1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2012 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
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
16  *       distribution.
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.
20  *
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.
32  *
33  *  version: DPDK.L.1.2.3-3
34  */
35 
36 #include <stdio.h>
37 #include <stdlib.h>
38 #include <string.h>
39 #include <unistd.h>
40 #include <stdint.h>
41 #include <stdarg.h>
42 #include <inttypes.h>
43 #include <inttypes.h>
44 #include <sys/queue.h>
45 #include <errno.h>
46 #include <netinet/ip.h>
47 
48 #include <rte_common.h>
49 #include <rte_memory.h>
50 #include <rte_memzone.h>
51 #include <rte_tailq.h>
52 #include <rte_eal.h>
53 #include <rte_byteorder.h>
54 #include <rte_launch.h>
55 #include <rte_per_lcore.h>
56 #include <rte_lcore.h>
57 #include <rte_branch_prediction.h>
58 #include <rte_atomic.h>
59 #include <rte_ring.h>
60 #include <rte_log.h>
61 #include <rte_debug.h>
62 #include <rte_mempool.h>
63 #include <rte_memcpy.h>
64 #include <rte_mbuf.h>
65 #include <rte_ether.h>
66 #include <rte_interrupts.h>
67 #include <rte_pci.h>
68 #include <rte_ethdev.h>
69 #include <rte_byteorder.h>
70 #include <rte_malloc.h>
71 #include <rte_hash_crc.h>
72 #include <rte_fbk_hash.h>
73 #include <rte_string_fns.h>
74 
75 #include "common.h"
76 #include "args.h"
77 #include "init.h"
78 #include "main.h"
79 
80 /*
81  * When doing reads from the NIC or the client queues,
82  * use this batch size
83  */
84 #define PACKET_READ_SIZE 32
85 
86 /*
87  * Local buffers to put packets in, used to send packets in bursts to the
88  * clients
89  */
90 struct client_rx_buf {
91 	struct rte_mbuf *buffer[PACKET_READ_SIZE];
92 	uint16_t count;
93 };
94 
95 /* One buffer per client rx queue - dynamically allocate array */
96 static struct client_rx_buf *cl_rx_buf;
97 
98 static const char *
99 get_printable_mac_addr(uint8_t port)
100 {
101 	static const char err_address[] = "00:00:00:00:00:00";
102 	static char addresses[RTE_MAX_ETHPORTS][sizeof(err_address)];
103 
104 	if (unlikely(port >= RTE_MAX_ETHPORTS))
105 		return err_address;
106 	if (unlikely(addresses[port][0]=='\0')){
107 		struct ether_addr mac;
108 		rte_eth_macaddr_get(port, &mac);
109 		rte_snprintf(addresses[port], sizeof(addresses[port]),
110 				"%02x:%02x:%02x:%02x:%02x:%02x\n",
111 				mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2],
112 				mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]);
113 	}
114 	return addresses[port];
115 }
116 
117 /*
118  * This function displays the recorded statistics for each port
119  * and for each client. It uses ANSI terminal codes to clear
120  * screen when called. It is called from a single non-master
121  * thread in the server process, when the process is run with more
122  * than one lcore enabled.
123  */
124 static void
125 do_stats_display(void)
126 {
127 	unsigned i, j;
128 	const char clr[] = { 27, '[', '2', 'J', '\0' };
129 	const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
130 	uint64_t port_tx[RTE_MAX_ETHPORTS], port_tx_drop[RTE_MAX_ETHPORTS];
131 	uint64_t client_tx[MAX_CLIENTS], client_tx_drop[MAX_CLIENTS];
132 
133 	/* to get TX stats, we need to do some summing calculations */
134 	memset(port_tx, 0, sizeof(port_tx));
135 	memset(port_tx_drop, 0, sizeof(port_tx_drop));
136 	memset(client_tx, 0, sizeof(client_tx));
137 	memset(client_tx_drop, 0, sizeof(client_tx_drop));
138 
139 	for (i = 0; i < num_clients; i++){
140 		const volatile struct tx_stats *tx = &ports->tx_stats[i];
141 		for (j = 0; j < ports->num_ports; j++){
142 			/* assign to local variables here, save re-reading volatile vars */
143 			const uint64_t tx_val = tx->tx[j];
144 			const uint64_t drop_val = tx->tx_drop[j];
145 			port_tx[j] += tx_val;
146 			port_tx_drop[j] += drop_val;
147 			client_tx[i] += tx_val;
148 			client_tx_drop[i] += drop_val;
149 		}
150 	}
151 
152 	/* Clear screen and move to top left */
153 	printf("%s%s", clr, topLeft);
154 
155 	printf("PORTS\n");
156 	printf("-----\n");
157 	for (i = 0; i < ports->num_ports; i++)
158 		printf("Port %u: '%s'\t", (unsigned)ports->id[i],
159 				get_printable_mac_addr(ports->id[i]));
160 	printf("\n\n");
161 	for (i = 0; i < ports->num_ports; i++){
162 		printf("Port %u - rx: %9"PRIu64"\t"
163 				"tx: %9"PRIu64"\n",
164 				(unsigned)ports->id[i], ports->rx_stats.rx[i],
165 				port_tx[i]);
166 	}
167 
168 	printf("\nCLIENTS\n");
169 	printf("-------\n");
170 	for (i = 0; i < num_clients; i++){
171 		const unsigned long long rx = clients[i].stats.rx;
172 		const unsigned long long rx_drop = clients[i].stats.rx_drop;
173 		printf("Client %2u - rx: %9llu, rx_drop: %9llu\n"
174 				"            tx: %9"PRIu64", tx_drop: %9"PRIu64"\n",
175 				i, rx, rx_drop, client_tx[i], client_tx_drop[i]);
176 	}
177 
178 	printf("\n");
179 }
180 
181 /*
182  * The function called from each non-master lcore used by the process.
183  * The test_and_set function is used to randomly pick a single lcore on which
184  * the code to display the statistics will run. Otherwise, the code just
185  * repeatedly sleeps.
186  */
187 static int
188 sleep_lcore(__attribute__((unused)) void *dummy)
189 {
190 	/* Used to pick a display thread - static, so zero-initialised */
191 	static rte_atomic32_t display_stats;
192 
193 	/* Only one core should display stats */
194 	if (rte_atomic32_test_and_set(&display_stats)) {
195 		const unsigned sleeptime = 1;
196 		printf("Core %u displaying statistics\n", rte_lcore_id());
197 
198 		/* Longer initial pause so above printf is seen */
199 		sleep(sleeptime * 3);
200 
201 		/* Loop forever: sleep always returns 0 or <= param */
202 		while (sleep(sleeptime) <= sleeptime)
203 			do_stats_display();
204 	}
205 	else {
206 		const unsigned sleeptime = 100;
207 		printf("Putting core %u to sleep\n", rte_lcore_id());
208 		while (sleep(sleeptime) <= sleeptime)
209 			; /* loop doing nothing */
210 	}
211 	return 0;
212 }
213 
214 /*
215  * Function to set all the client statistic values to zero.
216  * Called at program startup.
217  */
218 static void
219 clear_stats(void)
220 {
221 	unsigned i;
222 
223 	for (i = 0; i < num_clients; i++)
224 		clients[i].stats.rx = clients[i].stats.rx_drop = 0;
225 }
226 
227 /*
228  * send a burst of traffic to a client, assuming there are packets
229  * available to be sent to this client
230  */
231 static void
232 flush_rx_queue(uint16_t client)
233 {
234 	uint16_t j;
235 	struct client *cl;
236 
237 	if (cl_rx_buf[client].count == 0)
238 		return;
239 
240 	cl = &clients[client];
241 	if (rte_ring_enqueue_bulk(cl->rx_q, (void **)cl_rx_buf[client].buffer,
242 			cl_rx_buf[client].count) != 0){
243 		for (j = 0; j < cl_rx_buf[client].count; j++)
244 			rte_pktmbuf_free(cl_rx_buf[client].buffer[j]);
245 		cl->stats.rx_drop += cl_rx_buf[client].count;
246 	}
247 	else
248 		cl->stats.rx += cl_rx_buf[client].count;
249 
250 	cl_rx_buf[client].count = 0;
251 }
252 
253 /*
254  * marks a packet down to be sent to a particular client process
255  */
256 static inline void
257 enqueue_rx_packet(uint8_t client, struct rte_mbuf *buf)
258 {
259 	cl_rx_buf[client].buffer[cl_rx_buf[client].count++] = buf;
260 }
261 
262 /*
263  * This function takes a group of packets and routes them
264  * individually to the client process. Very simply round-robins the packets
265  * without checking any of the packet contents.
266  */
267 static void
268 process_packets(uint32_t port_num __rte_unused,
269 		struct rte_mbuf *pkts[], uint16_t rx_count)
270 {
271 	uint16_t i;
272 	uint8_t client = 0;
273 
274 	for (i = 0; i < rx_count; i++) {
275 		enqueue_rx_packet(client, pkts[i]);
276 
277 		if (++client == num_clients)
278 			client = 0;
279 	}
280 
281 	for (i = 0; i < num_clients; i++)
282 		flush_rx_queue(i);
283 }
284 
285 /*
286  * Function called by the master lcore of the DPDK process.
287  */
288 static void
289 do_packet_forwarding(void)
290 {
291 	unsigned port_num = 0; /* indexes the port[] array */
292 
293 	for (;;) {
294 		struct rte_mbuf *buf[PACKET_READ_SIZE];
295 		uint16_t rx_count;
296 
297 		/* read a port */
298 		rx_count = rte_eth_rx_burst(ports->id[port_num], 0, \
299 				buf, PACKET_READ_SIZE);
300 		ports->rx_stats.rx[port_num] += rx_count;
301 
302 		/* Now process the NIC packets read */
303 		if (likely(rx_count > 0))
304 			process_packets(port_num, buf, rx_count);
305 
306 		/* move to next port */
307 		if (++port_num == ports->num_ports)
308 			port_num = 0;
309 	}
310 }
311 
312 int
313 MAIN(int argc, char *argv[])
314 {
315 	/* initialise the system */
316 	if (init(argc, argv) < 0 )
317 		return -1;
318 	RTE_LOG(INFO, APP, "Finished Process Init.\n");
319 
320 	cl_rx_buf = calloc(num_clients, sizeof(cl_rx_buf[0]));
321 
322 	/* clear statistics */
323 	clear_stats();
324 
325 	/* put all other cores to sleep bar master */
326 	rte_eal_mp_remote_launch(sleep_lcore, NULL, SKIP_MASTER);
327 
328 	do_packet_forwarding();
329 	return 0;
330 }
331