1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
22 #include "async-append.h"
26 #include "connectivity.h"
31 #include "dynamic-string.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
44 #include "ofp-print.h"
47 #include "ofproto/bond.h"
48 #include "ofproto/ofproto.h"
50 #include "poll-loop.h"
55 #include "socket-util.h"
57 #include "stream-ssl.h"
59 #include "system-stats.h"
64 #include "lib/vswitch-idl.h"
65 #include "xenserver.h"
66 #include "openvswitch/vlog.h"
67 #include "sflow_api.h"
68 #include "vlan-bitmap.h"
71 VLOG_DEFINE_THIS_MODULE(bridge);
73 COVERAGE_DEFINE(bridge_reconfigure);
76 /* These members are always valid.
78 * They are immutable: they never change between iface_create() and
80 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
85 struct netdev *netdev; /* Network device. */
86 ofp_port_t ofp_port; /* OpenFlow port number. */
89 /* These members are valid only within bridge_reconfigure(). */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct ovs_list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 /* Port mirroring. */
131 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
134 struct hmap mappings; /* "struct" indexed by UUID */
136 /* Used during reconfiguration. */
137 struct shash wanted_ports;
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
146 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
147 struct bridge *bridge;
153 /* All bridges, indexed by name. */
154 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
156 /* OVSDB IDL used to obtain configuration. */
157 static struct ovsdb_idl *idl;
159 /* We want to complete daemonization, fully detaching from our parent process,
160 * only after we have completed our initial configuration, committed our state
161 * to the database, and received confirmation back from the database server
162 * that it applied the commit. This allows our parent process to know that,
163 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
164 * to have already been filled in. (It is only "very likely" rather than
165 * certain because there is always a slim possibility that the transaction will
166 * fail or that some other client has added new bridges, ports, etc. while
167 * ovs-vswitchd was configuring using an old configuration.)
169 * We only need to do this once for our initial configuration at startup, so
170 * 'initial_config_done' tracks whether we've already done it. While we are
171 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
172 * itself and is otherwise NULL. */
173 static bool initial_config_done;
174 static struct ovsdb_idl_txn *daemonize_txn;
176 /* Most recently processed IDL sequence number. */
177 static unsigned int idl_seqno;
179 /* Track changes to port connectivity. */
180 static uint64_t connectivity_seqno = LLONG_MIN;
182 /* Status update to database.
184 * Some information in the database must be kept as up-to-date as possible to
185 * allow controllers to respond rapidly to network outages. Those status are
186 * updated via the 'status_txn'.
188 * We use the global connectivity sequence number to detect the status change.
189 * Also, to prevent the status update from sending too much to the database,
190 * we check the return status of each update transaction and do not start new
191 * update if the previous transaction status is 'TXN_INCOMPLETE'.
193 * 'statux_txn' is NULL if there is no ongoing status update.
195 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
196 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
197 * which will cause the main thread wake up soon and retry the status update.
199 static struct ovsdb_idl_txn *status_txn;
200 static bool status_txn_try_again;
202 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
203 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
204 #define STATUS_CHECK_AGAIN_MSEC 100
206 /* Each time this timer expires, the bridge fetches interface and mirror
207 * statistics and pushes them into the database. */
208 static int stats_timer_interval;
209 static long long int stats_timer = LLONG_MIN;
211 /* Each time this timer expires, the bridge fetches the list of port/VLAN
212 * membership that has been modified by the AA.
214 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
215 static long long int aa_refresh_timer = LLONG_MIN;
217 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
218 * expensive. This can cause bridge_reconfigure() to take a long time during
219 * which no other work can be done. To deal with this problem, we limit port
220 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
221 * call to bridge_reconfigure(). If there is more work to do after the limit
222 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
223 * This allows the rest of the code to catch up on important things like
224 * forwarding packets. */
225 #define OFP_PORT_ACTION_WINDOW 10
227 static void add_del_bridges(const struct ovsrec_open_vswitch *);
228 static void bridge_run__(void);
229 static void bridge_create(const struct ovsrec_bridge *);
230 static void bridge_destroy(struct bridge *);
231 static struct bridge *bridge_lookup(const char *name);
232 static unixctl_cb_func bridge_unixctl_dump_flows;
233 static unixctl_cb_func bridge_unixctl_reconnect;
234 static size_t bridge_get_controllers(const struct bridge *br,
235 struct ovsrec_controller ***controllersp);
236 static void bridge_collect_wanted_ports(struct bridge *,
237 const unsigned long *splinter_vlans,
238 struct shash *wanted_ports);
239 static void bridge_delete_ofprotos(void);
240 static void bridge_delete_or_reconfigure_ports(struct bridge *);
241 static void bridge_del_ports(struct bridge *,
242 const struct shash *wanted_ports);
243 static void bridge_add_ports(struct bridge *,
244 const struct shash *wanted_ports);
246 static void bridge_configure_datapath_id(struct bridge *);
247 static void bridge_configure_netflow(struct bridge *);
248 static void bridge_configure_forward_bpdu(struct bridge *);
249 static void bridge_configure_mac_table(struct bridge *);
250 static void bridge_configure_mcast_snooping(struct bridge *);
251 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
252 static void bridge_configure_ipfix(struct bridge *);
253 static void bridge_configure_spanning_tree(struct bridge *);
254 static void bridge_configure_tables(struct bridge *);
255 static void bridge_configure_dp_desc(struct bridge *);
256 static void bridge_configure_aa(struct bridge *);
257 static void bridge_aa_refresh_queued(struct bridge *);
258 static bool bridge_aa_need_refresh(struct bridge *);
259 static void bridge_configure_remotes(struct bridge *,
260 const struct sockaddr_in *managers,
262 static void bridge_pick_local_hw_addr(struct bridge *,
263 uint8_t ea[ETH_ADDR_LEN],
264 struct iface **hw_addr_iface);
265 static uint64_t bridge_pick_datapath_id(struct bridge *,
266 const uint8_t bridge_ea[ETH_ADDR_LEN],
267 struct iface *hw_addr_iface);
268 static uint64_t dpid_from_hash(const void *, size_t nbytes);
269 static bool bridge_has_bond_fake_iface(const struct bridge *,
271 static bool port_is_bond_fake_iface(const struct port *);
273 static unixctl_cb_func qos_unixctl_show;
275 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
276 static void port_del_ifaces(struct port *);
277 static void port_destroy(struct port *);
278 static struct port *port_lookup(const struct bridge *, const char *name);
279 static void port_configure(struct port *);
280 static struct lacp_settings *port_configure_lacp(struct port *,
281 struct lacp_settings *);
282 static void port_configure_bond(struct port *, struct bond_settings *);
283 static bool port_is_synthetic(const struct port *);
285 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
286 static void run_system_stats(void);
288 static void bridge_configure_mirrors(struct bridge *);
289 static struct mirror *mirror_create(struct bridge *,
290 const struct ovsrec_mirror *);
291 static void mirror_destroy(struct mirror *);
292 static bool mirror_configure(struct mirror *);
293 static void mirror_refresh_stats(struct mirror *);
295 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
296 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
297 const struct ovsrec_port *);
298 static bool iface_is_internal(const struct ovsrec_interface *iface,
299 const struct ovsrec_bridge *br);
300 static const char *iface_get_type(const struct ovsrec_interface *,
301 const struct ovsrec_bridge *);
302 static void iface_destroy(struct iface *);
303 static void iface_destroy__(struct iface *);
304 static struct iface *iface_lookup(const struct bridge *, const char *name);
305 static struct iface *iface_find(const char *name);
306 static struct iface *iface_from_ofp_port(const struct bridge *,
307 ofp_port_t ofp_port);
308 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
309 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
310 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
311 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
312 static void iface_configure_cfm(struct iface *);
313 static void iface_refresh_cfm_stats(struct iface *);
314 static void iface_refresh_stats(struct iface *);
315 static void iface_refresh_netdev_status(struct iface *);
316 static void iface_refresh_ofproto_status(struct iface *);
317 static bool iface_is_synthetic(const struct iface *);
318 static ofp_port_t iface_get_requested_ofp_port(
319 const struct ovsrec_interface *);
320 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
323 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
325 * This is deprecated. It is only for compatibility with broken device drivers
326 * in old versions of Linux that do not properly support VLANs when VLAN
327 * devices are not used. When broken device drivers are no longer in
328 * widespread use, we will delete these interfaces. */
330 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
331 static bool vlan_splinters_enabled_anywhere;
333 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
334 static unsigned long int *collect_splinter_vlans(
335 const struct ovsrec_open_vswitch *);
336 static void configure_splinter_port(struct port *);
337 static void add_vlan_splinter_ports(struct bridge *,
338 const unsigned long int *splinter_vlans,
339 struct shash *ports);
341 static void discover_types(const struct ovsrec_open_vswitch *cfg);
344 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
346 struct shash iface_hints;
347 static bool initialized = false;
354 shash_init(&iface_hints);
357 for (i = 0; i < cfg->n_bridges; i++) {
358 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
361 for (j = 0; j < br_cfg->n_ports; j++) {
362 struct ovsrec_port *port_cfg = br_cfg->ports[j];
365 for (k = 0; k < port_cfg->n_interfaces; k++) {
366 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
367 struct iface_hint *iface_hint;
369 iface_hint = xmalloc(sizeof *iface_hint);
370 iface_hint->br_name = br_cfg->name;
371 iface_hint->br_type = br_cfg->datapath_type;
372 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
374 shash_add(&iface_hints, if_cfg->name, iface_hint);
380 ofproto_init(&iface_hints);
382 shash_destroy_free_data(&iface_hints);
386 /* Public functions. */
388 /* Initializes the bridge module, configuring it to obtain its configuration
389 * from an OVSDB server accessed over 'remote', which should be a string in a
390 * form acceptable to ovsdb_idl_create(). */
392 bridge_init(const char *remote)
394 /* Create connection to database. */
395 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
396 idl_seqno = ovsdb_idl_get_seqno(idl);
397 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
398 ovsdb_idl_verify_write_only(idl);
400 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
401 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
402 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
403 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
404 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
405 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
406 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
407 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
408 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
410 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
411 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
412 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
413 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
414 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
415 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
416 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
418 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
419 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
420 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
421 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
422 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
423 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
424 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
425 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
426 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
427 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
428 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
429 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
430 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
440 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
441 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
442 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
443 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
444 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
445 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
446 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
448 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
449 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
450 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
451 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
453 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
455 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
457 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
458 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
460 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
461 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
462 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
463 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
465 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
466 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
467 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
468 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
469 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
471 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
473 /* Register unixctl commands. */
474 unixctl_command_register("qos/show", "interface", 1, 1,
475 qos_unixctl_show, NULL);
476 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
477 bridge_unixctl_dump_flows, NULL);
478 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
479 bridge_unixctl_reconnect, NULL);
492 struct bridge *br, *next_br;
494 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
497 ovsdb_idl_destroy(idl);
500 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
501 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
502 * responsible for freeing '*managersp' (with free()).
504 * You may be asking yourself "why does ovs-vswitchd care?", because
505 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
506 * should not be and in fact is not directly involved in that. But
507 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
508 * it has to tell in-band control where the managers are to enable that.
509 * (Thus, only managers connected in-band are collected.)
512 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
513 struct sockaddr_in **managersp, size_t *n_managersp)
515 struct sockaddr_in *managers = NULL;
516 size_t n_managers = 0;
520 /* Collect all of the potential targets from the "targets" columns of the
521 * rows pointed to by "manager_options", excluding any that are
524 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
525 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
527 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
528 sset_find_and_delete(&targets, m->target);
530 sset_add(&targets, m->target);
534 /* Now extract the targets' IP addresses. */
535 if (!sset_is_empty(&targets)) {
538 managers = xmalloc(sset_count(&targets) * sizeof *managers);
539 SSET_FOR_EACH (target, &targets) {
541 struct sockaddr_storage ss;
542 struct sockaddr_in in;
545 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
547 && sa.ss.ss_family == AF_INET) {
548 managers[n_managers++] = sa.in;
552 sset_destroy(&targets);
554 *managersp = managers;
555 *n_managersp = n_managers;
559 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
561 unsigned long int *splinter_vlans;
562 struct sockaddr_in *managers;
563 struct bridge *br, *next;
564 int sflow_bridge_number;
567 COVERAGE_INC(bridge_reconfigure);
569 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
570 OFPROTO_FLOW_LIMIT_DEFAULT));
571 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
572 OFPROTO_MAX_IDLE_DEFAULT));
573 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
575 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
578 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
579 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
582 discover_types(ovs_cfg);
585 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
586 * to 'ovs_cfg', with only very minimal configuration otherwise.
588 * This is mostly an update to bridge data structures. Nothing is pushed
589 * down to ofproto or lower layers. */
590 add_del_bridges(ovs_cfg);
591 splinter_vlans = collect_splinter_vlans(ovs_cfg);
592 HMAP_FOR_EACH (br, node, &all_bridges) {
593 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
594 bridge_del_ports(br, &br->wanted_ports);
596 free(splinter_vlans);
598 /* Start pushing configuration changes down to the ofproto layer:
600 * - Delete ofprotos that are no longer configured.
602 * - Delete ports that are no longer configured.
604 * - Reconfigure existing ports to their desired configurations, or
605 * delete them if not possible.
607 * We have to do all the deletions before we can do any additions, because
608 * the ports to be added might require resources that will be freed up by
609 * deletions (they might especially overlap in name). */
610 bridge_delete_ofprotos();
611 HMAP_FOR_EACH (br, node, &all_bridges) {
613 bridge_delete_or_reconfigure_ports(br);
617 /* Finish pushing configuration changes to the ofproto layer:
619 * - Create ofprotos that are missing.
621 * - Add ports that are missing. */
622 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
626 error = ofproto_create(br->name, br->type, &br->ofproto);
628 VLOG_ERR("failed to create bridge %s: %s", br->name,
629 ovs_strerror(error));
630 shash_destroy(&br->wanted_ports);
633 /* Trigger storing datapath version. */
634 seq_change(connectivity_seq_get());
638 HMAP_FOR_EACH (br, node, &all_bridges) {
639 bridge_add_ports(br, &br->wanted_ports);
640 shash_destroy(&br->wanted_ports);
643 reconfigure_system_stats(ovs_cfg);
645 /* Complete the configuration. */
646 sflow_bridge_number = 0;
647 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
648 HMAP_FOR_EACH (br, node, &all_bridges) {
651 /* We need the datapath ID early to allow LACP ports to use it as the
652 * default system ID. */
653 bridge_configure_datapath_id(br);
655 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
658 port_configure(port);
660 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
661 iface_set_ofport(iface->cfg, iface->ofp_port);
662 /* Clear eventual previous errors */
663 ovsrec_interface_set_error(iface->cfg, NULL);
664 iface_configure_cfm(iface);
665 iface_configure_qos(iface, port->cfg->qos);
666 iface_set_mac(br, port, iface);
667 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
669 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
673 bridge_configure_mirrors(br);
674 bridge_configure_forward_bpdu(br);
675 bridge_configure_mac_table(br);
676 bridge_configure_mcast_snooping(br);
677 bridge_configure_remotes(br, managers, n_managers);
678 bridge_configure_netflow(br);
679 bridge_configure_sflow(br, &sflow_bridge_number);
680 bridge_configure_ipfix(br);
681 bridge_configure_spanning_tree(br);
682 bridge_configure_tables(br);
683 bridge_configure_dp_desc(br);
684 bridge_configure_aa(br);
688 /* The ofproto-dpif provider does some final reconfiguration in its
689 * ->type_run() function. We have to call it before notifying the database
690 * client that reconfiguration is complete, otherwise there is a very
691 * narrow race window in which e.g. ofproto/trace will not recognize the
692 * new configuration (sometimes this causes unit test failures). */
696 /* Delete ofprotos which aren't configured or have the wrong type. Create
697 * ofprotos which don't exist but need to. */
699 bridge_delete_ofprotos(void)
706 /* Delete ofprotos with no bridge or with the wrong type. */
709 ofproto_enumerate_types(&types);
710 SSET_FOR_EACH (type, &types) {
713 ofproto_enumerate_names(type, &names);
714 SSET_FOR_EACH (name, &names) {
715 br = bridge_lookup(name);
716 if (!br || strcmp(type, br->type)) {
717 ofproto_delete(name, type);
721 sset_destroy(&names);
722 sset_destroy(&types);
726 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
728 if (*n >= *allocated) {
729 ports = x2nrealloc(ports, allocated, sizeof *ports);
731 ports[(*n)++] = port;
736 bridge_delete_or_reconfigure_ports(struct bridge *br)
738 struct ofproto_port ofproto_port;
739 struct ofproto_port_dump dump;
741 struct sset ofproto_ports;
742 struct port *port, *port_next;
744 /* List of "ofp_port"s to delete. We make a list instead of deleting them
745 * right away because ofproto implementations aren't necessarily able to
746 * iterate through a changing list of ports in an entirely robust way. */
753 sset_init(&ofproto_ports);
755 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
756 * that are not configured in the database. (This commonly happens when
757 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
759 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
760 * that have the wrong OpenFlow port number (and arrange to add them back
761 * with the correct OpenFlow port number). */
762 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
763 ofp_port_t requested_ofp_port;
766 sset_add(&ofproto_ports, ofproto_port.name);
768 iface = iface_lookup(br, ofproto_port.name);
770 /* No such iface is configured, so we should delete this
773 * As a corner case exception, keep the port if it's a bond fake
775 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
776 && !strcmp(ofproto_port.type, "internal")) {
782 if (strcmp(ofproto_port.type, iface->type)
783 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
784 /* The interface is the wrong type or can't be configured.
789 /* If the requested OpenFlow port for 'iface' changed, and it's not
790 * already the correct port, then we might want to temporarily delete
791 * this interface, so we can add it back again with the new OpenFlow
793 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
794 if (iface->ofp_port != OFPP_LOCAL &&
795 requested_ofp_port != OFPP_NONE &&
796 requested_ofp_port != iface->ofp_port) {
797 ofp_port_t victim_request;
798 struct iface *victim;
800 /* Check for an existing OpenFlow port currently occupying
801 * 'iface''s requested port number. If there isn't one, then
802 * delete this port. Otherwise we need to consider further. */
803 victim = iface_from_ofp_port(br, requested_ofp_port);
808 /* 'victim' is a port currently using 'iface''s requested port
809 * number. Unless 'victim' specifically requested that port
810 * number, too, then we can delete both 'iface' and 'victim'
811 * temporarily. (We'll add both of them back again later with new
812 * OpenFlow port numbers.)
814 * If 'victim' did request port number 'requested_ofp_port', just
815 * like 'iface', then that's a configuration inconsistency that we
816 * can't resolve. We might as well let it keep its current port
818 victim_request = iface_get_requested_ofp_port(victim->cfg);
819 if (victim_request != requested_ofp_port) {
820 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
821 iface_destroy(victim);
830 iface_destroy(iface);
831 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
833 for (i = 0; i < n; i++) {
834 ofproto_port_del(br->ofproto, del[i]);
838 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
839 * that we didn't see when we iterated through the datapath, i.e. ports
840 * that disappeared underneath use. This is an unusual situation, but it
841 * can happen in some cases:
843 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
844 * idea but could happen).
846 * - The port represented a device that disappeared, e.g. a tuntap
847 * device destroyed via "tunctl -d", a physical Ethernet device
848 * whose module was just unloaded via "rmmod", or a virtual NIC for a
849 * VM whose VM was just terminated. */
850 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
851 struct iface *iface, *iface_next;
853 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
854 if (!sset_contains(&ofproto_ports, iface->name)) {
855 iface_destroy__(iface);
859 if (list_is_empty(&port->ifaces)) {
863 sset_destroy(&ofproto_ports);
867 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
868 bool with_requested_port)
870 struct shash_node *port_node;
872 SHASH_FOR_EACH (port_node, wanted_ports) {
873 const struct ovsrec_port *port_cfg = port_node->data;
876 for (i = 0; i < port_cfg->n_interfaces; i++) {
877 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
878 ofp_port_t requested_ofp_port;
880 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
881 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
882 struct iface *iface = iface_lookup(br, iface_cfg->name);
885 iface_create(br, iface_cfg, port_cfg);
893 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
895 /* First add interfaces that request a particular port number. */
896 bridge_add_ports__(br, wanted_ports, true);
898 /* Then add interfaces that want automatic port number assignment.
899 * We add these afterward to avoid accidentally taking a specifically
900 * requested port number. */
901 bridge_add_ports__(br, wanted_ports, false);
905 port_configure(struct port *port)
907 const struct ovsrec_port *cfg = port->cfg;
908 struct bond_settings bond_settings;
909 struct lacp_settings lacp_settings;
910 struct ofproto_bundle_settings s;
913 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
914 configure_splinter_port(port);
923 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
924 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
925 s.slaves[s.n_slaves++] = iface->ofp_port;
930 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
934 /* Get VLAN trunks. */
937 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
941 if (cfg->vlan_mode) {
942 if (!strcmp(cfg->vlan_mode, "access")) {
943 s.vlan_mode = PORT_VLAN_ACCESS;
944 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
945 s.vlan_mode = PORT_VLAN_TRUNK;
946 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
947 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
948 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
949 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
951 /* This "can't happen" because ovsdb-server should prevent it. */
952 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
953 "back to trunk mode", port->name, cfg->vlan_mode);
954 s.vlan_mode = PORT_VLAN_TRUNK;
958 s.vlan_mode = PORT_VLAN_ACCESS;
960 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
964 s.vlan_mode = PORT_VLAN_TRUNK;
967 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
970 /* Get LACP settings. */
971 s.lacp = port_configure_lacp(port, &lacp_settings);
975 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
976 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
977 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
980 s.lacp_slaves = NULL;
983 /* Get bond settings. */
984 if (s.n_slaves > 1) {
985 s.bond = &bond_settings;
986 port_configure_bond(port, &bond_settings);
989 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
990 netdev_set_miimon_interval(iface->netdev, 0);
995 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1000 free(s.lacp_slaves);
1003 /* Pick local port hardware address and datapath ID for 'br'. */
1005 bridge_configure_datapath_id(struct bridge *br)
1007 uint8_t ea[ETH_ADDR_LEN];
1009 struct iface *local_iface;
1010 struct iface *hw_addr_iface;
1013 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1014 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1016 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1018 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1019 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1020 "Ethernet address: %s",
1021 br->name, ovs_strerror(error));
1024 memcpy(br->ea, ea, ETH_ADDR_LEN);
1026 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1027 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1028 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1029 ofproto_set_datapath_id(br->ofproto, dpid);
1032 dpid_string = xasprintf("%016"PRIx64, dpid);
1033 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1037 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1040 bridge_get_allowed_versions(struct bridge *br)
1042 if (!br->cfg->n_protocols)
1045 return ofputil_versions_from_strings(br->cfg->protocols,
1046 br->cfg->n_protocols);
1049 /* Set NetFlow configuration on 'br'. */
1051 bridge_configure_netflow(struct bridge *br)
1053 struct ovsrec_netflow *cfg = br->cfg->netflow;
1054 struct netflow_options opts;
1057 ofproto_set_netflow(br->ofproto, NULL);
1061 memset(&opts, 0, sizeof opts);
1063 /* Get default NetFlow configuration from datapath.
1064 * Apply overrides from 'cfg'. */
1065 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1066 if (cfg->engine_type) {
1067 opts.engine_type = *cfg->engine_type;
1069 if (cfg->engine_id) {
1070 opts.engine_id = *cfg->engine_id;
1073 /* Configure active timeout interval. */
1074 opts.active_timeout = cfg->active_timeout;
1075 if (!opts.active_timeout) {
1076 opts.active_timeout = -1;
1077 } else if (opts.active_timeout < 0) {
1078 VLOG_WARN("bridge %s: active timeout interval set to negative "
1079 "value, using default instead (%d seconds)", br->name,
1080 NF_ACTIVE_TIMEOUT_DEFAULT);
1081 opts.active_timeout = -1;
1084 /* Add engine ID to interface number to disambiguate bridgs? */
1085 opts.add_id_to_iface = cfg->add_id_to_interface;
1086 if (opts.add_id_to_iface) {
1087 if (opts.engine_id > 0x7f) {
1088 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1089 "another vswitch, choose an engine id less than 128",
1092 if (hmap_count(&br->ports) > 508) {
1093 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1094 "another port when more than 508 ports are used",
1100 sset_init(&opts.collectors);
1101 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1104 if (ofproto_set_netflow(br->ofproto, &opts)) {
1105 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1107 sset_destroy(&opts.collectors);
1110 /* Set sFlow configuration on 'br'. */
1112 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1114 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1115 struct ovsrec_controller **controllers;
1116 struct ofproto_sflow_options oso;
1117 size_t n_controllers;
1121 ofproto_set_sflow(br->ofproto, NULL);
1125 memset(&oso, 0, sizeof oso);
1127 sset_init(&oso.targets);
1128 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1130 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1131 if (cfg->sampling) {
1132 oso.sampling_rate = *cfg->sampling;
1135 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1137 oso.polling_interval = *cfg->polling;
1140 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1142 oso.header_len = *cfg->header;
1145 oso.sub_id = (*sflow_bridge_number)++;
1146 oso.agent_device = cfg->agent;
1148 oso.control_ip = NULL;
1149 n_controllers = bridge_get_controllers(br, &controllers);
1150 for (i = 0; i < n_controllers; i++) {
1151 if (controllers[i]->local_ip) {
1152 oso.control_ip = controllers[i]->local_ip;
1156 ofproto_set_sflow(br->ofproto, &oso);
1158 sset_destroy(&oso.targets);
1161 /* Returns whether a IPFIX row is valid. */
1163 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1165 return ipfix && ipfix->n_targets > 0;
1168 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1170 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1171 const struct bridge *br)
1173 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1176 /* Set IPFIX configuration on 'br'. */
1178 bridge_configure_ipfix(struct bridge *br)
1180 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1181 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1182 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1183 struct ofproto_ipfix_bridge_exporter_options be_opts;
1184 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1185 size_t n_fe_opts = 0;
1187 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1188 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1193 if (!valid_be_cfg && n_fe_opts == 0) {
1194 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1199 memset(&be_opts, 0, sizeof be_opts);
1201 sset_init(&be_opts.targets);
1202 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1204 if (be_cfg->sampling) {
1205 be_opts.sampling_rate = *be_cfg->sampling;
1207 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1209 if (be_cfg->obs_domain_id) {
1210 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1212 if (be_cfg->obs_point_id) {
1213 be_opts.obs_point_id = *be_cfg->obs_point_id;
1215 if (be_cfg->cache_active_timeout) {
1216 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1218 if (be_cfg->cache_max_flows) {
1219 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1222 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1223 "enable-tunnel-sampling", true);
1225 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1226 "enable-input-sampling", false);
1228 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1229 "enable-output-sampling", false);
1232 if (n_fe_opts > 0) {
1233 struct ofproto_ipfix_flow_exporter_options *opts;
1234 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1236 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1237 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1238 opts->collector_set_id = fe_cfg->id;
1239 sset_init(&opts->targets);
1240 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1241 fe_cfg->ipfix->n_targets);
1242 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1243 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1244 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1245 ? *fe_cfg->ipfix->cache_max_flows : 0;
1251 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1255 sset_destroy(&be_opts.targets);
1258 if (n_fe_opts > 0) {
1259 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1261 for (i = 0; i < n_fe_opts; i++) {
1262 sset_destroy(&opts->targets);
1270 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1271 struct ofproto_port_stp_settings *port_s,
1272 int *port_num_counter, unsigned long *port_num_bitmap)
1274 const char *config_str;
1275 struct iface *iface;
1277 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1278 port_s->enable = false;
1281 port_s->enable = true;
1284 /* STP over bonds is not supported. */
1285 if (!list_is_singleton(&port->ifaces)) {
1286 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1288 port_s->enable = false;
1292 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1294 /* Internal ports shouldn't participate in spanning tree, so
1296 if (!strcmp(iface->type, "internal")) {
1297 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1298 port_s->enable = false;
1302 /* STP on mirror output ports is not supported. */
1303 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1304 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1305 port_s->enable = false;
1309 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1311 unsigned long int port_num = strtoul(config_str, NULL, 0);
1312 int port_idx = port_num - 1;
1314 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1315 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1316 port_s->enable = false;
1320 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1321 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1322 port->name, port_num);
1323 port_s->enable = false;
1326 bitmap_set1(port_num_bitmap, port_idx);
1327 port_s->port_num = port_idx;
1329 if (*port_num_counter >= STP_MAX_PORTS) {
1330 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1331 port_s->enable = false;
1335 port_s->port_num = (*port_num_counter)++;
1338 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1340 port_s->path_cost = strtoul(config_str, NULL, 10);
1342 enum netdev_features current;
1345 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1346 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1347 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1350 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1352 port_s->priority = strtoul(config_str, NULL, 0);
1354 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1359 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1360 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1362 const char *config_str;
1363 struct iface *iface;
1365 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1366 port_s->enable = false;
1369 port_s->enable = true;
1372 /* RSTP over bonds is not supported. */
1373 if (!list_is_singleton(&port->ifaces)) {
1374 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1376 port_s->enable = false;
1380 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1382 /* Internal ports shouldn't participate in spanning tree, so
1384 if (!strcmp(iface->type, "internal")) {
1385 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1386 port_s->enable = false;
1390 /* RSTP on mirror output ports is not supported. */
1391 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1392 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1393 port_s->enable = false;
1397 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1399 unsigned long int port_num = strtoul(config_str, NULL, 0);
1400 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1401 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1402 port_s->enable = false;
1405 port_s->port_num = port_num;
1407 if (*port_num_counter >= RSTP_MAX_PORTS) {
1408 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1409 port_s->enable = false;
1412 /* If rstp-port-num is not specified, use 0.
1413 * rstp_port_set_port_number() will look for the first free one. */
1414 port_s->port_num = 0;
1417 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1419 port_s->path_cost = strtoul(config_str, NULL, 10);
1421 enum netdev_features current;
1424 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1425 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1426 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1429 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1431 port_s->priority = strtoul(config_str, NULL, 0);
1433 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1436 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1438 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1440 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1443 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1444 "rstp-admin-port-state", true);
1446 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1447 "rstp-port-admin-edge", false);
1448 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1449 "rstp-port-auto-edge", true);
1450 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1451 "rstp-port-mcheck", false);
1454 /* Set spanning tree configuration on 'br'. */
1456 bridge_configure_stp(struct bridge *br, bool enable_stp)
1459 ofproto_set_stp(br->ofproto, NULL);
1461 struct ofproto_stp_settings br_s;
1462 const char *config_str;
1464 int port_num_counter;
1465 unsigned long *port_num_bitmap;
1467 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1469 uint8_t ea[ETH_ADDR_LEN];
1471 if (eth_addr_from_string(config_str, ea)) {
1472 br_s.system_id = eth_addr_to_uint64(ea);
1474 br_s.system_id = eth_addr_to_uint64(br->ea);
1475 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1476 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1479 br_s.system_id = eth_addr_to_uint64(br->ea);
1482 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1484 br_s.priority = strtoul(config_str, NULL, 0);
1486 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1489 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1491 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1493 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1496 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1498 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1500 br_s.max_age = STP_DEFAULT_MAX_AGE;
1503 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1505 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1507 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1510 /* Configure STP on the bridge. */
1511 if (ofproto_set_stp(br->ofproto, &br_s)) {
1512 VLOG_ERR("bridge %s: could not enable STP", br->name);
1516 /* Users must either set the port number with the "stp-port-num"
1517 * configuration on all ports or none. If manual configuration
1518 * is not done, then we allocate them sequentially. */
1519 port_num_counter = 0;
1520 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1521 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1522 struct ofproto_port_stp_settings port_s;
1523 struct iface *iface;
1525 port_configure_stp(br->ofproto, port, &port_s,
1526 &port_num_counter, port_num_bitmap);
1528 /* As bonds are not supported, just apply configuration to
1529 * all interfaces. */
1530 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1531 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1533 VLOG_ERR("port %s: could not enable STP", port->name);
1539 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1540 && port_num_counter) {
1541 VLOG_ERR("bridge %s: must manually configure all STP port "
1542 "IDs or none, disabling", br->name);
1543 ofproto_set_stp(br->ofproto, NULL);
1545 bitmap_free(port_num_bitmap);
1550 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1553 ofproto_set_rstp(br->ofproto, NULL);
1555 struct ofproto_rstp_settings br_s;
1556 const char *config_str;
1558 int port_num_counter;
1560 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1562 uint8_t ea[ETH_ADDR_LEN];
1564 if (eth_addr_from_string(config_str, ea)) {
1565 br_s.address = eth_addr_to_uint64(ea);
1568 br_s.address = eth_addr_to_uint64(br->ea);
1569 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1570 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1574 br_s.address = eth_addr_to_uint64(br->ea);
1577 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1579 br_s.priority = strtoul(config_str, NULL, 0);
1581 br_s.priority = RSTP_DEFAULT_PRIORITY;
1584 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1586 br_s.ageing_time = strtoul(config_str, NULL, 0);
1588 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1591 config_str = smap_get(&br->cfg->other_config,
1592 "rstp-force-protocol-version");
1594 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1596 br_s.force_protocol_version = FPV_DEFAULT;
1599 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1601 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1603 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1606 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1608 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1610 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1613 config_str = smap_get(&br->cfg->other_config,
1614 "rstp-transmit-hold-count");
1616 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1618 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1621 /* Configure RSTP on the bridge. */
1622 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1623 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1627 port_num_counter = 0;
1628 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1629 struct ofproto_port_rstp_settings port_s;
1630 struct iface *iface;
1632 port_configure_rstp(br->ofproto, port, &port_s,
1635 /* As bonds are not supported, just apply configuration to
1636 * all interfaces. */
1637 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1638 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1640 VLOG_ERR("port %s: could not enable RSTP", port->name);
1649 bridge_configure_spanning_tree(struct bridge *br)
1651 bool enable_rstp = br->cfg->rstp_enable;
1652 bool enable_stp = br->cfg->stp_enable;
1654 if (enable_rstp && enable_stp) {
1655 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1656 "configured; enabling RSTP", br->name);
1660 bridge_configure_stp(br, enable_stp);
1661 bridge_configure_rstp(br, enable_rstp);
1665 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1667 const struct port *port = port_lookup(br, name);
1668 return port && port_is_bond_fake_iface(port);
1672 port_is_bond_fake_iface(const struct port *port)
1674 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1678 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1680 struct bridge *br, *next;
1681 struct shash new_br;
1684 /* Collect new bridges' names and types. */
1685 shash_init(&new_br);
1686 for (i = 0; i < cfg->n_bridges; i++) {
1687 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1688 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1690 if (strchr(br_cfg->name, '/')) {
1691 /* Prevent remote ovsdb-server users from accessing arbitrary
1692 * directories, e.g. consider a bridge named "../../../etc/". */
1693 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1695 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1696 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1700 /* Get rid of deleted bridges or those whose types have changed.
1701 * Update 'cfg' of bridges that still exist. */
1702 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1703 br->cfg = shash_find_data(&new_br, br->name);
1704 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1705 br->cfg->datapath_type))) {
1710 /* Add new bridges. */
1711 for (i = 0; i < cfg->n_bridges; i++) {
1712 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1713 struct bridge *br = bridge_lookup(br_cfg->name);
1715 bridge_create(br_cfg);
1719 shash_destroy(&new_br);
1722 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1723 * Returns 0 if successful, otherwise a positive errno value. */
1725 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1726 struct netdev *netdev, char **errp)
1728 return netdev_set_config(netdev, &iface_cfg->options, errp);
1731 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1732 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1734 * If successful, returns 0 and stores the network device in '*netdevp'. On
1735 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1737 iface_do_create(const struct bridge *br,
1738 const struct ovsrec_interface *iface_cfg,
1739 const struct ovsrec_port *port_cfg,
1740 ofp_port_t *ofp_portp, struct netdev **netdevp,
1743 struct netdev *netdev = NULL;
1746 if (netdev_is_reserved_name(iface_cfg->name)) {
1747 VLOG_WARN("could not create interface %s, name is reserved",
1753 error = netdev_open(iface_cfg->name,
1754 iface_get_type(iface_cfg, br->cfg), &netdev);
1756 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1757 iface_cfg->name, ovs_strerror(error));
1761 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1766 *ofp_portp = iface_pick_ofport(iface_cfg);
1767 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1772 VLOG_INFO("bridge %s: added interface %s on port %d",
1773 br->name, iface_cfg->name, *ofp_portp);
1775 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1776 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1784 netdev_close(netdev);
1788 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1789 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1790 * automatically allocated for the iface. Takes ownership of and
1791 * deallocates 'if_cfg'.
1793 * Return true if an iface is successfully created, false otherwise. */
1795 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1796 const struct ovsrec_port *port_cfg)
1798 struct netdev *netdev;
1799 struct iface *iface;
1800 ofp_port_t ofp_port;
1805 /* Do the bits that can fail up front. */
1806 ovs_assert(!iface_lookup(br, iface_cfg->name));
1807 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1809 iface_clear_db_record(iface_cfg, errp);
1814 /* Get or create the port structure. */
1815 port = port_lookup(br, port_cfg->name);
1817 port = port_create(br, port_cfg);
1820 /* Create the iface structure. */
1821 iface = xzalloc(sizeof *iface);
1822 list_push_back(&port->ifaces, &iface->port_elem);
1823 hmap_insert(&br->iface_by_name, &iface->name_node,
1824 hash_string(iface_cfg->name, 0));
1826 iface->name = xstrdup(iface_cfg->name);
1827 iface->ofp_port = ofp_port;
1828 iface->netdev = netdev;
1829 iface->type = iface_get_type(iface_cfg, br->cfg);
1830 iface->cfg = iface_cfg;
1831 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1832 hash_ofp_port(ofp_port));
1834 /* Populate initial status in database. */
1835 iface_refresh_stats(iface);
1836 iface_refresh_netdev_status(iface);
1838 /* Add bond fake iface if necessary. */
1839 if (port_is_bond_fake_iface(port)) {
1840 struct ofproto_port ofproto_port;
1842 if (ofproto_port_query_by_name(br->ofproto, port->name,
1844 struct netdev *netdev;
1847 error = netdev_open(port->name, "internal", &netdev);
1849 ofp_port_t fake_ofp_port = OFPP_NONE;
1850 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1851 netdev_close(netdev);
1853 VLOG_WARN("could not open network device %s (%s)",
1854 port->name, ovs_strerror(error));
1857 /* Already exists, nothing to do. */
1858 ofproto_port_destroy(&ofproto_port);
1865 /* Set forward BPDU option. */
1867 bridge_configure_forward_bpdu(struct bridge *br)
1869 ofproto_set_forward_bpdu(br->ofproto,
1870 smap_get_bool(&br->cfg->other_config,
1875 /* Set MAC learning table configuration for 'br'. */
1877 bridge_configure_mac_table(struct bridge *br)
1879 const char *idle_time_str;
1882 const char *mac_table_size_str;
1885 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1886 idle_time = (idle_time_str && atoi(idle_time_str)
1887 ? atoi(idle_time_str)
1888 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1890 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1891 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1892 ? atoi(mac_table_size_str)
1895 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1898 /* Set multicast snooping table configuration for 'br'. */
1900 bridge_configure_mcast_snooping(struct bridge *br)
1902 if (!br->cfg->mcast_snooping_enable) {
1903 ofproto_set_mcast_snooping(br->ofproto, NULL);
1906 struct ofproto_mcast_snooping_settings br_s;
1907 const char *idle_time_str;
1908 const char *max_entries_str;
1910 idle_time_str = smap_get(&br->cfg->other_config,
1911 "mcast-snooping-aging-time");
1912 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1913 ? atoi(idle_time_str)
1914 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1916 max_entries_str = smap_get(&br->cfg->other_config,
1917 "mcast-snooping-table-size");
1918 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1919 ? atoi(max_entries_str)
1920 : MCAST_DEFAULT_MAX_ENTRIES);
1922 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1923 "mcast-snooping-disable-flood-unregistered",
1926 /* Configure multicast snooping on the bridge */
1927 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1928 VLOG_ERR("bridge %s: could not enable multicast snooping",
1933 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1934 struct ofproto_mcast_snooping_port_settings port_s;
1935 port_s.flood = smap_get_bool(&port->cfg->other_config,
1936 "mcast-snooping-flood", false);
1937 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1938 "mcast-snooping-flood-reports", false);
1939 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1940 VLOG_ERR("port %s: could not configure mcast snooping",
1948 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1949 const struct port *fake_br, struct iface **hw_addr_iface)
1951 struct hmapx mirror_output_ports;
1953 bool found_addr = false;
1957 /* Mirror output ports don't participate in picking the local hardware
1958 * address. ofproto can't help us find out whether a given port is a
1959 * mirror output because we haven't configured mirrors yet, so we need to
1960 * accumulate them ourselves. */
1961 hmapx_init(&mirror_output_ports);
1962 for (i = 0; i < br->cfg->n_mirrors; i++) {
1963 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1964 if (m->output_port) {
1965 hmapx_add(&mirror_output_ports, m->output_port);
1969 /* Otherwise choose the minimum non-local MAC address among all of the
1971 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1972 uint8_t iface_ea[ETH_ADDR_LEN];
1973 struct iface *candidate;
1974 struct iface *iface;
1976 /* Mirror output ports don't participate. */
1977 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1981 /* Choose the MAC address to represent the port. */
1983 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1984 /* Find the interface with this Ethernet address (if any) so that
1985 * we can provide the correct devname to the caller. */
1986 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1987 uint8_t candidate_ea[ETH_ADDR_LEN];
1988 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1989 && eth_addr_equals(iface_ea, candidate_ea)) {
1994 /* Choose the interface whose MAC address will represent the port.
1995 * The Linux kernel bonding code always chooses the MAC address of
1996 * the first slave added to a bond, and the Fedora networking
1997 * scripts always add slaves to a bond in alphabetical order, so
1998 * for compatibility we choose the interface with the name that is
1999 * first in alphabetical order. */
2000 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2001 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2006 /* The local port doesn't count (since we're trying to choose its
2007 * MAC address anyway). */
2008 if (iface->ofp_port == OFPP_LOCAL) {
2012 /* For fake bridges we only choose from ports with the same tag */
2013 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2014 if (!port->cfg->tag) {
2017 if (*port->cfg->tag != *fake_br->cfg->tag) {
2023 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2029 /* Compare against our current choice. */
2030 if (!eth_addr_is_multicast(iface_ea) &&
2031 !eth_addr_is_local(iface_ea) &&
2032 !eth_addr_is_reserved(iface_ea) &&
2033 !eth_addr_is_zero(iface_ea) &&
2034 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2036 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2037 *hw_addr_iface = iface;
2043 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2044 *hw_addr_iface = NULL;
2047 hmapx_destroy(&mirror_output_ports);
2051 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2052 struct iface **hw_addr_iface)
2055 *hw_addr_iface = NULL;
2057 /* Did the user request a particular MAC? */
2058 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2059 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2060 if (eth_addr_is_multicast(ea)) {
2061 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2062 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2063 } else if (eth_addr_is_zero(ea)) {
2064 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2070 /* Find a local hw address */
2071 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2074 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2075 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2076 * an interface on 'br', then that interface must be passed in as
2077 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2078 * 'hw_addr_iface' must be passed in as a null pointer. */
2080 bridge_pick_datapath_id(struct bridge *br,
2081 const uint8_t bridge_ea[ETH_ADDR_LEN],
2082 struct iface *hw_addr_iface)
2085 * The procedure for choosing a bridge MAC address will, in the most
2086 * ordinary case, also choose a unique MAC that we can use as a datapath
2087 * ID. In some special cases, though, multiple bridges will end up with
2088 * the same MAC address. This is OK for the bridges, but it will confuse
2089 * the OpenFlow controller, because each datapath needs a unique datapath
2092 * Datapath IDs must be unique. It is also very desirable that they be
2093 * stable from one run to the next, so that policy set on a datapath
2096 const char *datapath_id;
2099 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2100 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2104 if (!hw_addr_iface) {
2106 * A purely internal bridge, that is, one that has no non-virtual
2107 * network devices on it at all, is difficult because it has no
2108 * natural unique identifier at all.
2110 * When the host is a XenServer, we handle this case by hashing the
2111 * host's UUID with the name of the bridge. Names of bridges are
2112 * persistent across XenServer reboots, although they can be reused if
2113 * an internal network is destroyed and then a new one is later
2114 * created, so this is fairly effective.
2116 * When the host is not a XenServer, we punt by using a random MAC
2117 * address on each run.
2119 const char *host_uuid = xenserver_get_host_uuid();
2121 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2122 dpid = dpid_from_hash(combined, strlen(combined));
2128 return eth_addr_to_uint64(bridge_ea);
2132 dpid_from_hash(const void *data, size_t n)
2134 uint8_t hash[SHA1_DIGEST_SIZE];
2136 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2137 sha1_bytes(data, n, hash);
2138 eth_addr_mark_random(hash);
2139 return eth_addr_to_uint64(hash);
2143 iface_refresh_netdev_status(struct iface *iface)
2147 enum netdev_features current;
2148 enum netdev_flags flags;
2149 const char *link_state;
2150 uint8_t mac[ETH_ADDR_LEN];
2151 int64_t bps, mtu_64, ifindex64, link_resets;
2154 if (iface_is_synthetic(iface)) {
2158 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2159 && !status_txn_try_again) {
2163 iface->change_seq = netdev_get_change_seq(iface->netdev);
2167 if (!netdev_get_status(iface->netdev, &smap)) {
2168 ovsrec_interface_set_status(iface->cfg, &smap);
2170 ovsrec_interface_set_status(iface->cfg, NULL);
2173 smap_destroy(&smap);
2175 error = netdev_get_flags(iface->netdev, &flags);
2177 const char *state = flags & NETDEV_UP ? "up" : "down";
2179 ovsrec_interface_set_admin_state(iface->cfg, state);
2181 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2184 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2185 ovsrec_interface_set_link_state(iface->cfg, link_state);
2187 link_resets = netdev_get_carrier_resets(iface->netdev);
2188 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2190 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2191 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2193 ovsrec_interface_set_duplex(iface->cfg,
2194 netdev_features_is_full_duplex(current)
2196 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2198 ovsrec_interface_set_duplex(iface->cfg, NULL);
2199 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2202 error = netdev_get_mtu(iface->netdev, &mtu);
2205 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2207 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2210 error = netdev_get_etheraddr(iface->netdev, mac);
2212 char mac_string[32];
2214 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2215 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2217 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2220 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2221 * if there is no valid ifindex number. */
2222 ifindex64 = netdev_get_ifindex(iface->netdev);
2223 if (ifindex64 < 0) {
2226 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2230 iface_refresh_ofproto_status(struct iface *iface)
2234 if (iface_is_synthetic(iface)) {
2238 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2242 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2244 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2247 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2249 || status_txn_try_again) {
2250 iface_refresh_cfm_stats(iface);
2253 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2255 || status_txn_try_again) {
2259 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2260 iface->ofp_port, &smap);
2261 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2262 smap_destroy(&smap);
2266 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2269 iface_refresh_cfm_stats(struct iface *iface)
2271 const struct ovsrec_interface *cfg = iface->cfg;
2272 struct cfm_status status;
2275 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2276 iface->ofp_port, &status);
2278 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2279 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2280 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2281 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2282 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2283 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2285 const char *reasons[CFM_FAULT_N_REASONS];
2286 int64_t cfm_health = status.health;
2287 int64_t cfm_flap_count = status.flap_count;
2288 bool faulted = status.faults != 0;
2291 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2294 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2295 int reason = 1 << i;
2296 if (status.faults & reason) {
2297 reasons[j++] = cfm_fault_reason_to_str(reason);
2300 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2302 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2304 if (status.remote_opstate >= 0) {
2305 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2306 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2308 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2311 ovsrec_interface_set_cfm_remote_mpids(cfg,
2312 (const int64_t *)status.rmps,
2314 if (cfm_health >= 0) {
2315 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2317 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2325 iface_refresh_stats(struct iface *iface)
2327 #define IFACE_STATS \
2328 IFACE_STAT(rx_packets, "rx_packets") \
2329 IFACE_STAT(tx_packets, "tx_packets") \
2330 IFACE_STAT(rx_bytes, "rx_bytes") \
2331 IFACE_STAT(tx_bytes, "tx_bytes") \
2332 IFACE_STAT(rx_dropped, "rx_dropped") \
2333 IFACE_STAT(tx_dropped, "tx_dropped") \
2334 IFACE_STAT(rx_errors, "rx_errors") \
2335 IFACE_STAT(tx_errors, "tx_errors") \
2336 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2337 IFACE_STAT(rx_over_errors, "rx_over_err") \
2338 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2339 IFACE_STAT(collisions, "collisions")
2341 #define IFACE_STAT(MEMBER, NAME) + 1
2342 enum { N_IFACE_STATS = IFACE_STATS };
2344 int64_t values[N_IFACE_STATS];
2345 const char *keys[N_IFACE_STATS];
2348 struct netdev_stats stats;
2350 if (iface_is_synthetic(iface)) {
2354 /* Intentionally ignore return value, since errors will set 'stats' to
2355 * all-1s, and we will deal with that correctly below. */
2356 netdev_get_stats(iface->netdev, &stats);
2358 /* Copy statistics into keys[] and values[]. */
2360 #define IFACE_STAT(MEMBER, NAME) \
2361 if (stats.MEMBER != UINT64_MAX) { \
2363 values[n] = stats.MEMBER; \
2368 ovs_assert(n <= N_IFACE_STATS);
2370 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2375 br_refresh_datapath_info(struct bridge *br)
2377 const char *version;
2379 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2380 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2383 ovsrec_bridge_set_datapath_version(br->cfg,
2384 version ? version : "<unknown>");
2388 br_refresh_stp_status(struct bridge *br)
2390 struct smap smap = SMAP_INITIALIZER(&smap);
2391 struct ofproto *ofproto = br->ofproto;
2392 struct ofproto_stp_status status;
2394 if (ofproto_get_stp_status(ofproto, &status)) {
2398 if (!status.enabled) {
2399 ovsrec_bridge_set_status(br->cfg, NULL);
2403 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2404 STP_ID_ARGS(status.bridge_id));
2405 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2406 STP_ID_ARGS(status.designated_root));
2407 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2409 ovsrec_bridge_set_status(br->cfg, &smap);
2410 smap_destroy(&smap);
2414 port_refresh_stp_status(struct port *port)
2416 struct ofproto *ofproto = port->bridge->ofproto;
2417 struct iface *iface;
2418 struct ofproto_port_stp_status status;
2421 if (port_is_synthetic(port)) {
2425 /* STP doesn't currently support bonds. */
2426 if (!list_is_singleton(&port->ifaces)) {
2427 ovsrec_port_set_status(port->cfg, NULL);
2431 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2432 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2436 if (!status.enabled) {
2437 ovsrec_port_set_status(port->cfg, NULL);
2441 /* Set Status column. */
2443 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2444 smap_add(&smap, "stp_state", stp_state_name(status.state));
2445 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2446 smap_add(&smap, "stp_role", stp_role_name(status.role));
2447 ovsrec_port_set_status(port->cfg, &smap);
2448 smap_destroy(&smap);
2452 port_refresh_stp_stats(struct port *port)
2454 struct ofproto *ofproto = port->bridge->ofproto;
2455 struct iface *iface;
2456 struct ofproto_port_stp_stats stats;
2457 const char *keys[3];
2458 int64_t int_values[3];
2460 if (port_is_synthetic(port)) {
2464 /* STP doesn't currently support bonds. */
2465 if (!list_is_singleton(&port->ifaces)) {
2469 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2470 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2474 if (!stats.enabled) {
2475 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2479 /* Set Statistics column. */
2480 keys[0] = "stp_tx_count";
2481 int_values[0] = stats.tx_count;
2482 keys[1] = "stp_rx_count";
2483 int_values[1] = stats.rx_count;
2484 keys[2] = "stp_error_count";
2485 int_values[2] = stats.error_count;
2487 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2488 ARRAY_SIZE(int_values));
2492 br_refresh_rstp_status(struct bridge *br)
2494 struct smap smap = SMAP_INITIALIZER(&smap);
2495 struct ofproto *ofproto = br->ofproto;
2496 struct ofproto_rstp_status status;
2498 if (ofproto_get_rstp_status(ofproto, &status)) {
2501 if (!status.enabled) {
2502 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2505 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2506 RSTP_ID_ARGS(status.bridge_id));
2507 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2508 status.root_path_cost);
2509 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2510 RSTP_ID_ARGS(status.root_id));
2511 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2512 RSTP_ID_ARGS(status.designated_id));
2513 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2514 status.designated_port_id);
2515 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2516 status.bridge_port_id);
2517 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2518 smap_destroy(&smap);
2522 port_refresh_rstp_status(struct port *port)
2524 struct ofproto *ofproto = port->bridge->ofproto;
2525 struct iface *iface;
2526 struct ofproto_port_rstp_status status;
2527 const char *keys[4];
2528 int64_t int_values[4];
2531 if (port_is_synthetic(port)) {
2535 /* RSTP doesn't currently support bonds. */
2536 if (!list_is_singleton(&port->ifaces)) {
2537 ovsrec_port_set_rstp_status(port->cfg, NULL);
2541 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2542 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2546 if (!status.enabled) {
2547 ovsrec_port_set_rstp_status(port->cfg, NULL);
2548 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2551 /* Set Status column. */
2554 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2556 smap_add_format(&smap, "rstp_port_role", "%s",
2557 rstp_port_role_name(status.role));
2558 smap_add_format(&smap, "rstp_port_state", "%s",
2559 rstp_state_name(status.state));
2560 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2561 RSTP_ID_ARGS(status.designated_bridge_id));
2562 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2563 status.designated_port_id);
2564 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2565 status.designated_path_cost);
2567 ovsrec_port_set_rstp_status(port->cfg, &smap);
2568 smap_destroy(&smap);
2570 /* Set Statistics column. */
2571 keys[0] = "rstp_tx_count";
2572 int_values[0] = status.tx_count;
2573 keys[1] = "rstp_rx_count";
2574 int_values[1] = status.rx_count;
2575 keys[2] = "rstp_uptime";
2576 int_values[2] = status.uptime;
2577 keys[3] = "rstp_error_count";
2578 int_values[3] = status.error_count;
2579 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2580 ARRAY_SIZE(int_values));
2584 port_refresh_bond_status(struct port *port, bool force_update)
2586 uint8_t mac[ETH_ADDR_LEN];
2588 /* Return if port is not a bond */
2589 if (list_is_singleton(&port->ifaces)) {
2593 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2597 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2598 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2604 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2606 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2610 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2612 bool enable = enable_system_stats(cfg);
2614 system_stats_enable(enable);
2616 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2621 run_system_stats(void)
2623 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2626 stats = system_stats_run();
2628 struct ovsdb_idl_txn *txn;
2629 struct ovsdb_datum datum;
2631 txn = ovsdb_idl_txn_create(idl);
2632 ovsdb_datum_from_smap(&datum, stats);
2633 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2635 ovsdb_idl_txn_commit(txn);
2636 ovsdb_idl_txn_destroy(txn);
2643 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2646 case OFPCR12_ROLE_EQUAL:
2648 case OFPCR12_ROLE_MASTER:
2650 case OFPCR12_ROLE_SLAVE:
2652 case OFPCR12_ROLE_NOCHANGE:
2654 return "*** INVALID ROLE ***";
2659 refresh_controller_status(void)
2663 const struct ovsrec_controller *cfg;
2667 /* Accumulate status for controllers on all bridges. */
2668 HMAP_FOR_EACH (br, node, &all_bridges) {
2669 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2672 /* Update each controller in the database with current status. */
2673 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2674 struct ofproto_controller_info *cinfo =
2675 shash_find_data(&info, cfg->target);
2678 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2679 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2681 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2683 ovsrec_controller_set_is_connected(cfg, false);
2684 ovsrec_controller_set_role(cfg, NULL);
2685 ovsrec_controller_set_status(cfg, NULL);
2689 ofproto_free_ofproto_controller_info(&info);
2692 /* Update interface and mirror statistics if necessary. */
2694 run_stats_update(void)
2696 static struct ovsdb_idl_txn *stats_txn;
2697 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2704 /* Statistics update interval should always be greater than or equal to
2706 stats_interval = MAX(smap_get_int(&cfg->other_config,
2707 "stats-update-interval",
2709 if (stats_timer_interval != stats_interval) {
2710 stats_timer_interval = stats_interval;
2711 stats_timer = LLONG_MIN;
2714 if (time_msec() >= stats_timer) {
2715 enum ovsdb_idl_txn_status status;
2717 /* Rate limit the update. Do not start a new update if the
2718 * previous one is not done. */
2722 stats_txn = ovsdb_idl_txn_create(idl);
2723 HMAP_FOR_EACH (br, node, &all_bridges) {
2727 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2728 struct iface *iface;
2730 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2731 iface_refresh_stats(iface);
2733 port_refresh_stp_stats(port);
2735 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2736 mirror_refresh_stats(m);
2739 refresh_controller_status();
2742 status = ovsdb_idl_txn_commit(stats_txn);
2743 if (status != TXN_INCOMPLETE) {
2744 stats_timer = time_msec() + stats_timer_interval;
2745 ovsdb_idl_txn_destroy(stats_txn);
2751 /* Update bridge/port/interface status if necessary. */
2753 run_status_update(void)
2758 /* Rate limit the update. Do not start a new update if the
2759 * previous one is not done. */
2760 seq = seq_read(connectivity_seq_get());
2761 if (seq != connectivity_seqno || status_txn_try_again) {
2764 connectivity_seqno = seq;
2765 status_txn = ovsdb_idl_txn_create(idl);
2766 HMAP_FOR_EACH (br, node, &all_bridges) {
2769 br_refresh_stp_status(br);
2770 br_refresh_rstp_status(br);
2771 br_refresh_datapath_info(br);
2772 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2773 struct iface *iface;
2775 port_refresh_stp_status(port);
2776 port_refresh_rstp_status(port);
2777 port_refresh_bond_status(port, status_txn_try_again);
2778 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2779 iface_refresh_netdev_status(iface);
2780 iface_refresh_ofproto_status(iface);
2787 /* Commit the transaction and get the status. If the transaction finishes,
2788 * then destroy the transaction. Otherwise, keep it so that we can check
2789 * progress the next time that this function is called. */
2791 enum ovsdb_idl_txn_status status;
2793 status = ovsdb_idl_txn_commit(status_txn);
2794 if (status != TXN_INCOMPLETE) {
2795 ovsdb_idl_txn_destroy(status_txn);
2798 /* Sets the 'status_txn_try_again' if the transaction fails. */
2799 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2800 status_txn_try_again = false;
2802 status_txn_try_again = true;
2807 /* Refresh AA port status if necessary. */
2808 if (time_msec() >= aa_refresh_timer) {
2811 HMAP_FOR_EACH (br, node, &all_bridges) {
2812 if (bridge_aa_need_refresh(br)) {
2813 struct ovsdb_idl_txn *txn;
2815 txn = ovsdb_idl_txn_create(idl);
2816 bridge_aa_refresh_queued(br);
2817 ovsdb_idl_txn_commit(txn);
2818 ovsdb_idl_txn_destroy(txn);
2822 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2827 status_update_wait(void)
2829 /* This prevents the process from constantly waking up on
2830 * connectivity seq, when there is no connection to ovsdb. */
2831 if (!ovsdb_idl_has_lock(idl)) {
2835 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2836 * transaction to complete. If the status update to database needs to be
2837 * run again (transaction fails), registers a timeout in
2838 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2839 * sequence number. */
2841 ovsdb_idl_txn_wait(status_txn);
2842 } else if (status_txn_try_again) {
2843 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2845 seq_wait(connectivity_seq_get(), connectivity_seqno);
2856 /* Let each datapath type do the work that it needs to do. */
2858 ofproto_enumerate_types(&types);
2859 SSET_FOR_EACH (type, &types) {
2860 ofproto_type_run(type);
2862 sset_destroy(&types);
2864 /* Let each bridge do the work that it needs to do. */
2865 HMAP_FOR_EACH (br, node, &all_bridges) {
2866 ofproto_run(br->ofproto);
2873 static struct ovsrec_open_vswitch null_cfg;
2874 const struct ovsrec_open_vswitch *cfg;
2876 bool vlan_splinters_changed;
2878 ovsrec_open_vswitch_init(&null_cfg);
2882 if (ovsdb_idl_is_lock_contended(idl)) {
2883 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2884 struct bridge *br, *next_br;
2886 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2887 "disabling this process (pid %ld) until it goes away",
2888 (long int) getpid());
2890 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2893 /* Since we will not be running system_stats_run() in this process
2894 * with the current situation of multiple ovs-vswitchd daemons,
2895 * disable system stats collection. */
2896 system_stats_enable(false);
2898 } else if (!ovsdb_idl_has_lock(idl)) {
2901 cfg = ovsrec_open_vswitch_first(idl);
2903 /* Initialize the ofproto library. This only needs to run once, but
2904 * it must be done after the configuration is set. If the
2905 * initialization has already occurred, bridge_init_ofproto()
2906 * returns immediately. */
2907 bridge_init_ofproto(cfg);
2909 /* Once the value of flow-restore-wait is false, we no longer should
2910 * check its value from the database. */
2911 if (cfg && ofproto_get_flow_restore_wait()) {
2912 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2913 "flow-restore-wait", false));
2918 /* Re-configure SSL. We do this on every trip through the main loop,
2919 * instead of just when the database changes, because the contents of the
2920 * key and certificate files can change without the database changing.
2922 * We do this before bridge_reconfigure() because that function might
2923 * initiate SSL connections and thus requires SSL to be configured. */
2924 if (cfg && cfg->ssl) {
2925 const struct ovsrec_ssl *ssl = cfg->ssl;
2927 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2928 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2931 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2932 * usage has changed. */
2933 vlan_splinters_changed = false;
2934 if (vlan_splinters_enabled_anywhere) {
2937 HMAP_FOR_EACH (br, node, &all_bridges) {
2938 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2939 vlan_splinters_changed = true;
2945 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2946 struct ovsdb_idl_txn *txn;
2948 idl_seqno = ovsdb_idl_get_seqno(idl);
2949 txn = ovsdb_idl_txn_create(idl);
2950 bridge_reconfigure(cfg ? cfg : &null_cfg);
2953 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2954 discover_types(cfg);
2957 /* If we are completing our initial configuration for this run
2958 * of ovs-vswitchd, then keep the transaction around to monitor
2959 * it for completion. */
2960 if (initial_config_done) {
2961 /* Always sets the 'status_txn_try_again' to check again,
2962 * in case that this transaction fails. */
2963 status_txn_try_again = true;
2964 ovsdb_idl_txn_commit(txn);
2965 ovsdb_idl_txn_destroy(txn);
2967 initial_config_done = true;
2968 daemonize_txn = txn;
2972 if (daemonize_txn) {
2973 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2974 if (status != TXN_INCOMPLETE) {
2975 ovsdb_idl_txn_destroy(daemonize_txn);
2976 daemonize_txn = NULL;
2978 /* ovs-vswitchd has completed initialization, so allow the
2979 * process that forked us to exit successfully. */
2980 daemonize_complete();
2982 vlog_enable_async();
2984 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2989 run_status_update();
2999 ovsdb_idl_wait(idl);
3000 if (daemonize_txn) {
3001 ovsdb_idl_txn_wait(daemonize_txn);
3005 ofproto_enumerate_types(&types);
3006 SSET_FOR_EACH (type, &types) {
3007 ofproto_type_wait(type);
3009 sset_destroy(&types);
3011 if (!hmap_is_empty(&all_bridges)) {
3014 HMAP_FOR_EACH (br, node, &all_bridges) {
3015 ofproto_wait(br->ofproto);
3018 poll_timer_wait_until(stats_timer);
3021 status_update_wait();
3022 system_stats_wait();
3025 /* Adds some memory usage statistics for bridges into 'usage', for use with
3026 * memory_report(). */
3028 bridge_get_memory_usage(struct simap *usage)
3035 ofproto_enumerate_types(&types);
3036 SSET_FOR_EACH (type, &types) {
3037 ofproto_type_get_memory_usage(type, usage);
3039 sset_destroy(&types);
3041 HMAP_FOR_EACH (br, node, &all_bridges) {
3042 ofproto_get_memory_usage(br->ofproto, usage);
3046 /* QoS unixctl user interface functions. */
3048 struct qos_unixctl_show_cbdata {
3050 struct iface *iface;
3054 qos_unixctl_show_queue(unsigned int queue_id,
3055 const struct smap *details,
3056 struct iface *iface,
3059 struct netdev_queue_stats stats;
3060 struct smap_node *node;
3063 ds_put_cstr(ds, "\n");
3065 ds_put_format(ds, "Queue %u:\n", queue_id);
3067 ds_put_cstr(ds, "Default:\n");
3070 SMAP_FOR_EACH (node, details) {
3071 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3074 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3076 if (stats.tx_packets != UINT64_MAX) {
3077 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3080 if (stats.tx_bytes != UINT64_MAX) {
3081 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3084 if (stats.tx_errors != UINT64_MAX) {
3085 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3088 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3089 queue_id, ovs_strerror(error));
3094 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3095 const char *argv[], void *aux OVS_UNUSED)
3097 struct ds ds = DS_EMPTY_INITIALIZER;
3098 struct smap smap = SMAP_INITIALIZER(&smap);
3099 struct iface *iface;
3101 struct smap_node *node;
3103 iface = iface_find(argv[1]);
3105 unixctl_command_reply_error(conn, "no such interface");
3109 netdev_get_qos(iface->netdev, &type, &smap);
3111 if (*type != '\0') {
3112 struct netdev_queue_dump dump;
3113 struct smap details;
3114 unsigned int queue_id;
3116 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3118 SMAP_FOR_EACH (node, &smap) {
3119 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3122 smap_init(&details);
3123 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3124 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3126 smap_destroy(&details);
3128 unixctl_command_reply(conn, ds_cstr(&ds));
3130 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3131 unixctl_command_reply_error(conn, ds_cstr(&ds));
3134 smap_destroy(&smap);
3138 /* Bridge reconfiguration functions. */
3140 bridge_create(const struct ovsrec_bridge *br_cfg)
3144 ovs_assert(!bridge_lookup(br_cfg->name));
3145 br = xzalloc(sizeof *br);
3147 br->name = xstrdup(br_cfg->name);
3148 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3151 /* Derive the default Ethernet address from the bridge's UUID. This should
3152 * be unique and it will be stable between ovs-vswitchd runs. */
3153 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3154 eth_addr_mark_random(br->default_ea);
3156 hmap_init(&br->ports);
3157 hmap_init(&br->ifaces);
3158 hmap_init(&br->iface_by_name);
3159 hmap_init(&br->mirrors);
3161 hmap_init(&br->mappings);
3162 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3166 bridge_destroy(struct bridge *br)
3169 struct mirror *mirror, *next_mirror;
3170 struct port *port, *next_port;
3172 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3175 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3176 mirror_destroy(mirror);
3179 hmap_remove(&all_bridges, &br->node);
3180 ofproto_destroy(br->ofproto);
3181 hmap_destroy(&br->ifaces);
3182 hmap_destroy(&br->ports);
3183 hmap_destroy(&br->iface_by_name);
3184 hmap_destroy(&br->mirrors);
3185 hmap_destroy(&br->mappings);
3192 static struct bridge *
3193 bridge_lookup(const char *name)
3197 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3198 if (!strcmp(br->name, name)) {
3205 /* Handle requests for a listing of all flows known by the OpenFlow
3206 * stack, including those normally hidden. */
3208 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3209 const char *argv[], void *aux OVS_UNUSED)
3214 br = bridge_lookup(argv[1]);
3216 unixctl_command_reply_error(conn, "Unknown bridge");
3221 ofproto_get_all_flows(br->ofproto, &results);
3223 unixctl_command_reply(conn, ds_cstr(&results));
3224 ds_destroy(&results);
3227 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3228 * connections and reconnect. If BRIDGE is not specified, then all bridges
3229 * drop their controller connections and reconnect. */
3231 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3232 const char *argv[], void *aux OVS_UNUSED)
3236 br = bridge_lookup(argv[1]);
3238 unixctl_command_reply_error(conn, "Unknown bridge");
3241 ofproto_reconnect_controllers(br->ofproto);
3243 HMAP_FOR_EACH (br, node, &all_bridges) {
3244 ofproto_reconnect_controllers(br->ofproto);
3247 unixctl_command_reply(conn, NULL);
3251 bridge_get_controllers(const struct bridge *br,
3252 struct ovsrec_controller ***controllersp)
3254 struct ovsrec_controller **controllers;
3255 size_t n_controllers;
3257 controllers = br->cfg->controller;
3258 n_controllers = br->cfg->n_controller;
3260 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3266 *controllersp = controllers;
3268 return n_controllers;
3272 bridge_collect_wanted_ports(struct bridge *br,
3273 const unsigned long int *splinter_vlans,
3274 struct shash *wanted_ports)
3278 shash_init(wanted_ports);
3280 for (i = 0; i < br->cfg->n_ports; i++) {
3281 const char *name = br->cfg->ports[i]->name;
3282 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3283 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3287 if (bridge_get_controllers(br, NULL)
3288 && !shash_find(wanted_ports, br->name)) {
3289 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3290 br->name, br->name);
3292 ovsrec_interface_init(&br->synth_local_iface);
3293 ovsrec_port_init(&br->synth_local_port);
3295 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3296 br->synth_local_port.n_interfaces = 1;
3297 br->synth_local_port.name = br->name;
3299 br->synth_local_iface.name = br->name;
3300 br->synth_local_iface.type = "internal";
3302 br->synth_local_ifacep = &br->synth_local_iface;
3304 shash_add(wanted_ports, br->name, &br->synth_local_port);
3307 if (splinter_vlans) {
3308 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3312 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3313 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3314 * 'br' needs to complete its configuration. */
3316 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3318 struct shash_node *port_node;
3319 struct port *port, *next;
3321 /* Get rid of deleted ports.
3322 * Get rid of deleted interfaces on ports that still exist. */
3323 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3324 port->cfg = shash_find_data(wanted_ports, port->name);
3328 port_del_ifaces(port);
3332 /* Update iface->cfg and iface->type in interfaces that still exist. */
3333 SHASH_FOR_EACH (port_node, wanted_ports) {
3334 const struct ovsrec_port *port = port_node->data;
3337 for (i = 0; i < port->n_interfaces; i++) {
3338 const struct ovsrec_interface *cfg = port->interfaces[i];
3339 struct iface *iface = iface_lookup(br, cfg->name);
3340 const char *type = iface_get_type(cfg, br->cfg);
3345 } else if (!strcmp(type, "null")) {
3346 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3347 " may be removed in February 2013. Please email"
3348 " dev@openvswitch.org with concerns.",
3351 /* We will add new interfaces later. */
3357 /* Initializes 'oc' appropriately as a management service controller for
3360 * The caller must free oc->target when it is no longer needed. */
3362 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3363 struct ofproto_controller *oc)
3365 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3366 oc->max_backoff = 0;
3367 oc->probe_interval = 60;
3368 oc->band = OFPROTO_OUT_OF_BAND;
3370 oc->burst_limit = 0;
3371 oc->enable_async_msgs = true;
3375 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3377 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3378 struct ofproto_controller *oc)
3382 oc->target = c->target;
3383 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3384 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3385 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3386 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3387 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3388 oc->burst_limit = (c->controller_burst_limit
3389 ? *c->controller_burst_limit : 0);
3390 oc->enable_async_msgs = (!c->enable_async_messages
3391 || *c->enable_async_messages);
3392 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3393 if (dscp < 0 || dscp > 63) {
3394 dscp = DSCP_DEFAULT;
3399 /* Configures the IP stack for 'br''s local interface properly according to the
3400 * configuration in 'c'. */
3402 bridge_configure_local_iface_netdev(struct bridge *br,
3403 struct ovsrec_controller *c)
3405 struct netdev *netdev;
3406 struct in_addr mask, gateway;
3408 struct iface *local_iface;
3411 /* If there's no local interface or no IP address, give up. */
3412 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3413 if (!local_iface || !c->local_ip
3414 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3418 /* Bring up the local interface. */
3419 netdev = local_iface->netdev;
3420 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3422 /* Configure the IP address and netmask. */
3423 if (!c->local_netmask
3424 || !inet_pton(AF_INET, c->local_netmask, &mask)
3426 mask.s_addr = guess_netmask(ip.s_addr);
3428 if (!netdev_set_in4(netdev, ip, mask)) {
3429 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3430 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3433 /* Configure the default gateway. */
3434 if (c->local_gateway
3435 && inet_pton(AF_INET, c->local_gateway, &gateway)
3436 && gateway.s_addr) {
3437 if (!netdev_add_router(netdev, gateway)) {
3438 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3439 br->name, IP_ARGS(gateway.s_addr));
3444 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3445 * in either string are treated as equal to any number of slashes in the other,
3446 * e.g. "x///y" is equal to "x/y".
3448 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3449 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3450 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3451 * 'b' against a prefix of 'a'.
3454 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3456 const char *b_start = b;
3458 if (b - b_start >= b_stoplen) {
3460 } else if (*a != *b) {
3462 } else if (*a == '/') {
3463 a += strspn(a, "/");
3464 b += strspn(b, "/");
3465 } else if (*a == '\0') {
3475 bridge_configure_remotes(struct bridge *br,
3476 const struct sockaddr_in *managers, size_t n_managers)
3478 bool disable_in_band;
3480 struct ovsrec_controller **controllers;
3481 size_t n_controllers;
3483 enum ofproto_fail_mode fail_mode;
3485 struct ofproto_controller *ocs;
3489 /* Check if we should disable in-band control on this bridge. */
3490 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3493 /* Set OpenFlow queue ID for in-band control. */
3494 ofproto_set_in_band_queue(br->ofproto,
3495 smap_get_int(&br->cfg->other_config,
3496 "in-band-queue", -1));
3498 if (disable_in_band) {
3499 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3501 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3504 n_controllers = bridge_get_controllers(br, &controllers);
3506 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3509 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3510 for (i = 0; i < n_controllers; i++) {
3511 struct ovsrec_controller *c = controllers[i];
3513 if (!strncmp(c->target, "punix:", 6)
3514 || !strncmp(c->target, "unix:", 5)) {
3515 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3518 if (!strncmp(c->target, "unix:", 5)) {
3519 /* Connect to a listening socket */
3520 whitelist = xasprintf("unix:%s/", ovs_rundir());
3521 if (strchr(c->target, '/') &&
3522 !equal_pathnames(c->target, whitelist,
3523 strlen(whitelist))) {
3524 /* Absolute path specified, but not in ovs_rundir */
3525 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3526 "controller \"%s\" due to possibility for "
3527 "remote exploit. Instead, specify socket "
3528 "in whitelisted \"%s\" or connect to "
3529 "\"unix:%s/%s.mgmt\" (which is always "
3530 "available without special configuration).",
3531 br->name, c->target, whitelist,
3532 ovs_rundir(), br->name);
3537 whitelist = xasprintf("punix:%s/%s.controller",
3538 ovs_rundir(), br->name);
3539 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3540 /* Prevent remote ovsdb-server users from accessing
3541 * arbitrary Unix domain sockets and overwriting arbitrary
3543 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3544 "controller \"%s\" due to possibility of "
3545 "overwriting local files. Instead, specify "
3546 "whitelisted \"%s\" or connect to "
3547 "\"unix:%s/%s.mgmt\" (which is always "
3548 "available without special configuration).",
3549 br->name, c->target, whitelist,
3550 ovs_rundir(), br->name);
3559 bridge_configure_local_iface_netdev(br, c);
3560 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3561 if (disable_in_band) {
3562 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3567 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3568 bridge_get_allowed_versions(br));
3569 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3572 /* Set the fail-mode. */
3573 fail_mode = !br->cfg->fail_mode
3574 || !strcmp(br->cfg->fail_mode, "standalone")
3575 ? OFPROTO_FAIL_STANDALONE
3576 : OFPROTO_FAIL_SECURE;
3577 ofproto_set_fail_mode(br->ofproto, fail_mode);
3579 /* Configure OpenFlow controller connection snooping. */
3580 if (!ofproto_has_snoops(br->ofproto)) {
3584 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3585 ovs_rundir(), br->name));
3586 ofproto_set_snoops(br->ofproto, &snoops);
3587 sset_destroy(&snoops);
3592 bridge_configure_tables(struct bridge *br)
3594 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3598 n_tables = ofproto_get_n_tables(br->ofproto);
3600 for (i = 0; i < n_tables; i++) {
3601 struct ofproto_table_settings s;
3602 bool use_default_prefixes = true;
3605 s.max_flows = UINT_MAX;
3608 s.n_prefix_fields = 0;
3609 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3611 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3612 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3615 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3616 s.max_flows = *cfg->flow_limit;
3618 if (cfg->overflow_policy
3619 && !strcmp(cfg->overflow_policy, "evict")) {
3621 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3622 for (k = 0; k < cfg->n_groups; k++) {
3623 const char *string = cfg->groups[k];
3626 msg = mf_parse_subfield__(&s.groups[k], &string);
3628 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3629 "'groups' (%s)", br->name, i, msg);
3631 } else if (*string) {
3632 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3633 "element '%s' contains trailing garbage",
3634 br->name, i, cfg->groups[k]);
3640 /* Prefix lookup fields. */
3641 s.n_prefix_fields = 0;
3642 for (k = 0; k < cfg->n_prefixes; k++) {
3643 const char *name = cfg->prefixes[k];
3644 const struct mf_field *mf;
3646 if (strcmp(name, "none") == 0) {
3647 use_default_prefixes = false;
3648 s.n_prefix_fields = 0;
3651 mf = mf_from_name(name);
3653 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3657 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3658 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3659 "%s", br->name, name);
3662 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3663 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3664 "field not used: %s", br->name, name);
3667 use_default_prefixes = false;
3668 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3671 if (use_default_prefixes) {
3672 /* Use default values. */
3673 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3674 memcpy(s.prefix_fields, default_prefix_fields,
3675 sizeof default_prefix_fields);
3678 struct ds ds = DS_EMPTY_INITIALIZER;
3679 for (k = 0; k < s.n_prefix_fields; k++) {
3681 ds_put_char(&ds, ',');
3683 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3685 if (s.n_prefix_fields == 0) {
3686 ds_put_cstr(&ds, "none");
3688 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3689 br->name, i, ds_cstr(&ds));
3693 ofproto_configure_table(br->ofproto, i, &s);
3697 for (; j < br->cfg->n_flow_tables; j++) {
3698 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3699 "%"PRId64" not supported by this datapath", br->name,
3700 br->cfg->key_flow_tables[j]);
3705 bridge_configure_dp_desc(struct bridge *br)
3707 ofproto_set_dp_desc(br->ofproto,
3708 smap_get(&br->cfg->other_config, "dp-desc"));
3711 static struct aa_mapping *
3712 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3714 struct aa_mapping *m;
3716 HMAP_FOR_EACH_IN_BUCKET (m,
3718 hash_bytes(&isid, sizeof isid, 0),
3720 if (isid == m->isid) {
3727 static struct aa_mapping *
3728 bridge_aa_mapping_create(struct bridge *br,
3732 struct aa_mapping *m;
3734 m = xzalloc(sizeof *m);
3738 m->br_name = xstrdup(br->name);
3739 hmap_insert(&br->mappings,
3741 hash_bytes(&isid, sizeof isid, 0));
3747 bridge_aa_mapping_destroy(struct aa_mapping *m)
3750 struct bridge *br = m->bridge;
3753 ofproto_aa_mapping_unregister(br->ofproto, m);
3756 hmap_remove(&br->mappings, &m->hmap_node);
3765 bridge_aa_mapping_configure(struct aa_mapping *m)
3767 struct aa_mapping_settings s;
3773 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3779 bridge_configure_aa(struct bridge *br)
3781 const struct ovsdb_datum *mc;
3782 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3783 struct aa_settings aa_s;
3784 struct aa_mapping *m, *next;
3788 ofproto_set_aa(br->ofproto, NULL, NULL);
3792 memset(&aa_s, 0, sizeof aa_s);
3793 aa_s.system_description = auto_attach->system_description;
3794 aa_s.system_name = auto_attach->system_name;
3795 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3797 mc = ovsrec_autoattach_get_mappings(auto_attach,
3799 OVSDB_TYPE_INTEGER);
3800 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3801 union ovsdb_atom atom;
3803 atom.integer = m->isid;
3804 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3805 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3807 bridge_aa_mapping_destroy(m);
3811 /* Add new mappings and reconfigure existing ones. */
3812 for (i = 0; i < auto_attach->n_mappings; ++i) {
3813 struct aa_mapping *m =
3814 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3817 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3818 auto_attach->key_mappings[i],
3819 auto_attach->value_mappings[i]);
3820 m = bridge_aa_mapping_create(br,
3821 auto_attach->key_mappings[i],
3822 auto_attach->value_mappings[i]);
3824 if (!bridge_aa_mapping_configure(m)) {
3825 bridge_aa_mapping_destroy(m);
3832 bridge_aa_need_refresh(struct bridge *br)
3834 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3838 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3840 int64_t *trunks = NULL;
3842 bool found = false, reconfigure = false;
3844 for (i = 0; i < port->cfg->n_trunks; i++) {
3845 if (port->cfg->trunks[i] == m->vlan) {
3852 case BRIDGE_AA_VLAN_OPER_ADD:
3854 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3856 for (i = 0; i < port->cfg->n_trunks; i++) {
3857 trunks[i] = port->cfg->trunks[i];
3859 trunks[i++] = m->vlan;
3865 case BRIDGE_AA_VLAN_OPER_REMOVE:
3869 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3871 for (i = 0; i < port->cfg->n_trunks; i++) {
3872 if (port->cfg->trunks[i] != m->vlan) {
3873 trunks[j++] = port->cfg->trunks[i];
3882 case BRIDGE_AA_VLAN_OPER_UNDEF:
3884 VLOG_WARN("unrecognized operation %u", m->oper);
3889 /* VLAN switching under trunk mode cause the trunk port to switch all
3890 * VLANs, see ovs-vswitchd.conf.db
3893 static char *vlan_mode_access = "access";
3894 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3898 static char *vlan_mode_trunk = "trunk";
3899 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3902 ovsrec_port_set_trunks(port->cfg, trunks, i);
3904 /* Force reconfigure of the port. */
3905 port_configure(port);
3910 bridge_aa_refresh_queued(struct bridge *br)
3912 struct ovs_list *list = xmalloc(sizeof *list);
3913 struct bridge_aa_vlan *node;
3916 ofproto_aa_vlan_get_queued(br->ofproto, list);
3918 LIST_FOR_EACH(node, list_node, list) {
3921 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3924 port = port_lookup(br, node->port_name);
3926 bridge_aa_update_trunks(port, node);
3929 list_remove(&node->list_node);
3930 free(node->port_name);
3938 /* Port functions. */
3940 static struct port *
3941 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3945 port = xzalloc(sizeof *port);
3947 port->name = xstrdup(cfg->name);
3949 list_init(&port->ifaces);
3951 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3955 /* Deletes interfaces from 'port' that are no longer configured for it. */
3957 port_del_ifaces(struct port *port)
3959 struct iface *iface, *next;
3960 struct sset new_ifaces;
3963 /* Collect list of new interfaces. */
3964 sset_init(&new_ifaces);
3965 for (i = 0; i < port->cfg->n_interfaces; i++) {
3966 const char *name = port->cfg->interfaces[i]->name;
3967 const char *type = port->cfg->interfaces[i]->type;
3968 if (strcmp(type, "null")) {
3969 sset_add(&new_ifaces, name);
3973 /* Get rid of deleted interfaces. */
3974 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3975 if (!sset_contains(&new_ifaces, iface->name)) {
3976 iface_destroy(iface);
3980 sset_destroy(&new_ifaces);
3984 port_destroy(struct port *port)
3987 struct bridge *br = port->bridge;
3988 struct iface *iface, *next;
3991 ofproto_bundle_unregister(br->ofproto, port);
3994 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3995 iface_destroy__(iface);
3998 hmap_remove(&br->ports, &port->hmap_node);
4004 static struct port *
4005 port_lookup(const struct bridge *br, const char *name)
4009 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4011 if (!strcmp(port->name, name)) {
4019 enable_lacp(struct port *port, bool *activep)
4021 if (!port->cfg->lacp) {
4022 /* XXX when LACP implementation has been sufficiently tested, enable by
4023 * default and make active on bonded ports. */
4025 } else if (!strcmp(port->cfg->lacp, "off")) {
4027 } else if (!strcmp(port->cfg->lacp, "active")) {
4030 } else if (!strcmp(port->cfg->lacp, "passive")) {
4034 VLOG_WARN("port %s: unknown LACP mode %s",
4035 port->name, port->cfg->lacp);
4040 static struct lacp_settings *
4041 port_configure_lacp(struct port *port, struct lacp_settings *s)
4043 const char *lacp_time, *system_id;
4046 if (!enable_lacp(port, &s->active)) {
4050 s->name = port->name;
4052 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4054 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4055 ETH_ADDR_SCAN_ARGS(s->id))) {
4056 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4057 " address.", port->name, system_id);
4061 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4064 if (eth_addr_is_zero(s->id)) {
4065 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4069 /* Prefer bondable links if unspecified. */
4070 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4072 s->priority = (priority > 0 && priority <= UINT16_MAX
4074 : UINT16_MAX - !list_is_short(&port->ifaces));
4076 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4077 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4079 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4080 "lacp-fallback-ab", false);
4086 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4088 int priority, portid, key;
4090 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4091 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4093 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4095 if (portid <= 0 || portid > UINT16_MAX) {
4096 portid = ofp_to_u16(iface->ofp_port);
4099 if (priority <= 0 || priority > UINT16_MAX) {
4100 priority = UINT16_MAX;
4103 if (key < 0 || key > UINT16_MAX) {
4107 s->name = iface->name;
4109 s->priority = priority;
4114 port_configure_bond(struct port *port, struct bond_settings *s)
4116 const char *detect_s;
4117 struct iface *iface;
4119 int miimon_interval;
4121 s->name = port->name;
4123 if (port->cfg->bond_mode) {
4124 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4125 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4126 port->name, port->cfg->bond_mode,
4127 bond_mode_to_string(s->balance));
4130 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4132 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4133 * active-backup. At some point we should remove this warning. */
4134 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4135 " in previous versions, the default bond_mode was"
4136 " balance-slb", port->name,
4137 bond_mode_to_string(s->balance));
4139 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4140 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4141 "please use another bond type or disable flood_vlans",
4145 miimon_interval = smap_get_int(&port->cfg->other_config,
4146 "bond-miimon-interval", 0);
4147 if (miimon_interval <= 0) {
4148 miimon_interval = 200;
4151 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4152 if (!detect_s || !strcmp(detect_s, "carrier")) {
4153 miimon_interval = 0;
4154 } else if (strcmp(detect_s, "miimon")) {
4155 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4156 "defaulting to carrier", port->name, detect_s);
4157 miimon_interval = 0;
4160 s->up_delay = MAX(0, port->cfg->bond_updelay);
4161 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4162 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4163 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4164 "bond-rebalance-interval", 10000);
4165 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4166 s->rebalance_interval = 1000;
4169 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4170 "lacp-fallback-ab", false);
4172 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4173 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4176 mac_s = port->cfg->bond_active_slave;
4177 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4178 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4179 /* OVSDB did not store the last active interface */
4180 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4184 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4185 * instead of obtaining it from the database. */
4187 port_is_synthetic(const struct port *port)
4189 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4192 /* Interface functions. */
4195 iface_is_internal(const struct ovsrec_interface *iface,
4196 const struct ovsrec_bridge *br)
4198 /* The local port and "internal" ports are always "internal". */
4199 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4202 /* Returns the correct network device type for interface 'iface' in bridge
4205 iface_get_type(const struct ovsrec_interface *iface,
4206 const struct ovsrec_bridge *br)
4210 /* The local port always has type "internal". Other ports take
4211 * their type from the database and default to "system" if none is
4213 if (iface_is_internal(iface, br)) {
4216 type = iface->type[0] ? iface->type : "system";
4219 return ofproto_port_open_type(br->datapath_type, type);
4223 iface_destroy__(struct iface *iface)
4226 struct port *port = iface->port;
4227 struct bridge *br = port->bridge;
4229 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4230 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4233 if (iface->ofp_port != OFPP_NONE) {
4234 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4237 list_remove(&iface->port_elem);
4238 hmap_remove(&br->iface_by_name, &iface->name_node);
4240 /* The user is changing configuration here, so netdev_remove needs to be
4241 * used as opposed to netdev_close */
4242 netdev_remove(iface->netdev);
4250 iface_destroy(struct iface *iface)
4253 struct port *port = iface->port;
4255 iface_destroy__(iface);
4256 if (list_is_empty(&port->ifaces)) {
4262 static struct iface *
4263 iface_lookup(const struct bridge *br, const char *name)
4265 struct iface *iface;
4267 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4268 &br->iface_by_name) {
4269 if (!strcmp(iface->name, name)) {
4277 static struct iface *
4278 iface_find(const char *name)
4280 const struct bridge *br;
4282 HMAP_FOR_EACH (br, node, &all_bridges) {
4283 struct iface *iface = iface_lookup(br, name);
4292 static struct iface *
4293 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4295 struct iface *iface;
4297 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4299 if (iface->ofp_port == ofp_port) {
4306 /* Set Ethernet address of 'iface', if one is specified in the configuration
4309 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4311 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4312 struct iface *hw_addr_iface;
4314 if (strcmp(iface->type, "internal")) {
4318 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4320 } else if (port->cfg->fake_bridge) {
4321 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4322 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4327 if (iface->ofp_port == OFPP_LOCAL) {
4328 VLOG_ERR("interface %s: ignoring mac in Interface record "
4329 "(use Bridge record to set local port's mac)",
4331 } else if (eth_addr_is_multicast(mac)) {
4332 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4335 int error = netdev_set_etheraddr(iface->netdev, mac);
4337 VLOG_ERR("interface %s: setting MAC failed (%s)",
4338 iface->name, ovs_strerror(error));
4344 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4346 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4348 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4349 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4350 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4354 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4355 * sets the "ofport" field to -1.
4357 * This is appropriate when 'if_cfg''s interface cannot be created or is
4358 * otherwise invalid. */
4360 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4362 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4363 iface_set_ofport(if_cfg, OFPP_NONE);
4364 ovsrec_interface_set_error(if_cfg, errp);
4365 ovsrec_interface_set_status(if_cfg, NULL);
4366 ovsrec_interface_set_admin_state(if_cfg, NULL);
4367 ovsrec_interface_set_duplex(if_cfg, NULL);
4368 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4369 ovsrec_interface_set_link_state(if_cfg, NULL);
4370 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4371 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4372 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4373 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4374 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4375 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4376 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4377 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4382 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4384 union ovsdb_atom atom;
4386 atom.integer = target;
4387 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4391 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4393 struct ofpbuf queues_buf;
4395 ofpbuf_init(&queues_buf, 0);
4397 if (!qos || qos->type[0] == '\0') {
4398 netdev_set_qos(iface->netdev, NULL, NULL);
4400 const struct ovsdb_datum *queues;
4401 struct netdev_queue_dump dump;
4402 unsigned int queue_id;
4403 struct smap details;
4407 /* Configure top-level Qos for 'iface'. */
4408 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4410 /* Deconfigure queues that were deleted. */
4411 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4413 smap_init(&details);
4414 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4415 if (!queue_ids_include(queues, queue_id)) {
4416 netdev_delete_queue(iface->netdev, queue_id);
4419 smap_destroy(&details);
4421 /* Configure queues for 'iface'. */
4423 for (i = 0; i < qos->n_queues; i++) {
4424 const struct ovsrec_queue *queue = qos->value_queues[i];
4425 unsigned int queue_id = qos->key_queues[i];
4427 if (queue_id == 0) {
4431 if (queue->n_dscp == 1) {
4432 struct ofproto_port_queue *port_queue;
4434 port_queue = ofpbuf_put_uninit(&queues_buf,
4435 sizeof *port_queue);
4436 port_queue->queue = queue_id;
4437 port_queue->dscp = queue->dscp[0];
4440 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4443 struct smap details;
4445 smap_init(&details);
4446 netdev_set_queue(iface->netdev, 0, &details);
4447 smap_destroy(&details);
4451 if (iface->ofp_port != OFPP_NONE) {
4452 const struct ofproto_port_queue *port_queues = queues_buf.data;
4453 size_t n_queues = queues_buf.size / sizeof *port_queues;
4455 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4456 port_queues, n_queues);
4459 netdev_set_policing(iface->netdev,
4460 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4461 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4463 ofpbuf_uninit(&queues_buf);
4467 iface_configure_cfm(struct iface *iface)
4469 const struct ovsrec_interface *cfg = iface->cfg;
4470 const char *opstate_str;
4471 const char *cfm_ccm_vlan;
4472 struct cfm_settings s;
4473 struct smap netdev_args;
4475 if (!cfg->n_cfm_mpid) {
4476 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4480 s.check_tnl_key = false;
4481 smap_init(&netdev_args);
4482 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4483 const char *key = smap_get(&netdev_args, "key");
4484 const char *in_key = smap_get(&netdev_args, "in_key");
4486 s.check_tnl_key = (key && !strcmp(key, "flow"))
4487 || (in_key && !strcmp(in_key, "flow"));
4489 smap_destroy(&netdev_args);
4491 s.mpid = *cfg->cfm_mpid;
4492 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4493 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4494 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4496 if (s.interval <= 0) {
4500 if (!cfm_ccm_vlan) {
4502 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4503 s.ccm_vlan = CFM_RANDOM_VLAN;
4505 s.ccm_vlan = atoi(cfm_ccm_vlan);
4506 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4511 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4513 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4515 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4516 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4518 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4521 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4522 * instead of obtaining it from the database. */
4524 iface_is_synthetic(const struct iface *iface)
4526 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4530 iface_validate_ofport__(size_t n, int64_t *ofport)
4532 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4533 ? u16_to_ofp(*ofport)
4538 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4540 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4544 iface_pick_ofport(const struct ovsrec_interface *cfg)
4546 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4547 return (requested_ofport != OFPP_NONE
4549 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4552 /* Port mirroring. */
4554 static struct mirror *
4555 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4559 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4560 if (uuid_equals(uuid, &m->uuid)) {
4568 bridge_configure_mirrors(struct bridge *br)
4570 const struct ovsdb_datum *mc;
4571 unsigned long *flood_vlans;
4572 struct mirror *m, *next;
4575 /* Get rid of deleted mirrors. */
4576 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4577 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4578 union ovsdb_atom atom;
4580 atom.uuid = m->uuid;
4581 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4586 /* Add new mirrors and reconfigure existing ones. */
4587 for (i = 0; i < br->cfg->n_mirrors; i++) {
4588 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4589 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4591 m = mirror_create(br, cfg);
4594 if (!mirror_configure(m)) {
4599 /* Update flooded vlans (for RSPAN). */
4600 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4601 br->cfg->n_flood_vlans);
4602 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4603 bitmap_free(flood_vlans);
4606 static struct mirror *
4607 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4611 m = xzalloc(sizeof *m);
4612 m->uuid = cfg->header_.uuid;
4613 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4615 m->name = xstrdup(cfg->name);
4621 mirror_destroy(struct mirror *m)
4624 struct bridge *br = m->bridge;
4627 ofproto_mirror_unregister(br->ofproto, m);
4630 hmap_remove(&br->mirrors, &m->hmap_node);
4637 mirror_collect_ports(struct mirror *m,
4638 struct ovsrec_port **in_ports, int n_in_ports,
4639 void ***out_portsp, size_t *n_out_portsp)
4641 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4642 size_t n_out_ports = 0;
4645 for (i = 0; i < n_in_ports; i++) {
4646 const char *name = in_ports[i]->name;
4647 struct port *port = port_lookup(m->bridge, name);
4649 out_ports[n_out_ports++] = port;
4651 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4652 "port %s", m->bridge->name, m->name, name);
4655 *out_portsp = out_ports;
4656 *n_out_portsp = n_out_ports;
4660 mirror_configure(struct mirror *m)
4662 const struct ovsrec_mirror *cfg = m->cfg;
4663 struct ofproto_mirror_settings s;
4666 if (strcmp(cfg->name, m->name)) {
4668 m->name = xstrdup(cfg->name);
4672 /* Get output port or VLAN. */
4673 if (cfg->output_port) {
4674 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4675 if (!s.out_bundle) {
4676 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4677 m->bridge->name, m->name);
4680 s.out_vlan = UINT16_MAX;
4682 if (cfg->output_vlan) {
4683 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4684 "output vlan; ignoring output vlan",
4685 m->bridge->name, m->name);
4687 } else if (cfg->output_vlan) {
4688 /* The database should prevent invalid VLAN values. */
4689 s.out_bundle = NULL;
4690 s.out_vlan = *cfg->output_vlan;
4692 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4693 m->bridge->name, m->name);
4697 /* Get port selection. */
4698 if (cfg->select_all) {
4699 size_t n_ports = hmap_count(&m->bridge->ports);
4700 void **ports = xmalloc(n_ports * sizeof *ports);
4705 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4715 /* Get ports, dropping ports that don't exist.
4716 * The IDL ensures that there are no duplicates. */
4717 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4718 &s.srcs, &s.n_srcs);
4719 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4720 &s.dsts, &s.n_dsts);
4723 /* Get VLAN selection. */
4724 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4727 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4730 if (s.srcs != s.dsts) {
4739 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4741 * This is deprecated. It is only for compatibility with broken device drivers
4742 * in old versions of Linux that do not properly support VLANs when VLAN
4743 * devices are not used. When broken device drivers are no longer in
4744 * widespread use, we will delete these interfaces. */
4746 static struct ovsrec_port **recs;
4747 static size_t n_recs, allocated_recs;
4749 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4750 * splinters are reconfigured. */
4752 register_rec(struct ovsrec_port *rec)
4754 if (n_recs >= allocated_recs) {
4755 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4757 recs[n_recs++] = rec;
4760 /* Frees all of the ports registered with register_reg(). */
4762 free_registered_recs(void)
4766 for (i = 0; i < n_recs; i++) {
4767 struct ovsrec_port *port = recs[i];
4770 for (j = 0; j < port->n_interfaces; j++) {
4771 struct ovsrec_interface *iface = port->interfaces[j];
4776 smap_destroy(&port->other_config);
4777 free(port->interfaces);
4785 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4788 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4790 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4794 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4797 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4798 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4799 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4802 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4803 * use, returns NULL.
4805 * Updates 'vlan_splinters_enabled_anywhere'. */
4806 static unsigned long int *
4807 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4809 unsigned long int *splinter_vlans;
4810 struct sset splinter_ifaces;
4811 const char *real_dev_name;
4812 struct shash *real_devs;
4813 struct shash_node *node;
4817 /* Free space allocated for synthesized ports and interfaces, since we're
4818 * in the process of reconstructing all of them. */
4819 free_registered_recs();
4821 splinter_vlans = bitmap_allocate(4096);
4822 sset_init(&splinter_ifaces);
4823 vlan_splinters_enabled_anywhere = false;
4824 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4825 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4828 for (j = 0; j < br_cfg->n_ports; j++) {
4829 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4832 for (k = 0; k < port_cfg->n_interfaces; k++) {
4833 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4835 if (vlan_splinters_is_enabled(iface_cfg)) {
4836 vlan_splinters_enabled_anywhere = true;
4837 sset_add(&splinter_ifaces, iface_cfg->name);
4838 vlan_bitmap_from_array__(port_cfg->trunks,
4844 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4845 bitmap_set1(splinter_vlans, *port_cfg->tag);
4850 if (!vlan_splinters_enabled_anywhere) {
4851 free(splinter_vlans);
4852 sset_destroy(&splinter_ifaces);
4856 HMAP_FOR_EACH (br, node, &all_bridges) {
4858 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4862 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4863 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4864 * device to be created for it. */
4865 bitmap_set0(splinter_vlans, 0);
4866 bitmap_set0(splinter_vlans, 4095);
4868 /* Delete all VLAN devices that we don't need. */
4870 real_devs = vlandev_get_real_devs();
4871 SHASH_FOR_EACH (node, real_devs) {
4872 const struct vlan_real_dev *real_dev = node->data;
4873 const struct vlan_dev *vlan_dev;
4874 bool real_dev_has_splinters;
4876 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4878 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4879 if (!real_dev_has_splinters
4880 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4881 struct netdev *netdev;
4883 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4884 if (!netdev_get_in4(netdev, NULL, NULL) ||
4885 !netdev_get_in6(netdev, NULL)) {
4886 /* It has an IP address configured, so we don't own
4887 * it. Don't delete it. */
4889 vlandev_del(vlan_dev->name);
4891 netdev_close(netdev);
4898 /* Add all VLAN devices that we need. */
4899 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4902 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4903 if (!vlandev_get_name(real_dev_name, vid)) {
4904 vlandev_add(real_dev_name, vid);
4911 sset_destroy(&splinter_ifaces);
4913 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4914 free(splinter_vlans);
4917 return splinter_vlans;
4920 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4923 configure_splinter_port(struct port *port)
4925 struct ofproto *ofproto = port->bridge->ofproto;
4926 ofp_port_t realdev_ofp_port;
4927 const char *realdev_name;
4928 struct iface *vlandev, *realdev;
4930 ofproto_bundle_unregister(port->bridge->ofproto, port);
4932 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4935 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4936 realdev = iface_lookup(port->bridge, realdev_name);
4937 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4939 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4943 static struct ovsrec_port *
4944 synthesize_splinter_port(const char *real_dev_name,
4945 const char *vlan_dev_name, int vid)
4947 struct ovsrec_interface *iface;
4948 struct ovsrec_port *port;
4950 iface = xmalloc(sizeof *iface);
4951 ovsrec_interface_init(iface);
4952 iface->name = xstrdup(vlan_dev_name);
4953 iface->type = "system";
4955 port = xmalloc(sizeof *port);
4956 ovsrec_port_init(port);
4957 port->interfaces = xmemdup(&iface, sizeof iface);
4958 port->n_interfaces = 1;
4959 port->name = xstrdup(vlan_dev_name);
4960 port->vlan_mode = "splinter";
4961 port->tag = xmalloc(sizeof *port->tag);
4964 smap_add(&port->other_config, "realdev", real_dev_name);
4970 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4971 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4972 * 1-bit in the 'splinter_vlans' bitmap. */
4974 add_vlan_splinter_ports(struct bridge *br,
4975 const unsigned long int *splinter_vlans,
4976 struct shash *ports)
4980 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4981 * we're modifying 'ports'. */
4982 for (i = 0; i < br->cfg->n_ports; i++) {
4983 const char *name = br->cfg->ports[i]->name;
4984 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4987 for (j = 0; j < port_cfg->n_interfaces; j++) {
4988 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4990 if (vlan_splinters_is_enabled(iface_cfg)) {
4991 const char *real_dev_name;
4994 real_dev_name = iface_cfg->name;
4995 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4996 const char *vlan_dev_name;
4998 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5000 && !shash_find(ports, vlan_dev_name)) {
5001 shash_add(ports, vlan_dev_name,
5002 synthesize_splinter_port(
5003 real_dev_name, vlan_dev_name, vid));
5012 mirror_refresh_stats(struct mirror *m)
5014 struct ofproto *ofproto = m->bridge->ofproto;
5015 uint64_t tx_packets, tx_bytes;
5016 const char *keys[2];
5018 size_t stat_cnt = 0;
5020 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5021 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5025 if (tx_packets != UINT64_MAX) {
5026 keys[stat_cnt] = "tx_packets";
5027 values[stat_cnt] = tx_packets;
5030 if (tx_bytes != UINT64_MAX) {
5031 keys[stat_cnt] = "tx_bytes";
5032 values[stat_cnt] = tx_bytes;
5036 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5040 * Add registered netdev and dpif types to ovsdb to allow external
5041 * applications to query the capabilities of the Open vSwitch instance
5042 * running on the node.
5045 discover_types(const struct ovsrec_open_vswitch *cfg)
5049 /* Datapath types. */
5051 dp_enumerate_types(&types);
5052 const char **datapath_types = sset_array(&types);
5053 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5054 sset_count(&types));
5055 free(datapath_types);
5056 sset_destroy(&types);
5060 netdev_enumerate_types(&types);
5061 const char **iface_types = sset_array(&types);
5062 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5064 sset_destroy(&types);