blob: 20df5e2c1311f11ff16cc5d2b5e553f758af4c42 [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 {
Eyal Bari0f4b1842018-04-12 12:39:51 +030045 s = format (s, "VXLAN decap error - tunnel for vni %d does not exist",
John Loc42912d2016-11-07 18:30:47 -050046 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
Eyal Bari0fa56782018-06-04 12:25:05 +030052buf_fib_index (vlib_buffer_t *b, u32 is_ip4)
John Lo2b81eb82017-01-30 13:12:10 -050053{
Eyal Bari0fa56782018-06-04 12:25:05 +030054 u32 sw_if_index = vnet_buffer (b)->sw_if_index[VLIB_TX];
55 if (sw_if_index != (u32) ~ 0)
56 return sw_if_index;
John Lo2b81eb82017-01-30 13:12:10 -050057
Eyal Barifb663012017-10-19 15:27:51 +030058 u32 * fib_index_by_sw_if_index = is_ip4 ?
59 ip4_main.fib_index_by_sw_if_index : ip6_main.fib_index_by_sw_if_index;
Eyal Bari0fa56782018-06-04 12:25:05 +030060 sw_if_index = vnet_buffer (b)->sw_if_index[VLIB_RX];
John Lo2b81eb82017-01-30 13:12:10 -050061
Eyal Bari0fa56782018-06-04 12:25:05 +030062 return vec_elt (fib_index_by_sw_if_index, sw_if_index);
John Lo2b81eb82017-01-30 13:12:10 -050063}
64
Eyal Baridd47eca2018-07-08 08:15:56 +030065typedef vxlan4_tunnel_key_t last_tunnel_cache4;
Eyal Bari0f4b1842018-04-12 12:39:51 +030066
Eyal Bari0f4b1842018-04-12 12:39:51 +030067always_inline vxlan_tunnel_t *
68vxlan4_find_tunnel (vxlan_main_t * vxm, last_tunnel_cache4 * cache,
Eyal Bari0fa56782018-06-04 12:25:05 +030069 u32 fib_index, ip4_header_t * ip4_0, vxlan_header_t * vxlan0,
Eyal Bari0f4b1842018-04-12 12:39:51 +030070 vxlan_tunnel_t ** stats_t0)
71{
72 /* Make sure VXLAN tunnel exist according to packet SIP and VNI */
Eyal Baridd47eca2018-07-08 08:15:56 +030073 vxlan4_tunnel_key_t key4 = {
74 .key = {
75 [0] = ip4_0->src_address.as_u32,
76 [1] = (((u64) fib_index) << 32) | vxlan0->vni_reserved,
77 }
Eyal Bari0f4b1842018-04-12 12:39:51 +030078 };
79
Eyal Baridd47eca2018-07-08 08:15:56 +030080 if (PREDICT_FALSE (clib_bihash_key_compare_16_8 (key4.key, cache->key) == 0))
Eyal Bari0f4b1842018-04-12 12:39:51 +030081 {
Eyal Baridd47eca2018-07-08 08:15:56 +030082 int rv = clib_bihash_search_inline_16_8 (&vxm->vxlan4_tunnel_by_key, &key4);
83 if (PREDICT_FALSE (rv != 0))
Eyal Bari0f4b1842018-04-12 12:39:51 +030084 return 0;
85
Eyal Baridd47eca2018-07-08 08:15:56 +030086 *cache = key4;
Eyal Bari0f4b1842018-04-12 12:39:51 +030087 }
Eyal Baridd47eca2018-07-08 08:15:56 +030088 vxlan_tunnel_t * t0 = pool_elt_at_index (vxm->tunnels, cache->value);
Eyal Bari0fa56782018-06-04 12:25:05 +030089
Eyal Bari0f4b1842018-04-12 12:39:51 +030090 /* Validate VXLAN tunnel SIP against packet DIP */
91 if (PREDICT_TRUE (ip4_0->dst_address.as_u32 == t0->src.ip4.as_u32))
92 *stats_t0 = t0;
93 else
94 {
95 /* try multicast */
96 if (PREDICT_TRUE (!ip4_address_is_multicast (&ip4_0->dst_address)))
97 return 0;
98
Eyal Baridd47eca2018-07-08 08:15:56 +030099 key4.key[0] = ip4_0->dst_address.as_u32;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300100 /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */
Eyal Baridd47eca2018-07-08 08:15:56 +0300101 int rv = clib_bihash_search_inline_16_8 (&vxm->vxlan4_tunnel_by_key, &key4);
102 if (PREDICT_FALSE (rv != 0))
Eyal Bari0f4b1842018-04-12 12:39:51 +0300103 return 0;
Eyal Baridd47eca2018-07-08 08:15:56 +0300104
105 *stats_t0 = pool_elt_at_index (vxm->tunnels, key4.value);
Eyal Bari0f4b1842018-04-12 12:39:51 +0300106 }
107
108 return t0;
109}
110
Eyal Bari0fa56782018-06-04 12:25:05 +0300111typedef vxlan6_tunnel_key_t last_tunnel_cache6;
112
Eyal Bari0f4b1842018-04-12 12:39:51 +0300113always_inline vxlan_tunnel_t *
114vxlan6_find_tunnel (vxlan_main_t * vxm, last_tunnel_cache6 * cache,
Eyal Bari0fa56782018-06-04 12:25:05 +0300115 u32 fib_index, ip6_header_t * ip6_0, vxlan_header_t * vxlan0,
Eyal Bari0f4b1842018-04-12 12:39:51 +0300116 vxlan_tunnel_t ** stats_t0)
117{
118 /* Make sure VXLAN tunnel exist according to packet SIP and VNI */
Eyal Bari0fa56782018-06-04 12:25:05 +0300119
120 vxlan6_tunnel_key_t key6 = {
121 .key = {
122 [0] = ip6_0->src_address.as_u64[0],
123 [1] = ip6_0->src_address.as_u64[1],
Eyal Baridd47eca2018-07-08 08:15:56 +0300124 [2] = (((u64) fib_index) << 32) | vxlan0->vni_reserved,
Eyal Bari0fa56782018-06-04 12:25:05 +0300125 }
Eyal Bari0f4b1842018-04-12 12:39:51 +0300126 };
127
Eyal Baridd47eca2018-07-08 08:15:56 +0300128 if (PREDICT_FALSE (clib_bihash_key_compare_24_8 (key6.key, cache->key) == 0))
Eyal Bari0f4b1842018-04-12 12:39:51 +0300129 {
Eyal Baridd47eca2018-07-08 08:15:56 +0300130 int rv = clib_bihash_search_inline_24_8 (&vxm->vxlan6_tunnel_by_key, &key6);
Eyal Bari0fa56782018-06-04 12:25:05 +0300131 if (PREDICT_FALSE (rv != 0))
Eyal Bari0f4b1842018-04-12 12:39:51 +0300132 return 0;
133
Eyal Bari0fa56782018-06-04 12:25:05 +0300134 *cache = key6;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300135 }
Eyal Bari0fa56782018-06-04 12:25:05 +0300136 vxlan_tunnel_t * t0 = pool_elt_at_index (vxm->tunnels, cache->value);
Eyal Bari0f4b1842018-04-12 12:39:51 +0300137
138 /* Validate VXLAN tunnel SIP against packet DIP */
139 if (PREDICT_TRUE (ip6_address_is_equal (&ip6_0->dst_address, &t0->src.ip6)))
140 *stats_t0 = t0;
141 else
142 {
143 /* try multicast */
144 if (PREDICT_TRUE (!ip6_address_is_multicast (&ip6_0->dst_address)))
145 return 0;
146
Eyal Bari0f4b1842018-04-12 12:39:51 +0300147 /* Make sure mcast VXLAN tunnel exist by packet DIP and VNI */
Eyal Bari0fa56782018-06-04 12:25:05 +0300148 key6.key[0] = ip6_0->dst_address.as_u64[0];
149 key6.key[1] = ip6_0->dst_address.as_u64[1];
Eyal Baridd47eca2018-07-08 08:15:56 +0300150 int rv = clib_bihash_search_inline_24_8 (&vxm->vxlan6_tunnel_by_key, &key6);
Eyal Bari0fa56782018-06-04 12:25:05 +0300151 if (PREDICT_FALSE (rv != 0))
Eyal Bari0f4b1842018-04-12 12:39:51 +0300152 return 0;
Eyal Bari0fa56782018-06-04 12:25:05 +0300153
154 *stats_t0 = pool_elt_at_index (vxm->tunnels, key6.value);
Eyal Bari0f4b1842018-04-12 12:39:51 +0300155 }
156
157 return t0;
158}
159
Chris Luke99cb3352016-04-26 10:49:53 -0400160always_inline uword
Ed Warnickecb9cada2015-12-08 15:45:58 -0700161vxlan_input (vlib_main_t * vm,
162 vlib_node_runtime_t * node,
Chris Luke99cb3352016-04-26 10:49:53 -0400163 vlib_frame_t * from_frame,
John Lo37682e12016-11-30 12:51:39 -0500164 u32 is_ip4)
Ed Warnickecb9cada2015-12-08 15:45:58 -0700165{
Ed Warnickecb9cada2015-12-08 15:45:58 -0700166 vxlan_main_t * vxm = &vxlan_main;
167 vnet_main_t * vnm = vxm->vnet_main;
168 vnet_interface_main_t * im = &vnm->interface_main;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300169 vlib_combined_counter_main_t * rx_counter = im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX;
170 vlib_combined_counter_main_t * drop_counter = im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_DROP;
Eyal Baridd47eca2018-07-08 08:15:56 +0300171 last_tunnel_cache4 last4;
Eyal Bari0fa56782018-06-04 12:25:05 +0300172 last_tunnel_cache6 last6;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700173 u32 pkts_decapsulated = 0;
Damjan Marion586afd72017-04-05 19:18:20 +0200174 u32 thread_index = vlib_get_thread_index();
Ed Warnickecb9cada2015-12-08 15:45:58 -0700175
Dave Barachf9c231e2016-08-05 10:10:18 -0400176 if (is_ip4)
Eyal Baridd47eca2018-07-08 08:15:56 +0300177 memset (&last4, 0xff, sizeof last4);
Dave Barachf9c231e2016-08-05 10:10:18 -0400178 else
Eyal Bari0fa56782018-06-04 12:25:05 +0300179 memset (&last6, 0xff, sizeof last6);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700180
Eyal Barifb663012017-10-19 15:27:51 +0300181 u32 next_index = node->cached_next_index;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700182
Eyal Barifb663012017-10-19 15:27:51 +0300183 u32 * from = vlib_frame_vector_args (from_frame);
184 u32 n_left_from = from_frame->n_vectors;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700185
186 while (n_left_from > 0)
187 {
Eyal Barifb663012017-10-19 15:27:51 +0300188 u32 * to_next, n_left_to_next;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300189 vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
Eyal Barifb663012017-10-19 15:27:51 +0300190
Ed Warnickecb9cada2015-12-08 15:45:58 -0700191 while (n_left_from >= 4 && n_left_to_next >= 2)
192 {
Ed Warnickecb9cada2015-12-08 15:45:58 -0700193 /* Prefetch next iteration. */
194 {
195 vlib_buffer_t * p2, * p3;
196
197 p2 = vlib_get_buffer (vm, from[2]);
198 p3 = vlib_get_buffer (vm, from[3]);
199
200 vlib_prefetch_buffer_header (p2, LOAD);
201 vlib_prefetch_buffer_header (p3, LOAD);
202
203 CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
204 CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
205 }
206
Eyal Bari0f4b1842018-04-12 12:39:51 +0300207 u32 bi0 = to_next[0] = from[0];
208 u32 bi1 = to_next[1] = from[1];
Ed Warnickecb9cada2015-12-08 15:45:58 -0700209 from += 2;
210 to_next += 2;
211 n_left_to_next -= 2;
212 n_left_from -= 2;
213
Eyal Bari0f4b1842018-04-12 12:39:51 +0300214 vlib_buffer_t * b0, * b1;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700215 b0 = vlib_get_buffer (vm, bi0);
216 b1 = vlib_get_buffer (vm, bi1);
217
218 /* udp leaves current_data pointing at the vxlan header */
Eyal Barifb663012017-10-19 15:27:51 +0300219 void * cur0 = vlib_buffer_get_current (b0);
220 void * cur1 = vlib_buffer_get_current (b1);
221 vxlan_header_t * vxlan0 = cur0;
222 vxlan_header_t * vxlan1 = cur1;
223
224 ip4_header_t * ip4_0, * ip4_1;
225 ip6_header_t * ip6_0, * ip6_1;
Chris Luke99cb3352016-04-26 10:49:53 -0400226 if (is_ip4) {
Eyal Barifb663012017-10-19 15:27:51 +0300227 ip4_0 = cur0 - sizeof(udp_header_t) - sizeof(ip4_header_t);
228 ip4_1 = cur1 - sizeof(udp_header_t) - sizeof(ip4_header_t);
Chris Luke99cb3352016-04-26 10:49:53 -0400229 } else {
Eyal Barifb663012017-10-19 15:27:51 +0300230 ip6_0 = cur0 - sizeof(udp_header_t) - sizeof(ip6_header_t);
231 ip6_1 = cur1 - sizeof(udp_header_t) - sizeof(ip6_header_t);
Chris Luke99cb3352016-04-26 10:49:53 -0400232 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700233
Eyal Bari0f4b1842018-04-12 12:39:51 +0300234 /* pop vxlan */
Eyal Barifb663012017-10-19 15:27:51 +0300235 vlib_buffer_advance (b0, sizeof *vxlan0);
236 vlib_buffer_advance (b1, sizeof *vxlan1);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700237
Eyal Bari0fa56782018-06-04 12:25:05 +0300238 u32 fi0 = buf_fib_index(b0, is_ip4);
239 u32 fi1 = buf_fib_index(b1, is_ip4);
240
Eyal Bari0f4b1842018-04-12 12:39:51 +0300241 vxlan_tunnel_t * t0, * stats_t0;
242 vxlan_tunnel_t * t1, * stats_t1;
243 if (is_ip4)
244 {
Eyal Bari0fa56782018-06-04 12:25:05 +0300245 t0 = vxlan4_find_tunnel (vxm, &last4, fi0, ip4_0, vxlan0, &stats_t0);
246 t1 = vxlan4_find_tunnel (vxm, &last4, fi1, ip4_1, vxlan1, &stats_t1);
Eyal Bari0f4b1842018-04-12 12:39:51 +0300247 }
248 else
249 {
Eyal Bari0fa56782018-06-04 12:25:05 +0300250 t0 = vxlan6_find_tunnel (vxm, &last6, fi0, ip6_0, vxlan0, &stats_t0);
251 t1 = vxlan6_find_tunnel (vxm, &last6, fi1, ip6_1, vxlan1, &stats_t1);
Chris Luke99cb3352016-04-26 10:49:53 -0400252 }
Ed Warnickecb9cada2015-12-08 15:45:58 -0700253
Eyal Barifb663012017-10-19 15:27:51 +0300254 u32 len0 = vlib_buffer_length_in_chain (vm, b0);
Eyal Barifb663012017-10-19 15:27:51 +0300255 u32 len1 = vlib_buffer_length_in_chain (vm, b1);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700256
Eyal Bari0f4b1842018-04-12 12:39:51 +0300257 u32 next0, next1;
258 u8 error0 = 0, error1 = 0;
259 /* Validate VXLAN tunnel encap-fib index agaist packet */
Eyal Bari0fa56782018-06-04 12:25:05 +0300260 if (PREDICT_FALSE (t0 == 0 || vxlan0->flags != VXLAN_FLAGS_I))
Ed Warnickecb9cada2015-12-08 15:45:58 -0700261 {
Eyal Bari0f4b1842018-04-12 12:39:51 +0300262 next0 = VXLAN_INPUT_NEXT_DROP;
263
264 if (t0 != 0 && vxlan0->flags != VXLAN_FLAGS_I)
265 {
266 error0 = VXLAN_ERROR_BAD_FLAGS;
267 vlib_increment_combined_counter
268 (drop_counter, thread_index, stats_t0->sw_if_index, 1, len0);
269 }
270 else
271 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
272 b0->error = node->errors[error0];
273 }
274 else
275 {
276 next0 = t0->decap_next_index;
277
278 /* Required to make the l2 tag push / pop code work on l2 subifs */
279 if (PREDICT_TRUE(next0 == VXLAN_INPUT_NEXT_L2_INPUT))
280 vnet_update_l2_len (b0);
281
282 /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */
283 vnet_buffer(b0)->sw_if_index[VLIB_RX] = t0->sw_if_index;
284 vlib_increment_combined_counter
285 (rx_counter, thread_index, stats_t0->sw_if_index, 1, len0);
286 pkts_decapsulated++;
287 }
288
289 /* Validate VXLAN tunnel encap-fib index agaist packet */
Eyal Bari0fa56782018-06-04 12:25:05 +0300290 if (PREDICT_FALSE (t1 == 0 || vxlan1->flags != VXLAN_FLAGS_I))
Eyal Bari0f4b1842018-04-12 12:39:51 +0300291 {
292 next1 = VXLAN_INPUT_NEXT_DROP;
293
294 if (t1 != 0 && vxlan1->flags != VXLAN_FLAGS_I)
295 {
296 error1 = VXLAN_ERROR_BAD_FLAGS;
297 vlib_increment_combined_counter
298 (drop_counter, thread_index, stats_t1->sw_if_index, 1, len1);
299 }
300 else
301 error1 = VXLAN_ERROR_NO_SUCH_TUNNEL;
302 b1->error = node->errors[error1];
303 }
304 else
305 {
306 next1 = t1->decap_next_index;
307
308 /* Required to make the l2 tag push / pop code work on l2 subifs */
309 if (PREDICT_TRUE(next1 == VXLAN_INPUT_NEXT_L2_INPUT))
310 vnet_update_l2_len (b1);
311
312 /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */
313 vnet_buffer(b1)->sw_if_index[VLIB_RX] = t1->sw_if_index;
314 pkts_decapsulated++;
315
316 vlib_increment_combined_counter
317 (rx_counter, thread_index, stats_t1->sw_if_index, 1, len1);
318 }
319
320 if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
321 {
322 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
323 tr->next_index = next0;
324 tr->error = error0;
325 tr->tunnel_index = t0 == 0 ? ~0 : t0 - vxm->tunnels;
326 tr->vni = vnet_get_vni (vxlan0);
327 }
328 if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
329 {
330 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b1, sizeof (*tr));
Ed Warnickecb9cada2015-12-08 15:45:58 -0700331 tr->next_index = next1;
332 tr->error = error1;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300333 tr->tunnel_index = t1 == 0 ? ~0 : t1 - vxm->tunnels;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700334 tr->vni = vnet_get_vni (vxlan1);
335 }
336
337 vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
338 to_next, n_left_to_next,
339 bi0, bi1, next0, next1);
340 }
341
342 while (n_left_from > 0 && n_left_to_next > 0)
343 {
Eyal Bari0f4b1842018-04-12 12:39:51 +0300344 u32 bi0 = to_next[0] = from[0];
Ed Warnickecb9cada2015-12-08 15:45:58 -0700345 from += 1;
346 to_next += 1;
347 n_left_from -= 1;
348 n_left_to_next -= 1;
349
Eyal Bari0f4b1842018-04-12 12:39:51 +0300350 vlib_buffer_t * b0 = vlib_get_buffer (vm, bi0);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700351
352 /* udp leaves current_data pointing at the vxlan header */
Eyal Barifb663012017-10-19 15:27:51 +0300353 void * cur0 = vlib_buffer_get_current (b0);
354 vxlan_header_t * vxlan0 = cur0;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300355 ip4_header_t * ip4_0;
356 ip6_header_t * ip6_0;
Eyal Barifb663012017-10-19 15:27:51 +0300357 if (is_ip4)
358 ip4_0 = cur0 -sizeof(udp_header_t) - sizeof(ip4_header_t);
359 else
360 ip6_0 = cur0 -sizeof(udp_header_t) - sizeof(ip6_header_t);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700361
362 /* pop (ip, udp, vxlan) */
Eyal Barifb663012017-10-19 15:27:51 +0300363 vlib_buffer_advance (b0, sizeof(*vxlan0));
Ed Warnickecb9cada2015-12-08 15:45:58 -0700364
Eyal Bari0fa56782018-06-04 12:25:05 +0300365 u32 fi0 = buf_fib_index(b0, is_ip4);
366
Eyal Bari0f4b1842018-04-12 12:39:51 +0300367 vxlan_tunnel_t * t0, * stats_t0;
368 if (is_ip4)
Eyal Bari0fa56782018-06-04 12:25:05 +0300369 t0 = vxlan4_find_tunnel (vxm, &last4, fi0, ip4_0, vxlan0, &stats_t0);
Eyal Bari0f4b1842018-04-12 12:39:51 +0300370 else
Eyal Bari0fa56782018-06-04 12:25:05 +0300371 t0 = vxlan6_find_tunnel (vxm, &last6, fi0, ip6_0, vxlan0, &stats_t0);
Ed Warnickecb9cada2015-12-08 15:45:58 -0700372
Eyal Bari0f4b1842018-04-12 12:39:51 +0300373 uword len0 = vlib_buffer_length_in_chain (vm, b0);
John Loc42912d2016-11-07 18:30:47 -0500374
Eyal Bari0f4b1842018-04-12 12:39:51 +0300375 u32 next0;
376 u8 error0 = 0;
377 /* Validate VXLAN tunnel encap-fib index agaist packet */
Eyal Bari0fa56782018-06-04 12:25:05 +0300378 if (PREDICT_FALSE (t0 == 0 || vxlan0->flags != VXLAN_FLAGS_I))
Ed Warnickecb9cada2015-12-08 15:45:58 -0700379 {
Eyal Bari0f4b1842018-04-12 12:39:51 +0300380 next0 = VXLAN_INPUT_NEXT_DROP;
381
382 if (t0 != 0 && vxlan0->flags != VXLAN_FLAGS_I)
383 {
384 error0 = VXLAN_ERROR_BAD_FLAGS;
385 vlib_increment_combined_counter
386 (drop_counter, thread_index, stats_t0->sw_if_index, 1, len0);
387 }
388 else
389 error0 = VXLAN_ERROR_NO_SUCH_TUNNEL;
390 b0->error = node->errors[error0];
391 }
392 else
393 {
394 next0 = t0->decap_next_index;
395
396 /* Required to make the l2 tag push / pop code work on l2 subifs */
397 if (PREDICT_TRUE(next0 == VXLAN_INPUT_NEXT_L2_INPUT))
398 vnet_update_l2_len (b0);
399
400 /* Set packet input sw_if_index to unicast VXLAN tunnel for learning */
401 vnet_buffer(b0)->sw_if_index[VLIB_RX] = t0->sw_if_index;
402 pkts_decapsulated++;
403
404 vlib_increment_combined_counter
405 (rx_counter, thread_index, stats_t0->sw_if_index, 1, len0);
406 }
407
408 if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
409 {
410 vxlan_rx_trace_t *tr
Ed Warnickecb9cada2015-12-08 15:45:58 -0700411 = vlib_add_trace (vm, node, b0, sizeof (*tr));
412 tr->next_index = next0;
413 tr->error = error0;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300414 tr->tunnel_index = t0 == 0 ? ~0 : t0 - vxm->tunnels;
Ed Warnickecb9cada2015-12-08 15:45:58 -0700415 tr->vni = vnet_get_vni (vxlan0);
416 }
417 vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
418 to_next, n_left_to_next,
419 bi0, next0);
420 }
421
422 vlib_put_next_frame (vm, node, next_index, n_left_to_next);
423 }
Eyal Barifb663012017-10-19 15:27:51 +0300424 /* Do we still need this now that tunnel tx stats is kept? */
425 u32 node_idx = is_ip4 ? vxlan4_input_node.index : vxlan6_input_node.index;
426 vlib_node_increment_counter (vm, node_idx, VXLAN_ERROR_DECAPSULATED,
427 pkts_decapsulated);
428
Ed Warnickecb9cada2015-12-08 15:45:58 -0700429 return from_frame->n_vectors;
430}
431
Chris Luke99cb3352016-04-26 10:49:53 -0400432static uword
433vxlan4_input (vlib_main_t * vm,
434 vlib_node_runtime_t * node,
435 vlib_frame_t * from_frame)
436{
437 return vxlan_input(vm, node, from_frame, /* is_ip4 */ 1);
438}
439
440static uword
441vxlan6_input (vlib_main_t * vm,
442 vlib_node_runtime_t * node,
443 vlib_frame_t * from_frame)
444{
445 return vxlan_input(vm, node, from_frame, /* is_ip4 */ 0);
446}
447
Ed Warnickecb9cada2015-12-08 15:45:58 -0700448static char * vxlan_error_strings[] = {
449#define vxlan_error(n,s) s,
450#include <vnet/vxlan/vxlan_error.def>
451#undef vxlan_error
452#undef _
453};
454
Chris Luke99cb3352016-04-26 10:49:53 -0400455VLIB_REGISTER_NODE (vxlan4_input_node) = {
456 .function = vxlan4_input,
457 .name = "vxlan4-input",
458 /* Takes a vector of packets. */
459 .vector_size = sizeof (u32),
460
461 .n_errors = VXLAN_N_ERROR,
462 .error_strings = vxlan_error_strings,
463
464 .n_next_nodes = VXLAN_INPUT_N_NEXT,
465 .next_nodes = {
466#define _(s,n) [VXLAN_INPUT_NEXT_##s] = n,
467 foreach_vxlan_input_next
468#undef _
469 },
470
471//temp .format_buffer = format_vxlan_header,
472 .format_trace = format_vxlan_rx_trace,
473 // $$$$ .unformat_buffer = unformat_vxlan_header,
474};
475
Damjan Marion1c80e832016-05-11 23:07:18 +0200476VLIB_NODE_FUNCTION_MULTIARCH (vxlan4_input_node, vxlan4_input)
477
Chris Luke99cb3352016-04-26 10:49:53 -0400478VLIB_REGISTER_NODE (vxlan6_input_node) = {
479 .function = vxlan6_input,
480 .name = "vxlan6-input",
Ed Warnickecb9cada2015-12-08 15:45:58 -0700481 /* Takes a vector of packets. */
482 .vector_size = sizeof (u32),
483
484 .n_errors = VXLAN_N_ERROR,
485 .error_strings = vxlan_error_strings,
486
487 .n_next_nodes = VXLAN_INPUT_N_NEXT,
488 .next_nodes = {
489#define _(s,n) [VXLAN_INPUT_NEXT_##s] = n,
490 foreach_vxlan_input_next
491#undef _
492 },
493
494//temp .format_buffer = format_vxlan_header,
495 .format_trace = format_vxlan_rx_trace,
496 // $$$$ .unformat_buffer = unformat_vxlan_header,
497};
Damjan Marion1c80e832016-05-11 23:07:18 +0200498
499VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_input_node, vxlan6_input)
500
John Lo37682e12016-11-30 12:51:39 -0500501
502typedef enum {
503 IP_VXLAN_BYPASS_NEXT_DROP,
504 IP_VXLAN_BYPASS_NEXT_VXLAN,
505 IP_VXLAN_BYPASS_N_NEXT,
506} ip_vxan_bypass_next_t;
507
508always_inline uword
509ip_vxlan_bypass_inline (vlib_main_t * vm,
John Lo2b81eb82017-01-30 13:12:10 -0500510 vlib_node_runtime_t * node,
511 vlib_frame_t * frame,
512 u32 is_ip4)
John Lo37682e12016-11-30 12:51:39 -0500513{
514 vxlan_main_t * vxm = &vxlan_main;
515 u32 * from, * to_next, n_left_from, n_left_to_next, next_index;
516 vlib_node_runtime_t * error_node = vlib_node_get_runtime (vm, ip4_input_node.index);
517 ip4_address_t addr4; /* last IPv4 address matching a local VTEP address */
518 ip6_address_t addr6; /* last IPv6 address matching a local VTEP address */
519
520 from = vlib_frame_vector_args (frame);
521 n_left_from = frame->n_vectors;
522 next_index = node->cached_next_index;
523
524 if (node->flags & VLIB_NODE_FLAG_TRACE)
525 ip4_forward_next_trace (vm, node, frame, VLIB_TX);
526
527 if (is_ip4) addr4.data_u32 = ~0;
528 else ip6_address_set_zero (&addr6);
529
530 while (n_left_from > 0)
531 {
532 vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
533
534 while (n_left_from >= 4 && n_left_to_next >= 2)
535 {
536 vlib_buffer_t * b0, * b1;
John Lo2b81eb82017-01-30 13:12:10 -0500537 ip4_header_t * ip40, * ip41;
538 ip6_header_t * ip60, * ip61;
John Lo37682e12016-11-30 12:51:39 -0500539 udp_header_t * udp0, * udp1;
540 u32 bi0, ip_len0, udp_len0, flags0, next0;
541 u32 bi1, ip_len1, udp_len1, flags1, next1;
542 i32 len_diff0, len_diff1;
543 u8 error0, good_udp0, proto0;
544 u8 error1, good_udp1, proto1;
545
546 /* Prefetch next iteration. */
547 {
548 vlib_buffer_t * p2, * p3;
549
550 p2 = vlib_get_buffer (vm, from[2]);
551 p3 = vlib_get_buffer (vm, from[3]);
552
553 vlib_prefetch_buffer_header (p2, LOAD);
554 vlib_prefetch_buffer_header (p3, LOAD);
555
556 CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
557 CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
558 }
559
560 bi0 = to_next[0] = from[0];
561 bi1 = to_next[1] = from[1];
562 from += 2;
563 n_left_from -= 2;
564 to_next += 2;
565 n_left_to_next -= 2;
566
567 b0 = vlib_get_buffer (vm, bi0);
568 b1 = vlib_get_buffer (vm, bi1);
John Lo2b81eb82017-01-30 13:12:10 -0500569 if (is_ip4)
570 {
571 ip40 = vlib_buffer_get_current (b0);
572 ip41 = vlib_buffer_get_current (b1);
573 }
574 else
575 {
576 ip60 = vlib_buffer_get_current (b0);
577 ip61 = vlib_buffer_get_current (b1);
578 }
John Lo37682e12016-11-30 12:51:39 -0500579
580 /* Setup packet for next IP feature */
Damjan Marion7d98a122018-07-19 20:42:08 +0200581 vnet_feature_next(&next0, b0);
582 vnet_feature_next(&next1, b1);
John Lo37682e12016-11-30 12:51:39 -0500583
John Lo2b81eb82017-01-30 13:12:10 -0500584 if (is_ip4)
585 {
586 /* Treat IP frag packets as "experimental" protocol for now
587 until support of IP frag reassembly is implemented */
588 proto0 = ip4_is_fragment(ip40) ? 0xfe : ip40->protocol;
589 proto1 = ip4_is_fragment(ip41) ? 0xfe : ip41->protocol;
590 }
591 else
592 {
593 proto0 = ip60->protocol;
594 proto1 = ip61->protocol;
595 }
John Lo37682e12016-11-30 12:51:39 -0500596
597 /* Process packet 0 */
598 if (proto0 != IP_PROTOCOL_UDP)
599 goto exit0; /* not UDP packet */
600
John Lo2b81eb82017-01-30 13:12:10 -0500601 if (is_ip4)
602 udp0 = ip4_next_header (ip40);
603 else
604 udp0 = ip6_next_header (ip60);
605
John Lo37682e12016-11-30 12:51:39 -0500606 if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan))
607 goto exit0; /* not VXLAN packet */
608
John Lo2b81eb82017-01-30 13:12:10 -0500609 /* Validate DIP against VTEPs*/
Eyal Bari0f4b1842018-04-12 12:39:51 +0300610 if (is_ip4)
John Lo37682e12016-11-30 12:51:39 -0500611 {
John Lo2b81eb82017-01-30 13:12:10 -0500612 if (addr4.as_u32 != ip40->dst_address.as_u32)
John Lo37682e12016-11-30 12:51:39 -0500613 {
John Lo2b81eb82017-01-30 13:12:10 -0500614 if (!hash_get (vxm->vtep4, ip40->dst_address.as_u32))
615 goto exit0; /* no local VTEP for VXLAN packet */
616 addr4 = ip40->dst_address;
John Lo37682e12016-11-30 12:51:39 -0500617 }
618 }
John Lo2b81eb82017-01-30 13:12:10 -0500619 else
620 {
621 if (!ip6_address_is_equal (&addr6, &ip60->dst_address))
622 {
623 if (!hash_get_mem (vxm->vtep6, &ip60->dst_address))
624 goto exit0; /* no local VTEP for VXLAN packet */
625 addr6 = ip60->dst_address;
626 }
627 }
John Lo37682e12016-11-30 12:51:39 -0500628
629 flags0 = b0->flags;
Damjan Marion213b5aa2017-07-13 21:19:27 +0200630 good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
John Lo37682e12016-11-30 12:51:39 -0500631
632 /* Don't verify UDP checksum for packets with explicit zero checksum. */
633 good_udp0 |= udp0->checksum == 0;
634
635 /* Verify UDP length */
John Lo2b81eb82017-01-30 13:12:10 -0500636 if (is_ip4)
637 ip_len0 = clib_net_to_host_u16 (ip40->length);
638 else
639 ip_len0 = clib_net_to_host_u16 (ip60->payload_length);
John Lo37682e12016-11-30 12:51:39 -0500640 udp_len0 = clib_net_to_host_u16 (udp0->length);
John Lo37682e12016-11-30 12:51:39 -0500641 len_diff0 = ip_len0 - udp_len0;
642
643 /* Verify UDP checksum */
644 if (PREDICT_FALSE (!good_udp0))
645 {
Damjan Marion213b5aa2017-07-13 21:19:27 +0200646 if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0)
John Lo2b81eb82017-01-30 13:12:10 -0500647 {
648 if (is_ip4)
649 flags0 = ip4_tcp_udp_validate_checksum (vm, b0);
650 else
651 flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0);
652 good_udp0 =
Damjan Marion213b5aa2017-07-13 21:19:27 +0200653 (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
John Lo2b81eb82017-01-30 13:12:10 -0500654 }
John Lo37682e12016-11-30 12:51:39 -0500655 }
656
John Lo2b81eb82017-01-30 13:12:10 -0500657 if (is_ip4)
658 {
659 error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM;
660 error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH;
661 }
662 else
663 {
664 error0 = good_udp0 ? 0 : IP6_ERROR_UDP_CHECKSUM;
665 error0 = (len_diff0 >= 0) ? error0 : IP6_ERROR_UDP_LENGTH;
666 }
John Lo37682e12016-11-30 12:51:39 -0500667
Eyal Bari0f4b1842018-04-12 12:39:51 +0300668 next0 = error0 ?
John Lo37682e12016-11-30 12:51:39 -0500669 IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN;
670 b0->error = error0 ? error_node->errors[error0] : 0;
671
John Lo2b81eb82017-01-30 13:12:10 -0500672 /* vxlan-input node expect current at VXLAN header */
673 if (is_ip4)
674 vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t));
675 else
676 vlib_buffer_advance (b0, sizeof(ip6_header_t)+sizeof(udp_header_t));
677
John Lo37682e12016-11-30 12:51:39 -0500678 exit0:
679 /* Process packet 1 */
680 if (proto1 != IP_PROTOCOL_UDP)
681 goto exit1; /* not UDP packet */
682
John Lo2b81eb82017-01-30 13:12:10 -0500683 if (is_ip4)
684 udp1 = ip4_next_header (ip41);
685 else
686 udp1 = ip6_next_header (ip61);
687
John Lo37682e12016-11-30 12:51:39 -0500688 if (udp1->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan))
689 goto exit1; /* not VXLAN packet */
690
John Lo2b81eb82017-01-30 13:12:10 -0500691 /* Validate DIP against VTEPs*/
Eyal Bari0f4b1842018-04-12 12:39:51 +0300692 if (is_ip4)
John Lo37682e12016-11-30 12:51:39 -0500693 {
John Lo2b81eb82017-01-30 13:12:10 -0500694 if (addr4.as_u32 != ip41->dst_address.as_u32)
John Lo37682e12016-11-30 12:51:39 -0500695 {
John Lo2b81eb82017-01-30 13:12:10 -0500696 if (!hash_get (vxm->vtep4, ip41->dst_address.as_u32))
697 goto exit1; /* no local VTEP for VXLAN packet */
698 addr4 = ip41->dst_address;
699 }
John Lo37682e12016-11-30 12:51:39 -0500700 }
John Lo2b81eb82017-01-30 13:12:10 -0500701 else
702 {
703 if (!ip6_address_is_equal (&addr6, &ip61->dst_address))
704 {
705 if (!hash_get_mem (vxm->vtep6, &ip61->dst_address))
706 goto exit1; /* no local VTEP for VXLAN packet */
707 addr6 = ip61->dst_address;
708 }
709 }
John Lo37682e12016-11-30 12:51:39 -0500710
711 flags1 = b1->flags;
Damjan Marion213b5aa2017-07-13 21:19:27 +0200712 good_udp1 = (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
John Lo37682e12016-11-30 12:51:39 -0500713
714 /* Don't verify UDP checksum for packets with explicit zero checksum. */
715 good_udp1 |= udp1->checksum == 0;
716
717 /* Verify UDP length */
John Lo2b81eb82017-01-30 13:12:10 -0500718 if (is_ip4)
719 ip_len1 = clib_net_to_host_u16 (ip41->length);
720 else
721 ip_len1 = clib_net_to_host_u16 (ip61->payload_length);
John Lo37682e12016-11-30 12:51:39 -0500722 udp_len1 = clib_net_to_host_u16 (udp1->length);
John Lo37682e12016-11-30 12:51:39 -0500723 len_diff1 = ip_len1 - udp_len1;
724
725 /* Verify UDP checksum */
726 if (PREDICT_FALSE (!good_udp1))
727 {
Damjan Marion213b5aa2017-07-13 21:19:27 +0200728 if ((flags1 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0)
John Lo2b81eb82017-01-30 13:12:10 -0500729 {
730 if (is_ip4)
731 flags1 = ip4_tcp_udp_validate_checksum (vm, b1);
732 else
733 flags1 = ip6_tcp_udp_icmp_validate_checksum (vm, b1);
734 good_udp1 =
Damjan Marion213b5aa2017-07-13 21:19:27 +0200735 (flags1 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
John Lo2b81eb82017-01-30 13:12:10 -0500736 }
John Lo37682e12016-11-30 12:51:39 -0500737 }
738
John Lo2b81eb82017-01-30 13:12:10 -0500739 if (is_ip4)
740 {
741 error1 = good_udp1 ? 0 : IP4_ERROR_UDP_CHECKSUM;
742 error1 = (len_diff1 >= 0) ? error1 : IP4_ERROR_UDP_LENGTH;
743 }
744 else
745 {
Eyal Baria93ea422017-02-01 13:36:15 +0200746 error1 = good_udp1 ? 0 : IP6_ERROR_UDP_CHECKSUM;
747 error1 = (len_diff1 >= 0) ? error1 : IP6_ERROR_UDP_LENGTH;
John Lo2b81eb82017-01-30 13:12:10 -0500748 }
John Lo37682e12016-11-30 12:51:39 -0500749
Eyal Bari0f4b1842018-04-12 12:39:51 +0300750 next1 = error1 ?
John Lo37682e12016-11-30 12:51:39 -0500751 IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN;
752 b1->error = error1 ? error_node->errors[error1] : 0;
Eyal Bari0f4b1842018-04-12 12:39:51 +0300753
John Lo2b81eb82017-01-30 13:12:10 -0500754 /* vxlan-input node expect current at VXLAN header */
755 if (is_ip4)
756 vlib_buffer_advance (b1, sizeof(ip4_header_t)+sizeof(udp_header_t));
757 else
758 vlib_buffer_advance (b1, sizeof(ip6_header_t)+sizeof(udp_header_t));
759
John Lo37682e12016-11-30 12:51:39 -0500760 exit1:
761 vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
762 to_next, n_left_to_next,
763 bi0, bi1, next0, next1);
764 }
765
766 while (n_left_from > 0 && n_left_to_next > 0)
767 {
768 vlib_buffer_t * b0;
John Lo2b81eb82017-01-30 13:12:10 -0500769 ip4_header_t * ip40;
770 ip6_header_t * ip60;
John Lo37682e12016-11-30 12:51:39 -0500771 udp_header_t * udp0;
772 u32 bi0, ip_len0, udp_len0, flags0, next0;
773 i32 len_diff0;
774 u8 error0, good_udp0, proto0;
775
776 bi0 = to_next[0] = from[0];
777 from += 1;
778 n_left_from -= 1;
779 to_next += 1;
780 n_left_to_next -= 1;
781
782 b0 = vlib_get_buffer (vm, bi0);
John Lo2b81eb82017-01-30 13:12:10 -0500783 if (is_ip4)
784 ip40 = vlib_buffer_get_current (b0);
785 else
786 ip60 = vlib_buffer_get_current (b0);
John Lo37682e12016-11-30 12:51:39 -0500787
788 /* Setup packet for next IP feature */
Damjan Marion7d98a122018-07-19 20:42:08 +0200789 vnet_feature_next(&next0, b0);
John Lo37682e12016-11-30 12:51:39 -0500790
John Lo2b81eb82017-01-30 13:12:10 -0500791 if (is_ip4)
792 /* Treat IP4 frag packets as "experimental" protocol for now
793 until support of IP frag reassembly is implemented */
794 proto0 = ip4_is_fragment(ip40) ? 0xfe : ip40->protocol;
795 else
796 proto0 = ip60->protocol;
John Lo37682e12016-11-30 12:51:39 -0500797
798 if (proto0 != IP_PROTOCOL_UDP)
799 goto exit; /* not UDP packet */
800
John Lo2b81eb82017-01-30 13:12:10 -0500801 if (is_ip4)
802 udp0 = ip4_next_header (ip40);
803 else
804 udp0 = ip6_next_header (ip60);
805
John Lo37682e12016-11-30 12:51:39 -0500806 if (udp0->dst_port != clib_host_to_net_u16 (UDP_DST_PORT_vxlan))
807 goto exit; /* not VXLAN packet */
808
John Lo2b81eb82017-01-30 13:12:10 -0500809 /* Validate DIP against VTEPs*/
Eyal Bari0f4b1842018-04-12 12:39:51 +0300810 if (is_ip4)
John Lo37682e12016-11-30 12:51:39 -0500811 {
John Lo2b81eb82017-01-30 13:12:10 -0500812 if (addr4.as_u32 != ip40->dst_address.as_u32)
John Lo37682e12016-11-30 12:51:39 -0500813 {
John Lo2b81eb82017-01-30 13:12:10 -0500814 if (!hash_get (vxm->vtep4, ip40->dst_address.as_u32))
815 goto exit; /* no local VTEP for VXLAN packet */
816 addr4 = ip40->dst_address;
817 }
John Lo37682e12016-11-30 12:51:39 -0500818 }
John Lo2b81eb82017-01-30 13:12:10 -0500819 else
820 {
821 if (!ip6_address_is_equal (&addr6, &ip60->dst_address))
822 {
823 if (!hash_get_mem (vxm->vtep6, &ip60->dst_address))
824 goto exit; /* no local VTEP for VXLAN packet */
825 addr6 = ip60->dst_address;
826 }
827 }
John Lo37682e12016-11-30 12:51:39 -0500828
829 flags0 = b0->flags;
Damjan Marion213b5aa2017-07-13 21:19:27 +0200830 good_udp0 = (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
John Lo37682e12016-11-30 12:51:39 -0500831
832 /* Don't verify UDP checksum for packets with explicit zero checksum. */
833 good_udp0 |= udp0->checksum == 0;
834
835 /* Verify UDP length */
John Lo2b81eb82017-01-30 13:12:10 -0500836 if (is_ip4)
837 ip_len0 = clib_net_to_host_u16 (ip40->length);
838 else
839 ip_len0 = clib_net_to_host_u16 (ip60->payload_length);
John Lo37682e12016-11-30 12:51:39 -0500840 udp_len0 = clib_net_to_host_u16 (udp0->length);
John Lo37682e12016-11-30 12:51:39 -0500841 len_diff0 = ip_len0 - udp_len0;
842
843 /* Verify UDP checksum */
844 if (PREDICT_FALSE (!good_udp0))
845 {
Damjan Marion213b5aa2017-07-13 21:19:27 +0200846 if ((flags0 & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0)
John Lo2b81eb82017-01-30 13:12:10 -0500847 {
848 if (is_ip4)
849 flags0 = ip4_tcp_udp_validate_checksum (vm, b0);
850 else
851 flags0 = ip6_tcp_udp_icmp_validate_checksum (vm, b0);
852 good_udp0 =
Damjan Marion213b5aa2017-07-13 21:19:27 +0200853 (flags0 & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
John Lo2b81eb82017-01-30 13:12:10 -0500854 }
John Lo37682e12016-11-30 12:51:39 -0500855 }
856
John Lo2b81eb82017-01-30 13:12:10 -0500857 if (is_ip4)
858 {
859 error0 = good_udp0 ? 0 : IP4_ERROR_UDP_CHECKSUM;
860 error0 = (len_diff0 >= 0) ? error0 : IP4_ERROR_UDP_LENGTH;
861 }
862 else
863 {
864 error0 = good_udp0 ? 0 : IP6_ERROR_UDP_CHECKSUM;
865 error0 = (len_diff0 >= 0) ? error0 : IP6_ERROR_UDP_LENGTH;
866 }
John Lo37682e12016-11-30 12:51:39 -0500867
Eyal Bari0f4b1842018-04-12 12:39:51 +0300868 next0 = error0 ?
John Lo37682e12016-11-30 12:51:39 -0500869 IP_VXLAN_BYPASS_NEXT_DROP : IP_VXLAN_BYPASS_NEXT_VXLAN;
870 b0->error = error0 ? error_node->errors[error0] : 0;
871
John Lo2b81eb82017-01-30 13:12:10 -0500872 /* vxlan-input node expect current at VXLAN header */
873 if (is_ip4)
874 vlib_buffer_advance (b0, sizeof(ip4_header_t)+sizeof(udp_header_t));
875 else
876 vlib_buffer_advance (b0, sizeof(ip6_header_t)+sizeof(udp_header_t));
877
John Lo37682e12016-11-30 12:51:39 -0500878 exit:
879 vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
880 to_next, n_left_to_next,
881 bi0, next0);
882 }
883
884 vlib_put_next_frame (vm, node, next_index, n_left_to_next);
885 }
886
887 return frame->n_vectors;
888}
889
890static uword
891ip4_vxlan_bypass (vlib_main_t * vm,
892 vlib_node_runtime_t * node,
893 vlib_frame_t * frame)
894{
895 return ip_vxlan_bypass_inline (vm, node, frame, /* is_ip4 */ 1);
896}
897
898VLIB_REGISTER_NODE (ip4_vxlan_bypass_node) = {
899 .function = ip4_vxlan_bypass,
900 .name = "ip4-vxlan-bypass",
901 .vector_size = sizeof (u32),
902
903 .n_next_nodes = IP_VXLAN_BYPASS_N_NEXT,
904 .next_nodes = {
905 [IP_VXLAN_BYPASS_NEXT_DROP] = "error-drop",
906 [IP_VXLAN_BYPASS_NEXT_VXLAN] = "vxlan4-input",
907 },
908
909 .format_buffer = format_ip4_header,
910 .format_trace = format_ip4_forward_next_trace,
911};
912
913VLIB_NODE_FUNCTION_MULTIARCH (ip4_vxlan_bypass_node,ip4_vxlan_bypass)
914
John Lo37682e12016-11-30 12:51:39 -0500915/* Dummy init function to get us linked in. */
916clib_error_t * ip4_vxlan_bypass_init (vlib_main_t * vm)
917{ return 0; }
918
919VLIB_INIT_FUNCTION (ip4_vxlan_bypass_init);
John Lo2b81eb82017-01-30 13:12:10 -0500920
921static uword
922ip6_vxlan_bypass (vlib_main_t * vm,
923 vlib_node_runtime_t * node,
924 vlib_frame_t * frame)
925{
926 return ip_vxlan_bypass_inline (vm, node, frame, /* is_ip4 */ 0);
927}
928
929VLIB_REGISTER_NODE (ip6_vxlan_bypass_node) = {
930 .function = ip6_vxlan_bypass,
931 .name = "ip6-vxlan-bypass",
932 .vector_size = sizeof (u32),
933
934 .n_next_nodes = IP_VXLAN_BYPASS_N_NEXT,
935 .next_nodes = {
936 [IP_VXLAN_BYPASS_NEXT_DROP] = "error-drop",
937 [IP_VXLAN_BYPASS_NEXT_VXLAN] = "vxlan6-input",
938 },
939
940 .format_buffer = format_ip6_header,
941 .format_trace = format_ip6_forward_next_trace,
942};
943
944VLIB_NODE_FUNCTION_MULTIARCH (ip6_vxlan_bypass_node,ip6_vxlan_bypass)
945
946/* Dummy init function to get us linked in. */
947clib_error_t * ip6_vxlan_bypass_init (vlib_main_t * vm)
948{ return 0; }
949
950VLIB_INIT_FUNCTION (ip6_vxlan_bypass_init);
eyal bariaf86a482018-04-17 11:20:27 +0300951
952#define foreach_vxlan_flow_input_next \
953_(DROP, "error-drop") \
954_(L2_INPUT, "l2-input")
955
956typedef enum
957{
958#define _(s,n) VXLAN_FLOW_NEXT_##s,
959 foreach_vxlan_flow_input_next
960#undef _
961 VXLAN_FLOW_N_NEXT,
962} vxlan_flow_input_next_t;
963
964#define foreach_vxlan_flow_error \
965 _(NONE, "no error") \
966 _(IP_CHECKSUM_ERROR, "Rx ip checksum errors") \
967 _(IP_HEADER_ERROR, "Rx ip header errors") \
968 _(UDP_CHECKSUM_ERROR, "Rx udp checksum errors") \
969 _(UDP_LENGTH_ERROR, "Rx udp length errors")
970
971typedef enum
972{
973#define _(f,s) VXLAN_FLOW_ERROR_##f,
974 foreach_vxlan_flow_error
975#undef _
976 VXLAN_FLOW_N_ERROR,
977} vxlan_flow_error_t;
978
979static char *vxlan_flow_error_strings[] = {
980#define _(n,s) s,
981 foreach_vxlan_flow_error
982#undef _
983};
984
985
986static_always_inline u8
987vxlan_validate_udp_csum (vlib_main_t * vm, vlib_buffer_t *b)
988{
989 u32 flags = b->flags;
990 enum { offset = sizeof(ip4_header_t) + sizeof(udp_header_t) + sizeof(vxlan_header_t), };
991
992 /* Verify UDP checksum */
993 if ((flags & VNET_BUFFER_F_L4_CHECKSUM_COMPUTED) == 0)
994 {
995 vlib_buffer_advance (b, -offset);
996 flags = ip4_tcp_udp_validate_checksum (vm, b);
997 vlib_buffer_advance (b, offset);
998 }
999
1000 return (flags & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0;
1001}
1002
1003static_always_inline u8
1004vxlan_check_udp_csum (vlib_main_t * vm, vlib_buffer_t *b)
1005{
1006 ip4_vxlan_header_t * hdr = vlib_buffer_get_current(b) - sizeof *hdr;
1007 udp_header_t * udp = &hdr->udp;
1008 /* Don't verify UDP checksum for packets with explicit zero checksum. */
1009 u8 good_csum = (b->flags & VNET_BUFFER_F_L4_CHECKSUM_CORRECT) != 0 ||
1010 udp->checksum == 0;
1011
1012 return !good_csum;
1013}
1014
1015static_always_inline u8
1016vxlan_check_ip (vlib_buffer_t *b, u16 payload_len)
1017{
1018 ip4_vxlan_header_t * hdr = vlib_buffer_get_current(b) - sizeof *hdr;
1019 u16 ip_len = clib_net_to_host_u16 (hdr->ip4.length);
1020 u16 expected = payload_len + sizeof *hdr;
1021 return ip_len > expected || hdr->ip4.ttl == 0 || hdr->ip4.ip_version_and_header_length != 0x45;
1022}
1023
1024static_always_inline u8
1025vxlan_check_ip_udp_len (vlib_buffer_t *b)
1026{
1027 ip4_vxlan_header_t * hdr = vlib_buffer_get_current(b) - sizeof *hdr;
1028 u16 ip_len = clib_net_to_host_u16 (hdr->ip4.length);
1029 u16 udp_len = clib_net_to_host_u16 (hdr->udp.length);
1030 return udp_len > ip_len;
1031}
1032
1033static_always_inline u8
1034vxlan_err_code (u8 ip_err0, u8 udp_err0, u8 csum_err0)
1035{
1036 u8 error0 = VXLAN_FLOW_ERROR_NONE;
1037 if (ip_err0)
1038 error0 = VXLAN_FLOW_ERROR_IP_HEADER_ERROR;
1039 if (udp_err0)
1040 error0 = VXLAN_FLOW_ERROR_UDP_LENGTH_ERROR;
1041 if (csum_err0)
1042 error0 = VXLAN_FLOW_ERROR_UDP_CHECKSUM_ERROR;
1043 return error0;
1044}
1045
Eyal Bari93a6f252018-06-14 08:57:39 +03001046VLIB_NODE_FN (vxlan4_flow_input_node) (vlib_main_t * vm,
eyal bariaf86a482018-04-17 11:20:27 +03001047 vlib_node_runtime_t * node,
1048 vlib_frame_t * f)
1049{
1050 enum { payload_offset = sizeof(ip4_vxlan_header_t) };
1051
1052 vxlan_main_t * vxm = &vxlan_main;
1053 vnet_interface_main_t * im = &vnet_main.interface_main;
1054 vlib_combined_counter_main_t * rx_counter[VXLAN_FLOW_N_NEXT] = {
1055 [VXLAN_FLOW_NEXT_DROP] = im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_DROP,
1056 [VXLAN_FLOW_NEXT_L2_INPUT] = im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
1057 };
1058 u32 thread_index = vlib_get_thread_index();
1059
1060 u32 * from = vlib_frame_vector_args (f);
1061 u32 n_left_from = f->n_vectors;
1062 u32 next_index = VXLAN_FLOW_NEXT_L2_INPUT;
1063
1064 while (n_left_from > 0)
1065 {
1066 u32 n_left_to_next, *to_next;
1067
1068 vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
1069
1070 while (n_left_from > 3 && n_left_to_next > 3)
1071 {
1072 u32 bi0 = to_next[0] = from[0];
1073 u32 bi1 = to_next[1] = from[1];
1074 u32 bi2 = to_next[2] = from[2];
1075 u32 bi3 = to_next[3] = from[3];
1076 from+=4;
1077 n_left_from-=4;
1078 to_next+=4;
1079 n_left_to_next-=4;
1080
1081 vlib_buffer_t * b0 = vlib_get_buffer (vm, bi0);
1082 vlib_buffer_t * b1 = vlib_get_buffer (vm, bi1);
1083 vlib_buffer_t * b2 = vlib_get_buffer (vm, bi2);
1084 vlib_buffer_t * b3 = vlib_get_buffer (vm, bi3);
1085
1086 vlib_buffer_advance (b0, payload_offset);
1087 vlib_buffer_advance (b1, payload_offset);
1088 vlib_buffer_advance (b2, payload_offset);
1089 vlib_buffer_advance (b3, payload_offset);
1090
1091 u16 len0 = vlib_buffer_length_in_chain (vm, b0);
1092 u16 len1 = vlib_buffer_length_in_chain (vm, b1);
1093 u16 len2 = vlib_buffer_length_in_chain (vm, b2);
1094 u16 len3 = vlib_buffer_length_in_chain (vm, b3);
1095
1096 u32 next0 = VXLAN_FLOW_NEXT_L2_INPUT, next1 = VXLAN_FLOW_NEXT_L2_INPUT,
1097 next2 = VXLAN_FLOW_NEXT_L2_INPUT, next3 = VXLAN_FLOW_NEXT_L2_INPUT;
1098
1099 u8 ip_err0 = vxlan_check_ip (b0, len0);
1100 u8 ip_err1 = vxlan_check_ip (b1, len1);
1101 u8 ip_err2 = vxlan_check_ip (b2, len2);
1102 u8 ip_err3 = vxlan_check_ip (b3, len3);
1103 u8 ip_err = ip_err0 | ip_err1 | ip_err2 | ip_err3;
1104
1105 u8 udp_err0 = vxlan_check_ip_udp_len (b0);
1106 u8 udp_err1 = vxlan_check_ip_udp_len (b1);
1107 u8 udp_err2 = vxlan_check_ip_udp_len (b2);
1108 u8 udp_err3 = vxlan_check_ip_udp_len (b3);
1109 u8 udp_err = udp_err0 | udp_err1 | udp_err2 | udp_err3;
1110
1111 u8 csum_err0 = vxlan_check_udp_csum (vm, b0);
1112 u8 csum_err1 = vxlan_check_udp_csum (vm, b1);
1113 u8 csum_err2 = vxlan_check_udp_csum (vm, b2);
1114 u8 csum_err3 = vxlan_check_udp_csum (vm, b3);
1115 u8 csum_err = csum_err0 | csum_err1 | csum_err2 | csum_err3;
1116
1117 if (PREDICT_FALSE(csum_err))
1118 {
1119 if (csum_err0)
1120 csum_err0 = !vxlan_validate_udp_csum (vm, b0);
1121 if (csum_err1)
1122 csum_err1 = !vxlan_validate_udp_csum (vm, b1);
1123 if (csum_err2)
1124 csum_err2 = !vxlan_validate_udp_csum (vm, b2);
1125 if (csum_err3)
1126 csum_err3 = !vxlan_validate_udp_csum (vm, b3);
1127 csum_err = csum_err0 | csum_err1 | csum_err2 | csum_err3;
1128 }
1129
1130 if (PREDICT_FALSE(ip_err || udp_err || csum_err))
1131 {
1132 if (ip_err0 || udp_err0 || csum_err0)
1133 {
1134 next0 = VXLAN_FLOW_NEXT_DROP;
1135 u8 error0 = vxlan_err_code (ip_err0, udp_err0, csum_err0);
1136 b0->error = node->errors[error0];
1137 }
1138 if (ip_err1 || udp_err1 || csum_err1)
1139 {
1140 next1 = VXLAN_FLOW_NEXT_DROP;
1141 u8 error1 = vxlan_err_code (ip_err1, udp_err1, csum_err1);
1142 b1->error = node->errors[error1];
1143 }
1144 if (ip_err2 || udp_err2 || csum_err2)
1145 {
1146 next2 = VXLAN_FLOW_NEXT_DROP;
1147 u8 error2 = vxlan_err_code (ip_err2, udp_err2, csum_err2);
1148 b2->error = node->errors[error2];
1149 }
1150 if (ip_err3 || udp_err3 || csum_err3)
1151 {
1152 next3 = VXLAN_FLOW_NEXT_DROP;
1153 u8 error3 = vxlan_err_code (ip_err3, udp_err3, csum_err3);
1154 b3->error = node->errors[error3];
1155 }
1156 }
1157
1158 vnet_update_l2_len (b0);
1159 vnet_update_l2_len (b1);
1160 vnet_update_l2_len (b2);
1161 vnet_update_l2_len (b3);
1162
1163 ASSERT (b0->flow_id != 0);
1164 ASSERT (b1->flow_id != 0);
1165 ASSERT (b2->flow_id != 0);
1166 ASSERT (b3->flow_id != 0);
1167
1168 u32 t_index0 = b0->flow_id - vxm->flow_id_start;
1169 u32 t_index1 = b1->flow_id - vxm->flow_id_start;
1170 u32 t_index2 = b2->flow_id - vxm->flow_id_start;
1171 u32 t_index3 = b3->flow_id - vxm->flow_id_start;
1172
1173 vxlan_tunnel_t * t0 = &vxm->tunnels[t_index0];
1174 vxlan_tunnel_t * t1 = &vxm->tunnels[t_index1];
1175 vxlan_tunnel_t * t2 = &vxm->tunnels[t_index2];
1176 vxlan_tunnel_t * t3 = &vxm->tunnels[t_index3];
1177
1178 /* flow id consumed */
1179 b0->flow_id = 0;
1180 b1->flow_id = 0;
1181 b2->flow_id = 0;
1182 b3->flow_id = 0;
1183
1184 u32 sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX] = t0->sw_if_index;
1185 u32 sw_if_index1 = vnet_buffer (b1)->sw_if_index[VLIB_RX] = t1->sw_if_index;
1186 u32 sw_if_index2 = vnet_buffer (b2)->sw_if_index[VLIB_RX] = t2->sw_if_index;
1187 u32 sw_if_index3 = vnet_buffer (b3)->sw_if_index[VLIB_RX] = t3->sw_if_index;
1188
1189 vlib_increment_combined_counter (rx_counter[next0], thread_index, sw_if_index0, 1, len0);
1190 vlib_increment_combined_counter (rx_counter[next1], thread_index, sw_if_index1, 1, len1);
1191 vlib_increment_combined_counter (rx_counter[next2], thread_index, sw_if_index2, 1, len2);
1192 vlib_increment_combined_counter (rx_counter[next3], thread_index, sw_if_index3, 1, len3);
1193
1194 u32 flags = b0->flags | b1->flags | b2->flags | b3->flags;
1195
1196 if (PREDICT_FALSE(flags & VLIB_BUFFER_IS_TRACED))
1197 {
1198 if (b0->flags & VLIB_BUFFER_IS_TRACED)
1199 {
1200 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof *tr);
1201 u8 error0 = vxlan_err_code (ip_err0, udp_err0, csum_err0);
1202 *tr = (vxlan_rx_trace_t) {
1203 .next_index = next0, .error = error0, .tunnel_index = t_index0, .vni = t0->vni };
1204 }
1205 if (b1->flags & VLIB_BUFFER_IS_TRACED)
1206 {
1207 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b1, sizeof *tr);
1208 u8 error1 = vxlan_err_code (ip_err1, udp_err1, csum_err1);
1209 *tr = (vxlan_rx_trace_t) {
1210 .next_index = next1, .error = error1, .tunnel_index = t_index1, .vni = t1->vni };
1211 }
1212 if (b2->flags & VLIB_BUFFER_IS_TRACED)
1213 {
1214 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b2, sizeof *tr);
1215 u8 error2 = vxlan_err_code (ip_err2, udp_err2, csum_err2);
1216 *tr = (vxlan_rx_trace_t) {
1217 .next_index = next2, .error = error2, .tunnel_index = t_index2, .vni = t2->vni };
1218 }
1219 if (b3->flags & VLIB_BUFFER_IS_TRACED)
1220 {
1221 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b3, sizeof *tr);
1222 u8 error3 = vxlan_err_code (ip_err3, udp_err3, csum_err3);
1223 *tr = (vxlan_rx_trace_t) {
1224 .next_index = next3, .error = error3, .tunnel_index = t_index3, .vni = t3->vni };
1225 }
1226 }
1227 vlib_validate_buffer_enqueue_x4
1228 (vm, node, next_index, to_next, n_left_to_next,
1229 bi0, bi1, bi2, bi3, next0, next1, next2, next3);
1230 }
1231 while (n_left_from > 0 && n_left_to_next > 0)
1232 {
1233 u32 bi0 = to_next[0] = from[0];
1234 from++;
1235 n_left_from--;
1236 to_next++;
1237 n_left_to_next--;
1238
1239 vlib_buffer_t * b0 = vlib_get_buffer (vm, bi0);
1240 vlib_buffer_advance (b0, payload_offset);
1241
1242 u16 len0 = vlib_buffer_length_in_chain (vm, b0);
1243 u32 next0 = VXLAN_FLOW_NEXT_L2_INPUT;
1244
1245 u8 ip_err0 = vxlan_check_ip (b0, len0);
1246 u8 udp_err0 = vxlan_check_ip_udp_len (b0);
1247 u8 csum_err0 = vxlan_check_udp_csum (vm, b0);
1248
1249 if (csum_err0)
1250 csum_err0 = !vxlan_validate_udp_csum (vm, b0);
1251 if (ip_err0 || udp_err0 || csum_err0)
1252 {
1253 next0 = VXLAN_FLOW_NEXT_DROP;
1254 u8 error0 = vxlan_err_code (ip_err0, udp_err0, csum_err0);
1255 b0->error = node->errors[error0];
1256 }
1257
1258 vnet_update_l2_len (b0);
1259
1260 ASSERT (b0->flow_id != 0);
1261 u32 t_index0 = b0->flow_id - vxm->flow_id_start;
1262 vxlan_tunnel_t * t0 = &vxm->tunnels[t_index0];
1263 b0->flow_id = 0;
1264
1265 u32 sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX] = t0->sw_if_index;
1266 vlib_increment_combined_counter (rx_counter[next0], thread_index, sw_if_index0, 1, len0);
1267
1268 if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
1269 {
1270 vxlan_rx_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof *tr);
1271 u8 error0 = vxlan_err_code (ip_err0, udp_err0, csum_err0);
1272 *tr = (vxlan_rx_trace_t) {
1273 .next_index = next0, .error = error0, .tunnel_index = t_index0, .vni = t0->vni };
1274 }
1275 vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
1276 to_next, n_left_to_next,
1277 bi0, next0);
1278 }
1279
1280 vlib_put_next_frame (vm, node, next_index, n_left_to_next);
1281 }
1282
1283 return f->n_vectors;
1284}
1285
1286/* *INDENT-OFF* */
1287#ifndef CLIB_MULTIARCH_VARIANT
1288VLIB_REGISTER_NODE (vxlan4_flow_input_node) = {
1289 .name = "vxlan-flow-input",
eyal bariaf86a482018-04-17 11:20:27 +03001290 .type = VLIB_NODE_TYPE_INTERNAL,
1291 .vector_size = sizeof (u32),
1292
1293 .format_trace = format_vxlan_rx_trace,
1294
1295 .n_errors = VXLAN_FLOW_N_ERROR,
1296 .error_strings = vxlan_flow_error_strings,
1297
1298 .n_next_nodes = VXLAN_FLOW_N_NEXT,
1299 .next_nodes = {
1300#define _(s,n) [VXLAN_FLOW_NEXT_##s] = n,
1301 foreach_vxlan_flow_input_next
1302#undef _
1303 },
1304};
1305#endif
1306/* *INDENT-ON* */