1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
44 #include "socket-util.h"
46 #include "stream-ssl.h"
48 #include "system-stats.h"
53 #include "lib/vswitch-idl.h"
54 #include "xenserver.h"
56 #include "sflow_api.h"
57 #include "vlan-bitmap.h"
59 VLOG_DEFINE_THIS_MODULE(bridge);
61 COVERAGE_DEFINE(bridge_reconfigure);
63 /* Configuration of an uninstantiated iface. */
65 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
66 const struct ovsrec_interface *cfg; /* Interface record. */
67 const struct ovsrec_port *parent; /* Parent port record. */
70 /* OpenFlow port slated for removal from ofproto. */
72 struct list list_node; /* Node in bridge's ofpp_garbage. */
73 uint16_t ofp_port; /* Port to be deleted. */
77 /* These members are always valid. */
78 struct list port_elem; /* Element in struct port's "ifaces" list. */
79 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
80 struct port *port; /* Containing port. */
81 char *name; /* Host network device name. */
83 /* These members are valid only after bridge_reconfigure() causes them to
85 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
86 int ofp_port; /* OpenFlow port number, -1 if unknown. */
87 struct netdev *netdev; /* Network device. */
88 const char *type; /* Usually same as cfg->type. */
89 const struct ovsrec_interface *cfg;
93 struct uuid uuid; /* UUID of this "mirror" record in database. */
94 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
95 struct bridge *bridge;
97 const struct ovsrec_mirror *cfg;
101 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
102 struct bridge *bridge;
105 const struct ovsrec_port *cfg;
107 /* An ordinary bridge port has 1 interface.
108 * A bridge port for bonding has at least 2 interfaces. */
109 struct list ifaces; /* List of "struct iface"s. */
113 struct hmap_node node; /* In 'all_bridges'. */
114 char *name; /* User-specified arbitrary name. */
115 char *type; /* Datapath type. */
116 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
117 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
118 const struct ovsrec_bridge *cfg;
120 /* OpenFlow switch processing. */
121 struct ofproto *ofproto; /* OpenFlow switch. */
124 struct hmap ports; /* "struct port"s indexed by name. */
125 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
126 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
128 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
129 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
130 Indexed on 'cfg->name'. */
132 /* Port mirroring. */
133 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
135 /* Synthetic local port if necessary. */
136 struct ovsrec_port synth_local_port;
137 struct ovsrec_interface synth_local_iface;
138 struct ovsrec_interface *synth_local_ifacep;
141 /* All bridges, indexed by name. */
142 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
144 /* OVSDB IDL used to obtain configuration. */
145 static struct ovsdb_idl *idl;
147 /* Most recently processed IDL sequence number. */
148 static unsigned int idl_seqno;
150 /* Each time this timer expires, the bridge fetches systems and interface
151 * statistics and pushes them into the database. */
152 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
153 static long long int stats_timer = LLONG_MIN;
155 /* Stores the time after which rate limited statistics may be written to the
156 * database. Only updated when changes to the database require rate limiting.
158 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
159 static long long int db_limiter = LLONG_MIN;
161 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
162 * expensive. This can cause bridge_reconfigure() to take a long time during
163 * which no other work can be done. To deal with this problem, we limit port
164 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
165 * call to bridge_reconfigure(). If there is more work to do after the limit
166 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
167 * This allows the rest of the code to catch up on important things like
168 * forwarding packets. */
169 #define OFP_PORT_ACTION_WINDOW 10
170 static bool reconfiguring = false;
172 static void add_del_bridges(const struct ovsrec_open_vswitch *);
173 static void bridge_update_ofprotos(void);
174 static void bridge_create(const struct ovsrec_bridge *);
175 static void bridge_destroy(struct bridge *);
176 static struct bridge *bridge_lookup(const char *name);
177 static unixctl_cb_func bridge_unixctl_dump_flows;
178 static unixctl_cb_func bridge_unixctl_reconnect;
179 static size_t bridge_get_controllers(const struct bridge *br,
180 struct ovsrec_controller ***controllersp);
181 static void bridge_add_del_ports(struct bridge *,
182 const unsigned long int *splinter_vlans);
183 static void bridge_refresh_ofp_port(struct bridge *);
184 static void bridge_configure_datapath_id(struct bridge *);
185 static void bridge_configure_flow_eviction_threshold(struct bridge *);
186 static void bridge_configure_netflow(struct bridge *);
187 static void bridge_configure_forward_bpdu(struct bridge *);
188 static void bridge_configure_mac_idle_time(struct bridge *);
189 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
190 static void bridge_configure_stp(struct bridge *);
191 static void bridge_configure_tables(struct bridge *);
192 static void bridge_configure_remotes(struct bridge *,
193 const struct sockaddr_in *managers,
195 static void bridge_pick_local_hw_addr(struct bridge *,
196 uint8_t ea[ETH_ADDR_LEN],
197 struct iface **hw_addr_iface);
198 static uint64_t bridge_pick_datapath_id(struct bridge *,
199 const uint8_t bridge_ea[ETH_ADDR_LEN],
200 struct iface *hw_addr_iface);
201 static void bridge_queue_if_cfg(struct bridge *,
202 const struct ovsrec_interface *,
203 const struct ovsrec_port *);
204 static uint64_t dpid_from_hash(const void *, size_t nbytes);
205 static bool bridge_has_bond_fake_iface(const struct bridge *,
207 static bool port_is_bond_fake_iface(const struct port *);
209 static unixctl_cb_func qos_unixctl_show;
211 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
212 static void port_del_ifaces(struct port *);
213 static void port_destroy(struct port *);
214 static struct port *port_lookup(const struct bridge *, const char *name);
215 static void port_configure(struct port *);
216 static struct lacp_settings *port_configure_lacp(struct port *,
217 struct lacp_settings *);
218 static void port_configure_bond(struct port *, struct bond_settings *,
219 uint32_t *bond_stable_ids);
220 static bool port_is_synthetic(const struct port *);
222 static void bridge_configure_mirrors(struct bridge *);
223 static struct mirror *mirror_create(struct bridge *,
224 const struct ovsrec_mirror *);
225 static void mirror_destroy(struct mirror *);
226 static bool mirror_configure(struct mirror *);
227 static void mirror_refresh_stats(struct mirror *);
229 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
230 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
231 static const char *iface_get_type(const struct ovsrec_interface *,
232 const struct ovsrec_bridge *);
233 static void iface_destroy(struct iface *);
234 static struct iface *iface_lookup(const struct bridge *, const char *name);
235 static struct iface *iface_find(const char *name);
236 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
237 static struct iface *iface_from_ofp_port(const struct bridge *,
239 static void iface_set_mac(struct iface *);
240 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
241 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
242 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
243 static void iface_configure_cfm(struct iface *);
244 static void iface_refresh_cfm_stats(struct iface *);
245 static void iface_refresh_stats(struct iface *);
246 static void iface_refresh_status(struct iface *);
247 static bool iface_is_synthetic(const struct iface *);
248 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
250 static void shash_to_ovs_idl_map(struct shash *,
251 char ***keys, char ***values, size_t *n);
253 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
255 * This is deprecated. It is only for compatibility with broken device drivers
256 * in old versions of Linux that do not properly support VLANs when VLAN
257 * devices are not used. When broken device drivers are no longer in
258 * widespread use, we will delete these interfaces. */
260 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
261 static bool vlan_splinters_enabled_anywhere;
263 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
264 static unsigned long int *collect_splinter_vlans(
265 const struct ovsrec_open_vswitch *);
266 static void configure_splinter_port(struct port *);
267 static void add_vlan_splinter_ports(struct bridge *,
268 const unsigned long int *splinter_vlans,
269 struct shash *ports);
271 /* Public functions. */
273 /* Initializes the bridge module, configuring it to obtain its configuration
274 * from an OVSDB server accessed over 'remote', which should be a string in a
275 * form acceptable to ovsdb_idl_create(). */
277 bridge_init(const char *remote)
279 /* Create connection to database. */
280 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
281 idl_seqno = ovsdb_idl_get_seqno(idl);
282 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
284 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
285 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
288 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
289 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
290 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
292 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
293 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
294 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
296 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
297 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
298 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
299 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
315 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
317 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
319 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
320 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
322 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
324 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
326 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
327 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
329 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
335 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
336 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
339 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
341 /* Register unixctl commands. */
342 unixctl_command_register("qos/show", "interface", 1, 1,
343 qos_unixctl_show, NULL);
344 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
345 bridge_unixctl_dump_flows, NULL);
346 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
347 bridge_unixctl_reconnect, NULL);
357 struct bridge *br, *next_br;
359 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
362 ovsdb_idl_destroy(idl);
365 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
366 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
367 * responsible for freeing '*managersp' (with free()).
369 * You may be asking yourself "why does ovs-vswitchd care?", because
370 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
371 * should not be and in fact is not directly involved in that. But
372 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
373 * it has to tell in-band control where the managers are to enable that.
374 * (Thus, only managers connected in-band are collected.)
377 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
378 struct sockaddr_in **managersp, size_t *n_managersp)
380 struct sockaddr_in *managers = NULL;
381 size_t n_managers = 0;
385 /* Collect all of the potential targets from the "targets" columns of the
386 * rows pointed to by "manager_options", excluding any that are
389 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
390 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
392 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
393 sset_find_and_delete(&targets, m->target);
395 sset_add(&targets, m->target);
399 /* Now extract the targets' IP addresses. */
400 if (!sset_is_empty(&targets)) {
403 managers = xmalloc(sset_count(&targets) * sizeof *managers);
404 SSET_FOR_EACH (target, &targets) {
405 struct sockaddr_in *sin = &managers[n_managers];
407 if (stream_parse_target_with_default_ports(target,
415 sset_destroy(&targets);
417 *managersp = managers;
418 *n_managersp = n_managers;
422 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
424 unsigned long int *splinter_vlans;
427 COVERAGE_INC(bridge_reconfigure);
429 assert(!reconfiguring);
430 reconfiguring = true;
432 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
433 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
434 * configuration otherwise.
436 * This is mostly an update to bridge data structures. Nothing is pushed
437 * down to ofproto or lower layers. */
438 add_del_bridges(ovs_cfg);
439 splinter_vlans = collect_splinter_vlans(ovs_cfg);
440 HMAP_FOR_EACH (br, node, &all_bridges) {
441 bridge_add_del_ports(br, splinter_vlans);
443 free(splinter_vlans);
445 /* Delete datapaths that are no longer configured, and create ones which
446 * don't exist but should. */
447 bridge_update_ofprotos();
449 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
450 HMAP_FOR_EACH (br, node, &all_bridges) {
451 bridge_refresh_ofp_port(br);
454 /* Clear database records for "if_cfg"s which haven't been instantiated. */
455 HMAP_FOR_EACH (br, node, &all_bridges) {
456 struct if_cfg *if_cfg;
458 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
459 iface_clear_db_record(if_cfg->cfg);
465 bridge_reconfigure_ofp(void)
467 long long int deadline;
471 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
473 /* The kernel will reject any attempt to add a given port to a datapath if
474 * that port already belongs to a different datapath, so we must do all
475 * port deletions before any port additions. */
476 HMAP_FOR_EACH (br, node, &all_bridges) {
477 struct ofpp_garbage *garbage, *next;
479 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
480 ofproto_port_del(br->ofproto, garbage->ofp_port);
481 list_remove(&garbage->list_node);
485 if (time_msec() >= deadline) {
491 HMAP_FOR_EACH (br, node, &all_bridges) {
492 struct if_cfg *if_cfg, *next;
494 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
495 iface_create(br, if_cfg, -1);
497 if (time_msec() >= deadline) {
507 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
509 struct sockaddr_in *managers;
510 int sflow_bridge_number;
515 assert(reconfiguring);
516 done = bridge_reconfigure_ofp();
518 /* Complete the configuration. */
519 sflow_bridge_number = 0;
520 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
521 HMAP_FOR_EACH (br, node, &all_bridges) {
524 /* We need the datapath ID early to allow LACP ports to use it as the
525 * default system ID. */
526 bridge_configure_datapath_id(br);
528 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
531 port_configure(port);
533 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
534 iface_configure_cfm(iface);
535 iface_configure_qos(iface, port->cfg->qos);
536 iface_set_mac(iface);
539 bridge_configure_mirrors(br);
540 bridge_configure_flow_eviction_threshold(br);
541 bridge_configure_forward_bpdu(br);
542 bridge_configure_mac_idle_time(br);
543 bridge_configure_remotes(br, managers, n_managers);
544 bridge_configure_netflow(br);
545 bridge_configure_sflow(br, &sflow_bridge_number);
546 bridge_configure_stp(br);
547 bridge_configure_tables(br);
552 /* ovs-vswitchd has completed initialization, so allow the process that
553 * forked us to exit successfully. */
554 daemonize_complete();
555 reconfiguring = false;
561 /* Delete ofprotos which aren't configured or have the wrong type. Create
562 * ofprotos which don't exist but need to. */
564 bridge_update_ofprotos(void)
566 struct bridge *br, *next;
571 /* Delete ofprotos with no bridge or with the wrong type. */
574 ofproto_enumerate_types(&types);
575 SSET_FOR_EACH (type, &types) {
578 ofproto_enumerate_names(type, &names);
579 SSET_FOR_EACH (name, &names) {
580 br = bridge_lookup(name);
581 if (!br || strcmp(type, br->type)) {
582 ofproto_delete(name, type);
586 sset_destroy(&names);
587 sset_destroy(&types);
589 /* Add ofprotos for bridges which don't have one yet. */
590 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
598 /* Remove ports from any datapath with the same name as 'br'. If we
599 * don't do this, creating 'br''s ofproto will fail because a port with
600 * the same name as its local port already exists. */
601 HMAP_FOR_EACH (br2, node, &all_bridges) {
602 struct ofproto_port ofproto_port;
608 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
610 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
612 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
615 ofproto_port_destroy(&ofproto_port);
619 error = ofproto_create(br->name, br->type, &br->ofproto);
621 VLOG_ERR("failed to create bridge %s: %s", br->name,
629 port_configure(struct port *port)
631 const struct ovsrec_port *cfg = port->cfg;
632 struct bond_settings bond_settings;
633 struct lacp_settings lacp_settings;
634 struct ofproto_bundle_settings s;
637 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
638 configure_splinter_port(port);
647 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
648 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
649 s.slaves[s.n_slaves++] = iface->ofp_port;
654 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
658 /* Get VLAN trunks. */
661 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
665 if (cfg->vlan_mode) {
666 if (!strcmp(cfg->vlan_mode, "access")) {
667 s.vlan_mode = PORT_VLAN_ACCESS;
668 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
669 s.vlan_mode = PORT_VLAN_TRUNK;
670 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
671 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
672 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
673 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
675 /* This "can't happen" because ovsdb-server should prevent it. */
676 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
677 s.vlan_mode = PORT_VLAN_TRUNK;
681 s.vlan_mode = PORT_VLAN_ACCESS;
683 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
687 s.vlan_mode = PORT_VLAN_TRUNK;
690 s.use_priority_tags = !strcmp("true", ovsrec_port_get_other_config_value(
691 cfg, "priority-tags", ""));
693 /* Get LACP settings. */
694 s.lacp = port_configure_lacp(port, &lacp_settings);
698 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
699 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
700 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
703 s.lacp_slaves = NULL;
706 /* Get bond settings. */
707 if (s.n_slaves > 1) {
708 s.bond = &bond_settings;
709 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
710 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
713 s.bond_stable_ids = NULL;
715 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
716 netdev_set_miimon_interval(iface->netdev, 0);
721 ofproto_bundle_register(port->bridge->ofproto, port, &s);
727 free(s.bond_stable_ids);
730 /* Pick local port hardware address and datapath ID for 'br'. */
732 bridge_configure_datapath_id(struct bridge *br)
734 uint8_t ea[ETH_ADDR_LEN];
736 struct iface *local_iface;
737 struct iface *hw_addr_iface;
740 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
741 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
743 int error = netdev_set_etheraddr(local_iface->netdev, ea);
745 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
746 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
747 "Ethernet address: %s",
748 br->name, strerror(error));
751 memcpy(br->ea, ea, ETH_ADDR_LEN);
753 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
754 ofproto_set_datapath_id(br->ofproto, dpid);
756 dpid_string = xasprintf("%016"PRIx64, dpid);
757 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
761 /* Set NetFlow configuration on 'br'. */
763 bridge_configure_netflow(struct bridge *br)
765 struct ovsrec_netflow *cfg = br->cfg->netflow;
766 struct netflow_options opts;
769 ofproto_set_netflow(br->ofproto, NULL);
773 memset(&opts, 0, sizeof opts);
775 /* Get default NetFlow configuration from datapath.
776 * Apply overrides from 'cfg'. */
777 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
778 if (cfg->engine_type) {
779 opts.engine_type = *cfg->engine_type;
781 if (cfg->engine_id) {
782 opts.engine_id = *cfg->engine_id;
785 /* Configure active timeout interval. */
786 opts.active_timeout = cfg->active_timeout;
787 if (!opts.active_timeout) {
788 opts.active_timeout = -1;
789 } else if (opts.active_timeout < 0) {
790 VLOG_WARN("bridge %s: active timeout interval set to negative "
791 "value, using default instead (%d seconds)", br->name,
792 NF_ACTIVE_TIMEOUT_DEFAULT);
793 opts.active_timeout = -1;
796 /* Add engine ID to interface number to disambiguate bridgs? */
797 opts.add_id_to_iface = cfg->add_id_to_interface;
798 if (opts.add_id_to_iface) {
799 if (opts.engine_id > 0x7f) {
800 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
801 "another vswitch, choose an engine id less than 128",
804 if (hmap_count(&br->ports) > 508) {
805 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
806 "another port when more than 508 ports are used",
812 sset_init(&opts.collectors);
813 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
816 if (ofproto_set_netflow(br->ofproto, &opts)) {
817 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
819 sset_destroy(&opts.collectors);
822 /* Set sFlow configuration on 'br'. */
824 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
826 const struct ovsrec_sflow *cfg = br->cfg->sflow;
827 struct ovsrec_controller **controllers;
828 struct ofproto_sflow_options oso;
829 size_t n_controllers;
833 ofproto_set_sflow(br->ofproto, NULL);
837 memset(&oso, 0, sizeof oso);
839 sset_init(&oso.targets);
840 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
842 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
844 oso.sampling_rate = *cfg->sampling;
847 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
849 oso.polling_interval = *cfg->polling;
852 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
854 oso.header_len = *cfg->header;
857 oso.sub_id = (*sflow_bridge_number)++;
858 oso.agent_device = cfg->agent;
860 oso.control_ip = NULL;
861 n_controllers = bridge_get_controllers(br, &controllers);
862 for (i = 0; i < n_controllers; i++) {
863 if (controllers[i]->local_ip) {
864 oso.control_ip = controllers[i]->local_ip;
868 ofproto_set_sflow(br->ofproto, &oso);
870 sset_destroy(&oso.targets);
874 port_configure_stp(const struct ofproto *ofproto, struct port *port,
875 struct ofproto_port_stp_settings *port_s,
876 int *port_num_counter, unsigned long *port_num_bitmap)
878 const char *config_str;
881 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-enable",
883 if (config_str && !strcmp(config_str, "false")) {
884 port_s->enable = false;
887 port_s->enable = true;
890 /* STP over bonds is not supported. */
891 if (!list_is_singleton(&port->ifaces)) {
892 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
894 port_s->enable = false;
898 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
900 /* Internal ports shouldn't participate in spanning tree, so
902 if (!strcmp(iface->type, "internal")) {
903 VLOG_DBG("port %s: disable STP on internal ports", port->name);
904 port_s->enable = false;
908 /* STP on mirror output ports is not supported. */
909 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
910 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
911 port_s->enable = false;
915 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-port-num",
918 unsigned long int port_num = strtoul(config_str, NULL, 0);
919 int port_idx = port_num - 1;
921 if (port_num < 1 || port_num > STP_MAX_PORTS) {
922 VLOG_ERR("port %s: invalid stp-port-num", port->name);
923 port_s->enable = false;
927 if (bitmap_is_set(port_num_bitmap, port_idx)) {
928 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
929 port->name, port_num);
930 port_s->enable = false;
933 bitmap_set1(port_num_bitmap, port_idx);
934 port_s->port_num = port_idx;
936 if (*port_num_counter > STP_MAX_PORTS) {
937 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
938 port_s->enable = false;
942 port_s->port_num = (*port_num_counter)++;
945 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-path-cost",
948 port_s->path_cost = strtoul(config_str, NULL, 10);
950 enum netdev_features current;
952 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
953 /* Couldn't get speed, so assume 100Mb/s. */
954 port_s->path_cost = 19;
958 mbps = netdev_features_to_bps(current) / 1000000;
959 port_s->path_cost = stp_convert_speed_to_cost(mbps);
963 config_str = ovsrec_port_get_other_config_value(port->cfg,
967 port_s->priority = strtoul(config_str, NULL, 0);
969 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
973 /* Set spanning tree configuration on 'br'. */
975 bridge_configure_stp(struct bridge *br)
977 if (!br->cfg->stp_enable) {
978 ofproto_set_stp(br->ofproto, NULL);
980 struct ofproto_stp_settings br_s;
981 const char *config_str;
983 int port_num_counter;
984 unsigned long *port_num_bitmap;
986 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
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 = ovsrec_bridge_get_other_config_value(br->cfg,
1007 br_s.priority = strtoul(config_str, NULL, 0);
1009 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1012 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
1016 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1018 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1021 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
1025 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1027 br_s.max_age = STP_DEFAULT_MAX_AGE;
1030 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
1031 "stp-forward-delay",
1034 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1036 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1039 /* Configure STP on the bridge. */
1040 if (ofproto_set_stp(br->ofproto, &br_s)) {
1041 VLOG_ERR("bridge %s: could not enable STP", br->name);
1045 /* Users must either set the port number with the "stp-port-num"
1046 * configuration on all ports or none. If manual configuration
1047 * is not done, then we allocate them sequentially. */
1048 port_num_counter = 0;
1049 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1050 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1051 struct ofproto_port_stp_settings port_s;
1052 struct iface *iface;
1054 port_configure_stp(br->ofproto, port, &port_s,
1055 &port_num_counter, port_num_bitmap);
1057 /* As bonds are not supported, just apply configuration to
1058 * all interfaces. */
1059 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1060 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1062 VLOG_ERR("port %s: could not enable STP", port->name);
1068 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1069 && port_num_counter) {
1070 VLOG_ERR("bridge %s: must manually configure all STP port "
1071 "IDs or none, disabling", br->name);
1072 ofproto_set_stp(br->ofproto, NULL);
1074 bitmap_free(port_num_bitmap);
1079 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1081 const struct port *port = port_lookup(br, name);
1082 return port && port_is_bond_fake_iface(port);
1086 port_is_bond_fake_iface(const struct port *port)
1088 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1092 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1094 struct bridge *br, *next;
1095 struct shash new_br;
1098 /* Collect new bridges' names and types. */
1099 shash_init(&new_br);
1100 for (i = 0; i < cfg->n_bridges; i++) {
1101 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1102 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1104 if (strchr(br_cfg->name, '/')) {
1105 /* Prevent remote ovsdb-server users from accessing arbitrary
1106 * directories, e.g. consider a bridge named "../../../etc/". */
1107 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1109 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1110 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1114 /* Get rid of deleted bridges or those whose types have changed.
1115 * Update 'cfg' of bridges that still exist. */
1116 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1117 br->cfg = shash_find_data(&new_br, br->name);
1118 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1119 br->cfg->datapath_type))) {
1124 /* Add new bridges. */
1125 for (i = 0; i < cfg->n_bridges; i++) {
1126 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1127 struct bridge *br = bridge_lookup(br_cfg->name);
1129 bridge_create(br_cfg);
1133 shash_destroy(&new_br);
1137 iface_set_ofp_port(struct iface *iface, int ofp_port)
1139 struct bridge *br = iface->port->bridge;
1141 assert(iface->ofp_port < 0 && ofp_port >= 0);
1142 iface->ofp_port = ofp_port;
1143 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1144 iface_set_ofport(iface->cfg, ofp_port);
1147 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1148 * Returns 0 if successful, otherwise a positive errno value. */
1150 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1151 struct netdev *netdev)
1157 shash_from_ovs_idl_map(iface_cfg->key_options,
1158 iface_cfg->value_options,
1159 iface_cfg->n_options, &args);
1160 error = netdev_set_config(netdev, &args);
1161 shash_destroy(&args);
1164 VLOG_WARN("could not configure network device %s (%s)",
1165 iface_cfg->name, strerror(error));
1171 bridge_ofproto_port_del(struct bridge *br, struct ofproto_port ofproto_port)
1173 int error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1175 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1176 br->name, ofproto_port.name, strerror(error));
1178 VLOG_INFO("bridge %s: removed interface %s (%d)", br->name,
1179 ofproto_port.name, ofproto_port.ofp_port);
1183 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1184 * consistent with the ofp_ports in "br->ofproto". */
1186 bridge_refresh_ofp_port(struct bridge *br)
1188 struct ofproto_port_dump dump;
1189 struct ofproto_port ofproto_port;
1190 struct port *port, *port_next;
1192 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1193 hmap_clear(&br->ifaces);
1194 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1195 struct iface *iface;
1197 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1198 iface->ofp_port = -1;
1202 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1203 * already exist in the datapath and promote them to full fledged "struct
1204 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1205 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1206 struct iface *iface = iface_lookup(br, ofproto_port.name);
1208 if (iface->ofp_port >= 0) {
1209 VLOG_WARN("bridge %s: interface %s reported twice",
1210 br->name, ofproto_port.name);
1211 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1212 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1213 br->name, ofproto_port.ofp_port);
1214 } else if (!strcmp(ofproto_port.type, iface->type)) {
1215 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1217 /* Port has incorrect type so delete it later. */
1220 struct if_cfg *if_cfg = if_cfg_lookup(br, ofproto_port.name);
1223 iface_create(br, if_cfg, ofproto_port.ofp_port);
1224 } else if (bridge_has_bond_fake_iface(br, ofproto_port.name)
1225 && strcmp(ofproto_port.type, "internal")) {
1226 /* Bond fake iface with the wrong type. */
1227 bridge_ofproto_port_del(br, ofproto_port);
1229 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1230 garbage->ofp_port = ofproto_port.ofp_port;
1231 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1236 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1237 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1238 * later they can be added to ofproto. */
1239 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1240 struct iface *iface, *iface_next;
1242 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1243 if (iface->ofp_port < 0) {
1244 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1245 iface_destroy(iface);
1249 if (list_is_empty(&port->ifaces)) {
1255 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1256 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1257 * automatically allocated for the iface. Takes ownership of and
1258 * deallocates 'if_cfg'.
1260 * Return true if an iface is successfully created, false otherwise. */
1262 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1264 struct iface *iface;
1269 assert(!iface_lookup(br, if_cfg->cfg->name));
1271 port = port_lookup(br, if_cfg->parent->name);
1273 port = port_create(br, if_cfg->parent);
1276 iface = xzalloc(sizeof *iface);
1278 iface->name = xstrdup(if_cfg->cfg->name);
1279 iface->ofp_port = -1;
1280 iface->netdev = NULL;
1281 iface->cfg = if_cfg->cfg;
1282 hmap_insert(&br->iface_by_name, &iface->name_node,
1283 hash_string(iface->name, 0));
1284 list_push_back(&port->ifaces, &iface->port_elem);
1285 iface->type = iface_get_type(iface->cfg, br->cfg);
1286 if (ofp_port >= 0) {
1287 iface_set_ofp_port(iface, ofp_port);
1290 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1294 error = netdev_open(iface->name, iface->type, &iface->netdev);
1296 VLOG_WARN("could not open network device %s (%s)", iface->name,
1301 && port->cfg->vlan_mode
1302 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1303 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1306 /* Configure the netdev. */
1307 if (iface->netdev) {
1308 int error = iface_set_netdev_config(iface->cfg, iface->netdev);
1310 netdev_close(iface->netdev);
1311 iface->netdev = NULL;
1315 /* Add the port, if necessary. */
1316 if (iface->netdev && iface->ofp_port < 0) {
1317 uint16_t new_ofp_port;
1320 error = ofproto_port_add(br->ofproto, iface->netdev, &new_ofp_port);
1322 VLOG_INFO("bridge %s: added interface %s (%d)", br->name,
1323 iface->name, new_ofp_port);
1324 iface_set_ofp_port(iface, new_ofp_port);
1326 netdev_close(iface->netdev);
1327 iface->netdev = NULL;
1331 /* Initially populate stats columns. */
1332 if (iface->netdev) {
1333 iface_refresh_stats(iface);
1334 iface_refresh_status(iface);
1337 /* Delete the iface if we failed. */
1338 ok = iface->netdev && iface->ofp_port >= 0;
1340 VLOG_DBG("bridge %s: interface %s is on port %d",
1341 br->name, iface->name, iface->ofp_port);
1343 struct ofproto_port ofproto_port;
1345 if (iface->netdev) {
1346 VLOG_ERR("bridge %s: missing %s interface, dropping",
1347 br->name, iface->name);
1349 /* We already reported a related error, don't bother
1350 * duplicating it. */
1352 if (!ofproto_port_query_by_name(br->ofproto, port->name,
1354 VLOG_INFO("bridge %s: removed interface %s (%d)",
1355 br->name, port->name, ofproto_port.ofp_port);
1356 bridge_ofproto_port_del(br, ofproto_port);
1357 ofproto_port_destroy(&ofproto_port);
1359 iface_clear_db_record(iface->cfg);
1360 iface_destroy(iface);
1364 if (list_is_empty(&port->ifaces)) {
1370 /* Add bond fake iface if necessary. */
1371 if (port_is_bond_fake_iface(port)) {
1372 struct ofproto_port ofproto_port;
1374 if (ofproto_port_query_by_name(br->ofproto, port->name,
1376 struct netdev *netdev;
1379 error = netdev_open(port->name, "internal", &netdev);
1381 ofproto_port_add(br->ofproto, netdev, NULL);
1382 netdev_close(netdev);
1384 VLOG_WARN("could not open network device %s (%s)",
1385 port->name, strerror(error));
1388 /* Already exists, nothing to do. */
1389 ofproto_port_destroy(&ofproto_port);
1396 /* Set Flow eviction threshold */
1398 bridge_configure_flow_eviction_threshold(struct bridge *br)
1400 const char *threshold_str;
1404 ovsrec_bridge_get_other_config_value(br->cfg,
1405 "flow-eviction-threshold",
1407 if (threshold_str) {
1408 threshold = strtoul(threshold_str, NULL, 10);
1410 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1412 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1415 /* Set forward BPDU option. */
1417 bridge_configure_forward_bpdu(struct bridge *br)
1419 const char *forward_bpdu_str;
1420 bool forward_bpdu = false;
1422 forward_bpdu_str = ovsrec_bridge_get_other_config_value(br->cfg,
1425 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1426 forward_bpdu = true;
1428 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1431 /* Set MAC aging time for 'br'. */
1433 bridge_configure_mac_idle_time(struct bridge *br)
1435 const char *idle_time_str;
1438 idle_time_str = ovsrec_bridge_get_other_config_value(br->cfg,
1441 idle_time = (idle_time_str && atoi(idle_time_str)
1442 ? atoi(idle_time_str)
1443 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1444 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1448 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1449 struct iface **hw_addr_iface)
1451 struct hmapx mirror_output_ports;
1454 bool found_addr = false;
1458 *hw_addr_iface = NULL;
1460 /* Did the user request a particular MAC? */
1461 hwaddr = ovsrec_bridge_get_other_config_value(br->cfg, "hwaddr", NULL);
1462 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1463 if (eth_addr_is_multicast(ea)) {
1464 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1465 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1466 } else if (eth_addr_is_zero(ea)) {
1467 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1473 /* Mirror output ports don't participate in picking the local hardware
1474 * address. ofproto can't help us find out whether a given port is a
1475 * mirror output because we haven't configured mirrors yet, so we need to
1476 * accumulate them ourselves. */
1477 hmapx_init(&mirror_output_ports);
1478 for (i = 0; i < br->cfg->n_mirrors; i++) {
1479 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1480 if (m->output_port) {
1481 hmapx_add(&mirror_output_ports, m->output_port);
1485 /* Otherwise choose the minimum non-local MAC address among all of the
1487 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1488 uint8_t iface_ea[ETH_ADDR_LEN];
1489 struct iface *candidate;
1490 struct iface *iface;
1492 /* Mirror output ports don't participate. */
1493 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1497 /* Choose the MAC address to represent the port. */
1499 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1500 /* Find the interface with this Ethernet address (if any) so that
1501 * we can provide the correct devname to the caller. */
1502 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1503 uint8_t candidate_ea[ETH_ADDR_LEN];
1504 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1505 && eth_addr_equals(iface_ea, candidate_ea)) {
1510 /* Choose the interface whose MAC address will represent the port.
1511 * The Linux kernel bonding code always chooses the MAC address of
1512 * the first slave added to a bond, and the Fedora networking
1513 * scripts always add slaves to a bond in alphabetical order, so
1514 * for compatibility we choose the interface with the name that is
1515 * first in alphabetical order. */
1516 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1517 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1522 /* The local port doesn't count (since we're trying to choose its
1523 * MAC address anyway). */
1524 if (iface->ofp_port == OFPP_LOCAL) {
1529 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1535 /* Compare against our current choice. */
1536 if (!eth_addr_is_multicast(iface_ea) &&
1537 !eth_addr_is_local(iface_ea) &&
1538 !eth_addr_is_reserved(iface_ea) &&
1539 !eth_addr_is_zero(iface_ea) &&
1540 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1542 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1543 *hw_addr_iface = iface;
1548 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1549 br->name, ETH_ADDR_ARGS(ea));
1551 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1552 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1553 *hw_addr_iface = NULL;
1554 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1555 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1558 hmapx_destroy(&mirror_output_ports);
1561 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1562 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1563 * an interface on 'br', then that interface must be passed in as
1564 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1565 * 'hw_addr_iface' must be passed in as a null pointer. */
1567 bridge_pick_datapath_id(struct bridge *br,
1568 const uint8_t bridge_ea[ETH_ADDR_LEN],
1569 struct iface *hw_addr_iface)
1572 * The procedure for choosing a bridge MAC address will, in the most
1573 * ordinary case, also choose a unique MAC that we can use as a datapath
1574 * ID. In some special cases, though, multiple bridges will end up with
1575 * the same MAC address. This is OK for the bridges, but it will confuse
1576 * the OpenFlow controller, because each datapath needs a unique datapath
1579 * Datapath IDs must be unique. It is also very desirable that they be
1580 * stable from one run to the next, so that policy set on a datapath
1583 const char *datapath_id;
1586 datapath_id = ovsrec_bridge_get_other_config_value(br->cfg, "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, &sh)) {
1650 char **keys, **values;
1652 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1653 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1658 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1661 shash_destroy_free_data(&sh);
1663 error = netdev_get_flags(iface->netdev, &flags);
1665 ovsrec_interface_set_admin_state(iface->cfg,
1666 flags & NETDEV_UP ? "up" : "down");
1669 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1672 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1674 ovsrec_interface_set_duplex(iface->cfg,
1675 netdev_features_is_full_duplex(current)
1677 /* warning: uint64_t -> int64_t conversion */
1678 bps = netdev_features_to_bps(current);
1679 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1682 ovsrec_interface_set_duplex(iface->cfg, NULL);
1683 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1686 error = netdev_get_mtu(iface->netdev, &mtu);
1689 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1692 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1696 /* Writes 'iface''s CFM statistics to the database. */
1698 iface_refresh_cfm_stats(struct iface *iface)
1700 const struct ovsrec_interface *cfg = iface->cfg;
1702 const uint64_t *rmps;
1706 if (iface_is_synthetic(iface)) {
1710 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1713 const char *reasons[CFM_FAULT_N_REASONS];
1714 bool fault_bool = fault;
1718 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1719 int reason = 1 << i;
1720 if (fault & reason) {
1721 reasons[j++] = cfm_fault_reason_to_str(reason);
1725 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1726 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1728 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1729 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1732 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1733 iface->ofp_port, &rmps, &n_rmps);
1735 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1738 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1741 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1744 int64_t cfm_health = health;
1745 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1747 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1752 iface_refresh_stats(struct iface *iface)
1754 #define IFACE_STATS \
1755 IFACE_STAT(rx_packets, "rx_packets") \
1756 IFACE_STAT(tx_packets, "tx_packets") \
1757 IFACE_STAT(rx_bytes, "rx_bytes") \
1758 IFACE_STAT(tx_bytes, "tx_bytes") \
1759 IFACE_STAT(rx_dropped, "rx_dropped") \
1760 IFACE_STAT(tx_dropped, "tx_dropped") \
1761 IFACE_STAT(rx_errors, "rx_errors") \
1762 IFACE_STAT(tx_errors, "tx_errors") \
1763 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1764 IFACE_STAT(rx_over_errors, "rx_over_err") \
1765 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1766 IFACE_STAT(collisions, "collisions")
1768 #define IFACE_STAT(MEMBER, NAME) NAME,
1769 static char *keys[] = { IFACE_STATS };
1771 int64_t values[ARRAY_SIZE(keys)];
1774 struct netdev_stats stats;
1776 if (iface_is_synthetic(iface)) {
1780 /* Intentionally ignore return value, since errors will set 'stats' to
1781 * all-1s, and we will deal with that correctly below. */
1782 netdev_get_stats(iface->netdev, &stats);
1784 /* Copy statistics into values[] array. */
1786 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1789 assert(i == ARRAY_SIZE(keys));
1791 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1797 br_refresh_stp_status(struct bridge *br)
1799 struct ofproto *ofproto = br->ofproto;
1800 struct ofproto_stp_status status;
1801 char *keys[3], *values[3];
1804 if (ofproto_get_stp_status(ofproto, &status)) {
1808 if (!status.enabled) {
1809 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1813 keys[0] = "stp_bridge_id",
1814 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1815 keys[1] = "stp_designated_root",
1816 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1817 keys[2] = "stp_root_path_cost",
1818 values[2] = xasprintf("%d", status.root_path_cost);
1820 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1822 for (i = 0; i < ARRAY_SIZE(values); i++) {
1828 port_refresh_stp_status(struct port *port)
1830 struct ofproto *ofproto = port->bridge->ofproto;
1831 struct iface *iface;
1832 struct ofproto_port_stp_status status;
1834 char *str_values[4];
1835 int64_t int_values[3];
1838 if (port_is_synthetic(port)) {
1842 /* STP doesn't currently support bonds. */
1843 if (!list_is_singleton(&port->ifaces)) {
1844 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1848 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1850 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1854 if (!status.enabled) {
1855 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1856 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1860 /* Set Status column. */
1861 keys[0] = "stp_port_id";
1862 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1863 keys[1] = "stp_state";
1864 str_values[1] = xstrdup(stp_state_name(status.state));
1865 keys[2] = "stp_sec_in_state";
1866 str_values[2] = xasprintf("%u", status.sec_in_state);
1867 keys[3] = "stp_role";
1868 str_values[3] = xstrdup(stp_role_name(status.role));
1870 ovsrec_port_set_status(port->cfg, keys, str_values,
1871 ARRAY_SIZE(str_values));
1873 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1874 free(str_values[i]);
1877 /* Set Statistics column. */
1878 keys[0] = "stp_tx_count";
1879 int_values[0] = status.tx_count;
1880 keys[1] = "stp_rx_count";
1881 int_values[1] = status.rx_count;
1882 keys[2] = "stp_error_count";
1883 int_values[2] = status.error_count;
1885 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1886 ARRAY_SIZE(int_values));
1890 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1894 /* Use other-config:enable-system-stats by preference. */
1895 enable = ovsrec_open_vswitch_get_other_config_value(cfg,
1896 "enable-statistics",
1899 return !strcmp(enable, "true");
1902 /* Disable by default. */
1907 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1909 struct ovsdb_datum datum;
1913 if (enable_system_stats(cfg)) {
1914 get_system_stats(&stats);
1917 ovsdb_datum_from_shash(&datum, &stats);
1918 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1922 static inline const char *
1923 nx_role_to_str(enum nx_role role)
1928 case NX_ROLE_MASTER:
1933 return "*** INVALID ROLE ***";
1938 refresh_controller_status(void)
1942 const struct ovsrec_controller *cfg;
1946 /* Accumulate status for controllers on all bridges. */
1947 HMAP_FOR_EACH (br, node, &all_bridges) {
1948 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1951 /* Update each controller in the database with current status. */
1952 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1953 struct ofproto_controller_info *cinfo =
1954 shash_find_data(&info, cfg->target);
1957 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1958 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1959 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1960 (char **) cinfo->pairs.values,
1963 ovsrec_controller_set_is_connected(cfg, false);
1964 ovsrec_controller_set_role(cfg, NULL);
1965 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1969 ofproto_free_ofproto_controller_info(&info);
1973 refresh_cfm_stats(void)
1975 static struct ovsdb_idl_txn *txn = NULL;
1980 txn = ovsdb_idl_txn_create(idl);
1982 HMAP_FOR_EACH (br, node, &all_bridges) {
1983 struct iface *iface;
1985 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1986 iface_refresh_cfm_stats(iface);
1991 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1992 ovsdb_idl_txn_destroy(txn);
1997 /* Performs periodic activity required by bridges that needs to be done with
1998 * the least possible latency.
2000 * It makes sense to call this function a couple of times per poll loop, to
2001 * provide a significant performance boost on some benchmarks with ofprotos
2002 * that use the ofproto-dpif implementation. */
2004 bridge_run_fast(void)
2008 HMAP_FOR_EACH (br, node, &all_bridges) {
2009 ofproto_run_fast(br->ofproto);
2016 static const struct ovsrec_open_vswitch null_cfg;
2017 const struct ovsrec_open_vswitch *cfg;
2018 struct ovsdb_idl_txn *reconf_txn = NULL;
2020 bool vlan_splinters_changed;
2023 /* (Re)configure if necessary. */
2024 if (!reconfiguring) {
2027 if (ovsdb_idl_is_lock_contended(idl)) {
2028 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2029 struct bridge *br, *next_br;
2031 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2032 "disabling this process until it goes away");
2034 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2038 } else if (!ovsdb_idl_has_lock(idl)) {
2042 cfg = ovsrec_open_vswitch_first(idl);
2044 /* Let each bridge do the work that it needs to do. */
2045 HMAP_FOR_EACH (br, node, &all_bridges) {
2046 ofproto_run(br->ofproto);
2049 /* Re-configure SSL. We do this on every trip through the main loop,
2050 * instead of just when the database changes, because the contents of the
2051 * key and certificate files can change without the database changing.
2053 * We do this before bridge_reconfigure() because that function might
2054 * initiate SSL connections and thus requires SSL to be configured. */
2055 if (cfg && cfg->ssl) {
2056 const struct ovsrec_ssl *ssl = cfg->ssl;
2058 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2059 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2062 if (!reconfiguring) {
2063 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2064 * usage has changed. */
2065 vlan_splinters_changed = false;
2066 if (vlan_splinters_enabled_anywhere) {
2067 HMAP_FOR_EACH (br, node, &all_bridges) {
2068 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2069 vlan_splinters_changed = true;
2075 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2076 idl_seqno = ovsdb_idl_get_seqno(idl);
2078 reconf_txn = ovsdb_idl_txn_create(idl);
2079 bridge_reconfigure(cfg);
2081 /* We still need to reconfigure to avoid dangling pointers to
2082 * now-destroyed ovsrec structures inside bridge data. */
2083 bridge_reconfigure(&null_cfg);
2088 if (reconfiguring) {
2091 reconf_txn = ovsdb_idl_txn_create(idl);
2093 if (bridge_reconfigure_continue(cfg)) {
2094 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2097 bridge_reconfigure_continue(&null_cfg);
2102 ovsdb_idl_txn_commit(reconf_txn);
2103 ovsdb_idl_txn_destroy(reconf_txn);
2107 /* Refresh system and interface stats if necessary. */
2108 if (time_msec() >= stats_timer) {
2110 struct ovsdb_idl_txn *txn;
2112 txn = ovsdb_idl_txn_create(idl);
2113 HMAP_FOR_EACH (br, node, &all_bridges) {
2117 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2118 struct iface *iface;
2120 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2121 iface_refresh_stats(iface);
2122 iface_refresh_status(iface);
2126 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2127 mirror_refresh_stats(m);
2131 refresh_system_stats(cfg);
2132 refresh_controller_status();
2133 ovsdb_idl_txn_commit(txn);
2134 ovsdb_idl_txn_destroy(txn); /* XXX */
2137 stats_timer = time_msec() + STATS_INTERVAL;
2140 if (time_msec() >= db_limiter) {
2141 struct ovsdb_idl_txn *txn;
2143 txn = ovsdb_idl_txn_create(idl);
2144 HMAP_FOR_EACH (br, node, &all_bridges) {
2145 struct iface *iface;
2148 br_refresh_stp_status(br);
2150 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2151 port_refresh_stp_status(port);
2154 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2155 const char *link_state;
2156 int64_t link_resets;
2159 if (iface_is_synthetic(iface)) {
2163 current = ofproto_port_is_lacp_current(br->ofproto,
2167 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2169 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2172 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2173 ovsrec_interface_set_link_state(iface->cfg, link_state);
2175 link_resets = netdev_get_carrier_resets(iface->netdev);
2176 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2180 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2181 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2183 ovsdb_idl_txn_destroy(txn);
2186 refresh_cfm_stats();
2192 ovsdb_idl_wait(idl);
2194 if (reconfiguring) {
2195 poll_immediate_wake();
2198 if (!hmap_is_empty(&all_bridges)) {
2201 HMAP_FOR_EACH (br, node, &all_bridges) {
2202 ofproto_wait(br->ofproto);
2204 poll_timer_wait_until(stats_timer);
2206 if (db_limiter > time_msec()) {
2207 poll_timer_wait_until(db_limiter);
2212 /* QoS unixctl user interface functions. */
2214 struct qos_unixctl_show_cbdata {
2216 struct iface *iface;
2220 qos_unixctl_show_cb(unsigned int queue_id,
2221 const struct shash *details,
2224 struct qos_unixctl_show_cbdata *data = aux;
2225 struct ds *ds = data->ds;
2226 struct iface *iface = data->iface;
2227 struct netdev_queue_stats stats;
2228 struct shash_node *node;
2231 ds_put_cstr(ds, "\n");
2233 ds_put_format(ds, "Queue %u:\n", queue_id);
2235 ds_put_cstr(ds, "Default:\n");
2238 SHASH_FOR_EACH (node, details) {
2239 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2242 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2244 if (stats.tx_packets != UINT64_MAX) {
2245 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2248 if (stats.tx_bytes != UINT64_MAX) {
2249 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2252 if (stats.tx_errors != UINT64_MAX) {
2253 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2256 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2257 queue_id, strerror(error));
2262 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2263 const char *argv[], void *aux OVS_UNUSED)
2265 struct ds ds = DS_EMPTY_INITIALIZER;
2266 struct shash sh = SHASH_INITIALIZER(&sh);
2267 struct iface *iface;
2269 struct shash_node *node;
2270 struct qos_unixctl_show_cbdata data;
2273 iface = iface_find(argv[1]);
2275 unixctl_command_reply_error(conn, "no such interface");
2279 netdev_get_qos(iface->netdev, &type, &sh);
2281 if (*type != '\0') {
2282 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2284 SHASH_FOR_EACH (node, &sh) {
2285 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2290 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2293 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2295 unixctl_command_reply(conn, ds_cstr(&ds));
2297 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2298 unixctl_command_reply_error(conn, ds_cstr(&ds));
2301 shash_destroy_free_data(&sh);
2305 /* Bridge reconfiguration functions. */
2307 bridge_create(const struct ovsrec_bridge *br_cfg)
2311 assert(!bridge_lookup(br_cfg->name));
2312 br = xzalloc(sizeof *br);
2314 br->name = xstrdup(br_cfg->name);
2315 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2318 /* Derive the default Ethernet address from the bridge's UUID. This should
2319 * be unique and it will be stable between ovs-vswitchd runs. */
2320 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2321 eth_addr_mark_random(br->default_ea);
2323 hmap_init(&br->ports);
2324 hmap_init(&br->ifaces);
2325 hmap_init(&br->iface_by_name);
2326 hmap_init(&br->mirrors);
2328 hmap_init(&br->if_cfg_todo);
2329 list_init(&br->ofpp_garbage);
2331 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2335 bridge_destroy(struct bridge *br)
2338 struct mirror *mirror, *next_mirror;
2339 struct port *port, *next_port;
2340 struct if_cfg *if_cfg, *next_if_cfg;
2341 struct ofpp_garbage *garbage, *next_garbage;
2343 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2346 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2347 mirror_destroy(mirror);
2349 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2350 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2353 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2354 &br->ofpp_garbage) {
2355 list_remove(&garbage->list_node);
2359 hmap_remove(&all_bridges, &br->node);
2360 ofproto_destroy(br->ofproto);
2361 hmap_destroy(&br->ifaces);
2362 hmap_destroy(&br->ports);
2363 hmap_destroy(&br->iface_by_name);
2364 hmap_destroy(&br->mirrors);
2365 hmap_destroy(&br->if_cfg_todo);
2372 static struct bridge *
2373 bridge_lookup(const char *name)
2377 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2378 if (!strcmp(br->name, name)) {
2385 /* Handle requests for a listing of all flows known by the OpenFlow
2386 * stack, including those normally hidden. */
2388 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2389 const char *argv[], void *aux OVS_UNUSED)
2394 br = bridge_lookup(argv[1]);
2396 unixctl_command_reply_error(conn, "Unknown bridge");
2401 ofproto_get_all_flows(br->ofproto, &results);
2403 unixctl_command_reply(conn, ds_cstr(&results));
2404 ds_destroy(&results);
2407 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2408 * connections and reconnect. If BRIDGE is not specified, then all bridges
2409 * drop their controller connections and reconnect. */
2411 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2412 const char *argv[], void *aux OVS_UNUSED)
2416 br = bridge_lookup(argv[1]);
2418 unixctl_command_reply_error(conn, "Unknown bridge");
2421 ofproto_reconnect_controllers(br->ofproto);
2423 HMAP_FOR_EACH (br, node, &all_bridges) {
2424 ofproto_reconnect_controllers(br->ofproto);
2427 unixctl_command_reply(conn, NULL);
2431 bridge_get_controllers(const struct bridge *br,
2432 struct ovsrec_controller ***controllersp)
2434 struct ovsrec_controller **controllers;
2435 size_t n_controllers;
2437 controllers = br->cfg->controller;
2438 n_controllers = br->cfg->n_controller;
2440 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2446 *controllersp = controllers;
2448 return n_controllers;
2452 bridge_queue_if_cfg(struct bridge *br,
2453 const struct ovsrec_interface *cfg,
2454 const struct ovsrec_port *parent)
2456 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2459 if_cfg->parent = parent;
2460 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2461 hash_string(if_cfg->cfg->name, 0));
2464 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2465 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2466 * 'br' needs to complete its configuration. */
2468 bridge_add_del_ports(struct bridge *br,
2469 const unsigned long int *splinter_vlans)
2471 struct shash_node *port_node;
2472 struct port *port, *next;
2473 struct shash new_ports;
2476 assert(hmap_is_empty(&br->if_cfg_todo));
2478 /* Collect new ports. */
2479 shash_init(&new_ports);
2480 for (i = 0; i < br->cfg->n_ports; i++) {
2481 const char *name = br->cfg->ports[i]->name;
2482 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2483 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2487 if (bridge_get_controllers(br, NULL)
2488 && !shash_find(&new_ports, br->name)) {
2489 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2490 br->name, br->name);
2492 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2493 br->synth_local_port.n_interfaces = 1;
2494 br->synth_local_port.name = br->name;
2496 br->synth_local_iface.name = br->name;
2497 br->synth_local_iface.type = "internal";
2499 br->synth_local_ifacep = &br->synth_local_iface;
2501 shash_add(&new_ports, br->name, &br->synth_local_port);
2504 if (splinter_vlans) {
2505 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2508 /* Get rid of deleted ports.
2509 * Get rid of deleted interfaces on ports that still exist. */
2510 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2511 port->cfg = shash_find_data(&new_ports, port->name);
2515 port_del_ifaces(port);
2519 /* Update iface->cfg and iface->type in interfaces that still exist.
2520 * Add new interfaces to creation queue. */
2521 SHASH_FOR_EACH (port_node, &new_ports) {
2522 const struct ovsrec_port *port = port_node->data;
2525 for (i = 0; i < port->n_interfaces; i++) {
2526 const struct ovsrec_interface *cfg = port->interfaces[i];
2527 struct iface *iface = iface_lookup(br, cfg->name);
2531 iface->type = iface_get_type(cfg, br->cfg);
2533 bridge_queue_if_cfg(br, cfg, port);
2538 shash_destroy(&new_ports);
2541 /* Initializes 'oc' appropriately as a management service controller for
2544 * The caller must free oc->target when it is no longer needed. */
2546 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2547 struct ofproto_controller *oc)
2549 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2550 oc->max_backoff = 0;
2551 oc->probe_interval = 60;
2552 oc->band = OFPROTO_OUT_OF_BAND;
2554 oc->burst_limit = 0;
2555 oc->enable_async_msgs = true;
2558 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2560 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2561 struct ofproto_controller *oc)
2563 const char *config_str;
2565 oc->target = c->target;
2566 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2567 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2568 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2569 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2570 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2571 oc->burst_limit = (c->controller_burst_limit
2572 ? *c->controller_burst_limit : 0);
2573 oc->enable_async_msgs = (!c->enable_async_messages
2574 || *c->enable_async_messages);
2575 config_str = ovsrec_controller_get_other_config_value(c, "dscp", NULL);
2577 oc->dscp = DSCP_DEFAULT;
2579 int dscp = atoi(config_str);
2581 if (dscp >= 0 && dscp <= 63) {
2587 /* Configures the IP stack for 'br''s local interface properly according to the
2588 * configuration in 'c'. */
2590 bridge_configure_local_iface_netdev(struct bridge *br,
2591 struct ovsrec_controller *c)
2593 struct netdev *netdev;
2594 struct in_addr mask, gateway;
2596 struct iface *local_iface;
2599 /* If there's no local interface or no IP address, give up. */
2600 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2601 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2605 /* Bring up the local interface. */
2606 netdev = local_iface->netdev;
2607 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2609 /* Configure the IP address and netmask. */
2610 if (!c->local_netmask
2611 || !inet_aton(c->local_netmask, &mask)
2613 mask.s_addr = guess_netmask(ip.s_addr);
2615 if (!netdev_set_in4(netdev, ip, mask)) {
2616 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2617 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2620 /* Configure the default gateway. */
2621 if (c->local_gateway
2622 && inet_aton(c->local_gateway, &gateway)
2623 && gateway.s_addr) {
2624 if (!netdev_add_router(netdev, gateway)) {
2625 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2626 br->name, IP_ARGS(&gateway.s_addr));
2631 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2632 * in either string are treated as equal to any number of slashes in the other,
2633 * e.g. "x///y" is equal to "x/y". */
2635 equal_pathnames(const char *a, const char *b)
2639 a += strspn(a, "/");
2640 b += strspn(b, "/");
2641 } else if (*a == '\0') {
2652 bridge_configure_remotes(struct bridge *br,
2653 const struct sockaddr_in *managers, size_t n_managers)
2655 const char *disable_ib_str, *queue_id_str;
2656 bool disable_in_band = false;
2659 struct ovsrec_controller **controllers;
2660 size_t n_controllers;
2662 enum ofproto_fail_mode fail_mode;
2664 struct ofproto_controller *ocs;
2668 /* Check if we should disable in-band control on this bridge. */
2669 disable_ib_str = ovsrec_bridge_get_other_config_value(br->cfg,
2672 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2673 disable_in_band = true;
2676 /* Set OpenFlow queue ID for in-band control. */
2677 queue_id_str = ovsrec_bridge_get_other_config_value(br->cfg,
2680 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2681 ofproto_set_in_band_queue(br->ofproto, queue_id);
2683 if (disable_in_band) {
2684 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2686 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2689 n_controllers = bridge_get_controllers(br, &controllers);
2691 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2694 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2695 for (i = 0; i < n_controllers; i++) {
2696 struct ovsrec_controller *c = controllers[i];
2698 if (!strncmp(c->target, "punix:", 6)
2699 || !strncmp(c->target, "unix:", 5)) {
2700 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2703 whitelist = xasprintf("unix:%s/%s.controller",
2704 ovs_rundir(), br->name);
2705 if (!equal_pathnames(c->target, whitelist)) {
2706 /* Prevent remote ovsdb-server users from accessing arbitrary
2707 * Unix domain sockets and overwriting arbitrary local
2709 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2710 "controller \"%s\" due to possibility for remote "
2711 "exploit. Instead, specify whitelisted \"%s\" or "
2712 "connect to \"unix:%s/%s.mgmt\" (which is always "
2713 "available without special configuration).",
2714 br->name, c->target, whitelist,
2715 ovs_rundir(), br->name);
2723 bridge_configure_local_iface_netdev(br, c);
2724 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2725 if (disable_in_band) {
2726 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2731 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2732 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2735 /* Set the fail-mode. */
2736 fail_mode = !br->cfg->fail_mode
2737 || !strcmp(br->cfg->fail_mode, "standalone")
2738 ? OFPROTO_FAIL_STANDALONE
2739 : OFPROTO_FAIL_SECURE;
2740 ofproto_set_fail_mode(br->ofproto, fail_mode);
2742 /* Configure OpenFlow controller connection snooping. */
2743 if (!ofproto_has_snoops(br->ofproto)) {
2747 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2748 ovs_rundir(), br->name));
2749 ofproto_set_snoops(br->ofproto, &snoops);
2750 sset_destroy(&snoops);
2755 bridge_configure_tables(struct bridge *br)
2757 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2761 n_tables = ofproto_get_n_tables(br->ofproto);
2763 for (i = 0; i < n_tables; i++) {
2764 struct ofproto_table_settings s;
2767 s.max_flows = UINT_MAX;
2771 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2772 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2775 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2776 s.max_flows = *cfg->flow_limit;
2778 if (cfg->overflow_policy
2779 && !strcmp(cfg->overflow_policy, "evict")) {
2782 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2783 for (k = 0; k < cfg->n_groups; k++) {
2784 const char *string = cfg->groups[k];
2787 msg = mf_parse_subfield__(&s.groups[k], &string);
2789 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2790 "'groups' (%s)", br->name, i, msg);
2792 } else if (*string) {
2793 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2794 "element '%s' contains trailing garbage",
2795 br->name, i, cfg->groups[k]);
2803 ofproto_configure_table(br->ofproto, i, &s);
2807 for (; j < br->cfg->n_flow_tables; j++) {
2808 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2809 "%"PRId64" not supported by this datapath", br->name,
2810 br->cfg->key_flow_tables[j]);
2814 /* Port functions. */
2816 static struct port *
2817 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2821 port = xzalloc(sizeof *port);
2823 port->name = xstrdup(cfg->name);
2825 list_init(&port->ifaces);
2827 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2831 /* Deletes interfaces from 'port' that are no longer configured for it. */
2833 port_del_ifaces(struct port *port)
2835 struct iface *iface, *next;
2836 struct sset new_ifaces;
2839 /* Collect list of new interfaces. */
2840 sset_init(&new_ifaces);
2841 for (i = 0; i < port->cfg->n_interfaces; i++) {
2842 const char *name = port->cfg->interfaces[i]->name;
2843 const char *type = port->cfg->interfaces[i]->type;
2844 if (strcmp(type, "null")) {
2845 sset_add(&new_ifaces, name);
2849 /* Get rid of deleted interfaces. */
2850 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2851 if (!sset_contains(&new_ifaces, iface->name)) {
2852 iface_destroy(iface);
2856 sset_destroy(&new_ifaces);
2860 port_destroy(struct port *port)
2863 struct bridge *br = port->bridge;
2864 struct iface *iface, *next;
2867 ofproto_bundle_unregister(br->ofproto, port);
2870 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2871 iface_destroy(iface);
2874 hmap_remove(&br->ports, &port->hmap_node);
2880 static struct port *
2881 port_lookup(const struct bridge *br, const char *name)
2885 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2887 if (!strcmp(port->name, name)) {
2895 enable_lacp(struct port *port, bool *activep)
2897 if (!port->cfg->lacp) {
2898 /* XXX when LACP implementation has been sufficiently tested, enable by
2899 * default and make active on bonded ports. */
2901 } else if (!strcmp(port->cfg->lacp, "off")) {
2903 } else if (!strcmp(port->cfg->lacp, "active")) {
2906 } else if (!strcmp(port->cfg->lacp, "passive")) {
2910 VLOG_WARN("port %s: unknown LACP mode %s",
2911 port->name, port->cfg->lacp);
2916 static struct lacp_settings *
2917 port_configure_lacp(struct port *port, struct lacp_settings *s)
2919 const char *lacp_time, *system_id;
2922 if (!enable_lacp(port, &s->active)) {
2926 s->name = port->name;
2928 system_id = ovsrec_port_get_other_config_value(port->cfg, "lacp-system-id",
2931 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2932 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2933 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2934 " address.", port->name, system_id);
2938 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2941 if (eth_addr_is_zero(s->id)) {
2942 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2946 /* Prefer bondable links if unspecified. */
2947 priority = atoi(ovsrec_port_get_other_config_value(port->cfg,
2948 "lacp-system-priority",
2950 s->priority = (priority > 0 && priority <= UINT16_MAX
2952 : UINT16_MAX - !list_is_short(&port->ifaces));
2954 lacp_time = ovsrec_port_get_other_config_value(port->cfg, "lacp-time",
2956 s->fast = !strcasecmp(lacp_time, "fast");
2961 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2963 int priority, portid, key;
2965 portid = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2969 atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2970 "lacp-port-priority",
2972 key = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2973 "lacp-aggregation-key",
2976 if (portid <= 0 || portid > UINT16_MAX) {
2977 portid = iface->ofp_port;
2980 if (priority <= 0 || priority > UINT16_MAX) {
2981 priority = UINT16_MAX;
2984 if (key < 0 || key > UINT16_MAX) {
2988 s->name = iface->name;
2990 s->priority = priority;
2995 port_configure_bond(struct port *port, struct bond_settings *s,
2996 uint32_t *bond_stable_ids)
2998 const char *detect_s;
2999 struct iface *iface;
3000 int miimon_interval;
3003 s->name = port->name;
3005 if (port->cfg->bond_mode) {
3006 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3007 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3008 port->name, port->cfg->bond_mode,
3009 bond_mode_to_string(s->balance));
3012 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3014 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3015 * active-backup. At some point we should remove this warning. */
3016 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3017 " in previous versions, the default bond_mode was"
3018 " balance-slb", port->name,
3019 bond_mode_to_string(s->balance));
3021 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3022 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3023 "please use another bond type or disable flood_vlans",
3028 atoi(ovsrec_port_get_other_config_value(port->cfg,
3029 "bond-miimon-interval", "0"));
3030 if (miimon_interval <= 0) {
3031 miimon_interval = 200;
3034 detect_s = ovsrec_port_get_other_config_value(port->cfg,
3037 if (!strcmp(detect_s, "carrier")) {
3038 miimon_interval = 0;
3039 } else if (strcmp(detect_s, "miimon")) {
3040 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3041 "defaulting to carrier", port->name, detect_s);
3042 miimon_interval = 0;
3045 s->up_delay = MAX(0, port->cfg->bond_updelay);
3046 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3047 s->basis = atoi(ovsrec_port_get_other_config_value(port->cfg,
3050 s->rebalance_interval = atoi(
3051 ovsrec_port_get_other_config_value(port->cfg,
3052 "bond-rebalance-interval",
3054 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3055 s->rebalance_interval = 1000;
3058 s->fake_iface = port->cfg->bond_fake_iface;
3061 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3062 long long stable_id;
3065 atoll(ovsrec_interface_get_other_config_value(iface->cfg,
3068 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3069 stable_id = iface->ofp_port;
3071 bond_stable_ids[i++] = stable_id;
3073 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3077 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3078 * instead of obtaining it from the database. */
3080 port_is_synthetic(const struct port *port)
3082 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3085 /* Interface functions. */
3087 /* Returns the correct network device type for interface 'iface' in bridge
3090 iface_get_type(const struct ovsrec_interface *iface,
3091 const struct ovsrec_bridge *br)
3093 /* The local port always has type "internal". Other ports take their type
3094 * from the database and default to "system" if none is specified. */
3095 return (!strcmp(iface->name, br->name) ? "internal"
3096 : iface->type[0] ? iface->type
3101 iface_destroy(struct iface *iface)
3104 struct port *port = iface->port;
3105 struct bridge *br = port->bridge;
3107 if (br->ofproto && iface->ofp_port >= 0) {
3108 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3111 if (iface->ofp_port >= 0) {
3112 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3115 list_remove(&iface->port_elem);
3116 hmap_remove(&br->iface_by_name, &iface->name_node);
3118 netdev_close(iface->netdev);
3125 static struct iface *
3126 iface_lookup(const struct bridge *br, const char *name)
3128 struct iface *iface;
3130 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3131 &br->iface_by_name) {
3132 if (!strcmp(iface->name, name)) {
3140 static struct iface *
3141 iface_find(const char *name)
3143 const struct bridge *br;
3145 HMAP_FOR_EACH (br, node, &all_bridges) {
3146 struct iface *iface = iface_lookup(br, name);
3155 static struct if_cfg *
3156 if_cfg_lookup(const struct bridge *br, const char *name)
3158 struct if_cfg *if_cfg;
3160 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3162 if (!strcmp(if_cfg->cfg->name, name)) {
3170 static struct iface *
3171 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3173 struct iface *iface;
3175 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3176 hash_int(ofp_port, 0), &br->ifaces) {
3177 if (iface->ofp_port == ofp_port) {
3184 /* Set Ethernet address of 'iface', if one is specified in the configuration
3187 iface_set_mac(struct iface *iface)
3189 uint8_t ea[ETH_ADDR_LEN];
3191 if (!strcmp(iface->type, "internal")
3192 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3193 if (iface->ofp_port == OFPP_LOCAL) {
3194 VLOG_ERR("interface %s: ignoring mac in Interface record "
3195 "(use Bridge record to set local port's mac)",
3197 } else if (eth_addr_is_multicast(ea)) {
3198 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3201 int error = netdev_set_etheraddr(iface->netdev, ea);
3203 VLOG_ERR("interface %s: setting MAC failed (%s)",
3204 iface->name, strerror(error));
3210 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3212 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3214 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3215 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3219 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3220 * sets the "ofport" field to -1.
3222 * This is appropriate when 'if_cfg''s interface cannot be created or is
3223 * otherwise invalid. */
3225 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3227 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3228 iface_set_ofport(if_cfg, -1);
3229 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3230 ovsrec_interface_set_admin_state(if_cfg, NULL);
3231 ovsrec_interface_set_duplex(if_cfg, NULL);
3232 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3233 ovsrec_interface_set_link_state(if_cfg, NULL);
3234 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3235 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3236 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3237 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3238 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3239 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3243 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3245 * The value strings in '*shash' are taken directly from values[], not copied,
3246 * so the caller should not modify or free them. */
3248 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3249 struct shash *shash)
3254 for (i = 0; i < n; i++) {
3255 shash_add(shash, keys[i], values[i]);
3259 /* Creates 'keys' and 'values' arrays from 'shash'.
3261 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3262 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3263 * are populated with with strings taken directly from 'shash' and thus have
3264 * the same ownership of the key-value pairs in shash.
3267 shash_to_ovs_idl_map(struct shash *shash,
3268 char ***keys, char ***values, size_t *n)
3272 struct shash_node *sn;
3274 count = shash_count(shash);
3276 k = xmalloc(count * sizeof *k);
3277 v = xmalloc(count * sizeof *v);
3280 SHASH_FOR_EACH(sn, shash) {
3291 struct iface_delete_queues_cbdata {
3292 struct netdev *netdev;
3293 const struct ovsdb_datum *queues;
3297 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3299 union ovsdb_atom atom;
3301 atom.integer = target;
3302 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3306 iface_delete_queues(unsigned int queue_id,
3307 const struct shash *details OVS_UNUSED, void *cbdata_)
3309 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3311 if (!queue_ids_include(cbdata->queues, queue_id)) {
3312 netdev_delete_queue(cbdata->netdev, queue_id);
3317 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3319 struct ofpbuf queues_buf;
3321 ofpbuf_init(&queues_buf, 0);
3323 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3324 netdev_set_qos(iface->netdev, NULL, NULL);
3326 struct iface_delete_queues_cbdata cbdata;
3327 struct shash details;
3331 /* Configure top-level Qos for 'iface'. */
3332 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3333 qos->n_other_config, &details);
3334 netdev_set_qos(iface->netdev, qos->type, &details);
3335 shash_destroy(&details);
3337 /* Deconfigure queues that were deleted. */
3338 cbdata.netdev = iface->netdev;
3339 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3341 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3343 /* Configure queues for 'iface'. */
3345 for (i = 0; i < qos->n_queues; i++) {
3346 const struct ovsrec_queue *queue = qos->value_queues[i];
3347 unsigned int queue_id = qos->key_queues[i];
3349 if (queue_id == 0) {
3353 if (queue->n_dscp == 1) {
3354 struct ofproto_port_queue *port_queue;
3356 port_queue = ofpbuf_put_uninit(&queues_buf,
3357 sizeof *port_queue);
3358 port_queue->queue = queue_id;
3359 port_queue->dscp = queue->dscp[0];
3362 shash_from_ovs_idl_map(queue->key_other_config,
3363 queue->value_other_config,
3364 queue->n_other_config, &details);
3365 netdev_set_queue(iface->netdev, queue_id, &details);
3366 shash_destroy(&details);
3369 shash_init(&details);
3370 netdev_set_queue(iface->netdev, 0, &details);
3371 shash_destroy(&details);
3375 if (iface->ofp_port >= 0) {
3376 const struct ofproto_port_queue *port_queues = queues_buf.data;
3377 size_t n_queues = queues_buf.size / sizeof *port_queues;
3379 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3380 port_queues, n_queues);
3383 netdev_set_policing(iface->netdev,
3384 iface->cfg->ingress_policing_rate,
3385 iface->cfg->ingress_policing_burst);
3387 ofpbuf_uninit(&queues_buf);
3391 iface_configure_cfm(struct iface *iface)
3393 const struct ovsrec_interface *cfg = iface->cfg;
3394 const char *extended_str, *opstate_str;
3395 const char *cfm_ccm_vlan;
3396 struct cfm_settings s;
3398 if (!cfg->n_cfm_mpid) {
3399 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3403 s.mpid = *cfg->cfm_mpid;
3404 s.interval = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3407 cfm_ccm_vlan = ovsrec_interface_get_other_config_value(iface->cfg,
3410 s.ccm_pcp = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3413 if (s.interval <= 0) {
3417 if (!strcasecmp("random", cfm_ccm_vlan)) {
3418 s.ccm_vlan = CFM_RANDOM_VLAN;
3420 s.ccm_vlan = atoi(cfm_ccm_vlan);
3421 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3426 extended_str = ovsrec_interface_get_other_config_value(iface->cfg,
3429 s.extended = !strcasecmp("true", extended_str);
3431 opstate_str = ovsrec_interface_get_other_config_value(iface->cfg,
3434 s.opup = !strcasecmp("up", opstate_str);
3436 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3439 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3440 * instead of obtaining it from the database. */
3442 iface_is_synthetic(const struct iface *iface)
3444 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3448 /* Port mirroring. */
3450 static struct mirror *
3451 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3455 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3456 if (uuid_equals(uuid, &m->uuid)) {
3464 bridge_configure_mirrors(struct bridge *br)
3466 const struct ovsdb_datum *mc;
3467 unsigned long *flood_vlans;
3468 struct mirror *m, *next;
3471 /* Get rid of deleted mirrors. */
3472 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3473 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3474 union ovsdb_atom atom;
3476 atom.uuid = m->uuid;
3477 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3482 /* Add new mirrors and reconfigure existing ones. */
3483 for (i = 0; i < br->cfg->n_mirrors; i++) {
3484 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3485 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3487 m = mirror_create(br, cfg);
3490 if (!mirror_configure(m)) {
3495 /* Update flooded vlans (for RSPAN). */
3496 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3497 br->cfg->n_flood_vlans);
3498 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3499 bitmap_free(flood_vlans);
3502 static struct mirror *
3503 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3507 m = xzalloc(sizeof *m);
3508 m->uuid = cfg->header_.uuid;
3509 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3511 m->name = xstrdup(cfg->name);
3517 mirror_destroy(struct mirror *m)
3520 struct bridge *br = m->bridge;
3523 ofproto_mirror_unregister(br->ofproto, m);
3526 hmap_remove(&br->mirrors, &m->hmap_node);
3533 mirror_collect_ports(struct mirror *m,
3534 struct ovsrec_port **in_ports, int n_in_ports,
3535 void ***out_portsp, size_t *n_out_portsp)
3537 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3538 size_t n_out_ports = 0;
3541 for (i = 0; i < n_in_ports; i++) {
3542 const char *name = in_ports[i]->name;
3543 struct port *port = port_lookup(m->bridge, name);
3545 out_ports[n_out_ports++] = port;
3547 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3548 "port %s", m->bridge->name, m->name, name);
3551 *out_portsp = out_ports;
3552 *n_out_portsp = n_out_ports;
3556 mirror_configure(struct mirror *m)
3558 const struct ovsrec_mirror *cfg = m->cfg;
3559 struct ofproto_mirror_settings s;
3562 if (strcmp(cfg->name, m->name)) {
3564 m->name = xstrdup(cfg->name);
3568 /* Get output port or VLAN. */
3569 if (cfg->output_port) {
3570 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3571 if (!s.out_bundle) {
3572 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3573 m->bridge->name, m->name);
3576 s.out_vlan = UINT16_MAX;
3578 if (cfg->output_vlan) {
3579 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3580 "output vlan; ignoring output vlan",
3581 m->bridge->name, m->name);
3583 } else if (cfg->output_vlan) {
3584 /* The database should prevent invalid VLAN values. */
3585 s.out_bundle = NULL;
3586 s.out_vlan = *cfg->output_vlan;
3588 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3589 m->bridge->name, m->name);
3593 /* Get port selection. */
3594 if (cfg->select_all) {
3595 size_t n_ports = hmap_count(&m->bridge->ports);
3596 void **ports = xmalloc(n_ports * sizeof *ports);
3601 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3611 /* Get ports, dropping ports that don't exist.
3612 * The IDL ensures that there are no duplicates. */
3613 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3614 &s.srcs, &s.n_srcs);
3615 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3616 &s.dsts, &s.n_dsts);
3619 /* Get VLAN selection. */
3620 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3623 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3626 if (s.srcs != s.dsts) {
3635 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3637 * This is deprecated. It is only for compatibility with broken device drivers
3638 * in old versions of Linux that do not properly support VLANs when VLAN
3639 * devices are not used. When broken device drivers are no longer in
3640 * widespread use, we will delete these interfaces. */
3642 static void **blocks;
3643 static size_t n_blocks, allocated_blocks;
3645 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3646 * VLAN splinters are reconfigured. */
3648 register_block(void *block)
3650 if (n_blocks >= allocated_blocks) {
3651 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3653 blocks[n_blocks++] = block;
3656 /* Frees all of the blocks registered with register_block(). */
3658 free_registered_blocks(void)
3662 for (i = 0; i < n_blocks; i++) {
3668 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3671 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3675 value = ovsrec_interface_get_other_config_value(iface_cfg,
3676 "enable-vlan-splinters",
3678 return !strcmp(value, "true");
3681 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3684 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3685 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3686 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3689 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3690 * use, returns NULL.
3692 * Updates 'vlan_splinters_enabled_anywhere'. */
3693 static unsigned long int *
3694 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3696 unsigned long int *splinter_vlans;
3697 struct sset splinter_ifaces;
3698 const char *real_dev_name;
3699 struct shash *real_devs;
3700 struct shash_node *node;
3704 /* Free space allocated for synthesized ports and interfaces, since we're
3705 * in the process of reconstructing all of them. */
3706 free_registered_blocks();
3708 splinter_vlans = bitmap_allocate(4096);
3709 sset_init(&splinter_ifaces);
3710 vlan_splinters_enabled_anywhere = false;
3711 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3712 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3715 for (j = 0; j < br_cfg->n_ports; j++) {
3716 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3719 for (k = 0; k < port_cfg->n_interfaces; k++) {
3720 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3722 if (vlan_splinters_is_enabled(iface_cfg)) {
3723 vlan_splinters_enabled_anywhere = true;
3724 sset_add(&splinter_ifaces, iface_cfg->name);
3725 vlan_bitmap_from_array__(port_cfg->trunks,
3731 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3732 bitmap_set1(splinter_vlans, *port_cfg->tag);
3737 if (!vlan_splinters_enabled_anywhere) {
3738 free(splinter_vlans);
3739 sset_destroy(&splinter_ifaces);
3743 HMAP_FOR_EACH (br, node, &all_bridges) {
3745 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3749 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3750 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3751 * device to be created for it. */
3752 bitmap_set0(splinter_vlans, 0);
3753 bitmap_set0(splinter_vlans, 4095);
3755 /* Delete all VLAN devices that we don't need. */
3757 real_devs = vlandev_get_real_devs();
3758 SHASH_FOR_EACH (node, real_devs) {
3759 const struct vlan_real_dev *real_dev = node->data;
3760 const struct vlan_dev *vlan_dev;
3761 bool real_dev_has_splinters;
3763 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3765 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3766 if (!real_dev_has_splinters
3767 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3768 struct netdev *netdev;
3770 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3771 if (!netdev_get_in4(netdev, NULL, NULL) ||
3772 !netdev_get_in6(netdev, NULL)) {
3773 vlandev_del(vlan_dev->name);
3775 /* It has an IP address configured, so we don't own
3776 * it. Don't delete it. */
3778 netdev_close(netdev);
3785 /* Add all VLAN devices that we need. */
3786 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3789 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3790 if (!vlandev_get_name(real_dev_name, vid)) {
3791 vlandev_add(real_dev_name, vid);
3798 sset_destroy(&splinter_ifaces);
3800 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3801 free(splinter_vlans);
3804 return splinter_vlans;
3807 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3810 configure_splinter_port(struct port *port)
3812 struct ofproto *ofproto = port->bridge->ofproto;
3813 uint16_t realdev_ofp_port;
3814 const char *realdev_name;
3815 struct iface *vlandev, *realdev;
3817 ofproto_bundle_unregister(port->bridge->ofproto, port);
3819 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3822 realdev_name = ovsrec_port_get_other_config_value(port->cfg,
3824 realdev = iface_lookup(port->bridge, realdev_name);
3825 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3827 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3831 static struct ovsrec_port *
3832 synthesize_splinter_port(const char *real_dev_name,
3833 const char *vlan_dev_name, int vid)
3835 struct ovsrec_interface *iface;
3836 struct ovsrec_port *port;
3838 iface = xzalloc(sizeof *iface);
3839 iface->name = xstrdup(vlan_dev_name);
3840 iface->type = "system";
3842 port = xzalloc(sizeof *port);
3843 port->interfaces = xmemdup(&iface, sizeof iface);
3844 port->n_interfaces = 1;
3845 port->name = xstrdup(vlan_dev_name);
3846 port->vlan_mode = "splinter";
3847 port->tag = xmalloc(sizeof *port->tag);
3849 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3850 port->key_other_config[0] = "realdev";
3851 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3852 port->value_other_config[0] = xstrdup(real_dev_name);
3853 port->n_other_config = 1;
3855 register_block(iface);
3856 register_block(iface->name);
3857 register_block(port);
3858 register_block(port->interfaces);
3859 register_block(port->name);
3860 register_block(port->tag);
3861 register_block(port->key_other_config);
3862 register_block(port->value_other_config);
3863 register_block(port->value_other_config[0]);
3868 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3869 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3870 * 1-bit in the 'splinter_vlans' bitmap. */
3872 add_vlan_splinter_ports(struct bridge *br,
3873 const unsigned long int *splinter_vlans,
3874 struct shash *ports)
3878 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3879 * we're modifying 'ports'. */
3880 for (i = 0; i < br->cfg->n_ports; i++) {
3881 const char *name = br->cfg->ports[i]->name;
3882 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3885 for (j = 0; j < port_cfg->n_interfaces; j++) {
3886 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3888 if (vlan_splinters_is_enabled(iface_cfg)) {
3889 const char *real_dev_name;
3892 real_dev_name = iface_cfg->name;
3893 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3894 const char *vlan_dev_name;
3896 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3898 && !shash_find(ports, vlan_dev_name)) {
3899 shash_add(ports, vlan_dev_name,
3900 synthesize_splinter_port(
3901 real_dev_name, vlan_dev_name, vid));
3910 mirror_refresh_stats(struct mirror *m)
3912 struct ofproto *ofproto = m->bridge->ofproto;
3913 uint64_t tx_packets, tx_bytes;
3916 size_t stat_cnt = 0;
3918 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3919 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3923 if (tx_packets != UINT64_MAX) {
3924 keys[stat_cnt] = "tx_packets";
3925 values[stat_cnt] = tx_packets;
3928 if (tx_bytes != UINT64_MAX) {
3929 keys[stat_cnt] = "tx_bytes";
3930 values[stat_cnt] = tx_bytes;
3934 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);