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.
80 - Multi arch support: x86_64 and POWER8.
81 - Link state information is provided.
90 These options can be modified in the ``.config`` file.
92 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
94 Toggle compilation of librte_pmd_mlx4 itself.
96 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
98 Toggle debugging code and stricter compilation flags. Enabling this option
99 adds additional run-time checks and debugging messages at the cost of
102 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG_BROKEN_VERBS`` (default **n**)
104 Mellanox OFED versions earlier than 4.2 may return false errors from
105 Verbs object destruction APIs after the device is plugged out.
106 Enabling this option replaces assertion checks that cause the program
107 to abort with harmless debugging messages as a workaround.
108 Relevant only when CONFIG_RTE_LIBRTE_MLX4_DEBUG is enabled.
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 Environment variables
124 ~~~~~~~~~~~~~~~~~~~~~
126 - ``MLX4_INLINE_RECV_SIZE``
128 A nonzero value enables inline receive for packets up to that size. May
129 significantly improve performance in some cases but lower it in
130 others. Requires careful testing.
132 Run-time configuration
133 ~~~~~~~~~~~~~~~~~~~~~~
135 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
136 because it is affected by their state. Forcing them down prevents packets
139 - **ethtool** operations on related kernel interfaces also affect the PMD.
141 - ``port`` parameter [int]
143 This parameter provides a physical port to probe and can be specified multiple
144 times for additional ports. All ports are probed by default if left
147 Kernel module parameters
148 ~~~~~~~~~~~~~~~~~~~~~~~~
150 The **mlx4_core** kernel module has several parameters that affect the
151 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
154 - **num_vfs** (integer or triplet, optionally prefixed by device address
157 Create the given number of VFs on the specified devices.
159 - **log_num_mgm_entry_size** (integer)
161 Device-managed flow steering (DMFS) is required by DPDK applications. It is
162 enabled by using a negative value, the last four bits of which have a
165 - **-1**: force device-managed flow steering (DMFS).
166 - **-7**: configure optimized steering mode to improve performance with the
167 following limitation: VLAN filtering is not supported with this mode.
168 This is the recommended mode in case VLAN filter is not needed.
173 This driver relies on external libraries and kernel drivers for resources
174 allocations and initialization. The following dependencies are not part of
175 DPDK and must be installed separately:
179 User space verbs framework used by librte_pmd_mlx4. This library provides
180 a generic interface between the kernel and low-level user space drivers
183 It allows slow and privileged operations (context initialization, hardware
184 resources allocations) to be managed by the kernel and fast operations to
185 never leave user space.
189 Low-level user space driver library for Mellanox ConnectX-3 devices,
190 it is automatically loaded by libibverbs.
192 This library basically implements send/receive calls to the hardware
195 - **Kernel modules** (mlnx-ofed-kernel)
197 They provide the kernel-side verbs API and low level device drivers that
198 manage actual hardware initialization and resources sharing with user
201 Unlike most other PMDs, these modules must remain loaded and bound to
204 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
205 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
206 - mlx4_ib: InifiniBand device driver.
207 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
209 - **Firmware update**
211 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
213 Because each release provides new features, these updates must be applied to
214 match the kernel modules and libraries they come with.
218 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
221 Currently supported by DPDK:
223 - Mellanox OFED **4.1**.
224 - Firmware version **2.36.5000** and above.
226 Getting Mellanox OFED
227 ~~~~~~~~~~~~~~~~~~~~~
229 While these libraries and kernel modules are available on OpenFabrics
230 Alliance's `website <https://www.openfabrics.org/>`_ and provided by package
231 managers on most distributions, this PMD requires Ethernet extensions that
232 may not be supported at the moment (this is a work in progress).
235 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers>`_
236 includes the necessary support and should be used in the meantime. For DPDK,
237 only libibverbs, libmlx4, mlnx-ofed-kernel packages and firmware updates are
238 required from that distribution.
242 Several versions of Mellanox OFED are available. Installing the version
243 this DPDK release was developed and tested against is strongly
244 recommended. Please check the `prerequisites`_.
249 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
254 1. Download latest Mellanox OFED. For more info check the `prerequisites`_.
256 2. Install the required libraries and kernel modules either by installing
257 only the required set, or by installing the entire Mellanox OFED:
261 .. code-block:: console
265 For SR-IOV hypervisors use:
267 .. code-block:: console
269 ./mlnxofedinstall --enable-sriov -hypervisor
271 For SR-IOV virtual machine use:
273 .. code-block:: console
275 ./mlnxofedinstall --guest
277 3. Verify the firmware is the correct one:
279 .. code-block:: console
283 4. Set all ports links to Ethernet, follow instructions on the screen:
285 .. code-block:: console
289 Or in the manual way:
291 .. code-block:: console
293 PCI=<NIC PCI address>
294 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port0"
295 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port1"
297 5. In case of bare metal or hypervisor, configure optimized steering mode
298 by adding the following line to ``/etc/modprobe.d/mlx4_core.conf``:
300 .. code-block:: console
302 options mlx4_core log_num_mgm_entry_size=-7
306 If VLAN filtering is used, set log_num_mgm_entry_size=-1.
307 Performance degradation can occur on this case.
309 6. Restart the driver:
311 .. code-block:: console
313 /etc/init.d/openibd restart
317 .. code-block:: console
319 service openibd restart
321 7. Compile DPDK and you are ready to go. See instructions on
322 :ref:`Development Kit Build System <Development_Kit_Build_System>`
327 1. Verify the optimized steering mode is configured:
329 .. code-block:: console
331 cat /sys/module/mlx4_core/parameters/log_num_mgm_entry_size
333 2. Use environment variable MLX4_INLINE_RECV_SIZE=64 to get maximum
334 performance for 64B messages.
336 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
337 for better performance. For VMs, verify that the right CPU
338 and NUMA node are pinned according to the above. Run:
340 .. code-block:: console
344 to identify the NUMA node to which the PCIe adapter is connected.
346 4. If more than one adapter is used, and root complex capabilities allow
347 to put both adapters on the same NUMA node without PCI bandwidth degradation,
348 it is recommended to locate both adapters on the same NUMA node.
349 This in order to forward packets from one to the other without
350 NUMA performance penalty.
352 5. Disable pause frames:
354 .. code-block:: console
356 ethtool -A <netdev> rx off tx off
358 6. Verify IO non-posted prefetch is disabled by default. This can be checked
359 via the BIOS configuration. Please contact you server provider for more
360 information about the settings.
364 On some machines, depends on the machine integrator, it is beneficial
365 to set the PCI max read request parameter to 1K. This can be
366 done in the following way:
368 To query the read request size use:
370 .. code-block:: console
372 setpci -s <NIC PCI address> 68.w
374 If the output is different than 3XXX, set it by:
376 .. code-block:: console
378 setpci -s <NIC PCI address> 68.w=3XXX
380 The XXX can be different on different systems. Make sure to configure
381 according to the setpci output.
386 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
387 devices managed by librte_pmd_mlx4.
389 #. Load the kernel modules:
391 .. code-block:: console
393 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
395 Alternatively if MLNX_OFED is fully installed, the following script can
398 .. code-block:: console
400 /etc/init.d/openibd restart
404 User space I/O kernel modules (uio and igb_uio) are not used and do
405 not have to be loaded.
407 #. Make sure Ethernet interfaces are in working order and linked to kernel
408 verbs. Related sysfs entries should be present:
410 .. code-block:: console
412 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
416 .. code-block:: console
423 #. Optionally, retrieve their PCI bus addresses for whitelisting:
425 .. code-block:: console
428 for intf in eth2 eth3 eth4 eth5;
430 (cd "/sys/class/net/${intf}/device/" && pwd -P);
433 sed -n 's,.*/\(.*\),-w \1,p'
437 .. code-block:: console
446 There are only two distinct PCI bus addresses because the Mellanox
447 ConnectX-3 adapters installed on this system are dual port.
449 #. Request huge pages:
451 .. code-block:: console
453 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
455 #. Start testpmd with basic parameters:
457 .. code-block:: console
459 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
463 .. code-block:: console
466 EAL: PCI device 0000:83:00.0 on NUMA socket 1
467 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
468 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
469 PMD: librte_pmd_mlx4: 2 port(s) detected
470 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
471 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
472 EAL: PCI device 0000:84:00.0 on NUMA socket 1
473 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
474 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
475 PMD: librte_pmd_mlx4: 2 port(s) detected
476 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
477 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
478 Interactive-mode selected
479 Configuring Port 0 (socket 0)
480 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
481 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
482 Port 0: 00:02:C9:B5:B7:50
483 Configuring Port 1 (socket 0)
484 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
485 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
486 Port 1: 00:02:C9:B5:B7:51
487 Configuring Port 2 (socket 0)
488 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
489 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
490 Port 2: 00:02:C9:B5:BA:B0
491 Configuring Port 3 (socket 0)
492 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
493 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
494 Port 3: 00:02:C9:B5:BA:B1
495 Checking link statuses...
496 Port 0 Link Up - speed 10000 Mbps - full-duplex
497 Port 1 Link Up - speed 40000 Mbps - full-duplex
498 Port 2 Link Up - speed 10000 Mbps - full-duplex
499 Port 3 Link Up - speed 40000 Mbps - full-duplex