1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
22 #include "async-append.h"
26 #include "connectivity.h"
31 #include "dynamic-string.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
44 #include "ofp-print.h"
47 #include "ofproto/bond.h"
48 #include "ofproto/ofproto.h"
50 #include "poll-loop.h"
55 #include "socket-util.h"
57 #include "stream-ssl.h"
59 #include "system-stats.h"
64 #include "lib/vswitch-idl.h"
65 #include "xenserver.h"
66 #include "openvswitch/vlog.h"
67 #include "sflow_api.h"
68 #include "vlan-bitmap.h"
71 VLOG_DEFINE_THIS_MODULE(bridge);
73 COVERAGE_DEFINE(bridge_reconfigure);
76 /* These members are always valid.
78 * They are immutable: they never change between iface_create() and
80 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
85 struct netdev *netdev; /* Network device. */
86 ofp_port_t ofp_port; /* OpenFlow port number. */
89 /* These members are valid only within bridge_reconfigure(). */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct ovs_list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 /* Port mirroring. */
131 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
134 struct hmap mappings; /* "struct" indexed by UUID */
136 /* Used during reconfiguration. */
137 struct shash wanted_ports;
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
146 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
147 struct bridge *bridge;
153 /* All bridges, indexed by name. */
154 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
156 /* OVSDB IDL used to obtain configuration. */
157 static struct ovsdb_idl *idl;
159 /* We want to complete daemonization, fully detaching from our parent process,
160 * only after we have completed our initial configuration, committed our state
161 * to the database, and received confirmation back from the database server
162 * that it applied the commit. This allows our parent process to know that,
163 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
164 * to have already been filled in. (It is only "very likely" rather than
165 * certain because there is always a slim possibility that the transaction will
166 * fail or that some other client has added new bridges, ports, etc. while
167 * ovs-vswitchd was configuring using an old configuration.)
169 * We only need to do this once for our initial configuration at startup, so
170 * 'initial_config_done' tracks whether we've already done it. While we are
171 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
172 * itself and is otherwise NULL. */
173 static bool initial_config_done;
174 static struct ovsdb_idl_txn *daemonize_txn;
176 /* Most recently processed IDL sequence number. */
177 static unsigned int idl_seqno;
179 /* Track changes to port connectivity. */
180 static uint64_t connectivity_seqno = LLONG_MIN;
182 /* Status update to database.
184 * Some information in the database must be kept as up-to-date as possible to
185 * allow controllers to respond rapidly to network outages. Those status are
186 * updated via the 'status_txn'.
188 * We use the global connectivity sequence number to detect the status change.
189 * Also, to prevent the status update from sending too much to the database,
190 * we check the return status of each update transaction and do not start new
191 * update if the previous transaction status is 'TXN_INCOMPLETE'.
193 * 'statux_txn' is NULL if there is no ongoing status update.
195 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
196 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
197 * which will cause the main thread wake up soon and retry the status update.
199 static struct ovsdb_idl_txn *status_txn;
200 static bool status_txn_try_again;
202 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
203 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
204 #define STATUS_CHECK_AGAIN_MSEC 100
206 /* Statistics update to database. */
207 static struct ovsdb_idl_txn *stats_txn;
209 /* Each time this timer expires, the bridge fetches interface and mirror
210 * statistics and pushes them into the database. */
211 static int stats_timer_interval;
212 static long long int stats_timer = LLONG_MIN;
214 /* Each time this timer expires, the bridge fetches the list of port/VLAN
215 * membership that has been modified by the AA.
217 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
218 static long long int aa_refresh_timer = LLONG_MIN;
220 static void add_del_bridges(const struct ovsrec_open_vswitch *);
221 static void bridge_run__(void);
222 static void bridge_create(const struct ovsrec_bridge *);
223 static void bridge_destroy(struct bridge *);
224 static struct bridge *bridge_lookup(const char *name);
225 static unixctl_cb_func bridge_unixctl_dump_flows;
226 static unixctl_cb_func bridge_unixctl_reconnect;
227 static size_t bridge_get_controllers(const struct bridge *br,
228 struct ovsrec_controller ***controllersp);
229 static void bridge_collect_wanted_ports(struct bridge *,
230 const unsigned long *splinter_vlans,
231 struct shash *wanted_ports);
232 static void bridge_delete_ofprotos(void);
233 static void bridge_delete_or_reconfigure_ports(struct bridge *);
234 static void bridge_del_ports(struct bridge *,
235 const struct shash *wanted_ports);
236 static void bridge_add_ports(struct bridge *,
237 const struct shash *wanted_ports);
239 static void bridge_configure_datapath_id(struct bridge *);
240 static void bridge_configure_netflow(struct bridge *);
241 static void bridge_configure_forward_bpdu(struct bridge *);
242 static void bridge_configure_mac_table(struct bridge *);
243 static void bridge_configure_mcast_snooping(struct bridge *);
244 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
245 static void bridge_configure_ipfix(struct bridge *);
246 static void bridge_configure_spanning_tree(struct bridge *);
247 static void bridge_configure_tables(struct bridge *);
248 static void bridge_configure_dp_desc(struct bridge *);
249 static void bridge_configure_aa(struct bridge *);
250 static void bridge_aa_refresh_queued(struct bridge *);
251 static bool bridge_aa_need_refresh(struct bridge *);
252 static void bridge_configure_remotes(struct bridge *,
253 const struct sockaddr_in *managers,
255 static void bridge_pick_local_hw_addr(struct bridge *,
256 uint8_t ea[ETH_ADDR_LEN],
257 struct iface **hw_addr_iface);
258 static uint64_t bridge_pick_datapath_id(struct bridge *,
259 const uint8_t bridge_ea[ETH_ADDR_LEN],
260 struct iface *hw_addr_iface);
261 static uint64_t dpid_from_hash(const void *, size_t nbytes);
262 static bool bridge_has_bond_fake_iface(const struct bridge *,
264 static bool port_is_bond_fake_iface(const struct port *);
266 static unixctl_cb_func qos_unixctl_show;
268 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
269 static void port_del_ifaces(struct port *);
270 static void port_destroy(struct port *);
271 static struct port *port_lookup(const struct bridge *, const char *name);
272 static void port_configure(struct port *);
273 static struct lacp_settings *port_configure_lacp(struct port *,
274 struct lacp_settings *);
275 static void port_configure_bond(struct port *, struct bond_settings *);
276 static bool port_is_synthetic(const struct port *);
278 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
279 static void run_system_stats(void);
281 static void bridge_configure_mirrors(struct bridge *);
282 static struct mirror *mirror_create(struct bridge *,
283 const struct ovsrec_mirror *);
284 static void mirror_destroy(struct mirror *);
285 static bool mirror_configure(struct mirror *);
286 static void mirror_refresh_stats(struct mirror *);
288 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
289 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
290 const struct ovsrec_port *);
291 static bool iface_is_internal(const struct ovsrec_interface *iface,
292 const struct ovsrec_bridge *br);
293 static const char *iface_get_type(const struct ovsrec_interface *,
294 const struct ovsrec_bridge *);
295 static void iface_destroy(struct iface *);
296 static void iface_destroy__(struct iface *);
297 static struct iface *iface_lookup(const struct bridge *, const char *name);
298 static struct iface *iface_find(const char *name);
299 static struct iface *iface_from_ofp_port(const struct bridge *,
300 ofp_port_t ofp_port);
301 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
302 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
303 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
304 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
305 static void iface_configure_cfm(struct iface *);
306 static void iface_refresh_cfm_stats(struct iface *);
307 static void iface_refresh_stats(struct iface *);
308 static void iface_refresh_netdev_status(struct iface *);
309 static void iface_refresh_ofproto_status(struct iface *);
310 static bool iface_is_synthetic(const struct iface *);
311 static ofp_port_t iface_get_requested_ofp_port(
312 const struct ovsrec_interface *);
313 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
316 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
318 * This is deprecated. It is only for compatibility with broken device drivers
319 * in old versions of Linux that do not properly support VLANs when VLAN
320 * devices are not used. When broken device drivers are no longer in
321 * widespread use, we will delete these interfaces. */
323 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
324 static bool vlan_splinters_enabled_anywhere;
326 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
327 static unsigned long int *collect_splinter_vlans(
328 const struct ovsrec_open_vswitch *);
329 static void configure_splinter_port(struct port *);
330 static void add_vlan_splinter_ports(struct bridge *,
331 const unsigned long int *splinter_vlans,
332 struct shash *ports);
334 static void discover_types(const struct ovsrec_open_vswitch *cfg);
337 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
339 struct shash iface_hints;
340 static bool initialized = false;
347 shash_init(&iface_hints);
350 for (i = 0; i < cfg->n_bridges; i++) {
351 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
354 for (j = 0; j < br_cfg->n_ports; j++) {
355 struct ovsrec_port *port_cfg = br_cfg->ports[j];
358 for (k = 0; k < port_cfg->n_interfaces; k++) {
359 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
360 struct iface_hint *iface_hint;
362 iface_hint = xmalloc(sizeof *iface_hint);
363 iface_hint->br_name = br_cfg->name;
364 iface_hint->br_type = br_cfg->datapath_type;
365 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
367 shash_add(&iface_hints, if_cfg->name, iface_hint);
373 ofproto_init(&iface_hints);
375 shash_destroy_free_data(&iface_hints);
379 /* Public functions. */
381 /* Initializes the bridge module, configuring it to obtain its configuration
382 * from an OVSDB server accessed over 'remote', which should be a string in a
383 * form acceptable to ovsdb_idl_create(). */
385 bridge_init(const char *remote)
387 /* Create connection to database. */
388 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
389 idl_seqno = ovsdb_idl_get_seqno(idl);
390 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
391 ovsdb_idl_verify_write_only(idl);
393 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
394 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
395 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
396 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
397 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
399 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
400 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
401 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
403 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
404 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
405 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
406 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
407 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
408 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
409 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
411 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
412 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
413 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
414 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
415 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
416 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
417 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
418 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
419 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
420 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
421 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
422 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
423 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
424 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
425 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
426 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
427 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
428 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
429 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
430 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
439 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
441 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
442 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
443 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
444 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
446 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
448 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
450 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
451 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
453 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
454 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
455 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
456 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
458 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
459 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
460 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
461 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
462 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
464 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
466 /* Register unixctl commands. */
467 unixctl_command_register("qos/show", "interface", 1, 1,
468 qos_unixctl_show, NULL);
469 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
470 bridge_unixctl_dump_flows, NULL);
471 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
472 bridge_unixctl_reconnect, NULL);
485 struct bridge *br, *next_br;
487 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
490 ovsdb_idl_destroy(idl);
493 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
494 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
495 * responsible for freeing '*managersp' (with free()).
497 * You may be asking yourself "why does ovs-vswitchd care?", because
498 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
499 * should not be and in fact is not directly involved in that. But
500 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
501 * it has to tell in-band control where the managers are to enable that.
502 * (Thus, only managers connected in-band and with non-loopback addresses
506 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
507 struct sockaddr_in **managersp, size_t *n_managersp)
509 struct sockaddr_in *managers = NULL;
510 size_t n_managers = 0;
514 /* Collect all of the potential targets from the "targets" columns of the
515 * rows pointed to by "manager_options", excluding any that are
518 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
519 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
521 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
522 sset_find_and_delete(&targets, m->target);
524 sset_add(&targets, m->target);
528 /* Now extract the targets' IP addresses. */
529 if (!sset_is_empty(&targets)) {
532 managers = xmalloc(sset_count(&targets) * sizeof *managers);
533 SSET_FOR_EACH (target, &targets) {
535 struct sockaddr_storage ss;
536 struct sockaddr_in in;
539 /* Ignore loopback. */
540 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
542 && sa.ss.ss_family == AF_INET
543 && sa.in.sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
544 managers[n_managers++] = sa.in;
548 sset_destroy(&targets);
550 *managersp = managers;
551 *n_managersp = n_managers;
555 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
557 unsigned long int *splinter_vlans;
558 struct sockaddr_in *managers;
559 struct bridge *br, *next;
560 int sflow_bridge_number;
563 COVERAGE_INC(bridge_reconfigure);
565 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
566 OFPROTO_FLOW_LIMIT_DEFAULT));
567 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
568 OFPROTO_MAX_IDLE_DEFAULT));
569 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
571 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
574 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
575 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
577 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
578 * to 'ovs_cfg', with only very minimal configuration otherwise.
580 * This is mostly an update to bridge data structures. Nothing is pushed
581 * down to ofproto or lower layers. */
582 add_del_bridges(ovs_cfg);
583 splinter_vlans = collect_splinter_vlans(ovs_cfg);
584 HMAP_FOR_EACH (br, node, &all_bridges) {
585 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
586 bridge_del_ports(br, &br->wanted_ports);
588 free(splinter_vlans);
590 /* Start pushing configuration changes down to the ofproto layer:
592 * - Delete ofprotos that are no longer configured.
594 * - Delete ports that are no longer configured.
596 * - Reconfigure existing ports to their desired configurations, or
597 * delete them if not possible.
599 * We have to do all the deletions before we can do any additions, because
600 * the ports to be added might require resources that will be freed up by
601 * deletions (they might especially overlap in name). */
602 bridge_delete_ofprotos();
603 HMAP_FOR_EACH (br, node, &all_bridges) {
605 bridge_delete_or_reconfigure_ports(br);
609 /* Finish pushing configuration changes to the ofproto layer:
611 * - Create ofprotos that are missing.
613 * - Add ports that are missing. */
614 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
618 error = ofproto_create(br->name, br->type, &br->ofproto);
620 VLOG_ERR("failed to create bridge %s: %s", br->name,
621 ovs_strerror(error));
622 shash_destroy(&br->wanted_ports);
625 /* Trigger storing datapath version. */
626 seq_change(connectivity_seq_get());
630 HMAP_FOR_EACH (br, node, &all_bridges) {
631 bridge_add_ports(br, &br->wanted_ports);
632 shash_destroy(&br->wanted_ports);
635 reconfigure_system_stats(ovs_cfg);
637 /* Complete the configuration. */
638 sflow_bridge_number = 0;
639 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
640 HMAP_FOR_EACH (br, node, &all_bridges) {
643 /* We need the datapath ID early to allow LACP ports to use it as the
644 * default system ID. */
645 bridge_configure_datapath_id(br);
647 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
650 port_configure(port);
652 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
653 iface_set_ofport(iface->cfg, iface->ofp_port);
654 /* Clear eventual previous errors */
655 ovsrec_interface_set_error(iface->cfg, NULL);
656 iface_configure_cfm(iface);
657 iface_configure_qos(iface, port->cfg->qos);
658 iface_set_mac(br, port, iface);
659 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
661 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
665 bridge_configure_mirrors(br);
666 bridge_configure_forward_bpdu(br);
667 bridge_configure_mac_table(br);
668 bridge_configure_mcast_snooping(br);
669 bridge_configure_remotes(br, managers, n_managers);
670 bridge_configure_netflow(br);
671 bridge_configure_sflow(br, &sflow_bridge_number);
672 bridge_configure_ipfix(br);
673 bridge_configure_spanning_tree(br);
674 bridge_configure_tables(br);
675 bridge_configure_dp_desc(br);
676 bridge_configure_aa(br);
680 /* The ofproto-dpif provider does some final reconfiguration in its
681 * ->type_run() function. We have to call it before notifying the database
682 * client that reconfiguration is complete, otherwise there is a very
683 * narrow race window in which e.g. ofproto/trace will not recognize the
684 * new configuration (sometimes this causes unit test failures). */
688 /* Delete ofprotos which aren't configured or have the wrong type. Create
689 * ofprotos which don't exist but need to. */
691 bridge_delete_ofprotos(void)
698 /* Delete ofprotos with no bridge or with the wrong type. */
701 ofproto_enumerate_types(&types);
702 SSET_FOR_EACH (type, &types) {
705 ofproto_enumerate_names(type, &names);
706 SSET_FOR_EACH (name, &names) {
707 br = bridge_lookup(name);
708 if (!br || strcmp(type, br->type)) {
709 ofproto_delete(name, type);
713 sset_destroy(&names);
714 sset_destroy(&types);
718 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
720 if (*n >= *allocated) {
721 ports = x2nrealloc(ports, allocated, sizeof *ports);
723 ports[(*n)++] = port;
728 bridge_delete_or_reconfigure_ports(struct bridge *br)
730 struct ofproto_port ofproto_port;
731 struct ofproto_port_dump dump;
733 struct sset ofproto_ports;
734 struct port *port, *port_next;
736 /* List of "ofp_port"s to delete. We make a list instead of deleting them
737 * right away because ofproto implementations aren't necessarily able to
738 * iterate through a changing list of ports in an entirely robust way. */
745 sset_init(&ofproto_ports);
747 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
748 * that are not configured in the database. (This commonly happens when
749 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
751 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
752 * that have the wrong OpenFlow port number (and arrange to add them back
753 * with the correct OpenFlow port number). */
754 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
755 ofp_port_t requested_ofp_port;
758 sset_add(&ofproto_ports, ofproto_port.name);
760 iface = iface_lookup(br, ofproto_port.name);
762 /* No such iface is configured, so we should delete this
765 * As a corner case exception, keep the port if it's a bond fake
767 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
768 && !strcmp(ofproto_port.type, "internal")) {
774 if (strcmp(ofproto_port.type, iface->type)
775 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
776 /* The interface is the wrong type or can't be configured.
781 /* If the requested OpenFlow port for 'iface' changed, and it's not
782 * already the correct port, then we might want to temporarily delete
783 * this interface, so we can add it back again with the new OpenFlow
785 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
786 if (iface->ofp_port != OFPP_LOCAL &&
787 requested_ofp_port != OFPP_NONE &&
788 requested_ofp_port != iface->ofp_port) {
789 ofp_port_t victim_request;
790 struct iface *victim;
792 /* Check for an existing OpenFlow port currently occupying
793 * 'iface''s requested port number. If there isn't one, then
794 * delete this port. Otherwise we need to consider further. */
795 victim = iface_from_ofp_port(br, requested_ofp_port);
800 /* 'victim' is a port currently using 'iface''s requested port
801 * number. Unless 'victim' specifically requested that port
802 * number, too, then we can delete both 'iface' and 'victim'
803 * temporarily. (We'll add both of them back again later with new
804 * OpenFlow port numbers.)
806 * If 'victim' did request port number 'requested_ofp_port', just
807 * like 'iface', then that's a configuration inconsistency that we
808 * can't resolve. We might as well let it keep its current port
810 victim_request = iface_get_requested_ofp_port(victim->cfg);
811 if (victim_request != requested_ofp_port) {
812 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
813 iface_destroy(victim);
822 iface_destroy(iface);
823 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
825 for (i = 0; i < n; i++) {
826 ofproto_port_del(br->ofproto, del[i]);
830 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
831 * that we didn't see when we iterated through the datapath, i.e. ports
832 * that disappeared underneath use. This is an unusual situation, but it
833 * can happen in some cases:
835 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
836 * idea but could happen).
838 * - The port represented a device that disappeared, e.g. a tuntap
839 * device destroyed via "tunctl -d", a physical Ethernet device
840 * whose module was just unloaded via "rmmod", or a virtual NIC for a
841 * VM whose VM was just terminated. */
842 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
843 struct iface *iface, *iface_next;
845 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
846 if (!sset_contains(&ofproto_ports, iface->name)) {
847 iface_destroy__(iface);
851 if (list_is_empty(&port->ifaces)) {
855 sset_destroy(&ofproto_ports);
859 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
860 bool with_requested_port)
862 struct shash_node *port_node;
864 SHASH_FOR_EACH (port_node, wanted_ports) {
865 const struct ovsrec_port *port_cfg = port_node->data;
868 for (i = 0; i < port_cfg->n_interfaces; i++) {
869 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
870 ofp_port_t requested_ofp_port;
872 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
873 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
874 struct iface *iface = iface_lookup(br, iface_cfg->name);
877 iface_create(br, iface_cfg, port_cfg);
885 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
887 /* First add interfaces that request a particular port number. */
888 bridge_add_ports__(br, wanted_ports, true);
890 /* Then add interfaces that want automatic port number assignment.
891 * We add these afterward to avoid accidentally taking a specifically
892 * requested port number. */
893 bridge_add_ports__(br, wanted_ports, false);
897 port_configure(struct port *port)
899 const struct ovsrec_port *cfg = port->cfg;
900 struct bond_settings bond_settings;
901 struct lacp_settings lacp_settings;
902 struct ofproto_bundle_settings s;
905 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
906 configure_splinter_port(port);
915 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
916 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
917 s.slaves[s.n_slaves++] = iface->ofp_port;
922 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
926 /* Get VLAN trunks. */
929 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
933 if (cfg->vlan_mode) {
934 if (!strcmp(cfg->vlan_mode, "access")) {
935 s.vlan_mode = PORT_VLAN_ACCESS;
936 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
937 s.vlan_mode = PORT_VLAN_TRUNK;
938 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
939 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
940 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
941 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
943 /* This "can't happen" because ovsdb-server should prevent it. */
944 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
945 "back to trunk mode", port->name, cfg->vlan_mode);
946 s.vlan_mode = PORT_VLAN_TRUNK;
950 s.vlan_mode = PORT_VLAN_ACCESS;
952 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
956 s.vlan_mode = PORT_VLAN_TRUNK;
959 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
962 /* Get LACP settings. */
963 s.lacp = port_configure_lacp(port, &lacp_settings);
967 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
968 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
969 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
972 s.lacp_slaves = NULL;
975 /* Get bond settings. */
976 if (s.n_slaves > 1) {
977 s.bond = &bond_settings;
978 port_configure_bond(port, &bond_settings);
981 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
982 netdev_set_miimon_interval(iface->netdev, 0);
987 ofproto_bundle_register(port->bridge->ofproto, port, &s);
995 /* Pick local port hardware address and datapath ID for 'br'. */
997 bridge_configure_datapath_id(struct bridge *br)
999 uint8_t ea[ETH_ADDR_LEN];
1001 struct iface *local_iface;
1002 struct iface *hw_addr_iface;
1005 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1006 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1008 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1010 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1011 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1012 "Ethernet address: %s",
1013 br->name, ovs_strerror(error));
1016 memcpy(br->ea, ea, ETH_ADDR_LEN);
1018 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1019 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1020 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1021 ofproto_set_datapath_id(br->ofproto, dpid);
1024 dpid_string = xasprintf("%016"PRIx64, dpid);
1025 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1029 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1032 bridge_get_allowed_versions(struct bridge *br)
1034 if (!br->cfg->n_protocols)
1037 return ofputil_versions_from_strings(br->cfg->protocols,
1038 br->cfg->n_protocols);
1041 /* Set NetFlow configuration on 'br'. */
1043 bridge_configure_netflow(struct bridge *br)
1045 struct ovsrec_netflow *cfg = br->cfg->netflow;
1046 struct netflow_options opts;
1049 ofproto_set_netflow(br->ofproto, NULL);
1053 memset(&opts, 0, sizeof opts);
1055 /* Get default NetFlow configuration from datapath.
1056 * Apply overrides from 'cfg'. */
1057 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1058 if (cfg->engine_type) {
1059 opts.engine_type = *cfg->engine_type;
1061 if (cfg->engine_id) {
1062 opts.engine_id = *cfg->engine_id;
1065 /* Configure active timeout interval. */
1066 opts.active_timeout = cfg->active_timeout;
1067 if (!opts.active_timeout) {
1068 opts.active_timeout = -1;
1069 } else if (opts.active_timeout < 0) {
1070 VLOG_WARN("bridge %s: active timeout interval set to negative "
1071 "value, using default instead (%d seconds)", br->name,
1072 NF_ACTIVE_TIMEOUT_DEFAULT);
1073 opts.active_timeout = -1;
1076 /* Add engine ID to interface number to disambiguate bridgs? */
1077 opts.add_id_to_iface = cfg->add_id_to_interface;
1078 if (opts.add_id_to_iface) {
1079 if (opts.engine_id > 0x7f) {
1080 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1081 "another vswitch, choose an engine id less than 128",
1084 if (hmap_count(&br->ports) > 508) {
1085 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1086 "another port when more than 508 ports are used",
1092 sset_init(&opts.collectors);
1093 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1096 if (ofproto_set_netflow(br->ofproto, &opts)) {
1097 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1099 sset_destroy(&opts.collectors);
1102 /* Set sFlow configuration on 'br'. */
1104 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1106 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1107 struct ovsrec_controller **controllers;
1108 struct ofproto_sflow_options oso;
1109 size_t n_controllers;
1113 ofproto_set_sflow(br->ofproto, NULL);
1117 memset(&oso, 0, sizeof oso);
1119 sset_init(&oso.targets);
1120 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1122 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1123 if (cfg->sampling) {
1124 oso.sampling_rate = *cfg->sampling;
1127 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1129 oso.polling_interval = *cfg->polling;
1132 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1134 oso.header_len = *cfg->header;
1137 oso.sub_id = (*sflow_bridge_number)++;
1138 oso.agent_device = cfg->agent;
1140 oso.control_ip = NULL;
1141 n_controllers = bridge_get_controllers(br, &controllers);
1142 for (i = 0; i < n_controllers; i++) {
1143 if (controllers[i]->local_ip) {
1144 oso.control_ip = controllers[i]->local_ip;
1148 ofproto_set_sflow(br->ofproto, &oso);
1150 sset_destroy(&oso.targets);
1153 /* Returns whether a IPFIX row is valid. */
1155 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1157 return ipfix && ipfix->n_targets > 0;
1160 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1162 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1163 const struct bridge *br)
1165 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1168 /* Set IPFIX configuration on 'br'. */
1170 bridge_configure_ipfix(struct bridge *br)
1172 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1173 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1174 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1175 struct ofproto_ipfix_bridge_exporter_options be_opts;
1176 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1177 size_t n_fe_opts = 0;
1179 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1180 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1185 if (!valid_be_cfg && n_fe_opts == 0) {
1186 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1191 memset(&be_opts, 0, sizeof be_opts);
1193 sset_init(&be_opts.targets);
1194 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1196 if (be_cfg->sampling) {
1197 be_opts.sampling_rate = *be_cfg->sampling;
1199 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1201 if (be_cfg->obs_domain_id) {
1202 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1204 if (be_cfg->obs_point_id) {
1205 be_opts.obs_point_id = *be_cfg->obs_point_id;
1207 if (be_cfg->cache_active_timeout) {
1208 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1210 if (be_cfg->cache_max_flows) {
1211 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1214 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1215 "enable-tunnel-sampling", true);
1217 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1218 "enable-input-sampling", false);
1220 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1221 "enable-output-sampling", false);
1224 if (n_fe_opts > 0) {
1225 struct ofproto_ipfix_flow_exporter_options *opts;
1226 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1228 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1229 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1230 opts->collector_set_id = fe_cfg->id;
1231 sset_init(&opts->targets);
1232 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1233 fe_cfg->ipfix->n_targets);
1234 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1235 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1236 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1237 ? *fe_cfg->ipfix->cache_max_flows : 0;
1243 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1247 sset_destroy(&be_opts.targets);
1250 if (n_fe_opts > 0) {
1251 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1253 for (i = 0; i < n_fe_opts; i++) {
1254 sset_destroy(&opts->targets);
1262 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1263 struct ofproto_port_stp_settings *port_s,
1264 int *port_num_counter, unsigned long *port_num_bitmap)
1266 const char *config_str;
1267 struct iface *iface;
1269 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1270 port_s->enable = false;
1273 port_s->enable = true;
1276 /* STP over bonds is not supported. */
1277 if (!list_is_singleton(&port->ifaces)) {
1278 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1280 port_s->enable = false;
1284 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1286 /* Internal ports shouldn't participate in spanning tree, so
1288 if (!strcmp(iface->type, "internal")) {
1289 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1290 port_s->enable = false;
1294 /* STP on mirror output ports is not supported. */
1295 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1296 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1297 port_s->enable = false;
1301 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1303 unsigned long int port_num = strtoul(config_str, NULL, 0);
1304 int port_idx = port_num - 1;
1306 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1307 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1308 port_s->enable = false;
1312 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1313 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1314 port->name, port_num);
1315 port_s->enable = false;
1318 bitmap_set1(port_num_bitmap, port_idx);
1319 port_s->port_num = port_idx;
1321 if (*port_num_counter >= STP_MAX_PORTS) {
1322 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1323 port_s->enable = false;
1327 port_s->port_num = (*port_num_counter)++;
1330 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1332 port_s->path_cost = strtoul(config_str, NULL, 10);
1334 enum netdev_features current;
1337 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1338 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1339 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1342 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1344 port_s->priority = strtoul(config_str, NULL, 0);
1346 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1351 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1352 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1354 const char *config_str;
1355 struct iface *iface;
1357 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1358 port_s->enable = false;
1361 port_s->enable = true;
1364 /* RSTP over bonds is not supported. */
1365 if (!list_is_singleton(&port->ifaces)) {
1366 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1368 port_s->enable = false;
1372 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1374 /* Internal ports shouldn't participate in spanning tree, so
1376 if (!strcmp(iface->type, "internal")) {
1377 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1378 port_s->enable = false;
1382 /* RSTP on mirror output ports is not supported. */
1383 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1384 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1385 port_s->enable = false;
1389 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1391 unsigned long int port_num = strtoul(config_str, NULL, 0);
1392 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1393 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1394 port_s->enable = false;
1397 port_s->port_num = port_num;
1399 if (*port_num_counter >= RSTP_MAX_PORTS) {
1400 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1401 port_s->enable = false;
1404 /* If rstp-port-num is not specified, use 0.
1405 * rstp_port_set_port_number() will look for the first free one. */
1406 port_s->port_num = 0;
1409 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1411 port_s->path_cost = strtoul(config_str, NULL, 10);
1413 enum netdev_features current;
1416 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1417 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1418 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1421 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1423 port_s->priority = strtoul(config_str, NULL, 0);
1425 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1428 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1430 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1432 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1435 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1436 "rstp-admin-port-state", true);
1438 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1439 "rstp-port-admin-edge", false);
1440 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1441 "rstp-port-auto-edge", true);
1442 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1443 "rstp-port-mcheck", false);
1446 /* Set spanning tree configuration on 'br'. */
1448 bridge_configure_stp(struct bridge *br, bool enable_stp)
1451 ofproto_set_stp(br->ofproto, NULL);
1453 struct ofproto_stp_settings br_s;
1454 const char *config_str;
1456 int port_num_counter;
1457 unsigned long *port_num_bitmap;
1459 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1461 uint8_t ea[ETH_ADDR_LEN];
1463 if (eth_addr_from_string(config_str, ea)) {
1464 br_s.system_id = eth_addr_to_uint64(ea);
1466 br_s.system_id = eth_addr_to_uint64(br->ea);
1467 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1468 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1471 br_s.system_id = eth_addr_to_uint64(br->ea);
1474 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1476 br_s.priority = strtoul(config_str, NULL, 0);
1478 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1481 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1483 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1485 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1488 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1490 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1492 br_s.max_age = STP_DEFAULT_MAX_AGE;
1495 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1497 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1499 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1502 /* Configure STP on the bridge. */
1503 if (ofproto_set_stp(br->ofproto, &br_s)) {
1504 VLOG_ERR("bridge %s: could not enable STP", br->name);
1508 /* Users must either set the port number with the "stp-port-num"
1509 * configuration on all ports or none. If manual configuration
1510 * is not done, then we allocate them sequentially. */
1511 port_num_counter = 0;
1512 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1513 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1514 struct ofproto_port_stp_settings port_s;
1515 struct iface *iface;
1517 port_configure_stp(br->ofproto, port, &port_s,
1518 &port_num_counter, port_num_bitmap);
1520 /* As bonds are not supported, just apply configuration to
1521 * all interfaces. */
1522 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1523 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1525 VLOG_ERR("port %s: could not enable STP", port->name);
1531 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1532 && port_num_counter) {
1533 VLOG_ERR("bridge %s: must manually configure all STP port "
1534 "IDs or none, disabling", br->name);
1535 ofproto_set_stp(br->ofproto, NULL);
1537 bitmap_free(port_num_bitmap);
1542 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1545 ofproto_set_rstp(br->ofproto, NULL);
1547 struct ofproto_rstp_settings br_s;
1548 const char *config_str;
1550 int port_num_counter;
1552 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1554 uint8_t ea[ETH_ADDR_LEN];
1556 if (eth_addr_from_string(config_str, ea)) {
1557 br_s.address = eth_addr_to_uint64(ea);
1560 br_s.address = eth_addr_to_uint64(br->ea);
1561 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1562 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1566 br_s.address = eth_addr_to_uint64(br->ea);
1569 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1571 br_s.priority = strtoul(config_str, NULL, 0);
1573 br_s.priority = RSTP_DEFAULT_PRIORITY;
1576 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1578 br_s.ageing_time = strtoul(config_str, NULL, 0);
1580 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1583 config_str = smap_get(&br->cfg->other_config,
1584 "rstp-force-protocol-version");
1586 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1588 br_s.force_protocol_version = FPV_DEFAULT;
1591 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1593 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1595 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1598 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1600 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1602 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1605 config_str = smap_get(&br->cfg->other_config,
1606 "rstp-transmit-hold-count");
1608 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1610 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1613 /* Configure RSTP on the bridge. */
1614 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1615 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1619 port_num_counter = 0;
1620 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1621 struct ofproto_port_rstp_settings port_s;
1622 struct iface *iface;
1624 port_configure_rstp(br->ofproto, port, &port_s,
1627 /* As bonds are not supported, just apply configuration to
1628 * all interfaces. */
1629 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1630 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1632 VLOG_ERR("port %s: could not enable RSTP", port->name);
1641 bridge_configure_spanning_tree(struct bridge *br)
1643 bool enable_rstp = br->cfg->rstp_enable;
1644 bool enable_stp = br->cfg->stp_enable;
1646 if (enable_rstp && enable_stp) {
1647 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1648 "configured; enabling RSTP", br->name);
1652 bridge_configure_stp(br, enable_stp);
1653 bridge_configure_rstp(br, enable_rstp);
1657 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1659 const struct port *port = port_lookup(br, name);
1660 return port && port_is_bond_fake_iface(port);
1664 port_is_bond_fake_iface(const struct port *port)
1666 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1670 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1672 struct bridge *br, *next;
1673 struct shash new_br;
1676 /* Collect new bridges' names and types. */
1677 shash_init(&new_br);
1678 for (i = 0; i < cfg->n_bridges; i++) {
1679 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1680 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1682 if (strchr(br_cfg->name, '/')) {
1683 /* Prevent remote ovsdb-server users from accessing arbitrary
1684 * directories, e.g. consider a bridge named "../../../etc/". */
1685 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1687 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1688 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1692 /* Get rid of deleted bridges or those whose types have changed.
1693 * Update 'cfg' of bridges that still exist. */
1694 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1695 br->cfg = shash_find_data(&new_br, br->name);
1696 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1697 br->cfg->datapath_type))) {
1702 /* Add new bridges. */
1703 for (i = 0; i < cfg->n_bridges; i++) {
1704 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1705 struct bridge *br = bridge_lookup(br_cfg->name);
1707 bridge_create(br_cfg);
1711 shash_destroy(&new_br);
1714 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1715 * Returns 0 if successful, otherwise a positive errno value. */
1717 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1718 struct netdev *netdev, char **errp)
1720 return netdev_set_config(netdev, &iface_cfg->options, errp);
1723 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1724 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1726 * If successful, returns 0 and stores the network device in '*netdevp'. On
1727 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1729 iface_do_create(const struct bridge *br,
1730 const struct ovsrec_interface *iface_cfg,
1731 const struct ovsrec_port *port_cfg,
1732 ofp_port_t *ofp_portp, struct netdev **netdevp,
1735 struct netdev *netdev = NULL;
1738 if (netdev_is_reserved_name(iface_cfg->name)) {
1739 VLOG_WARN("could not create interface %s, name is reserved",
1745 error = netdev_open(iface_cfg->name,
1746 iface_get_type(iface_cfg, br->cfg), &netdev);
1748 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1749 iface_cfg->name, ovs_strerror(error));
1753 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1758 *ofp_portp = iface_pick_ofport(iface_cfg);
1759 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1764 VLOG_INFO("bridge %s: added interface %s on port %d",
1765 br->name, iface_cfg->name, *ofp_portp);
1767 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1768 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1776 netdev_close(netdev);
1780 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1781 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1782 * automatically allocated for the iface. Takes ownership of and
1783 * deallocates 'if_cfg'.
1785 * Return true if an iface is successfully created, false otherwise. */
1787 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1788 const struct ovsrec_port *port_cfg)
1790 struct netdev *netdev;
1791 struct iface *iface;
1792 ofp_port_t ofp_port;
1797 /* Do the bits that can fail up front. */
1798 ovs_assert(!iface_lookup(br, iface_cfg->name));
1799 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1801 iface_clear_db_record(iface_cfg, errp);
1806 /* Get or create the port structure. */
1807 port = port_lookup(br, port_cfg->name);
1809 port = port_create(br, port_cfg);
1812 /* Create the iface structure. */
1813 iface = xzalloc(sizeof *iface);
1814 list_push_back(&port->ifaces, &iface->port_elem);
1815 hmap_insert(&br->iface_by_name, &iface->name_node,
1816 hash_string(iface_cfg->name, 0));
1818 iface->name = xstrdup(iface_cfg->name);
1819 iface->ofp_port = ofp_port;
1820 iface->netdev = netdev;
1821 iface->type = iface_get_type(iface_cfg, br->cfg);
1822 iface->cfg = iface_cfg;
1823 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1824 hash_ofp_port(ofp_port));
1826 /* Populate initial status in database. */
1827 iface_refresh_stats(iface);
1828 iface_refresh_netdev_status(iface);
1830 /* Add bond fake iface if necessary. */
1831 if (port_is_bond_fake_iface(port)) {
1832 struct ofproto_port ofproto_port;
1834 if (ofproto_port_query_by_name(br->ofproto, port->name,
1836 struct netdev *netdev;
1839 error = netdev_open(port->name, "internal", &netdev);
1841 ofp_port_t fake_ofp_port = OFPP_NONE;
1842 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1843 netdev_close(netdev);
1845 VLOG_WARN("could not open network device %s (%s)",
1846 port->name, ovs_strerror(error));
1849 /* Already exists, nothing to do. */
1850 ofproto_port_destroy(&ofproto_port);
1857 /* Set forward BPDU option. */
1859 bridge_configure_forward_bpdu(struct bridge *br)
1861 ofproto_set_forward_bpdu(br->ofproto,
1862 smap_get_bool(&br->cfg->other_config,
1867 /* Set MAC learning table configuration for 'br'. */
1869 bridge_configure_mac_table(struct bridge *br)
1871 const char *idle_time_str;
1874 const char *mac_table_size_str;
1877 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1878 idle_time = (idle_time_str && atoi(idle_time_str)
1879 ? atoi(idle_time_str)
1880 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1882 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1883 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1884 ? atoi(mac_table_size_str)
1887 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1890 /* Set multicast snooping table configuration for 'br'. */
1892 bridge_configure_mcast_snooping(struct bridge *br)
1894 if (!br->cfg->mcast_snooping_enable) {
1895 ofproto_set_mcast_snooping(br->ofproto, NULL);
1898 struct ofproto_mcast_snooping_settings br_s;
1899 const char *idle_time_str;
1900 const char *max_entries_str;
1902 idle_time_str = smap_get(&br->cfg->other_config,
1903 "mcast-snooping-aging-time");
1904 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1905 ? atoi(idle_time_str)
1906 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1908 max_entries_str = smap_get(&br->cfg->other_config,
1909 "mcast-snooping-table-size");
1910 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1911 ? atoi(max_entries_str)
1912 : MCAST_DEFAULT_MAX_ENTRIES);
1914 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1915 "mcast-snooping-disable-flood-unregistered",
1918 /* Configure multicast snooping on the bridge */
1919 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1920 VLOG_ERR("bridge %s: could not enable multicast snooping",
1925 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1926 struct ofproto_mcast_snooping_port_settings port_s;
1927 port_s.flood = smap_get_bool(&port->cfg->other_config,
1928 "mcast-snooping-flood", false);
1929 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1930 "mcast-snooping-flood-reports", false);
1931 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1932 VLOG_ERR("port %s: could not configure mcast snooping",
1940 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1941 const struct port *fake_br, struct iface **hw_addr_iface)
1943 struct hmapx mirror_output_ports;
1945 bool found_addr = false;
1949 /* Mirror output ports don't participate in picking the local hardware
1950 * address. ofproto can't help us find out whether a given port is a
1951 * mirror output because we haven't configured mirrors yet, so we need to
1952 * accumulate them ourselves. */
1953 hmapx_init(&mirror_output_ports);
1954 for (i = 0; i < br->cfg->n_mirrors; i++) {
1955 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1956 if (m->output_port) {
1957 hmapx_add(&mirror_output_ports, m->output_port);
1961 /* Otherwise choose the minimum non-local MAC address among all of the
1963 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1964 uint8_t iface_ea[ETH_ADDR_LEN];
1965 struct iface *candidate;
1966 struct iface *iface;
1968 /* Mirror output ports don't participate. */
1969 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1973 /* Choose the MAC address to represent the port. */
1975 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1976 /* Find the interface with this Ethernet address (if any) so that
1977 * we can provide the correct devname to the caller. */
1978 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1979 uint8_t candidate_ea[ETH_ADDR_LEN];
1980 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1981 && eth_addr_equals(iface_ea, candidate_ea)) {
1986 /* Choose the interface whose MAC address will represent the port.
1987 * The Linux kernel bonding code always chooses the MAC address of
1988 * the first slave added to a bond, and the Fedora networking
1989 * scripts always add slaves to a bond in alphabetical order, so
1990 * for compatibility we choose the interface with the name that is
1991 * first in alphabetical order. */
1992 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1993 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1998 /* The local port doesn't count (since we're trying to choose its
1999 * MAC address anyway). */
2000 if (iface->ofp_port == OFPP_LOCAL) {
2004 /* For fake bridges we only choose from ports with the same tag */
2005 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2006 if (!port->cfg->tag) {
2009 if (*port->cfg->tag != *fake_br->cfg->tag) {
2015 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2021 /* Compare against our current choice. */
2022 if (!eth_addr_is_multicast(iface_ea) &&
2023 !eth_addr_is_local(iface_ea) &&
2024 !eth_addr_is_reserved(iface_ea) &&
2025 !eth_addr_is_zero(iface_ea) &&
2026 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2028 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2029 *hw_addr_iface = iface;
2035 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2036 *hw_addr_iface = NULL;
2039 hmapx_destroy(&mirror_output_ports);
2043 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2044 struct iface **hw_addr_iface)
2047 *hw_addr_iface = NULL;
2049 /* Did the user request a particular MAC? */
2050 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2051 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2052 if (eth_addr_is_multicast(ea)) {
2053 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2054 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2055 } else if (eth_addr_is_zero(ea)) {
2056 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2062 /* Find a local hw address */
2063 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2066 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2067 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2068 * an interface on 'br', then that interface must be passed in as
2069 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2070 * 'hw_addr_iface' must be passed in as a null pointer. */
2072 bridge_pick_datapath_id(struct bridge *br,
2073 const uint8_t bridge_ea[ETH_ADDR_LEN],
2074 struct iface *hw_addr_iface)
2077 * The procedure for choosing a bridge MAC address will, in the most
2078 * ordinary case, also choose a unique MAC that we can use as a datapath
2079 * ID. In some special cases, though, multiple bridges will end up with
2080 * the same MAC address. This is OK for the bridges, but it will confuse
2081 * the OpenFlow controller, because each datapath needs a unique datapath
2084 * Datapath IDs must be unique. It is also very desirable that they be
2085 * stable from one run to the next, so that policy set on a datapath
2088 const char *datapath_id;
2091 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2092 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2096 if (!hw_addr_iface) {
2098 * A purely internal bridge, that is, one that has no non-virtual
2099 * network devices on it at all, is difficult because it has no
2100 * natural unique identifier at all.
2102 * When the host is a XenServer, we handle this case by hashing the
2103 * host's UUID with the name of the bridge. Names of bridges are
2104 * persistent across XenServer reboots, although they can be reused if
2105 * an internal network is destroyed and then a new one is later
2106 * created, so this is fairly effective.
2108 * When the host is not a XenServer, we punt by using a random MAC
2109 * address on each run.
2111 const char *host_uuid = xenserver_get_host_uuid();
2113 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2114 dpid = dpid_from_hash(combined, strlen(combined));
2120 return eth_addr_to_uint64(bridge_ea);
2124 dpid_from_hash(const void *data, size_t n)
2126 uint8_t hash[SHA1_DIGEST_SIZE];
2128 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2129 sha1_bytes(data, n, hash);
2130 eth_addr_mark_random(hash);
2131 return eth_addr_to_uint64(hash);
2135 iface_refresh_netdev_status(struct iface *iface)
2139 enum netdev_features current;
2140 enum netdev_flags flags;
2141 const char *link_state;
2142 uint8_t mac[ETH_ADDR_LEN];
2143 int64_t bps, mtu_64, ifindex64, link_resets;
2146 if (iface_is_synthetic(iface)) {
2150 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2151 && !status_txn_try_again) {
2155 iface->change_seq = netdev_get_change_seq(iface->netdev);
2159 if (!netdev_get_status(iface->netdev, &smap)) {
2160 ovsrec_interface_set_status(iface->cfg, &smap);
2162 ovsrec_interface_set_status(iface->cfg, NULL);
2165 smap_destroy(&smap);
2167 error = netdev_get_flags(iface->netdev, &flags);
2169 const char *state = flags & NETDEV_UP ? "up" : "down";
2171 ovsrec_interface_set_admin_state(iface->cfg, state);
2173 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2176 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2177 ovsrec_interface_set_link_state(iface->cfg, link_state);
2179 link_resets = netdev_get_carrier_resets(iface->netdev);
2180 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2182 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2183 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2185 ovsrec_interface_set_duplex(iface->cfg,
2186 netdev_features_is_full_duplex(current)
2188 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2190 ovsrec_interface_set_duplex(iface->cfg, NULL);
2191 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2194 error = netdev_get_mtu(iface->netdev, &mtu);
2197 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2199 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2202 error = netdev_get_etheraddr(iface->netdev, mac);
2204 char mac_string[32];
2206 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2207 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2209 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2212 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2213 * if there is no valid ifindex number. */
2214 ifindex64 = netdev_get_ifindex(iface->netdev);
2215 if (ifindex64 < 0) {
2218 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2222 iface_refresh_ofproto_status(struct iface *iface)
2226 if (iface_is_synthetic(iface)) {
2230 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2234 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2236 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2239 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2241 || status_txn_try_again) {
2242 iface_refresh_cfm_stats(iface);
2245 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2247 || status_txn_try_again) {
2251 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2252 iface->ofp_port, &smap);
2253 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2254 smap_destroy(&smap);
2258 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2261 iface_refresh_cfm_stats(struct iface *iface)
2263 const struct ovsrec_interface *cfg = iface->cfg;
2264 struct cfm_status status;
2267 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2268 iface->ofp_port, &status);
2270 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2271 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2272 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2273 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2274 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2275 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2277 const char *reasons[CFM_FAULT_N_REASONS];
2278 int64_t cfm_health = status.health;
2279 int64_t cfm_flap_count = status.flap_count;
2280 bool faulted = status.faults != 0;
2283 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2286 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2287 int reason = 1 << i;
2288 if (status.faults & reason) {
2289 reasons[j++] = cfm_fault_reason_to_str(reason);
2292 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2294 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2296 if (status.remote_opstate >= 0) {
2297 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2298 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2300 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2303 ovsrec_interface_set_cfm_remote_mpids(cfg,
2304 (const int64_t *)status.rmps,
2306 if (cfm_health >= 0) {
2307 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2309 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2317 iface_refresh_stats(struct iface *iface)
2319 #define IFACE_STATS \
2320 IFACE_STAT(rx_packets, "rx_packets") \
2321 IFACE_STAT(tx_packets, "tx_packets") \
2322 IFACE_STAT(rx_bytes, "rx_bytes") \
2323 IFACE_STAT(tx_bytes, "tx_bytes") \
2324 IFACE_STAT(rx_dropped, "rx_dropped") \
2325 IFACE_STAT(tx_dropped, "tx_dropped") \
2326 IFACE_STAT(rx_errors, "rx_errors") \
2327 IFACE_STAT(tx_errors, "tx_errors") \
2328 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2329 IFACE_STAT(rx_over_errors, "rx_over_err") \
2330 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2331 IFACE_STAT(collisions, "collisions")
2333 #define IFACE_STAT(MEMBER, NAME) + 1
2334 enum { N_IFACE_STATS = IFACE_STATS };
2336 int64_t values[N_IFACE_STATS];
2337 const char *keys[N_IFACE_STATS];
2340 struct netdev_stats stats;
2342 if (iface_is_synthetic(iface)) {
2346 /* Intentionally ignore return value, since errors will set 'stats' to
2347 * all-1s, and we will deal with that correctly below. */
2348 netdev_get_stats(iface->netdev, &stats);
2350 /* Copy statistics into keys[] and values[]. */
2352 #define IFACE_STAT(MEMBER, NAME) \
2353 if (stats.MEMBER != UINT64_MAX) { \
2355 values[n] = stats.MEMBER; \
2360 ovs_assert(n <= N_IFACE_STATS);
2362 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2367 br_refresh_datapath_info(struct bridge *br)
2369 const char *version;
2371 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2372 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2375 ovsrec_bridge_set_datapath_version(br->cfg,
2376 version ? version : "<unknown>");
2380 br_refresh_stp_status(struct bridge *br)
2382 struct smap smap = SMAP_INITIALIZER(&smap);
2383 struct ofproto *ofproto = br->ofproto;
2384 struct ofproto_stp_status status;
2386 if (ofproto_get_stp_status(ofproto, &status)) {
2390 if (!status.enabled) {
2391 ovsrec_bridge_set_status(br->cfg, NULL);
2395 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2396 STP_ID_ARGS(status.bridge_id));
2397 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2398 STP_ID_ARGS(status.designated_root));
2399 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2401 ovsrec_bridge_set_status(br->cfg, &smap);
2402 smap_destroy(&smap);
2406 port_refresh_stp_status(struct port *port)
2408 struct ofproto *ofproto = port->bridge->ofproto;
2409 struct iface *iface;
2410 struct ofproto_port_stp_status status;
2413 if (port_is_synthetic(port)) {
2417 /* STP doesn't currently support bonds. */
2418 if (!list_is_singleton(&port->ifaces)) {
2419 ovsrec_port_set_status(port->cfg, NULL);
2423 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2424 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2428 if (!status.enabled) {
2429 ovsrec_port_set_status(port->cfg, NULL);
2433 /* Set Status column. */
2435 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2436 smap_add(&smap, "stp_state", stp_state_name(status.state));
2437 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2438 smap_add(&smap, "stp_role", stp_role_name(status.role));
2439 ovsrec_port_set_status(port->cfg, &smap);
2440 smap_destroy(&smap);
2444 port_refresh_stp_stats(struct port *port)
2446 struct ofproto *ofproto = port->bridge->ofproto;
2447 struct iface *iface;
2448 struct ofproto_port_stp_stats stats;
2449 const char *keys[3];
2450 int64_t int_values[3];
2452 if (port_is_synthetic(port)) {
2456 /* STP doesn't currently support bonds. */
2457 if (!list_is_singleton(&port->ifaces)) {
2461 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2462 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2466 if (!stats.enabled) {
2467 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2471 /* Set Statistics column. */
2472 keys[0] = "stp_tx_count";
2473 int_values[0] = stats.tx_count;
2474 keys[1] = "stp_rx_count";
2475 int_values[1] = stats.rx_count;
2476 keys[2] = "stp_error_count";
2477 int_values[2] = stats.error_count;
2479 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2480 ARRAY_SIZE(int_values));
2484 br_refresh_rstp_status(struct bridge *br)
2486 struct smap smap = SMAP_INITIALIZER(&smap);
2487 struct ofproto *ofproto = br->ofproto;
2488 struct ofproto_rstp_status status;
2490 if (ofproto_get_rstp_status(ofproto, &status)) {
2493 if (!status.enabled) {
2494 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2497 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2498 RSTP_ID_ARGS(status.bridge_id));
2499 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2500 status.root_path_cost);
2501 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2502 RSTP_ID_ARGS(status.root_id));
2503 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2504 RSTP_ID_ARGS(status.designated_id));
2505 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2506 status.designated_port_id);
2507 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2508 status.bridge_port_id);
2509 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2510 smap_destroy(&smap);
2514 port_refresh_rstp_status(struct port *port)
2516 struct ofproto *ofproto = port->bridge->ofproto;
2517 struct iface *iface;
2518 struct ofproto_port_rstp_status status;
2519 const char *keys[4];
2520 int64_t int_values[4];
2523 if (port_is_synthetic(port)) {
2527 /* RSTP doesn't currently support bonds. */
2528 if (!list_is_singleton(&port->ifaces)) {
2529 ovsrec_port_set_rstp_status(port->cfg, NULL);
2533 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2534 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2538 if (!status.enabled) {
2539 ovsrec_port_set_rstp_status(port->cfg, NULL);
2540 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2543 /* Set Status column. */
2546 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2548 smap_add_format(&smap, "rstp_port_role", "%s",
2549 rstp_port_role_name(status.role));
2550 smap_add_format(&smap, "rstp_port_state", "%s",
2551 rstp_state_name(status.state));
2552 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2553 RSTP_ID_ARGS(status.designated_bridge_id));
2554 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2555 status.designated_port_id);
2556 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2557 status.designated_path_cost);
2559 ovsrec_port_set_rstp_status(port->cfg, &smap);
2560 smap_destroy(&smap);
2562 /* Set Statistics column. */
2563 keys[0] = "rstp_tx_count";
2564 int_values[0] = status.tx_count;
2565 keys[1] = "rstp_rx_count";
2566 int_values[1] = status.rx_count;
2567 keys[2] = "rstp_uptime";
2568 int_values[2] = status.uptime;
2569 keys[3] = "rstp_error_count";
2570 int_values[3] = status.error_count;
2571 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2572 ARRAY_SIZE(int_values));
2576 port_refresh_bond_status(struct port *port, bool force_update)
2578 uint8_t mac[ETH_ADDR_LEN];
2580 /* Return if port is not a bond */
2581 if (list_is_singleton(&port->ifaces)) {
2585 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2589 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2590 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2596 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2598 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2602 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2604 bool enable = enable_system_stats(cfg);
2606 system_stats_enable(enable);
2608 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2613 run_system_stats(void)
2615 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2618 stats = system_stats_run();
2620 struct ovsdb_idl_txn *txn;
2621 struct ovsdb_datum datum;
2623 txn = ovsdb_idl_txn_create(idl);
2624 ovsdb_datum_from_smap(&datum, stats);
2625 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2627 ovsdb_idl_txn_commit(txn);
2628 ovsdb_idl_txn_destroy(txn);
2635 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2638 case OFPCR12_ROLE_EQUAL:
2640 case OFPCR12_ROLE_MASTER:
2642 case OFPCR12_ROLE_SLAVE:
2644 case OFPCR12_ROLE_NOCHANGE:
2646 return "*** INVALID ROLE ***";
2651 refresh_controller_status(void)
2655 const struct ovsrec_controller *cfg;
2659 /* Accumulate status for controllers on all bridges. */
2660 HMAP_FOR_EACH (br, node, &all_bridges) {
2661 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2664 /* Update each controller in the database with current status. */
2665 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2666 struct ofproto_controller_info *cinfo =
2667 shash_find_data(&info, cfg->target);
2670 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2671 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2673 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2675 ovsrec_controller_set_is_connected(cfg, false);
2676 ovsrec_controller_set_role(cfg, NULL);
2677 ovsrec_controller_set_status(cfg, NULL);
2681 ofproto_free_ofproto_controller_info(&info);
2684 /* Update interface and mirror statistics if necessary. */
2686 run_stats_update(void)
2688 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2695 /* Statistics update interval should always be greater than or equal to
2697 stats_interval = MAX(smap_get_int(&cfg->other_config,
2698 "stats-update-interval",
2700 if (stats_timer_interval != stats_interval) {
2701 stats_timer_interval = stats_interval;
2702 stats_timer = LLONG_MIN;
2705 if (time_msec() >= stats_timer) {
2706 enum ovsdb_idl_txn_status status;
2708 /* Rate limit the update. Do not start a new update if the
2709 * previous one is not done. */
2713 stats_txn = ovsdb_idl_txn_create(idl);
2714 HMAP_FOR_EACH (br, node, &all_bridges) {
2718 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2719 struct iface *iface;
2721 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2722 iface_refresh_stats(iface);
2724 port_refresh_stp_stats(port);
2726 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2727 mirror_refresh_stats(m);
2730 refresh_controller_status();
2733 status = ovsdb_idl_txn_commit(stats_txn);
2734 if (status != TXN_INCOMPLETE) {
2735 stats_timer = time_msec() + stats_timer_interval;
2736 ovsdb_idl_txn_destroy(stats_txn);
2743 stats_update_wait(void)
2745 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2746 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2748 ovsdb_idl_txn_wait(stats_txn);
2750 poll_timer_wait_until(stats_timer);
2754 /* Update bridge/port/interface status if necessary. */
2756 run_status_update(void)
2761 /* Rate limit the update. Do not start a new update if the
2762 * previous one is not done. */
2763 seq = seq_read(connectivity_seq_get());
2764 if (seq != connectivity_seqno || status_txn_try_again) {
2767 connectivity_seqno = seq;
2768 status_txn = ovsdb_idl_txn_create(idl);
2769 HMAP_FOR_EACH (br, node, &all_bridges) {
2772 br_refresh_stp_status(br);
2773 br_refresh_rstp_status(br);
2774 br_refresh_datapath_info(br);
2775 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2776 struct iface *iface;
2778 port_refresh_stp_status(port);
2779 port_refresh_rstp_status(port);
2780 port_refresh_bond_status(port, status_txn_try_again);
2781 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2782 iface_refresh_netdev_status(iface);
2783 iface_refresh_ofproto_status(iface);
2790 /* Commit the transaction and get the status. If the transaction finishes,
2791 * then destroy the transaction. Otherwise, keep it so that we can check
2792 * progress the next time that this function is called. */
2794 enum ovsdb_idl_txn_status status;
2796 status = ovsdb_idl_txn_commit(status_txn);
2797 if (status != TXN_INCOMPLETE) {
2798 ovsdb_idl_txn_destroy(status_txn);
2801 /* Sets the 'status_txn_try_again' if the transaction fails. */
2802 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2803 status_txn_try_again = false;
2805 status_txn_try_again = true;
2810 /* Refresh AA port status if necessary. */
2811 if (time_msec() >= aa_refresh_timer) {
2814 HMAP_FOR_EACH (br, node, &all_bridges) {
2815 if (bridge_aa_need_refresh(br)) {
2816 struct ovsdb_idl_txn *txn;
2818 txn = ovsdb_idl_txn_create(idl);
2819 bridge_aa_refresh_queued(br);
2820 ovsdb_idl_txn_commit(txn);
2821 ovsdb_idl_txn_destroy(txn);
2825 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2830 status_update_wait(void)
2832 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2833 * transaction to complete. If the status update to database needs to be
2834 * run again (transaction fails), registers a timeout in
2835 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2836 * sequence number. */
2838 ovsdb_idl_txn_wait(status_txn);
2839 } else if (status_txn_try_again) {
2840 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2842 seq_wait(connectivity_seq_get(), connectivity_seqno);
2853 /* Let each datapath type do the work that it needs to do. */
2855 ofproto_enumerate_types(&types);
2856 SSET_FOR_EACH (type, &types) {
2857 ofproto_type_run(type);
2859 sset_destroy(&types);
2861 /* Let each bridge do the work that it needs to do. */
2862 HMAP_FOR_EACH (br, node, &all_bridges) {
2863 ofproto_run(br->ofproto);
2870 static struct ovsrec_open_vswitch null_cfg;
2871 const struct ovsrec_open_vswitch *cfg;
2873 bool vlan_splinters_changed;
2875 ovsrec_open_vswitch_init(&null_cfg);
2879 if (ovsdb_idl_is_lock_contended(idl)) {
2880 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2881 struct bridge *br, *next_br;
2883 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2884 "disabling this process (pid %ld) until it goes away",
2885 (long int) getpid());
2887 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2890 /* Since we will not be running system_stats_run() in this process
2891 * with the current situation of multiple ovs-vswitchd daemons,
2892 * disable system stats collection. */
2893 system_stats_enable(false);
2895 } else if (!ovsdb_idl_has_lock(idl)
2896 || !ovsdb_idl_has_ever_connected(idl)) {
2897 /* Returns if not holding the lock or not done retrieving db
2901 cfg = ovsrec_open_vswitch_first(idl);
2903 /* Initialize the ofproto library. This only needs to run once, but
2904 * it must be done after the configuration is set. If the
2905 * initialization has already occurred, bridge_init_ofproto()
2906 * returns immediately. */
2907 bridge_init_ofproto(cfg);
2909 /* Once the value of flow-restore-wait is false, we no longer should
2910 * check its value from the database. */
2911 if (cfg && ofproto_get_flow_restore_wait()) {
2912 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2913 "flow-restore-wait", false));
2918 /* Re-configure SSL. We do this on every trip through the main loop,
2919 * instead of just when the database changes, because the contents of the
2920 * key and certificate files can change without the database changing.
2922 * We do this before bridge_reconfigure() because that function might
2923 * initiate SSL connections and thus requires SSL to be configured. */
2924 if (cfg && cfg->ssl) {
2925 const struct ovsrec_ssl *ssl = cfg->ssl;
2927 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2928 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2931 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2932 * usage has changed. */
2933 vlan_splinters_changed = false;
2934 if (vlan_splinters_enabled_anywhere) {
2937 HMAP_FOR_EACH (br, node, &all_bridges) {
2938 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2939 vlan_splinters_changed = true;
2945 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2946 struct ovsdb_idl_txn *txn;
2948 idl_seqno = ovsdb_idl_get_seqno(idl);
2949 txn = ovsdb_idl_txn_create(idl);
2950 bridge_reconfigure(cfg ? cfg : &null_cfg);
2953 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2954 discover_types(cfg);
2957 /* If we are completing our initial configuration for this run
2958 * of ovs-vswitchd, then keep the transaction around to monitor
2959 * it for completion. */
2960 if (initial_config_done) {
2961 /* Always sets the 'status_txn_try_again' to check again,
2962 * in case that this transaction fails. */
2963 status_txn_try_again = true;
2964 ovsdb_idl_txn_commit(txn);
2965 ovsdb_idl_txn_destroy(txn);
2967 initial_config_done = true;
2968 daemonize_txn = txn;
2972 if (daemonize_txn) {
2973 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2974 if (status != TXN_INCOMPLETE) {
2975 ovsdb_idl_txn_destroy(daemonize_txn);
2976 daemonize_txn = NULL;
2978 /* ovs-vswitchd has completed initialization, so allow the
2979 * process that forked us to exit successfully. */
2980 daemonize_complete();
2982 vlog_enable_async();
2984 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2989 run_status_update();
2999 ovsdb_idl_wait(idl);
3000 if (daemonize_txn) {
3001 ovsdb_idl_txn_wait(daemonize_txn);
3005 ofproto_enumerate_types(&types);
3006 SSET_FOR_EACH (type, &types) {
3007 ofproto_type_wait(type);
3009 sset_destroy(&types);
3011 if (!hmap_is_empty(&all_bridges)) {
3014 HMAP_FOR_EACH (br, node, &all_bridges) {
3015 ofproto_wait(br->ofproto);
3017 stats_update_wait();
3018 status_update_wait();
3021 system_stats_wait();
3024 /* Adds some memory usage statistics for bridges into 'usage', for use with
3025 * memory_report(). */
3027 bridge_get_memory_usage(struct simap *usage)
3034 ofproto_enumerate_types(&types);
3035 SSET_FOR_EACH (type, &types) {
3036 ofproto_type_get_memory_usage(type, usage);
3038 sset_destroy(&types);
3040 HMAP_FOR_EACH (br, node, &all_bridges) {
3041 ofproto_get_memory_usage(br->ofproto, usage);
3045 /* QoS unixctl user interface functions. */
3047 struct qos_unixctl_show_cbdata {
3049 struct iface *iface;
3053 qos_unixctl_show_queue(unsigned int queue_id,
3054 const struct smap *details,
3055 struct iface *iface,
3058 struct netdev_queue_stats stats;
3059 struct smap_node *node;
3062 ds_put_cstr(ds, "\n");
3064 ds_put_format(ds, "Queue %u:\n", queue_id);
3066 ds_put_cstr(ds, "Default:\n");
3069 SMAP_FOR_EACH (node, details) {
3070 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3073 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3075 if (stats.tx_packets != UINT64_MAX) {
3076 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3079 if (stats.tx_bytes != UINT64_MAX) {
3080 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3083 if (stats.tx_errors != UINT64_MAX) {
3084 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3087 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3088 queue_id, ovs_strerror(error));
3093 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3094 const char *argv[], void *aux OVS_UNUSED)
3096 struct ds ds = DS_EMPTY_INITIALIZER;
3097 struct smap smap = SMAP_INITIALIZER(&smap);
3098 struct iface *iface;
3100 struct smap_node *node;
3102 iface = iface_find(argv[1]);
3104 unixctl_command_reply_error(conn, "no such interface");
3108 netdev_get_qos(iface->netdev, &type, &smap);
3110 if (*type != '\0') {
3111 struct netdev_queue_dump dump;
3112 struct smap details;
3113 unsigned int queue_id;
3115 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3117 SMAP_FOR_EACH (node, &smap) {
3118 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3121 smap_init(&details);
3122 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3123 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3125 smap_destroy(&details);
3127 unixctl_command_reply(conn, ds_cstr(&ds));
3129 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3130 unixctl_command_reply_error(conn, ds_cstr(&ds));
3133 smap_destroy(&smap);
3137 /* Bridge reconfiguration functions. */
3139 bridge_create(const struct ovsrec_bridge *br_cfg)
3143 ovs_assert(!bridge_lookup(br_cfg->name));
3144 br = xzalloc(sizeof *br);
3146 br->name = xstrdup(br_cfg->name);
3147 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3150 /* Derive the default Ethernet address from the bridge's UUID. This should
3151 * be unique and it will be stable between ovs-vswitchd runs. */
3152 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3153 eth_addr_mark_random(br->default_ea);
3155 hmap_init(&br->ports);
3156 hmap_init(&br->ifaces);
3157 hmap_init(&br->iface_by_name);
3158 hmap_init(&br->mirrors);
3160 hmap_init(&br->mappings);
3161 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3165 bridge_destroy(struct bridge *br)
3168 struct mirror *mirror, *next_mirror;
3169 struct port *port, *next_port;
3171 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3174 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3175 mirror_destroy(mirror);
3178 hmap_remove(&all_bridges, &br->node);
3179 ofproto_destroy(br->ofproto);
3180 hmap_destroy(&br->ifaces);
3181 hmap_destroy(&br->ports);
3182 hmap_destroy(&br->iface_by_name);
3183 hmap_destroy(&br->mirrors);
3184 hmap_destroy(&br->mappings);
3191 static struct bridge *
3192 bridge_lookup(const char *name)
3196 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3197 if (!strcmp(br->name, name)) {
3204 /* Handle requests for a listing of all flows known by the OpenFlow
3205 * stack, including those normally hidden. */
3207 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3208 const char *argv[], void *aux OVS_UNUSED)
3213 br = bridge_lookup(argv[1]);
3215 unixctl_command_reply_error(conn, "Unknown bridge");
3220 ofproto_get_all_flows(br->ofproto, &results);
3222 unixctl_command_reply(conn, ds_cstr(&results));
3223 ds_destroy(&results);
3226 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3227 * connections and reconnect. If BRIDGE is not specified, then all bridges
3228 * drop their controller connections and reconnect. */
3230 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3231 const char *argv[], void *aux OVS_UNUSED)
3235 br = bridge_lookup(argv[1]);
3237 unixctl_command_reply_error(conn, "Unknown bridge");
3240 ofproto_reconnect_controllers(br->ofproto);
3242 HMAP_FOR_EACH (br, node, &all_bridges) {
3243 ofproto_reconnect_controllers(br->ofproto);
3246 unixctl_command_reply(conn, NULL);
3250 bridge_get_controllers(const struct bridge *br,
3251 struct ovsrec_controller ***controllersp)
3253 struct ovsrec_controller **controllers;
3254 size_t n_controllers;
3256 controllers = br->cfg->controller;
3257 n_controllers = br->cfg->n_controller;
3259 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3265 *controllersp = controllers;
3267 return n_controllers;
3271 bridge_collect_wanted_ports(struct bridge *br,
3272 const unsigned long int *splinter_vlans,
3273 struct shash *wanted_ports)
3277 shash_init(wanted_ports);
3279 for (i = 0; i < br->cfg->n_ports; i++) {
3280 const char *name = br->cfg->ports[i]->name;
3281 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3282 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3286 if (bridge_get_controllers(br, NULL)
3287 && !shash_find(wanted_ports, br->name)) {
3288 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3289 br->name, br->name);
3291 ovsrec_interface_init(&br->synth_local_iface);
3292 ovsrec_port_init(&br->synth_local_port);
3294 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3295 br->synth_local_port.n_interfaces = 1;
3296 br->synth_local_port.name = br->name;
3298 br->synth_local_iface.name = br->name;
3299 br->synth_local_iface.type = "internal";
3301 br->synth_local_ifacep = &br->synth_local_iface;
3303 shash_add(wanted_ports, br->name, &br->synth_local_port);
3306 if (splinter_vlans) {
3307 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3311 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3312 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3313 * 'br' needs to complete its configuration. */
3315 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3317 struct shash_node *port_node;
3318 struct port *port, *next;
3320 /* Get rid of deleted ports.
3321 * Get rid of deleted interfaces on ports that still exist. */
3322 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3323 port->cfg = shash_find_data(wanted_ports, port->name);
3327 port_del_ifaces(port);
3331 /* Update iface->cfg and iface->type in interfaces that still exist. */
3332 SHASH_FOR_EACH (port_node, wanted_ports) {
3333 const struct ovsrec_port *port = port_node->data;
3336 for (i = 0; i < port->n_interfaces; i++) {
3337 const struct ovsrec_interface *cfg = port->interfaces[i];
3338 struct iface *iface = iface_lookup(br, cfg->name);
3339 const char *type = iface_get_type(cfg, br->cfg);
3344 } else if (!strcmp(type, "null")) {
3345 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3346 " may be removed in February 2013. Please email"
3347 " dev@openvswitch.org with concerns.",
3350 /* We will add new interfaces later. */
3356 /* Initializes 'oc' appropriately as a management service controller for
3359 * The caller must free oc->target when it is no longer needed. */
3361 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3362 struct ofproto_controller *oc)
3364 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3365 oc->max_backoff = 0;
3366 oc->probe_interval = 60;
3367 oc->band = OFPROTO_OUT_OF_BAND;
3369 oc->burst_limit = 0;
3370 oc->enable_async_msgs = true;
3374 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3376 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3377 struct ofproto_controller *oc)
3381 oc->target = c->target;
3382 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3383 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3384 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3385 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3386 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3387 oc->burst_limit = (c->controller_burst_limit
3388 ? *c->controller_burst_limit : 0);
3389 oc->enable_async_msgs = (!c->enable_async_messages
3390 || *c->enable_async_messages);
3391 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3392 if (dscp < 0 || dscp > 63) {
3393 dscp = DSCP_DEFAULT;
3398 /* Configures the IP stack for 'br''s local interface properly according to the
3399 * configuration in 'c'. */
3401 bridge_configure_local_iface_netdev(struct bridge *br,
3402 struct ovsrec_controller *c)
3404 struct netdev *netdev;
3405 struct in_addr mask, gateway;
3407 struct iface *local_iface;
3410 /* If there's no local interface or no IP address, give up. */
3411 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3412 if (!local_iface || !c->local_ip
3413 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3417 /* Bring up the local interface. */
3418 netdev = local_iface->netdev;
3419 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3421 /* Configure the IP address and netmask. */
3422 if (!c->local_netmask
3423 || !inet_pton(AF_INET, c->local_netmask, &mask)
3425 mask.s_addr = guess_netmask(ip.s_addr);
3427 if (!netdev_set_in4(netdev, ip, mask)) {
3428 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3429 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3432 /* Configure the default gateway. */
3433 if (c->local_gateway
3434 && inet_pton(AF_INET, c->local_gateway, &gateway)
3435 && gateway.s_addr) {
3436 if (!netdev_add_router(netdev, gateway)) {
3437 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3438 br->name, IP_ARGS(gateway.s_addr));
3443 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3444 * in either string are treated as equal to any number of slashes in the other,
3445 * e.g. "x///y" is equal to "x/y".
3447 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3448 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3449 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3450 * 'b' against a prefix of 'a'.
3453 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3455 const char *b_start = b;
3457 if (b - b_start >= b_stoplen) {
3459 } else if (*a != *b) {
3461 } else if (*a == '/') {
3462 a += strspn(a, "/");
3463 b += strspn(b, "/");
3464 } else if (*a == '\0') {
3474 bridge_configure_remotes(struct bridge *br,
3475 const struct sockaddr_in *managers, size_t n_managers)
3477 bool disable_in_band;
3479 struct ovsrec_controller **controllers;
3480 size_t n_controllers;
3482 enum ofproto_fail_mode fail_mode;
3484 struct ofproto_controller *ocs;
3488 /* Check if we should disable in-band control on this bridge. */
3489 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3492 /* Set OpenFlow queue ID for in-band control. */
3493 ofproto_set_in_band_queue(br->ofproto,
3494 smap_get_int(&br->cfg->other_config,
3495 "in-band-queue", -1));
3497 if (disable_in_band) {
3498 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3500 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3503 n_controllers = bridge_get_controllers(br, &controllers);
3505 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3508 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3509 for (i = 0; i < n_controllers; i++) {
3510 struct ovsrec_controller *c = controllers[i];
3512 if (!strncmp(c->target, "punix:", 6)
3513 || !strncmp(c->target, "unix:", 5)) {
3514 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3517 if (!strncmp(c->target, "unix:", 5)) {
3518 /* Connect to a listening socket */
3519 whitelist = xasprintf("unix:%s/", ovs_rundir());
3520 if (strchr(c->target, '/') &&
3521 !equal_pathnames(c->target, whitelist,
3522 strlen(whitelist))) {
3523 /* Absolute path specified, but not in ovs_rundir */
3524 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3525 "controller \"%s\" due to possibility for "
3526 "remote exploit. Instead, specify socket "
3527 "in whitelisted \"%s\" or connect to "
3528 "\"unix:%s/%s.mgmt\" (which is always "
3529 "available without special configuration).",
3530 br->name, c->target, whitelist,
3531 ovs_rundir(), br->name);
3536 whitelist = xasprintf("punix:%s/%s.controller",
3537 ovs_rundir(), br->name);
3538 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3539 /* Prevent remote ovsdb-server users from accessing
3540 * arbitrary Unix domain sockets and overwriting arbitrary
3542 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3543 "controller \"%s\" due to possibility of "
3544 "overwriting local files. Instead, specify "
3545 "whitelisted \"%s\" or connect to "
3546 "\"unix:%s/%s.mgmt\" (which is always "
3547 "available without special configuration).",
3548 br->name, c->target, whitelist,
3549 ovs_rundir(), br->name);
3558 bridge_configure_local_iface_netdev(br, c);
3559 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3560 if (disable_in_band) {
3561 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3566 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3567 bridge_get_allowed_versions(br));
3568 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3571 /* Set the fail-mode. */
3572 fail_mode = !br->cfg->fail_mode
3573 || !strcmp(br->cfg->fail_mode, "standalone")
3574 ? OFPROTO_FAIL_STANDALONE
3575 : OFPROTO_FAIL_SECURE;
3576 ofproto_set_fail_mode(br->ofproto, fail_mode);
3578 /* Configure OpenFlow controller connection snooping. */
3579 if (!ofproto_has_snoops(br->ofproto)) {
3583 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3584 ovs_rundir(), br->name));
3585 ofproto_set_snoops(br->ofproto, &snoops);
3586 sset_destroy(&snoops);
3591 bridge_configure_tables(struct bridge *br)
3593 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3597 n_tables = ofproto_get_n_tables(br->ofproto);
3599 for (i = 0; i < n_tables; i++) {
3600 struct ofproto_table_settings s;
3601 bool use_default_prefixes = true;
3604 s.max_flows = UINT_MAX;
3607 s.n_prefix_fields = 0;
3608 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3610 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3611 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3614 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3615 s.max_flows = *cfg->flow_limit;
3617 if (cfg->overflow_policy
3618 && !strcmp(cfg->overflow_policy, "evict")) {
3620 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3621 for (k = 0; k < cfg->n_groups; k++) {
3622 const char *string = cfg->groups[k];
3625 msg = mf_parse_subfield__(&s.groups[k], &string);
3627 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3628 "'groups' (%s)", br->name, i, msg);
3630 } else if (*string) {
3631 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3632 "element '%s' contains trailing garbage",
3633 br->name, i, cfg->groups[k]);
3639 /* Prefix lookup fields. */
3640 s.n_prefix_fields = 0;
3641 for (k = 0; k < cfg->n_prefixes; k++) {
3642 const char *name = cfg->prefixes[k];
3643 const struct mf_field *mf;
3645 if (strcmp(name, "none") == 0) {
3646 use_default_prefixes = false;
3647 s.n_prefix_fields = 0;
3650 mf = mf_from_name(name);
3652 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3656 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3657 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3658 "%s", br->name, name);
3661 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3662 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3663 "field not used: %s", br->name, name);
3666 use_default_prefixes = false;
3667 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3670 if (use_default_prefixes) {
3671 /* Use default values. */
3672 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3673 memcpy(s.prefix_fields, default_prefix_fields,
3674 sizeof default_prefix_fields);
3677 struct ds ds = DS_EMPTY_INITIALIZER;
3678 for (k = 0; k < s.n_prefix_fields; k++) {
3680 ds_put_char(&ds, ',');
3682 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3684 if (s.n_prefix_fields == 0) {
3685 ds_put_cstr(&ds, "none");
3687 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3688 br->name, i, ds_cstr(&ds));
3692 ofproto_configure_table(br->ofproto, i, &s);
3696 for (; j < br->cfg->n_flow_tables; j++) {
3697 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3698 "%"PRId64" not supported by this datapath", br->name,
3699 br->cfg->key_flow_tables[j]);
3704 bridge_configure_dp_desc(struct bridge *br)
3706 ofproto_set_dp_desc(br->ofproto,
3707 smap_get(&br->cfg->other_config, "dp-desc"));
3710 static struct aa_mapping *
3711 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3713 struct aa_mapping *m;
3715 HMAP_FOR_EACH_IN_BUCKET (m,
3717 hash_bytes(&isid, sizeof isid, 0),
3719 if (isid == m->isid) {
3726 static struct aa_mapping *
3727 bridge_aa_mapping_create(struct bridge *br,
3731 struct aa_mapping *m;
3733 m = xzalloc(sizeof *m);
3737 m->br_name = xstrdup(br->name);
3738 hmap_insert(&br->mappings,
3740 hash_bytes(&isid, sizeof isid, 0));
3746 bridge_aa_mapping_destroy(struct aa_mapping *m)
3749 struct bridge *br = m->bridge;
3752 ofproto_aa_mapping_unregister(br->ofproto, m);
3755 hmap_remove(&br->mappings, &m->hmap_node);
3764 bridge_aa_mapping_configure(struct aa_mapping *m)
3766 struct aa_mapping_settings s;
3772 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3778 bridge_configure_aa(struct bridge *br)
3780 const struct ovsdb_datum *mc;
3781 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3782 struct aa_settings aa_s;
3783 struct aa_mapping *m, *next;
3787 ofproto_set_aa(br->ofproto, NULL, NULL);
3791 memset(&aa_s, 0, sizeof aa_s);
3792 aa_s.system_description = auto_attach->system_description;
3793 aa_s.system_name = auto_attach->system_name;
3794 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3796 mc = ovsrec_autoattach_get_mappings(auto_attach,
3798 OVSDB_TYPE_INTEGER);
3799 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3800 union ovsdb_atom atom;
3802 atom.integer = m->isid;
3803 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3804 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3806 bridge_aa_mapping_destroy(m);
3810 /* Add new mappings and reconfigure existing ones. */
3811 for (i = 0; i < auto_attach->n_mappings; ++i) {
3812 struct aa_mapping *m =
3813 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3816 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3817 auto_attach->key_mappings[i],
3818 auto_attach->value_mappings[i]);
3819 m = bridge_aa_mapping_create(br,
3820 auto_attach->key_mappings[i],
3821 auto_attach->value_mappings[i]);
3823 if (!bridge_aa_mapping_configure(m)) {
3824 bridge_aa_mapping_destroy(m);
3831 bridge_aa_need_refresh(struct bridge *br)
3833 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3837 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3839 int64_t *trunks = NULL;
3841 bool found = false, reconfigure = false;
3843 for (i = 0; i < port->cfg->n_trunks; i++) {
3844 if (port->cfg->trunks[i] == m->vlan) {
3851 case BRIDGE_AA_VLAN_OPER_ADD:
3853 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3855 for (i = 0; i < port->cfg->n_trunks; i++) {
3856 trunks[i] = port->cfg->trunks[i];
3858 trunks[i++] = m->vlan;
3864 case BRIDGE_AA_VLAN_OPER_REMOVE:
3868 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3870 for (i = 0; i < port->cfg->n_trunks; i++) {
3871 if (port->cfg->trunks[i] != m->vlan) {
3872 trunks[j++] = port->cfg->trunks[i];
3881 case BRIDGE_AA_VLAN_OPER_UNDEF:
3883 VLOG_WARN("unrecognized operation %u", m->oper);
3888 /* VLAN switching under trunk mode cause the trunk port to switch all
3889 * VLANs, see ovs-vswitchd.conf.db
3892 static char *vlan_mode_access = "access";
3893 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3897 static char *vlan_mode_trunk = "trunk";
3898 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3901 ovsrec_port_set_trunks(port->cfg, trunks, i);
3903 /* Force reconfigure of the port. */
3904 port_configure(port);
3909 bridge_aa_refresh_queued(struct bridge *br)
3911 struct ovs_list *list = xmalloc(sizeof *list);
3912 struct bridge_aa_vlan *node, *next;
3915 ofproto_aa_vlan_get_queued(br->ofproto, list);
3917 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
3920 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3923 port = port_lookup(br, node->port_name);
3925 bridge_aa_update_trunks(port, node);
3928 list_remove(&node->list_node);
3929 free(node->port_name);
3937 /* Port functions. */
3939 static struct port *
3940 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3944 port = xzalloc(sizeof *port);
3946 port->name = xstrdup(cfg->name);
3948 list_init(&port->ifaces);
3950 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3954 /* Deletes interfaces from 'port' that are no longer configured for it. */
3956 port_del_ifaces(struct port *port)
3958 struct iface *iface, *next;
3959 struct sset new_ifaces;
3962 /* Collect list of new interfaces. */
3963 sset_init(&new_ifaces);
3964 for (i = 0; i < port->cfg->n_interfaces; i++) {
3965 const char *name = port->cfg->interfaces[i]->name;
3966 const char *type = port->cfg->interfaces[i]->type;
3967 if (strcmp(type, "null")) {
3968 sset_add(&new_ifaces, name);
3972 /* Get rid of deleted interfaces. */
3973 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3974 if (!sset_contains(&new_ifaces, iface->name)) {
3975 iface_destroy(iface);
3979 sset_destroy(&new_ifaces);
3983 port_destroy(struct port *port)
3986 struct bridge *br = port->bridge;
3987 struct iface *iface, *next;
3990 ofproto_bundle_unregister(br->ofproto, port);
3993 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3994 iface_destroy__(iface);
3997 hmap_remove(&br->ports, &port->hmap_node);
4003 static struct port *
4004 port_lookup(const struct bridge *br, const char *name)
4008 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4010 if (!strcmp(port->name, name)) {
4018 enable_lacp(struct port *port, bool *activep)
4020 if (!port->cfg->lacp) {
4021 /* XXX when LACP implementation has been sufficiently tested, enable by
4022 * default and make active on bonded ports. */
4024 } else if (!strcmp(port->cfg->lacp, "off")) {
4026 } else if (!strcmp(port->cfg->lacp, "active")) {
4029 } else if (!strcmp(port->cfg->lacp, "passive")) {
4033 VLOG_WARN("port %s: unknown LACP mode %s",
4034 port->name, port->cfg->lacp);
4039 static struct lacp_settings *
4040 port_configure_lacp(struct port *port, struct lacp_settings *s)
4042 const char *lacp_time, *system_id;
4045 if (!enable_lacp(port, &s->active)) {
4049 s->name = port->name;
4051 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4053 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4054 ETH_ADDR_SCAN_ARGS(s->id))) {
4055 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4056 " address.", port->name, system_id);
4060 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4063 if (eth_addr_is_zero(s->id)) {
4064 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4068 /* Prefer bondable links if unspecified. */
4069 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4071 s->priority = (priority > 0 && priority <= UINT16_MAX
4073 : UINT16_MAX - !list_is_short(&port->ifaces));
4075 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4076 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4078 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4079 "lacp-fallback-ab", false);
4085 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4087 int priority, portid, key;
4089 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4090 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4092 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4094 if (portid <= 0 || portid > UINT16_MAX) {
4095 portid = ofp_to_u16(iface->ofp_port);
4098 if (priority <= 0 || priority > UINT16_MAX) {
4099 priority = UINT16_MAX;
4102 if (key < 0 || key > UINT16_MAX) {
4106 s->name = iface->name;
4108 s->priority = priority;
4113 port_configure_bond(struct port *port, struct bond_settings *s)
4115 const char *detect_s;
4116 struct iface *iface;
4118 int miimon_interval;
4120 s->name = port->name;
4122 if (port->cfg->bond_mode) {
4123 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4124 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4125 port->name, port->cfg->bond_mode,
4126 bond_mode_to_string(s->balance));
4129 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4131 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4132 * active-backup. At some point we should remove this warning. */
4133 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4134 " in previous versions, the default bond_mode was"
4135 " balance-slb", port->name,
4136 bond_mode_to_string(s->balance));
4138 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4139 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4140 "please use another bond type or disable flood_vlans",
4144 miimon_interval = smap_get_int(&port->cfg->other_config,
4145 "bond-miimon-interval", 0);
4146 if (miimon_interval <= 0) {
4147 miimon_interval = 200;
4150 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4151 if (!detect_s || !strcmp(detect_s, "carrier")) {
4152 miimon_interval = 0;
4153 } else if (strcmp(detect_s, "miimon")) {
4154 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4155 "defaulting to carrier", port->name, detect_s);
4156 miimon_interval = 0;
4159 s->up_delay = MAX(0, port->cfg->bond_updelay);
4160 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4161 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4162 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4163 "bond-rebalance-interval", 10000);
4164 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4165 s->rebalance_interval = 1000;
4168 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4169 "lacp-fallback-ab", false);
4171 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4172 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4175 mac_s = port->cfg->bond_active_slave;
4176 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4177 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4178 /* OVSDB did not store the last active interface */
4179 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4183 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4184 * instead of obtaining it from the database. */
4186 port_is_synthetic(const struct port *port)
4188 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4191 /* Interface functions. */
4194 iface_is_internal(const struct ovsrec_interface *iface,
4195 const struct ovsrec_bridge *br)
4197 /* The local port and "internal" ports are always "internal". */
4198 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4201 /* Returns the correct network device type for interface 'iface' in bridge
4204 iface_get_type(const struct ovsrec_interface *iface,
4205 const struct ovsrec_bridge *br)
4209 /* The local port always has type "internal". Other ports take
4210 * their type from the database and default to "system" if none is
4212 if (iface_is_internal(iface, br)) {
4215 type = iface->type[0] ? iface->type : "system";
4218 return ofproto_port_open_type(br->datapath_type, type);
4222 iface_destroy__(struct iface *iface)
4225 struct port *port = iface->port;
4226 struct bridge *br = port->bridge;
4228 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4229 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4232 if (iface->ofp_port != OFPP_NONE) {
4233 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4236 list_remove(&iface->port_elem);
4237 hmap_remove(&br->iface_by_name, &iface->name_node);
4239 /* The user is changing configuration here, so netdev_remove needs to be
4240 * used as opposed to netdev_close */
4241 netdev_remove(iface->netdev);
4249 iface_destroy(struct iface *iface)
4252 struct port *port = iface->port;
4254 iface_destroy__(iface);
4255 if (list_is_empty(&port->ifaces)) {
4261 static struct iface *
4262 iface_lookup(const struct bridge *br, const char *name)
4264 struct iface *iface;
4266 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4267 &br->iface_by_name) {
4268 if (!strcmp(iface->name, name)) {
4276 static struct iface *
4277 iface_find(const char *name)
4279 const struct bridge *br;
4281 HMAP_FOR_EACH (br, node, &all_bridges) {
4282 struct iface *iface = iface_lookup(br, name);
4291 static struct iface *
4292 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4294 struct iface *iface;
4296 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4298 if (iface->ofp_port == ofp_port) {
4305 /* Set Ethernet address of 'iface', if one is specified in the configuration
4308 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4310 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4311 struct iface *hw_addr_iface;
4313 if (strcmp(iface->type, "internal")) {
4317 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4319 } else if (port->cfg->fake_bridge) {
4320 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4321 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4326 if (iface->ofp_port == OFPP_LOCAL) {
4327 VLOG_ERR("interface %s: ignoring mac in Interface record "
4328 "(use Bridge record to set local port's mac)",
4330 } else if (eth_addr_is_multicast(mac)) {
4331 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4334 int error = netdev_set_etheraddr(iface->netdev, mac);
4336 VLOG_ERR("interface %s: setting MAC failed (%s)",
4337 iface->name, ovs_strerror(error));
4343 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4345 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4347 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4348 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4349 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4353 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4354 * sets the "ofport" field to -1.
4356 * This is appropriate when 'if_cfg''s interface cannot be created or is
4357 * otherwise invalid. */
4359 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4361 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4362 iface_set_ofport(if_cfg, OFPP_NONE);
4363 ovsrec_interface_set_error(if_cfg, errp);
4364 ovsrec_interface_set_status(if_cfg, NULL);
4365 ovsrec_interface_set_admin_state(if_cfg, NULL);
4366 ovsrec_interface_set_duplex(if_cfg, NULL);
4367 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4368 ovsrec_interface_set_link_state(if_cfg, NULL);
4369 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4370 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4371 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4372 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4373 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4374 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4375 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4376 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4381 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4383 union ovsdb_atom atom;
4385 atom.integer = target;
4386 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4390 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4392 struct ofpbuf queues_buf;
4394 ofpbuf_init(&queues_buf, 0);
4396 if (!qos || qos->type[0] == '\0') {
4397 netdev_set_qos(iface->netdev, NULL, NULL);
4399 const struct ovsdb_datum *queues;
4400 struct netdev_queue_dump dump;
4401 unsigned int queue_id;
4402 struct smap details;
4406 /* Configure top-level Qos for 'iface'. */
4407 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4409 /* Deconfigure queues that were deleted. */
4410 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4412 smap_init(&details);
4413 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4414 if (!queue_ids_include(queues, queue_id)) {
4415 netdev_delete_queue(iface->netdev, queue_id);
4418 smap_destroy(&details);
4420 /* Configure queues for 'iface'. */
4422 for (i = 0; i < qos->n_queues; i++) {
4423 const struct ovsrec_queue *queue = qos->value_queues[i];
4424 unsigned int queue_id = qos->key_queues[i];
4426 if (queue_id == 0) {
4430 if (queue->n_dscp == 1) {
4431 struct ofproto_port_queue *port_queue;
4433 port_queue = ofpbuf_put_uninit(&queues_buf,
4434 sizeof *port_queue);
4435 port_queue->queue = queue_id;
4436 port_queue->dscp = queue->dscp[0];
4439 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4442 struct smap details;
4444 smap_init(&details);
4445 netdev_set_queue(iface->netdev, 0, &details);
4446 smap_destroy(&details);
4450 if (iface->ofp_port != OFPP_NONE) {
4451 const struct ofproto_port_queue *port_queues = queues_buf.data;
4452 size_t n_queues = queues_buf.size / sizeof *port_queues;
4454 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4455 port_queues, n_queues);
4458 netdev_set_policing(iface->netdev,
4459 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4460 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4462 ofpbuf_uninit(&queues_buf);
4466 iface_configure_cfm(struct iface *iface)
4468 const struct ovsrec_interface *cfg = iface->cfg;
4469 const char *opstate_str;
4470 const char *cfm_ccm_vlan;
4471 struct cfm_settings s;
4472 struct smap netdev_args;
4474 if (!cfg->n_cfm_mpid) {
4475 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4479 s.check_tnl_key = false;
4480 smap_init(&netdev_args);
4481 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4482 const char *key = smap_get(&netdev_args, "key");
4483 const char *in_key = smap_get(&netdev_args, "in_key");
4485 s.check_tnl_key = (key && !strcmp(key, "flow"))
4486 || (in_key && !strcmp(in_key, "flow"));
4488 smap_destroy(&netdev_args);
4490 s.mpid = *cfg->cfm_mpid;
4491 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4492 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4493 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4495 if (s.interval <= 0) {
4499 if (!cfm_ccm_vlan) {
4501 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4502 s.ccm_vlan = CFM_RANDOM_VLAN;
4504 s.ccm_vlan = atoi(cfm_ccm_vlan);
4505 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4510 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4512 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4514 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4515 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4517 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4520 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4521 * instead of obtaining it from the database. */
4523 iface_is_synthetic(const struct iface *iface)
4525 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4529 iface_validate_ofport__(size_t n, int64_t *ofport)
4531 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4532 ? u16_to_ofp(*ofport)
4537 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4539 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4543 iface_pick_ofport(const struct ovsrec_interface *cfg)
4545 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4546 return (requested_ofport != OFPP_NONE
4548 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4551 /* Port mirroring. */
4553 static struct mirror *
4554 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4558 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4559 if (uuid_equals(uuid, &m->uuid)) {
4567 bridge_configure_mirrors(struct bridge *br)
4569 const struct ovsdb_datum *mc;
4570 unsigned long *flood_vlans;
4571 struct mirror *m, *next;
4574 /* Get rid of deleted mirrors. */
4575 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4576 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4577 union ovsdb_atom atom;
4579 atom.uuid = m->uuid;
4580 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4585 /* Add new mirrors and reconfigure existing ones. */
4586 for (i = 0; i < br->cfg->n_mirrors; i++) {
4587 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4588 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4590 m = mirror_create(br, cfg);
4593 if (!mirror_configure(m)) {
4598 /* Update flooded vlans (for RSPAN). */
4599 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4600 br->cfg->n_flood_vlans);
4601 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4602 bitmap_free(flood_vlans);
4605 static struct mirror *
4606 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4610 m = xzalloc(sizeof *m);
4611 m->uuid = cfg->header_.uuid;
4612 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4614 m->name = xstrdup(cfg->name);
4620 mirror_destroy(struct mirror *m)
4623 struct bridge *br = m->bridge;
4626 ofproto_mirror_unregister(br->ofproto, m);
4629 hmap_remove(&br->mirrors, &m->hmap_node);
4636 mirror_collect_ports(struct mirror *m,
4637 struct ovsrec_port **in_ports, int n_in_ports,
4638 void ***out_portsp, size_t *n_out_portsp)
4640 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4641 size_t n_out_ports = 0;
4644 for (i = 0; i < n_in_ports; i++) {
4645 const char *name = in_ports[i]->name;
4646 struct port *port = port_lookup(m->bridge, name);
4648 out_ports[n_out_ports++] = port;
4650 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4651 "port %s", m->bridge->name, m->name, name);
4654 *out_portsp = out_ports;
4655 *n_out_portsp = n_out_ports;
4659 mirror_configure(struct mirror *m)
4661 const struct ovsrec_mirror *cfg = m->cfg;
4662 struct ofproto_mirror_settings s;
4665 if (strcmp(cfg->name, m->name)) {
4667 m->name = xstrdup(cfg->name);
4671 /* Get output port or VLAN. */
4672 if (cfg->output_port) {
4673 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4674 if (!s.out_bundle) {
4675 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4676 m->bridge->name, m->name);
4679 s.out_vlan = UINT16_MAX;
4681 if (cfg->output_vlan) {
4682 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4683 "output vlan; ignoring output vlan",
4684 m->bridge->name, m->name);
4686 } else if (cfg->output_vlan) {
4687 /* The database should prevent invalid VLAN values. */
4688 s.out_bundle = NULL;
4689 s.out_vlan = *cfg->output_vlan;
4691 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4692 m->bridge->name, m->name);
4696 /* Get port selection. */
4697 if (cfg->select_all) {
4698 size_t n_ports = hmap_count(&m->bridge->ports);
4699 void **ports = xmalloc(n_ports * sizeof *ports);
4704 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4714 /* Get ports, dropping ports that don't exist.
4715 * The IDL ensures that there are no duplicates. */
4716 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4717 &s.srcs, &s.n_srcs);
4718 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4719 &s.dsts, &s.n_dsts);
4722 /* Get VLAN selection. */
4723 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4726 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4729 if (s.srcs != s.dsts) {
4738 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4740 * This is deprecated. It is only for compatibility with broken device drivers
4741 * in old versions of Linux that do not properly support VLANs when VLAN
4742 * devices are not used. When broken device drivers are no longer in
4743 * widespread use, we will delete these interfaces. */
4745 static struct ovsrec_port **recs;
4746 static size_t n_recs, allocated_recs;
4748 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4749 * splinters are reconfigured. */
4751 register_rec(struct ovsrec_port *rec)
4753 if (n_recs >= allocated_recs) {
4754 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4756 recs[n_recs++] = rec;
4759 /* Frees all of the ports registered with register_reg(). */
4761 free_registered_recs(void)
4765 for (i = 0; i < n_recs; i++) {
4766 struct ovsrec_port *port = recs[i];
4769 for (j = 0; j < port->n_interfaces; j++) {
4770 struct ovsrec_interface *iface = port->interfaces[j];
4775 smap_destroy(&port->other_config);
4776 free(port->interfaces);
4784 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4787 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4789 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4793 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4796 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4797 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4798 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4801 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4802 * use, returns NULL.
4804 * Updates 'vlan_splinters_enabled_anywhere'. */
4805 static unsigned long int *
4806 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4808 unsigned long int *splinter_vlans;
4809 struct sset splinter_ifaces;
4810 const char *real_dev_name;
4811 struct shash *real_devs;
4812 struct shash_node *node;
4816 /* Free space allocated for synthesized ports and interfaces, since we're
4817 * in the process of reconstructing all of them. */
4818 free_registered_recs();
4820 splinter_vlans = bitmap_allocate(4096);
4821 sset_init(&splinter_ifaces);
4822 vlan_splinters_enabled_anywhere = false;
4823 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4824 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4827 for (j = 0; j < br_cfg->n_ports; j++) {
4828 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4831 for (k = 0; k < port_cfg->n_interfaces; k++) {
4832 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4834 if (vlan_splinters_is_enabled(iface_cfg)) {
4835 vlan_splinters_enabled_anywhere = true;
4836 sset_add(&splinter_ifaces, iface_cfg->name);
4837 vlan_bitmap_from_array__(port_cfg->trunks,
4843 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4844 bitmap_set1(splinter_vlans, *port_cfg->tag);
4849 if (!vlan_splinters_enabled_anywhere) {
4850 free(splinter_vlans);
4851 sset_destroy(&splinter_ifaces);
4855 HMAP_FOR_EACH (br, node, &all_bridges) {
4857 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4861 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4862 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4863 * device to be created for it. */
4864 bitmap_set0(splinter_vlans, 0);
4865 bitmap_set0(splinter_vlans, 4095);
4867 /* Delete all VLAN devices that we don't need. */
4869 real_devs = vlandev_get_real_devs();
4870 SHASH_FOR_EACH (node, real_devs) {
4871 const struct vlan_real_dev *real_dev = node->data;
4872 const struct vlan_dev *vlan_dev;
4873 bool real_dev_has_splinters;
4875 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4877 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4878 if (!real_dev_has_splinters
4879 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4880 struct netdev *netdev;
4882 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4883 if (!netdev_get_in4(netdev, NULL, NULL) ||
4884 !netdev_get_in6(netdev, NULL)) {
4885 /* It has an IP address configured, so we don't own
4886 * it. Don't delete it. */
4888 vlandev_del(vlan_dev->name);
4890 netdev_close(netdev);
4897 /* Add all VLAN devices that we need. */
4898 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4901 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4902 if (!vlandev_get_name(real_dev_name, vid)) {
4903 vlandev_add(real_dev_name, vid);
4910 sset_destroy(&splinter_ifaces);
4912 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4913 free(splinter_vlans);
4916 return splinter_vlans;
4919 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4922 configure_splinter_port(struct port *port)
4924 struct ofproto *ofproto = port->bridge->ofproto;
4925 ofp_port_t realdev_ofp_port;
4926 const char *realdev_name;
4927 struct iface *vlandev, *realdev;
4929 ofproto_bundle_unregister(port->bridge->ofproto, port);
4931 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4934 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4935 realdev = iface_lookup(port->bridge, realdev_name);
4936 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4938 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4942 static struct ovsrec_port *
4943 synthesize_splinter_port(const char *real_dev_name,
4944 const char *vlan_dev_name, int vid)
4946 struct ovsrec_interface *iface;
4947 struct ovsrec_port *port;
4949 iface = xmalloc(sizeof *iface);
4950 ovsrec_interface_init(iface);
4951 iface->name = xstrdup(vlan_dev_name);
4952 iface->type = "system";
4954 port = xmalloc(sizeof *port);
4955 ovsrec_port_init(port);
4956 port->interfaces = xmemdup(&iface, sizeof iface);
4957 port->n_interfaces = 1;
4958 port->name = xstrdup(vlan_dev_name);
4959 port->vlan_mode = "splinter";
4960 port->tag = xmalloc(sizeof *port->tag);
4963 smap_add(&port->other_config, "realdev", real_dev_name);
4969 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4970 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4971 * 1-bit in the 'splinter_vlans' bitmap. */
4973 add_vlan_splinter_ports(struct bridge *br,
4974 const unsigned long int *splinter_vlans,
4975 struct shash *ports)
4979 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4980 * we're modifying 'ports'. */
4981 for (i = 0; i < br->cfg->n_ports; i++) {
4982 const char *name = br->cfg->ports[i]->name;
4983 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4986 for (j = 0; j < port_cfg->n_interfaces; j++) {
4987 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4989 if (vlan_splinters_is_enabled(iface_cfg)) {
4990 const char *real_dev_name;
4993 real_dev_name = iface_cfg->name;
4994 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4995 const char *vlan_dev_name;
4997 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4999 && !shash_find(ports, vlan_dev_name)) {
5000 shash_add(ports, vlan_dev_name,
5001 synthesize_splinter_port(
5002 real_dev_name, vlan_dev_name, vid));
5011 mirror_refresh_stats(struct mirror *m)
5013 struct ofproto *ofproto = m->bridge->ofproto;
5014 uint64_t tx_packets, tx_bytes;
5015 const char *keys[2];
5017 size_t stat_cnt = 0;
5019 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5020 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5024 if (tx_packets != UINT64_MAX) {
5025 keys[stat_cnt] = "tx_packets";
5026 values[stat_cnt] = tx_packets;
5029 if (tx_bytes != UINT64_MAX) {
5030 keys[stat_cnt] = "tx_bytes";
5031 values[stat_cnt] = tx_bytes;
5035 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5039 * Add registered netdev and dpif types to ovsdb to allow external
5040 * applications to query the capabilities of the Open vSwitch instance
5041 * running on the node.
5044 discover_types(const struct ovsrec_open_vswitch *cfg)
5048 /* Datapath types. */
5050 dp_enumerate_types(&types);
5051 const char **datapath_types = sset_array(&types);
5052 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5053 sset_count(&types));
5054 free(datapath_types);
5055 sset_destroy(&types);
5059 netdev_enumerate_types(&types);
5060 const char **iface_types = sset_array(&types);
5061 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5063 sset_destroy(&types);