2 * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include "ofproto/ofproto-dpif.h"
20 #include "ofproto/ofproto-provider.h"
27 #include "byte-order.h"
28 #include "connectivity.h"
33 #include "dynamic-string.h"
34 #include "fail-open.h"
35 #include "guarded-list.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
42 #include "multipath.h"
43 #include "netdev-vport.h"
48 #include "odp-execute.h"
51 #include "ofp-actions.h"
52 #include "ofp-parse.h"
53 #include "ofp-print.h"
54 #include "ofproto-dpif-ipfix.h"
55 #include "ofproto-dpif-mirror.h"
56 #include "ofproto-dpif-monitor.h"
57 #include "ofproto-dpif-rid.h"
58 #include "ofproto-dpif-sflow.h"
59 #include "ofproto-dpif-upcall.h"
60 #include "ofproto-dpif-xlate.h"
61 #include "poll-loop.h"
62 #include "ovs-router.h"
68 #include "unaligned.h"
70 #include "vlan-bitmap.h"
73 VLOG_DEFINE_THIS_MODULE(ofproto_dpif);
75 COVERAGE_DEFINE(ofproto_dpif_expired);
76 COVERAGE_DEFINE(packet_in_overflow);
85 * - Do include packets and bytes from datapath flows which have not
86 * recently been processed by a revalidator. */
87 struct ovs_mutex stats_mutex;
88 struct dpif_flow_stats stats OVS_GUARDED;
90 /* If non-zero then the recirculation id that has
91 * been allocated for use with this rule.
92 * The recirculation id and associated internal flow should
93 * be freed when the rule is freed */
97 /* RULE_CAST() depends on this. */
98 BUILD_ASSERT_DECL(offsetof(struct rule_dpif, up) == 0);
100 static void rule_get_stats(struct rule *, uint64_t *packets, uint64_t *bytes,
101 long long int *used);
102 static struct rule_dpif *rule_dpif_cast(const struct rule *);
103 static void rule_expire(struct rule_dpif *);
110 * - Do include packets and bytes from datapath flows which have not
111 * recently been processed by a revalidator. */
112 struct ovs_mutex stats_mutex;
113 uint64_t packet_count OVS_GUARDED; /* Number of packets received. */
114 uint64_t byte_count OVS_GUARDED; /* Number of bytes received. */
118 struct hmap_node hmap_node; /* In struct ofproto's "bundles" hmap. */
119 struct ofproto_dpif *ofproto; /* Owning ofproto. */
120 void *aux; /* Key supplied by ofproto's client. */
121 char *name; /* Identifier for log messages. */
124 struct list ports; /* Contains "struct ofport"s. */
125 enum port_vlan_mode vlan_mode; /* VLAN mode */
126 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
127 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
128 * NULL if all VLANs are trunked. */
129 struct lacp *lacp; /* LACP if LACP is enabled, otherwise NULL. */
130 struct bond *bond; /* Nonnull iff more than one port. */
131 bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */
134 bool floodable; /* True if no port has OFPUTIL_PC_NO_FLOOD set. */
137 static void bundle_remove(struct ofport *);
138 static void bundle_update(struct ofbundle *);
139 static void bundle_destroy(struct ofbundle *);
140 static void bundle_del_port(struct ofport_dpif *);
141 static void bundle_run(struct ofbundle *);
142 static void bundle_wait(struct ofbundle *);
143 static void bundle_flush_macs(struct ofbundle *, bool);
144 static void bundle_move(struct ofbundle *, struct ofbundle *);
146 static void stp_run(struct ofproto_dpif *ofproto);
147 static void stp_wait(struct ofproto_dpif *ofproto);
148 static int set_stp_port(struct ofport *,
149 const struct ofproto_port_stp_settings *);
151 static void rstp_run(struct ofproto_dpif *ofproto);
152 static void set_rstp_port(struct ofport *,
153 const struct ofproto_port_rstp_settings *);
156 struct hmap_node odp_port_node; /* In dpif_backer's "odp_to_ofport_map". */
160 struct ofbundle *bundle; /* Bundle that contains this port, if any. */
161 struct list bundle_node; /* In struct ofbundle's "ports" list. */
162 struct cfm *cfm; /* Connectivity Fault Management, if any. */
163 struct bfd *bfd; /* BFD, if any. */
164 bool may_enable; /* May be enabled in bonds. */
165 bool is_tunnel; /* This port is a tunnel. */
166 bool is_layer3; /* This is a layer 3 port. */
167 long long int carrier_seq; /* Carrier status changes. */
168 struct ofport_dpif *peer; /* Peer if patch port. */
171 struct stp_port *stp_port; /* Spanning Tree Protocol, if any. */
172 enum stp_state stp_state; /* Always STP_DISABLED if STP not in use. */
173 long long int stp_state_entered;
175 /* Rapid Spanning Tree. */
176 struct rstp_port *rstp_port; /* Rapid Spanning Tree Protocol, if any. */
177 enum rstp_state rstp_state; /* Always RSTP_DISABLED if RSTP not in use. */
179 /* Queue to DSCP mapping. */
180 struct ofproto_port_queue *qdscp;
183 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
185 * This is deprecated. It is only for compatibility with broken device
186 * drivers in old versions of Linux that do not properly support VLANs when
187 * VLAN devices are not used. When broken device drivers are no longer in
188 * widespread use, we will delete these interfaces. */
189 ofp_port_t realdev_ofp_port;
193 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
195 * This is deprecated. It is only for compatibility with broken device drivers
196 * in old versions of Linux that do not properly support VLANs when VLAN
197 * devices are not used. When broken device drivers are no longer in
198 * widespread use, we will delete these interfaces. */
199 struct vlan_splinter {
200 struct hmap_node realdev_vid_node;
201 struct hmap_node vlandev_node;
202 ofp_port_t realdev_ofp_port;
203 ofp_port_t vlandev_ofp_port;
207 static void vsp_remove(struct ofport_dpif *);
208 static void vsp_add(struct ofport_dpif *, ofp_port_t realdev_ofp_port, int vid);
210 static odp_port_t ofp_port_to_odp_port(const struct ofproto_dpif *,
213 static ofp_port_t odp_port_to_ofp_port(const struct ofproto_dpif *,
216 static struct ofport_dpif *
217 ofport_dpif_cast(const struct ofport *ofport)
219 return ofport ? CONTAINER_OF(ofport, struct ofport_dpif, up) : NULL;
222 static void port_run(struct ofport_dpif *);
223 static int set_bfd(struct ofport *, const struct smap *);
224 static int set_cfm(struct ofport *, const struct cfm_settings *);
225 static void ofport_update_peer(struct ofport_dpif *);
227 /* Reasons that we might need to revalidate every datapath flow, and
228 * corresponding coverage counters.
230 * A value of 0 means that there is no need to revalidate.
232 * It would be nice to have some cleaner way to integrate with coverage
233 * counters, but with only a few reasons I guess this is good enough for
235 enum revalidate_reason {
236 REV_RECONFIGURE = 1, /* Switch configuration changed. */
237 REV_STP, /* Spanning tree protocol port status change. */
238 REV_RSTP, /* RSTP port status change. */
239 REV_BOND, /* Bonding changed. */
240 REV_PORT_TOGGLED, /* Port enabled or disabled by CFM, LACP, ...*/
241 REV_FLOW_TABLE, /* Flow table changed. */
242 REV_MAC_LEARNING, /* Mac learning changed. */
243 REV_MCAST_SNOOPING, /* Multicast snooping changed. */
245 COVERAGE_DEFINE(rev_reconfigure);
246 COVERAGE_DEFINE(rev_stp);
247 COVERAGE_DEFINE(rev_rstp);
248 COVERAGE_DEFINE(rev_bond);
249 COVERAGE_DEFINE(rev_port_toggled);
250 COVERAGE_DEFINE(rev_flow_table);
251 COVERAGE_DEFINE(rev_mac_learning);
252 COVERAGE_DEFINE(rev_mcast_snooping);
254 /* All datapaths of a given type share a single dpif backer instance. */
261 struct ovs_rwlock odp_to_ofport_lock;
262 struct hmap odp_to_ofport_map OVS_GUARDED; /* Contains "struct ofport"s. */
264 struct simap tnl_backers; /* Set of dpif ports backing tunnels. */
266 enum revalidate_reason need_revalidate; /* Revalidate all flows. */
268 bool recv_set_enable; /* Enables or disables receiving packets. */
271 struct recirc_id_pool *rid_pool; /* Recirculation ID pool. */
272 bool enable_recirc; /* True if the datapath supports recirculation */
274 /* True if the datapath supports variable-length
275 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
276 * False if the datapath supports only 8-byte (or shorter) userdata. */
277 bool variable_length_userdata;
279 /* True if the datapath supports masked data in OVS_ACTION_ATTR_SET
281 bool masked_set_action;
283 /* Maximum number of MPLS label stack entries that the datapath supports
285 size_t max_mpls_depth;
287 /* Version string of the datapath stored in OVSDB. */
288 char *dp_version_string;
290 /* True if the datapath supports tnl_push and pop actions. */
291 bool enable_tnl_push_pop;
292 struct atomic_count tnl_count;
295 /* All existing ofproto_backer instances, indexed by ofproto->up.type. */
296 static struct shash all_dpif_backers = SHASH_INITIALIZER(&all_dpif_backers);
298 struct ofproto_dpif {
299 struct hmap_node all_ofproto_dpifs_node; /* In 'all_ofproto_dpifs'. */
301 struct dpif_backer *backer;
303 uint64_t dump_seq; /* Last read of udpif_dump_seq(). */
305 /* Special OpenFlow rules. */
306 struct rule_dpif *miss_rule; /* Sends flow table misses to controller. */
307 struct rule_dpif *no_packet_in_rule; /* Drops flow table misses. */
308 struct rule_dpif *drop_frags_rule; /* Used in OFPC_FRAG_DROP mode. */
311 struct netflow *netflow;
312 struct dpif_sflow *sflow;
313 struct dpif_ipfix *ipfix;
314 struct hmap bundles; /* Contains "struct ofbundle"s. */
315 struct mac_learning *ml;
316 struct mcast_snooping *ms;
317 bool has_bonded_bundles;
319 struct mbridge *mbridge;
321 struct ovs_mutex stats_mutex;
322 struct netdev_stats stats OVS_GUARDED; /* To account packets generated and
323 * consumed in userspace. */
327 long long int stp_last_tick;
329 /* Rapid Spanning Tree. */
331 long long int rstp_last_tick;
333 /* VLAN splinters. */
334 struct ovs_mutex vsp_mutex;
335 struct hmap realdev_vid_map OVS_GUARDED; /* (realdev,vid) -> vlandev. */
336 struct hmap vlandev_map OVS_GUARDED; /* vlandev -> (realdev,vid). */
339 struct sset ports; /* Set of standard port names. */
340 struct sset ghost_ports; /* Ports with no datapath port. */
341 struct sset port_poll_set; /* Queued names for port_poll() reply. */
342 int port_poll_errno; /* Last errno for port_poll() reply. */
343 uint64_t change_seq; /* Connectivity status changes. */
346 struct guarded_list pins; /* Contains "struct ofputil_packet_in"s. */
347 struct seq *pins_seq; /* For notifying 'pins' reception. */
351 /* All existing ofproto_dpif instances, indexed by ->up.name. */
352 static struct hmap all_ofproto_dpifs = HMAP_INITIALIZER(&all_ofproto_dpifs);
354 static bool ofproto_use_tnl_push_pop = true;
355 static void ofproto_unixctl_init(void);
357 static inline struct ofproto_dpif *
358 ofproto_dpif_cast(const struct ofproto *ofproto)
360 ovs_assert(ofproto->ofproto_class == &ofproto_dpif_class);
361 return CONTAINER_OF(ofproto, struct ofproto_dpif, up);
365 ofproto_dpif_get_max_mpls_depth(const struct ofproto_dpif *ofproto)
367 return ofproto->backer->max_mpls_depth;
371 ofproto_dpif_get_enable_recirc(const struct ofproto_dpif *ofproto)
373 return ofproto->backer->enable_recirc;
376 static struct ofport_dpif *get_ofp_port(const struct ofproto_dpif *ofproto,
377 ofp_port_t ofp_port);
378 static void ofproto_trace(struct ofproto_dpif *, struct flow *,
379 const struct ofpbuf *packet,
380 const struct ofpact[], size_t ofpacts_len,
383 /* Global variables. */
384 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
386 /* Initial mappings of port to bridge mappings. */
387 static struct shash init_ofp_ports = SHASH_INITIALIZER(&init_ofp_ports);
389 /* Executes 'fm'. The caller retains ownership of 'fm' and everything in
392 ofproto_dpif_flow_mod(struct ofproto_dpif *ofproto,
393 struct ofputil_flow_mod *fm)
395 ofproto_flow_mod(&ofproto->up, fm);
398 /* Appends 'pin' to the queue of "packet ins" to be sent to the controller.
399 * Takes ownership of 'pin' and pin->packet. */
401 ofproto_dpif_send_packet_in(struct ofproto_dpif *ofproto,
402 struct ofproto_packet_in *pin)
404 if (!guarded_list_push_back(&ofproto->pins, &pin->list_node, 1024)) {
405 COVERAGE_INC(packet_in_overflow);
406 free(CONST_CAST(void *, pin->up.packet));
410 /* Wakes up main thread for packet-in I/O. */
411 seq_change(ofproto->pins_seq);
414 /* The default "table-miss" behaviour for OpenFlow1.3+ is to drop the
415 * packet rather than to send the packet to the controller.
417 * This function returns false to indicate that a packet_in message
418 * for a "table-miss" should be sent to at least one controller.
419 * False otherwise. */
421 ofproto_dpif_wants_packet_in_on_miss(struct ofproto_dpif *ofproto)
423 return connmgr_wants_packet_in_on_miss(ofproto->up.connmgr);
426 /* Factory functions. */
429 init(const struct shash *iface_hints)
431 struct shash_node *node;
433 /* Make a local copy, since we don't own 'iface_hints' elements. */
434 SHASH_FOR_EACH(node, iface_hints) {
435 const struct iface_hint *orig_hint = node->data;
436 struct iface_hint *new_hint = xmalloc(sizeof *new_hint);
438 new_hint->br_name = xstrdup(orig_hint->br_name);
439 new_hint->br_type = xstrdup(orig_hint->br_type);
440 new_hint->ofp_port = orig_hint->ofp_port;
442 shash_add(&init_ofp_ports, node->name, new_hint);
447 enumerate_types(struct sset *types)
449 dp_enumerate_types(types);
453 enumerate_names(const char *type, struct sset *names)
455 struct ofproto_dpif *ofproto;
458 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
459 if (strcmp(type, ofproto->up.type)) {
462 sset_add(names, ofproto->up.name);
469 del(const char *type, const char *name)
474 error = dpif_open(name, type, &dpif);
476 error = dpif_delete(dpif);
483 port_open_type(const char *datapath_type, const char *port_type)
485 return dpif_port_open_type(datapath_type, port_type);
488 /* Type functions. */
490 static void process_dpif_port_changes(struct dpif_backer *);
491 static void process_dpif_all_ports_changed(struct dpif_backer *);
492 static void process_dpif_port_change(struct dpif_backer *,
493 const char *devname);
494 static void process_dpif_port_error(struct dpif_backer *, int error);
496 static struct ofproto_dpif *
497 lookup_ofproto_dpif_by_port_name(const char *name)
499 struct ofproto_dpif *ofproto;
501 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
502 if (sset_contains(&ofproto->ports, name)) {
511 type_run(const char *type)
513 struct dpif_backer *backer;
515 backer = shash_find_data(&all_dpif_backers, type);
517 /* This is not necessarily a problem, since backers are only
518 * created on demand. */
523 if (dpif_run(backer->dpif)) {
524 backer->need_revalidate = REV_RECONFIGURE;
527 udpif_run(backer->udpif);
529 /* If vswitchd started with other_config:flow_restore_wait set as "true",
530 * and the configuration has now changed to "false", enable receiving
531 * packets from the datapath. */
532 if (!backer->recv_set_enable && !ofproto_get_flow_restore_wait()) {
535 backer->recv_set_enable = true;
537 error = dpif_recv_set(backer->dpif, backer->recv_set_enable);
539 VLOG_ERR("Failed to enable receiving packets in dpif.");
542 dpif_flow_flush(backer->dpif);
543 backer->need_revalidate = REV_RECONFIGURE;
546 if (backer->recv_set_enable) {
547 udpif_set_threads(backer->udpif, n_handlers, n_revalidators);
550 dpif_poll_threads_set(backer->dpif, n_dpdk_rxqs, pmd_cpu_mask);
552 if (backer->need_revalidate) {
553 struct ofproto_dpif *ofproto;
554 struct simap_node *node;
555 struct simap tmp_backers;
557 /* Handle tunnel garbage collection. */
558 simap_init(&tmp_backers);
559 simap_swap(&backer->tnl_backers, &tmp_backers);
561 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
562 struct ofport_dpif *iter;
564 if (backer != ofproto->backer) {
568 HMAP_FOR_EACH (iter, up.hmap_node, &ofproto->up.ports) {
569 char namebuf[NETDEV_VPORT_NAME_BUFSIZE];
572 if (!iter->is_tunnel) {
576 dp_port = netdev_vport_get_dpif_port(iter->up.netdev,
577 namebuf, sizeof namebuf);
578 node = simap_find(&tmp_backers, dp_port);
580 simap_put(&backer->tnl_backers, dp_port, node->data);
581 simap_delete(&tmp_backers, node);
582 node = simap_find(&backer->tnl_backers, dp_port);
584 node = simap_find(&backer->tnl_backers, dp_port);
586 odp_port_t odp_port = ODPP_NONE;
588 if (!dpif_port_add(backer->dpif, iter->up.netdev,
590 simap_put(&backer->tnl_backers, dp_port,
591 odp_to_u32(odp_port));
592 node = simap_find(&backer->tnl_backers, dp_port);
597 iter->odp_port = node ? u32_to_odp(node->data) : ODPP_NONE;
598 if (tnl_port_reconfigure(iter, iter->up.netdev,
600 ovs_native_tunneling_is_on(ofproto), dp_port)) {
601 backer->need_revalidate = REV_RECONFIGURE;
606 SIMAP_FOR_EACH (node, &tmp_backers) {
607 dpif_port_del(backer->dpif, u32_to_odp(node->data));
609 simap_destroy(&tmp_backers);
611 switch (backer->need_revalidate) {
612 case REV_RECONFIGURE: COVERAGE_INC(rev_reconfigure); break;
613 case REV_STP: COVERAGE_INC(rev_stp); break;
614 case REV_RSTP: COVERAGE_INC(rev_rstp); break;
615 case REV_BOND: COVERAGE_INC(rev_bond); break;
616 case REV_PORT_TOGGLED: COVERAGE_INC(rev_port_toggled); break;
617 case REV_FLOW_TABLE: COVERAGE_INC(rev_flow_table); break;
618 case REV_MAC_LEARNING: COVERAGE_INC(rev_mac_learning); break;
619 case REV_MCAST_SNOOPING: COVERAGE_INC(rev_mcast_snooping); break;
621 backer->need_revalidate = 0;
623 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
624 struct ofport_dpif *ofport;
625 struct ofbundle *bundle;
627 if (ofproto->backer != backer) {
632 xlate_ofproto_set(ofproto, ofproto->up.name,
633 ofproto->backer->dpif, ofproto->ml,
634 ofproto->stp, ofproto->rstp, ofproto->ms,
635 ofproto->mbridge, ofproto->sflow, ofproto->ipfix,
637 ofproto->up.forward_bpdu,
638 connmgr_has_in_band(ofproto->up.connmgr),
639 ofproto->backer->enable_recirc,
640 ofproto->backer->variable_length_userdata,
641 ofproto->backer->max_mpls_depth,
642 ofproto->backer->masked_set_action);
644 HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) {
645 xlate_bundle_set(ofproto, bundle, bundle->name,
646 bundle->vlan_mode, bundle->vlan,
647 bundle->trunks, bundle->use_priority_tags,
648 bundle->bond, bundle->lacp,
652 HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
653 int stp_port = ofport->stp_port
654 ? stp_port_no(ofport->stp_port)
656 xlate_ofport_set(ofproto, ofport->bundle, ofport,
657 ofport->up.ofp_port, ofport->odp_port,
658 ofport->up.netdev, ofport->cfm,
659 ofport->bfd, ofport->peer, stp_port,
660 ofport->rstp_port, ofport->qdscp,
661 ofport->n_qdscp, ofport->up.pp.config,
662 ofport->up.pp.state, ofport->is_tunnel,
668 udpif_revalidate(backer->udpif);
671 process_dpif_port_changes(backer);
676 /* Check for and handle port changes in 'backer''s dpif. */
678 process_dpif_port_changes(struct dpif_backer *backer)
684 error = dpif_port_poll(backer->dpif, &devname);
690 process_dpif_all_ports_changed(backer);
694 process_dpif_port_change(backer, devname);
699 process_dpif_port_error(backer, error);
706 process_dpif_all_ports_changed(struct dpif_backer *backer)
708 struct ofproto_dpif *ofproto;
709 struct dpif_port dpif_port;
710 struct dpif_port_dump dump;
711 struct sset devnames;
714 sset_init(&devnames);
715 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
716 if (ofproto->backer == backer) {
717 struct ofport *ofport;
719 HMAP_FOR_EACH (ofport, hmap_node, &ofproto->up.ports) {
720 sset_add(&devnames, netdev_get_name(ofport->netdev));
724 DPIF_PORT_FOR_EACH (&dpif_port, &dump, backer->dpif) {
725 sset_add(&devnames, dpif_port.name);
728 SSET_FOR_EACH (devname, &devnames) {
729 process_dpif_port_change(backer, devname);
731 sset_destroy(&devnames);
735 process_dpif_port_change(struct dpif_backer *backer, const char *devname)
737 struct ofproto_dpif *ofproto;
738 struct dpif_port port;
740 /* Don't report on the datapath's device. */
741 if (!strcmp(devname, dpif_base_name(backer->dpif))) {
745 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node,
746 &all_ofproto_dpifs) {
747 if (simap_contains(&ofproto->backer->tnl_backers, devname)) {
752 ofproto = lookup_ofproto_dpif_by_port_name(devname);
753 if (dpif_port_query_by_name(backer->dpif, devname, &port)) {
754 /* The port was removed. If we know the datapath,
755 * report it through poll_set(). If we don't, it may be
756 * notifying us of a removal we initiated, so ignore it.
757 * If there's a pending ENOBUFS, let it stand, since
758 * everything will be reevaluated. */
759 if (ofproto && ofproto->port_poll_errno != ENOBUFS) {
760 sset_add(&ofproto->port_poll_set, devname);
761 ofproto->port_poll_errno = 0;
763 } else if (!ofproto) {
764 /* The port was added, but we don't know with which
765 * ofproto we should associate it. Delete it. */
766 dpif_port_del(backer->dpif, port.port_no);
768 struct ofport_dpif *ofport;
770 ofport = ofport_dpif_cast(shash_find_data(
771 &ofproto->up.port_by_name, devname));
773 && ofport->odp_port != port.port_no
774 && !odp_port_to_ofport(backer, port.port_no))
776 /* 'ofport''s datapath port number has changed from
777 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
778 * structures to match. */
779 ovs_rwlock_wrlock(&backer->odp_to_ofport_lock);
780 hmap_remove(&backer->odp_to_ofport_map, &ofport->odp_port_node);
781 ofport->odp_port = port.port_no;
782 hmap_insert(&backer->odp_to_ofport_map, &ofport->odp_port_node,
783 hash_odp_port(port.port_no));
784 ovs_rwlock_unlock(&backer->odp_to_ofport_lock);
785 backer->need_revalidate = REV_RECONFIGURE;
788 dpif_port_destroy(&port);
791 /* Propagate 'error' to all ofprotos based on 'backer'. */
793 process_dpif_port_error(struct dpif_backer *backer, int error)
795 struct ofproto_dpif *ofproto;
797 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
798 if (ofproto->backer == backer) {
799 sset_clear(&ofproto->port_poll_set);
800 ofproto->port_poll_errno = error;
806 type_wait(const char *type)
808 struct dpif_backer *backer;
810 backer = shash_find_data(&all_dpif_backers, type);
812 /* This is not necessarily a problem, since backers are only
813 * created on demand. */
817 dpif_wait(backer->dpif);
820 /* Basic life-cycle. */
822 static int add_internal_flows(struct ofproto_dpif *);
824 static struct ofproto *
827 struct ofproto_dpif *ofproto = xmalloc(sizeof *ofproto);
832 dealloc(struct ofproto *ofproto_)
834 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
839 close_dpif_backer(struct dpif_backer *backer)
841 ovs_assert(backer->refcount > 0);
843 if (--backer->refcount) {
847 udpif_destroy(backer->udpif);
849 simap_destroy(&backer->tnl_backers);
850 ovs_rwlock_destroy(&backer->odp_to_ofport_lock);
851 hmap_destroy(&backer->odp_to_ofport_map);
852 shash_find_and_delete(&all_dpif_backers, backer->type);
853 recirc_id_pool_destroy(backer->rid_pool);
855 free(backer->dp_version_string);
856 dpif_close(backer->dpif);
860 /* Datapath port slated for removal from datapath. */
862 struct list list_node;
866 static bool check_variable_length_userdata(struct dpif_backer *backer);
867 static size_t check_max_mpls_depth(struct dpif_backer *backer);
868 static bool check_recirc(struct dpif_backer *backer);
869 static bool check_masked_set_action(struct dpif_backer *backer);
872 open_dpif_backer(const char *type, struct dpif_backer **backerp)
874 struct dpif_backer *backer;
875 struct dpif_port_dump port_dump;
876 struct dpif_port port;
877 struct shash_node *node;
878 struct list garbage_list;
879 struct odp_garbage *garbage, *next;
886 backer = shash_find_data(&all_dpif_backers, type);
893 backer_name = xasprintf("ovs-%s", type);
895 /* Remove any existing datapaths, since we assume we're the only
896 * userspace controlling the datapath. */
898 dp_enumerate_names(type, &names);
899 SSET_FOR_EACH(name, &names) {
900 struct dpif *old_dpif;
902 /* Don't remove our backer if it exists. */
903 if (!strcmp(name, backer_name)) {
907 if (dpif_open(name, type, &old_dpif)) {
908 VLOG_WARN("couldn't open old datapath %s to remove it", name);
910 dpif_delete(old_dpif);
911 dpif_close(old_dpif);
914 sset_destroy(&names);
916 backer = xmalloc(sizeof *backer);
918 error = dpif_create_and_open(backer_name, type, &backer->dpif);
921 VLOG_ERR("failed to open datapath of type %s: %s", type,
922 ovs_strerror(error));
926 backer->udpif = udpif_create(backer, backer->dpif);
928 backer->type = xstrdup(type);
929 backer->refcount = 1;
930 hmap_init(&backer->odp_to_ofport_map);
931 ovs_rwlock_init(&backer->odp_to_ofport_lock);
932 backer->need_revalidate = 0;
933 simap_init(&backer->tnl_backers);
934 backer->recv_set_enable = !ofproto_get_flow_restore_wait();
937 if (backer->recv_set_enable) {
938 dpif_flow_flush(backer->dpif);
941 /* Loop through the ports already on the datapath and remove any
942 * that we don't need anymore. */
943 list_init(&garbage_list);
944 dpif_port_dump_start(&port_dump, backer->dpif);
945 while (dpif_port_dump_next(&port_dump, &port)) {
946 node = shash_find(&init_ofp_ports, port.name);
947 if (!node && strcmp(port.name, dpif_base_name(backer->dpif))) {
948 garbage = xmalloc(sizeof *garbage);
949 garbage->odp_port = port.port_no;
950 list_push_front(&garbage_list, &garbage->list_node);
953 dpif_port_dump_done(&port_dump);
955 LIST_FOR_EACH_SAFE (garbage, next, list_node, &garbage_list) {
956 dpif_port_del(backer->dpif, garbage->odp_port);
957 list_remove(&garbage->list_node);
961 shash_add(&all_dpif_backers, type, backer);
963 backer->enable_recirc = check_recirc(backer);
964 backer->max_mpls_depth = check_max_mpls_depth(backer);
965 backer->masked_set_action = check_masked_set_action(backer);
966 backer->rid_pool = recirc_id_pool_create();
968 backer->enable_tnl_push_pop = dpif_supports_tnl_push_pop(backer->dpif);
969 atomic_count_init(&backer->tnl_count, 0);
971 error = dpif_recv_set(backer->dpif, backer->recv_set_enable);
973 VLOG_ERR("failed to listen on datapath of type %s: %s",
974 type, ovs_strerror(error));
975 close_dpif_backer(backer);
979 if (backer->recv_set_enable) {
980 udpif_set_threads(backer->udpif, n_handlers, n_revalidators);
983 /* This check fails if performed before udpif threads have been set,
984 * as the kernel module checks that the 'pid' in userspace action
986 backer->variable_length_userdata = check_variable_length_userdata(backer);
987 backer->dp_version_string = dpif_get_dp_version(backer->dpif);
993 ovs_native_tunneling_is_on(struct ofproto_dpif *ofproto)
995 return ofproto_use_tnl_push_pop && ofproto->backer->enable_tnl_push_pop &&
996 atomic_count_get(&ofproto->backer->tnl_count);
999 /* Tests whether 'backer''s datapath supports recirculation. Only newer
1000 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
1001 * features on older datapaths that don't support this feature.
1003 * Returns false if 'backer' definitely does not support recirculation, true if
1004 * it seems to support recirculation or if at least the error we get is
1007 check_recirc(struct dpif_backer *backer)
1010 struct odputil_keybuf keybuf;
1014 memset(&flow, 0, sizeof flow);
1018 ofpbuf_use_stack(&key, &keybuf, sizeof keybuf);
1019 odp_flow_key_from_flow(&key, &flow, NULL, 0, true);
1020 enable_recirc = dpif_probe_feature(backer->dpif, "recirculation", &key,
1023 if (enable_recirc) {
1024 VLOG_INFO("%s: Datapath supports recirculation",
1025 dpif_name(backer->dpif));
1027 VLOG_INFO("%s: Datapath does not support recirculation",
1028 dpif_name(backer->dpif));
1031 return enable_recirc;
1034 /* Tests whether 'backer''s datapath supports variable-length
1035 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
1036 * to disable some features on older datapaths that don't support this
1039 * Returns false if 'backer' definitely does not support variable-length
1040 * userdata, true if it seems to support them or if at least the error we get
1043 check_variable_length_userdata(struct dpif_backer *backer)
1045 struct eth_header *eth;
1046 struct ofpbuf actions;
1047 struct dpif_execute execute;
1048 struct ofpbuf packet;
1052 /* Compose a userspace action that will cause an ERANGE error on older
1053 * datapaths that don't support variable-length userdata.
1055 * We really test for using userdata longer than 8 bytes, but older
1056 * datapaths accepted these, silently truncating the userdata to 8 bytes.
1057 * The same older datapaths rejected userdata shorter than 8 bytes, so we
1058 * test for that instead as a proxy for longer userdata support. */
1059 ofpbuf_init(&actions, 64);
1060 start = nl_msg_start_nested(&actions, OVS_ACTION_ATTR_USERSPACE);
1061 nl_msg_put_u32(&actions, OVS_USERSPACE_ATTR_PID,
1062 dpif_port_get_pid(backer->dpif, ODPP_NONE, 0));
1063 nl_msg_put_unspec_zero(&actions, OVS_USERSPACE_ATTR_USERDATA, 4);
1064 nl_msg_end_nested(&actions, start);
1066 /* Compose a dummy ethernet packet. */
1067 ofpbuf_init(&packet, ETH_HEADER_LEN);
1068 eth = ofpbuf_put_zeros(&packet, ETH_HEADER_LEN);
1069 eth->eth_type = htons(0x1234);
1071 /* Execute the actions. On older datapaths this fails with ERANGE, on
1072 * newer datapaths it succeeds. */
1073 execute.actions = ofpbuf_data(&actions);
1074 execute.actions_len = ofpbuf_size(&actions);
1075 execute.packet = &packet;
1076 execute.md = PKT_METADATA_INITIALIZER(0);
1077 execute.needs_help = false;
1078 execute.probe = true;
1080 error = dpif_execute(backer->dpif, &execute);
1082 ofpbuf_uninit(&packet);
1083 ofpbuf_uninit(&actions);
1090 /* Variable-length userdata is not supported. */
1091 VLOG_WARN("%s: datapath does not support variable-length userdata "
1092 "feature (needs Linux 3.10+ or kernel module from OVS "
1093 "1..11+). The NXAST_SAMPLE action will be ignored.",
1094 dpif_name(backer->dpif));
1098 /* Something odd happened. We're not sure whether variable-length
1099 * userdata is supported. Default to "yes". */
1100 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
1101 dpif_name(backer->dpif), ovs_strerror(error));
1106 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
1108 * Returns the number of elements in a struct flow's mpls_lse field
1109 * if the datapath supports at least that many entries in an
1111 * Otherwise returns the number of MPLS push actions supported by
1114 check_max_mpls_depth(struct dpif_backer *backer)
1119 for (n = 0; n < FLOW_MAX_MPLS_LABELS; n++) {
1120 struct odputil_keybuf keybuf;
1123 memset(&flow, 0, sizeof flow);
1124 flow.dl_type = htons(ETH_TYPE_MPLS);
1125 flow_set_mpls_bos(&flow, n, 1);
1127 ofpbuf_use_stack(&key, &keybuf, sizeof keybuf);
1128 odp_flow_key_from_flow(&key, &flow, NULL, 0, false);
1129 if (!dpif_probe_feature(backer->dpif, "MPLS", &key, NULL)) {
1134 VLOG_INFO("%s: MPLS label stack length probed as %d",
1135 dpif_name(backer->dpif), n);
1139 /* Tests whether 'backer''s datapath supports masked data in
1140 * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
1141 * datapaths that don't support this feature. */
1143 check_masked_set_action(struct dpif_backer *backer)
1145 struct eth_header *eth;
1146 struct ofpbuf actions;
1147 struct dpif_execute execute;
1148 struct ofpbuf packet;
1150 struct ovs_key_ethernet key, mask;
1152 /* Compose a set action that will cause an EINVAL error on older
1153 * datapaths that don't support masked set actions.
1154 * Avoid using a full mask, as it could be translated to a non-masked
1155 * set action instead. */
1156 ofpbuf_init(&actions, 64);
1157 memset(&key, 0x53, sizeof key);
1158 memset(&mask, 0x7f, sizeof mask);
1159 commit_masked_set_action(&actions, OVS_KEY_ATTR_ETHERNET, &key, &mask,
1162 /* Compose a dummy ethernet packet. */
1163 ofpbuf_init(&packet, ETH_HEADER_LEN);
1164 eth = ofpbuf_put_zeros(&packet, ETH_HEADER_LEN);
1165 eth->eth_type = htons(0x1234);
1167 /* Execute the actions. On older datapaths this fails with EINVAL, on
1168 * newer datapaths it succeeds. */
1169 execute.actions = ofpbuf_data(&actions);
1170 execute.actions_len = ofpbuf_size(&actions);
1171 execute.packet = &packet;
1172 execute.md = PKT_METADATA_INITIALIZER(0);
1173 execute.needs_help = false;
1174 execute.probe = true;
1176 error = dpif_execute(backer->dpif, &execute);
1178 ofpbuf_uninit(&packet);
1179 ofpbuf_uninit(&actions);
1182 /* Masked set action is not supported. */
1183 VLOG_INFO("%s: datapath does not support masked set action feature.",
1184 dpif_name(backer->dpif));
1190 construct(struct ofproto *ofproto_)
1192 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1193 struct shash_node *node, *next;
1196 error = open_dpif_backer(ofproto->up.type, &ofproto->backer);
1201 ofproto->netflow = NULL;
1202 ofproto->sflow = NULL;
1203 ofproto->ipfix = NULL;
1204 ofproto->stp = NULL;
1205 ofproto->rstp = NULL;
1206 ofproto->dump_seq = 0;
1207 hmap_init(&ofproto->bundles);
1208 ofproto->ml = mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME);
1210 ofproto->mbridge = mbridge_create();
1211 ofproto->has_bonded_bundles = false;
1212 ofproto->lacp_enabled = false;
1213 ofproto_tunnel_init();
1214 ovs_mutex_init_adaptive(&ofproto->stats_mutex);
1215 ovs_mutex_init(&ofproto->vsp_mutex);
1217 guarded_list_init(&ofproto->pins);
1219 ofproto_unixctl_init();
1221 hmap_init(&ofproto->vlandev_map);
1222 hmap_init(&ofproto->realdev_vid_map);
1224 sset_init(&ofproto->ports);
1225 sset_init(&ofproto->ghost_ports);
1226 sset_init(&ofproto->port_poll_set);
1227 ofproto->port_poll_errno = 0;
1228 ofproto->change_seq = 0;
1229 ofproto->pins_seq = seq_create();
1230 ofproto->pins_seqno = seq_read(ofproto->pins_seq);
1233 SHASH_FOR_EACH_SAFE (node, next, &init_ofp_ports) {
1234 struct iface_hint *iface_hint = node->data;
1236 if (!strcmp(iface_hint->br_name, ofproto->up.name)) {
1237 /* Check if the datapath already has this port. */
1238 if (dpif_port_exists(ofproto->backer->dpif, node->name)) {
1239 sset_add(&ofproto->ports, node->name);
1242 free(iface_hint->br_name);
1243 free(iface_hint->br_type);
1245 shash_delete(&init_ofp_ports, node);
1249 hmap_insert(&all_ofproto_dpifs, &ofproto->all_ofproto_dpifs_node,
1250 hash_string(ofproto->up.name, 0));
1251 memset(&ofproto->stats, 0, sizeof ofproto->stats);
1253 ofproto_init_tables(ofproto_, N_TABLES);
1254 error = add_internal_flows(ofproto);
1256 ofproto->up.tables[TBL_INTERNAL].flags = OFTABLE_HIDDEN | OFTABLE_READONLY;
1262 add_internal_miss_flow(struct ofproto_dpif *ofproto, int id,
1263 const struct ofpbuf *ofpacts, struct rule_dpif **rulep)
1269 match_init_catchall(&match);
1270 match_set_reg(&match, 0, id);
1272 error = ofproto_dpif_add_internal_flow(ofproto, &match, 0, 0, ofpacts,
1274 *rulep = error ? NULL : rule_dpif_cast(rule);
1280 add_internal_flows(struct ofproto_dpif *ofproto)
1282 struct ofpact_controller *controller;
1283 uint64_t ofpacts_stub[128 / 8];
1284 struct ofpbuf ofpacts;
1285 struct rule *unused_rulep OVS_UNUSED;
1286 struct ofpact_resubmit *resubmit;
1291 ofpbuf_use_stack(&ofpacts, ofpacts_stub, sizeof ofpacts_stub);
1294 controller = ofpact_put_CONTROLLER(&ofpacts);
1295 controller->max_len = UINT16_MAX;
1296 controller->controller_id = 0;
1297 controller->reason = OFPR_NO_MATCH;
1298 ofpact_pad(&ofpacts);
1300 error = add_internal_miss_flow(ofproto, id++, &ofpacts,
1301 &ofproto->miss_rule);
1306 ofpbuf_clear(&ofpacts);
1307 error = add_internal_miss_flow(ofproto, id++, &ofpacts,
1308 &ofproto->no_packet_in_rule);
1313 error = add_internal_miss_flow(ofproto, id++, &ofpacts,
1314 &ofproto->drop_frags_rule);
1319 /* Drop any run away non-recirc rule lookups. Recirc_id has to be
1320 * zero when reaching this rule.
1322 * (priority=2), recirc_id=0, actions=drop
1324 ofpbuf_clear(&ofpacts);
1325 match_init_catchall(&match);
1326 match_set_recirc_id(&match, 0);
1327 error = ofproto_dpif_add_internal_flow(ofproto, &match, 2, 0, &ofpacts,
1333 /* Continue rule lookups for not-matched recirc rules from table 0.
1335 * (priority=1), actions=resubmit(, 0)
1337 resubmit = ofpact_put_RESUBMIT(&ofpacts);
1338 resubmit->in_port = OFPP_IN_PORT;
1339 resubmit->table_id = 0;
1341 match_init_catchall(&match);
1342 error = ofproto_dpif_add_internal_flow(ofproto, &match, 1, 0, &ofpacts,
1349 destruct(struct ofproto *ofproto_)
1351 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1352 struct ofproto_packet_in *pin, *next_pin;
1353 struct rule_dpif *rule;
1354 struct oftable *table;
1357 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1359 xlate_remove_ofproto(ofproto);
1362 /* Ensure that the upcall processing threads have no remaining references
1363 * to the ofproto or anything in it. */
1364 udpif_synchronize(ofproto->backer->udpif);
1366 hmap_remove(&all_ofproto_dpifs, &ofproto->all_ofproto_dpifs_node);
1368 OFPROTO_FOR_EACH_TABLE (table, &ofproto->up) {
1369 CLS_FOR_EACH (rule, up.cr, &table->cls) {
1370 ofproto_rule_delete(&ofproto->up, &rule->up);
1374 guarded_list_pop_all(&ofproto->pins, &pins);
1375 LIST_FOR_EACH_SAFE (pin, next_pin, list_node, &pins) {
1376 list_remove(&pin->list_node);
1377 free(CONST_CAST(void *, pin->up.packet));
1380 guarded_list_destroy(&ofproto->pins);
1382 mbridge_unref(ofproto->mbridge);
1384 netflow_unref(ofproto->netflow);
1385 dpif_sflow_unref(ofproto->sflow);
1386 dpif_ipfix_unref(ofproto->ipfix);
1387 hmap_destroy(&ofproto->bundles);
1388 mac_learning_unref(ofproto->ml);
1389 mcast_snooping_unref(ofproto->ms);
1391 hmap_destroy(&ofproto->vlandev_map);
1392 hmap_destroy(&ofproto->realdev_vid_map);
1394 sset_destroy(&ofproto->ports);
1395 sset_destroy(&ofproto->ghost_ports);
1396 sset_destroy(&ofproto->port_poll_set);
1398 ovs_mutex_destroy(&ofproto->stats_mutex);
1399 ovs_mutex_destroy(&ofproto->vsp_mutex);
1401 seq_destroy(ofproto->pins_seq);
1403 close_dpif_backer(ofproto->backer);
1407 run(struct ofproto *ofproto_)
1409 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1410 uint64_t new_seq, new_dump_seq;
1412 if (mbridge_need_revalidate(ofproto->mbridge)) {
1413 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1414 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
1415 mac_learning_flush(ofproto->ml);
1416 ovs_rwlock_unlock(&ofproto->ml->rwlock);
1417 mcast_snooping_mdb_flush(ofproto->ms);
1420 /* Always updates the ofproto->pins_seqno to avoid frequent wakeup during
1421 * flow restore. Even though nothing is processed during flow restore,
1422 * all queued 'pins' will be handled immediately when flow restore
1424 ofproto->pins_seqno = seq_read(ofproto->pins_seq);
1426 /* Do not perform any periodic activity required by 'ofproto' while
1427 * waiting for flow restore to complete. */
1428 if (!ofproto_get_flow_restore_wait()) {
1429 struct ofproto_packet_in *pin, *next_pin;
1432 guarded_list_pop_all(&ofproto->pins, &pins);
1433 LIST_FOR_EACH_SAFE (pin, next_pin, list_node, &pins) {
1434 connmgr_send_packet_in(ofproto->up.connmgr, pin);
1435 list_remove(&pin->list_node);
1436 free(CONST_CAST(void *, pin->up.packet));
1441 if (ofproto->netflow) {
1442 netflow_run(ofproto->netflow);
1444 if (ofproto->sflow) {
1445 dpif_sflow_run(ofproto->sflow);
1447 if (ofproto->ipfix) {
1448 dpif_ipfix_run(ofproto->ipfix);
1451 new_seq = seq_read(connectivity_seq_get());
1452 if (ofproto->change_seq != new_seq) {
1453 struct ofport_dpif *ofport;
1455 HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
1459 ofproto->change_seq = new_seq;
1461 if (ofproto->lacp_enabled || ofproto->has_bonded_bundles) {
1462 struct ofbundle *bundle;
1464 HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) {
1471 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
1472 if (mac_learning_run(ofproto->ml)) {
1473 ofproto->backer->need_revalidate = REV_MAC_LEARNING;
1475 ovs_rwlock_unlock(&ofproto->ml->rwlock);
1477 if (mcast_snooping_run(ofproto->ms)) {
1478 ofproto->backer->need_revalidate = REV_MCAST_SNOOPING;
1481 new_dump_seq = seq_read(udpif_dump_seq(ofproto->backer->udpif));
1482 if (ofproto->dump_seq != new_dump_seq) {
1483 struct rule *rule, *next_rule;
1485 /* We know stats are relatively fresh, so now is a good time to do some
1487 ofproto->dump_seq = new_dump_seq;
1489 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1491 ovs_mutex_lock(&ofproto_mutex);
1492 LIST_FOR_EACH_SAFE (rule, next_rule, expirable,
1493 &ofproto->up.expirable) {
1494 rule_expire(rule_dpif_cast(rule));
1496 ovs_mutex_unlock(&ofproto_mutex);
1498 /* All outstanding data in existing flows has been accounted, so it's a
1499 * good time to do bond rebalancing. */
1500 if (ofproto->has_bonded_bundles) {
1501 struct ofbundle *bundle;
1503 HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) {
1505 bond_rebalance(bundle->bond);
1514 wait(struct ofproto *ofproto_)
1516 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1518 if (ofproto_get_flow_restore_wait()) {
1522 if (ofproto->sflow) {
1523 dpif_sflow_wait(ofproto->sflow);
1525 if (ofproto->ipfix) {
1526 dpif_ipfix_wait(ofproto->ipfix);
1528 if (ofproto->lacp_enabled || ofproto->has_bonded_bundles) {
1529 struct ofbundle *bundle;
1531 HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) {
1532 bundle_wait(bundle);
1535 if (ofproto->netflow) {
1536 netflow_wait(ofproto->netflow);
1538 ovs_rwlock_rdlock(&ofproto->ml->rwlock);
1539 mac_learning_wait(ofproto->ml);
1540 ovs_rwlock_unlock(&ofproto->ml->rwlock);
1541 mcast_snooping_wait(ofproto->ms);
1543 if (ofproto->backer->need_revalidate) {
1544 /* Shouldn't happen, but if it does just go around again. */
1545 VLOG_DBG_RL(&rl, "need revalidate in ofproto_wait_cb()");
1546 poll_immediate_wake();
1549 seq_wait(udpif_dump_seq(ofproto->backer->udpif), ofproto->dump_seq);
1550 seq_wait(ofproto->pins_seq, ofproto->pins_seqno);
1554 type_get_memory_usage(const char *type, struct simap *usage)
1556 struct dpif_backer *backer;
1558 backer = shash_find_data(&all_dpif_backers, type);
1560 udpif_get_memory_usage(backer->udpif, usage);
1565 flush(struct ofproto *ofproto_)
1567 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1568 struct dpif_backer *backer = ofproto->backer;
1571 udpif_flush(backer->udpif);
1576 query_tables(struct ofproto *ofproto,
1577 struct ofputil_table_features *features,
1578 struct ofputil_table_stats *stats)
1580 strcpy(features->name, "classifier");
1585 for (i = 0; i < ofproto->n_tables; i++) {
1586 unsigned long missed, matched;
1588 atomic_read_relaxed(&ofproto->tables[i].n_matched, &matched);
1589 atomic_read_relaxed(&ofproto->tables[i].n_missed, &missed);
1591 stats[i].matched_count = matched;
1592 stats[i].lookup_count = matched + missed;
1597 static struct ofport *
1600 struct ofport_dpif *port = xmalloc(sizeof *port);
1605 port_dealloc(struct ofport *port_)
1607 struct ofport_dpif *port = ofport_dpif_cast(port_);
1612 port_construct(struct ofport *port_)
1614 struct ofport_dpif *port = ofport_dpif_cast(port_);
1615 struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
1616 const struct netdev *netdev = port->up.netdev;
1617 char namebuf[NETDEV_VPORT_NAME_BUFSIZE];
1618 struct dpif_port dpif_port;
1621 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1622 port->bundle = NULL;
1625 port->may_enable = false;
1626 port->stp_port = NULL;
1627 port->stp_state = STP_DISABLED;
1628 port->rstp_port = NULL;
1629 port->rstp_state = RSTP_DISABLED;
1630 port->is_tunnel = false;
1634 port->realdev_ofp_port = 0;
1635 port->vlandev_vid = 0;
1636 port->carrier_seq = netdev_get_carrier_resets(netdev);
1637 port->is_layer3 = netdev_vport_is_layer3(netdev);
1639 if (netdev_vport_is_patch(netdev)) {
1640 /* By bailing out here, we don't submit the port to the sFlow module
1641 * to be considered for counter polling export. This is correct
1642 * because the patch port represents an interface that sFlow considers
1643 * to be "internal" to the switch as a whole, and therefore not an
1644 * candidate for counter polling. */
1645 port->odp_port = ODPP_NONE;
1646 ofport_update_peer(port);
1650 error = dpif_port_query_by_name(ofproto->backer->dpif,
1651 netdev_vport_get_dpif_port(netdev, namebuf,
1658 port->odp_port = dpif_port.port_no;
1660 if (netdev_get_tunnel_config(netdev)) {
1661 atomic_count_inc(&ofproto->backer->tnl_count);
1662 tnl_port_add(port, port->up.netdev, port->odp_port,
1663 ovs_native_tunneling_is_on(ofproto), namebuf);
1664 port->is_tunnel = true;
1665 if (ofproto->ipfix) {
1666 dpif_ipfix_add_tunnel_port(ofproto->ipfix, port_, port->odp_port);
1669 /* Sanity-check that a mapping doesn't already exist. This
1670 * shouldn't happen for non-tunnel ports. */
1671 if (odp_port_to_ofp_port(ofproto, port->odp_port) != OFPP_NONE) {
1672 VLOG_ERR("port %s already has an OpenFlow port number",
1674 dpif_port_destroy(&dpif_port);
1678 ovs_rwlock_wrlock(&ofproto->backer->odp_to_ofport_lock);
1679 hmap_insert(&ofproto->backer->odp_to_ofport_map, &port->odp_port_node,
1680 hash_odp_port(port->odp_port));
1681 ovs_rwlock_unlock(&ofproto->backer->odp_to_ofport_lock);
1683 dpif_port_destroy(&dpif_port);
1685 if (ofproto->sflow) {
1686 dpif_sflow_add_port(ofproto->sflow, port_, port->odp_port);
1693 port_destruct(struct ofport *port_)
1695 struct ofport_dpif *port = ofport_dpif_cast(port_);
1696 struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
1697 const char *devname = netdev_get_name(port->up.netdev);
1698 char namebuf[NETDEV_VPORT_NAME_BUFSIZE];
1699 const char *dp_port_name;
1701 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1703 xlate_ofport_remove(port);
1706 dp_port_name = netdev_vport_get_dpif_port(port->up.netdev, namebuf,
1708 if (dpif_port_exists(ofproto->backer->dpif, dp_port_name)) {
1709 /* The underlying device is still there, so delete it. This
1710 * happens when the ofproto is being destroyed, since the caller
1711 * assumes that removal of attached ports will happen as part of
1713 if (!port->is_tunnel) {
1714 dpif_port_del(ofproto->backer->dpif, port->odp_port);
1719 port->peer->peer = NULL;
1723 if (port->odp_port != ODPP_NONE && !port->is_tunnel) {
1724 ovs_rwlock_wrlock(&ofproto->backer->odp_to_ofport_lock);
1725 hmap_remove(&ofproto->backer->odp_to_ofport_map, &port->odp_port_node);
1726 ovs_rwlock_unlock(&ofproto->backer->odp_to_ofport_lock);
1729 if (port->is_tunnel) {
1730 atomic_count_dec(&ofproto->backer->tnl_count);
1733 if (port->is_tunnel && ofproto->ipfix) {
1734 dpif_ipfix_del_tunnel_port(ofproto->ipfix, port->odp_port);
1738 sset_find_and_delete(&ofproto->ports, devname);
1739 sset_find_and_delete(&ofproto->ghost_ports, devname);
1740 bundle_remove(port_);
1741 set_cfm(port_, NULL);
1742 set_bfd(port_, NULL);
1743 if (port->stp_port) {
1744 stp_port_disable(port->stp_port);
1746 set_rstp_port(port_, NULL);
1747 if (ofproto->sflow) {
1748 dpif_sflow_del_port(ofproto->sflow, port->odp_port);
1755 port_modified(struct ofport *port_)
1757 struct ofport_dpif *port = ofport_dpif_cast(port_);
1758 char namebuf[NETDEV_VPORT_NAME_BUFSIZE];
1759 struct netdev *netdev = port->up.netdev;
1761 if (port->bundle && port->bundle->bond) {
1762 bond_slave_set_netdev(port->bundle->bond, port, netdev);
1766 cfm_set_netdev(port->cfm, netdev);
1770 bfd_set_netdev(port->bfd, netdev);
1773 ofproto_dpif_monitor_port_update(port, port->bfd, port->cfm,
1774 port->up.pp.hw_addr);
1776 netdev_vport_get_dpif_port(netdev, namebuf, sizeof namebuf);
1778 if (port->is_tunnel) {
1779 struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
1781 if (tnl_port_reconfigure(port, netdev, port->odp_port,
1782 ovs_native_tunneling_is_on(ofproto), namebuf)) {
1783 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1787 ofport_update_peer(port);
1791 port_reconfigured(struct ofport *port_, enum ofputil_port_config old_config)
1793 struct ofport_dpif *port = ofport_dpif_cast(port_);
1794 struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
1795 enum ofputil_port_config changed = old_config ^ port->up.pp.config;
1797 if (changed & (OFPUTIL_PC_NO_RECV | OFPUTIL_PC_NO_RECV_STP |
1798 OFPUTIL_PC_NO_FWD | OFPUTIL_PC_NO_FLOOD |
1799 OFPUTIL_PC_NO_PACKET_IN)) {
1800 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1802 if (changed & OFPUTIL_PC_NO_FLOOD && port->bundle) {
1803 bundle_update(port->bundle);
1809 set_sflow(struct ofproto *ofproto_,
1810 const struct ofproto_sflow_options *sflow_options)
1812 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1813 struct dpif_sflow *ds = ofproto->sflow;
1815 if (sflow_options) {
1817 struct ofport_dpif *ofport;
1819 ds = ofproto->sflow = dpif_sflow_create();
1820 HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
1821 dpif_sflow_add_port(ds, &ofport->up, ofport->odp_port);
1823 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1825 dpif_sflow_set_options(ds, sflow_options);
1828 dpif_sflow_unref(ds);
1829 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1830 ofproto->sflow = NULL;
1838 struct ofproto *ofproto_,
1839 const struct ofproto_ipfix_bridge_exporter_options *bridge_exporter_options,
1840 const struct ofproto_ipfix_flow_exporter_options *flow_exporters_options,
1841 size_t n_flow_exporters_options)
1843 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
1844 struct dpif_ipfix *di = ofproto->ipfix;
1845 bool has_options = bridge_exporter_options || flow_exporters_options;
1846 bool new_di = false;
1848 if (has_options && !di) {
1849 di = ofproto->ipfix = dpif_ipfix_create();
1854 /* Call set_options in any case to cleanly flush the flow
1855 * caches in the last exporters that are to be destroyed. */
1856 dpif_ipfix_set_options(
1857 di, bridge_exporter_options, flow_exporters_options,
1858 n_flow_exporters_options);
1860 /* Add tunnel ports only when a new ipfix created */
1861 if (new_di == true) {
1862 struct ofport_dpif *ofport;
1863 HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
1864 if (ofport->is_tunnel == true) {
1865 dpif_ipfix_add_tunnel_port(di, &ofport->up, ofport->odp_port);
1871 dpif_ipfix_unref(di);
1872 ofproto->ipfix = NULL;
1880 set_cfm(struct ofport *ofport_, const struct cfm_settings *s)
1882 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
1887 struct ofproto_dpif *ofproto;
1889 ofproto = ofproto_dpif_cast(ofport->up.ofproto);
1890 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1891 ofport->cfm = cfm_create(ofport->up.netdev);
1894 if (cfm_configure(ofport->cfm, s)) {
1901 cfm_unref(ofport->cfm);
1904 ofproto_dpif_monitor_port_update(ofport, ofport->bfd, ofport->cfm,
1905 ofport->up.pp.hw_addr);
1910 cfm_status_changed(struct ofport *ofport_)
1912 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
1914 return ofport->cfm ? cfm_check_status_change(ofport->cfm) : true;
1918 get_cfm_status(const struct ofport *ofport_,
1919 struct cfm_status *status)
1921 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
1925 cfm_get_status(ofport->cfm, status);
1934 set_bfd(struct ofport *ofport_, const struct smap *cfg)
1936 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport_->ofproto);
1937 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
1941 ofport->bfd = bfd_configure(old, netdev_get_name(ofport->up.netdev),
1942 cfg, ofport->up.netdev);
1943 if (ofport->bfd != old) {
1944 ofproto->backer->need_revalidate = REV_RECONFIGURE;
1946 ofproto_dpif_monitor_port_update(ofport, ofport->bfd, ofport->cfm,
1947 ofport->up.pp.hw_addr);
1952 bfd_status_changed(struct ofport *ofport_)
1954 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
1956 return ofport->bfd ? bfd_check_status_change(ofport->bfd) : true;
1960 get_bfd_status(struct ofport *ofport_, struct smap *smap)
1962 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
1966 bfd_get_status(ofport->bfd, smap);
1974 /* Spanning Tree. */
1976 /* Called while rstp_mutex is held. */
1978 rstp_send_bpdu_cb(struct ofpbuf *pkt, void *ofport_, void *ofproto_)
1980 struct ofproto_dpif *ofproto = ofproto_;
1981 struct ofport_dpif *ofport = ofport_;
1982 struct eth_header *eth = ofpbuf_l2(pkt);
1984 netdev_get_etheraddr(ofport->up.netdev, eth->eth_src);
1985 if (eth_addr_is_zero(eth->eth_src)) {
1986 VLOG_WARN_RL(&rl, "%s port %d: cannot send RSTP BPDU on a port which "
1987 "does not have a configured source MAC address.",
1988 ofproto->up.name, ofp_to_u16(ofport->up.ofp_port));
1990 ofproto_dpif_send_packet(ofport, pkt);
1996 send_bpdu_cb(struct ofpbuf *pkt, int port_num, void *ofproto_)
1998 struct ofproto_dpif *ofproto = ofproto_;
1999 struct stp_port *sp = stp_get_port(ofproto->stp, port_num);
2000 struct ofport_dpif *ofport;
2002 ofport = stp_port_get_aux(sp);
2004 VLOG_WARN_RL(&rl, "%s: cannot send BPDU on unknown port %d",
2005 ofproto->up.name, port_num);
2007 struct eth_header *eth = ofpbuf_l2(pkt);
2009 netdev_get_etheraddr(ofport->up.netdev, eth->eth_src);
2010 if (eth_addr_is_zero(eth->eth_src)) {
2011 VLOG_WARN_RL(&rl, "%s: cannot send BPDU on port %d "
2012 "with unknown MAC", ofproto->up.name, port_num);
2014 ofproto_dpif_send_packet(ofport, pkt);
2020 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
2022 set_rstp(struct ofproto *ofproto_, const struct ofproto_rstp_settings *s)
2024 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2026 /* Only revalidate flows if the configuration changed. */
2027 if (!s != !ofproto->rstp) {
2028 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2032 if (!ofproto->rstp) {
2033 ofproto->rstp = rstp_create(ofproto_->name, s->address,
2034 rstp_send_bpdu_cb, ofproto);
2035 ofproto->rstp_last_tick = time_msec();
2037 rstp_set_bridge_address(ofproto->rstp, s->address);
2038 rstp_set_bridge_priority(ofproto->rstp, s->priority);
2039 rstp_set_bridge_ageing_time(ofproto->rstp, s->ageing_time);
2040 rstp_set_bridge_force_protocol_version(ofproto->rstp,
2041 s->force_protocol_version);
2042 rstp_set_bridge_max_age(ofproto->rstp, s->bridge_max_age);
2043 rstp_set_bridge_forward_delay(ofproto->rstp, s->bridge_forward_delay);
2044 rstp_set_bridge_transmit_hold_count(ofproto->rstp,
2045 s->transmit_hold_count);
2047 struct ofport *ofport;
2048 HMAP_FOR_EACH (ofport, hmap_node, &ofproto->up.ports) {
2049 set_rstp_port(ofport, NULL);
2051 rstp_unref(ofproto->rstp);
2052 ofproto->rstp = NULL;
2057 get_rstp_status(struct ofproto *ofproto_, struct ofproto_rstp_status *s)
2059 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2061 if (ofproto->rstp) {
2063 s->root_id = rstp_get_root_id(ofproto->rstp);
2064 s->bridge_id = rstp_get_bridge_id(ofproto->rstp);
2065 s->designated_id = rstp_get_designated_id(ofproto->rstp);
2066 s->root_path_cost = rstp_get_root_path_cost(ofproto->rstp);
2067 s->designated_port_id = rstp_get_designated_port_id(ofproto->rstp);
2068 s->bridge_port_id = rstp_get_bridge_port_id(ofproto->rstp);
2075 update_rstp_port_state(struct ofport_dpif *ofport)
2077 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2078 enum rstp_state state;
2080 /* Figure out new state. */
2081 state = ofport->rstp_port ? rstp_port_get_state(ofport->rstp_port)
2085 if (ofport->rstp_state != state) {
2086 enum ofputil_port_state of_state;
2089 VLOG_DBG("port %s: RSTP state changed from %s to %s",
2090 netdev_get_name(ofport->up.netdev),
2091 rstp_state_name(ofport->rstp_state),
2092 rstp_state_name(state));
2094 if (rstp_learn_in_state(ofport->rstp_state)
2095 != rstp_learn_in_state(state)) {
2096 /* XXX: Learning action flows should also be flushed. */
2097 if (ofport->bundle) {
2098 if (!rstp_shift_root_learned_address(ofproto->rstp)
2099 || rstp_get_old_root_aux(ofproto->rstp) != ofport) {
2100 bundle_flush_macs(ofport->bundle, false);
2104 fwd_change = rstp_forward_in_state(ofport->rstp_state)
2105 != rstp_forward_in_state(state);
2107 ofproto->backer->need_revalidate = REV_RSTP;
2108 ofport->rstp_state = state;
2110 if (fwd_change && ofport->bundle) {
2111 bundle_update(ofport->bundle);
2114 /* Update the RSTP state bits in the OpenFlow port description. */
2115 of_state = ofport->up.pp.state & ~OFPUTIL_PS_STP_MASK;
2116 of_state |= (state == RSTP_LEARNING ? OFPUTIL_PS_STP_LEARN
2117 : state == RSTP_FORWARDING ? OFPUTIL_PS_STP_FORWARD
2118 : state == RSTP_DISCARDING ? OFPUTIL_PS_STP_LISTEN
2120 ofproto_port_set_state(&ofport->up, of_state);
2125 rstp_run(struct ofproto_dpif *ofproto)
2127 if (ofproto->rstp) {
2128 long long int now = time_msec();
2129 long long int elapsed = now - ofproto->rstp_last_tick;
2130 struct rstp_port *rp;
2131 struct ofport_dpif *ofport;
2133 /* Every second, decrease the values of the timers. */
2134 if (elapsed >= 1000) {
2135 rstp_tick_timers(ofproto->rstp);
2136 ofproto->rstp_last_tick = now;
2139 while ((ofport = rstp_get_next_changed_port_aux(ofproto->rstp, &rp))) {
2140 update_rstp_port_state(ofport);
2144 /* FIXME: This check should be done on-event (i.e., when setting
2145 * p->fdb_flush) and not periodically.
2147 while ((ofport = rstp_check_and_reset_fdb_flush(ofproto->rstp, &rp))) {
2148 if (!rstp_shift_root_learned_address(ofproto->rstp)
2149 || rstp_get_old_root_aux(ofproto->rstp) != ofport) {
2150 bundle_flush_macs(ofport->bundle, false);
2154 if (rstp_shift_root_learned_address(ofproto->rstp)) {
2155 bundle_move(((struct ofport_dpif *)rstp_get_old_root_aux(ofproto->rstp))->bundle,
2156 ((struct ofport_dpif *)rstp_get_new_root_aux(ofproto->rstp))->bundle);
2157 rstp_reset_root_changed(ofproto->rstp);
2162 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2164 set_stp(struct ofproto *ofproto_, const struct ofproto_stp_settings *s)
2166 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2168 /* Only revalidate flows if the configuration changed. */
2169 if (!s != !ofproto->stp) {
2170 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2174 if (!ofproto->stp) {
2175 ofproto->stp = stp_create(ofproto_->name, s->system_id,
2176 send_bpdu_cb, ofproto);
2177 ofproto->stp_last_tick = time_msec();
2180 stp_set_bridge_id(ofproto->stp, s->system_id);
2181 stp_set_bridge_priority(ofproto->stp, s->priority);
2182 stp_set_hello_time(ofproto->stp, s->hello_time);
2183 stp_set_max_age(ofproto->stp, s->max_age);
2184 stp_set_forward_delay(ofproto->stp, s->fwd_delay);
2186 struct ofport *ofport;
2188 HMAP_FOR_EACH (ofport, hmap_node, &ofproto->up.ports) {
2189 set_stp_port(ofport, NULL);
2192 stp_unref(ofproto->stp);
2193 ofproto->stp = NULL;
2200 get_stp_status(struct ofproto *ofproto_, struct ofproto_stp_status *s)
2202 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2206 s->bridge_id = stp_get_bridge_id(ofproto->stp);
2207 s->designated_root = stp_get_designated_root(ofproto->stp);
2208 s->root_path_cost = stp_get_root_path_cost(ofproto->stp);
2217 update_stp_port_state(struct ofport_dpif *ofport)
2219 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2220 enum stp_state state;
2222 /* Figure out new state. */
2223 state = ofport->stp_port ? stp_port_get_state(ofport->stp_port)
2227 if (ofport->stp_state != state) {
2228 enum ofputil_port_state of_state;
2231 VLOG_DBG("port %s: STP state changed from %s to %s",
2232 netdev_get_name(ofport->up.netdev),
2233 stp_state_name(ofport->stp_state),
2234 stp_state_name(state));
2235 if (stp_learn_in_state(ofport->stp_state)
2236 != stp_learn_in_state(state)) {
2237 /* xxx Learning action flows should also be flushed. */
2238 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
2239 mac_learning_flush(ofproto->ml);
2240 ovs_rwlock_unlock(&ofproto->ml->rwlock);
2241 mcast_snooping_mdb_flush(ofproto->ms);
2243 fwd_change = stp_forward_in_state(ofport->stp_state)
2244 != stp_forward_in_state(state);
2246 ofproto->backer->need_revalidate = REV_STP;
2247 ofport->stp_state = state;
2248 ofport->stp_state_entered = time_msec();
2250 if (fwd_change && ofport->bundle) {
2251 bundle_update(ofport->bundle);
2254 /* Update the STP state bits in the OpenFlow port description. */
2255 of_state = ofport->up.pp.state & ~OFPUTIL_PS_STP_MASK;
2256 of_state |= (state == STP_LISTENING ? OFPUTIL_PS_STP_LISTEN
2257 : state == STP_LEARNING ? OFPUTIL_PS_STP_LEARN
2258 : state == STP_FORWARDING ? OFPUTIL_PS_STP_FORWARD
2259 : state == STP_BLOCKING ? OFPUTIL_PS_STP_BLOCK
2261 ofproto_port_set_state(&ofport->up, of_state);
2265 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2266 * caller is responsible for assigning STP port numbers and ensuring
2267 * there are no duplicates. */
2269 set_stp_port(struct ofport *ofport_,
2270 const struct ofproto_port_stp_settings *s)
2272 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
2273 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2274 struct stp_port *sp = ofport->stp_port;
2276 if (!s || !s->enable) {
2278 ofport->stp_port = NULL;
2279 stp_port_disable(sp);
2280 update_stp_port_state(ofport);
2283 } else if (sp && stp_port_no(sp) != s->port_num
2284 && ofport == stp_port_get_aux(sp)) {
2285 /* The port-id changed, so disable the old one if it's not
2286 * already in use by another port. */
2287 stp_port_disable(sp);
2290 sp = ofport->stp_port = stp_get_port(ofproto->stp, s->port_num);
2292 /* Set name before enabling the port so that debugging messages can print
2294 stp_port_set_name(sp, netdev_get_name(ofport->up.netdev));
2295 stp_port_enable(sp);
2297 stp_port_set_aux(sp, ofport);
2298 stp_port_set_priority(sp, s->priority);
2299 stp_port_set_path_cost(sp, s->path_cost);
2301 update_stp_port_state(ofport);
2307 get_stp_port_status(struct ofport *ofport_,
2308 struct ofproto_port_stp_status *s)
2310 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
2311 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2312 struct stp_port *sp = ofport->stp_port;
2314 if (!ofproto->stp || !sp) {
2320 s->port_id = stp_port_get_id(sp);
2321 s->state = stp_port_get_state(sp);
2322 s->sec_in_state = (time_msec() - ofport->stp_state_entered) / 1000;
2323 s->role = stp_port_get_role(sp);
2329 get_stp_port_stats(struct ofport *ofport_,
2330 struct ofproto_port_stp_stats *s)
2332 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
2333 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2334 struct stp_port *sp = ofport->stp_port;
2336 if (!ofproto->stp || !sp) {
2342 stp_port_get_counts(sp, &s->tx_count, &s->rx_count, &s->error_count);
2348 stp_run(struct ofproto_dpif *ofproto)
2351 long long int now = time_msec();
2352 long long int elapsed = now - ofproto->stp_last_tick;
2353 struct stp_port *sp;
2356 stp_tick(ofproto->stp, MIN(INT_MAX, elapsed));
2357 ofproto->stp_last_tick = now;
2359 while (stp_get_changed_port(ofproto->stp, &sp)) {
2360 struct ofport_dpif *ofport = stp_port_get_aux(sp);
2363 update_stp_port_state(ofport);
2367 if (stp_check_and_reset_fdb_flush(ofproto->stp)) {
2368 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
2369 mac_learning_flush(ofproto->ml);
2370 ovs_rwlock_unlock(&ofproto->ml->rwlock);
2371 mcast_snooping_mdb_flush(ofproto->ms);
2377 stp_wait(struct ofproto_dpif *ofproto)
2380 poll_timer_wait(1000);
2384 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2385 * caller is responsible for assigning RSTP port numbers and ensuring
2386 * there are no duplicates. */
2388 set_rstp_port(struct ofport *ofport_,
2389 const struct ofproto_port_rstp_settings *s)
2391 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
2392 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2393 struct rstp_port *rp = ofport->rstp_port;
2395 if (!s || !s->enable) {
2397 rstp_port_unref(rp);
2398 ofport->rstp_port = NULL;
2399 update_rstp_port_state(ofport);
2404 /* Check if need to add a new port. */
2406 rp = ofport->rstp_port = rstp_add_port(ofproto->rstp);
2409 rstp_port_set(rp, s->port_num, s->priority, s->path_cost,
2410 s->admin_edge_port, s->auto_edge,
2411 s->admin_p2p_mac_state, s->admin_port_state, s->mcheck,
2413 update_rstp_port_state(ofport);
2414 /* Synchronize operational status. */
2415 rstp_port_set_mac_operational(rp, ofport->may_enable);
2419 get_rstp_port_status(struct ofport *ofport_,
2420 struct ofproto_port_rstp_status *s)
2422 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
2423 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2424 struct rstp_port *rp = ofport->rstp_port;
2426 if (!ofproto->rstp || !rp) {
2432 rstp_port_get_status(rp, &s->port_id, &s->state, &s->role,
2433 &s->designated_bridge_id, &s->designated_port_id,
2434 &s->designated_path_cost, &s->tx_count,
2435 &s->rx_count, &s->error_count, &s->uptime);
2440 set_queues(struct ofport *ofport_, const struct ofproto_port_queue *qdscp,
2443 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
2444 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
2446 if (ofport->n_qdscp != n_qdscp
2447 || (n_qdscp && memcmp(ofport->qdscp, qdscp,
2448 n_qdscp * sizeof *qdscp))) {
2449 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2450 free(ofport->qdscp);
2451 ofport->qdscp = n_qdscp
2452 ? xmemdup(qdscp, n_qdscp * sizeof *qdscp)
2454 ofport->n_qdscp = n_qdscp;
2462 /* Expires all MAC learning entries associated with 'bundle' and forces its
2463 * ofproto to revalidate every flow.
2465 * Normally MAC learning entries are removed only from the ofproto associated
2466 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2467 * are removed from every ofproto. When patch ports and SLB bonds are in use
2468 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2469 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2470 * with the host from which it migrated. */
2472 bundle_flush_macs(struct ofbundle *bundle, bool all_ofprotos)
2474 struct ofproto_dpif *ofproto = bundle->ofproto;
2475 struct mac_learning *ml = ofproto->ml;
2476 struct mac_entry *mac, *next_mac;
2478 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2479 ovs_rwlock_wrlock(&ml->rwlock);
2480 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2481 if (mac->port.p == bundle) {
2483 struct ofproto_dpif *o;
2485 HMAP_FOR_EACH (o, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
2487 struct mac_entry *e;
2489 ovs_rwlock_wrlock(&o->ml->rwlock);
2490 e = mac_learning_lookup(o->ml, mac->mac, mac->vlan);
2492 mac_learning_expire(o->ml, e);
2494 ovs_rwlock_unlock(&o->ml->rwlock);
2499 mac_learning_expire(ml, mac);
2502 ovs_rwlock_unlock(&ml->rwlock);
2506 bundle_move(struct ofbundle *old, struct ofbundle *new)
2508 struct ofproto_dpif *ofproto = old->ofproto;
2509 struct mac_learning *ml = ofproto->ml;
2510 struct mac_entry *mac, *next_mac;
2512 ovs_assert(new->ofproto == old->ofproto);
2514 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2515 ovs_rwlock_wrlock(&ml->rwlock);
2516 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2517 if (mac->port.p == old) {
2521 ovs_rwlock_unlock(&ml->rwlock);
2524 static struct ofbundle *
2525 bundle_lookup(const struct ofproto_dpif *ofproto, void *aux)
2527 struct ofbundle *bundle;
2529 HMAP_FOR_EACH_IN_BUCKET (bundle, hmap_node, hash_pointer(aux, 0),
2530 &ofproto->bundles) {
2531 if (bundle->aux == aux) {
2539 bundle_update(struct ofbundle *bundle)
2541 struct ofport_dpif *port;
2543 bundle->floodable = true;
2544 LIST_FOR_EACH (port, bundle_node, &bundle->ports) {
2545 if (port->up.pp.config & OFPUTIL_PC_NO_FLOOD
2547 || (bundle->ofproto->stp && !stp_forward_in_state(port->stp_state))
2548 || (bundle->ofproto->rstp && !rstp_forward_in_state(port->rstp_state))) {
2549 bundle->floodable = false;
2556 bundle_del_port(struct ofport_dpif *port)
2558 struct ofbundle *bundle = port->bundle;
2560 bundle->ofproto->backer->need_revalidate = REV_RECONFIGURE;
2562 list_remove(&port->bundle_node);
2563 port->bundle = NULL;
2566 lacp_slave_unregister(bundle->lacp, port);
2569 bond_slave_unregister(bundle->bond, port);
2572 bundle_update(bundle);
2576 bundle_add_port(struct ofbundle *bundle, ofp_port_t ofp_port,
2577 struct lacp_slave_settings *lacp)
2579 struct ofport_dpif *port;
2581 port = get_ofp_port(bundle->ofproto, ofp_port);
2586 if (port->bundle != bundle) {
2587 bundle->ofproto->backer->need_revalidate = REV_RECONFIGURE;
2589 bundle_remove(&port->up);
2592 port->bundle = bundle;
2593 list_push_back(&bundle->ports, &port->bundle_node);
2594 if (port->up.pp.config & OFPUTIL_PC_NO_FLOOD
2596 || (bundle->ofproto->stp && !stp_forward_in_state(port->stp_state))
2597 || (bundle->ofproto->rstp && !rstp_forward_in_state(port->rstp_state))) {
2598 bundle->floodable = false;
2602 bundle->ofproto->backer->need_revalidate = REV_RECONFIGURE;
2603 lacp_slave_register(bundle->lacp, port, lacp);
2610 bundle_destroy(struct ofbundle *bundle)
2612 struct ofproto_dpif *ofproto;
2613 struct ofport_dpif *port, *next_port;
2619 ofproto = bundle->ofproto;
2620 mbridge_unregister_bundle(ofproto->mbridge, bundle->aux);
2623 xlate_bundle_remove(bundle);
2626 LIST_FOR_EACH_SAFE (port, next_port, bundle_node, &bundle->ports) {
2627 bundle_del_port(port);
2630 bundle_flush_macs(bundle, true);
2631 hmap_remove(&ofproto->bundles, &bundle->hmap_node);
2633 free(bundle->trunks);
2634 lacp_unref(bundle->lacp);
2635 bond_unref(bundle->bond);
2640 bundle_set(struct ofproto *ofproto_, void *aux,
2641 const struct ofproto_bundle_settings *s)
2643 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2644 bool need_flush = false;
2645 struct ofport_dpif *port;
2646 struct ofbundle *bundle;
2647 unsigned long *trunks;
2653 bundle_destroy(bundle_lookup(ofproto, aux));
2657 ovs_assert(s->n_slaves == 1 || s->bond != NULL);
2658 ovs_assert((s->lacp != NULL) == (s->lacp_slaves != NULL));
2660 bundle = bundle_lookup(ofproto, aux);
2662 bundle = xmalloc(sizeof *bundle);
2664 bundle->ofproto = ofproto;
2665 hmap_insert(&ofproto->bundles, &bundle->hmap_node,
2666 hash_pointer(aux, 0));
2668 bundle->name = NULL;
2670 list_init(&bundle->ports);
2671 bundle->vlan_mode = PORT_VLAN_TRUNK;
2673 bundle->trunks = NULL;
2674 bundle->use_priority_tags = s->use_priority_tags;
2675 bundle->lacp = NULL;
2676 bundle->bond = NULL;
2678 bundle->floodable = true;
2679 mbridge_register_bundle(ofproto->mbridge, bundle);
2682 if (!bundle->name || strcmp(s->name, bundle->name)) {
2684 bundle->name = xstrdup(s->name);
2689 ofproto->lacp_enabled = true;
2690 if (!bundle->lacp) {
2691 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2692 bundle->lacp = lacp_create();
2694 lacp_configure(bundle->lacp, s->lacp);
2696 lacp_unref(bundle->lacp);
2697 bundle->lacp = NULL;
2700 /* Update set of ports. */
2702 for (i = 0; i < s->n_slaves; i++) {
2703 if (!bundle_add_port(bundle, s->slaves[i],
2704 s->lacp ? &s->lacp_slaves[i] : NULL)) {
2708 if (!ok || list_size(&bundle->ports) != s->n_slaves) {
2709 struct ofport_dpif *next_port;
2711 LIST_FOR_EACH_SAFE (port, next_port, bundle_node, &bundle->ports) {
2712 for (i = 0; i < s->n_slaves; i++) {
2713 if (s->slaves[i] == port->up.ofp_port) {
2718 bundle_del_port(port);
2722 ovs_assert(list_size(&bundle->ports) <= s->n_slaves);
2724 if (list_is_empty(&bundle->ports)) {
2725 bundle_destroy(bundle);
2729 /* Set VLAN tagging mode */
2730 if (s->vlan_mode != bundle->vlan_mode
2731 || s->use_priority_tags != bundle->use_priority_tags) {
2732 bundle->vlan_mode = s->vlan_mode;
2733 bundle->use_priority_tags = s->use_priority_tags;
2738 vlan = (s->vlan_mode == PORT_VLAN_TRUNK ? -1
2739 : s->vlan >= 0 && s->vlan <= 4095 ? s->vlan
2741 if (vlan != bundle->vlan) {
2742 bundle->vlan = vlan;
2746 /* Get trunked VLANs. */
2747 switch (s->vlan_mode) {
2748 case PORT_VLAN_ACCESS:
2752 case PORT_VLAN_TRUNK:
2753 trunks = CONST_CAST(unsigned long *, s->trunks);
2756 case PORT_VLAN_NATIVE_UNTAGGED:
2757 case PORT_VLAN_NATIVE_TAGGED:
2758 if (vlan != 0 && (!s->trunks
2759 || !bitmap_is_set(s->trunks, vlan)
2760 || bitmap_is_set(s->trunks, 0))) {
2761 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
2763 trunks = bitmap_clone(s->trunks, 4096);
2765 trunks = bitmap_allocate1(4096);
2767 bitmap_set1(trunks, vlan);
2768 bitmap_set0(trunks, 0);
2770 trunks = CONST_CAST(unsigned long *, s->trunks);
2777 if (!vlan_bitmap_equal(trunks, bundle->trunks)) {
2778 free(bundle->trunks);
2779 if (trunks == s->trunks) {
2780 bundle->trunks = vlan_bitmap_clone(trunks);
2782 bundle->trunks = trunks;
2787 if (trunks != s->trunks) {
2792 if (!list_is_short(&bundle->ports)) {
2793 bundle->ofproto->has_bonded_bundles = true;
2795 if (bond_reconfigure(bundle->bond, s->bond)) {
2796 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2799 bundle->bond = bond_create(s->bond, ofproto);
2800 ofproto->backer->need_revalidate = REV_RECONFIGURE;
2803 LIST_FOR_EACH (port, bundle_node, &bundle->ports) {
2804 bond_slave_register(bundle->bond, port,
2805 port->up.ofp_port, port->up.netdev);
2808 bond_unref(bundle->bond);
2809 bundle->bond = NULL;
2812 /* If we changed something that would affect MAC learning, un-learn
2813 * everything on this port and force flow revalidation. */
2815 bundle_flush_macs(bundle, false);
2822 bundle_remove(struct ofport *port_)
2824 struct ofport_dpif *port = ofport_dpif_cast(port_);
2825 struct ofbundle *bundle = port->bundle;
2828 bundle_del_port(port);
2829 if (list_is_empty(&bundle->ports)) {
2830 bundle_destroy(bundle);
2831 } else if (list_is_short(&bundle->ports)) {
2832 bond_unref(bundle->bond);
2833 bundle->bond = NULL;
2839 send_pdu_cb(void *port_, const void *pdu, size_t pdu_size)
2841 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2842 struct ofport_dpif *port = port_;
2843 uint8_t ea[ETH_ADDR_LEN];
2846 error = netdev_get_etheraddr(port->up.netdev, ea);
2848 struct ofpbuf packet;
2851 ofpbuf_init(&packet, 0);
2852 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2854 memcpy(packet_pdu, pdu, pdu_size);
2856 ofproto_dpif_send_packet(port, &packet);
2857 ofpbuf_uninit(&packet);
2859 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2860 "%s (%s)", port->bundle->name,
2861 netdev_get_name(port->up.netdev), ovs_strerror(error));
2866 bundle_send_learning_packets(struct ofbundle *bundle)
2868 struct ofproto_dpif *ofproto = bundle->ofproto;
2869 struct ofpbuf *learning_packet;
2870 int error, n_packets, n_errors;
2871 struct mac_entry *e;
2872 struct list packets;
2874 list_init(&packets);
2875 ovs_rwlock_rdlock(&ofproto->ml->rwlock);
2876 LIST_FOR_EACH (e, lru_node, &ofproto->ml->lrus) {
2877 if (e->port.p != bundle) {
2880 learning_packet = bond_compose_learning_packet(bundle->bond,
2883 /* Temporarily use 'frame' as a private pointer (see below). */
2884 ovs_assert(learning_packet->frame == ofpbuf_data(learning_packet));
2885 learning_packet->frame = port_void;
2886 list_push_back(&packets, &learning_packet->list_node);
2889 ovs_rwlock_unlock(&ofproto->ml->rwlock);
2891 error = n_packets = n_errors = 0;
2892 LIST_FOR_EACH (learning_packet, list_node, &packets) {
2894 void *port_void = learning_packet->frame;
2896 /* Restore 'frame'. */
2897 learning_packet->frame = ofpbuf_data(learning_packet);
2898 ret = ofproto_dpif_send_packet(port_void, learning_packet);
2905 ofpbuf_list_delete(&packets);
2908 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2909 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
2910 "packets, last error was: %s",
2911 bundle->name, n_errors, n_packets, ovs_strerror(error));
2913 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
2914 bundle->name, n_packets);
2919 bundle_run(struct ofbundle *bundle)
2922 lacp_run(bundle->lacp, send_pdu_cb);
2925 struct ofport_dpif *port;
2927 LIST_FOR_EACH (port, bundle_node, &bundle->ports) {
2928 bond_slave_set_may_enable(bundle->bond, port, port->may_enable);
2931 if (bond_run(bundle->bond, lacp_status(bundle->lacp))) {
2932 bundle->ofproto->backer->need_revalidate = REV_BOND;
2935 if (bond_should_send_learning_packets(bundle->bond)) {
2936 bundle_send_learning_packets(bundle);
2942 bundle_wait(struct ofbundle *bundle)
2945 lacp_wait(bundle->lacp);
2948 bond_wait(bundle->bond);
2955 mirror_set__(struct ofproto *ofproto_, void *aux,
2956 const struct ofproto_mirror_settings *s)
2958 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2959 struct ofbundle **srcs, **dsts;
2964 mirror_destroy(ofproto->mbridge, aux);
2968 srcs = xmalloc(s->n_srcs * sizeof *srcs);
2969 dsts = xmalloc(s->n_dsts * sizeof *dsts);
2971 for (i = 0; i < s->n_srcs; i++) {
2972 srcs[i] = bundle_lookup(ofproto, s->srcs[i]);
2975 for (i = 0; i < s->n_dsts; i++) {
2976 dsts[i] = bundle_lookup(ofproto, s->dsts[i]);
2979 error = mirror_set(ofproto->mbridge, aux, s->name, srcs, s->n_srcs, dsts,
2980 s->n_dsts, s->src_vlans,
2981 bundle_lookup(ofproto, s->out_bundle), s->out_vlan);
2988 mirror_get_stats__(struct ofproto *ofproto, void *aux,
2989 uint64_t *packets, uint64_t *bytes)
2991 return mirror_get_stats(ofproto_dpif_cast(ofproto)->mbridge, aux, packets,
2996 set_flood_vlans(struct ofproto *ofproto_, unsigned long *flood_vlans)
2998 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
2999 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
3000 if (mac_learning_set_flood_vlans(ofproto->ml, flood_vlans)) {
3001 mac_learning_flush(ofproto->ml);
3003 ovs_rwlock_unlock(&ofproto->ml->rwlock);
3008 is_mirror_output_bundle(const struct ofproto *ofproto_, void *aux)
3010 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3011 struct ofbundle *bundle = bundle_lookup(ofproto, aux);
3012 return bundle && mirror_bundle_out(ofproto->mbridge, bundle) != 0;
3016 forward_bpdu_changed(struct ofproto *ofproto_)
3018 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3019 ofproto->backer->need_revalidate = REV_RECONFIGURE;
3023 set_mac_table_config(struct ofproto *ofproto_, unsigned int idle_time,
3026 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3027 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
3028 mac_learning_set_idle_time(ofproto->ml, idle_time);
3029 mac_learning_set_max_entries(ofproto->ml, max_entries);
3030 ovs_rwlock_unlock(&ofproto->ml->rwlock);
3033 /* Configures multicast snooping on 'ofport' using the settings
3034 * defined in 's'. */
3036 set_mcast_snooping(struct ofproto *ofproto_,
3037 const struct ofproto_mcast_snooping_settings *s)
3039 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3041 /* Only revalidate flows if the configuration changed. */
3042 if (!s != !ofproto->ms) {
3043 ofproto->backer->need_revalidate = REV_RECONFIGURE;
3048 ofproto->ms = mcast_snooping_create();
3051 ovs_rwlock_wrlock(&ofproto->ms->rwlock);
3052 mcast_snooping_set_idle_time(ofproto->ms, s->idle_time);
3053 mcast_snooping_set_max_entries(ofproto->ms, s->max_entries);
3054 if (mcast_snooping_set_flood_unreg(ofproto->ms, s->flood_unreg)) {
3055 ofproto->backer->need_revalidate = REV_RECONFIGURE;
3057 ovs_rwlock_unlock(&ofproto->ms->rwlock);
3059 mcast_snooping_unref(ofproto->ms);
3066 /* Configures multicast snooping port's flood setting on 'ofproto'. */
3068 set_mcast_snooping_port(struct ofproto *ofproto_, void *aux, bool flood)
3070 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3071 struct ofbundle *bundle = bundle_lookup(ofproto, aux);
3074 ovs_rwlock_wrlock(&ofproto->ms->rwlock);
3075 mcast_snooping_set_port_flood(ofproto->ms, bundle->vlan, bundle,
3077 ovs_rwlock_unlock(&ofproto->ms->rwlock);
3085 static struct ofport_dpif *
3086 get_ofp_port(const struct ofproto_dpif *ofproto, ofp_port_t ofp_port)
3088 struct ofport *ofport = ofproto_get_port(&ofproto->up, ofp_port);
3089 return ofport ? ofport_dpif_cast(ofport) : NULL;
3093 ofproto_port_from_dpif_port(struct ofproto_dpif *ofproto,
3094 struct ofproto_port *ofproto_port,
3095 struct dpif_port *dpif_port)
3097 ofproto_port->name = dpif_port->name;
3098 ofproto_port->type = dpif_port->type;
3099 ofproto_port->ofp_port = odp_port_to_ofp_port(ofproto, dpif_port->port_no);
3103 ofport_update_peer(struct ofport_dpif *ofport)
3105 const struct ofproto_dpif *ofproto;
3106 struct dpif_backer *backer;
3109 if (!netdev_vport_is_patch(ofport->up.netdev)) {
3113 backer = ofproto_dpif_cast(ofport->up.ofproto)->backer;
3114 backer->need_revalidate = REV_RECONFIGURE;
3117 ofport->peer->peer = NULL;
3118 ofport->peer = NULL;
3121 peer_name = netdev_vport_patch_peer(ofport->up.netdev);
3126 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
3127 struct ofport *peer_ofport;
3128 struct ofport_dpif *peer;
3131 if (ofproto->backer != backer) {
3135 peer_ofport = shash_find_data(&ofproto->up.port_by_name, peer_name);
3140 peer = ofport_dpif_cast(peer_ofport);
3141 peer_peer = netdev_vport_patch_peer(peer->up.netdev);
3142 if (peer_peer && !strcmp(netdev_get_name(ofport->up.netdev),
3144 ofport->peer = peer;
3145 ofport->peer->peer = ofport;
3155 port_run(struct ofport_dpif *ofport)
3157 long long int carrier_seq = netdev_get_carrier_resets(ofport->up.netdev);
3158 bool carrier_changed = carrier_seq != ofport->carrier_seq;
3159 bool enable = netdev_get_carrier(ofport->up.netdev);
3160 bool cfm_enable = false;
3161 bool bfd_enable = false;
3163 ofport->carrier_seq = carrier_seq;
3166 int cfm_opup = cfm_get_opup(ofport->cfm);
3168 cfm_enable = !cfm_get_fault(ofport->cfm);
3170 if (cfm_opup >= 0) {
3171 cfm_enable = cfm_enable && cfm_opup;
3176 bfd_enable = bfd_forwarding(ofport->bfd);
3179 if (ofport->bfd || ofport->cfm) {
3180 enable = enable && (cfm_enable || bfd_enable);
3183 if (ofport->bundle) {
3184 enable = enable && lacp_slave_may_enable(ofport->bundle->lacp, ofport);
3185 if (carrier_changed) {
3186 lacp_slave_carrier_changed(ofport->bundle->lacp, ofport);
3190 if (ofport->may_enable != enable) {
3191 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
3193 ofproto->backer->need_revalidate = REV_PORT_TOGGLED;
3195 if (ofport->rstp_port) {
3196 rstp_port_set_mac_operational(ofport->rstp_port, enable);
3200 ofport->may_enable = enable;
3204 port_query_by_name(const struct ofproto *ofproto_, const char *devname,
3205 struct ofproto_port *ofproto_port)
3207 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3208 struct dpif_port dpif_port;
3211 if (sset_contains(&ofproto->ghost_ports, devname)) {
3212 const char *type = netdev_get_type_from_name(devname);
3214 /* We may be called before ofproto->up.port_by_name is populated with
3215 * the appropriate ofport. For this reason, we must get the name and
3216 * type from the netdev layer directly. */
3218 const struct ofport *ofport;
3220 ofport = shash_find_data(&ofproto->up.port_by_name, devname);
3221 ofproto_port->ofp_port = ofport ? ofport->ofp_port : OFPP_NONE;
3222 ofproto_port->name = xstrdup(devname);
3223 ofproto_port->type = xstrdup(type);
3229 if (!sset_contains(&ofproto->ports, devname)) {
3232 error = dpif_port_query_by_name(ofproto->backer->dpif,
3233 devname, &dpif_port);
3235 ofproto_port_from_dpif_port(ofproto, ofproto_port, &dpif_port);
3241 port_add(struct ofproto *ofproto_, struct netdev *netdev)
3243 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3244 const char *devname = netdev_get_name(netdev);
3245 char namebuf[NETDEV_VPORT_NAME_BUFSIZE];
3246 const char *dp_port_name;
3248 if (netdev_vport_is_patch(netdev)) {
3249 sset_add(&ofproto->ghost_ports, netdev_get_name(netdev));
3253 dp_port_name = netdev_vport_get_dpif_port(netdev, namebuf, sizeof namebuf);
3254 if (!dpif_port_exists(ofproto->backer->dpif, dp_port_name)) {
3255 odp_port_t port_no = ODPP_NONE;
3258 error = dpif_port_add(ofproto->backer->dpif, netdev, &port_no);
3262 if (netdev_get_tunnel_config(netdev)) {
3263 simap_put(&ofproto->backer->tnl_backers,
3264 dp_port_name, odp_to_u32(port_no));
3268 if (netdev_get_tunnel_config(netdev)) {
3269 sset_add(&ofproto->ghost_ports, devname);
3271 sset_add(&ofproto->ports, devname);
3277 port_del(struct ofproto *ofproto_, ofp_port_t ofp_port)
3279 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3280 struct ofport_dpif *ofport = get_ofp_port(ofproto, ofp_port);
3287 sset_find_and_delete(&ofproto->ghost_ports,
3288 netdev_get_name(ofport->up.netdev));
3289 ofproto->backer->need_revalidate = REV_RECONFIGURE;
3290 if (!ofport->is_tunnel && !netdev_vport_is_patch(ofport->up.netdev)) {
3291 error = dpif_port_del(ofproto->backer->dpif, ofport->odp_port);
3293 /* The caller is going to close ofport->up.netdev. If this is a
3294 * bonded port, then the bond is using that netdev, so remove it
3295 * from the bond. The client will need to reconfigure everything
3296 * after deleting ports, so then the slave will get re-added. */
3297 bundle_remove(&ofport->up);
3304 port_get_stats(const struct ofport *ofport_, struct netdev_stats *stats)
3306 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
3309 error = netdev_get_stats(ofport->up.netdev, stats);
3311 if (!error && ofport_->ofp_port == OFPP_LOCAL) {
3312 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
3314 ovs_mutex_lock(&ofproto->stats_mutex);
3315 /* ofproto->stats.tx_packets represents packets that we created
3316 * internally and sent to some port (e.g. packets sent with
3317 * ofproto_dpif_send_packet()). Account for them as if they had
3318 * come from OFPP_LOCAL and got forwarded. */
3320 if (stats->rx_packets != UINT64_MAX) {
3321 stats->rx_packets += ofproto->stats.tx_packets;
3324 if (stats->rx_bytes != UINT64_MAX) {
3325 stats->rx_bytes += ofproto->stats.tx_bytes;
3328 /* ofproto->stats.rx_packets represents packets that were received on
3329 * some port and we processed internally and dropped (e.g. STP).
3330 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3332 if (stats->tx_packets != UINT64_MAX) {
3333 stats->tx_packets += ofproto->stats.rx_packets;
3336 if (stats->tx_bytes != UINT64_MAX) {
3337 stats->tx_bytes += ofproto->stats.rx_bytes;
3339 ovs_mutex_unlock(&ofproto->stats_mutex);
3346 port_get_lacp_stats(const struct ofport *ofport_, struct lacp_slave_stats *stats)
3348 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
3349 if (ofport->bundle && ofport->bundle->lacp) {
3350 if (lacp_get_slave_stats(ofport->bundle->lacp, ofport, stats)) {
3357 struct port_dump_state {
3362 struct ofproto_port port;
3367 port_dump_start(const struct ofproto *ofproto_ OVS_UNUSED, void **statep)
3369 *statep = xzalloc(sizeof(struct port_dump_state));
3374 port_dump_next(const struct ofproto *ofproto_, void *state_,
3375 struct ofproto_port *port)
3377 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3378 struct port_dump_state *state = state_;
3379 const struct sset *sset;
3380 struct sset_node *node;
3382 if (state->has_port) {
3383 ofproto_port_destroy(&state->port);
3384 state->has_port = false;
3386 sset = state->ghost ? &ofproto->ghost_ports : &ofproto->ports;
3387 while ((node = sset_at_position(sset, &state->bucket, &state->offset))) {
3390 error = port_query_by_name(ofproto_, node->name, &state->port);
3392 *port = state->port;
3393 state->has_port = true;
3395 } else if (error != ENODEV) {
3400 if (!state->ghost) {
3401 state->ghost = true;
3404 return port_dump_next(ofproto_, state_, port);
3411 port_dump_done(const struct ofproto *ofproto_ OVS_UNUSED, void *state_)
3413 struct port_dump_state *state = state_;
3415 if (state->has_port) {
3416 ofproto_port_destroy(&state->port);
3423 port_poll(const struct ofproto *ofproto_, char **devnamep)
3425 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3427 if (ofproto->port_poll_errno) {
3428 int error = ofproto->port_poll_errno;
3429 ofproto->port_poll_errno = 0;
3433 if (sset_is_empty(&ofproto->port_poll_set)) {
3437 *devnamep = sset_pop(&ofproto->port_poll_set);
3442 port_poll_wait(const struct ofproto *ofproto_)
3444 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
3445 dpif_port_poll_wait(ofproto->backer->dpif);
3449 port_is_lacp_current(const struct ofport *ofport_)
3451 const struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
3452 return (ofport->bundle && ofport->bundle->lacp
3453 ? lacp_slave_is_current(ofport->bundle->lacp, ofport)
3457 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3458 * then delete it entirely. */
3460 rule_expire(struct rule_dpif *rule)
3461 OVS_REQUIRES(ofproto_mutex)
3463 uint16_t hard_timeout, idle_timeout;
3464 long long int now = time_msec();
3467 hard_timeout = rule->up.hard_timeout;
3468 idle_timeout = rule->up.idle_timeout;
3470 /* Has 'rule' expired? */
3472 long long int modified;
3474 ovs_mutex_lock(&rule->up.mutex);
3475 modified = rule->up.modified;
3476 ovs_mutex_unlock(&rule->up.mutex);
3478 if (now > modified + hard_timeout * 1000) {
3479 reason = OFPRR_HARD_TIMEOUT;
3483 if (reason < 0 && idle_timeout) {
3486 ovs_mutex_lock(&rule->stats_mutex);
3487 used = rule->stats.used;
3488 ovs_mutex_unlock(&rule->stats_mutex);
3490 if (now > used + idle_timeout * 1000) {
3491 reason = OFPRR_IDLE_TIMEOUT;
3496 COVERAGE_INC(ofproto_dpif_expired);
3497 ofproto_rule_expire(&rule->up, reason);
3501 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3502 * 'flow' must reflect the data in 'packet'. */
3504 ofproto_dpif_execute_actions(struct ofproto_dpif *ofproto,
3505 const struct flow *flow,
3506 struct rule_dpif *rule,
3507 const struct ofpact *ofpacts, size_t ofpacts_len,
3508 struct ofpbuf *packet)
3510 struct dpif_flow_stats stats;
3511 struct xlate_out xout;
3512 struct xlate_in xin;
3514 struct dpif_execute execute;
3517 ovs_assert((rule != NULL) != (ofpacts != NULL));
3519 dpif_flow_stats_extract(flow, packet, time_msec(), &stats);
3522 rule_dpif_credit_stats(rule, &stats);
3525 xlate_in_init(&xin, ofproto, flow, flow->in_port.ofp_port, rule,
3526 stats.tcp_flags, packet);
3527 xin.ofpacts = ofpacts;
3528 xin.ofpacts_len = ofpacts_len;
3529 xin.resubmit_stats = &stats;
3530 xlate_actions(&xin, &xout);
3532 execute.actions = ofpbuf_data(xout.odp_actions);
3533 execute.actions_len = ofpbuf_size(xout.odp_actions);
3535 execute.packet = packet;
3536 execute.md = pkt_metadata_from_flow(flow);
3537 execute.needs_help = (xout.slow & SLOW_ACTION) != 0;
3538 execute.probe = false;
3540 /* Fix up in_port. */
3541 in_port = flow->in_port.ofp_port;
3542 if (in_port == OFPP_NONE) {
3543 in_port = OFPP_LOCAL;
3545 execute.md.in_port.odp_port = ofp_port_to_odp_port(ofproto, in_port);
3547 error = dpif_execute(ofproto->backer->dpif, &execute);
3549 xlate_out_uninit(&xout);
3555 rule_dpif_credit_stats(struct rule_dpif *rule,
3556 const struct dpif_flow_stats *stats)
3558 ovs_mutex_lock(&rule->stats_mutex);
3559 rule->stats.n_packets += stats->n_packets;
3560 rule->stats.n_bytes += stats->n_bytes;
3561 rule->stats.used = MAX(rule->stats.used, stats->used);
3562 ovs_mutex_unlock(&rule->stats_mutex);
3566 rule_dpif_get_flow_cookie(const struct rule_dpif *rule)
3567 OVS_REQUIRES(rule->up.mutex)
3569 return rule->up.flow_cookie;
3573 rule_dpif_reduce_timeouts(struct rule_dpif *rule, uint16_t idle_timeout,
3574 uint16_t hard_timeout)
3576 ofproto_rule_reduce_timeouts(&rule->up, idle_timeout, hard_timeout);
3579 /* Returns 'rule''s actions. The returned actions are RCU-protected, and can
3580 * be read until the calling thread quiesces. */
3581 const struct rule_actions *
3582 rule_dpif_get_actions(const struct rule_dpif *rule)
3584 return rule_get_actions(&rule->up);
3587 /* Sets 'rule''s recirculation id. */
3589 rule_dpif_set_recirc_id(struct rule_dpif *rule, uint32_t id)
3590 OVS_REQUIRES(rule->up.mutex)
3592 ovs_assert(!rule->recirc_id);
3593 rule->recirc_id = id;
3596 /* Returns 'rule''s recirculation id. */
3598 rule_dpif_get_recirc_id(struct rule_dpif *rule)
3599 OVS_REQUIRES(rule->up.mutex)
3601 if (!rule->recirc_id) {
3602 struct ofproto_dpif *ofproto = ofproto_dpif_cast(rule->up.ofproto);
3604 rule_dpif_set_recirc_id(rule, ofproto_dpif_alloc_recirc_id(ofproto));
3606 return rule->recirc_id;
3609 /* Sets 'rule''s recirculation id. */
3611 rule_set_recirc_id(struct rule *rule_, uint32_t id)
3613 struct rule_dpif *rule = rule_dpif_cast(rule_);
3615 ovs_mutex_lock(&rule->up.mutex);
3616 rule_dpif_set_recirc_id(rule, id);
3617 ovs_mutex_unlock(&rule->up.mutex);
3620 /* Lookup 'flow' in table 0 of 'ofproto''s classifier.
3621 * If 'wc' is non-null, sets the fields that were relevant as part of
3622 * the lookup. Returns the table id where a match or miss occurred via
3623 * 'table_id'. This will be zero unless there was a miss and
3624 * OFPTC11_TABLE_MISS_CONTINUE is in effect for the sequence of tables
3625 * where misses occur, or TBL_INTERNAL if the rule has a non-zero
3626 * recirculation ID, and a match was found in the internal table, or if
3627 * there was no match and one of the special rules (drop_frags_rule,
3628 * miss_rule, or no_packet_in_rule) was returned.
3630 * The return value is the found rule, which is valid at least until the next
3631 * RCU quiescent period. If the rule needs to stay around longer,
3632 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3633 * on it before this returns. */
3635 rule_dpif_lookup(struct ofproto_dpif *ofproto, struct flow *flow,
3636 struct flow_wildcards *wc, bool take_ref,
3637 const struct dpif_flow_stats *stats, uint8_t *table_id)
3641 if (ofproto_dpif_get_enable_recirc(ofproto)) {
3642 /* Always exactly match recirc_id since datapath supports
3645 wc->masks.recirc_id = UINT32_MAX;
3647 *table_id = rule_dpif_lookup_get_init_table_id(flow);
3650 return rule_dpif_lookup_from_table(ofproto, flow, wc, take_ref, stats,
3651 table_id, flow->in_port.ofp_port, true,
3655 /* The returned rule (if any) is valid at least until the next RCU quiescent
3656 * period. If the rule needs to stay around longer, a non-zero 'take_ref'
3657 * must be passed in to cause a reference to be taken on it. */
3658 static struct rule_dpif *
3659 rule_dpif_lookup_in_table(struct ofproto_dpif *ofproto, uint8_t table_id,
3660 const struct flow *flow, struct flow_wildcards *wc,
3663 struct classifier *cls = &ofproto->up.tables[table_id].cls;
3664 const struct cls_rule *cls_rule;
3665 struct rule_dpif *rule;
3668 cls_rule = classifier_lookup(cls, flow, wc);
3670 rule = rule_dpif_cast(rule_from_cls_rule(cls_rule));
3672 /* Try again if the rule was released before we get the reference. */
3673 } while (rule && take_ref && !rule_dpif_try_ref(rule));
3678 /* Look up 'flow' in 'ofproto''s classifier starting from table '*table_id'.
3679 * Returns the rule that was found, which may be one of the special rules
3680 * according to packet miss hadling. If 'may_packet_in' is false, returning of
3681 * the miss_rule (which issues packet ins for the controller) is avoided.
3682 * Updates 'wc', if nonnull, to reflect the fields that were used during the
3685 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3686 * if none is found then the table miss configuration for that table is
3687 * honored, which can result in additional lookups in other OpenFlow tables.
3688 * In this case the function updates '*table_id' to reflect the final OpenFlow
3689 * table that was searched.
3691 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3694 * The rule is returned in '*rule', which is valid at least until the next
3695 * RCU quiescent period. If the '*rule' needs to stay around longer,
3696 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3697 * on it before this returns.
3699 * 'in_port' allows the lookup to take place as if the in port had the value
3700 * 'in_port'. This is needed for resubmit action support. */
3702 rule_dpif_lookup_from_table(struct ofproto_dpif *ofproto, struct flow *flow,
3703 struct flow_wildcards *wc, bool take_ref,
3704 const struct dpif_flow_stats *stats,
3705 uint8_t *table_id, ofp_port_t in_port,
3706 bool may_packet_in, bool honor_table_miss)
3708 ovs_be16 old_tp_src = flow->tp_src, old_tp_dst = flow->tp_dst;
3709 ofp_port_t old_in_port = flow->in_port.ofp_port;
3710 enum ofputil_table_miss miss_config;
3711 struct rule_dpif *rule;
3714 /* We always unwildcard nw_frag (for IP), so they
3715 * need not be unwildcarded here. */
3716 if (flow->nw_frag & FLOW_NW_FRAG_ANY
3717 && ofproto->up.frag_handling != OFPC_FRAG_NX_MATCH) {
3718 if (ofproto->up.frag_handling == OFPC_FRAG_NORMAL) {
3719 /* We must pretend that transport ports are unavailable. */
3720 flow->tp_src = htons(0);
3721 flow->tp_dst = htons(0);
3723 /* Must be OFPC_FRAG_DROP (we don't have OFPC_FRAG_REASM).
3724 * Use the drop_frags_rule (which cannot disappear). */
3725 rule = ofproto->drop_frags_rule;
3727 rule_dpif_ref(rule);
3730 struct oftable *tbl = &ofproto->up.tables[*table_id];
3733 atomic_add_relaxed(&tbl->n_matched, stats->n_packets, &orig);
3739 /* Look up a flow with 'in_port' as the input port. Then restore the
3740 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
3741 * have surprising behavior). */
3742 flow->in_port.ofp_port = in_port;
3744 /* Our current implementation depends on n_tables == N_TABLES, and
3745 * TBL_INTERNAL being the last table. */
3746 BUILD_ASSERT_DECL(N_TABLES == TBL_INTERNAL + 1);
3748 miss_config = OFPUTIL_TABLE_MISS_CONTINUE;
3750 for (next_id = *table_id;
3751 next_id < ofproto->up.n_tables;
3752 next_id++, next_id += (next_id == TBL_INTERNAL))
3754 *table_id = next_id;
3755 rule = rule_dpif_lookup_in_table(ofproto, next_id, flow, wc, take_ref);
3757 struct oftable *tbl = &ofproto->up.tables[next_id];
3760 atomic_add_relaxed(rule ? &tbl->n_matched : &tbl->n_missed,
3761 stats->n_packets, &orig);
3764 goto out; /* Match. */
3766 if (honor_table_miss) {
3767 miss_config = ofproto_table_get_miss_config(&ofproto->up,
3769 if (miss_config == OFPUTIL_TABLE_MISS_CONTINUE) {
3776 rule = ofproto->no_packet_in_rule;
3777 if (may_packet_in) {
3778 if (miss_config == OFPUTIL_TABLE_MISS_CONTINUE
3779 || miss_config == OFPUTIL_TABLE_MISS_CONTROLLER) {
3780 struct ofport_dpif *port;
3782 port = get_ofp_port(ofproto, old_in_port);
3784 VLOG_WARN_RL(&rl, "packet-in on unknown OpenFlow port %"PRIu16,
3786 } else if (!(port->up.pp.config & OFPUTIL_PC_NO_PACKET_IN)) {
3787 rule = ofproto->miss_rule;
3789 } else if (miss_config == OFPUTIL_TABLE_MISS_DEFAULT &&
3790 connmgr_wants_packet_in_on_miss(ofproto->up.connmgr)) {
3791 rule = ofproto->miss_rule;
3795 rule_dpif_ref(rule);
3798 /* Restore port numbers, as they may have been modified above. */
3799 flow->tp_src = old_tp_src;
3800 flow->tp_dst = old_tp_dst;
3801 /* Restore the old in port. */
3802 flow->in_port.ofp_port = old_in_port;
3808 complete_operation(struct rule_dpif *rule)
3809 OVS_REQUIRES(ofproto_mutex)
3811 struct ofproto_dpif *ofproto = ofproto_dpif_cast(rule->up.ofproto);
3813 ofproto->backer->need_revalidate = REV_FLOW_TABLE;
3816 static struct rule_dpif *rule_dpif_cast(const struct rule *rule)
3818 return rule ? CONTAINER_OF(rule, struct rule_dpif, up) : NULL;
3821 static struct rule *
3824 struct rule_dpif *rule = xmalloc(sizeof *rule);
3829 rule_dealloc(struct rule *rule_)
3831 struct rule_dpif *rule = rule_dpif_cast(rule_);
3836 rule_construct(struct rule *rule_)
3837 OVS_NO_THREAD_SAFETY_ANALYSIS
3839 struct rule_dpif *rule = rule_dpif_cast(rule_);
3840 ovs_mutex_init_adaptive(&rule->stats_mutex);
3841 rule->stats.n_packets = 0;
3842 rule->stats.n_bytes = 0;
3843 rule->stats.used = rule->up.modified;
3844 rule->recirc_id = 0;
3850 rule_insert(struct rule *rule_)
3851 OVS_REQUIRES(ofproto_mutex)
3853 struct rule_dpif *rule = rule_dpif_cast(rule_);
3854 complete_operation(rule);
3859 rule_delete(struct rule *rule_)
3860 OVS_REQUIRES(ofproto_mutex)
3862 struct rule_dpif *rule = rule_dpif_cast(rule_);
3863 complete_operation(rule);
3867 rule_destruct(struct rule *rule_)
3869 struct rule_dpif *rule = rule_dpif_cast(rule_);
3871 ovs_mutex_destroy(&rule->stats_mutex);
3872 if (rule->recirc_id) {
3873 struct ofproto_dpif *ofproto = ofproto_dpif_cast(rule->up.ofproto);
3875 ofproto_dpif_free_recirc_id(ofproto, rule->recirc_id);
3880 rule_get_stats(struct rule *rule_, uint64_t *packets, uint64_t *bytes,
3881 long long int *used)
3883 struct rule_dpif *rule = rule_dpif_cast(rule_);
3885 ovs_mutex_lock(&rule->stats_mutex);
3886 *packets = rule->stats.n_packets;
3887 *bytes = rule->stats.n_bytes;
3888 *used = rule->stats.used;
3889 ovs_mutex_unlock(&rule->stats_mutex);
3893 rule_dpif_execute(struct rule_dpif *rule, const struct flow *flow,
3894 struct ofpbuf *packet)
3896 struct ofproto_dpif *ofproto = ofproto_dpif_cast(rule->up.ofproto);
3898 ofproto_dpif_execute_actions(ofproto, flow, rule, NULL, 0, packet);
3902 rule_execute(struct rule *rule, const struct flow *flow,
3903 struct ofpbuf *packet)
3905 rule_dpif_execute(rule_dpif_cast(rule), flow, packet);
3906 ofpbuf_delete(packet);
3911 rule_modify_actions(struct rule *rule_, bool reset_counters)
3912 OVS_REQUIRES(ofproto_mutex)
3914 struct rule_dpif *rule = rule_dpif_cast(rule_);
3916 if (reset_counters) {
3917 ovs_mutex_lock(&rule->stats_mutex);
3918 rule->stats.n_packets = 0;
3919 rule->stats.n_bytes = 0;
3920 ovs_mutex_unlock(&rule->stats_mutex);
3923 complete_operation(rule);
3926 static struct group_dpif *group_dpif_cast(const struct ofgroup *group)
3928 return group ? CONTAINER_OF(group, struct group_dpif, up) : NULL;
3931 static struct ofgroup *
3934 struct group_dpif *group = xzalloc(sizeof *group);
3939 group_dealloc(struct ofgroup *group_)
3941 struct group_dpif *group = group_dpif_cast(group_);
3946 group_construct_stats(struct group_dpif *group)
3947 OVS_REQUIRES(group->stats_mutex)
3949 struct ofputil_bucket *bucket;
3950 const struct list *buckets;
3952 group->packet_count = 0;
3953 group->byte_count = 0;
3955 group_dpif_get_buckets(group, &buckets);
3956 LIST_FOR_EACH (bucket, list_node, buckets) {
3957 bucket->stats.packet_count = 0;
3958 bucket->stats.byte_count = 0;
3963 group_dpif_credit_stats(struct group_dpif *group,
3964 struct ofputil_bucket *bucket,
3965 const struct dpif_flow_stats *stats)
3967 ovs_mutex_lock(&group->stats_mutex);
3968 group->packet_count += stats->n_packets;
3969 group->byte_count += stats->n_bytes;
3971 bucket->stats.packet_count += stats->n_packets;
3972 bucket->stats.byte_count += stats->n_bytes;
3973 } else { /* Credit to all buckets */
3974 const struct list *buckets;
3976 group_dpif_get_buckets(group, &buckets);
3977 LIST_FOR_EACH (bucket, list_node, buckets) {
3978 bucket->stats.packet_count += stats->n_packets;
3979 bucket->stats.byte_count += stats->n_bytes;
3982 ovs_mutex_unlock(&group->stats_mutex);
3986 group_construct(struct ofgroup *group_)
3988 struct group_dpif *group = group_dpif_cast(group_);
3989 const struct ofputil_bucket *bucket;
3991 /* Prevent group chaining because our locking structure makes it hard to
3992 * implement deadlock-free. (See xlate_group_resource_check().) */
3993 LIST_FOR_EACH (bucket, list_node, &group->up.buckets) {
3994 const struct ofpact *a;
3996 OFPACT_FOR_EACH (a, bucket->ofpacts, bucket->ofpacts_len) {
3997 if (a->type == OFPACT_GROUP) {
3998 return OFPERR_OFPGMFC_CHAINING_UNSUPPORTED;
4003 ovs_mutex_init_adaptive(&group->stats_mutex);
4004 ovs_mutex_lock(&group->stats_mutex);
4005 group_construct_stats(group);
4006 ovs_mutex_unlock(&group->stats_mutex);
4011 group_destruct(struct ofgroup *group_)
4013 struct group_dpif *group = group_dpif_cast(group_);
4014 ovs_mutex_destroy(&group->stats_mutex);
4018 group_modify(struct ofgroup *group_)
4020 struct ofproto_dpif *ofproto = ofproto_dpif_cast(group_->ofproto);
4022 ofproto->backer->need_revalidate = REV_FLOW_TABLE;
4028 group_get_stats(const struct ofgroup *group_, struct ofputil_group_stats *ogs)
4030 struct group_dpif *group = group_dpif_cast(group_);
4031 struct ofputil_bucket *bucket;
4032 const struct list *buckets;
4033 struct bucket_counter *bucket_stats;
4035 ovs_mutex_lock(&group->stats_mutex);
4036 ogs->packet_count = group->packet_count;
4037 ogs->byte_count = group->byte_count;
4039 group_dpif_get_buckets(group, &buckets);
4040 bucket_stats = ogs->bucket_stats;
4041 LIST_FOR_EACH (bucket, list_node, buckets) {
4042 bucket_stats->packet_count = bucket->stats.packet_count;
4043 bucket_stats->byte_count = bucket->stats.byte_count;
4046 ovs_mutex_unlock(&group->stats_mutex);
4051 /* If the group exists, this function increments the groups's reference count.
4053 * Make sure to call group_dpif_unref() after no longer needing to maintain
4054 * a reference to the group. */
4056 group_dpif_lookup(struct ofproto_dpif *ofproto, uint32_t group_id,
4057 struct group_dpif **group)
4059 struct ofgroup *ofgroup;
4062 found = ofproto_group_lookup(&ofproto->up, group_id, &ofgroup);
4063 *group = found ? group_dpif_cast(ofgroup) : NULL;
4069 group_dpif_get_buckets(const struct group_dpif *group,
4070 const struct list **buckets)
4072 *buckets = &group->up.buckets;
4075 enum ofp11_group_type
4076 group_dpif_get_type(const struct group_dpif *group)
4078 return group->up.type;
4081 /* Sends 'packet' out 'ofport'.
4082 * May modify 'packet'.
4083 * Returns 0 if successful, otherwise a positive errno value. */
4085 ofproto_dpif_send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet)
4087 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
4090 error = xlate_send_packet(ofport, packet);
4092 ovs_mutex_lock(&ofproto->stats_mutex);
4093 ofproto->stats.tx_packets++;
4094 ofproto->stats.tx_bytes += ofpbuf_size(packet);
4095 ovs_mutex_unlock(&ofproto->stats_mutex);
4099 /* Return the version string of the datapath that backs up
4103 get_datapath_version(const struct ofproto *ofproto_)
4105 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
4107 return ofproto->backer->dp_version_string;
4111 set_frag_handling(struct ofproto *ofproto_,
4112 enum ofp_config_flags frag_handling)
4114 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
4115 if (frag_handling != OFPC_FRAG_REASM) {
4116 ofproto->backer->need_revalidate = REV_RECONFIGURE;
4124 packet_out(struct ofproto *ofproto_, struct ofpbuf *packet,
4125 const struct flow *flow,
4126 const struct ofpact *ofpacts, size_t ofpacts_len)
4128 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
4130 ofproto_dpif_execute_actions(ofproto, flow, NULL, ofpacts,
4131 ofpacts_len, packet);
4138 set_netflow(struct ofproto *ofproto_,
4139 const struct netflow_options *netflow_options)
4141 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
4143 if (netflow_options) {
4144 if (!ofproto->netflow) {
4145 ofproto->netflow = netflow_create();
4146 ofproto->backer->need_revalidate = REV_RECONFIGURE;
4148 return netflow_set_options(ofproto->netflow, netflow_options);
4149 } else if (ofproto->netflow) {
4150 ofproto->backer->need_revalidate = REV_RECONFIGURE;
4151 netflow_unref(ofproto->netflow);
4152 ofproto->netflow = NULL;
4159 get_netflow_ids(const struct ofproto *ofproto_,
4160 uint8_t *engine_type, uint8_t *engine_id)
4162 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
4164 dpif_get_netflow_ids(ofproto->backer->dpif, engine_type, engine_id);
4167 static struct ofproto_dpif *
4168 ofproto_dpif_lookup(const char *name)
4170 struct ofproto_dpif *ofproto;
4172 HMAP_FOR_EACH_WITH_HASH (ofproto, all_ofproto_dpifs_node,
4173 hash_string(name, 0), &all_ofproto_dpifs) {
4174 if (!strcmp(ofproto->up.name, name)) {
4182 ofproto_unixctl_fdb_flush(struct unixctl_conn *conn, int argc,
4183 const char *argv[], void *aux OVS_UNUSED)
4185 struct ofproto_dpif *ofproto;
4188 ofproto = ofproto_dpif_lookup(argv[1]);
4190 unixctl_command_reply_error(conn, "no such bridge");
4193 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
4194 mac_learning_flush(ofproto->ml);
4195 ovs_rwlock_unlock(&ofproto->ml->rwlock);
4197 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
4198 ovs_rwlock_wrlock(&ofproto->ml->rwlock);
4199 mac_learning_flush(ofproto->ml);
4200 ovs_rwlock_unlock(&ofproto->ml->rwlock);
4204 unixctl_command_reply(conn, "table successfully flushed");
4208 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn *conn, int argc,
4209 const char *argv[], void *aux OVS_UNUSED)
4211 struct ofproto_dpif *ofproto;
4214 ofproto = ofproto_dpif_lookup(argv[1]);
4216 unixctl_command_reply_error(conn, "no such bridge");
4220 if (!mcast_snooping_enabled(ofproto->ms)) {
4221 unixctl_command_reply_error(conn, "multicast snooping is disabled");
4224 mcast_snooping_mdb_flush(ofproto->ms);
4226 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
4227 if (!mcast_snooping_enabled(ofproto->ms)) {
4230 mcast_snooping_mdb_flush(ofproto->ms);
4234 unixctl_command_reply(conn, "table successfully flushed");
4237 static struct ofport_dpif *
4238 ofbundle_get_a_port(const struct ofbundle *bundle)
4240 return CONTAINER_OF(list_front(&bundle->ports), struct ofport_dpif,
4245 ofproto_unixctl_fdb_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
4246 const char *argv[], void *aux OVS_UNUSED)
4248 struct ds ds = DS_EMPTY_INITIALIZER;
4249 const struct ofproto_dpif *ofproto;
4250 const struct mac_entry *e;
4252 ofproto = ofproto_dpif_lookup(argv[1]);
4254 unixctl_command_reply_error(conn, "no such bridge");
4258 ds_put_cstr(&ds, " port VLAN MAC Age\n");
4259 ovs_rwlock_rdlock(&ofproto->ml->rwlock);
4260 LIST_FOR_EACH (e, lru_node, &ofproto->ml->lrus) {
4261 struct ofbundle *bundle = e->port.p;
4262 char name[OFP_MAX_PORT_NAME_LEN];
4264 ofputil_port_to_string(ofbundle_get_a_port(bundle)->up.ofp_port,
4266 ds_put_format(&ds, "%5s %4d "ETH_ADDR_FMT" %3d\n",
4267 name, e->vlan, ETH_ADDR_ARGS(e->mac),
4268 mac_entry_age(ofproto->ml, e));
4270 ovs_rwlock_unlock(&ofproto->ml->rwlock);
4271 unixctl_command_reply(conn, ds_cstr(&ds));
4276 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn *conn,
4277 int argc OVS_UNUSED,
4279 void *aux OVS_UNUSED)
4281 struct ds ds = DS_EMPTY_INITIALIZER;
4282 const struct ofproto_dpif *ofproto;
4283 const struct ofbundle *bundle;
4284 const struct mcast_group *grp;
4285 struct mcast_group_bundle *b;
4286 struct mcast_mrouter_bundle *mrouter;
4288 ofproto = ofproto_dpif_lookup(argv[1]);
4290 unixctl_command_reply_error(conn, "no such bridge");
4294 if (!mcast_snooping_enabled(ofproto->ms)) {
4295 unixctl_command_reply_error(conn, "multicast snooping is disabled");
4299 ds_put_cstr(&ds, " port VLAN GROUP Age\n");
4300 ovs_rwlock_rdlock(&ofproto->ms->rwlock);
4301 LIST_FOR_EACH (grp, group_node, &ofproto->ms->group_lru) {
4302 LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) {
4303 char name[OFP_MAX_PORT_NAME_LEN];
4306 ofputil_port_to_string(ofbundle_get_a_port(bundle)->up.ofp_port,
4308 ds_put_format(&ds, "%5s %4d "IP_FMT" %3d\n",
4309 name, grp->vlan, IP_ARGS(grp->ip4),
4310 mcast_bundle_age(ofproto->ms, b));
4314 /* ports connected to multicast routers */
4315 LIST_FOR_EACH(mrouter, mrouter_node, &ofproto->ms->mrouter_lru) {
4316 char name[OFP_MAX_PORT_NAME_LEN];
4318 bundle = mrouter->port;
4319 ofputil_port_to_string(ofbundle_get_a_port(bundle)->up.ofp_port,
4321 ds_put_format(&ds, "%5s %4d querier %3d\n",
4322 name, mrouter->vlan,
4323 mcast_mrouter_age(ofproto->ms, mrouter));
4325 ovs_rwlock_unlock(&ofproto->ms->rwlock);
4326 unixctl_command_reply(conn, ds_cstr(&ds));
4331 struct xlate_out xout;
4332 struct xlate_in xin;
4333 const struct flow *key;
4335 struct flow_wildcards wc;
4340 trace_format_rule(struct ds *result, int level, const struct rule_dpif *rule)
4342 const struct rule_actions *actions;
4345 ds_put_char_multiple(result, '\t', level);
4347 ds_put_cstr(result, "No match\n");
4351 ovs_mutex_lock(&rule->up.mutex);
4352 cookie = rule->up.flow_cookie;
4353 ovs_mutex_unlock(&rule->up.mutex);
4355 ds_put_format(result, "Rule: table=%"PRIu8" cookie=%#"PRIx64" ",
4356 rule ? rule->up.table_id : 0, ntohll(cookie));
4357 cls_rule_format(&rule->up.cr, result);
4358 ds_put_char(result, '\n');
4360 actions = rule_dpif_get_actions(rule);
4362 ds_put_char_multiple(result, '\t', level);
4363 ds_put_cstr(result, "OpenFlow actions=");
4364 ofpacts_format(actions->ofpacts, actions->ofpacts_len, result);
4365 ds_put_char(result, '\n');
4369 trace_format_flow(struct ds *result, int level, const char *title,
4370 struct trace_ctx *trace)
4372 ds_put_char_multiple(result, '\t', level);
4373 ds_put_format(result, "%s: ", title);
4374 /* Do not report unchanged flows for resubmits. */
4375 if ((level > 0 && flow_equal(&trace->xin.flow, &trace->flow))
4376 || (level == 0 && flow_equal(&trace->xin.flow, trace->key))) {
4377 ds_put_cstr(result, "unchanged");
4379 flow_format(result, &trace->xin.flow);
4380 trace->flow = trace->xin.flow;
4382 ds_put_char(result, '\n');
4386 trace_format_regs(struct ds *result, int level, const char *title,
4387 struct trace_ctx *trace)
4391 ds_put_char_multiple(result, '\t', level);
4392 ds_put_format(result, "%s:", title);
4393 for (i = 0; i < FLOW_N_REGS; i++) {
4394 ds_put_format(result, " reg%"PRIuSIZE"=0x%"PRIx32, i, trace->flow.regs[i]);
4396 ds_put_char(result, '\n');
4400 trace_format_odp(struct ds *result, int level, const char *title,
4401 struct trace_ctx *trace)
4403 struct ofpbuf *odp_actions = trace->xout.odp_actions;
4405 ds_put_char_multiple(result, '\t', level);
4406 ds_put_format(result, "%s: ", title);
4407 format_odp_actions(result, ofpbuf_data(odp_actions),
4408 ofpbuf_size(odp_actions));
4409 ds_put_char(result, '\n');
4413 trace_format_megaflow(struct ds *result, int level, const char *title,
4414 struct trace_ctx *trace)
4418 ds_put_char_multiple(result, '\t', level);
4419 ds_put_format(result, "%s: ", title);
4420 flow_wildcards_or(&trace->wc, &trace->xout.wc, &trace->wc);
4421 match_init(&match, trace->key, &trace->wc);
4422 match_format(&match, result, OFP_DEFAULT_PRIORITY);
4423 ds_put_char(result, '\n');
4426 static void trace_report(struct xlate_in *xin, const char *s, int recurse);
4429 trace_resubmit(struct xlate_in *xin, struct rule_dpif *rule, int recurse)
4431 struct trace_ctx *trace = CONTAINER_OF(xin, struct trace_ctx, xin);
4432 struct ds *result = trace->result;
4435 if (rule == xin->ofproto->miss_rule) {
4436 trace_report(xin, "No match, flow generates \"packet in\"s.",
4438 } else if (rule == xin->ofproto->no_packet_in_rule) {
4439 trace_report(xin, "No match, packets dropped because "
4440 "OFPPC_NO_PACKET_IN is set on in_port.", recurse);
4441 } else if (rule == xin->ofproto->drop_frags_rule) {
4442 trace_report(xin, "Packets dropped because they are IP "
4443 "fragments and the fragment handling mode is "
4444 "\"drop\".", recurse);
4448 ds_put_char(result, '\n');
4450 trace_format_flow(result, recurse, "Resubmitted flow", trace);
4451 trace_format_regs(result, recurse, "Resubmitted regs", trace);
4452 trace_format_odp(result, recurse, "Resubmitted odp", trace);
4453 trace_format_megaflow(result, recurse, "Resubmitted megaflow", trace);
4455 trace_format_rule(result, recurse, rule);
4459 trace_report(struct xlate_in *xin, const char *s, int recurse)
4461 struct trace_ctx *trace = CONTAINER_OF(xin, struct trace_ctx, xin);
4462 struct ds *result = trace->result;
4464 ds_put_char_multiple(result, '\t', recurse);
4465 ds_put_cstr(result, s);
4466 ds_put_char(result, '\n');
4469 /* Parses the 'argc' elements of 'argv', ignoring argv[0]. The following
4470 * forms are supported:
4472 * - [dpname] odp_flow [-generate | packet]
4473 * - bridge br_flow [-generate | packet]
4475 * On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure
4476 * returns a nonnull malloced error message. */
4477 static char * WARN_UNUSED_RESULT
4478 parse_flow_and_packet(int argc, const char *argv[],
4479 struct ofproto_dpif **ofprotop, struct flow *flow,
4480 struct ofpbuf **packetp)
4482 const struct dpif_backer *backer = NULL;
4483 const char *error = NULL;
4485 struct simap port_names = SIMAP_INITIALIZER(&port_names);
4486 struct ofpbuf *packet;
4487 struct ofpbuf odp_key;
4488 struct ofpbuf odp_mask;
4490 ofpbuf_init(&odp_key, 0);
4491 ofpbuf_init(&odp_mask, 0);
4493 /* Handle "-generate" or a hex string as the last argument. */
4494 if (!strcmp(argv[argc - 1], "-generate")) {
4495 packet = ofpbuf_new(0);
4498 error = eth_from_hex(argv[argc - 1], &packet);
4501 } else if (argc == 4) {
4502 /* The 3-argument form must end in "-generate' or a hex string. */
4508 /* odp_flow can have its in_port specified as a name instead of port no.
4509 * We do not yet know whether a given flow is a odp_flow or a br_flow.
4510 * But, to know whether a flow is odp_flow through odp_flow_from_string(),
4511 * we need to create a simap of name to port no. */
4513 const char *dp_type;
4514 if (!strncmp(argv[1], "ovs-", 4)) {
4515 dp_type = argv[1] + 4;
4519 backer = shash_find_data(&all_dpif_backers, dp_type);
4520 } else if (argc == 2) {
4521 struct shash_node *node;
4522 if (shash_count(&all_dpif_backers) == 1) {
4523 node = shash_first(&all_dpif_backers);
4524 backer = node->data;
4527 error = "Syntax error";
4530 if (backer && backer->dpif) {
4531 struct dpif_port dpif_port;
4532 struct dpif_port_dump port_dump;
4533 DPIF_PORT_FOR_EACH (&dpif_port, &port_dump, backer->dpif) {
4534 simap_put(&port_names, dpif_port.name,
4535 odp_to_u32(dpif_port.port_no));
4539 /* Parse the flow and determine whether a datapath or
4540 * bridge is specified. If function odp_flow_key_from_string()
4541 * returns 0, the flow is a odp_flow. If function
4542 * parse_ofp_exact_flow() returns NULL, the flow is a br_flow. */
4543 if (!odp_flow_from_string(argv[argc - 1], &port_names,
4544 &odp_key, &odp_mask)) {
4546 error = "Cannot find the datapath";
4550 if (odp_flow_key_to_flow(ofpbuf_data(&odp_key), ofpbuf_size(&odp_key),
4551 flow) == ODP_FIT_ERROR) {
4552 error = "Failed to parse flow key";
4556 *ofprotop = xlate_lookup_ofproto(backer, flow,
4557 &flow->in_port.ofp_port);
4558 if (*ofprotop == NULL) {
4559 error = "Invalid datapath flow";
4563 vsp_adjust_flow(*ofprotop, flow, NULL);
4566 char *err = parse_ofp_exact_flow(flow, NULL, argv[argc - 1], NULL);
4569 m_err = xasprintf("Bad flow syntax: %s", err);
4574 error = "Must specify bridge name";
4578 *ofprotop = ofproto_dpif_lookup(argv[1]);
4580 error = "Unknown bridge name";
4586 /* Generate a packet, if requested. */
4588 if (!ofpbuf_size(packet)) {
4589 flow_compose(packet, flow);
4591 struct pkt_metadata md = pkt_metadata_from_flow(flow);
4593 /* Use the metadata from the flow and the packet argument
4594 * to reconstruct the flow. */
4595 flow_extract(packet, &md, flow);
4600 if (error && !m_err) {
4601 m_err = xstrdup(error);
4604 ofpbuf_delete(packet);
4608 ofpbuf_uninit(&odp_key);
4609 ofpbuf_uninit(&odp_mask);
4610 simap_destroy(&port_names);
4615 ofproto_unixctl_trace(struct unixctl_conn *conn, int argc, const char *argv[],
4616 void *aux OVS_UNUSED)
4618 struct ofproto_dpif *ofproto;
4619 struct ofpbuf *packet;
4623 error = parse_flow_and_packet(argc, argv, &ofproto, &flow, &packet);
4628 ofproto_trace(ofproto, &flow, packet, NULL, 0, &result);
4629 unixctl_command_reply(conn, ds_cstr(&result));
4630 ds_destroy(&result);
4631 ofpbuf_delete(packet);
4633 unixctl_command_reply_error(conn, error);
4639 ofproto_unixctl_trace_actions(struct unixctl_conn *conn, int argc,
4640 const char *argv[], void *aux OVS_UNUSED)
4642 enum ofputil_protocol usable_protocols;
4643 struct ofproto_dpif *ofproto;
4644 bool enforce_consistency;
4645 struct ofpbuf ofpacts;
4646 struct ofpbuf *packet;
4651 /* Three kinds of error return values! */
4657 ofpbuf_init(&ofpacts, 0);
4659 /* Parse actions. */
4660 error = ofpacts_parse_actions(argv[--argc], &ofpacts, &usable_protocols);
4662 unixctl_command_reply_error(conn, error);
4667 /* OpenFlow 1.1 and later suggest that the switch enforces certain forms of
4668 * consistency between the flow and the actions. With -consistent, we
4669 * enforce consistency even for a flow supported in OpenFlow 1.0. */
4670 if (!strcmp(argv[1], "-consistent")) {
4671 enforce_consistency = true;
4675 enforce_consistency = false;
4678 error = parse_flow_and_packet(argc, argv, &ofproto, &flow, &packet);
4680 unixctl_command_reply_error(conn, error);
4685 /* Do the same checks as handle_packet_out() in ofproto.c.
4687 * We pass a 'table_id' of 0 to ofpacts_check(), which isn't
4688 * strictly correct because these actions aren't in any table, but it's OK
4689 * because it 'table_id' is used only to check goto_table instructions, but
4690 * packet-outs take a list of actions and therefore it can't include
4693 * We skip the "meter" check here because meter is an instruction, not an
4694 * action, and thus cannot appear in ofpacts. */
4695 in_port = ofp_to_u16(flow.in_port.ofp_port);
4696 if (in_port >= ofproto->up.max_ports && in_port < ofp_to_u16(OFPP_MAX)) {
4697 unixctl_command_reply_error(conn, "invalid in_port");
4700 if (enforce_consistency) {
4701 retval = ofpacts_check_consistency(ofpbuf_data(&ofpacts), ofpbuf_size(&ofpacts),
4702 &flow, u16_to_ofp(ofproto->up.max_ports),
4703 0, 0, usable_protocols);
4705 retval = ofpacts_check(ofpbuf_data(&ofpacts), ofpbuf_size(&ofpacts), &flow,
4706 u16_to_ofp(ofproto->up.max_ports), 0, 0,
4712 ds_put_format(&result, "Bad actions: %s", ofperr_to_string(retval));
4713 unixctl_command_reply_error(conn, ds_cstr(&result));
4717 ofproto_trace(ofproto, &flow, packet,
4718 ofpbuf_data(&ofpacts), ofpbuf_size(&ofpacts), &result);
4719 unixctl_command_reply(conn, ds_cstr(&result));
4722 ds_destroy(&result);
4723 ofpbuf_delete(packet);
4724 ofpbuf_uninit(&ofpacts);
4727 /* Implements a "trace" through 'ofproto''s flow table, appending a textual
4728 * description of the results to 'ds'.
4730 * The trace follows a packet with the specified 'flow' through the flow
4731 * table. 'packet' may be nonnull to trace an actual packet, with consequent
4732 * side effects (if it is nonnull then its flow must be 'flow').
4734 * If 'ofpacts' is nonnull then its 'ofpacts_len' bytes specify the actions to
4735 * trace, otherwise the actions are determined by a flow table lookup. */
4737 ofproto_trace(struct ofproto_dpif *ofproto, struct flow *flow,
4738 const struct ofpbuf *packet,
4739 const struct ofpact ofpacts[], size_t ofpacts_len,
4742 struct trace_ctx trace;
4744 ds_put_format(ds, "Bridge: %s\n", ofproto->up.name);
4745 ds_put_cstr(ds, "Flow: ");
4746 flow_format(ds, flow);
4747 ds_put_char(ds, '\n');
4749 flow_wildcards_init_catchall(&trace.wc);
4752 trace.key = flow; /* Original flow key, used for megaflow. */
4753 trace.flow = *flow; /* May be modified by actions. */
4754 xlate_in_init(&trace.xin, ofproto, flow, flow->in_port.ofp_port, NULL,
4755 ntohs(flow->tcp_flags), packet);
4756 trace.xin.ofpacts = ofpacts;
4757 trace.xin.ofpacts_len = ofpacts_len;
4758 trace.xin.resubmit_hook = trace_resubmit;
4759 trace.xin.report_hook = trace_report;
4761 xlate_actions(&trace.xin, &trace.xout);
4763 ds_put_char(ds, '\n');
4764 trace_format_flow(ds, 0, "Final flow", &trace);
4765 trace_format_megaflow(ds, 0, "Megaflow", &trace);
4767 ds_put_cstr(ds, "Datapath actions: ");
4768 format_odp_actions(ds, ofpbuf_data(trace.xout.odp_actions),
4769 ofpbuf_size(trace.xout.odp_actions));
4771 if (trace.xout.slow) {
4772 enum slow_path_reason slow;
4774 ds_put_cstr(ds, "\nThis flow is handled by the userspace "
4775 "slow path because it:");
4777 slow = trace.xout.slow;
4779 enum slow_path_reason bit = rightmost_1bit(slow);
4781 ds_put_format(ds, "\n\t- %s.",
4782 slow_path_reason_to_explanation(bit));
4788 xlate_out_uninit(&trace.xout);
4791 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
4792 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
4793 * to destroy 'ofproto_shash' and free the returned value. */
4794 static const struct shash_node **
4795 get_ofprotos(struct shash *ofproto_shash)
4797 const struct ofproto_dpif *ofproto;
4799 HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
4800 char *name = xasprintf("%s@%s", ofproto->up.type, ofproto->up.name);
4801 shash_add_nocopy(ofproto_shash, name, ofproto);
4804 return shash_sort(ofproto_shash);
4808 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn *conn, int argc OVS_UNUSED,
4809 const char *argv[] OVS_UNUSED,
4810 void *aux OVS_UNUSED)
4812 struct ds ds = DS_EMPTY_INITIALIZER;
4813 struct shash ofproto_shash;
4814 const struct shash_node **sorted_ofprotos;
4817 shash_init(&ofproto_shash);
4818 sorted_ofprotos = get_ofprotos(&ofproto_shash);
4819 for (i = 0; i < shash_count(&ofproto_shash); i++) {
4820 const struct shash_node *node = sorted_ofprotos[i];
4821 ds_put_format(&ds, "%s\n", node->name);
4824 shash_destroy(&ofproto_shash);
4825 free(sorted_ofprotos);
4827 unixctl_command_reply(conn, ds_cstr(&ds));
4832 dpif_show_backer(const struct dpif_backer *backer, struct ds *ds)
4834 const struct shash_node **ofprotos;
4835 struct dpif_dp_stats dp_stats;
4836 struct shash ofproto_shash;
4839 dpif_get_dp_stats(backer->dpif, &dp_stats);
4841 ds_put_format(ds, "%s: hit:%"PRIu64" missed:%"PRIu64"\n",
4842 dpif_name(backer->dpif), dp_stats.n_hit, dp_stats.n_missed);
4844 shash_init(&ofproto_shash);
4845 ofprotos = get_ofprotos(&ofproto_shash);
4846 for (i = 0; i < shash_count(&ofproto_shash); i++) {
4847 struct ofproto_dpif *ofproto = ofprotos[i]->data;
4848 const struct shash_node **ports;
4851 if (ofproto->backer != backer) {
4855 ds_put_format(ds, "\t%s:\n", ofproto->up.name);
4857 ports = shash_sort(&ofproto->up.port_by_name);
4858 for (j = 0; j < shash_count(&ofproto->up.port_by_name); j++) {
4859 const struct shash_node *node = ports[j];
4860 struct ofport *ofport = node->data;
4862 odp_port_t odp_port;
4864 ds_put_format(ds, "\t\t%s %u/", netdev_get_name(ofport->netdev),
4867 odp_port = ofp_port_to_odp_port(ofproto, ofport->ofp_port);
4868 if (odp_port != ODPP_NONE) {
4869 ds_put_format(ds, "%"PRIu32":", odp_port);
4871 ds_put_cstr(ds, "none:");
4874 ds_put_format(ds, " (%s", netdev_get_type(ofport->netdev));
4877 if (!netdev_get_config(ofport->netdev, &config)) {
4878 const struct smap_node **nodes;
4881 nodes = smap_sort(&config);
4882 for (i = 0; i < smap_count(&config); i++) {
4883 const struct smap_node *node = nodes[i];
4884 ds_put_format(ds, "%c %s=%s", i ? ',' : ':',
4885 node->key, node->value);
4889 smap_destroy(&config);
4891 ds_put_char(ds, ')');
4892 ds_put_char(ds, '\n');
4896 shash_destroy(&ofproto_shash);
4901 ofproto_unixctl_dpif_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
4902 const char *argv[] OVS_UNUSED, void *aux OVS_UNUSED)
4904 struct ds ds = DS_EMPTY_INITIALIZER;
4905 const struct shash_node **backers;
4908 backers = shash_sort(&all_dpif_backers);
4909 for (i = 0; i < shash_count(&all_dpif_backers); i++) {
4910 dpif_show_backer(backers[i]->data, &ds);
4914 unixctl_command_reply(conn, ds_cstr(&ds));
4919 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn *conn,
4920 int argc OVS_UNUSED, const char *argv[],
4921 void *aux OVS_UNUSED)
4923 const struct ofproto_dpif *ofproto;
4925 struct ds ds = DS_EMPTY_INITIALIZER;
4926 bool verbosity = false;
4928 struct dpif_port dpif_port;
4929 struct dpif_port_dump port_dump;
4930 struct hmap portno_names;
4932 struct dpif_flow_dump *flow_dump;
4933 struct dpif_flow_dump_thread *flow_dump_thread;
4937 ofproto = ofproto_dpif_lookup(argv[argc - 1]);
4939 unixctl_command_reply_error(conn, "no such bridge");
4943 if (argc > 2 && !strcmp(argv[1], "-m")) {
4947 hmap_init(&portno_names);
4948 DPIF_PORT_FOR_EACH (&dpif_port, &port_dump, ofproto->backer->dpif) {
4949 odp_portno_names_set(&portno_names, dpif_port.port_no, dpif_port.name);
4953 flow_dump = dpif_flow_dump_create(ofproto->backer->dpif, false);
4954 flow_dump_thread = dpif_flow_dump_thread_create(flow_dump);
4955 while (dpif_flow_dump_next(flow_dump_thread, &f, 1)) {
4958 if (odp_flow_key_to_flow(f.key, f.key_len, &flow) == ODP_FIT_ERROR
4959 || xlate_lookup_ofproto(ofproto->backer, &flow, NULL) != ofproto) {
4964 odp_format_ufid(&f.ufid, &ds);
4965 ds_put_cstr(&ds, " ");
4967 odp_flow_format(f.key, f.key_len, f.mask, f.mask_len,
4968 &portno_names, &ds, verbosity);
4969 ds_put_cstr(&ds, ", ");
4970 dpif_flow_stats_format(&f.stats, &ds);
4971 ds_put_cstr(&ds, ", actions:");
4972 format_odp_actions(&ds, f.actions, f.actions_len);
4973 ds_put_char(&ds, '\n');
4975 dpif_flow_dump_thread_destroy(flow_dump_thread);
4976 error = dpif_flow_dump_destroy(flow_dump);
4980 ds_put_format(&ds, "dpif/dump_flows failed: %s", ovs_strerror(errno));
4981 unixctl_command_reply_error(conn, ds_cstr(&ds));
4983 unixctl_command_reply(conn, ds_cstr(&ds));
4985 odp_portno_names_destroy(&portno_names);
4986 hmap_destroy(&portno_names);
4991 ofproto_revalidate_all_backers(void)
4993 const struct shash_node **backers;
4996 backers = shash_sort(&all_dpif_backers);
4997 for (i = 0; i < shash_count(&all_dpif_backers); i++) {
4998 struct dpif_backer *backer = backers[i]->data;
4999 backer->need_revalidate = REV_RECONFIGURE;
5005 disable_tnl_push_pop(struct unixctl_conn *conn OVS_UNUSED, int argc OVS_UNUSED,
5006 const char *argv[], void *aux OVS_UNUSED)
5008 if (!strcasecmp(argv[1], "off")) {
5009 ofproto_use_tnl_push_pop = false;
5010 unixctl_command_reply(conn, "Tunnel push-pop off");
5011 ofproto_revalidate_all_backers();
5012 } else if (!strcasecmp(argv[1], "on")) {
5013 ofproto_use_tnl_push_pop = true;
5014 unixctl_command_reply(conn, "Tunnel push-pop on");
5015 ofproto_revalidate_all_backers();
5020 ofproto_unixctl_init(void)
5022 static bool registered;
5028 unixctl_command_register(
5030 "{[dp_name] odp_flow | bridge br_flow} [-generate|packet]",
5031 1, 3, ofproto_unixctl_trace, NULL);
5032 unixctl_command_register(
5033 "ofproto/trace-packet-out",
5034 "[-consistent] {[dp_name] odp_flow | bridge br_flow} [-generate|packet] actions",
5035 2, 6, ofproto_unixctl_trace_actions, NULL);
5036 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
5037 ofproto_unixctl_fdb_flush, NULL);
5038 unixctl_command_register("fdb/show", "bridge", 1, 1,
5039 ofproto_unixctl_fdb_show, NULL);
5040 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
5041 ofproto_unixctl_mcast_snooping_flush, NULL);
5042 unixctl_command_register("mdb/show", "bridge", 1, 1,
5043 ofproto_unixctl_mcast_snooping_show, NULL);
5044 unixctl_command_register("dpif/dump-dps", "", 0, 0,
5045 ofproto_unixctl_dpif_dump_dps, NULL);
5046 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show,
5048 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
5049 ofproto_unixctl_dpif_dump_flows, NULL);
5051 unixctl_command_register("ofproto/tnl-push-pop", "[on]|[off]", 1, 1,
5052 disable_tnl_push_pop, NULL);
5055 /* Returns true if 'table' is the table used for internal rules,
5056 * false otherwise. */
5058 table_is_internal(uint8_t table_id)
5060 return table_id == TBL_INTERNAL;
5063 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
5065 * This is deprecated. It is only for compatibility with broken device drivers
5066 * in old versions of Linux that do not properly support VLANs when VLAN
5067 * devices are not used. When broken device drivers are no longer in
5068 * widespread use, we will delete these interfaces. */
5071 set_realdev(struct ofport *ofport_, ofp_port_t realdev_ofp_port, int vid)
5073 struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport_->ofproto);
5074 struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
5076 if (realdev_ofp_port == ofport->realdev_ofp_port
5077 && vid == ofport->vlandev_vid) {
5081 ofproto->backer->need_revalidate = REV_RECONFIGURE;
5083 if (ofport->realdev_ofp_port) {
5086 if (realdev_ofp_port && ofport->bundle) {
5087 /* vlandevs are enslaved to their realdevs, so they are not allowed to
5088 * themselves be part of a bundle. */
5089 bundle_set(ofport_->ofproto, ofport->bundle, NULL);
5092 ofport->realdev_ofp_port = realdev_ofp_port;
5093 ofport->vlandev_vid = vid;
5095 if (realdev_ofp_port) {
5096 vsp_add(ofport, realdev_ofp_port, vid);
5103 hash_realdev_vid(ofp_port_t realdev_ofp_port, int vid)
5105 return hash_2words(ofp_to_u16(realdev_ofp_port), vid);
5109 ofproto_has_vlan_splinters(const struct ofproto_dpif *ofproto)
5110 OVS_EXCLUDED(ofproto->vsp_mutex)
5112 /* hmap_is_empty is thread safe. */
5113 return !hmap_is_empty(&ofproto->realdev_vid_map);
5118 vsp_realdev_to_vlandev__(const struct ofproto_dpif *ofproto,
5119 ofp_port_t realdev_ofp_port, ovs_be16 vlan_tci)
5120 OVS_REQUIRES(ofproto->vsp_mutex)
5122 if (!hmap_is_empty(&ofproto->realdev_vid_map)) {
5123 int vid = vlan_tci_to_vid(vlan_tci);
5124 const struct vlan_splinter *vsp;
5126 HMAP_FOR_EACH_WITH_HASH (vsp, realdev_vid_node,
5127 hash_realdev_vid(realdev_ofp_port, vid),
5128 &ofproto->realdev_vid_map) {
5129 if (vsp->realdev_ofp_port == realdev_ofp_port
5130 && vsp->vid == vid) {
5131 return vsp->vlandev_ofp_port;
5135 return realdev_ofp_port;
5138 /* Returns the OFP port number of the Linux VLAN device that corresponds to
5139 * 'vlan_tci' on the network device with port number 'realdev_ofp_port' in
5140 * 'struct ofport_dpif'. For example, given 'realdev_ofp_port' of eth0 and
5141 * 'vlan_tci' 9, it would return the port number of eth0.9.
5143 * Unless VLAN splinters are enabled for port 'realdev_ofp_port', this
5144 * function just returns its 'realdev_ofp_port' argument. */
5146 vsp_realdev_to_vlandev(const struct ofproto_dpif *ofproto,
5147 ofp_port_t realdev_ofp_port, ovs_be16 vlan_tci)
5148 OVS_EXCLUDED(ofproto->vsp_mutex)
5152 /* hmap_is_empty is thread safe, see if we can return immediately. */
5153 if (hmap_is_empty(&ofproto->realdev_vid_map)) {
5154 return realdev_ofp_port;
5156 ovs_mutex_lock(&ofproto->vsp_mutex);
5157 ret = vsp_realdev_to_vlandev__(ofproto, realdev_ofp_port, vlan_tci);
5158 ovs_mutex_unlock(&ofproto->vsp_mutex);
5162 static struct vlan_splinter *
5163 vlandev_find(const struct ofproto_dpif *ofproto, ofp_port_t vlandev_ofp_port)
5165 struct vlan_splinter *vsp;
5167 HMAP_FOR_EACH_WITH_HASH (vsp, vlandev_node,
5168 hash_ofp_port(vlandev_ofp_port),
5169 &ofproto->vlandev_map) {
5170 if (vsp->vlandev_ofp_port == vlandev_ofp_port) {
5178 /* Returns the OpenFlow port number of the "real" device underlying the Linux
5179 * VLAN device with OpenFlow port number 'vlandev_ofp_port' and stores the
5180 * VLAN VID of the Linux VLAN device in '*vid'. For example, given
5181 * 'vlandev_ofp_port' of eth0.9, it would return the OpenFlow port number of
5182 * eth0 and store 9 in '*vid'.
5184 * Returns 0 and does not modify '*vid' if 'vlandev_ofp_port' is not a Linux
5185 * VLAN device. Unless VLAN splinters are enabled, this is what this function
5188 vsp_vlandev_to_realdev(const struct ofproto_dpif *ofproto,
5189 ofp_port_t vlandev_ofp_port, int *vid)
5190 OVS_REQUIRES(ofproto->vsp_mutex)
5192 if (!hmap_is_empty(&ofproto->vlandev_map)) {
5193 const struct vlan_splinter *vsp;
5195 vsp = vlandev_find(ofproto, vlandev_ofp_port);
5200 return vsp->realdev_ofp_port;
5206 /* Given 'flow', a flow representing a packet received on 'ofproto', checks
5207 * whether 'flow->in_port' represents a Linux VLAN device. If so, changes
5208 * 'flow->in_port' to the "real" device backing the VLAN device, sets
5209 * 'flow->vlan_tci' to the VLAN VID, and returns true. Optionally pushes the
5210 * appropriate VLAN on 'packet' if provided. Otherwise (which is always the
5211 * case unless VLAN splinters are enabled), returns false without making any
5214 vsp_adjust_flow(const struct ofproto_dpif *ofproto, struct flow *flow,
5215 struct ofpbuf *packet)
5216 OVS_EXCLUDED(ofproto->vsp_mutex)
5221 /* hmap_is_empty is thread safe. */
5222 if (hmap_is_empty(&ofproto->vlandev_map)) {
5226 ovs_mutex_lock(&ofproto->vsp_mutex);
5227 realdev = vsp_vlandev_to_realdev(ofproto, flow->in_port.ofp_port, &vid);
5228 ovs_mutex_unlock(&ofproto->vsp_mutex);
5233 /* Cause the flow to be processed as if it came in on the real device with
5234 * the VLAN device's VLAN ID. */
5235 flow->in_port.ofp_port = realdev;
5236 flow->vlan_tci = htons((vid & VLAN_VID_MASK) | VLAN_CFI);
5239 /* Make the packet resemble the flow, so that it gets sent to an
5240 * OpenFlow controller properly, so that it looks correct for sFlow,
5241 * and so that flow_extract() will get the correct vlan_tci if it is
5242 * called on 'packet'. */
5243 eth_push_vlan(packet, htons(ETH_TYPE_VLAN), flow->vlan_tci);
5250 vsp_remove(struct ofport_dpif *port)
5252 struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
5253 struct vlan_splinter *vsp;
5255 ovs_mutex_lock(&ofproto->vsp_mutex);
5256 vsp = vlandev_find(ofproto, port->up.ofp_port);
5258 hmap_remove(&ofproto->vlandev_map, &vsp->vlandev_node);
5259 hmap_remove(&ofproto->realdev_vid_map, &vsp->realdev_vid_node);
5262 port->realdev_ofp_port = 0;
5264 VLOG_ERR("missing vlan device record");
5266 ovs_mutex_unlock(&ofproto->vsp_mutex);
5270 vsp_add(struct ofport_dpif *port, ofp_port_t realdev_ofp_port, int vid)
5272 struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
5274 ovs_mutex_lock(&ofproto->vsp_mutex);
5275 if (!vsp_vlandev_to_realdev(ofproto, port->up.ofp_port, NULL)
5276 && (vsp_realdev_to_vlandev__(ofproto, realdev_ofp_port, htons(vid))
5277 == realdev_ofp_port)) {
5278 struct vlan_splinter *vsp;
5280 vsp = xmalloc(sizeof *vsp);
5281 vsp->realdev_ofp_port = realdev_ofp_port;
5282 vsp->vlandev_ofp_port = port->up.ofp_port;
5285 port->realdev_ofp_port = realdev_ofp_port;
5287 hmap_insert(&ofproto->vlandev_map, &vsp->vlandev_node,
5288 hash_ofp_port(port->up.ofp_port));
5289 hmap_insert(&ofproto->realdev_vid_map, &vsp->realdev_vid_node,
5290 hash_realdev_vid(realdev_ofp_port, vid));
5292 VLOG_ERR("duplicate vlan device record");
5294 ovs_mutex_unlock(&ofproto->vsp_mutex);
5298 ofp_port_to_odp_port(const struct ofproto_dpif *ofproto, ofp_port_t ofp_port)
5300 const struct ofport_dpif *ofport = get_ofp_port(ofproto, ofp_port);
5301 return ofport ? ofport->odp_port : ODPP_NONE;
5304 struct ofport_dpif *
5305 odp_port_to_ofport(const struct dpif_backer *backer, odp_port_t odp_port)
5307 struct ofport_dpif *port;
5309 ovs_rwlock_rdlock(&backer->odp_to_ofport_lock);
5310 HMAP_FOR_EACH_IN_BUCKET (port, odp_port_node, hash_odp_port(odp_port),
5311 &backer->odp_to_ofport_map) {
5312 if (port->odp_port == odp_port) {
5313 ovs_rwlock_unlock(&backer->odp_to_ofport_lock);
5318 ovs_rwlock_unlock(&backer->odp_to_ofport_lock);
5323 odp_port_to_ofp_port(const struct ofproto_dpif *ofproto, odp_port_t odp_port)
5325 struct ofport_dpif *port;
5327 port = odp_port_to_ofport(ofproto->backer, odp_port);
5328 if (port && &ofproto->up == port->up.ofproto) {
5329 return port->up.ofp_port;
5336 ofproto_dpif_alloc_recirc_id(struct ofproto_dpif *ofproto)
5338 struct dpif_backer *backer = ofproto->backer;
5340 return recirc_id_alloc(backer->rid_pool);
5344 ofproto_dpif_free_recirc_id(struct ofproto_dpif *ofproto, uint32_t recirc_id)
5346 struct dpif_backer *backer = ofproto->backer;
5348 recirc_id_free(backer->rid_pool, recirc_id);
5352 ofproto_dpif_add_internal_flow(struct ofproto_dpif *ofproto,
5353 const struct match *match, int priority,
5354 uint16_t idle_timeout,
5355 const struct ofpbuf *ofpacts,
5356 struct rule **rulep)
5358 struct ofputil_flow_mod fm;
5359 struct rule_dpif *rule;
5363 fm.priority = priority;
5364 fm.new_cookie = htonll(0);
5365 fm.cookie = htonll(0);
5366 fm.cookie_mask = htonll(0);
5367 fm.modify_cookie = false;
5368 fm.table_id = TBL_INTERNAL;
5369 fm.command = OFPFC_ADD;
5370 fm.idle_timeout = idle_timeout;
5371 fm.hard_timeout = 0;
5375 fm.flags = OFPUTIL_FF_HIDDEN_FIELDS | OFPUTIL_FF_NO_READONLY;
5376 fm.ofpacts = ofpbuf_data(ofpacts);
5377 fm.ofpacts_len = ofpbuf_size(ofpacts);
5379 error = ofproto_flow_mod(&ofproto->up, &fm);
5381 VLOG_ERR_RL(&rl, "failed to add internal flow (%s)",
5382 ofperr_to_string(error));
5387 rule = rule_dpif_lookup_in_table(ofproto, TBL_INTERNAL, &fm.match.flow,
5388 &fm.match.wc, false);
5398 ofproto_dpif_delete_internal_flow(struct ofproto_dpif *ofproto,
5399 struct match *match, int priority)
5401 struct ofputil_flow_mod fm;
5405 fm.priority = priority;
5406 fm.new_cookie = htonll(0);
5407 fm.cookie = htonll(0);
5408 fm.cookie_mask = htonll(0);
5409 fm.modify_cookie = false;
5410 fm.table_id = TBL_INTERNAL;
5411 fm.flags = OFPUTIL_FF_HIDDEN_FIELDS | OFPUTIL_FF_NO_READONLY;
5412 fm.command = OFPFC_DELETE_STRICT;
5414 error = ofproto_flow_mod(&ofproto->up, &fm);
5416 VLOG_ERR_RL(&rl, "failed to delete internal flow (%s)",
5417 ofperr_to_string(error));
5424 const struct ofproto_class ofproto_dpif_class = {
5438 NULL, /* get_memory_usage. */
5439 type_get_memory_usage,
5457 port_is_lacp_current,
5458 port_get_lacp_stats,
5459 NULL, /* rule_choose_table */
5468 NULL, /* rule_premodify_actions */
5469 rule_modify_actions,
5485 get_stp_port_status,
5490 get_rstp_port_status,
5497 is_mirror_output_bundle,
5498 forward_bpdu_changed,
5499 set_mac_table_config,
5501 set_mcast_snooping_port,
5503 NULL, /* meter_get_features */
5504 NULL, /* meter_set */
5505 NULL, /* meter_get */
5506 NULL, /* meter_del */
5507 group_alloc, /* group_alloc */
5508 group_construct, /* group_construct */
5509 group_destruct, /* group_destruct */
5510 group_dealloc, /* group_dealloc */
5511 group_modify, /* group_modify */
5512 group_get_stats, /* group_get_stats */
5513 get_datapath_version, /* get_datapath_version */