1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
71 /* OpenFlow port slated for removal from ofproto. */
73 struct list list_node; /* Node in bridge's ofpp_garbage. */
74 uint16_t ofp_port; /* Port to be deleted. */
78 /* These members are always valid. */
79 struct list port_elem; /* Element in struct port's "ifaces" list. */
80 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
84 /* These members are valid only after bridge_reconfigure() causes them to
86 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
87 int ofp_port; /* OpenFlow port number, -1 if unknown. */
88 struct netdev *netdev; /* Network device. */
89 const char *type; /* Usually same as cfg->type. */
90 const struct ovsrec_interface *cfg;
94 struct uuid uuid; /* UUID of this "mirror" record in database. */
95 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
96 struct bridge *bridge;
98 const struct ovsrec_mirror *cfg;
102 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
103 struct bridge *bridge;
106 const struct ovsrec_port *cfg;
108 /* An ordinary bridge port has 1 interface.
109 * A bridge port for bonding has at least 2 interfaces. */
110 struct list ifaces; /* List of "struct iface"s. */
114 struct hmap_node node; /* In 'all_bridges'. */
115 char *name; /* User-specified arbitrary name. */
116 char *type; /* Datapath type. */
117 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
118 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
119 const struct ovsrec_bridge *cfg;
121 /* OpenFlow switch processing. */
122 struct ofproto *ofproto; /* OpenFlow switch. */
125 struct hmap ports; /* "struct port"s indexed by name. */
126 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
127 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
129 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
130 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
131 Indexed on 'cfg->name'. */
133 /* Port mirroring. */
134 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
136 /* Synthetic local port if necessary. */
137 struct ovsrec_port synth_local_port;
138 struct ovsrec_interface synth_local_iface;
139 struct ovsrec_interface *synth_local_ifacep;
142 /* All bridges, indexed by name. */
143 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
145 /* OVSDB IDL used to obtain configuration. */
146 static struct ovsdb_idl *idl;
148 /* Most recently processed IDL sequence number. */
149 static unsigned int idl_seqno;
151 /* Each time this timer expires, the bridge fetches systems and interface
152 * statistics and pushes them into the database. */
153 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int stats_timer = LLONG_MIN;
156 /* Stores the time after which rate limited statistics may be written to the
157 * database. Only updated when changes to the database require rate limiting.
159 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
160 static long long int db_limiter = LLONG_MIN;
162 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
163 * expensive. This can cause bridge_reconfigure() to take a long time during
164 * which no other work can be done. To deal with this problem, we limit port
165 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
166 * call to bridge_reconfigure(). If there is more work to do after the limit
167 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
168 * This allows the rest of the code to catch up on important things like
169 * forwarding packets. */
170 #define OFP_PORT_ACTION_WINDOW 10
171 static bool reconfiguring = false;
173 static void add_del_bridges(const struct ovsrec_open_vswitch *);
174 static void bridge_update_ofprotos(void);
175 static void bridge_create(const struct ovsrec_bridge *);
176 static void bridge_destroy(struct bridge *);
177 static struct bridge *bridge_lookup(const char *name);
178 static unixctl_cb_func bridge_unixctl_dump_flows;
179 static unixctl_cb_func bridge_unixctl_reconnect;
180 static size_t bridge_get_controllers(const struct bridge *br,
181 struct ovsrec_controller ***controllersp);
182 static void bridge_add_del_ports(struct bridge *,
183 const unsigned long int *splinter_vlans);
184 static void bridge_refresh_ofp_port(struct bridge *);
185 static void bridge_configure_datapath_id(struct bridge *);
186 static void bridge_configure_flow_eviction_threshold(struct bridge *);
187 static void bridge_configure_netflow(struct bridge *);
188 static void bridge_configure_forward_bpdu(struct bridge *);
189 static void bridge_configure_mac_idle_time(struct bridge *);
190 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
191 static void bridge_configure_stp(struct bridge *);
192 static void bridge_configure_tables(struct bridge *);
193 static void bridge_configure_remotes(struct bridge *,
194 const struct sockaddr_in *managers,
196 static void bridge_pick_local_hw_addr(struct bridge *,
197 uint8_t ea[ETH_ADDR_LEN],
198 struct iface **hw_addr_iface);
199 static uint64_t bridge_pick_datapath_id(struct bridge *,
200 const uint8_t bridge_ea[ETH_ADDR_LEN],
201 struct iface *hw_addr_iface);
202 static void bridge_queue_if_cfg(struct bridge *,
203 const struct ovsrec_interface *,
204 const struct ovsrec_port *);
205 static uint64_t dpid_from_hash(const void *, size_t nbytes);
206 static bool bridge_has_bond_fake_iface(const struct bridge *,
208 static bool port_is_bond_fake_iface(const struct port *);
210 static unixctl_cb_func qos_unixctl_show;
212 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
213 static void port_del_ifaces(struct port *);
214 static void port_destroy(struct port *);
215 static struct port *port_lookup(const struct bridge *, const char *name);
216 static void port_configure(struct port *);
217 static struct lacp_settings *port_configure_lacp(struct port *,
218 struct lacp_settings *);
219 static void port_configure_bond(struct port *, struct bond_settings *,
220 uint32_t *bond_stable_ids);
221 static bool port_is_synthetic(const struct port *);
223 static void bridge_configure_mirrors(struct bridge *);
224 static struct mirror *mirror_create(struct bridge *,
225 const struct ovsrec_mirror *);
226 static void mirror_destroy(struct mirror *);
227 static bool mirror_configure(struct mirror *);
228 static void mirror_refresh_stats(struct mirror *);
230 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
231 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
232 static const char *iface_get_type(const struct ovsrec_interface *,
233 const struct ovsrec_bridge *);
234 static void iface_destroy(struct iface *);
235 static struct iface *iface_lookup(const struct bridge *, const char *name);
236 static struct iface *iface_find(const char *name);
237 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
238 static struct iface *iface_from_ofp_port(const struct bridge *,
240 static void iface_set_mac(struct iface *);
241 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
242 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
243 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
244 static void iface_configure_cfm(struct iface *);
245 static void iface_refresh_cfm_stats(struct iface *);
246 static void iface_refresh_stats(struct iface *);
247 static void iface_refresh_status(struct iface *);
248 static bool iface_is_synthetic(const struct iface *);
250 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
252 * This is deprecated. It is only for compatibility with broken device drivers
253 * in old versions of Linux that do not properly support VLANs when VLAN
254 * devices are not used. When broken device drivers are no longer in
255 * widespread use, we will delete these interfaces. */
257 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
258 static bool vlan_splinters_enabled_anywhere;
260 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
261 static unsigned long int *collect_splinter_vlans(
262 const struct ovsrec_open_vswitch *);
263 static void configure_splinter_port(struct port *);
264 static void add_vlan_splinter_ports(struct bridge *,
265 const unsigned long int *splinter_vlans,
266 struct shash *ports);
268 /* Public functions. */
270 /* Initializes the bridge module, configuring it to obtain its configuration
271 * from an OVSDB server accessed over 'remote', which should be a string in a
272 * form acceptable to ovsdb_idl_create(). */
274 bridge_init(const char *remote)
276 /* Create connection to database. */
277 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
278 idl_seqno = ovsdb_idl_get_seqno(idl);
279 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
281 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
282 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
289 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
290 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
291 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
293 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
294 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
295 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
312 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
314 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
315 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
316 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
317 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
319 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
321 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
323 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
324 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
326 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
338 /* Register unixctl commands. */
339 unixctl_command_register("qos/show", "interface", 1, 1,
340 qos_unixctl_show, NULL);
341 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
342 bridge_unixctl_dump_flows, NULL);
343 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
344 bridge_unixctl_reconnect, NULL);
354 struct bridge *br, *next_br;
356 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
359 ovsdb_idl_destroy(idl);
362 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
363 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
364 * responsible for freeing '*managersp' (with free()).
366 * You may be asking yourself "why does ovs-vswitchd care?", because
367 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
368 * should not be and in fact is not directly involved in that. But
369 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
370 * it has to tell in-band control where the managers are to enable that.
371 * (Thus, only managers connected in-band are collected.)
374 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
375 struct sockaddr_in **managersp, size_t *n_managersp)
377 struct sockaddr_in *managers = NULL;
378 size_t n_managers = 0;
382 /* Collect all of the potential targets from the "targets" columns of the
383 * rows pointed to by "manager_options", excluding any that are
386 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
387 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
389 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
390 sset_find_and_delete(&targets, m->target);
392 sset_add(&targets, m->target);
396 /* Now extract the targets' IP addresses. */
397 if (!sset_is_empty(&targets)) {
400 managers = xmalloc(sset_count(&targets) * sizeof *managers);
401 SSET_FOR_EACH (target, &targets) {
402 struct sockaddr_in *sin = &managers[n_managers];
404 if (stream_parse_target_with_default_ports(target,
412 sset_destroy(&targets);
414 *managersp = managers;
415 *n_managersp = n_managers;
419 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
421 unsigned long int *splinter_vlans;
424 COVERAGE_INC(bridge_reconfigure);
426 assert(!reconfiguring);
427 reconfiguring = true;
429 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
430 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
431 * configuration otherwise.
433 * This is mostly an update to bridge data structures. Nothing is pushed
434 * down to ofproto or lower layers. */
435 add_del_bridges(ovs_cfg);
436 splinter_vlans = collect_splinter_vlans(ovs_cfg);
437 HMAP_FOR_EACH (br, node, &all_bridges) {
438 bridge_add_del_ports(br, splinter_vlans);
440 free(splinter_vlans);
442 /* Delete datapaths that are no longer configured, and create ones which
443 * don't exist but should. */
444 bridge_update_ofprotos();
446 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
447 HMAP_FOR_EACH (br, node, &all_bridges) {
448 bridge_refresh_ofp_port(br);
451 /* Clear database records for "if_cfg"s which haven't been instantiated. */
452 HMAP_FOR_EACH (br, node, &all_bridges) {
453 struct if_cfg *if_cfg;
455 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
456 iface_clear_db_record(if_cfg->cfg);
462 bridge_reconfigure_ofp(void)
464 long long int deadline;
468 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
470 /* The kernel will reject any attempt to add a given port to a datapath if
471 * that port already belongs to a different datapath, so we must do all
472 * port deletions before any port additions. */
473 HMAP_FOR_EACH (br, node, &all_bridges) {
474 struct ofpp_garbage *garbage, *next;
476 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
477 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
478 * internal datastructures may not be consistent. Eventually, when
479 * port additions and deletions are cheaper, these calls should be
482 ofproto_port_del(br->ofproto, garbage->ofp_port);
483 list_remove(&garbage->list_node);
487 if (time_msec() >= deadline) {
494 HMAP_FOR_EACH (br, node, &all_bridges) {
495 struct if_cfg *if_cfg, *next;
497 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
498 iface_create(br, if_cfg, -1);
500 if (time_msec() >= deadline) {
510 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
512 struct sockaddr_in *managers;
513 int sflow_bridge_number;
518 assert(reconfiguring);
519 done = bridge_reconfigure_ofp();
521 /* Complete the configuration. */
522 sflow_bridge_number = 0;
523 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
524 HMAP_FOR_EACH (br, node, &all_bridges) {
527 /* We need the datapath ID early to allow LACP ports to use it as the
528 * default system ID. */
529 bridge_configure_datapath_id(br);
531 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
534 port_configure(port);
536 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
537 iface_configure_cfm(iface);
538 iface_configure_qos(iface, port->cfg->qos);
539 iface_set_mac(iface);
542 bridge_configure_mirrors(br);
543 bridge_configure_flow_eviction_threshold(br);
544 bridge_configure_forward_bpdu(br);
545 bridge_configure_mac_idle_time(br);
546 bridge_configure_remotes(br, managers, n_managers);
547 bridge_configure_netflow(br);
548 bridge_configure_sflow(br, &sflow_bridge_number);
549 bridge_configure_stp(br);
550 bridge_configure_tables(br);
555 /* ovs-vswitchd has completed initialization, so allow the process that
556 * forked us to exit successfully. */
557 daemonize_complete();
558 reconfiguring = false;
564 /* Delete ofprotos which aren't configured or have the wrong type. Create
565 * ofprotos which don't exist but need to. */
567 bridge_update_ofprotos(void)
569 struct bridge *br, *next;
574 /* Delete ofprotos with no bridge or with the wrong type. */
577 ofproto_enumerate_types(&types);
578 SSET_FOR_EACH (type, &types) {
581 ofproto_enumerate_names(type, &names);
582 SSET_FOR_EACH (name, &names) {
583 br = bridge_lookup(name);
584 if (!br || strcmp(type, br->type)) {
585 ofproto_delete(name, type);
589 sset_destroy(&names);
590 sset_destroy(&types);
592 /* Add ofprotos for bridges which don't have one yet. */
593 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
601 /* Remove ports from any datapath with the same name as 'br'. If we
602 * don't do this, creating 'br''s ofproto will fail because a port with
603 * the same name as its local port already exists. */
604 HMAP_FOR_EACH (br2, node, &all_bridges) {
605 struct ofproto_port ofproto_port;
611 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
613 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
615 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
618 ofproto_port_destroy(&ofproto_port);
622 error = ofproto_create(br->name, br->type, &br->ofproto);
624 VLOG_ERR("failed to create bridge %s: %s", br->name,
632 port_configure(struct port *port)
634 const struct ovsrec_port *cfg = port->cfg;
635 struct bond_settings bond_settings;
636 struct lacp_settings lacp_settings;
637 struct ofproto_bundle_settings s;
640 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
641 configure_splinter_port(port);
650 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
651 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
652 s.slaves[s.n_slaves++] = iface->ofp_port;
657 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
661 /* Get VLAN trunks. */
664 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
668 if (cfg->vlan_mode) {
669 if (!strcmp(cfg->vlan_mode, "access")) {
670 s.vlan_mode = PORT_VLAN_ACCESS;
671 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
672 s.vlan_mode = PORT_VLAN_TRUNK;
673 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
674 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
675 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
676 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
678 /* This "can't happen" because ovsdb-server should prevent it. */
679 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
680 s.vlan_mode = PORT_VLAN_TRUNK;
684 s.vlan_mode = PORT_VLAN_ACCESS;
686 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
690 s.vlan_mode = PORT_VLAN_TRUNK;
693 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
696 /* Get LACP settings. */
697 s.lacp = port_configure_lacp(port, &lacp_settings);
701 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
702 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
703 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
706 s.lacp_slaves = NULL;
709 /* Get bond settings. */
710 if (s.n_slaves > 1) {
711 s.bond = &bond_settings;
712 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
713 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
716 s.bond_stable_ids = NULL;
718 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
719 netdev_set_miimon_interval(iface->netdev, 0);
724 ofproto_bundle_register(port->bridge->ofproto, port, &s);
730 free(s.bond_stable_ids);
733 /* Pick local port hardware address and datapath ID for 'br'. */
735 bridge_configure_datapath_id(struct bridge *br)
737 uint8_t ea[ETH_ADDR_LEN];
739 struct iface *local_iface;
740 struct iface *hw_addr_iface;
743 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
744 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
746 int error = netdev_set_etheraddr(local_iface->netdev, ea);
748 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
749 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
750 "Ethernet address: %s",
751 br->name, strerror(error));
754 memcpy(br->ea, ea, ETH_ADDR_LEN);
756 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
757 ofproto_set_datapath_id(br->ofproto, dpid);
759 dpid_string = xasprintf("%016"PRIx64, dpid);
760 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
764 /* Set NetFlow configuration on 'br'. */
766 bridge_configure_netflow(struct bridge *br)
768 struct ovsrec_netflow *cfg = br->cfg->netflow;
769 struct netflow_options opts;
772 ofproto_set_netflow(br->ofproto, NULL);
776 memset(&opts, 0, sizeof opts);
778 /* Get default NetFlow configuration from datapath.
779 * Apply overrides from 'cfg'. */
780 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
781 if (cfg->engine_type) {
782 opts.engine_type = *cfg->engine_type;
784 if (cfg->engine_id) {
785 opts.engine_id = *cfg->engine_id;
788 /* Configure active timeout interval. */
789 opts.active_timeout = cfg->active_timeout;
790 if (!opts.active_timeout) {
791 opts.active_timeout = -1;
792 } else if (opts.active_timeout < 0) {
793 VLOG_WARN("bridge %s: active timeout interval set to negative "
794 "value, using default instead (%d seconds)", br->name,
795 NF_ACTIVE_TIMEOUT_DEFAULT);
796 opts.active_timeout = -1;
799 /* Add engine ID to interface number to disambiguate bridgs? */
800 opts.add_id_to_iface = cfg->add_id_to_interface;
801 if (opts.add_id_to_iface) {
802 if (opts.engine_id > 0x7f) {
803 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
804 "another vswitch, choose an engine id less than 128",
807 if (hmap_count(&br->ports) > 508) {
808 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
809 "another port when more than 508 ports are used",
815 sset_init(&opts.collectors);
816 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
819 if (ofproto_set_netflow(br->ofproto, &opts)) {
820 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
822 sset_destroy(&opts.collectors);
825 /* Set sFlow configuration on 'br'. */
827 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
829 const struct ovsrec_sflow *cfg = br->cfg->sflow;
830 struct ovsrec_controller **controllers;
831 struct ofproto_sflow_options oso;
832 size_t n_controllers;
836 ofproto_set_sflow(br->ofproto, NULL);
840 memset(&oso, 0, sizeof oso);
842 sset_init(&oso.targets);
843 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
845 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
847 oso.sampling_rate = *cfg->sampling;
850 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
852 oso.polling_interval = *cfg->polling;
855 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
857 oso.header_len = *cfg->header;
860 oso.sub_id = (*sflow_bridge_number)++;
861 oso.agent_device = cfg->agent;
863 oso.control_ip = NULL;
864 n_controllers = bridge_get_controllers(br, &controllers);
865 for (i = 0; i < n_controllers; i++) {
866 if (controllers[i]->local_ip) {
867 oso.control_ip = controllers[i]->local_ip;
871 ofproto_set_sflow(br->ofproto, &oso);
873 sset_destroy(&oso.targets);
877 port_configure_stp(const struct ofproto *ofproto, struct port *port,
878 struct ofproto_port_stp_settings *port_s,
879 int *port_num_counter, unsigned long *port_num_bitmap)
881 const char *config_str;
884 if (smap_get_bool(&port->cfg->other_config, "stp-enable", false)) {
885 port_s->enable = false;
888 port_s->enable = true;
891 /* STP over bonds is not supported. */
892 if (!list_is_singleton(&port->ifaces)) {
893 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
895 port_s->enable = false;
899 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
901 /* Internal ports shouldn't participate in spanning tree, so
903 if (!strcmp(iface->type, "internal")) {
904 VLOG_DBG("port %s: disable STP on internal ports", port->name);
905 port_s->enable = false;
909 /* STP on mirror output ports is not supported. */
910 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
911 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
912 port_s->enable = false;
916 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
918 unsigned long int port_num = strtoul(config_str, NULL, 0);
919 int port_idx = port_num - 1;
921 if (port_num < 1 || port_num > STP_MAX_PORTS) {
922 VLOG_ERR("port %s: invalid stp-port-num", port->name);
923 port_s->enable = false;
927 if (bitmap_is_set(port_num_bitmap, port_idx)) {
928 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
929 port->name, port_num);
930 port_s->enable = false;
933 bitmap_set1(port_num_bitmap, port_idx);
934 port_s->port_num = port_idx;
936 if (*port_num_counter > STP_MAX_PORTS) {
937 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
938 port_s->enable = false;
942 port_s->port_num = (*port_num_counter)++;
945 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
947 port_s->path_cost = strtoul(config_str, NULL, 10);
949 enum netdev_features current;
951 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
952 /* Couldn't get speed, so assume 100Mb/s. */
953 port_s->path_cost = 19;
957 mbps = netdev_features_to_bps(current) / 1000000;
958 port_s->path_cost = stp_convert_speed_to_cost(mbps);
962 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
964 port_s->priority = strtoul(config_str, NULL, 0);
966 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
970 /* Set spanning tree configuration on 'br'. */
972 bridge_configure_stp(struct bridge *br)
974 if (!br->cfg->stp_enable) {
975 ofproto_set_stp(br->ofproto, NULL);
977 struct ofproto_stp_settings br_s;
978 const char *config_str;
980 int port_num_counter;
981 unsigned long *port_num_bitmap;
983 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
985 uint8_t ea[ETH_ADDR_LEN];
987 if (eth_addr_from_string(config_str, ea)) {
988 br_s.system_id = eth_addr_to_uint64(ea);
990 br_s.system_id = eth_addr_to_uint64(br->ea);
991 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
992 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
995 br_s.system_id = eth_addr_to_uint64(br->ea);
998 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1000 br_s.priority = strtoul(config_str, NULL, 0);
1002 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1005 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1007 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1009 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1012 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1014 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1016 br_s.max_age = STP_DEFAULT_MAX_AGE;
1019 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1021 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1023 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1026 /* Configure STP on the bridge. */
1027 if (ofproto_set_stp(br->ofproto, &br_s)) {
1028 VLOG_ERR("bridge %s: could not enable STP", br->name);
1032 /* Users must either set the port number with the "stp-port-num"
1033 * configuration on all ports or none. If manual configuration
1034 * is not done, then we allocate them sequentially. */
1035 port_num_counter = 0;
1036 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1037 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1038 struct ofproto_port_stp_settings port_s;
1039 struct iface *iface;
1041 port_configure_stp(br->ofproto, port, &port_s,
1042 &port_num_counter, port_num_bitmap);
1044 /* As bonds are not supported, just apply configuration to
1045 * all interfaces. */
1046 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1047 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1049 VLOG_ERR("port %s: could not enable STP", port->name);
1055 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1056 && port_num_counter) {
1057 VLOG_ERR("bridge %s: must manually configure all STP port "
1058 "IDs or none, disabling", br->name);
1059 ofproto_set_stp(br->ofproto, NULL);
1061 bitmap_free(port_num_bitmap);
1066 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1068 const struct port *port = port_lookup(br, name);
1069 return port && port_is_bond_fake_iface(port);
1073 port_is_bond_fake_iface(const struct port *port)
1075 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1079 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1081 struct bridge *br, *next;
1082 struct shash new_br;
1085 /* Collect new bridges' names and types. */
1086 shash_init(&new_br);
1087 for (i = 0; i < cfg->n_bridges; i++) {
1088 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1089 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1091 if (strchr(br_cfg->name, '/')) {
1092 /* Prevent remote ovsdb-server users from accessing arbitrary
1093 * directories, e.g. consider a bridge named "../../../etc/". */
1094 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1096 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1097 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1101 /* Get rid of deleted bridges or those whose types have changed.
1102 * Update 'cfg' of bridges that still exist. */
1103 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1104 br->cfg = shash_find_data(&new_br, br->name);
1105 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1106 br->cfg->datapath_type))) {
1111 /* Add new bridges. */
1112 for (i = 0; i < cfg->n_bridges; i++) {
1113 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1114 struct bridge *br = bridge_lookup(br_cfg->name);
1116 bridge_create(br_cfg);
1120 shash_destroy(&new_br);
1124 iface_set_ofp_port(struct iface *iface, int ofp_port)
1126 struct bridge *br = iface->port->bridge;
1128 assert(iface->ofp_port < 0 && ofp_port >= 0);
1129 iface->ofp_port = ofp_port;
1130 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1131 iface_set_ofport(iface->cfg, ofp_port);
1134 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1135 * Returns 0 if successful, otherwise a positive errno value. */
1137 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1138 struct netdev *netdev)
1142 error = netdev_set_config(netdev, &iface_cfg->options);
1144 VLOG_WARN("could not configure network device %s (%s)",
1145 iface_cfg->name, strerror(error));
1150 /* This function determines whether 'ofproto_port', which is attached to
1151 * br->ofproto's datapath, is one that we want in 'br'.
1153 * If it is, it returns true, after creating an iface (if necessary),
1154 * configuring the iface's netdev according to the iface's options, and setting
1155 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1157 * If, on the other hand, 'port' should be removed, it returns false. The
1158 * caller should later detach the port from br->ofproto. */
1160 bridge_refresh_one_ofp_port(struct bridge *br,
1161 const struct ofproto_port *ofproto_port)
1163 const char *name = ofproto_port->name;
1164 const char *type = ofproto_port->type;
1165 uint16_t ofp_port = ofproto_port->ofp_port;
1167 struct iface *iface = iface_lookup(br, name);
1169 /* Check that the name-to-number mapping is one-to-one. */
1170 if (iface->ofp_port >= 0) {
1171 VLOG_WARN("bridge %s: interface %s reported twice",
1174 } else if (iface_from_ofp_port(br, ofp_port)) {
1175 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1176 br->name, ofp_port);
1180 /* There's a configured interface named 'name'. */
1181 if (strcmp(type, iface->type)
1182 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1183 /* It's the wrong type, or it's the right type but can't be
1184 * configured as the user requested, so we must destroy it. */
1187 /* It's the right type and configured correctly. keep it. */
1188 iface_set_ofp_port(iface, ofp_port);
1191 } else if (bridge_has_bond_fake_iface(br, name)
1192 && !strcmp(type, "internal")) {
1193 /* It's a bond fake iface. Keep it. */
1196 /* There's no configured interface named 'name', but there might be an
1197 * interface of that name queued to be created.
1199 * If there is, and it has the correct type, then try to configure it
1200 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1201 * delete it and later try to re-add it. */
1202 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1204 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1205 && iface_create(br, if_cfg, ofp_port));
1209 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1210 * consistent with the ofp_ports in "br->ofproto". */
1212 bridge_refresh_ofp_port(struct bridge *br)
1214 struct ofproto_port_dump dump;
1215 struct ofproto_port ofproto_port;
1216 struct port *port, *port_next;
1218 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1219 hmap_clear(&br->ifaces);
1220 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1221 struct iface *iface;
1223 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1224 iface->ofp_port = -1;
1228 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1229 * already exist in the datapath and promote them to full fledged "struct
1230 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1231 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1232 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1233 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1234 garbage->ofp_port = ofproto_port.ofp_port;
1235 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1239 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1240 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1241 * later they can be added to ofproto. */
1242 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1243 struct iface *iface, *iface_next;
1245 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1246 if (iface->ofp_port < 0) {
1247 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1248 iface_destroy(iface);
1252 if (list_is_empty(&port->ifaces)) {
1258 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1259 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1260 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1263 * If successful, returns 0 and stores the network device in '*netdevp'. On
1264 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1266 iface_do_create(const struct bridge *br,
1267 const struct ovsrec_interface *iface_cfg,
1268 const struct ovsrec_port *port_cfg,
1269 int *ofp_portp, struct netdev **netdevp)
1271 struct netdev *netdev;
1274 error = netdev_open(iface_cfg->name,
1275 iface_get_type(iface_cfg, br->cfg), &netdev);
1277 VLOG_WARN("could not open network device %s (%s)",
1278 iface_cfg->name, strerror(error));
1282 error = iface_set_netdev_config(iface_cfg, netdev);
1287 if (*ofp_portp < 0) {
1290 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1294 *ofp_portp = ofp_port;
1296 VLOG_INFO("bridge %s: added interface %s on port %d",
1297 br->name, iface_cfg->name, *ofp_portp);
1299 VLOG_DBG("bridge %s: interface %s is on port %d",
1300 br->name, iface_cfg->name, *ofp_portp);
1303 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1304 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1312 netdev_close(netdev);
1316 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1317 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1318 * automatically allocated for the iface. Takes ownership of and
1319 * deallocates 'if_cfg'.
1321 * Return true if an iface is successfully created, false otherwise. */
1323 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1325 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1326 const struct ovsrec_port *port_cfg = if_cfg->parent;
1328 struct netdev *netdev;
1329 struct iface *iface;
1333 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1335 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1338 /* Do the bits that can fail up front.
1340 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1341 * internal datastructures may not be consistent. Eventually, when port
1342 * additions and deletions are cheaper, these calls should be removed. */
1344 assert(!iface_lookup(br, iface_cfg->name));
1345 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1348 iface_clear_db_record(iface_cfg);
1352 /* Get or create the port structure. */
1353 port = port_lookup(br, port_cfg->name);
1355 port = port_create(br, port_cfg);
1358 /* Create the iface structure. */
1359 iface = xzalloc(sizeof *iface);
1360 list_push_back(&port->ifaces, &iface->port_elem);
1361 hmap_insert(&br->iface_by_name, &iface->name_node,
1362 hash_string(iface_cfg->name, 0));
1364 iface->name = xstrdup(iface_cfg->name);
1365 iface->ofp_port = -1;
1366 iface->netdev = netdev;
1367 iface->type = iface_get_type(iface_cfg, br->cfg);
1368 iface->cfg = iface_cfg;
1370 iface_set_ofp_port(iface, ofp_port);
1372 /* Populate initial status in database. */
1373 iface_refresh_stats(iface);
1374 iface_refresh_status(iface);
1376 /* Add bond fake iface if necessary. */
1377 if (port_is_bond_fake_iface(port)) {
1378 struct ofproto_port ofproto_port;
1380 if (ofproto_port_query_by_name(br->ofproto, port->name,
1382 struct netdev *netdev;
1385 error = netdev_open(port->name, "internal", &netdev);
1387 ofproto_port_add(br->ofproto, netdev, NULL);
1388 netdev_close(netdev);
1390 VLOG_WARN("could not open network device %s (%s)",
1391 port->name, strerror(error));
1394 /* Already exists, nothing to do. */
1395 ofproto_port_destroy(&ofproto_port);
1402 /* Set Flow eviction threshold */
1404 bridge_configure_flow_eviction_threshold(struct bridge *br)
1406 const char *threshold_str;
1409 threshold_str = smap_get(&br->cfg->other_config,
1410 "flow-eviction-threshold");
1411 if (threshold_str) {
1412 threshold = strtoul(threshold_str, NULL, 10);
1414 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1416 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1419 /* Set forward BPDU option. */
1421 bridge_configure_forward_bpdu(struct bridge *br)
1423 ofproto_set_forward_bpdu(br->ofproto,
1424 smap_get_bool(&br->cfg->other_config,
1429 /* Set MAC aging time for 'br'. */
1431 bridge_configure_mac_idle_time(struct bridge *br)
1433 const char *idle_time_str;
1436 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1437 idle_time = (idle_time_str && atoi(idle_time_str)
1438 ? atoi(idle_time_str)
1439 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1440 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1444 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1445 struct iface **hw_addr_iface)
1447 struct hmapx mirror_output_ports;
1450 bool found_addr = false;
1454 *hw_addr_iface = NULL;
1456 /* Did the user request a particular MAC? */
1457 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1458 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1459 if (eth_addr_is_multicast(ea)) {
1460 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1461 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1462 } else if (eth_addr_is_zero(ea)) {
1463 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1469 /* Mirror output ports don't participate in picking the local hardware
1470 * address. ofproto can't help us find out whether a given port is a
1471 * mirror output because we haven't configured mirrors yet, so we need to
1472 * accumulate them ourselves. */
1473 hmapx_init(&mirror_output_ports);
1474 for (i = 0; i < br->cfg->n_mirrors; i++) {
1475 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1476 if (m->output_port) {
1477 hmapx_add(&mirror_output_ports, m->output_port);
1481 /* Otherwise choose the minimum non-local MAC address among all of the
1483 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1484 uint8_t iface_ea[ETH_ADDR_LEN];
1485 struct iface *candidate;
1486 struct iface *iface;
1488 /* Mirror output ports don't participate. */
1489 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1493 /* Choose the MAC address to represent the port. */
1495 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1496 /* Find the interface with this Ethernet address (if any) so that
1497 * we can provide the correct devname to the caller. */
1498 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1499 uint8_t candidate_ea[ETH_ADDR_LEN];
1500 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1501 && eth_addr_equals(iface_ea, candidate_ea)) {
1506 /* Choose the interface whose MAC address will represent the port.
1507 * The Linux kernel bonding code always chooses the MAC address of
1508 * the first slave added to a bond, and the Fedora networking
1509 * scripts always add slaves to a bond in alphabetical order, so
1510 * for compatibility we choose the interface with the name that is
1511 * first in alphabetical order. */
1512 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1513 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1518 /* The local port doesn't count (since we're trying to choose its
1519 * MAC address anyway). */
1520 if (iface->ofp_port == OFPP_LOCAL) {
1525 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1531 /* Compare against our current choice. */
1532 if (!eth_addr_is_multicast(iface_ea) &&
1533 !eth_addr_is_local(iface_ea) &&
1534 !eth_addr_is_reserved(iface_ea) &&
1535 !eth_addr_is_zero(iface_ea) &&
1536 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1538 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1539 *hw_addr_iface = iface;
1544 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1545 br->name, ETH_ADDR_ARGS(ea));
1547 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1548 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1549 *hw_addr_iface = NULL;
1550 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1551 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1554 hmapx_destroy(&mirror_output_ports);
1557 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1558 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1559 * an interface on 'br', then that interface must be passed in as
1560 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1561 * 'hw_addr_iface' must be passed in as a null pointer. */
1563 bridge_pick_datapath_id(struct bridge *br,
1564 const uint8_t bridge_ea[ETH_ADDR_LEN],
1565 struct iface *hw_addr_iface)
1568 * The procedure for choosing a bridge MAC address will, in the most
1569 * ordinary case, also choose a unique MAC that we can use as a datapath
1570 * ID. In some special cases, though, multiple bridges will end up with
1571 * the same MAC address. This is OK for the bridges, but it will confuse
1572 * the OpenFlow controller, because each datapath needs a unique datapath
1575 * Datapath IDs must be unique. It is also very desirable that they be
1576 * stable from one run to the next, so that policy set on a datapath
1579 const char *datapath_id;
1582 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1583 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1587 if (!hw_addr_iface) {
1589 * A purely internal bridge, that is, one that has no non-virtual
1590 * network devices on it at all, is difficult because it has no
1591 * natural unique identifier at all.
1593 * When the host is a XenServer, we handle this case by hashing the
1594 * host's UUID with the name of the bridge. Names of bridges are
1595 * persistent across XenServer reboots, although they can be reused if
1596 * an internal network is destroyed and then a new one is later
1597 * created, so this is fairly effective.
1599 * When the host is not a XenServer, we punt by using a random MAC
1600 * address on each run.
1602 const char *host_uuid = xenserver_get_host_uuid();
1604 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1605 dpid = dpid_from_hash(combined, strlen(combined));
1611 return eth_addr_to_uint64(bridge_ea);
1615 dpid_from_hash(const void *data, size_t n)
1617 uint8_t hash[SHA1_DIGEST_SIZE];
1619 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1620 sha1_bytes(data, n, hash);
1621 eth_addr_mark_random(hash);
1622 return eth_addr_to_uint64(hash);
1626 iface_refresh_status(struct iface *iface)
1630 enum netdev_features current;
1631 enum netdev_flags flags;
1637 if (iface_is_synthetic(iface)) {
1643 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1644 ovsrec_interface_set_status(iface->cfg, &smap);
1646 ovsrec_interface_set_status(iface->cfg, NULL);
1649 smap_destroy(&smap);
1651 error = netdev_get_flags(iface->netdev, &flags);
1653 ovsrec_interface_set_admin_state(iface->cfg,
1654 flags & NETDEV_UP ? "up" : "down");
1657 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1660 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1662 ovsrec_interface_set_duplex(iface->cfg,
1663 netdev_features_is_full_duplex(current)
1665 /* warning: uint64_t -> int64_t conversion */
1666 bps = netdev_features_to_bps(current);
1667 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1670 ovsrec_interface_set_duplex(iface->cfg, NULL);
1671 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1674 error = netdev_get_mtu(iface->netdev, &mtu);
1677 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1680 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1684 /* Writes 'iface''s CFM statistics to the database. */
1686 iface_refresh_cfm_stats(struct iface *iface)
1688 const struct ovsrec_interface *cfg = iface->cfg;
1690 const uint64_t *rmps;
1694 if (iface_is_synthetic(iface)) {
1698 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1701 const char *reasons[CFM_FAULT_N_REASONS];
1702 bool fault_bool = fault;
1706 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1707 int reason = 1 << i;
1708 if (fault & reason) {
1709 reasons[j++] = cfm_fault_reason_to_str(reason);
1713 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1714 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1716 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1717 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1720 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1721 iface->ofp_port, &rmps, &n_rmps);
1723 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1726 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1729 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1732 int64_t cfm_health = health;
1733 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1735 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1740 iface_refresh_stats(struct iface *iface)
1742 #define IFACE_STATS \
1743 IFACE_STAT(rx_packets, "rx_packets") \
1744 IFACE_STAT(tx_packets, "tx_packets") \
1745 IFACE_STAT(rx_bytes, "rx_bytes") \
1746 IFACE_STAT(tx_bytes, "tx_bytes") \
1747 IFACE_STAT(rx_dropped, "rx_dropped") \
1748 IFACE_STAT(tx_dropped, "tx_dropped") \
1749 IFACE_STAT(rx_errors, "rx_errors") \
1750 IFACE_STAT(tx_errors, "tx_errors") \
1751 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1752 IFACE_STAT(rx_over_errors, "rx_over_err") \
1753 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1754 IFACE_STAT(collisions, "collisions")
1756 #define IFACE_STAT(MEMBER, NAME) NAME,
1757 static char *keys[] = { IFACE_STATS };
1759 int64_t values[ARRAY_SIZE(keys)];
1762 struct netdev_stats stats;
1764 if (iface_is_synthetic(iface)) {
1768 /* Intentionally ignore return value, since errors will set 'stats' to
1769 * all-1s, and we will deal with that correctly below. */
1770 netdev_get_stats(iface->netdev, &stats);
1772 /* Copy statistics into values[] array. */
1774 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1777 assert(i == ARRAY_SIZE(keys));
1779 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1785 br_refresh_stp_status(struct bridge *br)
1787 struct smap smap = SMAP_INITIALIZER(&smap);
1788 struct ofproto *ofproto = br->ofproto;
1789 struct ofproto_stp_status status;
1791 if (ofproto_get_stp_status(ofproto, &status)) {
1795 if (!status.enabled) {
1796 ovsrec_bridge_set_status(br->cfg, NULL);
1800 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1801 STP_ID_ARGS(status.bridge_id));
1802 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1803 STP_ID_ARGS(status.designated_root));
1804 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1806 ovsrec_bridge_set_status(br->cfg, &smap);
1807 smap_destroy(&smap);
1811 port_refresh_stp_status(struct port *port)
1813 struct ofproto *ofproto = port->bridge->ofproto;
1814 struct iface *iface;
1815 struct ofproto_port_stp_status status;
1817 int64_t int_values[3];
1820 if (port_is_synthetic(port)) {
1824 /* STP doesn't currently support bonds. */
1825 if (!list_is_singleton(&port->ifaces)) {
1826 ovsrec_port_set_status(port->cfg, NULL);
1830 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1832 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1836 if (!status.enabled) {
1837 ovsrec_port_set_status(port->cfg, NULL);
1838 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1842 /* Set Status column. */
1844 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1845 smap_add(&smap, "stp_state", stp_state_name(status.state));
1846 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1847 smap_add(&smap, "stp_role", stp_role_name(status.role));
1848 ovsrec_port_set_status(port->cfg, &smap);
1849 smap_destroy(&smap);
1851 /* Set Statistics column. */
1852 keys[0] = "stp_tx_count";
1853 int_values[0] = status.tx_count;
1854 keys[1] = "stp_rx_count";
1855 int_values[1] = status.rx_count;
1856 keys[2] = "stp_error_count";
1857 int_values[2] = status.error_count;
1859 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1860 ARRAY_SIZE(int_values));
1864 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1866 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1870 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1872 struct ovsdb_datum datum;
1876 if (enable_system_stats(cfg)) {
1877 get_system_stats(&stats);
1880 ovsdb_datum_from_shash(&datum, &stats);
1881 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1885 static inline const char *
1886 nx_role_to_str(enum nx_role role)
1891 case NX_ROLE_MASTER:
1896 return "*** INVALID ROLE ***";
1901 refresh_controller_status(void)
1905 const struct ovsrec_controller *cfg;
1909 /* Accumulate status for controllers on all bridges. */
1910 HMAP_FOR_EACH (br, node, &all_bridges) {
1911 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1914 /* Update each controller in the database with current status. */
1915 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1916 struct ofproto_controller_info *cinfo =
1917 shash_find_data(&info, cfg->target);
1920 struct smap smap = SMAP_INITIALIZER(&smap);
1921 const char **values = cinfo->pairs.values;
1922 const char **keys = cinfo->pairs.keys;
1925 for (i = 0; i < cinfo->pairs.n; i++) {
1926 smap_add(&smap, keys[i], values[i]);
1929 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1930 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1931 ovsrec_controller_set_status(cfg, &smap);
1932 smap_destroy(&smap);
1934 ovsrec_controller_set_is_connected(cfg, false);
1935 ovsrec_controller_set_role(cfg, NULL);
1936 ovsrec_controller_set_status(cfg, NULL);
1940 ofproto_free_ofproto_controller_info(&info);
1944 refresh_cfm_stats(void)
1946 static struct ovsdb_idl_txn *txn = NULL;
1951 txn = ovsdb_idl_txn_create(idl);
1953 HMAP_FOR_EACH (br, node, &all_bridges) {
1954 struct iface *iface;
1956 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1957 iface_refresh_cfm_stats(iface);
1962 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1963 ovsdb_idl_txn_destroy(txn);
1968 /* Performs periodic activity required by bridges that needs to be done with
1969 * the least possible latency.
1971 * It makes sense to call this function a couple of times per poll loop, to
1972 * provide a significant performance boost on some benchmarks with ofprotos
1973 * that use the ofproto-dpif implementation. */
1975 bridge_run_fast(void)
1979 HMAP_FOR_EACH (br, node, &all_bridges) {
1980 ofproto_run_fast(br->ofproto);
1987 static const struct ovsrec_open_vswitch null_cfg;
1988 const struct ovsrec_open_vswitch *cfg;
1989 struct ovsdb_idl_txn *reconf_txn = NULL;
1991 bool vlan_splinters_changed;
1994 /* (Re)configure if necessary. */
1995 if (!reconfiguring) {
1998 if (ovsdb_idl_is_lock_contended(idl)) {
1999 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2000 struct bridge *br, *next_br;
2002 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2003 "disabling this process until it goes away");
2005 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2009 } else if (!ovsdb_idl_has_lock(idl)) {
2013 cfg = ovsrec_open_vswitch_first(idl);
2015 /* Let each bridge do the work that it needs to do. */
2016 HMAP_FOR_EACH (br, node, &all_bridges) {
2017 ofproto_run(br->ofproto);
2020 /* Re-configure SSL. We do this on every trip through the main loop,
2021 * instead of just when the database changes, because the contents of the
2022 * key and certificate files can change without the database changing.
2024 * We do this before bridge_reconfigure() because that function might
2025 * initiate SSL connections and thus requires SSL to be configured. */
2026 if (cfg && cfg->ssl) {
2027 const struct ovsrec_ssl *ssl = cfg->ssl;
2029 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2030 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2033 if (!reconfiguring) {
2034 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2035 * usage has changed. */
2036 vlan_splinters_changed = false;
2037 if (vlan_splinters_enabled_anywhere) {
2038 HMAP_FOR_EACH (br, node, &all_bridges) {
2039 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2040 vlan_splinters_changed = true;
2046 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2047 idl_seqno = ovsdb_idl_get_seqno(idl);
2049 reconf_txn = ovsdb_idl_txn_create(idl);
2050 bridge_reconfigure(cfg);
2052 /* We still need to reconfigure to avoid dangling pointers to
2053 * now-destroyed ovsrec structures inside bridge data. */
2054 bridge_reconfigure(&null_cfg);
2059 if (reconfiguring) {
2062 reconf_txn = ovsdb_idl_txn_create(idl);
2064 if (bridge_reconfigure_continue(cfg)) {
2065 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2068 bridge_reconfigure_continue(&null_cfg);
2073 ovsdb_idl_txn_commit(reconf_txn);
2074 ovsdb_idl_txn_destroy(reconf_txn);
2078 /* Refresh system and interface stats if necessary. */
2079 if (time_msec() >= stats_timer) {
2081 struct ovsdb_idl_txn *txn;
2083 txn = ovsdb_idl_txn_create(idl);
2084 HMAP_FOR_EACH (br, node, &all_bridges) {
2088 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2089 struct iface *iface;
2091 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2092 iface_refresh_stats(iface);
2093 iface_refresh_status(iface);
2097 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2098 mirror_refresh_stats(m);
2102 refresh_system_stats(cfg);
2103 refresh_controller_status();
2104 ovsdb_idl_txn_commit(txn);
2105 ovsdb_idl_txn_destroy(txn); /* XXX */
2108 stats_timer = time_msec() + STATS_INTERVAL;
2111 if (time_msec() >= db_limiter) {
2112 struct ovsdb_idl_txn *txn;
2114 txn = ovsdb_idl_txn_create(idl);
2115 HMAP_FOR_EACH (br, node, &all_bridges) {
2116 struct iface *iface;
2119 br_refresh_stp_status(br);
2121 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2122 port_refresh_stp_status(port);
2125 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2126 const char *link_state;
2127 int64_t link_resets;
2130 if (iface_is_synthetic(iface)) {
2134 current = ofproto_port_is_lacp_current(br->ofproto,
2138 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2140 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2143 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2144 ovsrec_interface_set_link_state(iface->cfg, link_state);
2146 link_resets = netdev_get_carrier_resets(iface->netdev);
2147 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2151 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2152 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2154 ovsdb_idl_txn_destroy(txn);
2157 refresh_cfm_stats();
2163 ovsdb_idl_wait(idl);
2165 if (reconfiguring) {
2166 poll_immediate_wake();
2169 if (!hmap_is_empty(&all_bridges)) {
2172 HMAP_FOR_EACH (br, node, &all_bridges) {
2173 ofproto_wait(br->ofproto);
2175 poll_timer_wait_until(stats_timer);
2177 if (db_limiter > time_msec()) {
2178 poll_timer_wait_until(db_limiter);
2183 /* Adds some memory usage statistics for bridges into 'usage', for use with
2184 * memory_report(). */
2186 bridge_get_memory_usage(struct simap *usage)
2190 HMAP_FOR_EACH (br, node, &all_bridges) {
2191 ofproto_get_memory_usage(br->ofproto, usage);
2195 /* QoS unixctl user interface functions. */
2197 struct qos_unixctl_show_cbdata {
2199 struct iface *iface;
2203 qos_unixctl_show_cb(unsigned int queue_id,
2204 const struct smap *details,
2207 struct qos_unixctl_show_cbdata *data = aux;
2208 struct ds *ds = data->ds;
2209 struct iface *iface = data->iface;
2210 struct netdev_queue_stats stats;
2211 struct smap_node *node;
2214 ds_put_cstr(ds, "\n");
2216 ds_put_format(ds, "Queue %u:\n", queue_id);
2218 ds_put_cstr(ds, "Default:\n");
2221 SMAP_FOR_EACH (node, details) {
2222 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2225 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2227 if (stats.tx_packets != UINT64_MAX) {
2228 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2231 if (stats.tx_bytes != UINT64_MAX) {
2232 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2235 if (stats.tx_errors != UINT64_MAX) {
2236 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2239 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2240 queue_id, strerror(error));
2245 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2246 const char *argv[], void *aux OVS_UNUSED)
2248 struct ds ds = DS_EMPTY_INITIALIZER;
2249 struct smap smap = SMAP_INITIALIZER(&smap);
2250 struct iface *iface;
2252 struct smap_node *node;
2253 struct qos_unixctl_show_cbdata data;
2256 iface = iface_find(argv[1]);
2258 unixctl_command_reply_error(conn, "no such interface");
2262 netdev_get_qos(iface->netdev, &type, &smap);
2264 if (*type != '\0') {
2265 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2267 SMAP_FOR_EACH (node, &smap) {
2268 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2273 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2276 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2278 unixctl_command_reply(conn, ds_cstr(&ds));
2280 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2281 unixctl_command_reply_error(conn, ds_cstr(&ds));
2284 smap_destroy(&smap);
2288 /* Bridge reconfiguration functions. */
2290 bridge_create(const struct ovsrec_bridge *br_cfg)
2294 assert(!bridge_lookup(br_cfg->name));
2295 br = xzalloc(sizeof *br);
2297 br->name = xstrdup(br_cfg->name);
2298 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2301 /* Derive the default Ethernet address from the bridge's UUID. This should
2302 * be unique and it will be stable between ovs-vswitchd runs. */
2303 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2304 eth_addr_mark_random(br->default_ea);
2306 hmap_init(&br->ports);
2307 hmap_init(&br->ifaces);
2308 hmap_init(&br->iface_by_name);
2309 hmap_init(&br->mirrors);
2311 hmap_init(&br->if_cfg_todo);
2312 list_init(&br->ofpp_garbage);
2314 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2318 bridge_destroy(struct bridge *br)
2321 struct mirror *mirror, *next_mirror;
2322 struct port *port, *next_port;
2323 struct if_cfg *if_cfg, *next_if_cfg;
2324 struct ofpp_garbage *garbage, *next_garbage;
2326 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2329 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2330 mirror_destroy(mirror);
2332 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2333 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2336 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2337 &br->ofpp_garbage) {
2338 list_remove(&garbage->list_node);
2342 hmap_remove(&all_bridges, &br->node);
2343 ofproto_destroy(br->ofproto);
2344 hmap_destroy(&br->ifaces);
2345 hmap_destroy(&br->ports);
2346 hmap_destroy(&br->iface_by_name);
2347 hmap_destroy(&br->mirrors);
2348 hmap_destroy(&br->if_cfg_todo);
2355 static struct bridge *
2356 bridge_lookup(const char *name)
2360 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2361 if (!strcmp(br->name, name)) {
2368 /* Handle requests for a listing of all flows known by the OpenFlow
2369 * stack, including those normally hidden. */
2371 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2372 const char *argv[], void *aux OVS_UNUSED)
2377 br = bridge_lookup(argv[1]);
2379 unixctl_command_reply_error(conn, "Unknown bridge");
2384 ofproto_get_all_flows(br->ofproto, &results);
2386 unixctl_command_reply(conn, ds_cstr(&results));
2387 ds_destroy(&results);
2390 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2391 * connections and reconnect. If BRIDGE is not specified, then all bridges
2392 * drop their controller connections and reconnect. */
2394 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2395 const char *argv[], void *aux OVS_UNUSED)
2399 br = bridge_lookup(argv[1]);
2401 unixctl_command_reply_error(conn, "Unknown bridge");
2404 ofproto_reconnect_controllers(br->ofproto);
2406 HMAP_FOR_EACH (br, node, &all_bridges) {
2407 ofproto_reconnect_controllers(br->ofproto);
2410 unixctl_command_reply(conn, NULL);
2414 bridge_get_controllers(const struct bridge *br,
2415 struct ovsrec_controller ***controllersp)
2417 struct ovsrec_controller **controllers;
2418 size_t n_controllers;
2420 controllers = br->cfg->controller;
2421 n_controllers = br->cfg->n_controller;
2423 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2429 *controllersp = controllers;
2431 return n_controllers;
2435 bridge_queue_if_cfg(struct bridge *br,
2436 const struct ovsrec_interface *cfg,
2437 const struct ovsrec_port *parent)
2439 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2442 if_cfg->parent = parent;
2443 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2444 hash_string(if_cfg->cfg->name, 0));
2447 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2448 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2449 * 'br' needs to complete its configuration. */
2451 bridge_add_del_ports(struct bridge *br,
2452 const unsigned long int *splinter_vlans)
2454 struct shash_node *port_node;
2455 struct port *port, *next;
2456 struct shash new_ports;
2459 assert(hmap_is_empty(&br->if_cfg_todo));
2461 /* Collect new ports. */
2462 shash_init(&new_ports);
2463 for (i = 0; i < br->cfg->n_ports; i++) {
2464 const char *name = br->cfg->ports[i]->name;
2465 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2466 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2470 if (bridge_get_controllers(br, NULL)
2471 && !shash_find(&new_ports, br->name)) {
2472 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2473 br->name, br->name);
2475 ovsrec_interface_init(&br->synth_local_iface);
2476 ovsrec_port_init(&br->synth_local_port);
2478 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2479 br->synth_local_port.n_interfaces = 1;
2480 br->synth_local_port.name = br->name;
2482 br->synth_local_iface.name = br->name;
2483 br->synth_local_iface.type = "internal";
2485 br->synth_local_ifacep = &br->synth_local_iface;
2487 shash_add(&new_ports, br->name, &br->synth_local_port);
2490 if (splinter_vlans) {
2491 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2494 /* Get rid of deleted ports.
2495 * Get rid of deleted interfaces on ports that still exist. */
2496 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2497 port->cfg = shash_find_data(&new_ports, port->name);
2501 port_del_ifaces(port);
2505 /* Update iface->cfg and iface->type in interfaces that still exist.
2506 * Add new interfaces to creation queue. */
2507 SHASH_FOR_EACH (port_node, &new_ports) {
2508 const struct ovsrec_port *port = port_node->data;
2511 for (i = 0; i < port->n_interfaces; i++) {
2512 const struct ovsrec_interface *cfg = port->interfaces[i];
2513 struct iface *iface = iface_lookup(br, cfg->name);
2514 const char *type = iface_get_type(cfg, br->cfg);
2519 } else if (strcmp(type, "null")) {
2520 bridge_queue_if_cfg(br, cfg, port);
2525 shash_destroy(&new_ports);
2528 /* Initializes 'oc' appropriately as a management service controller for
2531 * The caller must free oc->target when it is no longer needed. */
2533 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2534 struct ofproto_controller *oc)
2536 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2537 oc->max_backoff = 0;
2538 oc->probe_interval = 60;
2539 oc->band = OFPROTO_OUT_OF_BAND;
2541 oc->burst_limit = 0;
2542 oc->enable_async_msgs = true;
2545 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2547 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2548 struct ofproto_controller *oc)
2552 oc->target = c->target;
2553 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2554 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2555 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2556 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2557 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2558 oc->burst_limit = (c->controller_burst_limit
2559 ? *c->controller_burst_limit : 0);
2560 oc->enable_async_msgs = (!c->enable_async_messages
2561 || *c->enable_async_messages);
2562 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2563 if (dscp < 0 || dscp > 63) {
2564 dscp = DSCP_DEFAULT;
2569 /* Configures the IP stack for 'br''s local interface properly according to the
2570 * configuration in 'c'. */
2572 bridge_configure_local_iface_netdev(struct bridge *br,
2573 struct ovsrec_controller *c)
2575 struct netdev *netdev;
2576 struct in_addr mask, gateway;
2578 struct iface *local_iface;
2581 /* If there's no local interface or no IP address, give up. */
2582 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2583 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2587 /* Bring up the local interface. */
2588 netdev = local_iface->netdev;
2589 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2591 /* Configure the IP address and netmask. */
2592 if (!c->local_netmask
2593 || !inet_aton(c->local_netmask, &mask)
2595 mask.s_addr = guess_netmask(ip.s_addr);
2597 if (!netdev_set_in4(netdev, ip, mask)) {
2598 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2599 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2602 /* Configure the default gateway. */
2603 if (c->local_gateway
2604 && inet_aton(c->local_gateway, &gateway)
2605 && gateway.s_addr) {
2606 if (!netdev_add_router(netdev, gateway)) {
2607 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2608 br->name, IP_ARGS(&gateway.s_addr));
2613 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2614 * in either string are treated as equal to any number of slashes in the other,
2615 * e.g. "x///y" is equal to "x/y". */
2617 equal_pathnames(const char *a, const char *b)
2621 a += strspn(a, "/");
2622 b += strspn(b, "/");
2623 } else if (*a == '\0') {
2634 bridge_configure_remotes(struct bridge *br,
2635 const struct sockaddr_in *managers, size_t n_managers)
2637 bool disable_in_band;
2639 struct ovsrec_controller **controllers;
2640 size_t n_controllers;
2642 enum ofproto_fail_mode fail_mode;
2644 struct ofproto_controller *ocs;
2648 /* Check if we should disable in-band control on this bridge. */
2649 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2652 /* Set OpenFlow queue ID for in-band control. */
2653 ofproto_set_in_band_queue(br->ofproto,
2654 smap_get_int(&br->cfg->other_config,
2655 "in-band-queue", -1));
2657 if (disable_in_band) {
2658 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2660 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2663 n_controllers = bridge_get_controllers(br, &controllers);
2665 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2668 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2669 for (i = 0; i < n_controllers; i++) {
2670 struct ovsrec_controller *c = controllers[i];
2672 if (!strncmp(c->target, "punix:", 6)
2673 || !strncmp(c->target, "unix:", 5)) {
2674 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2677 whitelist = xasprintf("unix:%s/%s.controller",
2678 ovs_rundir(), br->name);
2679 if (!equal_pathnames(c->target, whitelist)) {
2680 /* Prevent remote ovsdb-server users from accessing arbitrary
2681 * Unix domain sockets and overwriting arbitrary local
2683 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2684 "controller \"%s\" due to possibility for remote "
2685 "exploit. Instead, specify whitelisted \"%s\" or "
2686 "connect to \"unix:%s/%s.mgmt\" (which is always "
2687 "available without special configuration).",
2688 br->name, c->target, whitelist,
2689 ovs_rundir(), br->name);
2697 bridge_configure_local_iface_netdev(br, c);
2698 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2699 if (disable_in_band) {
2700 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2705 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2706 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2709 /* Set the fail-mode. */
2710 fail_mode = !br->cfg->fail_mode
2711 || !strcmp(br->cfg->fail_mode, "standalone")
2712 ? OFPROTO_FAIL_STANDALONE
2713 : OFPROTO_FAIL_SECURE;
2714 ofproto_set_fail_mode(br->ofproto, fail_mode);
2716 /* Configure OpenFlow controller connection snooping. */
2717 if (!ofproto_has_snoops(br->ofproto)) {
2721 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2722 ovs_rundir(), br->name));
2723 ofproto_set_snoops(br->ofproto, &snoops);
2724 sset_destroy(&snoops);
2729 bridge_configure_tables(struct bridge *br)
2731 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2735 n_tables = ofproto_get_n_tables(br->ofproto);
2737 for (i = 0; i < n_tables; i++) {
2738 struct ofproto_table_settings s;
2741 s.max_flows = UINT_MAX;
2745 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2746 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2749 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2750 s.max_flows = *cfg->flow_limit;
2752 if (cfg->overflow_policy
2753 && !strcmp(cfg->overflow_policy, "evict")) {
2756 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2757 for (k = 0; k < cfg->n_groups; k++) {
2758 const char *string = cfg->groups[k];
2761 msg = mf_parse_subfield__(&s.groups[k], &string);
2763 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2764 "'groups' (%s)", br->name, i, msg);
2766 } else if (*string) {
2767 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2768 "element '%s' contains trailing garbage",
2769 br->name, i, cfg->groups[k]);
2777 ofproto_configure_table(br->ofproto, i, &s);
2781 for (; j < br->cfg->n_flow_tables; j++) {
2782 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2783 "%"PRId64" not supported by this datapath", br->name,
2784 br->cfg->key_flow_tables[j]);
2788 /* Port functions. */
2790 static struct port *
2791 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2795 port = xzalloc(sizeof *port);
2797 port->name = xstrdup(cfg->name);
2799 list_init(&port->ifaces);
2801 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2805 /* Deletes interfaces from 'port' that are no longer configured for it. */
2807 port_del_ifaces(struct port *port)
2809 struct iface *iface, *next;
2810 struct sset new_ifaces;
2813 /* Collect list of new interfaces. */
2814 sset_init(&new_ifaces);
2815 for (i = 0; i < port->cfg->n_interfaces; i++) {
2816 const char *name = port->cfg->interfaces[i]->name;
2817 const char *type = port->cfg->interfaces[i]->type;
2818 if (strcmp(type, "null")) {
2819 sset_add(&new_ifaces, name);
2823 /* Get rid of deleted interfaces. */
2824 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2825 if (!sset_contains(&new_ifaces, iface->name)) {
2826 iface_destroy(iface);
2830 sset_destroy(&new_ifaces);
2834 port_destroy(struct port *port)
2837 struct bridge *br = port->bridge;
2838 struct iface *iface, *next;
2841 ofproto_bundle_unregister(br->ofproto, port);
2844 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2845 iface_destroy(iface);
2848 hmap_remove(&br->ports, &port->hmap_node);
2854 static struct port *
2855 port_lookup(const struct bridge *br, const char *name)
2859 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2861 if (!strcmp(port->name, name)) {
2869 enable_lacp(struct port *port, bool *activep)
2871 if (!port->cfg->lacp) {
2872 /* XXX when LACP implementation has been sufficiently tested, enable by
2873 * default and make active on bonded ports. */
2875 } else if (!strcmp(port->cfg->lacp, "off")) {
2877 } else if (!strcmp(port->cfg->lacp, "active")) {
2880 } else if (!strcmp(port->cfg->lacp, "passive")) {
2884 VLOG_WARN("port %s: unknown LACP mode %s",
2885 port->name, port->cfg->lacp);
2890 static struct lacp_settings *
2891 port_configure_lacp(struct port *port, struct lacp_settings *s)
2893 const char *lacp_time, *system_id;
2896 if (!enable_lacp(port, &s->active)) {
2900 s->name = port->name;
2902 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2904 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2905 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2906 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2907 " address.", port->name, system_id);
2911 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2914 if (eth_addr_is_zero(s->id)) {
2915 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2919 /* Prefer bondable links if unspecified. */
2920 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2922 s->priority = (priority > 0 && priority <= UINT16_MAX
2924 : UINT16_MAX - !list_is_short(&port->ifaces));
2926 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2927 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2932 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2934 int priority, portid, key;
2936 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2937 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2939 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2941 if (portid <= 0 || portid > UINT16_MAX) {
2942 portid = iface->ofp_port;
2945 if (priority <= 0 || priority > UINT16_MAX) {
2946 priority = UINT16_MAX;
2949 if (key < 0 || key > UINT16_MAX) {
2953 s->name = iface->name;
2955 s->priority = priority;
2960 port_configure_bond(struct port *port, struct bond_settings *s,
2961 uint32_t *bond_stable_ids)
2963 const char *detect_s;
2964 struct iface *iface;
2965 int miimon_interval;
2968 s->name = port->name;
2970 if (port->cfg->bond_mode) {
2971 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2972 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2973 port->name, port->cfg->bond_mode,
2974 bond_mode_to_string(s->balance));
2977 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2979 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2980 * active-backup. At some point we should remove this warning. */
2981 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2982 " in previous versions, the default bond_mode was"
2983 " balance-slb", port->name,
2984 bond_mode_to_string(s->balance));
2986 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2987 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2988 "please use another bond type or disable flood_vlans",
2992 miimon_interval = smap_get_int(&port->cfg->other_config,
2993 "bond-miimon-interval", 0);
2994 if (miimon_interval <= 0) {
2995 miimon_interval = 200;
2998 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
2999 if (!detect_s || !strcmp(detect_s, "carrier")) {
3000 miimon_interval = 0;
3001 } else if (strcmp(detect_s, "miimon")) {
3002 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3003 "defaulting to carrier", port->name, detect_s);
3004 miimon_interval = 0;
3007 s->up_delay = MAX(0, port->cfg->bond_updelay);
3008 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3009 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3010 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3011 "bond-rebalance-interval", 10000);
3012 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3013 s->rebalance_interval = 1000;
3016 s->fake_iface = port->cfg->bond_fake_iface;
3019 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3020 long long stable_id;
3022 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3024 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3025 stable_id = iface->ofp_port;
3027 bond_stable_ids[i++] = stable_id;
3029 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3033 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3034 * instead of obtaining it from the database. */
3036 port_is_synthetic(const struct port *port)
3038 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3041 /* Interface functions. */
3043 /* Returns the correct network device type for interface 'iface' in bridge
3046 iface_get_type(const struct ovsrec_interface *iface,
3047 const struct ovsrec_bridge *br)
3049 /* The local port always has type "internal". Other ports take their type
3050 * from the database and default to "system" if none is specified. */
3051 return (!strcmp(iface->name, br->name) ? "internal"
3052 : iface->type[0] ? iface->type
3057 iface_destroy(struct iface *iface)
3060 struct port *port = iface->port;
3061 struct bridge *br = port->bridge;
3063 if (br->ofproto && iface->ofp_port >= 0) {
3064 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3067 if (iface->ofp_port >= 0) {
3068 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3071 list_remove(&iface->port_elem);
3072 hmap_remove(&br->iface_by_name, &iface->name_node);
3074 netdev_close(iface->netdev);
3081 static struct iface *
3082 iface_lookup(const struct bridge *br, const char *name)
3084 struct iface *iface;
3086 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3087 &br->iface_by_name) {
3088 if (!strcmp(iface->name, name)) {
3096 static struct iface *
3097 iface_find(const char *name)
3099 const struct bridge *br;
3101 HMAP_FOR_EACH (br, node, &all_bridges) {
3102 struct iface *iface = iface_lookup(br, name);
3111 static struct if_cfg *
3112 if_cfg_lookup(const struct bridge *br, const char *name)
3114 struct if_cfg *if_cfg;
3116 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3118 if (!strcmp(if_cfg->cfg->name, name)) {
3126 static struct iface *
3127 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3129 struct iface *iface;
3131 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3132 hash_int(ofp_port, 0), &br->ifaces) {
3133 if (iface->ofp_port == ofp_port) {
3140 /* Set Ethernet address of 'iface', if one is specified in the configuration
3143 iface_set_mac(struct iface *iface)
3145 uint8_t ea[ETH_ADDR_LEN];
3147 if (!strcmp(iface->type, "internal")
3148 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3149 if (iface->ofp_port == OFPP_LOCAL) {
3150 VLOG_ERR("interface %s: ignoring mac in Interface record "
3151 "(use Bridge record to set local port's mac)",
3153 } else if (eth_addr_is_multicast(ea)) {
3154 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3157 int error = netdev_set_etheraddr(iface->netdev, ea);
3159 VLOG_ERR("interface %s: setting MAC failed (%s)",
3160 iface->name, strerror(error));
3166 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3168 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3170 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3171 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3175 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3176 * sets the "ofport" field to -1.
3178 * This is appropriate when 'if_cfg''s interface cannot be created or is
3179 * otherwise invalid. */
3181 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3183 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3184 iface_set_ofport(if_cfg, -1);
3185 ovsrec_interface_set_status(if_cfg, NULL);
3186 ovsrec_interface_set_admin_state(if_cfg, NULL);
3187 ovsrec_interface_set_duplex(if_cfg, NULL);
3188 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3189 ovsrec_interface_set_link_state(if_cfg, NULL);
3190 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3191 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3192 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3193 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3194 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3195 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3199 struct iface_delete_queues_cbdata {
3200 struct netdev *netdev;
3201 const struct ovsdb_datum *queues;
3205 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3207 union ovsdb_atom atom;
3209 atom.integer = target;
3210 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3214 iface_delete_queues(unsigned int queue_id,
3215 const struct smap *details OVS_UNUSED, void *cbdata_)
3217 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3219 if (!queue_ids_include(cbdata->queues, queue_id)) {
3220 netdev_delete_queue(cbdata->netdev, queue_id);
3225 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3227 struct ofpbuf queues_buf;
3229 ofpbuf_init(&queues_buf, 0);
3231 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3232 netdev_set_qos(iface->netdev, NULL, NULL);
3234 struct iface_delete_queues_cbdata cbdata;
3238 /* Configure top-level Qos for 'iface'. */
3239 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3241 /* Deconfigure queues that were deleted. */
3242 cbdata.netdev = iface->netdev;
3243 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3245 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3247 /* Configure queues for 'iface'. */
3249 for (i = 0; i < qos->n_queues; i++) {
3250 const struct ovsrec_queue *queue = qos->value_queues[i];
3251 unsigned int queue_id = qos->key_queues[i];
3253 if (queue_id == 0) {
3257 if (queue->n_dscp == 1) {
3258 struct ofproto_port_queue *port_queue;
3260 port_queue = ofpbuf_put_uninit(&queues_buf,
3261 sizeof *port_queue);
3262 port_queue->queue = queue_id;
3263 port_queue->dscp = queue->dscp[0];
3266 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3269 struct smap details;
3271 smap_init(&details);
3272 netdev_set_queue(iface->netdev, 0, &details);
3273 smap_destroy(&details);
3277 if (iface->ofp_port >= 0) {
3278 const struct ofproto_port_queue *port_queues = queues_buf.data;
3279 size_t n_queues = queues_buf.size / sizeof *port_queues;
3281 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3282 port_queues, n_queues);
3285 netdev_set_policing(iface->netdev,
3286 iface->cfg->ingress_policing_rate,
3287 iface->cfg->ingress_policing_burst);
3289 ofpbuf_uninit(&queues_buf);
3293 iface_configure_cfm(struct iface *iface)
3295 const struct ovsrec_interface *cfg = iface->cfg;
3296 const char *opstate_str;
3297 const char *cfm_ccm_vlan;
3298 struct cfm_settings s;
3300 if (!cfg->n_cfm_mpid) {
3301 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3305 s.mpid = *cfg->cfm_mpid;
3306 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3307 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3308 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3310 if (s.interval <= 0) {
3314 if (!cfm_ccm_vlan) {
3316 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3317 s.ccm_vlan = CFM_RANDOM_VLAN;
3319 s.ccm_vlan = atoi(cfm_ccm_vlan);
3320 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3325 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3328 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3329 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3331 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3334 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3335 * instead of obtaining it from the database. */
3337 iface_is_synthetic(const struct iface *iface)
3339 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3343 /* Port mirroring. */
3345 static struct mirror *
3346 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3350 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3351 if (uuid_equals(uuid, &m->uuid)) {
3359 bridge_configure_mirrors(struct bridge *br)
3361 const struct ovsdb_datum *mc;
3362 unsigned long *flood_vlans;
3363 struct mirror *m, *next;
3366 /* Get rid of deleted mirrors. */
3367 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3368 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3369 union ovsdb_atom atom;
3371 atom.uuid = m->uuid;
3372 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3377 /* Add new mirrors and reconfigure existing ones. */
3378 for (i = 0; i < br->cfg->n_mirrors; i++) {
3379 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3380 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3382 m = mirror_create(br, cfg);
3385 if (!mirror_configure(m)) {
3390 /* Update flooded vlans (for RSPAN). */
3391 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3392 br->cfg->n_flood_vlans);
3393 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3394 bitmap_free(flood_vlans);
3397 static struct mirror *
3398 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3402 m = xzalloc(sizeof *m);
3403 m->uuid = cfg->header_.uuid;
3404 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3406 m->name = xstrdup(cfg->name);
3412 mirror_destroy(struct mirror *m)
3415 struct bridge *br = m->bridge;
3418 ofproto_mirror_unregister(br->ofproto, m);
3421 hmap_remove(&br->mirrors, &m->hmap_node);
3428 mirror_collect_ports(struct mirror *m,
3429 struct ovsrec_port **in_ports, int n_in_ports,
3430 void ***out_portsp, size_t *n_out_portsp)
3432 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3433 size_t n_out_ports = 0;
3436 for (i = 0; i < n_in_ports; i++) {
3437 const char *name = in_ports[i]->name;
3438 struct port *port = port_lookup(m->bridge, name);
3440 out_ports[n_out_ports++] = port;
3442 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3443 "port %s", m->bridge->name, m->name, name);
3446 *out_portsp = out_ports;
3447 *n_out_portsp = n_out_ports;
3451 mirror_configure(struct mirror *m)
3453 const struct ovsrec_mirror *cfg = m->cfg;
3454 struct ofproto_mirror_settings s;
3457 if (strcmp(cfg->name, m->name)) {
3459 m->name = xstrdup(cfg->name);
3463 /* Get output port or VLAN. */
3464 if (cfg->output_port) {
3465 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3466 if (!s.out_bundle) {
3467 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3468 m->bridge->name, m->name);
3471 s.out_vlan = UINT16_MAX;
3473 if (cfg->output_vlan) {
3474 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3475 "output vlan; ignoring output vlan",
3476 m->bridge->name, m->name);
3478 } else if (cfg->output_vlan) {
3479 /* The database should prevent invalid VLAN values. */
3480 s.out_bundle = NULL;
3481 s.out_vlan = *cfg->output_vlan;
3483 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3484 m->bridge->name, m->name);
3488 /* Get port selection. */
3489 if (cfg->select_all) {
3490 size_t n_ports = hmap_count(&m->bridge->ports);
3491 void **ports = xmalloc(n_ports * sizeof *ports);
3496 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3506 /* Get ports, dropping ports that don't exist.
3507 * The IDL ensures that there are no duplicates. */
3508 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3509 &s.srcs, &s.n_srcs);
3510 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3511 &s.dsts, &s.n_dsts);
3514 /* Get VLAN selection. */
3515 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3518 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3521 if (s.srcs != s.dsts) {
3530 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3532 * This is deprecated. It is only for compatibility with broken device drivers
3533 * in old versions of Linux that do not properly support VLANs when VLAN
3534 * devices are not used. When broken device drivers are no longer in
3535 * widespread use, we will delete these interfaces. */
3537 static struct ovsrec_port **recs;
3538 static size_t n_recs, allocated_recs;
3540 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3541 * splinters are reconfigured. */
3543 register_rec(struct ovsrec_port *rec)
3545 if (n_recs >= allocated_recs) {
3546 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3548 recs[n_recs++] = rec;
3551 /* Frees all of the ports registered with register_reg(). */
3553 free_registered_recs(void)
3557 for (i = 0; i < n_recs; i++) {
3558 struct ovsrec_port *port = recs[i];
3561 for (j = 0; j < port->n_interfaces; j++) {
3562 struct ovsrec_interface *iface = port->interfaces[j];
3567 smap_destroy(&port->other_config);
3568 free(port->interfaces);
3576 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3579 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3581 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3585 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3588 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3589 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3590 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3593 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3594 * use, returns NULL.
3596 * Updates 'vlan_splinters_enabled_anywhere'. */
3597 static unsigned long int *
3598 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3600 unsigned long int *splinter_vlans;
3601 struct sset splinter_ifaces;
3602 const char *real_dev_name;
3603 struct shash *real_devs;
3604 struct shash_node *node;
3608 /* Free space allocated for synthesized ports and interfaces, since we're
3609 * in the process of reconstructing all of them. */
3610 free_registered_recs();
3612 splinter_vlans = bitmap_allocate(4096);
3613 sset_init(&splinter_ifaces);
3614 vlan_splinters_enabled_anywhere = false;
3615 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3616 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3619 for (j = 0; j < br_cfg->n_ports; j++) {
3620 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3623 for (k = 0; k < port_cfg->n_interfaces; k++) {
3624 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3626 if (vlan_splinters_is_enabled(iface_cfg)) {
3627 vlan_splinters_enabled_anywhere = true;
3628 sset_add(&splinter_ifaces, iface_cfg->name);
3629 vlan_bitmap_from_array__(port_cfg->trunks,
3635 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3636 bitmap_set1(splinter_vlans, *port_cfg->tag);
3641 if (!vlan_splinters_enabled_anywhere) {
3642 free(splinter_vlans);
3643 sset_destroy(&splinter_ifaces);
3647 HMAP_FOR_EACH (br, node, &all_bridges) {
3649 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3653 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3654 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3655 * device to be created for it. */
3656 bitmap_set0(splinter_vlans, 0);
3657 bitmap_set0(splinter_vlans, 4095);
3659 /* Delete all VLAN devices that we don't need. */
3661 real_devs = vlandev_get_real_devs();
3662 SHASH_FOR_EACH (node, real_devs) {
3663 const struct vlan_real_dev *real_dev = node->data;
3664 const struct vlan_dev *vlan_dev;
3665 bool real_dev_has_splinters;
3667 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3669 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3670 if (!real_dev_has_splinters
3671 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3672 struct netdev *netdev;
3674 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3675 if (!netdev_get_in4(netdev, NULL, NULL) ||
3676 !netdev_get_in6(netdev, NULL)) {
3677 vlandev_del(vlan_dev->name);
3679 /* It has an IP address configured, so we don't own
3680 * it. Don't delete it. */
3682 netdev_close(netdev);
3689 /* Add all VLAN devices that we need. */
3690 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3693 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3694 if (!vlandev_get_name(real_dev_name, vid)) {
3695 vlandev_add(real_dev_name, vid);
3702 sset_destroy(&splinter_ifaces);
3704 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3705 free(splinter_vlans);
3708 return splinter_vlans;
3711 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3714 configure_splinter_port(struct port *port)
3716 struct ofproto *ofproto = port->bridge->ofproto;
3717 uint16_t realdev_ofp_port;
3718 const char *realdev_name;
3719 struct iface *vlandev, *realdev;
3721 ofproto_bundle_unregister(port->bridge->ofproto, port);
3723 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3726 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3727 realdev = iface_lookup(port->bridge, realdev_name);
3728 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3730 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3734 static struct ovsrec_port *
3735 synthesize_splinter_port(const char *real_dev_name,
3736 const char *vlan_dev_name, int vid)
3738 struct ovsrec_interface *iface;
3739 struct ovsrec_port *port;
3741 iface = xmalloc(sizeof *iface);
3742 ovsrec_interface_init(iface);
3743 iface->name = xstrdup(vlan_dev_name);
3744 iface->type = "system";
3746 port = xmalloc(sizeof *port);
3747 ovsrec_port_init(port);
3748 port->interfaces = xmemdup(&iface, sizeof iface);
3749 port->n_interfaces = 1;
3750 port->name = xstrdup(vlan_dev_name);
3751 port->vlan_mode = "splinter";
3752 port->tag = xmalloc(sizeof *port->tag);
3755 smap_add(&port->other_config, "realdev", real_dev_name);
3761 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3762 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3763 * 1-bit in the 'splinter_vlans' bitmap. */
3765 add_vlan_splinter_ports(struct bridge *br,
3766 const unsigned long int *splinter_vlans,
3767 struct shash *ports)
3771 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3772 * we're modifying 'ports'. */
3773 for (i = 0; i < br->cfg->n_ports; i++) {
3774 const char *name = br->cfg->ports[i]->name;
3775 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3778 for (j = 0; j < port_cfg->n_interfaces; j++) {
3779 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3781 if (vlan_splinters_is_enabled(iface_cfg)) {
3782 const char *real_dev_name;
3785 real_dev_name = iface_cfg->name;
3786 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3787 const char *vlan_dev_name;
3789 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3791 && !shash_find(ports, vlan_dev_name)) {
3792 shash_add(ports, vlan_dev_name,
3793 synthesize_splinter_port(
3794 real_dev_name, vlan_dev_name, vid));
3803 mirror_refresh_stats(struct mirror *m)
3805 struct ofproto *ofproto = m->bridge->ofproto;
3806 uint64_t tx_packets, tx_bytes;
3809 size_t stat_cnt = 0;
3811 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3812 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3816 if (tx_packets != UINT64_MAX) {
3817 keys[stat_cnt] = "tx_packets";
3818 values[stat_cnt] = tx_packets;
3821 if (tx_bytes != UINT64_MAX) {
3822 keys[stat_cnt] = "tx_bytes";
3823 values[stat_cnt] = tx_bytes;
3827 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);