Line data Source code
1 : /*
2 : * l2_fwd.c : layer 2 forwarding using l2fib
3 : *
4 : * Copyright (c) 2013 Cisco and/or its affiliates.
5 : * Licensed under the Apache License, Version 2.0 (the "License");
6 : * you may not use this file except in compliance with the License.
7 : * You may obtain a copy of the License at:
8 : *
9 : * http://www.apache.org/licenses/LICENSE-2.0
10 : *
11 : * Unless required by applicable law or agreed to in writing, software
12 : * distributed under the License is distributed on an "AS IS" BASIS,
13 : * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 : * See the License for the specific language governing permissions and
15 : * limitations under the License.
16 : */
17 :
18 : #include <vlib/vlib.h>
19 : #include <vnet/vnet.h>
20 : #include <vnet/pg/pg.h>
21 : #include <vnet/ethernet/ethernet.h>
22 : #include <vlib/cli.h>
23 :
24 : #include <vnet/l2/l2_input.h>
25 : #include <vnet/l2/l2_bvi.h>
26 : #include <vnet/l2/l2_fwd.h>
27 : #include <vnet/l2/l2_fib.h>
28 : #include <vnet/l2/feat_bitmap.h>
29 :
30 : #include <vppinfra/error.h>
31 : #include <vppinfra/hash.h>
32 : #include <vppinfra/sparse_vec.h>
33 :
34 :
35 : /**
36 : * @file
37 : * @brief Ethernet Forwarding.
38 : *
39 : * Code in this file handles forwarding Layer 2 packets. This file calls
40 : * the FIB lookup, packet learning and the packet flooding as necessary.
41 : * Packet is then sent to the next graph node.
42 : */
43 :
44 : typedef struct
45 : {
46 :
47 : /* Hash table */
48 : BVT (clib_bihash) * mac_table;
49 :
50 : /* next node index for the L3 input node of each ethertype */
51 : next_by_ethertype_t l3_next;
52 :
53 : /* Next nodes for each feature */
54 : u32 feat_next_node_index[32];
55 :
56 : /* convenience variables */
57 : vlib_main_t *vlib_main;
58 : vnet_main_t *vnet_main;
59 : } l2fwd_main_t;
60 :
61 : typedef struct
62 : {
63 : /* per-pkt trace data */
64 : u8 dst_and_src[12];
65 : u32 sw_if_index;
66 : u16 bd_index;
67 : l2fib_entry_result_t result;
68 : } l2fwd_trace_t;
69 :
70 : /* packet trace format function */
71 : static u8 *
72 5297 : format_l2fwd_trace (u8 * s, va_list * args)
73 : {
74 5297 : CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
75 5297 : CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
76 5297 : l2fwd_trace_t *t = va_arg (*args, l2fwd_trace_t *);
77 :
78 : s =
79 5297 : format (s,
80 : "l2-fwd: sw_if_index %d dst %U src %U bd_index %d result [0x%llx, %d] %U",
81 5297 : t->sw_if_index, format_ethernet_address, t->dst_and_src,
82 5297 : format_ethernet_address, t->dst_and_src + 6,
83 5297 : t->bd_index, t->result.raw,
84 : t->result.fields.sw_if_index, format_l2fib_entry_result_flags,
85 5297 : t->result.fields.flags);
86 5297 : return s;
87 : }
88 :
89 : #ifndef CLIB_MARCH_VARIANT
90 : l2fwd_main_t l2fwd_main;
91 : #else
92 : extern l2fwd_main_t l2fwd_main;
93 : #endif
94 :
95 : extern vlib_node_registration_t l2fwd_node;
96 :
97 : #define foreach_l2fwd_error \
98 : _(L2FWD, "L2 forward packets") \
99 : _(FLOOD, "L2 forward misses") \
100 : _(HIT, "L2 forward hits") \
101 : _(BVI_BAD_MAC, "BVI L3 MAC mismatch") \
102 : _(BVI_ETHERTYPE, "BVI packet with unhandled ethertype") \
103 : _(FILTER_DROP, "Filter Mac Drop") \
104 : _(REFLECT_DROP, "Reflection Drop") \
105 : _(STALE_DROP, "Stale entry Drop")
106 :
107 : typedef enum
108 : {
109 : #define _(sym,str) L2FWD_ERROR_##sym,
110 : foreach_l2fwd_error
111 : #undef _
112 : L2FWD_N_ERROR,
113 : } l2fwd_error_t;
114 :
115 : static char *l2fwd_error_strings[] = {
116 : #define _(sym,string) string,
117 : foreach_l2fwd_error
118 : #undef _
119 : };
120 :
121 : typedef enum
122 : {
123 : L2FWD_NEXT_L2_OUTPUT,
124 : L2FWD_NEXT_DROP,
125 : L2FWD_N_NEXT,
126 : } l2fwd_next_t;
127 :
128 : /** Forward one packet based on the mac table lookup result. */
129 :
130 : static_always_inline void
131 52446600 : l2fwd_process (vlib_main_t * vm,
132 : vlib_node_runtime_t * node,
133 : l2fwd_main_t * msm,
134 : vlib_error_main_t * em,
135 : vlib_buffer_t * b0,
136 : u32 sw_if_index0, l2fib_entry_result_t * result0, u16 * next0)
137 : {
138 52446600 : int try_flood = result0->raw == ~0;
139 : int flood_error;
140 :
141 52446600 : if (PREDICT_FALSE (try_flood))
142 : {
143 3264 : flood_error = L2FWD_ERROR_FLOOD;
144 : }
145 : else
146 : {
147 : /* lookup hit, forward packet */
148 : #ifdef COUNTERS
149 : em->counters[node_counter_base_index + L2FWD_ERROR_HIT] += 1;
150 : #endif
151 :
152 52443400 : vnet_buffer (b0)->sw_if_index[VLIB_TX] = result0->fields.sw_if_index;
153 52443400 : *next0 = L2FWD_NEXT_L2_OUTPUT;
154 52443400 : int l2fib_seq_num_valid = 1;
155 :
156 : /* check l2fib seq num for stale entries */
157 52443400 : if (!l2fib_entry_result_is_set_AGE_NOT (result0))
158 : {
159 52439800 : l2fib_seq_num_t in_sn = vnet_buffer (b0)->l2.l2fib_sn;
160 52439800 : l2fib_seq_num_t expected_sn = l2_fib_update_seq_num (in_sn,
161 52439800 : l2_input_seq_num
162 : (result0->fields.sw_if_index));
163 :
164 52439800 : l2fib_seq_num_valid = expected_sn == result0->fields.sn;
165 : }
166 :
167 52443400 : if (PREDICT_FALSE (!l2fib_seq_num_valid))
168 : {
169 128 : flood_error = L2FWD_ERROR_STALE_DROP;
170 128 : try_flood = 1;
171 : }
172 : /* perform reflection check */
173 52443200 : else if (PREDICT_FALSE (sw_if_index0 == result0->fields.sw_if_index))
174 : {
175 0 : b0->error = node->errors[L2FWD_ERROR_REFLECT_DROP];
176 0 : *next0 = L2FWD_NEXT_DROP;
177 : }
178 : /* perform filter check */
179 52443200 : else if (PREDICT_FALSE (l2fib_entry_result_is_set_FILTER (result0)))
180 : {
181 0 : b0->error = node->errors[L2FWD_ERROR_FILTER_DROP];
182 0 : *next0 = L2FWD_NEXT_DROP;
183 : }
184 : /* perform BVI check */
185 52443200 : else if (PREDICT_FALSE (l2fib_entry_result_is_set_BVI (result0)))
186 : {
187 : u32 rc;
188 1649 : rc = l2_to_bvi (vm,
189 : msm->vnet_main,
190 : b0,
191 1649 : vnet_buffer (b0)->sw_if_index[VLIB_TX],
192 : &msm->l3_next, next0);
193 :
194 1649 : if (PREDICT_FALSE (rc))
195 : {
196 0 : if (rc == TO_BVI_ERR_BAD_MAC)
197 : {
198 0 : b0->error = node->errors[L2FWD_ERROR_BVI_BAD_MAC];
199 0 : *next0 = L2FWD_NEXT_DROP;
200 : }
201 0 : else if (rc == TO_BVI_ERR_ETHERTYPE)
202 : {
203 0 : b0->error = node->errors[L2FWD_ERROR_BVI_ETHERTYPE];
204 0 : *next0 = L2FWD_NEXT_DROP;
205 : }
206 : }
207 : }
208 : }
209 :
210 : /* flood */
211 52446600 : if (PREDICT_FALSE (try_flood))
212 : {
213 : /*
214 : * lookup miss, so flood which is typically the next feature
215 : * unless some other feature is inserted before uu_flood
216 : */
217 3392 : if (vnet_buffer (b0)->l2.feature_bitmap &
218 : (L2INPUT_FEAT_UU_FLOOD | L2INPUT_FEAT_UU_FWD))
219 : {
220 3118 : *next0 = vnet_l2_feature_next (b0, msm->feat_next_node_index,
221 : L2INPUT_FEAT_FWD);
222 : }
223 : else
224 : {
225 : /* Flooding is disabled */
226 274 : b0->error = node->errors[flood_error];
227 274 : *next0 = L2FWD_NEXT_DROP;
228 : }
229 : }
230 52446600 : }
231 :
232 :
233 : static_always_inline uword
234 971318 : l2fwd_node_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
235 : vlib_frame_t * frame, int do_trace)
236 : {
237 : u32 n_left, *from;
238 971318 : l2fwd_main_t *msm = &l2fwd_main;
239 971318 : vlib_node_t *n = vlib_get_node (vm, l2fwd_node.index);
240 971318 : CLIB_UNUSED (u32 node_counter_base_index) = n->error_heap_index;
241 971318 : vlib_error_main_t *em = &vm->error_main;
242 : l2fib_entry_key_t cached_key;
243 : l2fib_entry_result_t cached_result;
244 : vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b;
245 : u16 nexts[VLIB_FRAME_SIZE], *next;
246 :
247 : /* Clear the one-entry cache in case mac table was updated */
248 971318 : cached_key.raw = ~0;
249 971318 : cached_result.raw = ~0;
250 :
251 971318 : from = vlib_frame_vector_args (frame);
252 971318 : n_left = frame->n_vectors; /* number of packets to process */
253 971318 : vlib_get_buffers (vm, from, bufs, n_left);
254 971318 : next = nexts;
255 971318 : b = bufs;
256 :
257 12954700 : while (n_left >= 8)
258 : {
259 : u32 sw_if_index0, sw_if_index1, sw_if_index2, sw_if_index3;
260 : const ethernet_header_t *h0, *h1, *h2, *h3;
261 : l2fib_entry_key_t key0, key1, key2, key3;
262 : l2fib_entry_result_t result0, result1, result2, result3;
263 :
264 : /* Prefetch next iteration. */
265 : {
266 11983400 : vlib_prefetch_buffer_header (b[4], LOAD);
267 11983400 : vlib_prefetch_buffer_header (b[5], LOAD);
268 11983400 : vlib_prefetch_buffer_header (b[6], LOAD);
269 11983400 : vlib_prefetch_buffer_header (b[7], LOAD);
270 :
271 11983400 : clib_prefetch_load (b[4]->data);
272 11983400 : clib_prefetch_load (b[5]->data);
273 11983400 : clib_prefetch_load (b[6]->data);
274 11983400 : clib_prefetch_load (b[7]->data);
275 : }
276 :
277 : /* RX interface handles */
278 11983400 : sw_if_index0 = vnet_buffer (b[0])->sw_if_index[VLIB_RX];
279 11983400 : sw_if_index1 = vnet_buffer (b[1])->sw_if_index[VLIB_RX];
280 11983400 : sw_if_index2 = vnet_buffer (b[2])->sw_if_index[VLIB_RX];
281 11983400 : sw_if_index3 = vnet_buffer (b[3])->sw_if_index[VLIB_RX];
282 :
283 11983400 : h0 = vlib_buffer_get_current (b[0]);
284 11983400 : h1 = vlib_buffer_get_current (b[1]);
285 11983400 : h2 = vlib_buffer_get_current (b[2]);
286 11983400 : h3 = vlib_buffer_get_current (b[3]);
287 :
288 : #ifdef COUNTERS
289 : em->counters[node_counter_base_index + L2FWD_ERROR_L2FWD] += 4;
290 : #endif
291 : /* *INDENT-OFF* */
292 11983400 : l2fib_lookup_4 (msm->mac_table, &cached_key, &cached_result,
293 11983400 : h0->dst_address, h1->dst_address,
294 11983400 : h2->dst_address, h3->dst_address,
295 11983400 : vnet_buffer (b[0])->l2.bd_index,
296 11983400 : vnet_buffer (b[1])->l2.bd_index,
297 11983400 : vnet_buffer (b[2])->l2.bd_index,
298 11983400 : vnet_buffer (b[3])->l2.bd_index,
299 : &key0, /* not used */
300 : &key1, /* not used */
301 : &key2, /* not used */
302 : &key3, /* not used */
303 : &result0,
304 : &result1,
305 : &result2,
306 : &result3);
307 : /* *INDENT-ON* */
308 11983400 : l2fwd_process (vm, node, msm, em, b[0], sw_if_index0, &result0, next);
309 11983400 : l2fwd_process (vm, node, msm, em, b[1], sw_if_index1, &result1,
310 : next + 1);
311 11983400 : l2fwd_process (vm, node, msm, em, b[2], sw_if_index2, &result2,
312 : next + 2);
313 11983400 : l2fwd_process (vm, node, msm, em, b[3], sw_if_index3, &result3,
314 : next + 3);
315 :
316 : /* verify speculative enqueues, maybe switch current next frame */
317 : /* if next0==next1==next_index then nothing special needs to be done */
318 11983400 : if (do_trace)
319 : {
320 1437 : if (b[0]->flags & VLIB_BUFFER_IS_TRACED)
321 : {
322 1437 : l2fwd_trace_t *t = vlib_add_trace (vm, node, b[0], sizeof (*t));
323 1437 : t->sw_if_index = sw_if_index0;
324 1437 : t->bd_index = vnet_buffer (b[0])->l2.bd_index;
325 1437 : clib_memcpy_fast (t->dst_and_src, h0->dst_address,
326 : sizeof (h0->dst_address) +
327 : sizeof (h0->src_address));
328 1437 : t->result = result0;
329 : }
330 1437 : if (b[1]->flags & VLIB_BUFFER_IS_TRACED)
331 : {
332 1437 : l2fwd_trace_t *t = vlib_add_trace (vm, node, b[1], sizeof (*t));
333 1437 : t->sw_if_index = sw_if_index1;
334 1437 : t->bd_index = vnet_buffer (b[1])->l2.bd_index;
335 1437 : clib_memcpy_fast (t->dst_and_src, h1->dst_address,
336 : sizeof (h1->dst_address) +
337 : sizeof (h1->src_address));
338 1437 : t->result = result1;
339 : }
340 1437 : if (b[2]->flags & VLIB_BUFFER_IS_TRACED)
341 : {
342 1437 : l2fwd_trace_t *t = vlib_add_trace (vm, node, b[2], sizeof (*t));
343 1437 : t->sw_if_index = sw_if_index2;
344 1437 : t->bd_index = vnet_buffer (b[2])->l2.bd_index;
345 1437 : clib_memcpy_fast (t->dst_and_src, h2->dst_address,
346 : sizeof (h2->dst_address) +
347 : sizeof (h2->src_address));
348 1437 : t->result = result2;
349 : }
350 1437 : if (b[3]->flags & VLIB_BUFFER_IS_TRACED)
351 : {
352 1437 : l2fwd_trace_t *t = vlib_add_trace (vm, node, b[3], sizeof (*t));
353 1437 : t->sw_if_index = sw_if_index3;
354 1437 : t->bd_index = vnet_buffer (b[3])->l2.bd_index;
355 1437 : clib_memcpy_fast (t->dst_and_src, h3->dst_address,
356 : sizeof (h3->dst_address) +
357 : sizeof (h3->src_address));
358 1437 : t->result = result3;
359 : }
360 : }
361 :
362 11983400 : next += 4;
363 11983400 : b += 4;
364 11983400 : n_left -= 4;
365 : }
366 :
367 5484420 : while (n_left > 0)
368 : {
369 : u32 sw_if_index0;
370 : ethernet_header_t *h0;
371 : l2fib_entry_key_t key0;
372 : l2fib_entry_result_t result0;
373 :
374 4513100 : sw_if_index0 = vnet_buffer (b[0])->sw_if_index[VLIB_RX];
375 :
376 4513100 : h0 = vlib_buffer_get_current (b[0]);
377 :
378 : /* process 1 pkt */
379 : #ifdef COUNTERS
380 : em->counters[node_counter_base_index + L2FWD_ERROR_L2FWD] += 1;
381 : #endif
382 4513100 : l2fib_lookup_1 (msm->mac_table, &cached_key, &cached_result,
383 4513100 : h0->dst_address, vnet_buffer (b[0])->l2.bd_index, &key0,
384 : /* not used */ &result0);
385 4513100 : l2fwd_process (vm, node, msm, em, b[0], sw_if_index0, &result0, next);
386 :
387 4513100 : if (do_trace && PREDICT_FALSE (b[0]->flags & VLIB_BUFFER_IS_TRACED))
388 : {
389 987 : l2fwd_trace_t *t = vlib_add_trace (vm, node, b[0], sizeof (*t));
390 987 : t->sw_if_index = sw_if_index0;
391 987 : t->bd_index = vnet_buffer (b[0])->l2.bd_index;
392 987 : clib_memcpy_fast (t->dst_and_src, h0->dst_address,
393 : sizeof (h0->dst_address) +
394 : sizeof (h0->src_address));
395 987 : t->result = result0;
396 : }
397 :
398 : /* verify speculative enqueue, maybe switch current next frame */
399 4513100 : next += 1;
400 4513100 : b += 1;
401 4513100 : n_left -= 1;
402 : }
403 :
404 971318 : vlib_buffer_enqueue_to_next (vm, node, from, nexts, frame->n_vectors);
405 :
406 971318 : return frame->n_vectors;
407 : }
408 :
409 973554 : VLIB_NODE_FN (l2fwd_node) (vlib_main_t * vm,
410 : vlib_node_runtime_t * node, vlib_frame_t * frame)
411 : {
412 971318 : if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)))
413 312 : return l2fwd_node_inline (vm, node, frame, 1 /* do_trace */ );
414 971006 : return l2fwd_node_inline (vm, node, frame, 0 /* do_trace */ );
415 : }
416 :
417 : /* *INDENT-OFF* */
418 178120 : VLIB_REGISTER_NODE (l2fwd_node) = {
419 : .name = "l2-fwd",
420 : .vector_size = sizeof (u32),
421 : .format_trace = format_l2fwd_trace,
422 : .type = VLIB_NODE_TYPE_INTERNAL,
423 :
424 : .n_errors = ARRAY_LEN(l2fwd_error_strings),
425 : .error_strings = l2fwd_error_strings,
426 :
427 : .n_next_nodes = L2FWD_N_NEXT,
428 :
429 : /* edit / add dispositions here */
430 : .next_nodes = {
431 : [L2FWD_NEXT_L2_OUTPUT] = "l2-output",
432 : [L2FWD_NEXT_DROP] = "error-drop",
433 : },
434 : };
435 : /* *INDENT-ON* */
436 :
437 : #ifndef CLIB_MARCH_VARIANT
438 : clib_error_t *
439 559 : l2fwd_init (vlib_main_t * vm)
440 : {
441 559 : l2fwd_main_t *mp = &l2fwd_main;
442 :
443 559 : mp->vlib_main = vm;
444 559 : mp->vnet_main = vnet_get_main ();
445 :
446 : /* Initialize the feature next-node indexes */
447 559 : feat_bitmap_init_next_nodes (vm,
448 : l2fwd_node.index,
449 : L2INPUT_N_FEAT,
450 : l2input_get_feat_names (),
451 559 : mp->feat_next_node_index);
452 :
453 : /* init the hash table ptr */
454 559 : mp->mac_table = get_mac_table ();
455 :
456 : /* Initialize the next nodes for each ethertype */
457 559 : next_by_ethertype_init (&mp->l3_next);
458 :
459 559 : return 0;
460 : }
461 :
462 19039 : VLIB_INIT_FUNCTION (l2fwd_init);
463 :
464 :
465 : /** Add the L3 input node for this ethertype to the next nodes structure. */
466 : void
467 3917 : l2fwd_register_input_type (vlib_main_t * vm,
468 : ethernet_type_t type, u32 node_index)
469 : {
470 3917 : l2fwd_main_t *mp = &l2fwd_main;
471 : u32 next_index;
472 :
473 3917 : next_index = vlib_node_add_next (vm, l2fwd_node.index, node_index);
474 :
475 3917 : next_by_ethertype_register (&mp->l3_next, type, next_index);
476 3917 : }
477 :
478 :
479 : /**
480 : * Set subinterface forward enable/disable.
481 : * The CLI format is:
482 : * set interface l2 forward <interface> [disable]
483 : */
484 : static clib_error_t *
485 0 : int_fwd (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd)
486 : {
487 0 : vnet_main_t *vnm = vnet_get_main ();
488 0 : clib_error_t *error = 0;
489 : u32 sw_if_index;
490 : u32 enable;
491 :
492 0 : if (!unformat_user (input, unformat_vnet_sw_interface, vnm, &sw_if_index))
493 : {
494 0 : error = clib_error_return (0, "unknown interface `%U'",
495 : format_unformat_error, input);
496 0 : goto done;
497 : }
498 :
499 0 : enable = 1;
500 0 : if (unformat (input, "disable"))
501 : {
502 0 : enable = 0;
503 : }
504 :
505 : /* set the interface flag */
506 0 : if (l2input_intf_config (sw_if_index))
507 : {
508 0 : l2input_intf_bitmap_enable (sw_if_index, L2INPUT_FEAT_XCONNECT, enable);
509 : }
510 : else
511 : {
512 0 : l2input_intf_bitmap_enable (sw_if_index, L2INPUT_FEAT_FWD, enable);
513 : }
514 :
515 0 : done:
516 0 : return error;
517 : }
518 :
519 : /*?
520 : * Layer 2 unicast forwarding can be enabled and disabled on each
521 : * interface and on each bridge-domain. Use this command to
522 : * manage interfaces. It is enabled by default.
523 : *
524 : * @cliexpar
525 : * Example of how to enable forwarding:
526 : * @cliexcmd{set interface l2 forward GigabitEthernet0/8/0}
527 : * Example of how to disable forwarding:
528 : * @cliexcmd{set interface l2 forward GigabitEthernet0/8/0 disable}
529 : ?*/
530 : /* *INDENT-OFF* */
531 272887 : VLIB_CLI_COMMAND (int_fwd_cli, static) = {
532 : .path = "set interface l2 forward",
533 : .short_help = "set interface l2 forward <interface> [disable]",
534 : .function = int_fwd,
535 : };
536 : /* *INDENT-ON* */
537 :
538 : #endif
539 :
540 : /*
541 : * fd.io coding-style-patch-verification: ON
542 : *
543 : * Local Variables:
544 : * eval: (c-set-style "gnu")
545 : * End:
546 : */
|
