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.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
41 #include "ofproto/ofproto.h"
42 #include "poll-loop.h"
46 #include "socket-util.h"
48 #include "stream-ssl.h"
50 #include "system-stats.h"
55 #include "lib/vswitch-idl.h"
56 #include "xenserver.h"
58 #include "sflow_api.h"
59 #include "vlan-bitmap.h"
61 VLOG_DEFINE_THIS_MODULE(bridge);
63 COVERAGE_DEFINE(bridge_reconfigure);
65 /* Configuration of an uninstantiated iface. */
67 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
68 const struct ovsrec_interface *cfg; /* Interface record. */
69 const struct ovsrec_port *parent; /* Parent port record. */
70 ofp_port_t ofport; /* Requested OpenFlow port number. */
73 /* OpenFlow port slated for removal from ofproto. */
75 struct list list_node; /* Node in bridge's ofpp_garbage. */
76 ofp_port_t ofp_port; /* Port to be deleted. */
80 /* These members are always valid. */
81 struct list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
86 /* These members are valid only after bridge_reconfigure() causes them to
88 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
89 ofp_port_t ofp_port; /* OpenFlow port number, */
90 /* OFPP_NONE if unknown. */
91 struct netdev *netdev; /* Network device. */
92 const char *type; /* Usually same as cfg->type. */
93 const struct ovsrec_interface *cfg;
97 struct uuid uuid; /* UUID of this "mirror" record in database. */
98 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
99 struct bridge *bridge;
101 const struct ovsrec_mirror *cfg;
105 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
106 struct bridge *bridge;
109 const struct ovsrec_port *cfg;
111 /* An ordinary bridge port has 1 interface.
112 * A bridge port for bonding has at least 2 interfaces. */
113 struct list ifaces; /* List of "struct iface"s. */
117 struct hmap_node node; /* In 'all_bridges'. */
118 char *name; /* User-specified arbitrary name. */
119 char *type; /* Datapath type. */
120 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
121 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
122 const struct ovsrec_bridge *cfg;
124 /* OpenFlow switch processing. */
125 struct ofproto *ofproto; /* OpenFlow switch. */
128 struct hmap ports; /* "struct port"s indexed by name. */
129 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
130 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
132 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
133 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
134 Indexed on 'cfg->name'. */
136 /* Port mirroring. */
137 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
145 /* All bridges, indexed by name. */
146 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
148 /* OVSDB IDL used to obtain configuration. */
149 static struct ovsdb_idl *idl;
151 /* We want to complete daemonization, fully detaching from our parent process,
152 * only after we have completed our initial configuration, committed our state
153 * to the database, and received confirmation back from the database server
154 * that it applied the commit. This allows our parent process to know that,
155 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
156 * to have already been filled in. (It is only "very likely" rather than
157 * certain because there is always a slim possibility that the transaction will
158 * fail or that some other client has added new bridges, ports, etc. while
159 * ovs-vswitchd was configuring using an old configuration.)
161 * We only need to do this once for our initial configuration at startup, so
162 * 'initial_config_done' tracks whether we've already done it. While we are
163 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
164 * itself and is otherwise NULL. */
165 static bool initial_config_done;
166 static struct ovsdb_idl_txn *daemonize_txn;
168 /* Most recently processed IDL sequence number. */
169 static unsigned int idl_seqno;
171 /* Each time this timer expires, the bridge fetches interface and mirror
172 * statistics and pushes them into the database. */
173 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
174 static long long int iface_stats_timer = LLONG_MIN;
176 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
177 * expensive. This can cause bridge_reconfigure() to take a long time during
178 * which no other work can be done. To deal with this problem, we limit port
179 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
180 * call to bridge_reconfigure(). If there is more work to do after the limit
181 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
182 * This allows the rest of the code to catch up on important things like
183 * forwarding packets. */
184 #define OFP_PORT_ACTION_WINDOW 10
185 static bool reconfiguring = false;
187 static void add_del_bridges(const struct ovsrec_open_vswitch *);
188 static void bridge_update_ofprotos(void);
189 static void bridge_create(const struct ovsrec_bridge *);
190 static void bridge_destroy(struct bridge *);
191 static struct bridge *bridge_lookup(const char *name);
192 static unixctl_cb_func bridge_unixctl_dump_flows;
193 static unixctl_cb_func bridge_unixctl_reconnect;
194 static size_t bridge_get_controllers(const struct bridge *br,
195 struct ovsrec_controller ***controllersp);
196 static void bridge_add_del_ports(struct bridge *,
197 const unsigned long int *splinter_vlans);
198 static void bridge_refresh_ofp_port(struct bridge *);
199 static void bridge_configure_flow_miss_model(const char *opt);
200 static void bridge_configure_datapath_id(struct bridge *);
201 static void bridge_configure_netflow(struct bridge *);
202 static void bridge_configure_forward_bpdu(struct bridge *);
203 static void bridge_configure_mac_table(struct bridge *);
204 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
205 static void bridge_configure_ipfix(struct bridge *);
206 static void bridge_configure_stp(struct bridge *);
207 static void bridge_configure_tables(struct bridge *);
208 static void bridge_configure_dp_desc(struct bridge *);
209 static void bridge_configure_remotes(struct bridge *,
210 const struct sockaddr_in *managers,
212 static void bridge_pick_local_hw_addr(struct bridge *,
213 uint8_t ea[ETH_ADDR_LEN],
214 struct iface **hw_addr_iface);
215 static uint64_t bridge_pick_datapath_id(struct bridge *,
216 const uint8_t bridge_ea[ETH_ADDR_LEN],
217 struct iface *hw_addr_iface);
218 static void bridge_queue_if_cfg(struct bridge *,
219 const struct ovsrec_interface *,
220 const struct ovsrec_port *);
221 static uint64_t dpid_from_hash(const void *, size_t nbytes);
222 static bool bridge_has_bond_fake_iface(const struct bridge *,
224 static bool port_is_bond_fake_iface(const struct port *);
226 static unixctl_cb_func qos_unixctl_show;
228 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
229 static void port_del_ifaces(struct port *);
230 static void port_destroy(struct port *);
231 static struct port *port_lookup(const struct bridge *, const char *name);
232 static void port_configure(struct port *);
233 static struct lacp_settings *port_configure_lacp(struct port *,
234 struct lacp_settings *);
235 static void port_configure_bond(struct port *, struct bond_settings *);
236 static bool port_is_synthetic(const struct port *);
238 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
239 static void run_system_stats(void);
241 static void bridge_configure_mirrors(struct bridge *);
242 static struct mirror *mirror_create(struct bridge *,
243 const struct ovsrec_mirror *);
244 static void mirror_destroy(struct mirror *);
245 static bool mirror_configure(struct mirror *);
246 static void mirror_refresh_stats(struct mirror *);
248 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
249 static bool iface_create(struct bridge *, struct if_cfg *,
250 ofp_port_t ofp_port);
251 static bool iface_is_internal(const struct ovsrec_interface *iface,
252 const struct ovsrec_bridge *br);
253 static const char *iface_get_type(const struct ovsrec_interface *,
254 const struct ovsrec_bridge *);
255 static void iface_destroy(struct iface *);
256 static struct iface *iface_lookup(const struct bridge *, const char *name);
257 static struct iface *iface_find(const char *name);
258 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
259 static struct iface *iface_from_ofp_port(const struct bridge *,
260 ofp_port_t ofp_port);
261 static void iface_set_mac(struct iface *);
262 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
263 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
264 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
265 static void iface_configure_cfm(struct iface *);
266 static void iface_refresh_cfm_stats(struct iface *);
267 static void iface_refresh_stats(struct iface *);
268 static void iface_refresh_status(struct iface *);
269 static bool iface_is_synthetic(const struct iface *);
270 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
272 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
274 * This is deprecated. It is only for compatibility with broken device drivers
275 * in old versions of Linux that do not properly support VLANs when VLAN
276 * devices are not used. When broken device drivers are no longer in
277 * widespread use, we will delete these interfaces. */
279 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
280 static bool vlan_splinters_enabled_anywhere;
282 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
283 static unsigned long int *collect_splinter_vlans(
284 const struct ovsrec_open_vswitch *);
285 static void configure_splinter_port(struct port *);
286 static void add_vlan_splinter_ports(struct bridge *,
287 const unsigned long int *splinter_vlans,
288 struct shash *ports);
291 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
293 struct shash iface_hints;
294 static bool initialized = false;
301 shash_init(&iface_hints);
304 for (i = 0; i < cfg->n_bridges; i++) {
305 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
308 for (j = 0; j < br_cfg->n_ports; j++) {
309 struct ovsrec_port *port_cfg = br_cfg->ports[j];
312 for (k = 0; k < port_cfg->n_interfaces; k++) {
313 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
314 struct iface_hint *iface_hint;
316 iface_hint = xmalloc(sizeof *iface_hint);
317 iface_hint->br_name = br_cfg->name;
318 iface_hint->br_type = br_cfg->datapath_type;
319 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
321 shash_add(&iface_hints, if_cfg->name, iface_hint);
327 ofproto_init(&iface_hints);
329 shash_destroy_free_data(&iface_hints);
333 /* Public functions. */
335 /* Initializes the bridge module, configuring it to obtain its configuration
336 * from an OVSDB server accessed over 'remote', which should be a string in a
337 * form acceptable to ovsdb_idl_create(). */
339 bridge_init(const char *remote)
341 /* Create connection to database. */
342 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
343 idl_seqno = ovsdb_idl_get_seqno(idl);
344 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
345 ovsdb_idl_verify_write_only(idl);
347 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
355 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
357 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
359 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
361 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
362 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
382 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
384 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
385 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
387 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
394 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
396 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
407 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
409 /* Register unixctl commands. */
410 unixctl_command_register("qos/show", "interface", 1, 1,
411 qos_unixctl_show, NULL);
412 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
413 bridge_unixctl_dump_flows, NULL);
414 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
415 bridge_unixctl_reconnect, NULL);
425 struct bridge *br, *next_br;
427 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
430 ovsdb_idl_destroy(idl);
433 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
434 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
435 * responsible for freeing '*managersp' (with free()).
437 * You may be asking yourself "why does ovs-vswitchd care?", because
438 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
439 * should not be and in fact is not directly involved in that. But
440 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
441 * it has to tell in-band control where the managers are to enable that.
442 * (Thus, only managers connected in-band are collected.)
445 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
446 struct sockaddr_in **managersp, size_t *n_managersp)
448 struct sockaddr_in *managers = NULL;
449 size_t n_managers = 0;
453 /* Collect all of the potential targets from the "targets" columns of the
454 * rows pointed to by "manager_options", excluding any that are
457 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
458 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
460 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
461 sset_find_and_delete(&targets, m->target);
463 sset_add(&targets, m->target);
467 /* Now extract the targets' IP addresses. */
468 if (!sset_is_empty(&targets)) {
471 managers = xmalloc(sset_count(&targets) * sizeof *managers);
472 SSET_FOR_EACH (target, &targets) {
473 struct sockaddr_in *sin = &managers[n_managers];
475 if (stream_parse_target_with_default_ports(target,
483 sset_destroy(&targets);
485 *managersp = managers;
486 *n_managersp = n_managers;
490 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
492 unsigned long int *splinter_vlans;
495 COVERAGE_INC(bridge_reconfigure);
497 ovs_assert(!reconfiguring);
498 reconfiguring = true;
500 ofproto_set_flow_eviction_threshold(
501 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
502 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
504 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
505 "force-miss-model"));
507 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
508 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
509 * configuration otherwise.
511 * This is mostly an update to bridge data structures. Nothing is pushed
512 * down to ofproto or lower layers. */
513 add_del_bridges(ovs_cfg);
514 splinter_vlans = collect_splinter_vlans(ovs_cfg);
515 HMAP_FOR_EACH (br, node, &all_bridges) {
516 bridge_add_del_ports(br, splinter_vlans);
518 free(splinter_vlans);
520 /* Delete datapaths that are no longer configured, and create ones which
521 * don't exist but should. */
522 bridge_update_ofprotos();
524 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
525 HMAP_FOR_EACH (br, node, &all_bridges) {
526 bridge_refresh_ofp_port(br);
529 /* Clear database records for "if_cfg"s which haven't been instantiated. */
530 HMAP_FOR_EACH (br, node, &all_bridges) {
531 struct if_cfg *if_cfg;
533 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
534 iface_clear_db_record(if_cfg->cfg);
538 reconfigure_system_stats(ovs_cfg);
542 bridge_reconfigure_ofp(void)
544 long long int deadline;
548 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
550 /* The kernel will reject any attempt to add a given port to a datapath if
551 * that port already belongs to a different datapath, so we must do all
552 * port deletions before any port additions. */
553 HMAP_FOR_EACH (br, node, &all_bridges) {
554 struct ofpp_garbage *garbage, *next;
556 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
557 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
558 * internal datastructures may not be consistent. Eventually, when
559 * port additions and deletions are cheaper, these calls should be
562 ofproto_port_del(br->ofproto, garbage->ofp_port);
563 list_remove(&garbage->list_node);
567 if (time_msec() >= deadline) {
574 HMAP_FOR_EACH (br, node, &all_bridges) {
575 struct if_cfg *if_cfg, *next;
577 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
578 iface_create(br, if_cfg, OFPP_NONE);
580 if (time_msec() >= deadline) {
590 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
592 struct sockaddr_in *managers;
593 int sflow_bridge_number;
598 ovs_assert(reconfiguring);
599 done = bridge_reconfigure_ofp();
601 /* Complete the configuration. */
602 sflow_bridge_number = 0;
603 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
604 HMAP_FOR_EACH (br, node, &all_bridges) {
607 /* We need the datapath ID early to allow LACP ports to use it as the
608 * default system ID. */
609 bridge_configure_datapath_id(br);
611 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
614 port_configure(port);
616 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
617 iface_configure_cfm(iface);
618 iface_configure_qos(iface, port->cfg->qos);
619 iface_set_mac(iface);
620 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
624 bridge_configure_mirrors(br);
625 bridge_configure_forward_bpdu(br);
626 bridge_configure_mac_table(br);
627 bridge_configure_remotes(br, managers, n_managers);
628 bridge_configure_netflow(br);
629 bridge_configure_sflow(br, &sflow_bridge_number);
630 bridge_configure_ipfix(br);
631 bridge_configure_stp(br);
632 bridge_configure_tables(br);
633 bridge_configure_dp_desc(br);
635 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
636 /* XXX: Remove this warning message eventually. */
637 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
638 " moved to the Open_vSwitch table. Ignoring its"
639 " setting in the bridge table.");
647 /* Delete ofprotos which aren't configured or have the wrong type. Create
648 * ofprotos which don't exist but need to. */
650 bridge_update_ofprotos(void)
652 struct bridge *br, *next;
657 /* Delete ofprotos with no bridge or with the wrong type. */
660 ofproto_enumerate_types(&types);
661 SSET_FOR_EACH (type, &types) {
664 ofproto_enumerate_names(type, &names);
665 SSET_FOR_EACH (name, &names) {
666 br = bridge_lookup(name);
667 if (!br || strcmp(type, br->type)) {
668 ofproto_delete(name, type);
672 sset_destroy(&names);
673 sset_destroy(&types);
675 /* Add ofprotos for bridges which don't have one yet. */
676 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
684 /* Remove ports from any datapath with the same name as 'br'. If we
685 * don't do this, creating 'br''s ofproto will fail because a port with
686 * the same name as its local port already exists. */
687 HMAP_FOR_EACH (br2, node, &all_bridges) {
688 struct ofproto_port ofproto_port;
694 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
696 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
698 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
699 ovs_strerror(error));
701 ofproto_port_destroy(&ofproto_port);
705 error = ofproto_create(br->name, br->type, &br->ofproto);
707 VLOG_ERR("failed to create bridge %s: %s", br->name,
708 ovs_strerror(error));
715 port_configure(struct port *port)
717 const struct ovsrec_port *cfg = port->cfg;
718 struct bond_settings bond_settings;
719 struct lacp_settings lacp_settings;
720 struct ofproto_bundle_settings s;
723 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
724 configure_splinter_port(port);
733 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
734 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
735 s.slaves[s.n_slaves++] = iface->ofp_port;
740 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
744 /* Get VLAN trunks. */
747 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
751 if (cfg->vlan_mode) {
752 if (!strcmp(cfg->vlan_mode, "access")) {
753 s.vlan_mode = PORT_VLAN_ACCESS;
754 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
755 s.vlan_mode = PORT_VLAN_TRUNK;
756 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
757 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
758 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
759 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
761 /* This "can't happen" because ovsdb-server should prevent it. */
762 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
763 s.vlan_mode = PORT_VLAN_TRUNK;
767 s.vlan_mode = PORT_VLAN_ACCESS;
769 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
773 s.vlan_mode = PORT_VLAN_TRUNK;
776 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
779 /* Get LACP settings. */
780 s.lacp = port_configure_lacp(port, &lacp_settings);
784 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
785 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
786 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
789 s.lacp_slaves = NULL;
792 /* Get bond settings. */
793 if (s.n_slaves > 1) {
794 s.bond = &bond_settings;
795 port_configure_bond(port, &bond_settings);
798 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
799 netdev_set_miimon_interval(iface->netdev, 0);
804 ofproto_bundle_register(port->bridge->ofproto, port, &s);
813 bridge_configure_flow_miss_model(const char *opt)
815 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
818 if (strcmp(opt, "with-facets")) {
819 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
820 } else if (strcmp(opt, "without-facets")) {
821 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
825 ofproto_set_flow_miss_model(model);
828 /* Pick local port hardware address and datapath ID for 'br'. */
830 bridge_configure_datapath_id(struct bridge *br)
832 uint8_t ea[ETH_ADDR_LEN];
834 struct iface *local_iface;
835 struct iface *hw_addr_iface;
838 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
839 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
841 int error = netdev_set_etheraddr(local_iface->netdev, ea);
843 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
844 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
845 "Ethernet address: %s",
846 br->name, ovs_strerror(error));
849 memcpy(br->ea, ea, ETH_ADDR_LEN);
851 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
852 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
853 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
854 ofproto_set_datapath_id(br->ofproto, dpid);
857 dpid_string = xasprintf("%016"PRIx64, dpid);
858 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
862 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
865 bridge_get_allowed_versions(struct bridge *br)
867 if (!br->cfg->n_protocols)
870 return ofputil_versions_from_strings(br->cfg->protocols,
871 br->cfg->n_protocols);
874 /* Set NetFlow configuration on 'br'. */
876 bridge_configure_netflow(struct bridge *br)
878 struct ovsrec_netflow *cfg = br->cfg->netflow;
879 struct netflow_options opts;
882 ofproto_set_netflow(br->ofproto, NULL);
886 memset(&opts, 0, sizeof opts);
888 /* Get default NetFlow configuration from datapath.
889 * Apply overrides from 'cfg'. */
890 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
891 if (cfg->engine_type) {
892 opts.engine_type = *cfg->engine_type;
894 if (cfg->engine_id) {
895 opts.engine_id = *cfg->engine_id;
898 /* Configure active timeout interval. */
899 opts.active_timeout = cfg->active_timeout;
900 if (!opts.active_timeout) {
901 opts.active_timeout = -1;
902 } else if (opts.active_timeout < 0) {
903 VLOG_WARN("bridge %s: active timeout interval set to negative "
904 "value, using default instead (%d seconds)", br->name,
905 NF_ACTIVE_TIMEOUT_DEFAULT);
906 opts.active_timeout = -1;
909 /* Add engine ID to interface number to disambiguate bridgs? */
910 opts.add_id_to_iface = cfg->add_id_to_interface;
911 if (opts.add_id_to_iface) {
912 if (opts.engine_id > 0x7f) {
913 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
914 "another vswitch, choose an engine id less than 128",
917 if (hmap_count(&br->ports) > 508) {
918 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
919 "another port when more than 508 ports are used",
925 sset_init(&opts.collectors);
926 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
929 if (ofproto_set_netflow(br->ofproto, &opts)) {
930 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
932 sset_destroy(&opts.collectors);
935 /* Set sFlow configuration on 'br'. */
937 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
939 const struct ovsrec_sflow *cfg = br->cfg->sflow;
940 struct ovsrec_controller **controllers;
941 struct ofproto_sflow_options oso;
942 size_t n_controllers;
946 ofproto_set_sflow(br->ofproto, NULL);
950 memset(&oso, 0, sizeof oso);
952 sset_init(&oso.targets);
953 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
955 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
957 oso.sampling_rate = *cfg->sampling;
960 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
962 oso.polling_interval = *cfg->polling;
965 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
967 oso.header_len = *cfg->header;
970 oso.sub_id = (*sflow_bridge_number)++;
971 oso.agent_device = cfg->agent;
973 oso.control_ip = NULL;
974 n_controllers = bridge_get_controllers(br, &controllers);
975 for (i = 0; i < n_controllers; i++) {
976 if (controllers[i]->local_ip) {
977 oso.control_ip = controllers[i]->local_ip;
981 ofproto_set_sflow(br->ofproto, &oso);
983 sset_destroy(&oso.targets);
986 /* Set IPFIX configuration on 'br'. */
988 bridge_configure_ipfix(struct bridge *br)
990 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
991 const struct ovsrec_flow_sample_collector_set *fe_cfg;
992 struct ofproto_ipfix_bridge_exporter_options be_opts;
993 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
994 size_t n_fe_opts = 0;
996 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
997 if (fe_cfg->bridge == br->cfg) {
1002 if (!be_cfg && n_fe_opts == 0) {
1003 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1008 memset(&be_opts, 0, sizeof be_opts);
1010 sset_init(&be_opts.targets);
1011 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1013 if (be_cfg->sampling) {
1014 be_opts.sampling_rate = *be_cfg->sampling;
1016 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1018 if (be_cfg->obs_domain_id) {
1019 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1021 if (be_cfg->obs_point_id) {
1022 be_opts.obs_point_id = *be_cfg->obs_point_id;
1026 if (n_fe_opts > 0) {
1027 struct ofproto_ipfix_flow_exporter_options *opts;
1028 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1030 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1031 if (fe_cfg->bridge == br->cfg) {
1032 opts->collector_set_id = fe_cfg->id;
1033 sset_init(&opts->targets);
1034 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1035 fe_cfg->ipfix->n_targets);
1041 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1045 sset_destroy(&be_opts.targets);
1048 if (n_fe_opts > 0) {
1049 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1051 for (i = 0; i < n_fe_opts; i++) {
1052 sset_destroy(&opts->targets);
1060 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1061 struct ofproto_port_stp_settings *port_s,
1062 int *port_num_counter, unsigned long *port_num_bitmap)
1064 const char *config_str;
1065 struct iface *iface;
1067 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1068 port_s->enable = false;
1071 port_s->enable = true;
1074 /* STP over bonds is not supported. */
1075 if (!list_is_singleton(&port->ifaces)) {
1076 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1078 port_s->enable = false;
1082 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1084 /* Internal ports shouldn't participate in spanning tree, so
1086 if (!strcmp(iface->type, "internal")) {
1087 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1088 port_s->enable = false;
1092 /* STP on mirror output ports is not supported. */
1093 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1094 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1095 port_s->enable = false;
1099 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1101 unsigned long int port_num = strtoul(config_str, NULL, 0);
1102 int port_idx = port_num - 1;
1104 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1105 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1106 port_s->enable = false;
1110 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1111 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1112 port->name, port_num);
1113 port_s->enable = false;
1116 bitmap_set1(port_num_bitmap, port_idx);
1117 port_s->port_num = port_idx;
1119 if (*port_num_counter >= STP_MAX_PORTS) {
1120 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1121 port_s->enable = false;
1125 port_s->port_num = (*port_num_counter)++;
1128 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1130 port_s->path_cost = strtoul(config_str, NULL, 10);
1132 enum netdev_features current;
1135 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1136 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1137 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1140 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1142 port_s->priority = strtoul(config_str, NULL, 0);
1144 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1148 /* Set spanning tree configuration on 'br'. */
1150 bridge_configure_stp(struct bridge *br)
1152 if (!br->cfg->stp_enable) {
1153 ofproto_set_stp(br->ofproto, NULL);
1155 struct ofproto_stp_settings br_s;
1156 const char *config_str;
1158 int port_num_counter;
1159 unsigned long *port_num_bitmap;
1161 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1163 uint8_t ea[ETH_ADDR_LEN];
1165 if (eth_addr_from_string(config_str, ea)) {
1166 br_s.system_id = eth_addr_to_uint64(ea);
1168 br_s.system_id = eth_addr_to_uint64(br->ea);
1169 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1170 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1173 br_s.system_id = eth_addr_to_uint64(br->ea);
1176 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1178 br_s.priority = strtoul(config_str, NULL, 0);
1180 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1183 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1185 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1187 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1190 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1192 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1194 br_s.max_age = STP_DEFAULT_MAX_AGE;
1197 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1199 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1201 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1204 /* Configure STP on the bridge. */
1205 if (ofproto_set_stp(br->ofproto, &br_s)) {
1206 VLOG_ERR("bridge %s: could not enable STP", br->name);
1210 /* Users must either set the port number with the "stp-port-num"
1211 * configuration on all ports or none. If manual configuration
1212 * is not done, then we allocate them sequentially. */
1213 port_num_counter = 0;
1214 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1215 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1216 struct ofproto_port_stp_settings port_s;
1217 struct iface *iface;
1219 port_configure_stp(br->ofproto, port, &port_s,
1220 &port_num_counter, port_num_bitmap);
1222 /* As bonds are not supported, just apply configuration to
1223 * all interfaces. */
1224 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1225 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1227 VLOG_ERR("port %s: could not enable STP", port->name);
1233 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1234 && port_num_counter) {
1235 VLOG_ERR("bridge %s: must manually configure all STP port "
1236 "IDs or none, disabling", br->name);
1237 ofproto_set_stp(br->ofproto, NULL);
1239 bitmap_free(port_num_bitmap);
1244 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1246 const struct port *port = port_lookup(br, name);
1247 return port && port_is_bond_fake_iface(port);
1251 port_is_bond_fake_iface(const struct port *port)
1253 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1257 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1259 struct bridge *br, *next;
1260 struct shash new_br;
1263 /* Collect new bridges' names and types. */
1264 shash_init(&new_br);
1265 for (i = 0; i < cfg->n_bridges; i++) {
1266 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1267 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1269 if (strchr(br_cfg->name, '/')) {
1270 /* Prevent remote ovsdb-server users from accessing arbitrary
1271 * directories, e.g. consider a bridge named "../../../etc/". */
1272 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1274 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1275 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1279 /* Get rid of deleted bridges or those whose types have changed.
1280 * Update 'cfg' of bridges that still exist. */
1281 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1282 br->cfg = shash_find_data(&new_br, br->name);
1283 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1284 br->cfg->datapath_type))) {
1289 /* Add new bridges. */
1290 for (i = 0; i < cfg->n_bridges; i++) {
1291 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1292 struct bridge *br = bridge_lookup(br_cfg->name);
1294 bridge_create(br_cfg);
1298 shash_destroy(&new_br);
1302 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1304 struct bridge *br = iface->port->bridge;
1306 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1307 iface->ofp_port = ofp_port;
1308 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1309 hash_ofp_port(ofp_port));
1310 iface_set_ofport(iface->cfg, ofp_port);
1313 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1314 * Returns 0 if successful, otherwise a positive errno value. */
1316 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1317 struct netdev *netdev)
1321 error = netdev_set_config(netdev, &iface_cfg->options);
1323 VLOG_WARN("could not configure network device %s (%s)",
1324 iface_cfg->name, ovs_strerror(error));
1329 /* This function determines whether 'ofproto_port', which is attached to
1330 * br->ofproto's datapath, is one that we want in 'br'.
1332 * If it is, it returns true, after creating an iface (if necessary),
1333 * configuring the iface's netdev according to the iface's options, and setting
1334 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1336 * If, on the other hand, 'port' should be removed, it returns false. The
1337 * caller should later detach the port from br->ofproto. */
1339 bridge_refresh_one_ofp_port(struct bridge *br,
1340 const struct ofproto_port *ofproto_port)
1342 const char *name = ofproto_port->name;
1343 const char *type = ofproto_port->type;
1344 ofp_port_t ofp_port = ofproto_port->ofp_port;
1346 struct iface *iface = iface_lookup(br, name);
1348 /* Check that the name-to-number mapping is one-to-one. */
1349 if (iface->ofp_port != OFPP_NONE) {
1350 VLOG_WARN("bridge %s: interface %s reported twice",
1353 } else if (iface_from_ofp_port(br, ofp_port)) {
1354 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1355 br->name, ofp_port);
1359 /* There's a configured interface named 'name'. */
1360 if (strcmp(type, iface->type)
1361 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1362 /* It's the wrong type, or it's the right type but can't be
1363 * configured as the user requested, so we must destroy it. */
1366 /* It's the right type and configured correctly. Keep it. */
1367 iface_set_ofp_port(iface, ofp_port);
1370 } else if (bridge_has_bond_fake_iface(br, name)
1371 && !strcmp(type, "internal")) {
1372 /* It's a bond fake iface. Keep it. */
1375 /* There's no configured interface named 'name', but there might be an
1376 * interface of that name queued to be created.
1378 * If there is, and it has the correct type, then try to configure it
1379 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1380 * delete it and later try to re-add it. */
1381 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1383 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1384 && iface_create(br, if_cfg, ofp_port));
1388 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1389 * consistent with the ofp_ports in "br->ofproto". */
1391 bridge_refresh_ofp_port(struct bridge *br)
1393 struct ofproto_port_dump dump;
1394 struct ofproto_port ofproto_port;
1395 struct port *port, *port_next;
1397 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1398 hmap_clear(&br->ifaces);
1399 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1400 struct iface *iface;
1402 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1403 iface->ofp_port = OFPP_NONE;
1407 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1408 * already exist in the datapath and promote them to full fledged "struct
1409 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1410 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1411 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1412 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1413 garbage->ofp_port = ofproto_port.ofp_port;
1414 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1418 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1419 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1420 * later they can be added to ofproto. */
1421 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1422 struct iface *iface, *iface_next;
1424 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1425 if (iface->ofp_port == OFPP_NONE) {
1426 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1427 iface_destroy(iface);
1431 if (list_is_empty(&port->ifaces)) {
1437 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1438 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1439 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1442 * If successful, returns 0 and stores the network device in '*netdevp'. On
1443 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1445 iface_do_create(const struct bridge *br,
1446 const struct if_cfg *if_cfg,
1447 ofp_port_t *ofp_portp, struct netdev **netdevp)
1449 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1450 const struct ovsrec_port *port_cfg = if_cfg->parent;
1451 struct netdev *netdev = NULL;
1454 if (netdev_is_reserved_name(iface_cfg->name)) {
1455 VLOG_WARN("could not create interface %s, name is reserved",
1461 error = netdev_open(iface_cfg->name,
1462 iface_get_type(iface_cfg, br->cfg), &netdev);
1464 VLOG_WARN("could not open network device %s (%s)",
1465 iface_cfg->name, ovs_strerror(error));
1469 error = iface_set_netdev_config(iface_cfg, netdev);
1474 if (*ofp_portp == OFPP_NONE) {
1475 ofp_port_t ofp_port = if_cfg->ofport;
1477 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1481 *ofp_portp = ofp_port;
1483 VLOG_INFO("bridge %s: added interface %s on port %d",
1484 br->name, iface_cfg->name, *ofp_portp);
1486 VLOG_DBG("bridge %s: interface %s is on port %d",
1487 br->name, iface_cfg->name, *ofp_portp);
1490 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1491 || iface_is_internal(iface_cfg, br->cfg)) {
1492 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1500 netdev_close(netdev);
1504 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1505 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1506 * automatically allocated for the iface. Takes ownership of and
1507 * deallocates 'if_cfg'.
1509 * Return true if an iface is successfully created, false otherwise. */
1511 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1513 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1514 const struct ovsrec_port *port_cfg = if_cfg->parent;
1516 struct netdev *netdev;
1517 struct iface *iface;
1522 /* Do the bits that can fail up front.
1524 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1525 * internal datastructures may not be consistent. Eventually, when port
1526 * additions and deletions are cheaper, these calls should be removed. */
1528 ovs_assert(!iface_lookup(br, iface_cfg->name));
1529 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1532 iface_set_ofport(iface_cfg, OFPP_NONE);
1533 iface_clear_db_record(iface_cfg);
1538 /* Get or create the port structure. */
1539 port = port_lookup(br, port_cfg->name);
1541 port = port_create(br, port_cfg);
1544 /* Create the iface structure. */
1545 iface = xzalloc(sizeof *iface);
1546 list_push_back(&port->ifaces, &iface->port_elem);
1547 hmap_insert(&br->iface_by_name, &iface->name_node,
1548 hash_string(iface_cfg->name, 0));
1550 iface->name = xstrdup(iface_cfg->name);
1551 iface->ofp_port = OFPP_NONE;
1552 iface->netdev = netdev;
1553 iface->type = iface_get_type(iface_cfg, br->cfg);
1554 iface->cfg = iface_cfg;
1556 iface_set_ofp_port(iface, ofp_port);
1558 /* Populate initial status in database. */
1559 iface_refresh_stats(iface);
1560 iface_refresh_status(iface);
1562 /* Add bond fake iface if necessary. */
1563 if (port_is_bond_fake_iface(port)) {
1564 struct ofproto_port ofproto_port;
1566 if (ofproto_port_query_by_name(br->ofproto, port->name,
1568 struct netdev *netdev;
1571 error = netdev_open(port->name, "internal", &netdev);
1573 ofp_port_t fake_ofp_port = if_cfg->ofport;
1575 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1576 netdev_close(netdev);
1578 VLOG_WARN("could not open network device %s (%s)",
1579 port->name, ovs_strerror(error));
1582 /* Already exists, nothing to do. */
1583 ofproto_port_destroy(&ofproto_port);
1588 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1594 /* Set forward BPDU option. */
1596 bridge_configure_forward_bpdu(struct bridge *br)
1598 ofproto_set_forward_bpdu(br->ofproto,
1599 smap_get_bool(&br->cfg->other_config,
1604 /* Set MAC learning table configuration for 'br'. */
1606 bridge_configure_mac_table(struct bridge *br)
1608 const char *idle_time_str;
1611 const char *mac_table_size_str;
1614 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1615 idle_time = (idle_time_str && atoi(idle_time_str)
1616 ? atoi(idle_time_str)
1617 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1619 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1620 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1621 ? atoi(mac_table_size_str)
1624 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1628 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1629 struct iface **hw_addr_iface)
1631 struct hmapx mirror_output_ports;
1634 bool found_addr = false;
1638 *hw_addr_iface = NULL;
1640 /* Did the user request a particular MAC? */
1641 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1642 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1643 if (eth_addr_is_multicast(ea)) {
1644 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1645 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1646 } else if (eth_addr_is_zero(ea)) {
1647 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1653 /* Mirror output ports don't participate in picking the local hardware
1654 * address. ofproto can't help us find out whether a given port is a
1655 * mirror output because we haven't configured mirrors yet, so we need to
1656 * accumulate them ourselves. */
1657 hmapx_init(&mirror_output_ports);
1658 for (i = 0; i < br->cfg->n_mirrors; i++) {
1659 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1660 if (m->output_port) {
1661 hmapx_add(&mirror_output_ports, m->output_port);
1665 /* Otherwise choose the minimum non-local MAC address among all of the
1667 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1668 uint8_t iface_ea[ETH_ADDR_LEN];
1669 struct iface *candidate;
1670 struct iface *iface;
1672 /* Mirror output ports don't participate. */
1673 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1677 /* Choose the MAC address to represent the port. */
1679 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1680 /* Find the interface with this Ethernet address (if any) so that
1681 * we can provide the correct devname to the caller. */
1682 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1683 uint8_t candidate_ea[ETH_ADDR_LEN];
1684 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1685 && eth_addr_equals(iface_ea, candidate_ea)) {
1690 /* Choose the interface whose MAC address will represent the port.
1691 * The Linux kernel bonding code always chooses the MAC address of
1692 * the first slave added to a bond, and the Fedora networking
1693 * scripts always add slaves to a bond in alphabetical order, so
1694 * for compatibility we choose the interface with the name that is
1695 * first in alphabetical order. */
1696 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1697 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1702 /* The local port doesn't count (since we're trying to choose its
1703 * MAC address anyway). */
1704 if (iface->ofp_port == OFPP_LOCAL) {
1709 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1715 /* Compare against our current choice. */
1716 if (!eth_addr_is_multicast(iface_ea) &&
1717 !eth_addr_is_local(iface_ea) &&
1718 !eth_addr_is_reserved(iface_ea) &&
1719 !eth_addr_is_zero(iface_ea) &&
1720 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1722 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1723 *hw_addr_iface = iface;
1729 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1730 *hw_addr_iface = NULL;
1733 hmapx_destroy(&mirror_output_ports);
1736 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1737 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1738 * an interface on 'br', then that interface must be passed in as
1739 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1740 * 'hw_addr_iface' must be passed in as a null pointer. */
1742 bridge_pick_datapath_id(struct bridge *br,
1743 const uint8_t bridge_ea[ETH_ADDR_LEN],
1744 struct iface *hw_addr_iface)
1747 * The procedure for choosing a bridge MAC address will, in the most
1748 * ordinary case, also choose a unique MAC that we can use as a datapath
1749 * ID. In some special cases, though, multiple bridges will end up with
1750 * the same MAC address. This is OK for the bridges, but it will confuse
1751 * the OpenFlow controller, because each datapath needs a unique datapath
1754 * Datapath IDs must be unique. It is also very desirable that they be
1755 * stable from one run to the next, so that policy set on a datapath
1758 const char *datapath_id;
1761 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1762 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1766 if (!hw_addr_iface) {
1768 * A purely internal bridge, that is, one that has no non-virtual
1769 * network devices on it at all, is difficult because it has no
1770 * natural unique identifier at all.
1772 * When the host is a XenServer, we handle this case by hashing the
1773 * host's UUID with the name of the bridge. Names of bridges are
1774 * persistent across XenServer reboots, although they can be reused if
1775 * an internal network is destroyed and then a new one is later
1776 * created, so this is fairly effective.
1778 * When the host is not a XenServer, we punt by using a random MAC
1779 * address on each run.
1781 const char *host_uuid = xenserver_get_host_uuid();
1783 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1784 dpid = dpid_from_hash(combined, strlen(combined));
1790 return eth_addr_to_uint64(bridge_ea);
1794 dpid_from_hash(const void *data, size_t n)
1796 uint8_t hash[SHA1_DIGEST_SIZE];
1798 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1799 sha1_bytes(data, n, hash);
1800 eth_addr_mark_random(hash);
1801 return eth_addr_to_uint64(hash);
1805 iface_refresh_status(struct iface *iface)
1809 enum netdev_features current;
1813 uint8_t mac[ETH_ADDR_LEN];
1817 if (iface_is_synthetic(iface)) {
1823 if (!netdev_get_status(iface->netdev, &smap)) {
1824 ovsrec_interface_set_status(iface->cfg, &smap);
1826 ovsrec_interface_set_status(iface->cfg, NULL);
1829 smap_destroy(&smap);
1831 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1832 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1834 ovsrec_interface_set_duplex(iface->cfg,
1835 netdev_features_is_full_duplex(current)
1837 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1839 ovsrec_interface_set_duplex(iface->cfg, NULL);
1840 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1843 error = netdev_get_mtu(iface->netdev, &mtu);
1846 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1848 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1851 error = netdev_get_etheraddr(iface->netdev, mac);
1853 char mac_string[32];
1855 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1856 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1858 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1861 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1862 * if there is no valid ifindex number. */
1863 ifindex64 = netdev_get_ifindex(iface->netdev);
1864 if (ifindex64 < 0) {
1867 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1870 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1873 iface_refresh_cfm_stats(struct iface *iface)
1875 const struct ovsrec_interface *cfg = iface->cfg;
1876 struct ofproto_cfm_status status;
1878 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1879 iface->ofp_port, &status)) {
1880 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1881 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1882 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1883 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1884 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1886 const char *reasons[CFM_FAULT_N_REASONS];
1887 int64_t cfm_health = status.health;
1888 bool faulted = status.faults != 0;
1891 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1894 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1895 int reason = 1 << i;
1896 if (status.faults & reason) {
1897 reasons[j++] = cfm_fault_reason_to_str(reason);
1900 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1902 if (status.remote_opstate >= 0) {
1903 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1904 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1906 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1909 ovsrec_interface_set_cfm_remote_mpids(cfg,
1910 (const int64_t *)status.rmps,
1912 if (cfm_health >= 0) {
1913 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1915 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1921 iface_refresh_stats(struct iface *iface)
1923 #define IFACE_STATS \
1924 IFACE_STAT(rx_packets, "rx_packets") \
1925 IFACE_STAT(tx_packets, "tx_packets") \
1926 IFACE_STAT(rx_bytes, "rx_bytes") \
1927 IFACE_STAT(tx_bytes, "tx_bytes") \
1928 IFACE_STAT(rx_dropped, "rx_dropped") \
1929 IFACE_STAT(tx_dropped, "tx_dropped") \
1930 IFACE_STAT(rx_errors, "rx_errors") \
1931 IFACE_STAT(tx_errors, "tx_errors") \
1932 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1933 IFACE_STAT(rx_over_errors, "rx_over_err") \
1934 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1935 IFACE_STAT(collisions, "collisions")
1937 #define IFACE_STAT(MEMBER, NAME) + 1
1938 enum { N_IFACE_STATS = IFACE_STATS };
1940 int64_t values[N_IFACE_STATS];
1941 char *keys[N_IFACE_STATS];
1944 struct netdev_stats stats;
1946 if (iface_is_synthetic(iface)) {
1950 /* Intentionally ignore return value, since errors will set 'stats' to
1951 * all-1s, and we will deal with that correctly below. */
1952 netdev_get_stats(iface->netdev, &stats);
1954 /* Copy statistics into keys[] and values[]. */
1956 #define IFACE_STAT(MEMBER, NAME) \
1957 if (stats.MEMBER != UINT64_MAX) { \
1959 values[n] = stats.MEMBER; \
1964 ovs_assert(n <= N_IFACE_STATS);
1966 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1971 br_refresh_stp_status(struct bridge *br)
1973 struct smap smap = SMAP_INITIALIZER(&smap);
1974 struct ofproto *ofproto = br->ofproto;
1975 struct ofproto_stp_status status;
1977 if (ofproto_get_stp_status(ofproto, &status)) {
1981 if (!status.enabled) {
1982 ovsrec_bridge_set_status(br->cfg, NULL);
1986 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1987 STP_ID_ARGS(status.bridge_id));
1988 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1989 STP_ID_ARGS(status.designated_root));
1990 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1992 ovsrec_bridge_set_status(br->cfg, &smap);
1993 smap_destroy(&smap);
1997 port_refresh_stp_status(struct port *port)
1999 struct ofproto *ofproto = port->bridge->ofproto;
2000 struct iface *iface;
2001 struct ofproto_port_stp_status status;
2003 int64_t int_values[3];
2006 if (port_is_synthetic(port)) {
2010 /* STP doesn't currently support bonds. */
2011 if (!list_is_singleton(&port->ifaces)) {
2012 ovsrec_port_set_status(port->cfg, NULL);
2016 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2018 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2022 if (!status.enabled) {
2023 ovsrec_port_set_status(port->cfg, NULL);
2024 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2028 /* Set Status column. */
2030 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2031 smap_add(&smap, "stp_state", stp_state_name(status.state));
2032 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2033 smap_add(&smap, "stp_role", stp_role_name(status.role));
2034 ovsrec_port_set_status(port->cfg, &smap);
2035 smap_destroy(&smap);
2037 /* Set Statistics column. */
2038 keys[0] = "stp_tx_count";
2039 int_values[0] = status.tx_count;
2040 keys[1] = "stp_rx_count";
2041 int_values[1] = status.rx_count;
2042 keys[2] = "stp_error_count";
2043 int_values[2] = status.error_count;
2045 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2046 ARRAY_SIZE(int_values));
2050 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2052 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2056 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2058 bool enable = enable_system_stats(cfg);
2060 system_stats_enable(enable);
2062 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2067 run_system_stats(void)
2069 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2072 stats = system_stats_run();
2074 struct ovsdb_idl_txn *txn;
2075 struct ovsdb_datum datum;
2077 txn = ovsdb_idl_txn_create(idl);
2078 ovsdb_datum_from_smap(&datum, stats);
2079 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2081 ovsdb_idl_txn_commit(txn);
2082 ovsdb_idl_txn_destroy(txn);
2088 static inline const char *
2089 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2092 case OFPCR12_ROLE_EQUAL:
2094 case OFPCR12_ROLE_MASTER:
2096 case OFPCR12_ROLE_SLAVE:
2098 case OFPCR12_ROLE_NOCHANGE:
2100 return "*** INVALID ROLE ***";
2105 refresh_controller_status(void)
2109 const struct ovsrec_controller *cfg;
2113 /* Accumulate status for controllers on all bridges. */
2114 HMAP_FOR_EACH (br, node, &all_bridges) {
2115 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2118 /* Update each controller in the database with current status. */
2119 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2120 struct ofproto_controller_info *cinfo =
2121 shash_find_data(&info, cfg->target);
2124 struct smap smap = SMAP_INITIALIZER(&smap);
2125 const char **values = cinfo->pairs.values;
2126 const char **keys = cinfo->pairs.keys;
2129 for (i = 0; i < cinfo->pairs.n; i++) {
2130 smap_add(&smap, keys[i], values[i]);
2133 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2134 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2136 ovsrec_controller_set_status(cfg, &smap);
2137 smap_destroy(&smap);
2139 ovsrec_controller_set_is_connected(cfg, false);
2140 ovsrec_controller_set_role(cfg, NULL);
2141 ovsrec_controller_set_status(cfg, NULL);
2145 ofproto_free_ofproto_controller_info(&info);
2150 * Some information in the database must be kept as up-to-date as possible to
2151 * allow controllers to respond rapidly to network outages. We call these
2152 * statistics "instant" stats.
2154 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2155 * assuming that they've changed. The only means we have to determine whether
2156 * they have changed are:
2158 * - Try to commit changes to the database. If nothing changed, then
2159 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2162 * - instant_stats_run() is called late in the run loop, after anything that
2163 * might change any of the instant stats.
2165 * We use these two facts together to avoid waking the process up every
2166 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2169 /* Minimum interval between writing updates to the instant stats to the
2171 #define INSTANT_INTERVAL_MSEC 100
2173 /* Current instant stats database transaction, NULL if there is no ongoing
2175 static struct ovsdb_idl_txn *instant_txn;
2177 /* Next time (in msec on monotonic clock) at which we will update the instant
2179 static long long int instant_next_txn = LLONG_MIN;
2181 /* True if the run loop has run since we last saw that the instant stats were
2182 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2183 * to refresh the instant stats. */
2184 static bool instant_stats_could_have_changed;
2187 instant_stats_run(void)
2189 enum ovsdb_idl_txn_status status;
2191 instant_stats_could_have_changed = true;
2196 if (time_msec() < instant_next_txn) {
2199 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2201 instant_txn = ovsdb_idl_txn_create(idl);
2202 HMAP_FOR_EACH (br, node, &all_bridges) {
2203 struct iface *iface;
2206 br_refresh_stp_status(br);
2208 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2209 port_refresh_stp_status(port);
2212 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2213 enum netdev_flags flags;
2215 const char *link_state;
2216 int64_t link_resets;
2219 if (iface_is_synthetic(iface)) {
2223 current = ofproto_port_is_lacp_current(br->ofproto,
2227 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2229 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2232 error = netdev_get_flags(iface->netdev, &flags);
2234 const char *state = flags & NETDEV_UP ? "up" : "down";
2235 ovsrec_interface_set_admin_state(iface->cfg, state);
2237 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2240 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2241 ovsrec_interface_set_link_state(iface->cfg, link_state);
2243 link_resets = netdev_get_carrier_resets(iface->netdev);
2244 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2246 iface_refresh_cfm_stats(iface);
2249 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2251 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2252 smap_destroy(&smap);
2258 status = ovsdb_idl_txn_commit(instant_txn);
2259 if (status != TXN_INCOMPLETE) {
2260 ovsdb_idl_txn_destroy(instant_txn);
2263 if (status == TXN_UNCHANGED) {
2264 instant_stats_could_have_changed = false;
2269 instant_stats_wait(void)
2272 ovsdb_idl_txn_wait(instant_txn);
2273 } else if (instant_stats_could_have_changed) {
2274 poll_timer_wait_until(instant_next_txn);
2278 /* Performs periodic activity required by bridges that needs to be done with
2279 * the least possible latency.
2281 * It makes sense to call this function a couple of times per poll loop, to
2282 * provide a significant performance boost on some benchmarks with ofprotos
2283 * that use the ofproto-dpif implementation. */
2285 bridge_run_fast(void)
2292 ofproto_enumerate_types(&types);
2293 SSET_FOR_EACH (type, &types) {
2294 ofproto_type_run_fast(type);
2296 sset_destroy(&types);
2298 HMAP_FOR_EACH (br, node, &all_bridges) {
2299 ofproto_run_fast(br->ofproto);
2306 static struct ovsrec_open_vswitch null_cfg;
2307 const struct ovsrec_open_vswitch *cfg;
2308 struct ovsdb_idl_txn *reconf_txn = NULL;
2312 bool vlan_splinters_changed;
2315 ovsrec_open_vswitch_init(&null_cfg);
2317 /* (Re)configure if necessary. */
2318 if (!reconfiguring) {
2321 if (ovsdb_idl_is_lock_contended(idl)) {
2322 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2323 struct bridge *br, *next_br;
2325 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2326 "disabling this process (pid %ld) until it goes away",
2327 (long int) getpid());
2329 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2332 /* Since we will not be running system_stats_run() in this process
2333 * with the current situation of multiple ovs-vswitchd daemons,
2334 * disable system stats collection. */
2335 system_stats_enable(false);
2337 } else if (!ovsdb_idl_has_lock(idl)) {
2341 cfg = ovsrec_open_vswitch_first(idl);
2343 /* Initialize the ofproto library. This only needs to run once, but
2344 * it must be done after the configuration is set. If the
2345 * initialization has already occurred, bridge_init_ofproto()
2346 * returns immediately. */
2347 bridge_init_ofproto(cfg);
2349 /* Once the value of flow-restore-wait is false, we no longer should
2350 * check its value from the database. */
2351 if (cfg && ofproto_get_flow_restore_wait()) {
2352 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2353 "flow-restore-wait", false));
2356 /* Let each datapath type do the work that it needs to do. */
2358 ofproto_enumerate_types(&types);
2359 SSET_FOR_EACH (type, &types) {
2360 ofproto_type_run(type);
2362 sset_destroy(&types);
2364 /* Let each bridge do the work that it needs to do. */
2365 HMAP_FOR_EACH (br, node, &all_bridges) {
2366 ofproto_run(br->ofproto);
2369 /* Re-configure SSL. We do this on every trip through the main loop,
2370 * instead of just when the database changes, because the contents of the
2371 * key and certificate files can change without the database changing.
2373 * We do this before bridge_reconfigure() because that function might
2374 * initiate SSL connections and thus requires SSL to be configured. */
2375 if (cfg && cfg->ssl) {
2376 const struct ovsrec_ssl *ssl = cfg->ssl;
2378 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2379 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2382 if (!reconfiguring) {
2383 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2384 * usage has changed. */
2385 vlan_splinters_changed = false;
2386 if (vlan_splinters_enabled_anywhere) {
2387 HMAP_FOR_EACH (br, node, &all_bridges) {
2388 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2389 vlan_splinters_changed = true;
2395 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2396 idl_seqno = ovsdb_idl_get_seqno(idl);
2398 reconf_txn = ovsdb_idl_txn_create(idl);
2399 bridge_reconfigure(cfg);
2401 /* We still need to reconfigure to avoid dangling pointers to
2402 * now-destroyed ovsrec structures inside bridge data. */
2403 bridge_reconfigure(&null_cfg);
2408 if (reconfiguring) {
2410 reconf_txn = ovsdb_idl_txn_create(idl);
2413 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2414 reconfiguring = false;
2417 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2420 /* If we are completing our initial configuration for this run
2421 * of ovs-vswitchd, then keep the transaction around to monitor
2422 * it for completion. */
2423 if (!initial_config_done) {
2424 initial_config_done = true;
2425 daemonize_txn = reconf_txn;
2432 ovsdb_idl_txn_commit(reconf_txn);
2433 ovsdb_idl_txn_destroy(reconf_txn);
2437 if (daemonize_txn) {
2438 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2439 if (status != TXN_INCOMPLETE) {
2440 ovsdb_idl_txn_destroy(daemonize_txn);
2441 daemonize_txn = NULL;
2443 /* ovs-vswitchd has completed initialization, so allow the
2444 * process that forked us to exit successfully. */
2445 daemonize_complete();
2447 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2451 /* Refresh interface and mirror stats if necessary. */
2452 if (time_msec() >= iface_stats_timer) {
2454 struct ovsdb_idl_txn *txn;
2456 txn = ovsdb_idl_txn_create(idl);
2457 HMAP_FOR_EACH (br, node, &all_bridges) {
2461 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2462 struct iface *iface;
2464 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2465 iface_refresh_stats(iface);
2466 iface_refresh_status(iface);
2470 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2471 mirror_refresh_stats(m);
2475 refresh_controller_status();
2476 ovsdb_idl_txn_commit(txn);
2477 ovsdb_idl_txn_destroy(txn); /* XXX */
2480 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2484 instant_stats_run();
2493 ovsdb_idl_wait(idl);
2494 if (daemonize_txn) {
2495 ovsdb_idl_txn_wait(daemonize_txn);
2498 if (reconfiguring) {
2499 poll_immediate_wake();
2503 ofproto_enumerate_types(&types);
2504 SSET_FOR_EACH (type, &types) {
2505 ofproto_type_wait(type);
2507 sset_destroy(&types);
2509 if (!hmap_is_empty(&all_bridges)) {
2512 HMAP_FOR_EACH (br, node, &all_bridges) {
2513 ofproto_wait(br->ofproto);
2515 poll_timer_wait_until(iface_stats_timer);
2518 system_stats_wait();
2519 instant_stats_wait();
2522 /* Adds some memory usage statistics for bridges into 'usage', for use with
2523 * memory_report(). */
2525 bridge_get_memory_usage(struct simap *usage)
2529 HMAP_FOR_EACH (br, node, &all_bridges) {
2530 ofproto_get_memory_usage(br->ofproto, usage);
2534 /* QoS unixctl user interface functions. */
2536 struct qos_unixctl_show_cbdata {
2538 struct iface *iface;
2542 qos_unixctl_show_cb(unsigned int queue_id,
2543 const struct smap *details,
2546 struct qos_unixctl_show_cbdata *data = aux;
2547 struct ds *ds = data->ds;
2548 struct iface *iface = data->iface;
2549 struct netdev_queue_stats stats;
2550 struct smap_node *node;
2553 ds_put_cstr(ds, "\n");
2555 ds_put_format(ds, "Queue %u:\n", queue_id);
2557 ds_put_cstr(ds, "Default:\n");
2560 SMAP_FOR_EACH (node, details) {
2561 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2564 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2566 if (stats.tx_packets != UINT64_MAX) {
2567 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2570 if (stats.tx_bytes != UINT64_MAX) {
2571 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2574 if (stats.tx_errors != UINT64_MAX) {
2575 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2578 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2579 queue_id, ovs_strerror(error));
2584 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2585 const char *argv[], void *aux OVS_UNUSED)
2587 struct ds ds = DS_EMPTY_INITIALIZER;
2588 struct smap smap = SMAP_INITIALIZER(&smap);
2589 struct iface *iface;
2591 struct smap_node *node;
2592 struct qos_unixctl_show_cbdata data;
2595 iface = iface_find(argv[1]);
2597 unixctl_command_reply_error(conn, "no such interface");
2601 netdev_get_qos(iface->netdev, &type, &smap);
2603 if (*type != '\0') {
2604 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2606 SMAP_FOR_EACH (node, &smap) {
2607 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2612 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2615 ds_put_format(&ds, "failed to dump queues: %s",
2616 ovs_strerror(error));
2618 unixctl_command_reply(conn, ds_cstr(&ds));
2620 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2621 unixctl_command_reply_error(conn, ds_cstr(&ds));
2624 smap_destroy(&smap);
2628 /* Bridge reconfiguration functions. */
2630 bridge_create(const struct ovsrec_bridge *br_cfg)
2634 ovs_assert(!bridge_lookup(br_cfg->name));
2635 br = xzalloc(sizeof *br);
2637 br->name = xstrdup(br_cfg->name);
2638 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2641 /* Derive the default Ethernet address from the bridge's UUID. This should
2642 * be unique and it will be stable between ovs-vswitchd runs. */
2643 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2644 eth_addr_mark_random(br->default_ea);
2646 hmap_init(&br->ports);
2647 hmap_init(&br->ifaces);
2648 hmap_init(&br->iface_by_name);
2649 hmap_init(&br->mirrors);
2651 hmap_init(&br->if_cfg_todo);
2652 list_init(&br->ofpp_garbage);
2654 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2658 bridge_destroy(struct bridge *br)
2661 struct mirror *mirror, *next_mirror;
2662 struct port *port, *next_port;
2663 struct if_cfg *if_cfg, *next_if_cfg;
2664 struct ofpp_garbage *garbage, *next_garbage;
2666 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2669 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2670 mirror_destroy(mirror);
2672 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2673 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2676 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2677 &br->ofpp_garbage) {
2678 list_remove(&garbage->list_node);
2682 hmap_remove(&all_bridges, &br->node);
2683 ofproto_destroy(br->ofproto);
2684 hmap_destroy(&br->ifaces);
2685 hmap_destroy(&br->ports);
2686 hmap_destroy(&br->iface_by_name);
2687 hmap_destroy(&br->mirrors);
2688 hmap_destroy(&br->if_cfg_todo);
2695 static struct bridge *
2696 bridge_lookup(const char *name)
2700 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2701 if (!strcmp(br->name, name)) {
2708 /* Handle requests for a listing of all flows known by the OpenFlow
2709 * stack, including those normally hidden. */
2711 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2712 const char *argv[], void *aux OVS_UNUSED)
2717 br = bridge_lookup(argv[1]);
2719 unixctl_command_reply_error(conn, "Unknown bridge");
2724 ofproto_get_all_flows(br->ofproto, &results);
2726 unixctl_command_reply(conn, ds_cstr(&results));
2727 ds_destroy(&results);
2730 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2731 * connections and reconnect. If BRIDGE is not specified, then all bridges
2732 * drop their controller connections and reconnect. */
2734 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2735 const char *argv[], void *aux OVS_UNUSED)
2739 br = bridge_lookup(argv[1]);
2741 unixctl_command_reply_error(conn, "Unknown bridge");
2744 ofproto_reconnect_controllers(br->ofproto);
2746 HMAP_FOR_EACH (br, node, &all_bridges) {
2747 ofproto_reconnect_controllers(br->ofproto);
2750 unixctl_command_reply(conn, NULL);
2754 bridge_get_controllers(const struct bridge *br,
2755 struct ovsrec_controller ***controllersp)
2757 struct ovsrec_controller **controllers;
2758 size_t n_controllers;
2760 controllers = br->cfg->controller;
2761 n_controllers = br->cfg->n_controller;
2763 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2769 *controllersp = controllers;
2771 return n_controllers;
2775 bridge_queue_if_cfg(struct bridge *br,
2776 const struct ovsrec_interface *cfg,
2777 const struct ovsrec_port *parent)
2779 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2782 if_cfg->parent = parent;
2783 if_cfg->ofport = iface_pick_ofport(cfg);
2784 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2785 hash_string(if_cfg->cfg->name, 0));
2788 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2789 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2790 * 'br' needs to complete its configuration. */
2792 bridge_add_del_ports(struct bridge *br,
2793 const unsigned long int *splinter_vlans)
2795 struct shash_node *port_node;
2796 struct port *port, *next;
2797 struct shash new_ports;
2800 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2802 /* Collect new ports. */
2803 shash_init(&new_ports);
2804 for (i = 0; i < br->cfg->n_ports; i++) {
2805 const char *name = br->cfg->ports[i]->name;
2806 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2807 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2811 if (bridge_get_controllers(br, NULL)
2812 && !shash_find(&new_ports, br->name)) {
2813 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2814 br->name, br->name);
2816 ovsrec_interface_init(&br->synth_local_iface);
2817 ovsrec_port_init(&br->synth_local_port);
2819 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2820 br->synth_local_port.n_interfaces = 1;
2821 br->synth_local_port.name = br->name;
2823 br->synth_local_iface.name = br->name;
2824 br->synth_local_iface.type = "internal";
2826 br->synth_local_ifacep = &br->synth_local_iface;
2828 shash_add(&new_ports, br->name, &br->synth_local_port);
2831 if (splinter_vlans) {
2832 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2835 /* Get rid of deleted ports.
2836 * Get rid of deleted interfaces on ports that still exist. */
2837 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2838 port->cfg = shash_find_data(&new_ports, port->name);
2842 port_del_ifaces(port);
2846 /* Update iface->cfg and iface->type in interfaces that still exist.
2847 * Add new interfaces to creation queue. */
2848 SHASH_FOR_EACH (port_node, &new_ports) {
2849 const struct ovsrec_port *port = port_node->data;
2852 for (i = 0; i < port->n_interfaces; i++) {
2853 const struct ovsrec_interface *cfg = port->interfaces[i];
2854 struct iface *iface = iface_lookup(br, cfg->name);
2855 const char *type = iface_get_type(cfg, br->cfg);
2860 } else if (!strcmp(type, "null")) {
2861 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2862 " may be removed in February 2013. Please email"
2863 " dev@openvswitch.org with concerns.",
2866 bridge_queue_if_cfg(br, cfg, port);
2871 shash_destroy(&new_ports);
2874 /* Initializes 'oc' appropriately as a management service controller for
2877 * The caller must free oc->target when it is no longer needed. */
2879 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2880 struct ofproto_controller *oc)
2882 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2883 oc->max_backoff = 0;
2884 oc->probe_interval = 60;
2885 oc->band = OFPROTO_OUT_OF_BAND;
2887 oc->burst_limit = 0;
2888 oc->enable_async_msgs = true;
2891 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2893 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2894 struct ofproto_controller *oc)
2898 oc->target = c->target;
2899 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2900 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2901 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2902 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2903 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2904 oc->burst_limit = (c->controller_burst_limit
2905 ? *c->controller_burst_limit : 0);
2906 oc->enable_async_msgs = (!c->enable_async_messages
2907 || *c->enable_async_messages);
2908 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2909 if (dscp < 0 || dscp > 63) {
2910 dscp = DSCP_DEFAULT;
2915 /* Configures the IP stack for 'br''s local interface properly according to the
2916 * configuration in 'c'. */
2918 bridge_configure_local_iface_netdev(struct bridge *br,
2919 struct ovsrec_controller *c)
2921 struct netdev *netdev;
2922 struct in_addr mask, gateway;
2924 struct iface *local_iface;
2927 /* If there's no local interface or no IP address, give up. */
2928 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2929 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2933 /* Bring up the local interface. */
2934 netdev = local_iface->netdev;
2935 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2937 /* Configure the IP address and netmask. */
2938 if (!c->local_netmask
2939 || !inet_aton(c->local_netmask, &mask)
2941 mask.s_addr = guess_netmask(ip.s_addr);
2943 if (!netdev_set_in4(netdev, ip, mask)) {
2944 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2945 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2948 /* Configure the default gateway. */
2949 if (c->local_gateway
2950 && inet_aton(c->local_gateway, &gateway)
2951 && gateway.s_addr) {
2952 if (!netdev_add_router(netdev, gateway)) {
2953 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2954 br->name, IP_ARGS(gateway.s_addr));
2959 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2960 * in either string are treated as equal to any number of slashes in the other,
2961 * e.g. "x///y" is equal to "x/y".
2963 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2964 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2965 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2966 * 'b' against a prefix of 'a'.
2969 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2971 const char *b_start = b;
2973 if (b - b_start >= b_stoplen) {
2975 } else if (*a != *b) {
2977 } else if (*a == '/') {
2978 a += strspn(a, "/");
2979 b += strspn(b, "/");
2980 } else if (*a == '\0') {
2990 bridge_configure_remotes(struct bridge *br,
2991 const struct sockaddr_in *managers, size_t n_managers)
2993 bool disable_in_band;
2995 struct ovsrec_controller **controllers;
2996 size_t n_controllers;
2998 enum ofproto_fail_mode fail_mode;
3000 struct ofproto_controller *ocs;
3004 /* Check if we should disable in-band control on this bridge. */
3005 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3008 /* Set OpenFlow queue ID for in-band control. */
3009 ofproto_set_in_band_queue(br->ofproto,
3010 smap_get_int(&br->cfg->other_config,
3011 "in-band-queue", -1));
3013 if (disable_in_band) {
3014 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3016 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3019 n_controllers = bridge_get_controllers(br, &controllers);
3021 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3024 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3025 for (i = 0; i < n_controllers; i++) {
3026 struct ovsrec_controller *c = controllers[i];
3028 if (!strncmp(c->target, "punix:", 6)
3029 || !strncmp(c->target, "unix:", 5)) {
3030 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3033 if (!strncmp(c->target, "unix:", 5)) {
3034 /* Connect to a listening socket */
3035 whitelist = xasprintf("unix:%s/", ovs_rundir());
3036 if (strchr(c->target, '/') &&
3037 !equal_pathnames(c->target, whitelist,
3038 strlen(whitelist))) {
3039 /* Absolute path specified, but not in ovs_rundir */
3040 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3041 "controller \"%s\" due to possibility for "
3042 "remote exploit. Instead, specify socket "
3043 "in whitelisted \"%s\" or connect to "
3044 "\"unix:%s/%s.mgmt\" (which is always "
3045 "available without special configuration).",
3046 br->name, c->target, whitelist,
3047 ovs_rundir(), br->name);
3052 whitelist = xasprintf("punix:%s/%s.controller",
3053 ovs_rundir(), br->name);
3054 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3055 /* Prevent remote ovsdb-server users from accessing
3056 * arbitrary Unix domain sockets and overwriting arbitrary
3058 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3059 "controller \"%s\" due to possibility of "
3060 "overwriting local files. Instead, specify "
3061 "whitelisted \"%s\" or connect to "
3062 "\"unix:%s/%s.mgmt\" (which is always "
3063 "available without special configuration).",
3064 br->name, c->target, whitelist,
3065 ovs_rundir(), br->name);
3074 bridge_configure_local_iface_netdev(br, c);
3075 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3076 if (disable_in_band) {
3077 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3082 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3083 bridge_get_allowed_versions(br));
3084 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3087 /* Set the fail-mode. */
3088 fail_mode = !br->cfg->fail_mode
3089 || !strcmp(br->cfg->fail_mode, "standalone")
3090 ? OFPROTO_FAIL_STANDALONE
3091 : OFPROTO_FAIL_SECURE;
3092 ofproto_set_fail_mode(br->ofproto, fail_mode);
3094 /* Configure OpenFlow controller connection snooping. */
3095 if (!ofproto_has_snoops(br->ofproto)) {
3099 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3100 ovs_rundir(), br->name));
3101 ofproto_set_snoops(br->ofproto, &snoops);
3102 sset_destroy(&snoops);
3107 bridge_configure_tables(struct bridge *br)
3109 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3113 n_tables = ofproto_get_n_tables(br->ofproto);
3115 for (i = 0; i < n_tables; i++) {
3116 struct ofproto_table_settings s;
3119 s.max_flows = UINT_MAX;
3123 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3124 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3127 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3128 s.max_flows = *cfg->flow_limit;
3130 if (cfg->overflow_policy
3131 && !strcmp(cfg->overflow_policy, "evict")) {
3134 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3135 for (k = 0; k < cfg->n_groups; k++) {
3136 const char *string = cfg->groups[k];
3139 msg = mf_parse_subfield__(&s.groups[k], &string);
3141 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3142 "'groups' (%s)", br->name, i, msg);
3144 } else if (*string) {
3145 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3146 "element '%s' contains trailing garbage",
3147 br->name, i, cfg->groups[k]);
3155 ofproto_configure_table(br->ofproto, i, &s);
3159 for (; j < br->cfg->n_flow_tables; j++) {
3160 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3161 "%"PRId64" not supported by this datapath", br->name,
3162 br->cfg->key_flow_tables[j]);
3167 bridge_configure_dp_desc(struct bridge *br)
3169 ofproto_set_dp_desc(br->ofproto,
3170 smap_get(&br->cfg->other_config, "dp-desc"));
3173 /* Port functions. */
3175 static struct port *
3176 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3180 port = xzalloc(sizeof *port);
3182 port->name = xstrdup(cfg->name);
3184 list_init(&port->ifaces);
3186 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3190 /* Deletes interfaces from 'port' that are no longer configured for it. */
3192 port_del_ifaces(struct port *port)
3194 struct iface *iface, *next;
3195 struct sset new_ifaces;
3198 /* Collect list of new interfaces. */
3199 sset_init(&new_ifaces);
3200 for (i = 0; i < port->cfg->n_interfaces; i++) {
3201 const char *name = port->cfg->interfaces[i]->name;
3202 const char *type = port->cfg->interfaces[i]->type;
3203 if (strcmp(type, "null")) {
3204 sset_add(&new_ifaces, name);
3208 /* Get rid of deleted interfaces. */
3209 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3210 if (!sset_contains(&new_ifaces, iface->name)) {
3211 iface_destroy(iface);
3215 sset_destroy(&new_ifaces);
3219 port_destroy(struct port *port)
3222 struct bridge *br = port->bridge;
3223 struct iface *iface, *next;
3226 ofproto_bundle_unregister(br->ofproto, port);
3229 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3230 iface_destroy(iface);
3233 hmap_remove(&br->ports, &port->hmap_node);
3239 static struct port *
3240 port_lookup(const struct bridge *br, const char *name)
3244 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3246 if (!strcmp(port->name, name)) {
3254 enable_lacp(struct port *port, bool *activep)
3256 if (!port->cfg->lacp) {
3257 /* XXX when LACP implementation has been sufficiently tested, enable by
3258 * default and make active on bonded ports. */
3260 } else if (!strcmp(port->cfg->lacp, "off")) {
3262 } else if (!strcmp(port->cfg->lacp, "active")) {
3265 } else if (!strcmp(port->cfg->lacp, "passive")) {
3269 VLOG_WARN("port %s: unknown LACP mode %s",
3270 port->name, port->cfg->lacp);
3275 static struct lacp_settings *
3276 port_configure_lacp(struct port *port, struct lacp_settings *s)
3278 const char *lacp_time, *system_id;
3281 if (!enable_lacp(port, &s->active)) {
3285 s->name = port->name;
3287 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3289 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3290 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3291 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3292 " address.", port->name, system_id);
3296 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3299 if (eth_addr_is_zero(s->id)) {
3300 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3304 /* Prefer bondable links if unspecified. */
3305 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3307 s->priority = (priority > 0 && priority <= UINT16_MAX
3309 : UINT16_MAX - !list_is_short(&port->ifaces));
3311 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3312 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3317 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3319 int priority, portid, key;
3321 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3322 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3324 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3326 if (portid <= 0 || portid > UINT16_MAX) {
3327 portid = ofp_to_u16(iface->ofp_port);
3330 if (priority <= 0 || priority > UINT16_MAX) {
3331 priority = UINT16_MAX;
3334 if (key < 0 || key > UINT16_MAX) {
3338 s->name = iface->name;
3340 s->priority = priority;
3345 port_configure_bond(struct port *port, struct bond_settings *s)
3347 const char *detect_s;
3348 struct iface *iface;
3349 int miimon_interval;
3351 s->name = port->name;
3353 if (port->cfg->bond_mode) {
3354 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3355 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3356 port->name, port->cfg->bond_mode,
3357 bond_mode_to_string(s->balance));
3360 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3362 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3363 * active-backup. At some point we should remove this warning. */
3364 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3365 " in previous versions, the default bond_mode was"
3366 " balance-slb", port->name,
3367 bond_mode_to_string(s->balance));
3369 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3370 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3371 "please use another bond type or disable flood_vlans",
3375 miimon_interval = smap_get_int(&port->cfg->other_config,
3376 "bond-miimon-interval", 0);
3377 if (miimon_interval <= 0) {
3378 miimon_interval = 200;
3381 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3382 if (!detect_s || !strcmp(detect_s, "carrier")) {
3383 miimon_interval = 0;
3384 } else if (strcmp(detect_s, "miimon")) {
3385 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3386 "defaulting to carrier", port->name, detect_s);
3387 miimon_interval = 0;
3390 s->up_delay = MAX(0, port->cfg->bond_updelay);
3391 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3392 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3393 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3394 "bond-rebalance-interval", 10000);
3395 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3396 s->rebalance_interval = 1000;
3399 s->fake_iface = port->cfg->bond_fake_iface;
3401 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3402 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3406 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3407 * instead of obtaining it from the database. */
3409 port_is_synthetic(const struct port *port)
3411 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3414 /* Interface functions. */
3417 iface_is_internal(const struct ovsrec_interface *iface,
3418 const struct ovsrec_bridge *br)
3420 /* The local port and "internal" ports are always "internal". */
3421 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3424 /* Returns the correct network device type for interface 'iface' in bridge
3427 iface_get_type(const struct ovsrec_interface *iface,
3428 const struct ovsrec_bridge *br)
3432 /* The local port always has type "internal". Other ports take
3433 * their type from the database and default to "system" if none is
3435 if (iface_is_internal(iface, br)) {
3438 type = iface->type[0] ? iface->type : "system";
3441 return ofproto_port_open_type(br->datapath_type, type);
3445 iface_destroy(struct iface *iface)
3448 struct port *port = iface->port;
3449 struct bridge *br = port->bridge;
3451 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3452 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3455 if (iface->ofp_port != OFPP_NONE) {
3456 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3459 list_remove(&iface->port_elem);
3460 hmap_remove(&br->iface_by_name, &iface->name_node);
3462 netdev_close(iface->netdev);
3469 static struct iface *
3470 iface_lookup(const struct bridge *br, const char *name)
3472 struct iface *iface;
3474 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3475 &br->iface_by_name) {
3476 if (!strcmp(iface->name, name)) {
3484 static struct iface *
3485 iface_find(const char *name)
3487 const struct bridge *br;
3489 HMAP_FOR_EACH (br, node, &all_bridges) {
3490 struct iface *iface = iface_lookup(br, name);
3499 static struct if_cfg *
3500 if_cfg_lookup(const struct bridge *br, const char *name)
3502 struct if_cfg *if_cfg;
3504 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3506 if (!strcmp(if_cfg->cfg->name, name)) {
3514 static struct iface *
3515 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3517 struct iface *iface;
3519 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3521 if (iface->ofp_port == ofp_port) {
3528 /* Set Ethernet address of 'iface', if one is specified in the configuration
3531 iface_set_mac(struct iface *iface)
3533 uint8_t ea[ETH_ADDR_LEN];
3535 if (!strcmp(iface->type, "internal")
3536 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3537 if (iface->ofp_port == OFPP_LOCAL) {
3538 VLOG_ERR("interface %s: ignoring mac in Interface record "
3539 "(use Bridge record to set local port's mac)",
3541 } else if (eth_addr_is_multicast(ea)) {
3542 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3545 int error = netdev_set_etheraddr(iface->netdev, ea);
3547 VLOG_ERR("interface %s: setting MAC failed (%s)",
3548 iface->name, ovs_strerror(error));
3554 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3556 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3559 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3560 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3561 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3565 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3566 * sets the "ofport" field to -1.
3568 * This is appropriate when 'if_cfg''s interface cannot be created or is
3569 * otherwise invalid. */
3571 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3573 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3574 ovsrec_interface_set_status(if_cfg, NULL);
3575 ovsrec_interface_set_admin_state(if_cfg, NULL);
3576 ovsrec_interface_set_duplex(if_cfg, NULL);
3577 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3578 ovsrec_interface_set_link_state(if_cfg, NULL);
3579 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3580 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3581 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3582 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3583 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3584 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3585 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3586 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3590 struct iface_delete_queues_cbdata {
3591 struct netdev *netdev;
3592 const struct ovsdb_datum *queues;
3596 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3598 union ovsdb_atom atom;
3600 atom.integer = target;
3601 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3605 iface_delete_queues(unsigned int queue_id,
3606 const struct smap *details OVS_UNUSED, void *cbdata_)
3608 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3610 if (!queue_ids_include(cbdata->queues, queue_id)) {
3611 netdev_delete_queue(cbdata->netdev, queue_id);
3616 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3618 struct ofpbuf queues_buf;
3620 ofpbuf_init(&queues_buf, 0);
3622 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3623 netdev_set_qos(iface->netdev, NULL, NULL);
3625 struct iface_delete_queues_cbdata cbdata;
3629 /* Configure top-level Qos for 'iface'. */
3630 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3632 /* Deconfigure queues that were deleted. */
3633 cbdata.netdev = iface->netdev;
3634 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3636 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3638 /* Configure queues for 'iface'. */
3640 for (i = 0; i < qos->n_queues; i++) {
3641 const struct ovsrec_queue *queue = qos->value_queues[i];
3642 unsigned int queue_id = qos->key_queues[i];
3644 if (queue_id == 0) {
3648 if (queue->n_dscp == 1) {
3649 struct ofproto_port_queue *port_queue;
3651 port_queue = ofpbuf_put_uninit(&queues_buf,
3652 sizeof *port_queue);
3653 port_queue->queue = queue_id;
3654 port_queue->dscp = queue->dscp[0];
3657 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3660 struct smap details;
3662 smap_init(&details);
3663 netdev_set_queue(iface->netdev, 0, &details);
3664 smap_destroy(&details);
3668 if (iface->ofp_port != OFPP_NONE) {
3669 const struct ofproto_port_queue *port_queues = queues_buf.data;
3670 size_t n_queues = queues_buf.size / sizeof *port_queues;
3672 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3673 port_queues, n_queues);
3676 netdev_set_policing(iface->netdev,
3677 iface->cfg->ingress_policing_rate,
3678 iface->cfg->ingress_policing_burst);
3680 ofpbuf_uninit(&queues_buf);
3684 iface_configure_cfm(struct iface *iface)
3686 const struct ovsrec_interface *cfg = iface->cfg;
3687 const char *opstate_str;
3688 const char *cfm_ccm_vlan;
3689 struct cfm_settings s;
3690 struct smap netdev_args;
3692 if (!cfg->n_cfm_mpid) {
3693 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3697 s.check_tnl_key = false;
3698 smap_init(&netdev_args);
3699 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3700 const char *key = smap_get(&netdev_args, "key");
3701 const char *in_key = smap_get(&netdev_args, "in_key");
3703 s.check_tnl_key = (key && !strcmp(key, "flow"))
3704 || (in_key && !strcmp(in_key, "flow"));
3706 smap_destroy(&netdev_args);
3708 s.mpid = *cfg->cfm_mpid;
3709 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3710 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3711 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3713 if (s.interval <= 0) {
3717 if (!cfm_ccm_vlan) {
3719 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3720 s.ccm_vlan = CFM_RANDOM_VLAN;
3722 s.ccm_vlan = atoi(cfm_ccm_vlan);
3723 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3728 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3730 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3732 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3733 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3735 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3738 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3739 * instead of obtaining it from the database. */
3741 iface_is_synthetic(const struct iface *iface)
3743 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3747 iface_pick_ofport(const struct ovsrec_interface *cfg)
3749 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3751 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3756 /* Port mirroring. */
3758 static struct mirror *
3759 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3763 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3764 if (uuid_equals(uuid, &m->uuid)) {
3772 bridge_configure_mirrors(struct bridge *br)
3774 const struct ovsdb_datum *mc;
3775 unsigned long *flood_vlans;
3776 struct mirror *m, *next;
3779 /* Get rid of deleted mirrors. */
3780 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3781 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3782 union ovsdb_atom atom;
3784 atom.uuid = m->uuid;
3785 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3790 /* Add new mirrors and reconfigure existing ones. */
3791 for (i = 0; i < br->cfg->n_mirrors; i++) {
3792 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3793 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3795 m = mirror_create(br, cfg);
3798 if (!mirror_configure(m)) {
3803 /* Update flooded vlans (for RSPAN). */
3804 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3805 br->cfg->n_flood_vlans);
3806 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3807 bitmap_free(flood_vlans);
3810 static struct mirror *
3811 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3815 m = xzalloc(sizeof *m);
3816 m->uuid = cfg->header_.uuid;
3817 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3819 m->name = xstrdup(cfg->name);
3825 mirror_destroy(struct mirror *m)
3828 struct bridge *br = m->bridge;
3831 ofproto_mirror_unregister(br->ofproto, m);
3834 hmap_remove(&br->mirrors, &m->hmap_node);
3841 mirror_collect_ports(struct mirror *m,
3842 struct ovsrec_port **in_ports, int n_in_ports,
3843 void ***out_portsp, size_t *n_out_portsp)
3845 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3846 size_t n_out_ports = 0;
3849 for (i = 0; i < n_in_ports; i++) {
3850 const char *name = in_ports[i]->name;
3851 struct port *port = port_lookup(m->bridge, name);
3853 out_ports[n_out_ports++] = port;
3855 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3856 "port %s", m->bridge->name, m->name, name);
3859 *out_portsp = out_ports;
3860 *n_out_portsp = n_out_ports;
3864 mirror_configure(struct mirror *m)
3866 const struct ovsrec_mirror *cfg = m->cfg;
3867 struct ofproto_mirror_settings s;
3870 if (strcmp(cfg->name, m->name)) {
3872 m->name = xstrdup(cfg->name);
3876 /* Get output port or VLAN. */
3877 if (cfg->output_port) {
3878 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3879 if (!s.out_bundle) {
3880 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3881 m->bridge->name, m->name);
3884 s.out_vlan = UINT16_MAX;
3886 if (cfg->output_vlan) {
3887 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3888 "output vlan; ignoring output vlan",
3889 m->bridge->name, m->name);
3891 } else if (cfg->output_vlan) {
3892 /* The database should prevent invalid VLAN values. */
3893 s.out_bundle = NULL;
3894 s.out_vlan = *cfg->output_vlan;
3896 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3897 m->bridge->name, m->name);
3901 /* Get port selection. */
3902 if (cfg->select_all) {
3903 size_t n_ports = hmap_count(&m->bridge->ports);
3904 void **ports = xmalloc(n_ports * sizeof *ports);
3909 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3919 /* Get ports, dropping ports that don't exist.
3920 * The IDL ensures that there are no duplicates. */
3921 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3922 &s.srcs, &s.n_srcs);
3923 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3924 &s.dsts, &s.n_dsts);
3927 /* Get VLAN selection. */
3928 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3931 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3934 if (s.srcs != s.dsts) {
3943 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3945 * This is deprecated. It is only for compatibility with broken device drivers
3946 * in old versions of Linux that do not properly support VLANs when VLAN
3947 * devices are not used. When broken device drivers are no longer in
3948 * widespread use, we will delete these interfaces. */
3950 static struct ovsrec_port **recs;
3951 static size_t n_recs, allocated_recs;
3953 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3954 * splinters are reconfigured. */
3956 register_rec(struct ovsrec_port *rec)
3958 if (n_recs >= allocated_recs) {
3959 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3961 recs[n_recs++] = rec;
3964 /* Frees all of the ports registered with register_reg(). */
3966 free_registered_recs(void)
3970 for (i = 0; i < n_recs; i++) {
3971 struct ovsrec_port *port = recs[i];
3974 for (j = 0; j < port->n_interfaces; j++) {
3975 struct ovsrec_interface *iface = port->interfaces[j];
3980 smap_destroy(&port->other_config);
3981 free(port->interfaces);
3989 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3992 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3994 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3998 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4001 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4002 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4003 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4006 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4007 * use, returns NULL.
4009 * Updates 'vlan_splinters_enabled_anywhere'. */
4010 static unsigned long int *
4011 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4013 unsigned long int *splinter_vlans;
4014 struct sset splinter_ifaces;
4015 const char *real_dev_name;
4016 struct shash *real_devs;
4017 struct shash_node *node;
4021 /* Free space allocated for synthesized ports and interfaces, since we're
4022 * in the process of reconstructing all of them. */
4023 free_registered_recs();
4025 splinter_vlans = bitmap_allocate(4096);
4026 sset_init(&splinter_ifaces);
4027 vlan_splinters_enabled_anywhere = false;
4028 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4029 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4032 for (j = 0; j < br_cfg->n_ports; j++) {
4033 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4036 for (k = 0; k < port_cfg->n_interfaces; k++) {
4037 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4039 if (vlan_splinters_is_enabled(iface_cfg)) {
4040 vlan_splinters_enabled_anywhere = true;
4041 sset_add(&splinter_ifaces, iface_cfg->name);
4042 vlan_bitmap_from_array__(port_cfg->trunks,
4048 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4049 bitmap_set1(splinter_vlans, *port_cfg->tag);
4054 if (!vlan_splinters_enabled_anywhere) {
4055 free(splinter_vlans);
4056 sset_destroy(&splinter_ifaces);
4060 HMAP_FOR_EACH (br, node, &all_bridges) {
4062 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4066 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4067 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4068 * device to be created for it. */
4069 bitmap_set0(splinter_vlans, 0);
4070 bitmap_set0(splinter_vlans, 4095);
4072 /* Delete all VLAN devices that we don't need. */
4074 real_devs = vlandev_get_real_devs();
4075 SHASH_FOR_EACH (node, real_devs) {
4076 const struct vlan_real_dev *real_dev = node->data;
4077 const struct vlan_dev *vlan_dev;
4078 bool real_dev_has_splinters;
4080 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4082 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4083 if (!real_dev_has_splinters
4084 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4085 struct netdev *netdev;
4087 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4088 if (!netdev_get_in4(netdev, NULL, NULL) ||
4089 !netdev_get_in6(netdev, NULL)) {
4090 vlandev_del(vlan_dev->name);
4092 /* It has an IP address configured, so we don't own
4093 * it. Don't delete it. */
4095 netdev_close(netdev);
4102 /* Add all VLAN devices that we need. */
4103 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4106 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4107 if (!vlandev_get_name(real_dev_name, vid)) {
4108 vlandev_add(real_dev_name, vid);
4115 sset_destroy(&splinter_ifaces);
4117 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4118 free(splinter_vlans);
4121 return splinter_vlans;
4124 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4127 configure_splinter_port(struct port *port)
4129 struct ofproto *ofproto = port->bridge->ofproto;
4130 ofp_port_t realdev_ofp_port;
4131 const char *realdev_name;
4132 struct iface *vlandev, *realdev;
4134 ofproto_bundle_unregister(port->bridge->ofproto, port);
4136 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4139 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4140 realdev = iface_lookup(port->bridge, realdev_name);
4141 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4143 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4147 static struct ovsrec_port *
4148 synthesize_splinter_port(const char *real_dev_name,
4149 const char *vlan_dev_name, int vid)
4151 struct ovsrec_interface *iface;
4152 struct ovsrec_port *port;
4154 iface = xmalloc(sizeof *iface);
4155 ovsrec_interface_init(iface);
4156 iface->name = xstrdup(vlan_dev_name);
4157 iface->type = "system";
4159 port = xmalloc(sizeof *port);
4160 ovsrec_port_init(port);
4161 port->interfaces = xmemdup(&iface, sizeof iface);
4162 port->n_interfaces = 1;
4163 port->name = xstrdup(vlan_dev_name);
4164 port->vlan_mode = "splinter";
4165 port->tag = xmalloc(sizeof *port->tag);
4168 smap_add(&port->other_config, "realdev", real_dev_name);
4174 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4175 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4176 * 1-bit in the 'splinter_vlans' bitmap. */
4178 add_vlan_splinter_ports(struct bridge *br,
4179 const unsigned long int *splinter_vlans,
4180 struct shash *ports)
4184 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4185 * we're modifying 'ports'. */
4186 for (i = 0; i < br->cfg->n_ports; i++) {
4187 const char *name = br->cfg->ports[i]->name;
4188 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4191 for (j = 0; j < port_cfg->n_interfaces; j++) {
4192 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4194 if (vlan_splinters_is_enabled(iface_cfg)) {
4195 const char *real_dev_name;
4198 real_dev_name = iface_cfg->name;
4199 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4200 const char *vlan_dev_name;
4202 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4204 && !shash_find(ports, vlan_dev_name)) {
4205 shash_add(ports, vlan_dev_name,
4206 synthesize_splinter_port(
4207 real_dev_name, vlan_dev_name, vid));
4216 mirror_refresh_stats(struct mirror *m)
4218 struct ofproto *ofproto = m->bridge->ofproto;
4219 uint64_t tx_packets, tx_bytes;
4222 size_t stat_cnt = 0;
4224 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4225 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4229 if (tx_packets != UINT64_MAX) {
4230 keys[stat_cnt] = "tx_packets";
4231 values[stat_cnt] = tx_packets;
4234 if (tx_bytes != UINT64_MAX) {
4235 keys[stat_cnt] = "tx_bytes";
4236 values[stat_cnt] = tx_bytes;
4240 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);