1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
71 /* OpenFlow port slated for removal from ofproto. */
73 struct list list_node; /* Node in bridge's ofpp_garbage. */
74 uint16_t ofp_port; /* Port to be deleted. */
78 /* These members are always valid. */
79 struct list port_elem; /* Element in struct port's "ifaces" list. */
80 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
84 /* These members are valid only after bridge_reconfigure() causes them to
86 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
87 int ofp_port; /* OpenFlow port number, -1 if unknown. */
88 struct netdev *netdev; /* Network device. */
89 const char *type; /* Usually same as cfg->type. */
90 const struct ovsrec_interface *cfg;
94 struct uuid uuid; /* UUID of this "mirror" record in database. */
95 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
96 struct bridge *bridge;
98 const struct ovsrec_mirror *cfg;
102 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
103 struct bridge *bridge;
106 const struct ovsrec_port *cfg;
108 /* An ordinary bridge port has 1 interface.
109 * A bridge port for bonding has at least 2 interfaces. */
110 struct list ifaces; /* List of "struct iface"s. */
114 struct hmap_node node; /* In 'all_bridges'. */
115 char *name; /* User-specified arbitrary name. */
116 char *type; /* Datapath type. */
117 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
118 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
119 const struct ovsrec_bridge *cfg;
121 /* OpenFlow switch processing. */
122 struct ofproto *ofproto; /* OpenFlow switch. */
125 struct hmap ports; /* "struct port"s indexed by name. */
126 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
127 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
129 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
130 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
131 Indexed on 'cfg->name'. */
133 /* Port mirroring. */
134 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
136 /* Synthetic local port if necessary. */
137 struct ovsrec_port synth_local_port;
138 struct ovsrec_interface synth_local_iface;
139 struct ovsrec_interface *synth_local_ifacep;
142 /* All bridges, indexed by name. */
143 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
145 /* OVSDB IDL used to obtain configuration. */
146 static struct ovsdb_idl *idl;
148 /* Most recently processed IDL sequence number. */
149 static unsigned int idl_seqno;
151 /* Each time this timer expires, the bridge fetches interface and mirror
152 * statistics and pushes them into the database. */
153 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int iface_stats_timer = LLONG_MIN;
156 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
157 * expensive. This can cause bridge_reconfigure() to take a long time during
158 * which no other work can be done. To deal with this problem, we limit port
159 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
160 * call to bridge_reconfigure(). If there is more work to do after the limit
161 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
162 * This allows the rest of the code to catch up on important things like
163 * forwarding packets. */
164 #define OFP_PORT_ACTION_WINDOW 10
165 static bool reconfiguring = false;
167 static void add_del_bridges(const struct ovsrec_open_vswitch *);
168 static void bridge_update_ofprotos(void);
169 static void bridge_create(const struct ovsrec_bridge *);
170 static void bridge_destroy(struct bridge *);
171 static struct bridge *bridge_lookup(const char *name);
172 static unixctl_cb_func bridge_unixctl_dump_flows;
173 static unixctl_cb_func bridge_unixctl_reconnect;
174 static size_t bridge_get_controllers(const struct bridge *br,
175 struct ovsrec_controller ***controllersp);
176 static void bridge_add_del_ports(struct bridge *,
177 const unsigned long int *splinter_vlans);
178 static void bridge_refresh_ofp_port(struct bridge *);
179 static void bridge_configure_datapath_id(struct bridge *);
180 static void bridge_configure_flow_eviction_threshold(struct bridge *);
181 static void bridge_configure_netflow(struct bridge *);
182 static void bridge_configure_forward_bpdu(struct bridge *);
183 static void bridge_configure_mac_idle_time(struct bridge *);
184 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
185 static void bridge_configure_stp(struct bridge *);
186 static void bridge_configure_tables(struct bridge *);
187 static void bridge_configure_remotes(struct bridge *,
188 const struct sockaddr_in *managers,
190 static void bridge_pick_local_hw_addr(struct bridge *,
191 uint8_t ea[ETH_ADDR_LEN],
192 struct iface **hw_addr_iface);
193 static uint64_t bridge_pick_datapath_id(struct bridge *,
194 const uint8_t bridge_ea[ETH_ADDR_LEN],
195 struct iface *hw_addr_iface);
196 static void bridge_queue_if_cfg(struct bridge *,
197 const struct ovsrec_interface *,
198 const struct ovsrec_port *);
199 static uint64_t dpid_from_hash(const void *, size_t nbytes);
200 static bool bridge_has_bond_fake_iface(const struct bridge *,
202 static bool port_is_bond_fake_iface(const struct port *);
204 static unixctl_cb_func qos_unixctl_show;
206 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
207 static void port_del_ifaces(struct port *);
208 static void port_destroy(struct port *);
209 static struct port *port_lookup(const struct bridge *, const char *name);
210 static void port_configure(struct port *);
211 static struct lacp_settings *port_configure_lacp(struct port *,
212 struct lacp_settings *);
213 static void port_configure_bond(struct port *, struct bond_settings *,
214 uint32_t *bond_stable_ids);
215 static bool port_is_synthetic(const struct port *);
217 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
218 static void run_system_stats(void);
220 static void bridge_configure_mirrors(struct bridge *);
221 static struct mirror *mirror_create(struct bridge *,
222 const struct ovsrec_mirror *);
223 static void mirror_destroy(struct mirror *);
224 static bool mirror_configure(struct mirror *);
225 static void mirror_refresh_stats(struct mirror *);
227 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
228 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
229 static const char *iface_get_type(const struct ovsrec_interface *,
230 const struct ovsrec_bridge *);
231 static void iface_destroy(struct iface *);
232 static struct iface *iface_lookup(const struct bridge *, const char *name);
233 static struct iface *iface_find(const char *name);
234 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
235 static struct iface *iface_from_ofp_port(const struct bridge *,
237 static void iface_set_mac(struct iface *);
238 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
239 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
240 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
241 static void iface_configure_cfm(struct iface *);
242 static void iface_refresh_cfm_stats(struct iface *);
243 static void iface_refresh_stats(struct iface *);
244 static void iface_refresh_status(struct iface *);
245 static bool iface_is_synthetic(const struct iface *);
247 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
249 * This is deprecated. It is only for compatibility with broken device drivers
250 * in old versions of Linux that do not properly support VLANs when VLAN
251 * devices are not used. When broken device drivers are no longer in
252 * widespread use, we will delete these interfaces. */
254 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
255 static bool vlan_splinters_enabled_anywhere;
257 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
258 static unsigned long int *collect_splinter_vlans(
259 const struct ovsrec_open_vswitch *);
260 static void configure_splinter_port(struct port *);
261 static void add_vlan_splinter_ports(struct bridge *,
262 const unsigned long int *splinter_vlans,
263 struct shash *ports);
265 /* Public functions. */
267 /* Initializes the bridge module, configuring it to obtain its configuration
268 * from an OVSDB server accessed over 'remote', which should be a string in a
269 * form acceptable to ovsdb_idl_create(). */
271 bridge_init(const char *remote)
273 /* Create connection to database. */
274 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
275 idl_seqno = ovsdb_idl_get_seqno(idl);
276 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
278 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
279 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
280 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
281 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
282 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
286 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
287 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
288 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
290 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
291 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
292 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
293 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
295 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
296 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
297 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
309 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
311 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
312 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
313 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
314 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
316 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
318 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
320 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
321 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
323 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
325 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
327 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
329 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
333 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
335 /* Register unixctl commands. */
336 unixctl_command_register("qos/show", "interface", 1, 1,
337 qos_unixctl_show, NULL);
338 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
339 bridge_unixctl_dump_flows, NULL);
340 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
341 bridge_unixctl_reconnect, NULL);
351 struct bridge *br, *next_br;
353 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
356 ovsdb_idl_destroy(idl);
359 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
360 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
361 * responsible for freeing '*managersp' (with free()).
363 * You may be asking yourself "why does ovs-vswitchd care?", because
364 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
365 * should not be and in fact is not directly involved in that. But
366 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
367 * it has to tell in-band control where the managers are to enable that.
368 * (Thus, only managers connected in-band are collected.)
371 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
372 struct sockaddr_in **managersp, size_t *n_managersp)
374 struct sockaddr_in *managers = NULL;
375 size_t n_managers = 0;
379 /* Collect all of the potential targets from the "targets" columns of the
380 * rows pointed to by "manager_options", excluding any that are
383 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
384 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
386 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
387 sset_find_and_delete(&targets, m->target);
389 sset_add(&targets, m->target);
393 /* Now extract the targets' IP addresses. */
394 if (!sset_is_empty(&targets)) {
397 managers = xmalloc(sset_count(&targets) * sizeof *managers);
398 SSET_FOR_EACH (target, &targets) {
399 struct sockaddr_in *sin = &managers[n_managers];
401 if (stream_parse_target_with_default_ports(target,
409 sset_destroy(&targets);
411 *managersp = managers;
412 *n_managersp = n_managers;
416 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
418 unsigned long int *splinter_vlans;
421 COVERAGE_INC(bridge_reconfigure);
423 assert(!reconfiguring);
424 reconfiguring = true;
426 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
427 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
428 * configuration otherwise.
430 * This is mostly an update to bridge data structures. Nothing is pushed
431 * down to ofproto or lower layers. */
432 add_del_bridges(ovs_cfg);
433 splinter_vlans = collect_splinter_vlans(ovs_cfg);
434 HMAP_FOR_EACH (br, node, &all_bridges) {
435 bridge_add_del_ports(br, splinter_vlans);
437 free(splinter_vlans);
439 /* Delete datapaths that are no longer configured, and create ones which
440 * don't exist but should. */
441 bridge_update_ofprotos();
443 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
444 HMAP_FOR_EACH (br, node, &all_bridges) {
445 bridge_refresh_ofp_port(br);
448 /* Clear database records for "if_cfg"s which haven't been instantiated. */
449 HMAP_FOR_EACH (br, node, &all_bridges) {
450 struct if_cfg *if_cfg;
452 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
453 iface_clear_db_record(if_cfg->cfg);
457 reconfigure_system_stats(ovs_cfg);
461 bridge_reconfigure_ofp(void)
463 long long int deadline;
467 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
469 /* The kernel will reject any attempt to add a given port to a datapath if
470 * that port already belongs to a different datapath, so we must do all
471 * port deletions before any port additions. */
472 HMAP_FOR_EACH (br, node, &all_bridges) {
473 struct ofpp_garbage *garbage, *next;
475 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
476 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
477 * internal datastructures may not be consistent. Eventually, when
478 * port additions and deletions are cheaper, these calls should be
481 ofproto_port_del(br->ofproto, garbage->ofp_port);
482 list_remove(&garbage->list_node);
486 if (time_msec() >= deadline) {
493 HMAP_FOR_EACH (br, node, &all_bridges) {
494 struct if_cfg *if_cfg, *next;
496 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
497 iface_create(br, if_cfg, -1);
499 if (time_msec() >= deadline) {
509 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
511 struct sockaddr_in *managers;
512 int sflow_bridge_number;
517 assert(reconfiguring);
518 done = bridge_reconfigure_ofp();
520 /* Complete the configuration. */
521 sflow_bridge_number = 0;
522 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
523 HMAP_FOR_EACH (br, node, &all_bridges) {
526 /* We need the datapath ID early to allow LACP ports to use it as the
527 * default system ID. */
528 bridge_configure_datapath_id(br);
530 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
533 port_configure(port);
535 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
536 iface_configure_cfm(iface);
537 iface_configure_qos(iface, port->cfg->qos);
538 iface_set_mac(iface);
541 bridge_configure_mirrors(br);
542 bridge_configure_flow_eviction_threshold(br);
543 bridge_configure_forward_bpdu(br);
544 bridge_configure_mac_idle_time(br);
545 bridge_configure_remotes(br, managers, n_managers);
546 bridge_configure_netflow(br);
547 bridge_configure_sflow(br, &sflow_bridge_number);
548 bridge_configure_stp(br);
549 bridge_configure_tables(br);
554 /* ovs-vswitchd has completed initialization, so allow the process that
555 * forked us to exit successfully. */
556 daemonize_complete();
557 reconfiguring = false;
559 VLOG_INFO("%s (Open vSwitch) %s", program_name, VERSION);
565 /* Delete ofprotos which aren't configured or have the wrong type. Create
566 * ofprotos which don't exist but need to. */
568 bridge_update_ofprotos(void)
570 struct bridge *br, *next;
575 /* Delete ofprotos with no bridge or with the wrong type. */
578 ofproto_enumerate_types(&types);
579 SSET_FOR_EACH (type, &types) {
582 ofproto_enumerate_names(type, &names);
583 SSET_FOR_EACH (name, &names) {
584 br = bridge_lookup(name);
585 if (!br || strcmp(type, br->type)) {
586 ofproto_delete(name, type);
590 sset_destroy(&names);
591 sset_destroy(&types);
593 /* Add ofprotos for bridges which don't have one yet. */
594 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
602 /* Remove ports from any datapath with the same name as 'br'. If we
603 * don't do this, creating 'br''s ofproto will fail because a port with
604 * the same name as its local port already exists. */
605 HMAP_FOR_EACH (br2, node, &all_bridges) {
606 struct ofproto_port ofproto_port;
612 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
614 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
616 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
619 ofproto_port_destroy(&ofproto_port);
623 error = ofproto_create(br->name, br->type, &br->ofproto);
625 VLOG_ERR("failed to create bridge %s: %s", br->name,
633 port_configure(struct port *port)
635 const struct ovsrec_port *cfg = port->cfg;
636 struct bond_settings bond_settings;
637 struct lacp_settings lacp_settings;
638 struct ofproto_bundle_settings s;
641 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
642 configure_splinter_port(port);
651 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
652 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
653 s.slaves[s.n_slaves++] = iface->ofp_port;
658 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
662 /* Get VLAN trunks. */
665 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
669 if (cfg->vlan_mode) {
670 if (!strcmp(cfg->vlan_mode, "access")) {
671 s.vlan_mode = PORT_VLAN_ACCESS;
672 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
673 s.vlan_mode = PORT_VLAN_TRUNK;
674 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
675 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
676 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
677 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
679 /* This "can't happen" because ovsdb-server should prevent it. */
680 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
681 s.vlan_mode = PORT_VLAN_TRUNK;
685 s.vlan_mode = PORT_VLAN_ACCESS;
687 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
691 s.vlan_mode = PORT_VLAN_TRUNK;
694 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
697 /* Get LACP settings. */
698 s.lacp = port_configure_lacp(port, &lacp_settings);
702 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
703 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
704 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
707 s.lacp_slaves = NULL;
710 /* Get bond settings. */
711 if (s.n_slaves > 1) {
712 s.bond = &bond_settings;
713 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
714 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
717 s.bond_stable_ids = NULL;
719 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
720 netdev_set_miimon_interval(iface->netdev, 0);
725 ofproto_bundle_register(port->bridge->ofproto, port, &s);
731 free(s.bond_stable_ids);
734 /* Pick local port hardware address and datapath ID for 'br'. */
736 bridge_configure_datapath_id(struct bridge *br)
738 uint8_t ea[ETH_ADDR_LEN];
740 struct iface *local_iface;
741 struct iface *hw_addr_iface;
744 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
745 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
747 int error = netdev_set_etheraddr(local_iface->netdev, ea);
749 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
750 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
751 "Ethernet address: %s",
752 br->name, strerror(error));
755 memcpy(br->ea, ea, ETH_ADDR_LEN);
757 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
758 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
759 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
760 ofproto_set_datapath_id(br->ofproto, dpid);
763 dpid_string = xasprintf("%016"PRIx64, dpid);
764 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
768 /* Set NetFlow configuration on 'br'. */
770 bridge_configure_netflow(struct bridge *br)
772 struct ovsrec_netflow *cfg = br->cfg->netflow;
773 struct netflow_options opts;
776 ofproto_set_netflow(br->ofproto, NULL);
780 memset(&opts, 0, sizeof opts);
782 /* Get default NetFlow configuration from datapath.
783 * Apply overrides from 'cfg'. */
784 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
785 if (cfg->engine_type) {
786 opts.engine_type = *cfg->engine_type;
788 if (cfg->engine_id) {
789 opts.engine_id = *cfg->engine_id;
792 /* Configure active timeout interval. */
793 opts.active_timeout = cfg->active_timeout;
794 if (!opts.active_timeout) {
795 opts.active_timeout = -1;
796 } else if (opts.active_timeout < 0) {
797 VLOG_WARN("bridge %s: active timeout interval set to negative "
798 "value, using default instead (%d seconds)", br->name,
799 NF_ACTIVE_TIMEOUT_DEFAULT);
800 opts.active_timeout = -1;
803 /* Add engine ID to interface number to disambiguate bridgs? */
804 opts.add_id_to_iface = cfg->add_id_to_interface;
805 if (opts.add_id_to_iface) {
806 if (opts.engine_id > 0x7f) {
807 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
808 "another vswitch, choose an engine id less than 128",
811 if (hmap_count(&br->ports) > 508) {
812 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
813 "another port when more than 508 ports are used",
819 sset_init(&opts.collectors);
820 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
823 if (ofproto_set_netflow(br->ofproto, &opts)) {
824 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
826 sset_destroy(&opts.collectors);
829 /* Set sFlow configuration on 'br'. */
831 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
833 const struct ovsrec_sflow *cfg = br->cfg->sflow;
834 struct ovsrec_controller **controllers;
835 struct ofproto_sflow_options oso;
836 size_t n_controllers;
840 ofproto_set_sflow(br->ofproto, NULL);
844 memset(&oso, 0, sizeof oso);
846 sset_init(&oso.targets);
847 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
849 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
851 oso.sampling_rate = *cfg->sampling;
854 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
856 oso.polling_interval = *cfg->polling;
859 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
861 oso.header_len = *cfg->header;
864 oso.sub_id = (*sflow_bridge_number)++;
865 oso.agent_device = cfg->agent;
867 oso.control_ip = NULL;
868 n_controllers = bridge_get_controllers(br, &controllers);
869 for (i = 0; i < n_controllers; i++) {
870 if (controllers[i]->local_ip) {
871 oso.control_ip = controllers[i]->local_ip;
875 ofproto_set_sflow(br->ofproto, &oso);
877 sset_destroy(&oso.targets);
881 port_configure_stp(const struct ofproto *ofproto, struct port *port,
882 struct ofproto_port_stp_settings *port_s,
883 int *port_num_counter, unsigned long *port_num_bitmap)
885 const char *config_str;
888 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
889 port_s->enable = false;
892 port_s->enable = true;
895 /* STP over bonds is not supported. */
896 if (!list_is_singleton(&port->ifaces)) {
897 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
899 port_s->enable = false;
903 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
905 /* Internal ports shouldn't participate in spanning tree, so
907 if (!strcmp(iface->type, "internal")) {
908 VLOG_DBG("port %s: disable STP on internal ports", port->name);
909 port_s->enable = false;
913 /* STP on mirror output ports is not supported. */
914 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
915 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
916 port_s->enable = false;
920 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
922 unsigned long int port_num = strtoul(config_str, NULL, 0);
923 int port_idx = port_num - 1;
925 if (port_num < 1 || port_num > STP_MAX_PORTS) {
926 VLOG_ERR("port %s: invalid stp-port-num", port->name);
927 port_s->enable = false;
931 if (bitmap_is_set(port_num_bitmap, port_idx)) {
932 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
933 port->name, port_num);
934 port_s->enable = false;
937 bitmap_set1(port_num_bitmap, port_idx);
938 port_s->port_num = port_idx;
940 if (*port_num_counter >= STP_MAX_PORTS) {
941 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
942 port_s->enable = false;
946 port_s->port_num = (*port_num_counter)++;
949 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
951 port_s->path_cost = strtoul(config_str, NULL, 10);
953 enum netdev_features current;
955 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
956 /* Couldn't get speed, so assume 100Mb/s. */
957 port_s->path_cost = 19;
961 mbps = netdev_features_to_bps(current) / 1000000;
962 port_s->path_cost = stp_convert_speed_to_cost(mbps);
966 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
968 port_s->priority = strtoul(config_str, NULL, 0);
970 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
974 /* Set spanning tree configuration on 'br'. */
976 bridge_configure_stp(struct bridge *br)
978 if (!br->cfg->stp_enable) {
979 ofproto_set_stp(br->ofproto, NULL);
981 struct ofproto_stp_settings br_s;
982 const char *config_str;
984 int port_num_counter;
985 unsigned long *port_num_bitmap;
987 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
989 uint8_t ea[ETH_ADDR_LEN];
991 if (eth_addr_from_string(config_str, ea)) {
992 br_s.system_id = eth_addr_to_uint64(ea);
994 br_s.system_id = eth_addr_to_uint64(br->ea);
995 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
996 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
999 br_s.system_id = eth_addr_to_uint64(br->ea);
1002 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1004 br_s.priority = strtoul(config_str, NULL, 0);
1006 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1009 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1011 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1013 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1016 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1018 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1020 br_s.max_age = STP_DEFAULT_MAX_AGE;
1023 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1025 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1027 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1030 /* Configure STP on the bridge. */
1031 if (ofproto_set_stp(br->ofproto, &br_s)) {
1032 VLOG_ERR("bridge %s: could not enable STP", br->name);
1036 /* Users must either set the port number with the "stp-port-num"
1037 * configuration on all ports or none. If manual configuration
1038 * is not done, then we allocate them sequentially. */
1039 port_num_counter = 0;
1040 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1041 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1042 struct ofproto_port_stp_settings port_s;
1043 struct iface *iface;
1045 port_configure_stp(br->ofproto, port, &port_s,
1046 &port_num_counter, port_num_bitmap);
1048 /* As bonds are not supported, just apply configuration to
1049 * all interfaces. */
1050 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1051 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1053 VLOG_ERR("port %s: could not enable STP", port->name);
1059 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1060 && port_num_counter) {
1061 VLOG_ERR("bridge %s: must manually configure all STP port "
1062 "IDs or none, disabling", br->name);
1063 ofproto_set_stp(br->ofproto, NULL);
1065 bitmap_free(port_num_bitmap);
1070 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1072 const struct port *port = port_lookup(br, name);
1073 return port && port_is_bond_fake_iface(port);
1077 port_is_bond_fake_iface(const struct port *port)
1079 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1083 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1085 struct bridge *br, *next;
1086 struct shash new_br;
1089 /* Collect new bridges' names and types. */
1090 shash_init(&new_br);
1091 for (i = 0; i < cfg->n_bridges; i++) {
1092 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1093 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1095 if (strchr(br_cfg->name, '/')) {
1096 /* Prevent remote ovsdb-server users from accessing arbitrary
1097 * directories, e.g. consider a bridge named "../../../etc/". */
1098 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1100 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1101 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1105 /* Get rid of deleted bridges or those whose types have changed.
1106 * Update 'cfg' of bridges that still exist. */
1107 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1108 br->cfg = shash_find_data(&new_br, br->name);
1109 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1110 br->cfg->datapath_type))) {
1115 /* Add new bridges. */
1116 for (i = 0; i < cfg->n_bridges; i++) {
1117 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1118 struct bridge *br = bridge_lookup(br_cfg->name);
1120 bridge_create(br_cfg);
1124 shash_destroy(&new_br);
1128 iface_set_ofp_port(struct iface *iface, int ofp_port)
1130 struct bridge *br = iface->port->bridge;
1132 assert(iface->ofp_port < 0 && ofp_port >= 0);
1133 iface->ofp_port = ofp_port;
1134 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1135 iface_set_ofport(iface->cfg, ofp_port);
1138 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1139 * Returns 0 if successful, otherwise a positive errno value. */
1141 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1142 struct netdev *netdev)
1146 error = netdev_set_config(netdev, &iface_cfg->options);
1148 VLOG_WARN("could not configure network device %s (%s)",
1149 iface_cfg->name, strerror(error));
1154 /* This function determines whether 'ofproto_port', which is attached to
1155 * br->ofproto's datapath, is one that we want in 'br'.
1157 * If it is, it returns true, after creating an iface (if necessary),
1158 * configuring the iface's netdev according to the iface's options, and setting
1159 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1161 * If, on the other hand, 'port' should be removed, it returns false. The
1162 * caller should later detach the port from br->ofproto. */
1164 bridge_refresh_one_ofp_port(struct bridge *br,
1165 const struct ofproto_port *ofproto_port)
1167 const char *name = ofproto_port->name;
1168 const char *type = ofproto_port->type;
1169 uint16_t ofp_port = ofproto_port->ofp_port;
1171 struct iface *iface = iface_lookup(br, name);
1173 /* Check that the name-to-number mapping is one-to-one. */
1174 if (iface->ofp_port >= 0) {
1175 VLOG_WARN("bridge %s: interface %s reported twice",
1178 } else if (iface_from_ofp_port(br, ofp_port)) {
1179 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1180 br->name, ofp_port);
1184 /* There's a configured interface named 'name'. */
1185 if (strcmp(type, iface->type)
1186 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1187 /* It's the wrong type, or it's the right type but can't be
1188 * configured as the user requested, so we must destroy it. */
1191 /* It's the right type and configured correctly. keep it. */
1192 iface_set_ofp_port(iface, ofp_port);
1195 } else if (bridge_has_bond_fake_iface(br, name)
1196 && !strcmp(type, "internal")) {
1197 /* It's a bond fake iface. Keep it. */
1200 /* There's no configured interface named 'name', but there might be an
1201 * interface of that name queued to be created.
1203 * If there is, and it has the correct type, then try to configure it
1204 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1205 * delete it and later try to re-add it. */
1206 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1208 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1209 && iface_create(br, if_cfg, ofp_port));
1213 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1214 * consistent with the ofp_ports in "br->ofproto". */
1216 bridge_refresh_ofp_port(struct bridge *br)
1218 struct ofproto_port_dump dump;
1219 struct ofproto_port ofproto_port;
1220 struct port *port, *port_next;
1222 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1223 hmap_clear(&br->ifaces);
1224 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1225 struct iface *iface;
1227 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1228 iface->ofp_port = -1;
1232 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1233 * already exist in the datapath and promote them to full fledged "struct
1234 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1235 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1236 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1237 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1238 garbage->ofp_port = ofproto_port.ofp_port;
1239 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1243 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1244 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1245 * later they can be added to ofproto. */
1246 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1247 struct iface *iface, *iface_next;
1249 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1250 if (iface->ofp_port < 0) {
1251 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1252 iface_destroy(iface);
1256 if (list_is_empty(&port->ifaces)) {
1262 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1263 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1264 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1267 * If successful, returns 0 and stores the network device in '*netdevp'. On
1268 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1270 iface_do_create(const struct bridge *br,
1271 const struct ovsrec_interface *iface_cfg,
1272 const struct ovsrec_port *port_cfg,
1273 int *ofp_portp, struct netdev **netdevp)
1275 struct netdev *netdev;
1278 error = netdev_open(iface_cfg->name,
1279 iface_get_type(iface_cfg, br->cfg), &netdev);
1281 VLOG_WARN("could not open network device %s (%s)",
1282 iface_cfg->name, strerror(error));
1286 error = iface_set_netdev_config(iface_cfg, netdev);
1291 if (*ofp_portp < 0) {
1294 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1298 *ofp_portp = ofp_port;
1300 VLOG_INFO("bridge %s: added interface %s on port %d",
1301 br->name, iface_cfg->name, *ofp_portp);
1303 VLOG_DBG("bridge %s: interface %s is on port %d",
1304 br->name, iface_cfg->name, *ofp_portp);
1307 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1308 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1316 netdev_close(netdev);
1320 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1321 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1322 * automatically allocated for the iface. Takes ownership of and
1323 * deallocates 'if_cfg'.
1325 * Return true if an iface is successfully created, false otherwise. */
1327 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1329 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1330 const struct ovsrec_port *port_cfg = if_cfg->parent;
1332 struct netdev *netdev;
1333 struct iface *iface;
1337 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1339 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1342 /* Do the bits that can fail up front.
1344 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1345 * internal datastructures may not be consistent. Eventually, when port
1346 * additions and deletions are cheaper, these calls should be removed. */
1348 assert(!iface_lookup(br, iface_cfg->name));
1349 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1352 iface_clear_db_record(iface_cfg);
1356 /* Get or create the port structure. */
1357 port = port_lookup(br, port_cfg->name);
1359 port = port_create(br, port_cfg);
1362 /* Create the iface structure. */
1363 iface = xzalloc(sizeof *iface);
1364 list_push_back(&port->ifaces, &iface->port_elem);
1365 hmap_insert(&br->iface_by_name, &iface->name_node,
1366 hash_string(iface_cfg->name, 0));
1368 iface->name = xstrdup(iface_cfg->name);
1369 iface->ofp_port = -1;
1370 iface->netdev = netdev;
1371 iface->type = iface_get_type(iface_cfg, br->cfg);
1372 iface->cfg = iface_cfg;
1374 iface_set_ofp_port(iface, ofp_port);
1376 /* Populate initial status in database. */
1377 iface_refresh_stats(iface);
1378 iface_refresh_status(iface);
1380 /* Add bond fake iface if necessary. */
1381 if (port_is_bond_fake_iface(port)) {
1382 struct ofproto_port ofproto_port;
1384 if (ofproto_port_query_by_name(br->ofproto, port->name,
1386 struct netdev *netdev;
1389 error = netdev_open(port->name, "internal", &netdev);
1391 ofproto_port_add(br->ofproto, netdev, NULL);
1392 netdev_close(netdev);
1394 VLOG_WARN("could not open network device %s (%s)",
1395 port->name, strerror(error));
1398 /* Already exists, nothing to do. */
1399 ofproto_port_destroy(&ofproto_port);
1406 /* Set Flow eviction threshold */
1408 bridge_configure_flow_eviction_threshold(struct bridge *br)
1410 const char *threshold_str;
1413 threshold_str = smap_get(&br->cfg->other_config,
1414 "flow-eviction-threshold");
1415 if (threshold_str) {
1416 threshold = strtoul(threshold_str, NULL, 10);
1418 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1420 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1423 /* Set forward BPDU option. */
1425 bridge_configure_forward_bpdu(struct bridge *br)
1427 ofproto_set_forward_bpdu(br->ofproto,
1428 smap_get_bool(&br->cfg->other_config,
1433 /* Set MAC aging time for 'br'. */
1435 bridge_configure_mac_idle_time(struct bridge *br)
1437 const char *idle_time_str;
1440 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1441 idle_time = (idle_time_str && atoi(idle_time_str)
1442 ? atoi(idle_time_str)
1443 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1444 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1448 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1449 struct iface **hw_addr_iface)
1451 struct hmapx mirror_output_ports;
1454 bool found_addr = false;
1458 *hw_addr_iface = NULL;
1460 /* Did the user request a particular MAC? */
1461 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1462 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1463 if (eth_addr_is_multicast(ea)) {
1464 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1465 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1466 } else if (eth_addr_is_zero(ea)) {
1467 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1473 /* Mirror output ports don't participate in picking the local hardware
1474 * address. ofproto can't help us find out whether a given port is a
1475 * mirror output because we haven't configured mirrors yet, so we need to
1476 * accumulate them ourselves. */
1477 hmapx_init(&mirror_output_ports);
1478 for (i = 0; i < br->cfg->n_mirrors; i++) {
1479 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1480 if (m->output_port) {
1481 hmapx_add(&mirror_output_ports, m->output_port);
1485 /* Otherwise choose the minimum non-local MAC address among all of the
1487 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1488 uint8_t iface_ea[ETH_ADDR_LEN];
1489 struct iface *candidate;
1490 struct iface *iface;
1492 /* Mirror output ports don't participate. */
1493 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1497 /* Choose the MAC address to represent the port. */
1499 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1500 /* Find the interface with this Ethernet address (if any) so that
1501 * we can provide the correct devname to the caller. */
1502 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1503 uint8_t candidate_ea[ETH_ADDR_LEN];
1504 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1505 && eth_addr_equals(iface_ea, candidate_ea)) {
1510 /* Choose the interface whose MAC address will represent the port.
1511 * The Linux kernel bonding code always chooses the MAC address of
1512 * the first slave added to a bond, and the Fedora networking
1513 * scripts always add slaves to a bond in alphabetical order, so
1514 * for compatibility we choose the interface with the name that is
1515 * first in alphabetical order. */
1516 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1517 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1522 /* The local port doesn't count (since we're trying to choose its
1523 * MAC address anyway). */
1524 if (iface->ofp_port == OFPP_LOCAL) {
1529 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1535 /* Compare against our current choice. */
1536 if (!eth_addr_is_multicast(iface_ea) &&
1537 !eth_addr_is_local(iface_ea) &&
1538 !eth_addr_is_reserved(iface_ea) &&
1539 !eth_addr_is_zero(iface_ea) &&
1540 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1542 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1543 *hw_addr_iface = iface;
1548 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1549 br->name, ETH_ADDR_ARGS(ea));
1551 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1552 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1553 *hw_addr_iface = NULL;
1554 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1555 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1558 hmapx_destroy(&mirror_output_ports);
1561 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1562 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1563 * an interface on 'br', then that interface must be passed in as
1564 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1565 * 'hw_addr_iface' must be passed in as a null pointer. */
1567 bridge_pick_datapath_id(struct bridge *br,
1568 const uint8_t bridge_ea[ETH_ADDR_LEN],
1569 struct iface *hw_addr_iface)
1572 * The procedure for choosing a bridge MAC address will, in the most
1573 * ordinary case, also choose a unique MAC that we can use as a datapath
1574 * ID. In some special cases, though, multiple bridges will end up with
1575 * the same MAC address. This is OK for the bridges, but it will confuse
1576 * the OpenFlow controller, because each datapath needs a unique datapath
1579 * Datapath IDs must be unique. It is also very desirable that they be
1580 * stable from one run to the next, so that policy set on a datapath
1583 const char *datapath_id;
1586 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1587 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1591 if (!hw_addr_iface) {
1593 * A purely internal bridge, that is, one that has no non-virtual
1594 * network devices on it at all, is difficult because it has no
1595 * natural unique identifier at all.
1597 * When the host is a XenServer, we handle this case by hashing the
1598 * host's UUID with the name of the bridge. Names of bridges are
1599 * persistent across XenServer reboots, although they can be reused if
1600 * an internal network is destroyed and then a new one is later
1601 * created, so this is fairly effective.
1603 * When the host is not a XenServer, we punt by using a random MAC
1604 * address on each run.
1606 const char *host_uuid = xenserver_get_host_uuid();
1608 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1609 dpid = dpid_from_hash(combined, strlen(combined));
1615 return eth_addr_to_uint64(bridge_ea);
1619 dpid_from_hash(const void *data, size_t n)
1621 uint8_t hash[SHA1_DIGEST_SIZE];
1623 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1624 sha1_bytes(data, n, hash);
1625 eth_addr_mark_random(hash);
1626 return eth_addr_to_uint64(hash);
1630 iface_refresh_status(struct iface *iface)
1634 enum netdev_features current;
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_features(iface->netdev, ¤t, NULL, NULL, NULL);
1656 ovsrec_interface_set_duplex(iface->cfg,
1657 netdev_features_is_full_duplex(current)
1659 /* warning: uint64_t -> int64_t conversion */
1660 bps = netdev_features_to_bps(current);
1661 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1664 ovsrec_interface_set_duplex(iface->cfg, NULL);
1665 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1668 error = netdev_get_mtu(iface->netdev, &mtu);
1671 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1674 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1678 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1681 iface_refresh_cfm_stats(struct iface *iface)
1683 const struct ovsrec_interface *cfg = iface->cfg;
1684 int fault, opup, error;
1685 const uint64_t *rmps;
1689 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1692 const char *reasons[CFM_FAULT_N_REASONS];
1693 bool fault_bool = fault;
1697 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1698 int reason = 1 << i;
1699 if (fault & reason) {
1700 reasons[j++] = cfm_fault_reason_to_str(reason);
1704 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1705 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1707 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1708 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1711 opup = ofproto_port_get_cfm_opup(iface->port->bridge->ofproto,
1714 ovsrec_interface_set_cfm_remote_opstate(cfg, opup ? "up" : "down");
1716 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1719 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1720 iface->ofp_port, &rmps, &n_rmps);
1722 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1725 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1728 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1731 int64_t cfm_health = health;
1732 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1734 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1739 iface_refresh_stats(struct iface *iface)
1741 #define IFACE_STATS \
1742 IFACE_STAT(rx_packets, "rx_packets") \
1743 IFACE_STAT(tx_packets, "tx_packets") \
1744 IFACE_STAT(rx_bytes, "rx_bytes") \
1745 IFACE_STAT(tx_bytes, "tx_bytes") \
1746 IFACE_STAT(rx_dropped, "rx_dropped") \
1747 IFACE_STAT(tx_dropped, "tx_dropped") \
1748 IFACE_STAT(rx_errors, "rx_errors") \
1749 IFACE_STAT(tx_errors, "tx_errors") \
1750 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1751 IFACE_STAT(rx_over_errors, "rx_over_err") \
1752 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1753 IFACE_STAT(collisions, "collisions")
1755 #define IFACE_STAT(MEMBER, NAME) NAME,
1756 static char *keys[] = { IFACE_STATS };
1758 int64_t values[ARRAY_SIZE(keys)];
1761 struct netdev_stats stats;
1763 if (iface_is_synthetic(iface)) {
1767 /* Intentionally ignore return value, since errors will set 'stats' to
1768 * all-1s, and we will deal with that correctly below. */
1769 netdev_get_stats(iface->netdev, &stats);
1771 /* Copy statistics into values[] array. */
1773 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1776 assert(i == ARRAY_SIZE(keys));
1778 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1784 br_refresh_stp_status(struct bridge *br)
1786 struct smap smap = SMAP_INITIALIZER(&smap);
1787 struct ofproto *ofproto = br->ofproto;
1788 struct ofproto_stp_status status;
1790 if (ofproto_get_stp_status(ofproto, &status)) {
1794 if (!status.enabled) {
1795 ovsrec_bridge_set_status(br->cfg, NULL);
1799 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1800 STP_ID_ARGS(status.bridge_id));
1801 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1802 STP_ID_ARGS(status.designated_root));
1803 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1805 ovsrec_bridge_set_status(br->cfg, &smap);
1806 smap_destroy(&smap);
1810 port_refresh_stp_status(struct port *port)
1812 struct ofproto *ofproto = port->bridge->ofproto;
1813 struct iface *iface;
1814 struct ofproto_port_stp_status status;
1816 int64_t int_values[3];
1819 if (port_is_synthetic(port)) {
1823 /* STP doesn't currently support bonds. */
1824 if (!list_is_singleton(&port->ifaces)) {
1825 ovsrec_port_set_status(port->cfg, NULL);
1829 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1831 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1835 if (!status.enabled) {
1836 ovsrec_port_set_status(port->cfg, NULL);
1837 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1841 /* Set Status column. */
1843 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1844 smap_add(&smap, "stp_state", stp_state_name(status.state));
1845 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1846 smap_add(&smap, "stp_role", stp_role_name(status.role));
1847 ovsrec_port_set_status(port->cfg, &smap);
1848 smap_destroy(&smap);
1850 /* Set Statistics column. */
1851 keys[0] = "stp_tx_count";
1852 int_values[0] = status.tx_count;
1853 keys[1] = "stp_rx_count";
1854 int_values[1] = status.rx_count;
1855 keys[2] = "stp_error_count";
1856 int_values[2] = status.error_count;
1858 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1859 ARRAY_SIZE(int_values));
1863 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1865 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1869 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1871 bool enable = enable_system_stats(cfg);
1873 system_stats_enable(enable);
1875 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1880 run_system_stats(void)
1882 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1885 stats = system_stats_run();
1887 struct ovsdb_idl_txn *txn;
1888 struct ovsdb_datum datum;
1890 txn = ovsdb_idl_txn_create(idl);
1891 ovsdb_datum_from_smap(&datum, stats);
1892 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1894 ovsdb_idl_txn_commit(txn);
1895 ovsdb_idl_txn_destroy(txn);
1901 static inline const char *
1902 nx_role_to_str(enum nx_role role)
1907 case NX_ROLE_MASTER:
1912 return "*** INVALID ROLE ***";
1917 refresh_controller_status(void)
1921 const struct ovsrec_controller *cfg;
1925 /* Accumulate status for controllers on all bridges. */
1926 HMAP_FOR_EACH (br, node, &all_bridges) {
1927 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1930 /* Update each controller in the database with current status. */
1931 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1932 struct ofproto_controller_info *cinfo =
1933 shash_find_data(&info, cfg->target);
1936 struct smap smap = SMAP_INITIALIZER(&smap);
1937 const char **values = cinfo->pairs.values;
1938 const char **keys = cinfo->pairs.keys;
1941 for (i = 0; i < cinfo->pairs.n; i++) {
1942 smap_add(&smap, keys[i], values[i]);
1945 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1946 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1947 ovsrec_controller_set_status(cfg, &smap);
1948 smap_destroy(&smap);
1950 ovsrec_controller_set_is_connected(cfg, false);
1951 ovsrec_controller_set_role(cfg, NULL);
1952 ovsrec_controller_set_status(cfg, NULL);
1956 ofproto_free_ofproto_controller_info(&info);
1960 refresh_instant_stats(void)
1962 static struct ovsdb_idl_txn *txn = NULL;
1967 txn = ovsdb_idl_txn_create(idl);
1969 HMAP_FOR_EACH (br, node, &all_bridges) {
1970 struct iface *iface;
1973 br_refresh_stp_status(br);
1975 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1976 port_refresh_stp_status(port);
1979 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1980 enum netdev_flags flags;
1981 const char *link_state;
1982 int64_t link_resets;
1985 if (iface_is_synthetic(iface)) {
1989 current = ofproto_port_is_lacp_current(br->ofproto,
1993 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1995 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1998 error = netdev_get_flags(iface->netdev, &flags);
2000 const char *state = flags & NETDEV_UP ? "up" : "down";
2001 ovsrec_interface_set_admin_state(iface->cfg, state);
2003 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2006 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2007 ovsrec_interface_set_link_state(iface->cfg, link_state);
2009 link_resets = netdev_get_carrier_resets(iface->netdev);
2010 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2012 iface_refresh_cfm_stats(iface);
2017 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
2018 ovsdb_idl_txn_destroy(txn);
2023 /* Performs periodic activity required by bridges that needs to be done with
2024 * the least possible latency.
2026 * It makes sense to call this function a couple of times per poll loop, to
2027 * provide a significant performance boost on some benchmarks with ofprotos
2028 * that use the ofproto-dpif implementation. */
2030 bridge_run_fast(void)
2034 HMAP_FOR_EACH (br, node, &all_bridges) {
2035 ofproto_run_fast(br->ofproto);
2042 static const struct ovsrec_open_vswitch null_cfg;
2043 const struct ovsrec_open_vswitch *cfg;
2044 struct ovsdb_idl_txn *reconf_txn = NULL;
2046 bool vlan_splinters_changed;
2049 ovsrec_open_vswitch_init((struct ovsrec_open_vswitch *) &null_cfg);
2051 /* (Re)configure if necessary. */
2052 if (!reconfiguring) {
2055 if (ovsdb_idl_is_lock_contended(idl)) {
2056 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2057 struct bridge *br, *next_br;
2059 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2060 "disabling this process until it goes away");
2062 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2066 } else if (!ovsdb_idl_has_lock(idl)) {
2070 cfg = ovsrec_open_vswitch_first(idl);
2072 /* Let each bridge do the work that it needs to do. */
2073 HMAP_FOR_EACH (br, node, &all_bridges) {
2074 ofproto_run(br->ofproto);
2077 /* Re-configure SSL. We do this on every trip through the main loop,
2078 * instead of just when the database changes, because the contents of the
2079 * key and certificate files can change without the database changing.
2081 * We do this before bridge_reconfigure() because that function might
2082 * initiate SSL connections and thus requires SSL to be configured. */
2083 if (cfg && cfg->ssl) {
2084 const struct ovsrec_ssl *ssl = cfg->ssl;
2086 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2087 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2090 if (!reconfiguring) {
2091 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2092 * usage has changed. */
2093 vlan_splinters_changed = false;
2094 if (vlan_splinters_enabled_anywhere) {
2095 HMAP_FOR_EACH (br, node, &all_bridges) {
2096 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2097 vlan_splinters_changed = true;
2103 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2104 idl_seqno = ovsdb_idl_get_seqno(idl);
2106 reconf_txn = ovsdb_idl_txn_create(idl);
2107 bridge_reconfigure(cfg);
2109 /* We still need to reconfigure to avoid dangling pointers to
2110 * now-destroyed ovsrec structures inside bridge data. */
2111 bridge_reconfigure(&null_cfg);
2116 if (reconfiguring) {
2119 reconf_txn = ovsdb_idl_txn_create(idl);
2121 if (bridge_reconfigure_continue(cfg)) {
2122 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2125 bridge_reconfigure_continue(&null_cfg);
2130 ovsdb_idl_txn_commit(reconf_txn);
2131 ovsdb_idl_txn_destroy(reconf_txn);
2135 /* Refresh interface and mirror stats if necessary. */
2136 if (time_msec() >= iface_stats_timer) {
2138 struct ovsdb_idl_txn *txn;
2140 txn = ovsdb_idl_txn_create(idl);
2141 HMAP_FOR_EACH (br, node, &all_bridges) {
2145 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2146 struct iface *iface;
2148 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2149 iface_refresh_stats(iface);
2150 iface_refresh_status(iface);
2154 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2155 mirror_refresh_stats(m);
2159 refresh_controller_status();
2160 ovsdb_idl_txn_commit(txn);
2161 ovsdb_idl_txn_destroy(txn); /* XXX */
2164 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2168 refresh_instant_stats();
2174 ovsdb_idl_wait(idl);
2176 if (reconfiguring) {
2177 poll_immediate_wake();
2180 if (!hmap_is_empty(&all_bridges)) {
2183 HMAP_FOR_EACH (br, node, &all_bridges) {
2184 ofproto_wait(br->ofproto);
2186 poll_timer_wait_until(iface_stats_timer);
2189 system_stats_wait();
2192 /* Adds some memory usage statistics for bridges into 'usage', for use with
2193 * memory_report(). */
2195 bridge_get_memory_usage(struct simap *usage)
2199 HMAP_FOR_EACH (br, node, &all_bridges) {
2200 ofproto_get_memory_usage(br->ofproto, usage);
2204 /* QoS unixctl user interface functions. */
2206 struct qos_unixctl_show_cbdata {
2208 struct iface *iface;
2212 qos_unixctl_show_cb(unsigned int queue_id,
2213 const struct smap *details,
2216 struct qos_unixctl_show_cbdata *data = aux;
2217 struct ds *ds = data->ds;
2218 struct iface *iface = data->iface;
2219 struct netdev_queue_stats stats;
2220 struct smap_node *node;
2223 ds_put_cstr(ds, "\n");
2225 ds_put_format(ds, "Queue %u:\n", queue_id);
2227 ds_put_cstr(ds, "Default:\n");
2230 SMAP_FOR_EACH (node, details) {
2231 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2234 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2236 if (stats.tx_packets != UINT64_MAX) {
2237 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2240 if (stats.tx_bytes != UINT64_MAX) {
2241 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2244 if (stats.tx_errors != UINT64_MAX) {
2245 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2248 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2249 queue_id, strerror(error));
2254 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2255 const char *argv[], void *aux OVS_UNUSED)
2257 struct ds ds = DS_EMPTY_INITIALIZER;
2258 struct smap smap = SMAP_INITIALIZER(&smap);
2259 struct iface *iface;
2261 struct smap_node *node;
2262 struct qos_unixctl_show_cbdata data;
2265 iface = iface_find(argv[1]);
2267 unixctl_command_reply_error(conn, "no such interface");
2271 netdev_get_qos(iface->netdev, &type, &smap);
2273 if (*type != '\0') {
2274 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2276 SMAP_FOR_EACH (node, &smap) {
2277 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2282 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2285 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2287 unixctl_command_reply(conn, ds_cstr(&ds));
2289 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2290 unixctl_command_reply_error(conn, ds_cstr(&ds));
2293 smap_destroy(&smap);
2297 /* Bridge reconfiguration functions. */
2299 bridge_create(const struct ovsrec_bridge *br_cfg)
2303 assert(!bridge_lookup(br_cfg->name));
2304 br = xzalloc(sizeof *br);
2306 br->name = xstrdup(br_cfg->name);
2307 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2310 /* Derive the default Ethernet address from the bridge's UUID. This should
2311 * be unique and it will be stable between ovs-vswitchd runs. */
2312 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2313 eth_addr_mark_random(br->default_ea);
2315 hmap_init(&br->ports);
2316 hmap_init(&br->ifaces);
2317 hmap_init(&br->iface_by_name);
2318 hmap_init(&br->mirrors);
2320 hmap_init(&br->if_cfg_todo);
2321 list_init(&br->ofpp_garbage);
2323 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2327 bridge_destroy(struct bridge *br)
2330 struct mirror *mirror, *next_mirror;
2331 struct port *port, *next_port;
2332 struct if_cfg *if_cfg, *next_if_cfg;
2333 struct ofpp_garbage *garbage, *next_garbage;
2335 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2338 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2339 mirror_destroy(mirror);
2341 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2342 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2345 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2346 &br->ofpp_garbage) {
2347 list_remove(&garbage->list_node);
2351 hmap_remove(&all_bridges, &br->node);
2352 ofproto_destroy(br->ofproto);
2353 hmap_destroy(&br->ifaces);
2354 hmap_destroy(&br->ports);
2355 hmap_destroy(&br->iface_by_name);
2356 hmap_destroy(&br->mirrors);
2357 hmap_destroy(&br->if_cfg_todo);
2364 static struct bridge *
2365 bridge_lookup(const char *name)
2369 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2370 if (!strcmp(br->name, name)) {
2377 /* Handle requests for a listing of all flows known by the OpenFlow
2378 * stack, including those normally hidden. */
2380 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2381 const char *argv[], void *aux OVS_UNUSED)
2386 br = bridge_lookup(argv[1]);
2388 unixctl_command_reply_error(conn, "Unknown bridge");
2393 ofproto_get_all_flows(br->ofproto, &results);
2395 unixctl_command_reply(conn, ds_cstr(&results));
2396 ds_destroy(&results);
2399 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2400 * connections and reconnect. If BRIDGE is not specified, then all bridges
2401 * drop their controller connections and reconnect. */
2403 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2404 const char *argv[], void *aux OVS_UNUSED)
2408 br = bridge_lookup(argv[1]);
2410 unixctl_command_reply_error(conn, "Unknown bridge");
2413 ofproto_reconnect_controllers(br->ofproto);
2415 HMAP_FOR_EACH (br, node, &all_bridges) {
2416 ofproto_reconnect_controllers(br->ofproto);
2419 unixctl_command_reply(conn, NULL);
2423 bridge_get_controllers(const struct bridge *br,
2424 struct ovsrec_controller ***controllersp)
2426 struct ovsrec_controller **controllers;
2427 size_t n_controllers;
2429 controllers = br->cfg->controller;
2430 n_controllers = br->cfg->n_controller;
2432 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2438 *controllersp = controllers;
2440 return n_controllers;
2444 bridge_queue_if_cfg(struct bridge *br,
2445 const struct ovsrec_interface *cfg,
2446 const struct ovsrec_port *parent)
2448 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2451 if_cfg->parent = parent;
2452 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2453 hash_string(if_cfg->cfg->name, 0));
2456 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2457 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2458 * 'br' needs to complete its configuration. */
2460 bridge_add_del_ports(struct bridge *br,
2461 const unsigned long int *splinter_vlans)
2463 struct shash_node *port_node;
2464 struct port *port, *next;
2465 struct shash new_ports;
2468 assert(hmap_is_empty(&br->if_cfg_todo));
2470 /* Collect new ports. */
2471 shash_init(&new_ports);
2472 for (i = 0; i < br->cfg->n_ports; i++) {
2473 const char *name = br->cfg->ports[i]->name;
2474 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2475 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2479 if (bridge_get_controllers(br, NULL)
2480 && !shash_find(&new_ports, br->name)) {
2481 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2482 br->name, br->name);
2484 ovsrec_interface_init(&br->synth_local_iface);
2485 ovsrec_port_init(&br->synth_local_port);
2487 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2488 br->synth_local_port.n_interfaces = 1;
2489 br->synth_local_port.name = br->name;
2491 br->synth_local_iface.name = br->name;
2492 br->synth_local_iface.type = "internal";
2494 br->synth_local_ifacep = &br->synth_local_iface;
2496 shash_add(&new_ports, br->name, &br->synth_local_port);
2499 if (splinter_vlans) {
2500 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2503 /* Get rid of deleted ports.
2504 * Get rid of deleted interfaces on ports that still exist. */
2505 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2506 port->cfg = shash_find_data(&new_ports, port->name);
2510 port_del_ifaces(port);
2514 /* Update iface->cfg and iface->type in interfaces that still exist.
2515 * Add new interfaces to creation queue. */
2516 SHASH_FOR_EACH (port_node, &new_ports) {
2517 const struct ovsrec_port *port = port_node->data;
2520 for (i = 0; i < port->n_interfaces; i++) {
2521 const struct ovsrec_interface *cfg = port->interfaces[i];
2522 struct iface *iface = iface_lookup(br, cfg->name);
2523 const char *type = iface_get_type(cfg, br->cfg);
2528 } else if (!strcmp(type, "null")) {
2529 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2530 " may be removed in February 2013. Please email"
2531 " dev@openvswitch.org with concerns.",
2534 bridge_queue_if_cfg(br, cfg, port);
2539 shash_destroy(&new_ports);
2542 /* Initializes 'oc' appropriately as a management service controller for
2545 * The caller must free oc->target when it is no longer needed. */
2547 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2548 struct ofproto_controller *oc)
2550 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2551 oc->max_backoff = 0;
2552 oc->probe_interval = 60;
2553 oc->band = OFPROTO_OUT_OF_BAND;
2555 oc->burst_limit = 0;
2556 oc->enable_async_msgs = true;
2559 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2561 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2562 struct ofproto_controller *oc)
2566 oc->target = c->target;
2567 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2568 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2569 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2570 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2571 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2572 oc->burst_limit = (c->controller_burst_limit
2573 ? *c->controller_burst_limit : 0);
2574 oc->enable_async_msgs = (!c->enable_async_messages
2575 || *c->enable_async_messages);
2576 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2577 if (dscp < 0 || dscp > 63) {
2578 dscp = DSCP_DEFAULT;
2583 /* Configures the IP stack for 'br''s local interface properly according to the
2584 * configuration in 'c'. */
2586 bridge_configure_local_iface_netdev(struct bridge *br,
2587 struct ovsrec_controller *c)
2589 struct netdev *netdev;
2590 struct in_addr mask, gateway;
2592 struct iface *local_iface;
2595 /* If there's no local interface or no IP address, give up. */
2596 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2597 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2601 /* Bring up the local interface. */
2602 netdev = local_iface->netdev;
2603 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2605 /* Configure the IP address and netmask. */
2606 if (!c->local_netmask
2607 || !inet_aton(c->local_netmask, &mask)
2609 mask.s_addr = guess_netmask(ip.s_addr);
2611 if (!netdev_set_in4(netdev, ip, mask)) {
2612 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2613 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2616 /* Configure the default gateway. */
2617 if (c->local_gateway
2618 && inet_aton(c->local_gateway, &gateway)
2619 && gateway.s_addr) {
2620 if (!netdev_add_router(netdev, gateway)) {
2621 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2622 br->name, IP_ARGS(&gateway.s_addr));
2627 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2628 * in either string are treated as equal to any number of slashes in the other,
2629 * e.g. "x///y" is equal to "x/y". */
2631 equal_pathnames(const char *a, const char *b)
2635 a += strspn(a, "/");
2636 b += strspn(b, "/");
2637 } else if (*a == '\0') {
2648 bridge_configure_remotes(struct bridge *br,
2649 const struct sockaddr_in *managers, size_t n_managers)
2651 bool disable_in_band;
2653 struct ovsrec_controller **controllers;
2654 size_t n_controllers;
2656 enum ofproto_fail_mode fail_mode;
2658 struct ofproto_controller *ocs;
2662 /* Check if we should disable in-band control on this bridge. */
2663 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2666 /* Set OpenFlow queue ID for in-band control. */
2667 ofproto_set_in_band_queue(br->ofproto,
2668 smap_get_int(&br->cfg->other_config,
2669 "in-band-queue", -1));
2671 if (disable_in_band) {
2672 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2674 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2677 n_controllers = bridge_get_controllers(br, &controllers);
2679 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2682 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2683 for (i = 0; i < n_controllers; i++) {
2684 struct ovsrec_controller *c = controllers[i];
2686 if (!strncmp(c->target, "punix:", 6)
2687 || !strncmp(c->target, "unix:", 5)) {
2688 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2691 whitelist = xasprintf("unix:%s/%s.controller",
2692 ovs_rundir(), br->name);
2693 if (!equal_pathnames(c->target, whitelist)) {
2694 /* Prevent remote ovsdb-server users from accessing arbitrary
2695 * Unix domain sockets and overwriting arbitrary local
2697 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2698 "controller \"%s\" due to possibility for remote "
2699 "exploit. Instead, specify whitelisted \"%s\" or "
2700 "connect to \"unix:%s/%s.mgmt\" (which is always "
2701 "available without special configuration).",
2702 br->name, c->target, whitelist,
2703 ovs_rundir(), br->name);
2711 bridge_configure_local_iface_netdev(br, c);
2712 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2713 if (disable_in_band) {
2714 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2719 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2720 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2723 /* Set the fail-mode. */
2724 fail_mode = !br->cfg->fail_mode
2725 || !strcmp(br->cfg->fail_mode, "standalone")
2726 ? OFPROTO_FAIL_STANDALONE
2727 : OFPROTO_FAIL_SECURE;
2728 ofproto_set_fail_mode(br->ofproto, fail_mode);
2730 /* Configure OpenFlow controller connection snooping. */
2731 if (!ofproto_has_snoops(br->ofproto)) {
2735 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2736 ovs_rundir(), br->name));
2737 ofproto_set_snoops(br->ofproto, &snoops);
2738 sset_destroy(&snoops);
2743 bridge_configure_tables(struct bridge *br)
2745 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2749 n_tables = ofproto_get_n_tables(br->ofproto);
2751 for (i = 0; i < n_tables; i++) {
2752 struct ofproto_table_settings s;
2755 s.max_flows = UINT_MAX;
2759 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2760 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2763 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2764 s.max_flows = *cfg->flow_limit;
2766 if (cfg->overflow_policy
2767 && !strcmp(cfg->overflow_policy, "evict")) {
2770 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2771 for (k = 0; k < cfg->n_groups; k++) {
2772 const char *string = cfg->groups[k];
2775 msg = mf_parse_subfield__(&s.groups[k], &string);
2777 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2778 "'groups' (%s)", br->name, i, msg);
2780 } else if (*string) {
2781 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2782 "element '%s' contains trailing garbage",
2783 br->name, i, cfg->groups[k]);
2791 ofproto_configure_table(br->ofproto, i, &s);
2795 for (; j < br->cfg->n_flow_tables; j++) {
2796 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2797 "%"PRId64" not supported by this datapath", br->name,
2798 br->cfg->key_flow_tables[j]);
2802 /* Port functions. */
2804 static struct port *
2805 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2809 port = xzalloc(sizeof *port);
2811 port->name = xstrdup(cfg->name);
2813 list_init(&port->ifaces);
2815 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2819 /* Deletes interfaces from 'port' that are no longer configured for it. */
2821 port_del_ifaces(struct port *port)
2823 struct iface *iface, *next;
2824 struct sset new_ifaces;
2827 /* Collect list of new interfaces. */
2828 sset_init(&new_ifaces);
2829 for (i = 0; i < port->cfg->n_interfaces; i++) {
2830 const char *name = port->cfg->interfaces[i]->name;
2831 const char *type = port->cfg->interfaces[i]->type;
2832 if (strcmp(type, "null")) {
2833 sset_add(&new_ifaces, name);
2837 /* Get rid of deleted interfaces. */
2838 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2839 if (!sset_contains(&new_ifaces, iface->name)) {
2840 iface_destroy(iface);
2844 sset_destroy(&new_ifaces);
2848 port_destroy(struct port *port)
2851 struct bridge *br = port->bridge;
2852 struct iface *iface, *next;
2855 ofproto_bundle_unregister(br->ofproto, port);
2858 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2859 iface_destroy(iface);
2862 hmap_remove(&br->ports, &port->hmap_node);
2868 static struct port *
2869 port_lookup(const struct bridge *br, const char *name)
2873 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2875 if (!strcmp(port->name, name)) {
2883 enable_lacp(struct port *port, bool *activep)
2885 if (!port->cfg->lacp) {
2886 /* XXX when LACP implementation has been sufficiently tested, enable by
2887 * default and make active on bonded ports. */
2889 } else if (!strcmp(port->cfg->lacp, "off")) {
2891 } else if (!strcmp(port->cfg->lacp, "active")) {
2894 } else if (!strcmp(port->cfg->lacp, "passive")) {
2898 VLOG_WARN("port %s: unknown LACP mode %s",
2899 port->name, port->cfg->lacp);
2904 static struct lacp_settings *
2905 port_configure_lacp(struct port *port, struct lacp_settings *s)
2907 const char *lacp_time, *system_id;
2910 if (!enable_lacp(port, &s->active)) {
2914 s->name = port->name;
2916 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2918 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2919 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2920 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2921 " address.", port->name, system_id);
2925 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2928 if (eth_addr_is_zero(s->id)) {
2929 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2933 /* Prefer bondable links if unspecified. */
2934 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2936 s->priority = (priority > 0 && priority <= UINT16_MAX
2938 : UINT16_MAX - !list_is_short(&port->ifaces));
2940 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2941 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2946 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2948 int priority, portid, key;
2950 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2951 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2953 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2955 if (portid <= 0 || portid > UINT16_MAX) {
2956 portid = iface->ofp_port;
2959 if (priority <= 0 || priority > UINT16_MAX) {
2960 priority = UINT16_MAX;
2963 if (key < 0 || key > UINT16_MAX) {
2967 s->name = iface->name;
2969 s->priority = priority;
2974 port_configure_bond(struct port *port, struct bond_settings *s,
2975 uint32_t *bond_stable_ids)
2977 const char *detect_s;
2978 struct iface *iface;
2979 int miimon_interval;
2982 s->name = port->name;
2984 if (port->cfg->bond_mode) {
2985 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2986 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2987 port->name, port->cfg->bond_mode,
2988 bond_mode_to_string(s->balance));
2991 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2993 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2994 * active-backup. At some point we should remove this warning. */
2995 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2996 " in previous versions, the default bond_mode was"
2997 " balance-slb", port->name,
2998 bond_mode_to_string(s->balance));
3000 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3001 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3002 "please use another bond type or disable flood_vlans",
3006 miimon_interval = smap_get_int(&port->cfg->other_config,
3007 "bond-miimon-interval", 0);
3008 if (miimon_interval <= 0) {
3009 miimon_interval = 200;
3012 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3013 if (!detect_s || !strcmp(detect_s, "carrier")) {
3014 miimon_interval = 0;
3015 } else if (strcmp(detect_s, "miimon")) {
3016 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3017 "defaulting to carrier", port->name, detect_s);
3018 miimon_interval = 0;
3021 s->up_delay = MAX(0, port->cfg->bond_updelay);
3022 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3023 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3024 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3025 "bond-rebalance-interval", 10000);
3026 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3027 s->rebalance_interval = 1000;
3030 s->fake_iface = port->cfg->bond_fake_iface;
3033 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3034 long long stable_id;
3036 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3038 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3039 stable_id = iface->ofp_port;
3041 bond_stable_ids[i++] = stable_id;
3043 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3047 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3048 * instead of obtaining it from the database. */
3050 port_is_synthetic(const struct port *port)
3052 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3055 /* Interface functions. */
3057 /* Returns the correct network device type for interface 'iface' in bridge
3060 iface_get_type(const struct ovsrec_interface *iface,
3061 const struct ovsrec_bridge *br)
3063 /* The local port always has type "internal". Other ports take their type
3064 * from the database and default to "system" if none is specified. */
3065 return (!strcmp(iface->name, br->name) ? "internal"
3066 : iface->type[0] ? iface->type
3071 iface_destroy(struct iface *iface)
3074 struct port *port = iface->port;
3075 struct bridge *br = port->bridge;
3077 if (br->ofproto && iface->ofp_port >= 0) {
3078 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3081 if (iface->ofp_port >= 0) {
3082 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3085 list_remove(&iface->port_elem);
3086 hmap_remove(&br->iface_by_name, &iface->name_node);
3088 netdev_close(iface->netdev);
3095 static struct iface *
3096 iface_lookup(const struct bridge *br, const char *name)
3098 struct iface *iface;
3100 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3101 &br->iface_by_name) {
3102 if (!strcmp(iface->name, name)) {
3110 static struct iface *
3111 iface_find(const char *name)
3113 const struct bridge *br;
3115 HMAP_FOR_EACH (br, node, &all_bridges) {
3116 struct iface *iface = iface_lookup(br, name);
3125 static struct if_cfg *
3126 if_cfg_lookup(const struct bridge *br, const char *name)
3128 struct if_cfg *if_cfg;
3130 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3132 if (!strcmp(if_cfg->cfg->name, name)) {
3140 static struct iface *
3141 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3143 struct iface *iface;
3145 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3146 hash_int(ofp_port, 0), &br->ifaces) {
3147 if (iface->ofp_port == ofp_port) {
3154 /* Set Ethernet address of 'iface', if one is specified in the configuration
3157 iface_set_mac(struct iface *iface)
3159 uint8_t ea[ETH_ADDR_LEN];
3161 if (!strcmp(iface->type, "internal")
3162 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3163 if (iface->ofp_port == OFPP_LOCAL) {
3164 VLOG_ERR("interface %s: ignoring mac in Interface record "
3165 "(use Bridge record to set local port's mac)",
3167 } else if (eth_addr_is_multicast(ea)) {
3168 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3171 int error = netdev_set_etheraddr(iface->netdev, ea);
3173 VLOG_ERR("interface %s: setting MAC failed (%s)",
3174 iface->name, strerror(error));
3180 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3182 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3184 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3185 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3189 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3190 * sets the "ofport" field to -1.
3192 * This is appropriate when 'if_cfg''s interface cannot be created or is
3193 * otherwise invalid. */
3195 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3197 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3198 iface_set_ofport(if_cfg, -1);
3199 ovsrec_interface_set_status(if_cfg, NULL);
3200 ovsrec_interface_set_admin_state(if_cfg, NULL);
3201 ovsrec_interface_set_duplex(if_cfg, NULL);
3202 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3203 ovsrec_interface_set_link_state(if_cfg, NULL);
3204 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3205 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3206 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3207 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3208 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3209 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3213 struct iface_delete_queues_cbdata {
3214 struct netdev *netdev;
3215 const struct ovsdb_datum *queues;
3219 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3221 union ovsdb_atom atom;
3223 atom.integer = target;
3224 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3228 iface_delete_queues(unsigned int queue_id,
3229 const struct smap *details OVS_UNUSED, void *cbdata_)
3231 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3233 if (!queue_ids_include(cbdata->queues, queue_id)) {
3234 netdev_delete_queue(cbdata->netdev, queue_id);
3239 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3241 struct ofpbuf queues_buf;
3243 ofpbuf_init(&queues_buf, 0);
3245 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3246 netdev_set_qos(iface->netdev, NULL, NULL);
3248 struct iface_delete_queues_cbdata cbdata;
3252 /* Configure top-level Qos for 'iface'. */
3253 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3255 /* Deconfigure queues that were deleted. */
3256 cbdata.netdev = iface->netdev;
3257 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3259 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3261 /* Configure queues for 'iface'. */
3263 for (i = 0; i < qos->n_queues; i++) {
3264 const struct ovsrec_queue *queue = qos->value_queues[i];
3265 unsigned int queue_id = qos->key_queues[i];
3267 if (queue_id == 0) {
3271 if (queue->n_dscp == 1) {
3272 struct ofproto_port_queue *port_queue;
3274 port_queue = ofpbuf_put_uninit(&queues_buf,
3275 sizeof *port_queue);
3276 port_queue->queue = queue_id;
3277 port_queue->dscp = queue->dscp[0];
3280 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3283 struct smap details;
3285 smap_init(&details);
3286 netdev_set_queue(iface->netdev, 0, &details);
3287 smap_destroy(&details);
3291 if (iface->ofp_port >= 0) {
3292 const struct ofproto_port_queue *port_queues = queues_buf.data;
3293 size_t n_queues = queues_buf.size / sizeof *port_queues;
3295 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3296 port_queues, n_queues);
3299 netdev_set_policing(iface->netdev,
3300 iface->cfg->ingress_policing_rate,
3301 iface->cfg->ingress_policing_burst);
3303 ofpbuf_uninit(&queues_buf);
3307 iface_configure_cfm(struct iface *iface)
3309 const struct ovsrec_interface *cfg = iface->cfg;
3310 const char *opstate_str;
3311 const char *cfm_ccm_vlan;
3312 struct cfm_settings s;
3314 if (!cfg->n_cfm_mpid) {
3315 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3319 s.mpid = *cfg->cfm_mpid;
3320 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3321 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3322 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3324 if (s.interval <= 0) {
3328 if (!cfm_ccm_vlan) {
3330 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3331 s.ccm_vlan = CFM_RANDOM_VLAN;
3333 s.ccm_vlan = atoi(cfm_ccm_vlan);
3334 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3339 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3342 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3343 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3345 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3348 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3349 * instead of obtaining it from the database. */
3351 iface_is_synthetic(const struct iface *iface)
3353 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3357 /* Port mirroring. */
3359 static struct mirror *
3360 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3364 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3365 if (uuid_equals(uuid, &m->uuid)) {
3373 bridge_configure_mirrors(struct bridge *br)
3375 const struct ovsdb_datum *mc;
3376 unsigned long *flood_vlans;
3377 struct mirror *m, *next;
3380 /* Get rid of deleted mirrors. */
3381 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3382 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3383 union ovsdb_atom atom;
3385 atom.uuid = m->uuid;
3386 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3391 /* Add new mirrors and reconfigure existing ones. */
3392 for (i = 0; i < br->cfg->n_mirrors; i++) {
3393 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3394 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3396 m = mirror_create(br, cfg);
3399 if (!mirror_configure(m)) {
3404 /* Update flooded vlans (for RSPAN). */
3405 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3406 br->cfg->n_flood_vlans);
3407 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3408 bitmap_free(flood_vlans);
3411 static struct mirror *
3412 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3416 m = xzalloc(sizeof *m);
3417 m->uuid = cfg->header_.uuid;
3418 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3420 m->name = xstrdup(cfg->name);
3426 mirror_destroy(struct mirror *m)
3429 struct bridge *br = m->bridge;
3432 ofproto_mirror_unregister(br->ofproto, m);
3435 hmap_remove(&br->mirrors, &m->hmap_node);
3442 mirror_collect_ports(struct mirror *m,
3443 struct ovsrec_port **in_ports, int n_in_ports,
3444 void ***out_portsp, size_t *n_out_portsp)
3446 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3447 size_t n_out_ports = 0;
3450 for (i = 0; i < n_in_ports; i++) {
3451 const char *name = in_ports[i]->name;
3452 struct port *port = port_lookup(m->bridge, name);
3454 out_ports[n_out_ports++] = port;
3456 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3457 "port %s", m->bridge->name, m->name, name);
3460 *out_portsp = out_ports;
3461 *n_out_portsp = n_out_ports;
3465 mirror_configure(struct mirror *m)
3467 const struct ovsrec_mirror *cfg = m->cfg;
3468 struct ofproto_mirror_settings s;
3471 if (strcmp(cfg->name, m->name)) {
3473 m->name = xstrdup(cfg->name);
3477 /* Get output port or VLAN. */
3478 if (cfg->output_port) {
3479 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3480 if (!s.out_bundle) {
3481 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3482 m->bridge->name, m->name);
3485 s.out_vlan = UINT16_MAX;
3487 if (cfg->output_vlan) {
3488 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3489 "output vlan; ignoring output vlan",
3490 m->bridge->name, m->name);
3492 } else if (cfg->output_vlan) {
3493 /* The database should prevent invalid VLAN values. */
3494 s.out_bundle = NULL;
3495 s.out_vlan = *cfg->output_vlan;
3497 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3498 m->bridge->name, m->name);
3502 /* Get port selection. */
3503 if (cfg->select_all) {
3504 size_t n_ports = hmap_count(&m->bridge->ports);
3505 void **ports = xmalloc(n_ports * sizeof *ports);
3510 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3520 /* Get ports, dropping ports that don't exist.
3521 * The IDL ensures that there are no duplicates. */
3522 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3523 &s.srcs, &s.n_srcs);
3524 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3525 &s.dsts, &s.n_dsts);
3528 /* Get VLAN selection. */
3529 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3532 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3535 if (s.srcs != s.dsts) {
3544 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3546 * This is deprecated. It is only for compatibility with broken device drivers
3547 * in old versions of Linux that do not properly support VLANs when VLAN
3548 * devices are not used. When broken device drivers are no longer in
3549 * widespread use, we will delete these interfaces. */
3551 static struct ovsrec_port **recs;
3552 static size_t n_recs, allocated_recs;
3554 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3555 * splinters are reconfigured. */
3557 register_rec(struct ovsrec_port *rec)
3559 if (n_recs >= allocated_recs) {
3560 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3562 recs[n_recs++] = rec;
3565 /* Frees all of the ports registered with register_reg(). */
3567 free_registered_recs(void)
3571 for (i = 0; i < n_recs; i++) {
3572 struct ovsrec_port *port = recs[i];
3575 for (j = 0; j < port->n_interfaces; j++) {
3576 struct ovsrec_interface *iface = port->interfaces[j];
3581 smap_destroy(&port->other_config);
3582 free(port->interfaces);
3590 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3593 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3595 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3599 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3602 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3603 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3604 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3607 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3608 * use, returns NULL.
3610 * Updates 'vlan_splinters_enabled_anywhere'. */
3611 static unsigned long int *
3612 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3614 unsigned long int *splinter_vlans;
3615 struct sset splinter_ifaces;
3616 const char *real_dev_name;
3617 struct shash *real_devs;
3618 struct shash_node *node;
3622 /* Free space allocated for synthesized ports and interfaces, since we're
3623 * in the process of reconstructing all of them. */
3624 free_registered_recs();
3626 splinter_vlans = bitmap_allocate(4096);
3627 sset_init(&splinter_ifaces);
3628 vlan_splinters_enabled_anywhere = false;
3629 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3630 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3633 for (j = 0; j < br_cfg->n_ports; j++) {
3634 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3637 for (k = 0; k < port_cfg->n_interfaces; k++) {
3638 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3640 if (vlan_splinters_is_enabled(iface_cfg)) {
3641 vlan_splinters_enabled_anywhere = true;
3642 sset_add(&splinter_ifaces, iface_cfg->name);
3643 vlan_bitmap_from_array__(port_cfg->trunks,
3649 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3650 bitmap_set1(splinter_vlans, *port_cfg->tag);
3655 if (!vlan_splinters_enabled_anywhere) {
3656 free(splinter_vlans);
3657 sset_destroy(&splinter_ifaces);
3661 HMAP_FOR_EACH (br, node, &all_bridges) {
3663 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3667 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3668 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3669 * device to be created for it. */
3670 bitmap_set0(splinter_vlans, 0);
3671 bitmap_set0(splinter_vlans, 4095);
3673 /* Delete all VLAN devices that we don't need. */
3675 real_devs = vlandev_get_real_devs();
3676 SHASH_FOR_EACH (node, real_devs) {
3677 const struct vlan_real_dev *real_dev = node->data;
3678 const struct vlan_dev *vlan_dev;
3679 bool real_dev_has_splinters;
3681 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3683 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3684 if (!real_dev_has_splinters
3685 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3686 struct netdev *netdev;
3688 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3689 if (!netdev_get_in4(netdev, NULL, NULL) ||
3690 !netdev_get_in6(netdev, NULL)) {
3691 vlandev_del(vlan_dev->name);
3693 /* It has an IP address configured, so we don't own
3694 * it. Don't delete it. */
3696 netdev_close(netdev);
3703 /* Add all VLAN devices that we need. */
3704 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3707 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3708 if (!vlandev_get_name(real_dev_name, vid)) {
3709 vlandev_add(real_dev_name, vid);
3716 sset_destroy(&splinter_ifaces);
3718 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3719 free(splinter_vlans);
3722 return splinter_vlans;
3725 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3728 configure_splinter_port(struct port *port)
3730 struct ofproto *ofproto = port->bridge->ofproto;
3731 uint16_t realdev_ofp_port;
3732 const char *realdev_name;
3733 struct iface *vlandev, *realdev;
3735 ofproto_bundle_unregister(port->bridge->ofproto, port);
3737 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3740 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3741 realdev = iface_lookup(port->bridge, realdev_name);
3742 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3744 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3748 static struct ovsrec_port *
3749 synthesize_splinter_port(const char *real_dev_name,
3750 const char *vlan_dev_name, int vid)
3752 struct ovsrec_interface *iface;
3753 struct ovsrec_port *port;
3755 iface = xmalloc(sizeof *iface);
3756 ovsrec_interface_init(iface);
3757 iface->name = xstrdup(vlan_dev_name);
3758 iface->type = "system";
3760 port = xmalloc(sizeof *port);
3761 ovsrec_port_init(port);
3762 port->interfaces = xmemdup(&iface, sizeof iface);
3763 port->n_interfaces = 1;
3764 port->name = xstrdup(vlan_dev_name);
3765 port->vlan_mode = "splinter";
3766 port->tag = xmalloc(sizeof *port->tag);
3769 smap_add(&port->other_config, "realdev", real_dev_name);
3775 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3776 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3777 * 1-bit in the 'splinter_vlans' bitmap. */
3779 add_vlan_splinter_ports(struct bridge *br,
3780 const unsigned long int *splinter_vlans,
3781 struct shash *ports)
3785 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3786 * we're modifying 'ports'. */
3787 for (i = 0; i < br->cfg->n_ports; i++) {
3788 const char *name = br->cfg->ports[i]->name;
3789 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3792 for (j = 0; j < port_cfg->n_interfaces; j++) {
3793 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3795 if (vlan_splinters_is_enabled(iface_cfg)) {
3796 const char *real_dev_name;
3799 real_dev_name = iface_cfg->name;
3800 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3801 const char *vlan_dev_name;
3803 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3805 && !shash_find(ports, vlan_dev_name)) {
3806 shash_add(ports, vlan_dev_name,
3807 synthesize_splinter_port(
3808 real_dev_name, vlan_dev_name, vid));
3817 mirror_refresh_stats(struct mirror *m)
3819 struct ofproto *ofproto = m->bridge->ofproto;
3820 uint64_t tx_packets, tx_bytes;
3823 size_t stat_cnt = 0;
3825 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3826 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3830 if (tx_packets != UINT64_MAX) {
3831 keys[stat_cnt] = "tx_packets";
3832 values[stat_cnt] = tx_packets;
3835 if (tx_bytes != UINT64_MAX) {
3836 keys[stat_cnt] = "tx_bytes";
3837 values[stat_cnt] = tx_bytes;
3841 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);