1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
71 /* OpenFlow port slated for removal from ofproto. */
73 struct list list_node; /* Node in bridge's ofpp_garbage. */
74 uint16_t ofp_port; /* Port to be deleted. */
78 /* These members are always valid. */
79 struct list port_elem; /* Element in struct port's "ifaces" list. */
80 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
84 /* These members are valid only after bridge_reconfigure() causes them to
86 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
87 int ofp_port; /* OpenFlow port number, -1 if unknown. */
88 struct netdev *netdev; /* Network device. */
89 const char *type; /* Usually same as cfg->type. */
90 const struct ovsrec_interface *cfg;
94 struct uuid uuid; /* UUID of this "mirror" record in database. */
95 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
96 struct bridge *bridge;
98 const struct ovsrec_mirror *cfg;
102 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
103 struct bridge *bridge;
106 const struct ovsrec_port *cfg;
108 /* An ordinary bridge port has 1 interface.
109 * A bridge port for bonding has at least 2 interfaces. */
110 struct list ifaces; /* List of "struct iface"s. */
114 struct hmap_node node; /* In 'all_bridges'. */
115 char *name; /* User-specified arbitrary name. */
116 char *type; /* Datapath type. */
117 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
118 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
119 const struct ovsrec_bridge *cfg;
121 /* OpenFlow switch processing. */
122 struct ofproto *ofproto; /* OpenFlow switch. */
125 struct hmap ports; /* "struct port"s indexed by name. */
126 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
127 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
129 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
130 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
131 Indexed on 'cfg->name'. */
133 /* Port mirroring. */
134 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
136 /* Synthetic local port if necessary. */
137 struct ovsrec_port synth_local_port;
138 struct ovsrec_interface synth_local_iface;
139 struct ovsrec_interface *synth_local_ifacep;
142 /* All bridges, indexed by name. */
143 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
145 /* OVSDB IDL used to obtain configuration. */
146 static struct ovsdb_idl *idl;
148 /* Most recently processed IDL sequence number. */
149 static unsigned int idl_seqno;
151 /* Each time this timer expires, the bridge fetches systems and interface
152 * statistics and pushes them into the database. */
153 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int stats_timer = LLONG_MIN;
156 /* Stores the time after which rate limited statistics may be written to the
157 * database. Only updated when changes to the database require rate limiting.
159 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
160 static long long int db_limiter = LLONG_MIN;
162 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
163 * expensive. This can cause bridge_reconfigure() to take a long time during
164 * which no other work can be done. To deal with this problem, we limit port
165 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
166 * call to bridge_reconfigure(). If there is more work to do after the limit
167 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
168 * This allows the rest of the code to catch up on important things like
169 * forwarding packets. */
170 #define OFP_PORT_ACTION_WINDOW 10
171 static bool reconfiguring = false;
173 static void add_del_bridges(const struct ovsrec_open_vswitch *);
174 static void bridge_update_ofprotos(void);
175 static void bridge_create(const struct ovsrec_bridge *);
176 static void bridge_destroy(struct bridge *);
177 static struct bridge *bridge_lookup(const char *name);
178 static unixctl_cb_func bridge_unixctl_dump_flows;
179 static unixctl_cb_func bridge_unixctl_reconnect;
180 static size_t bridge_get_controllers(const struct bridge *br,
181 struct ovsrec_controller ***controllersp);
182 static void bridge_add_del_ports(struct bridge *,
183 const unsigned long int *splinter_vlans);
184 static void bridge_refresh_ofp_port(struct bridge *);
185 static void bridge_configure_datapath_id(struct bridge *);
186 static void bridge_configure_flow_eviction_threshold(struct bridge *);
187 static void bridge_configure_netflow(struct bridge *);
188 static void bridge_configure_forward_bpdu(struct bridge *);
189 static void bridge_configure_mac_idle_time(struct bridge *);
190 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
191 static void bridge_configure_stp(struct bridge *);
192 static void bridge_configure_tables(struct bridge *);
193 static void bridge_configure_remotes(struct bridge *,
194 const struct sockaddr_in *managers,
196 static void bridge_pick_local_hw_addr(struct bridge *,
197 uint8_t ea[ETH_ADDR_LEN],
198 struct iface **hw_addr_iface);
199 static uint64_t bridge_pick_datapath_id(struct bridge *,
200 const uint8_t bridge_ea[ETH_ADDR_LEN],
201 struct iface *hw_addr_iface);
202 static void bridge_queue_if_cfg(struct bridge *,
203 const struct ovsrec_interface *,
204 const struct ovsrec_port *);
205 static uint64_t dpid_from_hash(const void *, size_t nbytes);
206 static bool bridge_has_bond_fake_iface(const struct bridge *,
208 static bool port_is_bond_fake_iface(const struct port *);
210 static unixctl_cb_func qos_unixctl_show;
212 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
213 static void port_del_ifaces(struct port *);
214 static void port_destroy(struct port *);
215 static struct port *port_lookup(const struct bridge *, const char *name);
216 static void port_configure(struct port *);
217 static struct lacp_settings *port_configure_lacp(struct port *,
218 struct lacp_settings *);
219 static void port_configure_bond(struct port *, struct bond_settings *,
220 uint32_t *bond_stable_ids);
221 static bool port_is_synthetic(const struct port *);
223 static void bridge_configure_mirrors(struct bridge *);
224 static struct mirror *mirror_create(struct bridge *,
225 const struct ovsrec_mirror *);
226 static void mirror_destroy(struct mirror *);
227 static bool mirror_configure(struct mirror *);
228 static void mirror_refresh_stats(struct mirror *);
230 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
231 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
232 static const char *iface_get_type(const struct ovsrec_interface *,
233 const struct ovsrec_bridge *);
234 static void iface_destroy(struct iface *);
235 static struct iface *iface_lookup(const struct bridge *, const char *name);
236 static struct iface *iface_find(const char *name);
237 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
238 static struct iface *iface_from_ofp_port(const struct bridge *,
240 static void iface_set_mac(struct iface *);
241 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
242 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
243 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
244 static void iface_configure_cfm(struct iface *);
245 static void iface_refresh_cfm_stats(struct iface *);
246 static void iface_refresh_stats(struct iface *);
247 static void iface_refresh_status(struct iface *);
248 static bool iface_is_synthetic(const struct iface *);
250 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
252 * This is deprecated. It is only for compatibility with broken device drivers
253 * in old versions of Linux that do not properly support VLANs when VLAN
254 * devices are not used. When broken device drivers are no longer in
255 * widespread use, we will delete these interfaces. */
257 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
258 static bool vlan_splinters_enabled_anywhere;
260 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
261 static unsigned long int *collect_splinter_vlans(
262 const struct ovsrec_open_vswitch *);
263 static void configure_splinter_port(struct port *);
264 static void add_vlan_splinter_ports(struct bridge *,
265 const unsigned long int *splinter_vlans,
266 struct shash *ports);
268 /* Public functions. */
270 /* Initializes the bridge module, configuring it to obtain its configuration
271 * from an OVSDB server accessed over 'remote', which should be a string in a
272 * form acceptable to ovsdb_idl_create(). */
274 bridge_init(const char *remote)
276 /* Create connection to database. */
277 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
278 idl_seqno = ovsdb_idl_get_seqno(idl);
279 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
281 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
282 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
289 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
290 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
291 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
293 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
294 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
295 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
312 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
314 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
315 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
316 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
317 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
319 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
321 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
323 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
324 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
326 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
338 /* Register unixctl commands. */
339 unixctl_command_register("qos/show", "interface", 1, 1,
340 qos_unixctl_show, NULL);
341 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
342 bridge_unixctl_dump_flows, NULL);
343 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
344 bridge_unixctl_reconnect, NULL);
354 struct bridge *br, *next_br;
356 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
359 ovsdb_idl_destroy(idl);
362 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
363 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
364 * responsible for freeing '*managersp' (with free()).
366 * You may be asking yourself "why does ovs-vswitchd care?", because
367 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
368 * should not be and in fact is not directly involved in that. But
369 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
370 * it has to tell in-band control where the managers are to enable that.
371 * (Thus, only managers connected in-band are collected.)
374 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
375 struct sockaddr_in **managersp, size_t *n_managersp)
377 struct sockaddr_in *managers = NULL;
378 size_t n_managers = 0;
382 /* Collect all of the potential targets from the "targets" columns of the
383 * rows pointed to by "manager_options", excluding any that are
386 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
387 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
389 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
390 sset_find_and_delete(&targets, m->target);
392 sset_add(&targets, m->target);
396 /* Now extract the targets' IP addresses. */
397 if (!sset_is_empty(&targets)) {
400 managers = xmalloc(sset_count(&targets) * sizeof *managers);
401 SSET_FOR_EACH (target, &targets) {
402 struct sockaddr_in *sin = &managers[n_managers];
404 if (stream_parse_target_with_default_ports(target,
412 sset_destroy(&targets);
414 *managersp = managers;
415 *n_managersp = n_managers;
419 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
421 unsigned long int *splinter_vlans;
424 COVERAGE_INC(bridge_reconfigure);
426 assert(!reconfiguring);
427 reconfiguring = true;
429 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
430 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
431 * configuration otherwise.
433 * This is mostly an update to bridge data structures. Nothing is pushed
434 * down to ofproto or lower layers. */
435 add_del_bridges(ovs_cfg);
436 splinter_vlans = collect_splinter_vlans(ovs_cfg);
437 HMAP_FOR_EACH (br, node, &all_bridges) {
438 bridge_add_del_ports(br, splinter_vlans);
440 free(splinter_vlans);
442 /* Delete datapaths that are no longer configured, and create ones which
443 * don't exist but should. */
444 bridge_update_ofprotos();
446 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
447 HMAP_FOR_EACH (br, node, &all_bridges) {
448 bridge_refresh_ofp_port(br);
451 /* Clear database records for "if_cfg"s which haven't been instantiated. */
452 HMAP_FOR_EACH (br, node, &all_bridges) {
453 struct if_cfg *if_cfg;
455 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
456 iface_clear_db_record(if_cfg->cfg);
462 bridge_reconfigure_ofp(void)
464 long long int deadline;
468 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
470 /* The kernel will reject any attempt to add a given port to a datapath if
471 * that port already belongs to a different datapath, so we must do all
472 * port deletions before any port additions. */
473 HMAP_FOR_EACH (br, node, &all_bridges) {
474 struct ofpp_garbage *garbage, *next;
476 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
477 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
478 * internal datastructures may not be consistent. Eventually, when
479 * port additions and deletions are cheaper, these calls should be
482 ofproto_port_del(br->ofproto, garbage->ofp_port);
483 list_remove(&garbage->list_node);
487 if (time_msec() >= deadline) {
494 HMAP_FOR_EACH (br, node, &all_bridges) {
495 struct if_cfg *if_cfg, *next;
497 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
498 iface_create(br, if_cfg, -1);
500 if (time_msec() >= deadline) {
510 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
512 struct sockaddr_in *managers;
513 int sflow_bridge_number;
518 assert(reconfiguring);
519 done = bridge_reconfigure_ofp();
521 /* Complete the configuration. */
522 sflow_bridge_number = 0;
523 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
524 HMAP_FOR_EACH (br, node, &all_bridges) {
527 /* We need the datapath ID early to allow LACP ports to use it as the
528 * default system ID. */
529 bridge_configure_datapath_id(br);
531 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
534 port_configure(port);
536 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
537 iface_configure_cfm(iface);
538 iface_configure_qos(iface, port->cfg->qos);
539 iface_set_mac(iface);
542 bridge_configure_mirrors(br);
543 bridge_configure_flow_eviction_threshold(br);
544 bridge_configure_forward_bpdu(br);
545 bridge_configure_mac_idle_time(br);
546 bridge_configure_remotes(br, managers, n_managers);
547 bridge_configure_netflow(br);
548 bridge_configure_sflow(br, &sflow_bridge_number);
549 bridge_configure_stp(br);
550 bridge_configure_tables(br);
555 /* ovs-vswitchd has completed initialization, so allow the process that
556 * forked us to exit successfully. */
557 daemonize_complete();
558 reconfiguring = false;
560 VLOG_INFO("%s (Open vSwitch) %s", program_name, VERSION);
566 /* Delete ofprotos which aren't configured or have the wrong type. Create
567 * ofprotos which don't exist but need to. */
569 bridge_update_ofprotos(void)
571 struct bridge *br, *next;
576 /* Delete ofprotos with no bridge or with the wrong type. */
579 ofproto_enumerate_types(&types);
580 SSET_FOR_EACH (type, &types) {
583 ofproto_enumerate_names(type, &names);
584 SSET_FOR_EACH (name, &names) {
585 br = bridge_lookup(name);
586 if (!br || strcmp(type, br->type)) {
587 ofproto_delete(name, type);
591 sset_destroy(&names);
592 sset_destroy(&types);
594 /* Add ofprotos for bridges which don't have one yet. */
595 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
603 /* Remove ports from any datapath with the same name as 'br'. If we
604 * don't do this, creating 'br''s ofproto will fail because a port with
605 * the same name as its local port already exists. */
606 HMAP_FOR_EACH (br2, node, &all_bridges) {
607 struct ofproto_port ofproto_port;
613 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
615 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
617 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
620 ofproto_port_destroy(&ofproto_port);
624 error = ofproto_create(br->name, br->type, &br->ofproto);
626 VLOG_ERR("failed to create bridge %s: %s", br->name,
634 port_configure(struct port *port)
636 const struct ovsrec_port *cfg = port->cfg;
637 struct bond_settings bond_settings;
638 struct lacp_settings lacp_settings;
639 struct ofproto_bundle_settings s;
642 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
643 configure_splinter_port(port);
652 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
653 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
654 s.slaves[s.n_slaves++] = iface->ofp_port;
659 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
663 /* Get VLAN trunks. */
666 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
670 if (cfg->vlan_mode) {
671 if (!strcmp(cfg->vlan_mode, "access")) {
672 s.vlan_mode = PORT_VLAN_ACCESS;
673 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
674 s.vlan_mode = PORT_VLAN_TRUNK;
675 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
676 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
677 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
678 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
680 /* This "can't happen" because ovsdb-server should prevent it. */
681 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
682 s.vlan_mode = PORT_VLAN_TRUNK;
686 s.vlan_mode = PORT_VLAN_ACCESS;
688 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
692 s.vlan_mode = PORT_VLAN_TRUNK;
695 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
698 /* Get LACP settings. */
699 s.lacp = port_configure_lacp(port, &lacp_settings);
703 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
704 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
705 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
708 s.lacp_slaves = NULL;
711 /* Get bond settings. */
712 if (s.n_slaves > 1) {
713 s.bond = &bond_settings;
714 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
715 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
718 s.bond_stable_ids = NULL;
720 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
721 netdev_set_miimon_interval(iface->netdev, 0);
726 ofproto_bundle_register(port->bridge->ofproto, port, &s);
732 free(s.bond_stable_ids);
735 /* Pick local port hardware address and datapath ID for 'br'. */
737 bridge_configure_datapath_id(struct bridge *br)
739 uint8_t ea[ETH_ADDR_LEN];
741 struct iface *local_iface;
742 struct iface *hw_addr_iface;
745 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
746 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
748 int error = netdev_set_etheraddr(local_iface->netdev, ea);
750 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
751 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
752 "Ethernet address: %s",
753 br->name, strerror(error));
756 memcpy(br->ea, ea, ETH_ADDR_LEN);
758 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
759 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
760 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
761 ofproto_set_datapath_id(br->ofproto, dpid);
764 dpid_string = xasprintf("%016"PRIx64, dpid);
765 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
769 /* Set NetFlow configuration on 'br'. */
771 bridge_configure_netflow(struct bridge *br)
773 struct ovsrec_netflow *cfg = br->cfg->netflow;
774 struct netflow_options opts;
777 ofproto_set_netflow(br->ofproto, NULL);
781 memset(&opts, 0, sizeof opts);
783 /* Get default NetFlow configuration from datapath.
784 * Apply overrides from 'cfg'. */
785 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
786 if (cfg->engine_type) {
787 opts.engine_type = *cfg->engine_type;
789 if (cfg->engine_id) {
790 opts.engine_id = *cfg->engine_id;
793 /* Configure active timeout interval. */
794 opts.active_timeout = cfg->active_timeout;
795 if (!opts.active_timeout) {
796 opts.active_timeout = -1;
797 } else if (opts.active_timeout < 0) {
798 VLOG_WARN("bridge %s: active timeout interval set to negative "
799 "value, using default instead (%d seconds)", br->name,
800 NF_ACTIVE_TIMEOUT_DEFAULT);
801 opts.active_timeout = -1;
804 /* Add engine ID to interface number to disambiguate bridgs? */
805 opts.add_id_to_iface = cfg->add_id_to_interface;
806 if (opts.add_id_to_iface) {
807 if (opts.engine_id > 0x7f) {
808 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
809 "another vswitch, choose an engine id less than 128",
812 if (hmap_count(&br->ports) > 508) {
813 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
814 "another port when more than 508 ports are used",
820 sset_init(&opts.collectors);
821 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
824 if (ofproto_set_netflow(br->ofproto, &opts)) {
825 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
827 sset_destroy(&opts.collectors);
830 /* Set sFlow configuration on 'br'. */
832 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
834 const struct ovsrec_sflow *cfg = br->cfg->sflow;
835 struct ovsrec_controller **controllers;
836 struct ofproto_sflow_options oso;
837 size_t n_controllers;
841 ofproto_set_sflow(br->ofproto, NULL);
845 memset(&oso, 0, sizeof oso);
847 sset_init(&oso.targets);
848 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
850 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
852 oso.sampling_rate = *cfg->sampling;
855 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
857 oso.polling_interval = *cfg->polling;
860 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
862 oso.header_len = *cfg->header;
865 oso.sub_id = (*sflow_bridge_number)++;
866 oso.agent_device = cfg->agent;
868 oso.control_ip = NULL;
869 n_controllers = bridge_get_controllers(br, &controllers);
870 for (i = 0; i < n_controllers; i++) {
871 if (controllers[i]->local_ip) {
872 oso.control_ip = controllers[i]->local_ip;
876 ofproto_set_sflow(br->ofproto, &oso);
878 sset_destroy(&oso.targets);
882 port_configure_stp(const struct ofproto *ofproto, struct port *port,
883 struct ofproto_port_stp_settings *port_s,
884 int *port_num_counter, unsigned long *port_num_bitmap)
886 const char *config_str;
889 if (smap_get_bool(&port->cfg->other_config, "stp-enable", false)) {
890 port_s->enable = false;
893 port_s->enable = true;
896 /* STP over bonds is not supported. */
897 if (!list_is_singleton(&port->ifaces)) {
898 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
900 port_s->enable = false;
904 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
906 /* Internal ports shouldn't participate in spanning tree, so
908 if (!strcmp(iface->type, "internal")) {
909 VLOG_DBG("port %s: disable STP on internal ports", port->name);
910 port_s->enable = false;
914 /* STP on mirror output ports is not supported. */
915 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
916 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
917 port_s->enable = false;
921 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
923 unsigned long int port_num = strtoul(config_str, NULL, 0);
924 int port_idx = port_num - 1;
926 if (port_num < 1 || port_num > STP_MAX_PORTS) {
927 VLOG_ERR("port %s: invalid stp-port-num", port->name);
928 port_s->enable = false;
932 if (bitmap_is_set(port_num_bitmap, port_idx)) {
933 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
934 port->name, port_num);
935 port_s->enable = false;
938 bitmap_set1(port_num_bitmap, port_idx);
939 port_s->port_num = port_idx;
941 if (*port_num_counter > STP_MAX_PORTS) {
942 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
943 port_s->enable = false;
947 port_s->port_num = (*port_num_counter)++;
950 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
952 port_s->path_cost = strtoul(config_str, NULL, 10);
954 enum netdev_features current;
956 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
957 /* Couldn't get speed, so assume 100Mb/s. */
958 port_s->path_cost = 19;
962 mbps = netdev_features_to_bps(current) / 1000000;
963 port_s->path_cost = stp_convert_speed_to_cost(mbps);
967 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
969 port_s->priority = strtoul(config_str, NULL, 0);
971 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
975 /* Set spanning tree configuration on 'br'. */
977 bridge_configure_stp(struct bridge *br)
979 if (!br->cfg->stp_enable) {
980 ofproto_set_stp(br->ofproto, NULL);
982 struct ofproto_stp_settings br_s;
983 const char *config_str;
985 int port_num_counter;
986 unsigned long *port_num_bitmap;
988 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
990 uint8_t ea[ETH_ADDR_LEN];
992 if (eth_addr_from_string(config_str, ea)) {
993 br_s.system_id = eth_addr_to_uint64(ea);
995 br_s.system_id = eth_addr_to_uint64(br->ea);
996 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
997 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1000 br_s.system_id = eth_addr_to_uint64(br->ea);
1003 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1005 br_s.priority = strtoul(config_str, NULL, 0);
1007 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1010 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1012 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1014 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1017 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1019 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1021 br_s.max_age = STP_DEFAULT_MAX_AGE;
1024 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1026 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1028 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1031 /* Configure STP on the bridge. */
1032 if (ofproto_set_stp(br->ofproto, &br_s)) {
1033 VLOG_ERR("bridge %s: could not enable STP", br->name);
1037 /* Users must either set the port number with the "stp-port-num"
1038 * configuration on all ports or none. If manual configuration
1039 * is not done, then we allocate them sequentially. */
1040 port_num_counter = 0;
1041 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1042 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1043 struct ofproto_port_stp_settings port_s;
1044 struct iface *iface;
1046 port_configure_stp(br->ofproto, port, &port_s,
1047 &port_num_counter, port_num_bitmap);
1049 /* As bonds are not supported, just apply configuration to
1050 * all interfaces. */
1051 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1052 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1054 VLOG_ERR("port %s: could not enable STP", port->name);
1060 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1061 && port_num_counter) {
1062 VLOG_ERR("bridge %s: must manually configure all STP port "
1063 "IDs or none, disabling", br->name);
1064 ofproto_set_stp(br->ofproto, NULL);
1066 bitmap_free(port_num_bitmap);
1071 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1073 const struct port *port = port_lookup(br, name);
1074 return port && port_is_bond_fake_iface(port);
1078 port_is_bond_fake_iface(const struct port *port)
1080 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1084 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1086 struct bridge *br, *next;
1087 struct shash new_br;
1090 /* Collect new bridges' names and types. */
1091 shash_init(&new_br);
1092 for (i = 0; i < cfg->n_bridges; i++) {
1093 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1094 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1096 if (strchr(br_cfg->name, '/')) {
1097 /* Prevent remote ovsdb-server users from accessing arbitrary
1098 * directories, e.g. consider a bridge named "../../../etc/". */
1099 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1101 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1102 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1106 /* Get rid of deleted bridges or those whose types have changed.
1107 * Update 'cfg' of bridges that still exist. */
1108 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1109 br->cfg = shash_find_data(&new_br, br->name);
1110 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1111 br->cfg->datapath_type))) {
1116 /* Add new bridges. */
1117 for (i = 0; i < cfg->n_bridges; i++) {
1118 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1119 struct bridge *br = bridge_lookup(br_cfg->name);
1121 bridge_create(br_cfg);
1125 shash_destroy(&new_br);
1129 iface_set_ofp_port(struct iface *iface, int ofp_port)
1131 struct bridge *br = iface->port->bridge;
1133 assert(iface->ofp_port < 0 && ofp_port >= 0);
1134 iface->ofp_port = ofp_port;
1135 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1136 iface_set_ofport(iface->cfg, ofp_port);
1139 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1140 * Returns 0 if successful, otherwise a positive errno value. */
1142 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1143 struct netdev *netdev)
1147 error = netdev_set_config(netdev, &iface_cfg->options);
1149 VLOG_WARN("could not configure network device %s (%s)",
1150 iface_cfg->name, strerror(error));
1155 /* This function determines whether 'ofproto_port', which is attached to
1156 * br->ofproto's datapath, is one that we want in 'br'.
1158 * If it is, it returns true, after creating an iface (if necessary),
1159 * configuring the iface's netdev according to the iface's options, and setting
1160 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1162 * If, on the other hand, 'port' should be removed, it returns false. The
1163 * caller should later detach the port from br->ofproto. */
1165 bridge_refresh_one_ofp_port(struct bridge *br,
1166 const struct ofproto_port *ofproto_port)
1168 const char *name = ofproto_port->name;
1169 const char *type = ofproto_port->type;
1170 uint16_t ofp_port = ofproto_port->ofp_port;
1172 struct iface *iface = iface_lookup(br, name);
1174 /* Check that the name-to-number mapping is one-to-one. */
1175 if (iface->ofp_port >= 0) {
1176 VLOG_WARN("bridge %s: interface %s reported twice",
1179 } else if (iface_from_ofp_port(br, ofp_port)) {
1180 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1181 br->name, ofp_port);
1185 /* There's a configured interface named 'name'. */
1186 if (strcmp(type, iface->type)
1187 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1188 /* It's the wrong type, or it's the right type but can't be
1189 * configured as the user requested, so we must destroy it. */
1192 /* It's the right type and configured correctly. keep it. */
1193 iface_set_ofp_port(iface, ofp_port);
1196 } else if (bridge_has_bond_fake_iface(br, name)
1197 && !strcmp(type, "internal")) {
1198 /* It's a bond fake iface. Keep it. */
1201 /* There's no configured interface named 'name', but there might be an
1202 * interface of that name queued to be created.
1204 * If there is, and it has the correct type, then try to configure it
1205 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1206 * delete it and later try to re-add it. */
1207 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1209 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1210 && iface_create(br, if_cfg, ofp_port));
1214 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1215 * consistent with the ofp_ports in "br->ofproto". */
1217 bridge_refresh_ofp_port(struct bridge *br)
1219 struct ofproto_port_dump dump;
1220 struct ofproto_port ofproto_port;
1221 struct port *port, *port_next;
1223 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1224 hmap_clear(&br->ifaces);
1225 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1226 struct iface *iface;
1228 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1229 iface->ofp_port = -1;
1233 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1234 * already exist in the datapath and promote them to full fledged "struct
1235 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1236 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1237 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1238 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1239 garbage->ofp_port = ofproto_port.ofp_port;
1240 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1244 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1245 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1246 * later they can be added to ofproto. */
1247 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1248 struct iface *iface, *iface_next;
1250 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1251 if (iface->ofp_port < 0) {
1252 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1253 iface_destroy(iface);
1257 if (list_is_empty(&port->ifaces)) {
1263 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1264 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1265 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1268 * If successful, returns 0 and stores the network device in '*netdevp'. On
1269 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1271 iface_do_create(const struct bridge *br,
1272 const struct ovsrec_interface *iface_cfg,
1273 const struct ovsrec_port *port_cfg,
1274 int *ofp_portp, struct netdev **netdevp)
1276 struct netdev *netdev;
1279 error = netdev_open(iface_cfg->name,
1280 iface_get_type(iface_cfg, br->cfg), &netdev);
1282 VLOG_WARN("could not open network device %s (%s)",
1283 iface_cfg->name, strerror(error));
1287 error = iface_set_netdev_config(iface_cfg, netdev);
1292 if (*ofp_portp < 0) {
1295 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1299 *ofp_portp = ofp_port;
1301 VLOG_INFO("bridge %s: added interface %s on port %d",
1302 br->name, iface_cfg->name, *ofp_portp);
1304 VLOG_DBG("bridge %s: interface %s is on port %d",
1305 br->name, iface_cfg->name, *ofp_portp);
1308 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1309 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1317 netdev_close(netdev);
1321 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1322 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1323 * automatically allocated for the iface. Takes ownership of and
1324 * deallocates 'if_cfg'.
1326 * Return true if an iface is successfully created, false otherwise. */
1328 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1330 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1331 const struct ovsrec_port *port_cfg = if_cfg->parent;
1333 struct netdev *netdev;
1334 struct iface *iface;
1338 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1340 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1343 /* Do the bits that can fail up front.
1345 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1346 * internal datastructures may not be consistent. Eventually, when port
1347 * additions and deletions are cheaper, these calls should be removed. */
1349 assert(!iface_lookup(br, iface_cfg->name));
1350 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1353 iface_clear_db_record(iface_cfg);
1357 /* Get or create the port structure. */
1358 port = port_lookup(br, port_cfg->name);
1360 port = port_create(br, port_cfg);
1363 /* Create the iface structure. */
1364 iface = xzalloc(sizeof *iface);
1365 list_push_back(&port->ifaces, &iface->port_elem);
1366 hmap_insert(&br->iface_by_name, &iface->name_node,
1367 hash_string(iface_cfg->name, 0));
1369 iface->name = xstrdup(iface_cfg->name);
1370 iface->ofp_port = -1;
1371 iface->netdev = netdev;
1372 iface->type = iface_get_type(iface_cfg, br->cfg);
1373 iface->cfg = iface_cfg;
1375 iface_set_ofp_port(iface, ofp_port);
1377 /* Populate initial status in database. */
1378 iface_refresh_stats(iface);
1379 iface_refresh_status(iface);
1381 /* Add bond fake iface if necessary. */
1382 if (port_is_bond_fake_iface(port)) {
1383 struct ofproto_port ofproto_port;
1385 if (ofproto_port_query_by_name(br->ofproto, port->name,
1387 struct netdev *netdev;
1390 error = netdev_open(port->name, "internal", &netdev);
1392 ofproto_port_add(br->ofproto, netdev, NULL);
1393 netdev_close(netdev);
1395 VLOG_WARN("could not open network device %s (%s)",
1396 port->name, strerror(error));
1399 /* Already exists, nothing to do. */
1400 ofproto_port_destroy(&ofproto_port);
1407 /* Set Flow eviction threshold */
1409 bridge_configure_flow_eviction_threshold(struct bridge *br)
1411 const char *threshold_str;
1414 threshold_str = smap_get(&br->cfg->other_config,
1415 "flow-eviction-threshold");
1416 if (threshold_str) {
1417 threshold = strtoul(threshold_str, NULL, 10);
1419 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1421 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1424 /* Set forward BPDU option. */
1426 bridge_configure_forward_bpdu(struct bridge *br)
1428 ofproto_set_forward_bpdu(br->ofproto,
1429 smap_get_bool(&br->cfg->other_config,
1434 /* Set MAC aging time for 'br'. */
1436 bridge_configure_mac_idle_time(struct bridge *br)
1438 const char *idle_time_str;
1441 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1442 idle_time = (idle_time_str && atoi(idle_time_str)
1443 ? atoi(idle_time_str)
1444 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1445 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1449 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1450 struct iface **hw_addr_iface)
1452 struct hmapx mirror_output_ports;
1455 bool found_addr = false;
1459 *hw_addr_iface = NULL;
1461 /* Did the user request a particular MAC? */
1462 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1463 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1464 if (eth_addr_is_multicast(ea)) {
1465 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1466 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1467 } else if (eth_addr_is_zero(ea)) {
1468 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1474 /* Mirror output ports don't participate in picking the local hardware
1475 * address. ofproto can't help us find out whether a given port is a
1476 * mirror output because we haven't configured mirrors yet, so we need to
1477 * accumulate them ourselves. */
1478 hmapx_init(&mirror_output_ports);
1479 for (i = 0; i < br->cfg->n_mirrors; i++) {
1480 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1481 if (m->output_port) {
1482 hmapx_add(&mirror_output_ports, m->output_port);
1486 /* Otherwise choose the minimum non-local MAC address among all of the
1488 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1489 uint8_t iface_ea[ETH_ADDR_LEN];
1490 struct iface *candidate;
1491 struct iface *iface;
1493 /* Mirror output ports don't participate. */
1494 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1498 /* Choose the MAC address to represent the port. */
1500 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1501 /* Find the interface with this Ethernet address (if any) so that
1502 * we can provide the correct devname to the caller. */
1503 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1504 uint8_t candidate_ea[ETH_ADDR_LEN];
1505 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1506 && eth_addr_equals(iface_ea, candidate_ea)) {
1511 /* Choose the interface whose MAC address will represent the port.
1512 * The Linux kernel bonding code always chooses the MAC address of
1513 * the first slave added to a bond, and the Fedora networking
1514 * scripts always add slaves to a bond in alphabetical order, so
1515 * for compatibility we choose the interface with the name that is
1516 * first in alphabetical order. */
1517 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1518 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1523 /* The local port doesn't count (since we're trying to choose its
1524 * MAC address anyway). */
1525 if (iface->ofp_port == OFPP_LOCAL) {
1530 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1536 /* Compare against our current choice. */
1537 if (!eth_addr_is_multicast(iface_ea) &&
1538 !eth_addr_is_local(iface_ea) &&
1539 !eth_addr_is_reserved(iface_ea) &&
1540 !eth_addr_is_zero(iface_ea) &&
1541 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1543 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1544 *hw_addr_iface = iface;
1549 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1550 br->name, ETH_ADDR_ARGS(ea));
1552 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1553 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1554 *hw_addr_iface = NULL;
1555 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1556 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1559 hmapx_destroy(&mirror_output_ports);
1562 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1563 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1564 * an interface on 'br', then that interface must be passed in as
1565 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1566 * 'hw_addr_iface' must be passed in as a null pointer. */
1568 bridge_pick_datapath_id(struct bridge *br,
1569 const uint8_t bridge_ea[ETH_ADDR_LEN],
1570 struct iface *hw_addr_iface)
1573 * The procedure for choosing a bridge MAC address will, in the most
1574 * ordinary case, also choose a unique MAC that we can use as a datapath
1575 * ID. In some special cases, though, multiple bridges will end up with
1576 * the same MAC address. This is OK for the bridges, but it will confuse
1577 * the OpenFlow controller, because each datapath needs a unique datapath
1580 * Datapath IDs must be unique. It is also very desirable that they be
1581 * stable from one run to the next, so that policy set on a datapath
1584 const char *datapath_id;
1587 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1588 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1592 if (!hw_addr_iface) {
1594 * A purely internal bridge, that is, one that has no non-virtual
1595 * network devices on it at all, is difficult because it has no
1596 * natural unique identifier at all.
1598 * When the host is a XenServer, we handle this case by hashing the
1599 * host's UUID with the name of the bridge. Names of bridges are
1600 * persistent across XenServer reboots, although they can be reused if
1601 * an internal network is destroyed and then a new one is later
1602 * created, so this is fairly effective.
1604 * When the host is not a XenServer, we punt by using a random MAC
1605 * address on each run.
1607 const char *host_uuid = xenserver_get_host_uuid();
1609 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1610 dpid = dpid_from_hash(combined, strlen(combined));
1616 return eth_addr_to_uint64(bridge_ea);
1620 dpid_from_hash(const void *data, size_t n)
1622 uint8_t hash[SHA1_DIGEST_SIZE];
1624 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1625 sha1_bytes(data, n, hash);
1626 eth_addr_mark_random(hash);
1627 return eth_addr_to_uint64(hash);
1631 iface_refresh_status(struct iface *iface)
1635 enum netdev_features current;
1636 enum netdev_flags flags;
1642 if (iface_is_synthetic(iface)) {
1648 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1649 ovsrec_interface_set_status(iface->cfg, &smap);
1651 ovsrec_interface_set_status(iface->cfg, NULL);
1654 smap_destroy(&smap);
1656 error = netdev_get_flags(iface->netdev, &flags);
1658 ovsrec_interface_set_admin_state(iface->cfg,
1659 flags & NETDEV_UP ? "up" : "down");
1662 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1665 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1667 ovsrec_interface_set_duplex(iface->cfg,
1668 netdev_features_is_full_duplex(current)
1670 /* warning: uint64_t -> int64_t conversion */
1671 bps = netdev_features_to_bps(current);
1672 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1675 ovsrec_interface_set_duplex(iface->cfg, NULL);
1676 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1679 error = netdev_get_mtu(iface->netdev, &mtu);
1682 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1685 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1689 /* Writes 'iface''s CFM statistics to the database. */
1691 iface_refresh_cfm_stats(struct iface *iface)
1693 const struct ovsrec_interface *cfg = iface->cfg;
1695 const uint64_t *rmps;
1699 if (iface_is_synthetic(iface)) {
1703 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1706 const char *reasons[CFM_FAULT_N_REASONS];
1707 bool fault_bool = fault;
1711 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1712 int reason = 1 << i;
1713 if (fault & reason) {
1714 reasons[j++] = cfm_fault_reason_to_str(reason);
1718 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1719 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1721 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1722 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1725 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1726 iface->ofp_port, &rmps, &n_rmps);
1728 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1731 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1734 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1737 int64_t cfm_health = health;
1738 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1740 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1745 iface_refresh_stats(struct iface *iface)
1747 #define IFACE_STATS \
1748 IFACE_STAT(rx_packets, "rx_packets") \
1749 IFACE_STAT(tx_packets, "tx_packets") \
1750 IFACE_STAT(rx_bytes, "rx_bytes") \
1751 IFACE_STAT(tx_bytes, "tx_bytes") \
1752 IFACE_STAT(rx_dropped, "rx_dropped") \
1753 IFACE_STAT(tx_dropped, "tx_dropped") \
1754 IFACE_STAT(rx_errors, "rx_errors") \
1755 IFACE_STAT(tx_errors, "tx_errors") \
1756 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1757 IFACE_STAT(rx_over_errors, "rx_over_err") \
1758 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1759 IFACE_STAT(collisions, "collisions")
1761 #define IFACE_STAT(MEMBER, NAME) NAME,
1762 static char *keys[] = { IFACE_STATS };
1764 int64_t values[ARRAY_SIZE(keys)];
1767 struct netdev_stats stats;
1769 if (iface_is_synthetic(iface)) {
1773 /* Intentionally ignore return value, since errors will set 'stats' to
1774 * all-1s, and we will deal with that correctly below. */
1775 netdev_get_stats(iface->netdev, &stats);
1777 /* Copy statistics into values[] array. */
1779 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1782 assert(i == ARRAY_SIZE(keys));
1784 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1790 br_refresh_stp_status(struct bridge *br)
1792 struct smap smap = SMAP_INITIALIZER(&smap);
1793 struct ofproto *ofproto = br->ofproto;
1794 struct ofproto_stp_status status;
1796 if (ofproto_get_stp_status(ofproto, &status)) {
1800 if (!status.enabled) {
1801 ovsrec_bridge_set_status(br->cfg, NULL);
1805 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1806 STP_ID_ARGS(status.bridge_id));
1807 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1808 STP_ID_ARGS(status.designated_root));
1809 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1811 ovsrec_bridge_set_status(br->cfg, &smap);
1812 smap_destroy(&smap);
1816 port_refresh_stp_status(struct port *port)
1818 struct ofproto *ofproto = port->bridge->ofproto;
1819 struct iface *iface;
1820 struct ofproto_port_stp_status status;
1822 int64_t int_values[3];
1825 if (port_is_synthetic(port)) {
1829 /* STP doesn't currently support bonds. */
1830 if (!list_is_singleton(&port->ifaces)) {
1831 ovsrec_port_set_status(port->cfg, NULL);
1835 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1837 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1841 if (!status.enabled) {
1842 ovsrec_port_set_status(port->cfg, NULL);
1843 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1847 /* Set Status column. */
1849 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1850 smap_add(&smap, "stp_state", stp_state_name(status.state));
1851 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1852 smap_add(&smap, "stp_role", stp_role_name(status.role));
1853 ovsrec_port_set_status(port->cfg, &smap);
1854 smap_destroy(&smap);
1856 /* Set Statistics column. */
1857 keys[0] = "stp_tx_count";
1858 int_values[0] = status.tx_count;
1859 keys[1] = "stp_rx_count";
1860 int_values[1] = status.rx_count;
1861 keys[2] = "stp_error_count";
1862 int_values[2] = status.error_count;
1864 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1865 ARRAY_SIZE(int_values));
1869 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1871 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1875 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1877 struct ovsdb_datum datum;
1881 if (enable_system_stats(cfg)) {
1882 get_system_stats(&stats);
1885 ovsdb_datum_from_shash(&datum, &stats);
1886 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1890 static inline const char *
1891 nx_role_to_str(enum nx_role role)
1896 case NX_ROLE_MASTER:
1901 return "*** INVALID ROLE ***";
1906 refresh_controller_status(void)
1910 const struct ovsrec_controller *cfg;
1914 /* Accumulate status for controllers on all bridges. */
1915 HMAP_FOR_EACH (br, node, &all_bridges) {
1916 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1919 /* Update each controller in the database with current status. */
1920 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1921 struct ofproto_controller_info *cinfo =
1922 shash_find_data(&info, cfg->target);
1925 struct smap smap = SMAP_INITIALIZER(&smap);
1926 const char **values = cinfo->pairs.values;
1927 const char **keys = cinfo->pairs.keys;
1930 for (i = 0; i < cinfo->pairs.n; i++) {
1931 smap_add(&smap, keys[i], values[i]);
1934 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1935 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1936 ovsrec_controller_set_status(cfg, &smap);
1937 smap_destroy(&smap);
1939 ovsrec_controller_set_is_connected(cfg, false);
1940 ovsrec_controller_set_role(cfg, NULL);
1941 ovsrec_controller_set_status(cfg, NULL);
1945 ofproto_free_ofproto_controller_info(&info);
1949 refresh_cfm_stats(void)
1951 static struct ovsdb_idl_txn *txn = NULL;
1956 txn = ovsdb_idl_txn_create(idl);
1958 HMAP_FOR_EACH (br, node, &all_bridges) {
1959 struct iface *iface;
1961 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1962 iface_refresh_cfm_stats(iface);
1967 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1968 ovsdb_idl_txn_destroy(txn);
1973 /* Performs periodic activity required by bridges that needs to be done with
1974 * the least possible latency.
1976 * It makes sense to call this function a couple of times per poll loop, to
1977 * provide a significant performance boost on some benchmarks with ofprotos
1978 * that use the ofproto-dpif implementation. */
1980 bridge_run_fast(void)
1984 HMAP_FOR_EACH (br, node, &all_bridges) {
1985 ofproto_run_fast(br->ofproto);
1992 static const struct ovsrec_open_vswitch null_cfg;
1993 const struct ovsrec_open_vswitch *cfg;
1994 struct ovsdb_idl_txn *reconf_txn = NULL;
1996 bool vlan_splinters_changed;
1999 /* (Re)configure if necessary. */
2000 if (!reconfiguring) {
2003 if (ovsdb_idl_is_lock_contended(idl)) {
2004 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2005 struct bridge *br, *next_br;
2007 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2008 "disabling this process until it goes away");
2010 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2014 } else if (!ovsdb_idl_has_lock(idl)) {
2018 cfg = ovsrec_open_vswitch_first(idl);
2020 /* Let each bridge do the work that it needs to do. */
2021 HMAP_FOR_EACH (br, node, &all_bridges) {
2022 ofproto_run(br->ofproto);
2025 /* Re-configure SSL. We do this on every trip through the main loop,
2026 * instead of just when the database changes, because the contents of the
2027 * key and certificate files can change without the database changing.
2029 * We do this before bridge_reconfigure() because that function might
2030 * initiate SSL connections and thus requires SSL to be configured. */
2031 if (cfg && cfg->ssl) {
2032 const struct ovsrec_ssl *ssl = cfg->ssl;
2034 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2035 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2038 if (!reconfiguring) {
2039 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2040 * usage has changed. */
2041 vlan_splinters_changed = false;
2042 if (vlan_splinters_enabled_anywhere) {
2043 HMAP_FOR_EACH (br, node, &all_bridges) {
2044 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2045 vlan_splinters_changed = true;
2051 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2052 idl_seqno = ovsdb_idl_get_seqno(idl);
2054 reconf_txn = ovsdb_idl_txn_create(idl);
2055 bridge_reconfigure(cfg);
2057 /* We still need to reconfigure to avoid dangling pointers to
2058 * now-destroyed ovsrec structures inside bridge data. */
2059 bridge_reconfigure(&null_cfg);
2064 if (reconfiguring) {
2067 reconf_txn = ovsdb_idl_txn_create(idl);
2069 if (bridge_reconfigure_continue(cfg)) {
2070 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2073 bridge_reconfigure_continue(&null_cfg);
2078 ovsdb_idl_txn_commit(reconf_txn);
2079 ovsdb_idl_txn_destroy(reconf_txn);
2083 /* Refresh system and interface stats if necessary. */
2084 if (time_msec() >= stats_timer) {
2086 struct ovsdb_idl_txn *txn;
2088 txn = ovsdb_idl_txn_create(idl);
2089 HMAP_FOR_EACH (br, node, &all_bridges) {
2093 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2094 struct iface *iface;
2096 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2097 iface_refresh_stats(iface);
2098 iface_refresh_status(iface);
2102 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2103 mirror_refresh_stats(m);
2107 refresh_system_stats(cfg);
2108 refresh_controller_status();
2109 ovsdb_idl_txn_commit(txn);
2110 ovsdb_idl_txn_destroy(txn); /* XXX */
2113 stats_timer = time_msec() + STATS_INTERVAL;
2116 if (time_msec() >= db_limiter) {
2117 struct ovsdb_idl_txn *txn;
2119 txn = ovsdb_idl_txn_create(idl);
2120 HMAP_FOR_EACH (br, node, &all_bridges) {
2121 struct iface *iface;
2124 br_refresh_stp_status(br);
2126 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2127 port_refresh_stp_status(port);
2130 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2131 const char *link_state;
2132 int64_t link_resets;
2135 if (iface_is_synthetic(iface)) {
2139 current = ofproto_port_is_lacp_current(br->ofproto,
2143 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2145 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2148 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2149 ovsrec_interface_set_link_state(iface->cfg, link_state);
2151 link_resets = netdev_get_carrier_resets(iface->netdev);
2152 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2156 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2157 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2159 ovsdb_idl_txn_destroy(txn);
2162 refresh_cfm_stats();
2168 ovsdb_idl_wait(idl);
2170 if (reconfiguring) {
2171 poll_immediate_wake();
2174 if (!hmap_is_empty(&all_bridges)) {
2177 HMAP_FOR_EACH (br, node, &all_bridges) {
2178 ofproto_wait(br->ofproto);
2180 poll_timer_wait_until(stats_timer);
2182 if (db_limiter > time_msec()) {
2183 poll_timer_wait_until(db_limiter);
2188 /* Adds some memory usage statistics for bridges into 'usage', for use with
2189 * memory_report(). */
2191 bridge_get_memory_usage(struct simap *usage)
2195 HMAP_FOR_EACH (br, node, &all_bridges) {
2196 ofproto_get_memory_usage(br->ofproto, usage);
2200 /* QoS unixctl user interface functions. */
2202 struct qos_unixctl_show_cbdata {
2204 struct iface *iface;
2208 qos_unixctl_show_cb(unsigned int queue_id,
2209 const struct smap *details,
2212 struct qos_unixctl_show_cbdata *data = aux;
2213 struct ds *ds = data->ds;
2214 struct iface *iface = data->iface;
2215 struct netdev_queue_stats stats;
2216 struct smap_node *node;
2219 ds_put_cstr(ds, "\n");
2221 ds_put_format(ds, "Queue %u:\n", queue_id);
2223 ds_put_cstr(ds, "Default:\n");
2226 SMAP_FOR_EACH (node, details) {
2227 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2230 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2232 if (stats.tx_packets != UINT64_MAX) {
2233 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2236 if (stats.tx_bytes != UINT64_MAX) {
2237 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2240 if (stats.tx_errors != UINT64_MAX) {
2241 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2244 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2245 queue_id, strerror(error));
2250 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2251 const char *argv[], void *aux OVS_UNUSED)
2253 struct ds ds = DS_EMPTY_INITIALIZER;
2254 struct smap smap = SMAP_INITIALIZER(&smap);
2255 struct iface *iface;
2257 struct smap_node *node;
2258 struct qos_unixctl_show_cbdata data;
2261 iface = iface_find(argv[1]);
2263 unixctl_command_reply_error(conn, "no such interface");
2267 netdev_get_qos(iface->netdev, &type, &smap);
2269 if (*type != '\0') {
2270 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2272 SMAP_FOR_EACH (node, &smap) {
2273 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2278 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2281 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2283 unixctl_command_reply(conn, ds_cstr(&ds));
2285 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2286 unixctl_command_reply_error(conn, ds_cstr(&ds));
2289 smap_destroy(&smap);
2293 /* Bridge reconfiguration functions. */
2295 bridge_create(const struct ovsrec_bridge *br_cfg)
2299 assert(!bridge_lookup(br_cfg->name));
2300 br = xzalloc(sizeof *br);
2302 br->name = xstrdup(br_cfg->name);
2303 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2306 /* Derive the default Ethernet address from the bridge's UUID. This should
2307 * be unique and it will be stable between ovs-vswitchd runs. */
2308 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2309 eth_addr_mark_random(br->default_ea);
2311 hmap_init(&br->ports);
2312 hmap_init(&br->ifaces);
2313 hmap_init(&br->iface_by_name);
2314 hmap_init(&br->mirrors);
2316 hmap_init(&br->if_cfg_todo);
2317 list_init(&br->ofpp_garbage);
2319 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2323 bridge_destroy(struct bridge *br)
2326 struct mirror *mirror, *next_mirror;
2327 struct port *port, *next_port;
2328 struct if_cfg *if_cfg, *next_if_cfg;
2329 struct ofpp_garbage *garbage, *next_garbage;
2331 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2334 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2335 mirror_destroy(mirror);
2337 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2338 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2341 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2342 &br->ofpp_garbage) {
2343 list_remove(&garbage->list_node);
2347 hmap_remove(&all_bridges, &br->node);
2348 ofproto_destroy(br->ofproto);
2349 hmap_destroy(&br->ifaces);
2350 hmap_destroy(&br->ports);
2351 hmap_destroy(&br->iface_by_name);
2352 hmap_destroy(&br->mirrors);
2353 hmap_destroy(&br->if_cfg_todo);
2360 static struct bridge *
2361 bridge_lookup(const char *name)
2365 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2366 if (!strcmp(br->name, name)) {
2373 /* Handle requests for a listing of all flows known by the OpenFlow
2374 * stack, including those normally hidden. */
2376 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2377 const char *argv[], void *aux OVS_UNUSED)
2382 br = bridge_lookup(argv[1]);
2384 unixctl_command_reply_error(conn, "Unknown bridge");
2389 ofproto_get_all_flows(br->ofproto, &results);
2391 unixctl_command_reply(conn, ds_cstr(&results));
2392 ds_destroy(&results);
2395 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2396 * connections and reconnect. If BRIDGE is not specified, then all bridges
2397 * drop their controller connections and reconnect. */
2399 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2400 const char *argv[], void *aux OVS_UNUSED)
2404 br = bridge_lookup(argv[1]);
2406 unixctl_command_reply_error(conn, "Unknown bridge");
2409 ofproto_reconnect_controllers(br->ofproto);
2411 HMAP_FOR_EACH (br, node, &all_bridges) {
2412 ofproto_reconnect_controllers(br->ofproto);
2415 unixctl_command_reply(conn, NULL);
2419 bridge_get_controllers(const struct bridge *br,
2420 struct ovsrec_controller ***controllersp)
2422 struct ovsrec_controller **controllers;
2423 size_t n_controllers;
2425 controllers = br->cfg->controller;
2426 n_controllers = br->cfg->n_controller;
2428 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2434 *controllersp = controllers;
2436 return n_controllers;
2440 bridge_queue_if_cfg(struct bridge *br,
2441 const struct ovsrec_interface *cfg,
2442 const struct ovsrec_port *parent)
2444 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2447 if_cfg->parent = parent;
2448 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2449 hash_string(if_cfg->cfg->name, 0));
2452 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2453 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2454 * 'br' needs to complete its configuration. */
2456 bridge_add_del_ports(struct bridge *br,
2457 const unsigned long int *splinter_vlans)
2459 struct shash_node *port_node;
2460 struct port *port, *next;
2461 struct shash new_ports;
2464 assert(hmap_is_empty(&br->if_cfg_todo));
2466 /* Collect new ports. */
2467 shash_init(&new_ports);
2468 for (i = 0; i < br->cfg->n_ports; i++) {
2469 const char *name = br->cfg->ports[i]->name;
2470 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2471 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2475 if (bridge_get_controllers(br, NULL)
2476 && !shash_find(&new_ports, br->name)) {
2477 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2478 br->name, br->name);
2480 ovsrec_interface_init(&br->synth_local_iface);
2481 ovsrec_port_init(&br->synth_local_port);
2483 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2484 br->synth_local_port.n_interfaces = 1;
2485 br->synth_local_port.name = br->name;
2487 br->synth_local_iface.name = br->name;
2488 br->synth_local_iface.type = "internal";
2490 br->synth_local_ifacep = &br->synth_local_iface;
2492 shash_add(&new_ports, br->name, &br->synth_local_port);
2495 if (splinter_vlans) {
2496 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2499 /* Get rid of deleted ports.
2500 * Get rid of deleted interfaces on ports that still exist. */
2501 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2502 port->cfg = shash_find_data(&new_ports, port->name);
2506 port_del_ifaces(port);
2510 /* Update iface->cfg and iface->type in interfaces that still exist.
2511 * Add new interfaces to creation queue. */
2512 SHASH_FOR_EACH (port_node, &new_ports) {
2513 const struct ovsrec_port *port = port_node->data;
2516 for (i = 0; i < port->n_interfaces; i++) {
2517 const struct ovsrec_interface *cfg = port->interfaces[i];
2518 struct iface *iface = iface_lookup(br, cfg->name);
2519 const char *type = iface_get_type(cfg, br->cfg);
2524 } else if (strcmp(type, "null")) {
2525 bridge_queue_if_cfg(br, cfg, port);
2530 shash_destroy(&new_ports);
2533 /* Initializes 'oc' appropriately as a management service controller for
2536 * The caller must free oc->target when it is no longer needed. */
2538 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2539 struct ofproto_controller *oc)
2541 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2542 oc->max_backoff = 0;
2543 oc->probe_interval = 60;
2544 oc->band = OFPROTO_OUT_OF_BAND;
2546 oc->burst_limit = 0;
2547 oc->enable_async_msgs = true;
2550 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2552 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2553 struct ofproto_controller *oc)
2557 oc->target = c->target;
2558 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2559 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2560 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2561 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2562 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2563 oc->burst_limit = (c->controller_burst_limit
2564 ? *c->controller_burst_limit : 0);
2565 oc->enable_async_msgs = (!c->enable_async_messages
2566 || *c->enable_async_messages);
2567 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2568 if (dscp < 0 || dscp > 63) {
2569 dscp = DSCP_DEFAULT;
2574 /* Configures the IP stack for 'br''s local interface properly according to the
2575 * configuration in 'c'. */
2577 bridge_configure_local_iface_netdev(struct bridge *br,
2578 struct ovsrec_controller *c)
2580 struct netdev *netdev;
2581 struct in_addr mask, gateway;
2583 struct iface *local_iface;
2586 /* If there's no local interface or no IP address, give up. */
2587 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2588 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2592 /* Bring up the local interface. */
2593 netdev = local_iface->netdev;
2594 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2596 /* Configure the IP address and netmask. */
2597 if (!c->local_netmask
2598 || !inet_aton(c->local_netmask, &mask)
2600 mask.s_addr = guess_netmask(ip.s_addr);
2602 if (!netdev_set_in4(netdev, ip, mask)) {
2603 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2604 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2607 /* Configure the default gateway. */
2608 if (c->local_gateway
2609 && inet_aton(c->local_gateway, &gateway)
2610 && gateway.s_addr) {
2611 if (!netdev_add_router(netdev, gateway)) {
2612 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2613 br->name, IP_ARGS(&gateway.s_addr));
2618 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2619 * in either string are treated as equal to any number of slashes in the other,
2620 * e.g. "x///y" is equal to "x/y". */
2622 equal_pathnames(const char *a, const char *b)
2626 a += strspn(a, "/");
2627 b += strspn(b, "/");
2628 } else if (*a == '\0') {
2639 bridge_configure_remotes(struct bridge *br,
2640 const struct sockaddr_in *managers, size_t n_managers)
2642 bool disable_in_band;
2644 struct ovsrec_controller **controllers;
2645 size_t n_controllers;
2647 enum ofproto_fail_mode fail_mode;
2649 struct ofproto_controller *ocs;
2653 /* Check if we should disable in-band control on this bridge. */
2654 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2657 /* Set OpenFlow queue ID for in-band control. */
2658 ofproto_set_in_band_queue(br->ofproto,
2659 smap_get_int(&br->cfg->other_config,
2660 "in-band-queue", -1));
2662 if (disable_in_band) {
2663 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2665 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2668 n_controllers = bridge_get_controllers(br, &controllers);
2670 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2673 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2674 for (i = 0; i < n_controllers; i++) {
2675 struct ovsrec_controller *c = controllers[i];
2677 if (!strncmp(c->target, "punix:", 6)
2678 || !strncmp(c->target, "unix:", 5)) {
2679 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2682 whitelist = xasprintf("unix:%s/%s.controller",
2683 ovs_rundir(), br->name);
2684 if (!equal_pathnames(c->target, whitelist)) {
2685 /* Prevent remote ovsdb-server users from accessing arbitrary
2686 * Unix domain sockets and overwriting arbitrary local
2688 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2689 "controller \"%s\" due to possibility for remote "
2690 "exploit. Instead, specify whitelisted \"%s\" or "
2691 "connect to \"unix:%s/%s.mgmt\" (which is always "
2692 "available without special configuration).",
2693 br->name, c->target, whitelist,
2694 ovs_rundir(), br->name);
2702 bridge_configure_local_iface_netdev(br, c);
2703 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2704 if (disable_in_band) {
2705 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2710 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2711 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2714 /* Set the fail-mode. */
2715 fail_mode = !br->cfg->fail_mode
2716 || !strcmp(br->cfg->fail_mode, "standalone")
2717 ? OFPROTO_FAIL_STANDALONE
2718 : OFPROTO_FAIL_SECURE;
2719 ofproto_set_fail_mode(br->ofproto, fail_mode);
2721 /* Configure OpenFlow controller connection snooping. */
2722 if (!ofproto_has_snoops(br->ofproto)) {
2726 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2727 ovs_rundir(), br->name));
2728 ofproto_set_snoops(br->ofproto, &snoops);
2729 sset_destroy(&snoops);
2734 bridge_configure_tables(struct bridge *br)
2736 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2740 n_tables = ofproto_get_n_tables(br->ofproto);
2742 for (i = 0; i < n_tables; i++) {
2743 struct ofproto_table_settings s;
2746 s.max_flows = UINT_MAX;
2750 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2751 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2754 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2755 s.max_flows = *cfg->flow_limit;
2757 if (cfg->overflow_policy
2758 && !strcmp(cfg->overflow_policy, "evict")) {
2761 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2762 for (k = 0; k < cfg->n_groups; k++) {
2763 const char *string = cfg->groups[k];
2766 msg = mf_parse_subfield__(&s.groups[k], &string);
2768 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2769 "'groups' (%s)", br->name, i, msg);
2771 } else if (*string) {
2772 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2773 "element '%s' contains trailing garbage",
2774 br->name, i, cfg->groups[k]);
2782 ofproto_configure_table(br->ofproto, i, &s);
2786 for (; j < br->cfg->n_flow_tables; j++) {
2787 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2788 "%"PRId64" not supported by this datapath", br->name,
2789 br->cfg->key_flow_tables[j]);
2793 /* Port functions. */
2795 static struct port *
2796 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2800 port = xzalloc(sizeof *port);
2802 port->name = xstrdup(cfg->name);
2804 list_init(&port->ifaces);
2806 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2810 /* Deletes interfaces from 'port' that are no longer configured for it. */
2812 port_del_ifaces(struct port *port)
2814 struct iface *iface, *next;
2815 struct sset new_ifaces;
2818 /* Collect list of new interfaces. */
2819 sset_init(&new_ifaces);
2820 for (i = 0; i < port->cfg->n_interfaces; i++) {
2821 const char *name = port->cfg->interfaces[i]->name;
2822 const char *type = port->cfg->interfaces[i]->type;
2823 if (strcmp(type, "null")) {
2824 sset_add(&new_ifaces, name);
2828 /* Get rid of deleted interfaces. */
2829 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2830 if (!sset_contains(&new_ifaces, iface->name)) {
2831 iface_destroy(iface);
2835 sset_destroy(&new_ifaces);
2839 port_destroy(struct port *port)
2842 struct bridge *br = port->bridge;
2843 struct iface *iface, *next;
2846 ofproto_bundle_unregister(br->ofproto, port);
2849 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2850 iface_destroy(iface);
2853 hmap_remove(&br->ports, &port->hmap_node);
2859 static struct port *
2860 port_lookup(const struct bridge *br, const char *name)
2864 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2866 if (!strcmp(port->name, name)) {
2874 enable_lacp(struct port *port, bool *activep)
2876 if (!port->cfg->lacp) {
2877 /* XXX when LACP implementation has been sufficiently tested, enable by
2878 * default and make active on bonded ports. */
2880 } else if (!strcmp(port->cfg->lacp, "off")) {
2882 } else if (!strcmp(port->cfg->lacp, "active")) {
2885 } else if (!strcmp(port->cfg->lacp, "passive")) {
2889 VLOG_WARN("port %s: unknown LACP mode %s",
2890 port->name, port->cfg->lacp);
2895 static struct lacp_settings *
2896 port_configure_lacp(struct port *port, struct lacp_settings *s)
2898 const char *lacp_time, *system_id;
2901 if (!enable_lacp(port, &s->active)) {
2905 s->name = port->name;
2907 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2909 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2910 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2911 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2912 " address.", port->name, system_id);
2916 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2919 if (eth_addr_is_zero(s->id)) {
2920 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2924 /* Prefer bondable links if unspecified. */
2925 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2927 s->priority = (priority > 0 && priority <= UINT16_MAX
2929 : UINT16_MAX - !list_is_short(&port->ifaces));
2931 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2932 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2937 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2939 int priority, portid, key;
2941 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2942 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2944 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2946 if (portid <= 0 || portid > UINT16_MAX) {
2947 portid = iface->ofp_port;
2950 if (priority <= 0 || priority > UINT16_MAX) {
2951 priority = UINT16_MAX;
2954 if (key < 0 || key > UINT16_MAX) {
2958 s->name = iface->name;
2960 s->priority = priority;
2965 port_configure_bond(struct port *port, struct bond_settings *s,
2966 uint32_t *bond_stable_ids)
2968 const char *detect_s;
2969 struct iface *iface;
2970 int miimon_interval;
2973 s->name = port->name;
2975 if (port->cfg->bond_mode) {
2976 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2977 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2978 port->name, port->cfg->bond_mode,
2979 bond_mode_to_string(s->balance));
2982 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2984 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2985 * active-backup. At some point we should remove this warning. */
2986 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2987 " in previous versions, the default bond_mode was"
2988 " balance-slb", port->name,
2989 bond_mode_to_string(s->balance));
2991 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2992 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2993 "please use another bond type or disable flood_vlans",
2997 miimon_interval = smap_get_int(&port->cfg->other_config,
2998 "bond-miimon-interval", 0);
2999 if (miimon_interval <= 0) {
3000 miimon_interval = 200;
3003 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3004 if (!detect_s || !strcmp(detect_s, "carrier")) {
3005 miimon_interval = 0;
3006 } else if (strcmp(detect_s, "miimon")) {
3007 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3008 "defaulting to carrier", port->name, detect_s);
3009 miimon_interval = 0;
3012 s->up_delay = MAX(0, port->cfg->bond_updelay);
3013 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3014 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3015 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3016 "bond-rebalance-interval", 10000);
3017 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3018 s->rebalance_interval = 1000;
3021 s->fake_iface = port->cfg->bond_fake_iface;
3024 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3025 long long stable_id;
3027 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3029 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3030 stable_id = iface->ofp_port;
3032 bond_stable_ids[i++] = stable_id;
3034 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3038 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3039 * instead of obtaining it from the database. */
3041 port_is_synthetic(const struct port *port)
3043 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3046 /* Interface functions. */
3048 /* Returns the correct network device type for interface 'iface' in bridge
3051 iface_get_type(const struct ovsrec_interface *iface,
3052 const struct ovsrec_bridge *br)
3054 /* The local port always has type "internal". Other ports take their type
3055 * from the database and default to "system" if none is specified. */
3056 return (!strcmp(iface->name, br->name) ? "internal"
3057 : iface->type[0] ? iface->type
3062 iface_destroy(struct iface *iface)
3065 struct port *port = iface->port;
3066 struct bridge *br = port->bridge;
3068 if (br->ofproto && iface->ofp_port >= 0) {
3069 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3072 if (iface->ofp_port >= 0) {
3073 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3076 list_remove(&iface->port_elem);
3077 hmap_remove(&br->iface_by_name, &iface->name_node);
3079 netdev_close(iface->netdev);
3086 static struct iface *
3087 iface_lookup(const struct bridge *br, const char *name)
3089 struct iface *iface;
3091 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3092 &br->iface_by_name) {
3093 if (!strcmp(iface->name, name)) {
3101 static struct iface *
3102 iface_find(const char *name)
3104 const struct bridge *br;
3106 HMAP_FOR_EACH (br, node, &all_bridges) {
3107 struct iface *iface = iface_lookup(br, name);
3116 static struct if_cfg *
3117 if_cfg_lookup(const struct bridge *br, const char *name)
3119 struct if_cfg *if_cfg;
3121 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3123 if (!strcmp(if_cfg->cfg->name, name)) {
3131 static struct iface *
3132 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3134 struct iface *iface;
3136 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3137 hash_int(ofp_port, 0), &br->ifaces) {
3138 if (iface->ofp_port == ofp_port) {
3145 /* Set Ethernet address of 'iface', if one is specified in the configuration
3148 iface_set_mac(struct iface *iface)
3150 uint8_t ea[ETH_ADDR_LEN];
3152 if (!strcmp(iface->type, "internal")
3153 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3154 if (iface->ofp_port == OFPP_LOCAL) {
3155 VLOG_ERR("interface %s: ignoring mac in Interface record "
3156 "(use Bridge record to set local port's mac)",
3158 } else if (eth_addr_is_multicast(ea)) {
3159 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3162 int error = netdev_set_etheraddr(iface->netdev, ea);
3164 VLOG_ERR("interface %s: setting MAC failed (%s)",
3165 iface->name, strerror(error));
3171 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3173 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3175 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3176 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3180 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3181 * sets the "ofport" field to -1.
3183 * This is appropriate when 'if_cfg''s interface cannot be created or is
3184 * otherwise invalid. */
3186 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3188 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3189 iface_set_ofport(if_cfg, -1);
3190 ovsrec_interface_set_status(if_cfg, NULL);
3191 ovsrec_interface_set_admin_state(if_cfg, NULL);
3192 ovsrec_interface_set_duplex(if_cfg, NULL);
3193 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3194 ovsrec_interface_set_link_state(if_cfg, NULL);
3195 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3196 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3197 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3198 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3199 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3200 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3204 struct iface_delete_queues_cbdata {
3205 struct netdev *netdev;
3206 const struct ovsdb_datum *queues;
3210 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3212 union ovsdb_atom atom;
3214 atom.integer = target;
3215 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3219 iface_delete_queues(unsigned int queue_id,
3220 const struct smap *details OVS_UNUSED, void *cbdata_)
3222 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3224 if (!queue_ids_include(cbdata->queues, queue_id)) {
3225 netdev_delete_queue(cbdata->netdev, queue_id);
3230 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3232 struct ofpbuf queues_buf;
3234 ofpbuf_init(&queues_buf, 0);
3236 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3237 netdev_set_qos(iface->netdev, NULL, NULL);
3239 struct iface_delete_queues_cbdata cbdata;
3243 /* Configure top-level Qos for 'iface'. */
3244 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3246 /* Deconfigure queues that were deleted. */
3247 cbdata.netdev = iface->netdev;
3248 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3250 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3252 /* Configure queues for 'iface'. */
3254 for (i = 0; i < qos->n_queues; i++) {
3255 const struct ovsrec_queue *queue = qos->value_queues[i];
3256 unsigned int queue_id = qos->key_queues[i];
3258 if (queue_id == 0) {
3262 if (queue->n_dscp == 1) {
3263 struct ofproto_port_queue *port_queue;
3265 port_queue = ofpbuf_put_uninit(&queues_buf,
3266 sizeof *port_queue);
3267 port_queue->queue = queue_id;
3268 port_queue->dscp = queue->dscp[0];
3271 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3274 struct smap details;
3276 smap_init(&details);
3277 netdev_set_queue(iface->netdev, 0, &details);
3278 smap_destroy(&details);
3282 if (iface->ofp_port >= 0) {
3283 const struct ofproto_port_queue *port_queues = queues_buf.data;
3284 size_t n_queues = queues_buf.size / sizeof *port_queues;
3286 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3287 port_queues, n_queues);
3290 netdev_set_policing(iface->netdev,
3291 iface->cfg->ingress_policing_rate,
3292 iface->cfg->ingress_policing_burst);
3294 ofpbuf_uninit(&queues_buf);
3298 iface_configure_cfm(struct iface *iface)
3300 const struct ovsrec_interface *cfg = iface->cfg;
3301 const char *opstate_str;
3302 const char *cfm_ccm_vlan;
3303 struct cfm_settings s;
3305 if (!cfg->n_cfm_mpid) {
3306 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3310 s.mpid = *cfg->cfm_mpid;
3311 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3312 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3313 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3315 if (s.interval <= 0) {
3319 if (!cfm_ccm_vlan) {
3321 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3322 s.ccm_vlan = CFM_RANDOM_VLAN;
3324 s.ccm_vlan = atoi(cfm_ccm_vlan);
3325 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3330 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3333 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3334 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3336 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3339 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3340 * instead of obtaining it from the database. */
3342 iface_is_synthetic(const struct iface *iface)
3344 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3348 /* Port mirroring. */
3350 static struct mirror *
3351 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3355 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3356 if (uuid_equals(uuid, &m->uuid)) {
3364 bridge_configure_mirrors(struct bridge *br)
3366 const struct ovsdb_datum *mc;
3367 unsigned long *flood_vlans;
3368 struct mirror *m, *next;
3371 /* Get rid of deleted mirrors. */
3372 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3373 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3374 union ovsdb_atom atom;
3376 atom.uuid = m->uuid;
3377 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3382 /* Add new mirrors and reconfigure existing ones. */
3383 for (i = 0; i < br->cfg->n_mirrors; i++) {
3384 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3385 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3387 m = mirror_create(br, cfg);
3390 if (!mirror_configure(m)) {
3395 /* Update flooded vlans (for RSPAN). */
3396 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3397 br->cfg->n_flood_vlans);
3398 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3399 bitmap_free(flood_vlans);
3402 static struct mirror *
3403 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3407 m = xzalloc(sizeof *m);
3408 m->uuid = cfg->header_.uuid;
3409 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3411 m->name = xstrdup(cfg->name);
3417 mirror_destroy(struct mirror *m)
3420 struct bridge *br = m->bridge;
3423 ofproto_mirror_unregister(br->ofproto, m);
3426 hmap_remove(&br->mirrors, &m->hmap_node);
3433 mirror_collect_ports(struct mirror *m,
3434 struct ovsrec_port **in_ports, int n_in_ports,
3435 void ***out_portsp, size_t *n_out_portsp)
3437 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3438 size_t n_out_ports = 0;
3441 for (i = 0; i < n_in_ports; i++) {
3442 const char *name = in_ports[i]->name;
3443 struct port *port = port_lookup(m->bridge, name);
3445 out_ports[n_out_ports++] = port;
3447 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3448 "port %s", m->bridge->name, m->name, name);
3451 *out_portsp = out_ports;
3452 *n_out_portsp = n_out_ports;
3456 mirror_configure(struct mirror *m)
3458 const struct ovsrec_mirror *cfg = m->cfg;
3459 struct ofproto_mirror_settings s;
3462 if (strcmp(cfg->name, m->name)) {
3464 m->name = xstrdup(cfg->name);
3468 /* Get output port or VLAN. */
3469 if (cfg->output_port) {
3470 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3471 if (!s.out_bundle) {
3472 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3473 m->bridge->name, m->name);
3476 s.out_vlan = UINT16_MAX;
3478 if (cfg->output_vlan) {
3479 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3480 "output vlan; ignoring output vlan",
3481 m->bridge->name, m->name);
3483 } else if (cfg->output_vlan) {
3484 /* The database should prevent invalid VLAN values. */
3485 s.out_bundle = NULL;
3486 s.out_vlan = *cfg->output_vlan;
3488 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3489 m->bridge->name, m->name);
3493 /* Get port selection. */
3494 if (cfg->select_all) {
3495 size_t n_ports = hmap_count(&m->bridge->ports);
3496 void **ports = xmalloc(n_ports * sizeof *ports);
3501 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3511 /* Get ports, dropping ports that don't exist.
3512 * The IDL ensures that there are no duplicates. */
3513 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3514 &s.srcs, &s.n_srcs);
3515 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3516 &s.dsts, &s.n_dsts);
3519 /* Get VLAN selection. */
3520 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3523 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3526 if (s.srcs != s.dsts) {
3535 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3537 * This is deprecated. It is only for compatibility with broken device drivers
3538 * in old versions of Linux that do not properly support VLANs when VLAN
3539 * devices are not used. When broken device drivers are no longer in
3540 * widespread use, we will delete these interfaces. */
3542 static struct ovsrec_port **recs;
3543 static size_t n_recs, allocated_recs;
3545 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3546 * splinters are reconfigured. */
3548 register_rec(struct ovsrec_port *rec)
3550 if (n_recs >= allocated_recs) {
3551 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3553 recs[n_recs++] = rec;
3556 /* Frees all of the ports registered with register_reg(). */
3558 free_registered_recs(void)
3562 for (i = 0; i < n_recs; i++) {
3563 struct ovsrec_port *port = recs[i];
3566 for (j = 0; j < port->n_interfaces; j++) {
3567 struct ovsrec_interface *iface = port->interfaces[j];
3572 smap_destroy(&port->other_config);
3573 free(port->interfaces);
3581 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3584 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3586 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3590 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3593 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3594 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3595 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3598 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3599 * use, returns NULL.
3601 * Updates 'vlan_splinters_enabled_anywhere'. */
3602 static unsigned long int *
3603 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3605 unsigned long int *splinter_vlans;
3606 struct sset splinter_ifaces;
3607 const char *real_dev_name;
3608 struct shash *real_devs;
3609 struct shash_node *node;
3613 /* Free space allocated for synthesized ports and interfaces, since we're
3614 * in the process of reconstructing all of them. */
3615 free_registered_recs();
3617 splinter_vlans = bitmap_allocate(4096);
3618 sset_init(&splinter_ifaces);
3619 vlan_splinters_enabled_anywhere = false;
3620 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3621 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3624 for (j = 0; j < br_cfg->n_ports; j++) {
3625 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3628 for (k = 0; k < port_cfg->n_interfaces; k++) {
3629 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3631 if (vlan_splinters_is_enabled(iface_cfg)) {
3632 vlan_splinters_enabled_anywhere = true;
3633 sset_add(&splinter_ifaces, iface_cfg->name);
3634 vlan_bitmap_from_array__(port_cfg->trunks,
3640 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3641 bitmap_set1(splinter_vlans, *port_cfg->tag);
3646 if (!vlan_splinters_enabled_anywhere) {
3647 free(splinter_vlans);
3648 sset_destroy(&splinter_ifaces);
3652 HMAP_FOR_EACH (br, node, &all_bridges) {
3654 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3658 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3659 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3660 * device to be created for it. */
3661 bitmap_set0(splinter_vlans, 0);
3662 bitmap_set0(splinter_vlans, 4095);
3664 /* Delete all VLAN devices that we don't need. */
3666 real_devs = vlandev_get_real_devs();
3667 SHASH_FOR_EACH (node, real_devs) {
3668 const struct vlan_real_dev *real_dev = node->data;
3669 const struct vlan_dev *vlan_dev;
3670 bool real_dev_has_splinters;
3672 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3674 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3675 if (!real_dev_has_splinters
3676 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3677 struct netdev *netdev;
3679 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3680 if (!netdev_get_in4(netdev, NULL, NULL) ||
3681 !netdev_get_in6(netdev, NULL)) {
3682 vlandev_del(vlan_dev->name);
3684 /* It has an IP address configured, so we don't own
3685 * it. Don't delete it. */
3687 netdev_close(netdev);
3694 /* Add all VLAN devices that we need. */
3695 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3698 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3699 if (!vlandev_get_name(real_dev_name, vid)) {
3700 vlandev_add(real_dev_name, vid);
3707 sset_destroy(&splinter_ifaces);
3709 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3710 free(splinter_vlans);
3713 return splinter_vlans;
3716 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3719 configure_splinter_port(struct port *port)
3721 struct ofproto *ofproto = port->bridge->ofproto;
3722 uint16_t realdev_ofp_port;
3723 const char *realdev_name;
3724 struct iface *vlandev, *realdev;
3726 ofproto_bundle_unregister(port->bridge->ofproto, port);
3728 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3731 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3732 realdev = iface_lookup(port->bridge, realdev_name);
3733 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3735 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3739 static struct ovsrec_port *
3740 synthesize_splinter_port(const char *real_dev_name,
3741 const char *vlan_dev_name, int vid)
3743 struct ovsrec_interface *iface;
3744 struct ovsrec_port *port;
3746 iface = xmalloc(sizeof *iface);
3747 ovsrec_interface_init(iface);
3748 iface->name = xstrdup(vlan_dev_name);
3749 iface->type = "system";
3751 port = xmalloc(sizeof *port);
3752 ovsrec_port_init(port);
3753 port->interfaces = xmemdup(&iface, sizeof iface);
3754 port->n_interfaces = 1;
3755 port->name = xstrdup(vlan_dev_name);
3756 port->vlan_mode = "splinter";
3757 port->tag = xmalloc(sizeof *port->tag);
3760 smap_add(&port->other_config, "realdev", real_dev_name);
3766 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3767 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3768 * 1-bit in the 'splinter_vlans' bitmap. */
3770 add_vlan_splinter_ports(struct bridge *br,
3771 const unsigned long int *splinter_vlans,
3772 struct shash *ports)
3776 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3777 * we're modifying 'ports'. */
3778 for (i = 0; i < br->cfg->n_ports; i++) {
3779 const char *name = br->cfg->ports[i]->name;
3780 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3783 for (j = 0; j < port_cfg->n_interfaces; j++) {
3784 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3786 if (vlan_splinters_is_enabled(iface_cfg)) {
3787 const char *real_dev_name;
3790 real_dev_name = iface_cfg->name;
3791 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3792 const char *vlan_dev_name;
3794 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3796 && !shash_find(ports, vlan_dev_name)) {
3797 shash_add(ports, vlan_dev_name,
3798 synthesize_splinter_port(
3799 real_dev_name, vlan_dev_name, vid));
3808 mirror_refresh_stats(struct mirror *m)
3810 struct ofproto *ofproto = m->bridge->ofproto;
3811 uint64_t tx_packets, tx_bytes;
3814 size_t stat_cnt = 0;
3816 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3817 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3821 if (tx_packets != UINT64_MAX) {
3822 keys[stat_cnt] = "tx_packets";
3823 values[stat_cnt] = tx_packets;
3826 if (tx_bytes != UINT64_MAX) {
3827 keys[stat_cnt] = "tx_bytes";
3828 values[stat_cnt] = tx_bytes;
3832 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);