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.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
37 #include "mac-learning.h"
38 #include "mcast-snooping.h"
39 #include "meta-flow.h"
42 #include "ofp-print.h"
45 #include "ofproto/bond.h"
46 #include "ofproto/ofproto.h"
48 #include "poll-loop.h"
53 #include "socket-util.h"
55 #include "stream-ssl.h"
57 #include "system-stats.h"
62 #include "lib/vswitch-idl.h"
63 #include "xenserver.h"
64 #include "openvswitch/vlog.h"
65 #include "sflow_api.h"
66 #include "vlan-bitmap.h"
69 VLOG_DEFINE_THIS_MODULE(bridge);
71 COVERAGE_DEFINE(bridge_reconfigure);
74 /* These members are always valid.
76 * They are immutable: they never change between iface_create() and
78 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
79 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
80 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
83 struct netdev *netdev; /* Network device. */
84 ofp_port_t ofp_port; /* OpenFlow port number. */
87 /* These members are valid only within bridge_reconfigure(). */
88 const char *type; /* Usually same as cfg->type. */
89 const struct ovsrec_interface *cfg;
93 struct uuid uuid; /* UUID of this "mirror" record in database. */
94 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
95 struct bridge *bridge;
97 const struct ovsrec_mirror *cfg;
101 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
102 struct bridge *bridge;
105 const struct ovsrec_port *cfg;
107 /* An ordinary bridge port has 1 interface.
108 * A bridge port for bonding has at least 2 interfaces. */
109 struct ovs_list ifaces; /* List of "struct iface"s. */
113 struct hmap_node node; /* In 'all_bridges'. */
114 char *name; /* User-specified arbitrary name. */
115 char *type; /* Datapath type. */
116 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
117 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
118 const struct ovsrec_bridge *cfg;
120 /* OpenFlow switch processing. */
121 struct ofproto *ofproto; /* OpenFlow switch. */
124 struct hmap ports; /* "struct port"s indexed by name. */
125 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
126 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
128 /* Port mirroring. */
129 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
132 struct hmap mappings; /* "struct" indexed by UUID */
134 /* Used during reconfiguration. */
135 struct shash wanted_ports;
137 /* Synthetic local port if necessary. */
138 struct ovsrec_port synth_local_port;
139 struct ovsrec_interface synth_local_iface;
140 struct ovsrec_interface *synth_local_ifacep;
144 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
145 struct bridge *bridge;
151 /* All bridges, indexed by name. */
152 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
154 /* OVSDB IDL used to obtain configuration. */
155 static struct ovsdb_idl *idl;
157 /* We want to complete daemonization, fully detaching from our parent process,
158 * only after we have completed our initial configuration, committed our state
159 * to the database, and received confirmation back from the database server
160 * that it applied the commit. This allows our parent process to know that,
161 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
162 * to have already been filled in. (It is only "very likely" rather than
163 * certain because there is always a slim possibility that the transaction will
164 * fail or that some other client has added new bridges, ports, etc. while
165 * ovs-vswitchd was configuring using an old configuration.)
167 * We only need to do this once for our initial configuration at startup, so
168 * 'initial_config_done' tracks whether we've already done it. While we are
169 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
170 * itself and is otherwise NULL. */
171 static bool initial_config_done;
172 static struct ovsdb_idl_txn *daemonize_txn;
174 /* Most recently processed IDL sequence number. */
175 static unsigned int idl_seqno;
177 /* Track changes to port connectivity. */
178 static uint64_t connectivity_seqno = LLONG_MIN;
180 /* Status update to database.
182 * Some information in the database must be kept as up-to-date as possible to
183 * allow controllers to respond rapidly to network outages. Those status are
184 * updated via the 'status_txn'.
186 * We use the global connectivity sequence number to detect the status change.
187 * Also, to prevent the status update from sending too much to the database,
188 * we check the return status of each update transaction and do not start new
189 * update if the previous transaction status is 'TXN_INCOMPLETE'.
191 * 'statux_txn' is NULL if there is no ongoing status update.
193 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
194 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
195 * which will cause the main thread wake up soon and retry the status update.
197 static struct ovsdb_idl_txn *status_txn;
198 static bool status_txn_try_again;
200 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
201 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
202 #define STATUS_CHECK_AGAIN_MSEC 100
204 /* Each time this timer expires, the bridge fetches interface and mirror
205 * statistics and pushes them into the database. */
206 static int stats_timer_interval;
207 static long long int stats_timer = LLONG_MIN;
209 /* Each time this timer expires, the bridge fetches the list of port/VLAN
210 * membership that has been modified by the AA.
212 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
213 static long long int aa_refresh_timer = LLONG_MIN;
215 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
216 * expensive. This can cause bridge_reconfigure() to take a long time during
217 * which no other work can be done. To deal with this problem, we limit port
218 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
219 * call to bridge_reconfigure(). If there is more work to do after the limit
220 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
221 * This allows the rest of the code to catch up on important things like
222 * forwarding packets. */
223 #define OFP_PORT_ACTION_WINDOW 10
225 static void add_del_bridges(const struct ovsrec_open_vswitch *);
226 static void bridge_run__(void);
227 static void bridge_create(const struct ovsrec_bridge *);
228 static void bridge_destroy(struct bridge *);
229 static struct bridge *bridge_lookup(const char *name);
230 static unixctl_cb_func bridge_unixctl_dump_flows;
231 static unixctl_cb_func bridge_unixctl_reconnect;
232 static size_t bridge_get_controllers(const struct bridge *br,
233 struct ovsrec_controller ***controllersp);
234 static void bridge_collect_wanted_ports(struct bridge *,
235 const unsigned long *splinter_vlans,
236 struct shash *wanted_ports);
237 static void bridge_delete_ofprotos(void);
238 static void bridge_delete_or_reconfigure_ports(struct bridge *);
239 static void bridge_del_ports(struct bridge *,
240 const struct shash *wanted_ports);
241 static void bridge_add_ports(struct bridge *,
242 const struct shash *wanted_ports);
244 static void bridge_configure_datapath_id(struct bridge *);
245 static void bridge_configure_netflow(struct bridge *);
246 static void bridge_configure_forward_bpdu(struct bridge *);
247 static void bridge_configure_mac_table(struct bridge *);
248 static void bridge_configure_mcast_snooping(struct bridge *);
249 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
250 static void bridge_configure_ipfix(struct bridge *);
251 static void bridge_configure_spanning_tree(struct bridge *);
252 static void bridge_configure_tables(struct bridge *);
253 static void bridge_configure_dp_desc(struct bridge *);
254 static void bridge_configure_aa(struct bridge *);
255 static void bridge_aa_refresh_queued(struct bridge *);
256 static bool bridge_aa_need_refresh(struct bridge *);
257 static void bridge_configure_remotes(struct bridge *,
258 const struct sockaddr_in *managers,
260 static void bridge_pick_local_hw_addr(struct bridge *,
261 uint8_t ea[ETH_ADDR_LEN],
262 struct iface **hw_addr_iface);
263 static uint64_t bridge_pick_datapath_id(struct bridge *,
264 const uint8_t bridge_ea[ETH_ADDR_LEN],
265 struct iface *hw_addr_iface);
266 static uint64_t dpid_from_hash(const void *, size_t nbytes);
267 static bool bridge_has_bond_fake_iface(const struct bridge *,
269 static bool port_is_bond_fake_iface(const struct port *);
271 static unixctl_cb_func qos_unixctl_show;
273 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
274 static void port_del_ifaces(struct port *);
275 static void port_destroy(struct port *);
276 static struct port *port_lookup(const struct bridge *, const char *name);
277 static void port_configure(struct port *);
278 static struct lacp_settings *port_configure_lacp(struct port *,
279 struct lacp_settings *);
280 static void port_configure_bond(struct port *, struct bond_settings *);
281 static bool port_is_synthetic(const struct port *);
283 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
284 static void run_system_stats(void);
286 static void bridge_configure_mirrors(struct bridge *);
287 static struct mirror *mirror_create(struct bridge *,
288 const struct ovsrec_mirror *);
289 static void mirror_destroy(struct mirror *);
290 static bool mirror_configure(struct mirror *);
291 static void mirror_refresh_stats(struct mirror *);
293 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
294 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
295 const struct ovsrec_port *);
296 static bool iface_is_internal(const struct ovsrec_interface *iface,
297 const struct ovsrec_bridge *br);
298 static const char *iface_get_type(const struct ovsrec_interface *,
299 const struct ovsrec_bridge *);
300 static void iface_destroy(struct iface *);
301 static void iface_destroy__(struct iface *);
302 static struct iface *iface_lookup(const struct bridge *, const char *name);
303 static struct iface *iface_find(const char *name);
304 static struct iface *iface_from_ofp_port(const struct bridge *,
305 ofp_port_t ofp_port);
306 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
307 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
308 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
309 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
310 static void iface_configure_cfm(struct iface *);
311 static void iface_refresh_cfm_stats(struct iface *);
312 static void iface_refresh_stats(struct iface *);
313 static void iface_refresh_netdev_status(struct iface *);
314 static void iface_refresh_ofproto_status(struct iface *);
315 static bool iface_is_synthetic(const struct iface *);
316 static ofp_port_t iface_get_requested_ofp_port(
317 const struct ovsrec_interface *);
318 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
320 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
322 * This is deprecated. It is only for compatibility with broken device drivers
323 * in old versions of Linux that do not properly support VLANs when VLAN
324 * devices are not used. When broken device drivers are no longer in
325 * widespread use, we will delete these interfaces. */
327 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
328 static bool vlan_splinters_enabled_anywhere;
330 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
331 static unsigned long int *collect_splinter_vlans(
332 const struct ovsrec_open_vswitch *);
333 static void configure_splinter_port(struct port *);
334 static void add_vlan_splinter_ports(struct bridge *,
335 const unsigned long int *splinter_vlans,
336 struct shash *ports);
339 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
341 struct shash iface_hints;
342 static bool initialized = false;
349 shash_init(&iface_hints);
352 for (i = 0; i < cfg->n_bridges; i++) {
353 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
356 for (j = 0; j < br_cfg->n_ports; j++) {
357 struct ovsrec_port *port_cfg = br_cfg->ports[j];
360 for (k = 0; k < port_cfg->n_interfaces; k++) {
361 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
362 struct iface_hint *iface_hint;
364 iface_hint = xmalloc(sizeof *iface_hint);
365 iface_hint->br_name = br_cfg->name;
366 iface_hint->br_type = br_cfg->datapath_type;
367 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
369 shash_add(&iface_hints, if_cfg->name, iface_hint);
375 ofproto_init(&iface_hints);
377 shash_destroy_free_data(&iface_hints);
381 /* Public functions. */
383 /* Initializes the bridge module, configuring it to obtain its configuration
384 * from an OVSDB server accessed over 'remote', which should be a string in a
385 * form acceptable to ovsdb_idl_create(). */
387 bridge_init(const char *remote)
389 /* Create connection to database. */
390 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
391 idl_seqno = ovsdb_idl_get_seqno(idl);
392 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
393 ovsdb_idl_verify_write_only(idl);
395 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
396 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
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 are collected.)
505 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
506 struct sockaddr_in **managersp, size_t *n_managersp)
508 struct sockaddr_in *managers = NULL;
509 size_t n_managers = 0;
513 /* Collect all of the potential targets from the "targets" columns of the
514 * rows pointed to by "manager_options", excluding any that are
517 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
518 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
520 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
521 sset_find_and_delete(&targets, m->target);
523 sset_add(&targets, m->target);
527 /* Now extract the targets' IP addresses. */
528 if (!sset_is_empty(&targets)) {
531 managers = xmalloc(sset_count(&targets) * sizeof *managers);
532 SSET_FOR_EACH (target, &targets) {
534 struct sockaddr_storage ss;
535 struct sockaddr_in in;
538 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
540 && sa.ss.ss_family == AF_INET) {
541 managers[n_managers++] = sa.in;
545 sset_destroy(&targets);
547 *managersp = managers;
548 *n_managersp = n_managers;
552 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
554 unsigned long int *splinter_vlans;
555 struct sockaddr_in *managers;
556 struct bridge *br, *next;
557 int sflow_bridge_number;
560 COVERAGE_INC(bridge_reconfigure);
562 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
563 OFPROTO_FLOW_LIMIT_DEFAULT));
564 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
565 OFPROTO_MAX_IDLE_DEFAULT));
566 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
568 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
571 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
572 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
574 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
575 * to 'ovs_cfg', with only very minimal configuration otherwise.
577 * This is mostly an update to bridge data structures. Nothing is pushed
578 * down to ofproto or lower layers. */
579 add_del_bridges(ovs_cfg);
580 splinter_vlans = collect_splinter_vlans(ovs_cfg);
581 HMAP_FOR_EACH (br, node, &all_bridges) {
582 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
583 bridge_del_ports(br, &br->wanted_ports);
585 free(splinter_vlans);
587 /* Start pushing configuration changes down to the ofproto layer:
589 * - Delete ofprotos that are no longer configured.
591 * - Delete ports that are no longer configured.
593 * - Reconfigure existing ports to their desired configurations, or
594 * delete them if not possible.
596 * We have to do all the deletions before we can do any additions, because
597 * the ports to be added might require resources that will be freed up by
598 * deletions (they might especially overlap in name). */
599 bridge_delete_ofprotos();
600 HMAP_FOR_EACH (br, node, &all_bridges) {
602 bridge_delete_or_reconfigure_ports(br);
606 /* Finish pushing configuration changes to the ofproto layer:
608 * - Create ofprotos that are missing.
610 * - Add ports that are missing. */
611 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
615 error = ofproto_create(br->name, br->type, &br->ofproto);
617 VLOG_ERR("failed to create bridge %s: %s", br->name,
618 ovs_strerror(error));
619 shash_destroy(&br->wanted_ports);
622 /* Trigger storing datapath version. */
623 seq_change(connectivity_seq_get());
627 HMAP_FOR_EACH (br, node, &all_bridges) {
628 bridge_add_ports(br, &br->wanted_ports);
629 shash_destroy(&br->wanted_ports);
632 reconfigure_system_stats(ovs_cfg);
634 /* Complete the configuration. */
635 sflow_bridge_number = 0;
636 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
637 HMAP_FOR_EACH (br, node, &all_bridges) {
640 /* We need the datapath ID early to allow LACP ports to use it as the
641 * default system ID. */
642 bridge_configure_datapath_id(br);
644 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
647 port_configure(port);
649 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
650 iface_set_ofport(iface->cfg, iface->ofp_port);
651 /* Clear eventual previous errors */
652 ovsrec_interface_set_error(iface->cfg, NULL);
653 iface_configure_cfm(iface);
654 iface_configure_qos(iface, port->cfg->qos);
655 iface_set_mac(br, port, iface);
656 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
658 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
662 bridge_configure_mirrors(br);
663 bridge_configure_forward_bpdu(br);
664 bridge_configure_mac_table(br);
665 bridge_configure_mcast_snooping(br);
666 bridge_configure_remotes(br, managers, n_managers);
667 bridge_configure_netflow(br);
668 bridge_configure_sflow(br, &sflow_bridge_number);
669 bridge_configure_ipfix(br);
670 bridge_configure_spanning_tree(br);
671 bridge_configure_tables(br);
672 bridge_configure_dp_desc(br);
673 bridge_configure_aa(br);
677 /* The ofproto-dpif provider does some final reconfiguration in its
678 * ->type_run() function. We have to call it before notifying the database
679 * client that reconfiguration is complete, otherwise there is a very
680 * narrow race window in which e.g. ofproto/trace will not recognize the
681 * new configuration (sometimes this causes unit test failures). */
685 /* Delete ofprotos which aren't configured or have the wrong type. Create
686 * ofprotos which don't exist but need to. */
688 bridge_delete_ofprotos(void)
695 /* Delete ofprotos with no bridge or with the wrong type. */
698 ofproto_enumerate_types(&types);
699 SSET_FOR_EACH (type, &types) {
702 ofproto_enumerate_names(type, &names);
703 SSET_FOR_EACH (name, &names) {
704 br = bridge_lookup(name);
705 if (!br || strcmp(type, br->type)) {
706 ofproto_delete(name, type);
710 sset_destroy(&names);
711 sset_destroy(&types);
715 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
717 if (*n >= *allocated) {
718 ports = x2nrealloc(ports, allocated, sizeof *ports);
720 ports[(*n)++] = port;
725 bridge_delete_or_reconfigure_ports(struct bridge *br)
727 struct ofproto_port ofproto_port;
728 struct ofproto_port_dump dump;
730 struct sset ofproto_ports;
731 struct port *port, *port_next;
733 /* List of "ofp_port"s to delete. We make a list instead of deleting them
734 * right away because ofproto implementations aren't necessarily able to
735 * iterate through a changing list of ports in an entirely robust way. */
742 sset_init(&ofproto_ports);
744 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
745 * that are not configured in the database. (This commonly happens when
746 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
748 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
749 * that have the wrong OpenFlow port number (and arrange to add them back
750 * with the correct OpenFlow port number). */
751 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
752 ofp_port_t requested_ofp_port;
755 sset_add(&ofproto_ports, ofproto_port.name);
757 iface = iface_lookup(br, ofproto_port.name);
759 /* No such iface is configured, so we should delete this
762 * As a corner case exception, keep the port if it's a bond fake
764 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
765 && !strcmp(ofproto_port.type, "internal")) {
771 if (strcmp(ofproto_port.type, iface->type)
772 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
773 /* The interface is the wrong type or can't be configured.
778 /* If the requested OpenFlow port for 'iface' changed, and it's not
779 * already the correct port, then we might want to temporarily delete
780 * this interface, so we can add it back again with the new OpenFlow
782 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
783 if (iface->ofp_port != OFPP_LOCAL &&
784 requested_ofp_port != OFPP_NONE &&
785 requested_ofp_port != iface->ofp_port) {
786 ofp_port_t victim_request;
787 struct iface *victim;
789 /* Check for an existing OpenFlow port currently occupying
790 * 'iface''s requested port number. If there isn't one, then
791 * delete this port. Otherwise we need to consider further. */
792 victim = iface_from_ofp_port(br, requested_ofp_port);
797 /* 'victim' is a port currently using 'iface''s requested port
798 * number. Unless 'victim' specifically requested that port
799 * number, too, then we can delete both 'iface' and 'victim'
800 * temporarily. (We'll add both of them back again later with new
801 * OpenFlow port numbers.)
803 * If 'victim' did request port number 'requested_ofp_port', just
804 * like 'iface', then that's a configuration inconsistency that we
805 * can't resolve. We might as well let it keep its current port
807 victim_request = iface_get_requested_ofp_port(victim->cfg);
808 if (victim_request != requested_ofp_port) {
809 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
810 iface_destroy(victim);
819 iface_destroy(iface);
820 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
822 for (i = 0; i < n; i++) {
823 ofproto_port_del(br->ofproto, del[i]);
827 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
828 * that we didn't see when we iterated through the datapath, i.e. ports
829 * that disappeared underneath use. This is an unusual situation, but it
830 * can happen in some cases:
832 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
833 * idea but could happen).
835 * - The port represented a device that disappeared, e.g. a tuntap
836 * device destroyed via "tunctl -d", a physical Ethernet device
837 * whose module was just unloaded via "rmmod", or a virtual NIC for a
838 * VM whose VM was just terminated. */
839 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
840 struct iface *iface, *iface_next;
842 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
843 if (!sset_contains(&ofproto_ports, iface->name)) {
844 iface_destroy__(iface);
848 if (list_is_empty(&port->ifaces)) {
852 sset_destroy(&ofproto_ports);
856 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
857 bool with_requested_port)
859 struct shash_node *port_node;
861 SHASH_FOR_EACH (port_node, wanted_ports) {
862 const struct ovsrec_port *port_cfg = port_node->data;
865 for (i = 0; i < port_cfg->n_interfaces; i++) {
866 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
867 ofp_port_t requested_ofp_port;
869 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
870 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
871 struct iface *iface = iface_lookup(br, iface_cfg->name);
874 iface_create(br, iface_cfg, port_cfg);
882 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
884 /* First add interfaces that request a particular port number. */
885 bridge_add_ports__(br, wanted_ports, true);
887 /* Then add interfaces that want automatic port number assignment.
888 * We add these afterward to avoid accidentally taking a specifically
889 * requested port number. */
890 bridge_add_ports__(br, wanted_ports, false);
894 port_configure(struct port *port)
896 const struct ovsrec_port *cfg = port->cfg;
897 struct bond_settings bond_settings;
898 struct lacp_settings lacp_settings;
899 struct ofproto_bundle_settings s;
902 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
903 configure_splinter_port(port);
912 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
913 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
914 s.slaves[s.n_slaves++] = iface->ofp_port;
919 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
923 /* Get VLAN trunks. */
926 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
930 if (cfg->vlan_mode) {
931 if (!strcmp(cfg->vlan_mode, "access")) {
932 s.vlan_mode = PORT_VLAN_ACCESS;
933 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
934 s.vlan_mode = PORT_VLAN_TRUNK;
935 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
936 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
937 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
938 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
940 /* This "can't happen" because ovsdb-server should prevent it. */
941 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
942 "back to trunk mode", port->name, cfg->vlan_mode);
943 s.vlan_mode = PORT_VLAN_TRUNK;
947 s.vlan_mode = PORT_VLAN_ACCESS;
949 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
953 s.vlan_mode = PORT_VLAN_TRUNK;
956 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
959 /* Get LACP settings. */
960 s.lacp = port_configure_lacp(port, &lacp_settings);
964 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
965 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
966 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
969 s.lacp_slaves = NULL;
972 /* Get bond settings. */
973 if (s.n_slaves > 1) {
974 s.bond = &bond_settings;
975 port_configure_bond(port, &bond_settings);
978 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
979 netdev_set_miimon_interval(iface->netdev, 0);
984 ofproto_bundle_register(port->bridge->ofproto, port, &s);
992 /* Pick local port hardware address and datapath ID for 'br'. */
994 bridge_configure_datapath_id(struct bridge *br)
996 uint8_t ea[ETH_ADDR_LEN];
998 struct iface *local_iface;
999 struct iface *hw_addr_iface;
1002 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1003 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1005 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1007 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1008 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1009 "Ethernet address: %s",
1010 br->name, ovs_strerror(error));
1013 memcpy(br->ea, ea, ETH_ADDR_LEN);
1015 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1016 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1017 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1018 ofproto_set_datapath_id(br->ofproto, dpid);
1021 dpid_string = xasprintf("%016"PRIx64, dpid);
1022 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1026 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1029 bridge_get_allowed_versions(struct bridge *br)
1031 if (!br->cfg->n_protocols)
1034 return ofputil_versions_from_strings(br->cfg->protocols,
1035 br->cfg->n_protocols);
1038 /* Set NetFlow configuration on 'br'. */
1040 bridge_configure_netflow(struct bridge *br)
1042 struct ovsrec_netflow *cfg = br->cfg->netflow;
1043 struct netflow_options opts;
1046 ofproto_set_netflow(br->ofproto, NULL);
1050 memset(&opts, 0, sizeof opts);
1052 /* Get default NetFlow configuration from datapath.
1053 * Apply overrides from 'cfg'. */
1054 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1055 if (cfg->engine_type) {
1056 opts.engine_type = *cfg->engine_type;
1058 if (cfg->engine_id) {
1059 opts.engine_id = *cfg->engine_id;
1062 /* Configure active timeout interval. */
1063 opts.active_timeout = cfg->active_timeout;
1064 if (!opts.active_timeout) {
1065 opts.active_timeout = -1;
1066 } else if (opts.active_timeout < 0) {
1067 VLOG_WARN("bridge %s: active timeout interval set to negative "
1068 "value, using default instead (%d seconds)", br->name,
1069 NF_ACTIVE_TIMEOUT_DEFAULT);
1070 opts.active_timeout = -1;
1073 /* Add engine ID to interface number to disambiguate bridgs? */
1074 opts.add_id_to_iface = cfg->add_id_to_interface;
1075 if (opts.add_id_to_iface) {
1076 if (opts.engine_id > 0x7f) {
1077 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1078 "another vswitch, choose an engine id less than 128",
1081 if (hmap_count(&br->ports) > 508) {
1082 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1083 "another port when more than 508 ports are used",
1089 sset_init(&opts.collectors);
1090 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1093 if (ofproto_set_netflow(br->ofproto, &opts)) {
1094 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1096 sset_destroy(&opts.collectors);
1099 /* Set sFlow configuration on 'br'. */
1101 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1103 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1104 struct ovsrec_controller **controllers;
1105 struct ofproto_sflow_options oso;
1106 size_t n_controllers;
1110 ofproto_set_sflow(br->ofproto, NULL);
1114 memset(&oso, 0, sizeof oso);
1116 sset_init(&oso.targets);
1117 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1119 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1120 if (cfg->sampling) {
1121 oso.sampling_rate = *cfg->sampling;
1124 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1126 oso.polling_interval = *cfg->polling;
1129 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1131 oso.header_len = *cfg->header;
1134 oso.sub_id = (*sflow_bridge_number)++;
1135 oso.agent_device = cfg->agent;
1137 oso.control_ip = NULL;
1138 n_controllers = bridge_get_controllers(br, &controllers);
1139 for (i = 0; i < n_controllers; i++) {
1140 if (controllers[i]->local_ip) {
1141 oso.control_ip = controllers[i]->local_ip;
1145 ofproto_set_sflow(br->ofproto, &oso);
1147 sset_destroy(&oso.targets);
1150 /* Returns whether a IPFIX row is valid. */
1152 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1154 return ipfix && ipfix->n_targets > 0;
1157 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1159 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1160 const struct bridge *br)
1162 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1165 /* Set IPFIX configuration on 'br'. */
1167 bridge_configure_ipfix(struct bridge *br)
1169 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1170 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1171 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1172 struct ofproto_ipfix_bridge_exporter_options be_opts;
1173 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1174 size_t n_fe_opts = 0;
1176 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1177 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1182 if (!valid_be_cfg && n_fe_opts == 0) {
1183 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1188 memset(&be_opts, 0, sizeof be_opts);
1190 sset_init(&be_opts.targets);
1191 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1193 if (be_cfg->sampling) {
1194 be_opts.sampling_rate = *be_cfg->sampling;
1196 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1198 if (be_cfg->obs_domain_id) {
1199 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1201 if (be_cfg->obs_point_id) {
1202 be_opts.obs_point_id = *be_cfg->obs_point_id;
1204 if (be_cfg->cache_active_timeout) {
1205 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1207 if (be_cfg->cache_max_flows) {
1208 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1211 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1212 "enable-tunnel-sampling", true);
1214 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1215 "enable-input-sampling", false);
1217 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1218 "enable-output-sampling", false);
1221 if (n_fe_opts > 0) {
1222 struct ofproto_ipfix_flow_exporter_options *opts;
1223 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1225 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1226 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1227 opts->collector_set_id = fe_cfg->id;
1228 sset_init(&opts->targets);
1229 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1230 fe_cfg->ipfix->n_targets);
1231 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1232 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1233 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1234 ? *fe_cfg->ipfix->cache_max_flows : 0;
1240 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1244 sset_destroy(&be_opts.targets);
1247 if (n_fe_opts > 0) {
1248 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1250 for (i = 0; i < n_fe_opts; i++) {
1251 sset_destroy(&opts->targets);
1259 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1260 struct ofproto_port_stp_settings *port_s,
1261 int *port_num_counter, unsigned long *port_num_bitmap)
1263 const char *config_str;
1264 struct iface *iface;
1266 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1267 port_s->enable = false;
1270 port_s->enable = true;
1273 /* STP over bonds is not supported. */
1274 if (!list_is_singleton(&port->ifaces)) {
1275 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1277 port_s->enable = false;
1281 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1283 /* Internal ports shouldn't participate in spanning tree, so
1285 if (!strcmp(iface->type, "internal")) {
1286 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1287 port_s->enable = false;
1291 /* STP on mirror output ports is not supported. */
1292 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1293 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1294 port_s->enable = false;
1298 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1300 unsigned long int port_num = strtoul(config_str, NULL, 0);
1301 int port_idx = port_num - 1;
1303 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1304 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1305 port_s->enable = false;
1309 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1310 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1311 port->name, port_num);
1312 port_s->enable = false;
1315 bitmap_set1(port_num_bitmap, port_idx);
1316 port_s->port_num = port_idx;
1318 if (*port_num_counter >= STP_MAX_PORTS) {
1319 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1320 port_s->enable = false;
1324 port_s->port_num = (*port_num_counter)++;
1327 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1329 port_s->path_cost = strtoul(config_str, NULL, 10);
1331 enum netdev_features current;
1334 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1335 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1336 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1339 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1341 port_s->priority = strtoul(config_str, NULL, 0);
1343 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1348 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1349 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1351 const char *config_str;
1352 struct iface *iface;
1354 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1355 port_s->enable = false;
1358 port_s->enable = true;
1361 /* RSTP over bonds is not supported. */
1362 if (!list_is_singleton(&port->ifaces)) {
1363 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1365 port_s->enable = false;
1369 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1371 /* Internal ports shouldn't participate in spanning tree, so
1373 if (!strcmp(iface->type, "internal")) {
1374 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1375 port_s->enable = false;
1379 /* RSTP on mirror output ports is not supported. */
1380 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1381 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1382 port_s->enable = false;
1386 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1388 unsigned long int port_num = strtoul(config_str, NULL, 0);
1389 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1390 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1391 port_s->enable = false;
1394 port_s->port_num = port_num;
1396 if (*port_num_counter >= RSTP_MAX_PORTS) {
1397 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1398 port_s->enable = false;
1401 /* If rstp-port-num is not specified, use 0.
1402 * rstp_port_set_port_number() will look for the first free one. */
1403 port_s->port_num = 0;
1406 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1408 port_s->path_cost = strtoul(config_str, NULL, 10);
1410 enum netdev_features current;
1413 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1414 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1415 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1418 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1420 port_s->priority = strtoul(config_str, NULL, 0);
1422 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1425 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1427 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1429 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1432 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1433 "rstp-admin-port-state", true);
1435 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1436 "rstp-port-admin-edge", false);
1437 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1438 "rstp-port-auto-edge", true);
1439 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1440 "rstp-port-mcheck", false);
1443 /* Set spanning tree configuration on 'br'. */
1445 bridge_configure_stp(struct bridge *br, bool enable_stp)
1448 ofproto_set_stp(br->ofproto, NULL);
1450 struct ofproto_stp_settings br_s;
1451 const char *config_str;
1453 int port_num_counter;
1454 unsigned long *port_num_bitmap;
1456 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1458 uint8_t ea[ETH_ADDR_LEN];
1460 if (eth_addr_from_string(config_str, ea)) {
1461 br_s.system_id = eth_addr_to_uint64(ea);
1463 br_s.system_id = eth_addr_to_uint64(br->ea);
1464 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1465 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1468 br_s.system_id = eth_addr_to_uint64(br->ea);
1471 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1473 br_s.priority = strtoul(config_str, NULL, 0);
1475 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1478 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1480 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1482 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1485 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1487 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1489 br_s.max_age = STP_DEFAULT_MAX_AGE;
1492 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1494 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1496 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1499 /* Configure STP on the bridge. */
1500 if (ofproto_set_stp(br->ofproto, &br_s)) {
1501 VLOG_ERR("bridge %s: could not enable STP", br->name);
1505 /* Users must either set the port number with the "stp-port-num"
1506 * configuration on all ports or none. If manual configuration
1507 * is not done, then we allocate them sequentially. */
1508 port_num_counter = 0;
1509 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1510 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1511 struct ofproto_port_stp_settings port_s;
1512 struct iface *iface;
1514 port_configure_stp(br->ofproto, port, &port_s,
1515 &port_num_counter, port_num_bitmap);
1517 /* As bonds are not supported, just apply configuration to
1518 * all interfaces. */
1519 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1520 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1522 VLOG_ERR("port %s: could not enable STP", port->name);
1528 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1529 && port_num_counter) {
1530 VLOG_ERR("bridge %s: must manually configure all STP port "
1531 "IDs or none, disabling", br->name);
1532 ofproto_set_stp(br->ofproto, NULL);
1534 bitmap_free(port_num_bitmap);
1539 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1542 ofproto_set_rstp(br->ofproto, NULL);
1544 struct ofproto_rstp_settings br_s;
1545 const char *config_str;
1547 int port_num_counter;
1549 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1551 uint8_t ea[ETH_ADDR_LEN];
1553 if (eth_addr_from_string(config_str, ea)) {
1554 br_s.address = eth_addr_to_uint64(ea);
1557 br_s.address = eth_addr_to_uint64(br->ea);
1558 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1559 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1563 br_s.address = eth_addr_to_uint64(br->ea);
1566 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1568 br_s.priority = strtoul(config_str, NULL, 0);
1570 br_s.priority = RSTP_DEFAULT_PRIORITY;
1573 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1575 br_s.ageing_time = strtoul(config_str, NULL, 0);
1577 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1580 config_str = smap_get(&br->cfg->other_config,
1581 "rstp-force-protocol-version");
1583 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1585 br_s.force_protocol_version = FPV_DEFAULT;
1588 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1590 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1592 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1595 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1597 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1599 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1602 config_str = smap_get(&br->cfg->other_config,
1603 "rstp-transmit-hold-count");
1605 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1607 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1610 /* Configure RSTP on the bridge. */
1611 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1612 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1616 port_num_counter = 0;
1617 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1618 struct ofproto_port_rstp_settings port_s;
1619 struct iface *iface;
1621 port_configure_rstp(br->ofproto, port, &port_s,
1624 /* As bonds are not supported, just apply configuration to
1625 * all interfaces. */
1626 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1627 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1629 VLOG_ERR("port %s: could not enable RSTP", port->name);
1638 bridge_configure_spanning_tree(struct bridge *br)
1640 bool enable_rstp = br->cfg->rstp_enable;
1641 bool enable_stp = br->cfg->stp_enable;
1643 if (enable_rstp && enable_stp) {
1644 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1645 "configured; enabling RSTP", br->name);
1649 bridge_configure_stp(br, enable_stp);
1650 bridge_configure_rstp(br, enable_rstp);
1654 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1656 const struct port *port = port_lookup(br, name);
1657 return port && port_is_bond_fake_iface(port);
1661 port_is_bond_fake_iface(const struct port *port)
1663 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1667 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1669 struct bridge *br, *next;
1670 struct shash new_br;
1673 /* Collect new bridges' names and types. */
1674 shash_init(&new_br);
1675 for (i = 0; i < cfg->n_bridges; i++) {
1676 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1677 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1679 if (strchr(br_cfg->name, '/')) {
1680 /* Prevent remote ovsdb-server users from accessing arbitrary
1681 * directories, e.g. consider a bridge named "../../../etc/". */
1682 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1684 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1685 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1689 /* Get rid of deleted bridges or those whose types have changed.
1690 * Update 'cfg' of bridges that still exist. */
1691 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1692 br->cfg = shash_find_data(&new_br, br->name);
1693 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1694 br->cfg->datapath_type))) {
1699 /* Add new bridges. */
1700 for (i = 0; i < cfg->n_bridges; i++) {
1701 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1702 struct bridge *br = bridge_lookup(br_cfg->name);
1704 bridge_create(br_cfg);
1708 shash_destroy(&new_br);
1711 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1712 * Returns 0 if successful, otherwise a positive errno value. */
1714 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1715 struct netdev *netdev, char **errp)
1717 return netdev_set_config(netdev, &iface_cfg->options, errp);
1720 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1721 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1723 * If successful, returns 0 and stores the network device in '*netdevp'. On
1724 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1726 iface_do_create(const struct bridge *br,
1727 const struct ovsrec_interface *iface_cfg,
1728 const struct ovsrec_port *port_cfg,
1729 ofp_port_t *ofp_portp, struct netdev **netdevp,
1732 struct netdev *netdev = NULL;
1735 if (netdev_is_reserved_name(iface_cfg->name)) {
1736 VLOG_WARN("could not create interface %s, name is reserved",
1742 error = netdev_open(iface_cfg->name,
1743 iface_get_type(iface_cfg, br->cfg), &netdev);
1745 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1746 iface_cfg->name, ovs_strerror(error));
1750 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1755 *ofp_portp = iface_pick_ofport(iface_cfg);
1756 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1761 VLOG_INFO("bridge %s: added interface %s on port %d",
1762 br->name, iface_cfg->name, *ofp_portp);
1764 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1765 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1773 netdev_close(netdev);
1777 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1778 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1779 * automatically allocated for the iface. Takes ownership of and
1780 * deallocates 'if_cfg'.
1782 * Return true if an iface is successfully created, false otherwise. */
1784 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1785 const struct ovsrec_port *port_cfg)
1787 struct netdev *netdev;
1788 struct iface *iface;
1789 ofp_port_t ofp_port;
1794 /* Do the bits that can fail up front. */
1795 ovs_assert(!iface_lookup(br, iface_cfg->name));
1796 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1798 iface_clear_db_record(iface_cfg, errp);
1803 /* Get or create the port structure. */
1804 port = port_lookup(br, port_cfg->name);
1806 port = port_create(br, port_cfg);
1809 /* Create the iface structure. */
1810 iface = xzalloc(sizeof *iface);
1811 list_push_back(&port->ifaces, &iface->port_elem);
1812 hmap_insert(&br->iface_by_name, &iface->name_node,
1813 hash_string(iface_cfg->name, 0));
1815 iface->name = xstrdup(iface_cfg->name);
1816 iface->ofp_port = ofp_port;
1817 iface->netdev = netdev;
1818 iface->type = iface_get_type(iface_cfg, br->cfg);
1819 iface->cfg = iface_cfg;
1820 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1821 hash_ofp_port(ofp_port));
1823 /* Populate initial status in database. */
1824 iface_refresh_stats(iface);
1825 iface_refresh_netdev_status(iface);
1827 /* Add bond fake iface if necessary. */
1828 if (port_is_bond_fake_iface(port)) {
1829 struct ofproto_port ofproto_port;
1831 if (ofproto_port_query_by_name(br->ofproto, port->name,
1833 struct netdev *netdev;
1836 error = netdev_open(port->name, "internal", &netdev);
1838 ofp_port_t fake_ofp_port = OFPP_NONE;
1839 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1840 netdev_close(netdev);
1842 VLOG_WARN("could not open network device %s (%s)",
1843 port->name, ovs_strerror(error));
1846 /* Already exists, nothing to do. */
1847 ofproto_port_destroy(&ofproto_port);
1854 /* Set forward BPDU option. */
1856 bridge_configure_forward_bpdu(struct bridge *br)
1858 ofproto_set_forward_bpdu(br->ofproto,
1859 smap_get_bool(&br->cfg->other_config,
1864 /* Set MAC learning table configuration for 'br'. */
1866 bridge_configure_mac_table(struct bridge *br)
1868 const char *idle_time_str;
1871 const char *mac_table_size_str;
1874 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1875 idle_time = (idle_time_str && atoi(idle_time_str)
1876 ? atoi(idle_time_str)
1877 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1879 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1880 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1881 ? atoi(mac_table_size_str)
1884 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1887 /* Set multicast snooping table configuration for 'br'. */
1889 bridge_configure_mcast_snooping(struct bridge *br)
1891 if (!br->cfg->mcast_snooping_enable) {
1892 ofproto_set_mcast_snooping(br->ofproto, NULL);
1895 struct ofproto_mcast_snooping_settings br_s;
1896 const char *idle_time_str;
1897 const char *max_entries_str;
1899 idle_time_str = smap_get(&br->cfg->other_config,
1900 "mcast-snooping-aging-time");
1901 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1902 ? atoi(idle_time_str)
1903 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1905 max_entries_str = smap_get(&br->cfg->other_config,
1906 "mcast-snooping-table-size");
1907 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1908 ? atoi(max_entries_str)
1909 : MCAST_DEFAULT_MAX_ENTRIES);
1911 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1912 "mcast-snooping-disable-flood-unregistered",
1915 /* Configure multicast snooping on the bridge */
1916 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1917 VLOG_ERR("bridge %s: could not enable multicast snooping",
1922 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1923 struct ofproto_mcast_snooping_port_settings port_s;
1924 port_s.flood = smap_get_bool(&port->cfg->other_config,
1925 "mcast-snooping-flood", false);
1926 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1927 "mcast-snooping-flood-reports", false);
1928 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1929 VLOG_ERR("port %s: could not configure mcast snooping",
1937 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1938 const struct port *fake_br, struct iface **hw_addr_iface)
1940 struct hmapx mirror_output_ports;
1942 bool found_addr = false;
1946 /* Mirror output ports don't participate in picking the local hardware
1947 * address. ofproto can't help us find out whether a given port is a
1948 * mirror output because we haven't configured mirrors yet, so we need to
1949 * accumulate them ourselves. */
1950 hmapx_init(&mirror_output_ports);
1951 for (i = 0; i < br->cfg->n_mirrors; i++) {
1952 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1953 if (m->output_port) {
1954 hmapx_add(&mirror_output_ports, m->output_port);
1958 /* Otherwise choose the minimum non-local MAC address among all of the
1960 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1961 uint8_t iface_ea[ETH_ADDR_LEN];
1962 struct iface *candidate;
1963 struct iface *iface;
1965 /* Mirror output ports don't participate. */
1966 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1970 /* Choose the MAC address to represent the port. */
1972 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1973 /* Find the interface with this Ethernet address (if any) so that
1974 * we can provide the correct devname to the caller. */
1975 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1976 uint8_t candidate_ea[ETH_ADDR_LEN];
1977 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1978 && eth_addr_equals(iface_ea, candidate_ea)) {
1983 /* Choose the interface whose MAC address will represent the port.
1984 * The Linux kernel bonding code always chooses the MAC address of
1985 * the first slave added to a bond, and the Fedora networking
1986 * scripts always add slaves to a bond in alphabetical order, so
1987 * for compatibility we choose the interface with the name that is
1988 * first in alphabetical order. */
1989 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1990 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1995 /* The local port doesn't count (since we're trying to choose its
1996 * MAC address anyway). */
1997 if (iface->ofp_port == OFPP_LOCAL) {
2001 /* For fake bridges we only choose from ports with the same tag */
2002 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2003 if (!port->cfg->tag) {
2006 if (*port->cfg->tag != *fake_br->cfg->tag) {
2012 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2018 /* Compare against our current choice. */
2019 if (!eth_addr_is_multicast(iface_ea) &&
2020 !eth_addr_is_local(iface_ea) &&
2021 !eth_addr_is_reserved(iface_ea) &&
2022 !eth_addr_is_zero(iface_ea) &&
2023 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2025 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2026 *hw_addr_iface = iface;
2032 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2033 *hw_addr_iface = NULL;
2036 hmapx_destroy(&mirror_output_ports);
2040 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2041 struct iface **hw_addr_iface)
2044 *hw_addr_iface = NULL;
2046 /* Did the user request a particular MAC? */
2047 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2048 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2049 if (eth_addr_is_multicast(ea)) {
2050 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2051 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2052 } else if (eth_addr_is_zero(ea)) {
2053 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2059 /* Find a local hw address */
2060 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2063 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2064 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2065 * an interface on 'br', then that interface must be passed in as
2066 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2067 * 'hw_addr_iface' must be passed in as a null pointer. */
2069 bridge_pick_datapath_id(struct bridge *br,
2070 const uint8_t bridge_ea[ETH_ADDR_LEN],
2071 struct iface *hw_addr_iface)
2074 * The procedure for choosing a bridge MAC address will, in the most
2075 * ordinary case, also choose a unique MAC that we can use as a datapath
2076 * ID. In some special cases, though, multiple bridges will end up with
2077 * the same MAC address. This is OK for the bridges, but it will confuse
2078 * the OpenFlow controller, because each datapath needs a unique datapath
2081 * Datapath IDs must be unique. It is also very desirable that they be
2082 * stable from one run to the next, so that policy set on a datapath
2085 const char *datapath_id;
2088 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2089 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2093 if (!hw_addr_iface) {
2095 * A purely internal bridge, that is, one that has no non-virtual
2096 * network devices on it at all, is difficult because it has no
2097 * natural unique identifier at all.
2099 * When the host is a XenServer, we handle this case by hashing the
2100 * host's UUID with the name of the bridge. Names of bridges are
2101 * persistent across XenServer reboots, although they can be reused if
2102 * an internal network is destroyed and then a new one is later
2103 * created, so this is fairly effective.
2105 * When the host is not a XenServer, we punt by using a random MAC
2106 * address on each run.
2108 const char *host_uuid = xenserver_get_host_uuid();
2110 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2111 dpid = dpid_from_hash(combined, strlen(combined));
2117 return eth_addr_to_uint64(bridge_ea);
2121 dpid_from_hash(const void *data, size_t n)
2123 uint8_t hash[SHA1_DIGEST_SIZE];
2125 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2126 sha1_bytes(data, n, hash);
2127 eth_addr_mark_random(hash);
2128 return eth_addr_to_uint64(hash);
2132 iface_refresh_netdev_status(struct iface *iface)
2136 enum netdev_features current;
2137 enum netdev_flags flags;
2138 const char *link_state;
2139 uint8_t mac[ETH_ADDR_LEN];
2140 int64_t bps, mtu_64, ifindex64, link_resets;
2143 if (iface_is_synthetic(iface)) {
2147 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2148 && !status_txn_try_again) {
2152 iface->change_seq = netdev_get_change_seq(iface->netdev);
2156 if (!netdev_get_status(iface->netdev, &smap)) {
2157 ovsrec_interface_set_status(iface->cfg, &smap);
2159 ovsrec_interface_set_status(iface->cfg, NULL);
2162 smap_destroy(&smap);
2164 error = netdev_get_flags(iface->netdev, &flags);
2166 const char *state = flags & NETDEV_UP ? "up" : "down";
2168 ovsrec_interface_set_admin_state(iface->cfg, state);
2170 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2173 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2174 ovsrec_interface_set_link_state(iface->cfg, link_state);
2176 link_resets = netdev_get_carrier_resets(iface->netdev);
2177 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2179 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2180 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2182 ovsrec_interface_set_duplex(iface->cfg,
2183 netdev_features_is_full_duplex(current)
2185 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2187 ovsrec_interface_set_duplex(iface->cfg, NULL);
2188 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2191 error = netdev_get_mtu(iface->netdev, &mtu);
2194 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2196 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2199 error = netdev_get_etheraddr(iface->netdev, mac);
2201 char mac_string[32];
2203 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2204 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2206 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2209 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2210 * if there is no valid ifindex number. */
2211 ifindex64 = netdev_get_ifindex(iface->netdev);
2212 if (ifindex64 < 0) {
2215 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2219 iface_refresh_ofproto_status(struct iface *iface)
2223 if (iface_is_synthetic(iface)) {
2227 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2231 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2233 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2236 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2238 || status_txn_try_again) {
2239 iface_refresh_cfm_stats(iface);
2242 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2244 || status_txn_try_again) {
2248 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2249 iface->ofp_port, &smap);
2250 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2251 smap_destroy(&smap);
2255 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2258 iface_refresh_cfm_stats(struct iface *iface)
2260 const struct ovsrec_interface *cfg = iface->cfg;
2261 struct cfm_status status;
2264 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2265 iface->ofp_port, &status);
2267 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2268 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2269 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2270 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2271 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2272 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2274 const char *reasons[CFM_FAULT_N_REASONS];
2275 int64_t cfm_health = status.health;
2276 int64_t cfm_flap_count = status.flap_count;
2277 bool faulted = status.faults != 0;
2280 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2283 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2284 int reason = 1 << i;
2285 if (status.faults & reason) {
2286 reasons[j++] = cfm_fault_reason_to_str(reason);
2289 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2291 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2293 if (status.remote_opstate >= 0) {
2294 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2295 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2297 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2300 ovsrec_interface_set_cfm_remote_mpids(cfg,
2301 (const int64_t *)status.rmps,
2303 if (cfm_health >= 0) {
2304 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2306 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2314 iface_refresh_stats(struct iface *iface)
2316 #define IFACE_STATS \
2317 IFACE_STAT(rx_packets, "rx_packets") \
2318 IFACE_STAT(tx_packets, "tx_packets") \
2319 IFACE_STAT(rx_bytes, "rx_bytes") \
2320 IFACE_STAT(tx_bytes, "tx_bytes") \
2321 IFACE_STAT(rx_dropped, "rx_dropped") \
2322 IFACE_STAT(tx_dropped, "tx_dropped") \
2323 IFACE_STAT(rx_errors, "rx_errors") \
2324 IFACE_STAT(tx_errors, "tx_errors") \
2325 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2326 IFACE_STAT(rx_over_errors, "rx_over_err") \
2327 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2328 IFACE_STAT(collisions, "collisions")
2330 #define IFACE_STAT(MEMBER, NAME) + 1
2331 enum { N_IFACE_STATS = IFACE_STATS };
2333 int64_t values[N_IFACE_STATS];
2334 const char *keys[N_IFACE_STATS];
2337 struct netdev_stats stats;
2339 if (iface_is_synthetic(iface)) {
2343 /* Intentionally ignore return value, since errors will set 'stats' to
2344 * all-1s, and we will deal with that correctly below. */
2345 netdev_get_stats(iface->netdev, &stats);
2347 /* Copy statistics into keys[] and values[]. */
2349 #define IFACE_STAT(MEMBER, NAME) \
2350 if (stats.MEMBER != UINT64_MAX) { \
2352 values[n] = stats.MEMBER; \
2357 ovs_assert(n <= N_IFACE_STATS);
2359 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2364 br_refresh_datapath_info(struct bridge *br)
2366 const char *version;
2368 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2369 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2372 ovsrec_bridge_set_datapath_version(br->cfg,
2373 version ? version : "<unknown>");
2377 br_refresh_stp_status(struct bridge *br)
2379 struct smap smap = SMAP_INITIALIZER(&smap);
2380 struct ofproto *ofproto = br->ofproto;
2381 struct ofproto_stp_status status;
2383 if (ofproto_get_stp_status(ofproto, &status)) {
2387 if (!status.enabled) {
2388 ovsrec_bridge_set_status(br->cfg, NULL);
2392 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2393 STP_ID_ARGS(status.bridge_id));
2394 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2395 STP_ID_ARGS(status.designated_root));
2396 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2398 ovsrec_bridge_set_status(br->cfg, &smap);
2399 smap_destroy(&smap);
2403 port_refresh_stp_status(struct port *port)
2405 struct ofproto *ofproto = port->bridge->ofproto;
2406 struct iface *iface;
2407 struct ofproto_port_stp_status status;
2410 if (port_is_synthetic(port)) {
2414 /* STP doesn't currently support bonds. */
2415 if (!list_is_singleton(&port->ifaces)) {
2416 ovsrec_port_set_status(port->cfg, NULL);
2420 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2421 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2425 if (!status.enabled) {
2426 ovsrec_port_set_status(port->cfg, NULL);
2430 /* Set Status column. */
2432 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2433 smap_add(&smap, "stp_state", stp_state_name(status.state));
2434 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2435 smap_add(&smap, "stp_role", stp_role_name(status.role));
2436 ovsrec_port_set_status(port->cfg, &smap);
2437 smap_destroy(&smap);
2441 port_refresh_stp_stats(struct port *port)
2443 struct ofproto *ofproto = port->bridge->ofproto;
2444 struct iface *iface;
2445 struct ofproto_port_stp_stats stats;
2446 const char *keys[3];
2447 int64_t int_values[3];
2449 if (port_is_synthetic(port)) {
2453 /* STP doesn't currently support bonds. */
2454 if (!list_is_singleton(&port->ifaces)) {
2458 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2459 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2463 if (!stats.enabled) {
2464 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2468 /* Set Statistics column. */
2469 keys[0] = "stp_tx_count";
2470 int_values[0] = stats.tx_count;
2471 keys[1] = "stp_rx_count";
2472 int_values[1] = stats.rx_count;
2473 keys[2] = "stp_error_count";
2474 int_values[2] = stats.error_count;
2476 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2477 ARRAY_SIZE(int_values));
2481 br_refresh_rstp_status(struct bridge *br)
2483 struct smap smap = SMAP_INITIALIZER(&smap);
2484 struct ofproto *ofproto = br->ofproto;
2485 struct ofproto_rstp_status status;
2487 if (ofproto_get_rstp_status(ofproto, &status)) {
2490 if (!status.enabled) {
2491 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2494 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2495 RSTP_ID_ARGS(status.bridge_id));
2496 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2497 status.root_path_cost);
2498 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2499 RSTP_ID_ARGS(status.root_id));
2500 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2501 RSTP_ID_ARGS(status.designated_id));
2502 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2503 status.designated_port_id);
2504 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2505 status.bridge_port_id);
2506 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2507 smap_destroy(&smap);
2511 port_refresh_rstp_status(struct port *port)
2513 struct ofproto *ofproto = port->bridge->ofproto;
2514 struct iface *iface;
2515 struct ofproto_port_rstp_status status;
2516 const char *keys[4];
2517 int64_t int_values[4];
2520 if (port_is_synthetic(port)) {
2524 /* RSTP doesn't currently support bonds. */
2525 if (!list_is_singleton(&port->ifaces)) {
2526 ovsrec_port_set_rstp_status(port->cfg, NULL);
2530 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2531 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2535 if (!status.enabled) {
2536 ovsrec_port_set_rstp_status(port->cfg, NULL);
2537 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2540 /* Set Status column. */
2543 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2545 smap_add_format(&smap, "rstp_port_role", "%s",
2546 rstp_port_role_name(status.role));
2547 smap_add_format(&smap, "rstp_port_state", "%s",
2548 rstp_state_name(status.state));
2549 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2550 RSTP_ID_ARGS(status.designated_bridge_id));
2551 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2552 status.designated_port_id);
2553 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2554 status.designated_path_cost);
2556 ovsrec_port_set_rstp_status(port->cfg, &smap);
2557 smap_destroy(&smap);
2559 /* Set Statistics column. */
2560 keys[0] = "rstp_tx_count";
2561 int_values[0] = status.tx_count;
2562 keys[1] = "rstp_rx_count";
2563 int_values[1] = status.rx_count;
2564 keys[2] = "rstp_uptime";
2565 int_values[2] = status.uptime;
2566 keys[3] = "rstp_error_count";
2567 int_values[3] = status.error_count;
2568 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2569 ARRAY_SIZE(int_values));
2573 port_refresh_bond_status(struct port *port, bool force_update)
2575 uint8_t mac[ETH_ADDR_LEN];
2577 /* Return if port is not a bond */
2578 if (list_is_singleton(&port->ifaces)) {
2582 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2586 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2587 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2593 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2595 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2599 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2601 bool enable = enable_system_stats(cfg);
2603 system_stats_enable(enable);
2605 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2610 run_system_stats(void)
2612 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2615 stats = system_stats_run();
2617 struct ovsdb_idl_txn *txn;
2618 struct ovsdb_datum datum;
2620 txn = ovsdb_idl_txn_create(idl);
2621 ovsdb_datum_from_smap(&datum, stats);
2622 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2624 ovsdb_idl_txn_commit(txn);
2625 ovsdb_idl_txn_destroy(txn);
2632 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2635 case OFPCR12_ROLE_EQUAL:
2637 case OFPCR12_ROLE_MASTER:
2639 case OFPCR12_ROLE_SLAVE:
2641 case OFPCR12_ROLE_NOCHANGE:
2643 return "*** INVALID ROLE ***";
2648 refresh_controller_status(void)
2652 const struct ovsrec_controller *cfg;
2656 /* Accumulate status for controllers on all bridges. */
2657 HMAP_FOR_EACH (br, node, &all_bridges) {
2658 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2661 /* Update each controller in the database with current status. */
2662 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2663 struct ofproto_controller_info *cinfo =
2664 shash_find_data(&info, cfg->target);
2667 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2668 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2670 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2672 ovsrec_controller_set_is_connected(cfg, false);
2673 ovsrec_controller_set_role(cfg, NULL);
2674 ovsrec_controller_set_status(cfg, NULL);
2678 ofproto_free_ofproto_controller_info(&info);
2681 /* Update interface and mirror statistics if necessary. */
2683 run_stats_update(void)
2685 static struct ovsdb_idl_txn *stats_txn;
2686 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2693 /* Statistics update interval should always be greater than or equal to
2695 stats_interval = MAX(smap_get_int(&cfg->other_config,
2696 "stats-update-interval",
2698 if (stats_timer_interval != stats_interval) {
2699 stats_timer_interval = stats_interval;
2700 stats_timer = LLONG_MIN;
2703 if (time_msec() >= stats_timer) {
2704 enum ovsdb_idl_txn_status status;
2706 /* Rate limit the update. Do not start a new update if the
2707 * previous one is not done. */
2711 stats_txn = ovsdb_idl_txn_create(idl);
2712 HMAP_FOR_EACH (br, node, &all_bridges) {
2716 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2717 struct iface *iface;
2719 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2720 iface_refresh_stats(iface);
2722 port_refresh_stp_stats(port);
2724 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2725 mirror_refresh_stats(m);
2728 refresh_controller_status();
2731 status = ovsdb_idl_txn_commit(stats_txn);
2732 if (status != TXN_INCOMPLETE) {
2733 stats_timer = time_msec() + stats_timer_interval;
2734 ovsdb_idl_txn_destroy(stats_txn);
2740 /* Update bridge/port/interface status if necessary. */
2742 run_status_update(void)
2747 /* Rate limit the update. Do not start a new update if the
2748 * previous one is not done. */
2749 seq = seq_read(connectivity_seq_get());
2750 if (seq != connectivity_seqno || status_txn_try_again) {
2753 connectivity_seqno = seq;
2754 status_txn = ovsdb_idl_txn_create(idl);
2755 HMAP_FOR_EACH (br, node, &all_bridges) {
2758 br_refresh_stp_status(br);
2759 br_refresh_rstp_status(br);
2760 br_refresh_datapath_info(br);
2761 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2762 struct iface *iface;
2764 port_refresh_stp_status(port);
2765 port_refresh_rstp_status(port);
2766 port_refresh_bond_status(port, status_txn_try_again);
2767 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2768 iface_refresh_netdev_status(iface);
2769 iface_refresh_ofproto_status(iface);
2776 /* Commit the transaction and get the status. If the transaction finishes,
2777 * then destroy the transaction. Otherwise, keep it so that we can check
2778 * progress the next time that this function is called. */
2780 enum ovsdb_idl_txn_status status;
2782 status = ovsdb_idl_txn_commit(status_txn);
2783 if (status != TXN_INCOMPLETE) {
2784 ovsdb_idl_txn_destroy(status_txn);
2787 /* Sets the 'status_txn_try_again' if the transaction fails. */
2788 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2789 status_txn_try_again = false;
2791 status_txn_try_again = true;
2796 /* Refresh AA port status if necessary. */
2797 if (time_msec() >= aa_refresh_timer) {
2800 HMAP_FOR_EACH (br, node, &all_bridges) {
2801 if (bridge_aa_need_refresh(br)) {
2802 struct ovsdb_idl_txn *txn;
2804 txn = ovsdb_idl_txn_create(idl);
2805 bridge_aa_refresh_queued(br);
2806 ovsdb_idl_txn_commit(txn);
2807 ovsdb_idl_txn_destroy(txn);
2811 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2816 status_update_wait(void)
2818 /* This prevents the process from constantly waking up on
2819 * connectivity seq, when there is no connection to ovsdb. */
2820 if (!ovsdb_idl_has_lock(idl)) {
2824 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2825 * transaction to complete. If the status update to database needs to be
2826 * run again (transaction fails), registers a timeout in
2827 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2828 * sequence number. */
2830 ovsdb_idl_txn_wait(status_txn);
2831 } else if (status_txn_try_again) {
2832 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2834 seq_wait(connectivity_seq_get(), connectivity_seqno);
2845 /* Let each datapath type do the work that it needs to do. */
2847 ofproto_enumerate_types(&types);
2848 SSET_FOR_EACH (type, &types) {
2849 ofproto_type_run(type);
2851 sset_destroy(&types);
2853 /* Let each bridge do the work that it needs to do. */
2854 HMAP_FOR_EACH (br, node, &all_bridges) {
2855 ofproto_run(br->ofproto);
2862 static struct ovsrec_open_vswitch null_cfg;
2863 const struct ovsrec_open_vswitch *cfg;
2865 bool vlan_splinters_changed;
2867 ovsrec_open_vswitch_init(&null_cfg);
2871 if (ovsdb_idl_is_lock_contended(idl)) {
2872 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2873 struct bridge *br, *next_br;
2875 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2876 "disabling this process (pid %ld) until it goes away",
2877 (long int) getpid());
2879 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2882 /* Since we will not be running system_stats_run() in this process
2883 * with the current situation of multiple ovs-vswitchd daemons,
2884 * disable system stats collection. */
2885 system_stats_enable(false);
2887 } else if (!ovsdb_idl_has_lock(idl)) {
2890 cfg = ovsrec_open_vswitch_first(idl);
2892 /* Initialize the ofproto library. This only needs to run once, but
2893 * it must be done after the configuration is set. If the
2894 * initialization has already occurred, bridge_init_ofproto()
2895 * returns immediately. */
2896 bridge_init_ofproto(cfg);
2898 /* Once the value of flow-restore-wait is false, we no longer should
2899 * check its value from the database. */
2900 if (cfg && ofproto_get_flow_restore_wait()) {
2901 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2902 "flow-restore-wait", false));
2907 /* Re-configure SSL. We do this on every trip through the main loop,
2908 * instead of just when the database changes, because the contents of the
2909 * key and certificate files can change without the database changing.
2911 * We do this before bridge_reconfigure() because that function might
2912 * initiate SSL connections and thus requires SSL to be configured. */
2913 if (cfg && cfg->ssl) {
2914 const struct ovsrec_ssl *ssl = cfg->ssl;
2916 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2917 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2920 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2921 * usage has changed. */
2922 vlan_splinters_changed = false;
2923 if (vlan_splinters_enabled_anywhere) {
2926 HMAP_FOR_EACH (br, node, &all_bridges) {
2927 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2928 vlan_splinters_changed = true;
2934 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2935 struct ovsdb_idl_txn *txn;
2937 idl_seqno = ovsdb_idl_get_seqno(idl);
2938 txn = ovsdb_idl_txn_create(idl);
2939 bridge_reconfigure(cfg ? cfg : &null_cfg);
2942 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2945 /* If we are completing our initial configuration for this run
2946 * of ovs-vswitchd, then keep the transaction around to monitor
2947 * it for completion. */
2948 if (initial_config_done) {
2949 /* Always sets the 'status_txn_try_again' to check again,
2950 * in case that this transaction fails. */
2951 status_txn_try_again = true;
2952 ovsdb_idl_txn_commit(txn);
2953 ovsdb_idl_txn_destroy(txn);
2955 initial_config_done = true;
2956 daemonize_txn = txn;
2960 if (daemonize_txn) {
2961 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2962 if (status != TXN_INCOMPLETE) {
2963 ovsdb_idl_txn_destroy(daemonize_txn);
2964 daemonize_txn = NULL;
2966 /* ovs-vswitchd has completed initialization, so allow the
2967 * process that forked us to exit successfully. */
2968 daemonize_complete();
2970 vlog_enable_async();
2972 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2977 run_status_update();
2987 ovsdb_idl_wait(idl);
2988 if (daemonize_txn) {
2989 ovsdb_idl_txn_wait(daemonize_txn);
2993 ofproto_enumerate_types(&types);
2994 SSET_FOR_EACH (type, &types) {
2995 ofproto_type_wait(type);
2997 sset_destroy(&types);
2999 if (!hmap_is_empty(&all_bridges)) {
3002 HMAP_FOR_EACH (br, node, &all_bridges) {
3003 ofproto_wait(br->ofproto);
3006 poll_timer_wait_until(stats_timer);
3009 status_update_wait();
3010 system_stats_wait();
3013 /* Adds some memory usage statistics for bridges into 'usage', for use with
3014 * memory_report(). */
3016 bridge_get_memory_usage(struct simap *usage)
3023 ofproto_enumerate_types(&types);
3024 SSET_FOR_EACH (type, &types) {
3025 ofproto_type_get_memory_usage(type, usage);
3027 sset_destroy(&types);
3029 HMAP_FOR_EACH (br, node, &all_bridges) {
3030 ofproto_get_memory_usage(br->ofproto, usage);
3034 /* QoS unixctl user interface functions. */
3036 struct qos_unixctl_show_cbdata {
3038 struct iface *iface;
3042 qos_unixctl_show_queue(unsigned int queue_id,
3043 const struct smap *details,
3044 struct iface *iface,
3047 struct netdev_queue_stats stats;
3048 struct smap_node *node;
3051 ds_put_cstr(ds, "\n");
3053 ds_put_format(ds, "Queue %u:\n", queue_id);
3055 ds_put_cstr(ds, "Default:\n");
3058 SMAP_FOR_EACH (node, details) {
3059 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3062 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3064 if (stats.tx_packets != UINT64_MAX) {
3065 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3068 if (stats.tx_bytes != UINT64_MAX) {
3069 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3072 if (stats.tx_errors != UINT64_MAX) {
3073 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3076 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3077 queue_id, ovs_strerror(error));
3082 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3083 const char *argv[], void *aux OVS_UNUSED)
3085 struct ds ds = DS_EMPTY_INITIALIZER;
3086 struct smap smap = SMAP_INITIALIZER(&smap);
3087 struct iface *iface;
3089 struct smap_node *node;
3091 iface = iface_find(argv[1]);
3093 unixctl_command_reply_error(conn, "no such interface");
3097 netdev_get_qos(iface->netdev, &type, &smap);
3099 if (*type != '\0') {
3100 struct netdev_queue_dump dump;
3101 struct smap details;
3102 unsigned int queue_id;
3104 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3106 SMAP_FOR_EACH (node, &smap) {
3107 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3110 smap_init(&details);
3111 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3112 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3114 smap_destroy(&details);
3116 unixctl_command_reply(conn, ds_cstr(&ds));
3118 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3119 unixctl_command_reply_error(conn, ds_cstr(&ds));
3122 smap_destroy(&smap);
3126 /* Bridge reconfiguration functions. */
3128 bridge_create(const struct ovsrec_bridge *br_cfg)
3132 ovs_assert(!bridge_lookup(br_cfg->name));
3133 br = xzalloc(sizeof *br);
3135 br->name = xstrdup(br_cfg->name);
3136 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3139 /* Derive the default Ethernet address from the bridge's UUID. This should
3140 * be unique and it will be stable between ovs-vswitchd runs. */
3141 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3142 eth_addr_mark_random(br->default_ea);
3144 hmap_init(&br->ports);
3145 hmap_init(&br->ifaces);
3146 hmap_init(&br->iface_by_name);
3147 hmap_init(&br->mirrors);
3149 hmap_init(&br->mappings);
3150 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3154 bridge_destroy(struct bridge *br)
3157 struct mirror *mirror, *next_mirror;
3158 struct port *port, *next_port;
3160 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3163 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3164 mirror_destroy(mirror);
3167 hmap_remove(&all_bridges, &br->node);
3168 ofproto_destroy(br->ofproto);
3169 hmap_destroy(&br->ifaces);
3170 hmap_destroy(&br->ports);
3171 hmap_destroy(&br->iface_by_name);
3172 hmap_destroy(&br->mirrors);
3173 hmap_destroy(&br->mappings);
3180 static struct bridge *
3181 bridge_lookup(const char *name)
3185 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3186 if (!strcmp(br->name, name)) {
3193 /* Handle requests for a listing of all flows known by the OpenFlow
3194 * stack, including those normally hidden. */
3196 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3197 const char *argv[], void *aux OVS_UNUSED)
3202 br = bridge_lookup(argv[1]);
3204 unixctl_command_reply_error(conn, "Unknown bridge");
3209 ofproto_get_all_flows(br->ofproto, &results);
3211 unixctl_command_reply(conn, ds_cstr(&results));
3212 ds_destroy(&results);
3215 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3216 * connections and reconnect. If BRIDGE is not specified, then all bridges
3217 * drop their controller connections and reconnect. */
3219 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3220 const char *argv[], void *aux OVS_UNUSED)
3224 br = bridge_lookup(argv[1]);
3226 unixctl_command_reply_error(conn, "Unknown bridge");
3229 ofproto_reconnect_controllers(br->ofproto);
3231 HMAP_FOR_EACH (br, node, &all_bridges) {
3232 ofproto_reconnect_controllers(br->ofproto);
3235 unixctl_command_reply(conn, NULL);
3239 bridge_get_controllers(const struct bridge *br,
3240 struct ovsrec_controller ***controllersp)
3242 struct ovsrec_controller **controllers;
3243 size_t n_controllers;
3245 controllers = br->cfg->controller;
3246 n_controllers = br->cfg->n_controller;
3248 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3254 *controllersp = controllers;
3256 return n_controllers;
3260 bridge_collect_wanted_ports(struct bridge *br,
3261 const unsigned long int *splinter_vlans,
3262 struct shash *wanted_ports)
3266 shash_init(wanted_ports);
3268 for (i = 0; i < br->cfg->n_ports; i++) {
3269 const char *name = br->cfg->ports[i]->name;
3270 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3271 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3275 if (bridge_get_controllers(br, NULL)
3276 && !shash_find(wanted_ports, br->name)) {
3277 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3278 br->name, br->name);
3280 ovsrec_interface_init(&br->synth_local_iface);
3281 ovsrec_port_init(&br->synth_local_port);
3283 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3284 br->synth_local_port.n_interfaces = 1;
3285 br->synth_local_port.name = br->name;
3287 br->synth_local_iface.name = br->name;
3288 br->synth_local_iface.type = "internal";
3290 br->synth_local_ifacep = &br->synth_local_iface;
3292 shash_add(wanted_ports, br->name, &br->synth_local_port);
3295 if (splinter_vlans) {
3296 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3300 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3301 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3302 * 'br' needs to complete its configuration. */
3304 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3306 struct shash_node *port_node;
3307 struct port *port, *next;
3309 /* Get rid of deleted ports.
3310 * Get rid of deleted interfaces on ports that still exist. */
3311 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3312 port->cfg = shash_find_data(wanted_ports, port->name);
3316 port_del_ifaces(port);
3320 /* Update iface->cfg and iface->type in interfaces that still exist. */
3321 SHASH_FOR_EACH (port_node, wanted_ports) {
3322 const struct ovsrec_port *port = port_node->data;
3325 for (i = 0; i < port->n_interfaces; i++) {
3326 const struct ovsrec_interface *cfg = port->interfaces[i];
3327 struct iface *iface = iface_lookup(br, cfg->name);
3328 const char *type = iface_get_type(cfg, br->cfg);
3333 } else if (!strcmp(type, "null")) {
3334 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3335 " may be removed in February 2013. Please email"
3336 " dev@openvswitch.org with concerns.",
3339 /* We will add new interfaces later. */
3345 /* Initializes 'oc' appropriately as a management service controller for
3348 * The caller must free oc->target when it is no longer needed. */
3350 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3351 struct ofproto_controller *oc)
3353 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3354 oc->max_backoff = 0;
3355 oc->probe_interval = 60;
3356 oc->band = OFPROTO_OUT_OF_BAND;
3358 oc->burst_limit = 0;
3359 oc->enable_async_msgs = true;
3363 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3365 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3366 struct ofproto_controller *oc)
3370 oc->target = c->target;
3371 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3372 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3373 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3374 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3375 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3376 oc->burst_limit = (c->controller_burst_limit
3377 ? *c->controller_burst_limit : 0);
3378 oc->enable_async_msgs = (!c->enable_async_messages
3379 || *c->enable_async_messages);
3380 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3381 if (dscp < 0 || dscp > 63) {
3382 dscp = DSCP_DEFAULT;
3387 /* Configures the IP stack for 'br''s local interface properly according to the
3388 * configuration in 'c'. */
3390 bridge_configure_local_iface_netdev(struct bridge *br,
3391 struct ovsrec_controller *c)
3393 struct netdev *netdev;
3394 struct in_addr mask, gateway;
3396 struct iface *local_iface;
3399 /* If there's no local interface or no IP address, give up. */
3400 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3401 if (!local_iface || !c->local_ip
3402 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3406 /* Bring up the local interface. */
3407 netdev = local_iface->netdev;
3408 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3410 /* Configure the IP address and netmask. */
3411 if (!c->local_netmask
3412 || !inet_pton(AF_INET, c->local_netmask, &mask)
3414 mask.s_addr = guess_netmask(ip.s_addr);
3416 if (!netdev_set_in4(netdev, ip, mask)) {
3417 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3418 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3421 /* Configure the default gateway. */
3422 if (c->local_gateway
3423 && inet_pton(AF_INET, c->local_gateway, &gateway)
3424 && gateway.s_addr) {
3425 if (!netdev_add_router(netdev, gateway)) {
3426 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3427 br->name, IP_ARGS(gateway.s_addr));
3432 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3433 * in either string are treated as equal to any number of slashes in the other,
3434 * e.g. "x///y" is equal to "x/y".
3436 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3437 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3438 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3439 * 'b' against a prefix of 'a'.
3442 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3444 const char *b_start = b;
3446 if (b - b_start >= b_stoplen) {
3448 } else if (*a != *b) {
3450 } else if (*a == '/') {
3451 a += strspn(a, "/");
3452 b += strspn(b, "/");
3453 } else if (*a == '\0') {
3463 bridge_configure_remotes(struct bridge *br,
3464 const struct sockaddr_in *managers, size_t n_managers)
3466 bool disable_in_band;
3468 struct ovsrec_controller **controllers;
3469 size_t n_controllers;
3471 enum ofproto_fail_mode fail_mode;
3473 struct ofproto_controller *ocs;
3477 /* Check if we should disable in-band control on this bridge. */
3478 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3481 /* Set OpenFlow queue ID for in-band control. */
3482 ofproto_set_in_band_queue(br->ofproto,
3483 smap_get_int(&br->cfg->other_config,
3484 "in-band-queue", -1));
3486 if (disable_in_band) {
3487 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3489 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3492 n_controllers = bridge_get_controllers(br, &controllers);
3494 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3497 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3498 for (i = 0; i < n_controllers; i++) {
3499 struct ovsrec_controller *c = controllers[i];
3501 if (!strncmp(c->target, "punix:", 6)
3502 || !strncmp(c->target, "unix:", 5)) {
3503 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3506 if (!strncmp(c->target, "unix:", 5)) {
3507 /* Connect to a listening socket */
3508 whitelist = xasprintf("unix:%s/", ovs_rundir());
3509 if (strchr(c->target, '/') &&
3510 !equal_pathnames(c->target, whitelist,
3511 strlen(whitelist))) {
3512 /* Absolute path specified, but not in ovs_rundir */
3513 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3514 "controller \"%s\" due to possibility for "
3515 "remote exploit. Instead, specify socket "
3516 "in whitelisted \"%s\" or connect to "
3517 "\"unix:%s/%s.mgmt\" (which is always "
3518 "available without special configuration).",
3519 br->name, c->target, whitelist,
3520 ovs_rundir(), br->name);
3525 whitelist = xasprintf("punix:%s/%s.controller",
3526 ovs_rundir(), br->name);
3527 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3528 /* Prevent remote ovsdb-server users from accessing
3529 * arbitrary Unix domain sockets and overwriting arbitrary
3531 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3532 "controller \"%s\" due to possibility of "
3533 "overwriting local files. Instead, specify "
3534 "whitelisted \"%s\" or connect to "
3535 "\"unix:%s/%s.mgmt\" (which is always "
3536 "available without special configuration).",
3537 br->name, c->target, whitelist,
3538 ovs_rundir(), br->name);
3547 bridge_configure_local_iface_netdev(br, c);
3548 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3549 if (disable_in_band) {
3550 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3555 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3556 bridge_get_allowed_versions(br));
3557 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3560 /* Set the fail-mode. */
3561 fail_mode = !br->cfg->fail_mode
3562 || !strcmp(br->cfg->fail_mode, "standalone")
3563 ? OFPROTO_FAIL_STANDALONE
3564 : OFPROTO_FAIL_SECURE;
3565 ofproto_set_fail_mode(br->ofproto, fail_mode);
3567 /* Configure OpenFlow controller connection snooping. */
3568 if (!ofproto_has_snoops(br->ofproto)) {
3572 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3573 ovs_rundir(), br->name));
3574 ofproto_set_snoops(br->ofproto, &snoops);
3575 sset_destroy(&snoops);
3580 bridge_configure_tables(struct bridge *br)
3582 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3586 n_tables = ofproto_get_n_tables(br->ofproto);
3588 for (i = 0; i < n_tables; i++) {
3589 struct ofproto_table_settings s;
3590 bool use_default_prefixes = true;
3593 s.max_flows = UINT_MAX;
3596 s.n_prefix_fields = 0;
3597 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3599 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3600 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3603 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3604 s.max_flows = *cfg->flow_limit;
3606 if (cfg->overflow_policy
3607 && !strcmp(cfg->overflow_policy, "evict")) {
3609 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3610 for (k = 0; k < cfg->n_groups; k++) {
3611 const char *string = cfg->groups[k];
3614 msg = mf_parse_subfield__(&s.groups[k], &string);
3616 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3617 "'groups' (%s)", br->name, i, msg);
3619 } else if (*string) {
3620 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3621 "element '%s' contains trailing garbage",
3622 br->name, i, cfg->groups[k]);
3628 /* Prefix lookup fields. */
3629 s.n_prefix_fields = 0;
3630 for (k = 0; k < cfg->n_prefixes; k++) {
3631 const char *name = cfg->prefixes[k];
3632 const struct mf_field *mf;
3634 if (strcmp(name, "none") == 0) {
3635 use_default_prefixes = false;
3636 s.n_prefix_fields = 0;
3639 mf = mf_from_name(name);
3641 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3645 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3646 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3647 "%s", br->name, name);
3650 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3651 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3652 "field not used: %s", br->name, name);
3655 use_default_prefixes = false;
3656 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3659 if (use_default_prefixes) {
3660 /* Use default values. */
3661 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3662 memcpy(s.prefix_fields, default_prefix_fields,
3663 sizeof default_prefix_fields);
3666 struct ds ds = DS_EMPTY_INITIALIZER;
3667 for (k = 0; k < s.n_prefix_fields; k++) {
3669 ds_put_char(&ds, ',');
3671 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3673 if (s.n_prefix_fields == 0) {
3674 ds_put_cstr(&ds, "none");
3676 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3677 br->name, i, ds_cstr(&ds));
3681 ofproto_configure_table(br->ofproto, i, &s);
3685 for (; j < br->cfg->n_flow_tables; j++) {
3686 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3687 "%"PRId64" not supported by this datapath", br->name,
3688 br->cfg->key_flow_tables[j]);
3693 bridge_configure_dp_desc(struct bridge *br)
3695 ofproto_set_dp_desc(br->ofproto,
3696 smap_get(&br->cfg->other_config, "dp-desc"));
3699 static struct aa_mapping *
3700 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3702 struct aa_mapping *m;
3704 HMAP_FOR_EACH_IN_BUCKET (m,
3706 hash_bytes(&isid, sizeof isid, 0),
3708 if (isid == m->isid) {
3715 static struct aa_mapping *
3716 bridge_aa_mapping_create(struct bridge *br,
3720 struct aa_mapping *m;
3722 m = xzalloc(sizeof *m);
3726 m->br_name = xstrdup(br->name);
3727 hmap_insert(&br->mappings,
3729 hash_bytes(&isid, sizeof isid, 0));
3735 bridge_aa_mapping_destroy(struct aa_mapping *m)
3738 struct bridge *br = m->bridge;
3741 ofproto_aa_mapping_unregister(br->ofproto, m);
3744 hmap_remove(&br->mappings, &m->hmap_node);
3753 bridge_aa_mapping_configure(struct aa_mapping *m)
3755 struct aa_mapping_settings s;
3761 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3767 bridge_configure_aa(struct bridge *br)
3769 const struct ovsdb_datum *mc;
3770 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3771 struct aa_settings aa_s;
3772 struct aa_mapping *m, *next;
3776 ofproto_set_aa(br->ofproto, NULL, NULL);
3780 memset(&aa_s, 0, sizeof aa_s);
3781 aa_s.system_description = auto_attach->system_description;
3782 aa_s.system_name = auto_attach->system_name;
3783 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3785 mc = ovsrec_autoattach_get_mappings(auto_attach,
3787 OVSDB_TYPE_INTEGER);
3788 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3789 union ovsdb_atom atom;
3791 atom.integer = m->isid;
3792 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3793 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3795 bridge_aa_mapping_destroy(m);
3799 /* Add new mappings and reconfigure existing ones. */
3800 for (i = 0; i < auto_attach->n_mappings; ++i) {
3801 struct aa_mapping *m =
3802 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3805 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3806 auto_attach->key_mappings[i],
3807 auto_attach->value_mappings[i]);
3808 m = bridge_aa_mapping_create(br,
3809 auto_attach->key_mappings[i],
3810 auto_attach->value_mappings[i]);
3812 if (!bridge_aa_mapping_configure(m)) {
3813 bridge_aa_mapping_destroy(m);
3820 bridge_aa_need_refresh(struct bridge *br)
3822 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3826 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3828 int64_t *trunks = NULL;
3830 bool found = false, reconfigure = false;
3832 for (i = 0; i < port->cfg->n_trunks; i++) {
3833 if (port->cfg->trunks[i] == m->vlan) {
3840 case BRIDGE_AA_VLAN_OPER_ADD:
3842 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3844 for (i = 0; i < port->cfg->n_trunks; i++) {
3845 trunks[i] = port->cfg->trunks[i];
3847 trunks[i++] = m->vlan;
3853 case BRIDGE_AA_VLAN_OPER_REMOVE:
3857 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3859 for (i = 0; i < port->cfg->n_trunks; i++) {
3860 if (port->cfg->trunks[i] != m->vlan) {
3861 trunks[j++] = port->cfg->trunks[i];
3870 case BRIDGE_AA_VLAN_OPER_UNDEF:
3872 VLOG_WARN("unrecognized operation %u", m->oper);
3877 /* VLAN switching under trunk mode cause the trunk port to switch all
3878 * VLANs, see ovs-vswitchd.conf.db
3881 static char *vlan_mode_access = "access";
3882 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3886 static char *vlan_mode_trunk = "trunk";
3887 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3890 ovsrec_port_set_trunks(port->cfg, trunks, i);
3892 /* Force reconfigure of the port. */
3893 port_configure(port);
3898 bridge_aa_refresh_queued(struct bridge *br)
3900 struct ovs_list *list = xmalloc(sizeof *list);
3901 struct bridge_aa_vlan *node;
3904 ofproto_aa_vlan_get_queued(br->ofproto, list);
3906 LIST_FOR_EACH(node, list_node, list) {
3909 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3912 port = port_lookup(br, node->port_name);
3914 bridge_aa_update_trunks(port, node);
3917 list_remove(&node->list_node);
3918 free(node->port_name);
3926 /* Port functions. */
3928 static struct port *
3929 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3933 port = xzalloc(sizeof *port);
3935 port->name = xstrdup(cfg->name);
3937 list_init(&port->ifaces);
3939 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3943 /* Deletes interfaces from 'port' that are no longer configured for it. */
3945 port_del_ifaces(struct port *port)
3947 struct iface *iface, *next;
3948 struct sset new_ifaces;
3951 /* Collect list of new interfaces. */
3952 sset_init(&new_ifaces);
3953 for (i = 0; i < port->cfg->n_interfaces; i++) {
3954 const char *name = port->cfg->interfaces[i]->name;
3955 const char *type = port->cfg->interfaces[i]->type;
3956 if (strcmp(type, "null")) {
3957 sset_add(&new_ifaces, name);
3961 /* Get rid of deleted interfaces. */
3962 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3963 if (!sset_contains(&new_ifaces, iface->name)) {
3964 iface_destroy(iface);
3968 sset_destroy(&new_ifaces);
3972 port_destroy(struct port *port)
3975 struct bridge *br = port->bridge;
3976 struct iface *iface, *next;
3979 ofproto_bundle_unregister(br->ofproto, port);
3982 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3983 iface_destroy__(iface);
3986 hmap_remove(&br->ports, &port->hmap_node);
3992 static struct port *
3993 port_lookup(const struct bridge *br, const char *name)
3997 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3999 if (!strcmp(port->name, name)) {
4007 enable_lacp(struct port *port, bool *activep)
4009 if (!port->cfg->lacp) {
4010 /* XXX when LACP implementation has been sufficiently tested, enable by
4011 * default and make active on bonded ports. */
4013 } else if (!strcmp(port->cfg->lacp, "off")) {
4015 } else if (!strcmp(port->cfg->lacp, "active")) {
4018 } else if (!strcmp(port->cfg->lacp, "passive")) {
4022 VLOG_WARN("port %s: unknown LACP mode %s",
4023 port->name, port->cfg->lacp);
4028 static struct lacp_settings *
4029 port_configure_lacp(struct port *port, struct lacp_settings *s)
4031 const char *lacp_time, *system_id;
4034 if (!enable_lacp(port, &s->active)) {
4038 s->name = port->name;
4040 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4042 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4043 ETH_ADDR_SCAN_ARGS(s->id))) {
4044 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4045 " address.", port->name, system_id);
4049 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4052 if (eth_addr_is_zero(s->id)) {
4053 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4057 /* Prefer bondable links if unspecified. */
4058 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4060 s->priority = (priority > 0 && priority <= UINT16_MAX
4062 : UINT16_MAX - !list_is_short(&port->ifaces));
4064 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4065 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4067 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4068 "lacp-fallback-ab", false);
4074 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4076 int priority, portid, key;
4078 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4079 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4081 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4083 if (portid <= 0 || portid > UINT16_MAX) {
4084 portid = ofp_to_u16(iface->ofp_port);
4087 if (priority <= 0 || priority > UINT16_MAX) {
4088 priority = UINT16_MAX;
4091 if (key < 0 || key > UINT16_MAX) {
4095 s->name = iface->name;
4097 s->priority = priority;
4102 port_configure_bond(struct port *port, struct bond_settings *s)
4104 const char *detect_s;
4105 struct iface *iface;
4107 int miimon_interval;
4109 s->name = port->name;
4111 if (port->cfg->bond_mode) {
4112 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4113 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4114 port->name, port->cfg->bond_mode,
4115 bond_mode_to_string(s->balance));
4118 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4120 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4121 * active-backup. At some point we should remove this warning. */
4122 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4123 " in previous versions, the default bond_mode was"
4124 " balance-slb", port->name,
4125 bond_mode_to_string(s->balance));
4127 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4128 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4129 "please use another bond type or disable flood_vlans",
4133 miimon_interval = smap_get_int(&port->cfg->other_config,
4134 "bond-miimon-interval", 0);
4135 if (miimon_interval <= 0) {
4136 miimon_interval = 200;
4139 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4140 if (!detect_s || !strcmp(detect_s, "carrier")) {
4141 miimon_interval = 0;
4142 } else if (strcmp(detect_s, "miimon")) {
4143 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4144 "defaulting to carrier", port->name, detect_s);
4145 miimon_interval = 0;
4148 s->up_delay = MAX(0, port->cfg->bond_updelay);
4149 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4150 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4151 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4152 "bond-rebalance-interval", 10000);
4153 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4154 s->rebalance_interval = 1000;
4157 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4158 "lacp-fallback-ab", false);
4160 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4161 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4164 mac_s = port->cfg->bond_active_slave;
4165 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4166 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4167 /* OVSDB did not store the last active interface */
4168 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4172 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4173 * instead of obtaining it from the database. */
4175 port_is_synthetic(const struct port *port)
4177 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4180 /* Interface functions. */
4183 iface_is_internal(const struct ovsrec_interface *iface,
4184 const struct ovsrec_bridge *br)
4186 /* The local port and "internal" ports are always "internal". */
4187 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4190 /* Returns the correct network device type for interface 'iface' in bridge
4193 iface_get_type(const struct ovsrec_interface *iface,
4194 const struct ovsrec_bridge *br)
4198 /* The local port always has type "internal". Other ports take
4199 * their type from the database and default to "system" if none is
4201 if (iface_is_internal(iface, br)) {
4204 type = iface->type[0] ? iface->type : "system";
4207 return ofproto_port_open_type(br->datapath_type, type);
4211 iface_destroy__(struct iface *iface)
4214 struct port *port = iface->port;
4215 struct bridge *br = port->bridge;
4217 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4218 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4221 if (iface->ofp_port != OFPP_NONE) {
4222 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4225 list_remove(&iface->port_elem);
4226 hmap_remove(&br->iface_by_name, &iface->name_node);
4228 /* The user is changing configuration here, so netdev_remove needs to be
4229 * used as opposed to netdev_close */
4230 netdev_remove(iface->netdev);
4238 iface_destroy(struct iface *iface)
4241 struct port *port = iface->port;
4243 iface_destroy__(iface);
4244 if (list_is_empty(&port->ifaces)) {
4250 static struct iface *
4251 iface_lookup(const struct bridge *br, const char *name)
4253 struct iface *iface;
4255 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4256 &br->iface_by_name) {
4257 if (!strcmp(iface->name, name)) {
4265 static struct iface *
4266 iface_find(const char *name)
4268 const struct bridge *br;
4270 HMAP_FOR_EACH (br, node, &all_bridges) {
4271 struct iface *iface = iface_lookup(br, name);
4280 static struct iface *
4281 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4283 struct iface *iface;
4285 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4287 if (iface->ofp_port == ofp_port) {
4294 /* Set Ethernet address of 'iface', if one is specified in the configuration
4297 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4299 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4300 struct iface *hw_addr_iface;
4302 if (strcmp(iface->type, "internal")) {
4306 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4308 } else if (port->cfg->fake_bridge) {
4309 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4310 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4315 if (iface->ofp_port == OFPP_LOCAL) {
4316 VLOG_ERR("interface %s: ignoring mac in Interface record "
4317 "(use Bridge record to set local port's mac)",
4319 } else if (eth_addr_is_multicast(mac)) {
4320 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4323 int error = netdev_set_etheraddr(iface->netdev, mac);
4325 VLOG_ERR("interface %s: setting MAC failed (%s)",
4326 iface->name, ovs_strerror(error));
4332 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4334 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4336 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4337 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4338 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4342 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4343 * sets the "ofport" field to -1.
4345 * This is appropriate when 'if_cfg''s interface cannot be created or is
4346 * otherwise invalid. */
4348 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4350 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4351 iface_set_ofport(if_cfg, OFPP_NONE);
4352 ovsrec_interface_set_error(if_cfg, errp);
4353 ovsrec_interface_set_status(if_cfg, NULL);
4354 ovsrec_interface_set_admin_state(if_cfg, NULL);
4355 ovsrec_interface_set_duplex(if_cfg, NULL);
4356 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4357 ovsrec_interface_set_link_state(if_cfg, NULL);
4358 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4359 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4360 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4361 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4362 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4363 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4364 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4365 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4370 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4372 union ovsdb_atom atom;
4374 atom.integer = target;
4375 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4379 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4381 struct ofpbuf queues_buf;
4383 ofpbuf_init(&queues_buf, 0);
4385 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
4386 netdev_set_qos(iface->netdev, NULL, NULL);
4388 const struct ovsdb_datum *queues;
4389 struct netdev_queue_dump dump;
4390 unsigned int queue_id;
4391 struct smap details;
4395 /* Configure top-level Qos for 'iface'. */
4396 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4398 /* Deconfigure queues that were deleted. */
4399 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4401 smap_init(&details);
4402 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4403 if (!queue_ids_include(queues, queue_id)) {
4404 netdev_delete_queue(iface->netdev, queue_id);
4407 smap_destroy(&details);
4409 /* Configure queues for 'iface'. */
4411 for (i = 0; i < qos->n_queues; i++) {
4412 const struct ovsrec_queue *queue = qos->value_queues[i];
4413 unsigned int queue_id = qos->key_queues[i];
4415 if (queue_id == 0) {
4419 if (queue->n_dscp == 1) {
4420 struct ofproto_port_queue *port_queue;
4422 port_queue = ofpbuf_put_uninit(&queues_buf,
4423 sizeof *port_queue);
4424 port_queue->queue = queue_id;
4425 port_queue->dscp = queue->dscp[0];
4428 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4431 struct smap details;
4433 smap_init(&details);
4434 netdev_set_queue(iface->netdev, 0, &details);
4435 smap_destroy(&details);
4439 if (iface->ofp_port != OFPP_NONE) {
4440 const struct ofproto_port_queue *port_queues = queues_buf.data;
4441 size_t n_queues = queues_buf.size / sizeof *port_queues;
4443 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4444 port_queues, n_queues);
4447 netdev_set_policing(iface->netdev,
4448 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4449 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4451 ofpbuf_uninit(&queues_buf);
4455 iface_configure_cfm(struct iface *iface)
4457 const struct ovsrec_interface *cfg = iface->cfg;
4458 const char *opstate_str;
4459 const char *cfm_ccm_vlan;
4460 struct cfm_settings s;
4461 struct smap netdev_args;
4463 if (!cfg->n_cfm_mpid) {
4464 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4468 s.check_tnl_key = false;
4469 smap_init(&netdev_args);
4470 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4471 const char *key = smap_get(&netdev_args, "key");
4472 const char *in_key = smap_get(&netdev_args, "in_key");
4474 s.check_tnl_key = (key && !strcmp(key, "flow"))
4475 || (in_key && !strcmp(in_key, "flow"));
4477 smap_destroy(&netdev_args);
4479 s.mpid = *cfg->cfm_mpid;
4480 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4481 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4482 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4484 if (s.interval <= 0) {
4488 if (!cfm_ccm_vlan) {
4490 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4491 s.ccm_vlan = CFM_RANDOM_VLAN;
4493 s.ccm_vlan = atoi(cfm_ccm_vlan);
4494 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4499 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4501 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4503 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4504 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4506 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4509 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4510 * instead of obtaining it from the database. */
4512 iface_is_synthetic(const struct iface *iface)
4514 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4518 iface_validate_ofport__(size_t n, int64_t *ofport)
4520 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4521 ? u16_to_ofp(*ofport)
4526 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4528 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4532 iface_pick_ofport(const struct ovsrec_interface *cfg)
4534 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4535 return (requested_ofport != OFPP_NONE
4537 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4540 /* Port mirroring. */
4542 static struct mirror *
4543 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4547 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4548 if (uuid_equals(uuid, &m->uuid)) {
4556 bridge_configure_mirrors(struct bridge *br)
4558 const struct ovsdb_datum *mc;
4559 unsigned long *flood_vlans;
4560 struct mirror *m, *next;
4563 /* Get rid of deleted mirrors. */
4564 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4565 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4566 union ovsdb_atom atom;
4568 atom.uuid = m->uuid;
4569 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4574 /* Add new mirrors and reconfigure existing ones. */
4575 for (i = 0; i < br->cfg->n_mirrors; i++) {
4576 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4577 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4579 m = mirror_create(br, cfg);
4582 if (!mirror_configure(m)) {
4587 /* Update flooded vlans (for RSPAN). */
4588 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4589 br->cfg->n_flood_vlans);
4590 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4591 bitmap_free(flood_vlans);
4594 static struct mirror *
4595 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4599 m = xzalloc(sizeof *m);
4600 m->uuid = cfg->header_.uuid;
4601 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4603 m->name = xstrdup(cfg->name);
4609 mirror_destroy(struct mirror *m)
4612 struct bridge *br = m->bridge;
4615 ofproto_mirror_unregister(br->ofproto, m);
4618 hmap_remove(&br->mirrors, &m->hmap_node);
4625 mirror_collect_ports(struct mirror *m,
4626 struct ovsrec_port **in_ports, int n_in_ports,
4627 void ***out_portsp, size_t *n_out_portsp)
4629 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4630 size_t n_out_ports = 0;
4633 for (i = 0; i < n_in_ports; i++) {
4634 const char *name = in_ports[i]->name;
4635 struct port *port = port_lookup(m->bridge, name);
4637 out_ports[n_out_ports++] = port;
4639 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4640 "port %s", m->bridge->name, m->name, name);
4643 *out_portsp = out_ports;
4644 *n_out_portsp = n_out_ports;
4648 mirror_configure(struct mirror *m)
4650 const struct ovsrec_mirror *cfg = m->cfg;
4651 struct ofproto_mirror_settings s;
4654 if (strcmp(cfg->name, m->name)) {
4656 m->name = xstrdup(cfg->name);
4660 /* Get output port or VLAN. */
4661 if (cfg->output_port) {
4662 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4663 if (!s.out_bundle) {
4664 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4665 m->bridge->name, m->name);
4668 s.out_vlan = UINT16_MAX;
4670 if (cfg->output_vlan) {
4671 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4672 "output vlan; ignoring output vlan",
4673 m->bridge->name, m->name);
4675 } else if (cfg->output_vlan) {
4676 /* The database should prevent invalid VLAN values. */
4677 s.out_bundle = NULL;
4678 s.out_vlan = *cfg->output_vlan;
4680 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4681 m->bridge->name, m->name);
4685 /* Get port selection. */
4686 if (cfg->select_all) {
4687 size_t n_ports = hmap_count(&m->bridge->ports);
4688 void **ports = xmalloc(n_ports * sizeof *ports);
4693 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4703 /* Get ports, dropping ports that don't exist.
4704 * The IDL ensures that there are no duplicates. */
4705 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4706 &s.srcs, &s.n_srcs);
4707 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4708 &s.dsts, &s.n_dsts);
4711 /* Get VLAN selection. */
4712 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4715 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4718 if (s.srcs != s.dsts) {
4727 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4729 * This is deprecated. It is only for compatibility with broken device drivers
4730 * in old versions of Linux that do not properly support VLANs when VLAN
4731 * devices are not used. When broken device drivers are no longer in
4732 * widespread use, we will delete these interfaces. */
4734 static struct ovsrec_port **recs;
4735 static size_t n_recs, allocated_recs;
4737 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4738 * splinters are reconfigured. */
4740 register_rec(struct ovsrec_port *rec)
4742 if (n_recs >= allocated_recs) {
4743 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4745 recs[n_recs++] = rec;
4748 /* Frees all of the ports registered with register_reg(). */
4750 free_registered_recs(void)
4754 for (i = 0; i < n_recs; i++) {
4755 struct ovsrec_port *port = recs[i];
4758 for (j = 0; j < port->n_interfaces; j++) {
4759 struct ovsrec_interface *iface = port->interfaces[j];
4764 smap_destroy(&port->other_config);
4765 free(port->interfaces);
4773 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4776 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4778 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4782 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4785 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4786 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4787 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4790 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4791 * use, returns NULL.
4793 * Updates 'vlan_splinters_enabled_anywhere'. */
4794 static unsigned long int *
4795 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4797 unsigned long int *splinter_vlans;
4798 struct sset splinter_ifaces;
4799 const char *real_dev_name;
4800 struct shash *real_devs;
4801 struct shash_node *node;
4805 /* Free space allocated for synthesized ports and interfaces, since we're
4806 * in the process of reconstructing all of them. */
4807 free_registered_recs();
4809 splinter_vlans = bitmap_allocate(4096);
4810 sset_init(&splinter_ifaces);
4811 vlan_splinters_enabled_anywhere = false;
4812 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4813 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4816 for (j = 0; j < br_cfg->n_ports; j++) {
4817 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4820 for (k = 0; k < port_cfg->n_interfaces; k++) {
4821 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4823 if (vlan_splinters_is_enabled(iface_cfg)) {
4824 vlan_splinters_enabled_anywhere = true;
4825 sset_add(&splinter_ifaces, iface_cfg->name);
4826 vlan_bitmap_from_array__(port_cfg->trunks,
4832 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4833 bitmap_set1(splinter_vlans, *port_cfg->tag);
4838 if (!vlan_splinters_enabled_anywhere) {
4839 free(splinter_vlans);
4840 sset_destroy(&splinter_ifaces);
4844 HMAP_FOR_EACH (br, node, &all_bridges) {
4846 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4850 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4851 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4852 * device to be created for it. */
4853 bitmap_set0(splinter_vlans, 0);
4854 bitmap_set0(splinter_vlans, 4095);
4856 /* Delete all VLAN devices that we don't need. */
4858 real_devs = vlandev_get_real_devs();
4859 SHASH_FOR_EACH (node, real_devs) {
4860 const struct vlan_real_dev *real_dev = node->data;
4861 const struct vlan_dev *vlan_dev;
4862 bool real_dev_has_splinters;
4864 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4866 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4867 if (!real_dev_has_splinters
4868 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4869 struct netdev *netdev;
4871 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4872 if (!netdev_get_in4(netdev, NULL, NULL) ||
4873 !netdev_get_in6(netdev, NULL)) {
4874 /* It has an IP address configured, so we don't own
4875 * it. Don't delete it. */
4877 vlandev_del(vlan_dev->name);
4879 netdev_close(netdev);
4886 /* Add all VLAN devices that we need. */
4887 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4890 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4891 if (!vlandev_get_name(real_dev_name, vid)) {
4892 vlandev_add(real_dev_name, vid);
4899 sset_destroy(&splinter_ifaces);
4901 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4902 free(splinter_vlans);
4905 return splinter_vlans;
4908 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4911 configure_splinter_port(struct port *port)
4913 struct ofproto *ofproto = port->bridge->ofproto;
4914 ofp_port_t realdev_ofp_port;
4915 const char *realdev_name;
4916 struct iface *vlandev, *realdev;
4918 ofproto_bundle_unregister(port->bridge->ofproto, port);
4920 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4923 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4924 realdev = iface_lookup(port->bridge, realdev_name);
4925 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4927 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4931 static struct ovsrec_port *
4932 synthesize_splinter_port(const char *real_dev_name,
4933 const char *vlan_dev_name, int vid)
4935 struct ovsrec_interface *iface;
4936 struct ovsrec_port *port;
4938 iface = xmalloc(sizeof *iface);
4939 ovsrec_interface_init(iface);
4940 iface->name = xstrdup(vlan_dev_name);
4941 iface->type = "system";
4943 port = xmalloc(sizeof *port);
4944 ovsrec_port_init(port);
4945 port->interfaces = xmemdup(&iface, sizeof iface);
4946 port->n_interfaces = 1;
4947 port->name = xstrdup(vlan_dev_name);
4948 port->vlan_mode = "splinter";
4949 port->tag = xmalloc(sizeof *port->tag);
4952 smap_add(&port->other_config, "realdev", real_dev_name);
4958 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4959 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4960 * 1-bit in the 'splinter_vlans' bitmap. */
4962 add_vlan_splinter_ports(struct bridge *br,
4963 const unsigned long int *splinter_vlans,
4964 struct shash *ports)
4968 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4969 * we're modifying 'ports'. */
4970 for (i = 0; i < br->cfg->n_ports; i++) {
4971 const char *name = br->cfg->ports[i]->name;
4972 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4975 for (j = 0; j < port_cfg->n_interfaces; j++) {
4976 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4978 if (vlan_splinters_is_enabled(iface_cfg)) {
4979 const char *real_dev_name;
4982 real_dev_name = iface_cfg->name;
4983 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4984 const char *vlan_dev_name;
4986 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4988 && !shash_find(ports, vlan_dev_name)) {
4989 shash_add(ports, vlan_dev_name,
4990 synthesize_splinter_port(
4991 real_dev_name, vlan_dev_name, vid));
5000 mirror_refresh_stats(struct mirror *m)
5002 struct ofproto *ofproto = m->bridge->ofproto;
5003 uint64_t tx_packets, tx_bytes;
5004 const char *keys[2];
5006 size_t stat_cnt = 0;
5008 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5009 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5013 if (tx_packets != UINT64_MAX) {
5014 keys[stat_cnt] = "tx_packets";
5015 values[stat_cnt] = tx_packets;
5018 if (tx_bytes != UINT64_MAX) {
5019 keys[stat_cnt] = "tx_bytes";
5020 values[stat_cnt] = tx_bytes;
5024 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);