]> git.proxmox.com Git - ceph.git/blob - ceph/src/seastar/dpdk/examples/l3fwd/l3fwd_em.c
projects / ceph.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
update sources to ceph Nautilus 14.2.1
[ceph.git] / ceph / src / seastar / dpdk / examples / l3fwd / l3fwd_em.c
1 /*-
2 * BSD LICENSE
3 *
4 * Copyright(c) 2010-2016 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 #include <stdbool.h>
45 #include <netinet/in.h>
46
47 #include <rte_debug.h>
48 #include <rte_ether.h>
49 #include <rte_ethdev.h>
50 #include <rte_mempool.h>
51 #include <rte_cycles.h>
52 #include <rte_mbuf.h>
53 #include <rte_ip.h>
54 #include <rte_tcp.h>
55 #include <rte_udp.h>
56 #include <rte_hash.h>
57
58 #include "l3fwd.h"
59
60 #if defined(RTE_MACHINE_CPUFLAG_SSE4_2) || defined(RTE_MACHINE_CPUFLAG_CRC32)
61 #define EM_HASH_CRC 1
62 #endif
63
64 #ifdef EM_HASH_CRC
65 #include <rte_hash_crc.h>
66 #define DEFAULT_HASH_FUNC rte_hash_crc
67 #else
68 #include <rte_jhash.h>
69 #define DEFAULT_HASH_FUNC rte_jhash
70 #endif
71
72 #define IPV6_ADDR_LEN 16
73
74 struct ipv4_5tuple {
75 uint32_t ip_dst;
76 uint32_t ip_src;
77 uint16_t port_dst;
78 uint16_t port_src;
79 uint8_t proto;
80 } __attribute__((__packed__));
81
82 union ipv4_5tuple_host {
83 struct {
84 uint8_t pad0;
85 uint8_t proto;
86 uint16_t pad1;
87 uint32_t ip_src;
88 uint32_t ip_dst;
89 uint16_t port_src;
90 uint16_t port_dst;
91 };
92 xmm_t xmm;
93 };
94
95 #define XMM_NUM_IN_IPV6_5TUPLE 3
96
97 struct ipv6_5tuple {
98 uint8_t ip_dst[IPV6_ADDR_LEN];
99 uint8_t ip_src[IPV6_ADDR_LEN];
100 uint16_t port_dst;
101 uint16_t port_src;
102 uint8_t proto;
103 } __attribute__((__packed__));
104
105 union ipv6_5tuple_host {
106 struct {
107 uint16_t pad0;
108 uint8_t proto;
109 uint8_t pad1;
110 uint8_t ip_src[IPV6_ADDR_LEN];
111 uint8_t ip_dst[IPV6_ADDR_LEN];
112 uint16_t port_src;
113 uint16_t port_dst;
114 uint64_t reserve;
115 };
116 xmm_t xmm[XMM_NUM_IN_IPV6_5TUPLE];
117 };
118
119
120
121 struct ipv4_l3fwd_em_route {
122 struct ipv4_5tuple key;
123 uint8_t if_out;
124 };
125
126 struct ipv6_l3fwd_em_route {
127 struct ipv6_5tuple key;
128 uint8_t if_out;
129 };
130
131 static struct ipv4_l3fwd_em_route ipv4_l3fwd_em_route_array[] = {
132 {{IPv4(101, 0, 0, 0), IPv4(100, 10, 0, 1), 101, 11, IPPROTO_TCP}, 0},
133 {{IPv4(201, 0, 0, 0), IPv4(200, 20, 0, 1), 102, 12, IPPROTO_TCP}, 1},
134 {{IPv4(111, 0, 0, 0), IPv4(100, 30, 0, 1), 101, 11, IPPROTO_TCP}, 2},
135 {{IPv4(211, 0, 0, 0), IPv4(200, 40, 0, 1), 102, 12, IPPROTO_TCP}, 3},
136 };
137
138 static struct ipv6_l3fwd_em_route ipv6_l3fwd_em_route_array[] = {
139 {{
140 {0xfe, 0x80, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
141 {0xfe, 0x80, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38, 0x05},
142 101, 11, IPPROTO_TCP}, 0},
143
144 {{
145 {0xfe, 0x90, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
146 {0xfe, 0x90, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38, 0x05},
147 102, 12, IPPROTO_TCP}, 1},
148
149 {{
150 {0xfe, 0xa0, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
151 {0xfe, 0xa0, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38, 0x05},
152 101, 11, IPPROTO_TCP}, 2},
153
154 {{
155 {0xfe, 0xb0, 0, 0, 0, 0, 0, 0, 0x02, 0x1e, 0x67, 0xff, 0xfe, 0, 0, 0},
156 {0xfe, 0xb0, 0, 0, 0, 0, 0, 0, 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38, 0x05},
157 102, 12, IPPROTO_TCP}, 3},
158 };
159
160 struct rte_hash *ipv4_l3fwd_em_lookup_struct[NB_SOCKETS];
161 struct rte_hash *ipv6_l3fwd_em_lookup_struct[NB_SOCKETS];
162
163 static inline uint32_t
164 ipv4_hash_crc(const void *data, __rte_unused uint32_t data_len,
165 uint32_t init_val)
166 {
167 const union ipv4_5tuple_host *k;
168 uint32_t t;
169 const uint32_t *p;
170
171 k = data;
172 t = k->proto;
173 p = (const uint32_t *)&k->port_src;
174
175 #ifdef EM_HASH_CRC
176 init_val = rte_hash_crc_4byte(t, init_val);
177 init_val = rte_hash_crc_4byte(k->ip_src, init_val);
178 init_val = rte_hash_crc_4byte(k->ip_dst, init_val);
179 init_val = rte_hash_crc_4byte(*p, init_val);
180 #else
181 init_val = rte_jhash_1word(t, init_val);
182 init_val = rte_jhash_1word(k->ip_src, init_val);
183 init_val = rte_jhash_1word(k->ip_dst, init_val);
184 init_val = rte_jhash_1word(*p, init_val);
185 #endif
186
187 return init_val;
188 }
189
190 static inline uint32_t
191 ipv6_hash_crc(const void *data, __rte_unused uint32_t data_len,
192 uint32_t init_val)
193 {
194 const union ipv6_5tuple_host *k;
195 uint32_t t;
196 const uint32_t *p;
197 #ifdef EM_HASH_CRC
198 const uint32_t *ip_src0, *ip_src1, *ip_src2, *ip_src3;
199 const uint32_t *ip_dst0, *ip_dst1, *ip_dst2, *ip_dst3;
200 #endif
201
202 k = data;
203 t = k->proto;
204 p = (const uint32_t *)&k->port_src;
205
206 #ifdef EM_HASH_CRC
207 ip_src0 = (const uint32_t *) k->ip_src;
208 ip_src1 = (const uint32_t *)(k->ip_src+4);
209 ip_src2 = (const uint32_t *)(k->ip_src+8);
210 ip_src3 = (const uint32_t *)(k->ip_src+12);
211 ip_dst0 = (const uint32_t *) k->ip_dst;
212 ip_dst1 = (const uint32_t *)(k->ip_dst+4);
213 ip_dst2 = (const uint32_t *)(k->ip_dst+8);
214 ip_dst3 = (const uint32_t *)(k->ip_dst+12);
215 init_val = rte_hash_crc_4byte(t, init_val);
216 init_val = rte_hash_crc_4byte(*ip_src0, init_val);
217 init_val = rte_hash_crc_4byte(*ip_src1, init_val);
218 init_val = rte_hash_crc_4byte(*ip_src2, init_val);
219 init_val = rte_hash_crc_4byte(*ip_src3, init_val);
220 init_val = rte_hash_crc_4byte(*ip_dst0, init_val);
221 init_val = rte_hash_crc_4byte(*ip_dst1, init_val);
222 init_val = rte_hash_crc_4byte(*ip_dst2, init_val);
223 init_val = rte_hash_crc_4byte(*ip_dst3, init_val);
224 init_val = rte_hash_crc_4byte(*p, init_val);
225 #else
226 init_val = rte_jhash_1word(t, init_val);
227 init_val = rte_jhash(k->ip_src,
228 sizeof(uint8_t) * IPV6_ADDR_LEN, init_val);
229 init_val = rte_jhash(k->ip_dst,
230 sizeof(uint8_t) * IPV6_ADDR_LEN, init_val);
231 init_val = rte_jhash_1word(*p, init_val);
232 #endif
233 return init_val;
234 }
235
236 #define IPV4_L3FWD_EM_NUM_ROUTES \
237 (sizeof(ipv4_l3fwd_em_route_array) / sizeof(ipv4_l3fwd_em_route_array[0]))
238
239 #define IPV6_L3FWD_EM_NUM_ROUTES \
240 (sizeof(ipv6_l3fwd_em_route_array) / sizeof(ipv6_l3fwd_em_route_array[0]))
241
242 static uint8_t ipv4_l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
243 static uint8_t ipv6_l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
244
245 static rte_xmm_t mask0;
246 static rte_xmm_t mask1;
247 static rte_xmm_t mask2;
248
249 #if defined(__SSE2__)
250 static inline xmm_t
251 em_mask_key(void *key, xmm_t mask)
252 {
253 __m128i data = _mm_loadu_si128((__m128i *)(key));
254
255 return _mm_and_si128(data, mask);
256 }
257 #elif defined(RTE_MACHINE_CPUFLAG_NEON)
258 static inline xmm_t
259 em_mask_key(void *key, xmm_t mask)
260 {
261 int32x4_t data = vld1q_s32((int32_t *)key);
262
263 return vandq_s32(data, mask);
264 }
265 #elif defined(RTE_MACHINE_CPUFLAG_ALTIVEC)
266 static inline xmm_t
267 em_mask_key(void *key, xmm_t mask)
268 {
269 xmm_t data = vec_ld(0, (xmm_t *)(key));
270
271 return vec_and(data, mask);
272 }
273 #else
274 #error No vector engine (SSE, NEON, ALTIVEC) available, check your toolchain
275 #endif
276
277 static inline uint8_t
278 em_get_ipv4_dst_port(void *ipv4_hdr, uint8_t portid, void *lookup_struct)
279 {
280 int ret = 0;
281 union ipv4_5tuple_host key;
282 struct rte_hash *ipv4_l3fwd_lookup_struct =
283 (struct rte_hash *)lookup_struct;
284
285 ipv4_hdr = (uint8_t *)ipv4_hdr + offsetof(struct ipv4_hdr, time_to_live);
286
287 /*
288 * Get 5 tuple: dst port, src port, dst IP address,
289 * src IP address and protocol.
290 */
291 key.xmm = em_mask_key(ipv4_hdr, mask0.x);
292
293 /* Find destination port */
294 ret = rte_hash_lookup(ipv4_l3fwd_lookup_struct, (const void *)&key);
295 return (uint8_t)((ret < 0) ? portid : ipv4_l3fwd_out_if[ret]);
296 }
297
298 static inline uint8_t
299 em_get_ipv6_dst_port(void *ipv6_hdr, uint8_t portid, void *lookup_struct)
300 {
301 int ret = 0;
302 union ipv6_5tuple_host key;
303 struct rte_hash *ipv6_l3fwd_lookup_struct =
304 (struct rte_hash *)lookup_struct;
305
306 ipv6_hdr = (uint8_t *)ipv6_hdr + offsetof(struct ipv6_hdr, payload_len);
307 void *data0 = ipv6_hdr;
308 void *data1 = ((uint8_t *)ipv6_hdr) + sizeof(xmm_t);
309 void *data2 = ((uint8_t *)ipv6_hdr) + sizeof(xmm_t) + sizeof(xmm_t);
310
311 /* Get part of 5 tuple: src IP address lower 96 bits and protocol */
312 key.xmm[0] = em_mask_key(data0, mask1.x);
313
314 /*
315 * Get part of 5 tuple: dst IP address lower 96 bits
316 * and src IP address higher 32 bits.
317 */
318 key.xmm[1] = *(xmm_t *)data1;
319
320 /*
321 * Get part of 5 tuple: dst port and src port
322 * and dst IP address higher 32 bits.
323 */
324 key.xmm[2] = em_mask_key(data2, mask2.x);
325
326 /* Find destination port */
327 ret = rte_hash_lookup(ipv6_l3fwd_lookup_struct, (const void *)&key);
328 return (uint8_t)((ret < 0) ? portid : ipv6_l3fwd_out_if[ret]);
329 }
330
331 #if defined(__SSE4_1__)
332 #if defined(NO_HASH_MULTI_LOOKUP)
333 #include "l3fwd_em_sse.h"
334 #else
335 #include "l3fwd_em_hlm_sse.h"
336 #endif
337 #else
338 #include "l3fwd_em.h"
339 #endif
340
341 static void
342 convert_ipv4_5tuple(struct ipv4_5tuple *key1,
343 union ipv4_5tuple_host *key2)
344 {
345 key2->ip_dst = rte_cpu_to_be_32(key1->ip_dst);
346 key2->ip_src = rte_cpu_to_be_32(key1->ip_src);
347 key2->port_dst = rte_cpu_to_be_16(key1->port_dst);
348 key2->port_src = rte_cpu_to_be_16(key1->port_src);
349 key2->proto = key1->proto;
350 key2->pad0 = 0;
351 key2->pad1 = 0;
352 }
353
354 static void
355 convert_ipv6_5tuple(struct ipv6_5tuple *key1,
356 union ipv6_5tuple_host *key2)
357 {
358 uint32_t i;
359
360 for (i = 0; i < 16; i++) {
361 key2->ip_dst[i] = key1->ip_dst[i];
362 key2->ip_src[i] = key1->ip_src[i];
363 }
364 key2->port_dst = rte_cpu_to_be_16(key1->port_dst);
365 key2->port_src = rte_cpu_to_be_16(key1->port_src);
366 key2->proto = key1->proto;
367 key2->pad0 = 0;
368 key2->pad1 = 0;
369 key2->reserve = 0;
370 }
371
372 #define BYTE_VALUE_MAX 256
373 #define ALL_32_BITS 0xffffffff
374 #define BIT_8_TO_15 0x0000ff00
375
376 static inline void
377 populate_ipv4_few_flow_into_table(const struct rte_hash *h)
378 {
379 uint32_t i;
380 int32_t ret;
381
382 mask0 = (rte_xmm_t){.u32 = {BIT_8_TO_15, ALL_32_BITS,
383 ALL_32_BITS, ALL_32_BITS} };
384
385 for (i = 0; i < IPV4_L3FWD_EM_NUM_ROUTES; i++) {
386 struct ipv4_l3fwd_em_route entry;
387 union ipv4_5tuple_host newkey;
388
389 entry = ipv4_l3fwd_em_route_array[i];
390 convert_ipv4_5tuple(&entry.key, &newkey);
391 ret = rte_hash_add_key(h, (void *) &newkey);
392 if (ret < 0) {
393 rte_exit(EXIT_FAILURE, "Unable to add entry %" PRIu32
394 " to the l3fwd hash.\n", i);
395 }
396 ipv4_l3fwd_out_if[ret] = entry.if_out;
397 }
398 printf("Hash: Adding 0x%" PRIx64 " keys\n",
399 (uint64_t)IPV4_L3FWD_EM_NUM_ROUTES);
400 }
401
402 #define BIT_16_TO_23 0x00ff0000
403 static inline void
404 populate_ipv6_few_flow_into_table(const struct rte_hash *h)
405 {
406 uint32_t i;
407 int32_t ret;
408
409 mask1 = (rte_xmm_t){.u32 = {BIT_16_TO_23, ALL_32_BITS,
410 ALL_32_BITS, ALL_32_BITS} };
411
412 mask2 = (rte_xmm_t){.u32 = {ALL_32_BITS, ALL_32_BITS, 0, 0} };
413
414 for (i = 0; i < IPV6_L3FWD_EM_NUM_ROUTES; i++) {
415 struct ipv6_l3fwd_em_route entry;
416 union ipv6_5tuple_host newkey;
417
418 entry = ipv6_l3fwd_em_route_array[i];
419 convert_ipv6_5tuple(&entry.key, &newkey);
420 ret = rte_hash_add_key(h, (void *) &newkey);
421 if (ret < 0) {
422 rte_exit(EXIT_FAILURE, "Unable to add entry %" PRIu32
423 " to the l3fwd hash.\n", i);
424 }
425 ipv6_l3fwd_out_if[ret] = entry.if_out;
426 }
427 printf("Hash: Adding 0x%" PRIx64 "keys\n",
428 (uint64_t)IPV6_L3FWD_EM_NUM_ROUTES);
429 }
430
431 #define NUMBER_PORT_USED 4
432 static inline void
433 populate_ipv4_many_flow_into_table(const struct rte_hash *h,
434 unsigned int nr_flow)
435 {
436 unsigned i;
437
438 mask0 = (rte_xmm_t){.u32 = {BIT_8_TO_15, ALL_32_BITS,
439 ALL_32_BITS, ALL_32_BITS} };
440
441 for (i = 0; i < nr_flow; i++) {
442 struct ipv4_l3fwd_em_route entry;
443 union ipv4_5tuple_host newkey;
444
445 uint8_t a = (uint8_t)
446 ((i/NUMBER_PORT_USED)%BYTE_VALUE_MAX);
447 uint8_t b = (uint8_t)
448 (((i/NUMBER_PORT_USED)/BYTE_VALUE_MAX)%BYTE_VALUE_MAX);
449 uint8_t c = (uint8_t)
450 ((i/NUMBER_PORT_USED)/(BYTE_VALUE_MAX*BYTE_VALUE_MAX));
451
452 /* Create the ipv4 exact match flow */
453 memset(&entry, 0, sizeof(entry));
454 switch (i & (NUMBER_PORT_USED - 1)) {
455 case 0:
456 entry = ipv4_l3fwd_em_route_array[0];
457 entry.key.ip_dst = IPv4(101, c, b, a);
458 break;
459 case 1:
460 entry = ipv4_l3fwd_em_route_array[1];
461 entry.key.ip_dst = IPv4(201, c, b, a);
462 break;
463 case 2:
464 entry = ipv4_l3fwd_em_route_array[2];
465 entry.key.ip_dst = IPv4(111, c, b, a);
466 break;
467 case 3:
468 entry = ipv4_l3fwd_em_route_array[3];
469 entry.key.ip_dst = IPv4(211, c, b, a);
470 break;
471 };
472 convert_ipv4_5tuple(&entry.key, &newkey);
473 int32_t ret = rte_hash_add_key(h, (void *) &newkey);
474
475 if (ret < 0)
476 rte_exit(EXIT_FAILURE, "Unable to add entry %u\n", i);
477
478 ipv4_l3fwd_out_if[ret] = (uint8_t) entry.if_out;
479
480 }
481 printf("Hash: Adding 0x%x keys\n", nr_flow);
482 }
483
484 static inline void
485 populate_ipv6_many_flow_into_table(const struct rte_hash *h,
486 unsigned int nr_flow)
487 {
488 unsigned i;
489
490 mask1 = (rte_xmm_t){.u32 = {BIT_16_TO_23, ALL_32_BITS,
491 ALL_32_BITS, ALL_32_BITS} };
492 mask2 = (rte_xmm_t){.u32 = {ALL_32_BITS, ALL_32_BITS, 0, 0} };
493
494 for (i = 0; i < nr_flow; i++) {
495 struct ipv6_l3fwd_em_route entry;
496 union ipv6_5tuple_host newkey;
497
498 uint8_t a = (uint8_t)
499 ((i/NUMBER_PORT_USED)%BYTE_VALUE_MAX);
500 uint8_t b = (uint8_t)
501 (((i/NUMBER_PORT_USED)/BYTE_VALUE_MAX)%BYTE_VALUE_MAX);
502 uint8_t c = (uint8_t)
503 ((i/NUMBER_PORT_USED)/(BYTE_VALUE_MAX*BYTE_VALUE_MAX));
504
505 /* Create the ipv6 exact match flow */
506 memset(&entry, 0, sizeof(entry));
507 switch (i & (NUMBER_PORT_USED - 1)) {
508 case 0:
509 entry = ipv6_l3fwd_em_route_array[0];
510 break;
511 case 1:
512 entry = ipv6_l3fwd_em_route_array[1];
513 break;
514 case 2:
515 entry = ipv6_l3fwd_em_route_array[2];
516 break;
517 case 3:
518 entry = ipv6_l3fwd_em_route_array[3];
519 break;
520 };
521 entry.key.ip_dst[13] = c;
522 entry.key.ip_dst[14] = b;
523 entry.key.ip_dst[15] = a;
524 convert_ipv6_5tuple(&entry.key, &newkey);
525 int32_t ret = rte_hash_add_key(h, (void *) &newkey);
526
527 if (ret < 0)
528 rte_exit(EXIT_FAILURE, "Unable to add entry %u\n", i);
529
530 ipv6_l3fwd_out_if[ret] = (uint8_t) entry.if_out;
531
532 }
533 printf("Hash: Adding 0x%x keys\n", nr_flow);
534 }
535
536 /* Requirements:
537 * 1. IP packets without extension;
538 * 2. L4 payload should be either TCP or UDP.
539 */
540 int
541 em_check_ptype(int portid)
542 {
543 int i, ret;
544 int ptype_l3_ipv4_ext = 0;
545 int ptype_l3_ipv6_ext = 0;
546 int ptype_l4_tcp = 0;
547 int ptype_l4_udp = 0;
548 uint32_t ptype_mask = RTE_PTYPE_L3_MASK | RTE_PTYPE_L4_MASK;
549
550 ret = rte_eth_dev_get_supported_ptypes(portid, ptype_mask, NULL, 0);
551 if (ret <= 0)
552 return 0;
553
554 uint32_t ptypes[ret];
555
556 ret = rte_eth_dev_get_supported_ptypes(portid, ptype_mask, ptypes, ret);
557 for (i = 0; i < ret; ++i) {
558 switch (ptypes[i]) {
559 case RTE_PTYPE_L3_IPV4_EXT:
560 ptype_l3_ipv4_ext = 1;
561 break;
562 case RTE_PTYPE_L3_IPV6_EXT:
563 ptype_l3_ipv6_ext = 1;
564 break;
565 case RTE_PTYPE_L4_TCP:
566 ptype_l4_tcp = 1;
567 break;
568 case RTE_PTYPE_L4_UDP:
569 ptype_l4_udp = 1;
570 break;
571 }
572 }
573
574 if (ptype_l3_ipv4_ext == 0)
575 printf("port %d cannot parse RTE_PTYPE_L3_IPV4_EXT\n", portid);
576 if (ptype_l3_ipv6_ext == 0)
577 printf("port %d cannot parse RTE_PTYPE_L3_IPV6_EXT\n", portid);
578 if (!ptype_l3_ipv4_ext || !ptype_l3_ipv6_ext)
579 return 0;
580
581 if (ptype_l4_tcp == 0)
582 printf("port %d cannot parse RTE_PTYPE_L4_TCP\n", portid);
583 if (ptype_l4_udp == 0)
584 printf("port %d cannot parse RTE_PTYPE_L4_UDP\n", portid);
585 if (ptype_l4_tcp && ptype_l4_udp)
586 return 1;
587
588 return 0;
589 }
590
591 static inline void
592 em_parse_ptype(struct rte_mbuf *m)
593 {
594 struct ether_hdr *eth_hdr;
595 uint32_t packet_type = RTE_PTYPE_UNKNOWN;
596 uint16_t ether_type;
597 void *l3;
598 int hdr_len;
599 struct ipv4_hdr *ipv4_hdr;
600 struct ipv6_hdr *ipv6_hdr;
601
602 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
603 ether_type = eth_hdr->ether_type;
604 l3 = (uint8_t *)eth_hdr + sizeof(struct ether_hdr);
605 if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4)) {
606 ipv4_hdr = (struct ipv4_hdr *)l3;
607 hdr_len = (ipv4_hdr->version_ihl & IPV4_HDR_IHL_MASK) *
608 IPV4_IHL_MULTIPLIER;
609 if (hdr_len == sizeof(struct ipv4_hdr)) {
610 packet_type |= RTE_PTYPE_L3_IPV4;
611 if (ipv4_hdr->next_proto_id == IPPROTO_TCP)
612 packet_type |= RTE_PTYPE_L4_TCP;
613 else if (ipv4_hdr->next_proto_id == IPPROTO_UDP)
614 packet_type |= RTE_PTYPE_L4_UDP;
615 } else
616 packet_type |= RTE_PTYPE_L3_IPV4_EXT;
617 } else if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4)) {
618 ipv6_hdr = (struct ipv6_hdr *)l3;
619 if (ipv6_hdr->proto == IPPROTO_TCP)
620 packet_type |= RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_TCP;
621 else if (ipv6_hdr->proto == IPPROTO_UDP)
622 packet_type |= RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_UDP;
623 else
624 packet_type |= RTE_PTYPE_L3_IPV6_EXT_UNKNOWN;
625 }
626
627 m->packet_type = packet_type;
628 }
629
630 uint16_t
631 em_cb_parse_ptype(uint8_t port __rte_unused, uint16_t queue __rte_unused,
632 struct rte_mbuf *pkts[], uint16_t nb_pkts,
633 uint16_t max_pkts __rte_unused,
634 void *user_param __rte_unused)
635 {
636 unsigned i;
637
638 for (i = 0; i < nb_pkts; ++i)
639 em_parse_ptype(pkts[i]);
640
641 return nb_pkts;
642 }
643
644 /* main processing loop */
645 int
646 em_main_loop(__attribute__((unused)) void *dummy)
647 {
648 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
649 unsigned lcore_id;
650 uint64_t prev_tsc, diff_tsc, cur_tsc;
651 int i, nb_rx;
652 uint8_t portid, queueid;
653 struct lcore_conf *qconf;
654 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) /
655 US_PER_S * BURST_TX_DRAIN_US;
656
657 prev_tsc = 0;
658
659 lcore_id = rte_lcore_id();
660 qconf = &lcore_conf[lcore_id];
661
662 if (qconf->n_rx_queue == 0) {
663 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
664 return 0;
665 }
666
667 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
668
669 for (i = 0; i < qconf->n_rx_queue; i++) {
670
671 portid = qconf->rx_queue_list[i].port_id;
672 queueid = qconf->rx_queue_list[i].queue_id;
673 RTE_LOG(INFO, L3FWD,
674 " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
675 lcore_id, portid, queueid);
676 }
677
678 while (!force_quit) {
679
680 cur_tsc = rte_rdtsc();
681
682 /*
683 * TX burst queue drain
684 */
685 diff_tsc = cur_tsc - prev_tsc;
686 if (unlikely(diff_tsc > drain_tsc)) {
687
688 for (i = 0; i < qconf->n_tx_port; ++i) {
689 portid = qconf->tx_port_id[i];
690 if (qconf->tx_mbufs[portid].len == 0)
691 continue;
692 send_burst(qconf,
693 qconf->tx_mbufs[portid].len,
694 portid);
695 qconf->tx_mbufs[portid].len = 0;
696 }
697
698 prev_tsc = cur_tsc;
699 }
700
701 /*
702 * Read packet from RX queues
703 */
704 for (i = 0; i < qconf->n_rx_queue; ++i) {
705 portid = qconf->rx_queue_list[i].port_id;
706 queueid = qconf->rx_queue_list[i].queue_id;
707 nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst,
708 MAX_PKT_BURST);
709 if (nb_rx == 0)
710 continue;
711
712 #if defined(__SSE4_1__)
713 l3fwd_em_send_packets(nb_rx, pkts_burst,
714 portid, qconf);
715 #else
716 l3fwd_em_no_opt_send_packets(nb_rx, pkts_burst,
717 portid, qconf);
718 #endif /* __SSE_4_1__ */
719 }
720 }
721
722 return 0;
723 }
724
725 /*
726 * Initialize exact match (hash) parameters.
727 */
728 void
729 setup_hash(const int socketid)
730 {
731 struct rte_hash_parameters ipv4_l3fwd_hash_params = {
732 .name = NULL,
733 .entries = L3FWD_HASH_ENTRIES,
734 .key_len = sizeof(union ipv4_5tuple_host),
735 .hash_func = ipv4_hash_crc,
736 .hash_func_init_val = 0,
737 };
738
739 struct rte_hash_parameters ipv6_l3fwd_hash_params = {
740 .name = NULL,
741 .entries = L3FWD_HASH_ENTRIES,
742 .key_len = sizeof(union ipv6_5tuple_host),
743 .hash_func = ipv6_hash_crc,
744 .hash_func_init_val = 0,
745 };
746
747 char s[64];
748
749 /* create ipv4 hash */
750 snprintf(s, sizeof(s), "ipv4_l3fwd_hash_%d", socketid);
751 ipv4_l3fwd_hash_params.name = s;
752 ipv4_l3fwd_hash_params.socket_id = socketid;
753 ipv4_l3fwd_em_lookup_struct[socketid] =
754 rte_hash_create(&ipv4_l3fwd_hash_params);
755 if (ipv4_l3fwd_em_lookup_struct[socketid] == NULL)
756 rte_exit(EXIT_FAILURE,
757 "Unable to create the l3fwd hash on socket %d\n",
758 socketid);
759
760 /* create ipv6 hash */
761 snprintf(s, sizeof(s), "ipv6_l3fwd_hash_%d", socketid);
762 ipv6_l3fwd_hash_params.name = s;
763 ipv6_l3fwd_hash_params.socket_id = socketid;
764 ipv6_l3fwd_em_lookup_struct[socketid] =
765 rte_hash_create(&ipv6_l3fwd_hash_params);
766 if (ipv6_l3fwd_em_lookup_struct[socketid] == NULL)
767 rte_exit(EXIT_FAILURE,
768 "Unable to create the l3fwd hash on socket %d\n",
769 socketid);
770
771 if (hash_entry_number != HASH_ENTRY_NUMBER_DEFAULT) {
772 /* For testing hash matching with a large number of flows we
773 * generate millions of IP 5-tuples with an incremented dst
774 * address to initialize the hash table. */
775 if (ipv6 == 0) {
776 /* populate the ipv4 hash */
777 populate_ipv4_many_flow_into_table(
778 ipv4_l3fwd_em_lookup_struct[socketid],
779 hash_entry_number);
780 } else {
781 /* populate the ipv6 hash */
782 populate_ipv6_many_flow_into_table(
783 ipv6_l3fwd_em_lookup_struct[socketid],
784 hash_entry_number);
785 }
786 } else {
787 /*
788 * Use data in ipv4/ipv6 l3fwd lookup table
789 * directly to initialize the hash table.
790 */
791 if (ipv6 == 0) {
792 /* populate the ipv4 hash */
793 populate_ipv4_few_flow_into_table(
794 ipv4_l3fwd_em_lookup_struct[socketid]);
795 } else {
796 /* populate the ipv6 hash */
797 populate_ipv6_few_flow_into_table(
798 ipv6_l3fwd_em_lookup_struct[socketid]);
799 }
800 }
801 }
802
803 /* Return ipv4/ipv6 em fwd lookup struct. */
804 void *
805 em_get_ipv4_l3fwd_lookup_struct(const int socketid)
806 {
807 return ipv4_l3fwd_em_lookup_struct[socketid];
808 }
809
810 void *
811 em_get_ipv6_l3fwd_lookup_struct(const int socketid)
812 {
813 return ipv6_l3fwd_em_lookup_struct[socketid];
814 }
Ceph