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.
82 - Scattered packets are supported for TX and RX.
91 These options can be modified in the ``.config`` file.
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_DEBUG_BROKEN_VERBS`` (default **n**)
105 Mellanox OFED versions earlier than 4.2 may return false errors from
106 Verbs object destruction APIs after the device is plugged out.
107 Enabling this option replaces assertion checks that cause the program
108 to abort with harmless debugging messages as a workaround.
109 Relevant only when CONFIG_RTE_LIBRTE_MLX4_DEBUG is enabled.
111 - ``CONFIG_RTE_LIBRTE_MLX4_MAX_INLINE`` (default **0**)
113 Amount of data to be inlined during TX operations. Improves latency but
116 - ``CONFIG_RTE_LIBRTE_MLX4_TX_MP_CACHE`` (default **8**)
118 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
119 which buffers are to be transmitted must be associated to memory regions
120 (MRs). This is a slow operation that must be cached.
122 This value is always 1 for RX queues since they use a single MP.
124 Environment variables
125 ~~~~~~~~~~~~~~~~~~~~~
127 - ``MLX4_INLINE_RECV_SIZE``
129 A nonzero value enables inline receive for packets up to that size. May
130 significantly improve performance in some cases but lower it in
131 others. Requires careful testing.
133 Run-time configuration
134 ~~~~~~~~~~~~~~~~~~~~~~
136 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
137 because it is affected by their state. Forcing them down prevents packets
140 - **ethtool** operations on related kernel interfaces also affect the PMD.
142 - ``port`` parameter [int]
144 This parameter provides a physical port to probe and can be specified multiple
145 times for additional ports. All ports are probed by default if left
148 Kernel module parameters
149 ~~~~~~~~~~~~~~~~~~~~~~~~
151 The **mlx4_core** kernel module has several parameters that affect the
152 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
155 - **num_vfs** (integer or triplet, optionally prefixed by device address
158 Create the given number of VFs on the specified devices.
160 - **log_num_mgm_entry_size** (integer)
162 Device-managed flow steering (DMFS) is required by DPDK applications. It is
163 enabled by using a negative value, the last four bits of which have a
166 - **-1**: force device-managed flow steering (DMFS).
167 - **-7**: configure optimized steering mode to improve performance with the
168 following limitation: VLAN filtering is not supported with this mode.
169 This is the recommended mode in case VLAN filter is not needed.
174 This driver relies on external libraries and kernel drivers for resources
175 allocations and initialization. The following dependencies are not part of
176 DPDK and must be installed separately:
180 User space verbs framework used by librte_pmd_mlx4. This library provides
181 a generic interface between the kernel and low-level user space drivers
184 It allows slow and privileged operations (context initialization, hardware
185 resources allocations) to be managed by the kernel and fast operations to
186 never leave user space.
190 Low-level user space driver library for Mellanox ConnectX-3 devices,
191 it is automatically loaded by libibverbs.
193 This library basically implements send/receive calls to the hardware
196 - **Kernel modules** (mlnx-ofed-kernel)
198 They provide the kernel-side verbs API and low level device drivers that
199 manage actual hardware initialization and resources sharing with user
202 Unlike most other PMDs, these modules must remain loaded and bound to
205 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
206 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
207 - mlx4_ib: InifiniBand device driver.
208 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
210 - **Firmware update**
212 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
214 Because each release provides new features, these updates must be applied to
215 match the kernel modules and libraries they come with.
219 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
222 Currently supported by DPDK:
224 - Mellanox OFED **4.1**.
225 - Firmware version **2.36.5000** and above.
227 Getting Mellanox OFED
228 ~~~~~~~~~~~~~~~~~~~~~
230 While these libraries and kernel modules are available on OpenFabrics
231 Alliance's `website <https://www.openfabrics.org/>`_ and provided by package
232 managers on most distributions, this PMD requires Ethernet extensions that
233 may not be supported at the moment (this is a work in progress).
236 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers>`_
237 includes the necessary support and should be used in the meantime. For DPDK,
238 only libibverbs, libmlx4, mlnx-ofed-kernel packages and firmware updates are
239 required from that distribution.
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`_.
250 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
255 1. Download latest Mellanox OFED. For more info check the `prerequisites`_.
257 2. Install the required libraries and kernel modules either by installing
258 only the required set, or by installing the entire Mellanox OFED:
262 .. code-block:: console
266 For SR-IOV hypervisors use:
268 .. code-block:: console
270 ./mlnxofedinstall --enable-sriov -hypervisor
272 For SR-IOV virtual machine use:
274 .. code-block:: console
276 ./mlnxofedinstall --guest
278 3. Verify the firmware is the correct one:
280 .. code-block:: console
284 4. Set all ports links to Ethernet, follow instructions on the screen:
286 .. code-block:: console
290 Or in the manual way:
292 .. code-block:: console
294 PCI=<NIC PCI address>
295 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port0"
296 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port1"
298 5. In case of bare metal or hypervisor, configure optimized steering mode
299 by adding the following line to ``/etc/modprobe.d/mlx4_core.conf``:
301 .. code-block:: console
303 options mlx4_core log_num_mgm_entry_size=-7
307 If VLAN filtering is used, set log_num_mgm_entry_size=-1.
308 Performance degradation can occur on this case.
310 6. Restart the driver:
312 .. code-block:: console
314 /etc/init.d/openibd restart
318 .. code-block:: console
320 service openibd restart
322 7. Compile DPDK and you are ready to go. See instructions on
323 :ref:`Development Kit Build System <Development_Kit_Build_System>`
328 1. Verify the optimized steering mode is configured:
330 .. code-block:: console
332 cat /sys/module/mlx4_core/parameters/log_num_mgm_entry_size
334 2. Use environment variable MLX4_INLINE_RECV_SIZE=64 to get maximum
335 performance for 64B messages.
337 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
338 for better performance. For VMs, verify that the right CPU
339 and NUMA node are pinned according to the above. Run:
341 .. code-block:: console
345 to identify the NUMA node to which the PCIe adapter is connected.
347 4. If more than one adapter is used, and root complex capabilities allow
348 to put both adapters on the same NUMA node without PCI bandwidth degradation,
349 it is recommended to locate both adapters on the same NUMA node.
350 This in order to forward packets from one to the other without
351 NUMA performance penalty.
353 5. Disable pause frames:
355 .. code-block:: console
357 ethtool -A <netdev> rx off tx off
359 6. Verify IO non-posted prefetch is disabled by default. This can be checked
360 via the BIOS configuration. Please contact you server provider for more
361 information about the settings.
365 On some machines, depends on the machine integrator, it is beneficial
366 to set the PCI max read request parameter to 1K. This can be
367 done in the following way:
369 To query the read request size use:
371 .. code-block:: console
373 setpci -s <NIC PCI address> 68.w
375 If the output is different than 3XXX, set it by:
377 .. code-block:: console
379 setpci -s <NIC PCI address> 68.w=3XXX
381 The XXX can be different on different systems. Make sure to configure
382 according to the setpci output.
387 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
388 devices managed by librte_pmd_mlx4.
390 #. Load the kernel modules:
392 .. code-block:: console
394 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
396 Alternatively if MLNX_OFED is fully installed, the following script can
399 .. code-block:: console
401 /etc/init.d/openibd restart
405 User space I/O kernel modules (uio and igb_uio) are not used and do
406 not have to be loaded.
408 #. Make sure Ethernet interfaces are in working order and linked to kernel
409 verbs. Related sysfs entries should be present:
411 .. code-block:: console
413 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
417 .. code-block:: console
424 #. Optionally, retrieve their PCI bus addresses for whitelisting:
426 .. code-block:: console
429 for intf in eth2 eth3 eth4 eth5;
431 (cd "/sys/class/net/${intf}/device/" && pwd -P);
434 sed -n 's,.*/\(.*\),-w \1,p'
438 .. code-block:: console
447 There are only two distinct PCI bus addresses because the Mellanox
448 ConnectX-3 adapters installed on this system are dual port.
450 #. Request huge pages:
452 .. code-block:: console
454 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
456 #. Start testpmd with basic parameters:
458 .. code-block:: console
460 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
464 .. code-block:: console
467 EAL: PCI device 0000:83:00.0 on NUMA socket 1
468 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
469 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
470 PMD: librte_pmd_mlx4: 2 port(s) detected
471 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
472 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
473 EAL: PCI device 0000:84:00.0 on NUMA socket 1
474 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
475 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
476 PMD: librte_pmd_mlx4: 2 port(s) detected
477 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
478 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
479 Interactive-mode selected
480 Configuring Port 0 (socket 0)
481 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
482 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
483 Port 0: 00:02:C9:B5:B7:50
484 Configuring Port 1 (socket 0)
485 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
486 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
487 Port 1: 00:02:C9:B5:B7:51
488 Configuring Port 2 (socket 0)
489 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
490 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
491 Port 2: 00:02:C9:B5:BA:B0
492 Configuring Port 3 (socket 0)
493 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
494 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
495 Port 3: 00:02:C9:B5:BA:B1
496 Checking link statuses...
497 Port 0 Link Up - speed 10000 Mbps - full-duplex
498 Port 1 Link Up - speed 40000 Mbps - full-duplex
499 Port 2 Link Up - speed 10000 Mbps - full-duplex
500 Port 3 Link Up - speed 40000 Mbps - full-duplex