ceph/src/seastar/dpdk/doc/guides/nics/nfb.rst

   1 ..  SPDX-License-Identifier: BSD-3-Clause
   2     Copyright 2019 Cesnet
   3     Copyright 2019 Netcope Technologies
   4
   5 NFB poll mode driver library
   6 =================================
   7
   8 The NFB poll mode driver library implements support for the Netcope
   9 FPGA Boards (**NFB-***), FPGA-based programmable NICs.
  10 The NFB PMD uses interface provided by the libnfb library to communicate
  11 with the NFB cards over the nfb layer.
  12
  13 More information about the
  14 `NFB cards <http://www.netcope.com/en/products/fpga-boards>`_
  15 and used technology
  16 (`Netcope Development Kit <http://www.netcope.com/en/products/fpga-development-kit>`_)
  17 can be found on the `Netcope Technologies website <http://www.netcope.com/>`_.
  18
  19 .. note::
  20
  21    This driver has external dependencies.
  22    Therefore it is disabled in default configuration files.
  23    It can be enabled by setting ``CONFIG_RTE_LIBRTE_NFB_PMD=y``
  24    and recompiling.
  25
  26 .. note::
  27
  28    Currently the driver is supported only on x86_64 architectures.
  29    Only x86_64 versions of the external libraries are provided.
  30
  31 Prerequisites
  32 -------------
  33
  34 This PMD requires kernel modules which are responsible for initialization and
  35 allocation of resources needed for nfb layer function.
  36 Communication between PMD and kernel modules is mediated by libnfb library.
  37 These kernel modules and library are not part of DPDK and must be installed
  38 separately:
  39
  40 *  **libnfb library**
  41
  42    The library provides API for initialization of nfb transfers, receiving and
  43    transmitting data segments.
  44
  45 *  **Kernel modules**
  46
  47    * nfb
  48
  49    Kernel modules manage initialization of hardware, allocation and
  50    sharing of resources for user space applications.
  51
  52 Dependencies can be found here:
  53 `Netcope common <https://www.netcope.com/en/company/community-support/dpdk-libsze2#NFB>`_.
  54
  55 Versions of the packages
  56 ~~~~~~~~~~~~~~~~~~~~~~~~
  57
  58 The minimum version of the provided packages:
  59
  60 * for DPDK from 19.05
  61
  62 Configuration
  63 -------------
  64
  65 These configuration options can be modified before compilation in the
  66 ``.config`` file:
  67
  68 *  ``CONFIG_RTE_LIBRTE_NFB_PMD`` default value: **n**
  69
  70    Value **y** enables compilation of nfb PMD.
  71
  72 Using the NFB PMD
  73 ----------------------
  74
  75 Kernel modules have to be loaded before running the DPDK application.
  76
  77 NFB card architecture
  78 ---------------------
  79
  80 The NFB cards are multi-port multi-queue cards, where (generally) data from any
  81 Ethernet port may be sent to any queue.
  82 They are represented in DPDK as a single port.
  83
  84 NFB-200G2QL card employs an add-on cable which allows to connect it to two
  85 physical PCI-E slots at the same time (see the diagram below).
  86 This is done to allow 200 Gbps of traffic to be transferred through the PCI-E
  87 bus (note that a single PCI-E 3.0 x16 slot provides only 125 Gbps theoretical
  88 throughput).
  89
  90 Although each slot may be connected to a different CPU and therefore to a different
  91 NUMA node, the card is represented as a single port in DPDK. To work with data
  92 from the individual queues on the right NUMA node, connection of NUMA nodes on
  93 first and last queue (each NUMA node has half of the queues) need to be checked.
  94
  95 .. figure:: img/szedata2_nfb200g_architecture.*
  96     :align: center
  97
  98     NFB-200G2QL high-level diagram
  99
 100 Limitations
 101 -----------
 102
 103 Driver is usable only on Linux architecture, namely on CentOS.
 104
 105 Since a card is always represented as a single port, but can be connected to two
 106 NUMA nodes, there is need for manual check where master/slave is connected.
 107
 108 Example of usage
 109 ----------------
 110
 111 Read packets from 0. and 1. receive queue and write them to 0. and 1.
 112 transmit queue:
 113
 114 .. code-block:: console
 115
 116    $RTE_TARGET/app/testpmd -l 0-3 -n 2 \
 117    -- --port-topology=chained --rxq=2 --txq=2 --nb-cores=2 -i -a
 118
 119 Example output:
 120
 121 .. code-block:: console
 122
 123    [...]
 124    EAL: PCI device 0000:06:00.0 on NUMA socket -1
 125    EAL:   probe driver: 1b26:c1c1 net_nfb
 126    PMD: Initializing NFB device (0000:06:00.0)
 127    PMD: Available DMA queues RX: 8 TX: 8
 128    PMD: NFB device (0000:06:00.0) successfully initialized
 129    Interactive-mode selected
 130    Auto-start selected
 131    Configuring Port 0 (socket 0)
 132    Port 0: 00:11:17:00:00:00
 133    Checking link statuses...
 134    Port 0 Link Up - speed 10000 Mbps - full-duplex
 135    Done
 136    Start automatic packet forwarding
 137      io packet forwarding - CRC stripping disabled - packets/burst=32
 138      nb forwarding cores=2 - nb forwarding ports=1
 139      RX queues=2 - RX desc=128 - RX free threshold=0
 140      RX threshold registers: pthresh=0 hthresh=0 wthresh=0
 141      TX queues=2 - TX desc=512 - TX free threshold=0
 142      TX threshold registers: pthresh=0 hthresh=0 wthresh=0
 143      TX RS bit threshold=0 - TXQ flags=0x0
 144    testpmd>