xref: /dpdk/doc/guides/sample_app_ug/rxtx_callbacks.rst (revision 089e5ed727a15da2729cfee9b63533dd120bd04c)
1..  SPDX-License-Identifier: BSD-3-Clause
2    Copyright(c) 2015 Intel Corporation.
3
4RX/TX Callbacks Sample Application
5==================================
6
7The RX/TX Callbacks sample application is a packet forwarding application that
8demonstrates the use of user defined callbacks on received and transmitted
9packets. The application performs a simple latency check, using callbacks, to
10determine the time packets spend within the application.
11
12In the sample application a user defined callback is applied to all received
13packets to add a timestamp. A separate callback is applied to all packets
14prior to transmission to calculate the elapsed time, in CPU cycles.
15
16If hardware timestamping is supported by the NIC, the sample application will
17also display the average latency since the packet was timestamped in hardware,
18on top of the latency since the packet was received and processed by the RX
19callback.
20
21Compiling the Application
22-------------------------
23
24To compile the sample application see :doc:`compiling`.
25
26The application is located in the ``rxtx_callbacks`` sub-directory.
27
28The callbacks feature requires that the ``CONFIG_RTE_ETHDEV_RXTX_CALLBACKS``
29setting is on in the ``config/common_`` config file that applies to the
30target. This is generally on by default:
31
32.. code-block:: console
33
34    CONFIG_RTE_ETHDEV_RXTX_CALLBACKS=y
35
36Running the Application
37-----------------------
38
39To run the example in a ``linux`` environment:
40
41.. code-block:: console
42
43    ./build/rxtx_callbacks -l 1 -n 4 -- [-t]
44
45Use -t to enable hardware timestamping. If not supported by the NIC, an error
46will be displayed.
47
48Refer to *DPDK Getting Started Guide* for general information on running
49applications and the Environment Abstraction Layer (EAL) options.
50
51
52
53Explanation
54-----------
55
56The ``rxtx_callbacks`` application is mainly a simple forwarding application
57based on the :doc:`skeleton`. See that section of the documentation for more
58details of the forwarding part of the application.
59
60The sections below explain the additional RX/TX callback code.
61
62
63The Main Function
64~~~~~~~~~~~~~~~~~
65
66The ``main()`` function performs the application initialization and calls the
67execution threads for each lcore. This function is effectively identical to
68the ``main()`` function explained in :doc:`skeleton`.
69
70The ``lcore_main()`` function is also identical.
71
72The main difference is in the user defined ``port_init()`` function where the
73callbacks are added. This is explained in the next section:
74
75
76The Port Initialization  Function
77~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
78
79The main functional part of the port initialization is shown below with
80comments:
81
82.. code-block:: c
83
84    static inline int
85    port_init(uint16_t port, struct rte_mempool *mbuf_pool)
86    {
87        struct rte_eth_conf port_conf = port_conf_default;
88        const uint16_t rx_rings = 1, tx_rings = 1;
89        struct rte_ether_addr addr;
90        int retval;
91        uint16_t q;
92
93        /* Configure the Ethernet device. */
94        retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
95        if (retval != 0)
96            return retval;
97
98        /* Allocate and set up 1 RX queue per Ethernet port. */
99        for (q = 0; q < rx_rings; q++) {
100            retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE,
101                    rte_eth_dev_socket_id(port), NULL, mbuf_pool);
102            if (retval < 0)
103                return retval;
104        }
105
106        /* Allocate and set up 1 TX queue per Ethernet port. */
107        for (q = 0; q < tx_rings; q++) {
108            retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE,
109                    rte_eth_dev_socket_id(port), NULL);
110            if (retval < 0)
111                return retval;
112        }
113
114        /* Start the Ethernet port. */
115        retval = rte_eth_dev_start(port);
116        if (retval < 0)
117            return retval;
118
119        /* Enable RX in promiscuous mode for the Ethernet device. */
120        rte_eth_promiscuous_enable(port);
121
122
123        /* Add the callbacks for RX and TX.*/
124        rte_eth_add_rx_callback(port, 0, add_timestamps, NULL);
125        rte_eth_add_tx_callback(port, 0, calc_latency, NULL);
126
127        return 0;
128    }
129
130
131The RX and TX callbacks are added to the ports/queues as function pointers:
132
133.. code-block:: c
134
135        rte_eth_add_rx_callback(port, 0, add_timestamps, NULL);
136        rte_eth_add_tx_callback(port, 0, calc_latency,   NULL);
137
138More than one callback can be added and additional information can be passed
139to callback function pointers as a ``void*``. In the examples above ``NULL``
140is used.
141
142The ``add_timestamps()`` and ``calc_latency()`` functions are explained below.
143
144
145The add_timestamps() Callback
146~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
147
148The ``add_timestamps()`` callback is added to the RX port and is applied to
149all packets received:
150
151.. code-block:: c
152
153    static uint16_t
154    add_timestamps(uint16_t port __rte_unused, uint16_t qidx __rte_unused,
155            struct rte_mbuf **pkts, uint16_t nb_pkts, void *_ __rte_unused)
156    {
157        unsigned i;
158        uint64_t now = rte_rdtsc();
159
160        for (i = 0; i < nb_pkts; i++)
161            pkts[i]->udata64 = now;
162
163        return nb_pkts;
164    }
165
166The DPDK function ``rte_rdtsc()`` is used to add a cycle count timestamp to
167each packet (see the *cycles* section of the *DPDK API Documentation* for
168details).
169
170
171The calc_latency() Callback
172~~~~~~~~~~~~~~~~~~~~~~~~~~~
173
174The ``calc_latency()`` callback is added to the TX port and is applied to all
175packets prior to transmission:
176
177.. code-block:: c
178
179    static uint16_t
180    calc_latency(uint16_t port __rte_unused, uint16_t qidx __rte_unused,
181            struct rte_mbuf **pkts, uint16_t nb_pkts, void *_ __rte_unused)
182    {
183        uint64_t cycles = 0;
184        uint64_t now = rte_rdtsc();
185        unsigned i;
186
187        for (i = 0; i < nb_pkts; i++)
188            cycles += now - pkts[i]->udata64;
189
190        latency_numbers.total_cycles += cycles;
191        latency_numbers.total_pkts   += nb_pkts;
192
193        if (latency_numbers.total_pkts > (100 * 1000 * 1000ULL)) {
194            printf("Latency = %"PRIu64" cycles\n",
195                    latency_numbers.total_cycles / latency_numbers.total_pkts);
196
197            latency_numbers.total_cycles = latency_numbers.total_pkts = 0;
198        }
199
200        return nb_pkts;
201    }
202
203The ``calc_latency()`` function accumulates the total number of packets and
204the total number of cycles used. Once more than 100 million packets have been
205transmitted the average cycle count per packet is printed out and the counters
206are reset.
207