xref: /dpdk/doc/guides/nics/enic.rst (revision 68a03efeed657e6e05f281479b33b51102797e15)
1..  SPDX-License-Identifier: BSD-3-Clause
2    Copyright (c) 2017, Cisco Systems, Inc.
3    All rights reserved.
4
5ENIC Poll Mode Driver
6=====================
7
8ENIC PMD is the DPDK poll-mode driver for the Cisco System Inc. VIC Ethernet
9NICs. These adapters are also referred to as vNICs below. If you are running
10or would like to run DPDK software applications on Cisco UCS servers using
11Cisco VIC adapters the following documentation is relevant.
12
13How to obtain ENIC PMD integrated DPDK
14--------------------------------------
15
16ENIC PMD support is integrated into the DPDK suite. dpdk-<version>.tar.gz
17should be downloaded from https://core.dpdk.org/download/
18
19
20Configuration information
21-------------------------
22
23- **vNIC Configuration Parameters**
24
25  - **Number of Queues**
26
27    The maximum number of receive queues (RQs), work queues (WQs) and
28    completion queues (CQs) are configurable on a per vNIC basis
29    through the Cisco UCS Manager (CIMC or UCSM).
30
31    These values should be configured as follows:
32
33    - The number of WQs should be greater or equal to the value of the
34      expected nb_tx_q parameter in the call to
35      rte_eth_dev_configure()
36
37    - The number of RQs configured in the vNIC should be greater or
38      equal to *twice* the value of the expected nb_rx_q parameter in
39      the call to rte_eth_dev_configure().  With the addition of Rx
40      scatter, a pair of RQs on the vnic is needed for each receive
41      queue used by DPDK, even if Rx scatter is not being used.
42      Having a vNIC with only 1 RQ is not a valid configuration, and
43      will fail with an error message.
44
45    - The number of CQs should set so that there is one CQ for each
46      WQ, and one CQ for each pair of RQs.
47
48    For example: If the application requires 3 Rx queues, and 3 Tx
49    queues, the vNIC should be configured to have at least 3 WQs, 6
50    RQs (3 pairs), and 6 CQs (3 for use by WQs + 3 for use by the 3
51    pairs of RQs).
52
53  - **Size of Queues**
54
55    Likewise, the number of receive and transmit descriptors are configurable on
56    a per-vNIC basis via the UCS Manager and should be greater than or equal to
57    the nb_rx_desc and   nb_tx_desc parameters expected to be used in the calls
58    to rte_eth_rx_queue_setup() and rte_eth_tx_queue_setup() respectively.
59    An application requesting more than the set size will be limited to that
60    size.
61
62    Unless there is a lack of resources due to creating many vNICs, it
63    is recommended that the WQ and RQ sizes be set to the maximum.  This
64    gives the application the greatest amount of flexibility in its
65    queue configuration.
66
67    - *Note*: Since the introduction of Rx scatter, for performance
68      reasons, this PMD uses two RQs on the vNIC per receive queue in
69      DPDK.  One RQ holds descriptors for the start of a packet, and the
70      second RQ holds the descriptors for the rest of the fragments of
71      a packet.  This means that the nb_rx_desc parameter to
72      rte_eth_rx_queue_setup() can be a greater than 4096.  The exact
73      amount will depend on the size of the mbufs being used for
74      receives, and the MTU size.
75
76      For example: If the mbuf size is 2048, and the MTU is 9000, then
77      receiving a full size packet will take 5 descriptors, 1 from the
78      start-of-packet queue, and 4 from the second queue.  Assuming
79      that the RQ size was set to the maximum of 4096, then the
80      application can specify up to 1024 + 4096 as the nb_rx_desc
81      parameter to rte_eth_rx_queue_setup().
82
83  - **Interrupts**
84
85    At least one interrupt per vNIC interface should be configured in the UCS
86    manager regardless of the number receive/transmit queues. The ENIC PMD
87    uses this interrupt to get information about link status and errors
88    in the fast path.
89
90    In addition to the interrupt for link status and errors, when using Rx queue
91    interrupts, increase the number of configured interrupts so that there is at
92    least one interrupt for each Rx queue. For example, if the app uses 3 Rx
93    queues and wants to use per-queue interrupts, configure 4 (3 + 1) interrupts.
94
95  - **Receive Side Scaling**
96
97    In order to fully utilize RSS in DPDK, enable all RSS related settings in
98    CIMC or UCSM. These include the following items listed under
99    Receive Side Scaling:
100    TCP, IPv4, TCP-IPv4, IPv6, TCP-IPv6, IPv6 Extension, TCP-IPv6 Extension.
101
102
103SR-IOV mode utilization
104-----------------------
105
106UCS blade servers configured with dynamic vNIC connection policies in UCSM
107are capable of supporting SR-IOV. SR-IOV virtual functions (VFs) are
108specialized vNICs, distinct from regular Ethernet vNICs. These VFs can be
109directly assigned to virtual machines (VMs) as 'passthrough' devices.
110
111In UCS, SR-IOV VFs require the use of the Cisco Virtual Machine Fabric Extender
112(VM-FEX), which gives the VM a dedicated
113interface on the Fabric Interconnect (FI). Layer 2 switching is done at
114the FI. This may eliminate the requirement for software switching on the
115host to route intra-host VM traffic.
116
117Please refer to `Creating a Dynamic vNIC Connection Policy
118<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>`_
119for information on configuring SR-IOV adapter policies and port profiles
120using UCSM.
121
122Once the policies are in place and the host OS is rebooted, VFs should be
123visible on the host, E.g.:
124
125.. code-block:: console
126
127     # lspci | grep Cisco | grep Ethernet
128     0d:00.0 Ethernet controller: Cisco Systems Inc VIC Ethernet NIC (rev a2)
129     0d:00.1 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
130     0d:00.2 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
131     0d:00.3 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
132     0d:00.4 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
133     0d:00.5 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
134     0d:00.6 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
135     0d:00.7 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
136
137Enable Intel IOMMU on the host and install KVM and libvirt, and reboot again as
138required. Then, using libvirt, create a VM instance with an assigned device.
139Below is an example ``interface`` block (part of the domain configuration XML)
140that adds the host VF 0d:00:01 to the VM. ``profileid='pp-vlan-25'`` indicates
141the port profile that has been configured in UCSM.
142
143.. code-block:: console
144
145    <interface type='hostdev' managed='yes'>
146      <mac address='52:54:00:ac:ff:b6'/>
147      <driver name='vfio'/>
148      <source>
149        <address type='pci' domain='0x0000' bus='0x0d' slot='0x00' function='0x1'/>
150      </source>
151      <virtualport type='802.1Qbh'>
152        <parameters profileid='pp-vlan-25'/>
153      </virtualport>
154    </interface>
155
156
157Alternatively, the configuration can be done in a separate file using the
158``network`` keyword. These methods are described in the libvirt documentation for
159`Network XML format <https://libvirt.org/formatnetwork.html>`_.
160
161When the VM instance is started, libvirt will bind the host VF to
162vfio, complete provisioning on the FI and bring up the link.
163
164.. note::
165
166    It is not possible to use a VF directly from the host because it is not
167    fully provisioned until libvirt brings up the VM that it is assigned
168    to.
169
170In the VM instance, the VF will now be visible. E.g., here the VF 00:04.0 is
171seen on the VM instance and should be available for binding to a DPDK.
172
173.. code-block:: console
174
175     # lspci | grep Ether
176     00:04.0 Ethernet controller: Cisco Systems Inc VIC SR-IOV VF (rev a2)
177
178Follow the normal DPDK install procedure, binding the VF to either ``igb_uio``
179or ``vfio`` in non-IOMMU mode.
180
181In the VM, the kernel enic driver may be automatically bound to the VF during
182boot. Unbinding it currently hangs due to a known issue with the driver. To
183work around the issue, block the enic module as follows.
184Please see :ref:`Limitations <enic_limitations>` for limitations in
185the use of SR-IOV.
186
187.. code-block:: console
188
189     # cat /etc/modprobe.d/enic.conf
190     blacklist enic
191
192     # dracut --force
193
194.. note::
195
196    Passthrough does not require SR-IOV. If VM-FEX is not desired, the user
197    may create as many regular vNICs as necessary and assign them to VMs as
198    passthrough devices. Since these vNICs are not SR-IOV VFs, using them as
199    passthrough devices do not require libvirt, port profiles, and VM-FEX.
200
201
202.. _enic-generic-flow-api:
203
204Generic Flow API support
205------------------------
206
207Generic Flow API (also called "rte_flow" API) is supported. More advanced
208capabilities are available when "Advanced Filtering" is enabled on the adapter.
209Advanced filtering was added to 1300 series VIC firmware starting with version
2102.0.13 for C-series UCS servers and version 3.1.2 for UCSM managed blade
211servers. Advanced filtering is available on 1400 series adapters and beyond.
212To enable advanced filtering, the 'Advanced filter' radio button should be
213selected via CIMC or UCSM followed by a reboot of the server.
214
215- **1200 series VICs**
216
217  5-tuple exact flow support for 1200 series adapters. This allows:
218
219  - Attributes: ingress
220  - Items: ipv4, ipv6, udp, tcp (must exactly match src/dst IP
221    addresses and ports and all must be specified)
222  - Actions: queue and void
223  - Selectors: 'is'
224
225- **1300 and later series VICS with advanced filters disabled**
226
227  With advanced filters disabled, an IPv4 or IPv6 item must be specified
228  in the pattern.
229
230  - Attributes: ingress
231  - Items: eth, vlan, ipv4, ipv6, udp, tcp, vxlan, inner eth, vlan, ipv4, ipv6, udp, tcp
232  - Actions: queue and void
233  - Selectors: 'is', 'spec' and 'mask'. 'last' is not supported
234  - In total, up to 64 bytes of mask is allowed across all headers
235
236- **1300 and later series VICS with advanced filters enabled**
237
238  - Attributes: ingress
239  - Items: eth, vlan, ipv4, ipv6, udp, tcp, vxlan, raw, inner eth, vlan, ipv4, ipv6, udp, tcp
240  - Actions: queue, mark, drop, flag, rss, passthru, and void
241  - Selectors: 'is', 'spec' and 'mask'. 'last' is not supported
242  - In total, up to 64 bytes of mask is allowed across all headers
243
244- **1400 and later series VICs with Flow Manager API enabled**
245
246  - Attributes: ingress, egress
247  - Items: eth, vlan, ipv4, ipv6, sctp, udp, tcp, vxlan, raw, inner eth, vlan, ipv4, ipv6, sctp, udp, tcp
248  - Ingress Actions: count, drop, flag, jump, mark, port_id, passthru, queue, rss, vxlan_decap, vxlan_encap, and void
249  - Egress Actions: count, drop, jump, passthru, vxlan_encap, and void
250  - Selectors: 'is', 'spec' and 'mask'. 'last' is not supported
251  - In total, up to 64 bytes of mask is allowed across all headers
252
253The VIC performs packet matching after applying VLAN strip. If VLAN
254stripping is enabled, EtherType in the ETH item corresponds to the
255stripped VLAN header's EtherType. Stripping does not affect the VLAN
256item. TCI and EtherType in the VLAN item are matched against those in
257the (stripped) VLAN header whether stripping is enabled or disabled.
258
259More features may be added in future firmware and new versions of the VIC.
260Please refer to the release notes.
261
262.. _overlay_offload:
263
264Overlay Offload
265---------------
266
267Recent hardware models support overlay offload. When enabled, the NIC performs
268the following operations for VXLAN, NVGRE, and GENEVE packets. In all cases,
269inner and outer packets can be IPv4 or IPv6.
270
271- TSO for VXLAN and GENEVE packets.
272
273  Hardware supports NVGRE TSO, but DPDK currently has no NVGRE offload flags.
274
275- Tx checksum offloads.
276
277  The NIC fills in IPv4/UDP/TCP checksums for both inner and outer packets.
278
279- Rx checksum offloads.
280
281  The NIC validates IPv4/UDP/TCP checksums of both inner and outer packets.
282  Good checksum flags (e.g. ``PKT_RX_L4_CKSUM_GOOD``) indicate that the inner
283  packet has the correct checksum, and if applicable, the outer packet also
284  has the correct checksum. Bad checksum flags (e.g. ``PKT_RX_L4_CKSUM_BAD``)
285  indicate that the inner and/or outer packets have invalid checksum values.
286
287- Inner Rx packet type classification
288
289  PMD sets inner L3/L4 packet types (e.g. ``RTE_PTYPE_INNER_L4_TCP``), and
290  ``RTE_PTYPE_TUNNEL_GRENAT`` to indicate that the packet is tunneled.
291  PMD does not set L3/L4 packet types for outer packets.
292
293- Inner RSS
294
295  RSS hash calculation, therefore queue selection, is done on inner packets.
296
297In order to enable overlay offload, the 'Enable VXLAN' box should be checked
298via CIMC or UCSM followed by a reboot of the server. When PMD successfully
299enables overlay offload, it prints the following message on the console.
300
301.. code-block:: console
302
303    Overlay offload is enabled
304
305By default, PMD enables overlay offload if hardware supports it. To disable
306it, set ``devargs`` parameter ``disable-overlay=1``. For example::
307
308    -a 12:00.0,disable-overlay=1
309
310By default, the NIC uses 4789 as the VXLAN port. The user may change
311it through ``rte_eth_dev_udp_tunnel_port_{add,delete}``. However, as
312the current NIC has a single VXLAN port number, the user cannot
313configure multiple port numbers.
314
315Geneve headers with non-zero options are not supported by default. To
316use Geneve with options, update the VIC firmware to the latest version
317and then set ``devargs`` parameter ``geneve-opt=1``. When Geneve with
318options is enabled, flow API cannot be used as the features are
319currently mutually exclusive. When this feature is successfully
320enabled, PMD prints the following message.
321
322.. code-block:: console
323
324    Geneve with options is enabled
325
326
327Ingress VLAN Rewrite
328--------------------
329
330VIC adapters can tag, untag, or modify the VLAN headers of ingress
331packets. The ingress VLAN rewrite mode controls this behavior. By
332default, it is set to pass-through, where the NIC does not modify the
333VLAN header in any way so that the application can see the original
334header. This mode is sufficient for many applications, but may not be
335suitable for others. Such applications may change the mode by setting
336``devargs`` parameter ``ig-vlan-rewrite`` to one of the following.
337
338- ``pass``: Pass-through mode. The NIC does not modify the VLAN
339  header. This is the default mode.
340
341- ``priority``: Priority-tag default VLAN mode. If the ingress packet
342  is tagged with the default VLAN, the NIC replaces its VLAN header
343  with the priority tag (VLAN ID 0).
344
345- ``trunk``: Default trunk mode. The NIC tags untagged ingress packets
346  with the default VLAN. Tagged ingress packets are not modified. To
347  the application, every packet appears as tagged.
348
349- ``untag``: Untag default VLAN mode. If the ingress packet is tagged
350  with the default VLAN, the NIC removes or untags its VLAN header so
351  that the application sees an untagged packet. As a result, the
352  default VLAN becomes `untagged`. This mode can be useful for
353  applications such as OVS-DPDK performance benchmarks that utilize
354  only the default VLAN and want to see only untagged packets.
355
356
357Vectorized Rx Handler
358---------------------
359
360ENIC PMD includes a version of the receive handler that is vectorized using
361AVX2 SIMD instructions. It is meant for bulk, throughput oriented workloads
362where reducing cycles/packet in PMD is a priority. In order to use the
363vectorized handler, take the following steps.
364
365- Use a recent version of gcc, icc, or clang and build 64-bit DPDK. If
366  the compiler is known to support AVX2, DPDK build system
367  automatically compiles the vectorized handler. Otherwise, the
368  handler is not available.
369
370- Set ``devargs`` parameter ``enable-avx2-rx=1`` to explicitly request that
371  PMD consider the vectorized handler when selecting the receive handler.
372  For example::
373
374    -a 12:00.0,enable-avx2-rx=1
375
376  As the current implementation is intended for field trials, by default, the
377  vectorized handler is not considered (``enable-avx2-rx=0``).
378
379- Run on a UCS M4 or later server with CPUs that support AVX2.
380
381PMD selects the vectorized handler when the handler is compiled into
382the driver, the user requests its use via ``enable-avx2-rx=1``, CPU
383supports AVX2, and scatter Rx is not used. To verify that the
384vectorized handler is selected, enable debug logging
385(``--log-level=pmd,debug``) and check the following message.
386
387.. code-block:: console
388
389    enic_use_vector_rx_handler use the non-scatter avx2 Rx handler
390
39164B Completion Queue Entry
392--------------------------
393
394Recent VIC adapters support 64B completion queue entries, as well as
39516B entries that are available on all adapter models. ENIC PMD enables
396and uses 64B entries by default, if available. 64B entries generally
397lower CPU cycles per Rx packet, as they avoid partial DMA writes and
398reduce cache contention between DMA and polling CPU. The effect is
399most pronounced when multiple Rx queues are used on Intel platforms
400with Data Direct I/O Technology (DDIO).
401
402If 64B entries are not available, PMD uses 16B entries. The user may
403explicitly disable 64B entries and use 16B entries by setting
404``devarg`` parameter ``cq64=0``. For example::
405
406    -a 12:00.0,cq64=0
407
408To verify the selected entry size, enable debug logging
409(``--log-level=enic,debug``) and check the following messages.
410
411.. code-block:: console
412
413    PMD: rte_enic_pmd: Supported CQ entry sizes: 16 32
414    PMD: rte_enic_pmd: Using 16B CQ entry size
415
416.. _enic_limitations:
417
418Limitations
419-----------
420
421- **VLAN 0 Priority Tagging**
422
423  If a vNIC is configured in TRUNK mode by the UCS manager, the adapter will
424  priority tag egress packets according to 802.1Q if they were not already
425  VLAN tagged by software. If the adapter is connected to a properly configured
426  switch, there will be no unexpected behavior.
427
428  In test setups where an Ethernet port of a Cisco adapter in TRUNK mode is
429  connected point-to-point to another adapter port or connected though a router
430  instead of a switch, all ingress packets will be VLAN tagged. Programs such
431  as l3fwd may not account for VLAN tags in packets and may misbehave. One
432  solution is to enable VLAN stripping on ingress so the VLAN tag is removed
433  from the packet and put into the mbuf->vlan_tci field. Here is an example
434  of how to accomplish this:
435
436.. code-block:: console
437
438     vlan_offload = rte_eth_dev_get_vlan_offload(port);
439     vlan_offload |= ETH_VLAN_STRIP_OFFLOAD;
440     rte_eth_dev_set_vlan_offload(port, vlan_offload);
441
442Another alternative is modify the adapter's ingress VLAN rewrite mode so that
443packets with the default VLAN tag are stripped by the adapter and presented to
444DPDK as untagged packets. In this case mbuf->vlan_tci and the PKT_RX_VLAN and
445PKT_RX_VLAN_STRIPPED mbuf flags would not be set. This mode is enabled with the
446``devargs`` parameter ``ig-vlan-rewrite=untag``. For example::
447
448    -a 12:00.0,ig-vlan-rewrite=untag
449
450- **SR-IOV**
451
452  - KVM hypervisor support only. VMware has not been tested.
453  - Requires VM-FEX, and so is only available on UCS managed servers connected
454    to Fabric Interconnects. It is not on standalone C-Series servers.
455  - VF devices are not usable directly from the host. They can  only be used
456    as assigned devices on VM instances.
457  - Currently, unbind of the ENIC kernel mode driver 'enic.ko' on the VM
458    instance may hang. As a workaround, enic.ko should be blocked or removed
459    from the boot process.
460  - pci_generic cannot be used as the uio module in the VM. igb_uio or
461    vfio in non-IOMMU mode can be used.
462  - The number of RQs in UCSM dynamic vNIC configurations must be at least 2.
463  - The number of SR-IOV devices is limited to 256. Components on target system
464    might limit this number to fewer than 256.
465
466- **Flow API**
467
468  - The number of filters that can be specified with the Generic Flow API is
469    dependent on how many header fields are being masked. Use 'flow create' in
470    a loop to determine how many filters your VIC will support (not more than
471    1000 for 1300 series VICs). Filters are checked for matching in the order they
472    were added. Since there currently is no grouping or priority support,
473    'catch-all' filters should be added last.
474  - The supported range of IDs for the 'MARK' action is 0 - 0xFFFD.
475  - RSS and PASSTHRU actions only support "receive normally". They are limited
476    to supporting MARK + RSS and PASSTHRU + MARK to allow the application to mark
477    packets and then receive them normally. These require 1400 series VIC adapters
478    and latest firmware.
479  - RAW items are limited to matching UDP tunnel headers like VXLAN.
480  - For 1400 VICs, all flows using the RSS action on a port use same hash
481    configuration. The RETA is ignored. The queues used in the RSS group must be
482    sequential. There is a performance hit if the number of queues is not a power of 2.
483    Only level 0 (outer header) RSS is allowed.
484
485- **Statistics**
486
487  - ``rx_good_bytes`` (ibytes) always includes VLAN header (4B) and CRC bytes (4B).
488    This behavior applies to 1300 and older series VIC adapters.
489    1400 series VICs do not count CRC bytes, and count VLAN header only when VLAN
490    stripping is disabled.
491  - When the NIC drops a packet because the Rx queue has no free buffers,
492    ``rx_good_bytes`` still increments by 4B if the packet is not VLAN tagged or
493    VLAN stripping is disabled, or by 8B if the packet is VLAN tagged and stripping
494    is enabled.
495    This behavior applies to 1300 and older series VIC adapters. 1400 series VICs
496    do not increment this byte counter when packets are dropped.
497
498- **RSS Hashing**
499
500  - Hardware enables and disables UDP and TCP RSS hashing together. The driver
501    cannot control UDP and TCP hashing individually.
502
503How to build the suite
504----------------------
505
506The build instructions for the DPDK suite should be followed. By default
507the ENIC PMD library will be built into the DPDK library.
508
509Refer to the document :ref:`compiling and testing a PMD for a NIC
510<pmd_build_and_test>` for details.
511
512For configuring and using UIO and VFIO frameworks, please refer to the
513documentation that comes with DPDK suite.
514
515Supported Cisco VIC adapters
516----------------------------
517
518ENIC PMD supports all recent generations of Cisco VIC adapters including:
519
520- VIC 1200 series
521- VIC 1300 series
522- VIC 1400 series
523
524Supported Operating Systems
525---------------------------
526
527Any Linux distribution fulfilling the conditions described in Dependencies
528section of DPDK documentation.
529
530Supported features
531------------------
532
533- Unicast, multicast and broadcast transmission and reception
534- Receive queue polling
535- Port Hardware Statistics
536- Hardware VLAN acceleration
537- IP checksum offload
538- Receive side VLAN stripping
539- Multiple receive and transmit queues
540- Promiscuous mode
541- Setting RX VLAN (supported via UCSM/CIMC only)
542- VLAN filtering (supported via UCSM/CIMC only)
543- Execution of application by unprivileged system users
544- IPV4, IPV6 and TCP RSS hashing
545- UDP RSS hashing (1400 series and later adapters)
546- Scattered Rx
547- MTU update
548- SR-IOV on UCS managed servers connected to Fabric Interconnects
549- Flow API
550- Overlay offload
551
552  - Rx/Tx checksum offloads for VXLAN, NVGRE, GENEVE
553  - TSO for VXLAN and GENEVE packets
554  - Inner RSS
555
556Known bugs and unsupported features in this release
557---------------------------------------------------
558
559- Signature or flex byte based flow direction
560- Drop feature of flow direction
561- VLAN based flow direction
562- Non-IPV4 flow direction
563- Setting of extended VLAN
564- MTU update only works if Scattered Rx mode is disabled
565- Maximum receive packet length is ignored if Scattered Rx mode is used
566
567Prerequisites
568-------------
569
570- Prepare the system as recommended by DPDK suite.  This includes environment
571  variables, hugepages configuration, tool-chains and configuration.
572- Insert vfio-pci kernel module using the command 'modprobe vfio-pci' if the
573  user wants to use VFIO framework.
574- Insert uio kernel module using the command 'modprobe uio' if the user wants
575  to use UIO framework.
576- DPDK suite should be configured based on the user's decision to use VFIO or
577  UIO framework.
578- If the vNIC device(s) to be used is bound to the kernel mode Ethernet driver
579  use 'ip' to bring the interface down. The dpdk-devbind.py tool can
580  then be used to unbind the device's bus id from the ENIC kernel mode driver.
581- Bind the intended vNIC to vfio-pci in case the user wants ENIC PMD to use
582  VFIO framework using dpdk-devbind.py.
583- Bind the intended vNIC to igb_uio in case the user wants ENIC PMD to use
584  UIO framework using dpdk-devbind.py.
585
586At this point the system should be ready to run DPDK applications. Once the
587application runs to completion, the vNIC can be detached from vfio-pci or
588igb_uio if necessary.
589
590Root privilege is required to bind and unbind vNICs to/from VFIO/UIO.
591VFIO framework helps an unprivileged user to run the applications.
592For an unprivileged user to run the applications on DPDK and ENIC PMD,
593it may be necessary to increase the maximum locked memory of the user.
594The following command could be used to do this.
595
596.. code-block:: console
597
598    sudo sh -c "ulimit -l <value in Kilo Bytes>"
599
600The value depends on the memory configuration of the application, DPDK and
601PMD.  Typically, the limit has to be raised to higher than 2GB.
602e.g., 2621440
603
604Additional Reference
605--------------------
606
607- https://www.cisco.com/c/en/us/products/servers-unified-computing/index.html
608- https://www.cisco.com/c/en/us/products/interfaces-modules/unified-computing-system-adapters/index.html
609
610Contact Information
611-------------------
612
613Any questions or bugs should be reported to DPDK community and to the ENIC PMD
614maintainers:
615
616- John Daley <johndale@cisco.com>
617- Hyong Youb Kim <hyonkim@cisco.com>
618