2 Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
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 Intel Corporation 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 Link Bonding Poll Mode Driver Library
32 =====================================
34 In addition to Poll Mode Drivers (PMDs) for physical and virtual hardware,
35 DPDK also includes a pure-software library that
36 allows physical PMD's to be bonded together to create a single logical PMD.
38 .. figure:: img/bond-overview.*
43 The Link Bonding PMD library(librte_pmd_bond) supports bonding of groups of
44 ``rte_eth_dev`` ports of the same speed and duplex to provide
45 similar the capabilities to that found in Linux bonding driver to allow the
46 aggregation of multiple (slave) NICs into a single logical interface between a
47 server and a switch. The new bonded PMD will then process these interfaces
48 based on the mode of operation specified to provide support for features such
49 as redundant links, fault tolerance and/or load balancing.
51 The librte_pmd_bond library exports a C API which provides an API for the
52 creation of bonded devices as well as the configuration and management of the
53 bonded device and its slave devices.
57 The Link Bonding PMD Library is enabled by default in the build
58 configuration files, the library can be disabled by setting
59 ``CONFIG_RTE_LIBRTE_PMD_BOND=n`` and recompiling the DPDK.
61 Link Bonding Modes Overview
62 ---------------------------
64 Currently the Link Bonding PMD library supports 4 modes of operation:
66 * **Round-Robin (Mode 0):**
68 .. figure:: img/bond-mode-0.*
73 This mode provides load balancing and fault tolerance by transmission of
74 packets in sequential order from the first available slave device through
75 the last. Packets are bulk dequeued from devices then serviced in a
76 round-robin manner. This mode does not guarantee in order reception of
77 packets and down stream should be able to handle out of order packets.
79 * **Active Backup (Mode 1):**
81 .. figure:: img/bond-mode-1.*
83 Active Backup (Mode 1)
86 In this mode only one slave in the bond is active at any time, a different
87 slave becomes active if, and only if, the primary active slave fails,
88 thereby providing fault tolerance to slave failure. The single logical
89 bonded interface's MAC address is externally visible on only one NIC (port)
90 to avoid confusing the network switch.
92 * **Balance XOR (Mode 2):**
94 .. figure:: img/bond-mode-2.*
99 This mode provides transmit load balancing (based on the selected
100 transmission policy) and fault tolerance. The default policy (layer2) uses
101 a simple calculation based on the packet flow source and destination MAC
102 addresses as well as the number of active slaves available to the bonded
103 device to classify the packet to a specific slave to transmit on. Alternate
104 transmission policies supported are layer 2+3, this takes the IP source and
105 destination addresses into the calculation of the transmit slave port and
106 the final supported policy is layer 3+4, this uses IP source and
107 destination addresses as well as the TCP/UDP source and destination port.
110 The coloring differences of the packets are used to identify different flow
111 classification calculated by the selected transmit policy
114 * **Broadcast (Mode 3):**
116 .. figure:: img/bond-mode-3.*
121 This mode provides fault tolerance by transmission of packets on all slave
124 * **Link Aggregation 802.3AD (Mode 4):**
126 .. figure:: img/bond-mode-4.*
128 Link Aggregation 802.3AD (Mode 4)
131 This mode provides dynamic link aggregation according to the 802.3ad
132 specification. It negotiates and monitors aggregation groups that share the
133 same speed and duplex settings using the selected balance transmit policy
134 for balancing outgoing traffic.
136 DPDK implementation of this mode provide some additional requirements of
139 #. It needs to call ``rte_eth_tx_burst`` and ``rte_eth_rx_burst`` with
140 intervals period of less than 100ms.
142 #. Calls to ``rte_eth_tx_burst`` must have a buffer size of at least 2xN,
143 where N is the number of slaves. This is a space required for LACP
144 frames. Additionally LACP packets are included in the statistics, but
145 they are not returned to the application.
147 * **Transmit Load Balancing (Mode 5):**
149 .. figure:: img/bond-mode-5.*
151 Transmit Load Balancing (Mode 5)
154 This mode provides an adaptive transmit load balancing. It dynamically
155 changes the transmitting slave, according to the computed load. Statistics
156 are collected in 100ms intervals and scheduled every 10ms.
159 Implementation Details
160 ----------------------
162 The librte_pmd_bond bonded device are compatible with the Ethernet device API
163 exported by the Ethernet PMDs described in the *DPDK API Reference*.
165 The Link Bonding Library supports the creation of bonded devices at application
166 startup time during EAL initialization using the ``--vdev`` option as well as
167 programmatically via the C API ``rte_eth_bond_create`` function.
169 Bonded devices support the dynamical addition and removal of slave devices using
170 the ``rte_eth_bond_slave_add`` / ``rte_eth_bond_slave_remove`` APIs.
172 After a slave device is added to a bonded device slave is stopped using
173 ``rte_eth_dev_stop`` and then reconfigured using ``rte_eth_dev_configure``
174 the RX and TX queues are also reconfigured using ``rte_eth_tx_queue_setup`` /
175 ``rte_eth_rx_queue_setup`` with the parameters use to configure the bonding
176 device. If RSS is enabled for bonding device, this mode is also enabled on new
177 slave and configured as well.
179 Setting up multi-queue mode for bonding device to RSS, makes it fully
180 RSS-capable, so all slaves are synchronized with its configuration. This mode is
181 intended to provide RSS configuration on slaves transparent for client
182 application implementation.
184 Bonding device stores its own version of RSS settings i.e. RETA, RSS hash
185 function and RSS key, used to set up its slaves. That let to define the meaning
186 of RSS configuration of bonding device as desired configuration of whole bonding
187 (as one unit), without pointing any of slave inside. It is required to ensure
188 consistency and made it more error-proof.
190 RSS hash function set for bonding device, is a maximal set of RSS hash functions
191 supported by all bonded slaves. RETA size is a GCD of all its RETA's sizes, so
192 it can be easily used as a pattern providing expected behavior, even if slave
193 RETAs' sizes are different. If RSS Key is not set for bonded device, it's not
194 changed on the slaves and default key for device is used.
196 All settings are managed through the bonding port API and always are propagated
197 in one direction (from bonding to slaves).
199 Link Status Change Interrupts / Polling
200 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
202 Link bonding devices support the registration of a link status change callback,
203 using the ``rte_eth_dev_callback_register`` API, this will be called when the
204 status of the bonding device changes. For example in the case of a bonding
205 device which has 3 slaves, the link status will change to up when one slave
206 becomes active or change to down when all slaves become inactive. There is no
207 callback notification when a single slave changes state and the previous
208 conditions are not met. If a user wishes to monitor individual slaves then they
209 must register callbacks with that slave directly.
211 The link bonding library also supports devices which do not implement link
212 status change interrupts, this is achieved by polling the devices link status at
213 a defined period which is set using the ``rte_eth_bond_link_monitoring_set``
214 API, the default polling interval is 10ms. When a device is added as a slave to
215 a bonding device it is determined using the ``RTE_PCI_DRV_INTR_LSC`` flag
216 whether the device supports interrupts or whether the link status should be
217 monitored by polling it.
219 Requirements / Limitations
220 ~~~~~~~~~~~~~~~~~~~~~~~~~~
222 The current implementation only supports devices that support the same speed
223 and duplex to be added as a slaves to the same bonded device. The bonded device
224 inherits these attributes from the first active slave added to the bonded
225 device and then all further slaves added to the bonded device must support
228 A bonding device must have a minimum of one slave before the bonding device
229 itself can be started.
231 To use a bonding device dynamic RSS configuration feature effectively, it is
232 also required, that all slaves should be RSS-capable and support, at least one
233 common hash function available for each of them. Changing RSS key is only
234 possible, when all slave devices support the same key size.
236 To prevent inconsistency on how slaves process packets, once a device is added
237 to a bonding device, RSS configuration should be managed through the bonding
238 device API, and not directly on the slave.
240 Like all other PMD, all functions exported by a PMD are lock-free functions
241 that are assumed not to be invoked in parallel on different logical cores to
242 work on the same target object.
244 It should also be noted that the PMD receive function should not be invoked
245 directly on a slave devices after they have been to a bonded device since
246 packets read directly from the slave device will no longer be available to the
247 bonded device to read.
252 Link bonding devices are created using the ``rte_eth_bond_create`` API
253 which requires a unique device name, the bonding mode,
254 and the socket Id to allocate the bonding device's resources on.
255 The other configurable parameters for a bonded device are its slave devices,
256 its primary slave, a user defined MAC address and transmission policy to use if
257 the device is in balance XOR mode.
262 Bonding devices support up to a maximum of ``RTE_MAX_ETHPORTS`` slave devices
263 of the same speed and duplex. Ethernet devices can be added as a slave to a
264 maximum of one bonded device. Slave devices are reconfigured with the
265 configuration of the bonded device on being added to a bonded device.
267 The bonded also guarantees to return the MAC address of the slave device to its
268 original value of removal of a slave from it.
273 The primary slave is used to define the default port to use when a bonded
274 device is in active backup mode. A different port will only be used if, and
275 only if, the current primary port goes down. If the user does not specify a
276 primary port it will default to being the first port added to the bonded device.
281 The bonded device can be configured with a user specified MAC address, this
282 address will be inherited by the some/all slave devices depending on the
283 operating mode. If the device is in active backup mode then only the primary
284 device will have the user specified MAC, all other slaves will retain their
285 original MAC address. In mode 0, 2, 3, 4 all slaves devices are configure with
286 the bonded devices MAC address.
288 If a user defined MAC address is not defined then the bonded device will
289 default to using the primary slaves MAC address.
291 Balance XOR Transmit Policies
292 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
294 There are 3 supported transmission policies for bonded device running in
295 Balance XOR mode. Layer 2, Layer 2+3, Layer 3+4.
297 * **Layer 2:** Ethernet MAC address based balancing is the default
298 transmission policy for Balance XOR bonding mode. It uses a simple XOR
299 calculation on the source MAC address and destination MAC address of the
300 packet and then calculate the modulus of this value to calculate the slave
301 device to transmit the packet on.
303 * **Layer 2 + 3:** Ethernet MAC address & IP Address based balancing uses a
304 combination of source/destination MAC addresses and the source/destination
305 IP addresses of the data packet to decide which slave port the packet will
308 * **Layer 3 + 4:** IP Address & UDP Port based balancing uses a combination
309 of source/destination IP Address and the source/destination UDP ports of
310 the packet of the data packet to decide which slave port the packet will be
313 All these policies support 802.1Q VLAN Ethernet packets, as well as IPv4, IPv6
314 and UDP protocols for load balancing.
316 Using Link Bonding Devices
317 --------------------------
319 The librte_pmd_bond library supports two modes of device creation, the libraries
320 export full C API or using the EAL command line to statically configure link
321 bonding devices at application startup. Using the EAL option it is possible to
322 use link bonding functionality transparently without specific knowledge of the
323 libraries API, this can be used, for example, to add bonding functionality,
324 such as active backup, to an existing application which has no knowledge of
325 the link bonding C API.
327 Using the Poll Mode Driver from an Application
328 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
330 Using the librte_pmd_bond libraries API it is possible to dynamically create
331 and manage link bonding device from within any application. Link bonding
332 devices are created using the ``rte_eth_bond_create`` API which requires a
333 unique device name, the link bonding mode to initial the device in and finally
334 the socket Id which to allocate the devices resources onto. After successful
335 creation of a bonding device it must be configured using the generic Ethernet
336 device configure API ``rte_eth_dev_configure`` and then the RX and TX queues
337 which will be used must be setup using ``rte_eth_tx_queue_setup`` /
338 ``rte_eth_rx_queue_setup``.
340 Slave devices can be dynamically added and removed from a link bonding device
341 using the ``rte_eth_bond_slave_add`` / ``rte_eth_bond_slave_remove``
342 APIs but at least one slave device must be added to the link bonding device
343 before it can be started using ``rte_eth_dev_start``.
345 The link status of a bonded device is dictated by that of its slaves, if all
346 slave device link status are down or if all slaves are removed from the link
347 bonding device then the link status of the bonding device will go down.
349 It is also possible to configure / query the configuration of the control
350 parameters of a bonded device using the provided APIs
351 ``rte_eth_bond_mode_set/ get``, ``rte_eth_bond_primary_set/get``,
352 ``rte_eth_bond_mac_set/reset`` and ``rte_eth_bond_xmit_policy_set/get``.
354 Using Link Bonding Devices from the EAL Command Line
355 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
357 Link bonding devices can be created at application startup time using the
358 ``--vdev`` EAL command line option. The device name must start with the
359 eth_bond prefix followed by numbers or letters. The name must be unique for
360 each device. Each device can have multiple options arranged in a comma
361 separated list. Multiple devices definitions can be arranged by calling the
362 ``--vdev`` option multiple times.
364 Device names and bonding options must be separated by commas as shown below:
366 .. code-block:: console
368 $RTE_TARGET/app/testpmd -c f -n 4 --vdev 'eth_bond0,bond_opt0=..,bond opt1=..'--vdev 'eth_bond1,bond _opt0=..,bond_opt1=..'
370 Link Bonding EAL Options
371 ^^^^^^^^^^^^^^^^^^^^^^^^
373 There are multiple ways of definitions that can be assessed and combined as
374 long as the following two rules are respected:
376 * A unique device name, in the format of eth_bondX is provided,
377 where X can be any combination of numbers and/or letters,
378 and the name is no greater than 32 characters long.
380 * A least one slave device is provided with for each bonded device definition.
382 * The operation mode of the bonded device being created is provided.
384 The different options are:
386 * mode: Integer value defining the bonding mode of the device.
387 Currently supports modes 0,1,2,3,4,5 (round-robin, active backup, balance,
388 broadcast, link aggregation, transmit load balancing).
390 .. code-block:: console
394 * slave: Defines the PMD device which will be added as slave to the bonded
395 device. This option can be selected multiple times, for each device to be
396 added as a slave. Physical devices should be specified using their PCI
397 address, in the format domain:bus:devid.function
399 .. code-block:: console
401 slave=0000:0a:00.0,slave=0000:0a:00.1
403 * primary: Optional parameter which defines the primary slave port,
404 is used in active backup mode to select the primary slave for data TX/RX if
405 it is available. The primary port also is used to select the MAC address to
406 use when it is not defined by the user. This defaults to the first slave
407 added to the device if it is specified. The primary device must be a slave
408 of the bonded device.
410 .. code-block:: console
414 * socket_id: Optional parameter used to select which socket on a NUMA device
415 the bonded devices resources will be allocated on.
417 .. code-block:: console
421 * mac: Optional parameter to select a MAC address for link bonding device,
422 this overrides the value of the primary slave device.
424 .. code-block:: console
426 mac=00:1e:67:1d:fd:1d
428 * xmit_policy: Optional parameter which defines the transmission policy when
429 the bonded device is in balance mode. If not user specified this defaults
430 to l2 (layer 2) forwarding, the other transmission policies available are
431 l23 (layer 2+3) and l34 (layer 3+4)
433 .. code-block:: console
437 * lsc_poll_period_ms: Optional parameter which defines the polling interval
438 in milli-seconds at which devices which don't support lsc interrupts are
439 checked for a change in the devices link status
441 .. code-block:: console
443 lsc_poll_period_ms=100
445 * up_delay: Optional parameter which adds a delay in milli-seconds to the
446 propagation of a devices link status changing to up, by default this
449 .. code-block:: console
453 * down_delay: Optional parameter which adds a delay in milli-seconds to the
454 propagation of a devices link status changing to down, by default this
457 .. code-block:: console
464 Create a bonded device in round robin mode with two slaves specified by their PCI address:
466 .. code-block:: console
468 $RTE_TARGET/app/testpmd -c '0xf' -n 4 --vdev 'eth_bond0,mode=0, slave=0000:00a:00.01,slave=0000:004:00.00' -- --port-topology=chained
470 Create a bonded device in round robin mode with two slaves specified by their PCI address and an overriding MAC address:
472 .. code-block:: console
474 $RTE_TARGET/app/testpmd -c '0xf' -n 4 --vdev 'eth_bond0,mode=0, slave=0000:00a:00.01,slave=0000:004:00.00,mac=00:1e:67:1d:fd:1d' -- --port-topology=chained
476 Create a bonded device in active backup mode with two slaves specified, and a primary slave specified by their PCI addresses:
478 .. code-block:: console
480 $RTE_TARGET/app/testpmd -c '0xf' -n 4 --vdev 'eth_bond0,mode=1, slave=0000:00a:00.01,slave=0000:004:00.00,primary=0000:00a:00.01' -- --port-topology=chained
482 Create a bonded device in balance mode with two slaves specified by their PCI addresses, and a transmission policy of layer 3 + 4 forwarding:
484 .. code-block:: console
486 $RTE_TARGET/app/testpmd -c '0xf' -n 4 --vdev 'eth_bond0,mode=2, slave=0000:00a:00.01,slave=0000:004:00.00,xmit_policy=l34' -- --port-topology=chained