]>
Commit | Line | Data |
---|---|---|
7c673cae FG |
1 | /*- |
2 | * BSD LICENSE | |
3 | * | |
4 | * Copyright(c) 2010-2014 Intel Corporation. All rights reserved. | |
5 | * All rights reserved. | |
6 | * | |
7 | * Redistribution and use in source and binary forms, with or without | |
8 | * modification, are permitted provided that the following conditions | |
9 | * are met: | |
10 | * | |
11 | * * Redistributions of source code must retain the above copyright | |
12 | * notice, this list of conditions and the following disclaimer. | |
13 | * * Redistributions in binary form must reproduce the above copyright | |
14 | * notice, this list of conditions and the following disclaimer in | |
15 | * the documentation and/or other materials provided with the | |
16 | * distribution. | |
17 | * * Neither the name of Intel Corporation nor the names of its | |
18 | * contributors may be used to endorse or promote products derived | |
19 | * from this software without specific prior written permission. | |
20 | * | |
21 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
22 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
23 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
24 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
25 | * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
26 | * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
27 | * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
28 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
29 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
30 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
31 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
32 | */ | |
33 | ||
34 | #include <stdio.h> | |
35 | #include <stdlib.h> | |
36 | #include <stdint.h> | |
37 | #include <inttypes.h> | |
38 | #include <sys/types.h> | |
39 | #include <string.h> | |
40 | #include <sys/queue.h> | |
41 | #include <stdarg.h> | |
42 | #include <errno.h> | |
43 | #include <getopt.h> | |
44 | ||
45 | #include <rte_common.h> | |
46 | #include <rte_byteorder.h> | |
47 | #include <rte_log.h> | |
48 | #include <rte_memory.h> | |
49 | #include <rte_memcpy.h> | |
50 | #include <rte_memzone.h> | |
51 | #include <rte_eal.h> | |
52 | #include <rte_per_lcore.h> | |
53 | #include <rte_launch.h> | |
54 | #include <rte_atomic.h> | |
55 | #include <rte_cycles.h> | |
56 | #include <rte_prefetch.h> | |
57 | #include <rte_lcore.h> | |
58 | #include <rte_per_lcore.h> | |
59 | #include <rte_branch_prediction.h> | |
60 | #include <rte_interrupts.h> | |
61 | #include <rte_pci.h> | |
62 | #include <rte_random.h> | |
63 | #include <rte_debug.h> | |
64 | #include <rte_ether.h> | |
65 | #include <rte_ethdev.h> | |
66 | #include <rte_mempool.h> | |
67 | #include <rte_mbuf.h> | |
68 | #include <rte_malloc.h> | |
69 | #include <rte_fbk_hash.h> | |
70 | #include <rte_ip.h> | |
71 | ||
72 | #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1 | |
73 | ||
74 | #define MAX_PORTS 16 | |
75 | ||
76 | #define MCAST_CLONE_PORTS 2 | |
77 | #define MCAST_CLONE_SEGS 2 | |
78 | ||
79 | #define PKT_MBUF_DATA_SIZE RTE_MBUF_DEFAULT_BUF_SIZE | |
80 | #define NB_PKT_MBUF 8192 | |
81 | ||
82 | #define HDR_MBUF_DATA_SIZE (2 * RTE_PKTMBUF_HEADROOM) | |
83 | #define NB_HDR_MBUF (NB_PKT_MBUF * MAX_PORTS) | |
84 | ||
85 | #define NB_CLONE_MBUF (NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2) | |
86 | ||
87 | /* allow max jumbo frame 9.5 KB */ | |
88 | #define JUMBO_FRAME_MAX_SIZE 0x2600 | |
89 | ||
90 | #define MAX_PKT_BURST 32 | |
91 | #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */ | |
92 | ||
93 | /* Configure how many packets ahead to prefetch, when reading packets */ | |
94 | #define PREFETCH_OFFSET 3 | |
95 | ||
96 | /* | |
97 | * Construct Ethernet multicast address from IPv4 multicast address. | |
98 | * Citing RFC 1112, section 6.4: | |
99 | * "An IP host group address is mapped to an Ethernet multicast address | |
100 | * by placing the low-order 23-bits of the IP address into the low-order | |
101 | * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)." | |
102 | */ | |
103 | #define ETHER_ADDR_FOR_IPV4_MCAST(x) \ | |
104 | (rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16) | |
105 | ||
106 | /* | |
107 | * Configurable number of RX/TX ring descriptors | |
108 | */ | |
109 | #define RTE_TEST_RX_DESC_DEFAULT 128 | |
110 | #define RTE_TEST_TX_DESC_DEFAULT 512 | |
111 | static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; | |
112 | static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; | |
113 | ||
114 | /* ethernet addresses of ports */ | |
115 | static struct ether_addr ports_eth_addr[MAX_PORTS]; | |
116 | ||
117 | /* mask of enabled ports */ | |
118 | static uint32_t enabled_port_mask = 0; | |
119 | ||
120 | static uint8_t nb_ports = 0; | |
121 | ||
122 | static int rx_queue_per_lcore = 1; | |
123 | ||
124 | struct mbuf_table { | |
125 | uint16_t len; | |
126 | struct rte_mbuf *m_table[MAX_PKT_BURST]; | |
127 | }; | |
128 | ||
129 | #define MAX_RX_QUEUE_PER_LCORE 16 | |
130 | #define MAX_TX_QUEUE_PER_PORT 16 | |
131 | struct lcore_queue_conf { | |
132 | uint64_t tx_tsc; | |
133 | uint16_t n_rx_queue; | |
134 | uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE]; | |
135 | uint16_t tx_queue_id[MAX_PORTS]; | |
136 | struct mbuf_table tx_mbufs[MAX_PORTS]; | |
137 | } __rte_cache_aligned; | |
138 | static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE]; | |
139 | ||
140 | static const struct rte_eth_conf port_conf = { | |
141 | .rxmode = { | |
142 | .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE, | |
143 | .split_hdr_size = 0, | |
144 | .header_split = 0, /**< Header Split disabled */ | |
145 | .hw_ip_checksum = 0, /**< IP checksum offload disabled */ | |
146 | .hw_vlan_filter = 0, /**< VLAN filtering disabled */ | |
147 | .jumbo_frame = 1, /**< Jumbo Frame Support enabled */ | |
148 | .hw_strip_crc = 0, /**< CRC stripped by hardware */ | |
149 | }, | |
150 | .txmode = { | |
151 | .mq_mode = ETH_MQ_TX_NONE, | |
152 | }, | |
153 | }; | |
154 | ||
155 | static struct rte_mempool *packet_pool, *header_pool, *clone_pool; | |
156 | ||
157 | ||
158 | /* Multicast */ | |
159 | static struct rte_fbk_hash_params mcast_hash_params = { | |
160 | .name = "MCAST_HASH", | |
161 | .entries = 1024, | |
162 | .entries_per_bucket = 4, | |
163 | .socket_id = 0, | |
164 | .hash_func = NULL, | |
165 | .init_val = 0, | |
166 | }; | |
167 | ||
168 | struct rte_fbk_hash_table *mcast_hash = NULL; | |
169 | ||
170 | struct mcast_group_params { | |
171 | uint32_t ip; | |
172 | uint16_t port_mask; | |
173 | }; | |
174 | ||
175 | static struct mcast_group_params mcast_group_table[] = { | |
176 | {IPv4(224,0,0,101), 0x1}, | |
177 | {IPv4(224,0,0,102), 0x2}, | |
178 | {IPv4(224,0,0,103), 0x3}, | |
179 | {IPv4(224,0,0,104), 0x4}, | |
180 | {IPv4(224,0,0,105), 0x5}, | |
181 | {IPv4(224,0,0,106), 0x6}, | |
182 | {IPv4(224,0,0,107), 0x7}, | |
183 | {IPv4(224,0,0,108), 0x8}, | |
184 | {IPv4(224,0,0,109), 0x9}, | |
185 | {IPv4(224,0,0,110), 0xA}, | |
186 | {IPv4(224,0,0,111), 0xB}, | |
187 | {IPv4(224,0,0,112), 0xC}, | |
188 | {IPv4(224,0,0,113), 0xD}, | |
189 | {IPv4(224,0,0,114), 0xE}, | |
190 | {IPv4(224,0,0,115), 0xF}, | |
191 | }; | |
192 | ||
193 | #define N_MCAST_GROUPS \ | |
194 | (sizeof (mcast_group_table) / sizeof (mcast_group_table[0])) | |
195 | ||
196 | ||
197 | /* Send burst of packets on an output interface */ | |
198 | static void | |
199 | send_burst(struct lcore_queue_conf *qconf, uint8_t port) | |
200 | { | |
201 | struct rte_mbuf **m_table; | |
202 | uint16_t n, queueid; | |
203 | int ret; | |
204 | ||
205 | queueid = qconf->tx_queue_id[port]; | |
206 | m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table; | |
207 | n = qconf->tx_mbufs[port].len; | |
208 | ||
209 | ret = rte_eth_tx_burst(port, queueid, m_table, n); | |
210 | while (unlikely (ret < n)) { | |
211 | rte_pktmbuf_free(m_table[ret]); | |
212 | ret++; | |
213 | } | |
214 | ||
215 | qconf->tx_mbufs[port].len = 0; | |
216 | } | |
217 | ||
218 | /* Get number of bits set. */ | |
219 | static inline uint32_t | |
220 | bitcnt(uint32_t v) | |
221 | { | |
222 | uint32_t n; | |
223 | ||
224 | for (n = 0; v != 0; v &= v - 1, n++) | |
225 | ; | |
226 | ||
227 | return n; | |
228 | } | |
229 | ||
230 | /** | |
231 | * Create the output multicast packet based on the given input packet. | |
232 | * There are two approaches for creating outgoing packet, though both | |
233 | * are based on data zero-copy idea, they differ in few details: | |
234 | * First one creates a clone of the input packet, e.g - walk though all | |
235 | * segments of the input packet, and for each of them create a new packet | |
236 | * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone() | |
237 | * for more details). Then new mbuf is allocated for the packet header | |
238 | * and is prepended to the 'clone' mbuf. | |
239 | * Second approach doesn't make a clone, it just increment refcnt for all | |
240 | * input packet segments. Then it allocates new mbuf for the packet header | |
241 | * and prepends it to the input packet. | |
242 | * Basically first approach reuses only input packet's data, but creates | |
243 | * it's own copy of packet's metadata. Second approach reuses both input's | |
244 | * packet data and metadata. | |
245 | * The advantage of first approach - is that each outgoing packet has it's | |
246 | * own copy of metadata, so we can safely modify data pointer of the | |
247 | * input packet. That allows us to skip creation if the output packet for | |
248 | * the last destination port, but instead modify input packet's header inplace, | |
249 | * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times. | |
250 | * The advantage of second approach - less work for each outgoing packet, | |
251 | * e.g: we skip "clone" operation completely. Though it comes with a price - | |
252 | * input packet's metadata has to be intact. So for N destination ports we | |
253 | * need to invoke mcast_out_pkt N times. | |
254 | * So for small number of outgoing ports (and segments in the input packet) | |
255 | * first approach will be faster. | |
256 | * As number of outgoing ports (and/or input segments) will grow, | |
257 | * second way will become more preferable. | |
258 | * | |
259 | * @param pkt | |
260 | * Input packet mbuf. | |
261 | * @param use_clone | |
262 | * Control which of the two approaches described above should be used: | |
263 | * - 0 - use second approach: | |
264 | * Don't "clone" input packet. | |
265 | * Prepend new header directly to the input packet | |
266 | * - 1 - use first approach: | |
267 | * Make a "clone" of input packet first. | |
268 | * Prepend new header to the clone of the input packet | |
269 | * @return | |
270 | * - The pointer to the new outgoing packet. | |
271 | * - NULL if operation failed. | |
272 | */ | |
273 | static inline struct rte_mbuf * | |
274 | mcast_out_pkt(struct rte_mbuf *pkt, int use_clone) | |
275 | { | |
276 | struct rte_mbuf *hdr; | |
277 | ||
278 | /* Create new mbuf for the header. */ | |
279 | if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL)) | |
280 | return NULL; | |
281 | ||
282 | /* If requested, then make a new clone packet. */ | |
283 | if (use_clone != 0 && | |
284 | unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) { | |
285 | rte_pktmbuf_free(hdr); | |
286 | return NULL; | |
287 | } | |
288 | ||
289 | /* prepend new header */ | |
290 | hdr->next = pkt; | |
291 | ||
292 | ||
293 | /* update header's fields */ | |
294 | hdr->pkt_len = (uint16_t)(hdr->data_len + pkt->pkt_len); | |
295 | hdr->nb_segs = (uint8_t)(pkt->nb_segs + 1); | |
296 | ||
297 | /* copy metadata from source packet*/ | |
298 | hdr->port = pkt->port; | |
299 | hdr->vlan_tci = pkt->vlan_tci; | |
300 | hdr->vlan_tci_outer = pkt->vlan_tci_outer; | |
301 | hdr->tx_offload = pkt->tx_offload; | |
302 | hdr->hash = pkt->hash; | |
303 | ||
304 | hdr->ol_flags = pkt->ol_flags; | |
305 | ||
306 | __rte_mbuf_sanity_check(hdr, 1); | |
307 | return hdr; | |
308 | } | |
309 | ||
310 | /* | |
311 | * Write new Ethernet header to the outgoing packet, | |
312 | * and put it into the outgoing queue for the given port. | |
313 | */ | |
314 | static inline void | |
315 | mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr, | |
316 | struct lcore_queue_conf *qconf, uint8_t port) | |
317 | { | |
318 | struct ether_hdr *ethdr; | |
319 | uint16_t len; | |
320 | ||
321 | /* Construct Ethernet header. */ | |
322 | ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr)); | |
323 | RTE_ASSERT(ethdr != NULL); | |
324 | ||
325 | ether_addr_copy(dest_addr, ðdr->d_addr); | |
326 | ether_addr_copy(&ports_eth_addr[port], ðdr->s_addr); | |
327 | ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4); | |
328 | ||
329 | /* Put new packet into the output queue */ | |
330 | len = qconf->tx_mbufs[port].len; | |
331 | qconf->tx_mbufs[port].m_table[len] = pkt; | |
332 | qconf->tx_mbufs[port].len = ++len; | |
333 | ||
334 | /* Transmit packets */ | |
335 | if (unlikely(MAX_PKT_BURST == len)) | |
336 | send_burst(qconf, port); | |
337 | } | |
338 | ||
339 | /* Multicast forward of the input packet */ | |
340 | static inline void | |
341 | mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf) | |
342 | { | |
343 | struct rte_mbuf *mc; | |
344 | struct ipv4_hdr *iphdr; | |
345 | uint32_t dest_addr, port_mask, port_num, use_clone; | |
346 | int32_t hash; | |
347 | uint8_t port; | |
348 | union { | |
349 | uint64_t as_int; | |
350 | struct ether_addr as_addr; | |
351 | } dst_eth_addr; | |
352 | ||
353 | /* Remove the Ethernet header from the input packet */ | |
354 | iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr)); | |
355 | RTE_ASSERT(iphdr != NULL); | |
356 | ||
357 | dest_addr = rte_be_to_cpu_32(iphdr->dst_addr); | |
358 | ||
359 | /* | |
360 | * Check that it is a valid multicast address and | |
361 | * we have some active ports assigned to it. | |
362 | */ | |
363 | if(!IS_IPV4_MCAST(dest_addr) || | |
364 | (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 || | |
365 | (port_mask = hash & enabled_port_mask) == 0) { | |
366 | rte_pktmbuf_free(m); | |
367 | return; | |
368 | } | |
369 | ||
370 | /* Calculate number of destination ports. */ | |
371 | port_num = bitcnt(port_mask); | |
372 | ||
373 | /* Should we use rte_pktmbuf_clone() or not. */ | |
374 | use_clone = (port_num <= MCAST_CLONE_PORTS && | |
375 | m->nb_segs <= MCAST_CLONE_SEGS); | |
376 | ||
377 | /* Mark all packet's segments as referenced port_num times */ | |
378 | if (use_clone == 0) | |
379 | rte_pktmbuf_refcnt_update(m, (uint16_t)port_num); | |
380 | ||
381 | /* construct destination ethernet address */ | |
382 | dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr); | |
383 | ||
384 | for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) { | |
385 | ||
386 | /* Prepare output packet and send it out. */ | |
387 | if ((port_mask & 1) != 0) { | |
388 | if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL)) | |
389 | mcast_send_pkt(mc, &dst_eth_addr.as_addr, | |
390 | qconf, port); | |
391 | else if (use_clone == 0) | |
392 | rte_pktmbuf_free(m); | |
393 | } | |
394 | } | |
395 | ||
396 | /* | |
397 | * If we making clone packets, then, for the last destination port, | |
398 | * we can overwrite input packet's metadata. | |
399 | */ | |
400 | if (use_clone != 0) | |
401 | mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port); | |
402 | else | |
403 | rte_pktmbuf_free(m); | |
404 | } | |
405 | ||
406 | /* Send burst of outgoing packet, if timeout expires. */ | |
407 | static inline void | |
408 | send_timeout_burst(struct lcore_queue_conf *qconf) | |
409 | { | |
410 | uint64_t cur_tsc; | |
411 | uint8_t portid; | |
412 | const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US; | |
413 | ||
414 | cur_tsc = rte_rdtsc(); | |
415 | if (likely (cur_tsc < qconf->tx_tsc + drain_tsc)) | |
416 | return; | |
417 | ||
418 | for (portid = 0; portid < MAX_PORTS; portid++) { | |
419 | if (qconf->tx_mbufs[portid].len != 0) | |
420 | send_burst(qconf, portid); | |
421 | } | |
422 | qconf->tx_tsc = cur_tsc; | |
423 | } | |
424 | ||
425 | /* main processing loop */ | |
426 | static int | |
427 | main_loop(__rte_unused void *dummy) | |
428 | { | |
429 | struct rte_mbuf *pkts_burst[MAX_PKT_BURST]; | |
430 | unsigned lcore_id; | |
431 | int i, j, nb_rx; | |
432 | uint8_t portid; | |
433 | struct lcore_queue_conf *qconf; | |
434 | ||
435 | lcore_id = rte_lcore_id(); | |
436 | qconf = &lcore_queue_conf[lcore_id]; | |
437 | ||
438 | ||
439 | if (qconf->n_rx_queue == 0) { | |
440 | RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n", | |
441 | lcore_id); | |
442 | return 0; | |
443 | } | |
444 | ||
445 | RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n", | |
446 | lcore_id); | |
447 | ||
448 | for (i = 0; i < qconf->n_rx_queue; i++) { | |
449 | ||
450 | portid = qconf->rx_queue_list[i]; | |
451 | RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n", | |
452 | lcore_id, (int) portid); | |
453 | } | |
454 | ||
455 | while (1) { | |
456 | ||
457 | /* | |
458 | * Read packet from RX queues | |
459 | */ | |
460 | for (i = 0; i < qconf->n_rx_queue; i++) { | |
461 | ||
462 | portid = qconf->rx_queue_list[i]; | |
463 | nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst, | |
464 | MAX_PKT_BURST); | |
465 | ||
466 | /* Prefetch first packets */ | |
467 | for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) { | |
468 | rte_prefetch0(rte_pktmbuf_mtod( | |
469 | pkts_burst[j], void *)); | |
470 | } | |
471 | ||
472 | /* Prefetch and forward already prefetched packets */ | |
473 | for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) { | |
474 | rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[ | |
475 | j + PREFETCH_OFFSET], void *)); | |
476 | mcast_forward(pkts_burst[j], qconf); | |
477 | } | |
478 | ||
479 | /* Forward remaining prefetched packets */ | |
480 | for (; j < nb_rx; j++) { | |
481 | mcast_forward(pkts_burst[j], qconf); | |
482 | } | |
483 | } | |
484 | ||
485 | /* Send out packets from TX queues */ | |
486 | send_timeout_burst(qconf); | |
487 | } | |
488 | } | |
489 | ||
490 | /* display usage */ | |
491 | static void | |
492 | print_usage(const char *prgname) | |
493 | { | |
494 | printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n" | |
495 | " -p PORTMASK: hexadecimal bitmask of ports to configure\n" | |
496 | " -q NQ: number of queue (=ports) per lcore (default is 1)\n", | |
497 | prgname); | |
498 | } | |
499 | ||
500 | static uint32_t | |
501 | parse_portmask(const char *portmask) | |
502 | { | |
503 | char *end = NULL; | |
504 | unsigned long pm; | |
505 | ||
506 | /* parse hexadecimal string */ | |
507 | pm = strtoul(portmask, &end, 16); | |
508 | if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0')) | |
509 | return 0; | |
510 | ||
511 | return (uint32_t)pm; | |
512 | } | |
513 | ||
514 | static int | |
515 | parse_nqueue(const char *q_arg) | |
516 | { | |
517 | char *end = NULL; | |
518 | unsigned long n; | |
519 | ||
520 | /* parse numerical string */ | |
521 | errno = 0; | |
522 | n = strtoul(q_arg, &end, 0); | |
523 | if (errno != 0 || end == NULL || *end != '\0' || | |
524 | n == 0 || n >= MAX_RX_QUEUE_PER_LCORE) | |
525 | return -1; | |
526 | ||
527 | return n; | |
528 | } | |
529 | ||
530 | /* Parse the argument given in the command line of the application */ | |
531 | static int | |
532 | parse_args(int argc, char **argv) | |
533 | { | |
534 | int opt, ret; | |
535 | char **argvopt; | |
536 | int option_index; | |
537 | char *prgname = argv[0]; | |
538 | static struct option lgopts[] = { | |
539 | {NULL, 0, 0, 0} | |
540 | }; | |
541 | ||
542 | argvopt = argv; | |
543 | ||
544 | while ((opt = getopt_long(argc, argvopt, "p:q:", | |
545 | lgopts, &option_index)) != EOF) { | |
546 | ||
547 | switch (opt) { | |
548 | /* portmask */ | |
549 | case 'p': | |
550 | enabled_port_mask = parse_portmask(optarg); | |
551 | if (enabled_port_mask == 0) { | |
552 | printf("invalid portmask\n"); | |
553 | print_usage(prgname); | |
554 | return -1; | |
555 | } | |
556 | break; | |
557 | ||
558 | /* nqueue */ | |
559 | case 'q': | |
560 | rx_queue_per_lcore = parse_nqueue(optarg); | |
561 | if (rx_queue_per_lcore < 0) { | |
562 | printf("invalid queue number\n"); | |
563 | print_usage(prgname); | |
564 | return -1; | |
565 | } | |
566 | break; | |
567 | ||
568 | default: | |
569 | print_usage(prgname); | |
570 | return -1; | |
571 | } | |
572 | } | |
573 | ||
574 | if (optind >= 0) | |
575 | argv[optind-1] = prgname; | |
576 | ||
577 | ret = optind-1; | |
578 | optind = 0; /* reset getopt lib */ | |
579 | return ret; | |
580 | } | |
581 | ||
582 | static void | |
583 | print_ethaddr(const char *name, struct ether_addr *eth_addr) | |
584 | { | |
585 | char buf[ETHER_ADDR_FMT_SIZE]; | |
586 | ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr); | |
587 | printf("%s%s", name, buf); | |
588 | } | |
589 | ||
590 | static int | |
591 | init_mcast_hash(void) | |
592 | { | |
593 | uint32_t i; | |
594 | ||
595 | mcast_hash_params.socket_id = rte_socket_id(); | |
596 | mcast_hash = rte_fbk_hash_create(&mcast_hash_params); | |
597 | if (mcast_hash == NULL){ | |
598 | return -1; | |
599 | } | |
600 | ||
601 | for (i = 0; i < N_MCAST_GROUPS; i ++){ | |
602 | if (rte_fbk_hash_add_key(mcast_hash, | |
603 | mcast_group_table[i].ip, | |
604 | mcast_group_table[i].port_mask) < 0) { | |
605 | return -1; | |
606 | } | |
607 | } | |
608 | ||
609 | return 0; | |
610 | } | |
611 | ||
612 | /* Check the link status of all ports in up to 9s, and print them finally */ | |
613 | static void | |
614 | check_all_ports_link_status(uint8_t port_num, uint32_t port_mask) | |
615 | { | |
616 | #define CHECK_INTERVAL 100 /* 100ms */ | |
617 | #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */ | |
618 | uint8_t portid, count, all_ports_up, print_flag = 0; | |
619 | struct rte_eth_link link; | |
620 | ||
621 | printf("\nChecking link status"); | |
622 | fflush(stdout); | |
623 | for (count = 0; count <= MAX_CHECK_TIME; count++) { | |
624 | all_ports_up = 1; | |
625 | for (portid = 0; portid < port_num; portid++) { | |
626 | if ((port_mask & (1 << portid)) == 0) | |
627 | continue; | |
628 | memset(&link, 0, sizeof(link)); | |
629 | rte_eth_link_get_nowait(portid, &link); | |
630 | /* print link status if flag set */ | |
631 | if (print_flag == 1) { | |
632 | if (link.link_status) | |
633 | printf("Port %d Link Up - speed %u " | |
634 | "Mbps - %s\n", (uint8_t)portid, | |
635 | (unsigned)link.link_speed, | |
636 | (link.link_duplex == ETH_LINK_FULL_DUPLEX) ? | |
637 | ("full-duplex") : ("half-duplex\n")); | |
638 | else | |
639 | printf("Port %d Link Down\n", | |
640 | (uint8_t)portid); | |
641 | continue; | |
642 | } | |
643 | /* clear all_ports_up flag if any link down */ | |
644 | if (link.link_status == ETH_LINK_DOWN) { | |
645 | all_ports_up = 0; | |
646 | break; | |
647 | } | |
648 | } | |
649 | /* after finally printing all link status, get out */ | |
650 | if (print_flag == 1) | |
651 | break; | |
652 | ||
653 | if (all_ports_up == 0) { | |
654 | printf("."); | |
655 | fflush(stdout); | |
656 | rte_delay_ms(CHECK_INTERVAL); | |
657 | } | |
658 | ||
659 | /* set the print_flag if all ports up or timeout */ | |
660 | if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) { | |
661 | print_flag = 1; | |
662 | printf("done\n"); | |
663 | } | |
664 | } | |
665 | } | |
666 | ||
667 | int | |
668 | main(int argc, char **argv) | |
669 | { | |
670 | struct lcore_queue_conf *qconf; | |
671 | struct rte_eth_dev_info dev_info; | |
672 | struct rte_eth_txconf *txconf; | |
673 | int ret; | |
674 | uint16_t queueid; | |
675 | unsigned lcore_id = 0, rx_lcore_id = 0; | |
676 | uint32_t n_tx_queue, nb_lcores; | |
677 | uint8_t portid; | |
678 | ||
679 | /* init EAL */ | |
680 | ret = rte_eal_init(argc, argv); | |
681 | if (ret < 0) | |
682 | rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n"); | |
683 | argc -= ret; | |
684 | argv += ret; | |
685 | ||
686 | /* parse application arguments (after the EAL ones) */ | |
687 | ret = parse_args(argc, argv); | |
688 | if (ret < 0) | |
689 | rte_exit(EXIT_FAILURE, "Invalid IPV4_MULTICAST parameters\n"); | |
690 | ||
691 | /* create the mbuf pools */ | |
692 | packet_pool = rte_pktmbuf_pool_create("packet_pool", NB_PKT_MBUF, 32, | |
693 | 0, PKT_MBUF_DATA_SIZE, rte_socket_id()); | |
694 | ||
695 | if (packet_pool == NULL) | |
696 | rte_exit(EXIT_FAILURE, "Cannot init packet mbuf pool\n"); | |
697 | ||
698 | header_pool = rte_pktmbuf_pool_create("header_pool", NB_HDR_MBUF, 32, | |
699 | 0, HDR_MBUF_DATA_SIZE, rte_socket_id()); | |
700 | ||
701 | if (header_pool == NULL) | |
702 | rte_exit(EXIT_FAILURE, "Cannot init header mbuf pool\n"); | |
703 | ||
704 | clone_pool = rte_pktmbuf_pool_create("clone_pool", NB_CLONE_MBUF, 32, | |
705 | 0, 0, rte_socket_id()); | |
706 | ||
707 | if (clone_pool == NULL) | |
708 | rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n"); | |
709 | ||
710 | nb_ports = rte_eth_dev_count(); | |
711 | if (nb_ports == 0) | |
712 | rte_exit(EXIT_FAILURE, "No physical ports!\n"); | |
713 | if (nb_ports > MAX_PORTS) | |
714 | nb_ports = MAX_PORTS; | |
715 | ||
716 | nb_lcores = rte_lcore_count(); | |
717 | ||
718 | /* initialize all ports */ | |
719 | for (portid = 0; portid < nb_ports; portid++) { | |
720 | /* skip ports that are not enabled */ | |
721 | if ((enabled_port_mask & (1 << portid)) == 0) { | |
722 | printf("Skipping disabled port %d\n", portid); | |
723 | continue; | |
724 | } | |
725 | ||
726 | qconf = &lcore_queue_conf[rx_lcore_id]; | |
727 | ||
728 | /* get the lcore_id for this port */ | |
729 | while (rte_lcore_is_enabled(rx_lcore_id) == 0 || | |
730 | qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) { | |
731 | ||
732 | rx_lcore_id ++; | |
733 | qconf = &lcore_queue_conf[rx_lcore_id]; | |
734 | ||
735 | if (rx_lcore_id >= RTE_MAX_LCORE) | |
736 | rte_exit(EXIT_FAILURE, "Not enough cores\n"); | |
737 | } | |
738 | qconf->rx_queue_list[qconf->n_rx_queue] = portid; | |
739 | qconf->n_rx_queue++; | |
740 | ||
741 | /* init port */ | |
742 | printf("Initializing port %d on lcore %u... ", portid, | |
743 | rx_lcore_id); | |
744 | fflush(stdout); | |
745 | ||
746 | n_tx_queue = nb_lcores; | |
747 | if (n_tx_queue > MAX_TX_QUEUE_PER_PORT) | |
748 | n_tx_queue = MAX_TX_QUEUE_PER_PORT; | |
749 | ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue, | |
750 | &port_conf); | |
751 | if (ret < 0) | |
752 | rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n", | |
753 | ret, portid); | |
754 | ||
755 | rte_eth_macaddr_get(portid, &ports_eth_addr[portid]); | |
756 | print_ethaddr(" Address:", &ports_eth_addr[portid]); | |
757 | printf(", "); | |
758 | ||
759 | /* init one RX queue */ | |
760 | queueid = 0; | |
761 | printf("rxq=%hu ", queueid); | |
762 | fflush(stdout); | |
763 | ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd, | |
764 | rte_eth_dev_socket_id(portid), | |
765 | NULL, | |
766 | packet_pool); | |
767 | if (ret < 0) | |
768 | rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n", | |
769 | ret, portid); | |
770 | ||
771 | /* init one TX queue per couple (lcore,port) */ | |
772 | queueid = 0; | |
773 | ||
774 | RTE_LCORE_FOREACH(lcore_id) { | |
775 | if (rte_lcore_is_enabled(lcore_id) == 0) | |
776 | continue; | |
777 | printf("txq=%u,%hu ", lcore_id, queueid); | |
778 | fflush(stdout); | |
779 | ||
780 | rte_eth_dev_info_get(portid, &dev_info); | |
781 | txconf = &dev_info.default_txconf; | |
782 | txconf->txq_flags = 0; | |
783 | ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd, | |
784 | rte_lcore_to_socket_id(lcore_id), txconf); | |
785 | if (ret < 0) | |
786 | rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, " | |
787 | "port=%d\n", ret, portid); | |
788 | ||
789 | qconf = &lcore_queue_conf[lcore_id]; | |
790 | qconf->tx_queue_id[portid] = queueid; | |
791 | queueid++; | |
792 | } | |
793 | ||
794 | /* Start device */ | |
795 | ret = rte_eth_dev_start(portid); | |
796 | if (ret < 0) | |
797 | rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n", | |
798 | ret, portid); | |
799 | ||
800 | printf("done:\n"); | |
801 | } | |
802 | ||
803 | check_all_ports_link_status(nb_ports, enabled_port_mask); | |
804 | ||
805 | /* initialize the multicast hash */ | |
806 | int retval = init_mcast_hash(); | |
807 | if (retval != 0) | |
808 | rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n"); | |
809 | ||
810 | /* launch per-lcore init on every lcore */ | |
811 | rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER); | |
812 | RTE_LCORE_FOREACH_SLAVE(lcore_id) { | |
813 | if (rte_eal_wait_lcore(lcore_id) < 0) | |
814 | return -1; | |
815 | } | |
816 | ||
817 | return 0; | |
818 | } |