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. */
69 int64_t ofport; /* Requested OpenFlow port number. */
72 /* OpenFlow port slated for removal from ofproto. */
74 struct list list_node; /* Node in bridge's ofpp_garbage. */
75 uint16_t ofp_port; /* Port to be deleted. */
79 /* These members are always valid. */
80 struct list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct port *port; /* Containing port. */
83 char *name; /* Host network device name. */
85 /* These members are valid only after bridge_reconfigure() causes them to
87 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
88 int ofp_port; /* OpenFlow port number, -1 if unknown. */
89 struct netdev *netdev; /* Network device. */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
131 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
132 Indexed on 'cfg->name'. */
134 /* Port mirroring. */
135 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
137 /* Synthetic local port if necessary. */
138 struct ovsrec_port synth_local_port;
139 struct ovsrec_interface synth_local_iface;
140 struct ovsrec_interface *synth_local_ifacep;
143 /* All bridges, indexed by name. */
144 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
146 /* OVSDB IDL used to obtain configuration. */
147 static struct ovsdb_idl *idl;
149 /* Most recently processed IDL sequence number. */
150 static unsigned int idl_seqno;
152 /* Each time this timer expires, the bridge fetches interface and mirror
153 * statistics and pushes them into the database. */
154 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
155 static long long int iface_stats_timer = LLONG_MIN;
157 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
158 * expensive. This can cause bridge_reconfigure() to take a long time during
159 * which no other work can be done. To deal with this problem, we limit port
160 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
161 * call to bridge_reconfigure(). If there is more work to do after the limit
162 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
163 * This allows the rest of the code to catch up on important things like
164 * forwarding packets. */
165 #define OFP_PORT_ACTION_WINDOW 10
166 static bool reconfiguring = false;
168 static void add_del_bridges(const struct ovsrec_open_vswitch *);
169 static void bridge_update_ofprotos(void);
170 static void bridge_create(const struct ovsrec_bridge *);
171 static void bridge_destroy(struct bridge *);
172 static struct bridge *bridge_lookup(const char *name);
173 static unixctl_cb_func bridge_unixctl_dump_flows;
174 static unixctl_cb_func bridge_unixctl_reconnect;
175 static size_t bridge_get_controllers(const struct bridge *br,
176 struct ovsrec_controller ***controllersp);
177 static void bridge_add_del_ports(struct bridge *,
178 const unsigned long int *splinter_vlans);
179 static void bridge_refresh_ofp_port(struct bridge *);
180 static void bridge_configure_datapath_id(struct bridge *);
181 static void bridge_configure_flow_eviction_threshold(struct bridge *);
182 static void bridge_configure_netflow(struct bridge *);
183 static void bridge_configure_forward_bpdu(struct bridge *);
184 static void bridge_configure_mac_idle_time(struct bridge *);
185 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
186 static void bridge_configure_stp(struct bridge *);
187 static void bridge_configure_tables(struct bridge *);
188 static void bridge_configure_remotes(struct bridge *,
189 const struct sockaddr_in *managers,
191 static void bridge_pick_local_hw_addr(struct bridge *,
192 uint8_t ea[ETH_ADDR_LEN],
193 struct iface **hw_addr_iface);
194 static uint64_t bridge_pick_datapath_id(struct bridge *,
195 const uint8_t bridge_ea[ETH_ADDR_LEN],
196 struct iface *hw_addr_iface);
197 static void bridge_queue_if_cfg(struct bridge *,
198 const struct ovsrec_interface *,
199 const struct ovsrec_port *);
200 static uint64_t dpid_from_hash(const void *, size_t nbytes);
201 static bool bridge_has_bond_fake_iface(const struct bridge *,
203 static bool port_is_bond_fake_iface(const struct port *);
205 static unixctl_cb_func qos_unixctl_show;
207 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
208 static void port_del_ifaces(struct port *);
209 static void port_destroy(struct port *);
210 static struct port *port_lookup(const struct bridge *, const char *name);
211 static void port_configure(struct port *);
212 static struct lacp_settings *port_configure_lacp(struct port *,
213 struct lacp_settings *);
214 static void port_configure_bond(struct port *, struct bond_settings *,
215 uint32_t *bond_stable_ids);
216 static bool port_is_synthetic(const struct port *);
218 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
219 static void run_system_stats(void);
221 static void bridge_configure_mirrors(struct bridge *);
222 static struct mirror *mirror_create(struct bridge *,
223 const struct ovsrec_mirror *);
224 static void mirror_destroy(struct mirror *);
225 static bool mirror_configure(struct mirror *);
226 static void mirror_refresh_stats(struct mirror *);
228 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
229 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
230 static const char *iface_get_type(const struct ovsrec_interface *,
231 const struct ovsrec_bridge *);
232 static void iface_destroy(struct iface *);
233 static struct iface *iface_lookup(const struct bridge *, const char *name);
234 static struct iface *iface_find(const char *name);
235 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
236 static struct iface *iface_from_ofp_port(const struct bridge *,
238 static void iface_set_mac(struct iface *);
239 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
240 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
241 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
242 static void iface_configure_cfm(struct iface *);
243 static void iface_refresh_cfm_stats(struct iface *);
244 static void iface_refresh_stats(struct iface *);
245 static void iface_refresh_status(struct iface *);
246 static bool iface_is_synthetic(const struct iface *);
248 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
250 * This is deprecated. It is only for compatibility with broken device drivers
251 * in old versions of Linux that do not properly support VLANs when VLAN
252 * devices are not used. When broken device drivers are no longer in
253 * widespread use, we will delete these interfaces. */
255 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
256 static bool vlan_splinters_enabled_anywhere;
258 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
259 static unsigned long int *collect_splinter_vlans(
260 const struct ovsrec_open_vswitch *);
261 static void configure_splinter_port(struct port *);
262 static void add_vlan_splinter_ports(struct bridge *,
263 const unsigned long int *splinter_vlans,
264 struct shash *ports);
266 /* Public functions. */
268 /* Initializes the bridge module, configuring it to obtain its configuration
269 * from an OVSDB server accessed over 'remote', which should be a string in a
270 * form acceptable to ovsdb_idl_create(). */
272 bridge_init(const char *remote)
274 /* Create connection to database. */
275 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
276 idl_seqno = ovsdb_idl_get_seqno(idl);
277 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
278 ovsdb_idl_verify_write_only(idl);
280 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
281 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
282 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
288 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
289 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
290 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
292 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
293 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
294 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
297 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
312 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
314 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
315 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
316 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
317 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
319 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
321 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
323 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
324 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
326 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
338 /* Register unixctl commands. */
339 unixctl_command_register("qos/show", "interface", 1, 1,
340 qos_unixctl_show, NULL);
341 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
342 bridge_unixctl_dump_flows, NULL);
343 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
344 bridge_unixctl_reconnect, NULL);
354 struct bridge *br, *next_br;
356 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
359 ovsdb_idl_destroy(idl);
362 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
363 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
364 * responsible for freeing '*managersp' (with free()).
366 * You may be asking yourself "why does ovs-vswitchd care?", because
367 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
368 * should not be and in fact is not directly involved in that. But
369 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
370 * it has to tell in-band control where the managers are to enable that.
371 * (Thus, only managers connected in-band are collected.)
374 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
375 struct sockaddr_in **managersp, size_t *n_managersp)
377 struct sockaddr_in *managers = NULL;
378 size_t n_managers = 0;
382 /* Collect all of the potential targets from the "targets" columns of the
383 * rows pointed to by "manager_options", excluding any that are
386 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
387 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
389 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
390 sset_find_and_delete(&targets, m->target);
392 sset_add(&targets, m->target);
396 /* Now extract the targets' IP addresses. */
397 if (!sset_is_empty(&targets)) {
400 managers = xmalloc(sset_count(&targets) * sizeof *managers);
401 SSET_FOR_EACH (target, &targets) {
402 struct sockaddr_in *sin = &managers[n_managers];
404 if (stream_parse_target_with_default_ports(target,
412 sset_destroy(&targets);
414 *managersp = managers;
415 *n_managersp = n_managers;
419 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
421 unsigned long int *splinter_vlans;
424 COVERAGE_INC(bridge_reconfigure);
426 assert(!reconfiguring);
427 reconfiguring = true;
429 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
430 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
431 * configuration otherwise.
433 * This is mostly an update to bridge data structures. Nothing is pushed
434 * down to ofproto or lower layers. */
435 add_del_bridges(ovs_cfg);
436 splinter_vlans = collect_splinter_vlans(ovs_cfg);
437 HMAP_FOR_EACH (br, node, &all_bridges) {
438 bridge_add_del_ports(br, splinter_vlans);
440 free(splinter_vlans);
442 /* Delete datapaths that are no longer configured, and create ones which
443 * don't exist but should. */
444 bridge_update_ofprotos();
446 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
447 HMAP_FOR_EACH (br, node, &all_bridges) {
448 bridge_refresh_ofp_port(br);
451 /* Clear database records for "if_cfg"s which haven't been instantiated. */
452 HMAP_FOR_EACH (br, node, &all_bridges) {
453 struct if_cfg *if_cfg;
455 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
456 iface_clear_db_record(if_cfg->cfg);
460 reconfigure_system_stats(ovs_cfg);
464 bridge_reconfigure_ofp(void)
466 long long int deadline;
470 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
472 /* The kernel will reject any attempt to add a given port to a datapath if
473 * that port already belongs to a different datapath, so we must do all
474 * port deletions before any port additions. */
475 HMAP_FOR_EACH (br, node, &all_bridges) {
476 struct ofpp_garbage *garbage, *next;
478 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
479 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
480 * internal datastructures may not be consistent. Eventually, when
481 * port additions and deletions are cheaper, these calls should be
484 ofproto_port_del(br->ofproto, garbage->ofp_port);
485 list_remove(&garbage->list_node);
489 if (time_msec() >= deadline) {
496 HMAP_FOR_EACH (br, node, &all_bridges) {
497 struct if_cfg *if_cfg, *next;
499 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
500 iface_create(br, if_cfg, -1);
502 if (time_msec() >= deadline) {
512 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
514 struct sockaddr_in *managers;
515 int sflow_bridge_number;
520 assert(reconfiguring);
521 done = bridge_reconfigure_ofp();
523 /* Complete the configuration. */
524 sflow_bridge_number = 0;
525 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
526 HMAP_FOR_EACH (br, node, &all_bridges) {
529 /* We need the datapath ID early to allow LACP ports to use it as the
530 * default system ID. */
531 bridge_configure_datapath_id(br);
533 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
536 port_configure(port);
538 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
539 iface_configure_cfm(iface);
540 iface_configure_qos(iface, port->cfg->qos);
541 iface_set_mac(iface);
544 bridge_configure_mirrors(br);
545 bridge_configure_flow_eviction_threshold(br);
546 bridge_configure_forward_bpdu(br);
547 bridge_configure_mac_idle_time(br);
548 bridge_configure_remotes(br, managers, n_managers);
549 bridge_configure_netflow(br);
550 bridge_configure_sflow(br, &sflow_bridge_number);
551 bridge_configure_stp(br);
552 bridge_configure_tables(br);
557 /* ovs-vswitchd has completed initialization, so allow the process that
558 * forked us to exit successfully. */
559 daemonize_complete();
560 reconfiguring = false;
562 VLOG_INFO("%s (Open vSwitch) %s", program_name, VERSION);
568 /* Delete ofprotos which aren't configured or have the wrong type. Create
569 * ofprotos which don't exist but need to. */
571 bridge_update_ofprotos(void)
573 struct bridge *br, *next;
578 /* Delete ofprotos with no bridge or with the wrong type. */
581 ofproto_enumerate_types(&types);
582 SSET_FOR_EACH (type, &types) {
585 ofproto_enumerate_names(type, &names);
586 SSET_FOR_EACH (name, &names) {
587 br = bridge_lookup(name);
588 if (!br || strcmp(type, br->type)) {
589 ofproto_delete(name, type);
593 sset_destroy(&names);
594 sset_destroy(&types);
596 /* Add ofprotos for bridges which don't have one yet. */
597 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
605 /* Remove ports from any datapath with the same name as 'br'. If we
606 * don't do this, creating 'br''s ofproto will fail because a port with
607 * the same name as its local port already exists. */
608 HMAP_FOR_EACH (br2, node, &all_bridges) {
609 struct ofproto_port ofproto_port;
615 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
617 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
619 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
622 ofproto_port_destroy(&ofproto_port);
626 error = ofproto_create(br->name, br->type, &br->ofproto);
628 VLOG_ERR("failed to create bridge %s: %s", br->name,
636 port_configure(struct port *port)
638 const struct ovsrec_port *cfg = port->cfg;
639 struct bond_settings bond_settings;
640 struct lacp_settings lacp_settings;
641 struct ofproto_bundle_settings s;
644 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
645 configure_splinter_port(port);
654 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
655 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
656 s.slaves[s.n_slaves++] = iface->ofp_port;
661 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
665 /* Get VLAN trunks. */
668 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
672 if (cfg->vlan_mode) {
673 if (!strcmp(cfg->vlan_mode, "access")) {
674 s.vlan_mode = PORT_VLAN_ACCESS;
675 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
676 s.vlan_mode = PORT_VLAN_TRUNK;
677 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
678 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
679 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
680 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
682 /* This "can't happen" because ovsdb-server should prevent it. */
683 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
684 s.vlan_mode = PORT_VLAN_TRUNK;
688 s.vlan_mode = PORT_VLAN_ACCESS;
690 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
694 s.vlan_mode = PORT_VLAN_TRUNK;
697 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
700 /* Get LACP settings. */
701 s.lacp = port_configure_lacp(port, &lacp_settings);
705 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
706 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
707 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
710 s.lacp_slaves = NULL;
713 /* Get bond settings. */
714 if (s.n_slaves > 1) {
715 s.bond = &bond_settings;
716 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
717 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
720 s.bond_stable_ids = NULL;
722 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
723 netdev_set_miimon_interval(iface->netdev, 0);
728 ofproto_bundle_register(port->bridge->ofproto, port, &s);
734 free(s.bond_stable_ids);
737 /* Pick local port hardware address and datapath ID for 'br'. */
739 bridge_configure_datapath_id(struct bridge *br)
741 uint8_t ea[ETH_ADDR_LEN];
743 struct iface *local_iface;
744 struct iface *hw_addr_iface;
747 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
748 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
750 int error = netdev_set_etheraddr(local_iface->netdev, ea);
752 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
753 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
754 "Ethernet address: %s",
755 br->name, strerror(error));
758 memcpy(br->ea, ea, ETH_ADDR_LEN);
760 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
761 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
762 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
763 ofproto_set_datapath_id(br->ofproto, dpid);
766 dpid_string = xasprintf("%016"PRIx64, dpid);
767 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
771 /* Set NetFlow configuration on 'br'. */
773 bridge_configure_netflow(struct bridge *br)
775 struct ovsrec_netflow *cfg = br->cfg->netflow;
776 struct netflow_options opts;
779 ofproto_set_netflow(br->ofproto, NULL);
783 memset(&opts, 0, sizeof opts);
785 /* Get default NetFlow configuration from datapath.
786 * Apply overrides from 'cfg'. */
787 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
788 if (cfg->engine_type) {
789 opts.engine_type = *cfg->engine_type;
791 if (cfg->engine_id) {
792 opts.engine_id = *cfg->engine_id;
795 /* Configure active timeout interval. */
796 opts.active_timeout = cfg->active_timeout;
797 if (!opts.active_timeout) {
798 opts.active_timeout = -1;
799 } else if (opts.active_timeout < 0) {
800 VLOG_WARN("bridge %s: active timeout interval set to negative "
801 "value, using default instead (%d seconds)", br->name,
802 NF_ACTIVE_TIMEOUT_DEFAULT);
803 opts.active_timeout = -1;
806 /* Add engine ID to interface number to disambiguate bridgs? */
807 opts.add_id_to_iface = cfg->add_id_to_interface;
808 if (opts.add_id_to_iface) {
809 if (opts.engine_id > 0x7f) {
810 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
811 "another vswitch, choose an engine id less than 128",
814 if (hmap_count(&br->ports) > 508) {
815 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
816 "another port when more than 508 ports are used",
822 sset_init(&opts.collectors);
823 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
826 if (ofproto_set_netflow(br->ofproto, &opts)) {
827 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
829 sset_destroy(&opts.collectors);
832 /* Set sFlow configuration on 'br'. */
834 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
836 const struct ovsrec_sflow *cfg = br->cfg->sflow;
837 struct ovsrec_controller **controllers;
838 struct ofproto_sflow_options oso;
839 size_t n_controllers;
843 ofproto_set_sflow(br->ofproto, NULL);
847 memset(&oso, 0, sizeof oso);
849 sset_init(&oso.targets);
850 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
852 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
854 oso.sampling_rate = *cfg->sampling;
857 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
859 oso.polling_interval = *cfg->polling;
862 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
864 oso.header_len = *cfg->header;
867 oso.sub_id = (*sflow_bridge_number)++;
868 oso.agent_device = cfg->agent;
870 oso.control_ip = NULL;
871 n_controllers = bridge_get_controllers(br, &controllers);
872 for (i = 0; i < n_controllers; i++) {
873 if (controllers[i]->local_ip) {
874 oso.control_ip = controllers[i]->local_ip;
878 ofproto_set_sflow(br->ofproto, &oso);
880 sset_destroy(&oso.targets);
884 port_configure_stp(const struct ofproto *ofproto, struct port *port,
885 struct ofproto_port_stp_settings *port_s,
886 int *port_num_counter, unsigned long *port_num_bitmap)
888 const char *config_str;
891 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
892 port_s->enable = false;
895 port_s->enable = true;
898 /* STP over bonds is not supported. */
899 if (!list_is_singleton(&port->ifaces)) {
900 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
902 port_s->enable = false;
906 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
908 /* Internal ports shouldn't participate in spanning tree, so
910 if (!strcmp(iface->type, "internal")) {
911 VLOG_DBG("port %s: disable STP on internal ports", port->name);
912 port_s->enable = false;
916 /* STP on mirror output ports is not supported. */
917 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
918 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
919 port_s->enable = false;
923 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
925 unsigned long int port_num = strtoul(config_str, NULL, 0);
926 int port_idx = port_num - 1;
928 if (port_num < 1 || port_num > STP_MAX_PORTS) {
929 VLOG_ERR("port %s: invalid stp-port-num", port->name);
930 port_s->enable = false;
934 if (bitmap_is_set(port_num_bitmap, port_idx)) {
935 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
936 port->name, port_num);
937 port_s->enable = false;
940 bitmap_set1(port_num_bitmap, port_idx);
941 port_s->port_num = port_idx;
943 if (*port_num_counter >= STP_MAX_PORTS) {
944 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
945 port_s->enable = false;
949 port_s->port_num = (*port_num_counter)++;
952 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
954 port_s->path_cost = strtoul(config_str, NULL, 10);
956 enum netdev_features current;
958 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
959 /* Couldn't get speed, so assume 100Mb/s. */
960 port_s->path_cost = 19;
964 mbps = netdev_features_to_bps(current) / 1000000;
965 port_s->path_cost = stp_convert_speed_to_cost(mbps);
969 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
971 port_s->priority = strtoul(config_str, NULL, 0);
973 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
977 /* Set spanning tree configuration on 'br'. */
979 bridge_configure_stp(struct bridge *br)
981 if (!br->cfg->stp_enable) {
982 ofproto_set_stp(br->ofproto, NULL);
984 struct ofproto_stp_settings br_s;
985 const char *config_str;
987 int port_num_counter;
988 unsigned long *port_num_bitmap;
990 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
992 uint8_t ea[ETH_ADDR_LEN];
994 if (eth_addr_from_string(config_str, ea)) {
995 br_s.system_id = eth_addr_to_uint64(ea);
997 br_s.system_id = eth_addr_to_uint64(br->ea);
998 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
999 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1002 br_s.system_id = eth_addr_to_uint64(br->ea);
1005 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1007 br_s.priority = strtoul(config_str, NULL, 0);
1009 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1012 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1014 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1016 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1019 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1021 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1023 br_s.max_age = STP_DEFAULT_MAX_AGE;
1026 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1028 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1030 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1033 /* Configure STP on the bridge. */
1034 if (ofproto_set_stp(br->ofproto, &br_s)) {
1035 VLOG_ERR("bridge %s: could not enable STP", br->name);
1039 /* Users must either set the port number with the "stp-port-num"
1040 * configuration on all ports or none. If manual configuration
1041 * is not done, then we allocate them sequentially. */
1042 port_num_counter = 0;
1043 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1044 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1045 struct ofproto_port_stp_settings port_s;
1046 struct iface *iface;
1048 port_configure_stp(br->ofproto, port, &port_s,
1049 &port_num_counter, port_num_bitmap);
1051 /* As bonds are not supported, just apply configuration to
1052 * all interfaces. */
1053 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1054 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1056 VLOG_ERR("port %s: could not enable STP", port->name);
1062 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1063 && port_num_counter) {
1064 VLOG_ERR("bridge %s: must manually configure all STP port "
1065 "IDs or none, disabling", br->name);
1066 ofproto_set_stp(br->ofproto, NULL);
1068 bitmap_free(port_num_bitmap);
1073 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1075 const struct port *port = port_lookup(br, name);
1076 return port && port_is_bond_fake_iface(port);
1080 port_is_bond_fake_iface(const struct port *port)
1082 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1086 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1088 struct bridge *br, *next;
1089 struct shash new_br;
1092 /* Collect new bridges' names and types. */
1093 shash_init(&new_br);
1094 for (i = 0; i < cfg->n_bridges; i++) {
1095 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1096 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1098 if (strchr(br_cfg->name, '/')) {
1099 /* Prevent remote ovsdb-server users from accessing arbitrary
1100 * directories, e.g. consider a bridge named "../../../etc/". */
1101 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1103 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1104 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1108 /* Get rid of deleted bridges or those whose types have changed.
1109 * Update 'cfg' of bridges that still exist. */
1110 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1111 br->cfg = shash_find_data(&new_br, br->name);
1112 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1113 br->cfg->datapath_type))) {
1118 /* Add new bridges. */
1119 for (i = 0; i < cfg->n_bridges; i++) {
1120 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1121 struct bridge *br = bridge_lookup(br_cfg->name);
1123 bridge_create(br_cfg);
1127 shash_destroy(&new_br);
1131 iface_set_ofp_port(struct iface *iface, int ofp_port)
1133 struct bridge *br = iface->port->bridge;
1135 assert(iface->ofp_port < 0 && ofp_port >= 0);
1136 iface->ofp_port = ofp_port;
1137 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1138 iface_set_ofport(iface->cfg, ofp_port);
1141 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1142 * Returns 0 if successful, otherwise a positive errno value. */
1144 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1145 struct netdev *netdev)
1149 error = netdev_set_config(netdev, &iface_cfg->options);
1151 VLOG_WARN("could not configure network device %s (%s)",
1152 iface_cfg->name, strerror(error));
1157 /* This function determines whether 'ofproto_port', which is attached to
1158 * br->ofproto's datapath, is one that we want in 'br'.
1160 * If it is, it returns true, after creating an iface (if necessary),
1161 * configuring the iface's netdev according to the iface's options, and setting
1162 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1164 * If, on the other hand, 'port' should be removed, it returns false. The
1165 * caller should later detach the port from br->ofproto. */
1167 bridge_refresh_one_ofp_port(struct bridge *br,
1168 const struct ofproto_port *ofproto_port)
1170 const char *name = ofproto_port->name;
1171 const char *type = ofproto_port->type;
1172 uint16_t ofp_port = ofproto_port->ofp_port;
1174 struct iface *iface = iface_lookup(br, name);
1176 /* Check that the name-to-number mapping is one-to-one. */
1177 if (iface->ofp_port >= 0) {
1178 VLOG_WARN("bridge %s: interface %s reported twice",
1181 } else if (iface_from_ofp_port(br, ofp_port)) {
1182 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1183 br->name, ofp_port);
1187 /* There's a configured interface named 'name'. */
1188 if (strcmp(type, iface->type)
1189 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1190 /* It's the wrong type, or it's the right type but can't be
1191 * configured as the user requested, so we must destroy it. */
1194 /* It's the right type and configured correctly. keep it. */
1195 iface_set_ofp_port(iface, ofp_port);
1198 } else if (bridge_has_bond_fake_iface(br, name)
1199 && !strcmp(type, "internal")) {
1200 /* It's a bond fake iface. Keep it. */
1203 /* There's no configured interface named 'name', but there might be an
1204 * interface of that name queued to be created.
1206 * If there is, and it has the correct type, then try to configure it
1207 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1208 * delete it and later try to re-add it. */
1209 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1211 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1212 && iface_create(br, if_cfg, ofp_port));
1216 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1217 * consistent with the ofp_ports in "br->ofproto". */
1219 bridge_refresh_ofp_port(struct bridge *br)
1221 struct ofproto_port_dump dump;
1222 struct ofproto_port ofproto_port;
1223 struct port *port, *port_next;
1225 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1226 hmap_clear(&br->ifaces);
1227 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1228 struct iface *iface;
1230 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1231 iface->ofp_port = -1;
1235 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1236 * already exist in the datapath and promote them to full fledged "struct
1237 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1238 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1239 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1240 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1241 garbage->ofp_port = ofproto_port.ofp_port;
1242 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1246 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1247 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1248 * later they can be added to ofproto. */
1249 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1250 struct iface *iface, *iface_next;
1252 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1253 if (iface->ofp_port < 0) {
1254 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1255 iface_destroy(iface);
1259 if (list_is_empty(&port->ifaces)) {
1265 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1266 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1267 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1270 * If successful, returns 0 and stores the network device in '*netdevp'. On
1271 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1273 iface_do_create(const struct bridge *br,
1274 const struct if_cfg *if_cfg,
1275 int *ofp_portp, struct netdev **netdevp)
1277 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1278 const struct ovsrec_port *port_cfg = if_cfg->parent;
1279 struct netdev *netdev;
1282 error = netdev_open(iface_cfg->name,
1283 iface_get_type(iface_cfg, br->cfg), &netdev);
1285 VLOG_WARN("could not open network device %s (%s)",
1286 iface_cfg->name, strerror(error));
1290 error = iface_set_netdev_config(iface_cfg, netdev);
1295 if (*ofp_portp < 0) {
1296 uint16_t ofp_port = if_cfg->ofport;
1298 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1302 *ofp_portp = ofp_port;
1304 VLOG_INFO("bridge %s: added interface %s on port %d",
1305 br->name, iface_cfg->name, *ofp_portp);
1307 VLOG_DBG("bridge %s: interface %s is on port %d",
1308 br->name, iface_cfg->name, *ofp_portp);
1311 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1312 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1320 netdev_close(netdev);
1324 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1325 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1326 * automatically allocated for the iface. Takes ownership of and
1327 * deallocates 'if_cfg'.
1329 * Return true if an iface is successfully created, false otherwise. */
1331 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1333 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1334 const struct ovsrec_port *port_cfg = if_cfg->parent;
1336 struct netdev *netdev;
1337 struct iface *iface;
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, if_cfg, &ofp_port, &netdev);
1352 iface_clear_db_record(iface_cfg);
1357 /* Get or create the port structure. */
1358 port = port_lookup(br, port_cfg->name);
1360 port = port_create(br, port_cfg);
1363 /* Create the iface structure. */
1364 iface = xzalloc(sizeof *iface);
1365 list_push_back(&port->ifaces, &iface->port_elem);
1366 hmap_insert(&br->iface_by_name, &iface->name_node,
1367 hash_string(iface_cfg->name, 0));
1369 iface->name = xstrdup(iface_cfg->name);
1370 iface->ofp_port = -1;
1371 iface->netdev = netdev;
1372 iface->type = iface_get_type(iface_cfg, br->cfg);
1373 iface->cfg = iface_cfg;
1375 iface_set_ofp_port(iface, ofp_port);
1377 /* Populate initial status in database. */
1378 iface_refresh_stats(iface);
1379 iface_refresh_status(iface);
1381 /* Add bond fake iface if necessary. */
1382 if (port_is_bond_fake_iface(port)) {
1383 struct ofproto_port ofproto_port;
1385 if (ofproto_port_query_by_name(br->ofproto, port->name,
1387 struct netdev *netdev;
1390 error = netdev_open(port->name, "internal", &netdev);
1392 uint16_t ofp_port = if_cfg->ofport;
1394 ofproto_port_add(br->ofproto, netdev, &ofp_port);
1395 netdev_close(netdev);
1397 VLOG_WARN("could not open network device %s (%s)",
1398 port->name, strerror(error));
1401 /* Already exists, nothing to do. */
1402 ofproto_port_destroy(&ofproto_port);
1407 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1413 /* Set Flow eviction threshold */
1415 bridge_configure_flow_eviction_threshold(struct bridge *br)
1417 const char *threshold_str;
1420 threshold_str = smap_get(&br->cfg->other_config,
1421 "flow-eviction-threshold");
1422 if (threshold_str) {
1423 threshold = strtoul(threshold_str, NULL, 10);
1425 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1427 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1430 /* Set forward BPDU option. */
1432 bridge_configure_forward_bpdu(struct bridge *br)
1434 ofproto_set_forward_bpdu(br->ofproto,
1435 smap_get_bool(&br->cfg->other_config,
1440 /* Set MAC aging time for 'br'. */
1442 bridge_configure_mac_idle_time(struct bridge *br)
1444 const char *idle_time_str;
1447 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1448 idle_time = (idle_time_str && atoi(idle_time_str)
1449 ? atoi(idle_time_str)
1450 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1451 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1455 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1456 struct iface **hw_addr_iface)
1458 struct hmapx mirror_output_ports;
1461 bool found_addr = false;
1465 *hw_addr_iface = NULL;
1467 /* Did the user request a particular MAC? */
1468 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1469 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1470 if (eth_addr_is_multicast(ea)) {
1471 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1472 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1473 } else if (eth_addr_is_zero(ea)) {
1474 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1480 /* Mirror output ports don't participate in picking the local hardware
1481 * address. ofproto can't help us find out whether a given port is a
1482 * mirror output because we haven't configured mirrors yet, so we need to
1483 * accumulate them ourselves. */
1484 hmapx_init(&mirror_output_ports);
1485 for (i = 0; i < br->cfg->n_mirrors; i++) {
1486 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1487 if (m->output_port) {
1488 hmapx_add(&mirror_output_ports, m->output_port);
1492 /* Otherwise choose the minimum non-local MAC address among all of the
1494 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1495 uint8_t iface_ea[ETH_ADDR_LEN];
1496 struct iface *candidate;
1497 struct iface *iface;
1499 /* Mirror output ports don't participate. */
1500 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1504 /* Choose the MAC address to represent the port. */
1506 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1507 /* Find the interface with this Ethernet address (if any) so that
1508 * we can provide the correct devname to the caller. */
1509 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1510 uint8_t candidate_ea[ETH_ADDR_LEN];
1511 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1512 && eth_addr_equals(iface_ea, candidate_ea)) {
1517 /* Choose the interface whose MAC address will represent the port.
1518 * The Linux kernel bonding code always chooses the MAC address of
1519 * the first slave added to a bond, and the Fedora networking
1520 * scripts always add slaves to a bond in alphabetical order, so
1521 * for compatibility we choose the interface with the name that is
1522 * first in alphabetical order. */
1523 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1524 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1529 /* The local port doesn't count (since we're trying to choose its
1530 * MAC address anyway). */
1531 if (iface->ofp_port == OFPP_LOCAL) {
1536 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1542 /* Compare against our current choice. */
1543 if (!eth_addr_is_multicast(iface_ea) &&
1544 !eth_addr_is_local(iface_ea) &&
1545 !eth_addr_is_reserved(iface_ea) &&
1546 !eth_addr_is_zero(iface_ea) &&
1547 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1549 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1550 *hw_addr_iface = iface;
1555 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1556 br->name, ETH_ADDR_ARGS(ea));
1558 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1559 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1560 *hw_addr_iface = NULL;
1561 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1562 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1565 hmapx_destroy(&mirror_output_ports);
1568 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1569 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1570 * an interface on 'br', then that interface must be passed in as
1571 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1572 * 'hw_addr_iface' must be passed in as a null pointer. */
1574 bridge_pick_datapath_id(struct bridge *br,
1575 const uint8_t bridge_ea[ETH_ADDR_LEN],
1576 struct iface *hw_addr_iface)
1579 * The procedure for choosing a bridge MAC address will, in the most
1580 * ordinary case, also choose a unique MAC that we can use as a datapath
1581 * ID. In some special cases, though, multiple bridges will end up with
1582 * the same MAC address. This is OK for the bridges, but it will confuse
1583 * the OpenFlow controller, because each datapath needs a unique datapath
1586 * Datapath IDs must be unique. It is also very desirable that they be
1587 * stable from one run to the next, so that policy set on a datapath
1590 const char *datapath_id;
1593 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1594 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1598 if (!hw_addr_iface) {
1600 * A purely internal bridge, that is, one that has no non-virtual
1601 * network devices on it at all, is difficult because it has no
1602 * natural unique identifier at all.
1604 * When the host is a XenServer, we handle this case by hashing the
1605 * host's UUID with the name of the bridge. Names of bridges are
1606 * persistent across XenServer reboots, although they can be reused if
1607 * an internal network is destroyed and then a new one is later
1608 * created, so this is fairly effective.
1610 * When the host is not a XenServer, we punt by using a random MAC
1611 * address on each run.
1613 const char *host_uuid = xenserver_get_host_uuid();
1615 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1616 dpid = dpid_from_hash(combined, strlen(combined));
1622 return eth_addr_to_uint64(bridge_ea);
1626 dpid_from_hash(const void *data, size_t n)
1628 uint8_t hash[SHA1_DIGEST_SIZE];
1630 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1631 sha1_bytes(data, n, hash);
1632 eth_addr_mark_random(hash);
1633 return eth_addr_to_uint64(hash);
1637 iface_refresh_status(struct iface *iface)
1641 enum netdev_features current;
1647 if (iface_is_synthetic(iface)) {
1653 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1654 ovsrec_interface_set_status(iface->cfg, &smap);
1656 ovsrec_interface_set_status(iface->cfg, NULL);
1659 smap_destroy(&smap);
1661 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1663 ovsrec_interface_set_duplex(iface->cfg,
1664 netdev_features_is_full_duplex(current)
1666 /* warning: uint64_t -> int64_t conversion */
1667 bps = netdev_features_to_bps(current);
1668 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1671 ovsrec_interface_set_duplex(iface->cfg, NULL);
1672 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1675 error = netdev_get_mtu(iface->netdev, &mtu);
1678 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1681 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1685 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1688 iface_refresh_cfm_stats(struct iface *iface)
1690 const struct ovsrec_interface *cfg = iface->cfg;
1691 int fault, opup, error;
1692 const uint64_t *rmps;
1696 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1699 const char *reasons[CFM_FAULT_N_REASONS];
1700 bool fault_bool = fault;
1704 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1705 int reason = 1 << i;
1706 if (fault & reason) {
1707 reasons[j++] = cfm_fault_reason_to_str(reason);
1711 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1712 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1714 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1715 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1718 opup = ofproto_port_get_cfm_opup(iface->port->bridge->ofproto,
1721 ovsrec_interface_set_cfm_remote_opstate(cfg, opup ? "up" : "down");
1723 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1726 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1727 iface->ofp_port, &rmps, &n_rmps);
1729 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1732 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1735 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1738 int64_t cfm_health = health;
1739 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1741 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1746 iface_refresh_stats(struct iface *iface)
1748 #define IFACE_STATS \
1749 IFACE_STAT(rx_packets, "rx_packets") \
1750 IFACE_STAT(tx_packets, "tx_packets") \
1751 IFACE_STAT(rx_bytes, "rx_bytes") \
1752 IFACE_STAT(tx_bytes, "tx_bytes") \
1753 IFACE_STAT(rx_dropped, "rx_dropped") \
1754 IFACE_STAT(tx_dropped, "tx_dropped") \
1755 IFACE_STAT(rx_errors, "rx_errors") \
1756 IFACE_STAT(tx_errors, "tx_errors") \
1757 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1758 IFACE_STAT(rx_over_errors, "rx_over_err") \
1759 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1760 IFACE_STAT(collisions, "collisions")
1762 #define IFACE_STAT(MEMBER, NAME) NAME,
1763 static char *keys[] = { IFACE_STATS };
1765 int64_t values[ARRAY_SIZE(keys)];
1768 struct netdev_stats stats;
1770 if (iface_is_synthetic(iface)) {
1774 /* Intentionally ignore return value, since errors will set 'stats' to
1775 * all-1s, and we will deal with that correctly below. */
1776 netdev_get_stats(iface->netdev, &stats);
1778 /* Copy statistics into values[] array. */
1780 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1783 assert(i == ARRAY_SIZE(keys));
1785 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1791 br_refresh_stp_status(struct bridge *br)
1793 struct smap smap = SMAP_INITIALIZER(&smap);
1794 struct ofproto *ofproto = br->ofproto;
1795 struct ofproto_stp_status status;
1797 if (ofproto_get_stp_status(ofproto, &status)) {
1801 if (!status.enabled) {
1802 ovsrec_bridge_set_status(br->cfg, NULL);
1806 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1807 STP_ID_ARGS(status.bridge_id));
1808 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1809 STP_ID_ARGS(status.designated_root));
1810 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1812 ovsrec_bridge_set_status(br->cfg, &smap);
1813 smap_destroy(&smap);
1817 port_refresh_stp_status(struct port *port)
1819 struct ofproto *ofproto = port->bridge->ofproto;
1820 struct iface *iface;
1821 struct ofproto_port_stp_status status;
1823 int64_t int_values[3];
1826 if (port_is_synthetic(port)) {
1830 /* STP doesn't currently support bonds. */
1831 if (!list_is_singleton(&port->ifaces)) {
1832 ovsrec_port_set_status(port->cfg, NULL);
1836 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1838 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1842 if (!status.enabled) {
1843 ovsrec_port_set_status(port->cfg, NULL);
1844 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1848 /* Set Status column. */
1850 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1851 smap_add(&smap, "stp_state", stp_state_name(status.state));
1852 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1853 smap_add(&smap, "stp_role", stp_role_name(status.role));
1854 ovsrec_port_set_status(port->cfg, &smap);
1855 smap_destroy(&smap);
1857 /* Set Statistics column. */
1858 keys[0] = "stp_tx_count";
1859 int_values[0] = status.tx_count;
1860 keys[1] = "stp_rx_count";
1861 int_values[1] = status.rx_count;
1862 keys[2] = "stp_error_count";
1863 int_values[2] = status.error_count;
1865 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1866 ARRAY_SIZE(int_values));
1870 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1872 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1876 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1878 bool enable = enable_system_stats(cfg);
1880 system_stats_enable(enable);
1882 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1887 run_system_stats(void)
1889 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1892 stats = system_stats_run();
1894 struct ovsdb_idl_txn *txn;
1895 struct ovsdb_datum datum;
1897 txn = ovsdb_idl_txn_create(idl);
1898 ovsdb_datum_from_smap(&datum, stats);
1899 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1901 ovsdb_idl_txn_commit(txn);
1902 ovsdb_idl_txn_destroy(txn);
1908 static inline const char *
1909 nx_role_to_str(enum nx_role role)
1914 case NX_ROLE_MASTER:
1919 return "*** INVALID ROLE ***";
1924 refresh_controller_status(void)
1928 const struct ovsrec_controller *cfg;
1932 /* Accumulate status for controllers on all bridges. */
1933 HMAP_FOR_EACH (br, node, &all_bridges) {
1934 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1937 /* Update each controller in the database with current status. */
1938 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1939 struct ofproto_controller_info *cinfo =
1940 shash_find_data(&info, cfg->target);
1943 struct smap smap = SMAP_INITIALIZER(&smap);
1944 const char **values = cinfo->pairs.values;
1945 const char **keys = cinfo->pairs.keys;
1948 for (i = 0; i < cinfo->pairs.n; i++) {
1949 smap_add(&smap, keys[i], values[i]);
1952 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1953 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1954 ovsrec_controller_set_status(cfg, &smap);
1955 smap_destroy(&smap);
1957 ovsrec_controller_set_is_connected(cfg, false);
1958 ovsrec_controller_set_role(cfg, NULL);
1959 ovsrec_controller_set_status(cfg, NULL);
1963 ofproto_free_ofproto_controller_info(&info);
1967 refresh_instant_stats(void)
1969 static struct ovsdb_idl_txn *txn = NULL;
1974 txn = ovsdb_idl_txn_create(idl);
1976 HMAP_FOR_EACH (br, node, &all_bridges) {
1977 struct iface *iface;
1980 br_refresh_stp_status(br);
1982 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1983 port_refresh_stp_status(port);
1986 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1987 enum netdev_flags flags;
1988 const char *link_state;
1989 int64_t link_resets;
1992 if (iface_is_synthetic(iface)) {
1996 current = ofproto_port_is_lacp_current(br->ofproto,
2000 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2002 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2005 error = netdev_get_flags(iface->netdev, &flags);
2007 const char *state = flags & NETDEV_UP ? "up" : "down";
2008 ovsrec_interface_set_admin_state(iface->cfg, state);
2010 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2013 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2014 ovsrec_interface_set_link_state(iface->cfg, link_state);
2016 link_resets = netdev_get_carrier_resets(iface->netdev);
2017 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2019 iface_refresh_cfm_stats(iface);
2024 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
2025 ovsdb_idl_txn_destroy(txn);
2030 /* Performs periodic activity required by bridges that needs to be done with
2031 * the least possible latency.
2033 * It makes sense to call this function a couple of times per poll loop, to
2034 * provide a significant performance boost on some benchmarks with ofprotos
2035 * that use the ofproto-dpif implementation. */
2037 bridge_run_fast(void)
2041 HMAP_FOR_EACH (br, node, &all_bridges) {
2042 ofproto_run_fast(br->ofproto);
2049 static const struct ovsrec_open_vswitch null_cfg;
2050 const struct ovsrec_open_vswitch *cfg;
2051 struct ovsdb_idl_txn *reconf_txn = NULL;
2053 bool vlan_splinters_changed;
2056 ovsrec_open_vswitch_init((struct ovsrec_open_vswitch *) &null_cfg);
2058 /* (Re)configure if necessary. */
2059 if (!reconfiguring) {
2062 if (ovsdb_idl_is_lock_contended(idl)) {
2063 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2064 struct bridge *br, *next_br;
2066 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2067 "disabling this process until it goes away");
2069 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2073 } else if (!ovsdb_idl_has_lock(idl)) {
2077 cfg = ovsrec_open_vswitch_first(idl);
2079 /* Let each bridge do the work that it needs to do. */
2080 HMAP_FOR_EACH (br, node, &all_bridges) {
2081 ofproto_run(br->ofproto);
2084 /* Re-configure SSL. We do this on every trip through the main loop,
2085 * instead of just when the database changes, because the contents of the
2086 * key and certificate files can change without the database changing.
2088 * We do this before bridge_reconfigure() because that function might
2089 * initiate SSL connections and thus requires SSL to be configured. */
2090 if (cfg && cfg->ssl) {
2091 const struct ovsrec_ssl *ssl = cfg->ssl;
2093 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2094 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2097 if (!reconfiguring) {
2098 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2099 * usage has changed. */
2100 vlan_splinters_changed = false;
2101 if (vlan_splinters_enabled_anywhere) {
2102 HMAP_FOR_EACH (br, node, &all_bridges) {
2103 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2104 vlan_splinters_changed = true;
2110 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2111 idl_seqno = ovsdb_idl_get_seqno(idl);
2113 reconf_txn = ovsdb_idl_txn_create(idl);
2114 bridge_reconfigure(cfg);
2116 /* We still need to reconfigure to avoid dangling pointers to
2117 * now-destroyed ovsrec structures inside bridge data. */
2118 bridge_reconfigure(&null_cfg);
2123 if (reconfiguring) {
2126 reconf_txn = ovsdb_idl_txn_create(idl);
2128 if (bridge_reconfigure_continue(cfg)) {
2129 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2132 bridge_reconfigure_continue(&null_cfg);
2137 ovsdb_idl_txn_commit(reconf_txn);
2138 ovsdb_idl_txn_destroy(reconf_txn);
2142 /* Refresh interface and mirror stats if necessary. */
2143 if (time_msec() >= iface_stats_timer) {
2145 struct ovsdb_idl_txn *txn;
2147 txn = ovsdb_idl_txn_create(idl);
2148 HMAP_FOR_EACH (br, node, &all_bridges) {
2152 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2153 struct iface *iface;
2155 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2156 iface_refresh_stats(iface);
2157 iface_refresh_status(iface);
2161 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2162 mirror_refresh_stats(m);
2166 refresh_controller_status();
2167 ovsdb_idl_txn_commit(txn);
2168 ovsdb_idl_txn_destroy(txn); /* XXX */
2171 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2175 refresh_instant_stats();
2181 ovsdb_idl_wait(idl);
2183 if (reconfiguring) {
2184 poll_immediate_wake();
2187 if (!hmap_is_empty(&all_bridges)) {
2190 HMAP_FOR_EACH (br, node, &all_bridges) {
2191 ofproto_wait(br->ofproto);
2193 poll_timer_wait_until(iface_stats_timer);
2196 system_stats_wait();
2199 /* Adds some memory usage statistics for bridges into 'usage', for use with
2200 * memory_report(). */
2202 bridge_get_memory_usage(struct simap *usage)
2206 HMAP_FOR_EACH (br, node, &all_bridges) {
2207 ofproto_get_memory_usage(br->ofproto, usage);
2211 /* QoS unixctl user interface functions. */
2213 struct qos_unixctl_show_cbdata {
2215 struct iface *iface;
2219 qos_unixctl_show_cb(unsigned int queue_id,
2220 const struct smap *details,
2223 struct qos_unixctl_show_cbdata *data = aux;
2224 struct ds *ds = data->ds;
2225 struct iface *iface = data->iface;
2226 struct netdev_queue_stats stats;
2227 struct smap_node *node;
2230 ds_put_cstr(ds, "\n");
2232 ds_put_format(ds, "Queue %u:\n", queue_id);
2234 ds_put_cstr(ds, "Default:\n");
2237 SMAP_FOR_EACH (node, details) {
2238 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2241 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2243 if (stats.tx_packets != UINT64_MAX) {
2244 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2247 if (stats.tx_bytes != UINT64_MAX) {
2248 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2251 if (stats.tx_errors != UINT64_MAX) {
2252 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2255 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2256 queue_id, strerror(error));
2261 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2262 const char *argv[], void *aux OVS_UNUSED)
2264 struct ds ds = DS_EMPTY_INITIALIZER;
2265 struct smap smap = SMAP_INITIALIZER(&smap);
2266 struct iface *iface;
2268 struct smap_node *node;
2269 struct qos_unixctl_show_cbdata data;
2272 iface = iface_find(argv[1]);
2274 unixctl_command_reply_error(conn, "no such interface");
2278 netdev_get_qos(iface->netdev, &type, &smap);
2280 if (*type != '\0') {
2281 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2283 SMAP_FOR_EACH (node, &smap) {
2284 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2289 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2292 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2294 unixctl_command_reply(conn, ds_cstr(&ds));
2296 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2297 unixctl_command_reply_error(conn, ds_cstr(&ds));
2300 smap_destroy(&smap);
2304 /* Bridge reconfiguration functions. */
2306 bridge_create(const struct ovsrec_bridge *br_cfg)
2310 assert(!bridge_lookup(br_cfg->name));
2311 br = xzalloc(sizeof *br);
2313 br->name = xstrdup(br_cfg->name);
2314 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2317 /* Derive the default Ethernet address from the bridge's UUID. This should
2318 * be unique and it will be stable between ovs-vswitchd runs. */
2319 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2320 eth_addr_mark_random(br->default_ea);
2322 hmap_init(&br->ports);
2323 hmap_init(&br->ifaces);
2324 hmap_init(&br->iface_by_name);
2325 hmap_init(&br->mirrors);
2327 hmap_init(&br->if_cfg_todo);
2328 list_init(&br->ofpp_garbage);
2330 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2334 bridge_destroy(struct bridge *br)
2337 struct mirror *mirror, *next_mirror;
2338 struct port *port, *next_port;
2339 struct if_cfg *if_cfg, *next_if_cfg;
2340 struct ofpp_garbage *garbage, *next_garbage;
2342 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2345 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2346 mirror_destroy(mirror);
2348 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2349 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2352 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2353 &br->ofpp_garbage) {
2354 list_remove(&garbage->list_node);
2358 hmap_remove(&all_bridges, &br->node);
2359 ofproto_destroy(br->ofproto);
2360 hmap_destroy(&br->ifaces);
2361 hmap_destroy(&br->ports);
2362 hmap_destroy(&br->iface_by_name);
2363 hmap_destroy(&br->mirrors);
2364 hmap_destroy(&br->if_cfg_todo);
2371 static struct bridge *
2372 bridge_lookup(const char *name)
2376 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2377 if (!strcmp(br->name, name)) {
2384 /* Handle requests for a listing of all flows known by the OpenFlow
2385 * stack, including those normally hidden. */
2387 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2388 const char *argv[], void *aux OVS_UNUSED)
2393 br = bridge_lookup(argv[1]);
2395 unixctl_command_reply_error(conn, "Unknown bridge");
2400 ofproto_get_all_flows(br->ofproto, &results);
2402 unixctl_command_reply(conn, ds_cstr(&results));
2403 ds_destroy(&results);
2406 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2407 * connections and reconnect. If BRIDGE is not specified, then all bridges
2408 * drop their controller connections and reconnect. */
2410 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2411 const char *argv[], void *aux OVS_UNUSED)
2415 br = bridge_lookup(argv[1]);
2417 unixctl_command_reply_error(conn, "Unknown bridge");
2420 ofproto_reconnect_controllers(br->ofproto);
2422 HMAP_FOR_EACH (br, node, &all_bridges) {
2423 ofproto_reconnect_controllers(br->ofproto);
2426 unixctl_command_reply(conn, NULL);
2430 bridge_get_controllers(const struct bridge *br,
2431 struct ovsrec_controller ***controllersp)
2433 struct ovsrec_controller **controllers;
2434 size_t n_controllers;
2436 controllers = br->cfg->controller;
2437 n_controllers = br->cfg->n_controller;
2439 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2445 *controllersp = controllers;
2447 return n_controllers;
2451 bridge_queue_if_cfg(struct bridge *br,
2452 const struct ovsrec_interface *cfg,
2453 const struct ovsrec_port *parent)
2455 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2458 if_cfg->parent = parent;
2459 if_cfg->ofport = cfg->n_ofport_request ? *cfg->ofport_request : OFPP_NONE;
2460 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2461 hash_string(if_cfg->cfg->name, 0));
2464 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2465 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2466 * 'br' needs to complete its configuration. */
2468 bridge_add_del_ports(struct bridge *br,
2469 const unsigned long int *splinter_vlans)
2471 struct shash_node *port_node;
2472 struct port *port, *next;
2473 struct shash new_ports;
2476 assert(hmap_is_empty(&br->if_cfg_todo));
2478 /* Collect new ports. */
2479 shash_init(&new_ports);
2480 for (i = 0; i < br->cfg->n_ports; i++) {
2481 const char *name = br->cfg->ports[i]->name;
2482 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2483 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2487 if (bridge_get_controllers(br, NULL)
2488 && !shash_find(&new_ports, br->name)) {
2489 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2490 br->name, br->name);
2492 ovsrec_interface_init(&br->synth_local_iface);
2493 ovsrec_port_init(&br->synth_local_port);
2495 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2496 br->synth_local_port.n_interfaces = 1;
2497 br->synth_local_port.name = br->name;
2499 br->synth_local_iface.name = br->name;
2500 br->synth_local_iface.type = "internal";
2502 br->synth_local_ifacep = &br->synth_local_iface;
2504 shash_add(&new_ports, br->name, &br->synth_local_port);
2507 if (splinter_vlans) {
2508 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2511 /* Get rid of deleted ports.
2512 * Get rid of deleted interfaces on ports that still exist. */
2513 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2514 port->cfg = shash_find_data(&new_ports, port->name);
2518 port_del_ifaces(port);
2522 /* Update iface->cfg and iface->type in interfaces that still exist.
2523 * Add new interfaces to creation queue. */
2524 SHASH_FOR_EACH (port_node, &new_ports) {
2525 const struct ovsrec_port *port = port_node->data;
2528 for (i = 0; i < port->n_interfaces; i++) {
2529 const struct ovsrec_interface *cfg = port->interfaces[i];
2530 struct iface *iface = iface_lookup(br, cfg->name);
2531 const char *type = iface_get_type(cfg, br->cfg);
2536 } else if (!strcmp(type, "null")) {
2537 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2538 " may be removed in February 2013. Please email"
2539 " dev@openvswitch.org with concerns.",
2542 bridge_queue_if_cfg(br, cfg, port);
2547 shash_destroy(&new_ports);
2550 /* Initializes 'oc' appropriately as a management service controller for
2553 * The caller must free oc->target when it is no longer needed. */
2555 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2556 struct ofproto_controller *oc)
2558 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2559 oc->max_backoff = 0;
2560 oc->probe_interval = 60;
2561 oc->band = OFPROTO_OUT_OF_BAND;
2563 oc->burst_limit = 0;
2564 oc->enable_async_msgs = true;
2567 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2569 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2570 struct ofproto_controller *oc)
2574 oc->target = c->target;
2575 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2576 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2577 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2578 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2579 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2580 oc->burst_limit = (c->controller_burst_limit
2581 ? *c->controller_burst_limit : 0);
2582 oc->enable_async_msgs = (!c->enable_async_messages
2583 || *c->enable_async_messages);
2584 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2585 if (dscp < 0 || dscp > 63) {
2586 dscp = DSCP_DEFAULT;
2591 /* Configures the IP stack for 'br''s local interface properly according to the
2592 * configuration in 'c'. */
2594 bridge_configure_local_iface_netdev(struct bridge *br,
2595 struct ovsrec_controller *c)
2597 struct netdev *netdev;
2598 struct in_addr mask, gateway;
2600 struct iface *local_iface;
2603 /* If there's no local interface or no IP address, give up. */
2604 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2605 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2609 /* Bring up the local interface. */
2610 netdev = local_iface->netdev;
2611 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2613 /* Configure the IP address and netmask. */
2614 if (!c->local_netmask
2615 || !inet_aton(c->local_netmask, &mask)
2617 mask.s_addr = guess_netmask(ip.s_addr);
2619 if (!netdev_set_in4(netdev, ip, mask)) {
2620 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2621 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2624 /* Configure the default gateway. */
2625 if (c->local_gateway
2626 && inet_aton(c->local_gateway, &gateway)
2627 && gateway.s_addr) {
2628 if (!netdev_add_router(netdev, gateway)) {
2629 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2630 br->name, IP_ARGS(&gateway.s_addr));
2635 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2636 * in either string are treated as equal to any number of slashes in the other,
2637 * e.g. "x///y" is equal to "x/y". */
2639 equal_pathnames(const char *a, const char *b)
2643 a += strspn(a, "/");
2644 b += strspn(b, "/");
2645 } else if (*a == '\0') {
2656 bridge_configure_remotes(struct bridge *br,
2657 const struct sockaddr_in *managers, size_t n_managers)
2659 bool disable_in_band;
2661 struct ovsrec_controller **controllers;
2662 size_t n_controllers;
2664 enum ofproto_fail_mode fail_mode;
2666 struct ofproto_controller *ocs;
2670 /* Check if we should disable in-band control on this bridge. */
2671 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2674 /* Set OpenFlow queue ID for in-band control. */
2675 ofproto_set_in_band_queue(br->ofproto,
2676 smap_get_int(&br->cfg->other_config,
2677 "in-band-queue", -1));
2679 if (disable_in_band) {
2680 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2682 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2685 n_controllers = bridge_get_controllers(br, &controllers);
2687 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2690 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2691 for (i = 0; i < n_controllers; i++) {
2692 struct ovsrec_controller *c = controllers[i];
2694 if (!strncmp(c->target, "punix:", 6)
2695 || !strncmp(c->target, "unix:", 5)) {
2696 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2699 whitelist = xasprintf("unix:%s/%s.controller",
2700 ovs_rundir(), br->name);
2701 if (!equal_pathnames(c->target, whitelist)) {
2702 /* Prevent remote ovsdb-server users from accessing arbitrary
2703 * Unix domain sockets and overwriting arbitrary local
2705 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2706 "controller \"%s\" due to possibility for remote "
2707 "exploit. Instead, specify whitelisted \"%s\" or "
2708 "connect to \"unix:%s/%s.mgmt\" (which is always "
2709 "available without special configuration).",
2710 br->name, c->target, whitelist,
2711 ovs_rundir(), br->name);
2719 bridge_configure_local_iface_netdev(br, c);
2720 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2721 if (disable_in_band) {
2722 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2727 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2728 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2731 /* Set the fail-mode. */
2732 fail_mode = !br->cfg->fail_mode
2733 || !strcmp(br->cfg->fail_mode, "standalone")
2734 ? OFPROTO_FAIL_STANDALONE
2735 : OFPROTO_FAIL_SECURE;
2736 ofproto_set_fail_mode(br->ofproto, fail_mode);
2738 /* Configure OpenFlow controller connection snooping. */
2739 if (!ofproto_has_snoops(br->ofproto)) {
2743 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2744 ovs_rundir(), br->name));
2745 ofproto_set_snoops(br->ofproto, &snoops);
2746 sset_destroy(&snoops);
2751 bridge_configure_tables(struct bridge *br)
2753 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2757 n_tables = ofproto_get_n_tables(br->ofproto);
2759 for (i = 0; i < n_tables; i++) {
2760 struct ofproto_table_settings s;
2763 s.max_flows = UINT_MAX;
2767 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2768 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2771 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2772 s.max_flows = *cfg->flow_limit;
2774 if (cfg->overflow_policy
2775 && !strcmp(cfg->overflow_policy, "evict")) {
2778 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2779 for (k = 0; k < cfg->n_groups; k++) {
2780 const char *string = cfg->groups[k];
2783 msg = mf_parse_subfield__(&s.groups[k], &string);
2785 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2786 "'groups' (%s)", br->name, i, msg);
2788 } else if (*string) {
2789 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2790 "element '%s' contains trailing garbage",
2791 br->name, i, cfg->groups[k]);
2799 ofproto_configure_table(br->ofproto, i, &s);
2803 for (; j < br->cfg->n_flow_tables; j++) {
2804 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2805 "%"PRId64" not supported by this datapath", br->name,
2806 br->cfg->key_flow_tables[j]);
2810 /* Port functions. */
2812 static struct port *
2813 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2817 port = xzalloc(sizeof *port);
2819 port->name = xstrdup(cfg->name);
2821 list_init(&port->ifaces);
2823 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2827 /* Deletes interfaces from 'port' that are no longer configured for it. */
2829 port_del_ifaces(struct port *port)
2831 struct iface *iface, *next;
2832 struct sset new_ifaces;
2835 /* Collect list of new interfaces. */
2836 sset_init(&new_ifaces);
2837 for (i = 0; i < port->cfg->n_interfaces; i++) {
2838 const char *name = port->cfg->interfaces[i]->name;
2839 const char *type = port->cfg->interfaces[i]->type;
2840 if (strcmp(type, "null")) {
2841 sset_add(&new_ifaces, name);
2845 /* Get rid of deleted interfaces. */
2846 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2847 if (!sset_contains(&new_ifaces, iface->name)) {
2848 iface_destroy(iface);
2852 sset_destroy(&new_ifaces);
2856 port_destroy(struct port *port)
2859 struct bridge *br = port->bridge;
2860 struct iface *iface, *next;
2863 ofproto_bundle_unregister(br->ofproto, port);
2866 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2867 iface_destroy(iface);
2870 hmap_remove(&br->ports, &port->hmap_node);
2876 static struct port *
2877 port_lookup(const struct bridge *br, const char *name)
2881 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2883 if (!strcmp(port->name, name)) {
2891 enable_lacp(struct port *port, bool *activep)
2893 if (!port->cfg->lacp) {
2894 /* XXX when LACP implementation has been sufficiently tested, enable by
2895 * default and make active on bonded ports. */
2897 } else if (!strcmp(port->cfg->lacp, "off")) {
2899 } else if (!strcmp(port->cfg->lacp, "active")) {
2902 } else if (!strcmp(port->cfg->lacp, "passive")) {
2906 VLOG_WARN("port %s: unknown LACP mode %s",
2907 port->name, port->cfg->lacp);
2912 static struct lacp_settings *
2913 port_configure_lacp(struct port *port, struct lacp_settings *s)
2915 const char *lacp_time, *system_id;
2918 if (!enable_lacp(port, &s->active)) {
2922 s->name = port->name;
2924 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2926 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2927 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2928 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2929 " address.", port->name, system_id);
2933 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2936 if (eth_addr_is_zero(s->id)) {
2937 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2941 /* Prefer bondable links if unspecified. */
2942 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2944 s->priority = (priority > 0 && priority <= UINT16_MAX
2946 : UINT16_MAX - !list_is_short(&port->ifaces));
2948 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2949 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2954 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2956 int priority, portid, key;
2958 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2959 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2961 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2963 if (portid <= 0 || portid > UINT16_MAX) {
2964 portid = iface->ofp_port;
2967 if (priority <= 0 || priority > UINT16_MAX) {
2968 priority = UINT16_MAX;
2971 if (key < 0 || key > UINT16_MAX) {
2975 s->name = iface->name;
2977 s->priority = priority;
2982 port_configure_bond(struct port *port, struct bond_settings *s,
2983 uint32_t *bond_stable_ids)
2985 const char *detect_s;
2986 struct iface *iface;
2987 int miimon_interval;
2990 s->name = port->name;
2992 if (port->cfg->bond_mode) {
2993 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2994 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2995 port->name, port->cfg->bond_mode,
2996 bond_mode_to_string(s->balance));
2999 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3001 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3002 * active-backup. At some point we should remove this warning. */
3003 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3004 " in previous versions, the default bond_mode was"
3005 " balance-slb", port->name,
3006 bond_mode_to_string(s->balance));
3008 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3009 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3010 "please use another bond type or disable flood_vlans",
3014 miimon_interval = smap_get_int(&port->cfg->other_config,
3015 "bond-miimon-interval", 0);
3016 if (miimon_interval <= 0) {
3017 miimon_interval = 200;
3020 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3021 if (!detect_s || !strcmp(detect_s, "carrier")) {
3022 miimon_interval = 0;
3023 } else if (strcmp(detect_s, "miimon")) {
3024 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3025 "defaulting to carrier", port->name, detect_s);
3026 miimon_interval = 0;
3029 s->up_delay = MAX(0, port->cfg->bond_updelay);
3030 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3031 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3032 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3033 "bond-rebalance-interval", 10000);
3034 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3035 s->rebalance_interval = 1000;
3038 s->fake_iface = port->cfg->bond_fake_iface;
3041 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3042 long long stable_id;
3044 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3046 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3047 stable_id = iface->ofp_port;
3049 bond_stable_ids[i++] = stable_id;
3051 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3055 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3056 * instead of obtaining it from the database. */
3058 port_is_synthetic(const struct port *port)
3060 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3063 /* Interface functions. */
3065 /* Returns the correct network device type for interface 'iface' in bridge
3068 iface_get_type(const struct ovsrec_interface *iface,
3069 const struct ovsrec_bridge *br)
3071 /* The local port always has type "internal". Other ports take their type
3072 * from the database and default to "system" if none is specified. */
3073 return (!strcmp(iface->name, br->name) ? "internal"
3074 : iface->type[0] ? iface->type
3079 iface_destroy(struct iface *iface)
3082 struct port *port = iface->port;
3083 struct bridge *br = port->bridge;
3085 if (br->ofproto && iface->ofp_port >= 0) {
3086 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3089 if (iface->ofp_port >= 0) {
3090 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3093 list_remove(&iface->port_elem);
3094 hmap_remove(&br->iface_by_name, &iface->name_node);
3096 netdev_close(iface->netdev);
3103 static struct iface *
3104 iface_lookup(const struct bridge *br, const char *name)
3106 struct iface *iface;
3108 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3109 &br->iface_by_name) {
3110 if (!strcmp(iface->name, name)) {
3118 static struct iface *
3119 iface_find(const char *name)
3121 const struct bridge *br;
3123 HMAP_FOR_EACH (br, node, &all_bridges) {
3124 struct iface *iface = iface_lookup(br, name);
3133 static struct if_cfg *
3134 if_cfg_lookup(const struct bridge *br, const char *name)
3136 struct if_cfg *if_cfg;
3138 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3140 if (!strcmp(if_cfg->cfg->name, name)) {
3148 static struct iface *
3149 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3151 struct iface *iface;
3153 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3154 hash_int(ofp_port, 0), &br->ifaces) {
3155 if (iface->ofp_port == ofp_port) {
3162 /* Set Ethernet address of 'iface', if one is specified in the configuration
3165 iface_set_mac(struct iface *iface)
3167 uint8_t ea[ETH_ADDR_LEN];
3169 if (!strcmp(iface->type, "internal")
3170 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3171 if (iface->ofp_port == OFPP_LOCAL) {
3172 VLOG_ERR("interface %s: ignoring mac in Interface record "
3173 "(use Bridge record to set local port's mac)",
3175 } else if (eth_addr_is_multicast(ea)) {
3176 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3179 int error = netdev_set_etheraddr(iface->netdev, ea);
3181 VLOG_ERR("interface %s: setting MAC failed (%s)",
3182 iface->name, strerror(error));
3188 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3190 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3192 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3193 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3197 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3198 * sets the "ofport" field to -1.
3200 * This is appropriate when 'if_cfg''s interface cannot be created or is
3201 * otherwise invalid. */
3203 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3205 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3206 iface_set_ofport(if_cfg, -1);
3207 ovsrec_interface_set_status(if_cfg, NULL);
3208 ovsrec_interface_set_admin_state(if_cfg, NULL);
3209 ovsrec_interface_set_duplex(if_cfg, NULL);
3210 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3211 ovsrec_interface_set_link_state(if_cfg, NULL);
3212 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3213 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3214 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3215 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3216 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3217 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3221 struct iface_delete_queues_cbdata {
3222 struct netdev *netdev;
3223 const struct ovsdb_datum *queues;
3227 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3229 union ovsdb_atom atom;
3231 atom.integer = target;
3232 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3236 iface_delete_queues(unsigned int queue_id,
3237 const struct smap *details OVS_UNUSED, void *cbdata_)
3239 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3241 if (!queue_ids_include(cbdata->queues, queue_id)) {
3242 netdev_delete_queue(cbdata->netdev, queue_id);
3247 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3249 struct ofpbuf queues_buf;
3251 ofpbuf_init(&queues_buf, 0);
3253 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3254 netdev_set_qos(iface->netdev, NULL, NULL);
3256 struct iface_delete_queues_cbdata cbdata;
3260 /* Configure top-level Qos for 'iface'. */
3261 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3263 /* Deconfigure queues that were deleted. */
3264 cbdata.netdev = iface->netdev;
3265 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3267 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3269 /* Configure queues for 'iface'. */
3271 for (i = 0; i < qos->n_queues; i++) {
3272 const struct ovsrec_queue *queue = qos->value_queues[i];
3273 unsigned int queue_id = qos->key_queues[i];
3275 if (queue_id == 0) {
3279 if (queue->n_dscp == 1) {
3280 struct ofproto_port_queue *port_queue;
3282 port_queue = ofpbuf_put_uninit(&queues_buf,
3283 sizeof *port_queue);
3284 port_queue->queue = queue_id;
3285 port_queue->dscp = queue->dscp[0];
3288 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3291 struct smap details;
3293 smap_init(&details);
3294 netdev_set_queue(iface->netdev, 0, &details);
3295 smap_destroy(&details);
3299 if (iface->ofp_port >= 0) {
3300 const struct ofproto_port_queue *port_queues = queues_buf.data;
3301 size_t n_queues = queues_buf.size / sizeof *port_queues;
3303 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3304 port_queues, n_queues);
3307 netdev_set_policing(iface->netdev,
3308 iface->cfg->ingress_policing_rate,
3309 iface->cfg->ingress_policing_burst);
3311 ofpbuf_uninit(&queues_buf);
3315 iface_configure_cfm(struct iface *iface)
3317 const struct ovsrec_interface *cfg = iface->cfg;
3318 const char *opstate_str;
3319 const char *cfm_ccm_vlan;
3320 struct cfm_settings s;
3321 struct smap netdev_args;
3323 if (!cfg->n_cfm_mpid) {
3324 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3328 s.check_tnl_key = false;
3329 smap_init(&netdev_args);
3330 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3331 const char *key = smap_get(&netdev_args, "key");
3332 const char *in_key = smap_get(&netdev_args, "in_key");
3334 s.check_tnl_key = (key && !strcmp(key, "flow"))
3335 || (in_key && !strcmp(in_key, "flow"));
3337 smap_destroy(&netdev_args);
3339 s.mpid = *cfg->cfm_mpid;
3340 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3341 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3342 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3344 if (s.interval <= 0) {
3348 if (!cfm_ccm_vlan) {
3350 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3351 s.ccm_vlan = CFM_RANDOM_VLAN;
3353 s.ccm_vlan = atoi(cfm_ccm_vlan);
3354 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3359 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3362 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3363 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3365 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3368 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3369 * instead of obtaining it from the database. */
3371 iface_is_synthetic(const struct iface *iface)
3373 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3377 /* Port mirroring. */
3379 static struct mirror *
3380 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3384 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3385 if (uuid_equals(uuid, &m->uuid)) {
3393 bridge_configure_mirrors(struct bridge *br)
3395 const struct ovsdb_datum *mc;
3396 unsigned long *flood_vlans;
3397 struct mirror *m, *next;
3400 /* Get rid of deleted mirrors. */
3401 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3402 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3403 union ovsdb_atom atom;
3405 atom.uuid = m->uuid;
3406 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3411 /* Add new mirrors and reconfigure existing ones. */
3412 for (i = 0; i < br->cfg->n_mirrors; i++) {
3413 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3414 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3416 m = mirror_create(br, cfg);
3419 if (!mirror_configure(m)) {
3424 /* Update flooded vlans (for RSPAN). */
3425 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3426 br->cfg->n_flood_vlans);
3427 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3428 bitmap_free(flood_vlans);
3431 static struct mirror *
3432 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3436 m = xzalloc(sizeof *m);
3437 m->uuid = cfg->header_.uuid;
3438 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3440 m->name = xstrdup(cfg->name);
3446 mirror_destroy(struct mirror *m)
3449 struct bridge *br = m->bridge;
3452 ofproto_mirror_unregister(br->ofproto, m);
3455 hmap_remove(&br->mirrors, &m->hmap_node);
3462 mirror_collect_ports(struct mirror *m,
3463 struct ovsrec_port **in_ports, int n_in_ports,
3464 void ***out_portsp, size_t *n_out_portsp)
3466 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3467 size_t n_out_ports = 0;
3470 for (i = 0; i < n_in_ports; i++) {
3471 const char *name = in_ports[i]->name;
3472 struct port *port = port_lookup(m->bridge, name);
3474 out_ports[n_out_ports++] = port;
3476 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3477 "port %s", m->bridge->name, m->name, name);
3480 *out_portsp = out_ports;
3481 *n_out_portsp = n_out_ports;
3485 mirror_configure(struct mirror *m)
3487 const struct ovsrec_mirror *cfg = m->cfg;
3488 struct ofproto_mirror_settings s;
3491 if (strcmp(cfg->name, m->name)) {
3493 m->name = xstrdup(cfg->name);
3497 /* Get output port or VLAN. */
3498 if (cfg->output_port) {
3499 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3500 if (!s.out_bundle) {
3501 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3502 m->bridge->name, m->name);
3505 s.out_vlan = UINT16_MAX;
3507 if (cfg->output_vlan) {
3508 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3509 "output vlan; ignoring output vlan",
3510 m->bridge->name, m->name);
3512 } else if (cfg->output_vlan) {
3513 /* The database should prevent invalid VLAN values. */
3514 s.out_bundle = NULL;
3515 s.out_vlan = *cfg->output_vlan;
3517 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3518 m->bridge->name, m->name);
3522 /* Get port selection. */
3523 if (cfg->select_all) {
3524 size_t n_ports = hmap_count(&m->bridge->ports);
3525 void **ports = xmalloc(n_ports * sizeof *ports);
3530 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3540 /* Get ports, dropping ports that don't exist.
3541 * The IDL ensures that there are no duplicates. */
3542 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3543 &s.srcs, &s.n_srcs);
3544 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3545 &s.dsts, &s.n_dsts);
3548 /* Get VLAN selection. */
3549 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3552 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3555 if (s.srcs != s.dsts) {
3564 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3566 * This is deprecated. It is only for compatibility with broken device drivers
3567 * in old versions of Linux that do not properly support VLANs when VLAN
3568 * devices are not used. When broken device drivers are no longer in
3569 * widespread use, we will delete these interfaces. */
3571 static struct ovsrec_port **recs;
3572 static size_t n_recs, allocated_recs;
3574 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3575 * splinters are reconfigured. */
3577 register_rec(struct ovsrec_port *rec)
3579 if (n_recs >= allocated_recs) {
3580 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3582 recs[n_recs++] = rec;
3585 /* Frees all of the ports registered with register_reg(). */
3587 free_registered_recs(void)
3591 for (i = 0; i < n_recs; i++) {
3592 struct ovsrec_port *port = recs[i];
3595 for (j = 0; j < port->n_interfaces; j++) {
3596 struct ovsrec_interface *iface = port->interfaces[j];
3601 smap_destroy(&port->other_config);
3602 free(port->interfaces);
3610 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3613 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3615 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3619 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3622 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3623 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3624 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3627 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3628 * use, returns NULL.
3630 * Updates 'vlan_splinters_enabled_anywhere'. */
3631 static unsigned long int *
3632 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3634 unsigned long int *splinter_vlans;
3635 struct sset splinter_ifaces;
3636 const char *real_dev_name;
3637 struct shash *real_devs;
3638 struct shash_node *node;
3642 /* Free space allocated for synthesized ports and interfaces, since we're
3643 * in the process of reconstructing all of them. */
3644 free_registered_recs();
3646 splinter_vlans = bitmap_allocate(4096);
3647 sset_init(&splinter_ifaces);
3648 vlan_splinters_enabled_anywhere = false;
3649 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3650 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3653 for (j = 0; j < br_cfg->n_ports; j++) {
3654 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3657 for (k = 0; k < port_cfg->n_interfaces; k++) {
3658 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3660 if (vlan_splinters_is_enabled(iface_cfg)) {
3661 vlan_splinters_enabled_anywhere = true;
3662 sset_add(&splinter_ifaces, iface_cfg->name);
3663 vlan_bitmap_from_array__(port_cfg->trunks,
3669 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3670 bitmap_set1(splinter_vlans, *port_cfg->tag);
3675 if (!vlan_splinters_enabled_anywhere) {
3676 free(splinter_vlans);
3677 sset_destroy(&splinter_ifaces);
3681 HMAP_FOR_EACH (br, node, &all_bridges) {
3683 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3687 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3688 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3689 * device to be created for it. */
3690 bitmap_set0(splinter_vlans, 0);
3691 bitmap_set0(splinter_vlans, 4095);
3693 /* Delete all VLAN devices that we don't need. */
3695 real_devs = vlandev_get_real_devs();
3696 SHASH_FOR_EACH (node, real_devs) {
3697 const struct vlan_real_dev *real_dev = node->data;
3698 const struct vlan_dev *vlan_dev;
3699 bool real_dev_has_splinters;
3701 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3703 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3704 if (!real_dev_has_splinters
3705 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3706 struct netdev *netdev;
3708 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3709 if (!netdev_get_in4(netdev, NULL, NULL) ||
3710 !netdev_get_in6(netdev, NULL)) {
3711 vlandev_del(vlan_dev->name);
3713 /* It has an IP address configured, so we don't own
3714 * it. Don't delete it. */
3716 netdev_close(netdev);
3723 /* Add all VLAN devices that we need. */
3724 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3727 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3728 if (!vlandev_get_name(real_dev_name, vid)) {
3729 vlandev_add(real_dev_name, vid);
3736 sset_destroy(&splinter_ifaces);
3738 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3739 free(splinter_vlans);
3742 return splinter_vlans;
3745 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3748 configure_splinter_port(struct port *port)
3750 struct ofproto *ofproto = port->bridge->ofproto;
3751 uint16_t realdev_ofp_port;
3752 const char *realdev_name;
3753 struct iface *vlandev, *realdev;
3755 ofproto_bundle_unregister(port->bridge->ofproto, port);
3757 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3760 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3761 realdev = iface_lookup(port->bridge, realdev_name);
3762 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3764 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3768 static struct ovsrec_port *
3769 synthesize_splinter_port(const char *real_dev_name,
3770 const char *vlan_dev_name, int vid)
3772 struct ovsrec_interface *iface;
3773 struct ovsrec_port *port;
3775 iface = xmalloc(sizeof *iface);
3776 ovsrec_interface_init(iface);
3777 iface->name = xstrdup(vlan_dev_name);
3778 iface->type = "system";
3780 port = xmalloc(sizeof *port);
3781 ovsrec_port_init(port);
3782 port->interfaces = xmemdup(&iface, sizeof iface);
3783 port->n_interfaces = 1;
3784 port->name = xstrdup(vlan_dev_name);
3785 port->vlan_mode = "splinter";
3786 port->tag = xmalloc(sizeof *port->tag);
3789 smap_add(&port->other_config, "realdev", real_dev_name);
3795 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3796 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3797 * 1-bit in the 'splinter_vlans' bitmap. */
3799 add_vlan_splinter_ports(struct bridge *br,
3800 const unsigned long int *splinter_vlans,
3801 struct shash *ports)
3805 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3806 * we're modifying 'ports'. */
3807 for (i = 0; i < br->cfg->n_ports; i++) {
3808 const char *name = br->cfg->ports[i]->name;
3809 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3812 for (j = 0; j < port_cfg->n_interfaces; j++) {
3813 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3815 if (vlan_splinters_is_enabled(iface_cfg)) {
3816 const char *real_dev_name;
3819 real_dev_name = iface_cfg->name;
3820 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3821 const char *vlan_dev_name;
3823 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3825 && !shash_find(ports, vlan_dev_name)) {
3826 shash_add(ports, vlan_dev_name,
3827 synthesize_splinter_port(
3828 real_dev_name, vlan_dev_name, vid));
3837 mirror_refresh_stats(struct mirror *m)
3839 struct ofproto *ofproto = m->bridge->ofproto;
3840 uint64_t tx_packets, tx_bytes;
3843 size_t stat_cnt = 0;
3845 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3846 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3850 if (tx_packets != UINT64_MAX) {
3851 keys[stat_cnt] = "tx_packets";
3852 values[stat_cnt] = tx_packets;
3855 if (tx_bytes != UINT64_MAX) {
3856 keys[stat_cnt] = "tx_bytes";
3857 values[stat_cnt] = tx_bytes;
3861 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);