]> git.proxmox.com Git - ovs.git/blob - ovn/controller/physical.c
ovn-controller: Update check for parent port.
[ovs.git] / ovn / controller / physical.c
1 /* Copyright (c) 2015 Nicira, Inc.
2 *
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 *
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
14 */
15
16 #include <config.h>
17 #include "physical.h"
18 #include "lflow.h"
19 #include "match.h"
20 #include "ofctrl.h"
21 #include "ofp-actions.h"
22 #include "ofpbuf.h"
23 #include "ovn-controller.h"
24 #include "ovn/lib/ovn-sb-idl.h"
25 #include "openvswitch/vlog.h"
26 #include "shash.h"
27 #include "simap.h"
28 #include "smap.h"
29 #include "sset.h"
30 #include "vswitch-idl.h"
31
32 VLOG_DEFINE_THIS_MODULE(physical);
33
34 void
35 physical_register_ovs_idl(struct ovsdb_idl *ovs_idl)
36 {
37 ovsdb_idl_add_table(ovs_idl, &ovsrec_table_bridge);
38 ovsdb_idl_add_column(ovs_idl, &ovsrec_bridge_col_ports);
39
40 ovsdb_idl_add_table(ovs_idl, &ovsrec_table_port);
41 ovsdb_idl_add_column(ovs_idl, &ovsrec_port_col_name);
42 ovsdb_idl_add_column(ovs_idl, &ovsrec_port_col_interfaces);
43 ovsdb_idl_add_column(ovs_idl, &ovsrec_port_col_external_ids);
44
45 ovsdb_idl_add_table(ovs_idl, &ovsrec_table_interface);
46 ovsdb_idl_add_column(ovs_idl, &ovsrec_interface_col_name);
47 ovsdb_idl_add_column(ovs_idl, &ovsrec_interface_col_ofport);
48 ovsdb_idl_add_column(ovs_idl, &ovsrec_interface_col_external_ids);
49 }
50
51 /* Maps from a chassis to the OpenFlow port number of the tunnel that can be
52 * used to reach that chassis. */
53 struct chassis_tunnel {
54 struct hmap_node hmap_node;
55 const char *chassis_id;
56 ofp_port_t ofport;
57 enum chassis_tunnel_type type;
58 };
59
60 static struct chassis_tunnel *
61 chassis_tunnel_find(struct hmap *tunnels, const char *chassis_id)
62 {
63 struct chassis_tunnel *tun;
64 HMAP_FOR_EACH_WITH_HASH (tun, hmap_node, hash_string(chassis_id, 0),
65 tunnels) {
66 if (!strcmp(tun->chassis_id, chassis_id)) {
67 return tun;
68 }
69 }
70 return NULL;
71 }
72
73 static void
74 put_load(uint64_t value, enum mf_field_id dst, int ofs, int n_bits,
75 struct ofpbuf *ofpacts)
76 {
77 struct ofpact_set_field *sf = ofpact_put_SET_FIELD(ofpacts);
78 sf->field = mf_from_id(dst);
79 sf->flow_has_vlan = false;
80
81 ovs_be64 n_value = htonll(value);
82 bitwise_copy(&n_value, 8, 0, &sf->value, sf->field->n_bytes, ofs, n_bits);
83 bitwise_one(&sf->mask, sf->field->n_bytes, ofs, n_bits);
84 }
85
86 static void
87 put_move(enum mf_field_id src, int src_ofs,
88 enum mf_field_id dst, int dst_ofs,
89 int n_bits,
90 struct ofpbuf *ofpacts)
91 {
92 struct ofpact_reg_move *move = ofpact_put_REG_MOVE(ofpacts);
93 move->src.field = mf_from_id(src);
94 move->src.ofs = src_ofs;
95 move->src.n_bits = n_bits;
96 move->dst.field = mf_from_id(dst);
97 move->dst.ofs = dst_ofs;
98 move->dst.n_bits = n_bits;
99 }
100
101 static void
102 put_resubmit(uint8_t table_id, struct ofpbuf *ofpacts)
103 {
104 struct ofpact_resubmit *resubmit = ofpact_put_RESUBMIT(ofpacts);
105 resubmit->in_port = OFPP_IN_PORT;
106 resubmit->table_id = table_id;
107 }
108
109 static void
110 put_encapsulation(enum mf_field_id mff_ovn_geneve,
111 const struct chassis_tunnel *tun,
112 const struct sbrec_datapath_binding *datapath,
113 uint16_t outport, struct ofpbuf *ofpacts)
114 {
115 if (tun->type == GENEVE) {
116 put_load(datapath->tunnel_key, MFF_TUN_ID, 0, 24, ofpacts);
117 put_load(outport, mff_ovn_geneve, 0, 32, ofpacts);
118 put_move(MFF_LOG_INPORT, 0, mff_ovn_geneve, 16, 15, ofpacts);
119 } else if (tun->type == STT) {
120 put_load(datapath->tunnel_key | (outport << 24), MFF_TUN_ID, 0, 64,
121 ofpacts);
122 put_move(MFF_LOG_INPORT, 0, MFF_TUN_ID, 40, 15, ofpacts);
123 } else if (tun->type == VXLAN) {
124 put_load(datapath->tunnel_key, MFF_TUN_ID, 0, 24, ofpacts);
125 } else {
126 OVS_NOT_REACHED();
127 }
128 }
129
130 static void
131 put_stack(enum mf_field_id field, struct ofpact_stack *stack)
132 {
133 stack->subfield.field = mf_from_id(field);
134 stack->subfield.ofs = 0;
135 stack->subfield.n_bits = stack->subfield.field->n_bits;
136 }
137
138 void
139 physical_run(struct controller_ctx *ctx, enum mf_field_id mff_ovn_geneve,
140 const struct ovsrec_bridge *br_int, const char *this_chassis_id,
141 const struct simap *ct_zones, struct hmap *flow_table)
142 {
143 struct simap localvif_to_ofport = SIMAP_INITIALIZER(&localvif_to_ofport);
144 struct hmap tunnels = HMAP_INITIALIZER(&tunnels);
145 struct simap localnet_to_ofport = SIMAP_INITIALIZER(&localnet_to_ofport);
146
147 for (int i = 0; i < br_int->n_ports; i++) {
148 const struct ovsrec_port *port_rec = br_int->ports[i];
149 if (!strcmp(port_rec->name, br_int->name)) {
150 continue;
151 }
152
153 const char *chassis_id = smap_get(&port_rec->external_ids,
154 "ovn-chassis-id");
155 if (chassis_id && !strcmp(chassis_id, this_chassis_id)) {
156 continue;
157 }
158
159 const char *localnet = smap_get(&port_rec->external_ids,
160 "ovn-localnet-port");
161 const char *logpatch = smap_get(&port_rec->external_ids,
162 "ovn-logical-patch-port");
163
164 for (int j = 0; j < port_rec->n_interfaces; j++) {
165 const struct ovsrec_interface *iface_rec = port_rec->interfaces[j];
166
167 /* Get OpenFlow port number. */
168 if (!iface_rec->n_ofport) {
169 continue;
170 }
171 int64_t ofport = iface_rec->ofport[0];
172 if (ofport < 1 || ofport > ofp_to_u16(OFPP_MAX)) {
173 continue;
174 }
175
176 /* Record as patch to local net, logical patch port, chassis, or
177 * local logical port. */
178 bool is_patch = !strcmp(iface_rec->type, "patch");
179 if (is_patch && localnet) {
180 simap_put(&localnet_to_ofport, localnet, ofport);
181 break;
182 } else if (is_patch && logpatch) {
183 /* Logical patch ports can be handled just like VIFs. */
184 simap_put(&localvif_to_ofport, logpatch, ofport);
185 break;
186 } else if (chassis_id) {
187 enum chassis_tunnel_type tunnel_type;
188 if (!strcmp(iface_rec->type, "geneve")) {
189 tunnel_type = GENEVE;
190 if (!mff_ovn_geneve) {
191 continue;
192 }
193 } else if (!strcmp(iface_rec->type, "stt")) {
194 tunnel_type = STT;
195 } else if (!strcmp(iface_rec->type, "vxlan")) {
196 tunnel_type = VXLAN;
197 } else {
198 continue;
199 }
200
201 struct chassis_tunnel *tun = xmalloc(sizeof *tun);
202 hmap_insert(&tunnels, &tun->hmap_node,
203 hash_string(chassis_id, 0));
204 tun->chassis_id = chassis_id;
205 tun->ofport = u16_to_ofp(ofport);
206 tun->type = tunnel_type;
207 break;
208 } else {
209 const char *iface_id = smap_get(&iface_rec->external_ids,
210 "iface-id");
211 if (iface_id) {
212 simap_put(&localvif_to_ofport, iface_id, ofport);
213 }
214 }
215 }
216 }
217
218 struct ofpbuf ofpacts;
219 ofpbuf_init(&ofpacts, 0);
220
221 struct binding_elem {
222 struct ovs_list list_elem;
223 const struct sbrec_port_binding *binding;
224 };
225 /* The bindings for a given VLAN on a localnet port. */
226 struct localnet_vlan {
227 struct hmap_node node;
228 int tag;
229 struct ovs_list bindings;
230 };
231 /* A hash of localnet_vlans, hashed on VLAN ID, for a localnet port */
232 struct localnet_bindings {
233 ofp_port_t ofport;
234 struct hmap vlans;
235 };
236 /* Maps from network name to "struct localnet_bindings". */
237 struct shash localnet_inputs = SHASH_INITIALIZER(&localnet_inputs);
238
239 /* Contains bare "struct hmap_node"s whose hash values are the tunnel_key
240 * of datapaths with at least one local port binding. */
241 struct hmap local_datapaths = HMAP_INITIALIZER(&local_datapaths);
242
243 /* Set up flows in table 0 for physical-to-logical translation and in table
244 * 64 for logical-to-physical translation. */
245 const struct sbrec_port_binding *binding;
246 SBREC_PORT_BINDING_FOR_EACH (binding, ctx->ovnsb_idl) {
247 /* Find the OpenFlow port for the logical port, as 'ofport'. This is
248 * one of:
249 *
250 * - If the port is a VIF on the chassis we're managing, the
251 * OpenFlow port for the VIF. 'tun' will be NULL.
252 *
253 * In this or the next case, for a container nested inside a VM
254 * and accessible via a VLAN, 'tag' is the VLAN ID; otherwise
255 * 'tag' is 0.
256 *
257 * The same logic handles logical patch ports.
258 *
259 * - If the port is on a remote chassis, the OpenFlow port for a
260 * tunnel to the VIF's remote chassis. 'tun' identifies that
261 * tunnel.
262 *
263 * - If the port is a "localnet" port for a network that is
264 * attached to the chassis we're managing, the OpenFlow port for
265 * the localnet port (a patch port).
266 *
267 * The "localnet" port may be configured with a VLAN ID. If so,
268 * 'tag' will be set to that VLAN ID; otherwise 'tag' is 0.
269 */
270
271 int tag = 0;
272 ofp_port_t ofport;
273 if (!strcmp(binding->type, "localnet")) {
274 const char *network = smap_get(&binding->options, "network_name");
275 if (!network) {
276 continue;
277 }
278 ofport = u16_to_ofp(simap_get(&localnet_to_ofport, network));
279 if (ofport && binding->tag) {
280 tag = *binding->tag;
281 }
282 } else if (binding->parent_port && *binding->parent_port) {
283 if (!binding->tag) {
284 continue;
285 }
286 ofport = u16_to_ofp(simap_get(&localvif_to_ofport,
287 binding->parent_port));
288 if (ofport) {
289 tag = *binding->tag;
290 }
291 } else {
292 ofport = u16_to_ofp(simap_get(&localvif_to_ofport,
293 binding->logical_port));
294 }
295
296 const struct chassis_tunnel *tun = NULL;
297 if (!ofport) {
298 if (!binding->chassis) {
299 continue;
300 }
301 tun = chassis_tunnel_find(&tunnels, binding->chassis->name);
302 if (!tun) {
303 continue;
304 }
305 ofport = tun->ofport;
306 }
307
308 struct match match;
309 if (!tun) {
310 int zone_id = simap_get(ct_zones, binding->logical_port);
311 /* Packets that arrive from a vif can belong to a VM or
312 * to a container located inside that VM. Packets that
313 * arrive from containers have a tag (vlan) associated with them.
314 */
315
316 /* Table 0, Priority 150 and 100.
317 * ==============================
318 *
319 * Priority 150 is for tagged traffic. This may be containers in a
320 * VM or a VLAN on a local network. For such traffic, match on the
321 * tags and then strip the tag.
322 *
323 * Priority 100 is for traffic belonging to VMs or untagged locally
324 * connected networks.
325 *
326 * For both types of traffic: set MFF_LOG_INPORT to the logical
327 * input port, MFF_LOG_DATAPATH to the logical datapath, and
328 * resubmit into the logical ingress pipeline starting at table
329 * 16. */
330 if (!strcmp(binding->type, "localnet")) {
331 /* The same OpenFlow port may correspond to localnet ports
332 * attached to more than one logical datapath, so keep track of
333 * all associated bindings and add a flow at the end. */
334
335 const char *network
336 = smap_get(&binding->options, "network_name");
337 struct localnet_bindings *ln_bindings;
338 struct hmap_node *node;
339 struct localnet_vlan *ln_vlan;
340
341 ln_bindings = shash_find_data(&localnet_inputs, network);
342 if (!ln_bindings) {
343 ln_bindings = xmalloc(sizeof *ln_bindings);
344 ln_bindings->ofport = ofport;
345 hmap_init(&ln_bindings->vlans);
346 shash_add(&localnet_inputs, network, ln_bindings);
347 }
348 node = hmap_first_with_hash(&ln_bindings->vlans, tag);
349 if (node) {
350 ASSIGN_CONTAINER(ln_vlan, node, node);
351 } else {
352 ln_vlan = xmalloc(sizeof *ln_vlan);
353 ln_vlan->tag = tag;
354 list_init(&ln_vlan->bindings);
355 hmap_insert(&ln_bindings->vlans, &ln_vlan->node, tag);
356 }
357
358 struct binding_elem *b = xmalloc(sizeof *b);
359 b->binding = binding;
360 list_insert(&ln_vlan->bindings, &b->list_elem);
361 } else {
362 struct hmap_node *ld;
363 ld = hmap_first_with_hash(&local_datapaths,
364 binding->datapath->tunnel_key);
365 if (!ld) {
366 ld = xmalloc(sizeof *ld);
367 hmap_insert(&local_datapaths, ld,
368 binding->datapath->tunnel_key);
369 }
370
371 ofpbuf_clear(&ofpacts);
372 match_init_catchall(&match);
373 match_set_in_port(&match, ofport);
374 if (tag) {
375 match_set_dl_vlan(&match, htons(tag));
376 }
377
378 if (zone_id) {
379 put_load(zone_id, MFF_LOG_CT_ZONE, 0, 32, &ofpacts);
380 }
381
382 /* Set MFF_LOG_DATAPATH and MFF_LOG_INPORT. */
383 put_load(binding->datapath->tunnel_key, MFF_LOG_DATAPATH, 0, 64,
384 &ofpacts);
385 put_load(binding->tunnel_key, MFF_LOG_INPORT, 0, 32,
386 &ofpacts);
387
388 /* Strip vlans. */
389 if (tag) {
390 ofpact_put_STRIP_VLAN(&ofpacts);
391 }
392
393 /* Resubmit to first logical ingress pipeline table. */
394 put_resubmit(OFTABLE_LOG_INGRESS_PIPELINE, &ofpacts);
395 ofctrl_add_flow(flow_table, OFTABLE_PHY_TO_LOG,
396 tag ? 150 : 100, &match, &ofpacts);
397 }
398
399 /* Table 33, priority 100.
400 * =======================
401 *
402 * Implements output to local hypervisor. Each flow matches a
403 * logical output port on the local hypervisor, and resubmits to
404 * table 34.
405 */
406
407 match_init_catchall(&match);
408 ofpbuf_clear(&ofpacts);
409
410 /* Match MFF_LOG_DATAPATH, MFF_LOG_OUTPORT. */
411 match_set_metadata(&match, htonll(binding->datapath->tunnel_key));
412 match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0,
413 binding->tunnel_key);
414
415 if (zone_id) {
416 put_load(zone_id, MFF_LOG_CT_ZONE, 0, 32, &ofpacts);
417 }
418
419 /* Resubmit to table 34. */
420 put_resubmit(OFTABLE_DROP_LOOPBACK, &ofpacts);
421 ofctrl_add_flow(flow_table, OFTABLE_LOCAL_OUTPUT, 100, &match,
422 &ofpacts);
423
424 /* Table 64, Priority 100.
425 * =======================
426 *
427 * Deliver the packet to the local vif. */
428 match_init_catchall(&match);
429 ofpbuf_clear(&ofpacts);
430 match_set_metadata(&match, htonll(binding->datapath->tunnel_key));
431 match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0,
432 binding->tunnel_key);
433 if (tag) {
434 /* For containers sitting behind a local vif, tag the packets
435 * before delivering them. */
436 struct ofpact_vlan_vid *vlan_vid;
437 vlan_vid = ofpact_put_SET_VLAN_VID(&ofpacts);
438 vlan_vid->vlan_vid = tag;
439 vlan_vid->push_vlan_if_needed = true;
440
441 /* A packet might need to hair-pin back into its ingress
442 * OpenFlow port (to a different logical port, which we already
443 * checked back in table 34), so set the in_port to zero. */
444 put_stack(MFF_IN_PORT, ofpact_put_STACK_PUSH(&ofpacts));
445 put_load(0, MFF_IN_PORT, 0, 16, &ofpacts);
446 }
447 ofpact_put_OUTPUT(&ofpacts)->port = ofport;
448 if (tag) {
449 /* Revert the tag added to the packets headed to containers
450 * in the previous step. If we don't do this, the packets
451 * that are to be broadcasted to a VM in the same logical
452 * switch will also contain the tag. Also revert the zero'd
453 * in_port. */
454 ofpact_put_STRIP_VLAN(&ofpacts);
455 put_stack(MFF_IN_PORT, ofpact_put_STACK_POP(&ofpacts));
456 }
457 ofctrl_add_flow(flow_table, OFTABLE_LOG_TO_PHY, 100,
458 &match, &ofpacts);
459 } else {
460 /* Table 32, priority 100.
461 * =======================
462 *
463 * Implements output to remote hypervisors. Each flow matches an
464 * output port that includes a logical port on a remote hypervisor,
465 * and tunnels the packet to that hypervisor.
466 */
467
468 match_init_catchall(&match);
469 ofpbuf_clear(&ofpacts);
470
471 /* Match MFF_LOG_DATAPATH, MFF_LOG_OUTPORT. */
472 match_set_metadata(&match, htonll(binding->datapath->tunnel_key));
473 match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0,
474 binding->tunnel_key);
475
476 put_encapsulation(mff_ovn_geneve, tun, binding->datapath,
477 binding->tunnel_key, &ofpacts);
478
479 /* Output to tunnel. */
480 ofpact_put_OUTPUT(&ofpacts)->port = ofport;
481 ofctrl_add_flow(flow_table, OFTABLE_REMOTE_OUTPUT, 100,
482 &match, &ofpacts);
483 }
484
485 /* Table 34, Priority 100.
486 * =======================
487 *
488 * Drop packets whose logical inport and outport are the same. */
489 match_init_catchall(&match);
490 ofpbuf_clear(&ofpacts);
491 match_set_metadata(&match, htonll(binding->datapath->tunnel_key));
492 match_set_reg(&match, MFF_LOG_INPORT - MFF_REG0, binding->tunnel_key);
493 match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0, binding->tunnel_key);
494 ofctrl_add_flow(flow_table, OFTABLE_DROP_LOOPBACK, 100,
495 &match, &ofpacts);
496 }
497
498 /* Handle output to multicast groups, in tables 32 and 33. */
499 const struct sbrec_multicast_group *mc;
500 struct ofpbuf remote_ofpacts;
501 ofpbuf_init(&remote_ofpacts, 0);
502 SBREC_MULTICAST_GROUP_FOR_EACH (mc, ctx->ovnsb_idl) {
503 struct sset remote_chassis = SSET_INITIALIZER(&remote_chassis);
504 struct match match;
505
506 match_init_catchall(&match);
507 match_set_metadata(&match, htonll(mc->datapath->tunnel_key));
508 match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0, mc->tunnel_key);
509
510 /* Go through all of the ports in the multicast group:
511 *
512 * - For remote ports, add the chassis to 'remote_chassis'.
513 *
514 * - For local ports (other than logical patch ports), add actions
515 * to 'ofpacts' to set the output port and resubmit.
516 *
517 * - For logical patch ports, add actions to 'remote_ofpacts'
518 * instead. (If we put them in 'ofpacts', then the output
519 * would happen on every hypervisor in the multicast group,
520 * effectively duplicating the packet.)
521 */
522 ofpbuf_clear(&ofpacts);
523 ofpbuf_clear(&remote_ofpacts);
524 for (size_t i = 0; i < mc->n_ports; i++) {
525 struct sbrec_port_binding *port = mc->ports[i];
526
527 if (port->datapath != mc->datapath) {
528 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 1);
529 VLOG_WARN_RL(&rl, UUID_FMT": multicast group contains ports "
530 "in wrong datapath",
531 UUID_ARGS(&mc->header_.uuid));
532 continue;
533 }
534
535 int zone_id = simap_get(ct_zones, port->logical_port);
536 if (zone_id) {
537 put_load(zone_id, MFF_LOG_CT_ZONE, 0, 32, &ofpacts);
538 }
539
540 if (!strcmp(port->type, "patch")) {
541 put_load(port->tunnel_key, MFF_LOG_OUTPORT, 0, 32,
542 &remote_ofpacts);
543 put_resubmit(OFTABLE_DROP_LOOPBACK, &remote_ofpacts);
544 } else if (simap_contains(&localvif_to_ofport,
545 (port->parent_port && *port->parent_port)
546 ? port->parent_port : port->logical_port)) {
547 put_load(port->tunnel_key, MFF_LOG_OUTPORT, 0, 32, &ofpacts);
548 put_resubmit(OFTABLE_DROP_LOOPBACK, &ofpacts);
549 } else if (port->chassis) {
550 sset_add(&remote_chassis, port->chassis->name);
551 } else if (!strcmp(port->type, "localnet")) {
552 const char *network = smap_get(&port->options, "network_name");
553 if (!network) {
554 continue;
555 }
556 if (!simap_contains(&localnet_to_ofport, network)) {
557 continue;
558 }
559 put_load(port->tunnel_key, MFF_LOG_OUTPORT, 0, 32, &ofpacts);
560 put_resubmit(OFTABLE_DROP_LOOPBACK, &ofpacts);
561 }
562 }
563
564 /* Table 33, priority 100.
565 * =======================
566 *
567 * Handle output to the local logical ports in the multicast group, if
568 * any. */
569 bool local_ports = ofpacts.size > 0;
570 if (local_ports) {
571 /* Following delivery to local logical ports, restore the multicast
572 * group as the logical output port. */
573 put_load(mc->tunnel_key, MFF_LOG_OUTPORT, 0, 32, &ofpacts);
574
575 ofctrl_add_flow(flow_table, OFTABLE_LOCAL_OUTPUT, 100,
576 &match, &ofpacts);
577 }
578
579 /* Table 32, priority 100.
580 * =======================
581 *
582 * Handle output to the remote chassis in the multicast group, if
583 * any. */
584 if (!sset_is_empty(&remote_chassis) || remote_ofpacts.size > 0) {
585 if (remote_ofpacts.size > 0) {
586 /* Following delivery to logical patch ports, restore the
587 * multicast group as the logical output port. */
588 put_load(mc->tunnel_key, MFF_LOG_OUTPORT, 0, 32,
589 &remote_ofpacts);
590 }
591
592 const char *chassis;
593 const struct chassis_tunnel *prev = NULL;
594 SSET_FOR_EACH (chassis, &remote_chassis) {
595 const struct chassis_tunnel *tun
596 = chassis_tunnel_find(&tunnels, chassis);
597 if (!tun) {
598 continue;
599 }
600
601 if (!prev || tun->type != prev->type) {
602 put_encapsulation(mff_ovn_geneve, tun, mc->datapath,
603 mc->tunnel_key, &remote_ofpacts);
604 prev = tun;
605 }
606 ofpact_put_OUTPUT(&remote_ofpacts)->port = tun->ofport;
607 }
608
609 if (remote_ofpacts.size) {
610 if (local_ports) {
611 put_resubmit(OFTABLE_LOCAL_OUTPUT, &remote_ofpacts);
612 }
613 ofctrl_add_flow(flow_table, OFTABLE_REMOTE_OUTPUT, 100,
614 &match, &remote_ofpacts);
615 }
616 }
617 sset_destroy(&remote_chassis);
618 }
619 ofpbuf_uninit(&remote_ofpacts);
620
621 /* Table 0, priority 100.
622 * ======================
623 *
624 * Process packets that arrive from a remote hypervisor (by matching
625 * on tunnel in_port). */
626
627 /* Add flows for Geneve and STT encapsulations. These
628 * encapsulations have metadata about the ingress and egress logical
629 * ports. We set MFF_LOG_DATAPATH, MFF_LOG_INPORT, and
630 * MFF_LOG_OUTPORT from the tunnel key data, then resubmit to table
631 * 33 to handle packets to the local hypervisor. */
632 struct chassis_tunnel *tun;
633 HMAP_FOR_EACH (tun, hmap_node, &tunnels) {
634 struct match match = MATCH_CATCHALL_INITIALIZER;
635 match_set_in_port(&match, tun->ofport);
636
637 ofpbuf_clear(&ofpacts);
638 if (tun->type == GENEVE) {
639 put_move(MFF_TUN_ID, 0, MFF_LOG_DATAPATH, 0, 24, &ofpacts);
640 put_move(mff_ovn_geneve, 16, MFF_LOG_INPORT, 0, 15,
641 &ofpacts);
642 put_move(mff_ovn_geneve, 0, MFF_LOG_OUTPORT, 0, 16,
643 &ofpacts);
644 } else if (tun->type == STT) {
645 put_move(MFF_TUN_ID, 40, MFF_LOG_INPORT, 0, 15, &ofpacts);
646 put_move(MFF_TUN_ID, 24, MFF_LOG_OUTPORT, 0, 16, &ofpacts);
647 put_move(MFF_TUN_ID, 0, MFF_LOG_DATAPATH, 0, 24, &ofpacts);
648 } else if (tun->type == VXLAN) {
649 /* We'll handle VXLAN later. */
650 continue;
651 } else {
652 OVS_NOT_REACHED();
653 }
654
655 put_resubmit(OFTABLE_LOCAL_OUTPUT, &ofpacts);
656
657 ofctrl_add_flow(flow_table, OFTABLE_PHY_TO_LOG, 100, &match, &ofpacts);
658 }
659
660 /* Add flows for VXLAN encapsulations. Due to the limited amount of
661 * metadata, we only support VXLAN for connections to gateways. The
662 * VNI is used to populate MFF_LOG_DATAPATH. The gateway's logical
663 * port is set to MFF_LOG_INPORT. Then the packet is resubmitted to
664 * table 16 to determine the logical egress port.
665 *
666 * xxx Due to resubmitting to table 16, broadcasts will be re-sent to
667 * xxx all logical ports, including non-local ones which could cause
668 * xxx duplicate packets to be received by multiply-connected gateways. */
669 HMAP_FOR_EACH (tun, hmap_node, &tunnels) {
670 if (tun->type != VXLAN) {
671 continue;
672 }
673
674 SBREC_PORT_BINDING_FOR_EACH (binding, ctx->ovnsb_idl) {
675 struct match match = MATCH_CATCHALL_INITIALIZER;
676
677 if (!binding->chassis ||
678 strcmp(tun->chassis_id, binding->chassis->name)) {
679 continue;
680 }
681
682 match_set_in_port(&match, tun->ofport);
683 match_set_tun_id(&match, htonll(binding->datapath->tunnel_key));
684
685 ofpbuf_clear(&ofpacts);
686 put_move(MFF_TUN_ID, 0, MFF_LOG_DATAPATH, 0, 24, &ofpacts);
687 put_load(binding->tunnel_key, MFF_LOG_INPORT, 0, 15, &ofpacts);
688 put_resubmit(OFTABLE_LOG_INGRESS_PIPELINE, &ofpacts);
689
690 ofctrl_add_flow(flow_table, OFTABLE_PHY_TO_LOG, 100, &match,
691 &ofpacts);
692 }
693 }
694
695 /* Table 32, Priority 0.
696 * =======================
697 *
698 * Resubmit packets that are not directed at tunnels or part of a
699 * multicast group to the local output table. */
700 struct match match;
701 match_init_catchall(&match);
702 ofpbuf_clear(&ofpacts);
703 put_resubmit(OFTABLE_LOCAL_OUTPUT, &ofpacts);
704 ofctrl_add_flow(flow_table, OFTABLE_REMOTE_OUTPUT, 0, &match, &ofpacts);
705
706 /* Table 34, Priority 0.
707 * =======================
708 *
709 * Resubmit packets that don't output to the ingress port (already checked
710 * in table 33) to the logical egress pipeline, clearing the logical
711 * registers (for consistent behavior with packets that get tunneled). */
712 match_init_catchall(&match);
713 ofpbuf_clear(&ofpacts);
714 #define MFF_LOG_REG(ID) put_load(0, ID, 0, 32, &ofpacts);
715 MFF_LOG_REGS;
716 #undef MFF_LOG_REGS
717 put_resubmit(OFTABLE_LOG_EGRESS_PIPELINE, &ofpacts);
718 ofctrl_add_flow(flow_table, OFTABLE_DROP_LOOPBACK, 0, &match, &ofpacts);
719
720 ofpbuf_uninit(&ofpacts);
721 simap_destroy(&localvif_to_ofport);
722 struct chassis_tunnel *tun_next;
723 HMAP_FOR_EACH_SAFE (tun, tun_next, hmap_node, &tunnels) {
724 hmap_remove(&tunnels, &tun->hmap_node);
725 free(tun);
726 }
727 hmap_destroy(&tunnels);
728
729 /* Table 0, Priority 100.
730 * ======================
731 *
732 * We have now determined the full set of port bindings associated with
733 * each "localnet" network. Only create flows for datapaths that have
734 * another local binding. Otherwise, we know it would just be dropped.
735 */
736 struct shash_node *ln_bindings_node, *ln_bindings_node_next;
737 SHASH_FOR_EACH_SAFE (ln_bindings_node, ln_bindings_node_next,
738 &localnet_inputs) {
739 struct localnet_bindings *ln_bindings = ln_bindings_node->data;
740 struct localnet_vlan *ln_vlan, *ln_vlan_next;
741 HMAP_FOR_EACH_SAFE (ln_vlan, ln_vlan_next, node, &ln_bindings->vlans) {
742 struct match match;
743 match_init_catchall(&match);
744 match_set_in_port(&match, ln_bindings->ofport);
745 if (ln_vlan->tag) {
746 match_set_dl_vlan(&match, htons(ln_vlan->tag));
747 } else {
748 /* Match priority-tagged frames, e.g. VLAN ID 0.
749 *
750 * We'll add a second flow for frames that lack any 802.1Q
751 * header later. */
752 match_set_dl_tci_masked(&match, htons(VLAN_CFI),
753 htons(VLAN_VID_MASK | VLAN_CFI));
754 }
755
756 struct ofpbuf ofpacts;
757 ofpbuf_init(&ofpacts, 0);
758
759 ofpact_put_STRIP_VLAN(&ofpacts);
760 uint32_t ofpacts_orig_size = ofpacts.size;
761
762 struct binding_elem *b;
763 LIST_FOR_EACH_POP (b, list_elem, &ln_vlan->bindings) {
764 struct hmap_node *ld;
765 ld = hmap_first_with_hash(&local_datapaths,
766 b->binding->datapath->tunnel_key);
767 if (ld) {
768 /* Set MFF_LOG_DATAPATH and MFF_LOG_INPORT. */
769 put_load(b->binding->datapath->tunnel_key, MFF_LOG_DATAPATH,
770 0, 64, &ofpacts);
771 put_load(b->binding->tunnel_key, MFF_LOG_INPORT, 0, 32,
772 &ofpacts);
773 put_resubmit(OFTABLE_LOG_INGRESS_PIPELINE, &ofpacts);
774 }
775
776 free(b);
777 }
778
779 if (ofpacts.size > ofpacts_orig_size) {
780 ofctrl_add_flow(flow_table, 0, 100, &match, &ofpacts);
781
782 if (!ln_vlan->tag) {
783 /* Add a second flow for frames that lack any 802.1Q
784 * header. For these, drop the OFPACT_STRIP_VLAN
785 * action. */
786 ofpbuf_pull(&ofpacts, ofpacts_orig_size);
787 match_set_dl_tci_masked(&match, 0, htons(VLAN_CFI));
788 ofctrl_add_flow(flow_table, 0, 100, &match, &ofpacts);
789 }
790 }
791
792 ofpbuf_uninit(&ofpacts);
793
794 hmap_remove(&ln_bindings->vlans, &ln_vlan->node);
795 free(ln_vlan);
796 }
797 shash_delete(&localnet_inputs, ln_bindings_node);
798 hmap_destroy(&ln_bindings->vlans);
799 free(ln_bindings);
800 }
801 shash_destroy(&localnet_inputs);
802
803 struct hmap_node *node;
804 while ((node = hmap_first(&local_datapaths))) {
805 hmap_remove(&local_datapaths, node);
806 free(node);
807 }
808 hmap_destroy(&local_datapaths);
809
810 simap_destroy(&localnet_to_ofport);
811 }