examples: fix port mask parsing failure handling
[dpdk.git] / examples / distributor / main.c
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4
5 #include <stdint.h>
6 #include <inttypes.h>
7 #include <unistd.h>
8 #include <signal.h>
9 #include <getopt.h>
10
11 #include <rte_eal.h>
12 #include <rte_ethdev.h>
13 #include <rte_cycles.h>
14 #include <rte_malloc.h>
15 #include <rte_debug.h>
16 #include <rte_prefetch.h>
17 #include <rte_distributor.h>
18 #include <rte_pause.h>
19 #include <rte_power.h>
20
21 #define RX_RING_SIZE 1024
22 #define TX_RING_SIZE 1024
23 #define NUM_MBUFS ((64*1024)-1)
24 #define MBUF_CACHE_SIZE 128
25 #define BURST_SIZE 64
26 #define SCHED_RX_RING_SZ 8192
27 #define SCHED_TX_RING_SZ 65536
28 #define BURST_SIZE_TX 32
29
30 #define RTE_LOGTYPE_DISTRAPP RTE_LOGTYPE_USER1
31
32 #define ANSI_COLOR_RED     "\x1b[31m"
33 #define ANSI_COLOR_RESET   "\x1b[0m"
34
35 /* mask of enabled ports */
36 static uint32_t enabled_port_mask;
37 volatile uint8_t quit_signal;
38 volatile uint8_t quit_signal_rx;
39 volatile uint8_t quit_signal_dist;
40 volatile uint8_t quit_signal_work;
41 unsigned int power_lib_initialised;
42
43 static volatile struct app_stats {
44         struct {
45                 uint64_t rx_pkts;
46                 uint64_t returned_pkts;
47                 uint64_t enqueued_pkts;
48                 uint64_t enqdrop_pkts;
49         } rx __rte_cache_aligned;
50         int pad1 __rte_cache_aligned;
51
52         struct {
53                 uint64_t in_pkts;
54                 uint64_t ret_pkts;
55                 uint64_t sent_pkts;
56                 uint64_t enqdrop_pkts;
57         } dist __rte_cache_aligned;
58         int pad2 __rte_cache_aligned;
59
60         struct {
61                 uint64_t dequeue_pkts;
62                 uint64_t tx_pkts;
63                 uint64_t enqdrop_pkts;
64         } tx __rte_cache_aligned;
65         int pad3 __rte_cache_aligned;
66
67         uint64_t worker_pkts[64] __rte_cache_aligned;
68
69         int pad4 __rte_cache_aligned;
70
71         uint64_t worker_bursts[64][8] __rte_cache_aligned;
72
73         int pad5 __rte_cache_aligned;
74
75         uint64_t port_rx_pkts[64] __rte_cache_aligned;
76         uint64_t port_tx_pkts[64] __rte_cache_aligned;
77 } app_stats;
78
79 struct app_stats prev_app_stats;
80
81 static const struct rte_eth_conf port_conf_default = {
82         .rxmode = {
83                 .mq_mode = ETH_MQ_RX_RSS,
84                 .max_rx_pkt_len = RTE_ETHER_MAX_LEN,
85         },
86         .txmode = {
87                 .mq_mode = ETH_MQ_TX_NONE,
88         },
89         .rx_adv_conf = {
90                 .rss_conf = {
91                         .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
92                                 ETH_RSS_TCP | ETH_RSS_SCTP,
93                 }
94         },
95 };
96
97 struct output_buffer {
98         unsigned count;
99         struct rte_mbuf *mbufs[BURST_SIZE];
100 };
101
102 static void print_stats(void);
103
104 /*
105  * Initialises a given port using global settings and with the rx buffers
106  * coming from the mbuf_pool passed as parameter
107  */
108 static inline int
109 port_init(uint16_t port, struct rte_mempool *mbuf_pool)
110 {
111         struct rte_eth_conf port_conf = port_conf_default;
112         const uint16_t rxRings = 1, txRings = rte_lcore_count() - 1;
113         int retval;
114         uint16_t q;
115         uint16_t nb_rxd = RX_RING_SIZE;
116         uint16_t nb_txd = TX_RING_SIZE;
117         struct rte_eth_dev_info dev_info;
118         struct rte_eth_txconf txconf;
119
120         if (!rte_eth_dev_is_valid_port(port))
121                 return -1;
122
123         retval = rte_eth_dev_info_get(port, &dev_info);
124         if (retval != 0) {
125                 printf("Error during getting device (port %u) info: %s\n",
126                                 port, strerror(-retval));
127                 return retval;
128         }
129
130         if (dev_info.tx_offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE)
131                 port_conf.txmode.offloads |=
132                         DEV_TX_OFFLOAD_MBUF_FAST_FREE;
133
134         port_conf.rx_adv_conf.rss_conf.rss_hf &=
135                 dev_info.flow_type_rss_offloads;
136         if (port_conf.rx_adv_conf.rss_conf.rss_hf !=
137                         port_conf_default.rx_adv_conf.rss_conf.rss_hf) {
138                 printf("Port %u modified RSS hash function based on hardware support,"
139                         "requested:%#"PRIx64" configured:%#"PRIx64"\n",
140                         port,
141                         port_conf_default.rx_adv_conf.rss_conf.rss_hf,
142                         port_conf.rx_adv_conf.rss_conf.rss_hf);
143         }
144
145         retval = rte_eth_dev_configure(port, rxRings, txRings, &port_conf);
146         if (retval != 0)
147                 return retval;
148
149         retval = rte_eth_dev_adjust_nb_rx_tx_desc(port, &nb_rxd, &nb_txd);
150         if (retval != 0)
151                 return retval;
152
153         for (q = 0; q < rxRings; q++) {
154                 retval = rte_eth_rx_queue_setup(port, q, nb_rxd,
155                                                 rte_eth_dev_socket_id(port),
156                                                 NULL, mbuf_pool);
157                 if (retval < 0)
158                         return retval;
159         }
160
161         txconf = dev_info.default_txconf;
162         txconf.offloads = port_conf.txmode.offloads;
163         for (q = 0; q < txRings; q++) {
164                 retval = rte_eth_tx_queue_setup(port, q, nb_txd,
165                                                 rte_eth_dev_socket_id(port),
166                                                 &txconf);
167                 if (retval < 0)
168                         return retval;
169         }
170
171         retval = rte_eth_dev_start(port);
172         if (retval < 0)
173                 return retval;
174
175         struct rte_eth_link link;
176         do {
177                 retval = rte_eth_link_get_nowait(port, &link);
178                 if (retval < 0) {
179                         printf("Failed link get (port %u): %s\n",
180                                 port, rte_strerror(-retval));
181                         return retval;
182                 } else if (link.link_status)
183                         break;
184
185                 printf("Waiting for Link up on port %"PRIu16"\n", port);
186                 sleep(1);
187         } while (!link.link_status);
188
189         if (!link.link_status) {
190                 printf("Link down on port %"PRIu16"\n", port);
191                 return 0;
192         }
193
194         struct rte_ether_addr addr;
195         retval = rte_eth_macaddr_get(port, &addr);
196         if (retval < 0) {
197                 printf("Failed to get MAC address (port %u): %s\n",
198                                 port, rte_strerror(-retval));
199                 return retval;
200         }
201
202         printf("Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
203                         " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
204                         port,
205                         addr.addr_bytes[0], addr.addr_bytes[1],
206                         addr.addr_bytes[2], addr.addr_bytes[3],
207                         addr.addr_bytes[4], addr.addr_bytes[5]);
208
209         retval = rte_eth_promiscuous_enable(port);
210         if (retval != 0)
211                 return retval;
212
213         return 0;
214 }
215
216 struct lcore_params {
217         unsigned worker_id;
218         struct rte_distributor *d;
219         struct rte_ring *rx_dist_ring;
220         struct rte_ring *dist_tx_ring;
221         struct rte_mempool *mem_pool;
222 };
223
224 static int
225 lcore_rx(struct lcore_params *p)
226 {
227         const uint16_t nb_ports = rte_eth_dev_count_avail();
228         const int socket_id = rte_socket_id();
229         uint16_t port;
230         struct rte_mbuf *bufs[BURST_SIZE*2];
231
232         RTE_ETH_FOREACH_DEV(port) {
233                 /* skip ports that are not enabled */
234                 if ((enabled_port_mask & (1 << port)) == 0)
235                         continue;
236
237                 if (rte_eth_dev_socket_id(port) > 0 &&
238                                 rte_eth_dev_socket_id(port) != socket_id)
239                         printf("WARNING, port %u is on remote NUMA node to "
240                                         "RX thread.\n\tPerformance will not "
241                                         "be optimal.\n", port);
242         }
243
244         printf("\nCore %u doing packet RX.\n", rte_lcore_id());
245         port = 0;
246         while (!quit_signal_rx) {
247
248                 /* skip ports that are not enabled */
249                 if ((enabled_port_mask & (1 << port)) == 0) {
250                         if (++port == nb_ports)
251                                 port = 0;
252                         continue;
253                 }
254                 const uint16_t nb_rx = rte_eth_rx_burst(port, 0, bufs,
255                                 BURST_SIZE);
256                 if (unlikely(nb_rx == 0)) {
257                         if (++port == nb_ports)
258                                 port = 0;
259                         continue;
260                 }
261                 app_stats.rx.rx_pkts += nb_rx;
262
263 /*
264  * You can run the distributor on the rx core with this code. Returned
265  * packets are then send straight to the tx core.
266  */
267 #if 0
268         rte_distributor_process(d, bufs, nb_rx);
269         const uint16_t nb_ret = rte_distributor_returned_pktsd,
270                         bufs, BURST_SIZE*2);
271
272                 app_stats.rx.returned_pkts += nb_ret;
273                 if (unlikely(nb_ret == 0)) {
274                         if (++port == nb_ports)
275                                 port = 0;
276                         continue;
277                 }
278
279                 struct rte_ring *tx_ring = p->dist_tx_ring;
280                 uint16_t sent = rte_ring_enqueue_burst(tx_ring,
281                                 (void *)bufs, nb_ret, NULL);
282 #else
283                 uint16_t nb_ret = nb_rx;
284                 /*
285                  * Swap the following two lines if you want the rx traffic
286                  * to go directly to tx, no distribution.
287                  */
288                 struct rte_ring *out_ring = p->rx_dist_ring;
289                 /* struct rte_ring *out_ring = p->dist_tx_ring; */
290
291                 uint16_t sent = rte_ring_enqueue_burst(out_ring,
292                                 (void *)bufs, nb_ret, NULL);
293 #endif
294
295                 app_stats.rx.enqueued_pkts += sent;
296                 if (unlikely(sent < nb_ret)) {
297                         app_stats.rx.enqdrop_pkts +=  nb_ret - sent;
298                         RTE_LOG_DP(DEBUG, DISTRAPP,
299                                 "%s:Packet loss due to full ring\n", __func__);
300                         while (sent < nb_ret)
301                                 rte_pktmbuf_free(bufs[sent++]);
302                 }
303                 if (++port == nb_ports)
304                         port = 0;
305         }
306         if (power_lib_initialised)
307                 rte_power_exit(rte_lcore_id());
308         /* set worker & tx threads quit flag */
309         printf("\nCore %u exiting rx task.\n", rte_lcore_id());
310         quit_signal = 1;
311         return 0;
312 }
313
314 static inline void
315 flush_one_port(struct output_buffer *outbuf, uint8_t outp)
316 {
317         unsigned int nb_tx = rte_eth_tx_burst(outp, 0,
318                         outbuf->mbufs, outbuf->count);
319         app_stats.tx.tx_pkts += outbuf->count;
320
321         if (unlikely(nb_tx < outbuf->count)) {
322                 app_stats.tx.enqdrop_pkts +=  outbuf->count - nb_tx;
323                 do {
324                         rte_pktmbuf_free(outbuf->mbufs[nb_tx]);
325                 } while (++nb_tx < outbuf->count);
326         }
327         outbuf->count = 0;
328 }
329
330 static inline void
331 flush_all_ports(struct output_buffer *tx_buffers)
332 {
333         uint16_t outp;
334
335         RTE_ETH_FOREACH_DEV(outp) {
336                 /* skip ports that are not enabled */
337                 if ((enabled_port_mask & (1 << outp)) == 0)
338                         continue;
339
340                 if (tx_buffers[outp].count == 0)
341                         continue;
342
343                 flush_one_port(&tx_buffers[outp], outp);
344         }
345 }
346
347
348
349 static int
350 lcore_distributor(struct lcore_params *p)
351 {
352         struct rte_ring *in_r = p->rx_dist_ring;
353         struct rte_ring *out_r = p->dist_tx_ring;
354         struct rte_mbuf *bufs[BURST_SIZE * 4];
355         struct rte_distributor *d = p->d;
356
357         printf("\nCore %u acting as distributor core.\n", rte_lcore_id());
358         while (!quit_signal_dist) {
359                 const uint16_t nb_rx = rte_ring_dequeue_burst(in_r,
360                                 (void *)bufs, BURST_SIZE*1, NULL);
361                 if (nb_rx) {
362                         app_stats.dist.in_pkts += nb_rx;
363
364                         /* Distribute the packets */
365                         rte_distributor_process(d, bufs, nb_rx);
366                         /* Handle Returns */
367                         const uint16_t nb_ret =
368                                 rte_distributor_returned_pkts(d,
369                                         bufs, BURST_SIZE*2);
370
371                         if (unlikely(nb_ret == 0))
372                                 continue;
373                         app_stats.dist.ret_pkts += nb_ret;
374
375                         uint16_t sent = rte_ring_enqueue_burst(out_r,
376                                         (void *)bufs, nb_ret, NULL);
377                         app_stats.dist.sent_pkts += sent;
378                         if (unlikely(sent < nb_ret)) {
379                                 app_stats.dist.enqdrop_pkts += nb_ret - sent;
380                                 RTE_LOG(DEBUG, DISTRAPP,
381                                         "%s:Packet loss due to full out ring\n",
382                                         __func__);
383                                 while (sent < nb_ret)
384                                         rte_pktmbuf_free(bufs[sent++]);
385                         }
386                 }
387         }
388         printf("\nCore %u exiting distributor task.\n", rte_lcore_id());
389         quit_signal_work = 1;
390         if (power_lib_initialised)
391                 rte_power_exit(rte_lcore_id());
392         rte_distributor_flush(d);
393         /* Unblock any returns so workers can exit */
394         rte_distributor_clear_returns(d);
395         quit_signal_rx = 1;
396         return 0;
397 }
398
399
400 static int
401 lcore_tx(struct rte_ring *in_r)
402 {
403         static struct output_buffer tx_buffers[RTE_MAX_ETHPORTS];
404         const int socket_id = rte_socket_id();
405         uint16_t port;
406
407         RTE_ETH_FOREACH_DEV(port) {
408                 /* skip ports that are not enabled */
409                 if ((enabled_port_mask & (1 << port)) == 0)
410                         continue;
411
412                 if (rte_eth_dev_socket_id(port) > 0 &&
413                                 rte_eth_dev_socket_id(port) != socket_id)
414                         printf("WARNING, port %u is on remote NUMA node to "
415                                         "TX thread.\n\tPerformance will not "
416                                         "be optimal.\n", port);
417         }
418
419         printf("\nCore %u doing packet TX.\n", rte_lcore_id());
420         while (!quit_signal) {
421
422                 RTE_ETH_FOREACH_DEV(port) {
423                         /* skip ports that are not enabled */
424                         if ((enabled_port_mask & (1 << port)) == 0)
425                                 continue;
426
427                         struct rte_mbuf *bufs[BURST_SIZE_TX];
428                         const uint16_t nb_rx = rte_ring_dequeue_burst(in_r,
429                                         (void *)bufs, BURST_SIZE_TX, NULL);
430                         app_stats.tx.dequeue_pkts += nb_rx;
431
432                         /* if we get no traffic, flush anything we have */
433                         if (unlikely(nb_rx == 0)) {
434                                 flush_all_ports(tx_buffers);
435                                 continue;
436                         }
437
438                         /* for traffic we receive, queue it up for transmit */
439                         uint16_t i;
440                         rte_prefetch_non_temporal((void *)bufs[0]);
441                         rte_prefetch_non_temporal((void *)bufs[1]);
442                         rte_prefetch_non_temporal((void *)bufs[2]);
443                         for (i = 0; i < nb_rx; i++) {
444                                 struct output_buffer *outbuf;
445                                 uint8_t outp;
446                                 rte_prefetch_non_temporal((void *)bufs[i + 3]);
447                                 /*
448                                  * workers should update in_port to hold the
449                                  * output port value
450                                  */
451                                 outp = bufs[i]->port;
452                                 /* skip ports that are not enabled */
453                                 if ((enabled_port_mask & (1 << outp)) == 0)
454                                         continue;
455
456                                 outbuf = &tx_buffers[outp];
457                                 outbuf->mbufs[outbuf->count++] = bufs[i];
458                                 if (outbuf->count == BURST_SIZE_TX)
459                                         flush_one_port(outbuf, outp);
460                         }
461                 }
462         }
463         if (power_lib_initialised)
464                 rte_power_exit(rte_lcore_id());
465         printf("\nCore %u exiting tx task.\n", rte_lcore_id());
466         return 0;
467 }
468
469 static void
470 int_handler(int sig_num)
471 {
472         printf("Exiting on signal %d\n", sig_num);
473         /* set quit flag for rx thread to exit */
474         quit_signal_dist = 1;
475 }
476
477 static void
478 print_stats(void)
479 {
480         struct rte_eth_stats eth_stats;
481         unsigned int i, j;
482         const unsigned int num_workers = rte_lcore_count() - 4;
483
484         RTE_ETH_FOREACH_DEV(i) {
485                 rte_eth_stats_get(i, &eth_stats);
486                 app_stats.port_rx_pkts[i] = eth_stats.ipackets;
487                 app_stats.port_tx_pkts[i] = eth_stats.opackets;
488         }
489
490         printf("\n\nRX Thread:\n");
491         RTE_ETH_FOREACH_DEV(i) {
492                 printf("Port %u Pktsin : %5.2f\n", i,
493                                 (app_stats.port_rx_pkts[i] -
494                                 prev_app_stats.port_rx_pkts[i])/1000000.0);
495                 prev_app_stats.port_rx_pkts[i] = app_stats.port_rx_pkts[i];
496         }
497         printf(" - Received:    %5.2f\n",
498                         (app_stats.rx.rx_pkts -
499                         prev_app_stats.rx.rx_pkts)/1000000.0);
500         printf(" - Returned:    %5.2f\n",
501                         (app_stats.rx.returned_pkts -
502                         prev_app_stats.rx.returned_pkts)/1000000.0);
503         printf(" - Enqueued:    %5.2f\n",
504                         (app_stats.rx.enqueued_pkts -
505                         prev_app_stats.rx.enqueued_pkts)/1000000.0);
506         printf(" - Dropped:     %s%5.2f%s\n", ANSI_COLOR_RED,
507                         (app_stats.rx.enqdrop_pkts -
508                         prev_app_stats.rx.enqdrop_pkts)/1000000.0,
509                         ANSI_COLOR_RESET);
510
511         printf("Distributor thread:\n");
512         printf(" - In:          %5.2f\n",
513                         (app_stats.dist.in_pkts -
514                         prev_app_stats.dist.in_pkts)/1000000.0);
515         printf(" - Returned:    %5.2f\n",
516                         (app_stats.dist.ret_pkts -
517                         prev_app_stats.dist.ret_pkts)/1000000.0);
518         printf(" - Sent:        %5.2f\n",
519                         (app_stats.dist.sent_pkts -
520                         prev_app_stats.dist.sent_pkts)/1000000.0);
521         printf(" - Dropped      %s%5.2f%s\n", ANSI_COLOR_RED,
522                         (app_stats.dist.enqdrop_pkts -
523                         prev_app_stats.dist.enqdrop_pkts)/1000000.0,
524                         ANSI_COLOR_RESET);
525
526         printf("TX thread:\n");
527         printf(" - Dequeued:    %5.2f\n",
528                         (app_stats.tx.dequeue_pkts -
529                         prev_app_stats.tx.dequeue_pkts)/1000000.0);
530         RTE_ETH_FOREACH_DEV(i) {
531                 printf("Port %u Pktsout: %5.2f\n",
532                                 i, (app_stats.port_tx_pkts[i] -
533                                 prev_app_stats.port_tx_pkts[i])/1000000.0);
534                 prev_app_stats.port_tx_pkts[i] = app_stats.port_tx_pkts[i];
535         }
536         printf(" - Transmitted: %5.2f\n",
537                         (app_stats.tx.tx_pkts -
538                         prev_app_stats.tx.tx_pkts)/1000000.0);
539         printf(" - Dropped:     %s%5.2f%s\n", ANSI_COLOR_RED,
540                         (app_stats.tx.enqdrop_pkts -
541                         prev_app_stats.tx.enqdrop_pkts)/1000000.0,
542                         ANSI_COLOR_RESET);
543
544         prev_app_stats.rx.rx_pkts = app_stats.rx.rx_pkts;
545         prev_app_stats.rx.returned_pkts = app_stats.rx.returned_pkts;
546         prev_app_stats.rx.enqueued_pkts = app_stats.rx.enqueued_pkts;
547         prev_app_stats.rx.enqdrop_pkts = app_stats.rx.enqdrop_pkts;
548         prev_app_stats.dist.in_pkts = app_stats.dist.in_pkts;
549         prev_app_stats.dist.ret_pkts = app_stats.dist.ret_pkts;
550         prev_app_stats.dist.sent_pkts = app_stats.dist.sent_pkts;
551         prev_app_stats.dist.enqdrop_pkts = app_stats.dist.enqdrop_pkts;
552         prev_app_stats.tx.dequeue_pkts = app_stats.tx.dequeue_pkts;
553         prev_app_stats.tx.tx_pkts = app_stats.tx.tx_pkts;
554         prev_app_stats.tx.enqdrop_pkts = app_stats.tx.enqdrop_pkts;
555
556         for (i = 0; i < num_workers; i++) {
557                 printf("Worker %02u Pkts: %5.2f. Bursts(1-8): ", i,
558                                 (app_stats.worker_pkts[i] -
559                                 prev_app_stats.worker_pkts[i])/1000000.0);
560                 for (j = 0; j < 8; j++) {
561                         printf("%"PRIu64" ", app_stats.worker_bursts[i][j]);
562                         app_stats.worker_bursts[i][j] = 0;
563                 }
564                 printf("\n");
565                 prev_app_stats.worker_pkts[i] = app_stats.worker_pkts[i];
566         }
567 }
568
569 static int
570 lcore_worker(struct lcore_params *p)
571 {
572         struct rte_distributor *d = p->d;
573         const unsigned id = p->worker_id;
574         unsigned int num = 0;
575         unsigned int i;
576
577         /*
578          * for single port, xor_val will be zero so we won't modify the output
579          * port, otherwise we send traffic from 0 to 1, 2 to 3, and vice versa
580          */
581         const unsigned xor_val = (rte_eth_dev_count_avail() > 1);
582         struct rte_mbuf *buf[8] __rte_cache_aligned;
583
584         for (i = 0; i < 8; i++)
585                 buf[i] = NULL;
586
587         app_stats.worker_pkts[p->worker_id] = 1;
588
589         printf("\nCore %u acting as worker core.\n", rte_lcore_id());
590         while (!quit_signal_work) {
591                 num = rte_distributor_get_pkt(d, id, buf, buf, num);
592                 /* Do a little bit of work for each packet */
593                 for (i = 0; i < num; i++) {
594                         uint64_t t = rte_rdtsc()+100;
595
596                         while (rte_rdtsc() < t)
597                                 rte_pause();
598                         buf[i]->port ^= xor_val;
599                 }
600
601                 app_stats.worker_pkts[p->worker_id] += num;
602                 if (num > 0)
603                         app_stats.worker_bursts[p->worker_id][num-1]++;
604         }
605         if (power_lib_initialised)
606                 rte_power_exit(rte_lcore_id());
607         rte_free(p);
608         return 0;
609 }
610
611 static int
612 init_power_library(void)
613 {
614         int ret = 0, lcore_id;
615         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
616                 /* init power management library */
617                 ret = rte_power_init(lcore_id);
618                 if (ret) {
619                         RTE_LOG(ERR, POWER,
620                                 "Library initialization failed on core %u\n",
621                                 lcore_id);
622                         /*
623                          * Return on first failure, we'll fall back
624                          * to non-power operation
625                          */
626                         return ret;
627                 }
628         }
629         return ret;
630 }
631
632 /* display usage */
633 static void
634 print_usage(const char *prgname)
635 {
636         printf("%s [EAL options] -- -p PORTMASK\n"
637                         "  -p PORTMASK: hexadecimal bitmask of ports to configure\n",
638                         prgname);
639 }
640
641 static int
642 parse_portmask(const char *portmask)
643 {
644         char *end = NULL;
645         unsigned long pm;
646
647         /* parse hexadecimal string */
648         pm = strtoul(portmask, &end, 16);
649         if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
650                 return 0;
651
652         return pm;
653 }
654
655 /* Parse the argument given in the command line of the application */
656 static int
657 parse_args(int argc, char **argv)
658 {
659         int opt;
660         char **argvopt;
661         int option_index;
662         char *prgname = argv[0];
663         static struct option lgopts[] = {
664                 {NULL, 0, 0, 0}
665         };
666
667         argvopt = argv;
668
669         while ((opt = getopt_long(argc, argvopt, "p:",
670                         lgopts, &option_index)) != EOF) {
671
672                 switch (opt) {
673                 /* portmask */
674                 case 'p':
675                         enabled_port_mask = parse_portmask(optarg);
676                         if (enabled_port_mask == 0) {
677                                 printf("invalid portmask\n");
678                                 print_usage(prgname);
679                                 return -1;
680                         }
681                         break;
682
683                 default:
684                         print_usage(prgname);
685                         return -1;
686                 }
687         }
688
689         if (optind <= 1) {
690                 print_usage(prgname);
691                 return -1;
692         }
693
694         argv[optind-1] = prgname;
695
696         optind = 1; /* reset getopt lib */
697         return 0;
698 }
699
700 /* Main function, does initialization and calls the per-lcore functions */
701 int
702 main(int argc, char *argv[])
703 {
704         struct rte_mempool *mbuf_pool;
705         struct rte_distributor *d;
706         struct rte_ring *dist_tx_ring;
707         struct rte_ring *rx_dist_ring;
708         struct rte_power_core_capabilities lcore_cap;
709         unsigned int lcore_id, worker_id = 0;
710         int distr_core_id = -1, rx_core_id = -1, tx_core_id = -1;
711         unsigned nb_ports;
712         uint16_t portid;
713         uint16_t nb_ports_available;
714         uint64_t t, freq;
715
716         /* catch ctrl-c so we can print on exit */
717         signal(SIGINT, int_handler);
718
719         /* init EAL */
720         int ret = rte_eal_init(argc, argv);
721         if (ret < 0)
722                 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
723         argc -= ret;
724         argv += ret;
725
726         /* parse application arguments (after the EAL ones) */
727         ret = parse_args(argc, argv);
728         if (ret < 0)
729                 rte_exit(EXIT_FAILURE, "Invalid distributor parameters\n");
730
731         if (rte_lcore_count() < 5)
732                 rte_exit(EXIT_FAILURE, "Error, This application needs at "
733                                 "least 5 logical cores to run:\n"
734                                 "1 lcore for stats (can be core 0)\n"
735                                 "1 lcore for packet RX\n"
736                                 "1 lcore for distribution\n"
737                                 "1 lcore for packet TX\n"
738                                 "and at least 1 lcore for worker threads\n");
739
740         if (init_power_library() == 0)
741                 power_lib_initialised = 1;
742
743         nb_ports = rte_eth_dev_count_avail();
744         if (nb_ports == 0)
745                 rte_exit(EXIT_FAILURE, "Error: no ethernet ports detected\n");
746         if (nb_ports != 1 && (nb_ports & 1))
747                 rte_exit(EXIT_FAILURE, "Error: number of ports must be even, except "
748                                 "when using a single port\n");
749
750         mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL",
751                 NUM_MBUFS * nb_ports, MBUF_CACHE_SIZE, 0,
752                 RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
753         if (mbuf_pool == NULL)
754                 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
755         nb_ports_available = nb_ports;
756
757         /* initialize all ports */
758         RTE_ETH_FOREACH_DEV(portid) {
759                 /* skip ports that are not enabled */
760                 if ((enabled_port_mask & (1 << portid)) == 0) {
761                         printf("\nSkipping disabled port %d\n", portid);
762                         nb_ports_available--;
763                         continue;
764                 }
765                 /* init port */
766                 printf("Initializing port %u... done\n", portid);
767
768                 if (port_init(portid, mbuf_pool) != 0)
769                         rte_exit(EXIT_FAILURE, "Cannot initialize port %u\n",
770                                         portid);
771         }
772
773         if (!nb_ports_available) {
774                 rte_exit(EXIT_FAILURE,
775                                 "All available ports are disabled. Please set portmask.\n");
776         }
777
778         d = rte_distributor_create("PKT_DIST", rte_socket_id(),
779                         rte_lcore_count() - 4,
780                         RTE_DIST_ALG_BURST);
781         if (d == NULL)
782                 rte_exit(EXIT_FAILURE, "Cannot create distributor\n");
783
784         /*
785          * scheduler ring is read by the transmitter core, and written to
786          * by scheduler core
787          */
788         dist_tx_ring = rte_ring_create("Output_ring", SCHED_TX_RING_SZ,
789                         rte_socket_id(), RING_F_SC_DEQ | RING_F_SP_ENQ);
790         if (dist_tx_ring == NULL)
791                 rte_exit(EXIT_FAILURE, "Cannot create output ring\n");
792
793         rx_dist_ring = rte_ring_create("Input_ring", SCHED_RX_RING_SZ,
794                         rte_socket_id(), RING_F_SC_DEQ | RING_F_SP_ENQ);
795         if (rx_dist_ring == NULL)
796                 rte_exit(EXIT_FAILURE, "Cannot create output ring\n");
797
798         if (power_lib_initialised) {
799                 /*
800                  * Here we'll pre-assign lcore ids to the rx, tx and
801                  * distributor workloads if there's higher frequency
802                  * on those cores e.g. if Turbo Boost is enabled.
803                  * It's also worth mentioning that it will assign cores in a
804                  * specific order, so that if there's less than three
805                  * available, the higher frequency cores will go to the
806                  * distributor first, then rx, then tx.
807                  */
808                 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
809
810                         rte_power_get_capabilities(lcore_id, &lcore_cap);
811
812                         if (lcore_cap.priority != 1)
813                                 continue;
814
815                         if (distr_core_id < 0) {
816                                 distr_core_id = lcore_id;
817                                 printf("Distributor on priority core %d\n",
818                                         lcore_id);
819                                 continue;
820                         }
821                         if (rx_core_id < 0) {
822                                 rx_core_id = lcore_id;
823                                 printf("Rx on priority core %d\n",
824                                         lcore_id);
825                                 continue;
826                         }
827                         if (tx_core_id < 0) {
828                                 tx_core_id = lcore_id;
829                                 printf("Tx on priority core %d\n",
830                                         lcore_id);
831                                 continue;
832                         }
833                 }
834         }
835
836         /*
837          * If there's any of the key workloads left without an lcore_id
838          * after the high performing core assignment above, pre-assign
839          * them here.
840          */
841         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
842                 if (lcore_id == (unsigned int)distr_core_id ||
843                                 lcore_id == (unsigned int)rx_core_id ||
844                                 lcore_id == (unsigned int)tx_core_id)
845                         continue;
846                 if (distr_core_id < 0) {
847                         distr_core_id = lcore_id;
848                         printf("Distributor on core %d\n", lcore_id);
849                         continue;
850                 }
851                 if (rx_core_id < 0) {
852                         rx_core_id = lcore_id;
853                         printf("Rx on core %d\n", lcore_id);
854                         continue;
855                 }
856                 if (tx_core_id < 0) {
857                         tx_core_id = lcore_id;
858                         printf("Tx on core %d\n", lcore_id);
859                         continue;
860                 }
861         }
862
863         printf(" tx id %d, dist id %d, rx id %d\n",
864                         tx_core_id,
865                         distr_core_id,
866                         rx_core_id);
867
868         /*
869          * Kick off all the worker threads first, avoiding the pre-assigned
870          * lcore_ids for tx, rx and distributor workloads.
871          */
872         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
873                 if (lcore_id == (unsigned int)distr_core_id ||
874                                 lcore_id == (unsigned int)rx_core_id ||
875                                 lcore_id == (unsigned int)tx_core_id)
876                         continue;
877                 printf("Starting thread %d as worker, lcore_id %d\n",
878                                 worker_id, lcore_id);
879                 struct lcore_params *p =
880                         rte_malloc(NULL, sizeof(*p), 0);
881                 if (!p)
882                         rte_panic("malloc failure\n");
883                 *p = (struct lcore_params){worker_id++, d, rx_dist_ring,
884                         dist_tx_ring, mbuf_pool};
885
886                 rte_eal_remote_launch((lcore_function_t *)lcore_worker,
887                                 p, lcore_id);
888         }
889
890         /* Start tx core */
891         rte_eal_remote_launch((lcore_function_t *)lcore_tx,
892                         dist_tx_ring, tx_core_id);
893
894         /* Start distributor core */
895         struct lcore_params *pd =
896                 rte_malloc(NULL, sizeof(*pd), 0);
897         if (!pd)
898                 rte_panic("malloc failure\n");
899         *pd = (struct lcore_params){worker_id++, d,
900                 rx_dist_ring, dist_tx_ring, mbuf_pool};
901         rte_eal_remote_launch(
902                         (lcore_function_t *)lcore_distributor,
903                         pd, distr_core_id);
904
905         /* Start rx core */
906         struct lcore_params *pr =
907                 rte_malloc(NULL, sizeof(*pr), 0);
908         if (!pr)
909                 rte_panic("malloc failure\n");
910         *pr = (struct lcore_params){worker_id++, d, rx_dist_ring,
911                 dist_tx_ring, mbuf_pool};
912         rte_eal_remote_launch((lcore_function_t *)lcore_rx,
913                         pr, rx_core_id);
914
915         freq = rte_get_timer_hz();
916         t = rte_rdtsc() + freq;
917         while (!quit_signal_dist) {
918                 if (t < rte_rdtsc()) {
919                         print_stats();
920                         t = rte_rdtsc() + freq;
921                 }
922                 usleep(1000);
923         }
924
925         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
926                 if (rte_eal_wait_lcore(lcore_id) < 0)
927                         return -1;
928         }
929
930         print_stats();
931
932         rte_free(pd);
933         rte_free(pr);
934
935         return 0;
936 }