1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
71 /* OpenFlow port slated for removal from ofproto. */
73 struct list list_node; /* Node in bridge's ofpp_garbage. */
74 uint16_t ofp_port; /* Port to be deleted. */
78 /* These members are always valid. */
79 struct list port_elem; /* Element in struct port's "ifaces" list. */
80 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
84 /* These members are valid only after bridge_reconfigure() causes them to
86 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
87 int ofp_port; /* OpenFlow port number, -1 if unknown. */
88 struct netdev *netdev; /* Network device. */
89 const char *type; /* Usually same as cfg->type. */
90 const struct ovsrec_interface *cfg;
94 struct uuid uuid; /* UUID of this "mirror" record in database. */
95 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
96 struct bridge *bridge;
98 const struct ovsrec_mirror *cfg;
102 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
103 struct bridge *bridge;
106 const struct ovsrec_port *cfg;
108 /* An ordinary bridge port has 1 interface.
109 * A bridge port for bonding has at least 2 interfaces. */
110 struct list ifaces; /* List of "struct iface"s. */
114 struct hmap_node node; /* In 'all_bridges'. */
115 char *name; /* User-specified arbitrary name. */
116 char *type; /* Datapath type. */
117 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
118 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
119 const struct ovsrec_bridge *cfg;
121 /* OpenFlow switch processing. */
122 struct ofproto *ofproto; /* OpenFlow switch. */
125 struct hmap ports; /* "struct port"s indexed by name. */
126 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
127 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
129 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
130 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
131 Indexed on 'cfg->name'. */
133 /* Port mirroring. */
134 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
136 /* Synthetic local port if necessary. */
137 struct ovsrec_port synth_local_port;
138 struct ovsrec_interface synth_local_iface;
139 struct ovsrec_interface *synth_local_ifacep;
142 /* All bridges, indexed by name. */
143 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
145 /* OVSDB IDL used to obtain configuration. */
146 static struct ovsdb_idl *idl;
148 /* Most recently processed IDL sequence number. */
149 static unsigned int idl_seqno;
151 /* Each time this timer expires, the bridge fetches systems and interface
152 * statistics and pushes them into the database. */
153 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int stats_timer = LLONG_MIN;
156 /* Stores the time after which rate limited statistics may be written to the
157 * database. Only updated when changes to the database require rate limiting.
159 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
160 static long long int db_limiter = LLONG_MIN;
162 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
163 * expensive. This can cause bridge_reconfigure() to take a long time during
164 * which no other work can be done. To deal with this problem, we limit port
165 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
166 * call to bridge_reconfigure(). If there is more work to do after the limit
167 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
168 * This allows the rest of the code to catch up on important things like
169 * forwarding packets. */
170 #define OFP_PORT_ACTION_WINDOW 10
171 static bool reconfiguring = false;
173 static void add_del_bridges(const struct ovsrec_open_vswitch *);
174 static void bridge_update_ofprotos(void);
175 static void bridge_create(const struct ovsrec_bridge *);
176 static void bridge_destroy(struct bridge *);
177 static struct bridge *bridge_lookup(const char *name);
178 static unixctl_cb_func bridge_unixctl_dump_flows;
179 static unixctl_cb_func bridge_unixctl_reconnect;
180 static size_t bridge_get_controllers(const struct bridge *br,
181 struct ovsrec_controller ***controllersp);
182 static void bridge_add_del_ports(struct bridge *,
183 const unsigned long int *splinter_vlans);
184 static void bridge_refresh_ofp_port(struct bridge *);
185 static void bridge_configure_datapath_id(struct bridge *);
186 static void bridge_configure_flow_eviction_threshold(struct bridge *);
187 static void bridge_configure_netflow(struct bridge *);
188 static void bridge_configure_forward_bpdu(struct bridge *);
189 static void bridge_configure_mac_idle_time(struct bridge *);
190 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
191 static void bridge_configure_stp(struct bridge *);
192 static void bridge_configure_tables(struct bridge *);
193 static void bridge_configure_remotes(struct bridge *,
194 const struct sockaddr_in *managers,
196 static void bridge_pick_local_hw_addr(struct bridge *,
197 uint8_t ea[ETH_ADDR_LEN],
198 struct iface **hw_addr_iface);
199 static uint64_t bridge_pick_datapath_id(struct bridge *,
200 const uint8_t bridge_ea[ETH_ADDR_LEN],
201 struct iface *hw_addr_iface);
202 static void bridge_queue_if_cfg(struct bridge *,
203 const struct ovsrec_interface *,
204 const struct ovsrec_port *);
205 static uint64_t dpid_from_hash(const void *, size_t nbytes);
206 static bool bridge_has_bond_fake_iface(const struct bridge *,
208 static bool port_is_bond_fake_iface(const struct port *);
210 static unixctl_cb_func qos_unixctl_show;
212 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
213 static void port_del_ifaces(struct port *);
214 static void port_destroy(struct port *);
215 static struct port *port_lookup(const struct bridge *, const char *name);
216 static void port_configure(struct port *);
217 static struct lacp_settings *port_configure_lacp(struct port *,
218 struct lacp_settings *);
219 static void port_configure_bond(struct port *, struct bond_settings *,
220 uint32_t *bond_stable_ids);
221 static bool port_is_synthetic(const struct port *);
223 static void bridge_configure_mirrors(struct bridge *);
224 static struct mirror *mirror_create(struct bridge *,
225 const struct ovsrec_mirror *);
226 static void mirror_destroy(struct mirror *);
227 static bool mirror_configure(struct mirror *);
228 static void mirror_refresh_stats(struct mirror *);
230 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
231 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
232 static const char *iface_get_type(const struct ovsrec_interface *,
233 const struct ovsrec_bridge *);
234 static void iface_destroy(struct iface *);
235 static struct iface *iface_lookup(const struct bridge *, const char *name);
236 static struct iface *iface_find(const char *name);
237 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
238 static struct iface *iface_from_ofp_port(const struct bridge *,
240 static void iface_set_mac(struct iface *);
241 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
242 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
243 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
244 static void iface_configure_cfm(struct iface *);
245 static void iface_refresh_cfm_stats(struct iface *);
246 static void iface_refresh_stats(struct iface *);
247 static void iface_refresh_status(struct iface *);
248 static bool iface_is_synthetic(const struct iface *);
250 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
252 * This is deprecated. It is only for compatibility with broken device drivers
253 * in old versions of Linux that do not properly support VLANs when VLAN
254 * devices are not used. When broken device drivers are no longer in
255 * widespread use, we will delete these interfaces. */
257 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
258 static bool vlan_splinters_enabled_anywhere;
260 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
261 static unsigned long int *collect_splinter_vlans(
262 const struct ovsrec_open_vswitch *);
263 static void configure_splinter_port(struct port *);
264 static void add_vlan_splinter_ports(struct bridge *,
265 const unsigned long int *splinter_vlans,
266 struct shash *ports);
268 /* Public functions. */
270 /* Initializes the bridge module, configuring it to obtain its configuration
271 * from an OVSDB server accessed over 'remote', which should be a string in a
272 * form acceptable to ovsdb_idl_create(). */
274 bridge_init(const char *remote)
276 /* Create connection to database. */
277 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
278 idl_seqno = ovsdb_idl_get_seqno(idl);
279 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
281 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
282 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
289 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
290 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
291 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
293 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
294 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
295 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
312 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
314 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
315 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
316 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
317 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
319 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
321 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
323 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
324 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
326 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
338 /* Register unixctl commands. */
339 unixctl_command_register("qos/show", "interface", 1, 1,
340 qos_unixctl_show, NULL);
341 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
342 bridge_unixctl_dump_flows, NULL);
343 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
344 bridge_unixctl_reconnect, NULL);
354 struct bridge *br, *next_br;
356 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
359 ovsdb_idl_destroy(idl);
362 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
363 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
364 * responsible for freeing '*managersp' (with free()).
366 * You may be asking yourself "why does ovs-vswitchd care?", because
367 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
368 * should not be and in fact is not directly involved in that. But
369 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
370 * it has to tell in-band control where the managers are to enable that.
371 * (Thus, only managers connected in-band are collected.)
374 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
375 struct sockaddr_in **managersp, size_t *n_managersp)
377 struct sockaddr_in *managers = NULL;
378 size_t n_managers = 0;
382 /* Collect all of the potential targets from the "targets" columns of the
383 * rows pointed to by "manager_options", excluding any that are
386 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
387 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
389 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
390 sset_find_and_delete(&targets, m->target);
392 sset_add(&targets, m->target);
396 /* Now extract the targets' IP addresses. */
397 if (!sset_is_empty(&targets)) {
400 managers = xmalloc(sset_count(&targets) * sizeof *managers);
401 SSET_FOR_EACH (target, &targets) {
402 struct sockaddr_in *sin = &managers[n_managers];
404 if (stream_parse_target_with_default_ports(target,
412 sset_destroy(&targets);
414 *managersp = managers;
415 *n_managersp = n_managers;
419 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
421 unsigned long int *splinter_vlans;
424 COVERAGE_INC(bridge_reconfigure);
426 assert(!reconfiguring);
427 reconfiguring = true;
429 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
430 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
431 * configuration otherwise.
433 * This is mostly an update to bridge data structures. Nothing is pushed
434 * down to ofproto or lower layers. */
435 add_del_bridges(ovs_cfg);
436 splinter_vlans = collect_splinter_vlans(ovs_cfg);
437 HMAP_FOR_EACH (br, node, &all_bridges) {
438 bridge_add_del_ports(br, splinter_vlans);
440 free(splinter_vlans);
442 /* Delete datapaths that are no longer configured, and create ones which
443 * don't exist but should. */
444 bridge_update_ofprotos();
446 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
447 HMAP_FOR_EACH (br, node, &all_bridges) {
448 bridge_refresh_ofp_port(br);
451 /* Clear database records for "if_cfg"s which haven't been instantiated. */
452 HMAP_FOR_EACH (br, node, &all_bridges) {
453 struct if_cfg *if_cfg;
455 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
456 iface_clear_db_record(if_cfg->cfg);
462 bridge_reconfigure_ofp(void)
464 long long int deadline;
468 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
470 /* The kernel will reject any attempt to add a given port to a datapath if
471 * that port already belongs to a different datapath, so we must do all
472 * port deletions before any port additions. */
473 HMAP_FOR_EACH (br, node, &all_bridges) {
474 struct ofpp_garbage *garbage, *next;
476 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
477 ofproto_port_del(br->ofproto, garbage->ofp_port);
478 list_remove(&garbage->list_node);
482 if (time_msec() >= deadline) {
488 HMAP_FOR_EACH (br, node, &all_bridges) {
489 struct if_cfg *if_cfg, *next;
491 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
492 iface_create(br, if_cfg, -1);
494 if (time_msec() >= deadline) {
504 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
506 struct sockaddr_in *managers;
507 int sflow_bridge_number;
512 assert(reconfiguring);
513 done = bridge_reconfigure_ofp();
515 /* Complete the configuration. */
516 sflow_bridge_number = 0;
517 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
518 HMAP_FOR_EACH (br, node, &all_bridges) {
521 /* We need the datapath ID early to allow LACP ports to use it as the
522 * default system ID. */
523 bridge_configure_datapath_id(br);
525 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
528 port_configure(port);
530 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
531 iface_configure_cfm(iface);
532 iface_configure_qos(iface, port->cfg->qos);
533 iface_set_mac(iface);
536 bridge_configure_mirrors(br);
537 bridge_configure_flow_eviction_threshold(br);
538 bridge_configure_forward_bpdu(br);
539 bridge_configure_mac_idle_time(br);
540 bridge_configure_remotes(br, managers, n_managers);
541 bridge_configure_netflow(br);
542 bridge_configure_sflow(br, &sflow_bridge_number);
543 bridge_configure_stp(br);
544 bridge_configure_tables(br);
549 /* ovs-vswitchd has completed initialization, so allow the process that
550 * forked us to exit successfully. */
551 daemonize_complete();
552 reconfiguring = false;
558 /* Delete ofprotos which aren't configured or have the wrong type. Create
559 * ofprotos which don't exist but need to. */
561 bridge_update_ofprotos(void)
563 struct bridge *br, *next;
568 /* Delete ofprotos with no bridge or with the wrong type. */
571 ofproto_enumerate_types(&types);
572 SSET_FOR_EACH (type, &types) {
575 ofproto_enumerate_names(type, &names);
576 SSET_FOR_EACH (name, &names) {
577 br = bridge_lookup(name);
578 if (!br || strcmp(type, br->type)) {
579 ofproto_delete(name, type);
583 sset_destroy(&names);
584 sset_destroy(&types);
586 /* Add ofprotos for bridges which don't have one yet. */
587 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
595 /* Remove ports from any datapath with the same name as 'br'. If we
596 * don't do this, creating 'br''s ofproto will fail because a port with
597 * the same name as its local port already exists. */
598 HMAP_FOR_EACH (br2, node, &all_bridges) {
599 struct ofproto_port ofproto_port;
605 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
607 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
609 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
612 ofproto_port_destroy(&ofproto_port);
616 error = ofproto_create(br->name, br->type, &br->ofproto);
618 VLOG_ERR("failed to create bridge %s: %s", br->name,
626 port_configure(struct port *port)
628 const struct ovsrec_port *cfg = port->cfg;
629 struct bond_settings bond_settings;
630 struct lacp_settings lacp_settings;
631 struct ofproto_bundle_settings s;
634 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
635 configure_splinter_port(port);
644 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
645 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
646 s.slaves[s.n_slaves++] = iface->ofp_port;
651 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
655 /* Get VLAN trunks. */
658 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
662 if (cfg->vlan_mode) {
663 if (!strcmp(cfg->vlan_mode, "access")) {
664 s.vlan_mode = PORT_VLAN_ACCESS;
665 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
666 s.vlan_mode = PORT_VLAN_TRUNK;
667 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
668 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
669 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
670 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
672 /* This "can't happen" because ovsdb-server should prevent it. */
673 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
674 s.vlan_mode = PORT_VLAN_TRUNK;
678 s.vlan_mode = PORT_VLAN_ACCESS;
680 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
684 s.vlan_mode = PORT_VLAN_TRUNK;
687 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
690 /* Get LACP settings. */
691 s.lacp = port_configure_lacp(port, &lacp_settings);
695 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
696 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
697 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
700 s.lacp_slaves = NULL;
703 /* Get bond settings. */
704 if (s.n_slaves > 1) {
705 s.bond = &bond_settings;
706 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
707 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
710 s.bond_stable_ids = NULL;
712 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
713 netdev_set_miimon_interval(iface->netdev, 0);
718 ofproto_bundle_register(port->bridge->ofproto, port, &s);
724 free(s.bond_stable_ids);
727 /* Pick local port hardware address and datapath ID for 'br'. */
729 bridge_configure_datapath_id(struct bridge *br)
731 uint8_t ea[ETH_ADDR_LEN];
733 struct iface *local_iface;
734 struct iface *hw_addr_iface;
737 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
738 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
740 int error = netdev_set_etheraddr(local_iface->netdev, ea);
742 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
743 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
744 "Ethernet address: %s",
745 br->name, strerror(error));
748 memcpy(br->ea, ea, ETH_ADDR_LEN);
750 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
751 ofproto_set_datapath_id(br->ofproto, dpid);
753 dpid_string = xasprintf("%016"PRIx64, dpid);
754 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
758 /* Set NetFlow configuration on 'br'. */
760 bridge_configure_netflow(struct bridge *br)
762 struct ovsrec_netflow *cfg = br->cfg->netflow;
763 struct netflow_options opts;
766 ofproto_set_netflow(br->ofproto, NULL);
770 memset(&opts, 0, sizeof opts);
772 /* Get default NetFlow configuration from datapath.
773 * Apply overrides from 'cfg'. */
774 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
775 if (cfg->engine_type) {
776 opts.engine_type = *cfg->engine_type;
778 if (cfg->engine_id) {
779 opts.engine_id = *cfg->engine_id;
782 /* Configure active timeout interval. */
783 opts.active_timeout = cfg->active_timeout;
784 if (!opts.active_timeout) {
785 opts.active_timeout = -1;
786 } else if (opts.active_timeout < 0) {
787 VLOG_WARN("bridge %s: active timeout interval set to negative "
788 "value, using default instead (%d seconds)", br->name,
789 NF_ACTIVE_TIMEOUT_DEFAULT);
790 opts.active_timeout = -1;
793 /* Add engine ID to interface number to disambiguate bridgs? */
794 opts.add_id_to_iface = cfg->add_id_to_interface;
795 if (opts.add_id_to_iface) {
796 if (opts.engine_id > 0x7f) {
797 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
798 "another vswitch, choose an engine id less than 128",
801 if (hmap_count(&br->ports) > 508) {
802 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
803 "another port when more than 508 ports are used",
809 sset_init(&opts.collectors);
810 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
813 if (ofproto_set_netflow(br->ofproto, &opts)) {
814 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
816 sset_destroy(&opts.collectors);
819 /* Set sFlow configuration on 'br'. */
821 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
823 const struct ovsrec_sflow *cfg = br->cfg->sflow;
824 struct ovsrec_controller **controllers;
825 struct ofproto_sflow_options oso;
826 size_t n_controllers;
830 ofproto_set_sflow(br->ofproto, NULL);
834 memset(&oso, 0, sizeof oso);
836 sset_init(&oso.targets);
837 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
839 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
841 oso.sampling_rate = *cfg->sampling;
844 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
846 oso.polling_interval = *cfg->polling;
849 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
851 oso.header_len = *cfg->header;
854 oso.sub_id = (*sflow_bridge_number)++;
855 oso.agent_device = cfg->agent;
857 oso.control_ip = NULL;
858 n_controllers = bridge_get_controllers(br, &controllers);
859 for (i = 0; i < n_controllers; i++) {
860 if (controllers[i]->local_ip) {
861 oso.control_ip = controllers[i]->local_ip;
865 ofproto_set_sflow(br->ofproto, &oso);
867 sset_destroy(&oso.targets);
871 port_configure_stp(const struct ofproto *ofproto, struct port *port,
872 struct ofproto_port_stp_settings *port_s,
873 int *port_num_counter, unsigned long *port_num_bitmap)
875 const char *config_str;
878 if (smap_get_bool(&port->cfg->other_config, "stp-enable", false)) {
879 port_s->enable = false;
882 port_s->enable = true;
885 /* STP over bonds is not supported. */
886 if (!list_is_singleton(&port->ifaces)) {
887 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
889 port_s->enable = false;
893 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
895 /* Internal ports shouldn't participate in spanning tree, so
897 if (!strcmp(iface->type, "internal")) {
898 VLOG_DBG("port %s: disable STP on internal ports", port->name);
899 port_s->enable = false;
903 /* STP on mirror output ports is not supported. */
904 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
905 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
906 port_s->enable = false;
910 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
912 unsigned long int port_num = strtoul(config_str, NULL, 0);
913 int port_idx = port_num - 1;
915 if (port_num < 1 || port_num > STP_MAX_PORTS) {
916 VLOG_ERR("port %s: invalid stp-port-num", port->name);
917 port_s->enable = false;
921 if (bitmap_is_set(port_num_bitmap, port_idx)) {
922 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
923 port->name, port_num);
924 port_s->enable = false;
927 bitmap_set1(port_num_bitmap, port_idx);
928 port_s->port_num = port_idx;
930 if (*port_num_counter > STP_MAX_PORTS) {
931 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
932 port_s->enable = false;
936 port_s->port_num = (*port_num_counter)++;
939 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
941 port_s->path_cost = strtoul(config_str, NULL, 10);
943 enum netdev_features current;
945 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
946 /* Couldn't get speed, so assume 100Mb/s. */
947 port_s->path_cost = 19;
951 mbps = netdev_features_to_bps(current) / 1000000;
952 port_s->path_cost = stp_convert_speed_to_cost(mbps);
956 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
958 port_s->priority = strtoul(config_str, NULL, 0);
960 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
964 /* Set spanning tree configuration on 'br'. */
966 bridge_configure_stp(struct bridge *br)
968 if (!br->cfg->stp_enable) {
969 ofproto_set_stp(br->ofproto, NULL);
971 struct ofproto_stp_settings br_s;
972 const char *config_str;
974 int port_num_counter;
975 unsigned long *port_num_bitmap;
977 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
979 uint8_t ea[ETH_ADDR_LEN];
981 if (eth_addr_from_string(config_str, ea)) {
982 br_s.system_id = eth_addr_to_uint64(ea);
984 br_s.system_id = eth_addr_to_uint64(br->ea);
985 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
986 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
989 br_s.system_id = eth_addr_to_uint64(br->ea);
992 config_str = smap_get(&br->cfg->other_config, "stp-priority");
994 br_s.priority = strtoul(config_str, NULL, 0);
996 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
999 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1001 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1003 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1006 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1008 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1010 br_s.max_age = STP_DEFAULT_MAX_AGE;
1013 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1015 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1017 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1020 /* Configure STP on the bridge. */
1021 if (ofproto_set_stp(br->ofproto, &br_s)) {
1022 VLOG_ERR("bridge %s: could not enable STP", br->name);
1026 /* Users must either set the port number with the "stp-port-num"
1027 * configuration on all ports or none. If manual configuration
1028 * is not done, then we allocate them sequentially. */
1029 port_num_counter = 0;
1030 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1031 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1032 struct ofproto_port_stp_settings port_s;
1033 struct iface *iface;
1035 port_configure_stp(br->ofproto, port, &port_s,
1036 &port_num_counter, port_num_bitmap);
1038 /* As bonds are not supported, just apply configuration to
1039 * all interfaces. */
1040 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1041 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1043 VLOG_ERR("port %s: could not enable STP", port->name);
1049 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1050 && port_num_counter) {
1051 VLOG_ERR("bridge %s: must manually configure all STP port "
1052 "IDs or none, disabling", br->name);
1053 ofproto_set_stp(br->ofproto, NULL);
1055 bitmap_free(port_num_bitmap);
1060 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1062 const struct port *port = port_lookup(br, name);
1063 return port && port_is_bond_fake_iface(port);
1067 port_is_bond_fake_iface(const struct port *port)
1069 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1073 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1075 struct bridge *br, *next;
1076 struct shash new_br;
1079 /* Collect new bridges' names and types. */
1080 shash_init(&new_br);
1081 for (i = 0; i < cfg->n_bridges; i++) {
1082 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1083 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1085 if (strchr(br_cfg->name, '/')) {
1086 /* Prevent remote ovsdb-server users from accessing arbitrary
1087 * directories, e.g. consider a bridge named "../../../etc/". */
1088 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1090 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1091 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1095 /* Get rid of deleted bridges or those whose types have changed.
1096 * Update 'cfg' of bridges that still exist. */
1097 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1098 br->cfg = shash_find_data(&new_br, br->name);
1099 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1100 br->cfg->datapath_type))) {
1105 /* Add new bridges. */
1106 for (i = 0; i < cfg->n_bridges; i++) {
1107 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1108 struct bridge *br = bridge_lookup(br_cfg->name);
1110 bridge_create(br_cfg);
1114 shash_destroy(&new_br);
1118 iface_set_ofp_port(struct iface *iface, int ofp_port)
1120 struct bridge *br = iface->port->bridge;
1122 assert(iface->ofp_port < 0 && ofp_port >= 0);
1123 iface->ofp_port = ofp_port;
1124 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1125 iface_set_ofport(iface->cfg, ofp_port);
1128 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1129 * Returns 0 if successful, otherwise a positive errno value. */
1131 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1132 struct netdev *netdev)
1136 error = netdev_set_config(netdev, &iface_cfg->options);
1138 VLOG_WARN("could not configure network device %s (%s)",
1139 iface_cfg->name, strerror(error));
1144 /* This function determines whether 'ofproto_port', which is attached to
1145 * br->ofproto's datapath, is one that we want in 'br'.
1147 * If it is, it returns true, after creating an iface (if necessary),
1148 * configuring the iface's netdev according to the iface's options, and setting
1149 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1151 * If, on the other hand, 'port' should be removed, it returns false. The
1152 * caller should later detach the port from br->ofproto. */
1154 bridge_refresh_one_ofp_port(struct bridge *br,
1155 const struct ofproto_port *ofproto_port)
1157 const char *name = ofproto_port->name;
1158 const char *type = ofproto_port->type;
1159 uint16_t ofp_port = ofproto_port->ofp_port;
1161 struct iface *iface = iface_lookup(br, name);
1163 /* Check that the name-to-number mapping is one-to-one. */
1164 if (iface->ofp_port >= 0) {
1165 VLOG_WARN("bridge %s: interface %s reported twice",
1168 } else if (iface_from_ofp_port(br, ofp_port)) {
1169 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1170 br->name, ofp_port);
1174 /* There's a configured interface named 'name'. */
1175 if (strcmp(type, iface->type)
1176 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1177 /* It's the wrong type, or it's the right type but can't be
1178 * configured as the user requested, so we must destroy it. */
1181 /* It's the right type and configured correctly. keep it. */
1182 iface_set_ofp_port(iface, ofp_port);
1185 } else if (bridge_has_bond_fake_iface(br, name)
1186 && !strcmp(type, "internal")) {
1187 /* It's a bond fake iface. Keep it. */
1190 /* There's no configured interface named 'name', but there might be an
1191 * interface of that name queued to be created.
1193 * If there is, and it has the correct type, then try to configure it
1194 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1195 * delete it and later try to re-add it. */
1196 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1198 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1199 && iface_create(br, if_cfg, ofp_port));
1203 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1204 * consistent with the ofp_ports in "br->ofproto". */
1206 bridge_refresh_ofp_port(struct bridge *br)
1208 struct ofproto_port_dump dump;
1209 struct ofproto_port ofproto_port;
1210 struct port *port, *port_next;
1212 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1213 hmap_clear(&br->ifaces);
1214 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1215 struct iface *iface;
1217 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1218 iface->ofp_port = -1;
1222 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1223 * already exist in the datapath and promote them to full fledged "struct
1224 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1225 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1226 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1227 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1228 garbage->ofp_port = ofproto_port.ofp_port;
1229 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1233 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1234 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1235 * later they can be added to ofproto. */
1236 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1237 struct iface *iface, *iface_next;
1239 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1240 if (iface->ofp_port < 0) {
1241 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1242 iface_destroy(iface);
1246 if (list_is_empty(&port->ifaces)) {
1252 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1253 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1254 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1257 * If successful, returns 0 and stores the network device in '*netdevp'. On
1258 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1260 iface_do_create(const struct bridge *br,
1261 const struct ovsrec_interface *iface_cfg,
1262 const struct ovsrec_port *port_cfg,
1263 int *ofp_portp, struct netdev **netdevp)
1265 struct netdev *netdev;
1268 error = netdev_open(iface_cfg->name,
1269 iface_get_type(iface_cfg, br->cfg), &netdev);
1271 VLOG_WARN("could not open network device %s (%s)",
1272 iface_cfg->name, strerror(error));
1276 error = iface_set_netdev_config(iface_cfg, netdev);
1281 if (*ofp_portp < 0) {
1284 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1288 *ofp_portp = ofp_port;
1290 VLOG_INFO("bridge %s: added interface %s on port %d",
1291 br->name, iface_cfg->name, *ofp_portp);
1293 VLOG_DBG("bridge %s: interface %s is on port %d",
1294 br->name, iface_cfg->name, *ofp_portp);
1297 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1298 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1306 netdev_close(netdev);
1310 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1311 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1312 * automatically allocated for the iface. Takes ownership of and
1313 * deallocates 'if_cfg'.
1315 * Return true if an iface is successfully created, false otherwise. */
1317 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1319 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1320 const struct ovsrec_port *port_cfg = if_cfg->parent;
1322 struct netdev *netdev;
1323 struct iface *iface;
1327 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1329 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1332 /* Do the bits that can fail up front. */
1333 assert(!iface_lookup(br, iface_cfg->name));
1334 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1336 iface_clear_db_record(iface_cfg);
1340 /* Get or create the port structure. */
1341 port = port_lookup(br, port_cfg->name);
1343 port = port_create(br, port_cfg);
1346 /* Create the iface structure. */
1347 iface = xzalloc(sizeof *iface);
1348 list_push_back(&port->ifaces, &iface->port_elem);
1349 hmap_insert(&br->iface_by_name, &iface->name_node,
1350 hash_string(iface_cfg->name, 0));
1352 iface->name = xstrdup(iface_cfg->name);
1353 iface->ofp_port = -1;
1354 iface->netdev = netdev;
1355 iface->type = iface_get_type(iface_cfg, br->cfg);
1356 iface->cfg = iface_cfg;
1358 iface_set_ofp_port(iface, ofp_port);
1360 /* Populate initial status in database. */
1361 iface_refresh_stats(iface);
1362 iface_refresh_status(iface);
1364 /* Add bond fake iface if necessary. */
1365 if (port_is_bond_fake_iface(port)) {
1366 struct ofproto_port ofproto_port;
1368 if (ofproto_port_query_by_name(br->ofproto, port->name,
1370 struct netdev *netdev;
1373 error = netdev_open(port->name, "internal", &netdev);
1375 ofproto_port_add(br->ofproto, netdev, NULL);
1376 netdev_close(netdev);
1378 VLOG_WARN("could not open network device %s (%s)",
1379 port->name, strerror(error));
1382 /* Already exists, nothing to do. */
1383 ofproto_port_destroy(&ofproto_port);
1390 /* Set Flow eviction threshold */
1392 bridge_configure_flow_eviction_threshold(struct bridge *br)
1394 const char *threshold_str;
1397 threshold_str = smap_get(&br->cfg->other_config,
1398 "flow-eviction-threshold");
1399 if (threshold_str) {
1400 threshold = strtoul(threshold_str, NULL, 10);
1402 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1404 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1407 /* Set forward BPDU option. */
1409 bridge_configure_forward_bpdu(struct bridge *br)
1411 ofproto_set_forward_bpdu(br->ofproto,
1412 smap_get_bool(&br->cfg->other_config,
1417 /* Set MAC aging time for 'br'. */
1419 bridge_configure_mac_idle_time(struct bridge *br)
1421 const char *idle_time_str;
1424 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1425 idle_time = (idle_time_str && atoi(idle_time_str)
1426 ? atoi(idle_time_str)
1427 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1428 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1432 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1433 struct iface **hw_addr_iface)
1435 struct hmapx mirror_output_ports;
1438 bool found_addr = false;
1442 *hw_addr_iface = NULL;
1444 /* Did the user request a particular MAC? */
1445 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1446 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1447 if (eth_addr_is_multicast(ea)) {
1448 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1449 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1450 } else if (eth_addr_is_zero(ea)) {
1451 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1457 /* Mirror output ports don't participate in picking the local hardware
1458 * address. ofproto can't help us find out whether a given port is a
1459 * mirror output because we haven't configured mirrors yet, so we need to
1460 * accumulate them ourselves. */
1461 hmapx_init(&mirror_output_ports);
1462 for (i = 0; i < br->cfg->n_mirrors; i++) {
1463 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1464 if (m->output_port) {
1465 hmapx_add(&mirror_output_ports, m->output_port);
1469 /* Otherwise choose the minimum non-local MAC address among all of the
1471 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1472 uint8_t iface_ea[ETH_ADDR_LEN];
1473 struct iface *candidate;
1474 struct iface *iface;
1476 /* Mirror output ports don't participate. */
1477 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1481 /* Choose the MAC address to represent the port. */
1483 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1484 /* Find the interface with this Ethernet address (if any) so that
1485 * we can provide the correct devname to the caller. */
1486 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1487 uint8_t candidate_ea[ETH_ADDR_LEN];
1488 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1489 && eth_addr_equals(iface_ea, candidate_ea)) {
1494 /* Choose the interface whose MAC address will represent the port.
1495 * The Linux kernel bonding code always chooses the MAC address of
1496 * the first slave added to a bond, and the Fedora networking
1497 * scripts always add slaves to a bond in alphabetical order, so
1498 * for compatibility we choose the interface with the name that is
1499 * first in alphabetical order. */
1500 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1501 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1506 /* The local port doesn't count (since we're trying to choose its
1507 * MAC address anyway). */
1508 if (iface->ofp_port == OFPP_LOCAL) {
1513 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1519 /* Compare against our current choice. */
1520 if (!eth_addr_is_multicast(iface_ea) &&
1521 !eth_addr_is_local(iface_ea) &&
1522 !eth_addr_is_reserved(iface_ea) &&
1523 !eth_addr_is_zero(iface_ea) &&
1524 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1526 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1527 *hw_addr_iface = iface;
1532 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1533 br->name, ETH_ADDR_ARGS(ea));
1535 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1536 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1537 *hw_addr_iface = NULL;
1538 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1539 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1542 hmapx_destroy(&mirror_output_ports);
1545 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1546 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1547 * an interface on 'br', then that interface must be passed in as
1548 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1549 * 'hw_addr_iface' must be passed in as a null pointer. */
1551 bridge_pick_datapath_id(struct bridge *br,
1552 const uint8_t bridge_ea[ETH_ADDR_LEN],
1553 struct iface *hw_addr_iface)
1556 * The procedure for choosing a bridge MAC address will, in the most
1557 * ordinary case, also choose a unique MAC that we can use as a datapath
1558 * ID. In some special cases, though, multiple bridges will end up with
1559 * the same MAC address. This is OK for the bridges, but it will confuse
1560 * the OpenFlow controller, because each datapath needs a unique datapath
1563 * Datapath IDs must be unique. It is also very desirable that they be
1564 * stable from one run to the next, so that policy set on a datapath
1567 const char *datapath_id;
1570 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1571 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1575 if (!hw_addr_iface) {
1577 * A purely internal bridge, that is, one that has no non-virtual
1578 * network devices on it at all, is difficult because it has no
1579 * natural unique identifier at all.
1581 * When the host is a XenServer, we handle this case by hashing the
1582 * host's UUID with the name of the bridge. Names of bridges are
1583 * persistent across XenServer reboots, although they can be reused if
1584 * an internal network is destroyed and then a new one is later
1585 * created, so this is fairly effective.
1587 * When the host is not a XenServer, we punt by using a random MAC
1588 * address on each run.
1590 const char *host_uuid = xenserver_get_host_uuid();
1592 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1593 dpid = dpid_from_hash(combined, strlen(combined));
1599 return eth_addr_to_uint64(bridge_ea);
1603 dpid_from_hash(const void *data, size_t n)
1605 uint8_t hash[SHA1_DIGEST_SIZE];
1607 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1608 sha1_bytes(data, n, hash);
1609 eth_addr_mark_random(hash);
1610 return eth_addr_to_uint64(hash);
1614 iface_refresh_status(struct iface *iface)
1618 enum netdev_features current;
1619 enum netdev_flags flags;
1625 if (iface_is_synthetic(iface)) {
1631 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1632 ovsrec_interface_set_status(iface->cfg, &smap);
1634 ovsrec_interface_set_status(iface->cfg, NULL);
1637 smap_destroy(&smap);
1639 error = netdev_get_flags(iface->netdev, &flags);
1641 ovsrec_interface_set_admin_state(iface->cfg,
1642 flags & NETDEV_UP ? "up" : "down");
1645 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1648 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1650 ovsrec_interface_set_duplex(iface->cfg,
1651 netdev_features_is_full_duplex(current)
1653 /* warning: uint64_t -> int64_t conversion */
1654 bps = netdev_features_to_bps(current);
1655 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1658 ovsrec_interface_set_duplex(iface->cfg, NULL);
1659 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1662 error = netdev_get_mtu(iface->netdev, &mtu);
1665 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1668 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1672 /* Writes 'iface''s CFM statistics to the database. */
1674 iface_refresh_cfm_stats(struct iface *iface)
1676 const struct ovsrec_interface *cfg = iface->cfg;
1678 const uint64_t *rmps;
1682 if (iface_is_synthetic(iface)) {
1686 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1689 const char *reasons[CFM_FAULT_N_REASONS];
1690 bool fault_bool = fault;
1694 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1695 int reason = 1 << i;
1696 if (fault & reason) {
1697 reasons[j++] = cfm_fault_reason_to_str(reason);
1701 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1702 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1704 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1705 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1708 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1709 iface->ofp_port, &rmps, &n_rmps);
1711 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1714 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1717 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1720 int64_t cfm_health = health;
1721 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1723 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1728 iface_refresh_stats(struct iface *iface)
1730 #define IFACE_STATS \
1731 IFACE_STAT(rx_packets, "rx_packets") \
1732 IFACE_STAT(tx_packets, "tx_packets") \
1733 IFACE_STAT(rx_bytes, "rx_bytes") \
1734 IFACE_STAT(tx_bytes, "tx_bytes") \
1735 IFACE_STAT(rx_dropped, "rx_dropped") \
1736 IFACE_STAT(tx_dropped, "tx_dropped") \
1737 IFACE_STAT(rx_errors, "rx_errors") \
1738 IFACE_STAT(tx_errors, "tx_errors") \
1739 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1740 IFACE_STAT(rx_over_errors, "rx_over_err") \
1741 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1742 IFACE_STAT(collisions, "collisions")
1744 #define IFACE_STAT(MEMBER, NAME) NAME,
1745 static char *keys[] = { IFACE_STATS };
1747 int64_t values[ARRAY_SIZE(keys)];
1750 struct netdev_stats stats;
1752 if (iface_is_synthetic(iface)) {
1756 /* Intentionally ignore return value, since errors will set 'stats' to
1757 * all-1s, and we will deal with that correctly below. */
1758 netdev_get_stats(iface->netdev, &stats);
1760 /* Copy statistics into values[] array. */
1762 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1765 assert(i == ARRAY_SIZE(keys));
1767 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1773 br_refresh_stp_status(struct bridge *br)
1775 struct smap smap = SMAP_INITIALIZER(&smap);
1776 struct ofproto *ofproto = br->ofproto;
1777 struct ofproto_stp_status status;
1779 if (ofproto_get_stp_status(ofproto, &status)) {
1783 if (!status.enabled) {
1784 ovsrec_bridge_set_status(br->cfg, NULL);
1788 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1789 STP_ID_ARGS(status.bridge_id));
1790 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1791 STP_ID_ARGS(status.designated_root));
1792 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1794 ovsrec_bridge_set_status(br->cfg, &smap);
1795 smap_destroy(&smap);
1799 port_refresh_stp_status(struct port *port)
1801 struct ofproto *ofproto = port->bridge->ofproto;
1802 struct iface *iface;
1803 struct ofproto_port_stp_status status;
1805 int64_t int_values[3];
1808 if (port_is_synthetic(port)) {
1812 /* STP doesn't currently support bonds. */
1813 if (!list_is_singleton(&port->ifaces)) {
1814 ovsrec_port_set_status(port->cfg, NULL);
1818 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1820 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1824 if (!status.enabled) {
1825 ovsrec_port_set_status(port->cfg, NULL);
1826 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1830 /* Set Status column. */
1832 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1833 smap_add(&smap, "stp_state", stp_state_name(status.state));
1834 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1835 smap_add(&smap, "stp_role", stp_role_name(status.role));
1836 ovsrec_port_set_status(port->cfg, &smap);
1837 smap_destroy(&smap);
1839 /* Set Statistics column. */
1840 keys[0] = "stp_tx_count";
1841 int_values[0] = status.tx_count;
1842 keys[1] = "stp_rx_count";
1843 int_values[1] = status.rx_count;
1844 keys[2] = "stp_error_count";
1845 int_values[2] = status.error_count;
1847 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1848 ARRAY_SIZE(int_values));
1852 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1854 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1858 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1860 struct ovsdb_datum datum;
1864 if (enable_system_stats(cfg)) {
1865 get_system_stats(&stats);
1868 ovsdb_datum_from_shash(&datum, &stats);
1869 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1873 static inline const char *
1874 nx_role_to_str(enum nx_role role)
1879 case NX_ROLE_MASTER:
1884 return "*** INVALID ROLE ***";
1889 refresh_controller_status(void)
1893 const struct ovsrec_controller *cfg;
1897 /* Accumulate status for controllers on all bridges. */
1898 HMAP_FOR_EACH (br, node, &all_bridges) {
1899 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1902 /* Update each controller in the database with current status. */
1903 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1904 struct ofproto_controller_info *cinfo =
1905 shash_find_data(&info, cfg->target);
1908 struct smap smap = SMAP_INITIALIZER(&smap);
1909 const char **values = cinfo->pairs.values;
1910 const char **keys = cinfo->pairs.keys;
1913 for (i = 0; i < cinfo->pairs.n; i++) {
1914 smap_add(&smap, keys[i], values[i]);
1917 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1918 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1919 ovsrec_controller_set_status(cfg, &smap);
1920 smap_destroy(&smap);
1922 ovsrec_controller_set_is_connected(cfg, false);
1923 ovsrec_controller_set_role(cfg, NULL);
1924 ovsrec_controller_set_status(cfg, NULL);
1928 ofproto_free_ofproto_controller_info(&info);
1932 refresh_cfm_stats(void)
1934 static struct ovsdb_idl_txn *txn = NULL;
1939 txn = ovsdb_idl_txn_create(idl);
1941 HMAP_FOR_EACH (br, node, &all_bridges) {
1942 struct iface *iface;
1944 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1945 iface_refresh_cfm_stats(iface);
1950 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1951 ovsdb_idl_txn_destroy(txn);
1956 /* Performs periodic activity required by bridges that needs to be done with
1957 * the least possible latency.
1959 * It makes sense to call this function a couple of times per poll loop, to
1960 * provide a significant performance boost on some benchmarks with ofprotos
1961 * that use the ofproto-dpif implementation. */
1963 bridge_run_fast(void)
1967 HMAP_FOR_EACH (br, node, &all_bridges) {
1968 ofproto_run_fast(br->ofproto);
1975 static const struct ovsrec_open_vswitch null_cfg;
1976 const struct ovsrec_open_vswitch *cfg;
1977 struct ovsdb_idl_txn *reconf_txn = NULL;
1979 bool vlan_splinters_changed;
1982 /* (Re)configure if necessary. */
1983 if (!reconfiguring) {
1986 if (ovsdb_idl_is_lock_contended(idl)) {
1987 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1988 struct bridge *br, *next_br;
1990 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1991 "disabling this process until it goes away");
1993 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1997 } else if (!ovsdb_idl_has_lock(idl)) {
2001 cfg = ovsrec_open_vswitch_first(idl);
2003 /* Let each bridge do the work that it needs to do. */
2004 HMAP_FOR_EACH (br, node, &all_bridges) {
2005 ofproto_run(br->ofproto);
2008 /* Re-configure SSL. We do this on every trip through the main loop,
2009 * instead of just when the database changes, because the contents of the
2010 * key and certificate files can change without the database changing.
2012 * We do this before bridge_reconfigure() because that function might
2013 * initiate SSL connections and thus requires SSL to be configured. */
2014 if (cfg && cfg->ssl) {
2015 const struct ovsrec_ssl *ssl = cfg->ssl;
2017 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2018 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2021 if (!reconfiguring) {
2022 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2023 * usage has changed. */
2024 vlan_splinters_changed = false;
2025 if (vlan_splinters_enabled_anywhere) {
2026 HMAP_FOR_EACH (br, node, &all_bridges) {
2027 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2028 vlan_splinters_changed = true;
2034 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2035 idl_seqno = ovsdb_idl_get_seqno(idl);
2037 reconf_txn = ovsdb_idl_txn_create(idl);
2038 bridge_reconfigure(cfg);
2040 /* We still need to reconfigure to avoid dangling pointers to
2041 * now-destroyed ovsrec structures inside bridge data. */
2042 bridge_reconfigure(&null_cfg);
2047 if (reconfiguring) {
2050 reconf_txn = ovsdb_idl_txn_create(idl);
2052 if (bridge_reconfigure_continue(cfg)) {
2053 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2056 bridge_reconfigure_continue(&null_cfg);
2061 ovsdb_idl_txn_commit(reconf_txn);
2062 ovsdb_idl_txn_destroy(reconf_txn);
2066 /* Refresh system and interface stats if necessary. */
2067 if (time_msec() >= stats_timer) {
2069 struct ovsdb_idl_txn *txn;
2071 txn = ovsdb_idl_txn_create(idl);
2072 HMAP_FOR_EACH (br, node, &all_bridges) {
2076 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2077 struct iface *iface;
2079 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2080 iface_refresh_stats(iface);
2081 iface_refresh_status(iface);
2085 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2086 mirror_refresh_stats(m);
2090 refresh_system_stats(cfg);
2091 refresh_controller_status();
2092 ovsdb_idl_txn_commit(txn);
2093 ovsdb_idl_txn_destroy(txn); /* XXX */
2096 stats_timer = time_msec() + STATS_INTERVAL;
2099 if (time_msec() >= db_limiter) {
2100 struct ovsdb_idl_txn *txn;
2102 txn = ovsdb_idl_txn_create(idl);
2103 HMAP_FOR_EACH (br, node, &all_bridges) {
2104 struct iface *iface;
2107 br_refresh_stp_status(br);
2109 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2110 port_refresh_stp_status(port);
2113 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2114 const char *link_state;
2115 int64_t link_resets;
2118 if (iface_is_synthetic(iface)) {
2122 current = ofproto_port_is_lacp_current(br->ofproto,
2126 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2128 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2131 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2132 ovsrec_interface_set_link_state(iface->cfg, link_state);
2134 link_resets = netdev_get_carrier_resets(iface->netdev);
2135 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2139 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2140 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2142 ovsdb_idl_txn_destroy(txn);
2145 refresh_cfm_stats();
2151 ovsdb_idl_wait(idl);
2153 if (reconfiguring) {
2154 poll_immediate_wake();
2157 if (!hmap_is_empty(&all_bridges)) {
2160 HMAP_FOR_EACH (br, node, &all_bridges) {
2161 ofproto_wait(br->ofproto);
2163 poll_timer_wait_until(stats_timer);
2165 if (db_limiter > time_msec()) {
2166 poll_timer_wait_until(db_limiter);
2171 /* Adds some memory usage statistics for bridges into 'usage', for use with
2172 * memory_report(). */
2174 bridge_get_memory_usage(struct simap *usage)
2178 HMAP_FOR_EACH (br, node, &all_bridges) {
2179 ofproto_get_memory_usage(br->ofproto, usage);
2183 /* QoS unixctl user interface functions. */
2185 struct qos_unixctl_show_cbdata {
2187 struct iface *iface;
2191 qos_unixctl_show_cb(unsigned int queue_id,
2192 const struct smap *details,
2195 struct qos_unixctl_show_cbdata *data = aux;
2196 struct ds *ds = data->ds;
2197 struct iface *iface = data->iface;
2198 struct netdev_queue_stats stats;
2199 struct smap_node *node;
2202 ds_put_cstr(ds, "\n");
2204 ds_put_format(ds, "Queue %u:\n", queue_id);
2206 ds_put_cstr(ds, "Default:\n");
2209 SMAP_FOR_EACH (node, details) {
2210 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2213 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2215 if (stats.tx_packets != UINT64_MAX) {
2216 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2219 if (stats.tx_bytes != UINT64_MAX) {
2220 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2223 if (stats.tx_errors != UINT64_MAX) {
2224 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2227 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2228 queue_id, strerror(error));
2233 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2234 const char *argv[], void *aux OVS_UNUSED)
2236 struct ds ds = DS_EMPTY_INITIALIZER;
2237 struct smap smap = SMAP_INITIALIZER(&smap);
2238 struct iface *iface;
2240 struct smap_node *node;
2241 struct qos_unixctl_show_cbdata data;
2244 iface = iface_find(argv[1]);
2246 unixctl_command_reply_error(conn, "no such interface");
2250 netdev_get_qos(iface->netdev, &type, &smap);
2252 if (*type != '\0') {
2253 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2255 SMAP_FOR_EACH (node, &smap) {
2256 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2261 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2264 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2266 unixctl_command_reply(conn, ds_cstr(&ds));
2268 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2269 unixctl_command_reply_error(conn, ds_cstr(&ds));
2272 smap_destroy(&smap);
2276 /* Bridge reconfiguration functions. */
2278 bridge_create(const struct ovsrec_bridge *br_cfg)
2282 assert(!bridge_lookup(br_cfg->name));
2283 br = xzalloc(sizeof *br);
2285 br->name = xstrdup(br_cfg->name);
2286 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2289 /* Derive the default Ethernet address from the bridge's UUID. This should
2290 * be unique and it will be stable between ovs-vswitchd runs. */
2291 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2292 eth_addr_mark_random(br->default_ea);
2294 hmap_init(&br->ports);
2295 hmap_init(&br->ifaces);
2296 hmap_init(&br->iface_by_name);
2297 hmap_init(&br->mirrors);
2299 hmap_init(&br->if_cfg_todo);
2300 list_init(&br->ofpp_garbage);
2302 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2306 bridge_destroy(struct bridge *br)
2309 struct mirror *mirror, *next_mirror;
2310 struct port *port, *next_port;
2311 struct if_cfg *if_cfg, *next_if_cfg;
2312 struct ofpp_garbage *garbage, *next_garbage;
2314 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2317 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2318 mirror_destroy(mirror);
2320 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2321 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2324 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2325 &br->ofpp_garbage) {
2326 list_remove(&garbage->list_node);
2330 hmap_remove(&all_bridges, &br->node);
2331 ofproto_destroy(br->ofproto);
2332 hmap_destroy(&br->ifaces);
2333 hmap_destroy(&br->ports);
2334 hmap_destroy(&br->iface_by_name);
2335 hmap_destroy(&br->mirrors);
2336 hmap_destroy(&br->if_cfg_todo);
2343 static struct bridge *
2344 bridge_lookup(const char *name)
2348 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2349 if (!strcmp(br->name, name)) {
2356 /* Handle requests for a listing of all flows known by the OpenFlow
2357 * stack, including those normally hidden. */
2359 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2360 const char *argv[], void *aux OVS_UNUSED)
2365 br = bridge_lookup(argv[1]);
2367 unixctl_command_reply_error(conn, "Unknown bridge");
2372 ofproto_get_all_flows(br->ofproto, &results);
2374 unixctl_command_reply(conn, ds_cstr(&results));
2375 ds_destroy(&results);
2378 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2379 * connections and reconnect. If BRIDGE is not specified, then all bridges
2380 * drop their controller connections and reconnect. */
2382 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2383 const char *argv[], void *aux OVS_UNUSED)
2387 br = bridge_lookup(argv[1]);
2389 unixctl_command_reply_error(conn, "Unknown bridge");
2392 ofproto_reconnect_controllers(br->ofproto);
2394 HMAP_FOR_EACH (br, node, &all_bridges) {
2395 ofproto_reconnect_controllers(br->ofproto);
2398 unixctl_command_reply(conn, NULL);
2402 bridge_get_controllers(const struct bridge *br,
2403 struct ovsrec_controller ***controllersp)
2405 struct ovsrec_controller **controllers;
2406 size_t n_controllers;
2408 controllers = br->cfg->controller;
2409 n_controllers = br->cfg->n_controller;
2411 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2417 *controllersp = controllers;
2419 return n_controllers;
2423 bridge_queue_if_cfg(struct bridge *br,
2424 const struct ovsrec_interface *cfg,
2425 const struct ovsrec_port *parent)
2427 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2430 if_cfg->parent = parent;
2431 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2432 hash_string(if_cfg->cfg->name, 0));
2435 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2436 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2437 * 'br' needs to complete its configuration. */
2439 bridge_add_del_ports(struct bridge *br,
2440 const unsigned long int *splinter_vlans)
2442 struct shash_node *port_node;
2443 struct port *port, *next;
2444 struct shash new_ports;
2447 assert(hmap_is_empty(&br->if_cfg_todo));
2449 /* Collect new ports. */
2450 shash_init(&new_ports);
2451 for (i = 0; i < br->cfg->n_ports; i++) {
2452 const char *name = br->cfg->ports[i]->name;
2453 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2454 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2458 if (bridge_get_controllers(br, NULL)
2459 && !shash_find(&new_ports, br->name)) {
2460 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2461 br->name, br->name);
2463 ovsrec_interface_init(&br->synth_local_iface);
2464 ovsrec_port_init(&br->synth_local_port);
2466 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2467 br->synth_local_port.n_interfaces = 1;
2468 br->synth_local_port.name = br->name;
2470 br->synth_local_iface.name = br->name;
2471 br->synth_local_iface.type = "internal";
2473 br->synth_local_ifacep = &br->synth_local_iface;
2475 shash_add(&new_ports, br->name, &br->synth_local_port);
2478 if (splinter_vlans) {
2479 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2482 /* Get rid of deleted ports.
2483 * Get rid of deleted interfaces on ports that still exist. */
2484 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2485 port->cfg = shash_find_data(&new_ports, port->name);
2489 port_del_ifaces(port);
2493 /* Update iface->cfg and iface->type in interfaces that still exist.
2494 * Add new interfaces to creation queue. */
2495 SHASH_FOR_EACH (port_node, &new_ports) {
2496 const struct ovsrec_port *port = port_node->data;
2499 for (i = 0; i < port->n_interfaces; i++) {
2500 const struct ovsrec_interface *cfg = port->interfaces[i];
2501 struct iface *iface = iface_lookup(br, cfg->name);
2502 const char *type = iface_get_type(cfg, br->cfg);
2507 } else if (strcmp(type, "null")) {
2508 bridge_queue_if_cfg(br, cfg, port);
2513 shash_destroy(&new_ports);
2516 /* Initializes 'oc' appropriately as a management service controller for
2519 * The caller must free oc->target when it is no longer needed. */
2521 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2522 struct ofproto_controller *oc)
2524 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2525 oc->max_backoff = 0;
2526 oc->probe_interval = 60;
2527 oc->band = OFPROTO_OUT_OF_BAND;
2529 oc->burst_limit = 0;
2530 oc->enable_async_msgs = true;
2533 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2535 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2536 struct ofproto_controller *oc)
2540 oc->target = c->target;
2541 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2542 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2543 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2544 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2545 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2546 oc->burst_limit = (c->controller_burst_limit
2547 ? *c->controller_burst_limit : 0);
2548 oc->enable_async_msgs = (!c->enable_async_messages
2549 || *c->enable_async_messages);
2550 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2551 if (dscp < 0 || dscp > 63) {
2552 dscp = DSCP_DEFAULT;
2557 /* Configures the IP stack for 'br''s local interface properly according to the
2558 * configuration in 'c'. */
2560 bridge_configure_local_iface_netdev(struct bridge *br,
2561 struct ovsrec_controller *c)
2563 struct netdev *netdev;
2564 struct in_addr mask, gateway;
2566 struct iface *local_iface;
2569 /* If there's no local interface or no IP address, give up. */
2570 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2571 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2575 /* Bring up the local interface. */
2576 netdev = local_iface->netdev;
2577 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2579 /* Configure the IP address and netmask. */
2580 if (!c->local_netmask
2581 || !inet_aton(c->local_netmask, &mask)
2583 mask.s_addr = guess_netmask(ip.s_addr);
2585 if (!netdev_set_in4(netdev, ip, mask)) {
2586 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2587 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2590 /* Configure the default gateway. */
2591 if (c->local_gateway
2592 && inet_aton(c->local_gateway, &gateway)
2593 && gateway.s_addr) {
2594 if (!netdev_add_router(netdev, gateway)) {
2595 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2596 br->name, IP_ARGS(&gateway.s_addr));
2601 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2602 * in either string are treated as equal to any number of slashes in the other,
2603 * e.g. "x///y" is equal to "x/y". */
2605 equal_pathnames(const char *a, const char *b)
2609 a += strspn(a, "/");
2610 b += strspn(b, "/");
2611 } else if (*a == '\0') {
2622 bridge_configure_remotes(struct bridge *br,
2623 const struct sockaddr_in *managers, size_t n_managers)
2625 bool disable_in_band;
2627 struct ovsrec_controller **controllers;
2628 size_t n_controllers;
2630 enum ofproto_fail_mode fail_mode;
2632 struct ofproto_controller *ocs;
2636 /* Check if we should disable in-band control on this bridge. */
2637 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2640 /* Set OpenFlow queue ID for in-band control. */
2641 ofproto_set_in_band_queue(br->ofproto,
2642 smap_get_int(&br->cfg->other_config,
2643 "in-band-queue", -1));
2645 if (disable_in_band) {
2646 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2648 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2651 n_controllers = bridge_get_controllers(br, &controllers);
2653 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2656 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2657 for (i = 0; i < n_controllers; i++) {
2658 struct ovsrec_controller *c = controllers[i];
2660 if (!strncmp(c->target, "punix:", 6)
2661 || !strncmp(c->target, "unix:", 5)) {
2662 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2665 whitelist = xasprintf("unix:%s/%s.controller",
2666 ovs_rundir(), br->name);
2667 if (!equal_pathnames(c->target, whitelist)) {
2668 /* Prevent remote ovsdb-server users from accessing arbitrary
2669 * Unix domain sockets and overwriting arbitrary local
2671 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2672 "controller \"%s\" due to possibility for remote "
2673 "exploit. Instead, specify whitelisted \"%s\" or "
2674 "connect to \"unix:%s/%s.mgmt\" (which is always "
2675 "available without special configuration).",
2676 br->name, c->target, whitelist,
2677 ovs_rundir(), br->name);
2685 bridge_configure_local_iface_netdev(br, c);
2686 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2687 if (disable_in_band) {
2688 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2693 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2694 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2697 /* Set the fail-mode. */
2698 fail_mode = !br->cfg->fail_mode
2699 || !strcmp(br->cfg->fail_mode, "standalone")
2700 ? OFPROTO_FAIL_STANDALONE
2701 : OFPROTO_FAIL_SECURE;
2702 ofproto_set_fail_mode(br->ofproto, fail_mode);
2704 /* Configure OpenFlow controller connection snooping. */
2705 if (!ofproto_has_snoops(br->ofproto)) {
2709 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2710 ovs_rundir(), br->name));
2711 ofproto_set_snoops(br->ofproto, &snoops);
2712 sset_destroy(&snoops);
2717 bridge_configure_tables(struct bridge *br)
2719 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2723 n_tables = ofproto_get_n_tables(br->ofproto);
2725 for (i = 0; i < n_tables; i++) {
2726 struct ofproto_table_settings s;
2729 s.max_flows = UINT_MAX;
2733 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2734 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2737 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2738 s.max_flows = *cfg->flow_limit;
2740 if (cfg->overflow_policy
2741 && !strcmp(cfg->overflow_policy, "evict")) {
2744 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2745 for (k = 0; k < cfg->n_groups; k++) {
2746 const char *string = cfg->groups[k];
2749 msg = mf_parse_subfield__(&s.groups[k], &string);
2751 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2752 "'groups' (%s)", br->name, i, msg);
2754 } else if (*string) {
2755 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2756 "element '%s' contains trailing garbage",
2757 br->name, i, cfg->groups[k]);
2765 ofproto_configure_table(br->ofproto, i, &s);
2769 for (; j < br->cfg->n_flow_tables; j++) {
2770 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2771 "%"PRId64" not supported by this datapath", br->name,
2772 br->cfg->key_flow_tables[j]);
2776 /* Port functions. */
2778 static struct port *
2779 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2783 port = xzalloc(sizeof *port);
2785 port->name = xstrdup(cfg->name);
2787 list_init(&port->ifaces);
2789 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2793 /* Deletes interfaces from 'port' that are no longer configured for it. */
2795 port_del_ifaces(struct port *port)
2797 struct iface *iface, *next;
2798 struct sset new_ifaces;
2801 /* Collect list of new interfaces. */
2802 sset_init(&new_ifaces);
2803 for (i = 0; i < port->cfg->n_interfaces; i++) {
2804 const char *name = port->cfg->interfaces[i]->name;
2805 const char *type = port->cfg->interfaces[i]->type;
2806 if (strcmp(type, "null")) {
2807 sset_add(&new_ifaces, name);
2811 /* Get rid of deleted interfaces. */
2812 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2813 if (!sset_contains(&new_ifaces, iface->name)) {
2814 iface_destroy(iface);
2818 sset_destroy(&new_ifaces);
2822 port_destroy(struct port *port)
2825 struct bridge *br = port->bridge;
2826 struct iface *iface, *next;
2829 ofproto_bundle_unregister(br->ofproto, port);
2832 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2833 iface_destroy(iface);
2836 hmap_remove(&br->ports, &port->hmap_node);
2842 static struct port *
2843 port_lookup(const struct bridge *br, const char *name)
2847 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2849 if (!strcmp(port->name, name)) {
2857 enable_lacp(struct port *port, bool *activep)
2859 if (!port->cfg->lacp) {
2860 /* XXX when LACP implementation has been sufficiently tested, enable by
2861 * default and make active on bonded ports. */
2863 } else if (!strcmp(port->cfg->lacp, "off")) {
2865 } else if (!strcmp(port->cfg->lacp, "active")) {
2868 } else if (!strcmp(port->cfg->lacp, "passive")) {
2872 VLOG_WARN("port %s: unknown LACP mode %s",
2873 port->name, port->cfg->lacp);
2878 static struct lacp_settings *
2879 port_configure_lacp(struct port *port, struct lacp_settings *s)
2881 const char *lacp_time, *system_id;
2884 if (!enable_lacp(port, &s->active)) {
2888 s->name = port->name;
2890 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2892 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2893 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2894 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2895 " address.", port->name, system_id);
2899 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2902 if (eth_addr_is_zero(s->id)) {
2903 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2907 /* Prefer bondable links if unspecified. */
2908 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2910 s->priority = (priority > 0 && priority <= UINT16_MAX
2912 : UINT16_MAX - !list_is_short(&port->ifaces));
2914 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2915 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2920 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2922 int priority, portid, key;
2924 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2925 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2927 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2929 if (portid <= 0 || portid > UINT16_MAX) {
2930 portid = iface->ofp_port;
2933 if (priority <= 0 || priority > UINT16_MAX) {
2934 priority = UINT16_MAX;
2937 if (key < 0 || key > UINT16_MAX) {
2941 s->name = iface->name;
2943 s->priority = priority;
2948 port_configure_bond(struct port *port, struct bond_settings *s,
2949 uint32_t *bond_stable_ids)
2951 const char *detect_s;
2952 struct iface *iface;
2953 int miimon_interval;
2956 s->name = port->name;
2958 if (port->cfg->bond_mode) {
2959 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2960 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2961 port->name, port->cfg->bond_mode,
2962 bond_mode_to_string(s->balance));
2965 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2967 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2968 * active-backup. At some point we should remove this warning. */
2969 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2970 " in previous versions, the default bond_mode was"
2971 " balance-slb", port->name,
2972 bond_mode_to_string(s->balance));
2974 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2975 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2976 "please use another bond type or disable flood_vlans",
2980 miimon_interval = smap_get_int(&port->cfg->other_config,
2981 "bond-miimon-interval", 0);
2982 if (miimon_interval <= 0) {
2983 miimon_interval = 200;
2986 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
2987 if (!detect_s || !strcmp(detect_s, "carrier")) {
2988 miimon_interval = 0;
2989 } else if (strcmp(detect_s, "miimon")) {
2990 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2991 "defaulting to carrier", port->name, detect_s);
2992 miimon_interval = 0;
2995 s->up_delay = MAX(0, port->cfg->bond_updelay);
2996 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2997 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
2998 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
2999 "bond-rebalance-interval", 10000);
3000 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3001 s->rebalance_interval = 1000;
3004 s->fake_iface = port->cfg->bond_fake_iface;
3007 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3008 long long stable_id;
3010 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3012 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3013 stable_id = iface->ofp_port;
3015 bond_stable_ids[i++] = stable_id;
3017 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3021 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3022 * instead of obtaining it from the database. */
3024 port_is_synthetic(const struct port *port)
3026 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3029 /* Interface functions. */
3031 /* Returns the correct network device type for interface 'iface' in bridge
3034 iface_get_type(const struct ovsrec_interface *iface,
3035 const struct ovsrec_bridge *br)
3037 /* The local port always has type "internal". Other ports take their type
3038 * from the database and default to "system" if none is specified. */
3039 return (!strcmp(iface->name, br->name) ? "internal"
3040 : iface->type[0] ? iface->type
3045 iface_destroy(struct iface *iface)
3048 struct port *port = iface->port;
3049 struct bridge *br = port->bridge;
3051 if (br->ofproto && iface->ofp_port >= 0) {
3052 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3055 if (iface->ofp_port >= 0) {
3056 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3059 list_remove(&iface->port_elem);
3060 hmap_remove(&br->iface_by_name, &iface->name_node);
3062 netdev_close(iface->netdev);
3069 static struct iface *
3070 iface_lookup(const struct bridge *br, const char *name)
3072 struct iface *iface;
3074 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3075 &br->iface_by_name) {
3076 if (!strcmp(iface->name, name)) {
3084 static struct iface *
3085 iface_find(const char *name)
3087 const struct bridge *br;
3089 HMAP_FOR_EACH (br, node, &all_bridges) {
3090 struct iface *iface = iface_lookup(br, name);
3099 static struct if_cfg *
3100 if_cfg_lookup(const struct bridge *br, const char *name)
3102 struct if_cfg *if_cfg;
3104 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3106 if (!strcmp(if_cfg->cfg->name, name)) {
3114 static struct iface *
3115 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3117 struct iface *iface;
3119 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3120 hash_int(ofp_port, 0), &br->ifaces) {
3121 if (iface->ofp_port == ofp_port) {
3128 /* Set Ethernet address of 'iface', if one is specified in the configuration
3131 iface_set_mac(struct iface *iface)
3133 uint8_t ea[ETH_ADDR_LEN];
3135 if (!strcmp(iface->type, "internal")
3136 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3137 if (iface->ofp_port == OFPP_LOCAL) {
3138 VLOG_ERR("interface %s: ignoring mac in Interface record "
3139 "(use Bridge record to set local port's mac)",
3141 } else if (eth_addr_is_multicast(ea)) {
3142 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3145 int error = netdev_set_etheraddr(iface->netdev, ea);
3147 VLOG_ERR("interface %s: setting MAC failed (%s)",
3148 iface->name, strerror(error));
3154 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3156 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3158 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3159 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3163 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3164 * sets the "ofport" field to -1.
3166 * This is appropriate when 'if_cfg''s interface cannot be created or is
3167 * otherwise invalid. */
3169 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3171 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3172 iface_set_ofport(if_cfg, -1);
3173 ovsrec_interface_set_status(if_cfg, NULL);
3174 ovsrec_interface_set_admin_state(if_cfg, NULL);
3175 ovsrec_interface_set_duplex(if_cfg, NULL);
3176 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3177 ovsrec_interface_set_link_state(if_cfg, NULL);
3178 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3179 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3180 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3181 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3182 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3183 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3187 struct iface_delete_queues_cbdata {
3188 struct netdev *netdev;
3189 const struct ovsdb_datum *queues;
3193 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3195 union ovsdb_atom atom;
3197 atom.integer = target;
3198 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3202 iface_delete_queues(unsigned int queue_id,
3203 const struct smap *details OVS_UNUSED, void *cbdata_)
3205 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3207 if (!queue_ids_include(cbdata->queues, queue_id)) {
3208 netdev_delete_queue(cbdata->netdev, queue_id);
3213 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3215 struct ofpbuf queues_buf;
3217 ofpbuf_init(&queues_buf, 0);
3219 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3220 netdev_set_qos(iface->netdev, NULL, NULL);
3222 struct iface_delete_queues_cbdata cbdata;
3226 /* Configure top-level Qos for 'iface'. */
3227 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3229 /* Deconfigure queues that were deleted. */
3230 cbdata.netdev = iface->netdev;
3231 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3233 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3235 /* Configure queues for 'iface'. */
3237 for (i = 0; i < qos->n_queues; i++) {
3238 const struct ovsrec_queue *queue = qos->value_queues[i];
3239 unsigned int queue_id = qos->key_queues[i];
3241 if (queue_id == 0) {
3245 if (queue->n_dscp == 1) {
3246 struct ofproto_port_queue *port_queue;
3248 port_queue = ofpbuf_put_uninit(&queues_buf,
3249 sizeof *port_queue);
3250 port_queue->queue = queue_id;
3251 port_queue->dscp = queue->dscp[0];
3254 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3257 struct smap details;
3259 smap_init(&details);
3260 netdev_set_queue(iface->netdev, 0, &details);
3261 smap_destroy(&details);
3265 if (iface->ofp_port >= 0) {
3266 const struct ofproto_port_queue *port_queues = queues_buf.data;
3267 size_t n_queues = queues_buf.size / sizeof *port_queues;
3269 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3270 port_queues, n_queues);
3273 netdev_set_policing(iface->netdev,
3274 iface->cfg->ingress_policing_rate,
3275 iface->cfg->ingress_policing_burst);
3277 ofpbuf_uninit(&queues_buf);
3281 iface_configure_cfm(struct iface *iface)
3283 const struct ovsrec_interface *cfg = iface->cfg;
3284 const char *opstate_str;
3285 const char *cfm_ccm_vlan;
3286 struct cfm_settings s;
3288 if (!cfg->n_cfm_mpid) {
3289 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3293 s.mpid = *cfg->cfm_mpid;
3294 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3295 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3296 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3298 if (s.interval <= 0) {
3302 if (!cfm_ccm_vlan) {
3304 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3305 s.ccm_vlan = CFM_RANDOM_VLAN;
3307 s.ccm_vlan = atoi(cfm_ccm_vlan);
3308 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3313 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3316 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3317 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3319 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3322 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3323 * instead of obtaining it from the database. */
3325 iface_is_synthetic(const struct iface *iface)
3327 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3331 /* Port mirroring. */
3333 static struct mirror *
3334 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3338 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3339 if (uuid_equals(uuid, &m->uuid)) {
3347 bridge_configure_mirrors(struct bridge *br)
3349 const struct ovsdb_datum *mc;
3350 unsigned long *flood_vlans;
3351 struct mirror *m, *next;
3354 /* Get rid of deleted mirrors. */
3355 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3356 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3357 union ovsdb_atom atom;
3359 atom.uuid = m->uuid;
3360 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3365 /* Add new mirrors and reconfigure existing ones. */
3366 for (i = 0; i < br->cfg->n_mirrors; i++) {
3367 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3368 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3370 m = mirror_create(br, cfg);
3373 if (!mirror_configure(m)) {
3378 /* Update flooded vlans (for RSPAN). */
3379 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3380 br->cfg->n_flood_vlans);
3381 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3382 bitmap_free(flood_vlans);
3385 static struct mirror *
3386 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3390 m = xzalloc(sizeof *m);
3391 m->uuid = cfg->header_.uuid;
3392 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3394 m->name = xstrdup(cfg->name);
3400 mirror_destroy(struct mirror *m)
3403 struct bridge *br = m->bridge;
3406 ofproto_mirror_unregister(br->ofproto, m);
3409 hmap_remove(&br->mirrors, &m->hmap_node);
3416 mirror_collect_ports(struct mirror *m,
3417 struct ovsrec_port **in_ports, int n_in_ports,
3418 void ***out_portsp, size_t *n_out_portsp)
3420 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3421 size_t n_out_ports = 0;
3424 for (i = 0; i < n_in_ports; i++) {
3425 const char *name = in_ports[i]->name;
3426 struct port *port = port_lookup(m->bridge, name);
3428 out_ports[n_out_ports++] = port;
3430 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3431 "port %s", m->bridge->name, m->name, name);
3434 *out_portsp = out_ports;
3435 *n_out_portsp = n_out_ports;
3439 mirror_configure(struct mirror *m)
3441 const struct ovsrec_mirror *cfg = m->cfg;
3442 struct ofproto_mirror_settings s;
3445 if (strcmp(cfg->name, m->name)) {
3447 m->name = xstrdup(cfg->name);
3451 /* Get output port or VLAN. */
3452 if (cfg->output_port) {
3453 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3454 if (!s.out_bundle) {
3455 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3456 m->bridge->name, m->name);
3459 s.out_vlan = UINT16_MAX;
3461 if (cfg->output_vlan) {
3462 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3463 "output vlan; ignoring output vlan",
3464 m->bridge->name, m->name);
3466 } else if (cfg->output_vlan) {
3467 /* The database should prevent invalid VLAN values. */
3468 s.out_bundle = NULL;
3469 s.out_vlan = *cfg->output_vlan;
3471 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3472 m->bridge->name, m->name);
3476 /* Get port selection. */
3477 if (cfg->select_all) {
3478 size_t n_ports = hmap_count(&m->bridge->ports);
3479 void **ports = xmalloc(n_ports * sizeof *ports);
3484 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3494 /* Get ports, dropping ports that don't exist.
3495 * The IDL ensures that there are no duplicates. */
3496 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3497 &s.srcs, &s.n_srcs);
3498 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3499 &s.dsts, &s.n_dsts);
3502 /* Get VLAN selection. */
3503 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3506 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3509 if (s.srcs != s.dsts) {
3518 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3520 * This is deprecated. It is only for compatibility with broken device drivers
3521 * in old versions of Linux that do not properly support VLANs when VLAN
3522 * devices are not used. When broken device drivers are no longer in
3523 * widespread use, we will delete these interfaces. */
3525 static struct ovsrec_port **recs;
3526 static size_t n_recs, allocated_recs;
3528 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3529 * splinters are reconfigured. */
3531 register_rec(struct ovsrec_port *rec)
3533 if (n_recs >= allocated_recs) {
3534 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3536 recs[n_recs++] = rec;
3539 /* Frees all of the ports registered with register_reg(). */
3541 free_registered_recs(void)
3545 for (i = 0; i < n_recs; i++) {
3546 struct ovsrec_port *port = recs[i];
3549 for (j = 0; j < port->n_interfaces; j++) {
3550 struct ovsrec_interface *iface = port->interfaces[j];
3555 smap_destroy(&port->other_config);
3556 free(port->interfaces);
3564 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3567 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3569 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3573 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3576 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3577 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3578 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3581 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3582 * use, returns NULL.
3584 * Updates 'vlan_splinters_enabled_anywhere'. */
3585 static unsigned long int *
3586 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3588 unsigned long int *splinter_vlans;
3589 struct sset splinter_ifaces;
3590 const char *real_dev_name;
3591 struct shash *real_devs;
3592 struct shash_node *node;
3596 /* Free space allocated for synthesized ports and interfaces, since we're
3597 * in the process of reconstructing all of them. */
3598 free_registered_recs();
3600 splinter_vlans = bitmap_allocate(4096);
3601 sset_init(&splinter_ifaces);
3602 vlan_splinters_enabled_anywhere = false;
3603 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3604 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3607 for (j = 0; j < br_cfg->n_ports; j++) {
3608 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3611 for (k = 0; k < port_cfg->n_interfaces; k++) {
3612 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3614 if (vlan_splinters_is_enabled(iface_cfg)) {
3615 vlan_splinters_enabled_anywhere = true;
3616 sset_add(&splinter_ifaces, iface_cfg->name);
3617 vlan_bitmap_from_array__(port_cfg->trunks,
3623 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3624 bitmap_set1(splinter_vlans, *port_cfg->tag);
3629 if (!vlan_splinters_enabled_anywhere) {
3630 free(splinter_vlans);
3631 sset_destroy(&splinter_ifaces);
3635 HMAP_FOR_EACH (br, node, &all_bridges) {
3637 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3641 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3642 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3643 * device to be created for it. */
3644 bitmap_set0(splinter_vlans, 0);
3645 bitmap_set0(splinter_vlans, 4095);
3647 /* Delete all VLAN devices that we don't need. */
3649 real_devs = vlandev_get_real_devs();
3650 SHASH_FOR_EACH (node, real_devs) {
3651 const struct vlan_real_dev *real_dev = node->data;
3652 const struct vlan_dev *vlan_dev;
3653 bool real_dev_has_splinters;
3655 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3657 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3658 if (!real_dev_has_splinters
3659 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3660 struct netdev *netdev;
3662 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3663 if (!netdev_get_in4(netdev, NULL, NULL) ||
3664 !netdev_get_in6(netdev, NULL)) {
3665 vlandev_del(vlan_dev->name);
3667 /* It has an IP address configured, so we don't own
3668 * it. Don't delete it. */
3670 netdev_close(netdev);
3677 /* Add all VLAN devices that we need. */
3678 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3681 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3682 if (!vlandev_get_name(real_dev_name, vid)) {
3683 vlandev_add(real_dev_name, vid);
3690 sset_destroy(&splinter_ifaces);
3692 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3693 free(splinter_vlans);
3696 return splinter_vlans;
3699 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3702 configure_splinter_port(struct port *port)
3704 struct ofproto *ofproto = port->bridge->ofproto;
3705 uint16_t realdev_ofp_port;
3706 const char *realdev_name;
3707 struct iface *vlandev, *realdev;
3709 ofproto_bundle_unregister(port->bridge->ofproto, port);
3711 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3714 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3715 realdev = iface_lookup(port->bridge, realdev_name);
3716 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3718 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3722 static struct ovsrec_port *
3723 synthesize_splinter_port(const char *real_dev_name,
3724 const char *vlan_dev_name, int vid)
3726 struct ovsrec_interface *iface;
3727 struct ovsrec_port *port;
3729 iface = xmalloc(sizeof *iface);
3730 ovsrec_interface_init(iface);
3731 iface->name = xstrdup(vlan_dev_name);
3732 iface->type = "system";
3734 port = xmalloc(sizeof *port);
3735 ovsrec_port_init(port);
3736 port->interfaces = xmemdup(&iface, sizeof iface);
3737 port->n_interfaces = 1;
3738 port->name = xstrdup(vlan_dev_name);
3739 port->vlan_mode = "splinter";
3740 port->tag = xmalloc(sizeof *port->tag);
3743 smap_add(&port->other_config, "realdev", real_dev_name);
3749 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3750 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3751 * 1-bit in the 'splinter_vlans' bitmap. */
3753 add_vlan_splinter_ports(struct bridge *br,
3754 const unsigned long int *splinter_vlans,
3755 struct shash *ports)
3759 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3760 * we're modifying 'ports'. */
3761 for (i = 0; i < br->cfg->n_ports; i++) {
3762 const char *name = br->cfg->ports[i]->name;
3763 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3766 for (j = 0; j < port_cfg->n_interfaces; j++) {
3767 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3769 if (vlan_splinters_is_enabled(iface_cfg)) {
3770 const char *real_dev_name;
3773 real_dev_name = iface_cfg->name;
3774 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3775 const char *vlan_dev_name;
3777 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3779 && !shash_find(ports, vlan_dev_name)) {
3780 shash_add(ports, vlan_dev_name,
3781 synthesize_splinter_port(
3782 real_dev_name, vlan_dev_name, vid));
3791 mirror_refresh_stats(struct mirror *m)
3793 struct ofproto *ofproto = m->bridge->ofproto;
3794 uint64_t tx_packets, tx_bytes;
3797 size_t stat_cnt = 0;
3799 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3800 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3804 if (tx_packets != UINT64_MAX) {
3805 keys[stat_cnt] = "tx_packets";
3806 values[stat_cnt] = tx_packets;
3809 if (tx_bytes != UINT64_MAX) {
3810 keys[stat_cnt] = "tx_bytes";
3811 values[stat_cnt] = tx_bytes;
3815 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);