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 - Link state information is provided.
84 - Scattered packets are supported for TX and RX.
85 - Inner L3/L4 (IP, TCP and UDP) TX/RX checksum offloading and validation.
86 - Outer L3 (IP) TX/RX checksum offloading and validation for VXLAN frames.
92 - RSS hash key cannot be modified.
93 - RSS RETA cannot be configured
94 - RSS always includes L3 (IPv4/IPv6) and L4 (UDP/TCP). They cannot be
103 These options can be modified in the ``.config`` file.
105 - ``CONFIG_RTE_LIBRTE_MLX4_PMD`` (default **n**)
107 Toggle compilation of librte_pmd_mlx4 itself.
109 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG`` (default **n**)
111 Toggle debugging code and stricter compilation flags. Enabling this option
112 adds additional run-time checks and debugging messages at the cost of
115 - ``CONFIG_RTE_LIBRTE_MLX4_DEBUG_BROKEN_VERBS`` (default **n**)
117 Mellanox OFED versions earlier than 4.2 may return false errors from
118 Verbs object destruction APIs after the device is plugged out.
119 Enabling this option replaces assertion checks that cause the program
120 to abort with harmless debugging messages as a workaround.
121 Relevant only when CONFIG_RTE_LIBRTE_MLX4_DEBUG is enabled.
123 - ``CONFIG_RTE_LIBRTE_MLX4_MAX_INLINE`` (default **0**)
125 Amount of data to be inlined during TX operations. Improves latency but
128 - ``CONFIG_RTE_LIBRTE_MLX4_TX_MP_CACHE`` (default **8**)
130 Maximum number of cached memory pools (MPs) per TX queue. Each MP from
131 which buffers are to be transmitted must be associated to memory regions
132 (MRs). This is a slow operation that must be cached.
134 This value is always 1 for RX queues since they use a single MP.
136 Environment variables
137 ~~~~~~~~~~~~~~~~~~~~~
139 - ``MLX4_INLINE_RECV_SIZE``
141 A nonzero value enables inline receive for packets up to that size. May
142 significantly improve performance in some cases but lower it in
143 others. Requires careful testing.
145 Run-time configuration
146 ~~~~~~~~~~~~~~~~~~~~~~
148 - The only constraint when RSS mode is requested is to make sure the number
149 of RX queues is a power of two. This is a hardware requirement.
151 - librte_pmd_mlx4 brings kernel network interfaces up during initialization
152 because it is affected by their state. Forcing them down prevents packets
155 - **ethtool** operations on related kernel interfaces also affect the PMD.
157 - ``port`` parameter [int]
159 This parameter provides a physical port to probe and can be specified multiple
160 times for additional ports. All ports are probed by default if left
163 Kernel module parameters
164 ~~~~~~~~~~~~~~~~~~~~~~~~
166 The **mlx4_core** kernel module has several parameters that affect the
167 behavior and/or the performance of librte_pmd_mlx4. Some of them are described
170 - **num_vfs** (integer or triplet, optionally prefixed by device address
173 Create the given number of VFs on the specified devices.
175 - **log_num_mgm_entry_size** (integer)
177 Device-managed flow steering (DMFS) is required by DPDK applications. It is
178 enabled by using a negative value, the last four bits of which have a
181 - **-1**: force device-managed flow steering (DMFS).
182 - **-7**: configure optimized steering mode to improve performance with the
183 following limitation: VLAN filtering is not supported with this mode.
184 This is the recommended mode in case VLAN filter is not needed.
189 This driver relies on external libraries and kernel drivers for resources
190 allocations and initialization. The following dependencies are not part of
191 DPDK and must be installed separately:
195 User space verbs framework used by librte_pmd_mlx4. This library provides
196 a generic interface between the kernel and low-level user space drivers
199 It allows slow and privileged operations (context initialization, hardware
200 resources allocations) to be managed by the kernel and fast operations to
201 never leave user space.
205 Low-level user space driver library for Mellanox ConnectX-3 devices,
206 it is automatically loaded by libibverbs.
208 This library basically implements send/receive calls to the hardware
211 - **Kernel modules** (mlnx-ofed-kernel)
213 They provide the kernel-side verbs API and low level device drivers that
214 manage actual hardware initialization and resources sharing with user
217 Unlike most other PMDs, these modules must remain loaded and bound to
220 - mlx4_core: hardware driver managing Mellanox ConnectX-3 devices.
221 - mlx4_en: Ethernet device driver that provides kernel network interfaces.
222 - mlx4_ib: InifiniBand device driver.
223 - ib_uverbs: user space driver for verbs (entry point for libibverbs).
225 - **Firmware update**
227 Mellanox OFED releases include firmware updates for ConnectX-3 adapters.
229 Because each release provides new features, these updates must be applied to
230 match the kernel modules and libraries they come with.
234 Both libraries are BSD and GPL licensed. Linux kernel modules are GPL
237 Currently supported by DPDK:
239 - Mellanox OFED **4.1**.
240 - Firmware version **2.36.5000** and above.
242 Getting Mellanox OFED
243 ~~~~~~~~~~~~~~~~~~~~~
245 While these libraries and kernel modules are available on OpenFabrics
246 Alliance's `website <https://www.openfabrics.org/>`_ and provided by package
247 managers on most distributions, this PMD requires Ethernet extensions that
248 may not be supported at the moment (this is a work in progress).
251 <http://www.mellanox.com/page/products_dyn?product_family=26&mtag=linux_sw_drivers>`_
252 includes the necessary support and should be used in the meantime. For DPDK,
253 only libibverbs, libmlx4, mlnx-ofed-kernel packages and firmware updates are
254 required from that distribution.
258 Several versions of Mellanox OFED are available. Installing the version
259 this DPDK release was developed and tested against is strongly
260 recommended. Please check the `prerequisites`_.
265 * Mellanox(R) ConnectX(R)-3 Pro 40G MCX354A-FCC_Ax (2*40G)
270 1. Download latest Mellanox OFED. For more info check the `prerequisites`_.
272 2. Install the required libraries and kernel modules either by installing
273 only the required set, or by installing the entire Mellanox OFED:
277 .. code-block:: console
281 For SR-IOV hypervisors use:
283 .. code-block:: console
285 ./mlnxofedinstall --enable-sriov -hypervisor
287 For SR-IOV virtual machine use:
289 .. code-block:: console
291 ./mlnxofedinstall --guest
293 3. Verify the firmware is the correct one:
295 .. code-block:: console
299 4. Set all ports links to Ethernet, follow instructions on the screen:
301 .. code-block:: console
305 Or in the manual way:
307 .. code-block:: console
309 PCI=<NIC PCI address>
310 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port0"
311 echo eth > "/sys/bus/pci/devices/$PCI/mlx4_port1"
313 5. 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 6. Restart the driver:
327 .. code-block:: console
329 /etc/init.d/openibd restart
333 .. code-block:: console
335 service openibd restart
337 7. 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 environment variable MLX4_INLINE_RECV_SIZE=64 to get maximum
350 performance for 64B messages.
352 3. Use the CPU near local NUMA node to which the PCIe adapter is connected,
353 for better performance. For VMs, verify that the right CPU
354 and NUMA node are pinned according to the above. Run:
356 .. code-block:: console
360 to identify the NUMA node to which the PCIe adapter is connected.
362 4. If more than one adapter is used, and root complex capabilities allow
363 to put both adapters on the same NUMA node without PCI bandwidth degradation,
364 it is recommended to locate both adapters on the same NUMA node.
365 This in order to forward packets from one to the other without
366 NUMA performance penalty.
368 5. Disable pause frames:
370 .. code-block:: console
372 ethtool -A <netdev> rx off tx off
374 6. Verify IO non-posted prefetch is disabled by default. This can be checked
375 via the BIOS configuration. Please contact you server provider for more
376 information about the settings.
380 On some machines, depends on the machine integrator, it is beneficial
381 to set the PCI max read request parameter to 1K. This can be
382 done in the following way:
384 To query the read request size use:
386 .. code-block:: console
388 setpci -s <NIC PCI address> 68.w
390 If the output is different than 3XXX, set it by:
392 .. code-block:: console
394 setpci -s <NIC PCI address> 68.w=3XXX
396 The XXX can be different on different systems. Make sure to configure
397 according to the setpci output.
402 This section demonstrates how to launch **testpmd** with Mellanox ConnectX-3
403 devices managed by librte_pmd_mlx4.
405 #. Load the kernel modules:
407 .. code-block:: console
409 modprobe -a ib_uverbs mlx4_en mlx4_core mlx4_ib
411 Alternatively if MLNX_OFED is fully installed, the following script can
414 .. code-block:: console
416 /etc/init.d/openibd restart
420 User space I/O kernel modules (uio and igb_uio) are not used and do
421 not have to be loaded.
423 #. Make sure Ethernet interfaces are in working order and linked to kernel
424 verbs. Related sysfs entries should be present:
426 .. code-block:: console
428 ls -d /sys/class/net/*/device/infiniband_verbs/uverbs* | cut -d / -f 5
432 .. code-block:: console
439 #. Optionally, retrieve their PCI bus addresses for whitelisting:
441 .. code-block:: console
444 for intf in eth2 eth3 eth4 eth5;
446 (cd "/sys/class/net/${intf}/device/" && pwd -P);
449 sed -n 's,.*/\(.*\),-w \1,p'
453 .. code-block:: console
462 There are only two distinct PCI bus addresses because the Mellanox
463 ConnectX-3 adapters installed on this system are dual port.
465 #. Request huge pages:
467 .. code-block:: console
469 echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages/nr_hugepages
471 #. Start testpmd with basic parameters:
473 .. code-block:: console
475 testpmd -l 8-15 -n 4 -w 0000:83:00.0 -w 0000:84:00.0 -- --rxq=2 --txq=2 -i
479 .. code-block:: console
482 EAL: PCI device 0000:83:00.0 on NUMA socket 1
483 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
484 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_0" (VF: false)
485 PMD: librte_pmd_mlx4: 2 port(s) detected
486 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:b7:50
487 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:b7:51
488 EAL: PCI device 0000:84:00.0 on NUMA socket 1
489 EAL: probe driver: 15b3:1007 librte_pmd_mlx4
490 PMD: librte_pmd_mlx4: PCI information matches, using device "mlx4_1" (VF: false)
491 PMD: librte_pmd_mlx4: 2 port(s) detected
492 PMD: librte_pmd_mlx4: port 1 MAC address is 00:02:c9:b5:ba:b0
493 PMD: librte_pmd_mlx4: port 2 MAC address is 00:02:c9:b5:ba:b1
494 Interactive-mode selected
495 Configuring Port 0 (socket 0)
496 PMD: librte_pmd_mlx4: 0x867d60: TX queues number update: 0 -> 2
497 PMD: librte_pmd_mlx4: 0x867d60: RX queues number update: 0 -> 2
498 Port 0: 00:02:C9:B5:B7:50
499 Configuring Port 1 (socket 0)
500 PMD: librte_pmd_mlx4: 0x867da0: TX queues number update: 0 -> 2
501 PMD: librte_pmd_mlx4: 0x867da0: RX queues number update: 0 -> 2
502 Port 1: 00:02:C9:B5:B7:51
503 Configuring Port 2 (socket 0)
504 PMD: librte_pmd_mlx4: 0x867de0: TX queues number update: 0 -> 2
505 PMD: librte_pmd_mlx4: 0x867de0: RX queues number update: 0 -> 2
506 Port 2: 00:02:C9:B5:BA:B0
507 Configuring Port 3 (socket 0)
508 PMD: librte_pmd_mlx4: 0x867e20: TX queues number update: 0 -> 2
509 PMD: librte_pmd_mlx4: 0x867e20: RX queues number update: 0 -> 2
510 Port 3: 00:02:C9:B5:BA:B1
511 Checking link statuses...
512 Port 0 Link Up - speed 10000 Mbps - full-duplex
513 Port 1 Link Up - speed 40000 Mbps - full-duplex
514 Port 2 Link Up - speed 10000 Mbps - full-duplex
515 Port 3 Link Up - speed 40000 Mbps - full-duplex