2 Copyright (c) 2017, Cisco Systems, Inc.
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions
9 1. Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
12 2. Redistributions in binary form must reproduce the above copyright
13 notice, this list of conditions and the following disclaimer in
14 the documentation and/or other materials provided with the
17 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
20 FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
21 COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
22 INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
23 BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
24 LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
25 CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
27 ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28 POSSIBILITY OF SUCH DAMAGE.
33 ENIC PMD is the DPDK poll-mode driver for the Cisco System Inc. VIC Ethernet
34 NICs. These adapters are also referred to as vNICs below. If you are running
35 or would like to run DPDK software applications on Cisco UCS servers using
36 Cisco VIC adapters the following documentation is relevant.
38 How to obtain ENIC PMD integrated DPDK
39 --------------------------------------
41 ENIC PMD support is integrated into the DPDK suite. dpdk-<version>.tar.gz
42 should be downloaded from http://dpdk.org
45 Configuration information
46 -------------------------
48 - **DPDK Configuration Parameters**
50 The following configuration options are available for the ENIC PMD:
52 - **CONFIG_RTE_LIBRTE_ENIC_PMD** (default y): Enables or disables inclusion
53 of the ENIC PMD driver in the DPDK compilation.
55 - **vNIC Configuration Parameters**
57 - **Number of Queues**
59 The maximum number of receive queues (RQs), work queues (WQs) and
60 completion queues (CQs) are configurable on a per vNIC basis
61 through the Cisco UCS Manager (CIMC or UCSM).
63 These values should be configured as follows:
65 - The number of WQs should be greater or equal to the value of the
66 expected nb_tx_q parameter in the call to
67 rte_eth_dev_configure()
69 - The number of RQs configured in the vNIC should be greater or
70 equal to *twice* the value of the expected nb_rx_q parameter in
71 the call to rte_eth_dev_configure(). With the addition of Rx
72 scatter, a pair of RQs on the vnic is needed for each receive
73 queue used by DPDK, even if Rx scatter is not being used.
74 Having a vNIC with only 1 RQ is not a valid configuration, and
75 will fail with an error message.
77 - The number of CQs should set so that there is one CQ for each
78 WQ, and one CQ for each pair of RQs.
80 For example: If the application requires 3 Rx queues, and 3 Tx
81 queues, the vNIC should be configured to have at least 3 WQs, 6
82 RQs (3 pairs), and 6 CQs (3 for use by WQs + 3 for use by the 3
87 Likewise, the number of receive and transmit descriptors are configurable on
88 a per-vNIC basis via the UCS Manager and should be greater than or equal to
89 the nb_rx_desc and nb_tx_desc parameters expected to be used in the calls
90 to rte_eth_rx_queue_setup() and rte_eth_tx_queue_setup() respectively.
91 An application requesting more than the set size will be limited to that
94 Unless there is a lack of resources due to creating many vNICs, it
95 is recommended that the WQ and RQ sizes be set to the maximum. This
96 gives the application the greatest amount of flexibility in its
99 - *Note*: Since the introduction of Rx scatter, for performance
100 reasons, this PMD uses two RQs on the vNIC per receive queue in
101 DPDK. One RQ holds descriptors for the start of a packet, and the
102 second RQ holds the descriptors for the rest of the fragments of
103 a packet. This means that the nb_rx_desc parameter to
104 rte_eth_rx_queue_setup() can be a greater than 4096. The exact
105 amount will depend on the size of the mbufs being used for
106 receives, and the MTU size.
108 For example: If the mbuf size is 2048, and the MTU is 9000, then
109 receiving a full size packet will take 5 descriptors, 1 from the
110 start-of-packet queue, and 4 from the second queue. Assuming
111 that the RQ size was set to the maximum of 4096, then the
112 application can specify up to 1024 + 4096 as the nb_rx_desc
113 parameter to rte_eth_rx_queue_setup().
117 At least one interrupt per vNIC interface should be configured in the UCS
118 manager regardless of the number receive/transmit queues. The ENIC PMD
119 uses this interrupt to get information about link status and errors
122 In addition to the interrupt for link status and errors, when using Rx queue
123 interrupts, increase the number of configured interrupts so that there is at
124 least one interrupt for each Rx queue. For example, if the app uses 3 Rx
125 queues and wants to use per-queue interrupts, configure 4 (3 + 1) interrupts.
127 .. _enic-flow-director:
129 Flow director support
130 ---------------------
132 Advanced filtering support was added to 1300 series VIC firmware starting
133 with version 2.0.13 for C-series UCS servers and version 3.1.2 for UCSM
134 managed blade servers. In order to enable advanced filtering the 'Advanced
135 filter' radio button should be enabled via CIMC or UCSM followed by a reboot
138 With advanced filters, perfect matching of all fields of IPv4, IPv6 headers
139 as well as TCP, UDP and SCTP L4 headers is available through flow director.
140 Masking of these fields for partial match is also supported.
142 Without advanced filter support, the flow director is limited to IPv4
143 perfect filtering of the 5-tuple with no masking of fields supported.
145 SR-IOV mode utilization
146 -----------------------
148 UCS blade servers configured with dynamic vNIC connection policies in UCS
149 manager are capable of supporting assigned devices on virtual machines (VMs)
150 through a KVM hypervisor. Assigned devices, also known as 'passthrough'
151 devices, are SR-IOV virtual functions (VFs) on the host which are exposed
154 The Cisco Virtual Machine Fabric Extender (VM-FEX) gives the VM a dedicated
155 interface on the Fabric Interconnect (FI). Layer 2 switching is done at
156 the FI. This may eliminate the requirement for software switching on the
157 host to route intra-host VM traffic.
159 Please refer to `Creating a Dynamic vNIC Connection Policy
160 <http://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/sw/vm_fex/vmware/gui/config_guide/b_GUI_VMware_VM-FEX_UCSM_Configuration_Guide/b_GUI_VMware_VM-FEX_UCSM_Configuration_Guide_chapter_010.html#task_433E01651F69464783A68E66DA8A47A5>`_
161 for information on configuring SR-IOV adapter policies using UCS manager.
163 Once the policies are in place and the host OS is rebooted, VFs should be
164 visible on the host, E.g.:
166 .. code-block:: console
168 # lspci | grep Cisco | grep Ethernet
169 0d:00.0 Ethernet controller: Cisco Systems Inc VIC Ethernet NIC (rev a2)
170 0d:00.1 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
171 0d:00.2 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
172 0d:00.3 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
173 0d:00.4 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
174 0d:00.5 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
175 0d:00.6 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
176 0d:00.7 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
178 Enable Intel IOMMU on the host and install KVM and libvirt. A VM instance should
179 be created with an assigned device. When using libvirt, this configuration can
180 be done within the domain (i.e. VM) config file. For example this entry maps
181 host VF 0d:00:01 into the VM.
183 .. code-block:: console
185 <interface type='hostdev' managed='yes'>
186 <mac address='52:54:00:ac:ff:b6'/>
188 <address type='pci' domain='0x0000' bus='0x0d' slot='0x00' function='0x1'/>
191 Alternatively, the configuration can be done in a separate file using the
192 ``network`` keyword. These methods are described in the libvirt documentation for
193 `Network XML format <https://libvirt.org/formatnetwork.html>`_.
195 When the VM instance is started, the ENIC KVM driver will bind the host VF to
196 vfio, complete provisioning on the FI and bring up the link.
200 It is not possible to use a VF directly from the host because it is not
201 fully provisioned until the hypervisor brings up the VM that it is assigned
204 In the VM instance, the VF will now be visible. E.g., here the VF 00:04.0 is
205 seen on the VM instance and should be available for binding to a DPDK.
207 .. code-block:: console
210 00:04.0 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
212 Follow the normal DPDK install procedure, binding the VF to either ``igb_uio``
213 or ``vfio`` in non-IOMMU mode.
215 Please see :ref:`Limitations <enic_limitations>` for limitations in
218 .. _enic-genic-flow-api:
220 Generic Flow API support
221 ------------------------
223 Generic Flow API is supported. The baseline support is:
225 - **1200 series VICs**
227 5-tuple exact flow support for 1200 series adapters. This allows:
229 - Attributes: ingress
230 - Items: ipv4, ipv6, udp, tcp (must exactly match src/dst IP
231 addresses and ports and all must be specified)
232 - Actions: queue and void
235 - **1300 series VICS with advanced filters disabled**
237 With advanced filters disabled, an IPv4 or IPv6 item must be specified
240 - Attributes: ingress
241 - Items: eth, ipv4, ipv6, udp, tcp, vxlan, inner eth, ipv4, ipv6, udp, tcp
242 - Actions: queue and void
243 - Selectors: 'is', 'spec' and 'mask'. 'last' is not supported
244 - In total, up to 64 bytes of mask is allowed across all headers
246 - **1300 series VICS with advanced filters enabled**
248 - Attributes: ingress
249 - Items: eth, ipv4, ipv6, udp, tcp, vxlan, inner eth, ipv4, ipv6, udp, tcp
250 - Actions: queue, mark, flag and void
251 - Selectors: 'is', 'spec' and 'mask'. 'last' is not supported
252 - In total, up to 64 bytes of mask is allowed across all headers
254 More features may be added in future firmware and new versions of the VIC.
255 Please refer to the release notes.
257 .. _enic_limitations:
262 - **VLAN 0 Priority Tagging**
264 If a vNIC is configured in TRUNK mode by the UCS manager, the adapter will
265 priority tag egress packets according to 802.1Q if they were not already
266 VLAN tagged by software. If the adapter is connected to a properly configured
267 switch, there will be no unexpected behavior.
269 In test setups where an Ethernet port of a Cisco adapter in TRUNK mode is
270 connected point-to-point to another adapter port or connected though a router
271 instead of a switch, all ingress packets will be VLAN tagged. Programs such
272 as l3fwd which do not account for VLAN tags in packets will misbehave. The
273 solution is to enable VLAN stripping on ingress. The following code fragment is
274 an example of how to accomplish this:
276 .. code-block:: console
278 vlan_offload = rte_eth_dev_get_vlan_offload(port);
279 vlan_offload |= ETH_VLAN_STRIP_OFFLOAD;
280 rte_eth_dev_set_vlan_offload(port, vlan_offload);
282 - Limited flow director support on 1200 series and 1300 series Cisco VIC
283 adapters with old firmware. Please see :ref:`enic-flow-director`.
285 - Flow director features are not supported on generation 1 Cisco VIC adapters
290 - KVM hypervisor support only. VMware has not been tested.
291 - Requires VM-FEX, and so is only available on UCS managed servers connected
292 to Fabric Interconnects. It is not on standalone C-Series servers.
293 - VF devices are not usable directly from the host. They can only be used
294 as assigned devices on VM instances.
295 - Currently, unbind of the ENIC kernel mode driver 'enic.ko' on the VM
296 instance may hang. As a workaround, enic.ko should be blacklisted or removed
297 from the boot process.
298 - pci_generic cannot be used as the uio module in the VM. igb_uio or
299 vfio in non-IOMMU mode can be used.
300 - The number of RQs in UCSM dynamic vNIC configurations must be at least 2.
301 - The number of SR-IOV devices is limited to 256. Components on target system
302 might limit this number to fewer than 256.
306 - The number of filters that can be specified with the Generic Flow API is
307 dependent on how many header fields are being masked. Use 'flow create' in
308 a loop to determine how many filters your VIC will support (not more than
309 1000 for 1300 series VICs). Filters are checked for matching in the order they
310 were added. Since there currently is no grouping or priority support,
311 'catch-all' filters should be added last.
315 - ``rx_good_bytes`` (ibytes) always includes VLAN header (4B) and CRC bytes (4B).
316 - When the NIC drops a packet because the Rx queue has no free buffers,
317 ``rx_good_bytes`` still increments by 4B if the packet is not VLAN tagged or
318 VLAN stripping is disabled, or by 8B if the packet is VLAN tagged and stripping
321 How to build the suite
322 ----------------------
324 The build instructions for the DPDK suite should be followed. By default
325 the ENIC PMD library will be built into the DPDK library.
327 Refer to the document :ref:`compiling and testing a PMD for a NIC
328 <pmd_build_and_test>` for details.
330 For configuring and using UIO and VFIO frameworks, please refer to the
331 documentation that comes with DPDK suite.
333 Supported Cisco VIC adapters
334 ----------------------------
336 ENIC PMD supports all recent generations of Cisco VIC adapters including:
350 Supported Operating Systems
351 ---------------------------
353 Any Linux distribution fulfilling the conditions described in Dependencies
354 section of DPDK documentation.
359 - Unicast, multicast and broadcast transmission and reception
360 - Receive queue polling
361 - Port Hardware Statistics
362 - Hardware VLAN acceleration
363 - IP checksum offload
364 - Receive side VLAN stripping
365 - Multiple receive and transmit queues
366 - Flow Director ADD, UPDATE, DELETE, STATS operation support IPv4 and IPv6
368 - Setting RX VLAN (supported via UCSM/CIMC only)
369 - VLAN filtering (supported via UCSM/CIMC only)
370 - Execution of application by unprivileged system users
371 - IPV4, IPV6 and TCP RSS hashing
374 - SR-IOV on UCS managed servers connected to Fabric Interconnects
377 Known bugs and unsupported features in this release
378 ---------------------------------------------------
380 - Signature or flex byte based flow direction
381 - Drop feature of flow direction
382 - VLAN based flow direction
383 - Non-IPV4 flow direction
384 - Setting of extended VLAN
386 - MTU update only works if Scattered Rx mode is disabled
387 - Maximum receive packet length is ignored if Scattered Rx mode is used
392 - Prepare the system as recommended by DPDK suite. This includes environment
393 variables, hugepages configuration, tool-chains and configuration.
394 - Insert vfio-pci kernel module using the command 'modprobe vfio-pci' if the
395 user wants to use VFIO framework.
396 - Insert uio kernel module using the command 'modprobe uio' if the user wants
397 to use UIO framework.
398 - DPDK suite should be configured based on the user's decision to use VFIO or
400 - If the vNIC device(s) to be used is bound to the kernel mode Ethernet driver
401 use 'ip' to bring the interface down. The dpdk-devbind.py tool can
402 then be used to unbind the device's bus id from the ENIC kernel mode driver.
403 - Bind the intended vNIC to vfio-pci in case the user wants ENIC PMD to use
404 VFIO framework using dpdk-devbind.py.
405 - Bind the intended vNIC to igb_uio in case the user wants ENIC PMD to use
406 UIO framework using dpdk-devbind.py.
408 At this point the system should be ready to run DPDK applications. Once the
409 application runs to completion, the vNIC can be detached from vfio-pci or
410 igb_uio if necessary.
412 Root privilege is required to bind and unbind vNICs to/from VFIO/UIO.
413 VFIO framework helps an unprivileged user to run the applications.
414 For an unprivileged user to run the applications on DPDK and ENIC PMD,
415 it may be necessary to increase the maximum locked memory of the user.
416 The following command could be used to do this.
418 .. code-block:: console
420 sudo sh -c "ulimit -l <value in Kilo Bytes>"
422 The value depends on the memory configuration of the application, DPDK and
423 PMD. Typically, the limit has to be raised to higher than 2GB.
426 The compilation of any unused drivers can be disabled using the
427 configuration file in config/ directory (e.g., config/common_linuxapp).
428 This would help in bringing down the time taken for building the
429 libraries and the initialization time of the application.
434 - https://www.cisco.com/c/en/us/products/servers-unified-computing/index.html
435 - https://www.cisco.com/c/en/us/products/interfaces-modules/unified-computing-system-adapters/index.html
440 Any questions or bugs should be reported to DPDK community and to the ENIC PMD
443 - John Daley <johndale@cisco.com>
444 - Nelson Escobar <neescoba@cisco.com>