4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 #include <sys/queue.h>
44 #include <netinet/ip.h>
46 #include <rte_common.h>
47 #include <rte_memory.h>
48 #include <rte_memzone.h>
49 #include <rte_tailq.h>
51 #include <rte_byteorder.h>
52 #include <rte_launch.h>
53 #include <rte_per_lcore.h>
54 #include <rte_lcore.h>
55 #include <rte_branch_prediction.h>
56 #include <rte_atomic.h>
59 #include <rte_debug.h>
60 #include <rte_mempool.h>
61 #include <rte_memcpy.h>
63 #include <rte_ether.h>
64 #include <rte_interrupts.h>
66 #include <rte_ethdev.h>
67 #include <rte_byteorder.h>
68 #include <rte_malloc.h>
69 #include <rte_fbk_hash.h>
70 #include <rte_string_fns.h>
78 * When doing reads from the NIC or the client queues,
81 #define PACKET_READ_SIZE 32
84 * Local buffers to put packets in, used to send packets in bursts to the
87 struct client_rx_buf {
88 struct rte_mbuf *buffer[PACKET_READ_SIZE];
92 /* One buffer per client rx queue - dynamically allocate array */
93 static struct client_rx_buf *cl_rx_buf;
96 get_printable_mac_addr(uint8_t port)
98 static const char err_address[] = "00:00:00:00:00:00";
99 static char addresses[RTE_MAX_ETHPORTS][sizeof(err_address)];
101 if (unlikely(port >= RTE_MAX_ETHPORTS))
103 if (unlikely(addresses[port][0]=='\0')){
104 struct ether_addr mac;
105 rte_eth_macaddr_get(port, &mac);
106 snprintf(addresses[port], sizeof(addresses[port]),
107 "%02x:%02x:%02x:%02x:%02x:%02x\n",
108 mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2],
109 mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]);
111 return addresses[port];
115 * This function displays the recorded statistics for each port
116 * and for each client. It uses ANSI terminal codes to clear
117 * screen when called. It is called from a single non-master
118 * thread in the server process, when the process is run with more
119 * than one lcore enabled.
122 do_stats_display(void)
125 const char clr[] = { 27, '[', '2', 'J', '\0' };
126 const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
127 uint64_t port_tx[RTE_MAX_ETHPORTS], port_tx_drop[RTE_MAX_ETHPORTS];
128 uint64_t client_tx[MAX_CLIENTS], client_tx_drop[MAX_CLIENTS];
130 /* to get TX stats, we need to do some summing calculations */
131 memset(port_tx, 0, sizeof(port_tx));
132 memset(port_tx_drop, 0, sizeof(port_tx_drop));
133 memset(client_tx, 0, sizeof(client_tx));
134 memset(client_tx_drop, 0, sizeof(client_tx_drop));
136 for (i = 0; i < num_clients; i++){
137 const volatile struct tx_stats *tx = &ports->tx_stats[i];
138 for (j = 0; j < ports->num_ports; j++){
139 /* assign to local variables here, save re-reading volatile vars */
140 const uint64_t tx_val = tx->tx[ports->id[j]];
141 const uint64_t drop_val = tx->tx_drop[ports->id[j]];
142 port_tx[j] += tx_val;
143 port_tx_drop[j] += drop_val;
144 client_tx[i] += tx_val;
145 client_tx_drop[i] += drop_val;
149 /* Clear screen and move to top left */
150 printf("%s%s", clr, topLeft);
154 for (i = 0; i < ports->num_ports; i++)
155 printf("Port %u: '%s'\t", (unsigned)ports->id[i],
156 get_printable_mac_addr(ports->id[i]));
158 for (i = 0; i < ports->num_ports; i++){
159 printf("Port %u - rx: %9"PRIu64"\t"
161 (unsigned)ports->id[i], ports->rx_stats.rx[i],
165 printf("\nCLIENTS\n");
167 for (i = 0; i < num_clients; i++){
168 const unsigned long long rx = clients[i].stats.rx;
169 const unsigned long long rx_drop = clients[i].stats.rx_drop;
170 printf("Client %2u - rx: %9llu, rx_drop: %9llu\n"
171 " tx: %9"PRIu64", tx_drop: %9"PRIu64"\n",
172 i, rx, rx_drop, client_tx[i], client_tx_drop[i]);
179 * The function called from each non-master lcore used by the process.
180 * The test_and_set function is used to randomly pick a single lcore on which
181 * the code to display the statistics will run. Otherwise, the code just
185 sleep_lcore(__attribute__((unused)) void *dummy)
187 /* Used to pick a display thread - static, so zero-initialised */
188 static rte_atomic32_t display_stats;
190 /* Only one core should display stats */
191 if (rte_atomic32_test_and_set(&display_stats)) {
192 const unsigned sleeptime = 1;
193 printf("Core %u displaying statistics\n", rte_lcore_id());
195 /* Longer initial pause so above printf is seen */
196 sleep(sleeptime * 3);
198 /* Loop forever: sleep always returns 0 or <= param */
199 while (sleep(sleeptime) <= sleeptime)
206 * Function to set all the client statistic values to zero.
207 * Called at program startup.
214 for (i = 0; i < num_clients; i++)
215 clients[i].stats.rx = clients[i].stats.rx_drop = 0;
219 * send a burst of traffic to a client, assuming there are packets
220 * available to be sent to this client
223 flush_rx_queue(uint16_t client)
228 if (cl_rx_buf[client].count == 0)
231 cl = &clients[client];
232 if (rte_ring_enqueue_bulk(cl->rx_q, (void **)cl_rx_buf[client].buffer,
233 cl_rx_buf[client].count) != 0){
234 for (j = 0; j < cl_rx_buf[client].count; j++)
235 rte_pktmbuf_free(cl_rx_buf[client].buffer[j]);
236 cl->stats.rx_drop += cl_rx_buf[client].count;
239 cl->stats.rx += cl_rx_buf[client].count;
241 cl_rx_buf[client].count = 0;
245 * marks a packet down to be sent to a particular client process
248 enqueue_rx_packet(uint8_t client, struct rte_mbuf *buf)
250 cl_rx_buf[client].buffer[cl_rx_buf[client].count++] = buf;
254 * This function takes a group of packets and routes them
255 * individually to the client process. Very simply round-robins the packets
256 * without checking any of the packet contents.
259 process_packets(uint32_t port_num __rte_unused,
260 struct rte_mbuf *pkts[], uint16_t rx_count)
265 for (i = 0; i < rx_count; i++) {
266 enqueue_rx_packet(client, pkts[i]);
268 if (++client == num_clients)
272 for (i = 0; i < num_clients; i++)
277 * Function called by the master lcore of the DPDK process.
280 do_packet_forwarding(void)
282 unsigned port_num = 0; /* indexes the port[] array */
285 struct rte_mbuf *buf[PACKET_READ_SIZE];
289 rx_count = rte_eth_rx_burst(ports->id[port_num], 0, \
290 buf, PACKET_READ_SIZE);
291 ports->rx_stats.rx[port_num] += rx_count;
293 /* Now process the NIC packets read */
294 if (likely(rx_count > 0))
295 process_packets(port_num, buf, rx_count);
297 /* move to next port */
298 if (++port_num == ports->num_ports)
304 MAIN(int argc, char *argv[])
306 /* initialise the system */
307 if (init(argc, argv) < 0 )
309 RTE_LOG(INFO, APP, "Finished Process Init.\n");
311 cl_rx_buf = calloc(num_clients, sizeof(cl_rx_buf[0]));
313 /* clear statistics */
316 /* put all other cores to sleep bar master */
317 rte_eal_mp_remote_launch(sleep_lcore, NULL, SKIP_MASTER);
319 do_packet_forwarding();