1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks
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"
44 #include "socket-util.h"
46 #include "stream-ssl.h"
48 #include "system-stats.h"
53 #include "vswitchd/vswitch-idl.h"
54 #include "xenserver.h"
56 #include "sflow_api.h"
57 #include "vlan-bitmap.h"
59 VLOG_DEFINE_THIS_MODULE(bridge);
61 COVERAGE_DEFINE(bridge_reconfigure);
63 /* Configuration of an uninstantiated iface. */
65 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
66 const struct ovsrec_interface *cfg; /* Interface record. */
67 const struct ovsrec_port *parent; /* Parent port record. */
70 /* OpenFlow port slated for removal from ofproto. */
72 struct list list_node; /* Node in bridge's ofpp_garbage. */
73 uint16_t ofp_port; /* Port to be deleted. */
77 /* These members are always valid. */
78 struct list port_elem; /* Element in struct port's "ifaces" list. */
79 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
80 struct port *port; /* Containing port. */
81 char *name; /* Host network device name. */
83 /* These members are valid only after bridge_reconfigure() causes them to
85 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
86 int ofp_port; /* OpenFlow port number, -1 if unknown. */
87 struct netdev *netdev; /* Network device. */
88 const char *type; /* Usually same as cfg->type. */
89 const struct ovsrec_interface *cfg;
90 bool need_refresh; /* Refresh iface after create. */
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 bridge *bridge;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
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 /* Each time this timer expires, the bridge fetches systems and interface
149 * statistics and pushes them into the database. */
150 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
151 static long long int stats_timer = LLONG_MIN;
153 /* Stores the time after which rate limited statistics may be written to the
154 * database. Only updated when changes to the database require rate limiting.
156 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
157 static long long int db_limiter = LLONG_MIN;
159 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
160 * expensive. This can cause bridge_reconfigure() to take a long time during
161 * which no other work can be done. To deal with this problem, we limit port
162 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
163 * call to bridge_reconfigure(). If there is more work to do after the limit
164 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
165 * This allows the rest of the code to catch up on important things like
166 * forwarding packets. */
167 #define OFP_PORT_ACTION_WINDOW 10
168 static bool reconfiguring = false;
170 static void add_del_bridges(const struct ovsrec_open_vswitch *);
171 static void bridge_update_ofprotos(void);
172 static void bridge_create(const struct ovsrec_bridge *);
173 static void bridge_destroy(struct bridge *);
174 static struct bridge *bridge_lookup(const char *name);
175 static unixctl_cb_func bridge_unixctl_dump_flows;
176 static unixctl_cb_func bridge_unixctl_reconnect;
177 static size_t bridge_get_controllers(const struct bridge *br,
178 struct ovsrec_controller ***controllersp);
179 static void bridge_add_del_ports(struct bridge *,
180 const unsigned long int *splinter_vlans);
181 static void bridge_refresh_ofp_port(struct bridge *);
182 static void bridge_configure_datapath_id(struct bridge *);
183 static void bridge_configure_flow_eviction_threshold(struct bridge *);
184 static void bridge_configure_netflow(struct bridge *);
185 static void bridge_configure_forward_bpdu(struct bridge *);
186 static void bridge_configure_mac_idle_time(struct bridge *);
187 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
188 static void bridge_configure_stp(struct bridge *);
189 static void bridge_configure_tables(struct bridge *);
190 static void bridge_configure_remotes(struct bridge *,
191 const struct sockaddr_in *managers,
193 static void bridge_pick_local_hw_addr(struct bridge *,
194 uint8_t ea[ETH_ADDR_LEN],
195 struct iface **hw_addr_iface);
196 static uint64_t bridge_pick_datapath_id(struct bridge *,
197 const uint8_t bridge_ea[ETH_ADDR_LEN],
198 struct iface *hw_addr_iface);
199 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
201 static const char *get_port_other_config(const struct ovsrec_port *,
203 const char *default_value);
204 static void bridge_queue_if_cfg(struct bridge *,
205 const struct ovsrec_interface *,
206 const struct ovsrec_port *);
207 static uint64_t dpid_from_hash(const void *, size_t nbytes);
208 static bool bridge_has_bond_fake_iface(const struct bridge *,
210 static bool port_is_bond_fake_iface(const struct port *);
212 static unixctl_cb_func qos_unixctl_show;
214 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
215 static void port_del_ifaces(struct port *);
216 static void port_destroy(struct port *);
217 static struct port *port_lookup(const struct bridge *, const char *name);
218 static void port_configure(struct port *);
219 static struct lacp_settings *port_configure_lacp(struct port *,
220 struct lacp_settings *);
221 static void port_configure_bond(struct port *, struct bond_settings *,
222 uint32_t *bond_stable_ids);
223 static bool port_is_synthetic(const struct port *);
225 static void bridge_configure_mirrors(struct bridge *);
226 static struct mirror *mirror_create(struct bridge *,
227 const struct ovsrec_mirror *);
228 static void mirror_destroy(struct mirror *);
229 static bool mirror_configure(struct mirror *);
230 static void mirror_refresh_stats(struct mirror *);
232 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
233 static void iface_create(struct bridge *, struct if_cfg *, int ofp_port);
234 static void iface_refresh_type(struct iface *);
235 static void iface_destroy(struct iface *);
236 static struct iface *iface_lookup(const struct bridge *, const char *name);
237 static struct iface *iface_find(const char *name);
238 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
239 static struct iface *iface_from_ofp_port(const struct bridge *,
241 static void iface_set_mac(struct iface *);
242 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
243 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
244 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
245 static void iface_configure_cfm(struct iface *);
246 static void iface_refresh_cfm_stats(struct iface *);
247 static void iface_refresh_stats(struct iface *);
248 static void iface_refresh_status(struct iface *);
249 static bool iface_is_synthetic(const struct iface *);
250 static const char *get_interface_other_config(const struct ovsrec_interface *,
252 const char *default_value);
254 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
256 static void shash_to_ovs_idl_map(struct shash *,
257 char ***keys, char ***values, size_t *n);
259 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
261 * This is deprecated. It is only for compatibility with broken device drivers
262 * in old versions of Linux that do not properly support VLANs when VLAN
263 * devices are not used. When broken device drivers are no longer in
264 * widespread use, we will delete these interfaces. */
266 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
267 static bool vlan_splinters_enabled_anywhere;
269 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
270 static unsigned long int *collect_splinter_vlans(
271 const struct ovsrec_open_vswitch *);
272 static void configure_splinter_port(struct port *);
273 static void add_vlan_splinter_ports(struct bridge *,
274 const unsigned long int *splinter_vlans,
275 struct shash *ports);
277 /* Public functions. */
279 /* Initializes the bridge module, configuring it to obtain its configuration
280 * from an OVSDB server accessed over 'remote', which should be a string in a
281 * form acceptable to ovsdb_idl_create(). */
283 bridge_init(const char *remote)
285 /* Create connection to database. */
286 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
287 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
289 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
290 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
291 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
292 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
293 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
294 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
295 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
297 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
298 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
299 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
301 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
302 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
303 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
315 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
316 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
317 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
318 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
319 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
321 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
322 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
323 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
324 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
326 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
331 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
333 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
335 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
338 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
339 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
340 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
341 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
343 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
345 /* Register unixctl commands. */
346 unixctl_command_register("qos/show", "interface", 1, 1,
347 qos_unixctl_show, NULL);
348 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
349 bridge_unixctl_dump_flows, NULL);
350 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
351 bridge_unixctl_reconnect, NULL);
361 struct bridge *br, *next_br;
363 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
366 ovsdb_idl_destroy(idl);
369 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
370 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
371 * responsible for freeing '*managersp' (with free()).
373 * You may be asking yourself "why does ovs-vswitchd care?", because
374 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
375 * should not be and in fact is not directly involved in that. But
376 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
377 * it has to tell in-band control where the managers are to enable that.
378 * (Thus, only managers connected in-band are collected.)
381 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
382 struct sockaddr_in **managersp, size_t *n_managersp)
384 struct sockaddr_in *managers = NULL;
385 size_t n_managers = 0;
389 /* Collect all of the potential targets from the "targets" columns of the
390 * rows pointed to by "manager_options", excluding any that are
393 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
394 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
396 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
397 sset_find_and_delete(&targets, m->target);
399 sset_add(&targets, m->target);
403 /* Now extract the targets' IP addresses. */
404 if (!sset_is_empty(&targets)) {
407 managers = xmalloc(sset_count(&targets) * sizeof *managers);
408 SSET_FOR_EACH (target, &targets) {
409 struct sockaddr_in *sin = &managers[n_managers];
411 if (stream_parse_target_with_default_ports(target,
419 sset_destroy(&targets);
421 *managersp = managers;
422 *n_managersp = n_managers;
426 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
428 unsigned long int *splinter_vlans;
431 COVERAGE_INC(bridge_reconfigure);
433 assert(!reconfiguring);
434 reconfiguring = true;
436 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
437 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
438 * configuration otherwise.
440 * This is mostly an update to bridge data structures. Nothing is pushed
441 * down to ofproto or lower layers. */
442 add_del_bridges(ovs_cfg);
443 splinter_vlans = collect_splinter_vlans(ovs_cfg);
444 HMAP_FOR_EACH (br, node, &all_bridges) {
445 bridge_add_del_ports(br, splinter_vlans);
447 free(splinter_vlans);
449 /* Delete datapaths that are no longer configured, and create ones which
450 * don't exist but should. */
451 bridge_update_ofprotos();
453 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
454 HMAP_FOR_EACH (br, node, &all_bridges) {
455 bridge_refresh_ofp_port(br);
458 /* Clear database records for "if_cfg"s which haven't been instantiated. */
459 HMAP_FOR_EACH (br, node, &all_bridges) {
460 struct if_cfg *if_cfg;
462 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
463 iface_clear_db_record(if_cfg->cfg);
469 bridge_reconfigure_ofp(void)
471 long long int deadline;
475 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
477 /* The kernel will reject any attempt to add a given port to a datapath if
478 * that port already belongs to a different datapath, so we must do all
479 * port deletions before any port additions. */
480 HMAP_FOR_EACH (br, node, &all_bridges) {
481 struct ofpp_garbage *garbage, *next;
483 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
484 ofproto_port_del(br->ofproto, garbage->ofp_port);
485 list_remove(&garbage->list_node);
489 if (time_msec() >= deadline) {
495 HMAP_FOR_EACH (br, node, &all_bridges) {
496 struct if_cfg *if_cfg, *next;
498 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
499 iface_create(br, if_cfg, -1);
501 if (time_msec() >= deadline) {
511 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
513 struct sockaddr_in *managers;
514 int sflow_bridge_number;
519 assert(reconfiguring);
520 done = bridge_reconfigure_ofp();
522 /* Complete the configuration. */
523 sflow_bridge_number = 0;
524 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
525 HMAP_FOR_EACH (br, node, &all_bridges) {
528 /* We need the datapath ID early to allow LACP ports to use it as the
529 * default system ID. */
530 bridge_configure_datapath_id(br);
532 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
535 port_configure(port);
537 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
538 iface_configure_cfm(iface);
539 iface_configure_qos(iface, port->cfg->qos);
540 iface_set_mac(iface);
543 bridge_configure_mirrors(br);
544 bridge_configure_flow_eviction_threshold(br);
545 bridge_configure_forward_bpdu(br);
546 bridge_configure_mac_idle_time(br);
547 bridge_configure_remotes(br, managers, n_managers);
548 bridge_configure_netflow(br);
549 bridge_configure_sflow(br, &sflow_bridge_number);
550 bridge_configure_stp(br);
551 bridge_configure_tables(br);
556 /* ovs-vswitchd has completed initialization, so allow the process that
557 * forked us to exit successfully. */
558 daemonize_complete();
559 reconfiguring = false;
565 /* Delete ofprotos which aren't configured or have the wrong type. Create
566 * ofprotos which don't exist but need to. */
568 bridge_update_ofprotos(void)
570 struct bridge *br, *next;
575 /* Delete ofprotos with no bridge or with the wrong type. */
578 ofproto_enumerate_types(&types);
579 SSET_FOR_EACH (type, &types) {
582 ofproto_enumerate_names(type, &names);
583 SSET_FOR_EACH (name, &names) {
584 br = bridge_lookup(name);
585 if (!br || strcmp(type, br->type)) {
586 ofproto_delete(name, type);
590 sset_destroy(&names);
591 sset_destroy(&types);
593 /* Add ofprotos for bridges which don't have one yet. */
594 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
602 /* Remove ports from any datapath with the same name as 'br'. If we
603 * don't do this, creating 'br''s ofproto will fail because a port with
604 * the same name as its local port already exists. */
605 HMAP_FOR_EACH (br2, node, &all_bridges) {
606 struct ofproto_port ofproto_port;
612 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
614 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
616 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
619 ofproto_port_destroy(&ofproto_port);
623 error = ofproto_create(br->name, br->type, &br->ofproto);
625 VLOG_ERR("failed to create bridge %s: %s", br->name,
633 port_configure(struct port *port)
635 const struct ovsrec_port *cfg = port->cfg;
636 struct bond_settings bond_settings;
637 struct lacp_settings lacp_settings;
638 struct ofproto_bundle_settings s;
641 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
642 configure_splinter_port(port);
651 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
652 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
653 s.slaves[s.n_slaves++] = iface->ofp_port;
658 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
662 /* Get VLAN trunks. */
665 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
669 if (cfg->vlan_mode) {
670 if (!strcmp(cfg->vlan_mode, "access")) {
671 s.vlan_mode = PORT_VLAN_ACCESS;
672 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
673 s.vlan_mode = PORT_VLAN_TRUNK;
674 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
675 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
676 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
677 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
679 /* This "can't happen" because ovsdb-server should prevent it. */
680 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
681 s.vlan_mode = PORT_VLAN_TRUNK;
685 s.vlan_mode = PORT_VLAN_ACCESS;
687 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
691 s.vlan_mode = PORT_VLAN_TRUNK;
694 s.use_priority_tags = !strcmp("true", get_port_other_config(
695 cfg, "priority-tags", ""));
697 /* Get LACP settings. */
698 s.lacp = port_configure_lacp(port, &lacp_settings);
702 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
703 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
704 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
707 s.lacp_slaves = NULL;
710 /* Get bond settings. */
711 if (s.n_slaves > 1) {
712 s.bond = &bond_settings;
713 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
714 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
717 s.bond_stable_ids = NULL;
719 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
720 netdev_set_miimon_interval(iface->netdev, 0);
725 ofproto_bundle_register(port->bridge->ofproto, port, &s);
731 free(s.bond_stable_ids);
734 /* Pick local port hardware address and datapath ID for 'br'. */
736 bridge_configure_datapath_id(struct bridge *br)
738 uint8_t ea[ETH_ADDR_LEN];
740 struct iface *local_iface;
741 struct iface *hw_addr_iface;
744 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
745 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
747 int error = netdev_set_etheraddr(local_iface->netdev, ea);
749 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
750 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
751 "Ethernet address: %s",
752 br->name, strerror(error));
755 memcpy(br->ea, ea, ETH_ADDR_LEN);
757 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
758 ofproto_set_datapath_id(br->ofproto, dpid);
760 dpid_string = xasprintf("%016"PRIx64, dpid);
761 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
765 /* Set NetFlow configuration on 'br'. */
767 bridge_configure_netflow(struct bridge *br)
769 struct ovsrec_netflow *cfg = br->cfg->netflow;
770 struct netflow_options opts;
773 ofproto_set_netflow(br->ofproto, NULL);
777 memset(&opts, 0, sizeof opts);
779 /* Get default NetFlow configuration from datapath.
780 * Apply overrides from 'cfg'. */
781 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
782 if (cfg->engine_type) {
783 opts.engine_type = *cfg->engine_type;
785 if (cfg->engine_id) {
786 opts.engine_id = *cfg->engine_id;
789 /* Configure active timeout interval. */
790 opts.active_timeout = cfg->active_timeout;
791 if (!opts.active_timeout) {
792 opts.active_timeout = -1;
793 } else if (opts.active_timeout < 0) {
794 VLOG_WARN("bridge %s: active timeout interval set to negative "
795 "value, using default instead (%d seconds)", br->name,
796 NF_ACTIVE_TIMEOUT_DEFAULT);
797 opts.active_timeout = -1;
800 /* Add engine ID to interface number to disambiguate bridgs? */
801 opts.add_id_to_iface = cfg->add_id_to_interface;
802 if (opts.add_id_to_iface) {
803 if (opts.engine_id > 0x7f) {
804 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
805 "another vswitch, choose an engine id less than 128",
808 if (hmap_count(&br->ports) > 508) {
809 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
810 "another port when more than 508 ports are used",
816 sset_init(&opts.collectors);
817 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
820 if (ofproto_set_netflow(br->ofproto, &opts)) {
821 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
823 sset_destroy(&opts.collectors);
826 /* Set sFlow configuration on 'br'. */
828 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
830 const struct ovsrec_sflow *cfg = br->cfg->sflow;
831 struct ovsrec_controller **controllers;
832 struct ofproto_sflow_options oso;
833 size_t n_controllers;
837 ofproto_set_sflow(br->ofproto, NULL);
841 memset(&oso, 0, sizeof oso);
843 sset_init(&oso.targets);
844 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
846 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
848 oso.sampling_rate = *cfg->sampling;
851 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
853 oso.polling_interval = *cfg->polling;
856 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
858 oso.header_len = *cfg->header;
861 oso.sub_id = (*sflow_bridge_number)++;
862 oso.agent_device = cfg->agent;
864 oso.control_ip = NULL;
865 n_controllers = bridge_get_controllers(br, &controllers);
866 for (i = 0; i < n_controllers; i++) {
867 if (controllers[i]->local_ip) {
868 oso.control_ip = controllers[i]->local_ip;
872 ofproto_set_sflow(br->ofproto, &oso);
874 sset_destroy(&oso.targets);
878 port_configure_stp(const struct ofproto *ofproto, struct port *port,
879 struct ofproto_port_stp_settings *port_s,
880 int *port_num_counter, unsigned long *port_num_bitmap)
882 const char *config_str;
885 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
886 if (config_str && !strcmp(config_str, "false")) {
887 port_s->enable = false;
890 port_s->enable = true;
893 /* STP over bonds is not supported. */
894 if (!list_is_singleton(&port->ifaces)) {
895 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
897 port_s->enable = false;
901 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
903 /* Internal ports shouldn't participate in spanning tree, so
905 if (!strcmp(iface->type, "internal")) {
906 VLOG_DBG("port %s: disable STP on internal ports", port->name);
907 port_s->enable = false;
911 /* STP on mirror output ports is not supported. */
912 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
913 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
914 port_s->enable = false;
918 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
920 unsigned long int port_num = strtoul(config_str, NULL, 0);
921 int port_idx = port_num - 1;
923 if (port_num < 1 || port_num > STP_MAX_PORTS) {
924 VLOG_ERR("port %s: invalid stp-port-num", port->name);
925 port_s->enable = false;
929 if (bitmap_is_set(port_num_bitmap, port_idx)) {
930 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
931 port->name, port_num);
932 port_s->enable = false;
935 bitmap_set1(port_num_bitmap, port_idx);
936 port_s->port_num = port_idx;
938 if (*port_num_counter > STP_MAX_PORTS) {
939 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
940 port_s->enable = false;
944 port_s->port_num = (*port_num_counter)++;
947 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
949 port_s->path_cost = strtoul(config_str, NULL, 10);
953 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
954 /* Couldn't get speed, so assume 100Mb/s. */
955 port_s->path_cost = 19;
959 mbps = netdev_features_to_bps(current) / 1000000;
960 port_s->path_cost = stp_convert_speed_to_cost(mbps);
964 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
966 port_s->priority = strtoul(config_str, NULL, 0);
968 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
972 /* Set spanning tree configuration on 'br'. */
974 bridge_configure_stp(struct bridge *br)
976 if (!br->cfg->stp_enable) {
977 ofproto_set_stp(br->ofproto, NULL);
979 struct ofproto_stp_settings br_s;
980 const char *config_str;
982 int port_num_counter;
983 unsigned long *port_num_bitmap;
985 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
987 uint8_t ea[ETH_ADDR_LEN];
989 if (eth_addr_from_string(config_str, ea)) {
990 br_s.system_id = eth_addr_to_uint64(ea);
992 br_s.system_id = eth_addr_to_uint64(br->ea);
993 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
994 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
997 br_s.system_id = eth_addr_to_uint64(br->ea);
1000 config_str = bridge_get_other_config(br->cfg, "stp-priority");
1002 br_s.priority = strtoul(config_str, NULL, 0);
1004 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1007 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
1009 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1011 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1014 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
1016 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1018 br_s.max_age = STP_DEFAULT_MAX_AGE;
1021 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
1023 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1025 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1028 /* Configure STP on the bridge. */
1029 if (ofproto_set_stp(br->ofproto, &br_s)) {
1030 VLOG_ERR("bridge %s: could not enable STP", br->name);
1034 /* Users must either set the port number with the "stp-port-num"
1035 * configuration on all ports or none. If manual configuration
1036 * is not done, then we allocate them sequentially. */
1037 port_num_counter = 0;
1038 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1039 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1040 struct ofproto_port_stp_settings port_s;
1041 struct iface *iface;
1043 port_configure_stp(br->ofproto, port, &port_s,
1044 &port_num_counter, port_num_bitmap);
1046 /* As bonds are not supported, just apply configuration to
1047 * all interfaces. */
1048 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1049 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1051 VLOG_ERR("port %s: could not enable STP", port->name);
1057 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1058 && port_num_counter) {
1059 VLOG_ERR("bridge %s: must manually configure all STP port "
1060 "IDs or none, disabling", br->name);
1061 ofproto_set_stp(br->ofproto, NULL);
1063 bitmap_free(port_num_bitmap);
1068 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1070 const struct port *port = port_lookup(br, name);
1071 return port && port_is_bond_fake_iface(port);
1075 port_is_bond_fake_iface(const struct port *port)
1077 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1081 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1083 struct bridge *br, *next;
1084 struct shash new_br;
1087 /* Collect new bridges' names and types. */
1088 shash_init(&new_br);
1089 for (i = 0; i < cfg->n_bridges; i++) {
1090 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1091 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1093 if (strchr(br_cfg->name, '/')) {
1094 /* Prevent remote ovsdb-server users from accessing arbitrary
1095 * directories, e.g. consider a bridge named "../../../etc/". */
1096 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1098 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1099 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1103 /* Get rid of deleted bridges or those whose types have changed.
1104 * Update 'cfg' of bridges that still exist. */
1105 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1106 br->cfg = shash_find_data(&new_br, br->name);
1107 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1108 br->cfg->datapath_type))) {
1113 /* Add new bridges. */
1114 for (i = 0; i < cfg->n_bridges; i++) {
1115 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1116 struct bridge *br = bridge_lookup(br_cfg->name);
1118 bridge_create(br_cfg);
1122 shash_destroy(&new_br);
1126 iface_set_ofp_port(struct iface *iface, int ofp_port)
1128 struct bridge *br = iface->port->bridge;
1130 assert(iface->ofp_port < 0 && ofp_port >= 0);
1131 iface->ofp_port = ofp_port;
1132 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1133 iface_set_ofport(iface->cfg, ofp_port);
1137 bridge_ofproto_port_del(struct bridge *br, struct ofproto_port ofproto_port)
1139 int error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1141 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1142 br->name, ofproto_port.name, strerror(error));
1144 VLOG_INFO("bridge %s: removed interface %s (%d)", br->name,
1145 ofproto_port.name, ofproto_port.ofp_port);
1149 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1150 * consistent with the ofp_ports in "br->ofproto". */
1152 bridge_refresh_ofp_port(struct bridge *br)
1154 struct ofproto_port_dump dump;
1155 struct ofproto_port ofproto_port;
1156 struct port *port, *port_next;
1158 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1159 hmap_clear(&br->ifaces);
1160 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1161 struct iface *iface;
1163 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1164 iface->ofp_port = -1;
1168 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1169 * already exist in the datapath and promote them to full fledged "struct
1170 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1171 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1172 struct iface *iface = iface_lookup(br, ofproto_port.name);
1174 if (iface->ofp_port >= 0) {
1175 VLOG_WARN("bridge %s: interface %s reported twice",
1176 br->name, ofproto_port.name);
1177 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1178 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1179 br->name, ofproto_port.ofp_port);
1180 } else if (!strcmp(ofproto_port.type, iface->type)) {
1181 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1183 /* Port has incorrect type so delete it later. */
1186 struct if_cfg *if_cfg = if_cfg_lookup(br, ofproto_port.name);
1189 iface_create(br, if_cfg, ofproto_port.ofp_port);
1190 } else if (bridge_has_bond_fake_iface(br, ofproto_port.name)
1191 && strcmp(ofproto_port.type, "internal")) {
1192 /* Bond fake iface with the wrong type. */
1193 bridge_ofproto_port_del(br, ofproto_port);
1195 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1196 garbage->ofp_port = ofproto_port.ofp_port;
1197 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1202 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1203 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1204 * later they can be added to ofproto. */
1205 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1206 struct iface *iface, *iface_next;
1208 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1209 if (iface->ofp_port < 0) {
1210 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1211 iface_destroy(iface);
1215 if (list_is_empty(&port->ifaces)) {
1221 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1222 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1223 * automatically allocated for the iface. Takes ownership of and
1224 * deallocates 'if_cfg'. */
1226 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1228 struct ofproto_port ofproto_port;
1229 struct iface *iface;
1233 assert(!iface_lookup(br, if_cfg->cfg->name));
1235 port = port_lookup(br, if_cfg->parent->name);
1237 port = port_create(br, if_cfg->parent);
1240 iface = xzalloc(sizeof *iface);
1242 iface->name = xstrdup(if_cfg->cfg->name);
1243 iface->ofp_port = -1;
1244 iface->netdev = NULL;
1245 iface->cfg = if_cfg->cfg;
1246 iface->need_refresh = true;
1247 hmap_insert(&br->iface_by_name, &iface->name_node,
1248 hash_string(iface->name, 0));
1249 list_push_back(&port->ifaces, &iface->port_elem);
1250 iface_refresh_type(iface);
1251 if (ofp_port >= 0) {
1252 iface_set_ofp_port(iface, ofp_port);
1255 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1259 error = netdev_open(iface->name, iface->type, &iface->netdev);
1261 VLOG_WARN("could not open network device %s (%s)", iface->name,
1266 && port->cfg->vlan_mode
1267 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1268 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1271 /* Configure the netdev. */
1272 if (iface->netdev) {
1276 shash_from_ovs_idl_map(iface->cfg->key_options,
1277 iface->cfg->value_options,
1278 iface->cfg->n_options, &args);
1279 error = netdev_set_config(iface->netdev, &args);
1280 shash_destroy(&args);
1283 VLOG_WARN("could not configure network device %s (%s)",
1284 iface->name, strerror(error));
1285 netdev_close(iface->netdev);
1286 iface->netdev = NULL;
1290 /* Add the port, if necessary. */
1291 if (iface->netdev && iface->ofp_port < 0) {
1292 uint16_t new_ofp_port;
1295 error = ofproto_port_add(br->ofproto, iface->netdev, &new_ofp_port);
1297 VLOG_INFO("bridge %s: added interface %s (%d)", br->name,
1298 iface->name, ofp_port);
1299 iface_set_ofp_port(iface, new_ofp_port);
1301 netdev_close(iface->netdev);
1302 iface->netdev = NULL;
1306 /* Populate stats columns in new Interface rows. */
1307 if (iface->netdev && iface->need_refresh) {
1308 iface_refresh_stats(iface);
1309 iface_refresh_status(iface);
1310 iface->need_refresh = false;
1313 /* Delete the iface if we failed. */
1314 if (iface->netdev && iface->ofp_port >= 0) {
1315 VLOG_DBG("bridge %s: interface %s is on port %d",
1316 br->name, iface->name, iface->ofp_port);
1318 if (iface->netdev) {
1319 VLOG_ERR("bridge %s: missing %s interface, dropping",
1320 br->name, iface->name);
1322 /* We already reported a related error, don't bother
1323 * duplicating it. */
1325 if (!ofproto_port_query_by_name(br->ofproto, port->name,
1327 VLOG_INFO("bridge %s: removed interface %s (%d)",
1328 br->name, port->name, ofproto_port.ofp_port);
1329 bridge_ofproto_port_del(br, ofproto_port);
1330 ofproto_port_destroy(&ofproto_port);
1332 iface_clear_db_record(iface->cfg);
1333 iface_destroy(iface);
1336 if (list_is_empty(&port->ifaces)) {
1341 /* Add bond fake iface if necessary. */
1342 if (port_is_bond_fake_iface(port)) {
1343 if (ofproto_port_query_by_name(br->ofproto, port->name,
1345 struct netdev *netdev;
1348 error = netdev_open(port->name, "internal", &netdev);
1350 ofproto_port_add(br->ofproto, netdev, NULL);
1351 netdev_close(netdev);
1353 VLOG_WARN("could not open network device %s (%s)",
1354 port->name, strerror(error));
1357 /* Already exists, nothing to do. */
1358 ofproto_port_destroy(&ofproto_port);
1364 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1368 for (i = 0; i < n; i++) {
1369 if (!strcmp(keys[i], key)) {
1377 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1379 return get_ovsrec_key_value(br_cfg->key_other_config,
1380 br_cfg->value_other_config,
1381 br_cfg->n_other_config, key);
1384 /* Set Flow eviction threshold */
1386 bridge_configure_flow_eviction_threshold(struct bridge *br)
1388 const char *threshold_str;
1391 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1392 if (threshold_str) {
1393 threshold = strtoul(threshold_str, NULL, 10);
1395 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1397 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1400 /* Set forward BPDU option. */
1402 bridge_configure_forward_bpdu(struct bridge *br)
1404 const char *forward_bpdu_str;
1405 bool forward_bpdu = false;
1407 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1408 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1409 forward_bpdu = true;
1411 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1414 /* Set MAC aging time for 'br'. */
1416 bridge_configure_mac_idle_time(struct bridge *br)
1418 const char *idle_time_str;
1421 idle_time_str = bridge_get_other_config(br->cfg, "mac-aging-time");
1422 idle_time = (idle_time_str && atoi(idle_time_str)
1423 ? atoi(idle_time_str)
1424 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1425 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1429 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1430 struct iface **hw_addr_iface)
1432 struct hmapx mirror_output_ports;
1435 bool found_addr = false;
1439 *hw_addr_iface = NULL;
1441 /* Did the user request a particular MAC? */
1442 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1443 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1444 if (eth_addr_is_multicast(ea)) {
1445 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1446 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1447 } else if (eth_addr_is_zero(ea)) {
1448 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1454 /* Mirror output ports don't participate in picking the local hardware
1455 * address. ofproto can't help us find out whether a given port is a
1456 * mirror output because we haven't configured mirrors yet, so we need to
1457 * accumulate them ourselves. */
1458 hmapx_init(&mirror_output_ports);
1459 for (i = 0; i < br->cfg->n_mirrors; i++) {
1460 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1461 if (m->output_port) {
1462 hmapx_add(&mirror_output_ports, m->output_port);
1466 /* Otherwise choose the minimum non-local MAC address among all of the
1468 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1469 uint8_t iface_ea[ETH_ADDR_LEN];
1470 struct iface *candidate;
1471 struct iface *iface;
1473 /* Mirror output ports don't participate. */
1474 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1478 /* Choose the MAC address to represent the port. */
1480 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1481 /* Find the interface with this Ethernet address (if any) so that
1482 * we can provide the correct devname to the caller. */
1483 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1484 uint8_t candidate_ea[ETH_ADDR_LEN];
1485 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1486 && eth_addr_equals(iface_ea, candidate_ea)) {
1491 /* Choose the interface whose MAC address will represent the port.
1492 * The Linux kernel bonding code always chooses the MAC address of
1493 * the first slave added to a bond, and the Fedora networking
1494 * scripts always add slaves to a bond in alphabetical order, so
1495 * for compatibility we choose the interface with the name that is
1496 * first in alphabetical order. */
1497 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1498 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1503 /* The local port doesn't count (since we're trying to choose its
1504 * MAC address anyway). */
1505 if (iface->ofp_port == OFPP_LOCAL) {
1510 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1516 /* Compare against our current choice. */
1517 if (!eth_addr_is_multicast(iface_ea) &&
1518 !eth_addr_is_local(iface_ea) &&
1519 !eth_addr_is_reserved(iface_ea) &&
1520 !eth_addr_is_zero(iface_ea) &&
1521 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1523 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1524 *hw_addr_iface = iface;
1529 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1530 br->name, ETH_ADDR_ARGS(ea));
1532 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1533 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1534 *hw_addr_iface = NULL;
1535 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1536 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1539 hmapx_destroy(&mirror_output_ports);
1542 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1543 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1544 * an interface on 'br', then that interface must be passed in as
1545 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1546 * 'hw_addr_iface' must be passed in as a null pointer. */
1548 bridge_pick_datapath_id(struct bridge *br,
1549 const uint8_t bridge_ea[ETH_ADDR_LEN],
1550 struct iface *hw_addr_iface)
1553 * The procedure for choosing a bridge MAC address will, in the most
1554 * ordinary case, also choose a unique MAC that we can use as a datapath
1555 * ID. In some special cases, though, multiple bridges will end up with
1556 * the same MAC address. This is OK for the bridges, but it will confuse
1557 * the OpenFlow controller, because each datapath needs a unique datapath
1560 * Datapath IDs must be unique. It is also very desirable that they be
1561 * stable from one run to the next, so that policy set on a datapath
1564 const char *datapath_id;
1567 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1568 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1572 if (!hw_addr_iface) {
1574 * A purely internal bridge, that is, one that has no non-virtual
1575 * network devices on it at all, is difficult because it has no
1576 * natural unique identifier at all.
1578 * When the host is a XenServer, we handle this case by hashing the
1579 * host's UUID with the name of the bridge. Names of bridges are
1580 * persistent across XenServer reboots, although they can be reused if
1581 * an internal network is destroyed and then a new one is later
1582 * created, so this is fairly effective.
1584 * When the host is not a XenServer, we punt by using a random MAC
1585 * address on each run.
1587 const char *host_uuid = xenserver_get_host_uuid();
1589 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1590 dpid = dpid_from_hash(combined, strlen(combined));
1596 return eth_addr_to_uint64(bridge_ea);
1600 dpid_from_hash(const void *data, size_t n)
1602 uint8_t hash[SHA1_DIGEST_SIZE];
1604 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1605 sha1_bytes(data, n, hash);
1606 eth_addr_mark_random(hash);
1607 return eth_addr_to_uint64(hash);
1611 iface_refresh_status(struct iface *iface)
1615 enum netdev_flags flags;
1622 if (iface_is_synthetic(iface)) {
1628 if (!netdev_get_status(iface->netdev, &sh)) {
1630 char **keys, **values;
1632 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1633 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1638 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1641 shash_destroy_free_data(&sh);
1643 error = netdev_get_flags(iface->netdev, &flags);
1645 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1648 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1651 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1653 ovsrec_interface_set_duplex(iface->cfg,
1654 netdev_features_is_full_duplex(current)
1656 /* warning: uint64_t -> int64_t conversion */
1657 bps = netdev_features_to_bps(current);
1658 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1661 ovsrec_interface_set_duplex(iface->cfg, NULL);
1662 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1665 error = netdev_get_mtu(iface->netdev, &mtu);
1668 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1671 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1675 /* Writes 'iface''s CFM statistics to the database. */
1677 iface_refresh_cfm_stats(struct iface *iface)
1679 const struct ovsrec_interface *cfg = iface->cfg;
1681 const uint64_t *rmps;
1684 if (iface_is_synthetic(iface)) {
1688 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1691 const char *reasons[CFM_FAULT_N_REASONS];
1692 bool fault_bool = fault;
1696 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1697 int reason = 1 << i;
1698 if (fault & reason) {
1699 reasons[j++] = cfm_fault_reason_to_str(reason);
1703 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1704 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1706 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1707 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1710 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1711 iface->ofp_port, &rmps, &n_rmps);
1713 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1716 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1721 iface_refresh_stats(struct iface *iface)
1723 #define IFACE_STATS \
1724 IFACE_STAT(rx_packets, "rx_packets") \
1725 IFACE_STAT(tx_packets, "tx_packets") \
1726 IFACE_STAT(rx_bytes, "rx_bytes") \
1727 IFACE_STAT(tx_bytes, "tx_bytes") \
1728 IFACE_STAT(rx_dropped, "rx_dropped") \
1729 IFACE_STAT(tx_dropped, "tx_dropped") \
1730 IFACE_STAT(rx_errors, "rx_errors") \
1731 IFACE_STAT(tx_errors, "tx_errors") \
1732 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1733 IFACE_STAT(rx_over_errors, "rx_over_err") \
1734 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1735 IFACE_STAT(collisions, "collisions")
1737 #define IFACE_STAT(MEMBER, NAME) NAME,
1738 static char *keys[] = { IFACE_STATS };
1740 int64_t values[ARRAY_SIZE(keys)];
1743 struct netdev_stats stats;
1745 if (iface_is_synthetic(iface)) {
1749 /* Intentionally ignore return value, since errors will set 'stats' to
1750 * all-1s, and we will deal with that correctly below. */
1751 netdev_get_stats(iface->netdev, &stats);
1753 /* Copy statistics into values[] array. */
1755 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1758 assert(i == ARRAY_SIZE(keys));
1760 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1765 br_refresh_stp_status(struct bridge *br)
1767 struct ofproto *ofproto = br->ofproto;
1768 struct ofproto_stp_status status;
1769 char *keys[3], *values[3];
1772 if (ofproto_get_stp_status(ofproto, &status)) {
1776 if (!status.enabled) {
1777 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1781 keys[0] = "stp_bridge_id",
1782 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1783 keys[1] = "stp_designated_root",
1784 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1785 keys[2] = "stp_root_path_cost",
1786 values[2] = xasprintf("%d", status.root_path_cost);
1788 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1790 for (i = 0; i < ARRAY_SIZE(values); i++) {
1796 port_refresh_stp_status(struct port *port)
1798 struct ofproto *ofproto = port->bridge->ofproto;
1799 struct iface *iface;
1800 struct ofproto_port_stp_status status;
1802 char *str_values[4];
1803 int64_t int_values[3];
1806 if (port_is_synthetic(port)) {
1810 /* STP doesn't currently support bonds. */
1811 if (!list_is_singleton(&port->ifaces)) {
1812 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1816 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1818 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1822 if (!status.enabled) {
1823 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1824 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1828 /* Set Status column. */
1829 keys[0] = "stp_port_id";
1830 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1831 keys[1] = "stp_state";
1832 str_values[1] = xstrdup(stp_state_name(status.state));
1833 keys[2] = "stp_sec_in_state";
1834 str_values[2] = xasprintf("%u", status.sec_in_state);
1835 keys[3] = "stp_role";
1836 str_values[3] = xstrdup(stp_role_name(status.role));
1838 ovsrec_port_set_status(port->cfg, keys, str_values,
1839 ARRAY_SIZE(str_values));
1841 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1842 free(str_values[i]);
1845 /* Set Statistics column. */
1846 keys[0] = "stp_tx_count";
1847 int_values[0] = status.tx_count;
1848 keys[1] = "stp_rx_count";
1849 int_values[1] = status.rx_count;
1850 keys[2] = "stp_error_count";
1851 int_values[2] = status.error_count;
1853 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1854 ARRAY_SIZE(int_values));
1858 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1862 /* Use other-config:enable-system-stats by preference. */
1863 enable = get_ovsrec_key_value(cfg->key_other_config,
1864 cfg->value_other_config,
1865 cfg->n_other_config,
1866 "enable-statistics");
1868 return !strcmp(enable, "true");
1871 /* Disable by default. */
1876 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1878 struct ovsdb_datum datum;
1882 if (enable_system_stats(cfg)) {
1883 get_system_stats(&stats);
1886 ovsdb_datum_from_shash(&datum, &stats);
1887 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1891 static inline const char *
1892 nx_role_to_str(enum nx_role role)
1897 case NX_ROLE_MASTER:
1902 return "*** INVALID ROLE ***";
1907 refresh_controller_status(void)
1911 const struct ovsrec_controller *cfg;
1915 /* Accumulate status for controllers on all bridges. */
1916 HMAP_FOR_EACH (br, node, &all_bridges) {
1917 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1920 /* Update each controller in the database with current status. */
1921 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1922 struct ofproto_controller_info *cinfo =
1923 shash_find_data(&info, cfg->target);
1926 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1927 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1928 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1929 (char **) cinfo->pairs.values,
1932 ovsrec_controller_set_is_connected(cfg, false);
1933 ovsrec_controller_set_role(cfg, NULL);
1934 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1938 ofproto_free_ofproto_controller_info(&info);
1942 refresh_cfm_stats(void)
1944 static struct ovsdb_idl_txn *txn = NULL;
1949 txn = ovsdb_idl_txn_create(idl);
1951 HMAP_FOR_EACH (br, node, &all_bridges) {
1952 struct iface *iface;
1954 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1955 iface_refresh_cfm_stats(iface);
1960 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1961 ovsdb_idl_txn_destroy(txn);
1966 /* Performs periodic activity required by bridges that needs to be done with
1967 * the least possible latency.
1969 * It makes sense to call this function a couple of times per poll loop, to
1970 * provide a significant performance boost on some benchmarks with ofprotos
1971 * that use the ofproto-dpif implementation. */
1973 bridge_run_fast(void)
1977 HMAP_FOR_EACH (br, node, &all_bridges) {
1978 ofproto_run_fast(br->ofproto);
1985 static const struct ovsrec_open_vswitch null_cfg;
1986 const struct ovsrec_open_vswitch *cfg;
1987 struct ovsdb_idl_txn *reconf_txn = NULL;
1989 bool vlan_splinters_changed;
1990 bool database_changed;
1993 /* (Re)configure if necessary. */
1994 if (!reconfiguring) {
1995 database_changed = ovsdb_idl_run(idl);
1996 if (ovsdb_idl_is_lock_contended(idl)) {
1997 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1998 struct bridge *br, *next_br;
2000 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2001 "disabling this process until it goes away");
2003 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2007 } else if (!ovsdb_idl_has_lock(idl)) {
2011 cfg = ovsrec_open_vswitch_first(idl);
2013 /* Let each bridge do the work that it needs to do. */
2014 HMAP_FOR_EACH (br, node, &all_bridges) {
2015 ofproto_run(br->ofproto);
2018 /* Re-configure SSL. We do this on every trip through the main loop,
2019 * instead of just when the database changes, because the contents of the
2020 * key and certificate files can change without the database changing.
2022 * We do this before bridge_reconfigure() because that function might
2023 * initiate SSL connections and thus requires SSL to be configured. */
2024 if (cfg && cfg->ssl) {
2025 const struct ovsrec_ssl *ssl = cfg->ssl;
2027 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2028 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2031 if (!reconfiguring) {
2032 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2033 * usage has changed. */
2034 vlan_splinters_changed = false;
2035 if (vlan_splinters_enabled_anywhere) {
2036 HMAP_FOR_EACH (br, node, &all_bridges) {
2037 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2038 vlan_splinters_changed = true;
2044 if (database_changed || vlan_splinters_changed) {
2046 reconf_txn = ovsdb_idl_txn_create(idl);
2047 bridge_reconfigure(cfg);
2049 /* We still need to reconfigure to avoid dangling pointers to
2050 * now-destroyed ovsrec structures inside bridge data. */
2051 bridge_reconfigure(&null_cfg);
2056 if (reconfiguring) {
2059 reconf_txn = ovsdb_idl_txn_create(idl);
2061 if (bridge_reconfigure_continue(cfg)) {
2062 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2065 bridge_reconfigure_continue(&null_cfg);
2070 ovsdb_idl_txn_commit(reconf_txn);
2071 ovsdb_idl_txn_destroy(reconf_txn);
2075 /* Refresh system and interface stats if necessary. */
2076 if (time_msec() >= stats_timer) {
2078 struct ovsdb_idl_txn *txn;
2080 txn = ovsdb_idl_txn_create(idl);
2081 HMAP_FOR_EACH (br, node, &all_bridges) {
2085 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2086 struct iface *iface;
2088 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2089 iface_refresh_stats(iface);
2090 iface_refresh_status(iface);
2094 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2095 mirror_refresh_stats(m);
2099 refresh_system_stats(cfg);
2100 refresh_controller_status();
2101 ovsdb_idl_txn_commit(txn);
2102 ovsdb_idl_txn_destroy(txn); /* XXX */
2105 stats_timer = time_msec() + STATS_INTERVAL;
2108 if (time_msec() >= db_limiter) {
2109 struct ovsdb_idl_txn *txn;
2111 txn = ovsdb_idl_txn_create(idl);
2112 HMAP_FOR_EACH (br, node, &all_bridges) {
2113 struct iface *iface;
2116 br_refresh_stp_status(br);
2118 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2119 port_refresh_stp_status(port);
2122 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2123 const char *link_state;
2124 int64_t link_resets;
2127 if (iface_is_synthetic(iface)) {
2131 current = ofproto_port_is_lacp_current(br->ofproto,
2135 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2137 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2140 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2141 ovsrec_interface_set_link_state(iface->cfg, link_state);
2143 link_resets = netdev_get_carrier_resets(iface->netdev);
2144 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2148 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2149 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2151 ovsdb_idl_txn_destroy(txn);
2154 refresh_cfm_stats();
2160 ovsdb_idl_wait(idl);
2162 if (reconfiguring) {
2163 poll_immediate_wake();
2166 if (!hmap_is_empty(&all_bridges)) {
2169 HMAP_FOR_EACH (br, node, &all_bridges) {
2170 ofproto_wait(br->ofproto);
2172 poll_timer_wait_until(stats_timer);
2174 if (db_limiter > time_msec()) {
2175 poll_timer_wait_until(db_limiter);
2180 /* QoS unixctl user interface functions. */
2182 struct qos_unixctl_show_cbdata {
2184 struct iface *iface;
2188 qos_unixctl_show_cb(unsigned int queue_id,
2189 const struct shash *details,
2192 struct qos_unixctl_show_cbdata *data = aux;
2193 struct ds *ds = data->ds;
2194 struct iface *iface = data->iface;
2195 struct netdev_queue_stats stats;
2196 struct shash_node *node;
2199 ds_put_cstr(ds, "\n");
2201 ds_put_format(ds, "Queue %u:\n", queue_id);
2203 ds_put_cstr(ds, "Default:\n");
2206 SHASH_FOR_EACH (node, details) {
2207 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2210 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2212 if (stats.tx_packets != UINT64_MAX) {
2213 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2216 if (stats.tx_bytes != UINT64_MAX) {
2217 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2220 if (stats.tx_errors != UINT64_MAX) {
2221 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2224 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2225 queue_id, strerror(error));
2230 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2231 const char *argv[], void *aux OVS_UNUSED)
2233 struct ds ds = DS_EMPTY_INITIALIZER;
2234 struct shash sh = SHASH_INITIALIZER(&sh);
2235 struct iface *iface;
2237 struct shash_node *node;
2238 struct qos_unixctl_show_cbdata data;
2241 iface = iface_find(argv[1]);
2243 unixctl_command_reply_error(conn, "no such interface");
2247 netdev_get_qos(iface->netdev, &type, &sh);
2249 if (*type != '\0') {
2250 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2252 SHASH_FOR_EACH (node, &sh) {
2253 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2258 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2261 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2263 unixctl_command_reply(conn, ds_cstr(&ds));
2265 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2266 unixctl_command_reply_error(conn, ds_cstr(&ds));
2269 shash_destroy_free_data(&sh);
2273 /* Bridge reconfiguration functions. */
2275 bridge_create(const struct ovsrec_bridge *br_cfg)
2279 assert(!bridge_lookup(br_cfg->name));
2280 br = xzalloc(sizeof *br);
2282 br->name = xstrdup(br_cfg->name);
2283 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2286 /* Derive the default Ethernet address from the bridge's UUID. This should
2287 * be unique and it will be stable between ovs-vswitchd runs. */
2288 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2289 eth_addr_mark_random(br->default_ea);
2291 hmap_init(&br->ports);
2292 hmap_init(&br->ifaces);
2293 hmap_init(&br->iface_by_name);
2294 hmap_init(&br->mirrors);
2296 hmap_init(&br->if_cfg_todo);
2297 list_init(&br->ofpp_garbage);
2299 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2303 bridge_destroy(struct bridge *br)
2306 struct mirror *mirror, *next_mirror;
2307 struct port *port, *next_port;
2308 struct if_cfg *if_cfg, *next_if_cfg;
2309 struct ofpp_garbage *garbage, *next_garbage;
2311 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2314 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2315 mirror_destroy(mirror);
2317 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2318 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2321 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2322 &br->ofpp_garbage) {
2323 list_remove(&garbage->list_node);
2327 hmap_remove(&all_bridges, &br->node);
2328 ofproto_destroy(br->ofproto);
2329 hmap_destroy(&br->ifaces);
2330 hmap_destroy(&br->ports);
2331 hmap_destroy(&br->iface_by_name);
2332 hmap_destroy(&br->mirrors);
2333 hmap_destroy(&br->if_cfg_todo);
2340 static struct bridge *
2341 bridge_lookup(const char *name)
2345 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2346 if (!strcmp(br->name, name)) {
2353 /* Handle requests for a listing of all flows known by the OpenFlow
2354 * stack, including those normally hidden. */
2356 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2357 const char *argv[], void *aux OVS_UNUSED)
2362 br = bridge_lookup(argv[1]);
2364 unixctl_command_reply_error(conn, "Unknown bridge");
2369 ofproto_get_all_flows(br->ofproto, &results);
2371 unixctl_command_reply(conn, ds_cstr(&results));
2372 ds_destroy(&results);
2375 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2376 * connections and reconnect. If BRIDGE is not specified, then all bridges
2377 * drop their controller connections and reconnect. */
2379 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2380 const char *argv[], void *aux OVS_UNUSED)
2384 br = bridge_lookup(argv[1]);
2386 unixctl_command_reply_error(conn, "Unknown bridge");
2389 ofproto_reconnect_controllers(br->ofproto);
2391 HMAP_FOR_EACH (br, node, &all_bridges) {
2392 ofproto_reconnect_controllers(br->ofproto);
2395 unixctl_command_reply(conn, NULL);
2399 bridge_get_controllers(const struct bridge *br,
2400 struct ovsrec_controller ***controllersp)
2402 struct ovsrec_controller **controllers;
2403 size_t n_controllers;
2405 controllers = br->cfg->controller;
2406 n_controllers = br->cfg->n_controller;
2408 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2414 *controllersp = controllers;
2416 return n_controllers;
2420 bridge_queue_if_cfg(struct bridge *br,
2421 const struct ovsrec_interface *cfg,
2422 const struct ovsrec_port *parent)
2424 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2427 if_cfg->parent = parent;
2428 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2429 hash_string(if_cfg->cfg->name, 0));
2432 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2433 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2434 * 'br' needs to complete its configuration. */
2436 bridge_add_del_ports(struct bridge *br,
2437 const unsigned long int *splinter_vlans)
2439 struct shash_node *port_node;
2440 struct port *port, *next;
2441 struct shash new_ports;
2444 assert(hmap_is_empty(&br->if_cfg_todo));
2446 /* Collect new ports. */
2447 shash_init(&new_ports);
2448 for (i = 0; i < br->cfg->n_ports; i++) {
2449 const char *name = br->cfg->ports[i]->name;
2450 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2451 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2455 if (bridge_get_controllers(br, NULL)
2456 && !shash_find(&new_ports, br->name)) {
2457 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2458 br->name, br->name);
2460 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2461 br->synth_local_port.n_interfaces = 1;
2462 br->synth_local_port.name = br->name;
2464 br->synth_local_iface.name = br->name;
2465 br->synth_local_iface.type = "internal";
2467 br->synth_local_ifacep = &br->synth_local_iface;
2469 shash_add(&new_ports, br->name, &br->synth_local_port);
2472 if (splinter_vlans) {
2473 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2476 /* Get rid of deleted ports.
2477 * Get rid of deleted interfaces on ports that still exist. */
2478 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2479 port->cfg = shash_find_data(&new_ports, port->name);
2483 port_del_ifaces(port);
2487 SHASH_FOR_EACH (port_node, &new_ports) {
2488 const struct ovsrec_port *port = port_node->data;
2491 for (i = 0; i < port->n_interfaces; i++) {
2492 const struct ovsrec_interface *cfg = port->interfaces[i];
2493 struct iface *iface = iface_lookup(br, cfg->name);
2497 iface_refresh_type(iface);
2499 bridge_queue_if_cfg(br, cfg, port);
2504 shash_destroy(&new_ports);
2507 /* Initializes 'oc' appropriately as a management service controller for
2510 * The caller must free oc->target when it is no longer needed. */
2512 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2513 struct ofproto_controller *oc)
2515 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2516 oc->max_backoff = 0;
2517 oc->probe_interval = 60;
2518 oc->band = OFPROTO_OUT_OF_BAND;
2520 oc->burst_limit = 0;
2521 oc->enable_async_msgs = true;
2524 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2526 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2527 struct ofproto_controller *oc)
2529 oc->target = c->target;
2530 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2531 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2532 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2533 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2534 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2535 oc->burst_limit = (c->controller_burst_limit
2536 ? *c->controller_burst_limit : 0);
2537 oc->enable_async_msgs = (!c->enable_async_messages
2538 || *c->enable_async_messages);
2541 /* Configures the IP stack for 'br''s local interface properly according to the
2542 * configuration in 'c'. */
2544 bridge_configure_local_iface_netdev(struct bridge *br,
2545 struct ovsrec_controller *c)
2547 struct netdev *netdev;
2548 struct in_addr mask, gateway;
2550 struct iface *local_iface;
2553 /* If there's no local interface or no IP address, give up. */
2554 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2555 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2559 /* Bring up the local interface. */
2560 netdev = local_iface->netdev;
2561 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2563 /* Configure the IP address and netmask. */
2564 if (!c->local_netmask
2565 || !inet_aton(c->local_netmask, &mask)
2567 mask.s_addr = guess_netmask(ip.s_addr);
2569 if (!netdev_set_in4(netdev, ip, mask)) {
2570 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2571 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2574 /* Configure the default gateway. */
2575 if (c->local_gateway
2576 && inet_aton(c->local_gateway, &gateway)
2577 && gateway.s_addr) {
2578 if (!netdev_add_router(netdev, gateway)) {
2579 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2580 br->name, IP_ARGS(&gateway.s_addr));
2585 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2586 * in either string are treated as equal to any number of slashes in the other,
2587 * e.g. "x///y" is equal to "x/y". */
2589 equal_pathnames(const char *a, const char *b)
2593 a += strspn(a, "/");
2594 b += strspn(b, "/");
2595 } else if (*a == '\0') {
2606 bridge_configure_remotes(struct bridge *br,
2607 const struct sockaddr_in *managers, size_t n_managers)
2609 const char *disable_ib_str, *queue_id_str;
2610 bool disable_in_band = false;
2613 struct ovsrec_controller **controllers;
2614 size_t n_controllers;
2616 enum ofproto_fail_mode fail_mode;
2618 struct ofproto_controller *ocs;
2622 /* Check if we should disable in-band control on this bridge. */
2623 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2624 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2625 disable_in_band = true;
2628 /* Set OpenFlow queue ID for in-band control. */
2629 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2630 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2631 ofproto_set_in_band_queue(br->ofproto, queue_id);
2633 if (disable_in_band) {
2634 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2636 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2639 n_controllers = bridge_get_controllers(br, &controllers);
2641 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2644 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2645 for (i = 0; i < n_controllers; i++) {
2646 struct ovsrec_controller *c = controllers[i];
2648 if (!strncmp(c->target, "punix:", 6)
2649 || !strncmp(c->target, "unix:", 5)) {
2650 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2653 whitelist = xasprintf("unix:%s/%s.controller",
2654 ovs_rundir(), br->name);
2655 if (!equal_pathnames(c->target, whitelist)) {
2656 /* Prevent remote ovsdb-server users from accessing arbitrary
2657 * Unix domain sockets and overwriting arbitrary local
2659 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2660 "controller \"%s\" due to possibility for remote "
2661 "exploit. Instead, specify whitelisted \"%s\" or "
2662 "connect to \"unix:%s/%s.mgmt\" (which is always "
2663 "available without special configuration).",
2664 br->name, c->target, whitelist,
2665 ovs_rundir(), br->name);
2673 bridge_configure_local_iface_netdev(br, c);
2674 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2675 if (disable_in_band) {
2676 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2681 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2682 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2685 /* Set the fail-mode. */
2686 fail_mode = !br->cfg->fail_mode
2687 || !strcmp(br->cfg->fail_mode, "standalone")
2688 ? OFPROTO_FAIL_STANDALONE
2689 : OFPROTO_FAIL_SECURE;
2690 ofproto_set_fail_mode(br->ofproto, fail_mode);
2692 /* Configure OpenFlow controller connection snooping. */
2693 if (!ofproto_has_snoops(br->ofproto)) {
2697 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2698 ovs_rundir(), br->name));
2699 ofproto_set_snoops(br->ofproto, &snoops);
2700 sset_destroy(&snoops);
2705 bridge_configure_tables(struct bridge *br)
2707 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2711 n_tables = ofproto_get_n_tables(br->ofproto);
2713 for (i = 0; i < n_tables; i++) {
2714 struct ofproto_table_settings s;
2717 s.max_flows = UINT_MAX;
2721 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2722 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2725 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2726 s.max_flows = *cfg->flow_limit;
2728 if (cfg->overflow_policy
2729 && !strcmp(cfg->overflow_policy, "evict")) {
2732 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2733 for (k = 0; k < cfg->n_groups; k++) {
2734 const char *string = cfg->groups[k];
2737 msg = mf_parse_subfield__(&s.groups[k], &string);
2739 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2740 "'groups' (%s)", br->name, i, msg);
2742 } else if (*string) {
2743 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2744 "element '%s' contains trailing garbage",
2745 br->name, i, cfg->groups[k]);
2753 ofproto_configure_table(br->ofproto, i, &s);
2757 for (; j < br->cfg->n_flow_tables; j++) {
2758 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2759 "%"PRId64" not supported by this datapath", br->name,
2760 br->cfg->key_flow_tables[j]);
2764 /* Port functions. */
2766 static struct port *
2767 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2771 port = xzalloc(sizeof *port);
2773 port->name = xstrdup(cfg->name);
2775 list_init(&port->ifaces);
2777 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2782 get_port_other_config(const struct ovsrec_port *port, const char *key,
2783 const char *default_value)
2787 value = get_ovsrec_key_value(port->key_other_config,
2788 port->value_other_config,
2789 port->n_other_config, key);
2790 return value ? value : default_value;
2794 get_interface_other_config(const struct ovsrec_interface *iface,
2795 const char *key, const char *default_value)
2799 value = get_ovsrec_key_value(iface->key_other_config,
2800 iface->value_other_config,
2801 iface->n_other_config, key);
2802 return value ? value : default_value;
2805 /* Deletes interfaces from 'port' that are no longer configured for it. */
2807 port_del_ifaces(struct port *port)
2809 struct iface *iface, *next;
2810 struct sset new_ifaces;
2813 /* Collect list of new interfaces. */
2814 sset_init(&new_ifaces);
2815 for (i = 0; i < port->cfg->n_interfaces; i++) {
2816 const char *name = port->cfg->interfaces[i]->name;
2817 const char *type = port->cfg->interfaces[i]->name;
2818 if (strcmp(type, "null")) {
2819 sset_add(&new_ifaces, name);
2823 /* Get rid of deleted interfaces. */
2824 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2825 if (!sset_contains(&new_ifaces, iface->name)) {
2826 iface_destroy(iface);
2830 sset_destroy(&new_ifaces);
2834 port_destroy(struct port *port)
2837 struct bridge *br = port->bridge;
2838 struct iface *iface, *next;
2841 ofproto_bundle_unregister(br->ofproto, port);
2844 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2845 iface_destroy(iface);
2848 hmap_remove(&br->ports, &port->hmap_node);
2854 static struct port *
2855 port_lookup(const struct bridge *br, const char *name)
2859 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2861 if (!strcmp(port->name, name)) {
2869 enable_lacp(struct port *port, bool *activep)
2871 if (!port->cfg->lacp) {
2872 /* XXX when LACP implementation has been sufficiently tested, enable by
2873 * default and make active on bonded ports. */
2875 } else if (!strcmp(port->cfg->lacp, "off")) {
2877 } else if (!strcmp(port->cfg->lacp, "active")) {
2880 } else if (!strcmp(port->cfg->lacp, "passive")) {
2884 VLOG_WARN("port %s: unknown LACP mode %s",
2885 port->name, port->cfg->lacp);
2890 static struct lacp_settings *
2891 port_configure_lacp(struct port *port, struct lacp_settings *s)
2893 const char *lacp_time, *system_id;
2894 long long int custom_time;
2897 if (!enable_lacp(port, &s->active)) {
2901 s->name = port->name;
2903 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2905 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2906 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2907 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2908 " address.", port->name, system_id);
2912 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2915 if (eth_addr_is_zero(s->id)) {
2916 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2920 /* Prefer bondable links if unspecified. */
2921 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2923 s->priority = (priority > 0 && priority <= UINT16_MAX
2925 : UINT16_MAX - !list_is_short(&port->ifaces));
2927 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2931 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2932 custom_time = atoi(lacp_time);
2933 if (!strcmp(lacp_time, "fast")) {
2934 s->lacp_time = LACP_TIME_FAST;
2935 } else if (!strcmp(lacp_time, "slow")) {
2936 s->lacp_time = LACP_TIME_SLOW;
2937 } else if (custom_time > 0) {
2938 s->lacp_time = LACP_TIME_CUSTOM;
2939 s->custom_time = custom_time;
2941 s->lacp_time = LACP_TIME_SLOW;
2948 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2950 int priority, portid, key;
2952 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2953 priority = atoi(get_interface_other_config(iface->cfg,
2954 "lacp-port-priority", "0"));
2955 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2958 if (portid <= 0 || portid > UINT16_MAX) {
2959 portid = iface->ofp_port;
2962 if (priority <= 0 || priority > UINT16_MAX) {
2963 priority = UINT16_MAX;
2966 if (key < 0 || key > UINT16_MAX) {
2970 s->name = iface->name;
2972 s->priority = priority;
2977 port_configure_bond(struct port *port, struct bond_settings *s,
2978 uint32_t *bond_stable_ids)
2980 const char *detect_s;
2981 struct iface *iface;
2982 int miimon_interval;
2985 s->name = port->name;
2987 if (port->cfg->bond_mode) {
2988 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2989 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2990 port->name, port->cfg->bond_mode,
2991 bond_mode_to_string(s->balance));
2994 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2996 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2997 * active-backup. At some point we should remove this warning. */
2998 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2999 " in previous versions, the default bond_mode was"
3000 " balance-slb", port->name,
3001 bond_mode_to_string(s->balance));
3003 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3004 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3005 "please use another bond type or disable flood_vlans",
3009 miimon_interval = atoi(get_port_other_config(port->cfg,
3010 "bond-miimon-interval", "0"));
3011 if (miimon_interval <= 0) {
3012 miimon_interval = 200;
3015 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
3016 if (!strcmp(detect_s, "carrier")) {
3017 miimon_interval = 0;
3018 } else if (strcmp(detect_s, "miimon")) {
3019 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3020 "defaulting to carrier", port->name, detect_s);
3021 miimon_interval = 0;
3024 s->up_delay = MAX(0, port->cfg->bond_updelay);
3025 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3026 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
3027 s->rebalance_interval = atoi(
3028 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3029 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3030 s->rebalance_interval = 1000;
3033 s->fake_iface = port->cfg->bond_fake_iface;
3036 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3037 long long stable_id;
3039 stable_id = atoll(get_interface_other_config(iface->cfg,
3040 "bond-stable-id", "0"));
3041 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3042 stable_id = iface->ofp_port;
3044 bond_stable_ids[i++] = stable_id;
3046 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3050 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3051 * instead of obtaining it from the database. */
3053 port_is_synthetic(const struct port *port)
3055 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3058 /* Interface functions. */
3061 iface_refresh_type(struct iface *iface)
3063 /* Determine interface type. The local port always has type
3064 * "internal". Other ports take their type from the database and
3065 * default to "system" if none is specified. */
3066 iface->type = (!strcmp(iface->name, iface->port->bridge->name) ? "internal"
3067 : iface->cfg->type[0] ? iface->cfg->type
3072 iface_destroy(struct iface *iface)
3075 struct port *port = iface->port;
3076 struct bridge *br = port->bridge;
3078 if (br->ofproto && iface->ofp_port >= 0) {
3079 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3082 if (iface->ofp_port >= 0) {
3083 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3086 list_remove(&iface->port_elem);
3087 hmap_remove(&br->iface_by_name, &iface->name_node);
3089 netdev_close(iface->netdev);
3096 static struct iface *
3097 iface_lookup(const struct bridge *br, const char *name)
3099 struct iface *iface;
3101 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3102 &br->iface_by_name) {
3103 if (!strcmp(iface->name, name)) {
3111 static struct iface *
3112 iface_find(const char *name)
3114 const struct bridge *br;
3116 HMAP_FOR_EACH (br, node, &all_bridges) {
3117 struct iface *iface = iface_lookup(br, name);
3126 static struct if_cfg *
3127 if_cfg_lookup(const struct bridge *br, const char *name)
3129 struct if_cfg *if_cfg;
3131 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3133 if (!strcmp(if_cfg->cfg->name, name)) {
3141 static struct iface *
3142 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3144 struct iface *iface;
3146 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3147 hash_int(ofp_port, 0), &br->ifaces) {
3148 if (iface->ofp_port == ofp_port) {
3155 /* Set Ethernet address of 'iface', if one is specified in the configuration
3158 iface_set_mac(struct iface *iface)
3160 uint8_t ea[ETH_ADDR_LEN];
3162 if (!strcmp(iface->type, "internal")
3163 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3164 if (iface->ofp_port == OFPP_LOCAL) {
3165 VLOG_ERR("interface %s: ignoring mac in Interface record "
3166 "(use Bridge record to set local port's mac)",
3168 } else if (eth_addr_is_multicast(ea)) {
3169 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3172 int error = netdev_set_etheraddr(iface->netdev, ea);
3174 VLOG_ERR("interface %s: setting MAC failed (%s)",
3175 iface->name, strerror(error));
3181 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3183 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3185 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3186 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3190 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3191 * sets the "ofport" field to -1.
3193 * This is appropriate when 'if_cfg''s interface cannot be created or is
3194 * otherwise invalid. */
3196 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3198 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3199 iface_set_ofport(if_cfg, -1);
3200 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3201 ovsrec_interface_set_admin_state(if_cfg, NULL);
3202 ovsrec_interface_set_duplex(if_cfg, NULL);
3203 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3204 ovsrec_interface_set_link_state(if_cfg, NULL);
3205 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3206 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3207 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3208 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3209 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3210 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3214 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3216 * The value strings in '*shash' are taken directly from values[], not copied,
3217 * so the caller should not modify or free them. */
3219 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3220 struct shash *shash)
3225 for (i = 0; i < n; i++) {
3226 shash_add(shash, keys[i], values[i]);
3230 /* Creates 'keys' and 'values' arrays from 'shash'.
3232 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3233 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3234 * are populated with with strings taken directly from 'shash' and thus have
3235 * the same ownership of the key-value pairs in shash.
3238 shash_to_ovs_idl_map(struct shash *shash,
3239 char ***keys, char ***values, size_t *n)
3243 struct shash_node *sn;
3245 count = shash_count(shash);
3247 k = xmalloc(count * sizeof *k);
3248 v = xmalloc(count * sizeof *v);
3251 SHASH_FOR_EACH(sn, shash) {
3262 struct iface_delete_queues_cbdata {
3263 struct netdev *netdev;
3264 const struct ovsdb_datum *queues;
3268 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3270 union ovsdb_atom atom;
3272 atom.integer = target;
3273 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3277 iface_delete_queues(unsigned int queue_id,
3278 const struct shash *details OVS_UNUSED, void *cbdata_)
3280 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3282 if (!queue_ids_include(cbdata->queues, queue_id)) {
3283 netdev_delete_queue(cbdata->netdev, queue_id);
3288 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3290 struct ofpbuf queues_buf;
3292 ofpbuf_init(&queues_buf, 0);
3294 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3295 netdev_set_qos(iface->netdev, NULL, NULL);
3297 struct iface_delete_queues_cbdata cbdata;
3298 struct shash details;
3302 /* Configure top-level Qos for 'iface'. */
3303 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3304 qos->n_other_config, &details);
3305 netdev_set_qos(iface->netdev, qos->type, &details);
3306 shash_destroy(&details);
3308 /* Deconfigure queues that were deleted. */
3309 cbdata.netdev = iface->netdev;
3310 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3312 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3314 /* Configure queues for 'iface'. */
3316 for (i = 0; i < qos->n_queues; i++) {
3317 const struct ovsrec_queue *queue = qos->value_queues[i];
3318 unsigned int queue_id = qos->key_queues[i];
3320 if (queue_id == 0) {
3324 if (queue->n_dscp == 1) {
3325 struct ofproto_port_queue *port_queue;
3327 port_queue = ofpbuf_put_uninit(&queues_buf,
3328 sizeof *port_queue);
3329 port_queue->queue = queue_id;
3330 port_queue->dscp = queue->dscp[0];
3333 shash_from_ovs_idl_map(queue->key_other_config,
3334 queue->value_other_config,
3335 queue->n_other_config, &details);
3336 netdev_set_queue(iface->netdev, queue_id, &details);
3337 shash_destroy(&details);
3340 shash_init(&details);
3341 netdev_set_queue(iface->netdev, 0, &details);
3342 shash_destroy(&details);
3346 if (iface->ofp_port >= 0) {
3347 const struct ofproto_port_queue *port_queues = queues_buf.data;
3348 size_t n_queues = queues_buf.size / sizeof *port_queues;
3350 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3351 port_queues, n_queues);
3354 netdev_set_policing(iface->netdev,
3355 iface->cfg->ingress_policing_rate,
3356 iface->cfg->ingress_policing_burst);
3358 ofpbuf_uninit(&queues_buf);
3362 iface_configure_cfm(struct iface *iface)
3364 const struct ovsrec_interface *cfg = iface->cfg;
3365 const char *extended_str, *opstate_str;
3366 struct cfm_settings s;
3368 if (!cfg->n_cfm_mpid) {
3369 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3373 s.mpid = *cfg->cfm_mpid;
3374 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3376 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3378 s.ccm_pcp = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_pcp",
3380 if (s.interval <= 0) {
3384 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3386 s.extended = !strcasecmp("true", extended_str);
3388 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3389 s.opup = !strcasecmp("up", opstate_str);
3391 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3394 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3395 * instead of obtaining it from the database. */
3397 iface_is_synthetic(const struct iface *iface)
3399 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3402 /* Port mirroring. */
3404 static struct mirror *
3405 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3409 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3410 if (uuid_equals(uuid, &m->uuid)) {
3418 bridge_configure_mirrors(struct bridge *br)
3420 const struct ovsdb_datum *mc;
3421 unsigned long *flood_vlans;
3422 struct mirror *m, *next;
3425 /* Get rid of deleted mirrors. */
3426 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3427 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3428 union ovsdb_atom atom;
3430 atom.uuid = m->uuid;
3431 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3436 /* Add new mirrors and reconfigure existing ones. */
3437 for (i = 0; i < br->cfg->n_mirrors; i++) {
3438 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3439 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3441 m = mirror_create(br, cfg);
3444 if (!mirror_configure(m)) {
3449 /* Update flooded vlans (for RSPAN). */
3450 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3451 br->cfg->n_flood_vlans);
3452 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3453 bitmap_free(flood_vlans);
3456 static struct mirror *
3457 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3461 m = xzalloc(sizeof *m);
3462 m->uuid = cfg->header_.uuid;
3463 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3465 m->name = xstrdup(cfg->name);
3471 mirror_destroy(struct mirror *m)
3474 struct bridge *br = m->bridge;
3477 ofproto_mirror_unregister(br->ofproto, m);
3480 hmap_remove(&br->mirrors, &m->hmap_node);
3487 mirror_collect_ports(struct mirror *m,
3488 struct ovsrec_port **in_ports, int n_in_ports,
3489 void ***out_portsp, size_t *n_out_portsp)
3491 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3492 size_t n_out_ports = 0;
3495 for (i = 0; i < n_in_ports; i++) {
3496 const char *name = in_ports[i]->name;
3497 struct port *port = port_lookup(m->bridge, name);
3499 out_ports[n_out_ports++] = port;
3501 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3502 "port %s", m->bridge->name, m->name, name);
3505 *out_portsp = out_ports;
3506 *n_out_portsp = n_out_ports;
3510 mirror_configure(struct mirror *m)
3512 const struct ovsrec_mirror *cfg = m->cfg;
3513 struct ofproto_mirror_settings s;
3516 if (strcmp(cfg->name, m->name)) {
3518 m->name = xstrdup(cfg->name);
3522 /* Get output port or VLAN. */
3523 if (cfg->output_port) {
3524 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3525 if (!s.out_bundle) {
3526 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3527 m->bridge->name, m->name);
3530 s.out_vlan = UINT16_MAX;
3532 if (cfg->output_vlan) {
3533 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3534 "output vlan; ignoring output vlan",
3535 m->bridge->name, m->name);
3537 } else if (cfg->output_vlan) {
3538 /* The database should prevent invalid VLAN values. */
3539 s.out_bundle = NULL;
3540 s.out_vlan = *cfg->output_vlan;
3542 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3543 m->bridge->name, m->name);
3547 /* Get port selection. */
3548 if (cfg->select_all) {
3549 size_t n_ports = hmap_count(&m->bridge->ports);
3550 void **ports = xmalloc(n_ports * sizeof *ports);
3555 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3565 /* Get ports, dropping ports that don't exist.
3566 * The IDL ensures that there are no duplicates. */
3567 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3568 &s.srcs, &s.n_srcs);
3569 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3570 &s.dsts, &s.n_dsts);
3573 /* Get VLAN selection. */
3574 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3577 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3580 if (s.srcs != s.dsts) {
3589 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3591 * This is deprecated. It is only for compatibility with broken device drivers
3592 * in old versions of Linux that do not properly support VLANs when VLAN
3593 * devices are not used. When broken device drivers are no longer in
3594 * widespread use, we will delete these interfaces. */
3596 static void **blocks;
3597 static size_t n_blocks, allocated_blocks;
3599 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3600 * VLAN splinters are reconfigured. */
3602 register_block(void *block)
3604 if (n_blocks >= allocated_blocks) {
3605 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3607 blocks[n_blocks++] = block;
3610 /* Frees all of the blocks registered with register_block(). */
3612 free_registered_blocks(void)
3616 for (i = 0; i < n_blocks; i++) {
3622 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3625 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3629 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3630 return !strcmp(value, "true");
3633 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3636 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3637 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3638 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3641 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3642 * use, returns NULL.
3644 * Updates 'vlan_splinters_enabled_anywhere'. */
3645 static unsigned long int *
3646 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3648 unsigned long int *splinter_vlans;
3649 struct sset splinter_ifaces;
3650 const char *real_dev_name;
3651 struct shash *real_devs;
3652 struct shash_node *node;
3656 /* Free space allocated for synthesized ports and interfaces, since we're
3657 * in the process of reconstructing all of them. */
3658 free_registered_blocks();
3660 splinter_vlans = bitmap_allocate(4096);
3661 sset_init(&splinter_ifaces);
3662 vlan_splinters_enabled_anywhere = false;
3663 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3664 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3667 for (j = 0; j < br_cfg->n_ports; j++) {
3668 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3671 for (k = 0; k < port_cfg->n_interfaces; k++) {
3672 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3674 if (vlan_splinters_is_enabled(iface_cfg)) {
3675 vlan_splinters_enabled_anywhere = true;
3676 sset_add(&splinter_ifaces, iface_cfg->name);
3677 vlan_bitmap_from_array__(port_cfg->trunks,
3683 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3684 bitmap_set1(splinter_vlans, *port_cfg->tag);
3689 if (!vlan_splinters_enabled_anywhere) {
3690 free(splinter_vlans);
3691 sset_destroy(&splinter_ifaces);
3695 HMAP_FOR_EACH (br, node, &all_bridges) {
3697 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3701 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3702 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3703 * device to be created for it. */
3704 bitmap_set0(splinter_vlans, 0);
3705 bitmap_set0(splinter_vlans, 4095);
3707 /* Delete all VLAN devices that we don't need. */
3709 real_devs = vlandev_get_real_devs();
3710 SHASH_FOR_EACH (node, real_devs) {
3711 const struct vlan_real_dev *real_dev = node->data;
3712 const struct vlan_dev *vlan_dev;
3713 bool real_dev_has_splinters;
3715 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3717 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3718 if (!real_dev_has_splinters
3719 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3720 struct netdev *netdev;
3722 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3723 if (!netdev_get_in4(netdev, NULL, NULL) ||
3724 !netdev_get_in6(netdev, NULL)) {
3725 vlandev_del(vlan_dev->name);
3727 /* It has an IP address configured, so we don't own
3728 * it. Don't delete it. */
3730 netdev_close(netdev);
3737 /* Add all VLAN devices that we need. */
3738 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3741 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3742 if (!vlandev_get_name(real_dev_name, vid)) {
3743 vlandev_add(real_dev_name, vid);
3750 sset_destroy(&splinter_ifaces);
3752 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3753 free(splinter_vlans);
3756 return splinter_vlans;
3759 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3762 configure_splinter_port(struct port *port)
3764 struct ofproto *ofproto = port->bridge->ofproto;
3765 uint16_t realdev_ofp_port;
3766 const char *realdev_name;
3767 struct iface *vlandev, *realdev;
3769 ofproto_bundle_unregister(port->bridge->ofproto, port);
3771 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3774 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3775 realdev = iface_lookup(port->bridge, realdev_name);
3776 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3778 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3782 static struct ovsrec_port *
3783 synthesize_splinter_port(const char *real_dev_name,
3784 const char *vlan_dev_name, int vid)
3786 struct ovsrec_interface *iface;
3787 struct ovsrec_port *port;
3789 iface = xzalloc(sizeof *iface);
3790 iface->name = xstrdup(vlan_dev_name);
3791 iface->type = "system";
3793 port = xzalloc(sizeof *port);
3794 port->interfaces = xmemdup(&iface, sizeof iface);
3795 port->n_interfaces = 1;
3796 port->name = xstrdup(vlan_dev_name);
3797 port->vlan_mode = "splinter";
3798 port->tag = xmalloc(sizeof *port->tag);
3800 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3801 port->key_other_config[0] = "realdev";
3802 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3803 port->value_other_config[0] = xstrdup(real_dev_name);
3804 port->n_other_config = 1;
3806 register_block(iface);
3807 register_block(iface->name);
3808 register_block(port);
3809 register_block(port->interfaces);
3810 register_block(port->name);
3811 register_block(port->tag);
3812 register_block(port->key_other_config);
3813 register_block(port->value_other_config);
3814 register_block(port->value_other_config[0]);
3819 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3820 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3821 * 1-bit in the 'splinter_vlans' bitmap. */
3823 add_vlan_splinter_ports(struct bridge *br,
3824 const unsigned long int *splinter_vlans,
3825 struct shash *ports)
3829 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3830 * we're modifying 'ports'. */
3831 for (i = 0; i < br->cfg->n_ports; i++) {
3832 const char *name = br->cfg->ports[i]->name;
3833 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3836 for (j = 0; j < port_cfg->n_interfaces; j++) {
3837 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3839 if (vlan_splinters_is_enabled(iface_cfg)) {
3840 const char *real_dev_name;
3843 real_dev_name = iface_cfg->name;
3844 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3845 const char *vlan_dev_name;
3847 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3849 && !shash_find(ports, vlan_dev_name)) {
3850 shash_add(ports, vlan_dev_name,
3851 synthesize_splinter_port(
3852 real_dev_name, vlan_dev_name, vid));
3861 mirror_refresh_stats(struct mirror *m)
3863 struct ofproto *ofproto = m->bridge->ofproto;
3864 uint64_t tx_packets, tx_bytes;
3867 size_t stat_cnt = 0;
3869 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3870 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3874 if (tx_packets != UINT64_MAX) {
3875 keys[stat_cnt] = "tx_packets";
3876 values[stat_cnt] = tx_packets;
3879 if (tx_bytes != UINT64_MAX) {
3880 keys[stat_cnt] = "tx_bytes";
3881 values[stat_cnt] = tx_bytes;
3885 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);