1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
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. */
71 /* OpenFlow port slated for removal from ofproto. */
73 struct list list_node; /* Node in bridge's ofpp_garbage. */
74 uint16_t ofp_port; /* Port to be deleted. */
78 /* These members are always valid. */
79 struct list port_elem; /* Element in struct port's "ifaces" list. */
80 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
84 /* These members are valid only after bridge_reconfigure() causes them to
86 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
87 int ofp_port; /* OpenFlow port number, -1 if unknown. */
88 struct netdev *netdev; /* Network device. */
89 const char *type; /* Usually same as cfg->type. */
90 const struct ovsrec_interface *cfg;
94 struct uuid uuid; /* UUID of this "mirror" record in database. */
95 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
96 struct bridge *bridge;
98 const struct ovsrec_mirror *cfg;
102 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
103 struct bridge *bridge;
106 const struct ovsrec_port *cfg;
108 /* An ordinary bridge port has 1 interface.
109 * A bridge port for bonding has at least 2 interfaces. */
110 struct list ifaces; /* List of "struct iface"s. */
114 struct hmap_node node; /* In 'all_bridges'. */
115 char *name; /* User-specified arbitrary name. */
116 char *type; /* Datapath type. */
117 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
118 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
119 const struct ovsrec_bridge *cfg;
121 /* OpenFlow switch processing. */
122 struct ofproto *ofproto; /* OpenFlow switch. */
125 struct hmap ports; /* "struct port"s indexed by name. */
126 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
127 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
129 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
130 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
131 Indexed on 'cfg->name'. */
133 /* Port mirroring. */
134 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
136 /* Synthetic local port if necessary. */
137 struct ovsrec_port synth_local_port;
138 struct ovsrec_interface synth_local_iface;
139 struct ovsrec_interface *synth_local_ifacep;
142 /* All bridges, indexed by name. */
143 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
145 /* OVSDB IDL used to obtain configuration. */
146 static struct ovsdb_idl *idl;
148 /* Most recently processed IDL sequence number. */
149 static unsigned int idl_seqno;
151 /* Each time this timer expires, the bridge fetches interface and mirror
152 * statistics and pushes them into the database. */
153 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int iface_stats_timer = LLONG_MIN;
156 /* Stores the time after which rate limited statistics may be written to the
157 * database. Only updated when changes to the database require rate limiting.
159 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
160 static long long int db_limiter = LLONG_MIN;
162 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
163 * expensive. This can cause bridge_reconfigure() to take a long time during
164 * which no other work can be done. To deal with this problem, we limit port
165 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
166 * call to bridge_reconfigure(). If there is more work to do after the limit
167 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
168 * This allows the rest of the code to catch up on important things like
169 * forwarding packets. */
170 #define OFP_PORT_ACTION_WINDOW 10
171 static bool reconfiguring = false;
173 static void add_del_bridges(const struct ovsrec_open_vswitch *);
174 static void bridge_update_ofprotos(void);
175 static void bridge_create(const struct ovsrec_bridge *);
176 static void bridge_destroy(struct bridge *);
177 static struct bridge *bridge_lookup(const char *name);
178 static unixctl_cb_func bridge_unixctl_dump_flows;
179 static unixctl_cb_func bridge_unixctl_reconnect;
180 static size_t bridge_get_controllers(const struct bridge *br,
181 struct ovsrec_controller ***controllersp);
182 static void bridge_add_del_ports(struct bridge *,
183 const unsigned long int *splinter_vlans);
184 static void bridge_refresh_ofp_port(struct bridge *);
185 static void bridge_configure_datapath_id(struct bridge *);
186 static void bridge_configure_flow_eviction_threshold(struct bridge *);
187 static void bridge_configure_netflow(struct bridge *);
188 static void bridge_configure_forward_bpdu(struct bridge *);
189 static void bridge_configure_mac_idle_time(struct bridge *);
190 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
191 static void bridge_configure_stp(struct bridge *);
192 static void bridge_configure_tables(struct bridge *);
193 static void bridge_configure_remotes(struct bridge *,
194 const struct sockaddr_in *managers,
196 static void bridge_pick_local_hw_addr(struct bridge *,
197 uint8_t ea[ETH_ADDR_LEN],
198 struct iface **hw_addr_iface);
199 static uint64_t bridge_pick_datapath_id(struct bridge *,
200 const uint8_t bridge_ea[ETH_ADDR_LEN],
201 struct iface *hw_addr_iface);
202 static void bridge_queue_if_cfg(struct bridge *,
203 const struct ovsrec_interface *,
204 const struct ovsrec_port *);
205 static uint64_t dpid_from_hash(const void *, size_t nbytes);
206 static bool bridge_has_bond_fake_iface(const struct bridge *,
208 static bool port_is_bond_fake_iface(const struct port *);
210 static unixctl_cb_func qos_unixctl_show;
212 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
213 static void port_del_ifaces(struct port *);
214 static void port_destroy(struct port *);
215 static struct port *port_lookup(const struct bridge *, const char *name);
216 static void port_configure(struct port *);
217 static struct lacp_settings *port_configure_lacp(struct port *,
218 struct lacp_settings *);
219 static void port_configure_bond(struct port *, struct bond_settings *,
220 uint32_t *bond_stable_ids);
221 static bool port_is_synthetic(const struct port *);
223 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
224 static void run_system_stats(void);
226 static void bridge_configure_mirrors(struct bridge *);
227 static struct mirror *mirror_create(struct bridge *,
228 const struct ovsrec_mirror *);
229 static void mirror_destroy(struct mirror *);
230 static bool mirror_configure(struct mirror *);
231 static void mirror_refresh_stats(struct mirror *);
233 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
234 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
235 static const char *iface_get_type(const struct ovsrec_interface *,
236 const struct ovsrec_bridge *);
237 static void iface_destroy(struct iface *);
238 static struct iface *iface_lookup(const struct bridge *, const char *name);
239 static struct iface *iface_find(const char *name);
240 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
241 static struct iface *iface_from_ofp_port(const struct bridge *,
243 static void iface_set_mac(struct iface *);
244 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
245 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
246 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
247 static void iface_configure_cfm(struct iface *);
248 static void iface_refresh_cfm_stats(struct iface *);
249 static void iface_refresh_stats(struct iface *);
250 static void iface_refresh_status(struct iface *);
251 static bool iface_is_synthetic(const struct iface *);
253 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
255 * This is deprecated. It is only for compatibility with broken device drivers
256 * in old versions of Linux that do not properly support VLANs when VLAN
257 * devices are not used. When broken device drivers are no longer in
258 * widespread use, we will delete these interfaces. */
260 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
261 static bool vlan_splinters_enabled_anywhere;
263 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
264 static unsigned long int *collect_splinter_vlans(
265 const struct ovsrec_open_vswitch *);
266 static void configure_splinter_port(struct port *);
267 static void add_vlan_splinter_ports(struct bridge *,
268 const unsigned long int *splinter_vlans,
269 struct shash *ports);
271 /* Public functions. */
273 /* Initializes the bridge module, configuring it to obtain its configuration
274 * from an OVSDB server accessed over 'remote', which should be a string in a
275 * form acceptable to ovsdb_idl_create(). */
277 bridge_init(const char *remote)
279 /* Create connection to database. */
280 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
281 idl_seqno = ovsdb_idl_get_seqno(idl);
282 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
284 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
285 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
288 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
289 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
290 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
292 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
293 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
294 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
296 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
297 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
298 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
299 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
315 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
317 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
319 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
320 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
322 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
324 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
326 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
327 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
329 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
335 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
336 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
339 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
341 /* Register unixctl commands. */
342 unixctl_command_register("qos/show", "interface", 1, 1,
343 qos_unixctl_show, NULL);
344 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
345 bridge_unixctl_dump_flows, NULL);
346 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
347 bridge_unixctl_reconnect, NULL);
357 struct bridge *br, *next_br;
359 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
362 ovsdb_idl_destroy(idl);
365 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
366 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
367 * responsible for freeing '*managersp' (with free()).
369 * You may be asking yourself "why does ovs-vswitchd care?", because
370 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
371 * should not be and in fact is not directly involved in that. But
372 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
373 * it has to tell in-band control where the managers are to enable that.
374 * (Thus, only managers connected in-band are collected.)
377 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
378 struct sockaddr_in **managersp, size_t *n_managersp)
380 struct sockaddr_in *managers = NULL;
381 size_t n_managers = 0;
385 /* Collect all of the potential targets from the "targets" columns of the
386 * rows pointed to by "manager_options", excluding any that are
389 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
390 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
392 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
393 sset_find_and_delete(&targets, m->target);
395 sset_add(&targets, m->target);
399 /* Now extract the targets' IP addresses. */
400 if (!sset_is_empty(&targets)) {
403 managers = xmalloc(sset_count(&targets) * sizeof *managers);
404 SSET_FOR_EACH (target, &targets) {
405 struct sockaddr_in *sin = &managers[n_managers];
407 if (stream_parse_target_with_default_ports(target,
415 sset_destroy(&targets);
417 *managersp = managers;
418 *n_managersp = n_managers;
422 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
424 unsigned long int *splinter_vlans;
427 COVERAGE_INC(bridge_reconfigure);
429 assert(!reconfiguring);
430 reconfiguring = true;
432 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
433 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
434 * configuration otherwise.
436 * This is mostly an update to bridge data structures. Nothing is pushed
437 * down to ofproto or lower layers. */
438 add_del_bridges(ovs_cfg);
439 splinter_vlans = collect_splinter_vlans(ovs_cfg);
440 HMAP_FOR_EACH (br, node, &all_bridges) {
441 bridge_add_del_ports(br, splinter_vlans);
443 free(splinter_vlans);
445 /* Delete datapaths that are no longer configured, and create ones which
446 * don't exist but should. */
447 bridge_update_ofprotos();
449 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
450 HMAP_FOR_EACH (br, node, &all_bridges) {
451 bridge_refresh_ofp_port(br);
454 /* Clear database records for "if_cfg"s which haven't been instantiated. */
455 HMAP_FOR_EACH (br, node, &all_bridges) {
456 struct if_cfg *if_cfg;
458 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
459 iface_clear_db_record(if_cfg->cfg);
463 reconfigure_system_stats(ovs_cfg);
467 bridge_reconfigure_ofp(void)
469 long long int deadline;
473 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
475 /* The kernel will reject any attempt to add a given port to a datapath if
476 * that port already belongs to a different datapath, so we must do all
477 * port deletions before any port additions. */
478 HMAP_FOR_EACH (br, node, &all_bridges) {
479 struct ofpp_garbage *garbage, *next;
481 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
482 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
483 * internal datastructures may not be consistent. Eventually, when
484 * port additions and deletions are cheaper, these calls should be
487 ofproto_port_del(br->ofproto, garbage->ofp_port);
488 list_remove(&garbage->list_node);
492 if (time_msec() >= deadline) {
499 HMAP_FOR_EACH (br, node, &all_bridges) {
500 struct if_cfg *if_cfg, *next;
502 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
503 iface_create(br, if_cfg, -1);
505 if (time_msec() >= deadline) {
515 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
517 struct sockaddr_in *managers;
518 int sflow_bridge_number;
523 assert(reconfiguring);
524 done = bridge_reconfigure_ofp();
526 /* Complete the configuration. */
527 sflow_bridge_number = 0;
528 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
529 HMAP_FOR_EACH (br, node, &all_bridges) {
532 /* We need the datapath ID early to allow LACP ports to use it as the
533 * default system ID. */
534 bridge_configure_datapath_id(br);
536 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
539 port_configure(port);
541 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
542 iface_configure_cfm(iface);
543 iface_configure_qos(iface, port->cfg->qos);
544 iface_set_mac(iface);
547 bridge_configure_mirrors(br);
548 bridge_configure_flow_eviction_threshold(br);
549 bridge_configure_forward_bpdu(br);
550 bridge_configure_mac_idle_time(br);
551 bridge_configure_remotes(br, managers, n_managers);
552 bridge_configure_netflow(br);
553 bridge_configure_sflow(br, &sflow_bridge_number);
554 bridge_configure_stp(br);
555 bridge_configure_tables(br);
560 /* ovs-vswitchd has completed initialization, so allow the process that
561 * forked us to exit successfully. */
562 daemonize_complete();
563 reconfiguring = false;
565 VLOG_INFO("%s (Open vSwitch) %s", program_name, VERSION);
571 /* Delete ofprotos which aren't configured or have the wrong type. Create
572 * ofprotos which don't exist but need to. */
574 bridge_update_ofprotos(void)
576 struct bridge *br, *next;
581 /* Delete ofprotos with no bridge or with the wrong type. */
584 ofproto_enumerate_types(&types);
585 SSET_FOR_EACH (type, &types) {
588 ofproto_enumerate_names(type, &names);
589 SSET_FOR_EACH (name, &names) {
590 br = bridge_lookup(name);
591 if (!br || strcmp(type, br->type)) {
592 ofproto_delete(name, type);
596 sset_destroy(&names);
597 sset_destroy(&types);
599 /* Add ofprotos for bridges which don't have one yet. */
600 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
608 /* Remove ports from any datapath with the same name as 'br'. If we
609 * don't do this, creating 'br''s ofproto will fail because a port with
610 * the same name as its local port already exists. */
611 HMAP_FOR_EACH (br2, node, &all_bridges) {
612 struct ofproto_port ofproto_port;
618 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
620 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
622 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
625 ofproto_port_destroy(&ofproto_port);
629 error = ofproto_create(br->name, br->type, &br->ofproto);
631 VLOG_ERR("failed to create bridge %s: %s", br->name,
639 port_configure(struct port *port)
641 const struct ovsrec_port *cfg = port->cfg;
642 struct bond_settings bond_settings;
643 struct lacp_settings lacp_settings;
644 struct ofproto_bundle_settings s;
647 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
648 configure_splinter_port(port);
657 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
658 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
659 s.slaves[s.n_slaves++] = iface->ofp_port;
664 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
668 /* Get VLAN trunks. */
671 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
675 if (cfg->vlan_mode) {
676 if (!strcmp(cfg->vlan_mode, "access")) {
677 s.vlan_mode = PORT_VLAN_ACCESS;
678 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
679 s.vlan_mode = PORT_VLAN_TRUNK;
680 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
681 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
682 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
683 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
685 /* This "can't happen" because ovsdb-server should prevent it. */
686 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
687 s.vlan_mode = PORT_VLAN_TRUNK;
691 s.vlan_mode = PORT_VLAN_ACCESS;
693 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
697 s.vlan_mode = PORT_VLAN_TRUNK;
700 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
703 /* Get LACP settings. */
704 s.lacp = port_configure_lacp(port, &lacp_settings);
708 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
709 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
710 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
713 s.lacp_slaves = NULL;
716 /* Get bond settings. */
717 if (s.n_slaves > 1) {
718 s.bond = &bond_settings;
719 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
720 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
723 s.bond_stable_ids = NULL;
725 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
726 netdev_set_miimon_interval(iface->netdev, 0);
731 ofproto_bundle_register(port->bridge->ofproto, port, &s);
737 free(s.bond_stable_ids);
740 /* Pick local port hardware address and datapath ID for 'br'. */
742 bridge_configure_datapath_id(struct bridge *br)
744 uint8_t ea[ETH_ADDR_LEN];
746 struct iface *local_iface;
747 struct iface *hw_addr_iface;
750 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
751 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
753 int error = netdev_set_etheraddr(local_iface->netdev, ea);
755 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
756 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
757 "Ethernet address: %s",
758 br->name, strerror(error));
761 memcpy(br->ea, ea, ETH_ADDR_LEN);
763 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
764 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
765 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
766 ofproto_set_datapath_id(br->ofproto, dpid);
769 dpid_string = xasprintf("%016"PRIx64, dpid);
770 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
774 /* Set NetFlow configuration on 'br'. */
776 bridge_configure_netflow(struct bridge *br)
778 struct ovsrec_netflow *cfg = br->cfg->netflow;
779 struct netflow_options opts;
782 ofproto_set_netflow(br->ofproto, NULL);
786 memset(&opts, 0, sizeof opts);
788 /* Get default NetFlow configuration from datapath.
789 * Apply overrides from 'cfg'. */
790 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
791 if (cfg->engine_type) {
792 opts.engine_type = *cfg->engine_type;
794 if (cfg->engine_id) {
795 opts.engine_id = *cfg->engine_id;
798 /* Configure active timeout interval. */
799 opts.active_timeout = cfg->active_timeout;
800 if (!opts.active_timeout) {
801 opts.active_timeout = -1;
802 } else if (opts.active_timeout < 0) {
803 VLOG_WARN("bridge %s: active timeout interval set to negative "
804 "value, using default instead (%d seconds)", br->name,
805 NF_ACTIVE_TIMEOUT_DEFAULT);
806 opts.active_timeout = -1;
809 /* Add engine ID to interface number to disambiguate bridgs? */
810 opts.add_id_to_iface = cfg->add_id_to_interface;
811 if (opts.add_id_to_iface) {
812 if (opts.engine_id > 0x7f) {
813 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
814 "another vswitch, choose an engine id less than 128",
817 if (hmap_count(&br->ports) > 508) {
818 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
819 "another port when more than 508 ports are used",
825 sset_init(&opts.collectors);
826 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
829 if (ofproto_set_netflow(br->ofproto, &opts)) {
830 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
832 sset_destroy(&opts.collectors);
835 /* Set sFlow configuration on 'br'. */
837 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
839 const struct ovsrec_sflow *cfg = br->cfg->sflow;
840 struct ovsrec_controller **controllers;
841 struct ofproto_sflow_options oso;
842 size_t n_controllers;
846 ofproto_set_sflow(br->ofproto, NULL);
850 memset(&oso, 0, sizeof oso);
852 sset_init(&oso.targets);
853 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
855 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
857 oso.sampling_rate = *cfg->sampling;
860 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
862 oso.polling_interval = *cfg->polling;
865 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
867 oso.header_len = *cfg->header;
870 oso.sub_id = (*sflow_bridge_number)++;
871 oso.agent_device = cfg->agent;
873 oso.control_ip = NULL;
874 n_controllers = bridge_get_controllers(br, &controllers);
875 for (i = 0; i < n_controllers; i++) {
876 if (controllers[i]->local_ip) {
877 oso.control_ip = controllers[i]->local_ip;
881 ofproto_set_sflow(br->ofproto, &oso);
883 sset_destroy(&oso.targets);
887 port_configure_stp(const struct ofproto *ofproto, struct port *port,
888 struct ofproto_port_stp_settings *port_s,
889 int *port_num_counter, unsigned long *port_num_bitmap)
891 const char *config_str;
894 if (smap_get_bool(&port->cfg->other_config, "stp-enable", false)) {
895 port_s->enable = false;
898 port_s->enable = true;
901 /* STP over bonds is not supported. */
902 if (!list_is_singleton(&port->ifaces)) {
903 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
905 port_s->enable = false;
909 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
911 /* Internal ports shouldn't participate in spanning tree, so
913 if (!strcmp(iface->type, "internal")) {
914 VLOG_DBG("port %s: disable STP on internal ports", port->name);
915 port_s->enable = false;
919 /* STP on mirror output ports is not supported. */
920 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
921 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
922 port_s->enable = false;
926 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
928 unsigned long int port_num = strtoul(config_str, NULL, 0);
929 int port_idx = port_num - 1;
931 if (port_num < 1 || port_num > STP_MAX_PORTS) {
932 VLOG_ERR("port %s: invalid stp-port-num", port->name);
933 port_s->enable = false;
937 if (bitmap_is_set(port_num_bitmap, port_idx)) {
938 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
939 port->name, port_num);
940 port_s->enable = false;
943 bitmap_set1(port_num_bitmap, port_idx);
944 port_s->port_num = port_idx;
946 if (*port_num_counter > STP_MAX_PORTS) {
947 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
948 port_s->enable = false;
952 port_s->port_num = (*port_num_counter)++;
955 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
957 port_s->path_cost = strtoul(config_str, NULL, 10);
959 enum netdev_features current;
961 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
962 /* Couldn't get speed, so assume 100Mb/s. */
963 port_s->path_cost = 19;
967 mbps = netdev_features_to_bps(current) / 1000000;
968 port_s->path_cost = stp_convert_speed_to_cost(mbps);
972 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
974 port_s->priority = strtoul(config_str, NULL, 0);
976 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
980 /* Set spanning tree configuration on 'br'. */
982 bridge_configure_stp(struct bridge *br)
984 if (!br->cfg->stp_enable) {
985 ofproto_set_stp(br->ofproto, NULL);
987 struct ofproto_stp_settings br_s;
988 const char *config_str;
990 int port_num_counter;
991 unsigned long *port_num_bitmap;
993 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
995 uint8_t ea[ETH_ADDR_LEN];
997 if (eth_addr_from_string(config_str, ea)) {
998 br_s.system_id = eth_addr_to_uint64(ea);
1000 br_s.system_id = eth_addr_to_uint64(br->ea);
1001 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1002 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1005 br_s.system_id = eth_addr_to_uint64(br->ea);
1008 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1010 br_s.priority = strtoul(config_str, NULL, 0);
1012 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1015 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1017 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1019 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1022 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1024 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1026 br_s.max_age = STP_DEFAULT_MAX_AGE;
1029 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1031 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1033 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1036 /* Configure STP on the bridge. */
1037 if (ofproto_set_stp(br->ofproto, &br_s)) {
1038 VLOG_ERR("bridge %s: could not enable STP", br->name);
1042 /* Users must either set the port number with the "stp-port-num"
1043 * configuration on all ports or none. If manual configuration
1044 * is not done, then we allocate them sequentially. */
1045 port_num_counter = 0;
1046 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1047 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1048 struct ofproto_port_stp_settings port_s;
1049 struct iface *iface;
1051 port_configure_stp(br->ofproto, port, &port_s,
1052 &port_num_counter, port_num_bitmap);
1054 /* As bonds are not supported, just apply configuration to
1055 * all interfaces. */
1056 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1057 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1059 VLOG_ERR("port %s: could not enable STP", port->name);
1065 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1066 && port_num_counter) {
1067 VLOG_ERR("bridge %s: must manually configure all STP port "
1068 "IDs or none, disabling", br->name);
1069 ofproto_set_stp(br->ofproto, NULL);
1071 bitmap_free(port_num_bitmap);
1076 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1078 const struct port *port = port_lookup(br, name);
1079 return port && port_is_bond_fake_iface(port);
1083 port_is_bond_fake_iface(const struct port *port)
1085 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1089 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1091 struct bridge *br, *next;
1092 struct shash new_br;
1095 /* Collect new bridges' names and types. */
1096 shash_init(&new_br);
1097 for (i = 0; i < cfg->n_bridges; i++) {
1098 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1099 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1101 if (strchr(br_cfg->name, '/')) {
1102 /* Prevent remote ovsdb-server users from accessing arbitrary
1103 * directories, e.g. consider a bridge named "../../../etc/". */
1104 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1106 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1107 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1111 /* Get rid of deleted bridges or those whose types have changed.
1112 * Update 'cfg' of bridges that still exist. */
1113 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1114 br->cfg = shash_find_data(&new_br, br->name);
1115 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1116 br->cfg->datapath_type))) {
1121 /* Add new bridges. */
1122 for (i = 0; i < cfg->n_bridges; i++) {
1123 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1124 struct bridge *br = bridge_lookup(br_cfg->name);
1126 bridge_create(br_cfg);
1130 shash_destroy(&new_br);
1134 iface_set_ofp_port(struct iface *iface, int ofp_port)
1136 struct bridge *br = iface->port->bridge;
1138 assert(iface->ofp_port < 0 && ofp_port >= 0);
1139 iface->ofp_port = ofp_port;
1140 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1141 iface_set_ofport(iface->cfg, ofp_port);
1144 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1145 * Returns 0 if successful, otherwise a positive errno value. */
1147 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1148 struct netdev *netdev)
1152 error = netdev_set_config(netdev, &iface_cfg->options);
1154 VLOG_WARN("could not configure network device %s (%s)",
1155 iface_cfg->name, strerror(error));
1160 /* This function determines whether 'ofproto_port', which is attached to
1161 * br->ofproto's datapath, is one that we want in 'br'.
1163 * If it is, it returns true, after creating an iface (if necessary),
1164 * configuring the iface's netdev according to the iface's options, and setting
1165 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1167 * If, on the other hand, 'port' should be removed, it returns false. The
1168 * caller should later detach the port from br->ofproto. */
1170 bridge_refresh_one_ofp_port(struct bridge *br,
1171 const struct ofproto_port *ofproto_port)
1173 const char *name = ofproto_port->name;
1174 const char *type = ofproto_port->type;
1175 uint16_t ofp_port = ofproto_port->ofp_port;
1177 struct iface *iface = iface_lookup(br, name);
1179 /* Check that the name-to-number mapping is one-to-one. */
1180 if (iface->ofp_port >= 0) {
1181 VLOG_WARN("bridge %s: interface %s reported twice",
1184 } else if (iface_from_ofp_port(br, ofp_port)) {
1185 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1186 br->name, ofp_port);
1190 /* There's a configured interface named 'name'. */
1191 if (strcmp(type, iface->type)
1192 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1193 /* It's the wrong type, or it's the right type but can't be
1194 * configured as the user requested, so we must destroy it. */
1197 /* It's the right type and configured correctly. keep it. */
1198 iface_set_ofp_port(iface, ofp_port);
1201 } else if (bridge_has_bond_fake_iface(br, name)
1202 && !strcmp(type, "internal")) {
1203 /* It's a bond fake iface. Keep it. */
1206 /* There's no configured interface named 'name', but there might be an
1207 * interface of that name queued to be created.
1209 * If there is, and it has the correct type, then try to configure it
1210 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1211 * delete it and later try to re-add it. */
1212 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1214 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1215 && iface_create(br, if_cfg, ofp_port));
1219 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1220 * consistent with the ofp_ports in "br->ofproto". */
1222 bridge_refresh_ofp_port(struct bridge *br)
1224 struct ofproto_port_dump dump;
1225 struct ofproto_port ofproto_port;
1226 struct port *port, *port_next;
1228 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1229 hmap_clear(&br->ifaces);
1230 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1231 struct iface *iface;
1233 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1234 iface->ofp_port = -1;
1238 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1239 * already exist in the datapath and promote them to full fledged "struct
1240 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1241 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1242 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1243 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1244 garbage->ofp_port = ofproto_port.ofp_port;
1245 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1249 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1250 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1251 * later they can be added to ofproto. */
1252 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1253 struct iface *iface, *iface_next;
1255 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1256 if (iface->ofp_port < 0) {
1257 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1258 iface_destroy(iface);
1262 if (list_is_empty(&port->ifaces)) {
1268 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1269 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1270 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1273 * If successful, returns 0 and stores the network device in '*netdevp'. On
1274 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1276 iface_do_create(const struct bridge *br,
1277 const struct ovsrec_interface *iface_cfg,
1278 const struct ovsrec_port *port_cfg,
1279 int *ofp_portp, struct netdev **netdevp)
1281 struct netdev *netdev;
1284 error = netdev_open(iface_cfg->name,
1285 iface_get_type(iface_cfg, br->cfg), &netdev);
1287 VLOG_WARN("could not open network device %s (%s)",
1288 iface_cfg->name, strerror(error));
1292 error = iface_set_netdev_config(iface_cfg, netdev);
1297 if (*ofp_portp < 0) {
1300 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1304 *ofp_portp = ofp_port;
1306 VLOG_INFO("bridge %s: added interface %s on port %d",
1307 br->name, iface_cfg->name, *ofp_portp);
1309 VLOG_DBG("bridge %s: interface %s is on port %d",
1310 br->name, iface_cfg->name, *ofp_portp);
1313 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1314 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1322 netdev_close(netdev);
1326 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1327 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1328 * automatically allocated for the iface. Takes ownership of and
1329 * deallocates 'if_cfg'.
1331 * Return true if an iface is successfully created, false otherwise. */
1333 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1335 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1336 const struct ovsrec_port *port_cfg = if_cfg->parent;
1338 struct netdev *netdev;
1339 struct iface *iface;
1343 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1345 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1348 /* Do the bits that can fail up front.
1350 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1351 * internal datastructures may not be consistent. Eventually, when port
1352 * additions and deletions are cheaper, these calls should be removed. */
1354 assert(!iface_lookup(br, iface_cfg->name));
1355 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1358 iface_clear_db_record(iface_cfg);
1362 /* Get or create the port structure. */
1363 port = port_lookup(br, port_cfg->name);
1365 port = port_create(br, port_cfg);
1368 /* Create the iface structure. */
1369 iface = xzalloc(sizeof *iface);
1370 list_push_back(&port->ifaces, &iface->port_elem);
1371 hmap_insert(&br->iface_by_name, &iface->name_node,
1372 hash_string(iface_cfg->name, 0));
1374 iface->name = xstrdup(iface_cfg->name);
1375 iface->ofp_port = -1;
1376 iface->netdev = netdev;
1377 iface->type = iface_get_type(iface_cfg, br->cfg);
1378 iface->cfg = iface_cfg;
1380 iface_set_ofp_port(iface, ofp_port);
1382 /* Populate initial status in database. */
1383 iface_refresh_stats(iface);
1384 iface_refresh_status(iface);
1386 /* Add bond fake iface if necessary. */
1387 if (port_is_bond_fake_iface(port)) {
1388 struct ofproto_port ofproto_port;
1390 if (ofproto_port_query_by_name(br->ofproto, port->name,
1392 struct netdev *netdev;
1395 error = netdev_open(port->name, "internal", &netdev);
1397 ofproto_port_add(br->ofproto, netdev, NULL);
1398 netdev_close(netdev);
1400 VLOG_WARN("could not open network device %s (%s)",
1401 port->name, strerror(error));
1404 /* Already exists, nothing to do. */
1405 ofproto_port_destroy(&ofproto_port);
1412 /* Set Flow eviction threshold */
1414 bridge_configure_flow_eviction_threshold(struct bridge *br)
1416 const char *threshold_str;
1419 threshold_str = smap_get(&br->cfg->other_config,
1420 "flow-eviction-threshold");
1421 if (threshold_str) {
1422 threshold = strtoul(threshold_str, NULL, 10);
1424 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1426 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1429 /* Set forward BPDU option. */
1431 bridge_configure_forward_bpdu(struct bridge *br)
1433 ofproto_set_forward_bpdu(br->ofproto,
1434 smap_get_bool(&br->cfg->other_config,
1439 /* Set MAC aging time for 'br'. */
1441 bridge_configure_mac_idle_time(struct bridge *br)
1443 const char *idle_time_str;
1446 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1447 idle_time = (idle_time_str && atoi(idle_time_str)
1448 ? atoi(idle_time_str)
1449 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1450 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1454 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1455 struct iface **hw_addr_iface)
1457 struct hmapx mirror_output_ports;
1460 bool found_addr = false;
1464 *hw_addr_iface = NULL;
1466 /* Did the user request a particular MAC? */
1467 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1468 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1469 if (eth_addr_is_multicast(ea)) {
1470 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1471 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1472 } else if (eth_addr_is_zero(ea)) {
1473 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1479 /* Mirror output ports don't participate in picking the local hardware
1480 * address. ofproto can't help us find out whether a given port is a
1481 * mirror output because we haven't configured mirrors yet, so we need to
1482 * accumulate them ourselves. */
1483 hmapx_init(&mirror_output_ports);
1484 for (i = 0; i < br->cfg->n_mirrors; i++) {
1485 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1486 if (m->output_port) {
1487 hmapx_add(&mirror_output_ports, m->output_port);
1491 /* Otherwise choose the minimum non-local MAC address among all of the
1493 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1494 uint8_t iface_ea[ETH_ADDR_LEN];
1495 struct iface *candidate;
1496 struct iface *iface;
1498 /* Mirror output ports don't participate. */
1499 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1503 /* Choose the MAC address to represent the port. */
1505 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1506 /* Find the interface with this Ethernet address (if any) so that
1507 * we can provide the correct devname to the caller. */
1508 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1509 uint8_t candidate_ea[ETH_ADDR_LEN];
1510 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1511 && eth_addr_equals(iface_ea, candidate_ea)) {
1516 /* Choose the interface whose MAC address will represent the port.
1517 * The Linux kernel bonding code always chooses the MAC address of
1518 * the first slave added to a bond, and the Fedora networking
1519 * scripts always add slaves to a bond in alphabetical order, so
1520 * for compatibility we choose the interface with the name that is
1521 * first in alphabetical order. */
1522 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1523 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1528 /* The local port doesn't count (since we're trying to choose its
1529 * MAC address anyway). */
1530 if (iface->ofp_port == OFPP_LOCAL) {
1535 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1541 /* Compare against our current choice. */
1542 if (!eth_addr_is_multicast(iface_ea) &&
1543 !eth_addr_is_local(iface_ea) &&
1544 !eth_addr_is_reserved(iface_ea) &&
1545 !eth_addr_is_zero(iface_ea) &&
1546 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1548 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1549 *hw_addr_iface = iface;
1554 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1555 br->name, ETH_ADDR_ARGS(ea));
1557 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1558 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1559 *hw_addr_iface = NULL;
1560 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1561 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1564 hmapx_destroy(&mirror_output_ports);
1567 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1568 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1569 * an interface on 'br', then that interface must be passed in as
1570 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1571 * 'hw_addr_iface' must be passed in as a null pointer. */
1573 bridge_pick_datapath_id(struct bridge *br,
1574 const uint8_t bridge_ea[ETH_ADDR_LEN],
1575 struct iface *hw_addr_iface)
1578 * The procedure for choosing a bridge MAC address will, in the most
1579 * ordinary case, also choose a unique MAC that we can use as a datapath
1580 * ID. In some special cases, though, multiple bridges will end up with
1581 * the same MAC address. This is OK for the bridges, but it will confuse
1582 * the OpenFlow controller, because each datapath needs a unique datapath
1585 * Datapath IDs must be unique. It is also very desirable that they be
1586 * stable from one run to the next, so that policy set on a datapath
1589 const char *datapath_id;
1592 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1593 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1597 if (!hw_addr_iface) {
1599 * A purely internal bridge, that is, one that has no non-virtual
1600 * network devices on it at all, is difficult because it has no
1601 * natural unique identifier at all.
1603 * When the host is a XenServer, we handle this case by hashing the
1604 * host's UUID with the name of the bridge. Names of bridges are
1605 * persistent across XenServer reboots, although they can be reused if
1606 * an internal network is destroyed and then a new one is later
1607 * created, so this is fairly effective.
1609 * When the host is not a XenServer, we punt by using a random MAC
1610 * address on each run.
1612 const char *host_uuid = xenserver_get_host_uuid();
1614 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1615 dpid = dpid_from_hash(combined, strlen(combined));
1621 return eth_addr_to_uint64(bridge_ea);
1625 dpid_from_hash(const void *data, size_t n)
1627 uint8_t hash[SHA1_DIGEST_SIZE];
1629 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1630 sha1_bytes(data, n, hash);
1631 eth_addr_mark_random(hash);
1632 return eth_addr_to_uint64(hash);
1636 iface_refresh_status(struct iface *iface)
1640 enum netdev_features current;
1641 enum netdev_flags flags;
1647 if (iface_is_synthetic(iface)) {
1653 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1654 ovsrec_interface_set_status(iface->cfg, &smap);
1656 ovsrec_interface_set_status(iface->cfg, NULL);
1659 smap_destroy(&smap);
1661 error = netdev_get_flags(iface->netdev, &flags);
1663 ovsrec_interface_set_admin_state(iface->cfg,
1664 flags & NETDEV_UP ? "up" : "down");
1667 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1670 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1672 ovsrec_interface_set_duplex(iface->cfg,
1673 netdev_features_is_full_duplex(current)
1675 /* warning: uint64_t -> int64_t conversion */
1676 bps = netdev_features_to_bps(current);
1677 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1680 ovsrec_interface_set_duplex(iface->cfg, NULL);
1681 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1684 error = netdev_get_mtu(iface->netdev, &mtu);
1687 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1690 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1694 /* Writes 'iface''s CFM statistics to the database. */
1696 iface_refresh_cfm_stats(struct iface *iface)
1698 const struct ovsrec_interface *cfg = iface->cfg;
1700 const uint64_t *rmps;
1704 if (iface_is_synthetic(iface)) {
1708 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1711 const char *reasons[CFM_FAULT_N_REASONS];
1712 bool fault_bool = fault;
1716 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1717 int reason = 1 << i;
1718 if (fault & reason) {
1719 reasons[j++] = cfm_fault_reason_to_str(reason);
1723 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1724 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1726 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1727 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1730 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1731 iface->ofp_port, &rmps, &n_rmps);
1733 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1736 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1739 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1742 int64_t cfm_health = health;
1743 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1745 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1750 iface_refresh_stats(struct iface *iface)
1752 #define IFACE_STATS \
1753 IFACE_STAT(rx_packets, "rx_packets") \
1754 IFACE_STAT(tx_packets, "tx_packets") \
1755 IFACE_STAT(rx_bytes, "rx_bytes") \
1756 IFACE_STAT(tx_bytes, "tx_bytes") \
1757 IFACE_STAT(rx_dropped, "rx_dropped") \
1758 IFACE_STAT(tx_dropped, "tx_dropped") \
1759 IFACE_STAT(rx_errors, "rx_errors") \
1760 IFACE_STAT(tx_errors, "tx_errors") \
1761 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1762 IFACE_STAT(rx_over_errors, "rx_over_err") \
1763 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1764 IFACE_STAT(collisions, "collisions")
1766 #define IFACE_STAT(MEMBER, NAME) NAME,
1767 static char *keys[] = { IFACE_STATS };
1769 int64_t values[ARRAY_SIZE(keys)];
1772 struct netdev_stats stats;
1774 if (iface_is_synthetic(iface)) {
1778 /* Intentionally ignore return value, since errors will set 'stats' to
1779 * all-1s, and we will deal with that correctly below. */
1780 netdev_get_stats(iface->netdev, &stats);
1782 /* Copy statistics into values[] array. */
1784 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1787 assert(i == ARRAY_SIZE(keys));
1789 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1795 br_refresh_stp_status(struct bridge *br)
1797 struct smap smap = SMAP_INITIALIZER(&smap);
1798 struct ofproto *ofproto = br->ofproto;
1799 struct ofproto_stp_status status;
1801 if (ofproto_get_stp_status(ofproto, &status)) {
1805 if (!status.enabled) {
1806 ovsrec_bridge_set_status(br->cfg, NULL);
1810 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1811 STP_ID_ARGS(status.bridge_id));
1812 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1813 STP_ID_ARGS(status.designated_root));
1814 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1816 ovsrec_bridge_set_status(br->cfg, &smap);
1817 smap_destroy(&smap);
1821 port_refresh_stp_status(struct port *port)
1823 struct ofproto *ofproto = port->bridge->ofproto;
1824 struct iface *iface;
1825 struct ofproto_port_stp_status status;
1827 int64_t int_values[3];
1830 if (port_is_synthetic(port)) {
1834 /* STP doesn't currently support bonds. */
1835 if (!list_is_singleton(&port->ifaces)) {
1836 ovsrec_port_set_status(port->cfg, NULL);
1840 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1842 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1846 if (!status.enabled) {
1847 ovsrec_port_set_status(port->cfg, NULL);
1848 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1852 /* Set Status column. */
1854 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1855 smap_add(&smap, "stp_state", stp_state_name(status.state));
1856 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1857 smap_add(&smap, "stp_role", stp_role_name(status.role));
1858 ovsrec_port_set_status(port->cfg, &smap);
1859 smap_destroy(&smap);
1861 /* Set Statistics column. */
1862 keys[0] = "stp_tx_count";
1863 int_values[0] = status.tx_count;
1864 keys[1] = "stp_rx_count";
1865 int_values[1] = status.rx_count;
1866 keys[2] = "stp_error_count";
1867 int_values[2] = status.error_count;
1869 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1870 ARRAY_SIZE(int_values));
1874 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1876 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1880 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1882 bool enable = enable_system_stats(cfg);
1884 system_stats_enable(enable);
1886 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1891 run_system_stats(void)
1893 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1896 stats = system_stats_run();
1898 struct ovsdb_idl_txn *txn;
1899 struct ovsdb_datum datum;
1901 txn = ovsdb_idl_txn_create(idl);
1902 ovsdb_datum_from_smap(&datum, stats);
1903 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1905 ovsdb_idl_txn_commit(txn);
1906 ovsdb_idl_txn_destroy(txn);
1912 static inline const char *
1913 nx_role_to_str(enum nx_role role)
1918 case NX_ROLE_MASTER:
1923 return "*** INVALID ROLE ***";
1928 refresh_controller_status(void)
1932 const struct ovsrec_controller *cfg;
1936 /* Accumulate status for controllers on all bridges. */
1937 HMAP_FOR_EACH (br, node, &all_bridges) {
1938 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1941 /* Update each controller in the database with current status. */
1942 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1943 struct ofproto_controller_info *cinfo =
1944 shash_find_data(&info, cfg->target);
1947 struct smap smap = SMAP_INITIALIZER(&smap);
1948 const char **values = cinfo->pairs.values;
1949 const char **keys = cinfo->pairs.keys;
1952 for (i = 0; i < cinfo->pairs.n; i++) {
1953 smap_add(&smap, keys[i], values[i]);
1956 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1957 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1958 ovsrec_controller_set_status(cfg, &smap);
1959 smap_destroy(&smap);
1961 ovsrec_controller_set_is_connected(cfg, false);
1962 ovsrec_controller_set_role(cfg, NULL);
1963 ovsrec_controller_set_status(cfg, NULL);
1967 ofproto_free_ofproto_controller_info(&info);
1971 refresh_cfm_stats(void)
1973 static struct ovsdb_idl_txn *txn = NULL;
1978 txn = ovsdb_idl_txn_create(idl);
1980 HMAP_FOR_EACH (br, node, &all_bridges) {
1981 struct iface *iface;
1983 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1984 iface_refresh_cfm_stats(iface);
1989 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1990 ovsdb_idl_txn_destroy(txn);
1995 /* Performs periodic activity required by bridges that needs to be done with
1996 * the least possible latency.
1998 * It makes sense to call this function a couple of times per poll loop, to
1999 * provide a significant performance boost on some benchmarks with ofprotos
2000 * that use the ofproto-dpif implementation. */
2002 bridge_run_fast(void)
2006 HMAP_FOR_EACH (br, node, &all_bridges) {
2007 ofproto_run_fast(br->ofproto);
2014 static const struct ovsrec_open_vswitch null_cfg;
2015 const struct ovsrec_open_vswitch *cfg;
2016 struct ovsdb_idl_txn *reconf_txn = NULL;
2018 bool vlan_splinters_changed;
2021 ovsrec_open_vswitch_init((struct ovsrec_open_vswitch *) &null_cfg);
2023 /* (Re)configure if necessary. */
2024 if (!reconfiguring) {
2027 if (ovsdb_idl_is_lock_contended(idl)) {
2028 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2029 struct bridge *br, *next_br;
2031 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2032 "disabling this process until it goes away");
2034 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2038 } else if (!ovsdb_idl_has_lock(idl)) {
2042 cfg = ovsrec_open_vswitch_first(idl);
2044 /* Let each bridge do the work that it needs to do. */
2045 HMAP_FOR_EACH (br, node, &all_bridges) {
2046 ofproto_run(br->ofproto);
2049 /* Re-configure SSL. We do this on every trip through the main loop,
2050 * instead of just when the database changes, because the contents of the
2051 * key and certificate files can change without the database changing.
2053 * We do this before bridge_reconfigure() because that function might
2054 * initiate SSL connections and thus requires SSL to be configured. */
2055 if (cfg && cfg->ssl) {
2056 const struct ovsrec_ssl *ssl = cfg->ssl;
2058 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2059 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2062 if (!reconfiguring) {
2063 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2064 * usage has changed. */
2065 vlan_splinters_changed = false;
2066 if (vlan_splinters_enabled_anywhere) {
2067 HMAP_FOR_EACH (br, node, &all_bridges) {
2068 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2069 vlan_splinters_changed = true;
2075 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2076 idl_seqno = ovsdb_idl_get_seqno(idl);
2078 reconf_txn = ovsdb_idl_txn_create(idl);
2079 bridge_reconfigure(cfg);
2081 /* We still need to reconfigure to avoid dangling pointers to
2082 * now-destroyed ovsrec structures inside bridge data. */
2083 bridge_reconfigure(&null_cfg);
2088 if (reconfiguring) {
2091 reconf_txn = ovsdb_idl_txn_create(idl);
2093 if (bridge_reconfigure_continue(cfg)) {
2094 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2097 bridge_reconfigure_continue(&null_cfg);
2102 ovsdb_idl_txn_commit(reconf_txn);
2103 ovsdb_idl_txn_destroy(reconf_txn);
2107 /* Refresh interface and mirror stats if necessary. */
2108 if (time_msec() >= iface_stats_timer) {
2110 struct ovsdb_idl_txn *txn;
2112 txn = ovsdb_idl_txn_create(idl);
2113 HMAP_FOR_EACH (br, node, &all_bridges) {
2117 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2118 struct iface *iface;
2120 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2121 iface_refresh_stats(iface);
2122 iface_refresh_status(iface);
2126 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2127 mirror_refresh_stats(m);
2131 refresh_controller_status();
2132 ovsdb_idl_txn_commit(txn);
2133 ovsdb_idl_txn_destroy(txn); /* XXX */
2136 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2141 if (time_msec() >= db_limiter) {
2142 struct ovsdb_idl_txn *txn;
2144 txn = ovsdb_idl_txn_create(idl);
2145 HMAP_FOR_EACH (br, node, &all_bridges) {
2146 struct iface *iface;
2149 br_refresh_stp_status(br);
2151 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2152 port_refresh_stp_status(port);
2155 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2156 const char *link_state;
2157 int64_t link_resets;
2160 if (iface_is_synthetic(iface)) {
2164 current = ofproto_port_is_lacp_current(br->ofproto,
2168 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2170 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2173 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2174 ovsrec_interface_set_link_state(iface->cfg, link_state);
2176 link_resets = netdev_get_carrier_resets(iface->netdev);
2177 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2181 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2182 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2184 ovsdb_idl_txn_destroy(txn);
2187 refresh_cfm_stats();
2193 ovsdb_idl_wait(idl);
2195 if (reconfiguring) {
2196 poll_immediate_wake();
2199 if (!hmap_is_empty(&all_bridges)) {
2202 HMAP_FOR_EACH (br, node, &all_bridges) {
2203 ofproto_wait(br->ofproto);
2205 poll_timer_wait_until(iface_stats_timer);
2207 if (db_limiter > time_msec()) {
2208 poll_timer_wait_until(db_limiter);
2212 system_stats_wait();
2215 /* Adds some memory usage statistics for bridges into 'usage', for use with
2216 * memory_report(). */
2218 bridge_get_memory_usage(struct simap *usage)
2222 HMAP_FOR_EACH (br, node, &all_bridges) {
2223 ofproto_get_memory_usage(br->ofproto, usage);
2227 /* QoS unixctl user interface functions. */
2229 struct qos_unixctl_show_cbdata {
2231 struct iface *iface;
2235 qos_unixctl_show_cb(unsigned int queue_id,
2236 const struct smap *details,
2239 struct qos_unixctl_show_cbdata *data = aux;
2240 struct ds *ds = data->ds;
2241 struct iface *iface = data->iface;
2242 struct netdev_queue_stats stats;
2243 struct smap_node *node;
2246 ds_put_cstr(ds, "\n");
2248 ds_put_format(ds, "Queue %u:\n", queue_id);
2250 ds_put_cstr(ds, "Default:\n");
2253 SMAP_FOR_EACH (node, details) {
2254 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2257 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2259 if (stats.tx_packets != UINT64_MAX) {
2260 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2263 if (stats.tx_bytes != UINT64_MAX) {
2264 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2267 if (stats.tx_errors != UINT64_MAX) {
2268 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2271 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2272 queue_id, strerror(error));
2277 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2278 const char *argv[], void *aux OVS_UNUSED)
2280 struct ds ds = DS_EMPTY_INITIALIZER;
2281 struct smap smap = SMAP_INITIALIZER(&smap);
2282 struct iface *iface;
2284 struct smap_node *node;
2285 struct qos_unixctl_show_cbdata data;
2288 iface = iface_find(argv[1]);
2290 unixctl_command_reply_error(conn, "no such interface");
2294 netdev_get_qos(iface->netdev, &type, &smap);
2296 if (*type != '\0') {
2297 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2299 SMAP_FOR_EACH (node, &smap) {
2300 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2305 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2308 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2310 unixctl_command_reply(conn, ds_cstr(&ds));
2312 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2313 unixctl_command_reply_error(conn, ds_cstr(&ds));
2316 smap_destroy(&smap);
2320 /* Bridge reconfiguration functions. */
2322 bridge_create(const struct ovsrec_bridge *br_cfg)
2326 assert(!bridge_lookup(br_cfg->name));
2327 br = xzalloc(sizeof *br);
2329 br->name = xstrdup(br_cfg->name);
2330 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2333 /* Derive the default Ethernet address from the bridge's UUID. This should
2334 * be unique and it will be stable between ovs-vswitchd runs. */
2335 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2336 eth_addr_mark_random(br->default_ea);
2338 hmap_init(&br->ports);
2339 hmap_init(&br->ifaces);
2340 hmap_init(&br->iface_by_name);
2341 hmap_init(&br->mirrors);
2343 hmap_init(&br->if_cfg_todo);
2344 list_init(&br->ofpp_garbage);
2346 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2350 bridge_destroy(struct bridge *br)
2353 struct mirror *mirror, *next_mirror;
2354 struct port *port, *next_port;
2355 struct if_cfg *if_cfg, *next_if_cfg;
2356 struct ofpp_garbage *garbage, *next_garbage;
2358 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2361 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2362 mirror_destroy(mirror);
2364 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2365 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2368 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2369 &br->ofpp_garbage) {
2370 list_remove(&garbage->list_node);
2374 hmap_remove(&all_bridges, &br->node);
2375 ofproto_destroy(br->ofproto);
2376 hmap_destroy(&br->ifaces);
2377 hmap_destroy(&br->ports);
2378 hmap_destroy(&br->iface_by_name);
2379 hmap_destroy(&br->mirrors);
2380 hmap_destroy(&br->if_cfg_todo);
2387 static struct bridge *
2388 bridge_lookup(const char *name)
2392 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2393 if (!strcmp(br->name, name)) {
2400 /* Handle requests for a listing of all flows known by the OpenFlow
2401 * stack, including those normally hidden. */
2403 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2404 const char *argv[], void *aux OVS_UNUSED)
2409 br = bridge_lookup(argv[1]);
2411 unixctl_command_reply_error(conn, "Unknown bridge");
2416 ofproto_get_all_flows(br->ofproto, &results);
2418 unixctl_command_reply(conn, ds_cstr(&results));
2419 ds_destroy(&results);
2422 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2423 * connections and reconnect. If BRIDGE is not specified, then all bridges
2424 * drop their controller connections and reconnect. */
2426 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2427 const char *argv[], void *aux OVS_UNUSED)
2431 br = bridge_lookup(argv[1]);
2433 unixctl_command_reply_error(conn, "Unknown bridge");
2436 ofproto_reconnect_controllers(br->ofproto);
2438 HMAP_FOR_EACH (br, node, &all_bridges) {
2439 ofproto_reconnect_controllers(br->ofproto);
2442 unixctl_command_reply(conn, NULL);
2446 bridge_get_controllers(const struct bridge *br,
2447 struct ovsrec_controller ***controllersp)
2449 struct ovsrec_controller **controllers;
2450 size_t n_controllers;
2452 controllers = br->cfg->controller;
2453 n_controllers = br->cfg->n_controller;
2455 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2461 *controllersp = controllers;
2463 return n_controllers;
2467 bridge_queue_if_cfg(struct bridge *br,
2468 const struct ovsrec_interface *cfg,
2469 const struct ovsrec_port *parent)
2471 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2474 if_cfg->parent = parent;
2475 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2476 hash_string(if_cfg->cfg->name, 0));
2479 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2480 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2481 * 'br' needs to complete its configuration. */
2483 bridge_add_del_ports(struct bridge *br,
2484 const unsigned long int *splinter_vlans)
2486 struct shash_node *port_node;
2487 struct port *port, *next;
2488 struct shash new_ports;
2491 assert(hmap_is_empty(&br->if_cfg_todo));
2493 /* Collect new ports. */
2494 shash_init(&new_ports);
2495 for (i = 0; i < br->cfg->n_ports; i++) {
2496 const char *name = br->cfg->ports[i]->name;
2497 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2498 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2502 if (bridge_get_controllers(br, NULL)
2503 && !shash_find(&new_ports, br->name)) {
2504 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2505 br->name, br->name);
2507 ovsrec_interface_init(&br->synth_local_iface);
2508 ovsrec_port_init(&br->synth_local_port);
2510 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2511 br->synth_local_port.n_interfaces = 1;
2512 br->synth_local_port.name = br->name;
2514 br->synth_local_iface.name = br->name;
2515 br->synth_local_iface.type = "internal";
2517 br->synth_local_ifacep = &br->synth_local_iface;
2519 shash_add(&new_ports, br->name, &br->synth_local_port);
2522 if (splinter_vlans) {
2523 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2526 /* Get rid of deleted ports.
2527 * Get rid of deleted interfaces on ports that still exist. */
2528 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2529 port->cfg = shash_find_data(&new_ports, port->name);
2533 port_del_ifaces(port);
2537 /* Update iface->cfg and iface->type in interfaces that still exist.
2538 * Add new interfaces to creation queue. */
2539 SHASH_FOR_EACH (port_node, &new_ports) {
2540 const struct ovsrec_port *port = port_node->data;
2543 for (i = 0; i < port->n_interfaces; i++) {
2544 const struct ovsrec_interface *cfg = port->interfaces[i];
2545 struct iface *iface = iface_lookup(br, cfg->name);
2546 const char *type = iface_get_type(cfg, br->cfg);
2551 } else if (strcmp(type, "null")) {
2552 bridge_queue_if_cfg(br, cfg, port);
2557 shash_destroy(&new_ports);
2560 /* Initializes 'oc' appropriately as a management service controller for
2563 * The caller must free oc->target when it is no longer needed. */
2565 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2566 struct ofproto_controller *oc)
2568 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2569 oc->max_backoff = 0;
2570 oc->probe_interval = 60;
2571 oc->band = OFPROTO_OUT_OF_BAND;
2573 oc->burst_limit = 0;
2574 oc->enable_async_msgs = true;
2577 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2579 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2580 struct ofproto_controller *oc)
2584 oc->target = c->target;
2585 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2586 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2587 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2588 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2589 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2590 oc->burst_limit = (c->controller_burst_limit
2591 ? *c->controller_burst_limit : 0);
2592 oc->enable_async_msgs = (!c->enable_async_messages
2593 || *c->enable_async_messages);
2594 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2595 if (dscp < 0 || dscp > 63) {
2596 dscp = DSCP_DEFAULT;
2601 /* Configures the IP stack for 'br''s local interface properly according to the
2602 * configuration in 'c'. */
2604 bridge_configure_local_iface_netdev(struct bridge *br,
2605 struct ovsrec_controller *c)
2607 struct netdev *netdev;
2608 struct in_addr mask, gateway;
2610 struct iface *local_iface;
2613 /* If there's no local interface or no IP address, give up. */
2614 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2615 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2619 /* Bring up the local interface. */
2620 netdev = local_iface->netdev;
2621 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2623 /* Configure the IP address and netmask. */
2624 if (!c->local_netmask
2625 || !inet_aton(c->local_netmask, &mask)
2627 mask.s_addr = guess_netmask(ip.s_addr);
2629 if (!netdev_set_in4(netdev, ip, mask)) {
2630 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2631 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2634 /* Configure the default gateway. */
2635 if (c->local_gateway
2636 && inet_aton(c->local_gateway, &gateway)
2637 && gateway.s_addr) {
2638 if (!netdev_add_router(netdev, gateway)) {
2639 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2640 br->name, IP_ARGS(&gateway.s_addr));
2645 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2646 * in either string are treated as equal to any number of slashes in the other,
2647 * e.g. "x///y" is equal to "x/y". */
2649 equal_pathnames(const char *a, const char *b)
2653 a += strspn(a, "/");
2654 b += strspn(b, "/");
2655 } else if (*a == '\0') {
2666 bridge_configure_remotes(struct bridge *br,
2667 const struct sockaddr_in *managers, size_t n_managers)
2669 bool disable_in_band;
2671 struct ovsrec_controller **controllers;
2672 size_t n_controllers;
2674 enum ofproto_fail_mode fail_mode;
2676 struct ofproto_controller *ocs;
2680 /* Check if we should disable in-band control on this bridge. */
2681 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2684 /* Set OpenFlow queue ID for in-band control. */
2685 ofproto_set_in_band_queue(br->ofproto,
2686 smap_get_int(&br->cfg->other_config,
2687 "in-band-queue", -1));
2689 if (disable_in_band) {
2690 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2692 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2695 n_controllers = bridge_get_controllers(br, &controllers);
2697 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2700 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2701 for (i = 0; i < n_controllers; i++) {
2702 struct ovsrec_controller *c = controllers[i];
2704 if (!strncmp(c->target, "punix:", 6)
2705 || !strncmp(c->target, "unix:", 5)) {
2706 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2709 whitelist = xasprintf("unix:%s/%s.controller",
2710 ovs_rundir(), br->name);
2711 if (!equal_pathnames(c->target, whitelist)) {
2712 /* Prevent remote ovsdb-server users from accessing arbitrary
2713 * Unix domain sockets and overwriting arbitrary local
2715 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2716 "controller \"%s\" due to possibility for remote "
2717 "exploit. Instead, specify whitelisted \"%s\" or "
2718 "connect to \"unix:%s/%s.mgmt\" (which is always "
2719 "available without special configuration).",
2720 br->name, c->target, whitelist,
2721 ovs_rundir(), br->name);
2729 bridge_configure_local_iface_netdev(br, c);
2730 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2731 if (disable_in_band) {
2732 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2737 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2738 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2741 /* Set the fail-mode. */
2742 fail_mode = !br->cfg->fail_mode
2743 || !strcmp(br->cfg->fail_mode, "standalone")
2744 ? OFPROTO_FAIL_STANDALONE
2745 : OFPROTO_FAIL_SECURE;
2746 ofproto_set_fail_mode(br->ofproto, fail_mode);
2748 /* Configure OpenFlow controller connection snooping. */
2749 if (!ofproto_has_snoops(br->ofproto)) {
2753 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2754 ovs_rundir(), br->name));
2755 ofproto_set_snoops(br->ofproto, &snoops);
2756 sset_destroy(&snoops);
2761 bridge_configure_tables(struct bridge *br)
2763 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2767 n_tables = ofproto_get_n_tables(br->ofproto);
2769 for (i = 0; i < n_tables; i++) {
2770 struct ofproto_table_settings s;
2773 s.max_flows = UINT_MAX;
2777 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2778 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2781 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2782 s.max_flows = *cfg->flow_limit;
2784 if (cfg->overflow_policy
2785 && !strcmp(cfg->overflow_policy, "evict")) {
2788 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2789 for (k = 0; k < cfg->n_groups; k++) {
2790 const char *string = cfg->groups[k];
2793 msg = mf_parse_subfield__(&s.groups[k], &string);
2795 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2796 "'groups' (%s)", br->name, i, msg);
2798 } else if (*string) {
2799 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2800 "element '%s' contains trailing garbage",
2801 br->name, i, cfg->groups[k]);
2809 ofproto_configure_table(br->ofproto, i, &s);
2813 for (; j < br->cfg->n_flow_tables; j++) {
2814 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2815 "%"PRId64" not supported by this datapath", br->name,
2816 br->cfg->key_flow_tables[j]);
2820 /* Port functions. */
2822 static struct port *
2823 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2827 port = xzalloc(sizeof *port);
2829 port->name = xstrdup(cfg->name);
2831 list_init(&port->ifaces);
2833 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2837 /* Deletes interfaces from 'port' that are no longer configured for it. */
2839 port_del_ifaces(struct port *port)
2841 struct iface *iface, *next;
2842 struct sset new_ifaces;
2845 /* Collect list of new interfaces. */
2846 sset_init(&new_ifaces);
2847 for (i = 0; i < port->cfg->n_interfaces; i++) {
2848 const char *name = port->cfg->interfaces[i]->name;
2849 const char *type = port->cfg->interfaces[i]->type;
2850 if (strcmp(type, "null")) {
2851 sset_add(&new_ifaces, name);
2855 /* Get rid of deleted interfaces. */
2856 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2857 if (!sset_contains(&new_ifaces, iface->name)) {
2858 iface_destroy(iface);
2862 sset_destroy(&new_ifaces);
2866 port_destroy(struct port *port)
2869 struct bridge *br = port->bridge;
2870 struct iface *iface, *next;
2873 ofproto_bundle_unregister(br->ofproto, port);
2876 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2877 iface_destroy(iface);
2880 hmap_remove(&br->ports, &port->hmap_node);
2886 static struct port *
2887 port_lookup(const struct bridge *br, const char *name)
2891 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2893 if (!strcmp(port->name, name)) {
2901 enable_lacp(struct port *port, bool *activep)
2903 if (!port->cfg->lacp) {
2904 /* XXX when LACP implementation has been sufficiently tested, enable by
2905 * default and make active on bonded ports. */
2907 } else if (!strcmp(port->cfg->lacp, "off")) {
2909 } else if (!strcmp(port->cfg->lacp, "active")) {
2912 } else if (!strcmp(port->cfg->lacp, "passive")) {
2916 VLOG_WARN("port %s: unknown LACP mode %s",
2917 port->name, port->cfg->lacp);
2922 static struct lacp_settings *
2923 port_configure_lacp(struct port *port, struct lacp_settings *s)
2925 const char *lacp_time, *system_id;
2928 if (!enable_lacp(port, &s->active)) {
2932 s->name = port->name;
2934 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2936 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2937 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2938 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2939 " address.", port->name, system_id);
2943 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2946 if (eth_addr_is_zero(s->id)) {
2947 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2951 /* Prefer bondable links if unspecified. */
2952 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2954 s->priority = (priority > 0 && priority <= UINT16_MAX
2956 : UINT16_MAX - !list_is_short(&port->ifaces));
2958 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2959 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2964 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2966 int priority, portid, key;
2968 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2969 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2971 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2973 if (portid <= 0 || portid > UINT16_MAX) {
2974 portid = iface->ofp_port;
2977 if (priority <= 0 || priority > UINT16_MAX) {
2978 priority = UINT16_MAX;
2981 if (key < 0 || key > UINT16_MAX) {
2985 s->name = iface->name;
2987 s->priority = priority;
2992 port_configure_bond(struct port *port, struct bond_settings *s,
2993 uint32_t *bond_stable_ids)
2995 const char *detect_s;
2996 struct iface *iface;
2997 int miimon_interval;
3000 s->name = port->name;
3002 if (port->cfg->bond_mode) {
3003 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3004 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3005 port->name, port->cfg->bond_mode,
3006 bond_mode_to_string(s->balance));
3009 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3011 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3012 * active-backup. At some point we should remove this warning. */
3013 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3014 " in previous versions, the default bond_mode was"
3015 " balance-slb", port->name,
3016 bond_mode_to_string(s->balance));
3018 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3019 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3020 "please use another bond type or disable flood_vlans",
3024 miimon_interval = smap_get_int(&port->cfg->other_config,
3025 "bond-miimon-interval", 0);
3026 if (miimon_interval <= 0) {
3027 miimon_interval = 200;
3030 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3031 if (!detect_s || !strcmp(detect_s, "carrier")) {
3032 miimon_interval = 0;
3033 } else if (strcmp(detect_s, "miimon")) {
3034 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3035 "defaulting to carrier", port->name, detect_s);
3036 miimon_interval = 0;
3039 s->up_delay = MAX(0, port->cfg->bond_updelay);
3040 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3041 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3042 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3043 "bond-rebalance-interval", 10000);
3044 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3045 s->rebalance_interval = 1000;
3048 s->fake_iface = port->cfg->bond_fake_iface;
3051 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3052 long long stable_id;
3054 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3056 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3057 stable_id = iface->ofp_port;
3059 bond_stable_ids[i++] = stable_id;
3061 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3065 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3066 * instead of obtaining it from the database. */
3068 port_is_synthetic(const struct port *port)
3070 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3073 /* Interface functions. */
3075 /* Returns the correct network device type for interface 'iface' in bridge
3078 iface_get_type(const struct ovsrec_interface *iface,
3079 const struct ovsrec_bridge *br)
3081 /* The local port always has type "internal". Other ports take their type
3082 * from the database and default to "system" if none is specified. */
3083 return (!strcmp(iface->name, br->name) ? "internal"
3084 : iface->type[0] ? iface->type
3089 iface_destroy(struct iface *iface)
3092 struct port *port = iface->port;
3093 struct bridge *br = port->bridge;
3095 if (br->ofproto && iface->ofp_port >= 0) {
3096 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3099 if (iface->ofp_port >= 0) {
3100 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3103 list_remove(&iface->port_elem);
3104 hmap_remove(&br->iface_by_name, &iface->name_node);
3106 netdev_close(iface->netdev);
3113 static struct iface *
3114 iface_lookup(const struct bridge *br, const char *name)
3116 struct iface *iface;
3118 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3119 &br->iface_by_name) {
3120 if (!strcmp(iface->name, name)) {
3128 static struct iface *
3129 iface_find(const char *name)
3131 const struct bridge *br;
3133 HMAP_FOR_EACH (br, node, &all_bridges) {
3134 struct iface *iface = iface_lookup(br, name);
3143 static struct if_cfg *
3144 if_cfg_lookup(const struct bridge *br, const char *name)
3146 struct if_cfg *if_cfg;
3148 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3150 if (!strcmp(if_cfg->cfg->name, name)) {
3158 static struct iface *
3159 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3161 struct iface *iface;
3163 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3164 hash_int(ofp_port, 0), &br->ifaces) {
3165 if (iface->ofp_port == ofp_port) {
3172 /* Set Ethernet address of 'iface', if one is specified in the configuration
3175 iface_set_mac(struct iface *iface)
3177 uint8_t ea[ETH_ADDR_LEN];
3179 if (!strcmp(iface->type, "internal")
3180 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3181 if (iface->ofp_port == OFPP_LOCAL) {
3182 VLOG_ERR("interface %s: ignoring mac in Interface record "
3183 "(use Bridge record to set local port's mac)",
3185 } else if (eth_addr_is_multicast(ea)) {
3186 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3189 int error = netdev_set_etheraddr(iface->netdev, ea);
3191 VLOG_ERR("interface %s: setting MAC failed (%s)",
3192 iface->name, strerror(error));
3198 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3200 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3202 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3203 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3207 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3208 * sets the "ofport" field to -1.
3210 * This is appropriate when 'if_cfg''s interface cannot be created or is
3211 * otherwise invalid. */
3213 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3215 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3216 iface_set_ofport(if_cfg, -1);
3217 ovsrec_interface_set_status(if_cfg, NULL);
3218 ovsrec_interface_set_admin_state(if_cfg, NULL);
3219 ovsrec_interface_set_duplex(if_cfg, NULL);
3220 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3221 ovsrec_interface_set_link_state(if_cfg, NULL);
3222 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3223 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3224 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3225 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3226 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3227 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3231 struct iface_delete_queues_cbdata {
3232 struct netdev *netdev;
3233 const struct ovsdb_datum *queues;
3237 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3239 union ovsdb_atom atom;
3241 atom.integer = target;
3242 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3246 iface_delete_queues(unsigned int queue_id,
3247 const struct smap *details OVS_UNUSED, void *cbdata_)
3249 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3251 if (!queue_ids_include(cbdata->queues, queue_id)) {
3252 netdev_delete_queue(cbdata->netdev, queue_id);
3257 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3259 struct ofpbuf queues_buf;
3261 ofpbuf_init(&queues_buf, 0);
3263 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3264 netdev_set_qos(iface->netdev, NULL, NULL);
3266 struct iface_delete_queues_cbdata cbdata;
3270 /* Configure top-level Qos for 'iface'. */
3271 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3273 /* Deconfigure queues that were deleted. */
3274 cbdata.netdev = iface->netdev;
3275 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3277 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3279 /* Configure queues for 'iface'. */
3281 for (i = 0; i < qos->n_queues; i++) {
3282 const struct ovsrec_queue *queue = qos->value_queues[i];
3283 unsigned int queue_id = qos->key_queues[i];
3285 if (queue_id == 0) {
3289 if (queue->n_dscp == 1) {
3290 struct ofproto_port_queue *port_queue;
3292 port_queue = ofpbuf_put_uninit(&queues_buf,
3293 sizeof *port_queue);
3294 port_queue->queue = queue_id;
3295 port_queue->dscp = queue->dscp[0];
3298 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3301 struct smap details;
3303 smap_init(&details);
3304 netdev_set_queue(iface->netdev, 0, &details);
3305 smap_destroy(&details);
3309 if (iface->ofp_port >= 0) {
3310 const struct ofproto_port_queue *port_queues = queues_buf.data;
3311 size_t n_queues = queues_buf.size / sizeof *port_queues;
3313 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3314 port_queues, n_queues);
3317 netdev_set_policing(iface->netdev,
3318 iface->cfg->ingress_policing_rate,
3319 iface->cfg->ingress_policing_burst);
3321 ofpbuf_uninit(&queues_buf);
3325 iface_configure_cfm(struct iface *iface)
3327 const struct ovsrec_interface *cfg = iface->cfg;
3328 const char *opstate_str;
3329 const char *cfm_ccm_vlan;
3330 struct cfm_settings s;
3332 if (!cfg->n_cfm_mpid) {
3333 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3337 s.mpid = *cfg->cfm_mpid;
3338 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3339 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3340 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3342 if (s.interval <= 0) {
3346 if (!cfm_ccm_vlan) {
3348 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3349 s.ccm_vlan = CFM_RANDOM_VLAN;
3351 s.ccm_vlan = atoi(cfm_ccm_vlan);
3352 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3357 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3360 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3361 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3363 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3366 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3367 * instead of obtaining it from the database. */
3369 iface_is_synthetic(const struct iface *iface)
3371 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3375 /* Port mirroring. */
3377 static struct mirror *
3378 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3382 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3383 if (uuid_equals(uuid, &m->uuid)) {
3391 bridge_configure_mirrors(struct bridge *br)
3393 const struct ovsdb_datum *mc;
3394 unsigned long *flood_vlans;
3395 struct mirror *m, *next;
3398 /* Get rid of deleted mirrors. */
3399 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3400 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3401 union ovsdb_atom atom;
3403 atom.uuid = m->uuid;
3404 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3409 /* Add new mirrors and reconfigure existing ones. */
3410 for (i = 0; i < br->cfg->n_mirrors; i++) {
3411 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3412 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3414 m = mirror_create(br, cfg);
3417 if (!mirror_configure(m)) {
3422 /* Update flooded vlans (for RSPAN). */
3423 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3424 br->cfg->n_flood_vlans);
3425 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3426 bitmap_free(flood_vlans);
3429 static struct mirror *
3430 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3434 m = xzalloc(sizeof *m);
3435 m->uuid = cfg->header_.uuid;
3436 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3438 m->name = xstrdup(cfg->name);
3444 mirror_destroy(struct mirror *m)
3447 struct bridge *br = m->bridge;
3450 ofproto_mirror_unregister(br->ofproto, m);
3453 hmap_remove(&br->mirrors, &m->hmap_node);
3460 mirror_collect_ports(struct mirror *m,
3461 struct ovsrec_port **in_ports, int n_in_ports,
3462 void ***out_portsp, size_t *n_out_portsp)
3464 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3465 size_t n_out_ports = 0;
3468 for (i = 0; i < n_in_ports; i++) {
3469 const char *name = in_ports[i]->name;
3470 struct port *port = port_lookup(m->bridge, name);
3472 out_ports[n_out_ports++] = port;
3474 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3475 "port %s", m->bridge->name, m->name, name);
3478 *out_portsp = out_ports;
3479 *n_out_portsp = n_out_ports;
3483 mirror_configure(struct mirror *m)
3485 const struct ovsrec_mirror *cfg = m->cfg;
3486 struct ofproto_mirror_settings s;
3489 if (strcmp(cfg->name, m->name)) {
3491 m->name = xstrdup(cfg->name);
3495 /* Get output port or VLAN. */
3496 if (cfg->output_port) {
3497 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3498 if (!s.out_bundle) {
3499 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3500 m->bridge->name, m->name);
3503 s.out_vlan = UINT16_MAX;
3505 if (cfg->output_vlan) {
3506 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3507 "output vlan; ignoring output vlan",
3508 m->bridge->name, m->name);
3510 } else if (cfg->output_vlan) {
3511 /* The database should prevent invalid VLAN values. */
3512 s.out_bundle = NULL;
3513 s.out_vlan = *cfg->output_vlan;
3515 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3516 m->bridge->name, m->name);
3520 /* Get port selection. */
3521 if (cfg->select_all) {
3522 size_t n_ports = hmap_count(&m->bridge->ports);
3523 void **ports = xmalloc(n_ports * sizeof *ports);
3528 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3538 /* Get ports, dropping ports that don't exist.
3539 * The IDL ensures that there are no duplicates. */
3540 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3541 &s.srcs, &s.n_srcs);
3542 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3543 &s.dsts, &s.n_dsts);
3546 /* Get VLAN selection. */
3547 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3550 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3553 if (s.srcs != s.dsts) {
3562 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3564 * This is deprecated. It is only for compatibility with broken device drivers
3565 * in old versions of Linux that do not properly support VLANs when VLAN
3566 * devices are not used. When broken device drivers are no longer in
3567 * widespread use, we will delete these interfaces. */
3569 static struct ovsrec_port **recs;
3570 static size_t n_recs, allocated_recs;
3572 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3573 * splinters are reconfigured. */
3575 register_rec(struct ovsrec_port *rec)
3577 if (n_recs >= allocated_recs) {
3578 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3580 recs[n_recs++] = rec;
3583 /* Frees all of the ports registered with register_reg(). */
3585 free_registered_recs(void)
3589 for (i = 0; i < n_recs; i++) {
3590 struct ovsrec_port *port = recs[i];
3593 for (j = 0; j < port->n_interfaces; j++) {
3594 struct ovsrec_interface *iface = port->interfaces[j];
3599 smap_destroy(&port->other_config);
3600 free(port->interfaces);
3608 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3611 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3613 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3617 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3620 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3621 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3622 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3625 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3626 * use, returns NULL.
3628 * Updates 'vlan_splinters_enabled_anywhere'. */
3629 static unsigned long int *
3630 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3632 unsigned long int *splinter_vlans;
3633 struct sset splinter_ifaces;
3634 const char *real_dev_name;
3635 struct shash *real_devs;
3636 struct shash_node *node;
3640 /* Free space allocated for synthesized ports and interfaces, since we're
3641 * in the process of reconstructing all of them. */
3642 free_registered_recs();
3644 splinter_vlans = bitmap_allocate(4096);
3645 sset_init(&splinter_ifaces);
3646 vlan_splinters_enabled_anywhere = false;
3647 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3648 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3651 for (j = 0; j < br_cfg->n_ports; j++) {
3652 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3655 for (k = 0; k < port_cfg->n_interfaces; k++) {
3656 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3658 if (vlan_splinters_is_enabled(iface_cfg)) {
3659 vlan_splinters_enabled_anywhere = true;
3660 sset_add(&splinter_ifaces, iface_cfg->name);
3661 vlan_bitmap_from_array__(port_cfg->trunks,
3667 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3668 bitmap_set1(splinter_vlans, *port_cfg->tag);
3673 if (!vlan_splinters_enabled_anywhere) {
3674 free(splinter_vlans);
3675 sset_destroy(&splinter_ifaces);
3679 HMAP_FOR_EACH (br, node, &all_bridges) {
3681 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3685 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3686 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3687 * device to be created for it. */
3688 bitmap_set0(splinter_vlans, 0);
3689 bitmap_set0(splinter_vlans, 4095);
3691 /* Delete all VLAN devices that we don't need. */
3693 real_devs = vlandev_get_real_devs();
3694 SHASH_FOR_EACH (node, real_devs) {
3695 const struct vlan_real_dev *real_dev = node->data;
3696 const struct vlan_dev *vlan_dev;
3697 bool real_dev_has_splinters;
3699 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3701 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3702 if (!real_dev_has_splinters
3703 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3704 struct netdev *netdev;
3706 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3707 if (!netdev_get_in4(netdev, NULL, NULL) ||
3708 !netdev_get_in6(netdev, NULL)) {
3709 vlandev_del(vlan_dev->name);
3711 /* It has an IP address configured, so we don't own
3712 * it. Don't delete it. */
3714 netdev_close(netdev);
3721 /* Add all VLAN devices that we need. */
3722 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3725 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3726 if (!vlandev_get_name(real_dev_name, vid)) {
3727 vlandev_add(real_dev_name, vid);
3734 sset_destroy(&splinter_ifaces);
3736 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3737 free(splinter_vlans);
3740 return splinter_vlans;
3743 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3746 configure_splinter_port(struct port *port)
3748 struct ofproto *ofproto = port->bridge->ofproto;
3749 uint16_t realdev_ofp_port;
3750 const char *realdev_name;
3751 struct iface *vlandev, *realdev;
3753 ofproto_bundle_unregister(port->bridge->ofproto, port);
3755 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3758 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3759 realdev = iface_lookup(port->bridge, realdev_name);
3760 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3762 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3766 static struct ovsrec_port *
3767 synthesize_splinter_port(const char *real_dev_name,
3768 const char *vlan_dev_name, int vid)
3770 struct ovsrec_interface *iface;
3771 struct ovsrec_port *port;
3773 iface = xmalloc(sizeof *iface);
3774 ovsrec_interface_init(iface);
3775 iface->name = xstrdup(vlan_dev_name);
3776 iface->type = "system";
3778 port = xmalloc(sizeof *port);
3779 ovsrec_port_init(port);
3780 port->interfaces = xmemdup(&iface, sizeof iface);
3781 port->n_interfaces = 1;
3782 port->name = xstrdup(vlan_dev_name);
3783 port->vlan_mode = "splinter";
3784 port->tag = xmalloc(sizeof *port->tag);
3787 smap_add(&port->other_config, "realdev", real_dev_name);
3793 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3794 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3795 * 1-bit in the 'splinter_vlans' bitmap. */
3797 add_vlan_splinter_ports(struct bridge *br,
3798 const unsigned long int *splinter_vlans,
3799 struct shash *ports)
3803 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3804 * we're modifying 'ports'. */
3805 for (i = 0; i < br->cfg->n_ports; i++) {
3806 const char *name = br->cfg->ports[i]->name;
3807 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3810 for (j = 0; j < port_cfg->n_interfaces; j++) {
3811 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3813 if (vlan_splinters_is_enabled(iface_cfg)) {
3814 const char *real_dev_name;
3817 real_dev_name = iface_cfg->name;
3818 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3819 const char *vlan_dev_name;
3821 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3823 && !shash_find(ports, vlan_dev_name)) {
3824 shash_add(ports, vlan_dev_name,
3825 synthesize_splinter_port(
3826 real_dev_name, vlan_dev_name, vid));
3835 mirror_refresh_stats(struct mirror *m)
3837 struct ofproto *ofproto = m->bridge->ofproto;
3838 uint64_t tx_packets, tx_bytes;
3841 size_t stat_cnt = 0;
3843 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3844 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3848 if (tx_packets != UINT64_MAX) {
3849 keys[stat_cnt] = "tx_packets";
3850 values[stat_cnt] = tx_packets;
3853 if (tx_bytes != UINT64_MAX) {
3854 keys[stat_cnt] = "tx_bytes";
3855 values[stat_cnt] = tx_bytes;
3859 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);