1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
64 VLOG_DEFINE_THIS_MODULE(bridge);
66 COVERAGE_DEFINE(bridge_reconfigure);
69 /* These members are always valid.
71 * They are immutable: they never change between iface_create() and
73 struct list port_elem; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
76 struct port *port; /* Containing port. */
77 char *name; /* Host network device name. */
78 struct netdev *netdev; /* Network device. */
79 ofp_port_t ofp_port; /* OpenFlow port number. */
82 /* These members are valid only within bridge_reconfigure(). */
83 const char *type; /* Usually same as cfg->type. */
84 const struct ovsrec_interface *cfg;
88 struct uuid uuid; /* UUID of this "mirror" record in database. */
89 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
90 struct bridge *bridge;
92 const struct ovsrec_mirror *cfg;
96 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
97 struct bridge *bridge;
100 const struct ovsrec_port *cfg;
102 /* An ordinary bridge port has 1 interface.
103 * A bridge port for bonding has at least 2 interfaces. */
104 struct list ifaces; /* List of "struct iface"s. */
108 struct hmap_node node; /* In 'all_bridges'. */
109 char *name; /* User-specified arbitrary name. */
110 char *type; /* Datapath type. */
111 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
112 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
113 const struct ovsrec_bridge *cfg;
115 /* OpenFlow switch processing. */
116 struct ofproto *ofproto; /* OpenFlow switch. */
119 struct hmap ports; /* "struct port"s indexed by name. */
120 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
121 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
123 /* Port mirroring. */
124 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
126 /* Used during reconfiguration. */
127 struct shash wanted_ports;
129 /* Synthetic local port if necessary. */
130 struct ovsrec_port synth_local_port;
131 struct ovsrec_interface synth_local_iface;
132 struct ovsrec_interface *synth_local_ifacep;
135 /* All bridges, indexed by name. */
136 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
138 /* OVSDB IDL used to obtain configuration. */
139 static struct ovsdb_idl *idl;
141 /* We want to complete daemonization, fully detaching from our parent process,
142 * only after we have completed our initial configuration, committed our state
143 * to the database, and received confirmation back from the database server
144 * that it applied the commit. This allows our parent process to know that,
145 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
146 * to have already been filled in. (It is only "very likely" rather than
147 * certain because there is always a slim possibility that the transaction will
148 * fail or that some other client has added new bridges, ports, etc. while
149 * ovs-vswitchd was configuring using an old configuration.)
151 * We only need to do this once for our initial configuration at startup, so
152 * 'initial_config_done' tracks whether we've already done it. While we are
153 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
154 * itself and is otherwise NULL. */
155 static bool initial_config_done;
156 static struct ovsdb_idl_txn *daemonize_txn;
158 /* Most recently processed IDL sequence number. */
159 static unsigned int idl_seqno;
161 /* Track changes to port connectivity. */
162 static uint64_t connectivity_seqno = LLONG_MIN;
164 /* Status update to database.
166 * Some information in the database must be kept as up-to-date as possible to
167 * allow controllers to respond rapidly to network outages. Those status are
168 * updated via the 'status_txn'.
170 * We use the global connectivity sequence number to detect the status change.
171 * Also, to prevent the status update from sending too much to the database,
172 * we check the return status of each update transaction and do not start new
173 * update if the previous transaction status is 'TXN_INCOMPLETE'.
175 * 'status_txn' is NULL if there is no ongoing status update.
177 * If the previous database transaction was incomplete or failed (is not
178 * 'TXN_SUCCESS' or 'TXN_UNCHANGED'), 'force_status_commit' is set to true.
179 * This means that 'status_txn' must be committed next iteration of bridge_run()
180 * even if the connectivity or netdev sequence numbers do not change.
182 static struct ovsdb_idl_txn *status_txn;
183 static bool force_status_commit = true;
185 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
186 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
187 #define STATUS_CHECK_AGAIN_MSEC 100
189 /* Each time this timer expires, the bridge fetches interface and mirror
190 * statistics and pushes them into the database. */
191 static int stats_timer_interval;
192 static long long int stats_timer = LLONG_MIN;
194 /* Current stats database transaction, NULL if there is no ongoing
196 static struct ovsdb_idl_txn *stats_txn;
198 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
199 * expensive. This can cause bridge_reconfigure() to take a long time during
200 * which no other work can be done. To deal with this problem, we limit port
201 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
202 * call to bridge_reconfigure(). If there is more work to do after the limit
203 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
204 * This allows the rest of the code to catch up on important things like
205 * forwarding packets. */
206 #define OFP_PORT_ACTION_WINDOW 10
208 static void add_del_bridges(const struct ovsrec_open_vswitch *);
209 static void bridge_run__(void);
210 static void bridge_create(const struct ovsrec_bridge *);
211 static void bridge_destroy(struct bridge *);
212 static struct bridge *bridge_lookup(const char *name);
213 static unixctl_cb_func bridge_unixctl_dump_flows;
214 static unixctl_cb_func bridge_unixctl_reconnect;
215 static size_t bridge_get_controllers(const struct bridge *br,
216 struct ovsrec_controller ***controllersp);
217 static void bridge_collect_wanted_ports(struct bridge *,
218 const unsigned long *splinter_vlans,
219 struct shash *wanted_ports);
220 static void bridge_delete_ofprotos(void);
221 static void bridge_delete_or_reconfigure_ports(struct bridge *);
222 static void bridge_del_ports(struct bridge *,
223 const struct shash *wanted_ports);
224 static void bridge_add_ports(struct bridge *,
225 const struct shash *wanted_ports);
227 static void bridge_configure_datapath_id(struct bridge *);
228 static void bridge_configure_netflow(struct bridge *);
229 static void bridge_configure_forward_bpdu(struct bridge *);
230 static void bridge_configure_mac_table(struct bridge *);
231 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
232 static void bridge_configure_ipfix(struct bridge *);
233 static void bridge_configure_stp(struct bridge *);
234 static void bridge_configure_tables(struct bridge *);
235 static void bridge_configure_dp_desc(struct bridge *);
236 static void bridge_configure_remotes(struct bridge *,
237 const struct sockaddr_in *managers,
239 static void bridge_pick_local_hw_addr(struct bridge *,
240 uint8_t ea[ETH_ADDR_LEN],
241 struct iface **hw_addr_iface);
242 static uint64_t bridge_pick_datapath_id(struct bridge *,
243 const uint8_t bridge_ea[ETH_ADDR_LEN],
244 struct iface *hw_addr_iface);
245 static uint64_t dpid_from_hash(const void *, size_t nbytes);
246 static bool bridge_has_bond_fake_iface(const struct bridge *,
248 static bool port_is_bond_fake_iface(const struct port *);
250 static unixctl_cb_func qos_unixctl_show;
252 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
253 static void port_del_ifaces(struct port *);
254 static void port_destroy(struct port *);
255 static struct port *port_lookup(const struct bridge *, const char *name);
256 static void port_configure(struct port *);
257 static struct lacp_settings *port_configure_lacp(struct port *,
258 struct lacp_settings *);
259 static void port_configure_bond(struct port *, struct bond_settings *);
260 static bool port_is_synthetic(const struct port *);
262 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
263 static void run_system_stats(void);
265 static void bridge_configure_mirrors(struct bridge *);
266 static struct mirror *mirror_create(struct bridge *,
267 const struct ovsrec_mirror *);
268 static void mirror_destroy(struct mirror *);
269 static bool mirror_configure(struct mirror *);
270 static void mirror_refresh_stats(struct mirror *);
272 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
273 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
274 const struct ovsrec_port *);
275 static bool iface_is_internal(const struct ovsrec_interface *iface,
276 const struct ovsrec_bridge *br);
277 static const char *iface_get_type(const struct ovsrec_interface *,
278 const struct ovsrec_bridge *);
279 static void iface_destroy(struct iface *);
280 static void iface_destroy__(struct iface *);
281 static struct iface *iface_lookup(const struct bridge *, const char *name);
282 static struct iface *iface_find(const char *name);
283 static struct iface *iface_from_ofp_port(const struct bridge *,
284 ofp_port_t ofp_port);
285 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
286 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
287 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
288 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
289 static void iface_configure_cfm(struct iface *);
290 static void iface_refresh_cfm_stats(struct iface *);
291 static void iface_refresh_stats(struct iface *);
292 static void iface_refresh_netdev_status(struct iface *);
293 static void iface_refresh_ofproto_status(struct iface *);
294 static bool iface_is_synthetic(const struct iface *);
295 static ofp_port_t iface_get_requested_ofp_port(
296 const struct ovsrec_interface *);
297 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
299 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
301 * This is deprecated. It is only for compatibility with broken device drivers
302 * in old versions of Linux that do not properly support VLANs when VLAN
303 * devices are not used. When broken device drivers are no longer in
304 * widespread use, we will delete these interfaces. */
306 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
307 static bool vlan_splinters_enabled_anywhere;
309 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
310 static unsigned long int *collect_splinter_vlans(
311 const struct ovsrec_open_vswitch *);
312 static void configure_splinter_port(struct port *);
313 static void add_vlan_splinter_ports(struct bridge *,
314 const unsigned long int *splinter_vlans,
315 struct shash *ports);
318 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
320 struct shash iface_hints;
321 static bool initialized = false;
328 shash_init(&iface_hints);
331 for (i = 0; i < cfg->n_bridges; i++) {
332 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
335 for (j = 0; j < br_cfg->n_ports; j++) {
336 struct ovsrec_port *port_cfg = br_cfg->ports[j];
339 for (k = 0; k < port_cfg->n_interfaces; k++) {
340 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
341 struct iface_hint *iface_hint;
343 iface_hint = xmalloc(sizeof *iface_hint);
344 iface_hint->br_name = br_cfg->name;
345 iface_hint->br_type = br_cfg->datapath_type;
346 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
348 shash_add(&iface_hints, if_cfg->name, iface_hint);
354 ofproto_init(&iface_hints);
356 shash_destroy_free_data(&iface_hints);
360 /* Public functions. */
362 /* Initializes the bridge module, configuring it to obtain its configuration
363 * from an OVSDB server accessed over 'remote', which should be a string in a
364 * form acceptable to ovsdb_idl_create(). */
366 bridge_init(const char *remote)
368 /* Create connection to database. */
369 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
370 idl_seqno = ovsdb_idl_get_seqno(idl);
371 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
372 ovsdb_idl_verify_write_only(idl);
374 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
375 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
376 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
377 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
378 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
379 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
380 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
382 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
383 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
384 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
386 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
387 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
388 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
390 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
391 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
407 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
408 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
409 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
411 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
412 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
413 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
414 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
416 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
418 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
420 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
421 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
423 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
424 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
425 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
426 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
428 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
429 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
430 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
431 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
432 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
434 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
436 /* Register unixctl commands. */
437 unixctl_command_register("qos/show", "interface", 1, 1,
438 qos_unixctl_show, NULL);
439 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
440 bridge_unixctl_dump_flows, NULL);
441 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
442 bridge_unixctl_reconnect, NULL);
452 struct bridge *br, *next_br;
454 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
457 ovsdb_idl_destroy(idl);
460 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
461 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
462 * responsible for freeing '*managersp' (with free()).
464 * You may be asking yourself "why does ovs-vswitchd care?", because
465 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
466 * should not be and in fact is not directly involved in that. But
467 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
468 * it has to tell in-band control where the managers are to enable that.
469 * (Thus, only managers connected in-band are collected.)
472 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
473 struct sockaddr_in **managersp, size_t *n_managersp)
475 struct sockaddr_in *managers = NULL;
476 size_t n_managers = 0;
480 /* Collect all of the potential targets from the "targets" columns of the
481 * rows pointed to by "manager_options", excluding any that are
484 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
485 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
487 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
488 sset_find_and_delete(&targets, m->target);
490 sset_add(&targets, m->target);
494 /* Now extract the targets' IP addresses. */
495 if (!sset_is_empty(&targets)) {
498 managers = xmalloc(sset_count(&targets) * sizeof *managers);
499 SSET_FOR_EACH (target, &targets) {
500 struct sockaddr_storage ss;
502 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
504 && ss.ss_family == AF_INET) {
505 managers[n_managers++] = *(struct sockaddr_in *) &ss;
509 sset_destroy(&targets);
511 *managersp = managers;
512 *n_managersp = n_managers;
516 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
518 unsigned long int *splinter_vlans;
519 struct sockaddr_in *managers;
520 struct bridge *br, *next;
521 int sflow_bridge_number;
524 COVERAGE_INC(bridge_reconfigure);
526 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
527 OFPROTO_FLOW_LIMIT_DEFAULT));
528 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
529 OFPROTO_MAX_IDLE_DEFAULT));
532 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
533 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
535 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
536 * to 'ovs_cfg', with only very minimal configuration otherwise.
538 * This is mostly an update to bridge data structures. Nothing is pushed
539 * down to ofproto or lower layers. */
540 add_del_bridges(ovs_cfg);
541 splinter_vlans = collect_splinter_vlans(ovs_cfg);
542 HMAP_FOR_EACH (br, node, &all_bridges) {
543 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
544 bridge_del_ports(br, &br->wanted_ports);
546 free(splinter_vlans);
548 /* Start pushing configuration changes down to the ofproto layer:
550 * - Delete ofprotos that are no longer configured.
552 * - Delete ports that are no longer configured.
554 * - Reconfigure existing ports to their desired configurations, or
555 * delete them if not possible.
557 * We have to do all the deletions before we can do any additions, because
558 * the ports to be added might require resources that will be freed up by
559 * deletions (they might especially overlap in name). */
560 bridge_delete_ofprotos();
561 HMAP_FOR_EACH (br, node, &all_bridges) {
563 bridge_delete_or_reconfigure_ports(br);
567 /* Finish pushing configuration changes to the ofproto layer:
569 * - Create ofprotos that are missing.
571 * - Add ports that are missing. */
572 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
576 error = ofproto_create(br->name, br->type, &br->ofproto);
578 VLOG_ERR("failed to create bridge %s: %s", br->name,
579 ovs_strerror(error));
580 shash_destroy(&br->wanted_ports);
585 HMAP_FOR_EACH (br, node, &all_bridges) {
586 bridge_add_ports(br, &br->wanted_ports);
587 shash_destroy(&br->wanted_ports);
590 reconfigure_system_stats(ovs_cfg);
592 /* Complete the configuration. */
593 sflow_bridge_number = 0;
594 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
595 HMAP_FOR_EACH (br, node, &all_bridges) {
598 /* We need the datapath ID early to allow LACP ports to use it as the
599 * default system ID. */
600 bridge_configure_datapath_id(br);
602 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
605 port_configure(port);
607 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
608 iface_set_ofport(iface->cfg, iface->ofp_port);
609 iface_configure_cfm(iface);
610 iface_configure_qos(iface, port->cfg->qos);
611 iface_set_mac(br, port, iface);
612 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
616 bridge_configure_mirrors(br);
617 bridge_configure_forward_bpdu(br);
618 bridge_configure_mac_table(br);
619 bridge_configure_remotes(br, managers, n_managers);
620 bridge_configure_netflow(br);
621 bridge_configure_sflow(br, &sflow_bridge_number);
622 bridge_configure_ipfix(br);
623 bridge_configure_stp(br);
624 bridge_configure_tables(br);
625 bridge_configure_dp_desc(br);
629 /* The ofproto-dpif provider does some final reconfiguration in its
630 * ->type_run() function. We have to call it before notifying the database
631 * client that reconfiguration is complete, otherwise there is a very
632 * narrow race window in which e.g. ofproto/trace will not recognize the
633 * new configuration (sometimes this causes unit test failures). */
637 /* Delete ofprotos which aren't configured or have the wrong type. Create
638 * ofprotos which don't exist but need to. */
640 bridge_delete_ofprotos(void)
647 /* Delete ofprotos with no bridge or with the wrong type. */
650 ofproto_enumerate_types(&types);
651 SSET_FOR_EACH (type, &types) {
654 ofproto_enumerate_names(type, &names);
655 SSET_FOR_EACH (name, &names) {
656 br = bridge_lookup(name);
657 if (!br || strcmp(type, br->type)) {
658 ofproto_delete(name, type);
662 sset_destroy(&names);
663 sset_destroy(&types);
667 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
669 if (*n >= *allocated) {
670 ports = x2nrealloc(ports, allocated, sizeof *ports);
672 ports[(*n)++] = port;
677 bridge_delete_or_reconfigure_ports(struct bridge *br)
679 struct ofproto_port ofproto_port;
680 struct ofproto_port_dump dump;
682 struct sset ofproto_ports;
683 struct port *port, *port_next;
685 /* List of "ofp_port"s to delete. We make a list instead of deleting them
686 * right away because ofproto implementations aren't necessarily able to
687 * iterate through a changing list of ports in an entirely robust way. */
694 sset_init(&ofproto_ports);
696 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
697 * that are not configured in the database. (This commonly happens when
698 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
700 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
701 * that have the wrong OpenFlow port number (and arrange to add them back
702 * with the correct OpenFlow port number). */
703 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
704 ofp_port_t requested_ofp_port;
707 sset_add(&ofproto_ports, ofproto_port.name);
709 iface = iface_lookup(br, ofproto_port.name);
711 /* No such iface is configured, so we should delete this
714 * As a corner case exception, keep the port if it's a bond fake
716 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
717 && !strcmp(ofproto_port.type, "internal")) {
723 if (strcmp(ofproto_port.type, iface->type)
724 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
725 /* The interface is the wrong type or can't be configured.
730 /* If the requested OpenFlow port for 'iface' changed, and it's not
731 * already the correct port, then we might want to temporarily delete
732 * this interface, so we can add it back again with the new OpenFlow
734 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
735 if (iface->ofp_port != OFPP_LOCAL &&
736 requested_ofp_port != OFPP_NONE &&
737 requested_ofp_port != iface->ofp_port) {
738 ofp_port_t victim_request;
739 struct iface *victim;
741 /* Check for an existing OpenFlow port currently occupying
742 * 'iface''s requested port number. If there isn't one, then
743 * delete this port. Otherwise we need to consider further. */
744 victim = iface_from_ofp_port(br, requested_ofp_port);
749 /* 'victim' is a port currently using 'iface''s requested port
750 * number. Unless 'victim' specifically requested that port
751 * number, too, then we can delete both 'iface' and 'victim'
752 * temporarily. (We'll add both of them back again later with new
753 * OpenFlow port numbers.)
755 * If 'victim' did request port number 'requested_ofp_port', just
756 * like 'iface', then that's a configuration inconsistency that we
757 * can't resolve. We might as well let it keep its current port
759 victim_request = iface_get_requested_ofp_port(victim->cfg);
760 if (victim_request != requested_ofp_port) {
761 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
762 iface_destroy(victim);
771 iface_destroy(iface);
772 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
774 for (i = 0; i < n; i++) {
775 ofproto_port_del(br->ofproto, del[i]);
779 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
780 * that we didn't see when we iterated through the datapath, i.e. ports
781 * that disappeared underneath use. This is an unusual situation, but it
782 * can happen in some cases:
784 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
785 * idea but could happen).
787 * - The port represented a device that disappeared, e.g. a tuntap
788 * device destroyed via "tunctl -d", a physical Ethernet device
789 * whose module was just unloaded via "rmmod", or a virtual NIC for a
790 * VM whose VM was just terminated. */
791 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
792 struct iface *iface, *iface_next;
794 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
795 if (!sset_contains(&ofproto_ports, iface->name)) {
796 iface_destroy__(iface);
800 if (list_is_empty(&port->ifaces)) {
804 sset_destroy(&ofproto_ports);
808 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
809 bool with_requested_port)
811 struct shash_node *port_node;
813 SHASH_FOR_EACH (port_node, wanted_ports) {
814 const struct ovsrec_port *port_cfg = port_node->data;
817 for (i = 0; i < port_cfg->n_interfaces; i++) {
818 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
819 ofp_port_t requested_ofp_port;
821 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
822 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
823 struct iface *iface = iface_lookup(br, iface_cfg->name);
826 iface_create(br, iface_cfg, port_cfg);
834 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
836 /* First add interfaces that request a particular port number. */
837 bridge_add_ports__(br, wanted_ports, true);
839 /* Then add interfaces that want automatic port number assignment.
840 * We add these afterward to avoid accidentally taking a specifically
841 * requested port number. */
842 bridge_add_ports__(br, wanted_ports, false);
846 port_configure(struct port *port)
848 const struct ovsrec_port *cfg = port->cfg;
849 struct bond_settings bond_settings;
850 struct lacp_settings lacp_settings;
851 struct ofproto_bundle_settings s;
854 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
855 configure_splinter_port(port);
864 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
865 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
866 s.slaves[s.n_slaves++] = iface->ofp_port;
871 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
875 /* Get VLAN trunks. */
878 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
882 if (cfg->vlan_mode) {
883 if (!strcmp(cfg->vlan_mode, "access")) {
884 s.vlan_mode = PORT_VLAN_ACCESS;
885 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
886 s.vlan_mode = PORT_VLAN_TRUNK;
887 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
888 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
889 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
890 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
892 /* This "can't happen" because ovsdb-server should prevent it. */
893 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
894 "back to trunk mode", port->name, cfg->vlan_mode);
895 s.vlan_mode = PORT_VLAN_TRUNK;
899 s.vlan_mode = PORT_VLAN_ACCESS;
901 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
905 s.vlan_mode = PORT_VLAN_TRUNK;
908 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
911 /* Get LACP settings. */
912 s.lacp = port_configure_lacp(port, &lacp_settings);
916 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
917 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
918 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
921 s.lacp_slaves = NULL;
924 /* Get bond settings. */
925 if (s.n_slaves > 1) {
926 s.bond = &bond_settings;
927 port_configure_bond(port, &bond_settings);
930 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
931 netdev_set_miimon_interval(iface->netdev, 0);
936 ofproto_bundle_register(port->bridge->ofproto, port, &s);
944 /* Pick local port hardware address and datapath ID for 'br'. */
946 bridge_configure_datapath_id(struct bridge *br)
948 uint8_t ea[ETH_ADDR_LEN];
950 struct iface *local_iface;
951 struct iface *hw_addr_iface;
954 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
955 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
957 int error = netdev_set_etheraddr(local_iface->netdev, ea);
959 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
960 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
961 "Ethernet address: %s",
962 br->name, ovs_strerror(error));
965 memcpy(br->ea, ea, ETH_ADDR_LEN);
967 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
968 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
969 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
970 ofproto_set_datapath_id(br->ofproto, dpid);
973 dpid_string = xasprintf("%016"PRIx64, dpid);
974 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
978 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
981 bridge_get_allowed_versions(struct bridge *br)
983 if (!br->cfg->n_protocols)
986 return ofputil_versions_from_strings(br->cfg->protocols,
987 br->cfg->n_protocols);
990 /* Set NetFlow configuration on 'br'. */
992 bridge_configure_netflow(struct bridge *br)
994 struct ovsrec_netflow *cfg = br->cfg->netflow;
995 struct netflow_options opts;
998 ofproto_set_netflow(br->ofproto, NULL);
1002 memset(&opts, 0, sizeof opts);
1004 /* Get default NetFlow configuration from datapath.
1005 * Apply overrides from 'cfg'. */
1006 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1007 if (cfg->engine_type) {
1008 opts.engine_type = *cfg->engine_type;
1010 if (cfg->engine_id) {
1011 opts.engine_id = *cfg->engine_id;
1014 /* Configure active timeout interval. */
1015 opts.active_timeout = cfg->active_timeout;
1016 if (!opts.active_timeout) {
1017 opts.active_timeout = -1;
1018 } else if (opts.active_timeout < 0) {
1019 VLOG_WARN("bridge %s: active timeout interval set to negative "
1020 "value, using default instead (%d seconds)", br->name,
1021 NF_ACTIVE_TIMEOUT_DEFAULT);
1022 opts.active_timeout = -1;
1025 /* Add engine ID to interface number to disambiguate bridgs? */
1026 opts.add_id_to_iface = cfg->add_id_to_interface;
1027 if (opts.add_id_to_iface) {
1028 if (opts.engine_id > 0x7f) {
1029 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1030 "another vswitch, choose an engine id less than 128",
1033 if (hmap_count(&br->ports) > 508) {
1034 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1035 "another port when more than 508 ports are used",
1041 sset_init(&opts.collectors);
1042 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1045 if (ofproto_set_netflow(br->ofproto, &opts)) {
1046 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1048 sset_destroy(&opts.collectors);
1051 /* Set sFlow configuration on 'br'. */
1053 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1055 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1056 struct ovsrec_controller **controllers;
1057 struct ofproto_sflow_options oso;
1058 size_t n_controllers;
1062 ofproto_set_sflow(br->ofproto, NULL);
1066 memset(&oso, 0, sizeof oso);
1068 sset_init(&oso.targets);
1069 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1071 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1072 if (cfg->sampling) {
1073 oso.sampling_rate = *cfg->sampling;
1076 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1078 oso.polling_interval = *cfg->polling;
1081 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1083 oso.header_len = *cfg->header;
1086 oso.sub_id = (*sflow_bridge_number)++;
1087 oso.agent_device = cfg->agent;
1089 oso.control_ip = NULL;
1090 n_controllers = bridge_get_controllers(br, &controllers);
1091 for (i = 0; i < n_controllers; i++) {
1092 if (controllers[i]->local_ip) {
1093 oso.control_ip = controllers[i]->local_ip;
1097 ofproto_set_sflow(br->ofproto, &oso);
1099 sset_destroy(&oso.targets);
1102 /* Returns whether a IPFIX row is valid. */
1104 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1106 return ipfix && ipfix->n_targets > 0;
1109 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1111 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1112 const struct bridge *br)
1114 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1117 /* Set IPFIX configuration on 'br'. */
1119 bridge_configure_ipfix(struct bridge *br)
1121 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1122 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1123 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1124 struct ofproto_ipfix_bridge_exporter_options be_opts;
1125 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1126 size_t n_fe_opts = 0;
1128 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1129 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1134 if (!valid_be_cfg && n_fe_opts == 0) {
1135 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1140 memset(&be_opts, 0, sizeof be_opts);
1142 sset_init(&be_opts.targets);
1143 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1145 if (be_cfg->sampling) {
1146 be_opts.sampling_rate = *be_cfg->sampling;
1148 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1150 if (be_cfg->obs_domain_id) {
1151 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1153 if (be_cfg->obs_point_id) {
1154 be_opts.obs_point_id = *be_cfg->obs_point_id;
1156 if (be_cfg->cache_active_timeout) {
1157 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1159 if (be_cfg->cache_max_flows) {
1160 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1164 if (n_fe_opts > 0) {
1165 struct ofproto_ipfix_flow_exporter_options *opts;
1166 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1168 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1169 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1170 opts->collector_set_id = fe_cfg->id;
1171 sset_init(&opts->targets);
1172 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1173 fe_cfg->ipfix->n_targets);
1174 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1175 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1176 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1177 ? *fe_cfg->ipfix->cache_max_flows : 0;
1183 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1187 sset_destroy(&be_opts.targets);
1190 if (n_fe_opts > 0) {
1191 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1193 for (i = 0; i < n_fe_opts; i++) {
1194 sset_destroy(&opts->targets);
1202 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1203 struct ofproto_port_stp_settings *port_s,
1204 int *port_num_counter, unsigned long *port_num_bitmap)
1206 const char *config_str;
1207 struct iface *iface;
1209 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1210 port_s->enable = false;
1213 port_s->enable = true;
1216 /* STP over bonds is not supported. */
1217 if (!list_is_singleton(&port->ifaces)) {
1218 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1220 port_s->enable = false;
1224 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1226 /* Internal ports shouldn't participate in spanning tree, so
1228 if (!strcmp(iface->type, "internal")) {
1229 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1230 port_s->enable = false;
1234 /* STP on mirror output ports is not supported. */
1235 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1236 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1237 port_s->enable = false;
1241 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1243 unsigned long int port_num = strtoul(config_str, NULL, 0);
1244 int port_idx = port_num - 1;
1246 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1247 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1248 port_s->enable = false;
1252 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1253 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1254 port->name, port_num);
1255 port_s->enable = false;
1258 bitmap_set1(port_num_bitmap, port_idx);
1259 port_s->port_num = port_idx;
1261 if (*port_num_counter >= STP_MAX_PORTS) {
1262 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1263 port_s->enable = false;
1267 port_s->port_num = (*port_num_counter)++;
1270 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1272 port_s->path_cost = strtoul(config_str, NULL, 10);
1274 enum netdev_features current;
1277 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1278 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1279 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1282 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1284 port_s->priority = strtoul(config_str, NULL, 0);
1286 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1290 /* Set spanning tree configuration on 'br'. */
1292 bridge_configure_stp(struct bridge *br)
1294 if (!br->cfg->stp_enable) {
1295 ofproto_set_stp(br->ofproto, NULL);
1297 struct ofproto_stp_settings br_s;
1298 const char *config_str;
1300 int port_num_counter;
1301 unsigned long *port_num_bitmap;
1303 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1305 uint8_t ea[ETH_ADDR_LEN];
1307 if (eth_addr_from_string(config_str, ea)) {
1308 br_s.system_id = eth_addr_to_uint64(ea);
1310 br_s.system_id = eth_addr_to_uint64(br->ea);
1311 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1312 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1315 br_s.system_id = eth_addr_to_uint64(br->ea);
1318 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1320 br_s.priority = strtoul(config_str, NULL, 0);
1322 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1325 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1327 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1329 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1332 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1334 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1336 br_s.max_age = STP_DEFAULT_MAX_AGE;
1339 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1341 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1343 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1346 /* Configure STP on the bridge. */
1347 if (ofproto_set_stp(br->ofproto, &br_s)) {
1348 VLOG_ERR("bridge %s: could not enable STP", br->name);
1352 /* Users must either set the port number with the "stp-port-num"
1353 * configuration on all ports or none. If manual configuration
1354 * is not done, then we allocate them sequentially. */
1355 port_num_counter = 0;
1356 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1357 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1358 struct ofproto_port_stp_settings port_s;
1359 struct iface *iface;
1361 port_configure_stp(br->ofproto, port, &port_s,
1362 &port_num_counter, port_num_bitmap);
1364 /* As bonds are not supported, just apply configuration to
1365 * all interfaces. */
1366 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1367 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1369 VLOG_ERR("port %s: could not enable STP", port->name);
1375 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1376 && port_num_counter) {
1377 VLOG_ERR("bridge %s: must manually configure all STP port "
1378 "IDs or none, disabling", br->name);
1379 ofproto_set_stp(br->ofproto, NULL);
1381 bitmap_free(port_num_bitmap);
1386 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1388 const struct port *port = port_lookup(br, name);
1389 return port && port_is_bond_fake_iface(port);
1393 port_is_bond_fake_iface(const struct port *port)
1395 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1399 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1401 struct bridge *br, *next;
1402 struct shash new_br;
1405 /* Collect new bridges' names and types. */
1406 shash_init(&new_br);
1407 for (i = 0; i < cfg->n_bridges; i++) {
1408 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1409 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1411 if (strchr(br_cfg->name, '/')) {
1412 /* Prevent remote ovsdb-server users from accessing arbitrary
1413 * directories, e.g. consider a bridge named "../../../etc/". */
1414 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1416 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1417 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1421 /* Get rid of deleted bridges or those whose types have changed.
1422 * Update 'cfg' of bridges that still exist. */
1423 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1424 br->cfg = shash_find_data(&new_br, br->name);
1425 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1426 br->cfg->datapath_type))) {
1431 /* Add new bridges. */
1432 for (i = 0; i < cfg->n_bridges; i++) {
1433 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1434 struct bridge *br = bridge_lookup(br_cfg->name);
1436 bridge_create(br_cfg);
1440 shash_destroy(&new_br);
1443 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1444 * Returns 0 if successful, otherwise a positive errno value. */
1446 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1447 struct netdev *netdev)
1449 return netdev_set_config(netdev, &iface_cfg->options);
1452 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1453 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1455 * If successful, returns 0 and stores the network device in '*netdevp'. On
1456 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1458 iface_do_create(const struct bridge *br,
1459 const struct ovsrec_interface *iface_cfg,
1460 const struct ovsrec_port *port_cfg,
1461 ofp_port_t *ofp_portp, struct netdev **netdevp)
1463 struct netdev *netdev = NULL;
1466 if (netdev_is_reserved_name(iface_cfg->name)) {
1467 VLOG_WARN("could not create interface %s, name is reserved",
1473 error = netdev_open(iface_cfg->name,
1474 iface_get_type(iface_cfg, br->cfg), &netdev);
1476 VLOG_WARN("could not open network device %s (%s)",
1477 iface_cfg->name, ovs_strerror(error));
1481 error = iface_set_netdev_config(iface_cfg, netdev);
1486 *ofp_portp = iface_pick_ofport(iface_cfg);
1487 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1492 VLOG_INFO("bridge %s: added interface %s on port %d",
1493 br->name, iface_cfg->name, *ofp_portp);
1495 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1496 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1504 netdev_close(netdev);
1508 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1509 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1510 * automatically allocated for the iface. Takes ownership of and
1511 * deallocates 'if_cfg'.
1513 * Return true if an iface is successfully created, false otherwise. */
1515 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1516 const struct ovsrec_port *port_cfg)
1518 struct netdev *netdev;
1519 struct iface *iface;
1520 ofp_port_t ofp_port;
1524 /* Do the bits that can fail up front. */
1525 ovs_assert(!iface_lookup(br, iface_cfg->name));
1526 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1528 iface_clear_db_record(iface_cfg);
1532 /* Get or create the port structure. */
1533 port = port_lookup(br, port_cfg->name);
1535 port = port_create(br, port_cfg);
1538 /* Create the iface structure. */
1539 iface = xzalloc(sizeof *iface);
1540 list_push_back(&port->ifaces, &iface->port_elem);
1541 hmap_insert(&br->iface_by_name, &iface->name_node,
1542 hash_string(iface_cfg->name, 0));
1544 iface->name = xstrdup(iface_cfg->name);
1545 iface->ofp_port = ofp_port;
1546 iface->netdev = netdev;
1547 iface->type = iface_get_type(iface_cfg, br->cfg);
1548 iface->cfg = iface_cfg;
1549 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1550 hash_ofp_port(ofp_port));
1552 /* Populate initial status in database. */
1553 iface_refresh_stats(iface);
1555 /* Add bond fake iface if necessary. */
1556 if (port_is_bond_fake_iface(port)) {
1557 struct ofproto_port ofproto_port;
1559 if (ofproto_port_query_by_name(br->ofproto, port->name,
1561 struct netdev *netdev;
1564 error = netdev_open(port->name, "internal", &netdev);
1566 ofp_port_t fake_ofp_port = OFPP_NONE;
1567 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1568 netdev_close(netdev);
1570 VLOG_WARN("could not open network device %s (%s)",
1571 port->name, ovs_strerror(error));
1574 /* Already exists, nothing to do. */
1575 ofproto_port_destroy(&ofproto_port);
1582 /* Set forward BPDU option. */
1584 bridge_configure_forward_bpdu(struct bridge *br)
1586 ofproto_set_forward_bpdu(br->ofproto,
1587 smap_get_bool(&br->cfg->other_config,
1592 /* Set MAC learning table configuration for 'br'. */
1594 bridge_configure_mac_table(struct bridge *br)
1596 const char *idle_time_str;
1599 const char *mac_table_size_str;
1602 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1603 idle_time = (idle_time_str && atoi(idle_time_str)
1604 ? atoi(idle_time_str)
1605 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1607 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1608 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1609 ? atoi(mac_table_size_str)
1612 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1616 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1617 const struct port *fake_br, struct iface **hw_addr_iface)
1619 struct hmapx mirror_output_ports;
1621 bool found_addr = false;
1625 /* Mirror output ports don't participate in picking the local hardware
1626 * address. ofproto can't help us find out whether a given port is a
1627 * mirror output because we haven't configured mirrors yet, so we need to
1628 * accumulate them ourselves. */
1629 hmapx_init(&mirror_output_ports);
1630 for (i = 0; i < br->cfg->n_mirrors; i++) {
1631 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1632 if (m->output_port) {
1633 hmapx_add(&mirror_output_ports, m->output_port);
1637 /* Otherwise choose the minimum non-local MAC address among all of the
1639 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1640 uint8_t iface_ea[ETH_ADDR_LEN];
1641 struct iface *candidate;
1642 struct iface *iface;
1644 /* Mirror output ports don't participate. */
1645 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1649 /* Choose the MAC address to represent the port. */
1651 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1652 /* Find the interface with this Ethernet address (if any) so that
1653 * we can provide the correct devname to the caller. */
1654 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1655 uint8_t candidate_ea[ETH_ADDR_LEN];
1656 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1657 && eth_addr_equals(iface_ea, candidate_ea)) {
1662 /* Choose the interface whose MAC address will represent the port.
1663 * The Linux kernel bonding code always chooses the MAC address of
1664 * the first slave added to a bond, and the Fedora networking
1665 * scripts always add slaves to a bond in alphabetical order, so
1666 * for compatibility we choose the interface with the name that is
1667 * first in alphabetical order. */
1668 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1669 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1674 /* The local port doesn't count (since we're trying to choose its
1675 * MAC address anyway). */
1676 if (iface->ofp_port == OFPP_LOCAL) {
1680 /* For fake bridges we only choose from ports with the same tag */
1681 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1682 if (!port->cfg->tag) {
1685 if (*port->cfg->tag != *fake_br->cfg->tag) {
1691 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1697 /* Compare against our current choice. */
1698 if (!eth_addr_is_multicast(iface_ea) &&
1699 !eth_addr_is_local(iface_ea) &&
1700 !eth_addr_is_reserved(iface_ea) &&
1701 !eth_addr_is_zero(iface_ea) &&
1702 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1704 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1705 *hw_addr_iface = iface;
1711 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1712 *hw_addr_iface = NULL;
1715 hmapx_destroy(&mirror_output_ports);
1719 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1720 struct iface **hw_addr_iface)
1723 *hw_addr_iface = NULL;
1725 /* Did the user request a particular MAC? */
1726 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1727 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1728 if (eth_addr_is_multicast(ea)) {
1729 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1730 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1731 } else if (eth_addr_is_zero(ea)) {
1732 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1738 /* Find a local hw address */
1739 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1742 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1743 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1744 * an interface on 'br', then that interface must be passed in as
1745 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1746 * 'hw_addr_iface' must be passed in as a null pointer. */
1748 bridge_pick_datapath_id(struct bridge *br,
1749 const uint8_t bridge_ea[ETH_ADDR_LEN],
1750 struct iface *hw_addr_iface)
1753 * The procedure for choosing a bridge MAC address will, in the most
1754 * ordinary case, also choose a unique MAC that we can use as a datapath
1755 * ID. In some special cases, though, multiple bridges will end up with
1756 * the same MAC address. This is OK for the bridges, but it will confuse
1757 * the OpenFlow controller, because each datapath needs a unique datapath
1760 * Datapath IDs must be unique. It is also very desirable that they be
1761 * stable from one run to the next, so that policy set on a datapath
1764 const char *datapath_id;
1767 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1768 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1772 if (!hw_addr_iface) {
1774 * A purely internal bridge, that is, one that has no non-virtual
1775 * network devices on it at all, is difficult because it has no
1776 * natural unique identifier at all.
1778 * When the host is a XenServer, we handle this case by hashing the
1779 * host's UUID with the name of the bridge. Names of bridges are
1780 * persistent across XenServer reboots, although they can be reused if
1781 * an internal network is destroyed and then a new one is later
1782 * created, so this is fairly effective.
1784 * When the host is not a XenServer, we punt by using a random MAC
1785 * address on each run.
1787 const char *host_uuid = xenserver_get_host_uuid();
1789 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1790 dpid = dpid_from_hash(combined, strlen(combined));
1796 return eth_addr_to_uint64(bridge_ea);
1800 dpid_from_hash(const void *data, size_t n)
1802 uint8_t hash[SHA1_DIGEST_SIZE];
1804 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1805 sha1_bytes(data, n, hash);
1806 eth_addr_mark_random(hash);
1807 return eth_addr_to_uint64(hash);
1811 iface_refresh_netdev_status(struct iface *iface)
1815 enum netdev_features current;
1816 enum netdev_flags flags;
1817 const char *link_state;
1818 uint8_t mac[ETH_ADDR_LEN];
1819 int64_t bps, mtu_64, ifindex64, link_resets;
1822 if (iface_is_synthetic(iface)) {
1826 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
1827 && !force_status_commit) {
1831 iface->change_seq = netdev_get_change_seq(iface->netdev);
1835 if (!netdev_get_status(iface->netdev, &smap)) {
1836 ovsrec_interface_set_status(iface->cfg, &smap);
1838 ovsrec_interface_set_status(iface->cfg, NULL);
1841 smap_destroy(&smap);
1843 error = netdev_get_flags(iface->netdev, &flags);
1845 const char *state = flags & NETDEV_UP ? "up" : "down";
1847 ovsrec_interface_set_admin_state(iface->cfg, state);
1849 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1852 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1853 ovsrec_interface_set_link_state(iface->cfg, link_state);
1855 link_resets = netdev_get_carrier_resets(iface->netdev);
1856 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1858 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1859 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1861 ovsrec_interface_set_duplex(iface->cfg,
1862 netdev_features_is_full_duplex(current)
1864 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1866 ovsrec_interface_set_duplex(iface->cfg, NULL);
1867 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1870 error = netdev_get_mtu(iface->netdev, &mtu);
1873 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1875 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1878 error = netdev_get_etheraddr(iface->netdev, mac);
1880 char mac_string[32];
1882 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1883 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1885 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1888 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1889 * if there is no valid ifindex number. */
1890 ifindex64 = netdev_get_ifindex(iface->netdev);
1891 if (ifindex64 < 0) {
1894 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1898 iface_refresh_ofproto_status(struct iface *iface)
1903 if (iface_is_synthetic(iface)) {
1907 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1911 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1913 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1916 iface_refresh_cfm_stats(iface);
1919 error = ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1920 iface->ofp_port, force_status_commit,
1923 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1925 smap_destroy(&smap);
1928 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1931 iface_refresh_cfm_stats(struct iface *iface)
1933 const struct ovsrec_interface *cfg = iface->cfg;
1934 struct ofproto_cfm_status status;
1937 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1938 iface->ofp_port, force_status_commit,
1941 /* Do nothing if there is no status change since last update. */
1942 } else if (error > 0) {
1943 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1944 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1945 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1946 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1947 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1948 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1950 const char *reasons[CFM_FAULT_N_REASONS];
1951 int64_t cfm_health = status.health;
1952 int64_t cfm_flap_count = status.flap_count;
1953 bool faulted = status.faults != 0;
1956 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1959 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1960 int reason = 1 << i;
1961 if (status.faults & reason) {
1962 reasons[j++] = cfm_fault_reason_to_str(reason);
1965 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1967 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1969 if (status.remote_opstate >= 0) {
1970 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1971 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1973 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1976 ovsrec_interface_set_cfm_remote_mpids(cfg,
1977 (const int64_t *)status.rmps,
1979 if (cfm_health >= 0) {
1980 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1982 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1990 iface_refresh_stats(struct iface *iface)
1992 #define IFACE_STATS \
1993 IFACE_STAT(rx_packets, "rx_packets") \
1994 IFACE_STAT(tx_packets, "tx_packets") \
1995 IFACE_STAT(rx_bytes, "rx_bytes") \
1996 IFACE_STAT(tx_bytes, "tx_bytes") \
1997 IFACE_STAT(rx_dropped, "rx_dropped") \
1998 IFACE_STAT(tx_dropped, "tx_dropped") \
1999 IFACE_STAT(rx_errors, "rx_errors") \
2000 IFACE_STAT(tx_errors, "tx_errors") \
2001 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2002 IFACE_STAT(rx_over_errors, "rx_over_err") \
2003 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2004 IFACE_STAT(collisions, "collisions")
2006 #define IFACE_STAT(MEMBER, NAME) + 1
2007 enum { N_IFACE_STATS = IFACE_STATS };
2009 int64_t values[N_IFACE_STATS];
2010 char *keys[N_IFACE_STATS];
2013 struct netdev_stats stats;
2015 if (iface_is_synthetic(iface)) {
2019 /* Intentionally ignore return value, since errors will set 'stats' to
2020 * all-1s, and we will deal with that correctly below. */
2021 netdev_get_stats(iface->netdev, &stats);
2023 /* Copy statistics into keys[] and values[]. */
2025 #define IFACE_STAT(MEMBER, NAME) \
2026 if (stats.MEMBER != UINT64_MAX) { \
2028 values[n] = stats.MEMBER; \
2033 ovs_assert(n <= N_IFACE_STATS);
2035 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2040 br_refresh_stp_status(struct bridge *br)
2042 struct smap smap = SMAP_INITIALIZER(&smap);
2043 struct ofproto *ofproto = br->ofproto;
2044 struct ofproto_stp_status status;
2046 if (ofproto_get_stp_status(ofproto, &status)) {
2050 if (!status.enabled) {
2051 ovsrec_bridge_set_status(br->cfg, NULL);
2055 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2056 STP_ID_ARGS(status.bridge_id));
2057 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2058 STP_ID_ARGS(status.designated_root));
2059 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2061 ovsrec_bridge_set_status(br->cfg, &smap);
2062 smap_destroy(&smap);
2066 port_refresh_stp_status(struct port *port)
2068 struct ofproto *ofproto = port->bridge->ofproto;
2069 struct iface *iface;
2070 struct ofproto_port_stp_status status;
2073 if (port_is_synthetic(port)) {
2077 /* STP doesn't currently support bonds. */
2078 if (!list_is_singleton(&port->ifaces)) {
2079 ovsrec_port_set_status(port->cfg, NULL);
2083 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2084 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2088 if (!status.enabled) {
2089 ovsrec_port_set_status(port->cfg, NULL);
2093 /* Set Status column. */
2095 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2096 smap_add(&smap, "stp_state", stp_state_name(status.state));
2097 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2098 smap_add(&smap, "stp_role", stp_role_name(status.role));
2099 ovsrec_port_set_status(port->cfg, &smap);
2100 smap_destroy(&smap);
2104 port_refresh_stp_stats(struct port *port)
2106 struct ofproto *ofproto = port->bridge->ofproto;
2107 struct iface *iface;
2108 struct ofproto_port_stp_stats stats;
2110 int64_t int_values[3];
2112 if (port_is_synthetic(port)) {
2116 /* STP doesn't currently support bonds. */
2117 if (!list_is_singleton(&port->ifaces)) {
2121 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2122 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2126 if (!stats.enabled) {
2127 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2131 /* Set Statistics column. */
2132 keys[0] = "stp_tx_count";
2133 int_values[0] = stats.tx_count;
2134 keys[1] = "stp_rx_count";
2135 int_values[1] = stats.rx_count;
2136 keys[2] = "stp_error_count";
2137 int_values[2] = stats.error_count;
2139 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2140 ARRAY_SIZE(int_values));
2144 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2146 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2150 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2152 bool enable = enable_system_stats(cfg);
2154 system_stats_enable(enable);
2156 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2161 run_system_stats(void)
2163 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2166 stats = system_stats_run();
2168 struct ovsdb_idl_txn *txn;
2169 struct ovsdb_datum datum;
2171 txn = ovsdb_idl_txn_create(idl);
2172 ovsdb_datum_from_smap(&datum, stats);
2173 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2175 ovsdb_idl_txn_commit(txn);
2176 ovsdb_idl_txn_destroy(txn);
2183 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2186 case OFPCR12_ROLE_EQUAL:
2188 case OFPCR12_ROLE_MASTER:
2190 case OFPCR12_ROLE_SLAVE:
2192 case OFPCR12_ROLE_NOCHANGE:
2194 return "*** INVALID ROLE ***";
2199 refresh_controller_status(void)
2203 const struct ovsrec_controller *cfg;
2207 /* Accumulate status for controllers on all bridges. */
2208 HMAP_FOR_EACH (br, node, &all_bridges) {
2209 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2212 /* Update each controller in the database with current status. */
2213 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2214 struct ofproto_controller_info *cinfo =
2215 shash_find_data(&info, cfg->target);
2218 struct smap smap = SMAP_INITIALIZER(&smap);
2219 const char **values = cinfo->pairs.values;
2220 const char **keys = cinfo->pairs.keys;
2223 for (i = 0; i < cinfo->pairs.n; i++) {
2224 smap_add(&smap, keys[i], values[i]);
2227 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2228 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2230 ovsrec_controller_set_status(cfg, &smap);
2231 smap_destroy(&smap);
2233 ovsrec_controller_set_is_connected(cfg, false);
2234 ovsrec_controller_set_role(cfg, NULL);
2235 ovsrec_controller_set_status(cfg, NULL);
2239 ofproto_free_ofproto_controller_info(&info);
2249 /* Let each datapath type do the work that it needs to do. */
2251 ofproto_enumerate_types(&types);
2252 SSET_FOR_EACH (type, &types) {
2253 ofproto_type_run(type);
2255 sset_destroy(&types);
2257 /* Let each bridge do the work that it needs to do. */
2258 HMAP_FOR_EACH (br, node, &all_bridges) {
2259 ofproto_run(br->ofproto);
2266 static struct ovsrec_open_vswitch null_cfg;
2267 const struct ovsrec_open_vswitch *cfg;
2269 bool vlan_splinters_changed;
2273 ovsrec_open_vswitch_init(&null_cfg);
2277 if (ovsdb_idl_is_lock_contended(idl)) {
2278 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2279 struct bridge *br, *next_br;
2281 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2282 "disabling this process (pid %ld) until it goes away",
2283 (long int) getpid());
2285 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2288 /* Since we will not be running system_stats_run() in this process
2289 * with the current situation of multiple ovs-vswitchd daemons,
2290 * disable system stats collection. */
2291 system_stats_enable(false);
2293 } else if (!ovsdb_idl_has_lock(idl)) {
2296 cfg = ovsrec_open_vswitch_first(idl);
2298 /* Initialize the ofproto library. This only needs to run once, but
2299 * it must be done after the configuration is set. If the
2300 * initialization has already occurred, bridge_init_ofproto()
2301 * returns immediately. */
2302 bridge_init_ofproto(cfg);
2304 /* Once the value of flow-restore-wait is false, we no longer should
2305 * check its value from the database. */
2306 if (cfg && ofproto_get_flow_restore_wait()) {
2307 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2308 "flow-restore-wait", false));
2313 /* Re-configure SSL. We do this on every trip through the main loop,
2314 * instead of just when the database changes, because the contents of the
2315 * key and certificate files can change without the database changing.
2317 * We do this before bridge_reconfigure() because that function might
2318 * initiate SSL connections and thus requires SSL to be configured. */
2319 if (cfg && cfg->ssl) {
2320 const struct ovsrec_ssl *ssl = cfg->ssl;
2322 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2323 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2326 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2327 * usage has changed. */
2328 vlan_splinters_changed = false;
2329 if (vlan_splinters_enabled_anywhere) {
2330 HMAP_FOR_EACH (br, node, &all_bridges) {
2331 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2332 vlan_splinters_changed = true;
2338 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2339 struct ovsdb_idl_txn *txn;
2341 idl_seqno = ovsdb_idl_get_seqno(idl);
2342 txn = ovsdb_idl_txn_create(idl);
2343 bridge_reconfigure(cfg ? cfg : &null_cfg);
2346 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2349 /* If we are completing our initial configuration for this run
2350 * of ovs-vswitchd, then keep the transaction around to monitor
2351 * it for completion. */
2352 if (initial_config_done) {
2353 ovsdb_idl_txn_commit(txn);
2354 ovsdb_idl_txn_destroy(txn);
2356 initial_config_done = true;
2357 daemonize_txn = txn;
2361 if (daemonize_txn) {
2362 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2363 if (status != TXN_INCOMPLETE) {
2364 ovsdb_idl_txn_destroy(daemonize_txn);
2365 daemonize_txn = NULL;
2367 /* ovs-vswitchd has completed initialization, so allow the
2368 * process that forked us to exit successfully. */
2369 daemonize_complete();
2371 vlog_enable_async();
2373 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2377 /* Statistics update interval should always be greater than or equal to
2380 stats_interval = MAX(smap_get_int(&cfg->other_config,
2381 "stats-update-interval",
2384 stats_interval = 5000;
2386 if (stats_timer_interval != stats_interval) {
2387 stats_timer_interval = stats_interval;
2388 stats_timer = LLONG_MIN;
2391 /* Refresh interface and mirror stats if necessary. */
2392 if (time_msec() >= stats_timer && cfg) {
2393 enum ovsdb_idl_txn_status status;
2395 /* Rate limit the update. Do not start a new update if the
2396 * previous one is not done. */
2398 stats_txn = ovsdb_idl_txn_create(idl);
2399 HMAP_FOR_EACH (br, node, &all_bridges) {
2403 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2404 struct iface *iface;
2406 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2407 iface_refresh_stats(iface);
2409 port_refresh_stp_stats(port);
2411 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2412 mirror_refresh_stats(m);
2415 refresh_controller_status();
2418 status = ovsdb_idl_txn_commit(stats_txn);
2419 if (status != TXN_INCOMPLETE) {
2420 stats_timer = time_msec() + stats_timer_interval;
2421 ovsdb_idl_txn_destroy(stats_txn);
2429 /* Check the need to update status. */
2430 seq = seq_read(connectivity_seq_get());
2431 if (seq != connectivity_seqno || force_status_commit) {
2432 connectivity_seqno = seq;
2433 status_txn = ovsdb_idl_txn_create(idl);
2434 HMAP_FOR_EACH (br, node, &all_bridges) {
2437 br_refresh_stp_status(br);
2438 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2439 struct iface *iface;
2441 port_refresh_stp_status(port);
2442 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2443 iface_refresh_netdev_status(iface);
2444 iface_refresh_ofproto_status(iface);
2452 enum ovsdb_idl_txn_status status;
2454 status = ovsdb_idl_txn_commit(status_txn);
2456 /* If the transaction is incomplete or fails, 'status_txn'
2457 * needs to be committed next iteration of bridge_run() even if
2458 * connectivity or netdev sequence numbers do not change. */
2459 if (status == TXN_SUCCESS || status == TXN_UNCHANGED)
2461 force_status_commit = false;
2463 force_status_commit = true;
2466 /* Do not destroy "status_txn" if the transaction is
2467 * "TXN_INCOMPLETE". */
2468 if (status != TXN_INCOMPLETE) {
2469 ovsdb_idl_txn_destroy(status_txn);
2483 ovsdb_idl_wait(idl);
2484 if (daemonize_txn) {
2485 ovsdb_idl_txn_wait(daemonize_txn);
2489 ofproto_enumerate_types(&types);
2490 SSET_FOR_EACH (type, &types) {
2491 ofproto_type_wait(type);
2493 sset_destroy(&types);
2495 if (!hmap_is_empty(&all_bridges)) {
2498 HMAP_FOR_EACH (br, node, &all_bridges) {
2499 ofproto_wait(br->ofproto);
2502 poll_timer_wait_until(stats_timer);
2505 /* If the status database transaction is 'TXN_INCOMPLETE' or is
2506 * unsuccessful, register a timeout in 'STATUS_CHECK_AGAIN_MSEC'. Else,
2507 * wait on the global connectivity sequence number. Note, this also helps
2508 * batch multiple status changes into one transaction. */
2509 if (force_status_commit) {
2510 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2512 seq_wait(connectivity_seq_get(), connectivity_seqno);
2515 system_stats_wait();
2518 /* Adds some memory usage statistics for bridges into 'usage', for use with
2519 * memory_report(). */
2521 bridge_get_memory_usage(struct simap *usage)
2528 ofproto_enumerate_types(&types);
2529 SSET_FOR_EACH (type, &types) {
2530 ofproto_type_get_memory_usage(type, usage);
2532 sset_destroy(&types);
2534 HMAP_FOR_EACH (br, node, &all_bridges) {
2535 ofproto_get_memory_usage(br->ofproto, usage);
2539 /* QoS unixctl user interface functions. */
2541 struct qos_unixctl_show_cbdata {
2543 struct iface *iface;
2547 qos_unixctl_show_queue(unsigned int queue_id,
2548 const struct smap *details,
2549 struct iface *iface,
2552 struct netdev_queue_stats stats;
2553 struct smap_node *node;
2556 ds_put_cstr(ds, "\n");
2558 ds_put_format(ds, "Queue %u:\n", queue_id);
2560 ds_put_cstr(ds, "Default:\n");
2563 SMAP_FOR_EACH (node, details) {
2564 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2567 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2569 if (stats.tx_packets != UINT64_MAX) {
2570 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2573 if (stats.tx_bytes != UINT64_MAX) {
2574 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2577 if (stats.tx_errors != UINT64_MAX) {
2578 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2581 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2582 queue_id, ovs_strerror(error));
2587 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2588 const char *argv[], void *aux OVS_UNUSED)
2590 struct ds ds = DS_EMPTY_INITIALIZER;
2591 struct smap smap = SMAP_INITIALIZER(&smap);
2592 struct iface *iface;
2594 struct smap_node *node;
2596 iface = iface_find(argv[1]);
2598 unixctl_command_reply_error(conn, "no such interface");
2602 netdev_get_qos(iface->netdev, &type, &smap);
2604 if (*type != '\0') {
2605 struct netdev_queue_dump dump;
2606 struct smap details;
2607 unsigned int queue_id;
2609 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2611 SMAP_FOR_EACH (node, &smap) {
2612 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2615 smap_init(&details);
2616 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2617 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2619 smap_destroy(&details);
2621 unixctl_command_reply(conn, ds_cstr(&ds));
2623 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2624 unixctl_command_reply_error(conn, ds_cstr(&ds));
2627 smap_destroy(&smap);
2631 /* Bridge reconfiguration functions. */
2633 bridge_create(const struct ovsrec_bridge *br_cfg)
2637 ovs_assert(!bridge_lookup(br_cfg->name));
2638 br = xzalloc(sizeof *br);
2640 br->name = xstrdup(br_cfg->name);
2641 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2644 /* Derive the default Ethernet address from the bridge's UUID. This should
2645 * be unique and it will be stable between ovs-vswitchd runs. */
2646 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2647 eth_addr_mark_random(br->default_ea);
2649 hmap_init(&br->ports);
2650 hmap_init(&br->ifaces);
2651 hmap_init(&br->iface_by_name);
2652 hmap_init(&br->mirrors);
2654 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2658 bridge_destroy(struct bridge *br)
2661 struct mirror *mirror, *next_mirror;
2662 struct port *port, *next_port;
2664 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2667 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2668 mirror_destroy(mirror);
2671 hmap_remove(&all_bridges, &br->node);
2672 ofproto_destroy(br->ofproto);
2673 hmap_destroy(&br->ifaces);
2674 hmap_destroy(&br->ports);
2675 hmap_destroy(&br->iface_by_name);
2676 hmap_destroy(&br->mirrors);
2683 static struct bridge *
2684 bridge_lookup(const char *name)
2688 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2689 if (!strcmp(br->name, name)) {
2696 /* Handle requests for a listing of all flows known by the OpenFlow
2697 * stack, including those normally hidden. */
2699 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2700 const char *argv[], void *aux OVS_UNUSED)
2705 br = bridge_lookup(argv[1]);
2707 unixctl_command_reply_error(conn, "Unknown bridge");
2712 ofproto_get_all_flows(br->ofproto, &results);
2714 unixctl_command_reply(conn, ds_cstr(&results));
2715 ds_destroy(&results);
2718 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2719 * connections and reconnect. If BRIDGE is not specified, then all bridges
2720 * drop their controller connections and reconnect. */
2722 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2723 const char *argv[], void *aux OVS_UNUSED)
2727 br = bridge_lookup(argv[1]);
2729 unixctl_command_reply_error(conn, "Unknown bridge");
2732 ofproto_reconnect_controllers(br->ofproto);
2734 HMAP_FOR_EACH (br, node, &all_bridges) {
2735 ofproto_reconnect_controllers(br->ofproto);
2738 unixctl_command_reply(conn, NULL);
2742 bridge_get_controllers(const struct bridge *br,
2743 struct ovsrec_controller ***controllersp)
2745 struct ovsrec_controller **controllers;
2746 size_t n_controllers;
2748 controllers = br->cfg->controller;
2749 n_controllers = br->cfg->n_controller;
2751 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2757 *controllersp = controllers;
2759 return n_controllers;
2763 bridge_collect_wanted_ports(struct bridge *br,
2764 const unsigned long int *splinter_vlans,
2765 struct shash *wanted_ports)
2769 shash_init(wanted_ports);
2771 for (i = 0; i < br->cfg->n_ports; i++) {
2772 const char *name = br->cfg->ports[i]->name;
2773 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2774 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2778 if (bridge_get_controllers(br, NULL)
2779 && !shash_find(wanted_ports, br->name)) {
2780 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2781 br->name, br->name);
2783 ovsrec_interface_init(&br->synth_local_iface);
2784 ovsrec_port_init(&br->synth_local_port);
2786 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2787 br->synth_local_port.n_interfaces = 1;
2788 br->synth_local_port.name = br->name;
2790 br->synth_local_iface.name = br->name;
2791 br->synth_local_iface.type = "internal";
2793 br->synth_local_ifacep = &br->synth_local_iface;
2795 shash_add(wanted_ports, br->name, &br->synth_local_port);
2798 if (splinter_vlans) {
2799 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2803 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2804 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2805 * 'br' needs to complete its configuration. */
2807 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2809 struct shash_node *port_node;
2810 struct port *port, *next;
2812 /* Get rid of deleted ports.
2813 * Get rid of deleted interfaces on ports that still exist. */
2814 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2815 port->cfg = shash_find_data(wanted_ports, port->name);
2819 port_del_ifaces(port);
2823 /* Update iface->cfg and iface->type in interfaces that still exist. */
2824 SHASH_FOR_EACH (port_node, wanted_ports) {
2825 const struct ovsrec_port *port = port_node->data;
2828 for (i = 0; i < port->n_interfaces; i++) {
2829 const struct ovsrec_interface *cfg = port->interfaces[i];
2830 struct iface *iface = iface_lookup(br, cfg->name);
2831 const char *type = iface_get_type(cfg, br->cfg);
2836 } else if (!strcmp(type, "null")) {
2837 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2838 " may be removed in February 2013. Please email"
2839 " dev@openvswitch.org with concerns.",
2842 /* We will add new interfaces later. */
2848 /* Initializes 'oc' appropriately as a management service controller for
2851 * The caller must free oc->target when it is no longer needed. */
2853 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2854 struct ofproto_controller *oc)
2856 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2857 oc->max_backoff = 0;
2858 oc->probe_interval = 60;
2859 oc->band = OFPROTO_OUT_OF_BAND;
2861 oc->burst_limit = 0;
2862 oc->enable_async_msgs = true;
2866 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2868 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2869 struct ofproto_controller *oc)
2873 oc->target = c->target;
2874 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2875 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2876 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2877 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2878 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2879 oc->burst_limit = (c->controller_burst_limit
2880 ? *c->controller_burst_limit : 0);
2881 oc->enable_async_msgs = (!c->enable_async_messages
2882 || *c->enable_async_messages);
2883 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2884 if (dscp < 0 || dscp > 63) {
2885 dscp = DSCP_DEFAULT;
2890 /* Configures the IP stack for 'br''s local interface properly according to the
2891 * configuration in 'c'. */
2893 bridge_configure_local_iface_netdev(struct bridge *br,
2894 struct ovsrec_controller *c)
2896 struct netdev *netdev;
2897 struct in_addr mask, gateway;
2899 struct iface *local_iface;
2902 /* If there's no local interface or no IP address, give up. */
2903 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2904 if (!local_iface || !c->local_ip
2905 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2909 /* Bring up the local interface. */
2910 netdev = local_iface->netdev;
2911 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2913 /* Configure the IP address and netmask. */
2914 if (!c->local_netmask
2915 || !inet_pton(AF_INET, c->local_netmask, &mask)
2917 mask.s_addr = guess_netmask(ip.s_addr);
2919 if (!netdev_set_in4(netdev, ip, mask)) {
2920 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2921 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2924 /* Configure the default gateway. */
2925 if (c->local_gateway
2926 && inet_pton(AF_INET, c->local_gateway, &gateway)
2927 && gateway.s_addr) {
2928 if (!netdev_add_router(netdev, gateway)) {
2929 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2930 br->name, IP_ARGS(gateway.s_addr));
2935 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2936 * in either string are treated as equal to any number of slashes in the other,
2937 * e.g. "x///y" is equal to "x/y".
2939 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2940 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2941 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2942 * 'b' against a prefix of 'a'.
2945 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2947 const char *b_start = b;
2949 if (b - b_start >= b_stoplen) {
2951 } else if (*a != *b) {
2953 } else if (*a == '/') {
2954 a += strspn(a, "/");
2955 b += strspn(b, "/");
2956 } else if (*a == '\0') {
2966 bridge_configure_remotes(struct bridge *br,
2967 const struct sockaddr_in *managers, size_t n_managers)
2969 bool disable_in_band;
2971 struct ovsrec_controller **controllers;
2972 size_t n_controllers;
2974 enum ofproto_fail_mode fail_mode;
2976 struct ofproto_controller *ocs;
2980 /* Check if we should disable in-band control on this bridge. */
2981 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2984 /* Set OpenFlow queue ID for in-band control. */
2985 ofproto_set_in_band_queue(br->ofproto,
2986 smap_get_int(&br->cfg->other_config,
2987 "in-band-queue", -1));
2989 if (disable_in_band) {
2990 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2992 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2995 n_controllers = bridge_get_controllers(br, &controllers);
2997 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3000 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3001 for (i = 0; i < n_controllers; i++) {
3002 struct ovsrec_controller *c = controllers[i];
3004 if (!strncmp(c->target, "punix:", 6)
3005 || !strncmp(c->target, "unix:", 5)) {
3006 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3009 if (!strncmp(c->target, "unix:", 5)) {
3010 /* Connect to a listening socket */
3011 whitelist = xasprintf("unix:%s/", ovs_rundir());
3012 if (strchr(c->target, '/') &&
3013 !equal_pathnames(c->target, whitelist,
3014 strlen(whitelist))) {
3015 /* Absolute path specified, but not in ovs_rundir */
3016 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3017 "controller \"%s\" due to possibility for "
3018 "remote exploit. Instead, specify socket "
3019 "in whitelisted \"%s\" or connect to "
3020 "\"unix:%s/%s.mgmt\" (which is always "
3021 "available without special configuration).",
3022 br->name, c->target, whitelist,
3023 ovs_rundir(), br->name);
3028 whitelist = xasprintf("punix:%s/%s.controller",
3029 ovs_rundir(), br->name);
3030 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3031 /* Prevent remote ovsdb-server users from accessing
3032 * arbitrary Unix domain sockets and overwriting arbitrary
3034 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3035 "controller \"%s\" due to possibility of "
3036 "overwriting local files. Instead, specify "
3037 "whitelisted \"%s\" or connect to "
3038 "\"unix:%s/%s.mgmt\" (which is always "
3039 "available without special configuration).",
3040 br->name, c->target, whitelist,
3041 ovs_rundir(), br->name);
3050 bridge_configure_local_iface_netdev(br, c);
3051 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3052 if (disable_in_band) {
3053 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3058 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3059 bridge_get_allowed_versions(br));
3060 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3063 /* Set the fail-mode. */
3064 fail_mode = !br->cfg->fail_mode
3065 || !strcmp(br->cfg->fail_mode, "standalone")
3066 ? OFPROTO_FAIL_STANDALONE
3067 : OFPROTO_FAIL_SECURE;
3068 ofproto_set_fail_mode(br->ofproto, fail_mode);
3070 /* Configure OpenFlow controller connection snooping. */
3071 if (!ofproto_has_snoops(br->ofproto)) {
3075 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3076 ovs_rundir(), br->name));
3077 ofproto_set_snoops(br->ofproto, &snoops);
3078 sset_destroy(&snoops);
3083 bridge_configure_tables(struct bridge *br)
3085 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3089 n_tables = ofproto_get_n_tables(br->ofproto);
3091 for (i = 0; i < n_tables; i++) {
3092 struct ofproto_table_settings s;
3095 s.max_flows = UINT_MAX;
3098 s.n_prefix_fields = 0;
3099 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3101 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3102 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3105 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3106 s.max_flows = *cfg->flow_limit;
3108 if (cfg->overflow_policy
3109 && !strcmp(cfg->overflow_policy, "evict")) {
3111 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3112 for (k = 0; k < cfg->n_groups; k++) {
3113 const char *string = cfg->groups[k];
3116 msg = mf_parse_subfield__(&s.groups[k], &string);
3118 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3119 "'groups' (%s)", br->name, i, msg);
3121 } else if (*string) {
3122 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3123 "element '%s' contains trailing garbage",
3124 br->name, i, cfg->groups[k]);
3130 /* Prefix lookup fields. */
3131 s.n_prefix_fields = 0;
3132 for (k = 0; k < cfg->n_prefixes; k++) {
3133 const char *name = cfg->prefixes[k];
3134 const struct mf_field *mf = mf_from_name(name);
3136 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3140 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3141 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3142 "%s", br->name, name);
3145 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3146 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3147 "field not used: %s", br->name, name);
3150 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3152 if (s.n_prefix_fields > 0) {
3154 struct ds ds = DS_EMPTY_INITIALIZER;
3155 for (k = 0; k < s.n_prefix_fields; k++) {
3157 ds_put_char(&ds, ',');
3159 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3161 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3162 br->name, i, ds_cstr(&ds));
3167 ofproto_configure_table(br->ofproto, i, &s);
3171 for (; j < br->cfg->n_flow_tables; j++) {
3172 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3173 "%"PRId64" not supported by this datapath", br->name,
3174 br->cfg->key_flow_tables[j]);
3179 bridge_configure_dp_desc(struct bridge *br)
3181 ofproto_set_dp_desc(br->ofproto,
3182 smap_get(&br->cfg->other_config, "dp-desc"));
3185 /* Port functions. */
3187 static struct port *
3188 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3192 port = xzalloc(sizeof *port);
3194 port->name = xstrdup(cfg->name);
3196 list_init(&port->ifaces);
3198 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3202 /* Deletes interfaces from 'port' that are no longer configured for it. */
3204 port_del_ifaces(struct port *port)
3206 struct iface *iface, *next;
3207 struct sset new_ifaces;
3210 /* Collect list of new interfaces. */
3211 sset_init(&new_ifaces);
3212 for (i = 0; i < port->cfg->n_interfaces; i++) {
3213 const char *name = port->cfg->interfaces[i]->name;
3214 const char *type = port->cfg->interfaces[i]->type;
3215 if (strcmp(type, "null")) {
3216 sset_add(&new_ifaces, name);
3220 /* Get rid of deleted interfaces. */
3221 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3222 if (!sset_contains(&new_ifaces, iface->name)) {
3223 iface_destroy(iface);
3227 sset_destroy(&new_ifaces);
3231 port_destroy(struct port *port)
3234 struct bridge *br = port->bridge;
3235 struct iface *iface, *next;
3238 ofproto_bundle_unregister(br->ofproto, port);
3241 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3242 iface_destroy__(iface);
3245 hmap_remove(&br->ports, &port->hmap_node);
3251 static struct port *
3252 port_lookup(const struct bridge *br, const char *name)
3256 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3258 if (!strcmp(port->name, name)) {
3266 enable_lacp(struct port *port, bool *activep)
3268 if (!port->cfg->lacp) {
3269 /* XXX when LACP implementation has been sufficiently tested, enable by
3270 * default and make active on bonded ports. */
3272 } else if (!strcmp(port->cfg->lacp, "off")) {
3274 } else if (!strcmp(port->cfg->lacp, "active")) {
3277 } else if (!strcmp(port->cfg->lacp, "passive")) {
3281 VLOG_WARN("port %s: unknown LACP mode %s",
3282 port->name, port->cfg->lacp);
3287 static struct lacp_settings *
3288 port_configure_lacp(struct port *port, struct lacp_settings *s)
3290 const char *lacp_time, *system_id;
3293 if (!enable_lacp(port, &s->active)) {
3297 s->name = port->name;
3299 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3301 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3302 ETH_ADDR_SCAN_ARGS(s->id))) {
3303 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3304 " address.", port->name, system_id);
3308 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3311 if (eth_addr_is_zero(s->id)) {
3312 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3316 /* Prefer bondable links if unspecified. */
3317 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3319 s->priority = (priority > 0 && priority <= UINT16_MAX
3321 : UINT16_MAX - !list_is_short(&port->ifaces));
3323 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3324 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3326 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3327 "lacp-fallback-ab", false);
3333 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3335 int priority, portid, key;
3337 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3338 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3340 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3342 if (portid <= 0 || portid > UINT16_MAX) {
3343 portid = ofp_to_u16(iface->ofp_port);
3346 if (priority <= 0 || priority > UINT16_MAX) {
3347 priority = UINT16_MAX;
3350 if (key < 0 || key > UINT16_MAX) {
3354 s->name = iface->name;
3356 s->priority = priority;
3361 port_configure_bond(struct port *port, struct bond_settings *s)
3363 const char *detect_s;
3364 struct iface *iface;
3365 int miimon_interval;
3367 s->name = port->name;
3369 if (port->cfg->bond_mode) {
3370 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3371 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3372 port->name, port->cfg->bond_mode,
3373 bond_mode_to_string(s->balance));
3376 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3378 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3379 * active-backup. At some point we should remove this warning. */
3380 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3381 " in previous versions, the default bond_mode was"
3382 " balance-slb", port->name,
3383 bond_mode_to_string(s->balance));
3385 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3386 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3387 "please use another bond type or disable flood_vlans",
3391 miimon_interval = smap_get_int(&port->cfg->other_config,
3392 "bond-miimon-interval", 0);
3393 if (miimon_interval <= 0) {
3394 miimon_interval = 200;
3397 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3398 if (!detect_s || !strcmp(detect_s, "carrier")) {
3399 miimon_interval = 0;
3400 } else if (strcmp(detect_s, "miimon")) {
3401 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3402 "defaulting to carrier", port->name, detect_s);
3403 miimon_interval = 0;
3406 s->up_delay = MAX(0, port->cfg->bond_updelay);
3407 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3408 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3409 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3410 "bond-rebalance-interval", 10000);
3411 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3412 s->rebalance_interval = 1000;
3415 s->fake_iface = port->cfg->bond_fake_iface;
3417 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3418 "lacp-fallback-ab", false);
3420 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3421 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3425 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3426 * instead of obtaining it from the database. */
3428 port_is_synthetic(const struct port *port)
3430 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3433 /* Interface functions. */
3436 iface_is_internal(const struct ovsrec_interface *iface,
3437 const struct ovsrec_bridge *br)
3439 /* The local port and "internal" ports are always "internal". */
3440 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3443 /* Returns the correct network device type for interface 'iface' in bridge
3446 iface_get_type(const struct ovsrec_interface *iface,
3447 const struct ovsrec_bridge *br)
3451 /* The local port always has type "internal". Other ports take
3452 * their type from the database and default to "system" if none is
3454 if (iface_is_internal(iface, br)) {
3457 type = iface->type[0] ? iface->type : "system";
3460 return ofproto_port_open_type(br->datapath_type, type);
3464 iface_destroy__(struct iface *iface)
3467 struct port *port = iface->port;
3468 struct bridge *br = port->bridge;
3470 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3471 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3474 if (iface->ofp_port != OFPP_NONE) {
3475 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3478 list_remove(&iface->port_elem);
3479 hmap_remove(&br->iface_by_name, &iface->name_node);
3481 netdev_close(iface->netdev);
3489 iface_destroy(struct iface *iface)
3492 struct port *port = iface->port;
3494 iface_destroy__(iface);
3495 if (list_is_empty(&port->ifaces)) {
3501 static struct iface *
3502 iface_lookup(const struct bridge *br, const char *name)
3504 struct iface *iface;
3506 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3507 &br->iface_by_name) {
3508 if (!strcmp(iface->name, name)) {
3516 static struct iface *
3517 iface_find(const char *name)
3519 const struct bridge *br;
3521 HMAP_FOR_EACH (br, node, &all_bridges) {
3522 struct iface *iface = iface_lookup(br, name);
3531 static struct iface *
3532 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3534 struct iface *iface;
3536 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3538 if (iface->ofp_port == ofp_port) {
3545 /* Set Ethernet address of 'iface', if one is specified in the configuration
3548 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3550 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3551 struct iface *hw_addr_iface;
3553 if (strcmp(iface->type, "internal")) {
3557 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3559 } else if (port->cfg->fake_bridge) {
3560 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3561 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3566 if (iface->ofp_port == OFPP_LOCAL) {
3567 VLOG_ERR("interface %s: ignoring mac in Interface record "
3568 "(use Bridge record to set local port's mac)",
3570 } else if (eth_addr_is_multicast(mac)) {
3571 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3574 int error = netdev_set_etheraddr(iface->netdev, mac);
3576 VLOG_ERR("interface %s: setting MAC failed (%s)",
3577 iface->name, ovs_strerror(error));
3583 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3585 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3587 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3588 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3589 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3593 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3594 * sets the "ofport" field to -1.
3596 * This is appropriate when 'if_cfg''s interface cannot be created or is
3597 * otherwise invalid. */
3599 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3601 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3602 iface_set_ofport(if_cfg, OFPP_NONE);
3603 ovsrec_interface_set_status(if_cfg, NULL);
3604 ovsrec_interface_set_admin_state(if_cfg, NULL);
3605 ovsrec_interface_set_duplex(if_cfg, NULL);
3606 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3607 ovsrec_interface_set_link_state(if_cfg, NULL);
3608 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3609 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3610 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3611 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3612 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3613 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3614 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3615 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3620 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3622 union ovsdb_atom atom;
3624 atom.integer = target;
3625 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3629 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3631 struct ofpbuf queues_buf;
3633 ofpbuf_init(&queues_buf, 0);
3635 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3636 netdev_set_qos(iface->netdev, NULL, NULL);
3638 const struct ovsdb_datum *queues;
3639 struct netdev_queue_dump dump;
3640 unsigned int queue_id;
3641 struct smap details;
3645 /* Configure top-level Qos for 'iface'. */
3646 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3648 /* Deconfigure queues that were deleted. */
3649 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3651 smap_init(&details);
3652 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3653 if (!queue_ids_include(queues, queue_id)) {
3654 netdev_delete_queue(iface->netdev, queue_id);
3657 smap_destroy(&details);
3659 /* Configure queues for 'iface'. */
3661 for (i = 0; i < qos->n_queues; i++) {
3662 const struct ovsrec_queue *queue = qos->value_queues[i];
3663 unsigned int queue_id = qos->key_queues[i];
3665 if (queue_id == 0) {
3669 if (queue->n_dscp == 1) {
3670 struct ofproto_port_queue *port_queue;
3672 port_queue = ofpbuf_put_uninit(&queues_buf,
3673 sizeof *port_queue);
3674 port_queue->queue = queue_id;
3675 port_queue->dscp = queue->dscp[0];
3678 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3681 struct smap details;
3683 smap_init(&details);
3684 netdev_set_queue(iface->netdev, 0, &details);
3685 smap_destroy(&details);
3689 if (iface->ofp_port != OFPP_NONE) {
3690 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3691 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3693 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3694 port_queues, n_queues);
3697 netdev_set_policing(iface->netdev,
3698 iface->cfg->ingress_policing_rate,
3699 iface->cfg->ingress_policing_burst);
3701 ofpbuf_uninit(&queues_buf);
3705 iface_configure_cfm(struct iface *iface)
3707 const struct ovsrec_interface *cfg = iface->cfg;
3708 const char *opstate_str;
3709 const char *cfm_ccm_vlan;
3710 struct cfm_settings s;
3711 struct smap netdev_args;
3713 if (!cfg->n_cfm_mpid) {
3714 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3718 s.check_tnl_key = false;
3719 smap_init(&netdev_args);
3720 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3721 const char *key = smap_get(&netdev_args, "key");
3722 const char *in_key = smap_get(&netdev_args, "in_key");
3724 s.check_tnl_key = (key && !strcmp(key, "flow"))
3725 || (in_key && !strcmp(in_key, "flow"));
3727 smap_destroy(&netdev_args);
3729 s.mpid = *cfg->cfm_mpid;
3730 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3731 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3732 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3734 if (s.interval <= 0) {
3738 if (!cfm_ccm_vlan) {
3740 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3741 s.ccm_vlan = CFM_RANDOM_VLAN;
3743 s.ccm_vlan = atoi(cfm_ccm_vlan);
3744 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3749 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3751 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3753 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3754 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3756 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3759 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3760 * instead of obtaining it from the database. */
3762 iface_is_synthetic(const struct iface *iface)
3764 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3768 iface_validate_ofport__(size_t n, int64_t *ofport)
3770 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3771 ? u16_to_ofp(*ofport)
3776 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3778 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3782 iface_pick_ofport(const struct ovsrec_interface *cfg)
3784 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3785 return (requested_ofport != OFPP_NONE
3787 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3790 /* Port mirroring. */
3792 static struct mirror *
3793 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3797 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3798 if (uuid_equals(uuid, &m->uuid)) {
3806 bridge_configure_mirrors(struct bridge *br)
3808 const struct ovsdb_datum *mc;
3809 unsigned long *flood_vlans;
3810 struct mirror *m, *next;
3813 /* Get rid of deleted mirrors. */
3814 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3815 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3816 union ovsdb_atom atom;
3818 atom.uuid = m->uuid;
3819 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3824 /* Add new mirrors and reconfigure existing ones. */
3825 for (i = 0; i < br->cfg->n_mirrors; i++) {
3826 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3827 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3829 m = mirror_create(br, cfg);
3832 if (!mirror_configure(m)) {
3837 /* Update flooded vlans (for RSPAN). */
3838 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3839 br->cfg->n_flood_vlans);
3840 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3841 bitmap_free(flood_vlans);
3844 static struct mirror *
3845 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3849 m = xzalloc(sizeof *m);
3850 m->uuid = cfg->header_.uuid;
3851 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3853 m->name = xstrdup(cfg->name);
3859 mirror_destroy(struct mirror *m)
3862 struct bridge *br = m->bridge;
3865 ofproto_mirror_unregister(br->ofproto, m);
3868 hmap_remove(&br->mirrors, &m->hmap_node);
3875 mirror_collect_ports(struct mirror *m,
3876 struct ovsrec_port **in_ports, int n_in_ports,
3877 void ***out_portsp, size_t *n_out_portsp)
3879 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3880 size_t n_out_ports = 0;
3883 for (i = 0; i < n_in_ports; i++) {
3884 const char *name = in_ports[i]->name;
3885 struct port *port = port_lookup(m->bridge, name);
3887 out_ports[n_out_ports++] = port;
3889 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3890 "port %s", m->bridge->name, m->name, name);
3893 *out_portsp = out_ports;
3894 *n_out_portsp = n_out_ports;
3898 mirror_configure(struct mirror *m)
3900 const struct ovsrec_mirror *cfg = m->cfg;
3901 struct ofproto_mirror_settings s;
3904 if (strcmp(cfg->name, m->name)) {
3906 m->name = xstrdup(cfg->name);
3910 /* Get output port or VLAN. */
3911 if (cfg->output_port) {
3912 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3913 if (!s.out_bundle) {
3914 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3915 m->bridge->name, m->name);
3918 s.out_vlan = UINT16_MAX;
3920 if (cfg->output_vlan) {
3921 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3922 "output vlan; ignoring output vlan",
3923 m->bridge->name, m->name);
3925 } else if (cfg->output_vlan) {
3926 /* The database should prevent invalid VLAN values. */
3927 s.out_bundle = NULL;
3928 s.out_vlan = *cfg->output_vlan;
3930 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3931 m->bridge->name, m->name);
3935 /* Get port selection. */
3936 if (cfg->select_all) {
3937 size_t n_ports = hmap_count(&m->bridge->ports);
3938 void **ports = xmalloc(n_ports * sizeof *ports);
3943 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3953 /* Get ports, dropping ports that don't exist.
3954 * The IDL ensures that there are no duplicates. */
3955 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3956 &s.srcs, &s.n_srcs);
3957 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3958 &s.dsts, &s.n_dsts);
3961 /* Get VLAN selection. */
3962 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3965 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3968 if (s.srcs != s.dsts) {
3977 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3979 * This is deprecated. It is only for compatibility with broken device drivers
3980 * in old versions of Linux that do not properly support VLANs when VLAN
3981 * devices are not used. When broken device drivers are no longer in
3982 * widespread use, we will delete these interfaces. */
3984 static struct ovsrec_port **recs;
3985 static size_t n_recs, allocated_recs;
3987 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3988 * splinters are reconfigured. */
3990 register_rec(struct ovsrec_port *rec)
3992 if (n_recs >= allocated_recs) {
3993 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3995 recs[n_recs++] = rec;
3998 /* Frees all of the ports registered with register_reg(). */
4000 free_registered_recs(void)
4004 for (i = 0; i < n_recs; i++) {
4005 struct ovsrec_port *port = recs[i];
4008 for (j = 0; j < port->n_interfaces; j++) {
4009 struct ovsrec_interface *iface = port->interfaces[j];
4014 smap_destroy(&port->other_config);
4015 free(port->interfaces);
4023 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4026 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4028 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4032 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4035 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4036 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4037 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4040 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4041 * use, returns NULL.
4043 * Updates 'vlan_splinters_enabled_anywhere'. */
4044 static unsigned long int *
4045 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4047 unsigned long int *splinter_vlans;
4048 struct sset splinter_ifaces;
4049 const char *real_dev_name;
4050 struct shash *real_devs;
4051 struct shash_node *node;
4055 /* Free space allocated for synthesized ports and interfaces, since we're
4056 * in the process of reconstructing all of them. */
4057 free_registered_recs();
4059 splinter_vlans = bitmap_allocate(4096);
4060 sset_init(&splinter_ifaces);
4061 vlan_splinters_enabled_anywhere = false;
4062 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4063 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4066 for (j = 0; j < br_cfg->n_ports; j++) {
4067 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4070 for (k = 0; k < port_cfg->n_interfaces; k++) {
4071 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4073 if (vlan_splinters_is_enabled(iface_cfg)) {
4074 vlan_splinters_enabled_anywhere = true;
4075 sset_add(&splinter_ifaces, iface_cfg->name);
4076 vlan_bitmap_from_array__(port_cfg->trunks,
4082 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4083 bitmap_set1(splinter_vlans, *port_cfg->tag);
4088 if (!vlan_splinters_enabled_anywhere) {
4089 free(splinter_vlans);
4090 sset_destroy(&splinter_ifaces);
4094 HMAP_FOR_EACH (br, node, &all_bridges) {
4096 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4100 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4101 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4102 * device to be created for it. */
4103 bitmap_set0(splinter_vlans, 0);
4104 bitmap_set0(splinter_vlans, 4095);
4106 /* Delete all VLAN devices that we don't need. */
4108 real_devs = vlandev_get_real_devs();
4109 SHASH_FOR_EACH (node, real_devs) {
4110 const struct vlan_real_dev *real_dev = node->data;
4111 const struct vlan_dev *vlan_dev;
4112 bool real_dev_has_splinters;
4114 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4116 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4117 if (!real_dev_has_splinters
4118 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4119 struct netdev *netdev;
4121 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4122 if (!netdev_get_in4(netdev, NULL, NULL) ||
4123 !netdev_get_in6(netdev, NULL)) {
4124 /* It has an IP address configured, so we don't own
4125 * it. Don't delete it. */
4127 vlandev_del(vlan_dev->name);
4129 netdev_close(netdev);
4136 /* Add all VLAN devices that we need. */
4137 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4140 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4141 if (!vlandev_get_name(real_dev_name, vid)) {
4142 vlandev_add(real_dev_name, vid);
4149 sset_destroy(&splinter_ifaces);
4151 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4152 free(splinter_vlans);
4155 return splinter_vlans;
4158 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4161 configure_splinter_port(struct port *port)
4163 struct ofproto *ofproto = port->bridge->ofproto;
4164 ofp_port_t realdev_ofp_port;
4165 const char *realdev_name;
4166 struct iface *vlandev, *realdev;
4168 ofproto_bundle_unregister(port->bridge->ofproto, port);
4170 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4173 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4174 realdev = iface_lookup(port->bridge, realdev_name);
4175 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4177 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4181 static struct ovsrec_port *
4182 synthesize_splinter_port(const char *real_dev_name,
4183 const char *vlan_dev_name, int vid)
4185 struct ovsrec_interface *iface;
4186 struct ovsrec_port *port;
4188 iface = xmalloc(sizeof *iface);
4189 ovsrec_interface_init(iface);
4190 iface->name = xstrdup(vlan_dev_name);
4191 iface->type = "system";
4193 port = xmalloc(sizeof *port);
4194 ovsrec_port_init(port);
4195 port->interfaces = xmemdup(&iface, sizeof iface);
4196 port->n_interfaces = 1;
4197 port->name = xstrdup(vlan_dev_name);
4198 port->vlan_mode = "splinter";
4199 port->tag = xmalloc(sizeof *port->tag);
4202 smap_add(&port->other_config, "realdev", real_dev_name);
4208 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4209 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4210 * 1-bit in the 'splinter_vlans' bitmap. */
4212 add_vlan_splinter_ports(struct bridge *br,
4213 const unsigned long int *splinter_vlans,
4214 struct shash *ports)
4218 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4219 * we're modifying 'ports'. */
4220 for (i = 0; i < br->cfg->n_ports; i++) {
4221 const char *name = br->cfg->ports[i]->name;
4222 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4225 for (j = 0; j < port_cfg->n_interfaces; j++) {
4226 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4228 if (vlan_splinters_is_enabled(iface_cfg)) {
4229 const char *real_dev_name;
4232 real_dev_name = iface_cfg->name;
4233 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4234 const char *vlan_dev_name;
4236 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4238 && !shash_find(ports, vlan_dev_name)) {
4239 shash_add(ports, vlan_dev_name,
4240 synthesize_splinter_port(
4241 real_dev_name, vlan_dev_name, vid));
4250 mirror_refresh_stats(struct mirror *m)
4252 struct ofproto *ofproto = m->bridge->ofproto;
4253 uint64_t tx_packets, tx_bytes;
4256 size_t stat_cnt = 0;
4258 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4259 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4263 if (tx_packets != UINT64_MAX) {
4264 keys[stat_cnt] = "tx_packets";
4265 values[stat_cnt] = tx_packets;
4268 if (tx_bytes != UINT64_MAX) {
4269 keys[stat_cnt] = "tx_bytes";
4270 values[stat_cnt] = tx_bytes;
4274 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);