xref: /dpdk/app/test-pmd/testpmd.c (revision 14ad4f01845331a0ae98c681efa3086eeed3343a)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4 
5 #include <stdarg.h>
6 #include <stdio.h>
7 #include <stdlib.h>
8 #include <signal.h>
9 #include <string.h>
10 #include <time.h>
11 #include <fcntl.h>
12 #include <sys/mman.h>
13 #include <sys/types.h>
14 #include <errno.h>
15 #include <stdbool.h>
16 
17 #include <sys/queue.h>
18 #include <sys/stat.h>
19 
20 #include <stdint.h>
21 #include <unistd.h>
22 #include <inttypes.h>
23 
24 #include <rte_common.h>
25 #include <rte_errno.h>
26 #include <rte_byteorder.h>
27 #include <rte_log.h>
28 #include <rte_debug.h>
29 #include <rte_cycles.h>
30 #include <rte_malloc_heap.h>
31 #include <rte_memory.h>
32 #include <rte_memcpy.h>
33 #include <rte_launch.h>
34 #include <rte_eal.h>
35 #include <rte_alarm.h>
36 #include <rte_per_lcore.h>
37 #include <rte_lcore.h>
38 #include <rte_atomic.h>
39 #include <rte_branch_prediction.h>
40 #include <rte_mempool.h>
41 #include <rte_malloc.h>
42 #include <rte_mbuf.h>
43 #include <rte_mbuf_pool_ops.h>
44 #include <rte_interrupts.h>
45 #include <rte_pci.h>
46 #include <rte_ether.h>
47 #include <rte_ethdev.h>
48 #include <rte_dev.h>
49 #include <rte_string_fns.h>
50 #ifdef RTE_LIBRTE_IXGBE_PMD
51 #include <rte_pmd_ixgbe.h>
52 #endif
53 #ifdef RTE_LIBRTE_PDUMP
54 #include <rte_pdump.h>
55 #endif
56 #include <rte_flow.h>
57 #include <rte_metrics.h>
58 #ifdef RTE_LIBRTE_BITRATE
59 #include <rte_bitrate.h>
60 #endif
61 #ifdef RTE_LIBRTE_LATENCY_STATS
62 #include <rte_latencystats.h>
63 #endif
64 
65 #include "testpmd.h"
66 
67 #ifndef MAP_HUGETLB
68 /* FreeBSD may not have MAP_HUGETLB (in fact, it probably doesn't) */
69 #define HUGE_FLAG (0x40000)
70 #else
71 #define HUGE_FLAG MAP_HUGETLB
72 #endif
73 
74 #ifndef MAP_HUGE_SHIFT
75 /* older kernels (or FreeBSD) will not have this define */
76 #define HUGE_SHIFT (26)
77 #else
78 #define HUGE_SHIFT MAP_HUGE_SHIFT
79 #endif
80 
81 #define EXTMEM_HEAP_NAME "extmem"
82 
83 uint16_t verbose_level = 0; /**< Silent by default. */
84 int testpmd_logtype; /**< Log type for testpmd logs */
85 
86 /* use master core for command line ? */
87 uint8_t interactive = 0;
88 uint8_t auto_start = 0;
89 uint8_t tx_first;
90 char cmdline_filename[PATH_MAX] = {0};
91 
92 /*
93  * NUMA support configuration.
94  * When set, the NUMA support attempts to dispatch the allocation of the
95  * RX and TX memory rings, and of the DMA memory buffers (mbufs) for the
96  * probed ports among the CPU sockets 0 and 1.
97  * Otherwise, all memory is allocated from CPU socket 0.
98  */
99 uint8_t numa_support = 1; /**< numa enabled by default */
100 
101 /*
102  * In UMA mode,all memory is allocated from socket 0 if --socket-num is
103  * not configured.
104  */
105 uint8_t socket_num = UMA_NO_CONFIG;
106 
107 /*
108  * Select mempool allocation type:
109  * - native: use regular DPDK memory
110  * - anon: use regular DPDK memory to create mempool, but populate using
111  *         anonymous memory (may not be IOVA-contiguous)
112  * - xmem: use externally allocated hugepage memory
113  */
114 uint8_t mp_alloc_type = MP_ALLOC_NATIVE;
115 
116 /*
117  * Store specified sockets on which memory pool to be used by ports
118  * is allocated.
119  */
120 uint8_t port_numa[RTE_MAX_ETHPORTS];
121 
122 /*
123  * Store specified sockets on which RX ring to be used by ports
124  * is allocated.
125  */
126 uint8_t rxring_numa[RTE_MAX_ETHPORTS];
127 
128 /*
129  * Store specified sockets on which TX ring to be used by ports
130  * is allocated.
131  */
132 uint8_t txring_numa[RTE_MAX_ETHPORTS];
133 
134 /*
135  * Record the Ethernet address of peer target ports to which packets are
136  * forwarded.
137  * Must be instantiated with the ethernet addresses of peer traffic generator
138  * ports.
139  */
140 struct rte_ether_addr peer_eth_addrs[RTE_MAX_ETHPORTS];
141 portid_t nb_peer_eth_addrs = 0;
142 
143 /*
144  * Probed Target Environment.
145  */
146 struct rte_port *ports;	       /**< For all probed ethernet ports. */
147 portid_t nb_ports;             /**< Number of probed ethernet ports. */
148 struct fwd_lcore **fwd_lcores; /**< For all probed logical cores. */
149 lcoreid_t nb_lcores;           /**< Number of probed logical cores. */
150 
151 portid_t ports_ids[RTE_MAX_ETHPORTS]; /**< Store all port ids. */
152 
153 /*
154  * Test Forwarding Configuration.
155  *    nb_fwd_lcores <= nb_cfg_lcores <= nb_lcores
156  *    nb_fwd_ports  <= nb_cfg_ports  <= nb_ports
157  */
158 lcoreid_t nb_cfg_lcores; /**< Number of configured logical cores. */
159 lcoreid_t nb_fwd_lcores; /**< Number of forwarding logical cores. */
160 portid_t  nb_cfg_ports;  /**< Number of configured ports. */
161 portid_t  nb_fwd_ports;  /**< Number of forwarding ports. */
162 
163 unsigned int fwd_lcores_cpuids[RTE_MAX_LCORE]; /**< CPU ids configuration. */
164 portid_t fwd_ports_ids[RTE_MAX_ETHPORTS];      /**< Port ids configuration. */
165 
166 struct fwd_stream **fwd_streams; /**< For each RX queue of each port. */
167 streamid_t nb_fwd_streams;       /**< Is equal to (nb_ports * nb_rxq). */
168 
169 /*
170  * Forwarding engines.
171  */
172 struct fwd_engine * fwd_engines[] = {
173 	&io_fwd_engine,
174 	&mac_fwd_engine,
175 	&mac_swap_engine,
176 	&flow_gen_engine,
177 	&rx_only_engine,
178 	&tx_only_engine,
179 	&csum_fwd_engine,
180 	&icmp_echo_engine,
181 	&noisy_vnf_engine,
182 #if defined RTE_LIBRTE_PMD_SOFTNIC
183 	&softnic_fwd_engine,
184 #endif
185 #ifdef RTE_LIBRTE_IEEE1588
186 	&ieee1588_fwd_engine,
187 #endif
188 	NULL,
189 };
190 
191 struct rte_mempool *mempools[RTE_MAX_NUMA_NODES];
192 uint16_t mempool_flags;
193 
194 struct fwd_config cur_fwd_config;
195 struct fwd_engine *cur_fwd_eng = &io_fwd_engine; /**< IO mode by default. */
196 uint32_t retry_enabled;
197 uint32_t burst_tx_delay_time = BURST_TX_WAIT_US;
198 uint32_t burst_tx_retry_num = BURST_TX_RETRIES;
199 
200 uint16_t mbuf_data_size = DEFAULT_MBUF_DATA_SIZE; /**< Mbuf data space size. */
201 uint32_t param_total_num_mbufs = 0;  /**< number of mbufs in all pools - if
202                                       * specified on command-line. */
203 uint16_t stats_period; /**< Period to show statistics (disabled by default) */
204 
205 /*
206  * In container, it cannot terminate the process which running with 'stats-period'
207  * option. Set flag to exit stats period loop after received SIGINT/SIGTERM.
208  */
209 uint8_t f_quit;
210 
211 /*
212  * Configuration of packet segments used by the "txonly" processing engine.
213  */
214 uint16_t tx_pkt_length = TXONLY_DEF_PACKET_LEN; /**< TXONLY packet length. */
215 uint16_t tx_pkt_seg_lengths[RTE_MAX_SEGS_PER_PKT] = {
216 	TXONLY_DEF_PACKET_LEN,
217 };
218 uint8_t  tx_pkt_nb_segs = 1; /**< Number of segments in TXONLY packets */
219 
220 enum tx_pkt_split tx_pkt_split = TX_PKT_SPLIT_OFF;
221 /**< Split policy for packets to TX. */
222 
223 uint8_t txonly_multi_flow;
224 /**< Whether multiple flows are generated in TXONLY mode. */
225 
226 uint16_t nb_pkt_per_burst = DEF_PKT_BURST; /**< Number of packets per burst. */
227 uint16_t mb_mempool_cache = DEF_MBUF_CACHE; /**< Size of mbuf mempool cache. */
228 
229 /* current configuration is in DCB or not,0 means it is not in DCB mode */
230 uint8_t dcb_config = 0;
231 
232 /* Whether the dcb is in testing status */
233 uint8_t dcb_test = 0;
234 
235 /*
236  * Configurable number of RX/TX queues.
237  */
238 queueid_t nb_rxq = 1; /**< Number of RX queues per port. */
239 queueid_t nb_txq = 1; /**< Number of TX queues per port. */
240 
241 /*
242  * Configurable number of RX/TX ring descriptors.
243  * Defaults are supplied by drivers via ethdev.
244  */
245 #define RTE_TEST_RX_DESC_DEFAULT 0
246 #define RTE_TEST_TX_DESC_DEFAULT 0
247 uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; /**< Number of RX descriptors. */
248 uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; /**< Number of TX descriptors. */
249 
250 #define RTE_PMD_PARAM_UNSET -1
251 /*
252  * Configurable values of RX and TX ring threshold registers.
253  */
254 
255 int8_t rx_pthresh = RTE_PMD_PARAM_UNSET;
256 int8_t rx_hthresh = RTE_PMD_PARAM_UNSET;
257 int8_t rx_wthresh = RTE_PMD_PARAM_UNSET;
258 
259 int8_t tx_pthresh = RTE_PMD_PARAM_UNSET;
260 int8_t tx_hthresh = RTE_PMD_PARAM_UNSET;
261 int8_t tx_wthresh = RTE_PMD_PARAM_UNSET;
262 
263 /*
264  * Configurable value of RX free threshold.
265  */
266 int16_t rx_free_thresh = RTE_PMD_PARAM_UNSET;
267 
268 /*
269  * Configurable value of RX drop enable.
270  */
271 int8_t rx_drop_en = RTE_PMD_PARAM_UNSET;
272 
273 /*
274  * Configurable value of TX free threshold.
275  */
276 int16_t tx_free_thresh = RTE_PMD_PARAM_UNSET;
277 
278 /*
279  * Configurable value of TX RS bit threshold.
280  */
281 int16_t tx_rs_thresh = RTE_PMD_PARAM_UNSET;
282 
283 /*
284  * Configurable value of buffered packets before sending.
285  */
286 uint16_t noisy_tx_sw_bufsz;
287 
288 /*
289  * Configurable value of packet buffer timeout.
290  */
291 uint16_t noisy_tx_sw_buf_flush_time;
292 
293 /*
294  * Configurable value for size of VNF internal memory area
295  * used for simulating noisy neighbour behaviour
296  */
297 uint64_t noisy_lkup_mem_sz;
298 
299 /*
300  * Configurable value of number of random writes done in
301  * VNF simulation memory area.
302  */
303 uint64_t noisy_lkup_num_writes;
304 
305 /*
306  * Configurable value of number of random reads done in
307  * VNF simulation memory area.
308  */
309 uint64_t noisy_lkup_num_reads;
310 
311 /*
312  * Configurable value of number of random reads/writes done in
313  * VNF simulation memory area.
314  */
315 uint64_t noisy_lkup_num_reads_writes;
316 
317 /*
318  * Receive Side Scaling (RSS) configuration.
319  */
320 uint64_t rss_hf = ETH_RSS_IP; /* RSS IP by default. */
321 
322 /*
323  * Port topology configuration
324  */
325 uint16_t port_topology = PORT_TOPOLOGY_PAIRED; /* Ports are paired by default */
326 
327 /*
328  * Avoids to flush all the RX streams before starts forwarding.
329  */
330 uint8_t no_flush_rx = 0; /* flush by default */
331 
332 /*
333  * Flow API isolated mode.
334  */
335 uint8_t flow_isolate_all;
336 
337 /*
338  * Avoids to check link status when starting/stopping a port.
339  */
340 uint8_t no_link_check = 0; /* check by default */
341 
342 /*
343  * Enable link status change notification
344  */
345 uint8_t lsc_interrupt = 1; /* enabled by default */
346 
347 /*
348  * Enable device removal notification.
349  */
350 uint8_t rmv_interrupt = 1; /* enabled by default */
351 
352 uint8_t hot_plug = 0; /**< hotplug disabled by default. */
353 
354 /* After attach, port setup is called on event or by iterator */
355 bool setup_on_probe_event = true;
356 
357 /* Pretty printing of ethdev events */
358 static const char * const eth_event_desc[] = {
359 	[RTE_ETH_EVENT_UNKNOWN] = "unknown",
360 	[RTE_ETH_EVENT_INTR_LSC] = "link state change",
361 	[RTE_ETH_EVENT_QUEUE_STATE] = "queue state",
362 	[RTE_ETH_EVENT_INTR_RESET] = "reset",
363 	[RTE_ETH_EVENT_VF_MBOX] = "VF mbox",
364 	[RTE_ETH_EVENT_IPSEC] = "IPsec",
365 	[RTE_ETH_EVENT_MACSEC] = "MACsec",
366 	[RTE_ETH_EVENT_INTR_RMV] = "device removal",
367 	[RTE_ETH_EVENT_NEW] = "device probed",
368 	[RTE_ETH_EVENT_DESTROY] = "device released",
369 	[RTE_ETH_EVENT_MAX] = NULL,
370 };
371 
372 /*
373  * Display or mask ether events
374  * Default to all events except VF_MBOX
375  */
376 uint32_t event_print_mask = (UINT32_C(1) << RTE_ETH_EVENT_UNKNOWN) |
377 			    (UINT32_C(1) << RTE_ETH_EVENT_INTR_LSC) |
378 			    (UINT32_C(1) << RTE_ETH_EVENT_QUEUE_STATE) |
379 			    (UINT32_C(1) << RTE_ETH_EVENT_INTR_RESET) |
380 			    (UINT32_C(1) << RTE_ETH_EVENT_IPSEC) |
381 			    (UINT32_C(1) << RTE_ETH_EVENT_MACSEC) |
382 			    (UINT32_C(1) << RTE_ETH_EVENT_INTR_RMV);
383 /*
384  * Decide if all memory are locked for performance.
385  */
386 int do_mlockall = 0;
387 
388 /*
389  * NIC bypass mode configuration options.
390  */
391 
392 #if defined RTE_LIBRTE_IXGBE_PMD && defined RTE_LIBRTE_IXGBE_BYPASS
393 /* The NIC bypass watchdog timeout. */
394 uint32_t bypass_timeout = RTE_PMD_IXGBE_BYPASS_TMT_OFF;
395 #endif
396 
397 
398 #ifdef RTE_LIBRTE_LATENCY_STATS
399 
400 /*
401  * Set when latency stats is enabled in the commandline
402  */
403 uint8_t latencystats_enabled;
404 
405 /*
406  * Lcore ID to serive latency statistics.
407  */
408 lcoreid_t latencystats_lcore_id = -1;
409 
410 #endif
411 
412 /*
413  * Ethernet device configuration.
414  */
415 struct rte_eth_rxmode rx_mode = {
416 	.max_rx_pkt_len = RTE_ETHER_MAX_LEN,
417 		/**< Default maximum frame length. */
418 };
419 
420 struct rte_eth_txmode tx_mode = {
421 	.offloads = DEV_TX_OFFLOAD_MBUF_FAST_FREE,
422 };
423 
424 struct rte_fdir_conf fdir_conf = {
425 	.mode = RTE_FDIR_MODE_NONE,
426 	.pballoc = RTE_FDIR_PBALLOC_64K,
427 	.status = RTE_FDIR_REPORT_STATUS,
428 	.mask = {
429 		.vlan_tci_mask = 0xFFEF,
430 		.ipv4_mask     = {
431 			.src_ip = 0xFFFFFFFF,
432 			.dst_ip = 0xFFFFFFFF,
433 		},
434 		.ipv6_mask     = {
435 			.src_ip = {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
436 			.dst_ip = {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
437 		},
438 		.src_port_mask = 0xFFFF,
439 		.dst_port_mask = 0xFFFF,
440 		.mac_addr_byte_mask = 0xFF,
441 		.tunnel_type_mask = 1,
442 		.tunnel_id_mask = 0xFFFFFFFF,
443 	},
444 	.drop_queue = 127,
445 };
446 
447 volatile int test_done = 1; /* stop packet forwarding when set to 1. */
448 
449 struct queue_stats_mappings tx_queue_stats_mappings_array[MAX_TX_QUEUE_STATS_MAPPINGS];
450 struct queue_stats_mappings rx_queue_stats_mappings_array[MAX_RX_QUEUE_STATS_MAPPINGS];
451 
452 struct queue_stats_mappings *tx_queue_stats_mappings = tx_queue_stats_mappings_array;
453 struct queue_stats_mappings *rx_queue_stats_mappings = rx_queue_stats_mappings_array;
454 
455 uint16_t nb_tx_queue_stats_mappings = 0;
456 uint16_t nb_rx_queue_stats_mappings = 0;
457 
458 /*
459  * Display zero values by default for xstats
460  */
461 uint8_t xstats_hide_zero;
462 
463 unsigned int num_sockets = 0;
464 unsigned int socket_ids[RTE_MAX_NUMA_NODES];
465 
466 #ifdef RTE_LIBRTE_BITRATE
467 /* Bitrate statistics */
468 struct rte_stats_bitrates *bitrate_data;
469 lcoreid_t bitrate_lcore_id;
470 uint8_t bitrate_enabled;
471 #endif
472 
473 struct gro_status gro_ports[RTE_MAX_ETHPORTS];
474 uint8_t gro_flush_cycles = GRO_DEFAULT_FLUSH_CYCLES;
475 
476 struct vxlan_encap_conf vxlan_encap_conf = {
477 	.select_ipv4 = 1,
478 	.select_vlan = 0,
479 	.select_tos_ttl = 0,
480 	.vni = "\x00\x00\x00",
481 	.udp_src = 0,
482 	.udp_dst = RTE_BE16(4789),
483 	.ipv4_src = RTE_IPV4(127, 0, 0, 1),
484 	.ipv4_dst = RTE_IPV4(255, 255, 255, 255),
485 	.ipv6_src = "\x00\x00\x00\x00\x00\x00\x00\x00"
486 		"\x00\x00\x00\x00\x00\x00\x00\x01",
487 	.ipv6_dst = "\x00\x00\x00\x00\x00\x00\x00\x00"
488 		"\x00\x00\x00\x00\x00\x00\x11\x11",
489 	.vlan_tci = 0,
490 	.ip_tos = 0,
491 	.ip_ttl = 255,
492 	.eth_src = "\x00\x00\x00\x00\x00\x00",
493 	.eth_dst = "\xff\xff\xff\xff\xff\xff",
494 };
495 
496 struct nvgre_encap_conf nvgre_encap_conf = {
497 	.select_ipv4 = 1,
498 	.select_vlan = 0,
499 	.tni = "\x00\x00\x00",
500 	.ipv4_src = RTE_IPV4(127, 0, 0, 1),
501 	.ipv4_dst = RTE_IPV4(255, 255, 255, 255),
502 	.ipv6_src = "\x00\x00\x00\x00\x00\x00\x00\x00"
503 		"\x00\x00\x00\x00\x00\x00\x00\x01",
504 	.ipv6_dst = "\x00\x00\x00\x00\x00\x00\x00\x00"
505 		"\x00\x00\x00\x00\x00\x00\x11\x11",
506 	.vlan_tci = 0,
507 	.eth_src = "\x00\x00\x00\x00\x00\x00",
508 	.eth_dst = "\xff\xff\xff\xff\xff\xff",
509 };
510 
511 /* Forward function declarations */
512 static void setup_attached_port(portid_t pi);
513 static void map_port_queue_stats_mapping_registers(portid_t pi,
514 						   struct rte_port *port);
515 static void check_all_ports_link_status(uint32_t port_mask);
516 static int eth_event_callback(portid_t port_id,
517 			      enum rte_eth_event_type type,
518 			      void *param, void *ret_param);
519 static void dev_event_callback(const char *device_name,
520 				enum rte_dev_event_type type,
521 				void *param);
522 
523 /*
524  * Check if all the ports are started.
525  * If yes, return positive value. If not, return zero.
526  */
527 static int all_ports_started(void);
528 
529 struct gso_status gso_ports[RTE_MAX_ETHPORTS];
530 uint16_t gso_max_segment_size = RTE_ETHER_MAX_LEN - RTE_ETHER_CRC_LEN;
531 
532 /*
533  * Helper function to check if socket is already discovered.
534  * If yes, return positive value. If not, return zero.
535  */
536 int
537 new_socket_id(unsigned int socket_id)
538 {
539 	unsigned int i;
540 
541 	for (i = 0; i < num_sockets; i++) {
542 		if (socket_ids[i] == socket_id)
543 			return 0;
544 	}
545 	return 1;
546 }
547 
548 /*
549  * Setup default configuration.
550  */
551 static void
552 set_default_fwd_lcores_config(void)
553 {
554 	unsigned int i;
555 	unsigned int nb_lc;
556 	unsigned int sock_num;
557 
558 	nb_lc = 0;
559 	for (i = 0; i < RTE_MAX_LCORE; i++) {
560 		if (!rte_lcore_is_enabled(i))
561 			continue;
562 		sock_num = rte_lcore_to_socket_id(i);
563 		if (new_socket_id(sock_num)) {
564 			if (num_sockets >= RTE_MAX_NUMA_NODES) {
565 				rte_exit(EXIT_FAILURE,
566 					 "Total sockets greater than %u\n",
567 					 RTE_MAX_NUMA_NODES);
568 			}
569 			socket_ids[num_sockets++] = sock_num;
570 		}
571 		if (i == rte_get_master_lcore())
572 			continue;
573 		fwd_lcores_cpuids[nb_lc++] = i;
574 	}
575 	nb_lcores = (lcoreid_t) nb_lc;
576 	nb_cfg_lcores = nb_lcores;
577 	nb_fwd_lcores = 1;
578 }
579 
580 static void
581 set_def_peer_eth_addrs(void)
582 {
583 	portid_t i;
584 
585 	for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
586 		peer_eth_addrs[i].addr_bytes[0] = RTE_ETHER_LOCAL_ADMIN_ADDR;
587 		peer_eth_addrs[i].addr_bytes[5] = i;
588 	}
589 }
590 
591 static void
592 set_default_fwd_ports_config(void)
593 {
594 	portid_t pt_id;
595 	int i = 0;
596 
597 	RTE_ETH_FOREACH_DEV(pt_id) {
598 		fwd_ports_ids[i++] = pt_id;
599 
600 		/* Update sockets info according to the attached device */
601 		int socket_id = rte_eth_dev_socket_id(pt_id);
602 		if (socket_id >= 0 && new_socket_id(socket_id)) {
603 			if (num_sockets >= RTE_MAX_NUMA_NODES) {
604 				rte_exit(EXIT_FAILURE,
605 					 "Total sockets greater than %u\n",
606 					 RTE_MAX_NUMA_NODES);
607 			}
608 			socket_ids[num_sockets++] = socket_id;
609 		}
610 	}
611 
612 	nb_cfg_ports = nb_ports;
613 	nb_fwd_ports = nb_ports;
614 }
615 
616 void
617 set_def_fwd_config(void)
618 {
619 	set_default_fwd_lcores_config();
620 	set_def_peer_eth_addrs();
621 	set_default_fwd_ports_config();
622 }
623 
624 /* extremely pessimistic estimation of memory required to create a mempool */
625 static int
626 calc_mem_size(uint32_t nb_mbufs, uint32_t mbuf_sz, size_t pgsz, size_t *out)
627 {
628 	unsigned int n_pages, mbuf_per_pg, leftover;
629 	uint64_t total_mem, mbuf_mem, obj_sz;
630 
631 	/* there is no good way to predict how much space the mempool will
632 	 * occupy because it will allocate chunks on the fly, and some of those
633 	 * will come from default DPDK memory while some will come from our
634 	 * external memory, so just assume 128MB will be enough for everyone.
635 	 */
636 	uint64_t hdr_mem = 128 << 20;
637 
638 	/* account for possible non-contiguousness */
639 	obj_sz = rte_mempool_calc_obj_size(mbuf_sz, 0, NULL);
640 	if (obj_sz > pgsz) {
641 		TESTPMD_LOG(ERR, "Object size is bigger than page size\n");
642 		return -1;
643 	}
644 
645 	mbuf_per_pg = pgsz / obj_sz;
646 	leftover = (nb_mbufs % mbuf_per_pg) > 0;
647 	n_pages = (nb_mbufs / mbuf_per_pg) + leftover;
648 
649 	mbuf_mem = n_pages * pgsz;
650 
651 	total_mem = RTE_ALIGN(hdr_mem + mbuf_mem, pgsz);
652 
653 	if (total_mem > SIZE_MAX) {
654 		TESTPMD_LOG(ERR, "Memory size too big\n");
655 		return -1;
656 	}
657 	*out = (size_t)total_mem;
658 
659 	return 0;
660 }
661 
662 static int
663 pagesz_flags(uint64_t page_sz)
664 {
665 	/* as per mmap() manpage, all page sizes are log2 of page size
666 	 * shifted by MAP_HUGE_SHIFT
667 	 */
668 	int log2 = rte_log2_u64(page_sz);
669 
670 	return (log2 << HUGE_SHIFT);
671 }
672 
673 static void *
674 alloc_mem(size_t memsz, size_t pgsz, bool huge)
675 {
676 	void *addr;
677 	int flags;
678 
679 	/* allocate anonymous hugepages */
680 	flags = MAP_ANONYMOUS | MAP_PRIVATE;
681 	if (huge)
682 		flags |= HUGE_FLAG | pagesz_flags(pgsz);
683 
684 	addr = mmap(NULL, memsz, PROT_READ | PROT_WRITE, flags, -1, 0);
685 	if (addr == MAP_FAILED)
686 		return NULL;
687 
688 	return addr;
689 }
690 
691 struct extmem_param {
692 	void *addr;
693 	size_t len;
694 	size_t pgsz;
695 	rte_iova_t *iova_table;
696 	unsigned int iova_table_len;
697 };
698 
699 static int
700 create_extmem(uint32_t nb_mbufs, uint32_t mbuf_sz, struct extmem_param *param,
701 		bool huge)
702 {
703 	uint64_t pgsizes[] = {RTE_PGSIZE_2M, RTE_PGSIZE_1G, /* x86_64, ARM */
704 			RTE_PGSIZE_16M, RTE_PGSIZE_16G};    /* POWER */
705 	unsigned int cur_page, n_pages, pgsz_idx;
706 	size_t mem_sz, cur_pgsz;
707 	rte_iova_t *iovas = NULL;
708 	void *addr;
709 	int ret;
710 
711 	for (pgsz_idx = 0; pgsz_idx < RTE_DIM(pgsizes); pgsz_idx++) {
712 		/* skip anything that is too big */
713 		if (pgsizes[pgsz_idx] > SIZE_MAX)
714 			continue;
715 
716 		cur_pgsz = pgsizes[pgsz_idx];
717 
718 		/* if we were told not to allocate hugepages, override */
719 		if (!huge)
720 			cur_pgsz = sysconf(_SC_PAGESIZE);
721 
722 		ret = calc_mem_size(nb_mbufs, mbuf_sz, cur_pgsz, &mem_sz);
723 		if (ret < 0) {
724 			TESTPMD_LOG(ERR, "Cannot calculate memory size\n");
725 			return -1;
726 		}
727 
728 		/* allocate our memory */
729 		addr = alloc_mem(mem_sz, cur_pgsz, huge);
730 
731 		/* if we couldn't allocate memory with a specified page size,
732 		 * that doesn't mean we can't do it with other page sizes, so
733 		 * try another one.
734 		 */
735 		if (addr == NULL)
736 			continue;
737 
738 		/* store IOVA addresses for every page in this memory area */
739 		n_pages = mem_sz / cur_pgsz;
740 
741 		iovas = malloc(sizeof(*iovas) * n_pages);
742 
743 		if (iovas == NULL) {
744 			TESTPMD_LOG(ERR, "Cannot allocate memory for iova addresses\n");
745 			goto fail;
746 		}
747 		/* lock memory if it's not huge pages */
748 		if (!huge)
749 			mlock(addr, mem_sz);
750 
751 		/* populate IOVA addresses */
752 		for (cur_page = 0; cur_page < n_pages; cur_page++) {
753 			rte_iova_t iova;
754 			size_t offset;
755 			void *cur;
756 
757 			offset = cur_pgsz * cur_page;
758 			cur = RTE_PTR_ADD(addr, offset);
759 
760 			/* touch the page before getting its IOVA */
761 			*(volatile char *)cur = 0;
762 
763 			iova = rte_mem_virt2iova(cur);
764 
765 			iovas[cur_page] = iova;
766 		}
767 
768 		break;
769 	}
770 	/* if we couldn't allocate anything */
771 	if (iovas == NULL)
772 		return -1;
773 
774 	param->addr = addr;
775 	param->len = mem_sz;
776 	param->pgsz = cur_pgsz;
777 	param->iova_table = iovas;
778 	param->iova_table_len = n_pages;
779 
780 	return 0;
781 fail:
782 	if (iovas)
783 		free(iovas);
784 	if (addr)
785 		munmap(addr, mem_sz);
786 
787 	return -1;
788 }
789 
790 static int
791 setup_extmem(uint32_t nb_mbufs, uint32_t mbuf_sz, bool huge)
792 {
793 	struct extmem_param param;
794 	int socket_id, ret;
795 
796 	memset(&param, 0, sizeof(param));
797 
798 	/* check if our heap exists */
799 	socket_id = rte_malloc_heap_get_socket(EXTMEM_HEAP_NAME);
800 	if (socket_id < 0) {
801 		/* create our heap */
802 		ret = rte_malloc_heap_create(EXTMEM_HEAP_NAME);
803 		if (ret < 0) {
804 			TESTPMD_LOG(ERR, "Cannot create heap\n");
805 			return -1;
806 		}
807 	}
808 
809 	ret = create_extmem(nb_mbufs, mbuf_sz, &param, huge);
810 	if (ret < 0) {
811 		TESTPMD_LOG(ERR, "Cannot create memory area\n");
812 		return -1;
813 	}
814 
815 	/* we now have a valid memory area, so add it to heap */
816 	ret = rte_malloc_heap_memory_add(EXTMEM_HEAP_NAME,
817 			param.addr, param.len, param.iova_table,
818 			param.iova_table_len, param.pgsz);
819 
820 	/* when using VFIO, memory is automatically mapped for DMA by EAL */
821 
822 	/* not needed any more */
823 	free(param.iova_table);
824 
825 	if (ret < 0) {
826 		TESTPMD_LOG(ERR, "Cannot add memory to heap\n");
827 		munmap(param.addr, param.len);
828 		return -1;
829 	}
830 
831 	/* success */
832 
833 	TESTPMD_LOG(DEBUG, "Allocated %zuMB of external memory\n",
834 			param.len >> 20);
835 
836 	return 0;
837 }
838 static void
839 dma_unmap_cb(struct rte_mempool *mp __rte_unused, void *opaque __rte_unused,
840 	     struct rte_mempool_memhdr *memhdr, unsigned mem_idx __rte_unused)
841 {
842 	uint16_t pid = 0;
843 	int ret;
844 
845 	RTE_ETH_FOREACH_DEV(pid) {
846 		struct rte_eth_dev *dev =
847 			&rte_eth_devices[pid];
848 
849 		ret = rte_dev_dma_unmap(dev->device, memhdr->addr, 0,
850 					memhdr->len);
851 		if (ret) {
852 			TESTPMD_LOG(DEBUG,
853 				    "unable to DMA unmap addr 0x%p "
854 				    "for device %s\n",
855 				    memhdr->addr, dev->data->name);
856 		}
857 	}
858 	ret = rte_extmem_unregister(memhdr->addr, memhdr->len);
859 	if (ret) {
860 		TESTPMD_LOG(DEBUG,
861 			    "unable to un-register addr 0x%p\n", memhdr->addr);
862 	}
863 }
864 
865 static void
866 dma_map_cb(struct rte_mempool *mp __rte_unused, void *opaque __rte_unused,
867 	   struct rte_mempool_memhdr *memhdr, unsigned mem_idx __rte_unused)
868 {
869 	uint16_t pid = 0;
870 	size_t page_size = sysconf(_SC_PAGESIZE);
871 	int ret;
872 
873 	ret = rte_extmem_register(memhdr->addr, memhdr->len, NULL, 0,
874 				  page_size);
875 	if (ret) {
876 		TESTPMD_LOG(DEBUG,
877 			    "unable to register addr 0x%p\n", memhdr->addr);
878 		return;
879 	}
880 	RTE_ETH_FOREACH_DEV(pid) {
881 		struct rte_eth_dev *dev =
882 			&rte_eth_devices[pid];
883 
884 		ret = rte_dev_dma_map(dev->device, memhdr->addr, 0,
885 				      memhdr->len);
886 		if (ret) {
887 			TESTPMD_LOG(DEBUG,
888 				    "unable to DMA map addr 0x%p "
889 				    "for device %s\n",
890 				    memhdr->addr, dev->data->name);
891 		}
892 	}
893 }
894 
895 /*
896  * Configuration initialisation done once at init time.
897  */
898 static struct rte_mempool *
899 mbuf_pool_create(uint16_t mbuf_seg_size, unsigned nb_mbuf,
900 		 unsigned int socket_id)
901 {
902 	char pool_name[RTE_MEMPOOL_NAMESIZE];
903 	struct rte_mempool *rte_mp = NULL;
904 	uint32_t mb_size;
905 
906 	mb_size = sizeof(struct rte_mbuf) + mbuf_seg_size;
907 	mbuf_poolname_build(socket_id, pool_name, sizeof(pool_name));
908 
909 	TESTPMD_LOG(INFO,
910 		"create a new mbuf pool <%s>: n=%u, size=%u, socket=%u\n",
911 		pool_name, nb_mbuf, mbuf_seg_size, socket_id);
912 
913 	switch (mp_alloc_type) {
914 	case MP_ALLOC_NATIVE:
915 		{
916 			/* wrapper to rte_mempool_create() */
917 			TESTPMD_LOG(INFO, "preferred mempool ops selected: %s\n",
918 					rte_mbuf_best_mempool_ops());
919 			rte_mp = rte_pktmbuf_pool_create(pool_name, nb_mbuf,
920 				mb_mempool_cache, 0, mbuf_seg_size, socket_id);
921 			break;
922 		}
923 	case MP_ALLOC_ANON:
924 		{
925 			rte_mp = rte_mempool_create_empty(pool_name, nb_mbuf,
926 				mb_size, (unsigned int) mb_mempool_cache,
927 				sizeof(struct rte_pktmbuf_pool_private),
928 				socket_id, mempool_flags);
929 			if (rte_mp == NULL)
930 				goto err;
931 
932 			if (rte_mempool_populate_anon(rte_mp) == 0) {
933 				rte_mempool_free(rte_mp);
934 				rte_mp = NULL;
935 				goto err;
936 			}
937 			rte_pktmbuf_pool_init(rte_mp, NULL);
938 			rte_mempool_obj_iter(rte_mp, rte_pktmbuf_init, NULL);
939 			rte_mempool_mem_iter(rte_mp, dma_map_cb, NULL);
940 			break;
941 		}
942 	case MP_ALLOC_XMEM:
943 	case MP_ALLOC_XMEM_HUGE:
944 		{
945 			int heap_socket;
946 			bool huge = mp_alloc_type == MP_ALLOC_XMEM_HUGE;
947 
948 			if (setup_extmem(nb_mbuf, mbuf_seg_size, huge) < 0)
949 				rte_exit(EXIT_FAILURE, "Could not create external memory\n");
950 
951 			heap_socket =
952 				rte_malloc_heap_get_socket(EXTMEM_HEAP_NAME);
953 			if (heap_socket < 0)
954 				rte_exit(EXIT_FAILURE, "Could not get external memory socket ID\n");
955 
956 			TESTPMD_LOG(INFO, "preferred mempool ops selected: %s\n",
957 					rte_mbuf_best_mempool_ops());
958 			rte_mp = rte_pktmbuf_pool_create(pool_name, nb_mbuf,
959 					mb_mempool_cache, 0, mbuf_seg_size,
960 					heap_socket);
961 			break;
962 		}
963 	default:
964 		{
965 			rte_exit(EXIT_FAILURE, "Invalid mempool creation mode\n");
966 		}
967 	}
968 
969 err:
970 	if (rte_mp == NULL) {
971 		rte_exit(EXIT_FAILURE,
972 			"Creation of mbuf pool for socket %u failed: %s\n",
973 			socket_id, rte_strerror(rte_errno));
974 	} else if (verbose_level > 0) {
975 		rte_mempool_dump(stdout, rte_mp);
976 	}
977 	return rte_mp;
978 }
979 
980 /*
981  * Check given socket id is valid or not with NUMA mode,
982  * if valid, return 0, else return -1
983  */
984 static int
985 check_socket_id(const unsigned int socket_id)
986 {
987 	static int warning_once = 0;
988 
989 	if (new_socket_id(socket_id)) {
990 		if (!warning_once && numa_support)
991 			printf("Warning: NUMA should be configured manually by"
992 			       " using --port-numa-config and"
993 			       " --ring-numa-config parameters along with"
994 			       " --numa.\n");
995 		warning_once = 1;
996 		return -1;
997 	}
998 	return 0;
999 }
1000 
1001 /*
1002  * Get the allowed maximum number of RX queues.
1003  * *pid return the port id which has minimal value of
1004  * max_rx_queues in all ports.
1005  */
1006 queueid_t
1007 get_allowed_max_nb_rxq(portid_t *pid)
1008 {
1009 	queueid_t allowed_max_rxq = MAX_QUEUE_ID;
1010 	portid_t pi;
1011 	struct rte_eth_dev_info dev_info;
1012 
1013 	RTE_ETH_FOREACH_DEV(pi) {
1014 		rte_eth_dev_info_get(pi, &dev_info);
1015 		if (dev_info.max_rx_queues < allowed_max_rxq) {
1016 			allowed_max_rxq = dev_info.max_rx_queues;
1017 			*pid = pi;
1018 		}
1019 	}
1020 	return allowed_max_rxq;
1021 }
1022 
1023 /*
1024  * Check input rxq is valid or not.
1025  * If input rxq is not greater than any of maximum number
1026  * of RX queues of all ports, it is valid.
1027  * if valid, return 0, else return -1
1028  */
1029 int
1030 check_nb_rxq(queueid_t rxq)
1031 {
1032 	queueid_t allowed_max_rxq;
1033 	portid_t pid = 0;
1034 
1035 	allowed_max_rxq = get_allowed_max_nb_rxq(&pid);
1036 	if (rxq > allowed_max_rxq) {
1037 		printf("Fail: input rxq (%u) can't be greater "
1038 		       "than max_rx_queues (%u) of port %u\n",
1039 		       rxq,
1040 		       allowed_max_rxq,
1041 		       pid);
1042 		return -1;
1043 	}
1044 	return 0;
1045 }
1046 
1047 /*
1048  * Get the allowed maximum number of TX queues.
1049  * *pid return the port id which has minimal value of
1050  * max_tx_queues in all ports.
1051  */
1052 queueid_t
1053 get_allowed_max_nb_txq(portid_t *pid)
1054 {
1055 	queueid_t allowed_max_txq = MAX_QUEUE_ID;
1056 	portid_t pi;
1057 	struct rte_eth_dev_info dev_info;
1058 
1059 	RTE_ETH_FOREACH_DEV(pi) {
1060 		rte_eth_dev_info_get(pi, &dev_info);
1061 		if (dev_info.max_tx_queues < allowed_max_txq) {
1062 			allowed_max_txq = dev_info.max_tx_queues;
1063 			*pid = pi;
1064 		}
1065 	}
1066 	return allowed_max_txq;
1067 }
1068 
1069 /*
1070  * Check input txq is valid or not.
1071  * If input txq is not greater than any of maximum number
1072  * of TX queues of all ports, it is valid.
1073  * if valid, return 0, else return -1
1074  */
1075 int
1076 check_nb_txq(queueid_t txq)
1077 {
1078 	queueid_t allowed_max_txq;
1079 	portid_t pid = 0;
1080 
1081 	allowed_max_txq = get_allowed_max_nb_txq(&pid);
1082 	if (txq > allowed_max_txq) {
1083 		printf("Fail: input txq (%u) can't be greater "
1084 		       "than max_tx_queues (%u) of port %u\n",
1085 		       txq,
1086 		       allowed_max_txq,
1087 		       pid);
1088 		return -1;
1089 	}
1090 	return 0;
1091 }
1092 
1093 static void
1094 init_config(void)
1095 {
1096 	portid_t pid;
1097 	struct rte_port *port;
1098 	struct rte_mempool *mbp;
1099 	unsigned int nb_mbuf_per_pool;
1100 	lcoreid_t  lc_id;
1101 	uint8_t port_per_socket[RTE_MAX_NUMA_NODES];
1102 	struct rte_gro_param gro_param;
1103 	uint32_t gso_types;
1104 	uint16_t data_size;
1105 	bool warning = 0;
1106 	int k;
1107 
1108 	memset(port_per_socket,0,RTE_MAX_NUMA_NODES);
1109 
1110 	/* Configuration of logical cores. */
1111 	fwd_lcores = rte_zmalloc("testpmd: fwd_lcores",
1112 				sizeof(struct fwd_lcore *) * nb_lcores,
1113 				RTE_CACHE_LINE_SIZE);
1114 	if (fwd_lcores == NULL) {
1115 		rte_exit(EXIT_FAILURE, "rte_zmalloc(%d (struct fwd_lcore *)) "
1116 							"failed\n", nb_lcores);
1117 	}
1118 	for (lc_id = 0; lc_id < nb_lcores; lc_id++) {
1119 		fwd_lcores[lc_id] = rte_zmalloc("testpmd: struct fwd_lcore",
1120 					       sizeof(struct fwd_lcore),
1121 					       RTE_CACHE_LINE_SIZE);
1122 		if (fwd_lcores[lc_id] == NULL) {
1123 			rte_exit(EXIT_FAILURE, "rte_zmalloc(struct fwd_lcore) "
1124 								"failed\n");
1125 		}
1126 		fwd_lcores[lc_id]->cpuid_idx = lc_id;
1127 	}
1128 
1129 	RTE_ETH_FOREACH_DEV(pid) {
1130 		port = &ports[pid];
1131 		/* Apply default TxRx configuration for all ports */
1132 		port->dev_conf.txmode = tx_mode;
1133 		port->dev_conf.rxmode = rx_mode;
1134 		rte_eth_dev_info_get(pid, &port->dev_info);
1135 
1136 		if (!(port->dev_info.tx_offload_capa &
1137 		      DEV_TX_OFFLOAD_MBUF_FAST_FREE))
1138 			port->dev_conf.txmode.offloads &=
1139 				~DEV_TX_OFFLOAD_MBUF_FAST_FREE;
1140 		if (!(port->dev_info.tx_offload_capa &
1141 			DEV_TX_OFFLOAD_MATCH_METADATA))
1142 			port->dev_conf.txmode.offloads &=
1143 				~DEV_TX_OFFLOAD_MATCH_METADATA;
1144 		if (numa_support) {
1145 			if (port_numa[pid] != NUMA_NO_CONFIG)
1146 				port_per_socket[port_numa[pid]]++;
1147 			else {
1148 				uint32_t socket_id = rte_eth_dev_socket_id(pid);
1149 
1150 				/*
1151 				 * if socket_id is invalid,
1152 				 * set to the first available socket.
1153 				 */
1154 				if (check_socket_id(socket_id) < 0)
1155 					socket_id = socket_ids[0];
1156 				port_per_socket[socket_id]++;
1157 			}
1158 		}
1159 
1160 		/* Apply Rx offloads configuration */
1161 		for (k = 0; k < port->dev_info.max_rx_queues; k++)
1162 			port->rx_conf[k].offloads =
1163 				port->dev_conf.rxmode.offloads;
1164 		/* Apply Tx offloads configuration */
1165 		for (k = 0; k < port->dev_info.max_tx_queues; k++)
1166 			port->tx_conf[k].offloads =
1167 				port->dev_conf.txmode.offloads;
1168 
1169 		/* set flag to initialize port/queue */
1170 		port->need_reconfig = 1;
1171 		port->need_reconfig_queues = 1;
1172 		port->tx_metadata = 0;
1173 
1174 		/* Check for maximum number of segments per MTU. Accordingly
1175 		 * update the mbuf data size.
1176 		 */
1177 		if (port->dev_info.rx_desc_lim.nb_mtu_seg_max != UINT16_MAX &&
1178 				port->dev_info.rx_desc_lim.nb_mtu_seg_max != 0) {
1179 			data_size = rx_mode.max_rx_pkt_len /
1180 				port->dev_info.rx_desc_lim.nb_mtu_seg_max;
1181 
1182 			if ((data_size + RTE_PKTMBUF_HEADROOM) >
1183 							mbuf_data_size) {
1184 				mbuf_data_size = data_size +
1185 						 RTE_PKTMBUF_HEADROOM;
1186 				warning = 1;
1187 			}
1188 		}
1189 	}
1190 
1191 	if (warning)
1192 		TESTPMD_LOG(WARNING, "Configured mbuf size %hu\n",
1193 			    mbuf_data_size);
1194 
1195 	/*
1196 	 * Create pools of mbuf.
1197 	 * If NUMA support is disabled, create a single pool of mbuf in
1198 	 * socket 0 memory by default.
1199 	 * Otherwise, create a pool of mbuf in the memory of sockets 0 and 1.
1200 	 *
1201 	 * Use the maximum value of nb_rxd and nb_txd here, then nb_rxd and
1202 	 * nb_txd can be configured at run time.
1203 	 */
1204 	if (param_total_num_mbufs)
1205 		nb_mbuf_per_pool = param_total_num_mbufs;
1206 	else {
1207 		nb_mbuf_per_pool = RTE_TEST_RX_DESC_MAX +
1208 			(nb_lcores * mb_mempool_cache) +
1209 			RTE_TEST_TX_DESC_MAX + MAX_PKT_BURST;
1210 		nb_mbuf_per_pool *= RTE_MAX_ETHPORTS;
1211 	}
1212 
1213 	if (numa_support) {
1214 		uint8_t i;
1215 
1216 		for (i = 0; i < num_sockets; i++)
1217 			mempools[i] = mbuf_pool_create(mbuf_data_size,
1218 						       nb_mbuf_per_pool,
1219 						       socket_ids[i]);
1220 	} else {
1221 		if (socket_num == UMA_NO_CONFIG)
1222 			mempools[0] = mbuf_pool_create(mbuf_data_size,
1223 						       nb_mbuf_per_pool, 0);
1224 		else
1225 			mempools[socket_num] = mbuf_pool_create
1226 							(mbuf_data_size,
1227 							 nb_mbuf_per_pool,
1228 							 socket_num);
1229 	}
1230 
1231 	init_port_config();
1232 
1233 	gso_types = DEV_TX_OFFLOAD_TCP_TSO | DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
1234 		DEV_TX_OFFLOAD_GRE_TNL_TSO | DEV_TX_OFFLOAD_UDP_TSO;
1235 	/*
1236 	 * Records which Mbuf pool to use by each logical core, if needed.
1237 	 */
1238 	for (lc_id = 0; lc_id < nb_lcores; lc_id++) {
1239 		mbp = mbuf_pool_find(
1240 			rte_lcore_to_socket_id(fwd_lcores_cpuids[lc_id]));
1241 
1242 		if (mbp == NULL)
1243 			mbp = mbuf_pool_find(0);
1244 		fwd_lcores[lc_id]->mbp = mbp;
1245 		/* initialize GSO context */
1246 		fwd_lcores[lc_id]->gso_ctx.direct_pool = mbp;
1247 		fwd_lcores[lc_id]->gso_ctx.indirect_pool = mbp;
1248 		fwd_lcores[lc_id]->gso_ctx.gso_types = gso_types;
1249 		fwd_lcores[lc_id]->gso_ctx.gso_size = RTE_ETHER_MAX_LEN -
1250 			RTE_ETHER_CRC_LEN;
1251 		fwd_lcores[lc_id]->gso_ctx.flag = 0;
1252 	}
1253 
1254 	/* Configuration of packet forwarding streams. */
1255 	if (init_fwd_streams() < 0)
1256 		rte_exit(EXIT_FAILURE, "FAIL from init_fwd_streams()\n");
1257 
1258 	fwd_config_setup();
1259 
1260 	/* create a gro context for each lcore */
1261 	gro_param.gro_types = RTE_GRO_TCP_IPV4;
1262 	gro_param.max_flow_num = GRO_MAX_FLUSH_CYCLES;
1263 	gro_param.max_item_per_flow = MAX_PKT_BURST;
1264 	for (lc_id = 0; lc_id < nb_lcores; lc_id++) {
1265 		gro_param.socket_id = rte_lcore_to_socket_id(
1266 				fwd_lcores_cpuids[lc_id]);
1267 		fwd_lcores[lc_id]->gro_ctx = rte_gro_ctx_create(&gro_param);
1268 		if (fwd_lcores[lc_id]->gro_ctx == NULL) {
1269 			rte_exit(EXIT_FAILURE,
1270 					"rte_gro_ctx_create() failed\n");
1271 		}
1272 	}
1273 
1274 #if defined RTE_LIBRTE_PMD_SOFTNIC
1275 	if (strcmp(cur_fwd_eng->fwd_mode_name, "softnic") == 0) {
1276 		RTE_ETH_FOREACH_DEV(pid) {
1277 			port = &ports[pid];
1278 			const char *driver = port->dev_info.driver_name;
1279 
1280 			if (strcmp(driver, "net_softnic") == 0)
1281 				port->softport.fwd_lcore_arg = fwd_lcores;
1282 		}
1283 	}
1284 #endif
1285 
1286 }
1287 
1288 
1289 void
1290 reconfig(portid_t new_port_id, unsigned socket_id)
1291 {
1292 	struct rte_port *port;
1293 
1294 	/* Reconfiguration of Ethernet ports. */
1295 	port = &ports[new_port_id];
1296 	rte_eth_dev_info_get(new_port_id, &port->dev_info);
1297 
1298 	/* set flag to initialize port/queue */
1299 	port->need_reconfig = 1;
1300 	port->need_reconfig_queues = 1;
1301 	port->socket_id = socket_id;
1302 
1303 	init_port_config();
1304 }
1305 
1306 
1307 int
1308 init_fwd_streams(void)
1309 {
1310 	portid_t pid;
1311 	struct rte_port *port;
1312 	streamid_t sm_id, nb_fwd_streams_new;
1313 	queueid_t q;
1314 
1315 	/* set socket id according to numa or not */
1316 	RTE_ETH_FOREACH_DEV(pid) {
1317 		port = &ports[pid];
1318 		if (nb_rxq > port->dev_info.max_rx_queues) {
1319 			printf("Fail: nb_rxq(%d) is greater than "
1320 				"max_rx_queues(%d)\n", nb_rxq,
1321 				port->dev_info.max_rx_queues);
1322 			return -1;
1323 		}
1324 		if (nb_txq > port->dev_info.max_tx_queues) {
1325 			printf("Fail: nb_txq(%d) is greater than "
1326 				"max_tx_queues(%d)\n", nb_txq,
1327 				port->dev_info.max_tx_queues);
1328 			return -1;
1329 		}
1330 		if (numa_support) {
1331 			if (port_numa[pid] != NUMA_NO_CONFIG)
1332 				port->socket_id = port_numa[pid];
1333 			else {
1334 				port->socket_id = rte_eth_dev_socket_id(pid);
1335 
1336 				/*
1337 				 * if socket_id is invalid,
1338 				 * set to the first available socket.
1339 				 */
1340 				if (check_socket_id(port->socket_id) < 0)
1341 					port->socket_id = socket_ids[0];
1342 			}
1343 		}
1344 		else {
1345 			if (socket_num == UMA_NO_CONFIG)
1346 				port->socket_id = 0;
1347 			else
1348 				port->socket_id = socket_num;
1349 		}
1350 	}
1351 
1352 	q = RTE_MAX(nb_rxq, nb_txq);
1353 	if (q == 0) {
1354 		printf("Fail: Cannot allocate fwd streams as number of queues is 0\n");
1355 		return -1;
1356 	}
1357 	nb_fwd_streams_new = (streamid_t)(nb_ports * q);
1358 	if (nb_fwd_streams_new == nb_fwd_streams)
1359 		return 0;
1360 	/* clear the old */
1361 	if (fwd_streams != NULL) {
1362 		for (sm_id = 0; sm_id < nb_fwd_streams; sm_id++) {
1363 			if (fwd_streams[sm_id] == NULL)
1364 				continue;
1365 			rte_free(fwd_streams[sm_id]);
1366 			fwd_streams[sm_id] = NULL;
1367 		}
1368 		rte_free(fwd_streams);
1369 		fwd_streams = NULL;
1370 	}
1371 
1372 	/* init new */
1373 	nb_fwd_streams = nb_fwd_streams_new;
1374 	if (nb_fwd_streams) {
1375 		fwd_streams = rte_zmalloc("testpmd: fwd_streams",
1376 			sizeof(struct fwd_stream *) * nb_fwd_streams,
1377 			RTE_CACHE_LINE_SIZE);
1378 		if (fwd_streams == NULL)
1379 			rte_exit(EXIT_FAILURE, "rte_zmalloc(%d"
1380 				 " (struct fwd_stream *)) failed\n",
1381 				 nb_fwd_streams);
1382 
1383 		for (sm_id = 0; sm_id < nb_fwd_streams; sm_id++) {
1384 			fwd_streams[sm_id] = rte_zmalloc("testpmd:"
1385 				" struct fwd_stream", sizeof(struct fwd_stream),
1386 				RTE_CACHE_LINE_SIZE);
1387 			if (fwd_streams[sm_id] == NULL)
1388 				rte_exit(EXIT_FAILURE, "rte_zmalloc"
1389 					 "(struct fwd_stream) failed\n");
1390 		}
1391 	}
1392 
1393 	return 0;
1394 }
1395 
1396 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
1397 static void
1398 pkt_burst_stats_display(const char *rx_tx, struct pkt_burst_stats *pbs)
1399 {
1400 	unsigned int total_burst;
1401 	unsigned int nb_burst;
1402 	unsigned int burst_stats[3];
1403 	uint16_t pktnb_stats[3];
1404 	uint16_t nb_pkt;
1405 	int burst_percent[3];
1406 
1407 	/*
1408 	 * First compute the total number of packet bursts and the
1409 	 * two highest numbers of bursts of the same number of packets.
1410 	 */
1411 	total_burst = 0;
1412 	burst_stats[0] = burst_stats[1] = burst_stats[2] = 0;
1413 	pktnb_stats[0] = pktnb_stats[1] = pktnb_stats[2] = 0;
1414 	for (nb_pkt = 0; nb_pkt < MAX_PKT_BURST; nb_pkt++) {
1415 		nb_burst = pbs->pkt_burst_spread[nb_pkt];
1416 		if (nb_burst == 0)
1417 			continue;
1418 		total_burst += nb_burst;
1419 		if (nb_burst > burst_stats[0]) {
1420 			burst_stats[1] = burst_stats[0];
1421 			pktnb_stats[1] = pktnb_stats[0];
1422 			burst_stats[0] = nb_burst;
1423 			pktnb_stats[0] = nb_pkt;
1424 		} else if (nb_burst > burst_stats[1]) {
1425 			burst_stats[1] = nb_burst;
1426 			pktnb_stats[1] = nb_pkt;
1427 		}
1428 	}
1429 	if (total_burst == 0)
1430 		return;
1431 	burst_percent[0] = (burst_stats[0] * 100) / total_burst;
1432 	printf("  %s-bursts : %u [%d%% of %d pkts", rx_tx, total_burst,
1433 	       burst_percent[0], (int) pktnb_stats[0]);
1434 	if (burst_stats[0] == total_burst) {
1435 		printf("]\n");
1436 		return;
1437 	}
1438 	if (burst_stats[0] + burst_stats[1] == total_burst) {
1439 		printf(" + %d%% of %d pkts]\n",
1440 		       100 - burst_percent[0], pktnb_stats[1]);
1441 		return;
1442 	}
1443 	burst_percent[1] = (burst_stats[1] * 100) / total_burst;
1444 	burst_percent[2] = 100 - (burst_percent[0] + burst_percent[1]);
1445 	if ((burst_percent[1] == 0) || (burst_percent[2] == 0)) {
1446 		printf(" + %d%% of others]\n", 100 - burst_percent[0]);
1447 		return;
1448 	}
1449 	printf(" + %d%% of %d pkts + %d%% of others]\n",
1450 	       burst_percent[1], (int) pktnb_stats[1], burst_percent[2]);
1451 }
1452 #endif /* RTE_TEST_PMD_RECORD_BURST_STATS */
1453 
1454 static void
1455 fwd_stream_stats_display(streamid_t stream_id)
1456 {
1457 	struct fwd_stream *fs;
1458 	static const char *fwd_top_stats_border = "-------";
1459 
1460 	fs = fwd_streams[stream_id];
1461 	if ((fs->rx_packets == 0) && (fs->tx_packets == 0) &&
1462 	    (fs->fwd_dropped == 0))
1463 		return;
1464 	printf("\n  %s Forward Stats for RX Port=%2d/Queue=%2d -> "
1465 	       "TX Port=%2d/Queue=%2d %s\n",
1466 	       fwd_top_stats_border, fs->rx_port, fs->rx_queue,
1467 	       fs->tx_port, fs->tx_queue, fwd_top_stats_border);
1468 	printf("  RX-packets: %-14"PRIu64" TX-packets: %-14"PRIu64
1469 	       " TX-dropped: %-14"PRIu64,
1470 	       fs->rx_packets, fs->tx_packets, fs->fwd_dropped);
1471 
1472 	/* if checksum mode */
1473 	if (cur_fwd_eng == &csum_fwd_engine) {
1474 		printf("  RX- bad IP checksum: %-14"PRIu64
1475 		       "  Rx- bad L4 checksum: %-14"PRIu64
1476 		       " Rx- bad outer L4 checksum: %-14"PRIu64"\n",
1477 			fs->rx_bad_ip_csum, fs->rx_bad_l4_csum,
1478 			fs->rx_bad_outer_l4_csum);
1479 	} else {
1480 		printf("\n");
1481 	}
1482 
1483 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
1484 	pkt_burst_stats_display("RX", &fs->rx_burst_stats);
1485 	pkt_burst_stats_display("TX", &fs->tx_burst_stats);
1486 #endif
1487 }
1488 
1489 void
1490 fwd_stats_display(void)
1491 {
1492 	static const char *fwd_stats_border = "----------------------";
1493 	static const char *acc_stats_border = "+++++++++++++++";
1494 	struct {
1495 		struct fwd_stream *rx_stream;
1496 		struct fwd_stream *tx_stream;
1497 		uint64_t tx_dropped;
1498 		uint64_t rx_bad_ip_csum;
1499 		uint64_t rx_bad_l4_csum;
1500 		uint64_t rx_bad_outer_l4_csum;
1501 	} ports_stats[RTE_MAX_ETHPORTS];
1502 	uint64_t total_rx_dropped = 0;
1503 	uint64_t total_tx_dropped = 0;
1504 	uint64_t total_rx_nombuf = 0;
1505 	struct rte_eth_stats stats;
1506 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
1507 	uint64_t fwd_cycles = 0;
1508 #endif
1509 	uint64_t total_recv = 0;
1510 	uint64_t total_xmit = 0;
1511 	struct rte_port *port;
1512 	streamid_t sm_id;
1513 	portid_t pt_id;
1514 	int i;
1515 
1516 	memset(ports_stats, 0, sizeof(ports_stats));
1517 
1518 	for (sm_id = 0; sm_id < cur_fwd_config.nb_fwd_streams; sm_id++) {
1519 		struct fwd_stream *fs = fwd_streams[sm_id];
1520 
1521 		if (cur_fwd_config.nb_fwd_streams >
1522 		    cur_fwd_config.nb_fwd_ports) {
1523 			fwd_stream_stats_display(sm_id);
1524 		} else {
1525 			ports_stats[fs->tx_port].tx_stream = fs;
1526 			ports_stats[fs->rx_port].rx_stream = fs;
1527 		}
1528 
1529 		ports_stats[fs->tx_port].tx_dropped += fs->fwd_dropped;
1530 
1531 		ports_stats[fs->rx_port].rx_bad_ip_csum += fs->rx_bad_ip_csum;
1532 		ports_stats[fs->rx_port].rx_bad_l4_csum += fs->rx_bad_l4_csum;
1533 		ports_stats[fs->rx_port].rx_bad_outer_l4_csum +=
1534 				fs->rx_bad_outer_l4_csum;
1535 
1536 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
1537 		fwd_cycles += fs->core_cycles;
1538 #endif
1539 	}
1540 	for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++) {
1541 		uint8_t j;
1542 
1543 		pt_id = fwd_ports_ids[i];
1544 		port = &ports[pt_id];
1545 
1546 		rte_eth_stats_get(pt_id, &stats);
1547 		stats.ipackets -= port->stats.ipackets;
1548 		stats.opackets -= port->stats.opackets;
1549 		stats.ibytes -= port->stats.ibytes;
1550 		stats.obytes -= port->stats.obytes;
1551 		stats.imissed -= port->stats.imissed;
1552 		stats.oerrors -= port->stats.oerrors;
1553 		stats.rx_nombuf -= port->stats.rx_nombuf;
1554 
1555 		total_recv += stats.ipackets;
1556 		total_xmit += stats.opackets;
1557 		total_rx_dropped += stats.imissed;
1558 		total_tx_dropped += ports_stats[pt_id].tx_dropped;
1559 		total_tx_dropped += stats.oerrors;
1560 		total_rx_nombuf  += stats.rx_nombuf;
1561 
1562 		printf("\n  %s Forward statistics for port %-2d %s\n",
1563 		       fwd_stats_border, pt_id, fwd_stats_border);
1564 
1565 		if (!port->rx_queue_stats_mapping_enabled &&
1566 		    !port->tx_queue_stats_mapping_enabled) {
1567 			printf("  RX-packets: %-14"PRIu64
1568 			       " RX-dropped: %-14"PRIu64
1569 			       "RX-total: %-"PRIu64"\n",
1570 			       stats.ipackets, stats.imissed,
1571 			       stats.ipackets + stats.imissed);
1572 
1573 			if (cur_fwd_eng == &csum_fwd_engine)
1574 				printf("  Bad-ipcsum: %-14"PRIu64
1575 				       " Bad-l4csum: %-14"PRIu64
1576 				       "Bad-outer-l4csum: %-14"PRIu64"\n",
1577 				       ports_stats[pt_id].rx_bad_ip_csum,
1578 				       ports_stats[pt_id].rx_bad_l4_csum,
1579 				       ports_stats[pt_id].rx_bad_outer_l4_csum);
1580 			if (stats.ierrors + stats.rx_nombuf > 0) {
1581 				printf("  RX-error: %-"PRIu64"\n",
1582 				       stats.ierrors);
1583 				printf("  RX-nombufs: %-14"PRIu64"\n",
1584 				       stats.rx_nombuf);
1585 			}
1586 
1587 			printf("  TX-packets: %-14"PRIu64
1588 			       " TX-dropped: %-14"PRIu64
1589 			       "TX-total: %-"PRIu64"\n",
1590 			       stats.opackets, ports_stats[pt_id].tx_dropped,
1591 			       stats.opackets + ports_stats[pt_id].tx_dropped);
1592 		} else {
1593 			printf("  RX-packets:             %14"PRIu64
1594 			       "    RX-dropped:%14"PRIu64
1595 			       "    RX-total:%14"PRIu64"\n",
1596 			       stats.ipackets, stats.imissed,
1597 			       stats.ipackets + stats.imissed);
1598 
1599 			if (cur_fwd_eng == &csum_fwd_engine)
1600 				printf("  Bad-ipcsum:%14"PRIu64
1601 				       "    Bad-l4csum:%14"PRIu64
1602 				       "    Bad-outer-l4csum: %-14"PRIu64"\n",
1603 				       ports_stats[pt_id].rx_bad_ip_csum,
1604 				       ports_stats[pt_id].rx_bad_l4_csum,
1605 				       ports_stats[pt_id].rx_bad_outer_l4_csum);
1606 			if ((stats.ierrors + stats.rx_nombuf) > 0) {
1607 				printf("  RX-error:%"PRIu64"\n", stats.ierrors);
1608 				printf("  RX-nombufs:             %14"PRIu64"\n",
1609 				       stats.rx_nombuf);
1610 			}
1611 
1612 			printf("  TX-packets:             %14"PRIu64
1613 			       "    TX-dropped:%14"PRIu64
1614 			       "    TX-total:%14"PRIu64"\n",
1615 			       stats.opackets, ports_stats[pt_id].tx_dropped,
1616 			       stats.opackets + ports_stats[pt_id].tx_dropped);
1617 		}
1618 
1619 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
1620 		if (ports_stats[pt_id].rx_stream)
1621 			pkt_burst_stats_display("RX",
1622 				&ports_stats[pt_id].rx_stream->rx_burst_stats);
1623 		if (ports_stats[pt_id].tx_stream)
1624 			pkt_burst_stats_display("TX",
1625 				&ports_stats[pt_id].tx_stream->tx_burst_stats);
1626 #endif
1627 
1628 		if (port->rx_queue_stats_mapping_enabled) {
1629 			printf("\n");
1630 			for (j = 0; j < RTE_ETHDEV_QUEUE_STAT_CNTRS; j++) {
1631 				printf("  Stats reg %2d RX-packets:%14"PRIu64
1632 				       "     RX-errors:%14"PRIu64
1633 				       "    RX-bytes:%14"PRIu64"\n",
1634 				       j, stats.q_ipackets[j],
1635 				       stats.q_errors[j], stats.q_ibytes[j]);
1636 			}
1637 			printf("\n");
1638 		}
1639 		if (port->tx_queue_stats_mapping_enabled) {
1640 			for (j = 0; j < RTE_ETHDEV_QUEUE_STAT_CNTRS; j++) {
1641 				printf("  Stats reg %2d TX-packets:%14"PRIu64
1642 				       "                                 TX-bytes:%14"
1643 				       PRIu64"\n",
1644 				       j, stats.q_opackets[j],
1645 				       stats.q_obytes[j]);
1646 			}
1647 		}
1648 
1649 		printf("  %s--------------------------------%s\n",
1650 		       fwd_stats_border, fwd_stats_border);
1651 	}
1652 
1653 	printf("\n  %s Accumulated forward statistics for all ports"
1654 	       "%s\n",
1655 	       acc_stats_border, acc_stats_border);
1656 	printf("  RX-packets: %-14"PRIu64" RX-dropped: %-14"PRIu64"RX-total: "
1657 	       "%-"PRIu64"\n"
1658 	       "  TX-packets: %-14"PRIu64" TX-dropped: %-14"PRIu64"TX-total: "
1659 	       "%-"PRIu64"\n",
1660 	       total_recv, total_rx_dropped, total_recv + total_rx_dropped,
1661 	       total_xmit, total_tx_dropped, total_xmit + total_tx_dropped);
1662 	if (total_rx_nombuf > 0)
1663 		printf("  RX-nombufs: %-14"PRIu64"\n", total_rx_nombuf);
1664 	printf("  %s++++++++++++++++++++++++++++++++++++++++++++++"
1665 	       "%s\n",
1666 	       acc_stats_border, acc_stats_border);
1667 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
1668 	if (total_recv > 0)
1669 		printf("\n  CPU cycles/packet=%u (total cycles="
1670 		       "%"PRIu64" / total RX packets=%"PRIu64")\n",
1671 		       (unsigned int)(fwd_cycles / total_recv),
1672 		       fwd_cycles, total_recv);
1673 #endif
1674 }
1675 
1676 void
1677 fwd_stats_reset(void)
1678 {
1679 	streamid_t sm_id;
1680 	portid_t pt_id;
1681 	int i;
1682 
1683 	for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++) {
1684 		pt_id = fwd_ports_ids[i];
1685 		rte_eth_stats_get(pt_id, &ports[pt_id].stats);
1686 	}
1687 	for (sm_id = 0; sm_id < cur_fwd_config.nb_fwd_streams; sm_id++) {
1688 		struct fwd_stream *fs = fwd_streams[sm_id];
1689 
1690 		fs->rx_packets = 0;
1691 		fs->tx_packets = 0;
1692 		fs->fwd_dropped = 0;
1693 		fs->rx_bad_ip_csum = 0;
1694 		fs->rx_bad_l4_csum = 0;
1695 		fs->rx_bad_outer_l4_csum = 0;
1696 
1697 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
1698 		memset(&fs->rx_burst_stats, 0, sizeof(fs->rx_burst_stats));
1699 		memset(&fs->tx_burst_stats, 0, sizeof(fs->tx_burst_stats));
1700 #endif
1701 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
1702 		fs->core_cycles = 0;
1703 #endif
1704 	}
1705 }
1706 
1707 static void
1708 flush_fwd_rx_queues(void)
1709 {
1710 	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1711 	portid_t  rxp;
1712 	portid_t port_id;
1713 	queueid_t rxq;
1714 	uint16_t  nb_rx;
1715 	uint16_t  i;
1716 	uint8_t   j;
1717 	uint64_t prev_tsc = 0, diff_tsc, cur_tsc, timer_tsc = 0;
1718 	uint64_t timer_period;
1719 
1720 	/* convert to number of cycles */
1721 	timer_period = rte_get_timer_hz(); /* 1 second timeout */
1722 
1723 	for (j = 0; j < 2; j++) {
1724 		for (rxp = 0; rxp < cur_fwd_config.nb_fwd_ports; rxp++) {
1725 			for (rxq = 0; rxq < nb_rxq; rxq++) {
1726 				port_id = fwd_ports_ids[rxp];
1727 				/**
1728 				* testpmd can stuck in the below do while loop
1729 				* if rte_eth_rx_burst() always returns nonzero
1730 				* packets. So timer is added to exit this loop
1731 				* after 1sec timer expiry.
1732 				*/
1733 				prev_tsc = rte_rdtsc();
1734 				do {
1735 					nb_rx = rte_eth_rx_burst(port_id, rxq,
1736 						pkts_burst, MAX_PKT_BURST);
1737 					for (i = 0; i < nb_rx; i++)
1738 						rte_pktmbuf_free(pkts_burst[i]);
1739 
1740 					cur_tsc = rte_rdtsc();
1741 					diff_tsc = cur_tsc - prev_tsc;
1742 					timer_tsc += diff_tsc;
1743 				} while ((nb_rx > 0) &&
1744 					(timer_tsc < timer_period));
1745 				timer_tsc = 0;
1746 			}
1747 		}
1748 		rte_delay_ms(10); /* wait 10 milli-seconds before retrying */
1749 	}
1750 }
1751 
1752 static void
1753 run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
1754 {
1755 	struct fwd_stream **fsm;
1756 	streamid_t nb_fs;
1757 	streamid_t sm_id;
1758 #ifdef RTE_LIBRTE_BITRATE
1759 	uint64_t tics_per_1sec;
1760 	uint64_t tics_datum;
1761 	uint64_t tics_current;
1762 	uint16_t i, cnt_ports;
1763 
1764 	cnt_ports = nb_ports;
1765 	tics_datum = rte_rdtsc();
1766 	tics_per_1sec = rte_get_timer_hz();
1767 #endif
1768 	fsm = &fwd_streams[fc->stream_idx];
1769 	nb_fs = fc->stream_nb;
1770 	do {
1771 		for (sm_id = 0; sm_id < nb_fs; sm_id++)
1772 			(*pkt_fwd)(fsm[sm_id]);
1773 #ifdef RTE_LIBRTE_BITRATE
1774 		if (bitrate_enabled != 0 &&
1775 				bitrate_lcore_id == rte_lcore_id()) {
1776 			tics_current = rte_rdtsc();
1777 			if (tics_current - tics_datum >= tics_per_1sec) {
1778 				/* Periodic bitrate calculation */
1779 				for (i = 0; i < cnt_ports; i++)
1780 					rte_stats_bitrate_calc(bitrate_data,
1781 						ports_ids[i]);
1782 				tics_datum = tics_current;
1783 			}
1784 		}
1785 #endif
1786 #ifdef RTE_LIBRTE_LATENCY_STATS
1787 		if (latencystats_enabled != 0 &&
1788 				latencystats_lcore_id == rte_lcore_id())
1789 			rte_latencystats_update();
1790 #endif
1791 
1792 	} while (! fc->stopped);
1793 }
1794 
1795 static int
1796 start_pkt_forward_on_core(void *fwd_arg)
1797 {
1798 	run_pkt_fwd_on_lcore((struct fwd_lcore *) fwd_arg,
1799 			     cur_fwd_config.fwd_eng->packet_fwd);
1800 	return 0;
1801 }
1802 
1803 /*
1804  * Run the TXONLY packet forwarding engine to send a single burst of packets.
1805  * Used to start communication flows in network loopback test configurations.
1806  */
1807 static int
1808 run_one_txonly_burst_on_core(void *fwd_arg)
1809 {
1810 	struct fwd_lcore *fwd_lc;
1811 	struct fwd_lcore tmp_lcore;
1812 
1813 	fwd_lc = (struct fwd_lcore *) fwd_arg;
1814 	tmp_lcore = *fwd_lc;
1815 	tmp_lcore.stopped = 1;
1816 	run_pkt_fwd_on_lcore(&tmp_lcore, tx_only_engine.packet_fwd);
1817 	return 0;
1818 }
1819 
1820 /*
1821  * Launch packet forwarding:
1822  *     - Setup per-port forwarding context.
1823  *     - launch logical cores with their forwarding configuration.
1824  */
1825 static void
1826 launch_packet_forwarding(lcore_function_t *pkt_fwd_on_lcore)
1827 {
1828 	port_fwd_begin_t port_fwd_begin;
1829 	unsigned int i;
1830 	unsigned int lc_id;
1831 	int diag;
1832 
1833 	port_fwd_begin = cur_fwd_config.fwd_eng->port_fwd_begin;
1834 	if (port_fwd_begin != NULL) {
1835 		for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
1836 			(*port_fwd_begin)(fwd_ports_ids[i]);
1837 	}
1838 	for (i = 0; i < cur_fwd_config.nb_fwd_lcores; i++) {
1839 		lc_id = fwd_lcores_cpuids[i];
1840 		if ((interactive == 0) || (lc_id != rte_lcore_id())) {
1841 			fwd_lcores[i]->stopped = 0;
1842 			diag = rte_eal_remote_launch(pkt_fwd_on_lcore,
1843 						     fwd_lcores[i], lc_id);
1844 			if (diag != 0)
1845 				printf("launch lcore %u failed - diag=%d\n",
1846 				       lc_id, diag);
1847 		}
1848 	}
1849 }
1850 
1851 /*
1852  * Launch packet forwarding configuration.
1853  */
1854 void
1855 start_packet_forwarding(int with_tx_first)
1856 {
1857 	port_fwd_begin_t port_fwd_begin;
1858 	port_fwd_end_t  port_fwd_end;
1859 	struct rte_port *port;
1860 	unsigned int i;
1861 	portid_t   pt_id;
1862 
1863 	if (strcmp(cur_fwd_eng->fwd_mode_name, "rxonly") == 0 && !nb_rxq)
1864 		rte_exit(EXIT_FAILURE, "rxq are 0, cannot use rxonly fwd mode\n");
1865 
1866 	if (strcmp(cur_fwd_eng->fwd_mode_name, "txonly") == 0 && !nb_txq)
1867 		rte_exit(EXIT_FAILURE, "txq are 0, cannot use txonly fwd mode\n");
1868 
1869 	if ((strcmp(cur_fwd_eng->fwd_mode_name, "rxonly") != 0 &&
1870 		strcmp(cur_fwd_eng->fwd_mode_name, "txonly") != 0) &&
1871 		(!nb_rxq || !nb_txq))
1872 		rte_exit(EXIT_FAILURE,
1873 			"Either rxq or txq are 0, cannot use %s fwd mode\n",
1874 			cur_fwd_eng->fwd_mode_name);
1875 
1876 	if (all_ports_started() == 0) {
1877 		printf("Not all ports were started\n");
1878 		return;
1879 	}
1880 	if (test_done == 0) {
1881 		printf("Packet forwarding already started\n");
1882 		return;
1883 	}
1884 
1885 
1886 	if(dcb_test) {
1887 		for (i = 0; i < nb_fwd_ports; i++) {
1888 			pt_id = fwd_ports_ids[i];
1889 			port = &ports[pt_id];
1890 			if (!port->dcb_flag) {
1891 				printf("In DCB mode, all forwarding ports must "
1892                                        "be configured in this mode.\n");
1893 				return;
1894 			}
1895 		}
1896 		if (nb_fwd_lcores == 1) {
1897 			printf("In DCB mode,the nb forwarding cores "
1898                                "should be larger than 1.\n");
1899 			return;
1900 		}
1901 	}
1902 	test_done = 0;
1903 
1904 	fwd_config_setup();
1905 
1906 	if(!no_flush_rx)
1907 		flush_fwd_rx_queues();
1908 
1909 	pkt_fwd_config_display(&cur_fwd_config);
1910 	rxtx_config_display();
1911 
1912 	fwd_stats_reset();
1913 	for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++) {
1914 		pt_id = fwd_ports_ids[i];
1915 		port = &ports[pt_id];
1916 		map_port_queue_stats_mapping_registers(pt_id, port);
1917 	}
1918 	if (with_tx_first) {
1919 		port_fwd_begin = tx_only_engine.port_fwd_begin;
1920 		if (port_fwd_begin != NULL) {
1921 			for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
1922 				(*port_fwd_begin)(fwd_ports_ids[i]);
1923 		}
1924 		while (with_tx_first--) {
1925 			launch_packet_forwarding(
1926 					run_one_txonly_burst_on_core);
1927 			rte_eal_mp_wait_lcore();
1928 		}
1929 		port_fwd_end = tx_only_engine.port_fwd_end;
1930 		if (port_fwd_end != NULL) {
1931 			for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
1932 				(*port_fwd_end)(fwd_ports_ids[i]);
1933 		}
1934 	}
1935 	launch_packet_forwarding(start_pkt_forward_on_core);
1936 }
1937 
1938 void
1939 stop_packet_forwarding(void)
1940 {
1941 	port_fwd_end_t port_fwd_end;
1942 	lcoreid_t lc_id;
1943 	portid_t pt_id;
1944 	int i;
1945 
1946 	if (test_done) {
1947 		printf("Packet forwarding not started\n");
1948 		return;
1949 	}
1950 	printf("Telling cores to stop...");
1951 	for (lc_id = 0; lc_id < cur_fwd_config.nb_fwd_lcores; lc_id++)
1952 		fwd_lcores[lc_id]->stopped = 1;
1953 	printf("\nWaiting for lcores to finish...\n");
1954 	rte_eal_mp_wait_lcore();
1955 	port_fwd_end = cur_fwd_config.fwd_eng->port_fwd_end;
1956 	if (port_fwd_end != NULL) {
1957 		for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++) {
1958 			pt_id = fwd_ports_ids[i];
1959 			(*port_fwd_end)(pt_id);
1960 		}
1961 	}
1962 
1963 	fwd_stats_display();
1964 
1965 	printf("\nDone.\n");
1966 	test_done = 1;
1967 }
1968 
1969 void
1970 dev_set_link_up(portid_t pid)
1971 {
1972 	if (rte_eth_dev_set_link_up(pid) < 0)
1973 		printf("\nSet link up fail.\n");
1974 }
1975 
1976 void
1977 dev_set_link_down(portid_t pid)
1978 {
1979 	if (rte_eth_dev_set_link_down(pid) < 0)
1980 		printf("\nSet link down fail.\n");
1981 }
1982 
1983 static int
1984 all_ports_started(void)
1985 {
1986 	portid_t pi;
1987 	struct rte_port *port;
1988 
1989 	RTE_ETH_FOREACH_DEV(pi) {
1990 		port = &ports[pi];
1991 		/* Check if there is a port which is not started */
1992 		if ((port->port_status != RTE_PORT_STARTED) &&
1993 			(port->slave_flag == 0))
1994 			return 0;
1995 	}
1996 
1997 	/* No port is not started */
1998 	return 1;
1999 }
2000 
2001 int
2002 port_is_stopped(portid_t port_id)
2003 {
2004 	struct rte_port *port = &ports[port_id];
2005 
2006 	if ((port->port_status != RTE_PORT_STOPPED) &&
2007 	    (port->slave_flag == 0))
2008 		return 0;
2009 	return 1;
2010 }
2011 
2012 int
2013 all_ports_stopped(void)
2014 {
2015 	portid_t pi;
2016 
2017 	RTE_ETH_FOREACH_DEV(pi) {
2018 		if (!port_is_stopped(pi))
2019 			return 0;
2020 	}
2021 
2022 	return 1;
2023 }
2024 
2025 int
2026 port_is_started(portid_t port_id)
2027 {
2028 	if (port_id_is_invalid(port_id, ENABLED_WARN))
2029 		return 0;
2030 
2031 	if (ports[port_id].port_status != RTE_PORT_STARTED)
2032 		return 0;
2033 
2034 	return 1;
2035 }
2036 
2037 int
2038 start_port(portid_t pid)
2039 {
2040 	int diag, need_check_link_status = -1;
2041 	portid_t pi;
2042 	queueid_t qi;
2043 	struct rte_port *port;
2044 	struct rte_ether_addr mac_addr;
2045 
2046 	if (port_id_is_invalid(pid, ENABLED_WARN))
2047 		return 0;
2048 
2049 	if(dcb_config)
2050 		dcb_test = 1;
2051 	RTE_ETH_FOREACH_DEV(pi) {
2052 		if (pid != pi && pid != (portid_t)RTE_PORT_ALL)
2053 			continue;
2054 
2055 		need_check_link_status = 0;
2056 		port = &ports[pi];
2057 		if (rte_atomic16_cmpset(&(port->port_status), RTE_PORT_STOPPED,
2058 						 RTE_PORT_HANDLING) == 0) {
2059 			printf("Port %d is now not stopped\n", pi);
2060 			continue;
2061 		}
2062 
2063 		if (port->need_reconfig > 0) {
2064 			port->need_reconfig = 0;
2065 
2066 			if (flow_isolate_all) {
2067 				int ret = port_flow_isolate(pi, 1);
2068 				if (ret) {
2069 					printf("Failed to apply isolated"
2070 					       " mode on port %d\n", pi);
2071 					return -1;
2072 				}
2073 			}
2074 			configure_rxtx_dump_callbacks(0);
2075 			printf("Configuring Port %d (socket %u)\n", pi,
2076 					port->socket_id);
2077 			/* configure port */
2078 			diag = rte_eth_dev_configure(pi, nb_rxq, nb_txq,
2079 						&(port->dev_conf));
2080 			if (diag != 0) {
2081 				if (rte_atomic16_cmpset(&(port->port_status),
2082 				RTE_PORT_HANDLING, RTE_PORT_STOPPED) == 0)
2083 					printf("Port %d can not be set back "
2084 							"to stopped\n", pi);
2085 				printf("Fail to configure port %d\n", pi);
2086 				/* try to reconfigure port next time */
2087 				port->need_reconfig = 1;
2088 				return -1;
2089 			}
2090 		}
2091 		if (port->need_reconfig_queues > 0) {
2092 			port->need_reconfig_queues = 0;
2093 			/* setup tx queues */
2094 			for (qi = 0; qi < nb_txq; qi++) {
2095 				if ((numa_support) &&
2096 					(txring_numa[pi] != NUMA_NO_CONFIG))
2097 					diag = rte_eth_tx_queue_setup(pi, qi,
2098 						port->nb_tx_desc[qi],
2099 						txring_numa[pi],
2100 						&(port->tx_conf[qi]));
2101 				else
2102 					diag = rte_eth_tx_queue_setup(pi, qi,
2103 						port->nb_tx_desc[qi],
2104 						port->socket_id,
2105 						&(port->tx_conf[qi]));
2106 
2107 				if (diag == 0)
2108 					continue;
2109 
2110 				/* Fail to setup tx queue, return */
2111 				if (rte_atomic16_cmpset(&(port->port_status),
2112 							RTE_PORT_HANDLING,
2113 							RTE_PORT_STOPPED) == 0)
2114 					printf("Port %d can not be set back "
2115 							"to stopped\n", pi);
2116 				printf("Fail to configure port %d tx queues\n",
2117 				       pi);
2118 				/* try to reconfigure queues next time */
2119 				port->need_reconfig_queues = 1;
2120 				return -1;
2121 			}
2122 			for (qi = 0; qi < nb_rxq; qi++) {
2123 				/* setup rx queues */
2124 				if ((numa_support) &&
2125 					(rxring_numa[pi] != NUMA_NO_CONFIG)) {
2126 					struct rte_mempool * mp =
2127 						mbuf_pool_find(rxring_numa[pi]);
2128 					if (mp == NULL) {
2129 						printf("Failed to setup RX queue:"
2130 							"No mempool allocation"
2131 							" on the socket %d\n",
2132 							rxring_numa[pi]);
2133 						return -1;
2134 					}
2135 
2136 					diag = rte_eth_rx_queue_setup(pi, qi,
2137 					     port->nb_rx_desc[qi],
2138 					     rxring_numa[pi],
2139 					     &(port->rx_conf[qi]),
2140 					     mp);
2141 				} else {
2142 					struct rte_mempool *mp =
2143 						mbuf_pool_find(port->socket_id);
2144 					if (mp == NULL) {
2145 						printf("Failed to setup RX queue:"
2146 							"No mempool allocation"
2147 							" on the socket %d\n",
2148 							port->socket_id);
2149 						return -1;
2150 					}
2151 					diag = rte_eth_rx_queue_setup(pi, qi,
2152 					     port->nb_rx_desc[qi],
2153 					     port->socket_id,
2154 					     &(port->rx_conf[qi]),
2155 					     mp);
2156 				}
2157 				if (diag == 0)
2158 					continue;
2159 
2160 				/* Fail to setup rx queue, return */
2161 				if (rte_atomic16_cmpset(&(port->port_status),
2162 							RTE_PORT_HANDLING,
2163 							RTE_PORT_STOPPED) == 0)
2164 					printf("Port %d can not be set back "
2165 							"to stopped\n", pi);
2166 				printf("Fail to configure port %d rx queues\n",
2167 				       pi);
2168 				/* try to reconfigure queues next time */
2169 				port->need_reconfig_queues = 1;
2170 				return -1;
2171 			}
2172 		}
2173 		configure_rxtx_dump_callbacks(verbose_level);
2174 		/* start port */
2175 		if (rte_eth_dev_start(pi) < 0) {
2176 			printf("Fail to start port %d\n", pi);
2177 
2178 			/* Fail to setup rx queue, return */
2179 			if (rte_atomic16_cmpset(&(port->port_status),
2180 				RTE_PORT_HANDLING, RTE_PORT_STOPPED) == 0)
2181 				printf("Port %d can not be set back to "
2182 							"stopped\n", pi);
2183 			continue;
2184 		}
2185 
2186 		if (rte_atomic16_cmpset(&(port->port_status),
2187 			RTE_PORT_HANDLING, RTE_PORT_STARTED) == 0)
2188 			printf("Port %d can not be set into started\n", pi);
2189 
2190 		rte_eth_macaddr_get(pi, &mac_addr);
2191 		printf("Port %d: %02X:%02X:%02X:%02X:%02X:%02X\n", pi,
2192 				mac_addr.addr_bytes[0], mac_addr.addr_bytes[1],
2193 				mac_addr.addr_bytes[2], mac_addr.addr_bytes[3],
2194 				mac_addr.addr_bytes[4], mac_addr.addr_bytes[5]);
2195 
2196 		/* at least one port started, need checking link status */
2197 		need_check_link_status = 1;
2198 	}
2199 
2200 	if (need_check_link_status == 1 && !no_link_check)
2201 		check_all_ports_link_status(RTE_PORT_ALL);
2202 	else if (need_check_link_status == 0)
2203 		printf("Please stop the ports first\n");
2204 
2205 	printf("Done\n");
2206 	return 0;
2207 }
2208 
2209 void
2210 stop_port(portid_t pid)
2211 {
2212 	portid_t pi;
2213 	struct rte_port *port;
2214 	int need_check_link_status = 0;
2215 
2216 	if (dcb_test) {
2217 		dcb_test = 0;
2218 		dcb_config = 0;
2219 	}
2220 
2221 	if (port_id_is_invalid(pid, ENABLED_WARN))
2222 		return;
2223 
2224 	printf("Stopping ports...\n");
2225 
2226 	RTE_ETH_FOREACH_DEV(pi) {
2227 		if (pid != pi && pid != (portid_t)RTE_PORT_ALL)
2228 			continue;
2229 
2230 		if (port_is_forwarding(pi) != 0 && test_done == 0) {
2231 			printf("Please remove port %d from forwarding configuration.\n", pi);
2232 			continue;
2233 		}
2234 
2235 		if (port_is_bonding_slave(pi)) {
2236 			printf("Please remove port %d from bonded device.\n", pi);
2237 			continue;
2238 		}
2239 
2240 		port = &ports[pi];
2241 		if (rte_atomic16_cmpset(&(port->port_status), RTE_PORT_STARTED,
2242 						RTE_PORT_HANDLING) == 0)
2243 			continue;
2244 
2245 		rte_eth_dev_stop(pi);
2246 
2247 		if (rte_atomic16_cmpset(&(port->port_status),
2248 			RTE_PORT_HANDLING, RTE_PORT_STOPPED) == 0)
2249 			printf("Port %d can not be set into stopped\n", pi);
2250 		need_check_link_status = 1;
2251 	}
2252 	if (need_check_link_status && !no_link_check)
2253 		check_all_ports_link_status(RTE_PORT_ALL);
2254 
2255 	printf("Done\n");
2256 }
2257 
2258 static void
2259 remove_invalid_ports_in(portid_t *array, portid_t *total)
2260 {
2261 	portid_t i;
2262 	portid_t new_total = 0;
2263 
2264 	for (i = 0; i < *total; i++)
2265 		if (!port_id_is_invalid(array[i], DISABLED_WARN)) {
2266 			array[new_total] = array[i];
2267 			new_total++;
2268 		}
2269 	*total = new_total;
2270 }
2271 
2272 static void
2273 remove_invalid_ports(void)
2274 {
2275 	remove_invalid_ports_in(ports_ids, &nb_ports);
2276 	remove_invalid_ports_in(fwd_ports_ids, &nb_fwd_ports);
2277 	nb_cfg_ports = nb_fwd_ports;
2278 }
2279 
2280 void
2281 close_port(portid_t pid)
2282 {
2283 	portid_t pi;
2284 	struct rte_port *port;
2285 
2286 	if (port_id_is_invalid(pid, ENABLED_WARN))
2287 		return;
2288 
2289 	printf("Closing ports...\n");
2290 
2291 	RTE_ETH_FOREACH_DEV(pi) {
2292 		if (pid != pi && pid != (portid_t)RTE_PORT_ALL)
2293 			continue;
2294 
2295 		if (port_is_forwarding(pi) != 0 && test_done == 0) {
2296 			printf("Please remove port %d from forwarding configuration.\n", pi);
2297 			continue;
2298 		}
2299 
2300 		if (port_is_bonding_slave(pi)) {
2301 			printf("Please remove port %d from bonded device.\n", pi);
2302 			continue;
2303 		}
2304 
2305 		port = &ports[pi];
2306 		if (rte_atomic16_cmpset(&(port->port_status),
2307 			RTE_PORT_CLOSED, RTE_PORT_CLOSED) == 1) {
2308 			printf("Port %d is already closed\n", pi);
2309 			continue;
2310 		}
2311 
2312 		if (rte_atomic16_cmpset(&(port->port_status),
2313 			RTE_PORT_STOPPED, RTE_PORT_HANDLING) == 0) {
2314 			printf("Port %d is now not stopped\n", pi);
2315 			continue;
2316 		}
2317 
2318 		if (port->flow_list)
2319 			port_flow_flush(pi);
2320 		rte_eth_dev_close(pi);
2321 
2322 		remove_invalid_ports();
2323 
2324 		if (rte_atomic16_cmpset(&(port->port_status),
2325 			RTE_PORT_HANDLING, RTE_PORT_CLOSED) == 0)
2326 			printf("Port %d cannot be set to closed\n", pi);
2327 	}
2328 
2329 	printf("Done\n");
2330 }
2331 
2332 void
2333 reset_port(portid_t pid)
2334 {
2335 	int diag;
2336 	portid_t pi;
2337 	struct rte_port *port;
2338 
2339 	if (port_id_is_invalid(pid, ENABLED_WARN))
2340 		return;
2341 
2342 	printf("Resetting ports...\n");
2343 
2344 	RTE_ETH_FOREACH_DEV(pi) {
2345 		if (pid != pi && pid != (portid_t)RTE_PORT_ALL)
2346 			continue;
2347 
2348 		if (port_is_forwarding(pi) != 0 && test_done == 0) {
2349 			printf("Please remove port %d from forwarding "
2350 			       "configuration.\n", pi);
2351 			continue;
2352 		}
2353 
2354 		if (port_is_bonding_slave(pi)) {
2355 			printf("Please remove port %d from bonded device.\n",
2356 			       pi);
2357 			continue;
2358 		}
2359 
2360 		diag = rte_eth_dev_reset(pi);
2361 		if (diag == 0) {
2362 			port = &ports[pi];
2363 			port->need_reconfig = 1;
2364 			port->need_reconfig_queues = 1;
2365 		} else {
2366 			printf("Failed to reset port %d. diag=%d\n", pi, diag);
2367 		}
2368 	}
2369 
2370 	printf("Done\n");
2371 }
2372 
2373 void
2374 attach_port(char *identifier)
2375 {
2376 	portid_t pi;
2377 	struct rte_dev_iterator iterator;
2378 
2379 	printf("Attaching a new port...\n");
2380 
2381 	if (identifier == NULL) {
2382 		printf("Invalid parameters are specified\n");
2383 		return;
2384 	}
2385 
2386 	if (rte_dev_probe(identifier) != 0) {
2387 		TESTPMD_LOG(ERR, "Failed to attach port %s\n", identifier);
2388 		return;
2389 	}
2390 
2391 	/* first attach mode: event */
2392 	if (setup_on_probe_event) {
2393 		/* new ports are detected on RTE_ETH_EVENT_NEW event */
2394 		for (pi = 0; pi < RTE_MAX_ETHPORTS; pi++)
2395 			if (ports[pi].port_status == RTE_PORT_HANDLING &&
2396 					ports[pi].need_setup != 0)
2397 				setup_attached_port(pi);
2398 		return;
2399 	}
2400 
2401 	/* second attach mode: iterator */
2402 	RTE_ETH_FOREACH_MATCHING_DEV(pi, identifier, &iterator) {
2403 		/* setup ports matching the devargs used for probing */
2404 		if (port_is_forwarding(pi))
2405 			continue; /* port was already attached before */
2406 		setup_attached_port(pi);
2407 	}
2408 }
2409 
2410 static void
2411 setup_attached_port(portid_t pi)
2412 {
2413 	unsigned int socket_id;
2414 
2415 	socket_id = (unsigned)rte_eth_dev_socket_id(pi);
2416 	/* if socket_id is invalid, set to the first available socket. */
2417 	if (check_socket_id(socket_id) < 0)
2418 		socket_id = socket_ids[0];
2419 	reconfig(pi, socket_id);
2420 	rte_eth_promiscuous_enable(pi);
2421 
2422 	ports_ids[nb_ports++] = pi;
2423 	fwd_ports_ids[nb_fwd_ports++] = pi;
2424 	nb_cfg_ports = nb_fwd_ports;
2425 	ports[pi].need_setup = 0;
2426 	ports[pi].port_status = RTE_PORT_STOPPED;
2427 
2428 	printf("Port %d is attached. Now total ports is %d\n", pi, nb_ports);
2429 	printf("Done\n");
2430 }
2431 
2432 void
2433 detach_port_device(portid_t port_id)
2434 {
2435 	struct rte_device *dev;
2436 	portid_t sibling;
2437 
2438 	printf("Removing a device...\n");
2439 
2440 	dev = rte_eth_devices[port_id].device;
2441 	if (dev == NULL) {
2442 		printf("Device already removed\n");
2443 		return;
2444 	}
2445 
2446 	if (ports[port_id].port_status != RTE_PORT_CLOSED) {
2447 		if (ports[port_id].port_status != RTE_PORT_STOPPED) {
2448 			printf("Port not stopped\n");
2449 			return;
2450 		}
2451 		printf("Port was not closed\n");
2452 		if (ports[port_id].flow_list)
2453 			port_flow_flush(port_id);
2454 	}
2455 
2456 	if (rte_dev_remove(dev) != 0) {
2457 		TESTPMD_LOG(ERR, "Failed to detach device %s\n", dev->name);
2458 		return;
2459 	}
2460 	RTE_ETH_FOREACH_DEV_OF(sibling, dev) {
2461 		/* reset mapping between old ports and removed device */
2462 		rte_eth_devices[sibling].device = NULL;
2463 		if (ports[sibling].port_status != RTE_PORT_CLOSED) {
2464 			/* sibling ports are forced to be closed */
2465 			ports[sibling].port_status = RTE_PORT_CLOSED;
2466 			printf("Port %u is closed\n", sibling);
2467 		}
2468 	}
2469 
2470 	remove_invalid_ports();
2471 
2472 	printf("Device of port %u is detached\n", port_id);
2473 	printf("Now total ports is %d\n", nb_ports);
2474 	printf("Done\n");
2475 	return;
2476 }
2477 
2478 void
2479 pmd_test_exit(void)
2480 {
2481 	portid_t pt_id;
2482 	int ret;
2483 	int i;
2484 
2485 	if (test_done == 0)
2486 		stop_packet_forwarding();
2487 
2488 	for (i = 0 ; i < RTE_MAX_NUMA_NODES ; i++) {
2489 		if (mempools[i]) {
2490 			if (mp_alloc_type == MP_ALLOC_ANON)
2491 				rte_mempool_mem_iter(mempools[i], dma_unmap_cb,
2492 						     NULL);
2493 		}
2494 	}
2495 	if (ports != NULL) {
2496 		no_link_check = 1;
2497 		RTE_ETH_FOREACH_DEV(pt_id) {
2498 			printf("\nStopping port %d...\n", pt_id);
2499 			fflush(stdout);
2500 			stop_port(pt_id);
2501 		}
2502 		RTE_ETH_FOREACH_DEV(pt_id) {
2503 			printf("\nShutting down port %d...\n", pt_id);
2504 			fflush(stdout);
2505 			close_port(pt_id);
2506 		}
2507 	}
2508 
2509 	if (hot_plug) {
2510 		ret = rte_dev_event_monitor_stop();
2511 		if (ret) {
2512 			RTE_LOG(ERR, EAL,
2513 				"fail to stop device event monitor.");
2514 			return;
2515 		}
2516 
2517 		ret = rte_dev_event_callback_unregister(NULL,
2518 			dev_event_callback, NULL);
2519 		if (ret < 0) {
2520 			RTE_LOG(ERR, EAL,
2521 				"fail to unregister device event callback.\n");
2522 			return;
2523 		}
2524 
2525 		ret = rte_dev_hotplug_handle_disable();
2526 		if (ret) {
2527 			RTE_LOG(ERR, EAL,
2528 				"fail to disable hotplug handling.\n");
2529 			return;
2530 		}
2531 	}
2532 	for (i = 0 ; i < RTE_MAX_NUMA_NODES ; i++) {
2533 		if (mempools[i])
2534 			rte_mempool_free(mempools[i]);
2535 	}
2536 
2537 	printf("\nBye...\n");
2538 }
2539 
2540 typedef void (*cmd_func_t)(void);
2541 struct pmd_test_command {
2542 	const char *cmd_name;
2543 	cmd_func_t cmd_func;
2544 };
2545 
2546 #define PMD_TEST_CMD_NB (sizeof(pmd_test_menu) / sizeof(pmd_test_menu[0]))
2547 
2548 /* Check the link status of all ports in up to 9s, and print them finally */
2549 static void
2550 check_all_ports_link_status(uint32_t port_mask)
2551 {
2552 #define CHECK_INTERVAL 100 /* 100ms */
2553 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
2554 	portid_t portid;
2555 	uint8_t count, all_ports_up, print_flag = 0;
2556 	struct rte_eth_link link;
2557 
2558 	printf("Checking link statuses...\n");
2559 	fflush(stdout);
2560 	for (count = 0; count <= MAX_CHECK_TIME; count++) {
2561 		all_ports_up = 1;
2562 		RTE_ETH_FOREACH_DEV(portid) {
2563 			if ((port_mask & (1 << portid)) == 0)
2564 				continue;
2565 			memset(&link, 0, sizeof(link));
2566 			rte_eth_link_get_nowait(portid, &link);
2567 			/* print link status if flag set */
2568 			if (print_flag == 1) {
2569 				if (link.link_status)
2570 					printf(
2571 					"Port%d Link Up. speed %u Mbps- %s\n",
2572 					portid, link.link_speed,
2573 				(link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
2574 					("full-duplex") : ("half-duplex\n"));
2575 				else
2576 					printf("Port %d Link Down\n", portid);
2577 				continue;
2578 			}
2579 			/* clear all_ports_up flag if any link down */
2580 			if (link.link_status == ETH_LINK_DOWN) {
2581 				all_ports_up = 0;
2582 				break;
2583 			}
2584 		}
2585 		/* after finally printing all link status, get out */
2586 		if (print_flag == 1)
2587 			break;
2588 
2589 		if (all_ports_up == 0) {
2590 			fflush(stdout);
2591 			rte_delay_ms(CHECK_INTERVAL);
2592 		}
2593 
2594 		/* set the print_flag if all ports up or timeout */
2595 		if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
2596 			print_flag = 1;
2597 		}
2598 
2599 		if (lsc_interrupt)
2600 			break;
2601 	}
2602 }
2603 
2604 /*
2605  * This callback is for remove a port for a device. It has limitation because
2606  * it is not for multiple port removal for a device.
2607  * TODO: the device detach invoke will plan to be removed from user side to
2608  * eal. And convert all PMDs to free port resources on ether device closing.
2609  */
2610 static void
2611 rmv_port_callback(void *arg)
2612 {
2613 	int need_to_start = 0;
2614 	int org_no_link_check = no_link_check;
2615 	portid_t port_id = (intptr_t)arg;
2616 
2617 	RTE_ETH_VALID_PORTID_OR_RET(port_id);
2618 
2619 	if (!test_done && port_is_forwarding(port_id)) {
2620 		need_to_start = 1;
2621 		stop_packet_forwarding();
2622 	}
2623 	no_link_check = 1;
2624 	stop_port(port_id);
2625 	no_link_check = org_no_link_check;
2626 	close_port(port_id);
2627 	detach_port_device(port_id);
2628 	if (need_to_start)
2629 		start_packet_forwarding(0);
2630 }
2631 
2632 /* This function is used by the interrupt thread */
2633 static int
2634 eth_event_callback(portid_t port_id, enum rte_eth_event_type type, void *param,
2635 		  void *ret_param)
2636 {
2637 	RTE_SET_USED(param);
2638 	RTE_SET_USED(ret_param);
2639 
2640 	if (type >= RTE_ETH_EVENT_MAX) {
2641 		fprintf(stderr, "\nPort %" PRIu16 ": %s called upon invalid event %d\n",
2642 			port_id, __func__, type);
2643 		fflush(stderr);
2644 	} else if (event_print_mask & (UINT32_C(1) << type)) {
2645 		printf("\nPort %" PRIu16 ": %s event\n", port_id,
2646 			eth_event_desc[type]);
2647 		fflush(stdout);
2648 	}
2649 
2650 	switch (type) {
2651 	case RTE_ETH_EVENT_NEW:
2652 		ports[port_id].need_setup = 1;
2653 		ports[port_id].port_status = RTE_PORT_HANDLING;
2654 		break;
2655 	case RTE_ETH_EVENT_INTR_RMV:
2656 		if (port_id_is_invalid(port_id, DISABLED_WARN))
2657 			break;
2658 		if (rte_eal_alarm_set(100000,
2659 				rmv_port_callback, (void *)(intptr_t)port_id))
2660 			fprintf(stderr, "Could not set up deferred device removal\n");
2661 		break;
2662 	default:
2663 		break;
2664 	}
2665 	return 0;
2666 }
2667 
2668 static int
2669 register_eth_event_callback(void)
2670 {
2671 	int ret;
2672 	enum rte_eth_event_type event;
2673 
2674 	for (event = RTE_ETH_EVENT_UNKNOWN;
2675 			event < RTE_ETH_EVENT_MAX; event++) {
2676 		ret = rte_eth_dev_callback_register(RTE_ETH_ALL,
2677 				event,
2678 				eth_event_callback,
2679 				NULL);
2680 		if (ret != 0) {
2681 			TESTPMD_LOG(ERR, "Failed to register callback for "
2682 					"%s event\n", eth_event_desc[event]);
2683 			return -1;
2684 		}
2685 	}
2686 
2687 	return 0;
2688 }
2689 
2690 /* This function is used by the interrupt thread */
2691 static void
2692 dev_event_callback(const char *device_name, enum rte_dev_event_type type,
2693 			     __rte_unused void *arg)
2694 {
2695 	uint16_t port_id;
2696 	int ret;
2697 
2698 	if (type >= RTE_DEV_EVENT_MAX) {
2699 		fprintf(stderr, "%s called upon invalid event %d\n",
2700 			__func__, type);
2701 		fflush(stderr);
2702 	}
2703 
2704 	switch (type) {
2705 	case RTE_DEV_EVENT_REMOVE:
2706 		RTE_LOG(DEBUG, EAL, "The device: %s has been removed!\n",
2707 			device_name);
2708 		ret = rte_eth_dev_get_port_by_name(device_name, &port_id);
2709 		if (ret) {
2710 			RTE_LOG(ERR, EAL, "can not get port by device %s!\n",
2711 				device_name);
2712 			return;
2713 		}
2714 		/*
2715 		 * Because the user's callback is invoked in eal interrupt
2716 		 * callback, the interrupt callback need to be finished before
2717 		 * it can be unregistered when detaching device. So finish
2718 		 * callback soon and use a deferred removal to detach device
2719 		 * is need. It is a workaround, once the device detaching be
2720 		 * moved into the eal in the future, the deferred removal could
2721 		 * be deleted.
2722 		 */
2723 		if (rte_eal_alarm_set(100000,
2724 				rmv_port_callback, (void *)(intptr_t)port_id))
2725 			RTE_LOG(ERR, EAL,
2726 				"Could not set up deferred device removal\n");
2727 		break;
2728 	case RTE_DEV_EVENT_ADD:
2729 		RTE_LOG(ERR, EAL, "The device: %s has been added!\n",
2730 			device_name);
2731 		/* TODO: After finish kernel driver binding,
2732 		 * begin to attach port.
2733 		 */
2734 		break;
2735 	default:
2736 		break;
2737 	}
2738 }
2739 
2740 static int
2741 set_tx_queue_stats_mapping_registers(portid_t port_id, struct rte_port *port)
2742 {
2743 	uint16_t i;
2744 	int diag;
2745 	uint8_t mapping_found = 0;
2746 
2747 	for (i = 0; i < nb_tx_queue_stats_mappings; i++) {
2748 		if ((tx_queue_stats_mappings[i].port_id == port_id) &&
2749 				(tx_queue_stats_mappings[i].queue_id < nb_txq )) {
2750 			diag = rte_eth_dev_set_tx_queue_stats_mapping(port_id,
2751 					tx_queue_stats_mappings[i].queue_id,
2752 					tx_queue_stats_mappings[i].stats_counter_id);
2753 			if (diag != 0)
2754 				return diag;
2755 			mapping_found = 1;
2756 		}
2757 	}
2758 	if (mapping_found)
2759 		port->tx_queue_stats_mapping_enabled = 1;
2760 	return 0;
2761 }
2762 
2763 static int
2764 set_rx_queue_stats_mapping_registers(portid_t port_id, struct rte_port *port)
2765 {
2766 	uint16_t i;
2767 	int diag;
2768 	uint8_t mapping_found = 0;
2769 
2770 	for (i = 0; i < nb_rx_queue_stats_mappings; i++) {
2771 		if ((rx_queue_stats_mappings[i].port_id == port_id) &&
2772 				(rx_queue_stats_mappings[i].queue_id < nb_rxq )) {
2773 			diag = rte_eth_dev_set_rx_queue_stats_mapping(port_id,
2774 					rx_queue_stats_mappings[i].queue_id,
2775 					rx_queue_stats_mappings[i].stats_counter_id);
2776 			if (diag != 0)
2777 				return diag;
2778 			mapping_found = 1;
2779 		}
2780 	}
2781 	if (mapping_found)
2782 		port->rx_queue_stats_mapping_enabled = 1;
2783 	return 0;
2784 }
2785 
2786 static void
2787 map_port_queue_stats_mapping_registers(portid_t pi, struct rte_port *port)
2788 {
2789 	int diag = 0;
2790 
2791 	diag = set_tx_queue_stats_mapping_registers(pi, port);
2792 	if (diag != 0) {
2793 		if (diag == -ENOTSUP) {
2794 			port->tx_queue_stats_mapping_enabled = 0;
2795 			printf("TX queue stats mapping not supported port id=%d\n", pi);
2796 		}
2797 		else
2798 			rte_exit(EXIT_FAILURE,
2799 					"set_tx_queue_stats_mapping_registers "
2800 					"failed for port id=%d diag=%d\n",
2801 					pi, diag);
2802 	}
2803 
2804 	diag = set_rx_queue_stats_mapping_registers(pi, port);
2805 	if (diag != 0) {
2806 		if (diag == -ENOTSUP) {
2807 			port->rx_queue_stats_mapping_enabled = 0;
2808 			printf("RX queue stats mapping not supported port id=%d\n", pi);
2809 		}
2810 		else
2811 			rte_exit(EXIT_FAILURE,
2812 					"set_rx_queue_stats_mapping_registers "
2813 					"failed for port id=%d diag=%d\n",
2814 					pi, diag);
2815 	}
2816 }
2817 
2818 static void
2819 rxtx_port_config(struct rte_port *port)
2820 {
2821 	uint16_t qid;
2822 	uint64_t offloads;
2823 
2824 	for (qid = 0; qid < nb_rxq; qid++) {
2825 		offloads = port->rx_conf[qid].offloads;
2826 		port->rx_conf[qid] = port->dev_info.default_rxconf;
2827 		port->rx_conf[qid].offloads |= offloads;
2828 
2829 		/* Check if any Rx parameters have been passed */
2830 		if (rx_pthresh != RTE_PMD_PARAM_UNSET)
2831 			port->rx_conf[qid].rx_thresh.pthresh = rx_pthresh;
2832 
2833 		if (rx_hthresh != RTE_PMD_PARAM_UNSET)
2834 			port->rx_conf[qid].rx_thresh.hthresh = rx_hthresh;
2835 
2836 		if (rx_wthresh != RTE_PMD_PARAM_UNSET)
2837 			port->rx_conf[qid].rx_thresh.wthresh = rx_wthresh;
2838 
2839 		if (rx_free_thresh != RTE_PMD_PARAM_UNSET)
2840 			port->rx_conf[qid].rx_free_thresh = rx_free_thresh;
2841 
2842 		if (rx_drop_en != RTE_PMD_PARAM_UNSET)
2843 			port->rx_conf[qid].rx_drop_en = rx_drop_en;
2844 
2845 		port->nb_rx_desc[qid] = nb_rxd;
2846 	}
2847 
2848 	for (qid = 0; qid < nb_txq; qid++) {
2849 		offloads = port->tx_conf[qid].offloads;
2850 		port->tx_conf[qid] = port->dev_info.default_txconf;
2851 		port->tx_conf[qid].offloads |= offloads;
2852 
2853 		/* Check if any Tx parameters have been passed */
2854 		if (tx_pthresh != RTE_PMD_PARAM_UNSET)
2855 			port->tx_conf[qid].tx_thresh.pthresh = tx_pthresh;
2856 
2857 		if (tx_hthresh != RTE_PMD_PARAM_UNSET)
2858 			port->tx_conf[qid].tx_thresh.hthresh = tx_hthresh;
2859 
2860 		if (tx_wthresh != RTE_PMD_PARAM_UNSET)
2861 			port->tx_conf[qid].tx_thresh.wthresh = tx_wthresh;
2862 
2863 		if (tx_rs_thresh != RTE_PMD_PARAM_UNSET)
2864 			port->tx_conf[qid].tx_rs_thresh = tx_rs_thresh;
2865 
2866 		if (tx_free_thresh != RTE_PMD_PARAM_UNSET)
2867 			port->tx_conf[qid].tx_free_thresh = tx_free_thresh;
2868 
2869 		port->nb_tx_desc[qid] = nb_txd;
2870 	}
2871 }
2872 
2873 void
2874 init_port_config(void)
2875 {
2876 	portid_t pid;
2877 	struct rte_port *port;
2878 
2879 	RTE_ETH_FOREACH_DEV(pid) {
2880 		port = &ports[pid];
2881 		port->dev_conf.fdir_conf = fdir_conf;
2882 		rte_eth_dev_info_get(pid, &port->dev_info);
2883 		if (nb_rxq > 1) {
2884 			port->dev_conf.rx_adv_conf.rss_conf.rss_key = NULL;
2885 			port->dev_conf.rx_adv_conf.rss_conf.rss_hf =
2886 				rss_hf & port->dev_info.flow_type_rss_offloads;
2887 		} else {
2888 			port->dev_conf.rx_adv_conf.rss_conf.rss_key = NULL;
2889 			port->dev_conf.rx_adv_conf.rss_conf.rss_hf = 0;
2890 		}
2891 
2892 		if (port->dcb_flag == 0) {
2893 			if( port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0)
2894 				port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_RSS;
2895 			else
2896 				port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE;
2897 		}
2898 
2899 		rxtx_port_config(port);
2900 
2901 		rte_eth_macaddr_get(pid, &port->eth_addr);
2902 
2903 		map_port_queue_stats_mapping_registers(pid, port);
2904 #if defined RTE_LIBRTE_IXGBE_PMD && defined RTE_LIBRTE_IXGBE_BYPASS
2905 		rte_pmd_ixgbe_bypass_init(pid);
2906 #endif
2907 
2908 		if (lsc_interrupt &&
2909 		    (rte_eth_devices[pid].data->dev_flags &
2910 		     RTE_ETH_DEV_INTR_LSC))
2911 			port->dev_conf.intr_conf.lsc = 1;
2912 		if (rmv_interrupt &&
2913 		    (rte_eth_devices[pid].data->dev_flags &
2914 		     RTE_ETH_DEV_INTR_RMV))
2915 			port->dev_conf.intr_conf.rmv = 1;
2916 	}
2917 }
2918 
2919 void set_port_slave_flag(portid_t slave_pid)
2920 {
2921 	struct rte_port *port;
2922 
2923 	port = &ports[slave_pid];
2924 	port->slave_flag = 1;
2925 }
2926 
2927 void clear_port_slave_flag(portid_t slave_pid)
2928 {
2929 	struct rte_port *port;
2930 
2931 	port = &ports[slave_pid];
2932 	port->slave_flag = 0;
2933 }
2934 
2935 uint8_t port_is_bonding_slave(portid_t slave_pid)
2936 {
2937 	struct rte_port *port;
2938 
2939 	port = &ports[slave_pid];
2940 	if ((rte_eth_devices[slave_pid].data->dev_flags &
2941 	    RTE_ETH_DEV_BONDED_SLAVE) || (port->slave_flag == 1))
2942 		return 1;
2943 	return 0;
2944 }
2945 
2946 const uint16_t vlan_tags[] = {
2947 		0,  1,  2,  3,  4,  5,  6,  7,
2948 		8,  9, 10, 11,  12, 13, 14, 15,
2949 		16, 17, 18, 19, 20, 21, 22, 23,
2950 		24, 25, 26, 27, 28, 29, 30, 31
2951 };
2952 
2953 static  int
2954 get_eth_dcb_conf(portid_t pid, struct rte_eth_conf *eth_conf,
2955 		 enum dcb_mode_enable dcb_mode,
2956 		 enum rte_eth_nb_tcs num_tcs,
2957 		 uint8_t pfc_en)
2958 {
2959 	uint8_t i;
2960 	int32_t rc;
2961 	struct rte_eth_rss_conf rss_conf;
2962 
2963 	/*
2964 	 * Builds up the correct configuration for dcb+vt based on the vlan tags array
2965 	 * given above, and the number of traffic classes available for use.
2966 	 */
2967 	if (dcb_mode == DCB_VT_ENABLED) {
2968 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
2969 				&eth_conf->rx_adv_conf.vmdq_dcb_conf;
2970 		struct rte_eth_vmdq_dcb_tx_conf *vmdq_tx_conf =
2971 				&eth_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2972 
2973 		/* VMDQ+DCB RX and TX configurations */
2974 		vmdq_rx_conf->enable_default_pool = 0;
2975 		vmdq_rx_conf->default_pool = 0;
2976 		vmdq_rx_conf->nb_queue_pools =
2977 			(num_tcs ==  ETH_4_TCS ? ETH_32_POOLS : ETH_16_POOLS);
2978 		vmdq_tx_conf->nb_queue_pools =
2979 			(num_tcs ==  ETH_4_TCS ? ETH_32_POOLS : ETH_16_POOLS);
2980 
2981 		vmdq_rx_conf->nb_pool_maps = vmdq_rx_conf->nb_queue_pools;
2982 		for (i = 0; i < vmdq_rx_conf->nb_pool_maps; i++) {
2983 			vmdq_rx_conf->pool_map[i].vlan_id = vlan_tags[i];
2984 			vmdq_rx_conf->pool_map[i].pools =
2985 				1 << (i % vmdq_rx_conf->nb_queue_pools);
2986 		}
2987 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++) {
2988 			vmdq_rx_conf->dcb_tc[i] = i % num_tcs;
2989 			vmdq_tx_conf->dcb_tc[i] = i % num_tcs;
2990 		}
2991 
2992 		/* set DCB mode of RX and TX of multiple queues */
2993 		eth_conf->rxmode.mq_mode = ETH_MQ_RX_VMDQ_DCB;
2994 		eth_conf->txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2995 	} else {
2996 		struct rte_eth_dcb_rx_conf *rx_conf =
2997 				&eth_conf->rx_adv_conf.dcb_rx_conf;
2998 		struct rte_eth_dcb_tx_conf *tx_conf =
2999 				&eth_conf->tx_adv_conf.dcb_tx_conf;
3000 
3001 		rc = rte_eth_dev_rss_hash_conf_get(pid, &rss_conf);
3002 		if (rc != 0)
3003 			return rc;
3004 
3005 		rx_conf->nb_tcs = num_tcs;
3006 		tx_conf->nb_tcs = num_tcs;
3007 
3008 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++) {
3009 			rx_conf->dcb_tc[i] = i % num_tcs;
3010 			tx_conf->dcb_tc[i] = i % num_tcs;
3011 		}
3012 
3013 		eth_conf->rxmode.mq_mode = ETH_MQ_RX_DCB_RSS;
3014 		eth_conf->rx_adv_conf.rss_conf = rss_conf;
3015 		eth_conf->txmode.mq_mode = ETH_MQ_TX_DCB;
3016 	}
3017 
3018 	if (pfc_en)
3019 		eth_conf->dcb_capability_en =
3020 				ETH_DCB_PG_SUPPORT | ETH_DCB_PFC_SUPPORT;
3021 	else
3022 		eth_conf->dcb_capability_en = ETH_DCB_PG_SUPPORT;
3023 
3024 	return 0;
3025 }
3026 
3027 int
3028 init_port_dcb_config(portid_t pid,
3029 		     enum dcb_mode_enable dcb_mode,
3030 		     enum rte_eth_nb_tcs num_tcs,
3031 		     uint8_t pfc_en)
3032 {
3033 	struct rte_eth_conf port_conf;
3034 	struct rte_port *rte_port;
3035 	int retval;
3036 	uint16_t i;
3037 
3038 	rte_port = &ports[pid];
3039 
3040 	memset(&port_conf, 0, sizeof(struct rte_eth_conf));
3041 	/* Enter DCB configuration status */
3042 	dcb_config = 1;
3043 
3044 	port_conf.rxmode = rte_port->dev_conf.rxmode;
3045 	port_conf.txmode = rte_port->dev_conf.txmode;
3046 
3047 	/*set configuration of DCB in vt mode and DCB in non-vt mode*/
3048 	retval = get_eth_dcb_conf(pid, &port_conf, dcb_mode, num_tcs, pfc_en);
3049 	if (retval < 0)
3050 		return retval;
3051 	port_conf.rxmode.offloads |= DEV_RX_OFFLOAD_VLAN_FILTER;
3052 
3053 	/* re-configure the device . */
3054 	retval = rte_eth_dev_configure(pid, nb_rxq, nb_rxq, &port_conf);
3055 	if (retval < 0)
3056 		return retval;
3057 	rte_eth_dev_info_get(pid, &rte_port->dev_info);
3058 
3059 	/* If dev_info.vmdq_pool_base is greater than 0,
3060 	 * the queue id of vmdq pools is started after pf queues.
3061 	 */
3062 	if (dcb_mode == DCB_VT_ENABLED &&
3063 	    rte_port->dev_info.vmdq_pool_base > 0) {
3064 		printf("VMDQ_DCB multi-queue mode is nonsensical"
3065 			" for port %d.", pid);
3066 		return -1;
3067 	}
3068 
3069 	/* Assume the ports in testpmd have the same dcb capability
3070 	 * and has the same number of rxq and txq in dcb mode
3071 	 */
3072 	if (dcb_mode == DCB_VT_ENABLED) {
3073 		if (rte_port->dev_info.max_vfs > 0) {
3074 			nb_rxq = rte_port->dev_info.nb_rx_queues;
3075 			nb_txq = rte_port->dev_info.nb_tx_queues;
3076 		} else {
3077 			nb_rxq = rte_port->dev_info.max_rx_queues;
3078 			nb_txq = rte_port->dev_info.max_tx_queues;
3079 		}
3080 	} else {
3081 		/*if vt is disabled, use all pf queues */
3082 		if (rte_port->dev_info.vmdq_pool_base == 0) {
3083 			nb_rxq = rte_port->dev_info.max_rx_queues;
3084 			nb_txq = rte_port->dev_info.max_tx_queues;
3085 		} else {
3086 			nb_rxq = (queueid_t)num_tcs;
3087 			nb_txq = (queueid_t)num_tcs;
3088 
3089 		}
3090 	}
3091 	rx_free_thresh = 64;
3092 
3093 	memcpy(&rte_port->dev_conf, &port_conf, sizeof(struct rte_eth_conf));
3094 
3095 	rxtx_port_config(rte_port);
3096 	/* VLAN filter */
3097 	rte_port->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_VLAN_FILTER;
3098 	for (i = 0; i < RTE_DIM(vlan_tags); i++)
3099 		rx_vft_set(pid, vlan_tags[i], 1);
3100 
3101 	rte_eth_macaddr_get(pid, &rte_port->eth_addr);
3102 	map_port_queue_stats_mapping_registers(pid, rte_port);
3103 
3104 	rte_port->dcb_flag = 1;
3105 
3106 	return 0;
3107 }
3108 
3109 static void
3110 init_port(void)
3111 {
3112 	/* Configuration of Ethernet ports. */
3113 	ports = rte_zmalloc("testpmd: ports",
3114 			    sizeof(struct rte_port) * RTE_MAX_ETHPORTS,
3115 			    RTE_CACHE_LINE_SIZE);
3116 	if (ports == NULL) {
3117 		rte_exit(EXIT_FAILURE,
3118 				"rte_zmalloc(%d struct rte_port) failed\n",
3119 				RTE_MAX_ETHPORTS);
3120 	}
3121 
3122 	/* Initialize ports NUMA structures */
3123 	memset(port_numa, NUMA_NO_CONFIG, RTE_MAX_ETHPORTS);
3124 	memset(rxring_numa, NUMA_NO_CONFIG, RTE_MAX_ETHPORTS);
3125 	memset(txring_numa, NUMA_NO_CONFIG, RTE_MAX_ETHPORTS);
3126 }
3127 
3128 static void
3129 force_quit(void)
3130 {
3131 	pmd_test_exit();
3132 	prompt_exit();
3133 }
3134 
3135 static void
3136 print_stats(void)
3137 {
3138 	uint8_t i;
3139 	const char clr[] = { 27, '[', '2', 'J', '\0' };
3140 	const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
3141 
3142 	/* Clear screen and move to top left */
3143 	printf("%s%s", clr, top_left);
3144 
3145 	printf("\nPort statistics ====================================");
3146 	for (i = 0; i < cur_fwd_config.nb_fwd_ports; i++)
3147 		nic_stats_display(fwd_ports_ids[i]);
3148 
3149 	fflush(stdout);
3150 }
3151 
3152 static void
3153 signal_handler(int signum)
3154 {
3155 	if (signum == SIGINT || signum == SIGTERM) {
3156 		printf("\nSignal %d received, preparing to exit...\n",
3157 				signum);
3158 #ifdef RTE_LIBRTE_PDUMP
3159 		/* uninitialize packet capture framework */
3160 		rte_pdump_uninit();
3161 #endif
3162 #ifdef RTE_LIBRTE_LATENCY_STATS
3163 		rte_latencystats_uninit();
3164 #endif
3165 		force_quit();
3166 		/* Set flag to indicate the force termination. */
3167 		f_quit = 1;
3168 		/* exit with the expected status */
3169 		signal(signum, SIG_DFL);
3170 		kill(getpid(), signum);
3171 	}
3172 }
3173 
3174 int
3175 main(int argc, char** argv)
3176 {
3177 	int diag;
3178 	portid_t port_id;
3179 	uint16_t count;
3180 	int ret;
3181 
3182 	signal(SIGINT, signal_handler);
3183 	signal(SIGTERM, signal_handler);
3184 
3185 	diag = rte_eal_init(argc, argv);
3186 	if (diag < 0)
3187 		rte_panic("Cannot init EAL\n");
3188 
3189 	testpmd_logtype = rte_log_register("testpmd");
3190 	if (testpmd_logtype < 0)
3191 		rte_panic("Cannot register log type");
3192 	rte_log_set_level(testpmd_logtype, RTE_LOG_DEBUG);
3193 
3194 	ret = register_eth_event_callback();
3195 	if (ret != 0)
3196 		rte_panic("Cannot register for ethdev events");
3197 
3198 #ifdef RTE_LIBRTE_PDUMP
3199 	/* initialize packet capture framework */
3200 	rte_pdump_init();
3201 #endif
3202 
3203 	count = 0;
3204 	RTE_ETH_FOREACH_DEV(port_id) {
3205 		ports_ids[count] = port_id;
3206 		count++;
3207 	}
3208 	nb_ports = (portid_t) count;
3209 	if (nb_ports == 0)
3210 		TESTPMD_LOG(WARNING, "No probed ethernet devices\n");
3211 
3212 	/* allocate port structures, and init them */
3213 	init_port();
3214 
3215 	set_def_fwd_config();
3216 	if (nb_lcores == 0)
3217 		rte_panic("Empty set of forwarding logical cores - check the "
3218 			  "core mask supplied in the command parameters\n");
3219 
3220 	/* Bitrate/latency stats disabled by default */
3221 #ifdef RTE_LIBRTE_BITRATE
3222 	bitrate_enabled = 0;
3223 #endif
3224 #ifdef RTE_LIBRTE_LATENCY_STATS
3225 	latencystats_enabled = 0;
3226 #endif
3227 
3228 	/* on FreeBSD, mlockall() is disabled by default */
3229 #ifdef RTE_EXEC_ENV_FREEBSD
3230 	do_mlockall = 0;
3231 #else
3232 	do_mlockall = 1;
3233 #endif
3234 
3235 	argc -= diag;
3236 	argv += diag;
3237 	if (argc > 1)
3238 		launch_args_parse(argc, argv);
3239 
3240 	if (do_mlockall && mlockall(MCL_CURRENT | MCL_FUTURE)) {
3241 		TESTPMD_LOG(NOTICE, "mlockall() failed with error \"%s\"\n",
3242 			strerror(errno));
3243 	}
3244 
3245 	if (tx_first && interactive)
3246 		rte_exit(EXIT_FAILURE, "--tx-first cannot be used on "
3247 				"interactive mode.\n");
3248 
3249 	if (tx_first && lsc_interrupt) {
3250 		printf("Warning: lsc_interrupt needs to be off when "
3251 				" using tx_first. Disabling.\n");
3252 		lsc_interrupt = 0;
3253 	}
3254 
3255 	if (!nb_rxq && !nb_txq)
3256 		printf("Warning: Either rx or tx queues should be non-zero\n");
3257 
3258 	if (nb_rxq > 1 && nb_rxq > nb_txq)
3259 		printf("Warning: nb_rxq=%d enables RSS configuration, "
3260 		       "but nb_txq=%d will prevent to fully test it.\n",
3261 		       nb_rxq, nb_txq);
3262 
3263 	init_config();
3264 
3265 	if (hot_plug) {
3266 		ret = rte_dev_hotplug_handle_enable();
3267 		if (ret) {
3268 			RTE_LOG(ERR, EAL,
3269 				"fail to enable hotplug handling.");
3270 			return -1;
3271 		}
3272 
3273 		ret = rte_dev_event_monitor_start();
3274 		if (ret) {
3275 			RTE_LOG(ERR, EAL,
3276 				"fail to start device event monitoring.");
3277 			return -1;
3278 		}
3279 
3280 		ret = rte_dev_event_callback_register(NULL,
3281 			dev_event_callback, NULL);
3282 		if (ret) {
3283 			RTE_LOG(ERR, EAL,
3284 				"fail  to register device event callback\n");
3285 			return -1;
3286 		}
3287 	}
3288 
3289 	if (start_port(RTE_PORT_ALL) != 0)
3290 		rte_exit(EXIT_FAILURE, "Start ports failed\n");
3291 
3292 	/* set all ports to promiscuous mode by default */
3293 	RTE_ETH_FOREACH_DEV(port_id)
3294 		rte_eth_promiscuous_enable(port_id);
3295 
3296 	/* Init metrics library */
3297 	rte_metrics_init(rte_socket_id());
3298 
3299 #ifdef RTE_LIBRTE_LATENCY_STATS
3300 	if (latencystats_enabled != 0) {
3301 		int ret = rte_latencystats_init(1, NULL);
3302 		if (ret)
3303 			printf("Warning: latencystats init()"
3304 				" returned error %d\n",	ret);
3305 		printf("Latencystats running on lcore %d\n",
3306 			latencystats_lcore_id);
3307 	}
3308 #endif
3309 
3310 	/* Setup bitrate stats */
3311 #ifdef RTE_LIBRTE_BITRATE
3312 	if (bitrate_enabled != 0) {
3313 		bitrate_data = rte_stats_bitrate_create();
3314 		if (bitrate_data == NULL)
3315 			rte_exit(EXIT_FAILURE,
3316 				"Could not allocate bitrate data.\n");
3317 		rte_stats_bitrate_reg(bitrate_data);
3318 	}
3319 #endif
3320 
3321 #ifdef RTE_LIBRTE_CMDLINE
3322 	if (strlen(cmdline_filename) != 0)
3323 		cmdline_read_from_file(cmdline_filename);
3324 
3325 	if (interactive == 1) {
3326 		if (auto_start) {
3327 			printf("Start automatic packet forwarding\n");
3328 			start_packet_forwarding(0);
3329 		}
3330 		prompt();
3331 		pmd_test_exit();
3332 	} else
3333 #endif
3334 	{
3335 		char c;
3336 		int rc;
3337 
3338 		f_quit = 0;
3339 
3340 		printf("No commandline core given, start packet forwarding\n");
3341 		start_packet_forwarding(tx_first);
3342 		if (stats_period != 0) {
3343 			uint64_t prev_time = 0, cur_time, diff_time = 0;
3344 			uint64_t timer_period;
3345 
3346 			/* Convert to number of cycles */
3347 			timer_period = stats_period * rte_get_timer_hz();
3348 
3349 			while (f_quit == 0) {
3350 				cur_time = rte_get_timer_cycles();
3351 				diff_time += cur_time - prev_time;
3352 
3353 				if (diff_time >= timer_period) {
3354 					print_stats();
3355 					/* Reset the timer */
3356 					diff_time = 0;
3357 				}
3358 				/* Sleep to avoid unnecessary checks */
3359 				prev_time = cur_time;
3360 				sleep(1);
3361 			}
3362 		}
3363 
3364 		printf("Press enter to exit\n");
3365 		rc = read(0, &c, 1);
3366 		pmd_test_exit();
3367 		if (rc < 0)
3368 			return 1;
3369 	}
3370 
3371 	return 0;
3372 }
3373