1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 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"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/ofproto.h"
43 #include "poll-loop.h"
47 #include "socket-util.h"
49 #include "stream-ssl.h"
51 #include "system-stats.h"
56 #include "lib/vswitch-idl.h"
57 #include "xenserver.h"
59 #include "sflow_api.h"
60 #include "vlan-bitmap.h"
62 VLOG_DEFINE_THIS_MODULE(bridge);
64 COVERAGE_DEFINE(bridge_reconfigure);
66 /* Configuration of an uninstantiated iface. */
68 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
69 const struct ovsrec_interface *cfg; /* Interface record. */
70 const struct ovsrec_port *parent; /* Parent port record. */
71 ofp_port_t ofport; /* Requested OpenFlow port number. */
74 /* OpenFlow port slated for removal from ofproto. */
76 struct list list_node; /* Node in bridge's ofpp_garbage. */
77 ofp_port_t ofp_port; /* Port to be deleted. */
81 /* These members are always valid. */
82 struct list port_elem; /* Element in struct port's "ifaces" list. */
83 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
87 /* These members are valid only after bridge_reconfigure() causes them to
89 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
90 ofp_port_t ofp_port; /* OpenFlow port number, */
91 /* OFPP_NONE if unknown. */
92 struct netdev *netdev; /* Network device. */
93 const char *type; /* Usually same as cfg->type. */
94 const struct ovsrec_interface *cfg;
98 struct uuid uuid; /* UUID of this "mirror" record in database. */
99 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
100 struct bridge *bridge;
102 const struct ovsrec_mirror *cfg;
106 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
107 struct bridge *bridge;
110 const struct ovsrec_port *cfg;
112 /* An ordinary bridge port has 1 interface.
113 * A bridge port for bonding has at least 2 interfaces. */
114 struct list ifaces; /* List of "struct iface"s. */
118 struct hmap_node node; /* In 'all_bridges'. */
119 char *name; /* User-specified arbitrary name. */
120 char *type; /* Datapath type. */
121 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
122 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
123 const struct ovsrec_bridge *cfg;
125 /* OpenFlow switch processing. */
126 struct ofproto *ofproto; /* OpenFlow switch. */
129 struct hmap ports; /* "struct port"s indexed by name. */
130 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
131 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
133 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
134 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
135 Indexed on 'cfg->name'. */
137 /* Port mirroring. */
138 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port;
142 struct ovsrec_interface synth_local_iface;
143 struct ovsrec_interface *synth_local_ifacep;
146 /* All bridges, indexed by name. */
147 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
149 /* OVSDB IDL used to obtain configuration. */
150 static struct ovsdb_idl *idl;
152 /* We want to complete daemonization, fully detaching from our parent process,
153 * only after we have completed our initial configuration, committed our state
154 * to the database, and received confirmation back from the database server
155 * that it applied the commit. This allows our parent process to know that,
156 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
157 * to have already been filled in. (It is only "very likely" rather than
158 * certain because there is always a slim possibility that the transaction will
159 * fail or that some other client has added new bridges, ports, etc. while
160 * ovs-vswitchd was configuring using an old configuration.)
162 * We only need to do this once for our initial configuration at startup, so
163 * 'initial_config_done' tracks whether we've already done it. While we are
164 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
165 * itself and is otherwise NULL. */
166 static bool initial_config_done;
167 static struct ovsdb_idl_txn *daemonize_txn;
169 /* Most recently processed IDL sequence number. */
170 static unsigned int idl_seqno;
172 /* Each time this timer expires, the bridge fetches interface and mirror
173 * statistics and pushes them into the database. */
174 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
175 static long long int iface_stats_timer = LLONG_MIN;
177 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
178 * expensive. This can cause bridge_reconfigure() to take a long time during
179 * which no other work can be done. To deal with this problem, we limit port
180 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
181 * call to bridge_reconfigure(). If there is more work to do after the limit
182 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
183 * This allows the rest of the code to catch up on important things like
184 * forwarding packets. */
185 #define OFP_PORT_ACTION_WINDOW 10
186 static bool reconfiguring = false;
188 static void add_del_bridges(const struct ovsrec_open_vswitch *);
189 static void bridge_update_ofprotos(void);
190 static void bridge_create(const struct ovsrec_bridge *);
191 static void bridge_destroy(struct bridge *);
192 static struct bridge *bridge_lookup(const char *name);
193 static unixctl_cb_func bridge_unixctl_dump_flows;
194 static unixctl_cb_func bridge_unixctl_reconnect;
195 static size_t bridge_get_controllers(const struct bridge *br,
196 struct ovsrec_controller ***controllersp);
197 static void bridge_add_del_ports(struct bridge *,
198 const unsigned long int *splinter_vlans);
199 static void bridge_refresh_ofp_port(struct bridge *);
200 static void bridge_configure_flow_miss_model(const char *opt);
201 static void bridge_configure_datapath_id(struct bridge *);
202 static void bridge_configure_netflow(struct bridge *);
203 static void bridge_configure_forward_bpdu(struct bridge *);
204 static void bridge_configure_mac_table(struct bridge *);
205 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
206 static void bridge_configure_ipfix(struct bridge *);
207 static void bridge_configure_stp(struct bridge *);
208 static void bridge_configure_tables(struct bridge *);
209 static void bridge_configure_dp_desc(struct bridge *);
210 static void bridge_configure_remotes(struct bridge *,
211 const struct sockaddr_in *managers,
213 static void bridge_pick_local_hw_addr(struct bridge *,
214 uint8_t ea[ETH_ADDR_LEN],
215 struct iface **hw_addr_iface);
216 static uint64_t bridge_pick_datapath_id(struct bridge *,
217 const uint8_t bridge_ea[ETH_ADDR_LEN],
218 struct iface *hw_addr_iface);
219 static void bridge_queue_if_cfg(struct bridge *,
220 const struct ovsrec_interface *,
221 const struct ovsrec_port *);
222 static uint64_t dpid_from_hash(const void *, size_t nbytes);
223 static bool bridge_has_bond_fake_iface(const struct bridge *,
225 static bool port_is_bond_fake_iface(const struct port *);
227 static unixctl_cb_func qos_unixctl_show;
229 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
230 static void port_del_ifaces(struct port *);
231 static void port_destroy(struct port *);
232 static struct port *port_lookup(const struct bridge *, const char *name);
233 static void port_configure(struct port *);
234 static struct lacp_settings *port_configure_lacp(struct port *,
235 struct lacp_settings *);
236 static void port_configure_bond(struct port *, struct bond_settings *);
237 static bool port_is_synthetic(const struct port *);
239 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
240 static void run_system_stats(void);
242 static void bridge_configure_mirrors(struct bridge *);
243 static struct mirror *mirror_create(struct bridge *,
244 const struct ovsrec_mirror *);
245 static void mirror_destroy(struct mirror *);
246 static bool mirror_configure(struct mirror *);
247 static void mirror_refresh_stats(struct mirror *);
249 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
250 static bool iface_create(struct bridge *, struct if_cfg *,
251 ofp_port_t ofp_port);
252 static bool iface_is_internal(const struct ovsrec_interface *iface,
253 const struct ovsrec_bridge *br);
254 static const char *iface_get_type(const struct ovsrec_interface *,
255 const struct ovsrec_bridge *);
256 static void iface_destroy(struct iface *);
257 static struct iface *iface_lookup(const struct bridge *, const char *name);
258 static struct iface *iface_find(const char *name);
259 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
260 static struct iface *iface_from_ofp_port(const struct bridge *,
261 ofp_port_t ofp_port);
262 static void iface_set_mac(struct iface *);
263 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
264 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
265 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
266 static void iface_configure_cfm(struct iface *);
267 static void iface_refresh_cfm_stats(struct iface *);
268 static void iface_refresh_stats(struct iface *);
269 static void iface_refresh_status(struct iface *);
270 static bool iface_is_synthetic(const struct iface *);
271 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
273 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
275 * This is deprecated. It is only for compatibility with broken device drivers
276 * in old versions of Linux that do not properly support VLANs when VLAN
277 * devices are not used. When broken device drivers are no longer in
278 * widespread use, we will delete these interfaces. */
280 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
281 static bool vlan_splinters_enabled_anywhere;
283 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
284 static unsigned long int *collect_splinter_vlans(
285 const struct ovsrec_open_vswitch *);
286 static void configure_splinter_port(struct port *);
287 static void add_vlan_splinter_ports(struct bridge *,
288 const unsigned long int *splinter_vlans,
289 struct shash *ports);
292 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
294 struct shash iface_hints;
295 static bool initialized = false;
302 shash_init(&iface_hints);
305 for (i = 0; i < cfg->n_bridges; i++) {
306 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
309 for (j = 0; j < br_cfg->n_ports; j++) {
310 struct ovsrec_port *port_cfg = br_cfg->ports[j];
313 for (k = 0; k < port_cfg->n_interfaces; k++) {
314 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
315 struct iface_hint *iface_hint;
317 iface_hint = xmalloc(sizeof *iface_hint);
318 iface_hint->br_name = br_cfg->name;
319 iface_hint->br_type = br_cfg->datapath_type;
320 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
322 shash_add(&iface_hints, if_cfg->name, iface_hint);
328 ofproto_init(&iface_hints);
330 shash_destroy_free_data(&iface_hints);
334 /* Public functions. */
336 /* Initializes the bridge module, configuring it to obtain its configuration
337 * from an OVSDB server accessed over 'remote', which should be a string in a
338 * form acceptable to ovsdb_idl_create(). */
340 bridge_init(const char *remote)
342 /* Create connection to database. */
343 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
344 idl_seqno = ovsdb_idl_get_seqno(idl);
345 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
346 ovsdb_idl_verify_write_only(idl);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
349 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
354 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
357 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
358 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
361 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
362 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
363 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
382 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
383 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
385 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
387 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
388 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
395 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
397 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
406 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
408 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
410 /* Register unixctl commands. */
411 unixctl_command_register("qos/show", "interface", 1, 1,
412 qos_unixctl_show, NULL);
413 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
414 bridge_unixctl_dump_flows, NULL);
415 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
416 bridge_unixctl_reconnect, NULL);
426 struct bridge *br, *next_br;
428 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
431 ovsdb_idl_destroy(idl);
434 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
435 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
436 * responsible for freeing '*managersp' (with free()).
438 * You may be asking yourself "why does ovs-vswitchd care?", because
439 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
440 * should not be and in fact is not directly involved in that. But
441 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
442 * it has to tell in-band control where the managers are to enable that.
443 * (Thus, only managers connected in-band are collected.)
446 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
447 struct sockaddr_in **managersp, size_t *n_managersp)
449 struct sockaddr_in *managers = NULL;
450 size_t n_managers = 0;
454 /* Collect all of the potential targets from the "targets" columns of the
455 * rows pointed to by "manager_options", excluding any that are
458 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
459 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
461 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
462 sset_find_and_delete(&targets, m->target);
464 sset_add(&targets, m->target);
468 /* Now extract the targets' IP addresses. */
469 if (!sset_is_empty(&targets)) {
472 managers = xmalloc(sset_count(&targets) * sizeof *managers);
473 SSET_FOR_EACH (target, &targets) {
474 struct sockaddr_in *sin = &managers[n_managers];
476 if (stream_parse_target_with_default_ports(target,
484 sset_destroy(&targets);
486 *managersp = managers;
487 *n_managersp = n_managers;
491 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
493 unsigned long int *splinter_vlans;
496 COVERAGE_INC(bridge_reconfigure);
498 ovs_assert(!reconfiguring);
499 reconfiguring = true;
501 ofproto_set_flow_eviction_threshold(
502 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
503 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
505 ofproto_set_n_handler_threads(
506 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0));
508 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
509 "force-miss-model"));
511 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
512 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
513 * configuration otherwise.
515 * This is mostly an update to bridge data structures. Nothing is pushed
516 * down to ofproto or lower layers. */
517 add_del_bridges(ovs_cfg);
518 splinter_vlans = collect_splinter_vlans(ovs_cfg);
519 HMAP_FOR_EACH (br, node, &all_bridges) {
520 bridge_add_del_ports(br, splinter_vlans);
522 free(splinter_vlans);
524 /* Delete datapaths that are no longer configured, and create ones which
525 * don't exist but should. */
526 bridge_update_ofprotos();
528 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
529 HMAP_FOR_EACH (br, node, &all_bridges) {
530 bridge_refresh_ofp_port(br);
533 /* Clear database records for "if_cfg"s which haven't been instantiated. */
534 HMAP_FOR_EACH (br, node, &all_bridges) {
535 struct if_cfg *if_cfg;
537 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
538 iface_clear_db_record(if_cfg->cfg);
542 reconfigure_system_stats(ovs_cfg);
546 bridge_reconfigure_ofp(void)
548 long long int deadline;
552 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
554 /* The kernel will reject any attempt to add a given port to a datapath if
555 * that port already belongs to a different datapath, so we must do all
556 * port deletions before any port additions. */
557 HMAP_FOR_EACH (br, node, &all_bridges) {
558 struct ofpp_garbage *garbage, *next;
560 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
561 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
562 * internal datastructures may not be consistent. Eventually, when
563 * port additions and deletions are cheaper, these calls should be
566 ofproto_port_del(br->ofproto, garbage->ofp_port);
567 list_remove(&garbage->list_node);
571 if (time_msec() >= deadline) {
578 HMAP_FOR_EACH (br, node, &all_bridges) {
579 struct if_cfg *if_cfg, *next;
581 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
582 iface_create(br, if_cfg, OFPP_NONE);
584 if (time_msec() >= deadline) {
594 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
596 struct sockaddr_in *managers;
597 int sflow_bridge_number;
602 ovs_assert(reconfiguring);
603 done = bridge_reconfigure_ofp();
605 /* Complete the configuration. */
606 sflow_bridge_number = 0;
607 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
608 HMAP_FOR_EACH (br, node, &all_bridges) {
611 /* We need the datapath ID early to allow LACP ports to use it as the
612 * default system ID. */
613 bridge_configure_datapath_id(br);
615 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
618 port_configure(port);
620 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
621 iface_configure_cfm(iface);
622 iface_configure_qos(iface, port->cfg->qos);
623 iface_set_mac(iface);
624 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
628 bridge_configure_mirrors(br);
629 bridge_configure_forward_bpdu(br);
630 bridge_configure_mac_table(br);
631 bridge_configure_remotes(br, managers, n_managers);
632 bridge_configure_netflow(br);
633 bridge_configure_sflow(br, &sflow_bridge_number);
634 bridge_configure_ipfix(br);
635 bridge_configure_stp(br);
636 bridge_configure_tables(br);
637 bridge_configure_dp_desc(br);
639 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
640 /* XXX: Remove this warning message eventually. */
641 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
642 " moved to the Open_vSwitch table. Ignoring its"
643 " setting in the bridge table.");
651 /* Delete ofprotos which aren't configured or have the wrong type. Create
652 * ofprotos which don't exist but need to. */
654 bridge_update_ofprotos(void)
656 struct bridge *br, *next;
661 /* Delete ofprotos with no bridge or with the wrong type. */
664 ofproto_enumerate_types(&types);
665 SSET_FOR_EACH (type, &types) {
668 ofproto_enumerate_names(type, &names);
669 SSET_FOR_EACH (name, &names) {
670 br = bridge_lookup(name);
671 if (!br || strcmp(type, br->type)) {
672 ofproto_delete(name, type);
676 sset_destroy(&names);
677 sset_destroy(&types);
679 /* Add ofprotos for bridges which don't have one yet. */
680 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
688 /* Remove ports from any datapath with the same name as 'br'. If we
689 * don't do this, creating 'br''s ofproto will fail because a port with
690 * the same name as its local port already exists. */
691 HMAP_FOR_EACH (br2, node, &all_bridges) {
692 struct ofproto_port ofproto_port;
698 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
700 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
702 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
703 ovs_strerror(error));
705 ofproto_port_destroy(&ofproto_port);
709 error = ofproto_create(br->name, br->type, &br->ofproto);
711 VLOG_ERR("failed to create bridge %s: %s", br->name,
712 ovs_strerror(error));
719 port_configure(struct port *port)
721 const struct ovsrec_port *cfg = port->cfg;
722 struct bond_settings bond_settings;
723 struct lacp_settings lacp_settings;
724 struct ofproto_bundle_settings s;
727 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
728 configure_splinter_port(port);
737 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
738 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
739 s.slaves[s.n_slaves++] = iface->ofp_port;
744 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
748 /* Get VLAN trunks. */
751 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
755 if (cfg->vlan_mode) {
756 if (!strcmp(cfg->vlan_mode, "access")) {
757 s.vlan_mode = PORT_VLAN_ACCESS;
758 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
759 s.vlan_mode = PORT_VLAN_TRUNK;
760 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
761 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
762 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
763 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
765 /* This "can't happen" because ovsdb-server should prevent it. */
766 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
767 s.vlan_mode = PORT_VLAN_TRUNK;
771 s.vlan_mode = PORT_VLAN_ACCESS;
773 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
777 s.vlan_mode = PORT_VLAN_TRUNK;
780 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
783 /* Get LACP settings. */
784 s.lacp = port_configure_lacp(port, &lacp_settings);
788 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
789 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
790 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
793 s.lacp_slaves = NULL;
796 /* Get bond settings. */
797 if (s.n_slaves > 1) {
798 s.bond = &bond_settings;
799 port_configure_bond(port, &bond_settings);
802 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
803 netdev_set_miimon_interval(iface->netdev, 0);
808 ofproto_bundle_register(port->bridge->ofproto, port, &s);
817 bridge_configure_flow_miss_model(const char *opt)
819 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
822 if (strcmp(opt, "with-facets")) {
823 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
824 } else if (strcmp(opt, "without-facets")) {
825 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
829 ofproto_set_flow_miss_model(model);
832 /* Pick local port hardware address and datapath ID for 'br'. */
834 bridge_configure_datapath_id(struct bridge *br)
836 uint8_t ea[ETH_ADDR_LEN];
838 struct iface *local_iface;
839 struct iface *hw_addr_iface;
842 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
843 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
845 int error = netdev_set_etheraddr(local_iface->netdev, ea);
847 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
848 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
849 "Ethernet address: %s",
850 br->name, ovs_strerror(error));
853 memcpy(br->ea, ea, ETH_ADDR_LEN);
855 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
856 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
857 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
858 ofproto_set_datapath_id(br->ofproto, dpid);
861 dpid_string = xasprintf("%016"PRIx64, dpid);
862 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
866 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
869 bridge_get_allowed_versions(struct bridge *br)
871 if (!br->cfg->n_protocols)
874 return ofputil_versions_from_strings(br->cfg->protocols,
875 br->cfg->n_protocols);
878 /* Set NetFlow configuration on 'br'. */
880 bridge_configure_netflow(struct bridge *br)
882 struct ovsrec_netflow *cfg = br->cfg->netflow;
883 struct netflow_options opts;
886 ofproto_set_netflow(br->ofproto, NULL);
890 memset(&opts, 0, sizeof opts);
892 /* Get default NetFlow configuration from datapath.
893 * Apply overrides from 'cfg'. */
894 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
895 if (cfg->engine_type) {
896 opts.engine_type = *cfg->engine_type;
898 if (cfg->engine_id) {
899 opts.engine_id = *cfg->engine_id;
902 /* Configure active timeout interval. */
903 opts.active_timeout = cfg->active_timeout;
904 if (!opts.active_timeout) {
905 opts.active_timeout = -1;
906 } else if (opts.active_timeout < 0) {
907 VLOG_WARN("bridge %s: active timeout interval set to negative "
908 "value, using default instead (%d seconds)", br->name,
909 NF_ACTIVE_TIMEOUT_DEFAULT);
910 opts.active_timeout = -1;
913 /* Add engine ID to interface number to disambiguate bridgs? */
914 opts.add_id_to_iface = cfg->add_id_to_interface;
915 if (opts.add_id_to_iface) {
916 if (opts.engine_id > 0x7f) {
917 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
918 "another vswitch, choose an engine id less than 128",
921 if (hmap_count(&br->ports) > 508) {
922 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
923 "another port when more than 508 ports are used",
929 sset_init(&opts.collectors);
930 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
933 if (ofproto_set_netflow(br->ofproto, &opts)) {
934 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
936 sset_destroy(&opts.collectors);
939 /* Set sFlow configuration on 'br'. */
941 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
943 const struct ovsrec_sflow *cfg = br->cfg->sflow;
944 struct ovsrec_controller **controllers;
945 struct ofproto_sflow_options oso;
946 size_t n_controllers;
950 ofproto_set_sflow(br->ofproto, NULL);
954 memset(&oso, 0, sizeof oso);
956 sset_init(&oso.targets);
957 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
959 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
961 oso.sampling_rate = *cfg->sampling;
964 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
966 oso.polling_interval = *cfg->polling;
969 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
971 oso.header_len = *cfg->header;
974 oso.sub_id = (*sflow_bridge_number)++;
975 oso.agent_device = cfg->agent;
977 oso.control_ip = NULL;
978 n_controllers = bridge_get_controllers(br, &controllers);
979 for (i = 0; i < n_controllers; i++) {
980 if (controllers[i]->local_ip) {
981 oso.control_ip = controllers[i]->local_ip;
985 ofproto_set_sflow(br->ofproto, &oso);
987 sset_destroy(&oso.targets);
990 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
992 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
993 const struct bridge *br)
995 return fscs->ipfix && fscs->bridge == br->cfg;
998 /* Set IPFIX configuration on 'br'. */
1000 bridge_configure_ipfix(struct bridge *br)
1002 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1003 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1004 struct ofproto_ipfix_bridge_exporter_options be_opts;
1005 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1006 size_t n_fe_opts = 0;
1008 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1009 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1014 if (!be_cfg && n_fe_opts == 0) {
1015 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1020 memset(&be_opts, 0, sizeof be_opts);
1022 sset_init(&be_opts.targets);
1023 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1025 if (be_cfg->sampling) {
1026 be_opts.sampling_rate = *be_cfg->sampling;
1028 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1030 if (be_cfg->obs_domain_id) {
1031 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1033 if (be_cfg->obs_point_id) {
1034 be_opts.obs_point_id = *be_cfg->obs_point_id;
1036 if (be_cfg->cache_active_timeout) {
1037 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1039 if (be_cfg->cache_max_flows) {
1040 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1044 if (n_fe_opts > 0) {
1045 struct ofproto_ipfix_flow_exporter_options *opts;
1046 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1048 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1049 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1050 opts->collector_set_id = fe_cfg->id;
1051 sset_init(&opts->targets);
1052 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1053 fe_cfg->ipfix->n_targets);
1054 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1055 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1056 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1057 ? *fe_cfg->ipfix->cache_max_flows : 0;
1063 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1067 sset_destroy(&be_opts.targets);
1070 if (n_fe_opts > 0) {
1071 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1073 for (i = 0; i < n_fe_opts; i++) {
1074 sset_destroy(&opts->targets);
1082 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1083 struct ofproto_port_stp_settings *port_s,
1084 int *port_num_counter, unsigned long *port_num_bitmap)
1086 const char *config_str;
1087 struct iface *iface;
1089 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1090 port_s->enable = false;
1093 port_s->enable = true;
1096 /* STP over bonds is not supported. */
1097 if (!list_is_singleton(&port->ifaces)) {
1098 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1100 port_s->enable = false;
1104 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1106 /* Internal ports shouldn't participate in spanning tree, so
1108 if (!strcmp(iface->type, "internal")) {
1109 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1110 port_s->enable = false;
1114 /* STP on mirror output ports is not supported. */
1115 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1116 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1117 port_s->enable = false;
1121 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1123 unsigned long int port_num = strtoul(config_str, NULL, 0);
1124 int port_idx = port_num - 1;
1126 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1127 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1128 port_s->enable = false;
1132 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1133 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1134 port->name, port_num);
1135 port_s->enable = false;
1138 bitmap_set1(port_num_bitmap, port_idx);
1139 port_s->port_num = port_idx;
1141 if (*port_num_counter >= STP_MAX_PORTS) {
1142 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1143 port_s->enable = false;
1147 port_s->port_num = (*port_num_counter)++;
1150 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1152 port_s->path_cost = strtoul(config_str, NULL, 10);
1154 enum netdev_features current;
1157 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1158 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1159 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1162 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1164 port_s->priority = strtoul(config_str, NULL, 0);
1166 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1170 /* Set spanning tree configuration on 'br'. */
1172 bridge_configure_stp(struct bridge *br)
1174 if (!br->cfg->stp_enable) {
1175 ofproto_set_stp(br->ofproto, NULL);
1177 struct ofproto_stp_settings br_s;
1178 const char *config_str;
1180 int port_num_counter;
1181 unsigned long *port_num_bitmap;
1183 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1185 uint8_t ea[ETH_ADDR_LEN];
1187 if (eth_addr_from_string(config_str, ea)) {
1188 br_s.system_id = eth_addr_to_uint64(ea);
1190 br_s.system_id = eth_addr_to_uint64(br->ea);
1191 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1192 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1195 br_s.system_id = eth_addr_to_uint64(br->ea);
1198 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1200 br_s.priority = strtoul(config_str, NULL, 0);
1202 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1205 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1207 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1209 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1212 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1214 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1216 br_s.max_age = STP_DEFAULT_MAX_AGE;
1219 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1221 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1223 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1226 /* Configure STP on the bridge. */
1227 if (ofproto_set_stp(br->ofproto, &br_s)) {
1228 VLOG_ERR("bridge %s: could not enable STP", br->name);
1232 /* Users must either set the port number with the "stp-port-num"
1233 * configuration on all ports or none. If manual configuration
1234 * is not done, then we allocate them sequentially. */
1235 port_num_counter = 0;
1236 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1237 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1238 struct ofproto_port_stp_settings port_s;
1239 struct iface *iface;
1241 port_configure_stp(br->ofproto, port, &port_s,
1242 &port_num_counter, port_num_bitmap);
1244 /* As bonds are not supported, just apply configuration to
1245 * all interfaces. */
1246 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1247 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1249 VLOG_ERR("port %s: could not enable STP", port->name);
1255 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1256 && port_num_counter) {
1257 VLOG_ERR("bridge %s: must manually configure all STP port "
1258 "IDs or none, disabling", br->name);
1259 ofproto_set_stp(br->ofproto, NULL);
1261 bitmap_free(port_num_bitmap);
1266 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1268 const struct port *port = port_lookup(br, name);
1269 return port && port_is_bond_fake_iface(port);
1273 port_is_bond_fake_iface(const struct port *port)
1275 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1279 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1281 struct bridge *br, *next;
1282 struct shash new_br;
1285 /* Collect new bridges' names and types. */
1286 shash_init(&new_br);
1287 for (i = 0; i < cfg->n_bridges; i++) {
1288 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1289 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1291 if (strchr(br_cfg->name, '/')) {
1292 /* Prevent remote ovsdb-server users from accessing arbitrary
1293 * directories, e.g. consider a bridge named "../../../etc/". */
1294 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1296 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1297 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1301 /* Get rid of deleted bridges or those whose types have changed.
1302 * Update 'cfg' of bridges that still exist. */
1303 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1304 br->cfg = shash_find_data(&new_br, br->name);
1305 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1306 br->cfg->datapath_type))) {
1311 /* Add new bridges. */
1312 for (i = 0; i < cfg->n_bridges; i++) {
1313 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1314 struct bridge *br = bridge_lookup(br_cfg->name);
1316 bridge_create(br_cfg);
1320 shash_destroy(&new_br);
1324 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1326 struct bridge *br = iface->port->bridge;
1328 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1329 iface->ofp_port = ofp_port;
1330 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1331 hash_ofp_port(ofp_port));
1332 iface_set_ofport(iface->cfg, ofp_port);
1335 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1336 * Returns 0 if successful, otherwise a positive errno value. */
1338 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1339 struct netdev *netdev)
1343 error = netdev_set_config(netdev, &iface_cfg->options);
1345 VLOG_WARN("could not configure network device %s (%s)",
1346 iface_cfg->name, ovs_strerror(error));
1351 /* This function determines whether 'ofproto_port', which is attached to
1352 * br->ofproto's datapath, is one that we want in 'br'.
1354 * If it is, it returns true, after creating an iface (if necessary),
1355 * configuring the iface's netdev according to the iface's options, and setting
1356 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1358 * If, on the other hand, 'port' should be removed, it returns false. The
1359 * caller should later detach the port from br->ofproto. */
1361 bridge_refresh_one_ofp_port(struct bridge *br,
1362 const struct ofproto_port *ofproto_port)
1364 const char *name = ofproto_port->name;
1365 const char *type = ofproto_port->type;
1366 ofp_port_t ofp_port = ofproto_port->ofp_port;
1368 struct iface *iface = iface_lookup(br, name);
1370 /* Check that the name-to-number mapping is one-to-one. */
1371 if (iface->ofp_port != OFPP_NONE) {
1372 VLOG_WARN("bridge %s: interface %s reported twice",
1375 } else if (iface_from_ofp_port(br, ofp_port)) {
1376 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1377 br->name, ofp_port);
1381 /* There's a configured interface named 'name'. */
1382 if (strcmp(type, iface->type)
1383 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1384 /* It's the wrong type, or it's the right type but can't be
1385 * configured as the user requested, so we must destroy it. */
1388 /* It's the right type and configured correctly. Keep it. */
1389 iface_set_ofp_port(iface, ofp_port);
1392 } else if (bridge_has_bond_fake_iface(br, name)
1393 && !strcmp(type, "internal")) {
1394 /* It's a bond fake iface. Keep it. */
1397 /* There's no configured interface named 'name', but there might be an
1398 * interface of that name queued to be created.
1400 * If there is, and it has the correct type, then try to configure it
1401 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1402 * delete it and later try to re-add it. */
1403 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1405 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1406 && iface_create(br, if_cfg, ofp_port));
1410 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1411 * consistent with the ofp_ports in "br->ofproto". */
1413 bridge_refresh_ofp_port(struct bridge *br)
1415 struct ofproto_port_dump dump;
1416 struct ofproto_port ofproto_port;
1417 struct port *port, *port_next;
1419 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1420 hmap_clear(&br->ifaces);
1421 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1422 struct iface *iface;
1424 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1425 iface->ofp_port = OFPP_NONE;
1429 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1430 * already exist in the datapath and promote them to full fledged "struct
1431 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1432 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1433 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1434 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1435 garbage->ofp_port = ofproto_port.ofp_port;
1436 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1440 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1441 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1442 * later they can be added to ofproto. */
1443 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1444 struct iface *iface, *iface_next;
1446 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1447 if (iface->ofp_port == OFPP_NONE) {
1448 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1449 iface_destroy(iface);
1453 if (list_is_empty(&port->ifaces)) {
1459 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1460 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1461 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1464 * If successful, returns 0 and stores the network device in '*netdevp'. On
1465 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1467 iface_do_create(const struct bridge *br,
1468 const struct if_cfg *if_cfg,
1469 ofp_port_t *ofp_portp, struct netdev **netdevp)
1471 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1472 const struct ovsrec_port *port_cfg = if_cfg->parent;
1473 struct netdev *netdev = NULL;
1476 if (netdev_is_reserved_name(iface_cfg->name)) {
1477 VLOG_WARN("could not create interface %s, name is reserved",
1483 error = netdev_open(iface_cfg->name,
1484 iface_get_type(iface_cfg, br->cfg), &netdev);
1486 VLOG_WARN("could not open network device %s (%s)",
1487 iface_cfg->name, ovs_strerror(error));
1491 error = iface_set_netdev_config(iface_cfg, netdev);
1496 if (*ofp_portp == OFPP_NONE) {
1497 ofp_port_t ofp_port = if_cfg->ofport;
1499 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1503 *ofp_portp = ofp_port;
1505 VLOG_INFO("bridge %s: added interface %s on port %d",
1506 br->name, iface_cfg->name, *ofp_portp);
1508 VLOG_DBG("bridge %s: interface %s is on port %d",
1509 br->name, iface_cfg->name, *ofp_portp);
1512 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1513 || iface_is_internal(iface_cfg, br->cfg)) {
1514 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1522 netdev_close(netdev);
1526 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1527 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1528 * automatically allocated for the iface. Takes ownership of and
1529 * deallocates 'if_cfg'.
1531 * Return true if an iface is successfully created, false otherwise. */
1533 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1535 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1536 const struct ovsrec_port *port_cfg = if_cfg->parent;
1538 struct netdev *netdev;
1539 struct iface *iface;
1544 /* Do the bits that can fail up front.
1546 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1547 * internal datastructures may not be consistent. Eventually, when port
1548 * additions and deletions are cheaper, these calls should be removed. */
1550 ovs_assert(!iface_lookup(br, iface_cfg->name));
1551 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1554 iface_set_ofport(iface_cfg, OFPP_NONE);
1555 iface_clear_db_record(iface_cfg);
1560 /* Get or create the port structure. */
1561 port = port_lookup(br, port_cfg->name);
1563 port = port_create(br, port_cfg);
1566 /* Create the iface structure. */
1567 iface = xzalloc(sizeof *iface);
1568 list_push_back(&port->ifaces, &iface->port_elem);
1569 hmap_insert(&br->iface_by_name, &iface->name_node,
1570 hash_string(iface_cfg->name, 0));
1572 iface->name = xstrdup(iface_cfg->name);
1573 iface->ofp_port = OFPP_NONE;
1574 iface->netdev = netdev;
1575 iface->type = iface_get_type(iface_cfg, br->cfg);
1576 iface->cfg = iface_cfg;
1578 iface_set_ofp_port(iface, ofp_port);
1580 /* Populate initial status in database. */
1581 iface_refresh_stats(iface);
1582 iface_refresh_status(iface);
1584 /* Add bond fake iface if necessary. */
1585 if (port_is_bond_fake_iface(port)) {
1586 struct ofproto_port ofproto_port;
1588 if (ofproto_port_query_by_name(br->ofproto, port->name,
1590 struct netdev *netdev;
1593 error = netdev_open(port->name, "internal", &netdev);
1595 ofp_port_t fake_ofp_port = if_cfg->ofport;
1597 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1598 netdev_close(netdev);
1600 VLOG_WARN("could not open network device %s (%s)",
1601 port->name, ovs_strerror(error));
1604 /* Already exists, nothing to do. */
1605 ofproto_port_destroy(&ofproto_port);
1610 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1616 /* Set forward BPDU option. */
1618 bridge_configure_forward_bpdu(struct bridge *br)
1620 ofproto_set_forward_bpdu(br->ofproto,
1621 smap_get_bool(&br->cfg->other_config,
1626 /* Set MAC learning table configuration for 'br'. */
1628 bridge_configure_mac_table(struct bridge *br)
1630 const char *idle_time_str;
1633 const char *mac_table_size_str;
1636 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1637 idle_time = (idle_time_str && atoi(idle_time_str)
1638 ? atoi(idle_time_str)
1639 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1641 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1642 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1643 ? atoi(mac_table_size_str)
1646 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1650 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1651 struct iface **hw_addr_iface)
1653 struct hmapx mirror_output_ports;
1656 bool found_addr = false;
1660 *hw_addr_iface = NULL;
1662 /* Did the user request a particular MAC? */
1663 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1664 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1665 if (eth_addr_is_multicast(ea)) {
1666 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1667 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1668 } else if (eth_addr_is_zero(ea)) {
1669 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1675 /* Mirror output ports don't participate in picking the local hardware
1676 * address. ofproto can't help us find out whether a given port is a
1677 * mirror output because we haven't configured mirrors yet, so we need to
1678 * accumulate them ourselves. */
1679 hmapx_init(&mirror_output_ports);
1680 for (i = 0; i < br->cfg->n_mirrors; i++) {
1681 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1682 if (m->output_port) {
1683 hmapx_add(&mirror_output_ports, m->output_port);
1687 /* Otherwise choose the minimum non-local MAC address among all of the
1689 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1690 uint8_t iface_ea[ETH_ADDR_LEN];
1691 struct iface *candidate;
1692 struct iface *iface;
1694 /* Mirror output ports don't participate. */
1695 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1699 /* Choose the MAC address to represent the port. */
1701 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1702 /* Find the interface with this Ethernet address (if any) so that
1703 * we can provide the correct devname to the caller. */
1704 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1705 uint8_t candidate_ea[ETH_ADDR_LEN];
1706 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1707 && eth_addr_equals(iface_ea, candidate_ea)) {
1712 /* Choose the interface whose MAC address will represent the port.
1713 * The Linux kernel bonding code always chooses the MAC address of
1714 * the first slave added to a bond, and the Fedora networking
1715 * scripts always add slaves to a bond in alphabetical order, so
1716 * for compatibility we choose the interface with the name that is
1717 * first in alphabetical order. */
1718 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1719 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1724 /* The local port doesn't count (since we're trying to choose its
1725 * MAC address anyway). */
1726 if (iface->ofp_port == OFPP_LOCAL) {
1731 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1737 /* Compare against our current choice. */
1738 if (!eth_addr_is_multicast(iface_ea) &&
1739 !eth_addr_is_local(iface_ea) &&
1740 !eth_addr_is_reserved(iface_ea) &&
1741 !eth_addr_is_zero(iface_ea) &&
1742 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1744 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1745 *hw_addr_iface = iface;
1751 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1752 *hw_addr_iface = NULL;
1755 hmapx_destroy(&mirror_output_ports);
1758 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1759 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1760 * an interface on 'br', then that interface must be passed in as
1761 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1762 * 'hw_addr_iface' must be passed in as a null pointer. */
1764 bridge_pick_datapath_id(struct bridge *br,
1765 const uint8_t bridge_ea[ETH_ADDR_LEN],
1766 struct iface *hw_addr_iface)
1769 * The procedure for choosing a bridge MAC address will, in the most
1770 * ordinary case, also choose a unique MAC that we can use as a datapath
1771 * ID. In some special cases, though, multiple bridges will end up with
1772 * the same MAC address. This is OK for the bridges, but it will confuse
1773 * the OpenFlow controller, because each datapath needs a unique datapath
1776 * Datapath IDs must be unique. It is also very desirable that they be
1777 * stable from one run to the next, so that policy set on a datapath
1780 const char *datapath_id;
1783 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1784 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1788 if (!hw_addr_iface) {
1790 * A purely internal bridge, that is, one that has no non-virtual
1791 * network devices on it at all, is difficult because it has no
1792 * natural unique identifier at all.
1794 * When the host is a XenServer, we handle this case by hashing the
1795 * host's UUID with the name of the bridge. Names of bridges are
1796 * persistent across XenServer reboots, although they can be reused if
1797 * an internal network is destroyed and then a new one is later
1798 * created, so this is fairly effective.
1800 * When the host is not a XenServer, we punt by using a random MAC
1801 * address on each run.
1803 const char *host_uuid = xenserver_get_host_uuid();
1805 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1806 dpid = dpid_from_hash(combined, strlen(combined));
1812 return eth_addr_to_uint64(bridge_ea);
1816 dpid_from_hash(const void *data, size_t n)
1818 uint8_t hash[SHA1_DIGEST_SIZE];
1820 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1821 sha1_bytes(data, n, hash);
1822 eth_addr_mark_random(hash);
1823 return eth_addr_to_uint64(hash);
1827 iface_refresh_status(struct iface *iface)
1831 enum netdev_features current;
1835 uint8_t mac[ETH_ADDR_LEN];
1839 if (iface_is_synthetic(iface)) {
1845 if (!netdev_get_status(iface->netdev, &smap)) {
1846 ovsrec_interface_set_status(iface->cfg, &smap);
1848 ovsrec_interface_set_status(iface->cfg, NULL);
1851 smap_destroy(&smap);
1853 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1854 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1856 ovsrec_interface_set_duplex(iface->cfg,
1857 netdev_features_is_full_duplex(current)
1859 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1861 ovsrec_interface_set_duplex(iface->cfg, NULL);
1862 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1865 error = netdev_get_mtu(iface->netdev, &mtu);
1868 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1870 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1873 error = netdev_get_etheraddr(iface->netdev, mac);
1875 char mac_string[32];
1877 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1878 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1880 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1883 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1884 * if there is no valid ifindex number. */
1885 ifindex64 = netdev_get_ifindex(iface->netdev);
1886 if (ifindex64 < 0) {
1889 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1892 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1895 iface_refresh_cfm_stats(struct iface *iface)
1897 const struct ovsrec_interface *cfg = iface->cfg;
1898 struct ofproto_cfm_status status;
1900 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1901 iface->ofp_port, &status)) {
1902 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1903 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1904 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1905 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1906 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1908 const char *reasons[CFM_FAULT_N_REASONS];
1909 int64_t cfm_health = status.health;
1910 bool faulted = status.faults != 0;
1913 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1916 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1917 int reason = 1 << i;
1918 if (status.faults & reason) {
1919 reasons[j++] = cfm_fault_reason_to_str(reason);
1922 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1924 if (status.remote_opstate >= 0) {
1925 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1926 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1928 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1931 ovsrec_interface_set_cfm_remote_mpids(cfg,
1932 (const int64_t *)status.rmps,
1934 if (cfm_health >= 0) {
1935 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1937 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1945 iface_refresh_stats(struct iface *iface)
1947 #define IFACE_STATS \
1948 IFACE_STAT(rx_packets, "rx_packets") \
1949 IFACE_STAT(tx_packets, "tx_packets") \
1950 IFACE_STAT(rx_bytes, "rx_bytes") \
1951 IFACE_STAT(tx_bytes, "tx_bytes") \
1952 IFACE_STAT(rx_dropped, "rx_dropped") \
1953 IFACE_STAT(tx_dropped, "tx_dropped") \
1954 IFACE_STAT(rx_errors, "rx_errors") \
1955 IFACE_STAT(tx_errors, "tx_errors") \
1956 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1957 IFACE_STAT(rx_over_errors, "rx_over_err") \
1958 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1959 IFACE_STAT(collisions, "collisions")
1961 #define IFACE_STAT(MEMBER, NAME) + 1
1962 enum { N_IFACE_STATS = IFACE_STATS };
1964 int64_t values[N_IFACE_STATS];
1965 char *keys[N_IFACE_STATS];
1968 struct netdev_stats stats;
1970 if (iface_is_synthetic(iface)) {
1974 /* Intentionally ignore return value, since errors will set 'stats' to
1975 * all-1s, and we will deal with that correctly below. */
1976 netdev_get_stats(iface->netdev, &stats);
1978 /* Copy statistics into keys[] and values[]. */
1980 #define IFACE_STAT(MEMBER, NAME) \
1981 if (stats.MEMBER != UINT64_MAX) { \
1983 values[n] = stats.MEMBER; \
1988 ovs_assert(n <= N_IFACE_STATS);
1990 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1995 br_refresh_stp_status(struct bridge *br)
1997 struct smap smap = SMAP_INITIALIZER(&smap);
1998 struct ofproto *ofproto = br->ofproto;
1999 struct ofproto_stp_status status;
2001 if (ofproto_get_stp_status(ofproto, &status)) {
2005 if (!status.enabled) {
2006 ovsrec_bridge_set_status(br->cfg, NULL);
2010 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2011 STP_ID_ARGS(status.bridge_id));
2012 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2013 STP_ID_ARGS(status.designated_root));
2014 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2016 ovsrec_bridge_set_status(br->cfg, &smap);
2017 smap_destroy(&smap);
2021 port_refresh_stp_status(struct port *port)
2023 struct ofproto *ofproto = port->bridge->ofproto;
2024 struct iface *iface;
2025 struct ofproto_port_stp_status status;
2027 int64_t int_values[3];
2030 if (port_is_synthetic(port)) {
2034 /* STP doesn't currently support bonds. */
2035 if (!list_is_singleton(&port->ifaces)) {
2036 ovsrec_port_set_status(port->cfg, NULL);
2040 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2042 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2046 if (!status.enabled) {
2047 ovsrec_port_set_status(port->cfg, NULL);
2048 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2052 /* Set Status column. */
2054 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2055 smap_add(&smap, "stp_state", stp_state_name(status.state));
2056 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2057 smap_add(&smap, "stp_role", stp_role_name(status.role));
2058 ovsrec_port_set_status(port->cfg, &smap);
2059 smap_destroy(&smap);
2061 /* Set Statistics column. */
2062 keys[0] = "stp_tx_count";
2063 int_values[0] = status.tx_count;
2064 keys[1] = "stp_rx_count";
2065 int_values[1] = status.rx_count;
2066 keys[2] = "stp_error_count";
2067 int_values[2] = status.error_count;
2069 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2070 ARRAY_SIZE(int_values));
2074 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2076 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2080 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2082 bool enable = enable_system_stats(cfg);
2084 system_stats_enable(enable);
2086 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2091 run_system_stats(void)
2093 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2096 stats = system_stats_run();
2098 struct ovsdb_idl_txn *txn;
2099 struct ovsdb_datum datum;
2101 txn = ovsdb_idl_txn_create(idl);
2102 ovsdb_datum_from_smap(&datum, stats);
2103 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2105 ovsdb_idl_txn_commit(txn);
2106 ovsdb_idl_txn_destroy(txn);
2112 static inline const char *
2113 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2116 case OFPCR12_ROLE_EQUAL:
2118 case OFPCR12_ROLE_MASTER:
2120 case OFPCR12_ROLE_SLAVE:
2122 case OFPCR12_ROLE_NOCHANGE:
2124 return "*** INVALID ROLE ***";
2129 refresh_controller_status(void)
2133 const struct ovsrec_controller *cfg;
2137 /* Accumulate status for controllers on all bridges. */
2138 HMAP_FOR_EACH (br, node, &all_bridges) {
2139 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2142 /* Update each controller in the database with current status. */
2143 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2144 struct ofproto_controller_info *cinfo =
2145 shash_find_data(&info, cfg->target);
2148 struct smap smap = SMAP_INITIALIZER(&smap);
2149 const char **values = cinfo->pairs.values;
2150 const char **keys = cinfo->pairs.keys;
2153 for (i = 0; i < cinfo->pairs.n; i++) {
2154 smap_add(&smap, keys[i], values[i]);
2157 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2158 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2160 ovsrec_controller_set_status(cfg, &smap);
2161 smap_destroy(&smap);
2163 ovsrec_controller_set_is_connected(cfg, false);
2164 ovsrec_controller_set_role(cfg, NULL);
2165 ovsrec_controller_set_status(cfg, NULL);
2169 ofproto_free_ofproto_controller_info(&info);
2174 * Some information in the database must be kept as up-to-date as possible to
2175 * allow controllers to respond rapidly to network outages. We call these
2176 * statistics "instant" stats.
2178 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2179 * assuming that they've changed. The only means we have to determine whether
2180 * they have changed are:
2182 * - Try to commit changes to the database. If nothing changed, then
2183 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2186 * - instant_stats_run() is called late in the run loop, after anything that
2187 * might change any of the instant stats.
2189 * We use these two facts together to avoid waking the process up every
2190 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2193 /* Minimum interval between writing updates to the instant stats to the
2195 #define INSTANT_INTERVAL_MSEC 100
2197 /* Current instant stats database transaction, NULL if there is no ongoing
2199 static struct ovsdb_idl_txn *instant_txn;
2201 /* Next time (in msec on monotonic clock) at which we will update the instant
2203 static long long int instant_next_txn = LLONG_MIN;
2205 /* True if the run loop has run since we last saw that the instant stats were
2206 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2207 * to refresh the instant stats. */
2208 static bool instant_stats_could_have_changed;
2211 instant_stats_run(void)
2213 enum ovsdb_idl_txn_status status;
2215 instant_stats_could_have_changed = true;
2220 if (time_msec() < instant_next_txn) {
2223 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2225 instant_txn = ovsdb_idl_txn_create(idl);
2226 HMAP_FOR_EACH (br, node, &all_bridges) {
2227 struct iface *iface;
2230 br_refresh_stp_status(br);
2232 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2233 port_refresh_stp_status(port);
2236 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2237 enum netdev_flags flags;
2239 const char *link_state;
2240 int64_t link_resets;
2243 if (iface_is_synthetic(iface)) {
2247 current = ofproto_port_is_lacp_current(br->ofproto,
2251 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2253 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2256 error = netdev_get_flags(iface->netdev, &flags);
2258 const char *state = flags & NETDEV_UP ? "up" : "down";
2259 ovsrec_interface_set_admin_state(iface->cfg, state);
2261 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2264 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2265 ovsrec_interface_set_link_state(iface->cfg, link_state);
2267 link_resets = netdev_get_carrier_resets(iface->netdev);
2268 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2270 iface_refresh_cfm_stats(iface);
2273 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2275 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2276 smap_destroy(&smap);
2281 status = ovsdb_idl_txn_commit(instant_txn);
2282 if (status != TXN_INCOMPLETE) {
2283 ovsdb_idl_txn_destroy(instant_txn);
2286 if (status == TXN_UNCHANGED) {
2287 instant_stats_could_have_changed = false;
2292 instant_stats_wait(void)
2295 ovsdb_idl_txn_wait(instant_txn);
2296 } else if (instant_stats_could_have_changed) {
2297 poll_timer_wait_until(instant_next_txn);
2301 /* Performs periodic activity required by bridges that needs to be done with
2302 * the least possible latency.
2304 * It makes sense to call this function a couple of times per poll loop, to
2305 * provide a significant performance boost on some benchmarks with ofprotos
2306 * that use the ofproto-dpif implementation. */
2308 bridge_run_fast(void)
2315 ofproto_enumerate_types(&types);
2316 SSET_FOR_EACH (type, &types) {
2317 ofproto_type_run_fast(type);
2319 sset_destroy(&types);
2321 HMAP_FOR_EACH (br, node, &all_bridges) {
2322 ofproto_run_fast(br->ofproto);
2329 static struct ovsrec_open_vswitch null_cfg;
2330 const struct ovsrec_open_vswitch *cfg;
2331 struct ovsdb_idl_txn *reconf_txn = NULL;
2335 bool vlan_splinters_changed;
2338 ovsrec_open_vswitch_init(&null_cfg);
2340 /* (Re)configure if necessary. */
2341 if (!reconfiguring) {
2344 if (ovsdb_idl_is_lock_contended(idl)) {
2345 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2346 struct bridge *br, *next_br;
2348 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2349 "disabling this process (pid %ld) until it goes away",
2350 (long int) getpid());
2352 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2355 /* Since we will not be running system_stats_run() in this process
2356 * with the current situation of multiple ovs-vswitchd daemons,
2357 * disable system stats collection. */
2358 system_stats_enable(false);
2360 } else if (!ovsdb_idl_has_lock(idl)) {
2364 cfg = ovsrec_open_vswitch_first(idl);
2366 /* Initialize the ofproto library. This only needs to run once, but
2367 * it must be done after the configuration is set. If the
2368 * initialization has already occurred, bridge_init_ofproto()
2369 * returns immediately. */
2370 bridge_init_ofproto(cfg);
2372 /* Once the value of flow-restore-wait is false, we no longer should
2373 * check its value from the database. */
2374 if (cfg && ofproto_get_flow_restore_wait()) {
2375 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2376 "flow-restore-wait", false));
2379 /* Let each datapath type do the work that it needs to do. */
2381 ofproto_enumerate_types(&types);
2382 SSET_FOR_EACH (type, &types) {
2383 ofproto_type_run(type);
2385 sset_destroy(&types);
2387 /* Let each bridge do the work that it needs to do. */
2388 HMAP_FOR_EACH (br, node, &all_bridges) {
2389 ofproto_run(br->ofproto);
2392 /* Re-configure SSL. We do this on every trip through the main loop,
2393 * instead of just when the database changes, because the contents of the
2394 * key and certificate files can change without the database changing.
2396 * We do this before bridge_reconfigure() because that function might
2397 * initiate SSL connections and thus requires SSL to be configured. */
2398 if (cfg && cfg->ssl) {
2399 const struct ovsrec_ssl *ssl = cfg->ssl;
2401 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2402 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2405 if (!reconfiguring) {
2406 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2407 * usage has changed. */
2408 vlan_splinters_changed = false;
2409 if (vlan_splinters_enabled_anywhere) {
2410 HMAP_FOR_EACH (br, node, &all_bridges) {
2411 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2412 vlan_splinters_changed = true;
2418 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2419 idl_seqno = ovsdb_idl_get_seqno(idl);
2421 reconf_txn = ovsdb_idl_txn_create(idl);
2422 bridge_reconfigure(cfg);
2424 /* We still need to reconfigure to avoid dangling pointers to
2425 * now-destroyed ovsrec structures inside bridge data. */
2426 bridge_reconfigure(&null_cfg);
2431 if (reconfiguring) {
2433 reconf_txn = ovsdb_idl_txn_create(idl);
2436 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2437 reconfiguring = false;
2440 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2443 /* If we are completing our initial configuration for this run
2444 * of ovs-vswitchd, then keep the transaction around to monitor
2445 * it for completion. */
2446 if (!initial_config_done) {
2447 initial_config_done = true;
2448 daemonize_txn = reconf_txn;
2455 ovsdb_idl_txn_commit(reconf_txn);
2456 ovsdb_idl_txn_destroy(reconf_txn);
2460 if (daemonize_txn) {
2461 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2462 if (status != TXN_INCOMPLETE) {
2463 ovsdb_idl_txn_destroy(daemonize_txn);
2464 daemonize_txn = NULL;
2466 /* ovs-vswitchd has completed initialization, so allow the
2467 * process that forked us to exit successfully. */
2468 daemonize_complete();
2470 vlog_enable_async();
2472 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2476 /* Refresh interface and mirror stats if necessary. */
2477 if (time_msec() >= iface_stats_timer) {
2479 struct ovsdb_idl_txn *txn;
2481 txn = ovsdb_idl_txn_create(idl);
2482 HMAP_FOR_EACH (br, node, &all_bridges) {
2486 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2487 struct iface *iface;
2489 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2490 iface_refresh_stats(iface);
2491 iface_refresh_status(iface);
2495 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2496 mirror_refresh_stats(m);
2500 refresh_controller_status();
2501 ovsdb_idl_txn_commit(txn);
2502 ovsdb_idl_txn_destroy(txn); /* XXX */
2505 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2509 instant_stats_run();
2518 ovsdb_idl_wait(idl);
2519 if (daemonize_txn) {
2520 ovsdb_idl_txn_wait(daemonize_txn);
2523 if (reconfiguring) {
2524 poll_immediate_wake();
2528 ofproto_enumerate_types(&types);
2529 SSET_FOR_EACH (type, &types) {
2530 ofproto_type_wait(type);
2532 sset_destroy(&types);
2534 if (!hmap_is_empty(&all_bridges)) {
2537 HMAP_FOR_EACH (br, node, &all_bridges) {
2538 ofproto_wait(br->ofproto);
2540 poll_timer_wait_until(iface_stats_timer);
2543 system_stats_wait();
2544 instant_stats_wait();
2547 /* Adds some memory usage statistics for bridges into 'usage', for use with
2548 * memory_report(). */
2550 bridge_get_memory_usage(struct simap *usage)
2554 HMAP_FOR_EACH (br, node, &all_bridges) {
2555 ofproto_get_memory_usage(br->ofproto, usage);
2559 /* QoS unixctl user interface functions. */
2561 struct qos_unixctl_show_cbdata {
2563 struct iface *iface;
2567 qos_unixctl_show_queue(unsigned int queue_id,
2568 const struct smap *details,
2569 struct iface *iface,
2572 struct netdev_queue_stats stats;
2573 struct smap_node *node;
2576 ds_put_cstr(ds, "\n");
2578 ds_put_format(ds, "Queue %u:\n", queue_id);
2580 ds_put_cstr(ds, "Default:\n");
2583 SMAP_FOR_EACH (node, details) {
2584 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2587 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2589 if (stats.tx_packets != UINT64_MAX) {
2590 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2593 if (stats.tx_bytes != UINT64_MAX) {
2594 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2597 if (stats.tx_errors != UINT64_MAX) {
2598 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2601 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2602 queue_id, ovs_strerror(error));
2607 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2608 const char *argv[], void *aux OVS_UNUSED)
2610 struct ds ds = DS_EMPTY_INITIALIZER;
2611 struct smap smap = SMAP_INITIALIZER(&smap);
2612 struct iface *iface;
2614 struct smap_node *node;
2616 iface = iface_find(argv[1]);
2618 unixctl_command_reply_error(conn, "no such interface");
2622 netdev_get_qos(iface->netdev, &type, &smap);
2624 if (*type != '\0') {
2625 struct netdev_queue_dump dump;
2626 struct smap details;
2627 unsigned int queue_id;
2629 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2631 SMAP_FOR_EACH (node, &smap) {
2632 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2635 smap_init(&details);
2636 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2637 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2639 smap_destroy(&details);
2641 unixctl_command_reply(conn, ds_cstr(&ds));
2643 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2644 unixctl_command_reply_error(conn, ds_cstr(&ds));
2647 smap_destroy(&smap);
2651 /* Bridge reconfiguration functions. */
2653 bridge_create(const struct ovsrec_bridge *br_cfg)
2657 ovs_assert(!bridge_lookup(br_cfg->name));
2658 br = xzalloc(sizeof *br);
2660 br->name = xstrdup(br_cfg->name);
2661 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2664 /* Derive the default Ethernet address from the bridge's UUID. This should
2665 * be unique and it will be stable between ovs-vswitchd runs. */
2666 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2667 eth_addr_mark_random(br->default_ea);
2669 hmap_init(&br->ports);
2670 hmap_init(&br->ifaces);
2671 hmap_init(&br->iface_by_name);
2672 hmap_init(&br->mirrors);
2674 hmap_init(&br->if_cfg_todo);
2675 list_init(&br->ofpp_garbage);
2677 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2681 bridge_destroy(struct bridge *br)
2684 struct mirror *mirror, *next_mirror;
2685 struct port *port, *next_port;
2686 struct if_cfg *if_cfg, *next_if_cfg;
2687 struct ofpp_garbage *garbage, *next_garbage;
2689 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2692 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2693 mirror_destroy(mirror);
2695 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2696 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2699 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2700 &br->ofpp_garbage) {
2701 list_remove(&garbage->list_node);
2705 hmap_remove(&all_bridges, &br->node);
2706 ofproto_destroy(br->ofproto);
2707 hmap_destroy(&br->ifaces);
2708 hmap_destroy(&br->ports);
2709 hmap_destroy(&br->iface_by_name);
2710 hmap_destroy(&br->mirrors);
2711 hmap_destroy(&br->if_cfg_todo);
2718 static struct bridge *
2719 bridge_lookup(const char *name)
2723 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2724 if (!strcmp(br->name, name)) {
2731 /* Handle requests for a listing of all flows known by the OpenFlow
2732 * stack, including those normally hidden. */
2734 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2735 const char *argv[], void *aux OVS_UNUSED)
2740 br = bridge_lookup(argv[1]);
2742 unixctl_command_reply_error(conn, "Unknown bridge");
2747 ofproto_get_all_flows(br->ofproto, &results);
2749 unixctl_command_reply(conn, ds_cstr(&results));
2750 ds_destroy(&results);
2753 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2754 * connections and reconnect. If BRIDGE is not specified, then all bridges
2755 * drop their controller connections and reconnect. */
2757 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2758 const char *argv[], void *aux OVS_UNUSED)
2762 br = bridge_lookup(argv[1]);
2764 unixctl_command_reply_error(conn, "Unknown bridge");
2767 ofproto_reconnect_controllers(br->ofproto);
2769 HMAP_FOR_EACH (br, node, &all_bridges) {
2770 ofproto_reconnect_controllers(br->ofproto);
2773 unixctl_command_reply(conn, NULL);
2777 bridge_get_controllers(const struct bridge *br,
2778 struct ovsrec_controller ***controllersp)
2780 struct ovsrec_controller **controllers;
2781 size_t n_controllers;
2783 controllers = br->cfg->controller;
2784 n_controllers = br->cfg->n_controller;
2786 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2792 *controllersp = controllers;
2794 return n_controllers;
2798 bridge_queue_if_cfg(struct bridge *br,
2799 const struct ovsrec_interface *cfg,
2800 const struct ovsrec_port *parent)
2802 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2805 if_cfg->parent = parent;
2806 if_cfg->ofport = iface_pick_ofport(cfg);
2807 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2808 hash_string(if_cfg->cfg->name, 0));
2811 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2812 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2813 * 'br' needs to complete its configuration. */
2815 bridge_add_del_ports(struct bridge *br,
2816 const unsigned long int *splinter_vlans)
2818 struct shash_node *port_node;
2819 struct port *port, *next;
2820 struct shash new_ports;
2823 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2825 /* Collect new ports. */
2826 shash_init(&new_ports);
2827 for (i = 0; i < br->cfg->n_ports; i++) {
2828 const char *name = br->cfg->ports[i]->name;
2829 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2830 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2834 if (bridge_get_controllers(br, NULL)
2835 && !shash_find(&new_ports, br->name)) {
2836 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2837 br->name, br->name);
2839 ovsrec_interface_init(&br->synth_local_iface);
2840 ovsrec_port_init(&br->synth_local_port);
2842 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2843 br->synth_local_port.n_interfaces = 1;
2844 br->synth_local_port.name = br->name;
2846 br->synth_local_iface.name = br->name;
2847 br->synth_local_iface.type = "internal";
2849 br->synth_local_ifacep = &br->synth_local_iface;
2851 shash_add(&new_ports, br->name, &br->synth_local_port);
2854 if (splinter_vlans) {
2855 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2858 /* Get rid of deleted ports.
2859 * Get rid of deleted interfaces on ports that still exist. */
2860 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2861 port->cfg = shash_find_data(&new_ports, port->name);
2865 port_del_ifaces(port);
2869 /* Update iface->cfg and iface->type in interfaces that still exist.
2870 * Add new interfaces to creation queue. */
2871 SHASH_FOR_EACH (port_node, &new_ports) {
2872 const struct ovsrec_port *port = port_node->data;
2875 for (i = 0; i < port->n_interfaces; i++) {
2876 const struct ovsrec_interface *cfg = port->interfaces[i];
2877 struct iface *iface = iface_lookup(br, cfg->name);
2878 const char *type = iface_get_type(cfg, br->cfg);
2883 } else if (!strcmp(type, "null")) {
2884 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2885 " may be removed in February 2013. Please email"
2886 " dev@openvswitch.org with concerns.",
2889 bridge_queue_if_cfg(br, cfg, port);
2894 shash_destroy(&new_ports);
2897 /* Initializes 'oc' appropriately as a management service controller for
2900 * The caller must free oc->target when it is no longer needed. */
2902 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2903 struct ofproto_controller *oc)
2905 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2906 oc->max_backoff = 0;
2907 oc->probe_interval = 60;
2908 oc->band = OFPROTO_OUT_OF_BAND;
2910 oc->burst_limit = 0;
2911 oc->enable_async_msgs = true;
2914 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2916 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2917 struct ofproto_controller *oc)
2921 oc->target = c->target;
2922 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2923 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2924 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2925 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2926 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2927 oc->burst_limit = (c->controller_burst_limit
2928 ? *c->controller_burst_limit : 0);
2929 oc->enable_async_msgs = (!c->enable_async_messages
2930 || *c->enable_async_messages);
2931 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2932 if (dscp < 0 || dscp > 63) {
2933 dscp = DSCP_DEFAULT;
2938 /* Configures the IP stack for 'br''s local interface properly according to the
2939 * configuration in 'c'. */
2941 bridge_configure_local_iface_netdev(struct bridge *br,
2942 struct ovsrec_controller *c)
2944 struct netdev *netdev;
2945 struct in_addr mask, gateway;
2947 struct iface *local_iface;
2950 /* If there's no local interface or no IP address, give up. */
2951 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2952 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2956 /* Bring up the local interface. */
2957 netdev = local_iface->netdev;
2958 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2960 /* Configure the IP address and netmask. */
2961 if (!c->local_netmask
2962 || !inet_aton(c->local_netmask, &mask)
2964 mask.s_addr = guess_netmask(ip.s_addr);
2966 if (!netdev_set_in4(netdev, ip, mask)) {
2967 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2968 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2971 /* Configure the default gateway. */
2972 if (c->local_gateway
2973 && inet_aton(c->local_gateway, &gateway)
2974 && gateway.s_addr) {
2975 if (!netdev_add_router(netdev, gateway)) {
2976 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2977 br->name, IP_ARGS(gateway.s_addr));
2982 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2983 * in either string are treated as equal to any number of slashes in the other,
2984 * e.g. "x///y" is equal to "x/y".
2986 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2987 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2988 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2989 * 'b' against a prefix of 'a'.
2992 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2994 const char *b_start = b;
2996 if (b - b_start >= b_stoplen) {
2998 } else if (*a != *b) {
3000 } else if (*a == '/') {
3001 a += strspn(a, "/");
3002 b += strspn(b, "/");
3003 } else if (*a == '\0') {
3013 bridge_configure_remotes(struct bridge *br,
3014 const struct sockaddr_in *managers, size_t n_managers)
3016 bool disable_in_band;
3018 struct ovsrec_controller **controllers;
3019 size_t n_controllers;
3021 enum ofproto_fail_mode fail_mode;
3023 struct ofproto_controller *ocs;
3027 /* Check if we should disable in-band control on this bridge. */
3028 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3031 /* Set OpenFlow queue ID for in-band control. */
3032 ofproto_set_in_band_queue(br->ofproto,
3033 smap_get_int(&br->cfg->other_config,
3034 "in-band-queue", -1));
3036 if (disable_in_band) {
3037 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3039 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3042 n_controllers = bridge_get_controllers(br, &controllers);
3044 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3047 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3048 for (i = 0; i < n_controllers; i++) {
3049 struct ovsrec_controller *c = controllers[i];
3051 if (!strncmp(c->target, "punix:", 6)
3052 || !strncmp(c->target, "unix:", 5)) {
3053 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3056 if (!strncmp(c->target, "unix:", 5)) {
3057 /* Connect to a listening socket */
3058 whitelist = xasprintf("unix:%s/", ovs_rundir());
3059 if (strchr(c->target, '/') &&
3060 !equal_pathnames(c->target, whitelist,
3061 strlen(whitelist))) {
3062 /* Absolute path specified, but not in ovs_rundir */
3063 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3064 "controller \"%s\" due to possibility for "
3065 "remote exploit. Instead, specify socket "
3066 "in whitelisted \"%s\" or connect to "
3067 "\"unix:%s/%s.mgmt\" (which is always "
3068 "available without special configuration).",
3069 br->name, c->target, whitelist,
3070 ovs_rundir(), br->name);
3075 whitelist = xasprintf("punix:%s/%s.controller",
3076 ovs_rundir(), br->name);
3077 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3078 /* Prevent remote ovsdb-server users from accessing
3079 * arbitrary Unix domain sockets and overwriting arbitrary
3081 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3082 "controller \"%s\" due to possibility of "
3083 "overwriting local files. Instead, specify "
3084 "whitelisted \"%s\" or connect to "
3085 "\"unix:%s/%s.mgmt\" (which is always "
3086 "available without special configuration).",
3087 br->name, c->target, whitelist,
3088 ovs_rundir(), br->name);
3097 bridge_configure_local_iface_netdev(br, c);
3098 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3099 if (disable_in_band) {
3100 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3105 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3106 bridge_get_allowed_versions(br));
3107 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3110 /* Set the fail-mode. */
3111 fail_mode = !br->cfg->fail_mode
3112 || !strcmp(br->cfg->fail_mode, "standalone")
3113 ? OFPROTO_FAIL_STANDALONE
3114 : OFPROTO_FAIL_SECURE;
3115 ofproto_set_fail_mode(br->ofproto, fail_mode);
3117 /* Configure OpenFlow controller connection snooping. */
3118 if (!ofproto_has_snoops(br->ofproto)) {
3122 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3123 ovs_rundir(), br->name));
3124 ofproto_set_snoops(br->ofproto, &snoops);
3125 sset_destroy(&snoops);
3130 bridge_configure_tables(struct bridge *br)
3132 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3136 n_tables = ofproto_get_n_tables(br->ofproto);
3138 for (i = 0; i < n_tables; i++) {
3139 struct ofproto_table_settings s;
3142 s.max_flows = UINT_MAX;
3146 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3147 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3150 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3151 s.max_flows = *cfg->flow_limit;
3153 if (cfg->overflow_policy
3154 && !strcmp(cfg->overflow_policy, "evict")) {
3157 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3158 for (k = 0; k < cfg->n_groups; k++) {
3159 const char *string = cfg->groups[k];
3162 msg = mf_parse_subfield__(&s.groups[k], &string);
3164 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3165 "'groups' (%s)", br->name, i, msg);
3167 } else if (*string) {
3168 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3169 "element '%s' contains trailing garbage",
3170 br->name, i, cfg->groups[k]);
3178 ofproto_configure_table(br->ofproto, i, &s);
3182 for (; j < br->cfg->n_flow_tables; j++) {
3183 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3184 "%"PRId64" not supported by this datapath", br->name,
3185 br->cfg->key_flow_tables[j]);
3190 bridge_configure_dp_desc(struct bridge *br)
3192 ofproto_set_dp_desc(br->ofproto,
3193 smap_get(&br->cfg->other_config, "dp-desc"));
3196 /* Port functions. */
3198 static struct port *
3199 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3203 port = xzalloc(sizeof *port);
3205 port->name = xstrdup(cfg->name);
3207 list_init(&port->ifaces);
3209 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3213 /* Deletes interfaces from 'port' that are no longer configured for it. */
3215 port_del_ifaces(struct port *port)
3217 struct iface *iface, *next;
3218 struct sset new_ifaces;
3221 /* Collect list of new interfaces. */
3222 sset_init(&new_ifaces);
3223 for (i = 0; i < port->cfg->n_interfaces; i++) {
3224 const char *name = port->cfg->interfaces[i]->name;
3225 const char *type = port->cfg->interfaces[i]->type;
3226 if (strcmp(type, "null")) {
3227 sset_add(&new_ifaces, name);
3231 /* Get rid of deleted interfaces. */
3232 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3233 if (!sset_contains(&new_ifaces, iface->name)) {
3234 iface_destroy(iface);
3238 sset_destroy(&new_ifaces);
3242 port_destroy(struct port *port)
3245 struct bridge *br = port->bridge;
3246 struct iface *iface, *next;
3249 ofproto_bundle_unregister(br->ofproto, port);
3252 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3253 iface_destroy(iface);
3256 hmap_remove(&br->ports, &port->hmap_node);
3262 static struct port *
3263 port_lookup(const struct bridge *br, const char *name)
3267 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3269 if (!strcmp(port->name, name)) {
3277 enable_lacp(struct port *port, bool *activep)
3279 if (!port->cfg->lacp) {
3280 /* XXX when LACP implementation has been sufficiently tested, enable by
3281 * default and make active on bonded ports. */
3283 } else if (!strcmp(port->cfg->lacp, "off")) {
3285 } else if (!strcmp(port->cfg->lacp, "active")) {
3288 } else if (!strcmp(port->cfg->lacp, "passive")) {
3292 VLOG_WARN("port %s: unknown LACP mode %s",
3293 port->name, port->cfg->lacp);
3298 static struct lacp_settings *
3299 port_configure_lacp(struct port *port, struct lacp_settings *s)
3301 const char *lacp_time, *system_id;
3304 if (!enable_lacp(port, &s->active)) {
3308 s->name = port->name;
3310 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3312 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3313 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3314 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3315 " address.", port->name, system_id);
3319 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3322 if (eth_addr_is_zero(s->id)) {
3323 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3327 /* Prefer bondable links if unspecified. */
3328 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3330 s->priority = (priority > 0 && priority <= UINT16_MAX
3332 : UINT16_MAX - !list_is_short(&port->ifaces));
3334 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3335 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3340 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3342 int priority, portid, key;
3344 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3345 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3347 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3349 if (portid <= 0 || portid > UINT16_MAX) {
3350 portid = ofp_to_u16(iface->ofp_port);
3353 if (priority <= 0 || priority > UINT16_MAX) {
3354 priority = UINT16_MAX;
3357 if (key < 0 || key > UINT16_MAX) {
3361 s->name = iface->name;
3363 s->priority = priority;
3368 port_configure_bond(struct port *port, struct bond_settings *s)
3370 const char *detect_s;
3371 struct iface *iface;
3372 int miimon_interval;
3374 s->name = port->name;
3376 if (port->cfg->bond_mode) {
3377 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3378 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3379 port->name, port->cfg->bond_mode,
3380 bond_mode_to_string(s->balance));
3383 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3385 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3386 * active-backup. At some point we should remove this warning. */
3387 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3388 " in previous versions, the default bond_mode was"
3389 " balance-slb", port->name,
3390 bond_mode_to_string(s->balance));
3392 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3393 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3394 "please use another bond type or disable flood_vlans",
3398 miimon_interval = smap_get_int(&port->cfg->other_config,
3399 "bond-miimon-interval", 0);
3400 if (miimon_interval <= 0) {
3401 miimon_interval = 200;
3404 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3405 if (!detect_s || !strcmp(detect_s, "carrier")) {
3406 miimon_interval = 0;
3407 } else if (strcmp(detect_s, "miimon")) {
3408 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3409 "defaulting to carrier", port->name, detect_s);
3410 miimon_interval = 0;
3413 s->up_delay = MAX(0, port->cfg->bond_updelay);
3414 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3415 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3416 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3417 "bond-rebalance-interval", 10000);
3418 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3419 s->rebalance_interval = 1000;
3422 s->fake_iface = port->cfg->bond_fake_iface;
3424 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3425 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3429 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3430 * instead of obtaining it from the database. */
3432 port_is_synthetic(const struct port *port)
3434 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3437 /* Interface functions. */
3440 iface_is_internal(const struct ovsrec_interface *iface,
3441 const struct ovsrec_bridge *br)
3443 /* The local port and "internal" ports are always "internal". */
3444 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3447 /* Returns the correct network device type for interface 'iface' in bridge
3450 iface_get_type(const struct ovsrec_interface *iface,
3451 const struct ovsrec_bridge *br)
3455 /* The local port always has type "internal". Other ports take
3456 * their type from the database and default to "system" if none is
3458 if (iface_is_internal(iface, br)) {
3461 type = iface->type[0] ? iface->type : "system";
3464 return ofproto_port_open_type(br->datapath_type, type);
3468 iface_destroy(struct iface *iface)
3471 struct port *port = iface->port;
3472 struct bridge *br = port->bridge;
3474 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3475 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3478 if (iface->ofp_port != OFPP_NONE) {
3479 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3482 list_remove(&iface->port_elem);
3483 hmap_remove(&br->iface_by_name, &iface->name_node);
3485 netdev_close(iface->netdev);
3492 static struct iface *
3493 iface_lookup(const struct bridge *br, const char *name)
3495 struct iface *iface;
3497 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3498 &br->iface_by_name) {
3499 if (!strcmp(iface->name, name)) {
3507 static struct iface *
3508 iface_find(const char *name)
3510 const struct bridge *br;
3512 HMAP_FOR_EACH (br, node, &all_bridges) {
3513 struct iface *iface = iface_lookup(br, name);
3522 static struct if_cfg *
3523 if_cfg_lookup(const struct bridge *br, const char *name)
3525 struct if_cfg *if_cfg;
3527 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3529 if (!strcmp(if_cfg->cfg->name, name)) {
3537 static struct iface *
3538 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3540 struct iface *iface;
3542 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3544 if (iface->ofp_port == ofp_port) {
3551 /* Set Ethernet address of 'iface', if one is specified in the configuration
3554 iface_set_mac(struct iface *iface)
3556 uint8_t ea[ETH_ADDR_LEN];
3558 if (!strcmp(iface->type, "internal")
3559 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3560 if (iface->ofp_port == OFPP_LOCAL) {
3561 VLOG_ERR("interface %s: ignoring mac in Interface record "
3562 "(use Bridge record to set local port's mac)",
3564 } else if (eth_addr_is_multicast(ea)) {
3565 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3568 int error = netdev_set_etheraddr(iface->netdev, ea);
3570 VLOG_ERR("interface %s: setting MAC failed (%s)",
3571 iface->name, ovs_strerror(error));
3577 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3579 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3582 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3583 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3584 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3588 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3589 * sets the "ofport" field to -1.
3591 * This is appropriate when 'if_cfg''s interface cannot be created or is
3592 * otherwise invalid. */
3594 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3596 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3597 ovsrec_interface_set_status(if_cfg, NULL);
3598 ovsrec_interface_set_admin_state(if_cfg, NULL);
3599 ovsrec_interface_set_duplex(if_cfg, NULL);
3600 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3601 ovsrec_interface_set_link_state(if_cfg, NULL);
3602 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3603 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3604 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3605 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3606 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3607 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3608 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3609 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3614 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3616 union ovsdb_atom atom;
3618 atom.integer = target;
3619 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3623 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3625 struct ofpbuf queues_buf;
3627 ofpbuf_init(&queues_buf, 0);
3629 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3630 netdev_set_qos(iface->netdev, NULL, NULL);
3632 const struct ovsdb_datum *queues;
3633 struct netdev_queue_dump dump;
3634 unsigned int queue_id;
3635 struct smap details;
3639 /* Configure top-level Qos for 'iface'. */
3640 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3642 /* Deconfigure queues that were deleted. */
3643 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3645 smap_init(&details);
3646 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3647 if (!queue_ids_include(queues, queue_id)) {
3648 netdev_delete_queue(iface->netdev, queue_id);
3651 smap_destroy(&details);
3653 /* Configure queues for 'iface'. */
3655 for (i = 0; i < qos->n_queues; i++) {
3656 const struct ovsrec_queue *queue = qos->value_queues[i];
3657 unsigned int queue_id = qos->key_queues[i];
3659 if (queue_id == 0) {
3663 if (queue->n_dscp == 1) {
3664 struct ofproto_port_queue *port_queue;
3666 port_queue = ofpbuf_put_uninit(&queues_buf,
3667 sizeof *port_queue);
3668 port_queue->queue = queue_id;
3669 port_queue->dscp = queue->dscp[0];
3672 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3675 struct smap details;
3677 smap_init(&details);
3678 netdev_set_queue(iface->netdev, 0, &details);
3679 smap_destroy(&details);
3683 if (iface->ofp_port != OFPP_NONE) {
3684 const struct ofproto_port_queue *port_queues = queues_buf.data;
3685 size_t n_queues = queues_buf.size / sizeof *port_queues;
3687 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3688 port_queues, n_queues);
3691 netdev_set_policing(iface->netdev,
3692 iface->cfg->ingress_policing_rate,
3693 iface->cfg->ingress_policing_burst);
3695 ofpbuf_uninit(&queues_buf);
3699 iface_configure_cfm(struct iface *iface)
3701 const struct ovsrec_interface *cfg = iface->cfg;
3702 const char *opstate_str;
3703 const char *cfm_ccm_vlan;
3704 struct cfm_settings s;
3705 struct smap netdev_args;
3707 if (!cfg->n_cfm_mpid) {
3708 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3712 s.check_tnl_key = false;
3713 smap_init(&netdev_args);
3714 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3715 const char *key = smap_get(&netdev_args, "key");
3716 const char *in_key = smap_get(&netdev_args, "in_key");
3718 s.check_tnl_key = (key && !strcmp(key, "flow"))
3719 || (in_key && !strcmp(in_key, "flow"));
3721 smap_destroy(&netdev_args);
3723 s.mpid = *cfg->cfm_mpid;
3724 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3725 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3726 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3728 if (s.interval <= 0) {
3732 if (!cfm_ccm_vlan) {
3734 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3735 s.ccm_vlan = CFM_RANDOM_VLAN;
3737 s.ccm_vlan = atoi(cfm_ccm_vlan);
3738 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3743 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3745 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3747 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3748 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3750 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3753 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3754 * instead of obtaining it from the database. */
3756 iface_is_synthetic(const struct iface *iface)
3758 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3762 iface_pick_ofport(const struct ovsrec_interface *cfg)
3764 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3766 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3771 /* Port mirroring. */
3773 static struct mirror *
3774 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3778 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3779 if (uuid_equals(uuid, &m->uuid)) {
3787 bridge_configure_mirrors(struct bridge *br)
3789 const struct ovsdb_datum *mc;
3790 unsigned long *flood_vlans;
3791 struct mirror *m, *next;
3794 /* Get rid of deleted mirrors. */
3795 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3796 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3797 union ovsdb_atom atom;
3799 atom.uuid = m->uuid;
3800 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3805 /* Add new mirrors and reconfigure existing ones. */
3806 for (i = 0; i < br->cfg->n_mirrors; i++) {
3807 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3808 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3810 m = mirror_create(br, cfg);
3813 if (!mirror_configure(m)) {
3818 /* Update flooded vlans (for RSPAN). */
3819 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3820 br->cfg->n_flood_vlans);
3821 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3822 bitmap_free(flood_vlans);
3825 static struct mirror *
3826 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3830 m = xzalloc(sizeof *m);
3831 m->uuid = cfg->header_.uuid;
3832 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3834 m->name = xstrdup(cfg->name);
3840 mirror_destroy(struct mirror *m)
3843 struct bridge *br = m->bridge;
3846 ofproto_mirror_unregister(br->ofproto, m);
3849 hmap_remove(&br->mirrors, &m->hmap_node);
3856 mirror_collect_ports(struct mirror *m,
3857 struct ovsrec_port **in_ports, int n_in_ports,
3858 void ***out_portsp, size_t *n_out_portsp)
3860 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3861 size_t n_out_ports = 0;
3864 for (i = 0; i < n_in_ports; i++) {
3865 const char *name = in_ports[i]->name;
3866 struct port *port = port_lookup(m->bridge, name);
3868 out_ports[n_out_ports++] = port;
3870 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3871 "port %s", m->bridge->name, m->name, name);
3874 *out_portsp = out_ports;
3875 *n_out_portsp = n_out_ports;
3879 mirror_configure(struct mirror *m)
3881 const struct ovsrec_mirror *cfg = m->cfg;
3882 struct ofproto_mirror_settings s;
3885 if (strcmp(cfg->name, m->name)) {
3887 m->name = xstrdup(cfg->name);
3891 /* Get output port or VLAN. */
3892 if (cfg->output_port) {
3893 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3894 if (!s.out_bundle) {
3895 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3896 m->bridge->name, m->name);
3899 s.out_vlan = UINT16_MAX;
3901 if (cfg->output_vlan) {
3902 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3903 "output vlan; ignoring output vlan",
3904 m->bridge->name, m->name);
3906 } else if (cfg->output_vlan) {
3907 /* The database should prevent invalid VLAN values. */
3908 s.out_bundle = NULL;
3909 s.out_vlan = *cfg->output_vlan;
3911 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3912 m->bridge->name, m->name);
3916 /* Get port selection. */
3917 if (cfg->select_all) {
3918 size_t n_ports = hmap_count(&m->bridge->ports);
3919 void **ports = xmalloc(n_ports * sizeof *ports);
3924 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3934 /* Get ports, dropping ports that don't exist.
3935 * The IDL ensures that there are no duplicates. */
3936 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3937 &s.srcs, &s.n_srcs);
3938 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3939 &s.dsts, &s.n_dsts);
3942 /* Get VLAN selection. */
3943 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3946 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3949 if (s.srcs != s.dsts) {
3958 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3960 * This is deprecated. It is only for compatibility with broken device drivers
3961 * in old versions of Linux that do not properly support VLANs when VLAN
3962 * devices are not used. When broken device drivers are no longer in
3963 * widespread use, we will delete these interfaces. */
3965 static struct ovsrec_port **recs;
3966 static size_t n_recs, allocated_recs;
3968 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3969 * splinters are reconfigured. */
3971 register_rec(struct ovsrec_port *rec)
3973 if (n_recs >= allocated_recs) {
3974 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3976 recs[n_recs++] = rec;
3979 /* Frees all of the ports registered with register_reg(). */
3981 free_registered_recs(void)
3985 for (i = 0; i < n_recs; i++) {
3986 struct ovsrec_port *port = recs[i];
3989 for (j = 0; j < port->n_interfaces; j++) {
3990 struct ovsrec_interface *iface = port->interfaces[j];
3995 smap_destroy(&port->other_config);
3996 free(port->interfaces);
4004 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4007 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4009 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4013 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4016 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4017 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4018 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4021 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4022 * use, returns NULL.
4024 * Updates 'vlan_splinters_enabled_anywhere'. */
4025 static unsigned long int *
4026 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4028 unsigned long int *splinter_vlans;
4029 struct sset splinter_ifaces;
4030 const char *real_dev_name;
4031 struct shash *real_devs;
4032 struct shash_node *node;
4036 /* Free space allocated for synthesized ports and interfaces, since we're
4037 * in the process of reconstructing all of them. */
4038 free_registered_recs();
4040 splinter_vlans = bitmap_allocate(4096);
4041 sset_init(&splinter_ifaces);
4042 vlan_splinters_enabled_anywhere = false;
4043 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4044 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4047 for (j = 0; j < br_cfg->n_ports; j++) {
4048 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4051 for (k = 0; k < port_cfg->n_interfaces; k++) {
4052 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4054 if (vlan_splinters_is_enabled(iface_cfg)) {
4055 vlan_splinters_enabled_anywhere = true;
4056 sset_add(&splinter_ifaces, iface_cfg->name);
4057 vlan_bitmap_from_array__(port_cfg->trunks,
4063 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4064 bitmap_set1(splinter_vlans, *port_cfg->tag);
4069 if (!vlan_splinters_enabled_anywhere) {
4070 free(splinter_vlans);
4071 sset_destroy(&splinter_ifaces);
4075 HMAP_FOR_EACH (br, node, &all_bridges) {
4077 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4081 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4082 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4083 * device to be created for it. */
4084 bitmap_set0(splinter_vlans, 0);
4085 bitmap_set0(splinter_vlans, 4095);
4087 /* Delete all VLAN devices that we don't need. */
4089 real_devs = vlandev_get_real_devs();
4090 SHASH_FOR_EACH (node, real_devs) {
4091 const struct vlan_real_dev *real_dev = node->data;
4092 const struct vlan_dev *vlan_dev;
4093 bool real_dev_has_splinters;
4095 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4097 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4098 if (!real_dev_has_splinters
4099 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4100 struct netdev *netdev;
4102 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4103 if (!netdev_get_in4(netdev, NULL, NULL) ||
4104 !netdev_get_in6(netdev, NULL)) {
4105 /* It has an IP address configured, so we don't own
4106 * it. Don't delete it. */
4108 vlandev_del(vlan_dev->name);
4110 netdev_close(netdev);
4117 /* Add all VLAN devices that we need. */
4118 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4121 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4122 if (!vlandev_get_name(real_dev_name, vid)) {
4123 vlandev_add(real_dev_name, vid);
4130 sset_destroy(&splinter_ifaces);
4132 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4133 free(splinter_vlans);
4136 return splinter_vlans;
4139 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4142 configure_splinter_port(struct port *port)
4144 struct ofproto *ofproto = port->bridge->ofproto;
4145 ofp_port_t realdev_ofp_port;
4146 const char *realdev_name;
4147 struct iface *vlandev, *realdev;
4149 ofproto_bundle_unregister(port->bridge->ofproto, port);
4151 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4154 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4155 realdev = iface_lookup(port->bridge, realdev_name);
4156 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4158 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4162 static struct ovsrec_port *
4163 synthesize_splinter_port(const char *real_dev_name,
4164 const char *vlan_dev_name, int vid)
4166 struct ovsrec_interface *iface;
4167 struct ovsrec_port *port;
4169 iface = xmalloc(sizeof *iface);
4170 ovsrec_interface_init(iface);
4171 iface->name = xstrdup(vlan_dev_name);
4172 iface->type = "system";
4174 port = xmalloc(sizeof *port);
4175 ovsrec_port_init(port);
4176 port->interfaces = xmemdup(&iface, sizeof iface);
4177 port->n_interfaces = 1;
4178 port->name = xstrdup(vlan_dev_name);
4179 port->vlan_mode = "splinter";
4180 port->tag = xmalloc(sizeof *port->tag);
4183 smap_add(&port->other_config, "realdev", real_dev_name);
4189 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4190 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4191 * 1-bit in the 'splinter_vlans' bitmap. */
4193 add_vlan_splinter_ports(struct bridge *br,
4194 const unsigned long int *splinter_vlans,
4195 struct shash *ports)
4199 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4200 * we're modifying 'ports'. */
4201 for (i = 0; i < br->cfg->n_ports; i++) {
4202 const char *name = br->cfg->ports[i]->name;
4203 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4206 for (j = 0; j < port_cfg->n_interfaces; j++) {
4207 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4209 if (vlan_splinters_is_enabled(iface_cfg)) {
4210 const char *real_dev_name;
4213 real_dev_name = iface_cfg->name;
4214 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4215 const char *vlan_dev_name;
4217 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4219 && !shash_find(ports, vlan_dev_name)) {
4220 shash_add(ports, vlan_dev_name,
4221 synthesize_splinter_port(
4222 real_dev_name, vlan_dev_name, vid));
4231 mirror_refresh_stats(struct mirror *m)
4233 struct ofproto *ofproto = m->bridge->ofproto;
4234 uint64_t tx_packets, tx_bytes;
4237 size_t stat_cnt = 0;
4239 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4240 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4244 if (tx_packets != UINT64_MAX) {
4245 keys[stat_cnt] = "tx_packets";
4246 values[stat_cnt] = tx_packets;
4249 if (tx_bytes != UINT64_MAX) {
4250 keys[stat_cnt] = "tx_bytes";
4251 values[stat_cnt] = tx_bytes;
4255 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);