound to
199 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
200 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
201 - mlx4_ib: InifiniBand device driver.
202 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
204 - **Firmware update**
206 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
208 Because each release provides new features, these updates must be applied to
209 match the kernel modules and libraries they come with.
213 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
216 Depending on system constraints and user preferences either RDMA core library
217 with a recent enough Linux kernel release (recommended) or Mellanox OFED,
218 which provides compatibility with older releases.
220 Current RDMA core package and Linux kernel (recommended)
221 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
223 - Minimal Linux kernel version: 4.14.
224 - Minimal RDMA core version: v15 (see `RDMA core installation documentation`_).
226 - Starting with rdma-core v21, static libraries can be built::
229 CFLAGS=-fPIC cmake -DIN_PLACE=1 -DENABLE_STATIC=1 -GNinja ..
232 .. _`RDMA core installation documentation`: https://raw.githubusercontent.com/linux-rdma/rdma-core/master/README.md
234 If rdma-core libraries are built but not installed, DPDK makefile can link them,
235 thanks to these environment variables:
237 - ``EXTRA_CFLAGS=-I/path/to/rdma-core/build/include``
238 - ``EXTRA_LDFLAGS=-L/path/to/rdma-core/build/lib``
239 - ``PKG_CONFIG_PATH=/path/to/rdma-core/build/lib/pkgconfig``
241 .. _Mellanox_OFED_as_a_fallback:
243 Mellanox OFED as a fallback
244 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
246 - `Mellanox OFED`_ version: **4.4, 4.5**.
247 - firmware version: **2.42.5000** and above.
249 .. _`Mellanox OFED`: http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers
253 Several versions of Mellanox OFED are available. Installing the version
254 this DPDK release was developed and tested against is strongly
255 recommended. Please check the `prerequisites`_.
257 Installing Mellanox OFED
258 ^^^^^^^^^^^^^^^^^^^^^^^^
260 1. Download latest Mellanox OFED.
262 2. Install the required libraries and kernel modules either by installing
263 only the required set, or by installing the entire Mellanox OFED:
267 .. code-block:: console
269 ./mlnxofedinstall --dpdk --upstream-libs
271 For SR-IOV hypervisors use:
273 .. code-block:: console
275 ./mlnxofedinstall --dpdk --upstream-libs --enable-sriov --hypervisor
277 For SR-IOV virtual machine use:
279 .. code-block:: console
281 ./mlnxofedinstall --dpdk --upstream-libs --guest
283 3. Verify the firmware is the correct one:
285 .. code-block:: console
289 4. Set all ports links to Ethernet, follow instructions on the screen:
291 .. code-block:: console
295 5. Continue with :ref:`section 2 of the Quick Start Guide <QSG_2>`.
300 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
307 1. Set all ports links to Ethernet
309 .. code-block:: console
311 PCI=<NIC PCI address>
312 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port0"
313 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port1"
317 If using Mellanox OFED one can permanently set the port link
318 to Ethernet using connectx_port_config tool provided by it.
319 :ref:`Mellanox_OFED_as_a_fallback`:
323 2. In case of bare metal or hypervisor, configure optimized steering mode
324 by adding the following line to ``/etc/modprobe.d/mlx4_core.conf``:
326 .. code-block:: console
328 options mlx4_core log_num_mgm_entry_size=-7
332 If VLAN filtering is used, set log_num_mgm_entry_size=-1.
333 Performance degradation can occur on this case.
335 3. Restart the driver:
337 .. code-block:: console
339 /etc/init.d/openibd restart
343 .. code-block:: console
345 service openibd restart
347 4. Compile DPDK and you are ready to go. See instructions on
348 :ref:`Development Kit Build System <Development_Kit_Build_System>`
353 1. Verify the optimized steering mode is configured:
355 .. code-block:: console
357 cat /sys/module/mlx4_core/parameters/log_num_mgm_entry_size
359 2. Use the CPU near local NUMA node to which the PCIe adapter is connected,
360 for better performance. For VMs, verify that the right CPU
361 and NUMA node are pinned according to the above. Run:
363 .. code-block:: console
367 to identify the NUMA node to which the PCIe adapter is connected.
369 3. If more than one adapter is used, and root complex capabilities allow
370 to put both adapters on the same NUMA node without PCI bandwidth degradation,
371 it is recommended to locate both adapters on the same NUMA node.
372 This in order to forward packets from one to the other without
373 NUMA performance penalty.
375 4. Disable pause frames:
377 .. code-block:: console
379 ethtool -A <netdev> rx off tx off
381 5. Verify IO non-posted prefetch is disabled by default. This can be checked
382 via the BIOS configuration. Please contact you server provider for more
383 information about the settings.
387 On some machines, depends on the machine integrator, it is beneficial
388 to set the PCI max read request parameter to 1K. This can be
389 done in the following way:
391 To query the read request size use:
393 .. code-block:: console
395 setpci -s <NIC PCI address> 68.w
397 If the output is different than 3XXX, set it by:
399 .. code-block:: console
401 setpci -s <NIC PCI address> 68.w=3XXX
403 The XXX can be different on different systems. Make sure to configure
404 according to the setpci output.
406 6. To minimize overhead of searching Memory Regions:
408 - '--socket-mem' is recommended to pin memory by predictable amount.
409 - Configure per-lcore cache when creating Mempools for packet buffer.
410 - Refrain from dynamically allocating/freeing memory in run-time.
415 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
416 devices managed by librte_pmd_mlx4.
418 #. Load the kernel modules:
420 .. code-block:: console
422 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
424 Alternatively if MLNX_OFED is fully installed, the following script can
427 .. code-block:: console
429 /etc/init.d/openibd restart
433 User space I/O kernel modules (uio and igb_uio) are not used and do
434 not have to be loaded.
436 #. Make sure Ethernet interfaces are in working order and linked to kernel
437 verbs. Related sysfs entries should be present:
439 .. code-block:: console
441 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
445 .. code-block:: console
452 #. Optionally, retrieve their PCI bus addresses for whitelisting:
454 .. code-block:: console
457 for intf in eth2 eth3 eth4 eth5;
459 (cd "/sys/class/net/${intf}/device/" && pwd -P);
462 sed -n 's,.*/\(.*\),-w \1,p'
466 .. code-block:: console
475 There are only two distinct PCI bus addresses because the Mellanox
476 ConnectX-3 adapters installed on this system are dual port.
478 #. Request huge pages:
480 .. code-block:: console
482 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
484 #. Start testpmd with basic parameters:
486 .. code-block:: console
488 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
492 .. code-block:: console
495 EAL: PCI device 0000:83:00.0 on NUMA socket 1
496 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
497 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
498 PMD: librte_pmd_mlx4: 2 port(s) detected
499 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
500 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
501 EAL: PCI device 0000:84:00.0 on NUMA socket 1
502 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
503 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
504 PMD: librte_pmd_mlx4: 2 port(s) detected
505 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
506 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
507 Interactive-mode selected
508 Configuring Port 0 (socket 0)
509 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
510 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
511 Port 0: 00:02:C9:B5:B7:50
512 Configuring Port 1 (socket 0)
513 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
514 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
515 Port 1: 00:02:C9:B5:B7:51
516 Configuring Port 2 (socket 0)
517 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
518 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
519 Port 2: 00:02:C9:B5:BA:B0
520 Configuring Port 3 (socket 0)
521 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
522 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
523 Port 3: 00:02:C9:B5:BA:B1
524 Checking link statuses...
525 Port 0 Link Up - speed 10000 Mbps - full-duplex
526 Port 1 Link Up - speed 40000 Mbps - full-duplex
527 Port 2 Link Up - speed 10000 Mbps - full-duplex
528 Port 3 Link Up - speed 40000 Mbps - full-duplex