ceph/src/spdk/dpdk/doc/guides/sample_app_ug/skeleton.rst

   1 ..  SPDX-License-Identifier: BSD-3-Clause
   2     Copyright(c) 2015 Intel Corporation.
   3
   4 Basic Forwarding Sample Application
   5 ===================================
   6
   7 The Basic Forwarding sample application is a simple *skeleton* example of a
   8 forwarding application.
   9
  10 It is intended as a demonstration of the basic components of a DPDK forwarding
  11 application. For more detailed implementations see the L2 and L3 forwarding
  12 sample applications.
  13
  14 Compiling the Application
  15 -------------------------
  16
  17 To compile the sample application see :doc:`compiling`.
  18
  19 The application is located in the ``skeleton`` sub-directory.
  20
  21 Running the Application
  22 -----------------------
  23
  24 To run the example in a ``linux`` environment:
  25
  26 .. code-block:: console
  27
  28     ./build/basicfwd -l 1 -n 4
  29
  30 Refer to *DPDK Getting Started Guide* for general information on running
  31 applications and the Environment Abstraction Layer (EAL) options.
  32
  33
  34 Explanation
  35 -----------
  36
  37 The following sections provide an explanation of the main components of the
  38 code.
  39
  40 All DPDK library functions used in the sample code are prefixed with ``rte_``
  41 and are explained in detail in the *DPDK API Documentation*.
  42
  43
  44 The Main Function
  45 ~~~~~~~~~~~~~~~~~
  46
  47 The ``main()`` function performs the initialization and calls the execution
  48 threads for each lcore.
  49
  50 The first task is to initialize the Environment Abstraction Layer (EAL).  The
  51 ``argc`` and ``argv`` arguments are provided to the ``rte_eal_init()``
  52 function. The value returned is the number of parsed arguments:
  53
  54 .. code-block:: c
  55
  56     int ret = rte_eal_init(argc, argv);
  57     if (ret < 0)
  58         rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
  59
  60
  61 The ``main()`` also allocates a mempool to hold the mbufs (Message Buffers)
  62 used by the application:
  63
  64 .. code-block:: c
  65
  66     mbuf_pool = rte_mempool_create("MBUF_POOL",
  67                                    NUM_MBUFS * nb_ports,
  68                                    MBUF_SIZE,
  69                                    MBUF_CACHE_SIZE,
  70                                    sizeof(struct rte_pktmbuf_pool_private),
  71                                    rte_pktmbuf_pool_init, NULL,
  72                                    rte_pktmbuf_init,      NULL,
  73                                    rte_socket_id(),
  74                                    0);
  75
  76 Mbufs are the packet buffer structure used by DPDK. They are explained in
  77 detail in the "Mbuf Library" section of the *DPDK Programmer's Guide*.
  78
  79 The ``main()`` function also initializes all the ports using the user defined
  80 ``port_init()`` function which is explained in the next section:
  81
  82 .. code-block:: c
  83
  84     RTE_ETH_FOREACH_DEV(portid) {
  85         if (port_init(portid, mbuf_pool) != 0) {
  86             rte_exit(EXIT_FAILURE,
  87                      "Cannot init port %" PRIu8 "\n", portid);
  88         }
  89     }
  90
  91
  92 Once the initialization is complete, the application is ready to launch a
  93 function on an lcore. In this example ``lcore_main()`` is called on a single
  94 lcore.
  95
  96
  97 .. code-block:: c
  98
  99         lcore_main();
 100
 101 The ``lcore_main()`` function is explained below.
 102
 103
 104
 105 The Port Initialization  Function
 106 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 107
 108 The main functional part of the port initialization used in the Basic
 109 Forwarding application is shown below:
 110
 111 .. code-block:: c
 112
 113     static inline int
 114     port_init(uint16_t port, struct rte_mempool *mbuf_pool)
 115     {
 116         struct rte_eth_conf port_conf = port_conf_default;
 117         const uint16_t rx_rings = 1, tx_rings = 1;
 118         struct ether_addr addr;
 119         int retval;
 120         uint16_t q;
 121
 122         if (!rte_eth_dev_is_valid_port(port))
 123             return -1;
 124
 125         /* Configure the Ethernet device. */
 126         retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
 127         if (retval != 0)
 128             return retval;
 129
 130         /* Allocate and set up 1 RX queue per Ethernet port. */
 131         for (q = 0; q < rx_rings; q++) {
 132             retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE,
 133                     rte_eth_dev_socket_id(port), NULL, mbuf_pool);
 134             if (retval < 0)
 135                 return retval;
 136         }
 137
 138         /* Allocate and set up 1 TX queue per Ethernet port. */
 139         for (q = 0; q < tx_rings; q++) {
 140             retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE,
 141                     rte_eth_dev_socket_id(port), NULL);
 142             if (retval < 0)
 143                 return retval;
 144         }
 145
 146         /* Start the Ethernet port. */
 147         retval = rte_eth_dev_start(port);
 148         if (retval < 0)
 149             return retval;
 150
 151         /* Enable RX in promiscuous mode for the Ethernet device. */
 152         rte_eth_promiscuous_enable(port);
 153
 154         return 0;
 155     }
 156
 157 The Ethernet ports are configured with default settings using the
 158 ``rte_eth_dev_configure()`` function and the ``port_conf_default`` struct:
 159
 160 .. code-block:: c
 161
 162     static const struct rte_eth_conf port_conf_default = {
 163         .rxmode = { .max_rx_pkt_len = ETHER_MAX_LEN }
 164     };
 165
 166 For this example the ports are set up with 1 RX and 1 TX queue using the
 167 ``rte_eth_rx_queue_setup()`` and ``rte_eth_tx_queue_setup()`` functions.
 168
 169 The Ethernet port is then started:
 170
 171 .. code-block:: c
 172
 173         retval  = rte_eth_dev_start(port);
 174
 175
 176 Finally the RX port is set in promiscuous mode:
 177
 178 .. code-block:: c
 179
 180         rte_eth_promiscuous_enable(port);
 181
 182
 183 The Lcores Main
 184 ~~~~~~~~~~~~~~~
 185
 186 As we saw above the ``main()`` function calls an application function on the
 187 available lcores. For the Basic Forwarding application the lcore function
 188 looks like the following:
 189
 190 .. code-block:: c
 191
 192     static __attribute__((noreturn)) void
 193     lcore_main(void)
 194     {
 195         uint16_t port;
 196
 197         /*
 198          * Check that the port is on the same NUMA node as the polling thread
 199          * for best performance.
 200          */
 201         RTE_ETH_FOREACH_DEV(port)
 202             if (rte_eth_dev_socket_id(port) > 0 &&
 203                     rte_eth_dev_socket_id(port) !=
 204                             (int)rte_socket_id())
 205                 printf("WARNING, port %u is on remote NUMA node to "
 206                         "polling thread.\n\tPerformance will "
 207                         "not be optimal.\n", port);
 208
 209         printf("\nCore %u forwarding packets. [Ctrl+C to quit]\n",
 210                 rte_lcore_id());
 211
 212         /* Run until the application is quit or killed. */
 213         for (;;) {
 214             /*
 215              * Receive packets on a port and forward them on the paired
 216              * port. The mapping is 0 -> 1, 1 -> 0, 2 -> 3, 3 -> 2, etc.
 217              */
 218             RTE_ETH_FOREACH_DEV(port) {
 219
 220                 /* Get burst of RX packets, from first port of pair. */
 221                 struct rte_mbuf *bufs[BURST_SIZE];
 222                 const uint16_t nb_rx = rte_eth_rx_burst(port, 0,
 223                         bufs, BURST_SIZE);
 224
 225                 if (unlikely(nb_rx == 0))
 226                     continue;
 227
 228                 /* Send burst of TX packets, to second port of pair. */
 229                 const uint16_t nb_tx = rte_eth_tx_burst(port ^ 1, 0,
 230                         bufs, nb_rx);
 231
 232                 /* Free any unsent packets. */
 233                 if (unlikely(nb_tx < nb_rx)) {
 234                     uint16_t buf;
 235                     for (buf = nb_tx; buf < nb_rx; buf++)
 236                         rte_pktmbuf_free(bufs[buf]);
 237                 }
 238             }
 239         }
 240     }
 241
 242
 243 The main work of the application is done within the loop:
 244
 245 .. code-block:: c
 246
 247         for (;;) {
 248             RTE_ETH_FOREACH_DEV(port) {
 249
 250                 /* Get burst of RX packets, from first port of pair. */
 251                 struct rte_mbuf *bufs[BURST_SIZE];
 252                 const uint16_t nb_rx = rte_eth_rx_burst(port, 0,
 253                         bufs, BURST_SIZE);
 254
 255                 if (unlikely(nb_rx == 0))
 256                     continue;
 257
 258                 /* Send burst of TX packets, to second port of pair. */
 259                 const uint16_t nb_tx = rte_eth_tx_burst(port ^ 1, 0,
 260                         bufs, nb_rx);
 261
 262                 /* Free any unsent packets. */
 263                 if (unlikely(nb_tx < nb_rx)) {
 264                     uint16_t buf;
 265                     for (buf = nb_tx; buf < nb_rx; buf++)
 266                         rte_pktmbuf_free(bufs[buf]);
 267                 }
 268             }
 269         }
 270
 271 Packets are received in bursts on the RX ports and transmitted in bursts on
 272 the TX ports. The ports are grouped in pairs with a simple mapping scheme
 273 using the an XOR on the port number::
 274
 275     0 -> 1
 276     1 -> 0
 277
 278     2 -> 3
 279     3 -> 2
 280
 281     etc.
 282
 283 The ``rte_eth_tx_burst()`` function frees the memory buffers of packets that
 284 are transmitted. If packets fail to transmit, ``(nb_tx < nb_rx)``, then they
 285 must be freed explicitly using ``rte_pktmbuf_free()``.
 286
 287 The forwarding loop can be interrupted and the application closed using
 288 ``Ctrl-C``.