xref: /dpdk/app/test-pmd/testpmd.c (revision dada9ef6edc59015b6674b5a95258787c71401b0)
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  */
34 
35 #include <stdarg.h>
36 #include <stdio.h>
37 #include <stdlib.h>
38 #include <signal.h>
39 #include <string.h>
40 #include <time.h>
41 #include <fcntl.h>
42 #include <sys/types.h>
43 #include <errno.h>
44 
45 #include <sys/queue.h>
46 #include <sys/stat.h>
47 
48 #include <stdint.h>
49 #include <unistd.h>
50 #include <inttypes.h>
51 
52 #include <rte_common.h>
53 #include <rte_byteorder.h>
54 #include <rte_log.h>
55 #include <rte_debug.h>
56 #include <rte_cycles.h>
57 #include <rte_memory.h>
58 #include <rte_memcpy.h>
59 #include <rte_memzone.h>
60 #include <rte_launch.h>
61 #include <rte_tailq.h>
62 #include <rte_eal.h>
63 #include <rte_per_lcore.h>
64 #include <rte_lcore.h>
65 #include <rte_atomic.h>
66 #include <rte_branch_prediction.h>
67 #include <rte_ring.h>
68 #include <rte_mempool.h>
69 #include <rte_malloc.h>
70 #include <rte_mbuf.h>
71 #include <rte_interrupts.h>
72 #include <rte_pci.h>
73 #include <rte_ether.h>
74 #include <rte_ethdev.h>
75 #include <rte_string_fns.h>
76 
77 #include "testpmd.h"
78 
79 uint16_t verbose_level = 0; /**< Silent by default. */
80 
81 /* use master core for command line ? */
82 uint8_t interactive = 0;
83 
84 /*
85  * NUMA support configuration.
86  * When set, the NUMA support attempts to dispatch the allocation of the
87  * RX and TX memory rings, and of the DMA memory buffers (mbufs) for the
88  * probed ports among the CPU sockets 0 and 1.
89  * Otherwise, all memory is allocated from CPU socket 0.
90  */
91 uint8_t numa_support = 0; /**< No numa support by default */
92 
93 /*
94  * Record the Ethernet address of peer target ports to which packets are
95  * forwarded.
96  * Must be instanciated with the ethernet addresses of peer traffic generator
97  * ports.
98  */
99 struct ether_addr peer_eth_addrs[RTE_MAX_ETHPORTS];
100 portid_t nb_peer_eth_addrs = 0;
101 
102 /*
103  * Probed Target Environment.
104  */
105 struct rte_port *ports;	       /**< For all probed ethernet ports. */
106 portid_t nb_ports;             /**< Number of probed ethernet ports. */
107 struct fwd_lcore **fwd_lcores; /**< For all probed logical cores. */
108 lcoreid_t nb_lcores;           /**< Number of probed logical cores. */
109 
110 /*
111  * Test Forwarding Configuration.
112  *    nb_fwd_lcores <= nb_cfg_lcores <= nb_lcores
113  *    nb_fwd_ports  <= nb_cfg_ports  <= nb_ports
114  */
115 lcoreid_t nb_cfg_lcores; /**< Number of configured logical cores. */
116 lcoreid_t nb_fwd_lcores; /**< Number of forwarding logical cores. */
117 portid_t  nb_cfg_ports;  /**< Number of configured ports. */
118 portid_t  nb_fwd_ports;  /**< Number of forwarding ports. */
119 
120 unsigned int fwd_lcores_cpuids[RTE_MAX_LCORE]; /**< CPU ids configuration. */
121 portid_t fwd_ports_ids[RTE_MAX_ETHPORTS];      /**< Port ids configuration. */
122 
123 struct fwd_stream **fwd_streams; /**< For each RX queue of each port. */
124 streamid_t nb_fwd_streams;       /**< Is equal to (nb_ports * nb_rxq). */
125 
126 /*
127  * Forwarding engines.
128  */
129 struct fwd_engine * fwd_engines[] = {
130 	&io_fwd_engine,
131 	&mac_fwd_engine,
132 	&rx_only_engine,
133 	&tx_only_engine,
134 	&csum_fwd_engine,
135 #ifdef RTE_LIBRTE_IEEE1588
136 	&ieee1588_fwd_engine,
137 #endif
138 	NULL,
139 };
140 
141 struct fwd_config cur_fwd_config;
142 struct fwd_engine *cur_fwd_eng = &io_fwd_engine; /**< IO mode by default. */
143 
144 uint16_t mbuf_data_size = DEFAULT_MBUF_DATA_SIZE; /**< Mbuf data space size. */
145 
146 /*
147  * Configuration of packet segments used by the "txonly" processing engine.
148  */
149 uint16_t tx_pkt_length = TXONLY_DEF_PACKET_LEN; /**< TXONLY packet length. */
150 uint16_t tx_pkt_seg_lengths[RTE_MAX_SEGS_PER_PKT] = {
151 	TXONLY_DEF_PACKET_LEN,
152 };
153 uint8_t  tx_pkt_nb_segs = 1; /**< Number of segments in TXONLY packets */
154 
155 uint16_t nb_pkt_per_burst = DEF_PKT_BURST; /**< Number of packets per burst. */
156 uint16_t mb_mempool_cache = DEF_PKT_BURST; /**< Size of mbuf mempool cache. */
157 
158 /*
159  * Ethernet Ports Configuration.
160  */
161 int promiscuous_on = 1; /**< Ports set in promiscuous mode by default. */
162 
163 /*
164  * Configurable number of RX/TX queues.
165  */
166 queueid_t nb_rxq = 1; /**< Number of RX queues per port. */
167 queueid_t nb_txq = 1; /**< Number of TX queues per port. */
168 
169 /*
170  * Configurable number of RX/TX ring descriptors.
171  */
172 #define RTE_TEST_RX_DESC_DEFAULT 128
173 #define RTE_TEST_TX_DESC_DEFAULT 512
174 uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; /**< Number of RX descriptors. */
175 uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; /**< Number of TX descriptors. */
176 
177 /*
178  * Configurable values of RX and TX ring threshold registers.
179  */
180 #define RX_PTHRESH 8 /**< Default value of RX prefetch threshold register. */
181 #define RX_HTHRESH 8 /**< Default value of RX host threshold register. */
182 #define RX_WTHRESH 4 /**< Default value of RX write-back threshold register. */
183 
184 #define TX_PTHRESH 36 /**< Default value of TX prefetch threshold register. */
185 #define TX_HTHRESH 0 /**< Default value of TX host threshold register. */
186 #define TX_WTHRESH 0 /**< Default value of TX write-back threshold register. */
187 
188 struct rte_eth_thresh rx_thresh = {
189 	.pthresh = RX_PTHRESH,
190 	.hthresh = RX_HTHRESH,
191 	.wthresh = RX_WTHRESH,
192 };
193 
194 struct rte_eth_thresh tx_thresh = {
195 	.pthresh = TX_PTHRESH,
196 	.hthresh = TX_HTHRESH,
197 	.wthresh = TX_WTHRESH,
198 };
199 
200 /*
201  * Configurable value of RX free threshold.
202  */
203 uint16_t rx_free_thresh = 0; /* Immediately free RX descriptors by default. */
204 
205 /*
206  * Configurable value of TX free threshold.
207  */
208 uint16_t tx_free_thresh = 0; /* Use default values. */
209 
210 /*
211  * Configurable value of TX RS bit threshold.
212  */
213 uint16_t tx_rs_thresh = 0; /* Use default values. */
214 
215 /*
216  * Receive Side Scaling (RSS) configuration.
217  */
218 uint16_t rss_hf = ETH_RSS_IPV4 | ETH_RSS_IPV6; /* RSS IP by default. */
219 
220 /*
221  * Port topology configuration
222  */
223 uint16_t port_topology = PORT_TOPOLOGY_PAIRED; /* Ports are paired by default */
224 
225 /*
226  * Ethernet device configuration.
227  */
228 struct rte_eth_rxmode rx_mode = {
229 	.max_rx_pkt_len = ETHER_MAX_LEN, /**< Default maximum frame length. */
230 	.split_hdr_size = 0,
231 	.header_split   = 0, /**< Header Split disabled. */
232 	.hw_ip_checksum = 0, /**< IP checksum offload disabled. */
233 	.hw_vlan_filter = 1, /**< VLAN filtering enabled. */
234 	.jumbo_frame    = 0, /**< Jumbo Frame Support disabled. */
235 	.hw_strip_crc   = 0, /**< CRC stripping by hardware disabled. */
236 };
237 
238 struct rte_fdir_conf fdir_conf = {
239 	.mode = RTE_FDIR_MODE_NONE,
240 	.pballoc = RTE_FDIR_PBALLOC_64K,
241 	.status = RTE_FDIR_REPORT_STATUS,
242 	.flexbytes_offset = 0x6,
243 	.drop_queue = 127,
244 };
245 
246 static volatile int test_done = 1; /* stop packet forwarding when set to 1. */
247 
248 /*
249  * Setup default configuration.
250  */
251 static void
252 set_default_fwd_lcores_config(void)
253 {
254 	unsigned int i;
255 	unsigned int nb_lc;
256 
257 	nb_lc = 0;
258 	for (i = 0; i < RTE_MAX_LCORE; i++) {
259 		if (! rte_lcore_is_enabled(i))
260 			continue;
261 		if (i == rte_get_master_lcore())
262 			continue;
263 		fwd_lcores_cpuids[nb_lc++] = i;
264 	}
265 	nb_lcores = (lcoreid_t) nb_lc;
266 	nb_cfg_lcores = nb_lcores;
267 	nb_fwd_lcores = 1;
268 }
269 
270 static void
271 set_def_peer_eth_addrs(void)
272 {
273 	portid_t i;
274 
275 	for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
276 		peer_eth_addrs[i].addr_bytes[0] = ETHER_LOCAL_ADMIN_ADDR;
277 		peer_eth_addrs[i].addr_bytes[5] = i;
278 	}
279 }
280 
281 static void
282 set_default_fwd_ports_config(void)
283 {
284 	portid_t pt_id;
285 
286 	for (pt_id = 0; pt_id < nb_ports; pt_id++)
287 		fwd_ports_ids[pt_id] = pt_id;
288 
289 	nb_cfg_ports = nb_ports;
290 	nb_fwd_ports = nb_ports;
291 }
292 
293 void
294 set_def_fwd_config(void)
295 {
296 	set_default_fwd_lcores_config();
297 	set_def_peer_eth_addrs();
298 	set_default_fwd_ports_config();
299 }
300 
301 /*
302  * Configuration initialisation done once at init time.
303  */
304 struct mbuf_ctor_arg {
305 	uint16_t seg_buf_offset; /**< offset of data in data segment of mbuf. */
306 	uint16_t seg_buf_size;   /**< size of data segment in mbuf. */
307 };
308 
309 struct mbuf_pool_ctor_arg {
310 	uint16_t seg_buf_size; /**< size of data segment in mbuf. */
311 };
312 
313 static void
314 testpmd_mbuf_ctor(struct rte_mempool *mp,
315 		  void *opaque_arg,
316 		  void *raw_mbuf,
317 		  __attribute__((unused)) unsigned i)
318 {
319 	struct mbuf_ctor_arg *mb_ctor_arg;
320 	struct rte_mbuf    *mb;
321 
322 	mb_ctor_arg = (struct mbuf_ctor_arg *) opaque_arg;
323 	mb = (struct rte_mbuf *) raw_mbuf;
324 
325 	mb->pool         = mp;
326 	mb->buf_addr     = (void *) ((char *)mb + mb_ctor_arg->seg_buf_offset);
327 	mb->buf_physaddr = (uint64_t) (rte_mempool_virt2phy(mp, mb) +
328 			mb_ctor_arg->seg_buf_offset);
329 	mb->buf_len      = mb_ctor_arg->seg_buf_size;
330 	mb->type         = RTE_MBUF_PKT;
331 	mb->ol_flags     = 0;
332 	mb->pkt.data     = (char *) mb->buf_addr + RTE_PKTMBUF_HEADROOM;
333 	mb->pkt.nb_segs  = 1;
334 	mb->pkt.l2_len = 0;
335 	mb->pkt.l3_len = 0;
336 	mb->pkt.vlan_tci = 0;
337 	mb->pkt.hash.rss = 0;
338 }
339 
340 static void
341 testpmd_mbuf_pool_ctor(struct rte_mempool *mp,
342 		       void *opaque_arg)
343 {
344 	struct mbuf_pool_ctor_arg      *mbp_ctor_arg;
345 	struct rte_pktmbuf_pool_private *mbp_priv;
346 
347 	if (mp->private_data_size < sizeof(struct rte_pktmbuf_pool_private)) {
348 		printf("%s(%s) private_data_size %d < %d\n",
349 		       __func__, mp->name, (int) mp->private_data_size,
350 		       (int) sizeof(struct rte_pktmbuf_pool_private));
351 		return;
352 	}
353 	mbp_ctor_arg = (struct mbuf_pool_ctor_arg *) opaque_arg;
354 	mbp_priv = (struct rte_pktmbuf_pool_private *)
355 		((char *)mp + sizeof(struct rte_mempool));
356 	mbp_priv->mbuf_data_room_size = mbp_ctor_arg->seg_buf_size;
357 }
358 
359 static void
360 mbuf_pool_create(uint16_t mbuf_seg_size, unsigned nb_mbuf,
361 		 unsigned int socket_id)
362 {
363 	char pool_name[RTE_MEMPOOL_NAMESIZE];
364 	struct rte_mempool *rte_mp;
365 	struct mbuf_pool_ctor_arg mbp_ctor_arg;
366 	struct mbuf_ctor_arg mb_ctor_arg;
367 	uint32_t mb_size;
368 
369 	mbp_ctor_arg.seg_buf_size = (uint16_t) (RTE_PKTMBUF_HEADROOM +
370 						mbuf_seg_size);
371 	mb_ctor_arg.seg_buf_offset =
372 		(uint16_t) CACHE_LINE_ROUNDUP(sizeof(struct rte_mbuf));
373 	mb_ctor_arg.seg_buf_size = mbp_ctor_arg.seg_buf_size;
374 	mb_size = mb_ctor_arg.seg_buf_offset + mb_ctor_arg.seg_buf_size;
375 	mbuf_poolname_build(socket_id, pool_name, sizeof(pool_name));
376 	rte_mp = rte_mempool_create(pool_name, nb_mbuf, (unsigned) mb_size,
377 				    (unsigned) mb_mempool_cache,
378 				    sizeof(struct rte_pktmbuf_pool_private),
379 				    testpmd_mbuf_pool_ctor, &mbp_ctor_arg,
380 				    testpmd_mbuf_ctor, &mb_ctor_arg,
381 				    socket_id, 0);
382 	if (rte_mp == NULL) {
383 		rte_exit(EXIT_FAILURE, "Creation of mbuf pool for socket %u failed\n",
384 		       socket_id);
385 	}
386 }
387 
388 static void
389 init_config(void)
390 {
391 	struct rte_port *port;
392 	struct rte_mempool *mbp;
393 	unsigned int nb_mbuf_per_pool;
394 	streamid_t sm_id;
395 	lcoreid_t  lc_id;
396 	portid_t   pt_id;
397 
398 	/* Configuration of logical cores. */
399 	fwd_lcores = rte_zmalloc("testpmd: fwd_lcores",
400 				sizeof(struct fwd_lcore *) * nb_lcores,
401 				CACHE_LINE_SIZE);
402 	if (fwd_lcores == NULL) {
403 		rte_exit(EXIT_FAILURE, "rte_zmalloc(%d (struct fwd_lcore *)) failed\n",
404 		       nb_lcores);
405 	}
406 	for (lc_id = 0; lc_id < nb_lcores; lc_id++) {
407 		fwd_lcores[lc_id] = rte_zmalloc("testpmd: struct fwd_lcore",
408 					       sizeof(struct fwd_lcore),
409 					       CACHE_LINE_SIZE);
410 		if (fwd_lcores[lc_id] == NULL) {
411 			rte_exit(EXIT_FAILURE, "rte_zmalloc(struct fwd_lcore) failed\n");
412 		}
413 		fwd_lcores[lc_id]->cpuid_idx = lc_id;
414 	}
415 
416 	/*
417 	 * Create pools of mbuf.
418 	 * If NUMA support is disabled, create a single pool of mbuf in
419 	 * socket 0 memory.
420 	 * Otherwise, create a pool of mbuf in the memory of sockets 0 and 1.
421 	 */
422 	nb_mbuf_per_pool = nb_rxd + (nb_lcores * mb_mempool_cache) +
423 		nb_txd + MAX_PKT_BURST;
424 	if (numa_support) {
425 		nb_mbuf_per_pool = nb_mbuf_per_pool * (nb_ports >> 1);
426 		mbuf_pool_create(mbuf_data_size, nb_mbuf_per_pool, 0);
427 		mbuf_pool_create(mbuf_data_size, nb_mbuf_per_pool, 1);
428 	} else {
429 		nb_mbuf_per_pool = (nb_mbuf_per_pool * nb_ports);
430 		mbuf_pool_create(mbuf_data_size, nb_mbuf_per_pool, 0);
431 	}
432 
433 	/*
434 	 * Records which Mbuf pool to use by each logical core, if needed.
435 	 */
436 	for (lc_id = 0; lc_id < nb_lcores; lc_id++) {
437 		mbp = mbuf_pool_find(rte_lcore_to_socket_id(lc_id));
438 		if (mbp == NULL)
439 			mbp = mbuf_pool_find(0);
440 		fwd_lcores[lc_id]->mbp = mbp;
441 	}
442 
443 	/* Configuration of Ethernet ports. */
444 	ports = rte_zmalloc("testpmd: ports",
445 			    sizeof(struct rte_port) * nb_ports,
446 			    CACHE_LINE_SIZE);
447 	if (ports == NULL) {
448 		rte_exit(EXIT_FAILURE, "rte_zmalloc(%d struct rte_port) failed\n",
449 		       nb_ports);
450 	}
451 	port = ports;
452 	for (pt_id = 0; pt_id < nb_ports; pt_id++, port++) {
453 		rte_eth_dev_info_get(pt_id, &port->dev_info);
454 		if (nb_rxq > port->dev_info.max_rx_queues) {
455 			rte_exit(EXIT_FAILURE, "Port %d: max RX queues %d < nb_rxq %d\n",
456 			       (int) pt_id,
457 			       (int) port->dev_info.max_rx_queues,
458 			       (int) nb_rxq);
459 		}
460 		if (nb_txq > port->dev_info.max_tx_queues) {
461 			rte_exit(EXIT_FAILURE, "Port %d: max TX queues %d < nb_txq %d\n",
462 			       (int) pt_id,
463 			       (int) port->dev_info.max_tx_queues,
464 			       (int) nb_txq);
465 		}
466 
467 		if (numa_support)
468 			port->socket_id = (pt_id < (nb_ports >> 1)) ? 0 : 1;
469 		else
470 			port->socket_id = 0;
471 	}
472 
473 	/* Configuration of packet forwarding streams. */
474 	nb_fwd_streams = (streamid_t) (nb_ports * nb_rxq);
475 	fwd_streams = rte_zmalloc("testpmd: fwd_streams",
476 				  sizeof(struct fwd_stream *) * nb_fwd_streams,
477 				  CACHE_LINE_SIZE);
478 	if (fwd_streams == NULL) {
479 		rte_exit(EXIT_FAILURE, "rte_zmalloc(%d (struct fwd_stream *)) failed\n",
480 		       nb_fwd_streams);
481 	}
482 	for (sm_id = 0; sm_id < nb_fwd_streams; sm_id++) {
483 		fwd_streams[sm_id] = rte_zmalloc("testpmd: struct fwd_stream",
484 						 sizeof(struct fwd_stream),
485 						 CACHE_LINE_SIZE);
486 		if (fwd_streams[sm_id] == NULL) {
487 			rte_exit(EXIT_FAILURE, "rte_zmalloc(struct fwd_stream) failed\n");
488 		}
489 	}
490 }
491 
492 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
493 static void
494 pkt_burst_stats_display(const char *rx_tx, struct pkt_burst_stats *pbs)
495 {
496 	unsigned int total_burst;
497 	unsigned int nb_burst;
498 	unsigned int burst_stats[3];
499 	uint16_t pktnb_stats[3];
500 	uint16_t nb_pkt;
501 	int burst_percent[3];
502 
503 	/*
504 	 * First compute the total number of packet bursts and the
505 	 * two highest numbers of bursts of the same number of packets.
506 	 */
507 	total_burst = 0;
508 	burst_stats[0] = burst_stats[1] = burst_stats[2] = 0;
509 	pktnb_stats[0] = pktnb_stats[1] = pktnb_stats[2] = 0;
510 	for (nb_pkt = 0; nb_pkt < MAX_PKT_BURST; nb_pkt++) {
511 		nb_burst = pbs->pkt_burst_spread[nb_pkt];
512 		if (nb_burst == 0)
513 			continue;
514 		total_burst += nb_burst;
515 		if (nb_burst > burst_stats[0]) {
516 			burst_stats[1] = burst_stats[0];
517 			pktnb_stats[1] = pktnb_stats[0];
518 			burst_stats[0] = nb_burst;
519 			pktnb_stats[0] = nb_pkt;
520 		}
521 	}
522 	if (total_burst == 0)
523 		return;
524 	burst_percent[0] = (burst_stats[0] * 100) / total_burst;
525 	printf("  %s-bursts : %u [%d%% of %d pkts", rx_tx, total_burst,
526 	       burst_percent[0], (int) pktnb_stats[0]);
527 	if (burst_stats[0] == total_burst) {
528 		printf("]\n");
529 		return;
530 	}
531 	if (burst_stats[0] + burst_stats[1] == total_burst) {
532 		printf(" + %d%% of %d pkts]\n",
533 		       100 - burst_percent[0], pktnb_stats[1]);
534 		return;
535 	}
536 	burst_percent[1] = (burst_stats[1] * 100) / total_burst;
537 	burst_percent[2] = 100 - (burst_percent[0] + burst_percent[1]);
538 	if ((burst_percent[1] == 0) || (burst_percent[2] == 0)) {
539 		printf(" + %d%% of others]\n", 100 - burst_percent[0]);
540 		return;
541 	}
542 	printf(" + %d%% of %d pkts + %d%% of others]\n",
543 	       burst_percent[1], (int) pktnb_stats[1], burst_percent[2]);
544 }
545 #endif /* RTE_TEST_PMD_RECORD_BURST_STATS */
546 
547 static void
548 fwd_port_stats_display(portid_t port_id, struct rte_eth_stats *stats)
549 {
550 	struct rte_port *port;
551 
552 	static const char *fwd_stats_border = "----------------------";
553 
554 	port = &ports[port_id];
555 	printf("\n  %s Forward statistics for port %-2d %s\n",
556 		fwd_stats_border, port_id, fwd_stats_border);
557 	printf("  RX-packets: %-14"PRIu64" RX-dropped: %-14"PRIu64"RX-total: "
558 	       "%-"PRIu64"\n",
559 	       stats->ipackets, stats->ierrors,
560 	       (uint64_t) (stats->ipackets + stats->ierrors));
561 
562 	if (cur_fwd_eng == &csum_fwd_engine)
563 		printf("  Bad-ipcsum: %-14"PRIu64" Bad-l4csum: %-14"PRIu64" \n",
564 				port->rx_bad_ip_csum, port->rx_bad_l4_csum);
565 
566 	printf("  TX-packets: %-14"PRIu64" TX-dropped: %-14"PRIu64"TX-total: "
567 	       "%-"PRIu64"\n",
568 	       stats->opackets, port->tx_dropped,
569 	       (uint64_t) (stats->opackets + port->tx_dropped));
570 
571 	if (stats->rx_nombuf > 0)
572 		printf("  RX-nombufs: %-14"PRIu64"\n", stats->rx_nombuf);
573 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
574 	if (port->rx_stream)
575 		pkt_burst_stats_display("RX", &port->rx_stream->rx_burst_stats);
576 	if (port->tx_stream)
577 		pkt_burst_stats_display("TX", &port->tx_stream->tx_burst_stats);
578 #endif
579 	/* stats fdir */
580 	if (fdir_conf.mode != RTE_FDIR_MODE_NONE)
581 		printf("  Fdirmiss: %-14"PRIu64"   Fdirmatch: %-14"PRIu64"\n",
582 		       stats->fdirmiss,
583 		       stats->fdirmatch);
584 
585 	printf("  %s--------------------------------%s\n",
586 	       fwd_stats_border, fwd_stats_border);
587 }
588 
589 static void
590 fwd_stream_stats_display(streamid_t stream_id)
591 {
592 	struct fwd_stream *fs;
593 	static const char *fwd_top_stats_border = "-------";
594 
595 	fs = fwd_streams[stream_id];
596 	if ((fs->rx_packets == 0) && (fs->tx_packets == 0) &&
597 	    (fs->fwd_dropped == 0))
598 		return;
599 	printf("\n  %s Forward Stats for RX Port=%2d/Queue=%2d -> "
600 	       "TX Port=%2d/Queue=%2d %s\n",
601 	       fwd_top_stats_border, fs->rx_port, fs->rx_queue,
602 	       fs->tx_port, fs->tx_queue, fwd_top_stats_border);
603 	printf("  RX-packets: %-14u TX-packets: %-14u TX-dropped: %-14u",
604 	       fs->rx_packets, fs->tx_packets, fs->fwd_dropped);
605 
606 	/* if checksum mode */
607 	if (cur_fwd_eng == &csum_fwd_engine) {
608 	       printf("  RX- bad IP checksum: %-14u  Rx- bad L4 checksum: %-14u\n",
609 	       fs->rx_bad_ip_csum, fs->rx_bad_l4_csum);
610 	}
611 
612 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
613 	pkt_burst_stats_display("RX", &fs->rx_burst_stats);
614 	pkt_burst_stats_display("TX", &fs->tx_burst_stats);
615 #endif
616 }
617 
618 static void
619 flush_all_rx_queues(void)
620 {
621 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
622 	portid_t  rxp;
623 	queueid_t rxq;
624 	uint16_t  nb_rx;
625 	uint16_t  i;
626 	uint8_t   j;
627 
628 	for (j = 0; j < 2; j++) {
629 		for (rxp = 0; rxp < nb_ports; rxp++) {
630 			for (rxq = 0; rxq < nb_rxq; rxq++) {
631 				do {
632 					nb_rx = rte_eth_rx_burst(rxp, rxq,
633 								 pkts_burst,
634 								 MAX_PKT_BURST);
635 					for (i = 0; i < nb_rx; i++)
636 						rte_pktmbuf_free(pkts_burst[i]);
637 				} while (nb_rx > 0);
638 			}
639 		}
640 		rte_delay_ms(10); /* wait 10 milli-seconds before retrying */
641 	}
642 }
643 
644 static void
645 run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
646 {
647 	struct fwd_stream **fsm;
648 	streamid_t nb_fs;
649 	streamid_t sm_id;
650 
651 	fsm = &fwd_streams[fc->stream_idx];
652 	nb_fs = fc->stream_nb;
653 	do {
654 		for (sm_id = 0; sm_id < nb_fs; sm_id++)
655 			(*pkt_fwd)(fsm[sm_id]);
656 	} while (! fc->stopped);
657 }
658 
659 static int
660 start_pkt_forward_on_core(void *fwd_arg)
661 {
662 	run_pkt_fwd_on_lcore((struct fwd_lcore *) fwd_arg,
663 			     cur_fwd_config.fwd_eng->packet_fwd);
664 	return 0;
665 }
666 
667 /*
668  * Run the TXONLY packet forwarding engine to send a single burst of packets.
669  * Used to start communication flows in network loopback test configurations.
670  */
671 static int
672 run_one_txonly_burst_on_core(void *fwd_arg)
673 {
674 	struct fwd_lcore *fwd_lc;
675 	struct fwd_lcore tmp_lcore;
676 
677 	fwd_lc = (struct fwd_lcore *) fwd_arg;
678 	tmp_lcore = *fwd_lc;
679 	tmp_lcore.stopped = 1;
680 	run_pkt_fwd_on_lcore(&tmp_lcore, tx_only_engine.packet_fwd);
681 	return 0;
682 }
683 
684 /*
685  * Launch packet forwarding:
686  *     - Setup per-port forwarding context.
687  *     - launch logical cores with their forwarding configuration.
688  */
689 static void
690 launch_packet_forwarding(lcore_function_t *pkt_fwd_on_lcore)
691 {
692 	port_fwd_begin_t port_fwd_begin;
693 	unsigned int i;
694 	unsigned int lc_id;
695 	int diag;
696 
697 	port_fwd_begin = cur_fwd_config.fwd_eng->port_fwd_begin;
698 	if (port_fwd_begin != NULL) {
699 		for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
700 			(*port_fwd_begin)(fwd_ports_ids[i]);
701 	}
702 	for (i = 0; i < cur_fwd_config.nb_fwd_lcores; i++) {
703 		lc_id = fwd_lcores_cpuids[i];
704 		if ((interactive == 0) || (lc_id != rte_lcore_id())) {
705 			fwd_lcores[i]->stopped = 0;
706 			diag = rte_eal_remote_launch(pkt_fwd_on_lcore,
707 						     fwd_lcores[i], lc_id);
708 			if (diag != 0)
709 				printf("launch lcore %u failed - diag=%d\n",
710 				       lc_id, diag);
711 		}
712 	}
713 }
714 
715 /*
716  * Launch packet forwarding configuration.
717  */
718 void
719 start_packet_forwarding(int with_tx_first)
720 {
721 	port_fwd_begin_t port_fwd_begin;
722 	port_fwd_end_t  port_fwd_end;
723 	struct rte_port *port;
724 	unsigned int i;
725 	portid_t   pt_id;
726 	streamid_t sm_id;
727 
728 	if (test_done == 0) {
729 		printf("Packet forwarding already started\n");
730 		return;
731 	}
732 	test_done = 0;
733 	flush_all_rx_queues();
734 	fwd_config_setup();
735 	rxtx_config_display();
736 
737 	for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++) {
738 		pt_id = fwd_ports_ids[i];
739 		port = &ports[pt_id];
740 		rte_eth_stats_get(pt_id, &port->stats);
741 		port->tx_dropped = 0;
742 	}
743 	for (sm_id = 0; sm_id < cur_fwd_config.nb_fwd_streams; sm_id++) {
744 		fwd_streams[sm_id]->rx_packets = 0;
745 		fwd_streams[sm_id]->tx_packets = 0;
746 		fwd_streams[sm_id]->fwd_dropped = 0;
747 		fwd_streams[sm_id]->rx_bad_ip_csum = 0;
748 		fwd_streams[sm_id]->rx_bad_l4_csum = 0;
749 
750 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
751 		memset(&fwd_streams[sm_id]->rx_burst_stats, 0,
752 		       sizeof(fwd_streams[sm_id]->rx_burst_stats));
753 		memset(&fwd_streams[sm_id]->tx_burst_stats, 0,
754 		       sizeof(fwd_streams[sm_id]->tx_burst_stats));
755 #endif
756 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
757 		fwd_streams[sm_id]->core_cycles = 0;
758 #endif
759 	}
760 	if (with_tx_first) {
761 		port_fwd_begin = tx_only_engine.port_fwd_begin;
762 		if (port_fwd_begin != NULL) {
763 			for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
764 				(*port_fwd_begin)(fwd_ports_ids[i]);
765 		}
766 		launch_packet_forwarding(run_one_txonly_burst_on_core);
767 		rte_eal_mp_wait_lcore();
768 		port_fwd_end = tx_only_engine.port_fwd_end;
769 		if (port_fwd_end != NULL) {
770 			for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
771 				(*port_fwd_end)(fwd_ports_ids[i]);
772 		}
773 	}
774 	launch_packet_forwarding(start_pkt_forward_on_core);
775 }
776 
777 void
778 stop_packet_forwarding(void)
779 {
780 	struct rte_eth_stats stats;
781 	struct rte_port *port;
782 	port_fwd_end_t  port_fwd_end;
783 	int i;
784 	portid_t   pt_id;
785 	streamid_t sm_id;
786 	lcoreid_t  lc_id;
787 	uint64_t total_recv;
788 	uint64_t total_xmit;
789 	uint64_t total_rx_dropped;
790 	uint64_t total_tx_dropped;
791 	uint64_t total_rx_nombuf;
792 	uint64_t tx_dropped;
793 	uint64_t rx_bad_ip_csum;
794 	uint64_t rx_bad_l4_csum;
795 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
796 	uint64_t fwd_cycles;
797 #endif
798 	static const char *acc_stats_border = "+++++++++++++++";
799 
800 	if (test_done) {
801 		printf("Packet forwarding not started\n");
802 		return;
803 	}
804 	printf("Telling cores to stop...");
805 	for (lc_id = 0; lc_id < cur_fwd_config.nb_fwd_lcores; lc_id++)
806 		fwd_lcores[lc_id]->stopped = 1;
807 	printf("\nWaiting for lcores to finish...\n");
808 	rte_eal_mp_wait_lcore();
809 	port_fwd_end = cur_fwd_config.fwd_eng->port_fwd_end;
810 	if (port_fwd_end != NULL) {
811 		for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++) {
812 			pt_id = fwd_ports_ids[i];
813 			(*port_fwd_end)(pt_id);
814 		}
815 	}
816 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
817 	fwd_cycles = 0;
818 #endif
819 	for (sm_id = 0; sm_id < cur_fwd_config.nb_fwd_streams; sm_id++) {
820 		if (cur_fwd_config.nb_fwd_streams >
821 		    cur_fwd_config.nb_fwd_ports) {
822 			fwd_stream_stats_display(sm_id);
823 			ports[fwd_streams[sm_id]->tx_port].tx_stream = NULL;
824 			ports[fwd_streams[sm_id]->rx_port].rx_stream = NULL;
825 		} else {
826 			ports[fwd_streams[sm_id]->tx_port].tx_stream =
827 				fwd_streams[sm_id];
828 			ports[fwd_streams[sm_id]->rx_port].rx_stream =
829 				fwd_streams[sm_id];
830 		}
831 		tx_dropped = ports[fwd_streams[sm_id]->tx_port].tx_dropped;
832 		tx_dropped = (uint64_t) (tx_dropped +
833 					 fwd_streams[sm_id]->fwd_dropped);
834 		ports[fwd_streams[sm_id]->tx_port].tx_dropped = tx_dropped;
835 
836 		rx_bad_ip_csum = ports[fwd_streams[sm_id]->rx_port].rx_bad_ip_csum;
837 		rx_bad_ip_csum = (uint64_t) (rx_bad_ip_csum +
838 					 fwd_streams[sm_id]->rx_bad_ip_csum);
839 		ports[fwd_streams[sm_id]->rx_port].rx_bad_ip_csum = rx_bad_ip_csum;
840 
841 		rx_bad_l4_csum = ports[fwd_streams[sm_id]->rx_port].rx_bad_l4_csum;
842 		rx_bad_l4_csum = (uint64_t) (rx_bad_l4_csum +
843 					 fwd_streams[sm_id]->rx_bad_l4_csum);
844 		ports[fwd_streams[sm_id]->rx_port].rx_bad_l4_csum = rx_bad_l4_csum;
845 
846 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
847 		fwd_cycles = (uint64_t) (fwd_cycles +
848 					 fwd_streams[sm_id]->core_cycles);
849 #endif
850 	}
851 	total_recv = 0;
852 	total_xmit = 0;
853 	total_rx_dropped = 0;
854 	total_tx_dropped = 0;
855 	total_rx_nombuf  = 0;
856 	for (i = 0; i < ((cur_fwd_config.nb_fwd_ports + 1) & ~0x1); i++) {
857 		pt_id = fwd_ports_ids[i];
858 
859 		port = &ports[pt_id];
860 		rte_eth_stats_get(pt_id, &stats);
861 		stats.ipackets -= port->stats.ipackets;
862 		port->stats.ipackets = 0;
863 		stats.opackets -= port->stats.opackets;
864 		port->stats.opackets = 0;
865 		stats.ibytes   -= port->stats.ibytes;
866 		port->stats.ibytes = 0;
867 		stats.obytes   -= port->stats.obytes;
868 		port->stats.obytes = 0;
869 		stats.ierrors  -= port->stats.ierrors;
870 		port->stats.ierrors = 0;
871 		stats.oerrors  -= port->stats.oerrors;
872 		port->stats.oerrors = 0;
873 		stats.rx_nombuf -= port->stats.rx_nombuf;
874 		port->stats.rx_nombuf = 0;
875 		stats.fdirmatch -= port->stats.fdirmatch;
876 		port->stats.rx_nombuf = 0;
877 		stats.fdirmiss -= port->stats.fdirmiss;
878 		port->stats.rx_nombuf = 0;
879 
880 		total_recv += stats.ipackets;
881 		total_xmit += stats.opackets;
882 		total_rx_dropped += stats.ierrors;
883 		total_tx_dropped += port->tx_dropped;
884 		total_rx_nombuf  += stats.rx_nombuf;
885 
886 		fwd_port_stats_display(pt_id, &stats);
887 	}
888 	printf("\n  %s Accumulated forward statistics for all ports"
889 	       "%s\n",
890 	       acc_stats_border, acc_stats_border);
891 	printf("  RX-packets: %-14"PRIu64" RX-dropped: %-14"PRIu64"RX-total: "
892 	       "%-"PRIu64"\n"
893 	       "  TX-packets: %-14"PRIu64" TX-dropped: %-14"PRIu64"TX-total: "
894 	       "%-"PRIu64"\n",
895 	       total_recv, total_rx_dropped, total_recv + total_rx_dropped,
896 	       total_xmit, total_tx_dropped, total_xmit + total_tx_dropped);
897 	if (total_rx_nombuf > 0)
898 		printf("  RX-nombufs: %-14"PRIu64"\n", total_rx_nombuf);
899 	printf("  %s++++++++++++++++++++++++++++++++++++++++++++++"
900 	       "%s\n",
901 	       acc_stats_border, acc_stats_border);
902 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
903 	if (total_recv > 0)
904 		printf("\n  CPU cycles/packet=%u (total cycles="
905 		       "%"PRIu64" / total RX packets=%"PRIu64")\n",
906 		       (unsigned int)(fwd_cycles / total_recv),
907 		       fwd_cycles, total_recv);
908 #endif
909 	printf("\nDone.\n");
910 	test_done = 1;
911 }
912 
913 void
914 pmd_test_exit(void)
915 {
916 	portid_t pt_id;
917 
918 	for (pt_id = 0; pt_id < nb_ports; pt_id++) {
919 		printf("Stopping port %d...", pt_id);
920 		fflush(stdout);
921 		rte_eth_dev_close(pt_id);
922 		printf("done\n");
923 	}
924 	printf("bye...\n");
925 }
926 
927 typedef void (*cmd_func_t)(void);
928 struct pmd_test_command {
929 	const char *cmd_name;
930 	cmd_func_t cmd_func;
931 };
932 
933 #define PMD_TEST_CMD_NB (sizeof(pmd_test_menu) / sizeof(pmd_test_menu[0]))
934 
935 static void
936 fatal_init_error(const char *func_name, uint8_t port_id, int diag)
937 {
938 	rte_panic("%s(port_id=%d) failed - diag=%d\n",
939 		  func_name, port_id, diag);
940 }
941 
942 static void
943 init_ports(void)
944 {
945 	struct rte_eth_link   link;
946 	struct rte_eth_conf   port_conf = {
947 		.intr_conf = {
948 			.lsc = 0,
949 		},
950 	};
951 	struct rte_eth_rxconf rx_conf;
952 	struct rte_eth_txconf tx_conf;
953 	struct rte_port *port;
954 	unsigned int sock_id;
955 	portid_t  pi;
956 	queueid_t qi;
957 	int diag;
958 
959 	port_conf.rxmode = rx_mode;
960 	port_conf.fdir_conf = fdir_conf;
961 
962 	if (nb_rxq > 0) { /* configure RSS */
963 		port_conf.rx_adv_conf.rss_conf.rss_key = NULL;
964 		/* use default hash key */
965 		port_conf.rx_adv_conf.rss_conf.rss_hf = rss_hf;
966 	} else
967 		port_conf.rx_adv_conf.rss_conf.rss_hf = 0;
968 	rx_conf.rx_thresh = rx_thresh;
969 	rx_conf.rx_free_thresh = rx_free_thresh;
970 	tx_conf.tx_thresh = tx_thresh;
971 	tx_conf.tx_rs_thresh = tx_rs_thresh;
972 	tx_conf.tx_free_thresh = tx_free_thresh;
973 
974 	for (pi = 0; pi < nb_ports; pi++) {
975 		port = &ports[pi];
976 		memcpy(&port->dev_conf, &port_conf, sizeof(port_conf));
977 		sock_id = port->socket_id;
978 		printf("Initializing port %d... ", pi);
979 		fflush(stdout);
980 		diag = rte_eth_dev_configure(pi, nb_rxq, nb_txq, &port_conf);
981 		if (diag != 0) {
982 			fatal_init_error("rte_eth_dev_configure", pi, diag);
983 			/* NOT REACHED */
984 		}
985 		rte_eth_macaddr_get(pi, &port->eth_addr);
986 		for (qi = 0; qi < nb_txq; qi++) {
987 			diag = rte_eth_tx_queue_setup(pi, qi, nb_txd,
988 						      sock_id,
989 						      &tx_conf);
990 			if (diag != 0) {
991 				fatal_init_error("rte_eth_tx_queue_setup",
992 						 pi, diag);
993 				/* NOT REACHED */
994 			}
995 		}
996 		for (qi = 0; qi < nb_rxq; qi++) {
997 			diag = rte_eth_rx_queue_setup(pi, qi, nb_rxd, sock_id,
998 						      &rx_conf,
999 						      mbuf_pool_find(sock_id));
1000 			if (diag != 0) {
1001 				fatal_init_error("rte_eth_rx_queue_setup",
1002 						 pi , diag);
1003 				/* NOT REACHED */
1004 			}
1005 		}
1006 
1007 		/* Start device */
1008 		diag = rte_eth_dev_start(pi);
1009 		if (diag != 0) {
1010 			fatal_init_error("rte_eth_dev_start", pi, diag);
1011 			/* NOT REACHED */
1012 		}
1013 		printf("done: ");
1014 		rte_eth_link_get(pi, &link);
1015 		if (link.link_status) {
1016 			printf(" Link Up - speed %u Mbps - %s\n",
1017 			       (unsigned) link.link_speed,
1018 			       (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1019 			       ("full-duplex") : ("half-duplex\n"));
1020 		} else {
1021 			printf(" Link Down\n");
1022 		}
1023 
1024 		/*
1025 		 * If enabled, put device in promiscuous mode.
1026 		 * This allows the PMD test in IO forwarding mode to forward
1027 		 * packets to itself through 2 cross-connected  ports of the
1028 		 * target machine.
1029 		 */
1030 		if (promiscuous_on)
1031 			rte_eth_promiscuous_enable(pi);
1032 	}
1033 }
1034 
1035 #ifdef RTE_EXEC_ENV_BAREMETAL
1036 #define main _main
1037 #endif
1038 
1039 int
1040 main(int argc, char** argv)
1041 {
1042 	int  diag;
1043 
1044 	diag = rte_eal_init(argc, argv);
1045 	if (diag < 0)
1046 		rte_panic("Cannot init EAL\n");
1047 
1048 #ifdef RTE_LIBRTE_IGB_PMD
1049 	if (rte_igb_pmd_init())
1050 		rte_panic("Cannot init igb PMD\n");
1051 #endif
1052 #ifdef RTE_LIBRTE_IXGBE_PMD
1053 	if (rte_ixgbe_pmd_init())
1054 		rte_panic("Cannot init ixgbe PMD\n");
1055 
1056 	if (rte_ixgbevf_pmd_init())
1057 		rte_panic("Cannot init ixgbevf PMD\n");
1058 #endif
1059 
1060 	if (rte_eal_pci_probe())
1061 		rte_panic("Cannot probe PCI\n");
1062 
1063 	nb_ports = (portid_t) rte_eth_dev_count();
1064 	if (nb_ports == 0)
1065 		rte_exit(EXIT_FAILURE, "No probed ethernet devices - check that "
1066 			  "CONFIG_RTE_LIBRTE_IGB_PMD=y and that "
1067 			  "CONFIG_RTE_LIBRTE_IXGBE_PMD=y in your "
1068 			  "configuration file\n");
1069 
1070 	set_def_fwd_config();
1071 	if (nb_lcores == 0)
1072 		rte_panic("Empty set of forwarding logical cores - check the "
1073 			  "core mask supplied in the command parameters\n");
1074 
1075 	argc -= diag;
1076 	argv += diag;
1077 	if (argc > 1)
1078 		launch_args_parse(argc, argv);
1079 
1080 	if (nb_rxq > nb_txq)
1081 		printf("Warning: nb_rxq=%d enables RSS configuration, "
1082 		       "but nb_txq=%d will prevent to fully test it.\n",
1083 		       nb_rxq, nb_txq);
1084 
1085 	init_config();
1086 
1087 	init_ports();
1088 
1089 	if (interactive == 1)
1090 		prompt();
1091 	else {
1092 		char c;
1093 		int rc;
1094 
1095 		printf("No commandline core given, start packet forwarding\n");
1096 		start_packet_forwarding(0);
1097 		printf("Press enter to exit\n");
1098 		rc = read(0, &c, 1);
1099 		if (rc < 0)
1100 			return 1;
1101 	}
1102 
1103 	return 0;
1104 }
1105