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 - RSS, also known as RCA, is supported. In this mode the number of
82 configured RX queues must be a power of two.
83 - VLAN filtering is supported.
84 - Link state information is provided.
85 - Multiple MAC addresses (unicast, multicast) can be configured.
86 - Scattered packets are supported for TX and RX.
87 - Inner L3/L4 (IP, TCP and UDP) TX/RX checksum offloading and validation.
88 - Outer L3 (IP) TX/RX checksum offloading and validation for VXLAN frames.
94 - RSS hash key cannot be modified.
95 - RSS RETA cannot be configured
96 - RSS always includes L3 (IPv4/IPv6) and L4 (UDP/TCP). They cannot be
105 These options can be modified in the ``.config`` file.
107 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
109 Toggle compilation of librte_pmd_mlx4 itself.
111 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
113 Toggle debugging code and stricter compilation flags. Enabling this option
114 adds additional run-time checks and debugging messages at the cost of
117 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG_BROKEN_VERBS`` (default **n**)
119 Mellanox OFED versions earlier than 4.2 may return false errors from
120 Verbs object destruction APIs after the device is plugged out.
121 Enabling this option replaces assertion checks that cause the program
122 to abort with harmless debugging messages as a workaround.
123 Relevant only when CONFIG_RTE_LIBRTE_MLX4_DEBUG is enabled.
125 - ``CONFIG_RTE_LIBRTE_MLX4_MAX_INLINE`` (default **0**)
127 Amount of data to be inlined during TX operations. Improves latency but
130 - ``CONFIG_RTE_LIBRTE_MLX4_TX_MP_CACHE`` (default **8**)
132 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
133 which buffers are to be transmitted must be associated to memory regions
134 (MRs). This is a slow operation that must be cached.
136 This value is always 1 for RX queues since they use a single MP.
138 Environment variables
139 ~~~~~~~~~~~~~~~~~~~~~
141 - ``MLX4_INLINE_RECV_SIZE``
143 A nonzero value enables inline receive for packets up to that size. May
144 significantly improve performance in some cases but lower it in
145 others. Requires careful testing.
147 Run-time configuration
148 ~~~~~~~~~~~~~~~~~~~~~~
150 - The only constraint when RSS mode is requested is to make sure the number
151 of RX queues is a power of two. This is a hardware requirement.
153 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
154 because it is affected by their state. Forcing them down prevents packets
157 - **ethtool** operations on related kernel interfaces also affect the PMD.
159 - ``port`` parameter [int]
161 This parameter provides a physical port to probe and can be specified multiple
162 times for additional ports. All ports are probed by default if left
165 Kernel module parameters
166 ~~~~~~~~~~~~~~~~~~~~~~~~
168 The **mlx4_core** kernel module has several parameters that affect the
169 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
172 - **num_vfs** (integer or triplet, optionally prefixed by device address
175 Create the given number of VFs on the specified devices.
177 - **log_num_mgm_entry_size** (integer)
179 Device-managed flow steering (DMFS) is required by DPDK applications. It is
180 enabled by using a negative value, the last four bits of which have a
183 - **-1**: force device-managed flow steering (DMFS).
184 - **-7**: configure optimized steering mode to improve performance with the
185 following limitation: VLAN filtering is not supported with this mode.
186 This is the recommended mode in case VLAN filter is not needed.
191 This driver relies on external libraries and kernel drivers for resources
192 allocations and initialization. The following dependencies are not part of
193 DPDK and must be installed separately:
197 User space verbs framework used by librte_pmd_mlx4. This library provides
198 a generic interface between the kernel and low-level user space drivers
201 It allows slow and privileged operations (context initialization, hardware
202 resources allocations) to be managed by the kernel and fast operations to
203 never leave user space.
207 Low-level user space driver library for Mellanox ConnectX-3 devices,
208 it is automatically loaded by libibverbs.
210 This library basically implements send/receive calls to the hardware
213 - **Kernel modules** (mlnx-ofed-kernel)
215 They provide the kernel-side verbs API and low level device drivers that
216 manage actual hardware initialization and resources sharing with user
219 Unlike most other PMDs, these modules must remain loaded and bound to
222 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
223 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
224 - mlx4_ib: InifiniBand device driver.
225 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
227 - **Firmware update**
229 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
231 Because each release provides new features, these updates must be applied to
232 match the kernel modules and libraries they come with.
236 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
239 Currently supported by DPDK:
241 - Mellanox OFED **4.1**.
242 - Firmware version **2.36.5000** and above.
244 Getting Mellanox OFED
245 ~~~~~~~~~~~~~~~~~~~~~
247 While these libraries and kernel modules are available on OpenFabrics
248 Alliance's `website <https://www.openfabrics.org/>`_ and provided by package
249 managers on most distributions, this PMD requires Ethernet extensions that
250 may not be supported at the moment (this is a work in progress).
253 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers>`_
254 includes the necessary support and should be used in the meantime. For DPDK,
255 only libibverbs, libmlx4, mlnx-ofed-kernel packages and firmware updates are
256 required from that distribution.
260 Several versions of Mellanox OFED are available. Installing the version
261 this DPDK release was developed and tested against is strongly
262 recommended. Please check the `prerequisites`_.
267 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
272 1. Download latest Mellanox OFED. For more info check the `prerequisites`_.
274 2. Install the required libraries and kernel modules either by installing
275 only the required set, or by installing the entire Mellanox OFED:
279 .. code-block:: console
283 For SR-IOV hypervisors use:
285 .. code-block:: console
287 ./mlnxofedinstall --enable-sriov -hypervisor
289 For SR-IOV virtual machine use:
291 .. code-block:: console
293 ./mlnxofedinstall --guest
295 3. Verify the firmware is the correct one:
297 .. code-block:: console
301 4. Set all ports links to Ethernet, follow instructions on the screen:
303 .. code-block:: console
307 Or in the manual way:
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"
315 5. In case of bare metal or hypervisor, configure optimized steering mode
316 by adding the following line to ``/etc/modprobe.d/mlx4_core.conf``:
318 .. code-block:: console
320 options mlx4_core log_num_mgm_entry_size=-7
324 If VLAN filtering is used, set log_num_mgm_entry_size=-1.
325 Performance degradation can occur on this case.
327 6. Restart the driver:
329 .. code-block:: console
331 /etc/init.d/openibd restart
335 .. code-block:: console
337 service openibd restart
339 7. Compile DPDK and you are ready to go. See instructions on
340 :ref:`Development Kit Build System <Development_Kit_Build_System>`
345 1. Verify the optimized steering mode is configured:
347 .. code-block:: console
349 cat /sys/module/mlx4_core/parameters/log_num_mgm_entry_size
351 2. Use environment variable MLX4_INLINE_RECV_SIZE=64 to get maximum
352 performance for 64B messages.
354 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
355 for better performance. For VMs, verify that the right CPU
356 and NUMA node are pinned according to the above. Run:
358 .. code-block:: console
362 to identify the NUMA node to which the PCIe adapter is connected.
364 4. If more than one adapter is used, and root complex capabilities allow
365 to put both adapters on the same NUMA node without PCI bandwidth degradation,
366 it is recommended to locate both adapters on the same NUMA node.
367 This in order to forward packets from one to the other without
368 NUMA performance penalty.
370 5. Disable pause frames:
372 .. code-block:: console
374 ethtool -A <netdev> rx off tx off
376 6. Verify IO non-posted prefetch is disabled by default. This can be checked
377 via the BIOS configuration. Please contact you server provider for more
378 information about the settings.
382 On some machines, depends on the machine integrator, it is beneficial
383 to set the PCI max read request parameter to 1K. This can be
384 done in the following way:
386 To query the read request size use:
388 .. code-block:: console
390 setpci -s <NIC PCI address> 68.w
392 If the output is different than 3XXX, set it by:
394 .. code-block:: console
396 setpci -s <NIC PCI address> 68.w=3XXX
398 The XXX can be different on different systems. Make sure to configure
399 according to the setpci output.
404 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
405 devices managed by librte_pmd_mlx4.
407 #. Load the kernel modules:
409 .. code-block:: console
411 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
413 Alternatively if MLNX_OFED is fully installed, the following script can
416 .. code-block:: console
418 /etc/init.d/openibd restart
422 User space I/O kernel modules (uio and igb_uio) are not used and do
423 not have to be loaded.
425 #. Make sure Ethernet interfaces are in working order and linked to kernel
426 verbs. Related sysfs entries should be present:
428 .. code-block:: console
430 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
434 .. code-block:: console
441 #. Optionally, retrieve their PCI bus addresses for whitelisting:
443 .. code-block:: console
446 for intf in eth2 eth3 eth4 eth5;
448 (cd "/sys/class/net/${intf}/device/" && pwd -P);
451 sed -n 's,.*/\(.*\),-w \1,p'
455 .. code-block:: console
464 There are only two distinct PCI bus addresses because the Mellanox
465 ConnectX-3 adapters installed on this system are dual port.
467 #. Request huge pages:
469 .. code-block:: console
471 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
473 #. Start testpmd with basic parameters:
475 .. code-block:: console
477 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
481 .. code-block:: console
484 EAL: PCI device 0000:83:00.0 on NUMA socket 1
485 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
486 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
487 PMD: librte_pmd_mlx4: 2 port(s) detected
488 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
489 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
490 EAL: PCI device 0000:84:00.0 on NUMA socket 1
491 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
492 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
493 PMD: librte_pmd_mlx4: 2 port(s) detected
494 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
495 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
496 Interactive-mode selected
497 Configuring Port 0 (socket 0)
498 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
499 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
500 Port 0: 00:02:C9:B5:B7:50
501 Configuring Port 1 (socket 0)
502 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
503 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
504 Port 1: 00:02:C9:B5:B7:51
505 Configuring Port 2 (socket 0)
506 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
507 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
508 Port 2: 00:02:C9:B5:BA:B0
509 Configuring Port 3 (socket 0)
510 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
511 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
512 Port 3: 00:02:C9:B5:BA:B1
513 Checking link statuses...
514 Port 0 Link Up - speed 10000 Mbps - full-duplex
515 Port 1 Link Up - speed 40000 Mbps - full-duplex
516 Port 2 Link Up - speed 10000 Mbps - full-duplex
517 Port 3 Link Up - speed 40000 Mbps - full-duplex