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.
27 #include "dynamic-string.h"
34 #include "mac-learning.h"
35 #include "meta-flow.h"
37 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
69 int64_t ofport; /* Requested OpenFlow port number. */
72 /* OpenFlow port slated for removal from ofproto. */
74 struct list list_node; /* Node in bridge's ofpp_garbage. */
75 uint16_t ofp_port; /* Port to be deleted. */
79 /* These members are always valid. */
80 struct list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct port *port; /* Containing port. */
83 char *name; /* Host network device name. */
85 /* These members are valid only after bridge_reconfigure() causes them to
87 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
88 int ofp_port; /* OpenFlow port number, -1 if unknown. */
89 struct netdev *netdev; /* Network device. */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
131 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
132 Indexed on 'cfg->name'. */
134 /* Port mirroring. */
135 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
137 /* Synthetic local port if necessary. */
138 struct ovsrec_port synth_local_port;
139 struct ovsrec_interface synth_local_iface;
140 struct ovsrec_interface *synth_local_ifacep;
143 /* All bridges, indexed by name. */
144 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
146 /* OVSDB IDL used to obtain configuration. */
147 static struct ovsdb_idl *idl;
149 /* Most recently processed IDL sequence number. */
150 static unsigned int idl_seqno;
152 /* Each time this timer expires, the bridge fetches interface and mirror
153 * statistics and pushes them into the database. */
154 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
155 static long long int iface_stats_timer = LLONG_MIN;
157 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
158 * expensive. This can cause bridge_reconfigure() to take a long time during
159 * which no other work can be done. To deal with this problem, we limit port
160 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
161 * call to bridge_reconfigure(). If there is more work to do after the limit
162 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
163 * This allows the rest of the code to catch up on important things like
164 * forwarding packets. */
165 #define OFP_PORT_ACTION_WINDOW 10
166 static bool reconfiguring = false;
168 static void add_del_bridges(const struct ovsrec_open_vswitch *);
169 static void bridge_update_ofprotos(void);
170 static void bridge_create(const struct ovsrec_bridge *);
171 static void bridge_destroy(struct bridge *);
172 static struct bridge *bridge_lookup(const char *name);
173 static unixctl_cb_func bridge_unixctl_dump_flows;
174 static unixctl_cb_func bridge_unixctl_reconnect;
175 static size_t bridge_get_controllers(const struct bridge *br,
176 struct ovsrec_controller ***controllersp);
177 static void bridge_add_del_ports(struct bridge *,
178 const unsigned long int *splinter_vlans);
179 static void bridge_refresh_ofp_port(struct bridge *);
180 static void bridge_configure_datapath_id(struct bridge *);
181 static void bridge_configure_flow_eviction_threshold(struct bridge *);
182 static void bridge_configure_netflow(struct bridge *);
183 static void bridge_configure_forward_bpdu(struct bridge *);
184 static void bridge_configure_mac_table(struct bridge *);
185 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
186 static void bridge_configure_stp(struct bridge *);
187 static void bridge_configure_tables(struct bridge *);
188 static void bridge_configure_dp_desc(struct bridge *);
189 static void bridge_configure_remotes(struct bridge *,
190 const struct sockaddr_in *managers,
192 static void bridge_pick_local_hw_addr(struct bridge *,
193 uint8_t ea[ETH_ADDR_LEN],
194 struct iface **hw_addr_iface);
195 static uint64_t bridge_pick_datapath_id(struct bridge *,
196 const uint8_t bridge_ea[ETH_ADDR_LEN],
197 struct iface *hw_addr_iface);
198 static void bridge_queue_if_cfg(struct bridge *,
199 const struct ovsrec_interface *,
200 const struct ovsrec_port *);
201 static uint64_t dpid_from_hash(const void *, size_t nbytes);
202 static bool bridge_has_bond_fake_iface(const struct bridge *,
204 static bool port_is_bond_fake_iface(const struct port *);
206 static unixctl_cb_func qos_unixctl_show;
208 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
209 static void port_del_ifaces(struct port *);
210 static void port_destroy(struct port *);
211 static struct port *port_lookup(const struct bridge *, const char *name);
212 static void port_configure(struct port *);
213 static struct lacp_settings *port_configure_lacp(struct port *,
214 struct lacp_settings *);
215 static void port_configure_bond(struct port *, struct bond_settings *);
216 static bool port_is_synthetic(const struct port *);
218 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
219 static void run_system_stats(void);
221 static void bridge_configure_mirrors(struct bridge *);
222 static struct mirror *mirror_create(struct bridge *,
223 const struct ovsrec_mirror *);
224 static void mirror_destroy(struct mirror *);
225 static bool mirror_configure(struct mirror *);
226 static void mirror_refresh_stats(struct mirror *);
228 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
229 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
230 static bool iface_is_internal(const struct ovsrec_interface *iface,
231 const struct ovsrec_bridge *br);
232 static const char *iface_get_type(const struct ovsrec_interface *,
233 const struct ovsrec_bridge *);
234 static void iface_destroy(struct iface *);
235 static struct iface *iface_lookup(const struct bridge *, const char *name);
236 static struct iface *iface_find(const char *name);
237 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
238 static struct iface *iface_from_ofp_port(const struct bridge *,
240 static void iface_set_mac(struct iface *);
241 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
242 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
243 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
244 static void iface_configure_cfm(struct iface *);
245 static void iface_refresh_cfm_stats(struct iface *);
246 static void iface_refresh_stats(struct iface *);
247 static void iface_refresh_status(struct iface *);
248 static bool iface_is_synthetic(const struct iface *);
249 static int64_t iface_pick_ofport(const struct ovsrec_interface *);
251 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
253 * This is deprecated. It is only for compatibility with broken device drivers
254 * in old versions of Linux that do not properly support VLANs when VLAN
255 * devices are not used. When broken device drivers are no longer in
256 * widespread use, we will delete these interfaces. */
258 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
259 static bool vlan_splinters_enabled_anywhere;
261 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
262 static unsigned long int *collect_splinter_vlans(
263 const struct ovsrec_open_vswitch *);
264 static void configure_splinter_port(struct port *);
265 static void add_vlan_splinter_ports(struct bridge *,
266 const unsigned long int *splinter_vlans,
267 struct shash *ports);
270 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
272 struct shash iface_hints;
273 static bool initialized = false;
280 shash_init(&iface_hints);
283 for (i = 0; i < cfg->n_bridges; i++) {
284 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
287 for (j = 0; j < br_cfg->n_ports; j++) {
288 struct ovsrec_port *port_cfg = br_cfg->ports[j];
291 for (k = 0; k < port_cfg->n_interfaces; k++) {
292 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
293 struct iface_hint *iface_hint;
295 iface_hint = xmalloc(sizeof *iface_hint);
296 iface_hint->br_name = br_cfg->name;
297 iface_hint->br_type = br_cfg->datapath_type;
298 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
300 shash_add(&iface_hints, if_cfg->name, iface_hint);
306 ofproto_init(&iface_hints);
308 shash_destroy_free_data(&iface_hints);
312 /* Public functions. */
314 /* Initializes the bridge module, configuring it to obtain its configuration
315 * from an OVSDB server accessed over 'remote', which should be a string in a
316 * form acceptable to ovsdb_idl_create(). */
318 bridge_init(const char *remote)
320 /* Create connection to database. */
321 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
322 idl_seqno = ovsdb_idl_get_seqno(idl);
323 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
324 ovsdb_idl_verify_write_only(idl);
326 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
327 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
328 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
329 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
330 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
331 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
332 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
334 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
335 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
338 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
339 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
340 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
341 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
343 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
344 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
345 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
346 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
347 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
348 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
349 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
350 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
351 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
352 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
353 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
354 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
355 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
356 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
357 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
358 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
359 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
361 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
362 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
363 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
364 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
366 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
368 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
370 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
371 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
373 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
375 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
377 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
378 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
379 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
380 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
381 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
383 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
385 /* Register unixctl commands. */
386 unixctl_command_register("qos/show", "interface", 1, 1,
387 qos_unixctl_show, NULL);
388 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
389 bridge_unixctl_dump_flows, NULL);
390 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
391 bridge_unixctl_reconnect, NULL);
401 struct bridge *br, *next_br;
403 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
406 ovsdb_idl_destroy(idl);
409 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
410 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
411 * responsible for freeing '*managersp' (with free()).
413 * You may be asking yourself "why does ovs-vswitchd care?", because
414 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
415 * should not be and in fact is not directly involved in that. But
416 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
417 * it has to tell in-band control where the managers are to enable that.
418 * (Thus, only managers connected in-band are collected.)
421 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
422 struct sockaddr_in **managersp, size_t *n_managersp)
424 struct sockaddr_in *managers = NULL;
425 size_t n_managers = 0;
429 /* Collect all of the potential targets from the "targets" columns of the
430 * rows pointed to by "manager_options", excluding any that are
433 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
434 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
436 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
437 sset_find_and_delete(&targets, m->target);
439 sset_add(&targets, m->target);
443 /* Now extract the targets' IP addresses. */
444 if (!sset_is_empty(&targets)) {
447 managers = xmalloc(sset_count(&targets) * sizeof *managers);
448 SSET_FOR_EACH (target, &targets) {
449 struct sockaddr_in *sin = &managers[n_managers];
451 if (stream_parse_target_with_default_ports(target,
459 sset_destroy(&targets);
461 *managersp = managers;
462 *n_managersp = n_managers;
466 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
468 unsigned long int *splinter_vlans;
471 COVERAGE_INC(bridge_reconfigure);
473 ovs_assert(!reconfiguring);
474 reconfiguring = true;
476 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
477 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
478 * configuration otherwise.
480 * This is mostly an update to bridge data structures. Nothing is pushed
481 * down to ofproto or lower layers. */
482 add_del_bridges(ovs_cfg);
483 splinter_vlans = collect_splinter_vlans(ovs_cfg);
484 HMAP_FOR_EACH (br, node, &all_bridges) {
485 bridge_add_del_ports(br, splinter_vlans);
487 free(splinter_vlans);
489 /* Delete datapaths that are no longer configured, and create ones which
490 * don't exist but should. */
491 bridge_update_ofprotos();
493 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
494 HMAP_FOR_EACH (br, node, &all_bridges) {
495 bridge_refresh_ofp_port(br);
498 /* Clear database records for "if_cfg"s which haven't been instantiated. */
499 HMAP_FOR_EACH (br, node, &all_bridges) {
500 struct if_cfg *if_cfg;
502 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
503 iface_clear_db_record(if_cfg->cfg);
507 reconfigure_system_stats(ovs_cfg);
511 bridge_reconfigure_ofp(void)
513 long long int deadline;
517 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
519 /* The kernel will reject any attempt to add a given port to a datapath if
520 * that port already belongs to a different datapath, so we must do all
521 * port deletions before any port additions. */
522 HMAP_FOR_EACH (br, node, &all_bridges) {
523 struct ofpp_garbage *garbage, *next;
525 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
526 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
527 * internal datastructures may not be consistent. Eventually, when
528 * port additions and deletions are cheaper, these calls should be
531 ofproto_port_del(br->ofproto, garbage->ofp_port);
532 list_remove(&garbage->list_node);
536 if (time_msec() >= deadline) {
543 HMAP_FOR_EACH (br, node, &all_bridges) {
544 struct if_cfg *if_cfg, *next;
546 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
547 iface_create(br, if_cfg, -1);
549 if (time_msec() >= deadline) {
559 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
561 struct sockaddr_in *managers;
562 int sflow_bridge_number;
567 ovs_assert(reconfiguring);
568 done = bridge_reconfigure_ofp();
570 /* Complete the configuration. */
571 sflow_bridge_number = 0;
572 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
573 HMAP_FOR_EACH (br, node, &all_bridges) {
576 /* We need the datapath ID early to allow LACP ports to use it as the
577 * default system ID. */
578 bridge_configure_datapath_id(br);
580 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
583 port_configure(port);
585 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
586 iface_configure_cfm(iface);
587 iface_configure_qos(iface, port->cfg->qos);
588 iface_set_mac(iface);
591 bridge_configure_mirrors(br);
592 bridge_configure_flow_eviction_threshold(br);
593 bridge_configure_forward_bpdu(br);
594 bridge_configure_mac_table(br);
595 bridge_configure_remotes(br, managers, n_managers);
596 bridge_configure_netflow(br);
597 bridge_configure_sflow(br, &sflow_bridge_number);
598 bridge_configure_stp(br);
599 bridge_configure_tables(br);
600 bridge_configure_dp_desc(br);
605 /* ovs-vswitchd has completed initialization, so allow the process that
606 * forked us to exit successfully. */
607 daemonize_complete();
608 reconfiguring = false;
610 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
616 /* Delete ofprotos which aren't configured or have the wrong type. Create
617 * ofprotos which don't exist but need to. */
619 bridge_update_ofprotos(void)
621 struct bridge *br, *next;
626 /* Delete ofprotos with no bridge or with the wrong type. */
629 ofproto_enumerate_types(&types);
630 SSET_FOR_EACH (type, &types) {
633 ofproto_enumerate_names(type, &names);
634 SSET_FOR_EACH (name, &names) {
635 br = bridge_lookup(name);
636 if (!br || strcmp(type, br->type)) {
637 ofproto_delete(name, type);
641 sset_destroy(&names);
642 sset_destroy(&types);
644 /* Add ofprotos for bridges which don't have one yet. */
645 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
653 /* Remove ports from any datapath with the same name as 'br'. If we
654 * don't do this, creating 'br''s ofproto will fail because a port with
655 * the same name as its local port already exists. */
656 HMAP_FOR_EACH (br2, node, &all_bridges) {
657 struct ofproto_port ofproto_port;
663 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
665 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
667 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
670 ofproto_port_destroy(&ofproto_port);
674 error = ofproto_create(br->name, br->type, &br->ofproto);
676 VLOG_ERR("failed to create bridge %s: %s", br->name,
684 port_configure(struct port *port)
686 const struct ovsrec_port *cfg = port->cfg;
687 struct bond_settings bond_settings;
688 struct lacp_settings lacp_settings;
689 struct ofproto_bundle_settings s;
692 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
693 configure_splinter_port(port);
702 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
703 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
704 s.slaves[s.n_slaves++] = iface->ofp_port;
709 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
713 /* Get VLAN trunks. */
716 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
720 if (cfg->vlan_mode) {
721 if (!strcmp(cfg->vlan_mode, "access")) {
722 s.vlan_mode = PORT_VLAN_ACCESS;
723 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
724 s.vlan_mode = PORT_VLAN_TRUNK;
725 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
726 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
727 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
728 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
730 /* This "can't happen" because ovsdb-server should prevent it. */
731 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
732 s.vlan_mode = PORT_VLAN_TRUNK;
736 s.vlan_mode = PORT_VLAN_ACCESS;
738 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
742 s.vlan_mode = PORT_VLAN_TRUNK;
745 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
748 /* Get LACP settings. */
749 s.lacp = port_configure_lacp(port, &lacp_settings);
753 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
754 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
755 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
758 s.lacp_slaves = NULL;
761 /* Get bond settings. */
762 if (s.n_slaves > 1) {
763 s.bond = &bond_settings;
764 port_configure_bond(port, &bond_settings);
767 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
768 netdev_set_miimon_interval(iface->netdev, 0);
773 ofproto_bundle_register(port->bridge->ofproto, port, &s);
781 /* Pick local port hardware address and datapath ID for 'br'. */
783 bridge_configure_datapath_id(struct bridge *br)
785 uint8_t ea[ETH_ADDR_LEN];
787 struct iface *local_iface;
788 struct iface *hw_addr_iface;
791 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
792 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
794 int error = netdev_set_etheraddr(local_iface->netdev, ea);
796 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
797 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
798 "Ethernet address: %s",
799 br->name, strerror(error));
802 memcpy(br->ea, ea, ETH_ADDR_LEN);
804 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
805 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
806 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
807 ofproto_set_datapath_id(br->ofproto, dpid);
810 dpid_string = xasprintf("%016"PRIx64, dpid);
811 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
815 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
818 bridge_get_allowed_versions(struct bridge *br)
820 if (!br->cfg->n_protocols)
823 return ofputil_versions_from_strings(br->cfg->protocols,
824 br->cfg->n_protocols);
827 /* Set NetFlow configuration on 'br'. */
829 bridge_configure_netflow(struct bridge *br)
831 struct ovsrec_netflow *cfg = br->cfg->netflow;
832 struct netflow_options opts;
835 ofproto_set_netflow(br->ofproto, NULL);
839 memset(&opts, 0, sizeof opts);
841 /* Get default NetFlow configuration from datapath.
842 * Apply overrides from 'cfg'. */
843 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
844 if (cfg->engine_type) {
845 opts.engine_type = *cfg->engine_type;
847 if (cfg->engine_id) {
848 opts.engine_id = *cfg->engine_id;
851 /* Configure active timeout interval. */
852 opts.active_timeout = cfg->active_timeout;
853 if (!opts.active_timeout) {
854 opts.active_timeout = -1;
855 } else if (opts.active_timeout < 0) {
856 VLOG_WARN("bridge %s: active timeout interval set to negative "
857 "value, using default instead (%d seconds)", br->name,
858 NF_ACTIVE_TIMEOUT_DEFAULT);
859 opts.active_timeout = -1;
862 /* Add engine ID to interface number to disambiguate bridgs? */
863 opts.add_id_to_iface = cfg->add_id_to_interface;
864 if (opts.add_id_to_iface) {
865 if (opts.engine_id > 0x7f) {
866 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
867 "another vswitch, choose an engine id less than 128",
870 if (hmap_count(&br->ports) > 508) {
871 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
872 "another port when more than 508 ports are used",
878 sset_init(&opts.collectors);
879 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
882 if (ofproto_set_netflow(br->ofproto, &opts)) {
883 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
885 sset_destroy(&opts.collectors);
888 /* Set sFlow configuration on 'br'. */
890 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
892 const struct ovsrec_sflow *cfg = br->cfg->sflow;
893 struct ovsrec_controller **controllers;
894 struct ofproto_sflow_options oso;
895 size_t n_controllers;
899 ofproto_set_sflow(br->ofproto, NULL);
903 memset(&oso, 0, sizeof oso);
905 sset_init(&oso.targets);
906 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
908 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
910 oso.sampling_rate = *cfg->sampling;
913 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
915 oso.polling_interval = *cfg->polling;
918 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
920 oso.header_len = *cfg->header;
923 oso.sub_id = (*sflow_bridge_number)++;
924 oso.agent_device = cfg->agent;
926 oso.control_ip = NULL;
927 n_controllers = bridge_get_controllers(br, &controllers);
928 for (i = 0; i < n_controllers; i++) {
929 if (controllers[i]->local_ip) {
930 oso.control_ip = controllers[i]->local_ip;
934 ofproto_set_sflow(br->ofproto, &oso);
936 sset_destroy(&oso.targets);
940 port_configure_stp(const struct ofproto *ofproto, struct port *port,
941 struct ofproto_port_stp_settings *port_s,
942 int *port_num_counter, unsigned long *port_num_bitmap)
944 const char *config_str;
947 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
948 port_s->enable = false;
951 port_s->enable = true;
954 /* STP over bonds is not supported. */
955 if (!list_is_singleton(&port->ifaces)) {
956 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
958 port_s->enable = false;
962 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
964 /* Internal ports shouldn't participate in spanning tree, so
966 if (!strcmp(iface->type, "internal")) {
967 VLOG_DBG("port %s: disable STP on internal ports", port->name);
968 port_s->enable = false;
972 /* STP on mirror output ports is not supported. */
973 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
974 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
975 port_s->enable = false;
979 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
981 unsigned long int port_num = strtoul(config_str, NULL, 0);
982 int port_idx = port_num - 1;
984 if (port_num < 1 || port_num > STP_MAX_PORTS) {
985 VLOG_ERR("port %s: invalid stp-port-num", port->name);
986 port_s->enable = false;
990 if (bitmap_is_set(port_num_bitmap, port_idx)) {
991 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
992 port->name, port_num);
993 port_s->enable = false;
996 bitmap_set1(port_num_bitmap, port_idx);
997 port_s->port_num = port_idx;
999 if (*port_num_counter >= STP_MAX_PORTS) {
1000 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1001 port_s->enable = false;
1005 port_s->port_num = (*port_num_counter)++;
1008 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1010 port_s->path_cost = strtoul(config_str, NULL, 10);
1012 enum netdev_features current;
1015 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1016 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1017 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1020 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1022 port_s->priority = strtoul(config_str, NULL, 0);
1024 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1028 /* Set spanning tree configuration on 'br'. */
1030 bridge_configure_stp(struct bridge *br)
1032 if (!br->cfg->stp_enable) {
1033 ofproto_set_stp(br->ofproto, NULL);
1035 struct ofproto_stp_settings br_s;
1036 const char *config_str;
1038 int port_num_counter;
1039 unsigned long *port_num_bitmap;
1041 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1043 uint8_t ea[ETH_ADDR_LEN];
1045 if (eth_addr_from_string(config_str, ea)) {
1046 br_s.system_id = eth_addr_to_uint64(ea);
1048 br_s.system_id = eth_addr_to_uint64(br->ea);
1049 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1050 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1053 br_s.system_id = eth_addr_to_uint64(br->ea);
1056 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1058 br_s.priority = strtoul(config_str, NULL, 0);
1060 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1063 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1065 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1067 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1070 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1072 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1074 br_s.max_age = STP_DEFAULT_MAX_AGE;
1077 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1079 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1081 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1084 /* Configure STP on the bridge. */
1085 if (ofproto_set_stp(br->ofproto, &br_s)) {
1086 VLOG_ERR("bridge %s: could not enable STP", br->name);
1090 /* Users must either set the port number with the "stp-port-num"
1091 * configuration on all ports or none. If manual configuration
1092 * is not done, then we allocate them sequentially. */
1093 port_num_counter = 0;
1094 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1095 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1096 struct ofproto_port_stp_settings port_s;
1097 struct iface *iface;
1099 port_configure_stp(br->ofproto, port, &port_s,
1100 &port_num_counter, port_num_bitmap);
1102 /* As bonds are not supported, just apply configuration to
1103 * all interfaces. */
1104 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1105 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1107 VLOG_ERR("port %s: could not enable STP", port->name);
1113 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1114 && port_num_counter) {
1115 VLOG_ERR("bridge %s: must manually configure all STP port "
1116 "IDs or none, disabling", br->name);
1117 ofproto_set_stp(br->ofproto, NULL);
1119 bitmap_free(port_num_bitmap);
1124 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1126 const struct port *port = port_lookup(br, name);
1127 return port && port_is_bond_fake_iface(port);
1131 port_is_bond_fake_iface(const struct port *port)
1133 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1137 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1139 struct bridge *br, *next;
1140 struct shash new_br;
1143 /* Collect new bridges' names and types. */
1144 shash_init(&new_br);
1145 for (i = 0; i < cfg->n_bridges; i++) {
1146 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1147 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1149 if (strchr(br_cfg->name, '/')) {
1150 /* Prevent remote ovsdb-server users from accessing arbitrary
1151 * directories, e.g. consider a bridge named "../../../etc/". */
1152 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1154 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1155 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1159 /* Get rid of deleted bridges or those whose types have changed.
1160 * Update 'cfg' of bridges that still exist. */
1161 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1162 br->cfg = shash_find_data(&new_br, br->name);
1163 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1164 br->cfg->datapath_type))) {
1169 /* Add new bridges. */
1170 for (i = 0; i < cfg->n_bridges; i++) {
1171 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1172 struct bridge *br = bridge_lookup(br_cfg->name);
1174 bridge_create(br_cfg);
1178 shash_destroy(&new_br);
1182 iface_set_ofp_port(struct iface *iface, int ofp_port)
1184 struct bridge *br = iface->port->bridge;
1186 ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
1187 iface->ofp_port = ofp_port;
1188 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1189 iface_set_ofport(iface->cfg, ofp_port);
1192 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1193 * Returns 0 if successful, otherwise a positive errno value. */
1195 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1196 struct netdev *netdev)
1200 error = netdev_set_config(netdev, &iface_cfg->options);
1202 VLOG_WARN("could not configure network device %s (%s)",
1203 iface_cfg->name, strerror(error));
1208 /* This function determines whether 'ofproto_port', which is attached to
1209 * br->ofproto's datapath, is one that we want in 'br'.
1211 * If it is, it returns true, after creating an iface (if necessary),
1212 * configuring the iface's netdev according to the iface's options, and setting
1213 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1215 * If, on the other hand, 'port' should be removed, it returns false. The
1216 * caller should later detach the port from br->ofproto. */
1218 bridge_refresh_one_ofp_port(struct bridge *br,
1219 const struct ofproto_port *ofproto_port)
1221 const char *name = ofproto_port->name;
1222 const char *type = ofproto_port->type;
1223 uint16_t ofp_port = ofproto_port->ofp_port;
1225 struct iface *iface = iface_lookup(br, name);
1227 /* Check that the name-to-number mapping is one-to-one. */
1228 if (iface->ofp_port >= 0) {
1229 VLOG_WARN("bridge %s: interface %s reported twice",
1232 } else if (iface_from_ofp_port(br, ofp_port)) {
1233 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1234 br->name, ofp_port);
1238 /* There's a configured interface named 'name'. */
1239 if (strcmp(type, iface->type)
1240 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1241 /* It's the wrong type, or it's the right type but can't be
1242 * configured as the user requested, so we must destroy it. */
1245 /* It's the right type and configured correctly. Keep it. */
1246 iface_set_ofp_port(iface, ofp_port);
1249 } else if (bridge_has_bond_fake_iface(br, name)
1250 && !strcmp(type, "internal")) {
1251 /* It's a bond fake iface. Keep it. */
1254 /* There's no configured interface named 'name', but there might be an
1255 * interface of that name queued to be created.
1257 * If there is, and it has the correct type, then try to configure it
1258 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1259 * delete it and later try to re-add it. */
1260 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1262 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1263 && iface_create(br, if_cfg, ofp_port));
1267 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1268 * consistent with the ofp_ports in "br->ofproto". */
1270 bridge_refresh_ofp_port(struct bridge *br)
1272 struct ofproto_port_dump dump;
1273 struct ofproto_port ofproto_port;
1274 struct port *port, *port_next;
1276 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1277 hmap_clear(&br->ifaces);
1278 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1279 struct iface *iface;
1281 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1282 iface->ofp_port = -1;
1286 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1287 * already exist in the datapath and promote them to full fledged "struct
1288 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1289 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1290 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1291 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1292 garbage->ofp_port = ofproto_port.ofp_port;
1293 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1297 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1298 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1299 * later they can be added to ofproto. */
1300 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1301 struct iface *iface, *iface_next;
1303 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1304 if (iface->ofp_port < 0) {
1305 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1306 iface_destroy(iface);
1310 if (list_is_empty(&port->ifaces)) {
1316 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1317 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1318 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1321 * If successful, returns 0 and stores the network device in '*netdevp'. On
1322 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1324 iface_do_create(const struct bridge *br,
1325 const struct if_cfg *if_cfg,
1326 int *ofp_portp, struct netdev **netdevp)
1328 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1329 const struct ovsrec_port *port_cfg = if_cfg->parent;
1330 struct netdev *netdev;
1333 error = netdev_open(iface_cfg->name,
1334 iface_get_type(iface_cfg, br->cfg), &netdev);
1336 VLOG_WARN("could not open network device %s (%s)",
1337 iface_cfg->name, strerror(error));
1341 error = iface_set_netdev_config(iface_cfg, netdev);
1346 if (*ofp_portp < 0) {
1347 uint16_t ofp_port = if_cfg->ofport;
1349 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1353 *ofp_portp = ofp_port;
1355 VLOG_INFO("bridge %s: added interface %s on port %d",
1356 br->name, iface_cfg->name, *ofp_portp);
1358 VLOG_DBG("bridge %s: interface %s is on port %d",
1359 br->name, iface_cfg->name, *ofp_portp);
1362 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1363 || iface_is_internal(iface_cfg, br->cfg)) {
1364 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1372 netdev_close(netdev);
1376 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1377 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1378 * automatically allocated for the iface. Takes ownership of and
1379 * deallocates 'if_cfg'.
1381 * Return true if an iface is successfully created, false otherwise. */
1383 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1385 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1386 const struct ovsrec_port *port_cfg = if_cfg->parent;
1388 struct netdev *netdev;
1389 struct iface *iface;
1394 /* Do the bits that can fail up front.
1396 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1397 * internal datastructures may not be consistent. Eventually, when port
1398 * additions and deletions are cheaper, these calls should be removed. */
1400 ovs_assert(!iface_lookup(br, iface_cfg->name));
1401 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1404 iface_set_ofport(iface_cfg, -1);
1405 iface_clear_db_record(iface_cfg);
1410 /* Get or create the port structure. */
1411 port = port_lookup(br, port_cfg->name);
1413 port = port_create(br, port_cfg);
1416 /* Create the iface structure. */
1417 iface = xzalloc(sizeof *iface);
1418 list_push_back(&port->ifaces, &iface->port_elem);
1419 hmap_insert(&br->iface_by_name, &iface->name_node,
1420 hash_string(iface_cfg->name, 0));
1422 iface->name = xstrdup(iface_cfg->name);
1423 iface->ofp_port = -1;
1424 iface->netdev = netdev;
1425 iface->type = iface_get_type(iface_cfg, br->cfg);
1426 iface->cfg = iface_cfg;
1428 iface_set_ofp_port(iface, ofp_port);
1430 /* Populate initial status in database. */
1431 iface_refresh_stats(iface);
1432 iface_refresh_status(iface);
1434 /* Add bond fake iface if necessary. */
1435 if (port_is_bond_fake_iface(port)) {
1436 struct ofproto_port ofproto_port;
1438 if (ofproto_port_query_by_name(br->ofproto, port->name,
1440 struct netdev *netdev;
1443 error = netdev_open(port->name, "internal", &netdev);
1445 uint16_t fake_ofp_port = if_cfg->ofport;
1447 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1448 netdev_close(netdev);
1450 VLOG_WARN("could not open network device %s (%s)",
1451 port->name, strerror(error));
1454 /* Already exists, nothing to do. */
1455 ofproto_port_destroy(&ofproto_port);
1460 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1466 /* Set Flow eviction threshold */
1468 bridge_configure_flow_eviction_threshold(struct bridge *br)
1470 const char *threshold_str;
1473 threshold_str = smap_get(&br->cfg->other_config,
1474 "flow-eviction-threshold");
1475 if (threshold_str) {
1476 threshold = strtoul(threshold_str, NULL, 10);
1478 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1480 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1483 /* Set forward BPDU option. */
1485 bridge_configure_forward_bpdu(struct bridge *br)
1487 ofproto_set_forward_bpdu(br->ofproto,
1488 smap_get_bool(&br->cfg->other_config,
1493 /* Set MAC learning table configuration for 'br'. */
1495 bridge_configure_mac_table(struct bridge *br)
1497 const char *idle_time_str;
1500 const char *mac_table_size_str;
1503 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1504 idle_time = (idle_time_str && atoi(idle_time_str)
1505 ? atoi(idle_time_str)
1506 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1508 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1509 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1510 ? atoi(mac_table_size_str)
1513 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1517 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1518 struct iface **hw_addr_iface)
1520 struct hmapx mirror_output_ports;
1523 bool found_addr = false;
1527 *hw_addr_iface = NULL;
1529 /* Did the user request a particular MAC? */
1530 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1531 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1532 if (eth_addr_is_multicast(ea)) {
1533 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1534 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1535 } else if (eth_addr_is_zero(ea)) {
1536 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1542 /* Mirror output ports don't participate in picking the local hardware
1543 * address. ofproto can't help us find out whether a given port is a
1544 * mirror output because we haven't configured mirrors yet, so we need to
1545 * accumulate them ourselves. */
1546 hmapx_init(&mirror_output_ports);
1547 for (i = 0; i < br->cfg->n_mirrors; i++) {
1548 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1549 if (m->output_port) {
1550 hmapx_add(&mirror_output_ports, m->output_port);
1554 /* Otherwise choose the minimum non-local MAC address among all of the
1556 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1557 uint8_t iface_ea[ETH_ADDR_LEN];
1558 struct iface *candidate;
1559 struct iface *iface;
1561 /* Mirror output ports don't participate. */
1562 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1566 /* Choose the MAC address to represent the port. */
1568 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1569 /* Find the interface with this Ethernet address (if any) so that
1570 * we can provide the correct devname to the caller. */
1571 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1572 uint8_t candidate_ea[ETH_ADDR_LEN];
1573 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1574 && eth_addr_equals(iface_ea, candidate_ea)) {
1579 /* Choose the interface whose MAC address will represent the port.
1580 * The Linux kernel bonding code always chooses the MAC address of
1581 * the first slave added to a bond, and the Fedora networking
1582 * scripts always add slaves to a bond in alphabetical order, so
1583 * for compatibility we choose the interface with the name that is
1584 * first in alphabetical order. */
1585 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1586 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1591 /* The local port doesn't count (since we're trying to choose its
1592 * MAC address anyway). */
1593 if (iface->ofp_port == OFPP_LOCAL) {
1598 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1604 /* Compare against our current choice. */
1605 if (!eth_addr_is_multicast(iface_ea) &&
1606 !eth_addr_is_local(iface_ea) &&
1607 !eth_addr_is_reserved(iface_ea) &&
1608 !eth_addr_is_zero(iface_ea) &&
1609 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1611 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1612 *hw_addr_iface = iface;
1618 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1619 *hw_addr_iface = NULL;
1622 hmapx_destroy(&mirror_output_ports);
1625 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1626 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1627 * an interface on 'br', then that interface must be passed in as
1628 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1629 * 'hw_addr_iface' must be passed in as a null pointer. */
1631 bridge_pick_datapath_id(struct bridge *br,
1632 const uint8_t bridge_ea[ETH_ADDR_LEN],
1633 struct iface *hw_addr_iface)
1636 * The procedure for choosing a bridge MAC address will, in the most
1637 * ordinary case, also choose a unique MAC that we can use as a datapath
1638 * ID. In some special cases, though, multiple bridges will end up with
1639 * the same MAC address. This is OK for the bridges, but it will confuse
1640 * the OpenFlow controller, because each datapath needs a unique datapath
1643 * Datapath IDs must be unique. It is also very desirable that they be
1644 * stable from one run to the next, so that policy set on a datapath
1647 const char *datapath_id;
1650 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1651 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1655 if (!hw_addr_iface) {
1657 * A purely internal bridge, that is, one that has no non-virtual
1658 * network devices on it at all, is difficult because it has no
1659 * natural unique identifier at all.
1661 * When the host is a XenServer, we handle this case by hashing the
1662 * host's UUID with the name of the bridge. Names of bridges are
1663 * persistent across XenServer reboots, although they can be reused if
1664 * an internal network is destroyed and then a new one is later
1665 * created, so this is fairly effective.
1667 * When the host is not a XenServer, we punt by using a random MAC
1668 * address on each run.
1670 const char *host_uuid = xenserver_get_host_uuid();
1672 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1673 dpid = dpid_from_hash(combined, strlen(combined));
1679 return eth_addr_to_uint64(bridge_ea);
1683 dpid_from_hash(const void *data, size_t n)
1685 uint8_t hash[SHA1_DIGEST_SIZE];
1687 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1688 sha1_bytes(data, n, hash);
1689 eth_addr_mark_random(hash);
1690 return eth_addr_to_uint64(hash);
1694 iface_refresh_status(struct iface *iface)
1698 enum netdev_features current;
1702 uint8_t mac[ETH_ADDR_LEN];
1705 if (iface_is_synthetic(iface)) {
1711 if (!netdev_get_status(iface->netdev, &smap)) {
1712 ovsrec_interface_set_status(iface->cfg, &smap);
1714 ovsrec_interface_set_status(iface->cfg, NULL);
1717 smap_destroy(&smap);
1719 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1720 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1722 ovsrec_interface_set_duplex(iface->cfg,
1723 netdev_features_is_full_duplex(current)
1725 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1727 ovsrec_interface_set_duplex(iface->cfg, NULL);
1728 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1731 error = netdev_get_mtu(iface->netdev, &mtu);
1734 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1736 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1739 error = netdev_get_etheraddr(iface->netdev, mac);
1741 char mac_string[32];
1743 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1744 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1746 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1750 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1753 iface_refresh_cfm_stats(struct iface *iface)
1755 const struct ovsrec_interface *cfg = iface->cfg;
1756 struct ofproto_cfm_status status;
1758 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1759 iface->ofp_port, &status)) {
1760 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1761 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1762 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1763 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1764 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1766 const char *reasons[CFM_FAULT_N_REASONS];
1767 int64_t cfm_health = status.health;
1768 bool faulted = status.faults != 0;
1771 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1774 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1775 int reason = 1 << i;
1776 if (status.faults & reason) {
1777 reasons[j++] = cfm_fault_reason_to_str(reason);
1780 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1782 if (status.remote_opstate >= 0) {
1783 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1784 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1786 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1789 ovsrec_interface_set_cfm_remote_mpids(cfg,
1790 (const int64_t *)status.rmps,
1792 if (cfm_health >= 0) {
1793 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1795 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1801 iface_refresh_stats(struct iface *iface)
1803 #define IFACE_STATS \
1804 IFACE_STAT(rx_packets, "rx_packets") \
1805 IFACE_STAT(tx_packets, "tx_packets") \
1806 IFACE_STAT(rx_bytes, "rx_bytes") \
1807 IFACE_STAT(tx_bytes, "tx_bytes") \
1808 IFACE_STAT(rx_dropped, "rx_dropped") \
1809 IFACE_STAT(tx_dropped, "tx_dropped") \
1810 IFACE_STAT(rx_errors, "rx_errors") \
1811 IFACE_STAT(tx_errors, "tx_errors") \
1812 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1813 IFACE_STAT(rx_over_errors, "rx_over_err") \
1814 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1815 IFACE_STAT(collisions, "collisions")
1817 #define IFACE_STAT(MEMBER, NAME) NAME,
1818 static char *keys[] = { IFACE_STATS };
1820 int64_t values[ARRAY_SIZE(keys)];
1823 struct netdev_stats stats;
1825 if (iface_is_synthetic(iface)) {
1829 /* Intentionally ignore return value, since errors will set 'stats' to
1830 * all-1s, and we will deal with that correctly below. */
1831 netdev_get_stats(iface->netdev, &stats);
1833 /* Copy statistics into values[] array. */
1835 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1838 ovs_assert(i == ARRAY_SIZE(keys));
1840 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1846 br_refresh_stp_status(struct bridge *br)
1848 struct smap smap = SMAP_INITIALIZER(&smap);
1849 struct ofproto *ofproto = br->ofproto;
1850 struct ofproto_stp_status status;
1852 if (ofproto_get_stp_status(ofproto, &status)) {
1856 if (!status.enabled) {
1857 ovsrec_bridge_set_status(br->cfg, NULL);
1861 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1862 STP_ID_ARGS(status.bridge_id));
1863 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1864 STP_ID_ARGS(status.designated_root));
1865 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1867 ovsrec_bridge_set_status(br->cfg, &smap);
1868 smap_destroy(&smap);
1872 port_refresh_stp_status(struct port *port)
1874 struct ofproto *ofproto = port->bridge->ofproto;
1875 struct iface *iface;
1876 struct ofproto_port_stp_status status;
1878 int64_t int_values[3];
1881 if (port_is_synthetic(port)) {
1885 /* STP doesn't currently support bonds. */
1886 if (!list_is_singleton(&port->ifaces)) {
1887 ovsrec_port_set_status(port->cfg, NULL);
1891 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1893 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1897 if (!status.enabled) {
1898 ovsrec_port_set_status(port->cfg, NULL);
1899 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1903 /* Set Status column. */
1905 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1906 smap_add(&smap, "stp_state", stp_state_name(status.state));
1907 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1908 smap_add(&smap, "stp_role", stp_role_name(status.role));
1909 ovsrec_port_set_status(port->cfg, &smap);
1910 smap_destroy(&smap);
1912 /* Set Statistics column. */
1913 keys[0] = "stp_tx_count";
1914 int_values[0] = status.tx_count;
1915 keys[1] = "stp_rx_count";
1916 int_values[1] = status.rx_count;
1917 keys[2] = "stp_error_count";
1918 int_values[2] = status.error_count;
1920 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1921 ARRAY_SIZE(int_values));
1925 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1927 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1931 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1933 bool enable = enable_system_stats(cfg);
1935 system_stats_enable(enable);
1937 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1942 run_system_stats(void)
1944 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1947 stats = system_stats_run();
1949 struct ovsdb_idl_txn *txn;
1950 struct ovsdb_datum datum;
1952 txn = ovsdb_idl_txn_create(idl);
1953 ovsdb_datum_from_smap(&datum, stats);
1954 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1956 ovsdb_idl_txn_commit(txn);
1957 ovsdb_idl_txn_destroy(txn);
1963 static inline const char *
1964 ofp12_controller_role_to_str(enum ofp12_controller_role role)
1967 case OFPCR12_ROLE_EQUAL:
1969 case OFPCR12_ROLE_MASTER:
1971 case OFPCR12_ROLE_SLAVE:
1973 case OFPCR12_ROLE_NOCHANGE:
1975 return "*** INVALID ROLE ***";
1980 refresh_controller_status(void)
1984 const struct ovsrec_controller *cfg;
1988 /* Accumulate status for controllers on all bridges. */
1989 HMAP_FOR_EACH (br, node, &all_bridges) {
1990 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1993 /* Update each controller in the database with current status. */
1994 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1995 struct ofproto_controller_info *cinfo =
1996 shash_find_data(&info, cfg->target);
1999 struct smap smap = SMAP_INITIALIZER(&smap);
2000 const char **values = cinfo->pairs.values;
2001 const char **keys = cinfo->pairs.keys;
2004 for (i = 0; i < cinfo->pairs.n; i++) {
2005 smap_add(&smap, keys[i], values[i]);
2008 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2009 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2011 ovsrec_controller_set_status(cfg, &smap);
2012 smap_destroy(&smap);
2014 ovsrec_controller_set_is_connected(cfg, false);
2015 ovsrec_controller_set_role(cfg, NULL);
2016 ovsrec_controller_set_status(cfg, NULL);
2020 ofproto_free_ofproto_controller_info(&info);
2025 * Some information in the database must be kept as up-to-date as possible to
2026 * allow controllers to respond rapidly to network outages. We call these
2027 * statistics "instant" stats.
2029 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2030 * assuming that they've changed. The only means we have to determine whether
2031 * they have changed are:
2033 * - Try to commit changes to the database. If nothing changed, then
2034 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2037 * - instant_stats_run() is called late in the run loop, after anything that
2038 * might change any of the instant stats.
2040 * We use these two facts together to avoid waking the process up every
2041 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2044 /* Minimum interval between writing updates to the instant stats to the
2046 #define INSTANT_INTERVAL_MSEC 100
2048 /* Current instant stats database transaction, NULL if there is no ongoing
2050 static struct ovsdb_idl_txn *instant_txn;
2052 /* Next time (in msec on monotonic clock) at which we will update the instant
2054 static long long int instant_next_txn = LLONG_MIN;
2056 /* True if the run loop has run since we last saw that the instant stats were
2057 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2058 * to refresh the instant stats. */
2059 static bool instant_stats_could_have_changed;
2062 instant_stats_run(void)
2064 enum ovsdb_idl_txn_status status;
2066 instant_stats_could_have_changed = true;
2071 if (time_msec() < instant_next_txn) {
2074 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2076 instant_txn = ovsdb_idl_txn_create(idl);
2077 HMAP_FOR_EACH (br, node, &all_bridges) {
2078 struct iface *iface;
2081 br_refresh_stp_status(br);
2083 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2084 port_refresh_stp_status(port);
2087 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2088 enum netdev_flags flags;
2089 const char *link_state;
2090 int64_t link_resets;
2093 if (iface_is_synthetic(iface)) {
2097 current = ofproto_port_is_lacp_current(br->ofproto,
2101 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2103 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2106 error = netdev_get_flags(iface->netdev, &flags);
2108 const char *state = flags & NETDEV_UP ? "up" : "down";
2109 ovsrec_interface_set_admin_state(iface->cfg, state);
2111 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2114 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2115 ovsrec_interface_set_link_state(iface->cfg, link_state);
2117 link_resets = netdev_get_carrier_resets(iface->netdev);
2118 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2120 iface_refresh_cfm_stats(iface);
2125 status = ovsdb_idl_txn_commit(instant_txn);
2126 if (status != TXN_INCOMPLETE) {
2127 ovsdb_idl_txn_destroy(instant_txn);
2130 if (status == TXN_UNCHANGED) {
2131 instant_stats_could_have_changed = false;
2136 instant_stats_wait(void)
2139 ovsdb_idl_txn_wait(instant_txn);
2140 } else if (instant_stats_could_have_changed) {
2141 poll_timer_wait_until(instant_next_txn);
2145 /* Performs periodic activity required by bridges that needs to be done with
2146 * the least possible latency.
2148 * It makes sense to call this function a couple of times per poll loop, to
2149 * provide a significant performance boost on some benchmarks with ofprotos
2150 * that use the ofproto-dpif implementation. */
2152 bridge_run_fast(void)
2159 ofproto_enumerate_types(&types);
2160 SSET_FOR_EACH (type, &types) {
2161 ofproto_type_run_fast(type);
2163 sset_destroy(&types);
2165 HMAP_FOR_EACH (br, node, &all_bridges) {
2166 ofproto_run_fast(br->ofproto);
2173 static struct ovsrec_open_vswitch null_cfg;
2174 const struct ovsrec_open_vswitch *cfg;
2175 struct ovsdb_idl_txn *reconf_txn = NULL;
2179 bool vlan_splinters_changed;
2182 ovsrec_open_vswitch_init(&null_cfg);
2184 /* (Re)configure if necessary. */
2185 if (!reconfiguring) {
2188 if (ovsdb_idl_is_lock_contended(idl)) {
2189 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2190 struct bridge *br, *next_br;
2192 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2193 "disabling this process until it goes away");
2195 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2199 } else if (!ovsdb_idl_has_lock(idl)) {
2203 cfg = ovsrec_open_vswitch_first(idl);
2205 /* Initialize the ofproto library. This only needs to run once, but
2206 * it must be done after the configuration is set. If the
2207 * initialization has already occurred, bridge_init_ofproto()
2208 * returns immediately. */
2209 bridge_init_ofproto(cfg);
2211 /* Let each datapath type do the work that it needs to do. */
2213 ofproto_enumerate_types(&types);
2214 SSET_FOR_EACH (type, &types) {
2215 ofproto_type_run(type);
2217 sset_destroy(&types);
2219 /* Let each bridge do the work that it needs to do. */
2220 HMAP_FOR_EACH (br, node, &all_bridges) {
2221 ofproto_run(br->ofproto);
2224 /* Re-configure SSL. We do this on every trip through the main loop,
2225 * instead of just when the database changes, because the contents of the
2226 * key and certificate files can change without the database changing.
2228 * We do this before bridge_reconfigure() because that function might
2229 * initiate SSL connections and thus requires SSL to be configured. */
2230 if (cfg && cfg->ssl) {
2231 const struct ovsrec_ssl *ssl = cfg->ssl;
2233 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2234 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2237 if (!reconfiguring) {
2238 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2239 * usage has changed. */
2240 vlan_splinters_changed = false;
2241 if (vlan_splinters_enabled_anywhere) {
2242 HMAP_FOR_EACH (br, node, &all_bridges) {
2243 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2244 vlan_splinters_changed = true;
2250 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2251 idl_seqno = ovsdb_idl_get_seqno(idl);
2253 reconf_txn = ovsdb_idl_txn_create(idl);
2254 bridge_reconfigure(cfg);
2256 /* We still need to reconfigure to avoid dangling pointers to
2257 * now-destroyed ovsrec structures inside bridge data. */
2258 bridge_reconfigure(&null_cfg);
2263 if (reconfiguring) {
2266 reconf_txn = ovsdb_idl_txn_create(idl);
2268 if (bridge_reconfigure_continue(cfg)) {
2269 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2272 bridge_reconfigure_continue(&null_cfg);
2277 ovsdb_idl_txn_commit(reconf_txn);
2278 ovsdb_idl_txn_destroy(reconf_txn);
2282 /* Refresh interface and mirror stats if necessary. */
2283 if (time_msec() >= iface_stats_timer) {
2285 struct ovsdb_idl_txn *txn;
2287 txn = ovsdb_idl_txn_create(idl);
2288 HMAP_FOR_EACH (br, node, &all_bridges) {
2292 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2293 struct iface *iface;
2295 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2296 iface_refresh_stats(iface);
2297 iface_refresh_status(iface);
2301 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2302 mirror_refresh_stats(m);
2306 refresh_controller_status();
2307 ovsdb_idl_txn_commit(txn);
2308 ovsdb_idl_txn_destroy(txn); /* XXX */
2311 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2315 instant_stats_run();
2324 ovsdb_idl_wait(idl);
2326 if (reconfiguring) {
2327 poll_immediate_wake();
2331 ofproto_enumerate_types(&types);
2332 SSET_FOR_EACH (type, &types) {
2333 ofproto_type_wait(type);
2335 sset_destroy(&types);
2337 if (!hmap_is_empty(&all_bridges)) {
2340 HMAP_FOR_EACH (br, node, &all_bridges) {
2341 ofproto_wait(br->ofproto);
2343 poll_timer_wait_until(iface_stats_timer);
2346 system_stats_wait();
2347 instant_stats_wait();
2350 /* Adds some memory usage statistics for bridges into 'usage', for use with
2351 * memory_report(). */
2353 bridge_get_memory_usage(struct simap *usage)
2357 HMAP_FOR_EACH (br, node, &all_bridges) {
2358 ofproto_get_memory_usage(br->ofproto, usage);
2362 /* QoS unixctl user interface functions. */
2364 struct qos_unixctl_show_cbdata {
2366 struct iface *iface;
2370 qos_unixctl_show_cb(unsigned int queue_id,
2371 const struct smap *details,
2374 struct qos_unixctl_show_cbdata *data = aux;
2375 struct ds *ds = data->ds;
2376 struct iface *iface = data->iface;
2377 struct netdev_queue_stats stats;
2378 struct smap_node *node;
2381 ds_put_cstr(ds, "\n");
2383 ds_put_format(ds, "Queue %u:\n", queue_id);
2385 ds_put_cstr(ds, "Default:\n");
2388 SMAP_FOR_EACH (node, details) {
2389 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2392 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2394 if (stats.tx_packets != UINT64_MAX) {
2395 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2398 if (stats.tx_bytes != UINT64_MAX) {
2399 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2402 if (stats.tx_errors != UINT64_MAX) {
2403 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2406 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2407 queue_id, strerror(error));
2412 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2413 const char *argv[], void *aux OVS_UNUSED)
2415 struct ds ds = DS_EMPTY_INITIALIZER;
2416 struct smap smap = SMAP_INITIALIZER(&smap);
2417 struct iface *iface;
2419 struct smap_node *node;
2420 struct qos_unixctl_show_cbdata data;
2423 iface = iface_find(argv[1]);
2425 unixctl_command_reply_error(conn, "no such interface");
2429 netdev_get_qos(iface->netdev, &type, &smap);
2431 if (*type != '\0') {
2432 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2434 SMAP_FOR_EACH (node, &smap) {
2435 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2440 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2443 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2445 unixctl_command_reply(conn, ds_cstr(&ds));
2447 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2448 unixctl_command_reply_error(conn, ds_cstr(&ds));
2451 smap_destroy(&smap);
2455 /* Bridge reconfiguration functions. */
2457 bridge_create(const struct ovsrec_bridge *br_cfg)
2461 ovs_assert(!bridge_lookup(br_cfg->name));
2462 br = xzalloc(sizeof *br);
2464 br->name = xstrdup(br_cfg->name);
2465 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2468 /* Derive the default Ethernet address from the bridge's UUID. This should
2469 * be unique and it will be stable between ovs-vswitchd runs. */
2470 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2471 eth_addr_mark_random(br->default_ea);
2473 hmap_init(&br->ports);
2474 hmap_init(&br->ifaces);
2475 hmap_init(&br->iface_by_name);
2476 hmap_init(&br->mirrors);
2478 hmap_init(&br->if_cfg_todo);
2479 list_init(&br->ofpp_garbage);
2481 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2485 bridge_destroy(struct bridge *br)
2488 struct mirror *mirror, *next_mirror;
2489 struct port *port, *next_port;
2490 struct if_cfg *if_cfg, *next_if_cfg;
2491 struct ofpp_garbage *garbage, *next_garbage;
2493 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2496 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2497 mirror_destroy(mirror);
2499 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2500 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2503 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2504 &br->ofpp_garbage) {
2505 list_remove(&garbage->list_node);
2509 hmap_remove(&all_bridges, &br->node);
2510 ofproto_destroy(br->ofproto);
2511 hmap_destroy(&br->ifaces);
2512 hmap_destroy(&br->ports);
2513 hmap_destroy(&br->iface_by_name);
2514 hmap_destroy(&br->mirrors);
2515 hmap_destroy(&br->if_cfg_todo);
2522 static struct bridge *
2523 bridge_lookup(const char *name)
2527 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2528 if (!strcmp(br->name, name)) {
2535 /* Handle requests for a listing of all flows known by the OpenFlow
2536 * stack, including those normally hidden. */
2538 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2539 const char *argv[], void *aux OVS_UNUSED)
2544 br = bridge_lookup(argv[1]);
2546 unixctl_command_reply_error(conn, "Unknown bridge");
2551 ofproto_get_all_flows(br->ofproto, &results);
2553 unixctl_command_reply(conn, ds_cstr(&results));
2554 ds_destroy(&results);
2557 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2558 * connections and reconnect. If BRIDGE is not specified, then all bridges
2559 * drop their controller connections and reconnect. */
2561 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2562 const char *argv[], void *aux OVS_UNUSED)
2566 br = bridge_lookup(argv[1]);
2568 unixctl_command_reply_error(conn, "Unknown bridge");
2571 ofproto_reconnect_controllers(br->ofproto);
2573 HMAP_FOR_EACH (br, node, &all_bridges) {
2574 ofproto_reconnect_controllers(br->ofproto);
2577 unixctl_command_reply(conn, NULL);
2581 bridge_get_controllers(const struct bridge *br,
2582 struct ovsrec_controller ***controllersp)
2584 struct ovsrec_controller **controllers;
2585 size_t n_controllers;
2587 controllers = br->cfg->controller;
2588 n_controllers = br->cfg->n_controller;
2590 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2596 *controllersp = controllers;
2598 return n_controllers;
2602 bridge_queue_if_cfg(struct bridge *br,
2603 const struct ovsrec_interface *cfg,
2604 const struct ovsrec_port *parent)
2606 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2609 if_cfg->parent = parent;
2610 if_cfg->ofport = iface_pick_ofport(cfg);
2611 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2612 hash_string(if_cfg->cfg->name, 0));
2615 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2616 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2617 * 'br' needs to complete its configuration. */
2619 bridge_add_del_ports(struct bridge *br,
2620 const unsigned long int *splinter_vlans)
2622 struct shash_node *port_node;
2623 struct port *port, *next;
2624 struct shash new_ports;
2627 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2629 /* Collect new ports. */
2630 shash_init(&new_ports);
2631 for (i = 0; i < br->cfg->n_ports; i++) {
2632 const char *name = br->cfg->ports[i]->name;
2633 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2634 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2638 if (bridge_get_controllers(br, NULL)
2639 && !shash_find(&new_ports, br->name)) {
2640 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2641 br->name, br->name);
2643 ovsrec_interface_init(&br->synth_local_iface);
2644 ovsrec_port_init(&br->synth_local_port);
2646 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2647 br->synth_local_port.n_interfaces = 1;
2648 br->synth_local_port.name = br->name;
2650 br->synth_local_iface.name = br->name;
2651 br->synth_local_iface.type = "internal";
2653 br->synth_local_ifacep = &br->synth_local_iface;
2655 shash_add(&new_ports, br->name, &br->synth_local_port);
2658 if (splinter_vlans) {
2659 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2662 /* Get rid of deleted ports.
2663 * Get rid of deleted interfaces on ports that still exist. */
2664 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2665 port->cfg = shash_find_data(&new_ports, port->name);
2669 port_del_ifaces(port);
2673 /* Update iface->cfg and iface->type in interfaces that still exist.
2674 * Add new interfaces to creation queue. */
2675 SHASH_FOR_EACH (port_node, &new_ports) {
2676 const struct ovsrec_port *port = port_node->data;
2679 for (i = 0; i < port->n_interfaces; i++) {
2680 const struct ovsrec_interface *cfg = port->interfaces[i];
2681 struct iface *iface = iface_lookup(br, cfg->name);
2682 const char *type = iface_get_type(cfg, br->cfg);
2687 } else if (!strcmp(type, "null")) {
2688 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2689 " may be removed in February 2013. Please email"
2690 " dev@openvswitch.org with concerns.",
2693 bridge_queue_if_cfg(br, cfg, port);
2698 shash_destroy(&new_ports);
2701 /* Initializes 'oc' appropriately as a management service controller for
2704 * The caller must free oc->target when it is no longer needed. */
2706 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2707 struct ofproto_controller *oc)
2709 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2710 oc->max_backoff = 0;
2711 oc->probe_interval = 60;
2712 oc->band = OFPROTO_OUT_OF_BAND;
2714 oc->burst_limit = 0;
2715 oc->enable_async_msgs = true;
2718 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2720 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2721 struct ofproto_controller *oc)
2725 oc->target = c->target;
2726 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2727 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2728 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2729 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2730 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2731 oc->burst_limit = (c->controller_burst_limit
2732 ? *c->controller_burst_limit : 0);
2733 oc->enable_async_msgs = (!c->enable_async_messages
2734 || *c->enable_async_messages);
2735 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2736 if (dscp < 0 || dscp > 63) {
2737 dscp = DSCP_DEFAULT;
2742 /* Configures the IP stack for 'br''s local interface properly according to the
2743 * configuration in 'c'. */
2745 bridge_configure_local_iface_netdev(struct bridge *br,
2746 struct ovsrec_controller *c)
2748 struct netdev *netdev;
2749 struct in_addr mask, gateway;
2751 struct iface *local_iface;
2754 /* If there's no local interface or no IP address, give up. */
2755 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2756 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2760 /* Bring up the local interface. */
2761 netdev = local_iface->netdev;
2762 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2764 /* Configure the IP address and netmask. */
2765 if (!c->local_netmask
2766 || !inet_aton(c->local_netmask, &mask)
2768 mask.s_addr = guess_netmask(ip.s_addr);
2770 if (!netdev_set_in4(netdev, ip, mask)) {
2771 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2772 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2775 /* Configure the default gateway. */
2776 if (c->local_gateway
2777 && inet_aton(c->local_gateway, &gateway)
2778 && gateway.s_addr) {
2779 if (!netdev_add_router(netdev, gateway)) {
2780 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2781 br->name, IP_ARGS(gateway.s_addr));
2786 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2787 * in either string are treated as equal to any number of slashes in the other,
2788 * e.g. "x///y" is equal to "x/y".
2790 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2791 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2792 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2793 * 'b' against a prefix of 'a'.
2796 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2798 const char *b_start = b;
2800 if (b - b_start >= b_stoplen) {
2802 } else if (*a != *b) {
2804 } else if (*a == '/') {
2805 a += strspn(a, "/");
2806 b += strspn(b, "/");
2807 } else if (*a == '\0') {
2817 bridge_configure_remotes(struct bridge *br,
2818 const struct sockaddr_in *managers, size_t n_managers)
2820 bool disable_in_band;
2822 struct ovsrec_controller **controllers;
2823 size_t n_controllers;
2825 enum ofproto_fail_mode fail_mode;
2827 struct ofproto_controller *ocs;
2831 /* Check if we should disable in-band control on this bridge. */
2832 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2835 /* Set OpenFlow queue ID for in-band control. */
2836 ofproto_set_in_band_queue(br->ofproto,
2837 smap_get_int(&br->cfg->other_config,
2838 "in-band-queue", -1));
2840 if (disable_in_band) {
2841 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2843 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2846 n_controllers = bridge_get_controllers(br, &controllers);
2848 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2851 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2852 for (i = 0; i < n_controllers; i++) {
2853 struct ovsrec_controller *c = controllers[i];
2855 if (!strncmp(c->target, "punix:", 6)
2856 || !strncmp(c->target, "unix:", 5)) {
2857 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2860 if (!strncmp(c->target, "unix:", 5)) {
2861 /* Connect to a listening socket */
2862 whitelist = xasprintf("unix:%s/", ovs_rundir());
2863 if (strchr(c->target, '/') &&
2864 !equal_pathnames(c->target, whitelist,
2865 strlen(whitelist))) {
2866 /* Absolute path specified, but not in ovs_rundir */
2867 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2868 "controller \"%s\" due to possibility for "
2869 "remote exploit. Instead, specify socket "
2870 "in whitelisted \"%s\" or connect to "
2871 "\"unix:%s/%s.mgmt\" (which is always "
2872 "available without special configuration).",
2873 br->name, c->target, whitelist,
2874 ovs_rundir(), br->name);
2879 whitelist = xasprintf("punix:%s/%s.controller",
2880 ovs_rundir(), br->name);
2881 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2882 /* Prevent remote ovsdb-server users from accessing
2883 * arbitrary Unix domain sockets and overwriting arbitrary
2885 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2886 "controller \"%s\" due to possibility of "
2887 "overwriting local files. Instead, specify "
2888 "whitelisted \"%s\" or connect to "
2889 "\"unix:%s/%s.mgmt\" (which is always "
2890 "available without special configuration).",
2891 br->name, c->target, whitelist,
2892 ovs_rundir(), br->name);
2901 bridge_configure_local_iface_netdev(br, c);
2902 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2903 if (disable_in_band) {
2904 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2909 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
2910 bridge_get_allowed_versions(br));
2911 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2914 /* Set the fail-mode. */
2915 fail_mode = !br->cfg->fail_mode
2916 || !strcmp(br->cfg->fail_mode, "standalone")
2917 ? OFPROTO_FAIL_STANDALONE
2918 : OFPROTO_FAIL_SECURE;
2919 ofproto_set_fail_mode(br->ofproto, fail_mode);
2921 /* Configure OpenFlow controller connection snooping. */
2922 if (!ofproto_has_snoops(br->ofproto)) {
2926 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2927 ovs_rundir(), br->name));
2928 ofproto_set_snoops(br->ofproto, &snoops);
2929 sset_destroy(&snoops);
2934 bridge_configure_tables(struct bridge *br)
2936 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2940 n_tables = ofproto_get_n_tables(br->ofproto);
2942 for (i = 0; i < n_tables; i++) {
2943 struct ofproto_table_settings s;
2946 s.max_flows = UINT_MAX;
2950 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2951 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2954 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2955 s.max_flows = *cfg->flow_limit;
2957 if (cfg->overflow_policy
2958 && !strcmp(cfg->overflow_policy, "evict")) {
2961 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2962 for (k = 0; k < cfg->n_groups; k++) {
2963 const char *string = cfg->groups[k];
2966 msg = mf_parse_subfield__(&s.groups[k], &string);
2968 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2969 "'groups' (%s)", br->name, i, msg);
2971 } else if (*string) {
2972 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2973 "element '%s' contains trailing garbage",
2974 br->name, i, cfg->groups[k]);
2982 ofproto_configure_table(br->ofproto, i, &s);
2986 for (; j < br->cfg->n_flow_tables; j++) {
2987 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2988 "%"PRId64" not supported by this datapath", br->name,
2989 br->cfg->key_flow_tables[j]);
2994 bridge_configure_dp_desc(struct bridge *br)
2996 ofproto_set_dp_desc(br->ofproto,
2997 smap_get(&br->cfg->other_config, "dp-desc"));
3000 /* Port functions. */
3002 static struct port *
3003 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3007 port = xzalloc(sizeof *port);
3009 port->name = xstrdup(cfg->name);
3011 list_init(&port->ifaces);
3013 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3017 /* Deletes interfaces from 'port' that are no longer configured for it. */
3019 port_del_ifaces(struct port *port)
3021 struct iface *iface, *next;
3022 struct sset new_ifaces;
3025 /* Collect list of new interfaces. */
3026 sset_init(&new_ifaces);
3027 for (i = 0; i < port->cfg->n_interfaces; i++) {
3028 const char *name = port->cfg->interfaces[i]->name;
3029 const char *type = port->cfg->interfaces[i]->type;
3030 if (strcmp(type, "null")) {
3031 sset_add(&new_ifaces, name);
3035 /* Get rid of deleted interfaces. */
3036 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3037 if (!sset_contains(&new_ifaces, iface->name)) {
3038 iface_destroy(iface);
3042 sset_destroy(&new_ifaces);
3046 port_destroy(struct port *port)
3049 struct bridge *br = port->bridge;
3050 struct iface *iface, *next;
3053 ofproto_bundle_unregister(br->ofproto, port);
3056 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3057 iface_destroy(iface);
3060 hmap_remove(&br->ports, &port->hmap_node);
3066 static struct port *
3067 port_lookup(const struct bridge *br, const char *name)
3071 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3073 if (!strcmp(port->name, name)) {
3081 enable_lacp(struct port *port, bool *activep)
3083 if (!port->cfg->lacp) {
3084 /* XXX when LACP implementation has been sufficiently tested, enable by
3085 * default and make active on bonded ports. */
3087 } else if (!strcmp(port->cfg->lacp, "off")) {
3089 } else if (!strcmp(port->cfg->lacp, "active")) {
3092 } else if (!strcmp(port->cfg->lacp, "passive")) {
3096 VLOG_WARN("port %s: unknown LACP mode %s",
3097 port->name, port->cfg->lacp);
3102 static struct lacp_settings *
3103 port_configure_lacp(struct port *port, struct lacp_settings *s)
3105 const char *lacp_time, *system_id;
3108 if (!enable_lacp(port, &s->active)) {
3112 s->name = port->name;
3114 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3116 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3117 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3118 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3119 " address.", port->name, system_id);
3123 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3126 if (eth_addr_is_zero(s->id)) {
3127 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3131 /* Prefer bondable links if unspecified. */
3132 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3134 s->priority = (priority > 0 && priority <= UINT16_MAX
3136 : UINT16_MAX - !list_is_short(&port->ifaces));
3138 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3139 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3144 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3146 int priority, portid, key;
3148 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3149 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3151 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3153 if (portid <= 0 || portid > UINT16_MAX) {
3154 portid = iface->ofp_port;
3157 if (priority <= 0 || priority > UINT16_MAX) {
3158 priority = UINT16_MAX;
3161 if (key < 0 || key > UINT16_MAX) {
3165 s->name = iface->name;
3167 s->priority = priority;
3172 port_configure_bond(struct port *port, struct bond_settings *s)
3174 const char *detect_s;
3175 struct iface *iface;
3176 int miimon_interval;
3178 s->name = port->name;
3180 if (port->cfg->bond_mode) {
3181 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3182 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3183 port->name, port->cfg->bond_mode,
3184 bond_mode_to_string(s->balance));
3187 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3189 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3190 * active-backup. At some point we should remove this warning. */
3191 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3192 " in previous versions, the default bond_mode was"
3193 " balance-slb", port->name,
3194 bond_mode_to_string(s->balance));
3196 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3197 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3198 "please use another bond type or disable flood_vlans",
3202 miimon_interval = smap_get_int(&port->cfg->other_config,
3203 "bond-miimon-interval", 0);
3204 if (miimon_interval <= 0) {
3205 miimon_interval = 200;
3208 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3209 if (!detect_s || !strcmp(detect_s, "carrier")) {
3210 miimon_interval = 0;
3211 } else if (strcmp(detect_s, "miimon")) {
3212 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3213 "defaulting to carrier", port->name, detect_s);
3214 miimon_interval = 0;
3217 s->up_delay = MAX(0, port->cfg->bond_updelay);
3218 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3219 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3220 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3221 "bond-rebalance-interval", 10000);
3222 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3223 s->rebalance_interval = 1000;
3226 s->fake_iface = port->cfg->bond_fake_iface;
3228 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3229 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3233 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3234 * instead of obtaining it from the database. */
3236 port_is_synthetic(const struct port *port)
3238 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3241 /* Interface functions. */
3244 iface_is_internal(const struct ovsrec_interface *iface,
3245 const struct ovsrec_bridge *br)
3247 /* The local port and "internal" ports are always "internal". */
3248 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3251 /* Returns the correct network device type for interface 'iface' in bridge
3254 iface_get_type(const struct ovsrec_interface *iface,
3255 const struct ovsrec_bridge *br)
3259 /* The local port always has type "internal". Other ports take
3260 * their type from the database and default to "system" if none is
3262 if (iface_is_internal(iface, br)) {
3265 type = iface->type[0] ? iface->type : "system";
3268 return ofproto_port_open_type(br->datapath_type, type);
3272 iface_destroy(struct iface *iface)
3275 struct port *port = iface->port;
3276 struct bridge *br = port->bridge;
3278 if (br->ofproto && iface->ofp_port >= 0) {
3279 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3282 if (iface->ofp_port >= 0) {
3283 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3286 list_remove(&iface->port_elem);
3287 hmap_remove(&br->iface_by_name, &iface->name_node);
3289 netdev_close(iface->netdev);
3296 static struct iface *
3297 iface_lookup(const struct bridge *br, const char *name)
3299 struct iface *iface;
3301 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3302 &br->iface_by_name) {
3303 if (!strcmp(iface->name, name)) {
3311 static struct iface *
3312 iface_find(const char *name)
3314 const struct bridge *br;
3316 HMAP_FOR_EACH (br, node, &all_bridges) {
3317 struct iface *iface = iface_lookup(br, name);
3326 static struct if_cfg *
3327 if_cfg_lookup(const struct bridge *br, const char *name)
3329 struct if_cfg *if_cfg;
3331 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3333 if (!strcmp(if_cfg->cfg->name, name)) {
3341 static struct iface *
3342 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3344 struct iface *iface;
3346 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3347 hash_int(ofp_port, 0), &br->ifaces) {
3348 if (iface->ofp_port == ofp_port) {
3355 /* Set Ethernet address of 'iface', if one is specified in the configuration
3358 iface_set_mac(struct iface *iface)
3360 uint8_t ea[ETH_ADDR_LEN];
3362 if (!strcmp(iface->type, "internal")
3363 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3364 if (iface->ofp_port == OFPP_LOCAL) {
3365 VLOG_ERR("interface %s: ignoring mac in Interface record "
3366 "(use Bridge record to set local port's mac)",
3368 } else if (eth_addr_is_multicast(ea)) {
3369 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3372 int error = netdev_set_etheraddr(iface->netdev, ea);
3374 VLOG_ERR("interface %s: setting MAC failed (%s)",
3375 iface->name, strerror(error));
3381 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3383 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3385 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3386 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3390 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3391 * sets the "ofport" field to -1.
3393 * This is appropriate when 'if_cfg''s interface cannot be created or is
3394 * otherwise invalid. */
3396 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3398 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3399 ovsrec_interface_set_status(if_cfg, NULL);
3400 ovsrec_interface_set_admin_state(if_cfg, NULL);
3401 ovsrec_interface_set_duplex(if_cfg, NULL);
3402 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3403 ovsrec_interface_set_link_state(if_cfg, NULL);
3404 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3405 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3406 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3407 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3408 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3409 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3410 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3414 struct iface_delete_queues_cbdata {
3415 struct netdev *netdev;
3416 const struct ovsdb_datum *queues;
3420 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3422 union ovsdb_atom atom;
3424 atom.integer = target;
3425 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3429 iface_delete_queues(unsigned int queue_id,
3430 const struct smap *details OVS_UNUSED, void *cbdata_)
3432 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3434 if (!queue_ids_include(cbdata->queues, queue_id)) {
3435 netdev_delete_queue(cbdata->netdev, queue_id);
3440 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3442 struct ofpbuf queues_buf;
3444 ofpbuf_init(&queues_buf, 0);
3446 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3447 netdev_set_qos(iface->netdev, NULL, NULL);
3449 struct iface_delete_queues_cbdata cbdata;
3453 /* Configure top-level Qos for 'iface'. */
3454 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3456 /* Deconfigure queues that were deleted. */
3457 cbdata.netdev = iface->netdev;
3458 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3460 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3462 /* Configure queues for 'iface'. */
3464 for (i = 0; i < qos->n_queues; i++) {
3465 const struct ovsrec_queue *queue = qos->value_queues[i];
3466 unsigned int queue_id = qos->key_queues[i];
3468 if (queue_id == 0) {
3472 if (queue->n_dscp == 1) {
3473 struct ofproto_port_queue *port_queue;
3475 port_queue = ofpbuf_put_uninit(&queues_buf,
3476 sizeof *port_queue);
3477 port_queue->queue = queue_id;
3478 port_queue->dscp = queue->dscp[0];
3481 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3484 struct smap details;
3486 smap_init(&details);
3487 netdev_set_queue(iface->netdev, 0, &details);
3488 smap_destroy(&details);
3492 if (iface->ofp_port >= 0) {
3493 const struct ofproto_port_queue *port_queues = queues_buf.data;
3494 size_t n_queues = queues_buf.size / sizeof *port_queues;
3496 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3497 port_queues, n_queues);
3500 netdev_set_policing(iface->netdev,
3501 iface->cfg->ingress_policing_rate,
3502 iface->cfg->ingress_policing_burst);
3504 ofpbuf_uninit(&queues_buf);
3508 iface_configure_cfm(struct iface *iface)
3510 const struct ovsrec_interface *cfg = iface->cfg;
3511 const char *opstate_str;
3512 const char *cfm_ccm_vlan;
3513 struct cfm_settings s;
3514 struct smap netdev_args;
3516 if (!cfg->n_cfm_mpid) {
3517 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3521 s.check_tnl_key = false;
3522 smap_init(&netdev_args);
3523 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3524 const char *key = smap_get(&netdev_args, "key");
3525 const char *in_key = smap_get(&netdev_args, "in_key");
3527 s.check_tnl_key = (key && !strcmp(key, "flow"))
3528 || (in_key && !strcmp(in_key, "flow"));
3530 smap_destroy(&netdev_args);
3532 s.mpid = *cfg->cfm_mpid;
3533 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3534 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3535 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3537 if (s.interval <= 0) {
3541 if (!cfm_ccm_vlan) {
3543 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3544 s.ccm_vlan = CFM_RANDOM_VLAN;
3546 s.ccm_vlan = atoi(cfm_ccm_vlan);
3547 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3552 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3555 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3556 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3558 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3561 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3562 * instead of obtaining it from the database. */
3564 iface_is_synthetic(const struct iface *iface)
3566 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3570 iface_pick_ofport(const struct ovsrec_interface *cfg)
3572 int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3573 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3577 /* Port mirroring. */
3579 static struct mirror *
3580 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3584 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3585 if (uuid_equals(uuid, &m->uuid)) {
3593 bridge_configure_mirrors(struct bridge *br)
3595 const struct ovsdb_datum *mc;
3596 unsigned long *flood_vlans;
3597 struct mirror *m, *next;
3600 /* Get rid of deleted mirrors. */
3601 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3602 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3603 union ovsdb_atom atom;
3605 atom.uuid = m->uuid;
3606 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3611 /* Add new mirrors and reconfigure existing ones. */
3612 for (i = 0; i < br->cfg->n_mirrors; i++) {
3613 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3614 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3616 m = mirror_create(br, cfg);
3619 if (!mirror_configure(m)) {
3624 /* Update flooded vlans (for RSPAN). */
3625 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3626 br->cfg->n_flood_vlans);
3627 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3628 bitmap_free(flood_vlans);
3631 static struct mirror *
3632 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3636 m = xzalloc(sizeof *m);
3637 m->uuid = cfg->header_.uuid;
3638 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3640 m->name = xstrdup(cfg->name);
3646 mirror_destroy(struct mirror *m)
3649 struct bridge *br = m->bridge;
3652 ofproto_mirror_unregister(br->ofproto, m);
3655 hmap_remove(&br->mirrors, &m->hmap_node);
3662 mirror_collect_ports(struct mirror *m,
3663 struct ovsrec_port **in_ports, int n_in_ports,
3664 void ***out_portsp, size_t *n_out_portsp)
3666 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3667 size_t n_out_ports = 0;
3670 for (i = 0; i < n_in_ports; i++) {
3671 const char *name = in_ports[i]->name;
3672 struct port *port = port_lookup(m->bridge, name);
3674 out_ports[n_out_ports++] = port;
3676 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3677 "port %s", m->bridge->name, m->name, name);
3680 *out_portsp = out_ports;
3681 *n_out_portsp = n_out_ports;
3685 mirror_configure(struct mirror *m)
3687 const struct ovsrec_mirror *cfg = m->cfg;
3688 struct ofproto_mirror_settings s;
3691 if (strcmp(cfg->name, m->name)) {
3693 m->name = xstrdup(cfg->name);
3697 /* Get output port or VLAN. */
3698 if (cfg->output_port) {
3699 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3700 if (!s.out_bundle) {
3701 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3702 m->bridge->name, m->name);
3705 s.out_vlan = UINT16_MAX;
3707 if (cfg->output_vlan) {
3708 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3709 "output vlan; ignoring output vlan",
3710 m->bridge->name, m->name);
3712 } else if (cfg->output_vlan) {
3713 /* The database should prevent invalid VLAN values. */
3714 s.out_bundle = NULL;
3715 s.out_vlan = *cfg->output_vlan;
3717 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3718 m->bridge->name, m->name);
3722 /* Get port selection. */
3723 if (cfg->select_all) {
3724 size_t n_ports = hmap_count(&m->bridge->ports);
3725 void **ports = xmalloc(n_ports * sizeof *ports);
3730 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3740 /* Get ports, dropping ports that don't exist.
3741 * The IDL ensures that there are no duplicates. */
3742 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3743 &s.srcs, &s.n_srcs);
3744 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3745 &s.dsts, &s.n_dsts);
3748 /* Get VLAN selection. */
3749 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3752 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3755 if (s.srcs != s.dsts) {
3764 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3766 * This is deprecated. It is only for compatibility with broken device drivers
3767 * in old versions of Linux that do not properly support VLANs when VLAN
3768 * devices are not used. When broken device drivers are no longer in
3769 * widespread use, we will delete these interfaces. */
3771 static struct ovsrec_port **recs;
3772 static size_t n_recs, allocated_recs;
3774 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3775 * splinters are reconfigured. */
3777 register_rec(struct ovsrec_port *rec)
3779 if (n_recs >= allocated_recs) {
3780 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3782 recs[n_recs++] = rec;
3785 /* Frees all of the ports registered with register_reg(). */
3787 free_registered_recs(void)
3791 for (i = 0; i < n_recs; i++) {
3792 struct ovsrec_port *port = recs[i];
3795 for (j = 0; j < port->n_interfaces; j++) {
3796 struct ovsrec_interface *iface = port->interfaces[j];
3801 smap_destroy(&port->other_config);
3802 free(port->interfaces);
3810 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3813 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3815 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3819 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3822 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3823 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3824 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3827 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3828 * use, returns NULL.
3830 * Updates 'vlan_splinters_enabled_anywhere'. */
3831 static unsigned long int *
3832 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3834 unsigned long int *splinter_vlans;
3835 struct sset splinter_ifaces;
3836 const char *real_dev_name;
3837 struct shash *real_devs;
3838 struct shash_node *node;
3842 /* Free space allocated for synthesized ports and interfaces, since we're
3843 * in the process of reconstructing all of them. */
3844 free_registered_recs();
3846 splinter_vlans = bitmap_allocate(4096);
3847 sset_init(&splinter_ifaces);
3848 vlan_splinters_enabled_anywhere = false;
3849 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3850 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3853 for (j = 0; j < br_cfg->n_ports; j++) {
3854 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3857 for (k = 0; k < port_cfg->n_interfaces; k++) {
3858 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3860 if (vlan_splinters_is_enabled(iface_cfg)) {
3861 vlan_splinters_enabled_anywhere = true;
3862 sset_add(&splinter_ifaces, iface_cfg->name);
3863 vlan_bitmap_from_array__(port_cfg->trunks,
3869 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3870 bitmap_set1(splinter_vlans, *port_cfg->tag);
3875 if (!vlan_splinters_enabled_anywhere) {
3876 free(splinter_vlans);
3877 sset_destroy(&splinter_ifaces);
3881 HMAP_FOR_EACH (br, node, &all_bridges) {
3883 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3887 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3888 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3889 * device to be created for it. */
3890 bitmap_set0(splinter_vlans, 0);
3891 bitmap_set0(splinter_vlans, 4095);
3893 /* Delete all VLAN devices that we don't need. */
3895 real_devs = vlandev_get_real_devs();
3896 SHASH_FOR_EACH (node, real_devs) {
3897 const struct vlan_real_dev *real_dev = node->data;
3898 const struct vlan_dev *vlan_dev;
3899 bool real_dev_has_splinters;
3901 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3903 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3904 if (!real_dev_has_splinters
3905 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3906 struct netdev *netdev;
3908 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3909 if (!netdev_get_in4(netdev, NULL, NULL) ||
3910 !netdev_get_in6(netdev, NULL)) {
3911 vlandev_del(vlan_dev->name);
3913 /* It has an IP address configured, so we don't own
3914 * it. Don't delete it. */
3916 netdev_close(netdev);
3923 /* Add all VLAN devices that we need. */
3924 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3927 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3928 if (!vlandev_get_name(real_dev_name, vid)) {
3929 vlandev_add(real_dev_name, vid);
3936 sset_destroy(&splinter_ifaces);
3938 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3939 free(splinter_vlans);
3942 return splinter_vlans;
3945 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3948 configure_splinter_port(struct port *port)
3950 struct ofproto *ofproto = port->bridge->ofproto;
3951 uint16_t realdev_ofp_port;
3952 const char *realdev_name;
3953 struct iface *vlandev, *realdev;
3955 ofproto_bundle_unregister(port->bridge->ofproto, port);
3957 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3960 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3961 realdev = iface_lookup(port->bridge, realdev_name);
3962 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3964 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3968 static struct ovsrec_port *
3969 synthesize_splinter_port(const char *real_dev_name,
3970 const char *vlan_dev_name, int vid)
3972 struct ovsrec_interface *iface;
3973 struct ovsrec_port *port;
3975 iface = xmalloc(sizeof *iface);
3976 ovsrec_interface_init(iface);
3977 iface->name = xstrdup(vlan_dev_name);
3978 iface->type = "system";
3980 port = xmalloc(sizeof *port);
3981 ovsrec_port_init(port);
3982 port->interfaces = xmemdup(&iface, sizeof iface);
3983 port->n_interfaces = 1;
3984 port->name = xstrdup(vlan_dev_name);
3985 port->vlan_mode = "splinter";
3986 port->tag = xmalloc(sizeof *port->tag);
3989 smap_add(&port->other_config, "realdev", real_dev_name);
3995 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3996 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3997 * 1-bit in the 'splinter_vlans' bitmap. */
3999 add_vlan_splinter_ports(struct bridge *br,
4000 const unsigned long int *splinter_vlans,
4001 struct shash *ports)
4005 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4006 * we're modifying 'ports'. */
4007 for (i = 0; i < br->cfg->n_ports; i++) {
4008 const char *name = br->cfg->ports[i]->name;
4009 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4012 for (j = 0; j < port_cfg->n_interfaces; j++) {
4013 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4015 if (vlan_splinters_is_enabled(iface_cfg)) {
4016 const char *real_dev_name;
4019 real_dev_name = iface_cfg->name;
4020 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4021 const char *vlan_dev_name;
4023 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4025 && !shash_find(ports, vlan_dev_name)) {
4026 shash_add(ports, vlan_dev_name,
4027 synthesize_splinter_port(
4028 real_dev_name, vlan_dev_name, vid));
4037 mirror_refresh_stats(struct mirror *m)
4039 struct ofproto *ofproto = m->bridge->ofproto;
4040 uint64_t tx_packets, tx_bytes;
4043 size_t stat_cnt = 0;
4045 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4046 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4050 if (tx_packets != UINT64_MAX) {
4051 keys[stat_cnt] = "tx_packets";
4052 values[stat_cnt] = tx_packets;
4055 if (tx_bytes != UINT64_MAX) {
4056 keys[stat_cnt] = "tx_bytes";
4057 values[stat_cnt] = tx_bytes;
4061 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);