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 systems and interface
152 * statistics and pushes them into the database. */
153 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int 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 bridge_configure_mirrors(struct bridge *);
224 static struct mirror *mirror_create(struct bridge *,
225 const struct ovsrec_mirror *);
226 static void mirror_destroy(struct mirror *);
227 static bool mirror_configure(struct mirror *);
228 static void mirror_refresh_stats(struct mirror *);
230 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
231 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
232 static const char *iface_get_type(const struct ovsrec_interface *,
233 const struct ovsrec_bridge *);
234 static void iface_destroy(struct iface *);
235 static struct iface *iface_lookup(const struct bridge *, const char *name);
236 static struct iface *iface_find(const char *name);
237 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
238 static struct iface *iface_from_ofp_port(const struct bridge *,
240 static void iface_set_mac(struct iface *);
241 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
242 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
243 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
244 static void iface_configure_cfm(struct iface *);
245 static void iface_refresh_cfm_stats(struct iface *);
246 static void iface_refresh_stats(struct iface *);
247 static void iface_refresh_status(struct iface *);
248 static bool iface_is_synthetic(const struct iface *);
250 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
252 * This is deprecated. It is only for compatibility with broken device drivers
253 * in old versions of Linux that do not properly support VLANs when VLAN
254 * devices are not used. When broken device drivers are no longer in
255 * widespread use, we will delete these interfaces. */
257 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
258 static bool vlan_splinters_enabled_anywhere;
260 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
261 static unsigned long int *collect_splinter_vlans(
262 const struct ovsrec_open_vswitch *);
263 static void configure_splinter_port(struct port *);
264 static void add_vlan_splinter_ports(struct bridge *,
265 const unsigned long int *splinter_vlans,
266 struct shash *ports);
268 /* Public functions. */
270 /* Initializes the bridge module, configuring it to obtain its configuration
271 * from an OVSDB server accessed over 'remote', which should be a string in a
272 * form acceptable to ovsdb_idl_create(). */
274 bridge_init(const char *remote)
276 /* Create connection to database. */
277 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
278 idl_seqno = ovsdb_idl_get_seqno(idl);
279 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
281 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
282 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
289 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
290 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
291 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
293 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
294 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
295 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
312 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
314 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
315 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
316 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
317 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
319 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
321 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
323 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
324 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
326 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
338 /* Register unixctl commands. */
339 unixctl_command_register("qos/show", "interface", 1, 1,
340 qos_unixctl_show, NULL);
341 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
342 bridge_unixctl_dump_flows, NULL);
343 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
344 bridge_unixctl_reconnect, NULL);
354 struct bridge *br, *next_br;
356 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
359 ovsdb_idl_destroy(idl);
362 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
363 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
364 * responsible for freeing '*managersp' (with free()).
366 * You may be asking yourself "why does ovs-vswitchd care?", because
367 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
368 * should not be and in fact is not directly involved in that. But
369 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
370 * it has to tell in-band control where the managers are to enable that.
371 * (Thus, only managers connected in-band are collected.)
374 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
375 struct sockaddr_in **managersp, size_t *n_managersp)
377 struct sockaddr_in *managers = NULL;
378 size_t n_managers = 0;
382 /* Collect all of the potential targets from the "targets" columns of the
383 * rows pointed to by "manager_options", excluding any that are
386 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
387 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
389 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
390 sset_find_and_delete(&targets, m->target);
392 sset_add(&targets, m->target);
396 /* Now extract the targets' IP addresses. */
397 if (!sset_is_empty(&targets)) {
400 managers = xmalloc(sset_count(&targets) * sizeof *managers);
401 SSET_FOR_EACH (target, &targets) {
402 struct sockaddr_in *sin = &managers[n_managers];
404 if (stream_parse_target_with_default_ports(target,
412 sset_destroy(&targets);
414 *managersp = managers;
415 *n_managersp = n_managers;
419 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
421 unsigned long int *splinter_vlans;
424 COVERAGE_INC(bridge_reconfigure);
426 assert(!reconfiguring);
427 reconfiguring = true;
429 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
430 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
431 * configuration otherwise.
433 * This is mostly an update to bridge data structures. Nothing is pushed
434 * down to ofproto or lower layers. */
435 add_del_bridges(ovs_cfg);
436 splinter_vlans = collect_splinter_vlans(ovs_cfg);
437 HMAP_FOR_EACH (br, node, &all_bridges) {
438 bridge_add_del_ports(br, splinter_vlans);
440 free(splinter_vlans);
442 /* Delete datapaths that are no longer configured, and create ones which
443 * don't exist but should. */
444 bridge_update_ofprotos();
446 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
447 HMAP_FOR_EACH (br, node, &all_bridges) {
448 bridge_refresh_ofp_port(br);
451 /* Clear database records for "if_cfg"s which haven't been instantiated. */
452 HMAP_FOR_EACH (br, node, &all_bridges) {
453 struct if_cfg *if_cfg;
455 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
456 iface_clear_db_record(if_cfg->cfg);
462 bridge_reconfigure_ofp(void)
464 long long int deadline;
468 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
470 /* The kernel will reject any attempt to add a given port to a datapath if
471 * that port already belongs to a different datapath, so we must do all
472 * port deletions before any port additions. */
473 HMAP_FOR_EACH (br, node, &all_bridges) {
474 struct ofpp_garbage *garbage, *next;
476 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
477 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
478 * internal datastructures may not be consistent. Eventually, when
479 * port additions and deletions are cheaper, these calls should be
482 ofproto_port_del(br->ofproto, garbage->ofp_port);
483 list_remove(&garbage->list_node);
487 if (time_msec() >= deadline) {
494 HMAP_FOR_EACH (br, node, &all_bridges) {
495 struct if_cfg *if_cfg, *next;
497 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
498 iface_create(br, if_cfg, -1);
500 if (time_msec() >= deadline) {
510 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
512 struct sockaddr_in *managers;
513 int sflow_bridge_number;
518 assert(reconfiguring);
519 done = bridge_reconfigure_ofp();
521 /* Complete the configuration. */
522 sflow_bridge_number = 0;
523 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
524 HMAP_FOR_EACH (br, node, &all_bridges) {
527 /* We need the datapath ID early to allow LACP ports to use it as the
528 * default system ID. */
529 bridge_configure_datapath_id(br);
531 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
534 port_configure(port);
536 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
537 iface_configure_cfm(iface);
538 iface_configure_qos(iface, port->cfg->qos);
539 iface_set_mac(iface);
542 bridge_configure_mirrors(br);
543 bridge_configure_flow_eviction_threshold(br);
544 bridge_configure_forward_bpdu(br);
545 bridge_configure_mac_idle_time(br);
546 bridge_configure_remotes(br, managers, n_managers);
547 bridge_configure_netflow(br);
548 bridge_configure_sflow(br, &sflow_bridge_number);
549 bridge_configure_stp(br);
550 bridge_configure_tables(br);
555 /* ovs-vswitchd has completed initialization, so allow the process that
556 * forked us to exit successfully. */
557 daemonize_complete();
558 reconfiguring = false;
564 /* Delete ofprotos which aren't configured or have the wrong type. Create
565 * ofprotos which don't exist but need to. */
567 bridge_update_ofprotos(void)
569 struct bridge *br, *next;
574 /* Delete ofprotos with no bridge or with the wrong type. */
577 ofproto_enumerate_types(&types);
578 SSET_FOR_EACH (type, &types) {
581 ofproto_enumerate_names(type, &names);
582 SSET_FOR_EACH (name, &names) {
583 br = bridge_lookup(name);
584 if (!br || strcmp(type, br->type)) {
585 ofproto_delete(name, type);
589 sset_destroy(&names);
590 sset_destroy(&types);
592 /* Add ofprotos for bridges which don't have one yet. */
593 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
601 /* Remove ports from any datapath with the same name as 'br'. If we
602 * don't do this, creating 'br''s ofproto will fail because a port with
603 * the same name as its local port already exists. */
604 HMAP_FOR_EACH (br2, node, &all_bridges) {
605 struct ofproto_port ofproto_port;
611 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
613 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
615 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
618 ofproto_port_destroy(&ofproto_port);
622 error = ofproto_create(br->name, br->type, &br->ofproto);
624 VLOG_ERR("failed to create bridge %s: %s", br->name,
632 port_configure(struct port *port)
634 const struct ovsrec_port *cfg = port->cfg;
635 struct bond_settings bond_settings;
636 struct lacp_settings lacp_settings;
637 struct ofproto_bundle_settings s;
640 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
641 configure_splinter_port(port);
650 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
651 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
652 s.slaves[s.n_slaves++] = iface->ofp_port;
657 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
661 /* Get VLAN trunks. */
664 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
668 if (cfg->vlan_mode) {
669 if (!strcmp(cfg->vlan_mode, "access")) {
670 s.vlan_mode = PORT_VLAN_ACCESS;
671 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
672 s.vlan_mode = PORT_VLAN_TRUNK;
673 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
674 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
675 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
676 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
678 /* This "can't happen" because ovsdb-server should prevent it. */
679 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
680 s.vlan_mode = PORT_VLAN_TRUNK;
684 s.vlan_mode = PORT_VLAN_ACCESS;
686 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
690 s.vlan_mode = PORT_VLAN_TRUNK;
693 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
696 /* Get LACP settings. */
697 s.lacp = port_configure_lacp(port, &lacp_settings);
701 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
702 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
703 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
706 s.lacp_slaves = NULL;
709 /* Get bond settings. */
710 if (s.n_slaves > 1) {
711 s.bond = &bond_settings;
712 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
713 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
716 s.bond_stable_ids = NULL;
718 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
719 netdev_set_miimon_interval(iface->netdev, 0);
724 ofproto_bundle_register(port->bridge->ofproto, port, &s);
730 free(s.bond_stable_ids);
733 /* Pick local port hardware address and datapath ID for 'br'. */
735 bridge_configure_datapath_id(struct bridge *br)
737 uint8_t ea[ETH_ADDR_LEN];
739 struct iface *local_iface;
740 struct iface *hw_addr_iface;
743 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
744 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
746 int error = netdev_set_etheraddr(local_iface->netdev, ea);
748 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
749 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
750 "Ethernet address: %s",
751 br->name, strerror(error));
754 memcpy(br->ea, ea, ETH_ADDR_LEN);
756 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
757 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
758 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
759 ofproto_set_datapath_id(br->ofproto, dpid);
762 dpid_string = xasprintf("%016"PRIx64, dpid);
763 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
767 /* Set NetFlow configuration on 'br'. */
769 bridge_configure_netflow(struct bridge *br)
771 struct ovsrec_netflow *cfg = br->cfg->netflow;
772 struct netflow_options opts;
775 ofproto_set_netflow(br->ofproto, NULL);
779 memset(&opts, 0, sizeof opts);
781 /* Get default NetFlow configuration from datapath.
782 * Apply overrides from 'cfg'. */
783 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
784 if (cfg->engine_type) {
785 opts.engine_type = *cfg->engine_type;
787 if (cfg->engine_id) {
788 opts.engine_id = *cfg->engine_id;
791 /* Configure active timeout interval. */
792 opts.active_timeout = cfg->active_timeout;
793 if (!opts.active_timeout) {
794 opts.active_timeout = -1;
795 } else if (opts.active_timeout < 0) {
796 VLOG_WARN("bridge %s: active timeout interval set to negative "
797 "value, using default instead (%d seconds)", br->name,
798 NF_ACTIVE_TIMEOUT_DEFAULT);
799 opts.active_timeout = -1;
802 /* Add engine ID to interface number to disambiguate bridgs? */
803 opts.add_id_to_iface = cfg->add_id_to_interface;
804 if (opts.add_id_to_iface) {
805 if (opts.engine_id > 0x7f) {
806 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
807 "another vswitch, choose an engine id less than 128",
810 if (hmap_count(&br->ports) > 508) {
811 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
812 "another port when more than 508 ports are used",
818 sset_init(&opts.collectors);
819 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
822 if (ofproto_set_netflow(br->ofproto, &opts)) {
823 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
825 sset_destroy(&opts.collectors);
828 /* Set sFlow configuration on 'br'. */
830 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
832 const struct ovsrec_sflow *cfg = br->cfg->sflow;
833 struct ovsrec_controller **controllers;
834 struct ofproto_sflow_options oso;
835 size_t n_controllers;
839 ofproto_set_sflow(br->ofproto, NULL);
843 memset(&oso, 0, sizeof oso);
845 sset_init(&oso.targets);
846 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
848 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
850 oso.sampling_rate = *cfg->sampling;
853 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
855 oso.polling_interval = *cfg->polling;
858 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
860 oso.header_len = *cfg->header;
863 oso.sub_id = (*sflow_bridge_number)++;
864 oso.agent_device = cfg->agent;
866 oso.control_ip = NULL;
867 n_controllers = bridge_get_controllers(br, &controllers);
868 for (i = 0; i < n_controllers; i++) {
869 if (controllers[i]->local_ip) {
870 oso.control_ip = controllers[i]->local_ip;
874 ofproto_set_sflow(br->ofproto, &oso);
876 sset_destroy(&oso.targets);
880 port_configure_stp(const struct ofproto *ofproto, struct port *port,
881 struct ofproto_port_stp_settings *port_s,
882 int *port_num_counter, unsigned long *port_num_bitmap)
884 const char *config_str;
887 if (smap_get_bool(&port->cfg->other_config, "stp-enable", false)) {
888 port_s->enable = false;
891 port_s->enable = true;
894 /* STP over bonds is not supported. */
895 if (!list_is_singleton(&port->ifaces)) {
896 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
898 port_s->enable = false;
902 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
904 /* Internal ports shouldn't participate in spanning tree, so
906 if (!strcmp(iface->type, "internal")) {
907 VLOG_DBG("port %s: disable STP on internal ports", port->name);
908 port_s->enable = false;
912 /* STP on mirror output ports is not supported. */
913 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
914 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
915 port_s->enable = false;
919 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
921 unsigned long int port_num = strtoul(config_str, NULL, 0);
922 int port_idx = port_num - 1;
924 if (port_num < 1 || port_num > STP_MAX_PORTS) {
925 VLOG_ERR("port %s: invalid stp-port-num", port->name);
926 port_s->enable = false;
930 if (bitmap_is_set(port_num_bitmap, port_idx)) {
931 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
932 port->name, port_num);
933 port_s->enable = false;
936 bitmap_set1(port_num_bitmap, port_idx);
937 port_s->port_num = port_idx;
939 if (*port_num_counter > STP_MAX_PORTS) {
940 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
941 port_s->enable = false;
945 port_s->port_num = (*port_num_counter)++;
948 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
950 port_s->path_cost = strtoul(config_str, NULL, 10);
952 enum netdev_features current;
954 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
955 /* Couldn't get speed, so assume 100Mb/s. */
956 port_s->path_cost = 19;
960 mbps = netdev_features_to_bps(current) / 1000000;
961 port_s->path_cost = stp_convert_speed_to_cost(mbps);
965 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
967 port_s->priority = strtoul(config_str, NULL, 0);
969 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
973 /* Set spanning tree configuration on 'br'. */
975 bridge_configure_stp(struct bridge *br)
977 if (!br->cfg->stp_enable) {
978 ofproto_set_stp(br->ofproto, NULL);
980 struct ofproto_stp_settings br_s;
981 const char *config_str;
983 int port_num_counter;
984 unsigned long *port_num_bitmap;
986 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
988 uint8_t ea[ETH_ADDR_LEN];
990 if (eth_addr_from_string(config_str, ea)) {
991 br_s.system_id = eth_addr_to_uint64(ea);
993 br_s.system_id = eth_addr_to_uint64(br->ea);
994 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
995 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
998 br_s.system_id = eth_addr_to_uint64(br->ea);
1001 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1003 br_s.priority = strtoul(config_str, NULL, 0);
1005 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1008 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1010 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1012 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1015 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1017 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1019 br_s.max_age = STP_DEFAULT_MAX_AGE;
1022 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1024 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1026 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1029 /* Configure STP on the bridge. */
1030 if (ofproto_set_stp(br->ofproto, &br_s)) {
1031 VLOG_ERR("bridge %s: could not enable STP", br->name);
1035 /* Users must either set the port number with the "stp-port-num"
1036 * configuration on all ports or none. If manual configuration
1037 * is not done, then we allocate them sequentially. */
1038 port_num_counter = 0;
1039 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1040 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1041 struct ofproto_port_stp_settings port_s;
1042 struct iface *iface;
1044 port_configure_stp(br->ofproto, port, &port_s,
1045 &port_num_counter, port_num_bitmap);
1047 /* As bonds are not supported, just apply configuration to
1048 * all interfaces. */
1049 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1050 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1052 VLOG_ERR("port %s: could not enable STP", port->name);
1058 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1059 && port_num_counter) {
1060 VLOG_ERR("bridge %s: must manually configure all STP port "
1061 "IDs or none, disabling", br->name);
1062 ofproto_set_stp(br->ofproto, NULL);
1064 bitmap_free(port_num_bitmap);
1069 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1071 const struct port *port = port_lookup(br, name);
1072 return port && port_is_bond_fake_iface(port);
1076 port_is_bond_fake_iface(const struct port *port)
1078 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1082 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1084 struct bridge *br, *next;
1085 struct shash new_br;
1088 /* Collect new bridges' names and types. */
1089 shash_init(&new_br);
1090 for (i = 0; i < cfg->n_bridges; i++) {
1091 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1092 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1094 if (strchr(br_cfg->name, '/')) {
1095 /* Prevent remote ovsdb-server users from accessing arbitrary
1096 * directories, e.g. consider a bridge named "../../../etc/". */
1097 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1099 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1100 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1104 /* Get rid of deleted bridges or those whose types have changed.
1105 * Update 'cfg' of bridges that still exist. */
1106 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1107 br->cfg = shash_find_data(&new_br, br->name);
1108 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1109 br->cfg->datapath_type))) {
1114 /* Add new bridges. */
1115 for (i = 0; i < cfg->n_bridges; i++) {
1116 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1117 struct bridge *br = bridge_lookup(br_cfg->name);
1119 bridge_create(br_cfg);
1123 shash_destroy(&new_br);
1127 iface_set_ofp_port(struct iface *iface, int ofp_port)
1129 struct bridge *br = iface->port->bridge;
1131 assert(iface->ofp_port < 0 && ofp_port >= 0);
1132 iface->ofp_port = ofp_port;
1133 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1134 iface_set_ofport(iface->cfg, ofp_port);
1137 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1138 * Returns 0 if successful, otherwise a positive errno value. */
1140 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1141 struct netdev *netdev)
1145 error = netdev_set_config(netdev, &iface_cfg->options);
1147 VLOG_WARN("could not configure network device %s (%s)",
1148 iface_cfg->name, strerror(error));
1153 /* This function determines whether 'ofproto_port', which is attached to
1154 * br->ofproto's datapath, is one that we want in 'br'.
1156 * If it is, it returns true, after creating an iface (if necessary),
1157 * configuring the iface's netdev according to the iface's options, and setting
1158 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1160 * If, on the other hand, 'port' should be removed, it returns false. The
1161 * caller should later detach the port from br->ofproto. */
1163 bridge_refresh_one_ofp_port(struct bridge *br,
1164 const struct ofproto_port *ofproto_port)
1166 const char *name = ofproto_port->name;
1167 const char *type = ofproto_port->type;
1168 uint16_t ofp_port = ofproto_port->ofp_port;
1170 struct iface *iface = iface_lookup(br, name);
1172 /* Check that the name-to-number mapping is one-to-one. */
1173 if (iface->ofp_port >= 0) {
1174 VLOG_WARN("bridge %s: interface %s reported twice",
1177 } else if (iface_from_ofp_port(br, ofp_port)) {
1178 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1179 br->name, ofp_port);
1183 /* There's a configured interface named 'name'. */
1184 if (strcmp(type, iface->type)
1185 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1186 /* It's the wrong type, or it's the right type but can't be
1187 * configured as the user requested, so we must destroy it. */
1190 /* It's the right type and configured correctly. keep it. */
1191 iface_set_ofp_port(iface, ofp_port);
1194 } else if (bridge_has_bond_fake_iface(br, name)
1195 && !strcmp(type, "internal")) {
1196 /* It's a bond fake iface. Keep it. */
1199 /* There's no configured interface named 'name', but there might be an
1200 * interface of that name queued to be created.
1202 * If there is, and it has the correct type, then try to configure it
1203 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1204 * delete it and later try to re-add it. */
1205 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1207 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1208 && iface_create(br, if_cfg, ofp_port));
1212 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1213 * consistent with the ofp_ports in "br->ofproto". */
1215 bridge_refresh_ofp_port(struct bridge *br)
1217 struct ofproto_port_dump dump;
1218 struct ofproto_port ofproto_port;
1219 struct port *port, *port_next;
1221 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1222 hmap_clear(&br->ifaces);
1223 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1224 struct iface *iface;
1226 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1227 iface->ofp_port = -1;
1231 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1232 * already exist in the datapath and promote them to full fledged "struct
1233 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1234 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1235 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1236 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1237 garbage->ofp_port = ofproto_port.ofp_port;
1238 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1242 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1243 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1244 * later they can be added to ofproto. */
1245 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1246 struct iface *iface, *iface_next;
1248 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1249 if (iface->ofp_port < 0) {
1250 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1251 iface_destroy(iface);
1255 if (list_is_empty(&port->ifaces)) {
1261 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1262 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1263 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1266 * If successful, returns 0 and stores the network device in '*netdevp'. On
1267 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1269 iface_do_create(const struct bridge *br,
1270 const struct ovsrec_interface *iface_cfg,
1271 const struct ovsrec_port *port_cfg,
1272 int *ofp_portp, struct netdev **netdevp)
1274 struct netdev *netdev;
1277 error = netdev_open(iface_cfg->name,
1278 iface_get_type(iface_cfg, br->cfg), &netdev);
1280 VLOG_WARN("could not open network device %s (%s)",
1281 iface_cfg->name, strerror(error));
1285 error = iface_set_netdev_config(iface_cfg, netdev);
1290 if (*ofp_portp < 0) {
1293 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1297 *ofp_portp = ofp_port;
1299 VLOG_INFO("bridge %s: added interface %s on port %d",
1300 br->name, iface_cfg->name, *ofp_portp);
1302 VLOG_DBG("bridge %s: interface %s is on port %d",
1303 br->name, iface_cfg->name, *ofp_portp);
1306 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1307 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1315 netdev_close(netdev);
1319 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1320 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1321 * automatically allocated for the iface. Takes ownership of and
1322 * deallocates 'if_cfg'.
1324 * Return true if an iface is successfully created, false otherwise. */
1326 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1328 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1329 const struct ovsrec_port *port_cfg = if_cfg->parent;
1331 struct netdev *netdev;
1332 struct iface *iface;
1336 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1338 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1341 /* Do the bits that can fail up front.
1343 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1344 * internal datastructures may not be consistent. Eventually, when port
1345 * additions and deletions are cheaper, these calls should be removed. */
1347 assert(!iface_lookup(br, iface_cfg->name));
1348 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1351 iface_clear_db_record(iface_cfg);
1355 /* Get or create the port structure. */
1356 port = port_lookup(br, port_cfg->name);
1358 port = port_create(br, port_cfg);
1361 /* Create the iface structure. */
1362 iface = xzalloc(sizeof *iface);
1363 list_push_back(&port->ifaces, &iface->port_elem);
1364 hmap_insert(&br->iface_by_name, &iface->name_node,
1365 hash_string(iface_cfg->name, 0));
1367 iface->name = xstrdup(iface_cfg->name);
1368 iface->ofp_port = -1;
1369 iface->netdev = netdev;
1370 iface->type = iface_get_type(iface_cfg, br->cfg);
1371 iface->cfg = iface_cfg;
1373 iface_set_ofp_port(iface, ofp_port);
1375 /* Populate initial status in database. */
1376 iface_refresh_stats(iface);
1377 iface_refresh_status(iface);
1379 /* Add bond fake iface if necessary. */
1380 if (port_is_bond_fake_iface(port)) {
1381 struct ofproto_port ofproto_port;
1383 if (ofproto_port_query_by_name(br->ofproto, port->name,
1385 struct netdev *netdev;
1388 error = netdev_open(port->name, "internal", &netdev);
1390 ofproto_port_add(br->ofproto, netdev, NULL);
1391 netdev_close(netdev);
1393 VLOG_WARN("could not open network device %s (%s)",
1394 port->name, strerror(error));
1397 /* Already exists, nothing to do. */
1398 ofproto_port_destroy(&ofproto_port);
1405 /* Set Flow eviction threshold */
1407 bridge_configure_flow_eviction_threshold(struct bridge *br)
1409 const char *threshold_str;
1412 threshold_str = smap_get(&br->cfg->other_config,
1413 "flow-eviction-threshold");
1414 if (threshold_str) {
1415 threshold = strtoul(threshold_str, NULL, 10);
1417 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1419 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1422 /* Set forward BPDU option. */
1424 bridge_configure_forward_bpdu(struct bridge *br)
1426 ofproto_set_forward_bpdu(br->ofproto,
1427 smap_get_bool(&br->cfg->other_config,
1432 /* Set MAC aging time for 'br'. */
1434 bridge_configure_mac_idle_time(struct bridge *br)
1436 const char *idle_time_str;
1439 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1440 idle_time = (idle_time_str && atoi(idle_time_str)
1441 ? atoi(idle_time_str)
1442 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1443 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1447 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1448 struct iface **hw_addr_iface)
1450 struct hmapx mirror_output_ports;
1453 bool found_addr = false;
1457 *hw_addr_iface = NULL;
1459 /* Did the user request a particular MAC? */
1460 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1461 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1462 if (eth_addr_is_multicast(ea)) {
1463 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1464 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1465 } else if (eth_addr_is_zero(ea)) {
1466 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1472 /* Mirror output ports don't participate in picking the local hardware
1473 * address. ofproto can't help us find out whether a given port is a
1474 * mirror output because we haven't configured mirrors yet, so we need to
1475 * accumulate them ourselves. */
1476 hmapx_init(&mirror_output_ports);
1477 for (i = 0; i < br->cfg->n_mirrors; i++) {
1478 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1479 if (m->output_port) {
1480 hmapx_add(&mirror_output_ports, m->output_port);
1484 /* Otherwise choose the minimum non-local MAC address among all of the
1486 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1487 uint8_t iface_ea[ETH_ADDR_LEN];
1488 struct iface *candidate;
1489 struct iface *iface;
1491 /* Mirror output ports don't participate. */
1492 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1496 /* Choose the MAC address to represent the port. */
1498 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1499 /* Find the interface with this Ethernet address (if any) so that
1500 * we can provide the correct devname to the caller. */
1501 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1502 uint8_t candidate_ea[ETH_ADDR_LEN];
1503 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1504 && eth_addr_equals(iface_ea, candidate_ea)) {
1509 /* Choose the interface whose MAC address will represent the port.
1510 * The Linux kernel bonding code always chooses the MAC address of
1511 * the first slave added to a bond, and the Fedora networking
1512 * scripts always add slaves to a bond in alphabetical order, so
1513 * for compatibility we choose the interface with the name that is
1514 * first in alphabetical order. */
1515 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1516 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1521 /* The local port doesn't count (since we're trying to choose its
1522 * MAC address anyway). */
1523 if (iface->ofp_port == OFPP_LOCAL) {
1528 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1534 /* Compare against our current choice. */
1535 if (!eth_addr_is_multicast(iface_ea) &&
1536 !eth_addr_is_local(iface_ea) &&
1537 !eth_addr_is_reserved(iface_ea) &&
1538 !eth_addr_is_zero(iface_ea) &&
1539 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1541 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1542 *hw_addr_iface = iface;
1547 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1548 br->name, ETH_ADDR_ARGS(ea));
1550 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1551 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1552 *hw_addr_iface = NULL;
1553 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1554 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1557 hmapx_destroy(&mirror_output_ports);
1560 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1561 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1562 * an interface on 'br', then that interface must be passed in as
1563 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1564 * 'hw_addr_iface' must be passed in as a null pointer. */
1566 bridge_pick_datapath_id(struct bridge *br,
1567 const uint8_t bridge_ea[ETH_ADDR_LEN],
1568 struct iface *hw_addr_iface)
1571 * The procedure for choosing a bridge MAC address will, in the most
1572 * ordinary case, also choose a unique MAC that we can use as a datapath
1573 * ID. In some special cases, though, multiple bridges will end up with
1574 * the same MAC address. This is OK for the bridges, but it will confuse
1575 * the OpenFlow controller, because each datapath needs a unique datapath
1578 * Datapath IDs must be unique. It is also very desirable that they be
1579 * stable from one run to the next, so that policy set on a datapath
1582 const char *datapath_id;
1585 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1586 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1590 if (!hw_addr_iface) {
1592 * A purely internal bridge, that is, one that has no non-virtual
1593 * network devices on it at all, is difficult because it has no
1594 * natural unique identifier at all.
1596 * When the host is a XenServer, we handle this case by hashing the
1597 * host's UUID with the name of the bridge. Names of bridges are
1598 * persistent across XenServer reboots, although they can be reused if
1599 * an internal network is destroyed and then a new one is later
1600 * created, so this is fairly effective.
1602 * When the host is not a XenServer, we punt by using a random MAC
1603 * address on each run.
1605 const char *host_uuid = xenserver_get_host_uuid();
1607 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1608 dpid = dpid_from_hash(combined, strlen(combined));
1614 return eth_addr_to_uint64(bridge_ea);
1618 dpid_from_hash(const void *data, size_t n)
1620 uint8_t hash[SHA1_DIGEST_SIZE];
1622 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1623 sha1_bytes(data, n, hash);
1624 eth_addr_mark_random(hash);
1625 return eth_addr_to_uint64(hash);
1629 iface_refresh_status(struct iface *iface)
1633 enum netdev_features current;
1634 enum netdev_flags flags;
1640 if (iface_is_synthetic(iface)) {
1646 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1647 ovsrec_interface_set_status(iface->cfg, &smap);
1649 ovsrec_interface_set_status(iface->cfg, NULL);
1652 smap_destroy(&smap);
1654 error = netdev_get_flags(iface->netdev, &flags);
1656 ovsrec_interface_set_admin_state(iface->cfg,
1657 flags & NETDEV_UP ? "up" : "down");
1660 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1663 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1665 ovsrec_interface_set_duplex(iface->cfg,
1666 netdev_features_is_full_duplex(current)
1668 /* warning: uint64_t -> int64_t conversion */
1669 bps = netdev_features_to_bps(current);
1670 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1673 ovsrec_interface_set_duplex(iface->cfg, NULL);
1674 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1677 error = netdev_get_mtu(iface->netdev, &mtu);
1680 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1683 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1687 /* Writes 'iface''s CFM statistics to the database. */
1689 iface_refresh_cfm_stats(struct iface *iface)
1691 const struct ovsrec_interface *cfg = iface->cfg;
1693 const uint64_t *rmps;
1697 if (iface_is_synthetic(iface)) {
1701 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1704 const char *reasons[CFM_FAULT_N_REASONS];
1705 bool fault_bool = fault;
1709 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1710 int reason = 1 << i;
1711 if (fault & reason) {
1712 reasons[j++] = cfm_fault_reason_to_str(reason);
1716 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1717 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1719 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1720 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1723 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1724 iface->ofp_port, &rmps, &n_rmps);
1726 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1729 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1732 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1735 int64_t cfm_health = health;
1736 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1738 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1743 iface_refresh_stats(struct iface *iface)
1745 #define IFACE_STATS \
1746 IFACE_STAT(rx_packets, "rx_packets") \
1747 IFACE_STAT(tx_packets, "tx_packets") \
1748 IFACE_STAT(rx_bytes, "rx_bytes") \
1749 IFACE_STAT(tx_bytes, "tx_bytes") \
1750 IFACE_STAT(rx_dropped, "rx_dropped") \
1751 IFACE_STAT(tx_dropped, "tx_dropped") \
1752 IFACE_STAT(rx_errors, "rx_errors") \
1753 IFACE_STAT(tx_errors, "tx_errors") \
1754 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1755 IFACE_STAT(rx_over_errors, "rx_over_err") \
1756 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1757 IFACE_STAT(collisions, "collisions")
1759 #define IFACE_STAT(MEMBER, NAME) NAME,
1760 static char *keys[] = { IFACE_STATS };
1762 int64_t values[ARRAY_SIZE(keys)];
1765 struct netdev_stats stats;
1767 if (iface_is_synthetic(iface)) {
1771 /* Intentionally ignore return value, since errors will set 'stats' to
1772 * all-1s, and we will deal with that correctly below. */
1773 netdev_get_stats(iface->netdev, &stats);
1775 /* Copy statistics into values[] array. */
1777 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1780 assert(i == ARRAY_SIZE(keys));
1782 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1788 br_refresh_stp_status(struct bridge *br)
1790 struct smap smap = SMAP_INITIALIZER(&smap);
1791 struct ofproto *ofproto = br->ofproto;
1792 struct ofproto_stp_status status;
1794 if (ofproto_get_stp_status(ofproto, &status)) {
1798 if (!status.enabled) {
1799 ovsrec_bridge_set_status(br->cfg, NULL);
1803 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1804 STP_ID_ARGS(status.bridge_id));
1805 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1806 STP_ID_ARGS(status.designated_root));
1807 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1809 ovsrec_bridge_set_status(br->cfg, &smap);
1810 smap_destroy(&smap);
1814 port_refresh_stp_status(struct port *port)
1816 struct ofproto *ofproto = port->bridge->ofproto;
1817 struct iface *iface;
1818 struct ofproto_port_stp_status status;
1820 int64_t int_values[3];
1823 if (port_is_synthetic(port)) {
1827 /* STP doesn't currently support bonds. */
1828 if (!list_is_singleton(&port->ifaces)) {
1829 ovsrec_port_set_status(port->cfg, NULL);
1833 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1835 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1839 if (!status.enabled) {
1840 ovsrec_port_set_status(port->cfg, NULL);
1841 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1845 /* Set Status column. */
1847 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1848 smap_add(&smap, "stp_state", stp_state_name(status.state));
1849 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1850 smap_add(&smap, "stp_role", stp_role_name(status.role));
1851 ovsrec_port_set_status(port->cfg, &smap);
1852 smap_destroy(&smap);
1854 /* Set Statistics column. */
1855 keys[0] = "stp_tx_count";
1856 int_values[0] = status.tx_count;
1857 keys[1] = "stp_rx_count";
1858 int_values[1] = status.rx_count;
1859 keys[2] = "stp_error_count";
1860 int_values[2] = status.error_count;
1862 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1863 ARRAY_SIZE(int_values));
1867 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1869 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1873 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1875 struct ovsdb_datum datum;
1879 if (enable_system_stats(cfg)) {
1880 get_system_stats(&stats);
1883 ovsdb_datum_from_shash(&datum, &stats);
1884 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1888 static inline const char *
1889 nx_role_to_str(enum nx_role role)
1894 case NX_ROLE_MASTER:
1899 return "*** INVALID ROLE ***";
1904 refresh_controller_status(void)
1908 const struct ovsrec_controller *cfg;
1912 /* Accumulate status for controllers on all bridges. */
1913 HMAP_FOR_EACH (br, node, &all_bridges) {
1914 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1917 /* Update each controller in the database with current status. */
1918 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1919 struct ofproto_controller_info *cinfo =
1920 shash_find_data(&info, cfg->target);
1923 struct smap smap = SMAP_INITIALIZER(&smap);
1924 const char **values = cinfo->pairs.values;
1925 const char **keys = cinfo->pairs.keys;
1928 for (i = 0; i < cinfo->pairs.n; i++) {
1929 smap_add(&smap, keys[i], values[i]);
1932 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1933 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1934 ovsrec_controller_set_status(cfg, &smap);
1935 smap_destroy(&smap);
1937 ovsrec_controller_set_is_connected(cfg, false);
1938 ovsrec_controller_set_role(cfg, NULL);
1939 ovsrec_controller_set_status(cfg, NULL);
1943 ofproto_free_ofproto_controller_info(&info);
1947 refresh_cfm_stats(void)
1949 static struct ovsdb_idl_txn *txn = NULL;
1954 txn = ovsdb_idl_txn_create(idl);
1956 HMAP_FOR_EACH (br, node, &all_bridges) {
1957 struct iface *iface;
1959 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1960 iface_refresh_cfm_stats(iface);
1965 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1966 ovsdb_idl_txn_destroy(txn);
1971 /* Performs periodic activity required by bridges that needs to be done with
1972 * the least possible latency.
1974 * It makes sense to call this function a couple of times per poll loop, to
1975 * provide a significant performance boost on some benchmarks with ofprotos
1976 * that use the ofproto-dpif implementation. */
1978 bridge_run_fast(void)
1982 HMAP_FOR_EACH (br, node, &all_bridges) {
1983 ofproto_run_fast(br->ofproto);
1990 static const struct ovsrec_open_vswitch null_cfg;
1991 const struct ovsrec_open_vswitch *cfg;
1992 struct ovsdb_idl_txn *reconf_txn = NULL;
1994 bool vlan_splinters_changed;
1997 /* (Re)configure if necessary. */
1998 if (!reconfiguring) {
2001 if (ovsdb_idl_is_lock_contended(idl)) {
2002 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2003 struct bridge *br, *next_br;
2005 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2006 "disabling this process until it goes away");
2008 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2012 } else if (!ovsdb_idl_has_lock(idl)) {
2016 cfg = ovsrec_open_vswitch_first(idl);
2018 /* Let each bridge do the work that it needs to do. */
2019 HMAP_FOR_EACH (br, node, &all_bridges) {
2020 ofproto_run(br->ofproto);
2023 /* Re-configure SSL. We do this on every trip through the main loop,
2024 * instead of just when the database changes, because the contents of the
2025 * key and certificate files can change without the database changing.
2027 * We do this before bridge_reconfigure() because that function might
2028 * initiate SSL connections and thus requires SSL to be configured. */
2029 if (cfg && cfg->ssl) {
2030 const struct ovsrec_ssl *ssl = cfg->ssl;
2032 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2033 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2036 if (!reconfiguring) {
2037 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2038 * usage has changed. */
2039 vlan_splinters_changed = false;
2040 if (vlan_splinters_enabled_anywhere) {
2041 HMAP_FOR_EACH (br, node, &all_bridges) {
2042 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2043 vlan_splinters_changed = true;
2049 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2050 idl_seqno = ovsdb_idl_get_seqno(idl);
2052 reconf_txn = ovsdb_idl_txn_create(idl);
2053 bridge_reconfigure(cfg);
2055 /* We still need to reconfigure to avoid dangling pointers to
2056 * now-destroyed ovsrec structures inside bridge data. */
2057 bridge_reconfigure(&null_cfg);
2062 if (reconfiguring) {
2065 reconf_txn = ovsdb_idl_txn_create(idl);
2067 if (bridge_reconfigure_continue(cfg)) {
2068 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2071 bridge_reconfigure_continue(&null_cfg);
2076 ovsdb_idl_txn_commit(reconf_txn);
2077 ovsdb_idl_txn_destroy(reconf_txn);
2081 /* Refresh system and interface stats if necessary. */
2082 if (time_msec() >= stats_timer) {
2084 struct ovsdb_idl_txn *txn;
2086 txn = ovsdb_idl_txn_create(idl);
2087 HMAP_FOR_EACH (br, node, &all_bridges) {
2091 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2092 struct iface *iface;
2094 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2095 iface_refresh_stats(iface);
2096 iface_refresh_status(iface);
2100 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2101 mirror_refresh_stats(m);
2105 refresh_system_stats(cfg);
2106 refresh_controller_status();
2107 ovsdb_idl_txn_commit(txn);
2108 ovsdb_idl_txn_destroy(txn); /* XXX */
2111 stats_timer = time_msec() + STATS_INTERVAL;
2114 if (time_msec() >= db_limiter) {
2115 struct ovsdb_idl_txn *txn;
2117 txn = ovsdb_idl_txn_create(idl);
2118 HMAP_FOR_EACH (br, node, &all_bridges) {
2119 struct iface *iface;
2122 br_refresh_stp_status(br);
2124 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2125 port_refresh_stp_status(port);
2128 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2129 const char *link_state;
2130 int64_t link_resets;
2133 if (iface_is_synthetic(iface)) {
2137 current = ofproto_port_is_lacp_current(br->ofproto,
2141 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2143 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2146 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2147 ovsrec_interface_set_link_state(iface->cfg, link_state);
2149 link_resets = netdev_get_carrier_resets(iface->netdev);
2150 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2154 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2155 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2157 ovsdb_idl_txn_destroy(txn);
2160 refresh_cfm_stats();
2166 ovsdb_idl_wait(idl);
2168 if (reconfiguring) {
2169 poll_immediate_wake();
2172 if (!hmap_is_empty(&all_bridges)) {
2175 HMAP_FOR_EACH (br, node, &all_bridges) {
2176 ofproto_wait(br->ofproto);
2178 poll_timer_wait_until(stats_timer);
2180 if (db_limiter > time_msec()) {
2181 poll_timer_wait_until(db_limiter);
2186 /* Adds some memory usage statistics for bridges into 'usage', for use with
2187 * memory_report(). */
2189 bridge_get_memory_usage(struct simap *usage)
2193 HMAP_FOR_EACH (br, node, &all_bridges) {
2194 ofproto_get_memory_usage(br->ofproto, usage);
2198 /* QoS unixctl user interface functions. */
2200 struct qos_unixctl_show_cbdata {
2202 struct iface *iface;
2206 qos_unixctl_show_cb(unsigned int queue_id,
2207 const struct smap *details,
2210 struct qos_unixctl_show_cbdata *data = aux;
2211 struct ds *ds = data->ds;
2212 struct iface *iface = data->iface;
2213 struct netdev_queue_stats stats;
2214 struct smap_node *node;
2217 ds_put_cstr(ds, "\n");
2219 ds_put_format(ds, "Queue %u:\n", queue_id);
2221 ds_put_cstr(ds, "Default:\n");
2224 SMAP_FOR_EACH (node, details) {
2225 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2228 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2230 if (stats.tx_packets != UINT64_MAX) {
2231 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2234 if (stats.tx_bytes != UINT64_MAX) {
2235 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2238 if (stats.tx_errors != UINT64_MAX) {
2239 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2242 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2243 queue_id, strerror(error));
2248 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2249 const char *argv[], void *aux OVS_UNUSED)
2251 struct ds ds = DS_EMPTY_INITIALIZER;
2252 struct smap smap = SMAP_INITIALIZER(&smap);
2253 struct iface *iface;
2255 struct smap_node *node;
2256 struct qos_unixctl_show_cbdata data;
2259 iface = iface_find(argv[1]);
2261 unixctl_command_reply_error(conn, "no such interface");
2265 netdev_get_qos(iface->netdev, &type, &smap);
2267 if (*type != '\0') {
2268 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2270 SMAP_FOR_EACH (node, &smap) {
2271 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2276 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2279 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2281 unixctl_command_reply(conn, ds_cstr(&ds));
2283 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2284 unixctl_command_reply_error(conn, ds_cstr(&ds));
2287 smap_destroy(&smap);
2291 /* Bridge reconfiguration functions. */
2293 bridge_create(const struct ovsrec_bridge *br_cfg)
2297 assert(!bridge_lookup(br_cfg->name));
2298 br = xzalloc(sizeof *br);
2300 br->name = xstrdup(br_cfg->name);
2301 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2304 /* Derive the default Ethernet address from the bridge's UUID. This should
2305 * be unique and it will be stable between ovs-vswitchd runs. */
2306 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2307 eth_addr_mark_random(br->default_ea);
2309 hmap_init(&br->ports);
2310 hmap_init(&br->ifaces);
2311 hmap_init(&br->iface_by_name);
2312 hmap_init(&br->mirrors);
2314 hmap_init(&br->if_cfg_todo);
2315 list_init(&br->ofpp_garbage);
2317 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2321 bridge_destroy(struct bridge *br)
2324 struct mirror *mirror, *next_mirror;
2325 struct port *port, *next_port;
2326 struct if_cfg *if_cfg, *next_if_cfg;
2327 struct ofpp_garbage *garbage, *next_garbage;
2329 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2332 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2333 mirror_destroy(mirror);
2335 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2336 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2339 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2340 &br->ofpp_garbage) {
2341 list_remove(&garbage->list_node);
2345 hmap_remove(&all_bridges, &br->node);
2346 ofproto_destroy(br->ofproto);
2347 hmap_destroy(&br->ifaces);
2348 hmap_destroy(&br->ports);
2349 hmap_destroy(&br->iface_by_name);
2350 hmap_destroy(&br->mirrors);
2351 hmap_destroy(&br->if_cfg_todo);
2358 static struct bridge *
2359 bridge_lookup(const char *name)
2363 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2364 if (!strcmp(br->name, name)) {
2371 /* Handle requests for a listing of all flows known by the OpenFlow
2372 * stack, including those normally hidden. */
2374 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2375 const char *argv[], void *aux OVS_UNUSED)
2380 br = bridge_lookup(argv[1]);
2382 unixctl_command_reply_error(conn, "Unknown bridge");
2387 ofproto_get_all_flows(br->ofproto, &results);
2389 unixctl_command_reply(conn, ds_cstr(&results));
2390 ds_destroy(&results);
2393 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2394 * connections and reconnect. If BRIDGE is not specified, then all bridges
2395 * drop their controller connections and reconnect. */
2397 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2398 const char *argv[], void *aux OVS_UNUSED)
2402 br = bridge_lookup(argv[1]);
2404 unixctl_command_reply_error(conn, "Unknown bridge");
2407 ofproto_reconnect_controllers(br->ofproto);
2409 HMAP_FOR_EACH (br, node, &all_bridges) {
2410 ofproto_reconnect_controllers(br->ofproto);
2413 unixctl_command_reply(conn, NULL);
2417 bridge_get_controllers(const struct bridge *br,
2418 struct ovsrec_controller ***controllersp)
2420 struct ovsrec_controller **controllers;
2421 size_t n_controllers;
2423 controllers = br->cfg->controller;
2424 n_controllers = br->cfg->n_controller;
2426 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2432 *controllersp = controllers;
2434 return n_controllers;
2438 bridge_queue_if_cfg(struct bridge *br,
2439 const struct ovsrec_interface *cfg,
2440 const struct ovsrec_port *parent)
2442 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2445 if_cfg->parent = parent;
2446 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2447 hash_string(if_cfg->cfg->name, 0));
2450 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2451 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2452 * 'br' needs to complete its configuration. */
2454 bridge_add_del_ports(struct bridge *br,
2455 const unsigned long int *splinter_vlans)
2457 struct shash_node *port_node;
2458 struct port *port, *next;
2459 struct shash new_ports;
2462 assert(hmap_is_empty(&br->if_cfg_todo));
2464 /* Collect new ports. */
2465 shash_init(&new_ports);
2466 for (i = 0; i < br->cfg->n_ports; i++) {
2467 const char *name = br->cfg->ports[i]->name;
2468 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2469 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2473 if (bridge_get_controllers(br, NULL)
2474 && !shash_find(&new_ports, br->name)) {
2475 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2476 br->name, br->name);
2478 ovsrec_interface_init(&br->synth_local_iface);
2479 ovsrec_port_init(&br->synth_local_port);
2481 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2482 br->synth_local_port.n_interfaces = 1;
2483 br->synth_local_port.name = br->name;
2485 br->synth_local_iface.name = br->name;
2486 br->synth_local_iface.type = "internal";
2488 br->synth_local_ifacep = &br->synth_local_iface;
2490 shash_add(&new_ports, br->name, &br->synth_local_port);
2493 if (splinter_vlans) {
2494 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2497 /* Get rid of deleted ports.
2498 * Get rid of deleted interfaces on ports that still exist. */
2499 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2500 port->cfg = shash_find_data(&new_ports, port->name);
2504 port_del_ifaces(port);
2508 /* Update iface->cfg and iface->type in interfaces that still exist.
2509 * Add new interfaces to creation queue. */
2510 SHASH_FOR_EACH (port_node, &new_ports) {
2511 const struct ovsrec_port *port = port_node->data;
2514 for (i = 0; i < port->n_interfaces; i++) {
2515 const struct ovsrec_interface *cfg = port->interfaces[i];
2516 struct iface *iface = iface_lookup(br, cfg->name);
2517 const char *type = iface_get_type(cfg, br->cfg);
2522 } else if (strcmp(type, "null")) {
2523 bridge_queue_if_cfg(br, cfg, port);
2528 shash_destroy(&new_ports);
2531 /* Initializes 'oc' appropriately as a management service controller for
2534 * The caller must free oc->target when it is no longer needed. */
2536 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2537 struct ofproto_controller *oc)
2539 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2540 oc->max_backoff = 0;
2541 oc->probe_interval = 60;
2542 oc->band = OFPROTO_OUT_OF_BAND;
2544 oc->burst_limit = 0;
2545 oc->enable_async_msgs = true;
2548 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2550 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2551 struct ofproto_controller *oc)
2555 oc->target = c->target;
2556 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2557 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2558 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2559 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2560 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2561 oc->burst_limit = (c->controller_burst_limit
2562 ? *c->controller_burst_limit : 0);
2563 oc->enable_async_msgs = (!c->enable_async_messages
2564 || *c->enable_async_messages);
2565 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2566 if (dscp < 0 || dscp > 63) {
2567 dscp = DSCP_DEFAULT;
2572 /* Configures the IP stack for 'br''s local interface properly according to the
2573 * configuration in 'c'. */
2575 bridge_configure_local_iface_netdev(struct bridge *br,
2576 struct ovsrec_controller *c)
2578 struct netdev *netdev;
2579 struct in_addr mask, gateway;
2581 struct iface *local_iface;
2584 /* If there's no local interface or no IP address, give up. */
2585 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2586 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2590 /* Bring up the local interface. */
2591 netdev = local_iface->netdev;
2592 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2594 /* Configure the IP address and netmask. */
2595 if (!c->local_netmask
2596 || !inet_aton(c->local_netmask, &mask)
2598 mask.s_addr = guess_netmask(ip.s_addr);
2600 if (!netdev_set_in4(netdev, ip, mask)) {
2601 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2602 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2605 /* Configure the default gateway. */
2606 if (c->local_gateway
2607 && inet_aton(c->local_gateway, &gateway)
2608 && gateway.s_addr) {
2609 if (!netdev_add_router(netdev, gateway)) {
2610 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2611 br->name, IP_ARGS(&gateway.s_addr));
2616 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2617 * in either string are treated as equal to any number of slashes in the other,
2618 * e.g. "x///y" is equal to "x/y". */
2620 equal_pathnames(const char *a, const char *b)
2624 a += strspn(a, "/");
2625 b += strspn(b, "/");
2626 } else if (*a == '\0') {
2637 bridge_configure_remotes(struct bridge *br,
2638 const struct sockaddr_in *managers, size_t n_managers)
2640 bool disable_in_band;
2642 struct ovsrec_controller **controllers;
2643 size_t n_controllers;
2645 enum ofproto_fail_mode fail_mode;
2647 struct ofproto_controller *ocs;
2651 /* Check if we should disable in-band control on this bridge. */
2652 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2655 /* Set OpenFlow queue ID for in-band control. */
2656 ofproto_set_in_band_queue(br->ofproto,
2657 smap_get_int(&br->cfg->other_config,
2658 "in-band-queue", -1));
2660 if (disable_in_band) {
2661 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2663 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2666 n_controllers = bridge_get_controllers(br, &controllers);
2668 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2671 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2672 for (i = 0; i < n_controllers; i++) {
2673 struct ovsrec_controller *c = controllers[i];
2675 if (!strncmp(c->target, "punix:", 6)
2676 || !strncmp(c->target, "unix:", 5)) {
2677 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2680 whitelist = xasprintf("unix:%s/%s.controller",
2681 ovs_rundir(), br->name);
2682 if (!equal_pathnames(c->target, whitelist)) {
2683 /* Prevent remote ovsdb-server users from accessing arbitrary
2684 * Unix domain sockets and overwriting arbitrary local
2686 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2687 "controller \"%s\" due to possibility for remote "
2688 "exploit. Instead, specify whitelisted \"%s\" or "
2689 "connect to \"unix:%s/%s.mgmt\" (which is always "
2690 "available without special configuration).",
2691 br->name, c->target, whitelist,
2692 ovs_rundir(), br->name);
2700 bridge_configure_local_iface_netdev(br, c);
2701 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2702 if (disable_in_band) {
2703 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2708 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2709 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2712 /* Set the fail-mode. */
2713 fail_mode = !br->cfg->fail_mode
2714 || !strcmp(br->cfg->fail_mode, "standalone")
2715 ? OFPROTO_FAIL_STANDALONE
2716 : OFPROTO_FAIL_SECURE;
2717 ofproto_set_fail_mode(br->ofproto, fail_mode);
2719 /* Configure OpenFlow controller connection snooping. */
2720 if (!ofproto_has_snoops(br->ofproto)) {
2724 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2725 ovs_rundir(), br->name));
2726 ofproto_set_snoops(br->ofproto, &snoops);
2727 sset_destroy(&snoops);
2732 bridge_configure_tables(struct bridge *br)
2734 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2738 n_tables = ofproto_get_n_tables(br->ofproto);
2740 for (i = 0; i < n_tables; i++) {
2741 struct ofproto_table_settings s;
2744 s.max_flows = UINT_MAX;
2748 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2749 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2752 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2753 s.max_flows = *cfg->flow_limit;
2755 if (cfg->overflow_policy
2756 && !strcmp(cfg->overflow_policy, "evict")) {
2759 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2760 for (k = 0; k < cfg->n_groups; k++) {
2761 const char *string = cfg->groups[k];
2764 msg = mf_parse_subfield__(&s.groups[k], &string);
2766 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2767 "'groups' (%s)", br->name, i, msg);
2769 } else if (*string) {
2770 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2771 "element '%s' contains trailing garbage",
2772 br->name, i, cfg->groups[k]);
2780 ofproto_configure_table(br->ofproto, i, &s);
2784 for (; j < br->cfg->n_flow_tables; j++) {
2785 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2786 "%"PRId64" not supported by this datapath", br->name,
2787 br->cfg->key_flow_tables[j]);
2791 /* Port functions. */
2793 static struct port *
2794 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2798 port = xzalloc(sizeof *port);
2800 port->name = xstrdup(cfg->name);
2802 list_init(&port->ifaces);
2804 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2808 /* Deletes interfaces from 'port' that are no longer configured for it. */
2810 port_del_ifaces(struct port *port)
2812 struct iface *iface, *next;
2813 struct sset new_ifaces;
2816 /* Collect list of new interfaces. */
2817 sset_init(&new_ifaces);
2818 for (i = 0; i < port->cfg->n_interfaces; i++) {
2819 const char *name = port->cfg->interfaces[i]->name;
2820 const char *type = port->cfg->interfaces[i]->type;
2821 if (strcmp(type, "null")) {
2822 sset_add(&new_ifaces, name);
2826 /* Get rid of deleted interfaces. */
2827 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2828 if (!sset_contains(&new_ifaces, iface->name)) {
2829 iface_destroy(iface);
2833 sset_destroy(&new_ifaces);
2837 port_destroy(struct port *port)
2840 struct bridge *br = port->bridge;
2841 struct iface *iface, *next;
2844 ofproto_bundle_unregister(br->ofproto, port);
2847 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2848 iface_destroy(iface);
2851 hmap_remove(&br->ports, &port->hmap_node);
2857 static struct port *
2858 port_lookup(const struct bridge *br, const char *name)
2862 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2864 if (!strcmp(port->name, name)) {
2872 enable_lacp(struct port *port, bool *activep)
2874 if (!port->cfg->lacp) {
2875 /* XXX when LACP implementation has been sufficiently tested, enable by
2876 * default and make active on bonded ports. */
2878 } else if (!strcmp(port->cfg->lacp, "off")) {
2880 } else if (!strcmp(port->cfg->lacp, "active")) {
2883 } else if (!strcmp(port->cfg->lacp, "passive")) {
2887 VLOG_WARN("port %s: unknown LACP mode %s",
2888 port->name, port->cfg->lacp);
2893 static struct lacp_settings *
2894 port_configure_lacp(struct port *port, struct lacp_settings *s)
2896 const char *lacp_time, *system_id;
2899 if (!enable_lacp(port, &s->active)) {
2903 s->name = port->name;
2905 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2907 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2908 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2909 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2910 " address.", port->name, system_id);
2914 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2917 if (eth_addr_is_zero(s->id)) {
2918 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2922 /* Prefer bondable links if unspecified. */
2923 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2925 s->priority = (priority > 0 && priority <= UINT16_MAX
2927 : UINT16_MAX - !list_is_short(&port->ifaces));
2929 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2930 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2935 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2937 int priority, portid, key;
2939 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2940 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2942 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2944 if (portid <= 0 || portid > UINT16_MAX) {
2945 portid = iface->ofp_port;
2948 if (priority <= 0 || priority > UINT16_MAX) {
2949 priority = UINT16_MAX;
2952 if (key < 0 || key > UINT16_MAX) {
2956 s->name = iface->name;
2958 s->priority = priority;
2963 port_configure_bond(struct port *port, struct bond_settings *s,
2964 uint32_t *bond_stable_ids)
2966 const char *detect_s;
2967 struct iface *iface;
2968 int miimon_interval;
2971 s->name = port->name;
2973 if (port->cfg->bond_mode) {
2974 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2975 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2976 port->name, port->cfg->bond_mode,
2977 bond_mode_to_string(s->balance));
2980 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2982 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2983 * active-backup. At some point we should remove this warning. */
2984 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2985 " in previous versions, the default bond_mode was"
2986 " balance-slb", port->name,
2987 bond_mode_to_string(s->balance));
2989 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2990 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2991 "please use another bond type or disable flood_vlans",
2995 miimon_interval = smap_get_int(&port->cfg->other_config,
2996 "bond-miimon-interval", 0);
2997 if (miimon_interval <= 0) {
2998 miimon_interval = 200;
3001 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3002 if (!detect_s || !strcmp(detect_s, "carrier")) {
3003 miimon_interval = 0;
3004 } else if (strcmp(detect_s, "miimon")) {
3005 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3006 "defaulting to carrier", port->name, detect_s);
3007 miimon_interval = 0;
3010 s->up_delay = MAX(0, port->cfg->bond_updelay);
3011 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3012 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3013 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3014 "bond-rebalance-interval", 10000);
3015 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3016 s->rebalance_interval = 1000;
3019 s->fake_iface = port->cfg->bond_fake_iface;
3022 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3023 long long stable_id;
3025 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3027 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3028 stable_id = iface->ofp_port;
3030 bond_stable_ids[i++] = stable_id;
3032 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3036 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3037 * instead of obtaining it from the database. */
3039 port_is_synthetic(const struct port *port)
3041 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3044 /* Interface functions. */
3046 /* Returns the correct network device type for interface 'iface' in bridge
3049 iface_get_type(const struct ovsrec_interface *iface,
3050 const struct ovsrec_bridge *br)
3052 /* The local port always has type "internal". Other ports take their type
3053 * from the database and default to "system" if none is specified. */
3054 return (!strcmp(iface->name, br->name) ? "internal"
3055 : iface->type[0] ? iface->type
3060 iface_destroy(struct iface *iface)
3063 struct port *port = iface->port;
3064 struct bridge *br = port->bridge;
3066 if (br->ofproto && iface->ofp_port >= 0) {
3067 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3070 if (iface->ofp_port >= 0) {
3071 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3074 list_remove(&iface->port_elem);
3075 hmap_remove(&br->iface_by_name, &iface->name_node);
3077 netdev_close(iface->netdev);
3084 static struct iface *
3085 iface_lookup(const struct bridge *br, const char *name)
3087 struct iface *iface;
3089 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3090 &br->iface_by_name) {
3091 if (!strcmp(iface->name, name)) {
3099 static struct iface *
3100 iface_find(const char *name)
3102 const struct bridge *br;
3104 HMAP_FOR_EACH (br, node, &all_bridges) {
3105 struct iface *iface = iface_lookup(br, name);
3114 static struct if_cfg *
3115 if_cfg_lookup(const struct bridge *br, const char *name)
3117 struct if_cfg *if_cfg;
3119 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3121 if (!strcmp(if_cfg->cfg->name, name)) {
3129 static struct iface *
3130 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3132 struct iface *iface;
3134 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3135 hash_int(ofp_port, 0), &br->ifaces) {
3136 if (iface->ofp_port == ofp_port) {
3143 /* Set Ethernet address of 'iface', if one is specified in the configuration
3146 iface_set_mac(struct iface *iface)
3148 uint8_t ea[ETH_ADDR_LEN];
3150 if (!strcmp(iface->type, "internal")
3151 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3152 if (iface->ofp_port == OFPP_LOCAL) {
3153 VLOG_ERR("interface %s: ignoring mac in Interface record "
3154 "(use Bridge record to set local port's mac)",
3156 } else if (eth_addr_is_multicast(ea)) {
3157 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3160 int error = netdev_set_etheraddr(iface->netdev, ea);
3162 VLOG_ERR("interface %s: setting MAC failed (%s)",
3163 iface->name, strerror(error));
3169 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3171 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3173 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3174 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3178 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3179 * sets the "ofport" field to -1.
3181 * This is appropriate when 'if_cfg''s interface cannot be created or is
3182 * otherwise invalid. */
3184 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3186 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3187 iface_set_ofport(if_cfg, -1);
3188 ovsrec_interface_set_status(if_cfg, NULL);
3189 ovsrec_interface_set_admin_state(if_cfg, NULL);
3190 ovsrec_interface_set_duplex(if_cfg, NULL);
3191 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3192 ovsrec_interface_set_link_state(if_cfg, NULL);
3193 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3194 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3195 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3196 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3197 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3198 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3202 struct iface_delete_queues_cbdata {
3203 struct netdev *netdev;
3204 const struct ovsdb_datum *queues;
3208 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3210 union ovsdb_atom atom;
3212 atom.integer = target;
3213 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3217 iface_delete_queues(unsigned int queue_id,
3218 const struct smap *details OVS_UNUSED, void *cbdata_)
3220 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3222 if (!queue_ids_include(cbdata->queues, queue_id)) {
3223 netdev_delete_queue(cbdata->netdev, queue_id);
3228 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3230 struct ofpbuf queues_buf;
3232 ofpbuf_init(&queues_buf, 0);
3234 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3235 netdev_set_qos(iface->netdev, NULL, NULL);
3237 struct iface_delete_queues_cbdata cbdata;
3241 /* Configure top-level Qos for 'iface'. */
3242 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3244 /* Deconfigure queues that were deleted. */
3245 cbdata.netdev = iface->netdev;
3246 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3248 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3250 /* Configure queues for 'iface'. */
3252 for (i = 0; i < qos->n_queues; i++) {
3253 const struct ovsrec_queue *queue = qos->value_queues[i];
3254 unsigned int queue_id = qos->key_queues[i];
3256 if (queue_id == 0) {
3260 if (queue->n_dscp == 1) {
3261 struct ofproto_port_queue *port_queue;
3263 port_queue = ofpbuf_put_uninit(&queues_buf,
3264 sizeof *port_queue);
3265 port_queue->queue = queue_id;
3266 port_queue->dscp = queue->dscp[0];
3269 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3272 struct smap details;
3274 smap_init(&details);
3275 netdev_set_queue(iface->netdev, 0, &details);
3276 smap_destroy(&details);
3280 if (iface->ofp_port >= 0) {
3281 const struct ofproto_port_queue *port_queues = queues_buf.data;
3282 size_t n_queues = queues_buf.size / sizeof *port_queues;
3284 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3285 port_queues, n_queues);
3288 netdev_set_policing(iface->netdev,
3289 iface->cfg->ingress_policing_rate,
3290 iface->cfg->ingress_policing_burst);
3292 ofpbuf_uninit(&queues_buf);
3296 iface_configure_cfm(struct iface *iface)
3298 const struct ovsrec_interface *cfg = iface->cfg;
3299 const char *opstate_str;
3300 const char *cfm_ccm_vlan;
3301 struct cfm_settings s;
3303 if (!cfg->n_cfm_mpid) {
3304 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3308 s.mpid = *cfg->cfm_mpid;
3309 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3310 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3311 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3313 if (s.interval <= 0) {
3317 if (!cfm_ccm_vlan) {
3319 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3320 s.ccm_vlan = CFM_RANDOM_VLAN;
3322 s.ccm_vlan = atoi(cfm_ccm_vlan);
3323 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3328 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3331 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3332 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3334 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3337 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3338 * instead of obtaining it from the database. */
3340 iface_is_synthetic(const struct iface *iface)
3342 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3346 /* Port mirroring. */
3348 static struct mirror *
3349 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3353 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3354 if (uuid_equals(uuid, &m->uuid)) {
3362 bridge_configure_mirrors(struct bridge *br)
3364 const struct ovsdb_datum *mc;
3365 unsigned long *flood_vlans;
3366 struct mirror *m, *next;
3369 /* Get rid of deleted mirrors. */
3370 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3371 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3372 union ovsdb_atom atom;
3374 atom.uuid = m->uuid;
3375 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3380 /* Add new mirrors and reconfigure existing ones. */
3381 for (i = 0; i < br->cfg->n_mirrors; i++) {
3382 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3383 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3385 m = mirror_create(br, cfg);
3388 if (!mirror_configure(m)) {
3393 /* Update flooded vlans (for RSPAN). */
3394 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3395 br->cfg->n_flood_vlans);
3396 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3397 bitmap_free(flood_vlans);
3400 static struct mirror *
3401 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3405 m = xzalloc(sizeof *m);
3406 m->uuid = cfg->header_.uuid;
3407 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3409 m->name = xstrdup(cfg->name);
3415 mirror_destroy(struct mirror *m)
3418 struct bridge *br = m->bridge;
3421 ofproto_mirror_unregister(br->ofproto, m);
3424 hmap_remove(&br->mirrors, &m->hmap_node);
3431 mirror_collect_ports(struct mirror *m,
3432 struct ovsrec_port **in_ports, int n_in_ports,
3433 void ***out_portsp, size_t *n_out_portsp)
3435 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3436 size_t n_out_ports = 0;
3439 for (i = 0; i < n_in_ports; i++) {
3440 const char *name = in_ports[i]->name;
3441 struct port *port = port_lookup(m->bridge, name);
3443 out_ports[n_out_ports++] = port;
3445 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3446 "port %s", m->bridge->name, m->name, name);
3449 *out_portsp = out_ports;
3450 *n_out_portsp = n_out_ports;
3454 mirror_configure(struct mirror *m)
3456 const struct ovsrec_mirror *cfg = m->cfg;
3457 struct ofproto_mirror_settings s;
3460 if (strcmp(cfg->name, m->name)) {
3462 m->name = xstrdup(cfg->name);
3466 /* Get output port or VLAN. */
3467 if (cfg->output_port) {
3468 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3469 if (!s.out_bundle) {
3470 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3471 m->bridge->name, m->name);
3474 s.out_vlan = UINT16_MAX;
3476 if (cfg->output_vlan) {
3477 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3478 "output vlan; ignoring output vlan",
3479 m->bridge->name, m->name);
3481 } else if (cfg->output_vlan) {
3482 /* The database should prevent invalid VLAN values. */
3483 s.out_bundle = NULL;
3484 s.out_vlan = *cfg->output_vlan;
3486 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3487 m->bridge->name, m->name);
3491 /* Get port selection. */
3492 if (cfg->select_all) {
3493 size_t n_ports = hmap_count(&m->bridge->ports);
3494 void **ports = xmalloc(n_ports * sizeof *ports);
3499 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3509 /* Get ports, dropping ports that don't exist.
3510 * The IDL ensures that there are no duplicates. */
3511 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3512 &s.srcs, &s.n_srcs);
3513 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3514 &s.dsts, &s.n_dsts);
3517 /* Get VLAN selection. */
3518 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3521 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3524 if (s.srcs != s.dsts) {
3533 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3535 * This is deprecated. It is only for compatibility with broken device drivers
3536 * in old versions of Linux that do not properly support VLANs when VLAN
3537 * devices are not used. When broken device drivers are no longer in
3538 * widespread use, we will delete these interfaces. */
3540 static struct ovsrec_port **recs;
3541 static size_t n_recs, allocated_recs;
3543 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3544 * splinters are reconfigured. */
3546 register_rec(struct ovsrec_port *rec)
3548 if (n_recs >= allocated_recs) {
3549 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3551 recs[n_recs++] = rec;
3554 /* Frees all of the ports registered with register_reg(). */
3556 free_registered_recs(void)
3560 for (i = 0; i < n_recs; i++) {
3561 struct ovsrec_port *port = recs[i];
3564 for (j = 0; j < port->n_interfaces; j++) {
3565 struct ovsrec_interface *iface = port->interfaces[j];
3570 smap_destroy(&port->other_config);
3571 free(port->interfaces);
3579 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3582 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3584 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3588 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3591 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3592 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3593 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3596 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3597 * use, returns NULL.
3599 * Updates 'vlan_splinters_enabled_anywhere'. */
3600 static unsigned long int *
3601 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3603 unsigned long int *splinter_vlans;
3604 struct sset splinter_ifaces;
3605 const char *real_dev_name;
3606 struct shash *real_devs;
3607 struct shash_node *node;
3611 /* Free space allocated for synthesized ports and interfaces, since we're
3612 * in the process of reconstructing all of them. */
3613 free_registered_recs();
3615 splinter_vlans = bitmap_allocate(4096);
3616 sset_init(&splinter_ifaces);
3617 vlan_splinters_enabled_anywhere = false;
3618 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3619 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3622 for (j = 0; j < br_cfg->n_ports; j++) {
3623 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3626 for (k = 0; k < port_cfg->n_interfaces; k++) {
3627 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3629 if (vlan_splinters_is_enabled(iface_cfg)) {
3630 vlan_splinters_enabled_anywhere = true;
3631 sset_add(&splinter_ifaces, iface_cfg->name);
3632 vlan_bitmap_from_array__(port_cfg->trunks,
3638 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3639 bitmap_set1(splinter_vlans, *port_cfg->tag);
3644 if (!vlan_splinters_enabled_anywhere) {
3645 free(splinter_vlans);
3646 sset_destroy(&splinter_ifaces);
3650 HMAP_FOR_EACH (br, node, &all_bridges) {
3652 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3656 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3657 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3658 * device to be created for it. */
3659 bitmap_set0(splinter_vlans, 0);
3660 bitmap_set0(splinter_vlans, 4095);
3662 /* Delete all VLAN devices that we don't need. */
3664 real_devs = vlandev_get_real_devs();
3665 SHASH_FOR_EACH (node, real_devs) {
3666 const struct vlan_real_dev *real_dev = node->data;
3667 const struct vlan_dev *vlan_dev;
3668 bool real_dev_has_splinters;
3670 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3672 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3673 if (!real_dev_has_splinters
3674 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3675 struct netdev *netdev;
3677 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3678 if (!netdev_get_in4(netdev, NULL, NULL) ||
3679 !netdev_get_in6(netdev, NULL)) {
3680 vlandev_del(vlan_dev->name);
3682 /* It has an IP address configured, so we don't own
3683 * it. Don't delete it. */
3685 netdev_close(netdev);
3692 /* Add all VLAN devices that we need. */
3693 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3696 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3697 if (!vlandev_get_name(real_dev_name, vid)) {
3698 vlandev_add(real_dev_name, vid);
3705 sset_destroy(&splinter_ifaces);
3707 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3708 free(splinter_vlans);
3711 return splinter_vlans;
3714 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3717 configure_splinter_port(struct port *port)
3719 struct ofproto *ofproto = port->bridge->ofproto;
3720 uint16_t realdev_ofp_port;
3721 const char *realdev_name;
3722 struct iface *vlandev, *realdev;
3724 ofproto_bundle_unregister(port->bridge->ofproto, port);
3726 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3729 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3730 realdev = iface_lookup(port->bridge, realdev_name);
3731 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3733 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3737 static struct ovsrec_port *
3738 synthesize_splinter_port(const char *real_dev_name,
3739 const char *vlan_dev_name, int vid)
3741 struct ovsrec_interface *iface;
3742 struct ovsrec_port *port;
3744 iface = xmalloc(sizeof *iface);
3745 ovsrec_interface_init(iface);
3746 iface->name = xstrdup(vlan_dev_name);
3747 iface->type = "system";
3749 port = xmalloc(sizeof *port);
3750 ovsrec_port_init(port);
3751 port->interfaces = xmemdup(&iface, sizeof iface);
3752 port->n_interfaces = 1;
3753 port->name = xstrdup(vlan_dev_name);
3754 port->vlan_mode = "splinter";
3755 port->tag = xmalloc(sizeof *port->tag);
3758 smap_add(&port->other_config, "realdev", real_dev_name);
3764 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3765 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3766 * 1-bit in the 'splinter_vlans' bitmap. */
3768 add_vlan_splinter_ports(struct bridge *br,
3769 const unsigned long int *splinter_vlans,
3770 struct shash *ports)
3774 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3775 * we're modifying 'ports'. */
3776 for (i = 0; i < br->cfg->n_ports; i++) {
3777 const char *name = br->cfg->ports[i]->name;
3778 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3781 for (j = 0; j < port_cfg->n_interfaces; j++) {
3782 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3784 if (vlan_splinters_is_enabled(iface_cfg)) {
3785 const char *real_dev_name;
3788 real_dev_name = iface_cfg->name;
3789 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3790 const char *vlan_dev_name;
3792 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3794 && !shash_find(ports, vlan_dev_name)) {
3795 shash_add(ports, vlan_dev_name,
3796 synthesize_splinter_port(
3797 real_dev_name, vlan_dev_name, vid));
3806 mirror_refresh_stats(struct mirror *m)
3808 struct ofproto *ofproto = m->bridge->ofproto;
3809 uint64_t tx_packets, tx_bytes;
3812 size_t stat_cnt = 0;
3814 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3815 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3819 if (tx_packets != UINT64_MAX) {
3820 keys[stat_cnt] = "tx_packets";
3821 values[stat_cnt] = tx_packets;
3824 if (tx_bytes != UINT64_MAX) {
3825 keys[stat_cnt] = "tx_bytes";
3826 values[stat_cnt] = tx_bytes;
3830 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);