blob: 514b2c9968e692990632d3e775749ac5c64877f1 [file] [log] [blame]
Ed Warnickecb9cada2015-12-08 15:45:58 -07001/*
2 * decap.c: vxlan tunnel decap packet processing
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/pg/pg.h>
20#include <vnet/vxlan/vxlan.h>
21
John Lo13e3d452016-08-09 19:20:51 -040022vlib_node_registration_t vxlan4_input_node;
23vlib_node_registration_t vxlan6_input_node;
Ed Warnickecb9cada2015-12-08 15:45:58 -070024
25typedef struct {
26 u32 next_index;
27 u32 tunnel_index;
28 u32 error;
29 u32 vni;
30} vxlan_rx_trace_t;
31
32static u8 * format_vxlan_rx_trace (u8 * s, va_list * args)
33{
34 CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
35 CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
36 vxlan_rx_trace_t * t = va_arg (*args, vxlan_rx_trace_t *);
37
38 if (t->tunnel_index != ~0)
39 {
John Loc42912d2016-11-07 18:30:47 -050040 s = format (s, "VXLAN decap from vxlan_tunnel%d vni %d next %d error %d",
Ed Warnickecb9cada2015-12-08 15:45:58 -070041 t->tunnel_index, t->vni, t->next_index, t->error);
42 }
43 else
44 {
John Loc42912d2016-11-07 18:30:47 -050045 s = format (s, "VXLAN decap error - tunnel for vni %d does not exist",
46 t->vni);
Ed Warnickecb9cada2015-12-08 15:45:58 -070047 }
48 return s;
49}
50
John Lo2b81eb82017-01-30 13:12:10 -050051always_inline u32
52validate_vxlan_fib (vlib_buffer_t *b, vxlan_tunnel_t *t, u32 is_ip4)
53{
54 u32 fib_index, sw_if_index;
55
56 sw_if_index = vnet_buffer (b)->sw_if_index[VLIB_RX];
57
58 if (is_ip4)
59 fib_index = (vnet_buffer (b)->sw_if_index[VLIB_TX] == (u32) ~ 0) ?
60 vec_elt (ip4_main.fib_index_by_sw_if_index, sw_if_index) :
61 vnet_buffer (b)->sw_if_index[VLIB_TX];
62 else
63 fib_index = (vnet_buffer (b)->sw_if_index[VLIB_TX] == (u32) ~ 0) ?
64 vec_elt (ip6_main.fib_index_by_sw_if_index, sw_if_index) :
65 vnet_buffer (b)->sw_if_index[VLIB_TX];
66
67 return (fib_index == t->encap_fib_index);
68}
69
Chris Luke99cb3352016-04-26 10:49:53 -040070always_inline uword
Ed Warnickecb9cada2015-12-08 15:45:58 -070071vxlan_input (vlib_main_t * vm,
72 vlib_node_runtime_t * node,
Chris Luke99cb3352016-04-26 10:49:53 -040073 vlib_frame_t * from_frame,
John Lo37682e12016-11-30 12:51:39 -050074 u32 is_ip4)
Ed Warnickecb9cada2015-12-08 15:45:58 -070075{
76 u32 n_left_from, next_index, * from, * to_next;
77 vxlan_main_t * vxm = &vxlan_main;
78 vnet_main_t * vnm = vxm->vnet_main;
79 vnet_interface_main_t * im = &vnm->interface_main;
80 u32 last_tunnel_index = ~0;
Chris Luke99cb3352016-04-26 10:49:53 -040081 vxlan4_tunnel_key_t last_key4;
82 vxlan6_tunnel_key_t last_key6;
Ed Warnickecb9cada2015-12-08 15:45:58 -070083 u32 pkts_decapsulated = 0;
84 u32 cpu_index = os_get_cpu_number();
85 u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
86
Dave Barachf9c231e2016-08-05 10:10:18 -040087 if (is_ip4)
88 last_key4.as_u64 = ~0;
89 else
90 memset (&last_key6, 0xff, sizeof (last_key6));
Ed Warnickecb9cada2015-12-08 15:45:58 -070091
92 from = vlib_frame_vector_args (from_frame);
93 n_left_from = from_frame->n_vectors;
94
95 next_index = node->cached_next_index;
96 stats_sw_if_index = node->runtime_data[0];
97 stats_n_packets = stats_n_bytes = 0;
98
99 while (n_left_from > 0)
100 {
101 u32 n_left_to_next;
102
103 vlib_get_next_frame (vm, node, next_index,
104 to_next, n_left_to_next);
105 while (n_left_from >= 4 && n_left_to_next >= 2)
106 {
107 u32 bi0, bi1;
108 vlib_buffer_t * b0, * b1;
109 u32 next0, next1;
Chris Luke99cb3352016-04-26 10:49:53 -0400110 ip4_header_t * ip4_0, * ip4_1;
111 ip6_header_t * ip6_0, * ip6_1;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700112 vxlan_header_t * vxlan0, * vxlan1;
113 uword * p0, * p1;
114 u32 tunnel_index0, tunnel_index1;
John Lo56912c82016-12-08 16:10:02 -0500115 vxlan_tunnel_t * t0, * t1, * mt0 = NULL, * mt1 = NULL;
Chris Luke99cb3352016-04-26 10:49:53 -0400116 vxlan4_tunnel_key_t key4_0, key4_1;
117 vxlan6_tunnel_key_t key6_0, key6_1;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700118 u32 error0, error1;
119 u32 sw_if_index0, sw_if_index1, len0, len1;
120
121 /* Prefetch next iteration. */
122 {
123 vlib_buffer_t * p2, * p3;
124
125 p2 = vlib_get_buffer (vm, from[2]);
126 p3 = vlib_get_buffer (vm, from[3]);
127
128 vlib_prefetch_buffer_header (p2, LOAD);
129 vlib_prefetch_buffer_header (p3, LOAD);
130
131 CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
132 CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
133 }
134
135 bi0 = from[0];
136 bi1 = from[1];
137 to_next[0] = bi0;
138 to_next[1] = bi1;
139 from += 2;
140 to_next += 2;
141 n_left_to_next -= 2;
142 n_left_from -= 2;
143
144 b0 = vlib_get_buffer (vm, bi0);
145 b1 = vlib_get_buffer (vm, bi1);
146
147 /* udp leaves current_data pointing at the vxlan header */
148 vxlan0 = vlib_buffer_get_current (b0);
149 vxlan1 = vlib_buffer_get_current (b1);
Chris Luke99cb3352016-04-26 10:49:53 -0400150 if (is_ip4) {
John Loc42912d2016-11-07 18:30:47 -0500151 vlib_buffer_advance
152 (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
153 vlib_buffer_advance
154 (b1, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
Chris Luke99cb3352016-04-26 10:49:53 -0400155 ip4_0 = vlib_buffer_get_current (b0);
156 ip4_1 = vlib_buffer_get_current (b1);
157 } else {
158 vlib_buffer_advance
159 (b0, -(word)(sizeof(udp_header_t)+sizeof(ip6_header_t)));
160 vlib_buffer_advance
161 (b1, -(word)(sizeof(udp_header_t)+sizeof(ip6_header_t)));
162 ip6_0 = vlib_buffer_get_current (b0);
163 ip6_1 = vlib_buffer_get_current (b1);
164 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700165
166 /* pop (ip, udp, vxlan) */
Chris Luke99cb3352016-04-26 10:49:53 -0400167 if (is_ip4) {
168 vlib_buffer_advance
169 (b0, sizeof(*ip4_0)+sizeof(udp_header_t)+sizeof(*vxlan0));
John Loc42912d2016-11-07 18:30:47 -0500170 vlib_buffer_advance
Chris Luke99cb3352016-04-26 10:49:53 -0400171 (b1, sizeof(*ip4_1)+sizeof(udp_header_t)+sizeof(*vxlan1));
172 } else {
John Loc42912d2016-11-07 18:30:47 -0500173 vlib_buffer_advance
Chris Luke99cb3352016-04-26 10:49:53 -0400174 (b0, sizeof(*ip6_0)+sizeof(udp_header_t)+sizeof(*vxlan0));
175 vlib_buffer_advance
176 (b1, sizeof(*ip6_1)+sizeof(udp_header_t)+sizeof(*vxlan1));
177 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700178
179 tunnel_index0 = ~0;
180 error0 = 0;
181
182 tunnel_index1 = ~0;
183 error1 = 0;
184
John Loc42912d2016-11-07 18:30:47 -0500185 if (PREDICT_FALSE (vxlan0->flags != VXLAN_FLAGS_I))
186 {
187 error0 = VXLAN_ERROR_BAD_FLAGS;
188 next0 = VXLAN_INPUT_NEXT_DROP;
189 goto trace0;
190 }
191
Chris Luke99cb3352016-04-26 10:49:53 -0400192 if (is_ip4) {
193 key4_0.src = ip4_0->src_address.as_u32;
194 key4_0.vni = vxlan0->vni_reserved;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700195
John Lo56912c82016-12-08 16:10:02 -0500196 /* Make sure VXLAN tunnel exist according to packet SIP and VNI */
197 if (PREDICT_FALSE (key4_0.as_u64 != last_key4.as_u64))
Chris Luke99cb3352016-04-26 10:49:53 -0400198 {
199 p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64);
John Lo56912c82016-12-08 16:10:02 -0500200 if (PREDICT_FALSE (p0 == NULL))
Chris Luke99cb3352016-04-26 10:49:53 -0400201 {
202 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
203 next0 = VXLAN_INPUT_NEXT_DROP;
204 goto trace0;
205 }
Chris Luke99cb3352016-04-26 10:49:53 -0400206 last_key4.as_u64 = key4_0.as_u64;
207 tunnel_index0 = last_tunnel_index = p0[0];
208 }
209 else
210 tunnel_index0 = last_tunnel_index;
John Lo56912c82016-12-08 16:10:02 -0500211 t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0);
212
John Lo2b81eb82017-01-30 13:12:10 -0500213 /* Validate VXLAN tunnel encap-fib index agaist packet */
214 if (PREDICT_FALSE (validate_vxlan_fib (b0, t0, is_ip4) == 0))
215 {
216 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
217 next0 = VXLAN_INPUT_NEXT_DROP;
218 goto trace0;
219 }
220
221 /* Validate VXLAN tunnel SIP against packet DIP */
John Lo56912c82016-12-08 16:10:02 -0500222 if (PREDICT_TRUE (ip4_0->dst_address.as_u32 == t0->src.ip4.as_u32))
223 goto next0; /* valid packet */
224 if (PREDICT_FALSE (ip4_address_is_multicast (&ip4_0->dst_address)))
225 {
226 key4_0.src = ip4_0->dst_address.as_u32;
227 key4_0.vni = vxlan0->vni_reserved;
228 /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */
229 p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64);
230 if (PREDICT_TRUE (p0 != NULL))
231 {
232 mt0 = pool_elt_at_index (vxm->tunnels, p0[0]);
233 goto next0; /* valid packet */
234 }
235 }
236 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
237 next0 = VXLAN_INPUT_NEXT_DROP;
238 goto trace0;
239
240 } else /* !is_ip4 */ {
Chris Luke99cb3352016-04-26 10:49:53 -0400241 key6_0.src.as_u64[0] = ip6_0->src_address.as_u64[0];
242 key6_0.src.as_u64[1] = ip6_0->src_address.as_u64[1];
243 key6_0.vni = vxlan0->vni_reserved;
244
John Lo56912c82016-12-08 16:10:02 -0500245 /* Make sure VXLAN tunnel exist according to packet SIP and VNI */
246 if (PREDICT_FALSE (memcmp(&key6_0, &last_key6, sizeof(last_key6)) != 0))
Chris Lukec7949152016-05-31 10:42:14 -0400247 {
248 p0 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_0);
John Lo56912c82016-12-08 16:10:02 -0500249 if (PREDICT_FALSE (p0 == NULL))
Chris Lukec7949152016-05-31 10:42:14 -0400250 {
251 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
252 next0 = VXLAN_INPUT_NEXT_DROP;
253 goto trace0;
254 }
Chris Lukec7949152016-05-31 10:42:14 -0400255 clib_memcpy (&last_key6, &key6_0, sizeof(key6_0));
256 tunnel_index0 = last_tunnel_index = p0[0];
257 }
258 else
259 tunnel_index0 = last_tunnel_index;
John Lo56912c82016-12-08 16:10:02 -0500260 t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0);
261
John Lo2b81eb82017-01-30 13:12:10 -0500262 /* Validate VXLAN tunnel encap-fib index agaist packet */
263 if (PREDICT_FALSE (validate_vxlan_fib (b0, t0, is_ip4) == 0))
264 {
265 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
266 next0 = VXLAN_INPUT_NEXT_DROP;
267 goto trace0;
268 }
269
John Lo56912c82016-12-08 16:10:02 -0500270 /* Validate VXLAN tunnel SIP against packet DIP */
271 if (PREDICT_TRUE (ip6_address_is_equal (&ip6_0->dst_address,
272 &t0->src.ip6)))
273 goto next0; /* valid packet */
274 if (PREDICT_FALSE (ip6_address_is_multicast (&ip6_0->dst_address)))
275 {
276 key6_0.src.as_u64[0] = ip6_0->dst_address.as_u64[0];
277 key6_0.src.as_u64[1] = ip6_0->dst_address.as_u64[1];
278 key6_0.vni = vxlan0->vni_reserved;
John Lo2b81eb82017-01-30 13:12:10 -0500279 p0 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_0);
John Lo56912c82016-12-08 16:10:02 -0500280 if (PREDICT_TRUE (p0 != NULL))
281 {
282 mt0 = pool_elt_at_index (vxm->tunnels, p0[0]);
283 goto next0; /* valid packet */
284 }
285 }
286 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
287 next0 = VXLAN_INPUT_NEXT_DROP;
288 goto trace0;
Chris Luke99cb3352016-04-26 10:49:53 -0400289 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700290
John Lo56912c82016-12-08 16:10:02 -0500291 next0:
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800292 next0 = t0->decap_next_index;
Chris Luke99cb3352016-04-26 10:49:53 -0400293 sw_if_index0 = t0->sw_if_index;
294 len0 = vlib_buffer_length_in_chain (vm, b0);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700295
296 /* Required to make the l2 tag push / pop code work on l2 subifs */
Chris Luke99cb3352016-04-26 10:49:53 -0400297 if (PREDICT_TRUE(next0 == VXLAN_INPUT_NEXT_L2_INPUT))
298 vnet_update_l2_len (b0);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700299
John Lo56912c82016-12-08 16:10:02 -0500300 /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700301 vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0;
John Lo56912c82016-12-08 16:10:02 -0500302 sw_if_index0 = (mt0) ? mt0->sw_if_index : sw_if_index0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700303
304 pkts_decapsulated ++;
Chris Luke99cb3352016-04-26 10:49:53 -0400305 stats_n_packets += 1;
306 stats_n_bytes += len0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700307
308 /* Batch stats increment on the same vxlan tunnel so counter
309 is not incremented per packet */
310 if (PREDICT_FALSE (sw_if_index0 != stats_sw_if_index))
311 {
312 stats_n_packets -= 1;
313 stats_n_bytes -= len0;
314 if (stats_n_packets)
315 vlib_increment_combined_counter
316 (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
317 cpu_index, stats_sw_if_index,
318 stats_n_packets, stats_n_bytes);
319 stats_n_packets = 1;
320 stats_n_bytes = len0;
321 stats_sw_if_index = sw_if_index0;
322 }
323
324 trace0:
325 b0->error = error0 ? node->errors[error0] : 0;
326
327 if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
328 {
329 vxlan_rx_trace_t *tr
330 = vlib_add_trace (vm, node, b0, sizeof (*tr));
331 tr->next_index = next0;
332 tr->error = error0;
333 tr->tunnel_index = tunnel_index0;
334 tr->vni = vnet_get_vni (vxlan0);
335 }
336
John Loc42912d2016-11-07 18:30:47 -0500337 if (PREDICT_FALSE (vxlan1->flags != VXLAN_FLAGS_I))
338 {
339 error1 = VXLAN_ERROR_BAD_FLAGS;
340 next1 = VXLAN_INPUT_NEXT_DROP;
341 goto trace1;
342 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700343
Chris Luke99cb3352016-04-26 10:49:53 -0400344 if (is_ip4) {
345 key4_1.src = ip4_1->src_address.as_u32;
346 key4_1.vni = vxlan1->vni_reserved;
347
John Lo56912c82016-12-08 16:10:02 -0500348 /* Make sure unicast VXLAN tunnel exist by packet SIP and VNI */
349 if (PREDICT_FALSE (key4_1.as_u64 != last_key4.as_u64))
Chris Luke99cb3352016-04-26 10:49:53 -0400350 {
351 p1 = hash_get (vxm->vxlan4_tunnel_by_key, key4_1.as_u64);
John Lo56912c82016-12-08 16:10:02 -0500352 if (PREDICT_FALSE (p1 == NULL))
Chris Luke99cb3352016-04-26 10:49:53 -0400353 {
354 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
355 next1 = VXLAN_INPUT_NEXT_DROP;
356 goto trace1;
357 }
Chris Luke99cb3352016-04-26 10:49:53 -0400358 last_key4.as_u64 = key4_1.as_u64;
359 tunnel_index1 = last_tunnel_index = p1[0];
360 }
361 else
362 tunnel_index1 = last_tunnel_index;
John Lo56912c82016-12-08 16:10:02 -0500363 t1 = pool_elt_at_index (vxm->tunnels, tunnel_index1);
364
John Lo2b81eb82017-01-30 13:12:10 -0500365 /* Validate VXLAN tunnel encap-fib index agaist packet */
366 if (PREDICT_FALSE (validate_vxlan_fib (b1, t1, is_ip4) == 0))
367 {
368 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
369 next1 = VXLAN_INPUT_NEXT_DROP;
370 goto trace1;
371 }
372
John Lo56912c82016-12-08 16:10:02 -0500373 /* Validate VXLAN tunnel SIP against packet DIP */
374 if (PREDICT_TRUE (ip4_1->dst_address.as_u32 == t1->src.ip4.as_u32))
375 goto next1; /* valid packet */
376 if (PREDICT_FALSE (ip4_address_is_multicast (&ip4_1->dst_address)))
377 {
378 key4_1.src = ip4_1->dst_address.as_u32;
379 key4_1.vni = vxlan1->vni_reserved;
380 /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */
381 p1 = hash_get (vxm->vxlan4_tunnel_by_key, key4_1.as_u64);
382 if (PREDICT_TRUE (p1 != NULL))
383 {
384 mt1 = pool_elt_at_index (vxm->tunnels, p1[0]);
385 goto next1; /* valid packet */
386 }
387 }
388 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
389 next1 = VXLAN_INPUT_NEXT_DROP;
390 goto trace1;
391
392 } else /* !is_ip4 */ {
Chris Luke99cb3352016-04-26 10:49:53 -0400393 key6_1.src.as_u64[0] = ip6_1->src_address.as_u64[0];
394 key6_1.src.as_u64[1] = ip6_1->src_address.as_u64[1];
395 key6_1.vni = vxlan1->vni_reserved;
396
John Lo56912c82016-12-08 16:10:02 -0500397 /* Make sure VXLAN tunnel exist according to packet SIP and VNI */
Chris Luke99cb3352016-04-26 10:49:53 -0400398 if (PREDICT_FALSE (memcmp(&key6_1, &last_key6, sizeof(last_key6)) != 0))
Chris Lukec7949152016-05-31 10:42:14 -0400399 {
400 p1 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_1);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700401
John Lo56912c82016-12-08 16:10:02 -0500402 if (PREDICT_FALSE (p1 == NULL))
Chris Lukec7949152016-05-31 10:42:14 -0400403 {
404 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
405 next1 = VXLAN_INPUT_NEXT_DROP;
406 goto trace1;
407 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700408
Chris Lukec7949152016-05-31 10:42:14 -0400409 clib_memcpy (&last_key6, &key6_1, sizeof(key6_1));
410 tunnel_index1 = last_tunnel_index = p1[0];
411 }
412 else
413 tunnel_index1 = last_tunnel_index;
John Lo56912c82016-12-08 16:10:02 -0500414 t1 = pool_elt_at_index (vxm->tunnels, tunnel_index1);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700415
John Lo2b81eb82017-01-30 13:12:10 -0500416 /* Validate VXLAN tunnel encap-fib index agaist packet */
417 if (PREDICT_FALSE (validate_vxlan_fib (b1, t1, is_ip4) == 0))
418 {
419 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
420 next1 = VXLAN_INPUT_NEXT_DROP;
421 goto trace1;
422 }
423
John Lo56912c82016-12-08 16:10:02 -0500424 /* Validate VXLAN tunnel SIP against packet DIP */
425 if (PREDICT_TRUE (ip6_address_is_equal (&ip6_1->dst_address,
426 &t1->src.ip6)))
427 goto next1; /* valid packet */
428 if (PREDICT_FALSE (ip6_address_is_multicast (&ip6_1->dst_address)))
429 {
430 key6_1.src.as_u64[0] = ip6_1->dst_address.as_u64[0];
431 key6_1.src.as_u64[1] = ip6_1->dst_address.as_u64[1];
432 key6_1.vni = vxlan1->vni_reserved;
John Lo2b81eb82017-01-30 13:12:10 -0500433 p1 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_1);
John Lo56912c82016-12-08 16:10:02 -0500434 if (PREDICT_TRUE (p1 != NULL))
435 {
436 mt1 = pool_elt_at_index (vxm->tunnels, p1[0]);
437 goto next1; /* valid packet */
438 }
439 }
440 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
441 next1 = VXLAN_INPUT_NEXT_DROP;
442 goto trace1;
443 }
444
445 next1:
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800446 next1 = t1->decap_next_index;
Chris Luke99cb3352016-04-26 10:49:53 -0400447 sw_if_index1 = t1->sw_if_index;
448 len1 = vlib_buffer_length_in_chain (vm, b1);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700449
450 /* Required to make the l2 tag push / pop code work on l2 subifs */
Chris Luke99cb3352016-04-26 10:49:53 -0400451 if (PREDICT_TRUE(next1 == VXLAN_INPUT_NEXT_L2_INPUT))
452 vnet_update_l2_len (b1);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700453
John Lo56912c82016-12-08 16:10:02 -0500454 /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700455 vnet_buffer(b1)->sw_if_index[VLIB_RX] = sw_if_index1;
John Lo56912c82016-12-08 16:10:02 -0500456 sw_if_index1 = (mt1) ? mt1->sw_if_index : sw_if_index1;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700457
458 pkts_decapsulated ++;
Chris Luke99cb3352016-04-26 10:49:53 -0400459 stats_n_packets += 1;
460 stats_n_bytes += len1;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700461
462 /* Batch stats increment on the same vxlan tunnel so counter
463 is not incremented per packet */
464 if (PREDICT_FALSE (sw_if_index1 != stats_sw_if_index))
465 {
466 stats_n_packets -= 1;
467 stats_n_bytes -= len1;
468 if (stats_n_packets)
469 vlib_increment_combined_counter
470 (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
471 cpu_index, stats_sw_if_index,
472 stats_n_packets, stats_n_bytes);
473 stats_n_packets = 1;
474 stats_n_bytes = len1;
475 stats_sw_if_index = sw_if_index1;
476 }
477
478 trace1:
479 b1->error = error1 ? node->errors[error1] : 0;
480
481 if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
482 {
483 vxlan_rx_trace_t *tr
484 = vlib_add_trace (vm, node, b1, sizeof (*tr));
485 tr->next_index = next1;
486 tr->error = error1;
487 tr->tunnel_index = tunnel_index1;
488 tr->vni = vnet_get_vni (vxlan1);
489 }
490
491 vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
492 to_next, n_left_to_next,
493 bi0, bi1, next0, next1);
494 }
495
496 while (n_left_from > 0 && n_left_to_next > 0)
497 {
498 u32 bi0;
499 vlib_buffer_t * b0;
500 u32 next0;
Chris Luke99cb3352016-04-26 10:49:53 -0400501 ip4_header_t * ip4_0;
502 ip6_header_t * ip6_0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700503 vxlan_header_t * vxlan0;
504 uword * p0;
505 u32 tunnel_index0;
John Lo56912c82016-12-08 16:10:02 -0500506 vxlan_tunnel_t * t0, * mt0 = NULL;
Chris Luke99cb3352016-04-26 10:49:53 -0400507 vxlan4_tunnel_key_t key4_0;
508 vxlan6_tunnel_key_t key6_0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700509 u32 error0;
510 u32 sw_if_index0, len0;
511
512 bi0 = from[0];
513 to_next[0] = bi0;
514 from += 1;
515 to_next += 1;
516 n_left_from -= 1;
517 n_left_to_next -= 1;
518
519 b0 = vlib_get_buffer (vm, bi0);
520
521 /* udp leaves current_data pointing at the vxlan header */
522 vxlan0 = vlib_buffer_get_current (b0);
Chris Luke99cb3352016-04-26 10:49:53 -0400523 if (is_ip4) {
John Loc42912d2016-11-07 18:30:47 -0500524 vlib_buffer_advance
525 (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
Chris Luke99cb3352016-04-26 10:49:53 -0400526 ip4_0 = vlib_buffer_get_current (b0);
527 } else {
528 vlib_buffer_advance
529 (b0, -(word)(sizeof(udp_header_t)+sizeof(ip6_header_t)));
530 ip6_0 = vlib_buffer_get_current (b0);
531 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700532
533 /* pop (ip, udp, vxlan) */
Chris Luke99cb3352016-04-26 10:49:53 -0400534 if (is_ip4) {
535 vlib_buffer_advance
536 (b0, sizeof(*ip4_0)+sizeof(udp_header_t)+sizeof(*vxlan0));
537 } else {
John Loc42912d2016-11-07 18:30:47 -0500538 vlib_buffer_advance
Chris Luke99cb3352016-04-26 10:49:53 -0400539 (b0, sizeof(*ip6_0)+sizeof(udp_header_t)+sizeof(*vxlan0));
540 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700541
542 tunnel_index0 = ~0;
543 error0 = 0;
544
John Loc42912d2016-11-07 18:30:47 -0500545 if (PREDICT_FALSE (vxlan0->flags != VXLAN_FLAGS_I))
546 {
547 error0 = VXLAN_ERROR_BAD_FLAGS;
548 next0 = VXLAN_INPUT_NEXT_DROP;
549 goto trace00;
550 }
551
Chris Luke99cb3352016-04-26 10:49:53 -0400552 if (is_ip4) {
553 key4_0.src = ip4_0->src_address.as_u32;
554 key4_0.vni = vxlan0->vni_reserved;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700555
John Lo56912c82016-12-08 16:10:02 -0500556 /* Make sure unicast VXLAN tunnel exist by packet SIP and VNI */
Chris Luke99cb3352016-04-26 10:49:53 -0400557 if (PREDICT_FALSE (key4_0.as_u64 != last_key4.as_u64))
558 {
559 p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64);
John Lo56912c82016-12-08 16:10:02 -0500560 if (PREDICT_FALSE (p0 == NULL))
Chris Luke99cb3352016-04-26 10:49:53 -0400561 {
562 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
563 next0 = VXLAN_INPUT_NEXT_DROP;
564 goto trace00;
565 }
Chris Luke99cb3352016-04-26 10:49:53 -0400566 last_key4.as_u64 = key4_0.as_u64;
567 tunnel_index0 = last_tunnel_index = p0[0];
568 }
569 else
570 tunnel_index0 = last_tunnel_index;
John Lo56912c82016-12-08 16:10:02 -0500571 t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0);
572
John Lo2b81eb82017-01-30 13:12:10 -0500573 /* Validate VXLAN tunnel encap-fib index agaist packet */
574 if (PREDICT_FALSE (validate_vxlan_fib (b0, t0, is_ip4) == 0))
575 {
576 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
577 next0 = VXLAN_INPUT_NEXT_DROP;
578 goto trace00;
579 }
580
John Lo56912c82016-12-08 16:10:02 -0500581 /* Validate VXLAN tunnel SIP against packet DIP */
582 if (PREDICT_TRUE (ip4_0->dst_address.as_u32 == t0->src.ip4.as_u32))
583 goto next00; /* valid packet */
584 if (PREDICT_FALSE (ip4_address_is_multicast (&ip4_0->dst_address)))
585 {
586 key4_0.src = ip4_0->dst_address.as_u32;
587 key4_0.vni = vxlan0->vni_reserved;
588 /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */
589 p0 = hash_get (vxm->vxlan4_tunnel_by_key, key4_0.as_u64);
590 if (PREDICT_TRUE (p0 != NULL))
591 {
592 mt0 = pool_elt_at_index (vxm->tunnels, p0[0]);
593 goto next00; /* valid packet */
594 }
595 }
596 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
597 next0 = VXLAN_INPUT_NEXT_DROP;
598 goto trace00;
599
Chris Luke99cb3352016-04-26 10:49:53 -0400600 } else /* !is_ip4 */ {
601 key6_0.src.as_u64[0] = ip6_0->src_address.as_u64[0];
602 key6_0.src.as_u64[1] = ip6_0->src_address.as_u64[1];
603 key6_0.vni = vxlan0->vni_reserved;
604
John Lo56912c82016-12-08 16:10:02 -0500605 /* Make sure VXLAN tunnel exist according to packet SIP and VNI */
Chris Luke99cb3352016-04-26 10:49:53 -0400606 if (PREDICT_FALSE (memcmp(&key6_0, &last_key6, sizeof(last_key6)) != 0))
Chris Lukec7949152016-05-31 10:42:14 -0400607 {
608 p0 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_0);
John Lo56912c82016-12-08 16:10:02 -0500609 if (PREDICT_FALSE (p0 == NULL))
Chris Lukec7949152016-05-31 10:42:14 -0400610 {
611 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
612 next0 = VXLAN_INPUT_NEXT_DROP;
613 goto trace00;
614 }
Chris Lukec7949152016-05-31 10:42:14 -0400615 clib_memcpy (&last_key6, &key6_0, sizeof(key6_0));
616 tunnel_index0 = last_tunnel_index = p0[0];
617 }
618 else
619 tunnel_index0 = last_tunnel_index;
John Lo56912c82016-12-08 16:10:02 -0500620 t0 = pool_elt_at_index (vxm->tunnels, tunnel_index0);
621
John Lo2b81eb82017-01-30 13:12:10 -0500622 /* Validate VXLAN tunnel encap-fib index agaist packet */
623 if (PREDICT_FALSE (validate_vxlan_fib (b0, t0, is_ip4) == 0))
624 {
625 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
626 next0 = VXLAN_INPUT_NEXT_DROP;
627 goto trace00;
628 }
629
John Lo56912c82016-12-08 16:10:02 -0500630 /* Validate VXLAN tunnel SIP against packet DIP */
631 if (PREDICT_TRUE (ip6_address_is_equal (&ip6_0->dst_address,
632 &t0->src.ip6)))
633 goto next00; /* valid packet */
634 if (PREDICT_FALSE (ip6_address_is_multicast (&ip6_0->dst_address)))
635 {
636 key6_0.src.as_u64[0] = ip6_0->dst_address.as_u64[0];
637 key6_0.src.as_u64[1] = ip6_0->dst_address.as_u64[1];
638 key6_0.vni = vxlan0->vni_reserved;
John Lo2b81eb82017-01-30 13:12:10 -0500639 p0 = hash_get_mem (vxm->vxlan6_tunnel_by_key, &key6_0);
John Lo56912c82016-12-08 16:10:02 -0500640 if (PREDICT_TRUE (p0 != NULL))
641 {
642 mt0 = pool_elt_at_index (vxm->tunnels, p0[0]);
643 goto next00; /* valid packet */
644 }
645 }
646 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
647 next0 = VXLAN_INPUT_NEXT_DROP;
648 goto trace00;
Chris Luke99cb3352016-04-26 10:49:53 -0400649 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700650
John Lo56912c82016-12-08 16:10:02 -0500651 next00:
Hongjun Nibeb4bf72016-11-25 00:03:46 +0800652 next0 = t0->decap_next_index;
Chris Luke99cb3352016-04-26 10:49:53 -0400653 sw_if_index0 = t0->sw_if_index;
654 len0 = vlib_buffer_length_in_chain (vm, b0);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700655
656 /* Required to make the l2 tag push / pop code work on l2 subifs */
Chris Luke99cb3352016-04-26 10:49:53 -0400657 if (PREDICT_TRUE(next0 == VXLAN_INPUT_NEXT_L2_INPUT))
658 vnet_update_l2_len (b0);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700659
John Lo56912c82016-12-08 16:10:02 -0500660 /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */
Ed Warnickecb9cada2015-12-08 15:45:58 -0700661 vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0;
John Lo56912c82016-12-08 16:10:02 -0500662 sw_if_index0 = (mt0) ? mt0->sw_if_index : sw_if_index0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700663
664 pkts_decapsulated ++;
Chris Luke99cb3352016-04-26 10:49:53 -0400665 stats_n_packets += 1;
666 stats_n_bytes += len0;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700667
668 /* Batch stats increment on the same vxlan tunnel so counter
669 is not incremented per packet */
670 if (PREDICT_FALSE (sw_if_index0 != stats_sw_if_index))
671 {
672 stats_n_packets -= 1;
673 stats_n_bytes -= len0;
674 if (stats_n_packets)
675 vlib_increment_combined_counter
676 (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
677 cpu_index, stats_sw_if_index,
678 stats_n_packets, stats_n_bytes);
679 stats_n_packets = 1;
680 stats_n_bytes = len0;
681 stats_sw_if_index = sw_if_index0;
682 }
683
684 trace00:
685 b0->error = error0 ? node->errors[error0] : 0;
686
687 if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
688 {
689 vxlan_rx_trace_t *tr
690 = vlib_add_trace (vm, node, b0, sizeof (*tr));
691 tr->next_index = next0;
692 tr->error = error0;
693 tr->tunnel_index = tunnel_index0;
694 tr->vni = vnet_get_vni (vxlan0);
695 }
696 vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
697 to_next, n_left_to_next,
698 bi0, next0);
699 }
700
701 vlib_put_next_frame (vm, node, next_index, n_left_to_next);
702 }
703 /* Do we still need this now that tunnel tx stats is kept? */
John Lo13e3d452016-08-09 19:20:51 -0400704 vlib_node_increment_counter (vm, is_ip4?
705 vxlan4_input_node.index:vxlan6_input_node.index,
Ed Warnickecb9cada2015-12-08 15:45:58 -0700706 VXLAN_ERROR_DECAPSULATED,
707 pkts_decapsulated);
708
709 /* Increment any remaining batch stats */
710 if (stats_n_packets)
711 {
712 vlib_increment_combined_counter
713 (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
714 cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
715 node->runtime_data[0] = stats_sw_if_index;
716 }
717
718 return from_frame->n_vectors;
719}
720
Chris Luke99cb3352016-04-26 10:49:53 -0400721static uword
722vxlan4_input (vlib_main_t * vm,
723 vlib_node_runtime_t * node,
724 vlib_frame_t * from_frame)
725{
726 return vxlan_input(vm, node, from_frame, /* is_ip4 */ 1);
727}
728
729static uword
730vxlan6_input (vlib_main_t * vm,
731 vlib_node_runtime_t * node,
732 vlib_frame_t * from_frame)
733{
734 return vxlan_input(vm, node, from_frame, /* is_ip4 */ 0);
735}
736
Ed Warnickecb9cada2015-12-08 15:45:58 -0700737static char * vxlan_error_strings[] = {
738#define vxlan_error(n,s) s,
739#include <vnet/vxlan/vxlan_error.def>
740#undef vxlan_error
741#undef _
742};
743
Chris Luke99cb3352016-04-26 10:49:53 -0400744VLIB_REGISTER_NODE (vxlan4_input_node) = {
745 .function = vxlan4_input,
746 .name = "vxlan4-input",
747 /* Takes a vector of packets. */
748 .vector_size = sizeof (u32),
749
750 .n_errors = VXLAN_N_ERROR,
751 .error_strings = vxlan_error_strings,
752
753 .n_next_nodes = VXLAN_INPUT_N_NEXT,
754 .next_nodes = {
755#define _(s,n) [VXLAN_INPUT_NEXT_##s] = n,
756 foreach_vxlan_input_next
757#undef _
758 },
759
760//temp .format_buffer = format_vxlan_header,
761 .format_trace = format_vxlan_rx_trace,
762 // $$$$ .unformat_buffer = unformat_vxlan_header,
763};
764
Damjan Marion1c80e832016-05-11 23:07:18 +0200765VLIB_NODE_FUNCTION_MULTIARCH (vxlan4_input_node, vxlan4_input)
766
Chris Luke99cb3352016-04-26 10:49:53 -0400767VLIB_REGISTER_NODE (vxlan6_input_node) = {
768 .function = vxlan6_input,
769 .name = "vxlan6-input",
Ed Warnickecb9cada2015-12-08 15:45:58 -0700770 /* Takes a vector of packets. */
771 .vector_size = sizeof (u32),
772
773 .n_errors = VXLAN_N_ERROR,
774 .error_strings = vxlan_error_strings,
775
776 .n_next_nodes = VXLAN_INPUT_N_NEXT,
777 .next_nodes = {
778#define _(s,n) [VXLAN_INPUT_NEXT_##s] = n,
779 foreach_vxlan_input_next
780#undef _
781 },
782
783//temp .format_buffer = format_vxlan_header,
784 .format_trace = format_vxlan_rx_trace,
785 // $$$$ .unformat_buffer = unformat_vxlan_header,
786};
Damjan Marion1c80e832016-05-11 23:07:18 +0200787
788VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_input_node, vxlan6_input)
789
John Lo37682e12016-11-30 12:51:39 -0500790
791typedef enum {
792 IP_VXLAN_BYPASS_NEXT_DROP,
793 IP_VXLAN_BYPASS_NEXT_VXLAN,
794 IP_VXLAN_BYPASS_N_NEXT,
795} ip_vxan_bypass_next_t;
796
797always_inline uword
798ip_vxlan_bypass_inline (vlib_main_t * vm,
John Lo2b81eb82017-01-30 13:12:10 -0500799 vlib_node_runtime_t * node,
800 vlib_frame_t * frame,
801 u32 is_ip4)
John Lo37682e12016-11-30 12:51:39 -0500802{
803 vxlan_main_t * vxm = &vxlan_main;
804 u32 * from, * to_next, n_left_from, n_left_to_next, next_index;
805 vlib_node_runtime_t * error_node = vlib_node_get_runtime (vm, ip4_input_node.index);
806 ip4_address_t addr4; /* last IPv4 address matching a local VTEP address */
807 ip6_address_t addr6; /* last IPv6 address matching a local VTEP address */
808
809 from = vlib_frame_vector_args (frame);
810 n_left_from = frame->n_vectors;
811 next_index = node->cached_next_index;
812
813 if (node->flags & VLIB_NODE_FLAG_TRACE)
814 ip4_forward_next_trace (vm, node, frame, VLIB_TX);
815
816 if (is_ip4) addr4.data_u32 = ~0;
817 else ip6_address_set_zero (&addr6);
818
819 while (n_left_from > 0)
820 {
821 vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
822
823 while (n_left_from >= 4 && n_left_to_next >= 2)
824 {
825 vlib_buffer_t * b0, * b1;
John Lo2b81eb82017-01-30 13:12:10 -0500826 ip4_header_t * ip40, * ip41;
827 ip6_header_t * ip60, * ip61;
John Lo37682e12016-11-30 12:51:39 -0500828 udp_header_t * udp0, * udp1;
829 u32 bi0, ip_len0, udp_len0, flags0, next0;
830 u32 bi1, ip_len1, udp_len1, flags1, next1;
831 i32 len_diff0, len_diff1;
832 u8 error0, good_udp0, proto0;
833 u8 error1, good_udp1, proto1;
834
835 /* Prefetch next iteration. */
836 {
837 vlib_buffer_t * p2, * p3;
838
839 p2 = vlib_get_buffer (vm, from[2]);
840 p3 = vlib_get_buffer (vm, from[3]);
841
842 vlib_prefetch_buffer_header (p2, LOAD);
843 vlib_prefetch_buffer_header (p3, LOAD);
844
845 CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
846 CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
847 }
848
849 bi0 = to_next[0] = from[0];
850 bi1 = to_next[1] = from[1];
851 from += 2;
852 n_left_from -= 2;
853 to_next += 2;
854 n_left_to_next -= 2;
855
856 b0 = vlib_get_buffer (vm, bi0);
857 b1 = vlib_get_buffer (vm, bi1);
John Lo2b81eb82017-01-30 13:12:10 -0500858 if (is_ip4)
859 {
860 ip40 = vlib_buffer_get_current (b0);
861 ip41 = vlib_buffer_get_current (b1);
862 }
863 else
864 {
865 ip60 = vlib_buffer_get_current (b0);
866 ip61 = vlib_buffer_get_current (b1);
867 }
John Lo37682e12016-11-30 12:51:39 -0500868
869 /* Setup packet for next IP feature */
870 vnet_feature_next(vnet_buffer(b0)->sw_if_index[VLIB_RX], &next0, b0);
871 vnet_feature_next(vnet_buffer(b1)->sw_if_index[VLIB_RX], &next1, b1);
872
John Lo2b81eb82017-01-30 13:12:10 -0500873 if (is_ip4)
874 {
875 /* Treat IP frag packets as "experimental" protocol for now
876 until support of IP frag reassembly is implemented */
877 proto0 = ip4_is_fragment(ip40) ? 0xfe : ip40->protocol;
878 proto1 = ip4_is_fragment(ip41) ? 0xfe : ip41->protocol;
879 }
880 else
881 {
882 proto0 = ip60->protocol;
883 proto1 = ip61->protocol;
884 }
John Lo37682e12016-11-30 12:51:39 -0500885
886 /* Process packet 0 */
887 if (proto0 != IP_PROTOCOL_UDP)
888 goto exit0; /* not UDP packet */
889
John Lo2b81eb82017-01-30 13:12:10 -0500890 if (is_ip4)
891 udp0 = ip4_next_header (ip40);
892 else
893 udp0 = ip6_next_header (ip60);
894
John Lo37682e12016-11-30 12:51:39 -0500895 if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan))
896 goto exit0; /* not VXLAN packet */
897
John Lo2b81eb82017-01-30 13:12:10 -0500898 /* Validate DIP against VTEPs*/
John Lo37682e12016-11-30 12:51:39 -0500899 if (is_ip4)
900 {
John Lo2b81eb82017-01-30 13:12:10 -0500901 if (addr4.as_u32 != ip40->dst_address.as_u32)
John Lo37682e12016-11-30 12:51:39 -0500902 {
John Lo2b81eb82017-01-30 13:12:10 -0500903 if (!hash_get (vxm->vtep4, ip40->dst_address.as_u32))
904 goto exit0; /* no local VTEP for VXLAN packet */
905 addr4 = ip40->dst_address;
John Lo37682e12016-11-30 12:51:39 -0500906 }
907 }
John Lo2b81eb82017-01-30 13:12:10 -0500908 else
909 {
910 if (!ip6_address_is_equal (&addr6, &ip60->dst_address))
911 {
912 if (!hash_get_mem (vxm->vtep6, &ip60->dst_address))
913 goto exit0; /* no local VTEP for VXLAN packet */
914 addr6 = ip60->dst_address;
915 }
916 }
John Lo37682e12016-11-30 12:51:39 -0500917
918 flags0 = b0->flags;
919 good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0;
920
921 /* Don't verify UDP checksum for packets with explicit zero checksum. */
922 good_udp0 |= udp0->checksum == 0;
923
924 /* Verify UDP length */
John Lo2b81eb82017-01-30 13:12:10 -0500925 if (is_ip4)
926 ip_len0 = clib_net_to_host_u16 (ip40->length);
927 else
928 ip_len0 = clib_net_to_host_u16 (ip60->payload_length);
John Lo37682e12016-11-30 12:51:39 -0500929 udp_len0 = clib_net_to_host_u16 (udp0->length);
John Lo37682e12016-11-30 12:51:39 -0500930 len_diff0 = ip_len0 - udp_len0;
931
932 /* Verify UDP checksum */
933 if (PREDICT_FALSE (!good_udp0))
934 {
John Lo2b81eb82017-01-30 13:12:10 -0500935 if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0)
936 {
937 if (is_ip4)
938 flags0 = ip4_tcp_udp_validate_checksum (vm, b0);
939 else
940 flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0);
941 good_udp0 =
942 (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0;
943 }
John Lo37682e12016-11-30 12:51:39 -0500944 }
945
John Lo2b81eb82017-01-30 13:12:10 -0500946 if (is_ip4)
947 {
948 error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM;
949 error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH;
950 }
951 else
952 {
953 error0 = good_udp0 ? 0 : IP6_ERROR_UDP_CHECKSUM;
954 error0 = (len_diff0 >= 0) ? error0 : IP6_ERROR_UDP_LENGTH;
955 }
John Lo37682e12016-11-30 12:51:39 -0500956
957 next0 = error0 ?
958 IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN;
959 b0->error = error0 ? error_node->errors[error0] : 0;
960
John Lo2b81eb82017-01-30 13:12:10 -0500961 /* vxlan-input node expect current at VXLAN header */
962 if (is_ip4)
963 vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t));
964 else
965 vlib_buffer_advance (b0, sizeof(ip6_header_t)+sizeof(udp_header_t));
966
John Lo37682e12016-11-30 12:51:39 -0500967 exit0:
968 /* Process packet 1 */
969 if (proto1 != IP_PROTOCOL_UDP)
970 goto exit1; /* not UDP packet */
971
John Lo2b81eb82017-01-30 13:12:10 -0500972 if (is_ip4)
973 udp1 = ip4_next_header (ip41);
974 else
975 udp1 = ip6_next_header (ip61);
976
John Lo37682e12016-11-30 12:51:39 -0500977 if (udp1->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan))
978 goto exit1; /* not VXLAN packet */
979
John Lo2b81eb82017-01-30 13:12:10 -0500980 /* Validate DIP against VTEPs*/
John Lo37682e12016-11-30 12:51:39 -0500981 if (is_ip4)
982 {
John Lo2b81eb82017-01-30 13:12:10 -0500983 if (addr4.as_u32 != ip41->dst_address.as_u32)
John Lo37682e12016-11-30 12:51:39 -0500984 {
John Lo2b81eb82017-01-30 13:12:10 -0500985 if (!hash_get (vxm->vtep4, ip41->dst_address.as_u32))
986 goto exit1; /* no local VTEP for VXLAN packet */
987 addr4 = ip41->dst_address;
988 }
John Lo37682e12016-11-30 12:51:39 -0500989 }
John Lo2b81eb82017-01-30 13:12:10 -0500990 else
991 {
992 if (!ip6_address_is_equal (&addr6, &ip61->dst_address))
993 {
994 if (!hash_get_mem (vxm->vtep6, &ip61->dst_address))
995 goto exit1; /* no local VTEP for VXLAN packet */
996 addr6 = ip61->dst_address;
997 }
998 }
John Lo37682e12016-11-30 12:51:39 -0500999
1000 flags1 = b1->flags;
1001 good_udp1 = (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0;
1002
1003 /* Don't verify UDP checksum for packets with explicit zero checksum. */
1004 good_udp1 |= udp1->checksum == 0;
1005
1006 /* Verify UDP length */
John Lo2b81eb82017-01-30 13:12:10 -05001007 if (is_ip4)
1008 ip_len1 = clib_net_to_host_u16 (ip41->length);
1009 else
1010 ip_len1 = clib_net_to_host_u16 (ip61->payload_length);
John Lo37682e12016-11-30 12:51:39 -05001011 udp_len1 = clib_net_to_host_u16 (udp1->length);
John Lo37682e12016-11-30 12:51:39 -05001012 len_diff1 = ip_len1 - udp_len1;
1013
1014 /* Verify UDP checksum */
1015 if (PREDICT_FALSE (!good_udp1))
1016 {
John Lo2b81eb82017-01-30 13:12:10 -05001017 if ((flags1 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0)
1018 {
1019 if (is_ip4)
1020 flags1 = ip4_tcp_udp_validate_checksum (vm, b1);
1021 else
1022 flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, b1);
1023 good_udp1 =
1024 (flags1 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0;
1025 }
John Lo37682e12016-11-30 12:51:39 -05001026 }
1027
John Lo2b81eb82017-01-30 13:12:10 -05001028 if (is_ip4)
1029 {
1030 error1 = good_udp1 ? 0 : IP4_ERROR_UDP_CHECKSUM;
1031 error1 = (len_diff1 >= 0) ? error1 : IP4_ERROR_UDP_LENGTH;
1032 }
1033 else
1034 {
Eyal Baria93ea422017-02-01 13:36:15 +02001035 error1 = good_udp1 ? 0 : IP6_ERROR_UDP_CHECKSUM;
1036 error1 = (len_diff1 >= 0) ? error1 : IP6_ERROR_UDP_LENGTH;
John Lo2b81eb82017-01-30 13:12:10 -05001037 }
John Lo37682e12016-11-30 12:51:39 -05001038
1039 next1 = error1 ?
1040 IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN;
1041 b1->error = error1 ? error_node->errors[error1] : 0;
1042
John Lo2b81eb82017-01-30 13:12:10 -05001043 /* vxlan-input node expect current at VXLAN header */
1044 if (is_ip4)
1045 vlib_buffer_advance (b1, sizeof(ip4_header_t)+sizeof(udp_header_t));
1046 else
1047 vlib_buffer_advance (b1, sizeof(ip6_header_t)+sizeof(udp_header_t));
1048
John Lo37682e12016-11-30 12:51:39 -05001049 exit1:
1050 vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
1051 to_next, n_left_to_next,
1052 bi0, bi1, next0, next1);
1053 }
1054
1055 while (n_left_from > 0 && n_left_to_next > 0)
1056 {
1057 vlib_buffer_t * b0;
John Lo2b81eb82017-01-30 13:12:10 -05001058 ip4_header_t * ip40;
1059 ip6_header_t * ip60;
John Lo37682e12016-11-30 12:51:39 -05001060 udp_header_t * udp0;
1061 u32 bi0, ip_len0, udp_len0, flags0, next0;
1062 i32 len_diff0;
1063 u8 error0, good_udp0, proto0;
1064
1065 bi0 = to_next[0] = from[0];
1066 from += 1;
1067 n_left_from -= 1;
1068 to_next += 1;
1069 n_left_to_next -= 1;
1070
1071 b0 = vlib_get_buffer (vm, bi0);
John Lo2b81eb82017-01-30 13:12:10 -05001072 if (is_ip4)
1073 ip40 = vlib_buffer_get_current (b0);
1074 else
1075 ip60 = vlib_buffer_get_current (b0);
John Lo37682e12016-11-30 12:51:39 -05001076
1077 /* Setup packet for next IP feature */
1078 vnet_feature_next(vnet_buffer(b0)->sw_if_index[VLIB_RX], &next0, b0);
1079
John Lo2b81eb82017-01-30 13:12:10 -05001080 if (is_ip4)
1081 /* Treat IP4 frag packets as "experimental" protocol for now
1082 until support of IP frag reassembly is implemented */
1083 proto0 = ip4_is_fragment(ip40) ? 0xfe : ip40->protocol;
1084 else
1085 proto0 = ip60->protocol;
John Lo37682e12016-11-30 12:51:39 -05001086
1087 if (proto0 != IP_PROTOCOL_UDP)
1088 goto exit; /* not UDP packet */
1089
John Lo2b81eb82017-01-30 13:12:10 -05001090 if (is_ip4)
1091 udp0 = ip4_next_header (ip40);
1092 else
1093 udp0 = ip6_next_header (ip60);
1094
John Lo37682e12016-11-30 12:51:39 -05001095 if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan))
1096 goto exit; /* not VXLAN packet */
1097
John Lo2b81eb82017-01-30 13:12:10 -05001098 /* Validate DIP against VTEPs*/
John Lo37682e12016-11-30 12:51:39 -05001099 if (is_ip4)
1100 {
John Lo2b81eb82017-01-30 13:12:10 -05001101 if (addr4.as_u32 != ip40->dst_address.as_u32)
John Lo37682e12016-11-30 12:51:39 -05001102 {
John Lo2b81eb82017-01-30 13:12:10 -05001103 if (!hash_get (vxm->vtep4, ip40->dst_address.as_u32))
1104 goto exit; /* no local VTEP for VXLAN packet */
1105 addr4 = ip40->dst_address;
1106 }
John Lo37682e12016-11-30 12:51:39 -05001107 }
John Lo2b81eb82017-01-30 13:12:10 -05001108 else
1109 {
1110 if (!ip6_address_is_equal (&addr6, &ip60->dst_address))
1111 {
1112 if (!hash_get_mem (vxm->vtep6, &ip60->dst_address))
1113 goto exit; /* no local VTEP for VXLAN packet */
1114 addr6 = ip60->dst_address;
1115 }
1116 }
John Lo37682e12016-11-30 12:51:39 -05001117
1118 flags0 = b0->flags;
1119 good_udp0 = (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0;
1120
1121 /* Don't verify UDP checksum for packets with explicit zero checksum. */
1122 good_udp0 |= udp0->checksum == 0;
1123
1124 /* Verify UDP length */
John Lo2b81eb82017-01-30 13:12:10 -05001125 if (is_ip4)
1126 ip_len0 = clib_net_to_host_u16 (ip40->length);
1127 else
1128 ip_len0 = clib_net_to_host_u16 (ip60->payload_length);
John Lo37682e12016-11-30 12:51:39 -05001129 udp_len0 = clib_net_to_host_u16 (udp0->length);
John Lo37682e12016-11-30 12:51:39 -05001130 len_diff0 = ip_len0 - udp_len0;
1131
1132 /* Verify UDP checksum */
1133 if (PREDICT_FALSE (!good_udp0))
1134 {
John Lo2b81eb82017-01-30 13:12:10 -05001135 if ((flags0 & IP_BUFFER_L4_CHECKSUM_COMPUTED) == 0)
1136 {
1137 if (is_ip4)
1138 flags0 = ip4_tcp_udp_validate_checksum (vm, b0);
1139 else
1140 flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0);
1141 good_udp0 =
1142 (flags0 & IP_BUFFER_L4_CHECKSUM_CORRECT) != 0;
1143 }
John Lo37682e12016-11-30 12:51:39 -05001144 }
1145
John Lo2b81eb82017-01-30 13:12:10 -05001146 if (is_ip4)
1147 {
1148 error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM;
1149 error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH;
1150 }
1151 else
1152 {
1153 error0 = good_udp0 ? 0 : IP6_ERROR_UDP_CHECKSUM;
1154 error0 = (len_diff0 >= 0) ? error0 : IP6_ERROR_UDP_LENGTH;
1155 }
John Lo37682e12016-11-30 12:51:39 -05001156
1157 next0 = error0 ?
1158 IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN;
1159 b0->error = error0 ? error_node->errors[error0] : 0;
1160
John Lo2b81eb82017-01-30 13:12:10 -05001161 /* vxlan-input node expect current at VXLAN header */
1162 if (is_ip4)
1163 vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t));
1164 else
1165 vlib_buffer_advance (b0, sizeof(ip6_header_t)+sizeof(udp_header_t));
1166
John Lo37682e12016-11-30 12:51:39 -05001167 exit:
1168 vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
1169 to_next, n_left_to_next,
1170 bi0, next0);
1171 }
1172
1173 vlib_put_next_frame (vm, node, next_index, n_left_to_next);
1174 }
1175
1176 return frame->n_vectors;
1177}
1178
1179static uword
1180ip4_vxlan_bypass (vlib_main_t * vm,
1181 vlib_node_runtime_t * node,
1182 vlib_frame_t * frame)
1183{
1184 return ip_vxlan_bypass_inline (vm, node, frame, /* is_ip4 */ 1);
1185}
1186
1187VLIB_REGISTER_NODE (ip4_vxlan_bypass_node) = {
1188 .function = ip4_vxlan_bypass,
1189 .name = "ip4-vxlan-bypass",
1190 .vector_size = sizeof (u32),
1191
1192 .n_next_nodes = IP_VXLAN_BYPASS_N_NEXT,
1193 .next_nodes = {
1194 [IP_VXLAN_BYPASS_NEXT_DROP] = "error-drop",
1195 [IP_VXLAN_BYPASS_NEXT_VXLAN] = "vxlan4-input",
1196 },
1197
1198 .format_buffer = format_ip4_header,
1199 .format_trace = format_ip4_forward_next_trace,
1200};
1201
1202VLIB_NODE_FUNCTION_MULTIARCH (ip4_vxlan_bypass_node,ip4_vxlan_bypass)
1203
John Lo37682e12016-11-30 12:51:39 -05001204/* Dummy init function to get us linked in. */
1205clib_error_t * ip4_vxlan_bypass_init (vlib_main_t * vm)
1206{ return 0; }
1207
1208VLIB_INIT_FUNCTION (ip4_vxlan_bypass_init);
John Lo2b81eb82017-01-30 13:12:10 -05001209
1210static uword
1211ip6_vxlan_bypass (vlib_main_t * vm,
1212 vlib_node_runtime_t * node,
1213 vlib_frame_t * frame)
1214{
1215 return ip_vxlan_bypass_inline (vm, node, frame, /* is_ip4 */ 0);
1216}
1217
1218VLIB_REGISTER_NODE (ip6_vxlan_bypass_node) = {
1219 .function = ip6_vxlan_bypass,
1220 .name = "ip6-vxlan-bypass",
1221 .vector_size = sizeof (u32),
1222
1223 .n_next_nodes = IP_VXLAN_BYPASS_N_NEXT,
1224 .next_nodes = {
1225 [IP_VXLAN_BYPASS_NEXT_DROP] = "error-drop",
1226 [IP_VXLAN_BYPASS_NEXT_VXLAN] = "vxlan6-input",
1227 },
1228
1229 .format_buffer = format_ip6_header,
1230 .format_trace = format_ip6_forward_next_trace,
1231};
1232
1233VLIB_NODE_FUNCTION_MULTIARCH (ip6_vxlan_bypass_node,ip6_vxlan_bypass)
1234
1235/* Dummy init function to get us linked in. */
1236clib_error_t * ip6_vxlan_bypass_init (vlib_main_t * vm)
1237{ return 0; }
1238
1239VLIB_INIT_FUNCTION (ip6_vxlan_bypass_init);