1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks
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"
36 #include "ofp-print.h"
38 #include "ofproto/ofproto.h"
39 #include "poll-loop.h"
42 #include "socket-util.h"
44 #include "stream-ssl.h"
46 #include "system-stats.h"
51 #include "vswitchd/vswitch-idl.h"
52 #include "xenserver.h"
54 #include "sflow_api.h"
55 #include "vlan-bitmap.h"
57 VLOG_DEFINE_THIS_MODULE(bridge);
59 COVERAGE_DEFINE(bridge_reconfigure);
62 /* These members are always valid. */
63 struct list port_elem; /* Element in struct port's "ifaces" list. */
64 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
65 struct port *port; /* Containing port. */
66 char *name; /* Host network device name. */
67 tag_type tag; /* Tag associated with this interface. */
69 /* These members are valid only after bridge_reconfigure() causes them to
71 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
72 int ofp_port; /* OpenFlow port number, -1 if unknown. */
73 struct netdev *netdev; /* Network device. */
74 const char *type; /* Usually same as cfg->type. */
75 const struct ovsrec_interface *cfg;
79 struct uuid uuid; /* UUID of this "mirror" record in database. */
80 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
81 struct bridge *bridge;
83 const struct ovsrec_mirror *cfg;
87 struct bridge *bridge;
88 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
91 const struct ovsrec_port *cfg;
93 /* An ordinary bridge port has 1 interface.
94 * A bridge port for bonding has at least 2 interfaces. */
95 struct list ifaces; /* List of "struct iface"s. */
99 struct hmap_node node; /* In 'all_bridges'. */
100 char *name; /* User-specified arbitrary name. */
101 char *type; /* Datapath type. */
102 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
103 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
104 const struct ovsrec_bridge *cfg;
106 /* OpenFlow switch processing. */
107 struct ofproto *ofproto; /* OpenFlow switch. */
110 struct hmap ports; /* "struct port"s indexed by name. */
111 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
112 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
114 /* Port mirroring. */
115 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
117 /* Synthetic local port if necessary. */
118 struct ovsrec_port synth_local_port;
119 struct ovsrec_interface synth_local_iface;
120 struct ovsrec_interface *synth_local_ifacep;
123 /* All bridges, indexed by name. */
124 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
126 /* OVSDB IDL used to obtain configuration. */
127 static struct ovsdb_idl *idl;
129 /* Each time this timer expires, the bridge fetches systems and interface
130 * statistics and pushes them into the database. */
131 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
132 static long long int stats_timer = LLONG_MIN;
134 /* Stores the time after which rate limited statistics may be written to the
135 * database. Only updated when changes to the database require rate limiting.
137 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
138 static long long int db_limiter = LLONG_MIN;
140 static void add_del_bridges(const struct ovsrec_open_vswitch *);
141 static void bridge_del_ofprotos(void);
142 static bool bridge_add_ofprotos(struct bridge *);
143 static void bridge_create(const struct ovsrec_bridge *);
144 static void bridge_destroy(struct bridge *);
145 static struct bridge *bridge_lookup(const char *name);
146 static unixctl_cb_func bridge_unixctl_dump_flows;
147 static unixctl_cb_func bridge_unixctl_reconnect;
148 static size_t bridge_get_controllers(const struct bridge *br,
149 struct ovsrec_controller ***controllersp);
150 static void bridge_add_del_ports(struct bridge *,
151 const unsigned long int *splinter_vlans);
152 static void bridge_add_ofproto_ports(struct bridge *);
153 static void bridge_del_ofproto_ports(struct bridge *);
154 static void bridge_refresh_ofp_port(struct bridge *);
155 static void bridge_configure_datapath_id(struct bridge *);
156 static void bridge_configure_flow_eviction_threshold(struct bridge *);
157 static void bridge_configure_netflow(struct bridge *);
158 static void bridge_configure_forward_bpdu(struct bridge *);
159 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
160 static void bridge_configure_stp(struct bridge *);
161 static void bridge_configure_remotes(struct bridge *,
162 const struct sockaddr_in *managers,
164 static void bridge_pick_local_hw_addr(struct bridge *,
165 uint8_t ea[ETH_ADDR_LEN],
166 struct iface **hw_addr_iface);
167 static uint64_t bridge_pick_datapath_id(struct bridge *,
168 const uint8_t bridge_ea[ETH_ADDR_LEN],
169 struct iface *hw_addr_iface);
170 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
172 static const char *get_port_other_config(const struct ovsrec_port *,
174 const char *default_value);
175 static uint64_t dpid_from_hash(const void *, size_t nbytes);
176 static bool bridge_has_bond_fake_iface(const struct bridge *,
178 static bool port_is_bond_fake_iface(const struct port *);
180 static unixctl_cb_func qos_unixctl_show;
182 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
183 static void port_add_ifaces(struct port *);
184 static void port_del_ifaces(struct port *);
185 static void port_destroy(struct port *);
186 static struct port *port_lookup(const struct bridge *, const char *name);
187 static void port_configure(struct port *);
188 static struct lacp_settings *port_configure_lacp(struct port *,
189 struct lacp_settings *);
190 static void port_configure_bond(struct port *, struct bond_settings *,
191 uint32_t *bond_stable_ids);
192 static bool port_is_synthetic(const struct port *);
194 static void bridge_configure_mirrors(struct bridge *);
195 static struct mirror *mirror_create(struct bridge *,
196 const struct ovsrec_mirror *);
197 static void mirror_destroy(struct mirror *);
198 static bool mirror_configure(struct mirror *);
199 static void mirror_refresh_stats(struct mirror *);
201 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
202 static struct iface *iface_create(struct port *port,
203 const struct ovsrec_interface *if_cfg);
204 static void iface_destroy(struct iface *);
205 static struct iface *iface_lookup(const struct bridge *, const char *name);
206 static struct iface *iface_find(const char *name);
207 static struct iface *iface_from_ofp_port(const struct bridge *,
209 static void iface_set_mac(struct iface *);
210 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
211 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
212 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
213 static void iface_configure_cfm(struct iface *);
214 static void iface_refresh_cfm_stats(struct iface *);
215 static void iface_refresh_stats(struct iface *);
216 static void iface_refresh_status(struct iface *);
217 static bool iface_is_synthetic(const struct iface *);
218 static const char *get_interface_other_config(const struct ovsrec_interface *,
220 const char *default_value);
222 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
224 static void shash_to_ovs_idl_map(struct shash *,
225 char ***keys, char ***values, size_t *n);
227 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
229 * This is deprecated. It is only for compatibility with broken device drivers
230 * in old versions of Linux that do not properly support VLANs when VLAN
231 * devices are not used. When broken device drivers are no longer in
232 * widespread use, we will delete these interfaces. */
234 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
235 static bool vlan_splinters_enabled_anywhere;
237 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
238 static unsigned long int *collect_splinter_vlans(
239 const struct ovsrec_open_vswitch *);
240 static void configure_splinter_port(struct port *);
241 static void add_vlan_splinter_ports(struct bridge *,
242 const unsigned long int *splinter_vlans,
243 struct shash *ports);
245 /* Public functions. */
247 /* Initializes the bridge module, configuring it to obtain its configuration
248 * from an OVSDB server accessed over 'remote', which should be a string in a
249 * form acceptable to ovsdb_idl_create(). */
251 bridge_init(const char *remote)
253 /* Create connection to database. */
254 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
255 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
257 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
258 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
259 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
260 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
261 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
262 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
263 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
265 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
266 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
267 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
269 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
270 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
271 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
272 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
274 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
275 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
276 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
277 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
283 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
284 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
285 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
286 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
288 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
289 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
290 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
291 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
293 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
297 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
298 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
300 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
302 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
305 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
306 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
307 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
308 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
310 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
312 /* Register unixctl commands. */
313 unixctl_command_register("qos/show", "interface", 1, 1,
314 qos_unixctl_show, NULL);
315 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
316 bridge_unixctl_dump_flows, NULL);
317 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
318 bridge_unixctl_reconnect, NULL);
328 struct bridge *br, *next_br;
330 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
333 ovsdb_idl_destroy(idl);
336 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
337 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
338 * responsible for freeing '*managersp' (with free()).
340 * You may be asking yourself "why does ovs-vswitchd care?", because
341 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
342 * should not be and in fact is not directly involved in that. But
343 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
344 * it has to tell in-band control where the managers are to enable that.
345 * (Thus, only managers connected in-band are collected.)
348 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
349 struct sockaddr_in **managersp, size_t *n_managersp)
351 struct sockaddr_in *managers = NULL;
352 size_t n_managers = 0;
356 /* Collect all of the potential targets from the "targets" columns of the
357 * rows pointed to by "manager_options", excluding any that are
360 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
361 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
363 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
364 sset_find_and_delete(&targets, m->target);
366 sset_add(&targets, m->target);
370 /* Now extract the targets' IP addresses. */
371 if (!sset_is_empty(&targets)) {
374 managers = xmalloc(sset_count(&targets) * sizeof *managers);
375 SSET_FOR_EACH (target, &targets) {
376 struct sockaddr_in *sin = &managers[n_managers];
378 if (stream_parse_target_with_default_ports(target,
386 sset_destroy(&targets);
388 *managersp = managers;
389 *n_managersp = n_managers;
393 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
395 unsigned long int *splinter_vlans;
396 struct sockaddr_in *managers;
397 struct bridge *br, *next;
398 int sflow_bridge_number;
401 COVERAGE_INC(bridge_reconfigure);
403 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
404 * iface"s according to 'ovs_cfg', with only very minimal configuration
407 * This is mostly an update to bridge data structures. Very little is
408 * pushed down to ofproto or lower layers. */
409 add_del_bridges(ovs_cfg);
410 splinter_vlans = collect_splinter_vlans(ovs_cfg);
411 HMAP_FOR_EACH (br, node, &all_bridges) {
412 bridge_add_del_ports(br, splinter_vlans);
414 free(splinter_vlans);
416 /* Delete all datapaths and datapath ports that are no longer configured.
418 * The kernel will reject any attempt to add a given port to a datapath if
419 * that port already belongs to a different datapath, so we must do all
420 * port deletions before any port additions. A datapath always has a
421 * "local port" so we must delete not-configured datapaths too. */
422 bridge_del_ofprotos();
423 HMAP_FOR_EACH (br, node, &all_bridges) {
425 bridge_del_ofproto_ports(br);
429 /* Create datapaths and datapath ports that are missing.
431 * After this is done, we have our final set of bridges, ports, and
432 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
433 * has at least one iface, every "struct iface" has a valid ofp_port and
435 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
436 if (!br->ofproto && !bridge_add_ofprotos(br)) {
440 HMAP_FOR_EACH (br, node, &all_bridges) {
441 bridge_refresh_ofp_port(br);
442 bridge_add_ofproto_ports(br);
445 /* Complete the configuration. */
446 sflow_bridge_number = 0;
447 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
448 HMAP_FOR_EACH (br, node, &all_bridges) {
451 /* We need the datapath ID early to allow LACP ports to use it as the
452 * default system ID. */
453 bridge_configure_datapath_id(br);
455 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
458 port_configure(port);
460 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
461 iface_configure_cfm(iface);
462 iface_configure_qos(iface, port->cfg->qos);
463 iface_set_mac(iface);
466 bridge_configure_mirrors(br);
467 bridge_configure_flow_eviction_threshold(br);
468 bridge_configure_forward_bpdu(br);
469 bridge_configure_remotes(br, managers, n_managers);
470 bridge_configure_netflow(br);
471 bridge_configure_sflow(br, &sflow_bridge_number);
472 bridge_configure_stp(br);
476 /* ovs-vswitchd has completed initialization, so allow the process that
477 * forked us to exit successfully. */
478 daemonize_complete();
481 /* Iterate over all ofprotos and delete any of them that do not have a
482 * configured bridge or that are the wrong type. */
484 bridge_del_ofprotos(void)
492 ofproto_enumerate_types(&types);
493 SSET_FOR_EACH (type, &types) {
496 ofproto_enumerate_names(type, &names);
497 SSET_FOR_EACH (name, &names) {
498 struct bridge *br = bridge_lookup(name);
499 if (!br || strcmp(type, br->type)) {
500 ofproto_delete(name, type);
504 sset_destroy(&names);
505 sset_destroy(&types);
509 bridge_add_ofprotos(struct bridge *br)
511 int error = ofproto_create(br->name, br->type, &br->ofproto);
513 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
520 port_configure(struct port *port)
522 const struct ovsrec_port *cfg = port->cfg;
523 struct bond_settings bond_settings;
524 struct lacp_settings lacp_settings;
525 struct ofproto_bundle_settings s;
528 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
529 configure_splinter_port(port);
538 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
539 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
540 s.slaves[s.n_slaves++] = iface->ofp_port;
545 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
549 /* Get VLAN trunks. */
552 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
556 if (cfg->vlan_mode) {
557 if (!strcmp(cfg->vlan_mode, "access")) {
558 s.vlan_mode = PORT_VLAN_ACCESS;
559 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
560 s.vlan_mode = PORT_VLAN_TRUNK;
561 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
562 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
563 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
564 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
566 /* This "can't happen" because ovsdb-server should prevent it. */
567 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
568 s.vlan_mode = PORT_VLAN_TRUNK;
572 s.vlan_mode = PORT_VLAN_ACCESS;
574 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
578 s.vlan_mode = PORT_VLAN_TRUNK;
581 s.use_priority_tags = !strcmp("true", get_port_other_config(
582 cfg, "priority-tags", ""));
584 /* Get LACP settings. */
585 s.lacp = port_configure_lacp(port, &lacp_settings);
589 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
590 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
591 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
594 s.lacp_slaves = NULL;
597 /* Get bond settings. */
598 if (s.n_slaves > 1) {
599 s.bond = &bond_settings;
600 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
601 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
604 s.bond_stable_ids = NULL;
606 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
607 netdev_set_miimon_interval(iface->netdev, 0);
612 ofproto_bundle_register(port->bridge->ofproto, port, &s);
618 free(s.bond_stable_ids);
621 /* Pick local port hardware address and datapath ID for 'br'. */
623 bridge_configure_datapath_id(struct bridge *br)
625 uint8_t ea[ETH_ADDR_LEN];
627 struct iface *local_iface;
628 struct iface *hw_addr_iface;
631 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
632 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
634 int error = netdev_set_etheraddr(local_iface->netdev, ea);
636 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
637 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
638 "Ethernet address: %s",
639 br->name, strerror(error));
642 memcpy(br->ea, ea, ETH_ADDR_LEN);
644 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
645 ofproto_set_datapath_id(br->ofproto, dpid);
647 dpid_string = xasprintf("%016"PRIx64, dpid);
648 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
652 /* Set NetFlow configuration on 'br'. */
654 bridge_configure_netflow(struct bridge *br)
656 struct ovsrec_netflow *cfg = br->cfg->netflow;
657 struct netflow_options opts;
660 ofproto_set_netflow(br->ofproto, NULL);
664 memset(&opts, 0, sizeof opts);
666 /* Get default NetFlow configuration from datapath.
667 * Apply overrides from 'cfg'. */
668 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
669 if (cfg->engine_type) {
670 opts.engine_type = *cfg->engine_type;
672 if (cfg->engine_id) {
673 opts.engine_id = *cfg->engine_id;
676 /* Configure active timeout interval. */
677 opts.active_timeout = cfg->active_timeout;
678 if (!opts.active_timeout) {
679 opts.active_timeout = -1;
680 } else if (opts.active_timeout < 0) {
681 VLOG_WARN("bridge %s: active timeout interval set to negative "
682 "value, using default instead (%d seconds)", br->name,
683 NF_ACTIVE_TIMEOUT_DEFAULT);
684 opts.active_timeout = -1;
687 /* Add engine ID to interface number to disambiguate bridgs? */
688 opts.add_id_to_iface = cfg->add_id_to_interface;
689 if (opts.add_id_to_iface) {
690 if (opts.engine_id > 0x7f) {
691 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
692 "another vswitch, choose an engine id less than 128",
695 if (hmap_count(&br->ports) > 508) {
696 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
697 "another port when more than 508 ports are used",
703 sset_init(&opts.collectors);
704 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
707 if (ofproto_set_netflow(br->ofproto, &opts)) {
708 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
710 sset_destroy(&opts.collectors);
713 /* Set sFlow configuration on 'br'. */
715 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
717 const struct ovsrec_sflow *cfg = br->cfg->sflow;
718 struct ovsrec_controller **controllers;
719 struct ofproto_sflow_options oso;
720 size_t n_controllers;
724 ofproto_set_sflow(br->ofproto, NULL);
728 memset(&oso, 0, sizeof oso);
730 sset_init(&oso.targets);
731 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
733 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
735 oso.sampling_rate = *cfg->sampling;
738 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
740 oso.polling_interval = *cfg->polling;
743 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
745 oso.header_len = *cfg->header;
748 oso.sub_id = (*sflow_bridge_number)++;
749 oso.agent_device = cfg->agent;
751 oso.control_ip = NULL;
752 n_controllers = bridge_get_controllers(br, &controllers);
753 for (i = 0; i < n_controllers; i++) {
754 if (controllers[i]->local_ip) {
755 oso.control_ip = controllers[i]->local_ip;
759 ofproto_set_sflow(br->ofproto, &oso);
761 sset_destroy(&oso.targets);
765 port_configure_stp(const struct ofproto *ofproto, struct port *port,
766 struct ofproto_port_stp_settings *port_s,
767 int *port_num_counter, unsigned long *port_num_bitmap)
769 const char *config_str;
772 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
773 if (config_str && !strcmp(config_str, "false")) {
774 port_s->enable = false;
777 port_s->enable = true;
780 /* STP over bonds is not supported. */
781 if (!list_is_singleton(&port->ifaces)) {
782 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
784 port_s->enable = false;
788 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
790 /* Internal ports shouldn't participate in spanning tree, so
792 if (!strcmp(iface->type, "internal")) {
793 VLOG_DBG("port %s: disable STP on internal ports", port->name);
794 port_s->enable = false;
798 /* STP on mirror output ports is not supported. */
799 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
800 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
801 port_s->enable = false;
805 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
807 unsigned long int port_num = strtoul(config_str, NULL, 0);
808 int port_idx = port_num - 1;
810 if (port_num < 1 || port_num > STP_MAX_PORTS) {
811 VLOG_ERR("port %s: invalid stp-port-num", port->name);
812 port_s->enable = false;
816 if (bitmap_is_set(port_num_bitmap, port_idx)) {
817 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
818 port->name, port_num);
819 port_s->enable = false;
822 bitmap_set1(port_num_bitmap, port_idx);
823 port_s->port_num = port_idx;
825 if (*port_num_counter > STP_MAX_PORTS) {
826 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
827 port_s->enable = false;
831 port_s->port_num = (*port_num_counter)++;
834 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
836 port_s->path_cost = strtoul(config_str, NULL, 10);
840 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
841 /* Couldn't get speed, so assume 100Mb/s. */
842 port_s->path_cost = 19;
846 mbps = netdev_features_to_bps(current) / 1000000;
847 port_s->path_cost = stp_convert_speed_to_cost(mbps);
851 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
853 port_s->priority = strtoul(config_str, NULL, 0);
855 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
859 /* Set spanning tree configuration on 'br'. */
861 bridge_configure_stp(struct bridge *br)
863 if (!br->cfg->stp_enable) {
864 ofproto_set_stp(br->ofproto, NULL);
866 struct ofproto_stp_settings br_s;
867 const char *config_str;
869 int port_num_counter;
870 unsigned long *port_num_bitmap;
872 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
874 uint8_t ea[ETH_ADDR_LEN];
876 if (eth_addr_from_string(config_str, ea)) {
877 br_s.system_id = eth_addr_to_uint64(ea);
879 br_s.system_id = eth_addr_to_uint64(br->ea);
880 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
881 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
884 br_s.system_id = eth_addr_to_uint64(br->ea);
887 config_str = bridge_get_other_config(br->cfg, "stp-priority");
889 br_s.priority = strtoul(config_str, NULL, 0);
891 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
894 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
896 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
898 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
901 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
903 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
905 br_s.max_age = STP_DEFAULT_MAX_AGE;
908 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
910 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
912 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
915 /* Configure STP on the bridge. */
916 if (ofproto_set_stp(br->ofproto, &br_s)) {
917 VLOG_ERR("bridge %s: could not enable STP", br->name);
921 /* Users must either set the port number with the "stp-port-num"
922 * configuration on all ports or none. If manual configuration
923 * is not done, then we allocate them sequentially. */
924 port_num_counter = 0;
925 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
926 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
927 struct ofproto_port_stp_settings port_s;
930 port_configure_stp(br->ofproto, port, &port_s,
931 &port_num_counter, port_num_bitmap);
933 /* As bonds are not supported, just apply configuration to
935 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
936 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
938 VLOG_ERR("port %s: could not enable STP", port->name);
944 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
945 && port_num_counter) {
946 VLOG_ERR("bridge %s: must manually configure all STP port "
947 "IDs or none, disabling", br->name);
948 ofproto_set_stp(br->ofproto, NULL);
950 bitmap_free(port_num_bitmap);
955 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
957 const struct port *port = port_lookup(br, name);
958 return port && port_is_bond_fake_iface(port);
962 port_is_bond_fake_iface(const struct port *port)
964 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
968 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
970 struct bridge *br, *next;
974 /* Collect new bridges' names and types. */
976 for (i = 0; i < cfg->n_bridges; i++) {
977 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
978 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
980 if (strchr(br_cfg->name, '/')) {
981 /* Prevent remote ovsdb-server users from accessing arbitrary
982 * directories, e.g. consider a bridge named "../../../etc/". */
983 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
985 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
986 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
990 /* Get rid of deleted bridges or those whose types have changed.
991 * Update 'cfg' of bridges that still exist. */
992 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
993 br->cfg = shash_find_data(&new_br, br->name);
994 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
995 br->cfg->datapath_type))) {
1000 /* Add new bridges. */
1001 for (i = 0; i < cfg->n_bridges; i++) {
1002 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1003 struct bridge *br = bridge_lookup(br_cfg->name);
1005 bridge_create(br_cfg);
1009 shash_destroy(&new_br);
1012 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1015 * The kernel will reject any attempt to add a given port to a datapath if that
1016 * port already belongs to a different datapath, so we must do all port
1017 * deletions before any port additions. */
1019 bridge_del_ofproto_ports(struct bridge *br)
1021 struct ofproto_port_dump dump;
1022 struct ofproto_port ofproto_port;
1024 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1025 const char *name = ofproto_port.name;
1026 struct iface *iface;
1030 /* Ignore the local port. We can't change it anyhow. */
1031 if (!strcmp(name, br->name)) {
1035 /* Get the type that 'ofproto_port' should have (ordinarily the
1036 * type of its corresponding iface) or NULL if it should be
1038 iface = iface_lookup(br, name);
1039 type = (iface ? iface->type
1040 : bridge_has_bond_fake_iface(br, name) ? "internal"
1043 /* If it's the wrong type then delete the ofproto port. */
1045 && !strcmp(ofproto_port.type, type)
1046 && (!iface || !iface->netdev
1047 || !strcmp(netdev_get_type(iface->netdev), type))) {
1050 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1052 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1053 br->name, name, strerror(error));
1056 netdev_close(iface->netdev);
1057 iface->netdev = NULL;
1063 iface_set_ofp_port(struct iface *iface, int ofp_port)
1065 struct bridge *br = iface->port->bridge;
1067 assert(iface->ofp_port < 0 && ofp_port >= 0);
1068 iface->ofp_port = ofp_port;
1069 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1070 iface_set_ofport(iface->cfg, ofp_port);
1074 bridge_refresh_ofp_port(struct bridge *br)
1076 struct ofproto_port_dump dump;
1077 struct ofproto_port ofproto_port;
1080 /* Clear all the "ofp_port"es. */
1081 hmap_clear(&br->ifaces);
1082 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1083 struct iface *iface;
1085 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1086 iface->ofp_port = -1;
1090 /* Obtain the correct "ofp_port"s from ofproto. */
1091 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1092 struct iface *iface = iface_lookup(br, ofproto_port.name);
1094 if (iface->ofp_port >= 0) {
1095 VLOG_WARN("bridge %s: interface %s reported twice",
1096 br->name, ofproto_port.name);
1097 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1098 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1099 br->name, ofproto_port.ofp_port);
1101 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1107 /* Add an ofproto port for any "struct iface" that doesn't have one.
1108 * Delete any "struct iface" for which this fails.
1109 * Delete any "struct port" that thereby ends up with no ifaces. */
1111 bridge_add_ofproto_ports(struct bridge *br)
1113 struct port *port, *next_port;
1115 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1116 struct iface *iface, *next_iface;
1117 struct ofproto_port ofproto_port;
1119 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1122 /* Open the netdev. */
1123 if (!iface->netdev) {
1124 error = netdev_open(iface->name, iface->type, &iface->netdev);
1126 VLOG_WARN("could not open network device %s (%s)",
1127 iface->name, strerror(error));
1131 && port->cfg->vlan_mode
1132 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1133 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1139 /* Configure the netdev. */
1140 if (iface->netdev) {
1144 shash_from_ovs_idl_map(iface->cfg->key_options,
1145 iface->cfg->value_options,
1146 iface->cfg->n_options, &args);
1147 error = netdev_set_config(iface->netdev, &args);
1148 shash_destroy(&args);
1151 VLOG_WARN("could not configure network device %s (%s)",
1152 iface->name, strerror(error));
1153 netdev_close(iface->netdev);
1154 iface->netdev = NULL;
1158 /* Add the port, if necessary. */
1159 if (iface->netdev && iface->ofp_port < 0) {
1163 error = ofproto_port_add(br->ofproto, iface->netdev,
1166 iface_set_ofp_port(iface, ofp_port);
1168 netdev_close(iface->netdev);
1169 iface->netdev = NULL;
1173 /* Populate stats columns in new Interface rows. */
1174 if (iface->netdev && !iface->cfg->mtu) {
1175 iface_refresh_stats(iface);
1176 iface_refresh_status(iface);
1179 /* Delete the iface if we failed. */
1180 if (iface->netdev && iface->ofp_port >= 0) {
1181 VLOG_DBG("bridge %s: interface %s is on port %d",
1182 br->name, iface->name, iface->ofp_port);
1184 if (iface->netdev) {
1185 VLOG_ERR("bridge %s: missing %s interface, dropping",
1186 br->name, iface->name);
1188 /* We already reported a related error, don't bother
1189 * duplicating it. */
1191 iface_clear_db_record(iface->cfg);
1192 iface_destroy(iface);
1195 if (list_is_empty(&port->ifaces)) {
1196 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1201 /* Add bond fake iface if necessary. */
1202 if (port_is_bond_fake_iface(port)) {
1203 if (ofproto_port_query_by_name(br->ofproto, port->name,
1205 struct netdev *netdev;
1208 error = netdev_open(port->name, "internal", &netdev);
1210 ofproto_port_add(br->ofproto, netdev, NULL);
1211 netdev_close(netdev);
1213 VLOG_WARN("could not open network device %s (%s)",
1214 port->name, strerror(error));
1217 /* Already exists, nothing to do. */
1218 ofproto_port_destroy(&ofproto_port);
1225 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1229 for (i = 0; i < n; i++) {
1230 if (!strcmp(keys[i], key)) {
1238 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1240 return get_ovsrec_key_value(br_cfg->key_other_config,
1241 br_cfg->value_other_config,
1242 br_cfg->n_other_config, key);
1245 /* Set Flow eviction threshold */
1247 bridge_configure_flow_eviction_threshold(struct bridge *br)
1249 const char *threshold_str;
1252 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1253 if (threshold_str) {
1254 threshold = strtoul(threshold_str, NULL, 10);
1256 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1258 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1261 /* Set forward BPDU option. */
1263 bridge_configure_forward_bpdu(struct bridge *br)
1265 const char *forward_bpdu_str;
1266 bool forward_bpdu = false;
1268 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1269 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1270 forward_bpdu = true;
1272 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1276 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1277 struct iface **hw_addr_iface)
1279 struct hmapx mirror_output_ports;
1282 bool found_addr = false;
1286 *hw_addr_iface = NULL;
1288 /* Did the user request a particular MAC? */
1289 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1290 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1291 if (eth_addr_is_multicast(ea)) {
1292 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1293 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1294 } else if (eth_addr_is_zero(ea)) {
1295 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1301 /* Mirror output ports don't participate in picking the local hardware
1302 * address. ofproto can't help us find out whether a given port is a
1303 * mirror output because we haven't configured mirrors yet, so we need to
1304 * accumulate them ourselves. */
1305 hmapx_init(&mirror_output_ports);
1306 for (i = 0; i < br->cfg->n_mirrors; i++) {
1307 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1308 if (m->output_port) {
1309 hmapx_add(&mirror_output_ports, m->output_port);
1313 /* Otherwise choose the minimum non-local MAC address among all of the
1315 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1316 uint8_t iface_ea[ETH_ADDR_LEN];
1317 struct iface *candidate;
1318 struct iface *iface;
1320 /* Mirror output ports don't participate. */
1321 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1325 /* Choose the MAC address to represent the port. */
1327 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1328 /* Find the interface with this Ethernet address (if any) so that
1329 * we can provide the correct devname to the caller. */
1330 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1331 uint8_t candidate_ea[ETH_ADDR_LEN];
1332 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1333 && eth_addr_equals(iface_ea, candidate_ea)) {
1338 /* Choose the interface whose MAC address will represent the port.
1339 * The Linux kernel bonding code always chooses the MAC address of
1340 * the first slave added to a bond, and the Fedora networking
1341 * scripts always add slaves to a bond in alphabetical order, so
1342 * for compatibility we choose the interface with the name that is
1343 * first in alphabetical order. */
1344 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1345 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1350 /* The local port doesn't count (since we're trying to choose its
1351 * MAC address anyway). */
1352 if (iface->ofp_port == OFPP_LOCAL) {
1357 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1363 /* Compare against our current choice. */
1364 if (!eth_addr_is_multicast(iface_ea) &&
1365 !eth_addr_is_local(iface_ea) &&
1366 !eth_addr_is_reserved(iface_ea) &&
1367 !eth_addr_is_zero(iface_ea) &&
1368 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1370 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1371 *hw_addr_iface = iface;
1376 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1377 br->name, ETH_ADDR_ARGS(ea));
1379 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1380 *hw_addr_iface = NULL;
1381 VLOG_WARN("bridge %s: using default bridge Ethernet "
1382 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1385 hmapx_destroy(&mirror_output_ports);
1388 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1389 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1390 * an interface on 'br', then that interface must be passed in as
1391 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1392 * 'hw_addr_iface' must be passed in as a null pointer. */
1394 bridge_pick_datapath_id(struct bridge *br,
1395 const uint8_t bridge_ea[ETH_ADDR_LEN],
1396 struct iface *hw_addr_iface)
1399 * The procedure for choosing a bridge MAC address will, in the most
1400 * ordinary case, also choose a unique MAC that we can use as a datapath
1401 * ID. In some special cases, though, multiple bridges will end up with
1402 * the same MAC address. This is OK for the bridges, but it will confuse
1403 * the OpenFlow controller, because each datapath needs a unique datapath
1406 * Datapath IDs must be unique. It is also very desirable that they be
1407 * stable from one run to the next, so that policy set on a datapath
1410 const char *datapath_id;
1413 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1414 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1418 if (!hw_addr_iface) {
1420 * A purely internal bridge, that is, one that has no non-virtual
1421 * network devices on it at all, is difficult because it has no
1422 * natural unique identifier at all.
1424 * When the host is a XenServer, we handle this case by hashing the
1425 * host's UUID with the name of the bridge. Names of bridges are
1426 * persistent across XenServer reboots, although they can be reused if
1427 * an internal network is destroyed and then a new one is later
1428 * created, so this is fairly effective.
1430 * When the host is not a XenServer, we punt by using a random MAC
1431 * address on each run.
1433 const char *host_uuid = xenserver_get_host_uuid();
1435 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1436 dpid = dpid_from_hash(combined, strlen(combined));
1442 return eth_addr_to_uint64(bridge_ea);
1446 dpid_from_hash(const void *data, size_t n)
1448 uint8_t hash[SHA1_DIGEST_SIZE];
1450 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1451 sha1_bytes(data, n, hash);
1452 eth_addr_mark_random(hash);
1453 return eth_addr_to_uint64(hash);
1457 iface_refresh_status(struct iface *iface)
1461 enum netdev_flags flags;
1468 if (iface_is_synthetic(iface)) {
1474 if (!netdev_get_status(iface->netdev, &sh)) {
1476 char **keys, **values;
1478 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1479 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1484 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1487 shash_destroy_free_data(&sh);
1489 error = netdev_get_flags(iface->netdev, &flags);
1491 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1494 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1497 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1499 ovsrec_interface_set_duplex(iface->cfg,
1500 netdev_features_is_full_duplex(current)
1502 /* warning: uint64_t -> int64_t conversion */
1503 bps = netdev_features_to_bps(current);
1504 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1507 ovsrec_interface_set_duplex(iface->cfg, NULL);
1508 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1511 error = netdev_get_mtu(iface->netdev, &mtu);
1514 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1517 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1521 /* Writes 'iface''s CFM statistics to the database. */
1523 iface_refresh_cfm_stats(struct iface *iface)
1525 const struct ovsrec_interface *cfg = iface->cfg;
1527 const uint64_t *rmps;
1530 if (iface_is_synthetic(iface)) {
1534 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1537 bool fault_bool = fault;
1538 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1540 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1543 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1544 iface->ofp_port, &rmps, &n_rmps);
1546 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1549 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1554 iface_refresh_stats(struct iface *iface)
1556 #define IFACE_STATS \
1557 IFACE_STAT(rx_packets, "rx_packets") \
1558 IFACE_STAT(tx_packets, "tx_packets") \
1559 IFACE_STAT(rx_bytes, "rx_bytes") \
1560 IFACE_STAT(tx_bytes, "tx_bytes") \
1561 IFACE_STAT(rx_dropped, "rx_dropped") \
1562 IFACE_STAT(tx_dropped, "tx_dropped") \
1563 IFACE_STAT(rx_errors, "rx_errors") \
1564 IFACE_STAT(tx_errors, "tx_errors") \
1565 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1566 IFACE_STAT(rx_over_errors, "rx_over_err") \
1567 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1568 IFACE_STAT(collisions, "collisions")
1570 #define IFACE_STAT(MEMBER, NAME) NAME,
1571 static char *keys[] = { IFACE_STATS };
1573 int64_t values[ARRAY_SIZE(keys)];
1576 struct netdev_stats stats;
1578 if (iface_is_synthetic(iface)) {
1582 /* Intentionally ignore return value, since errors will set 'stats' to
1583 * all-1s, and we will deal with that correctly below. */
1584 netdev_get_stats(iface->netdev, &stats);
1586 /* Copy statistics into values[] array. */
1588 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1591 assert(i == ARRAY_SIZE(keys));
1593 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1598 br_refresh_stp_status(struct bridge *br)
1600 struct ofproto *ofproto = br->ofproto;
1601 struct ofproto_stp_status status;
1602 char *keys[3], *values[3];
1605 if (ofproto_get_stp_status(ofproto, &status)) {
1609 if (!status.enabled) {
1610 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1614 keys[0] = "stp_bridge_id",
1615 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1616 keys[1] = "stp_designated_root",
1617 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1618 keys[2] = "stp_root_path_cost",
1619 values[2] = xasprintf("%d", status.root_path_cost);
1621 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1623 for (i = 0; i < ARRAY_SIZE(values); i++) {
1629 port_refresh_stp_status(struct port *port)
1631 struct ofproto *ofproto = port->bridge->ofproto;
1632 struct iface *iface;
1633 struct ofproto_port_stp_status status;
1635 char *str_values[4];
1636 int64_t int_values[3];
1639 if (port_is_synthetic(port)) {
1643 /* STP doesn't currently support bonds. */
1644 if (!list_is_singleton(&port->ifaces)) {
1645 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1649 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1651 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1655 if (!status.enabled) {
1656 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1657 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1661 /* Set Status column. */
1662 keys[0] = "stp_port_id";
1663 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1664 keys[1] = "stp_state";
1665 str_values[1] = xstrdup(stp_state_name(status.state));
1666 keys[2] = "stp_sec_in_state";
1667 str_values[2] = xasprintf("%u", status.sec_in_state);
1668 keys[3] = "stp_role";
1669 str_values[3] = xstrdup(stp_role_name(status.role));
1671 ovsrec_port_set_status(port->cfg, keys, str_values,
1672 ARRAY_SIZE(str_values));
1674 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1675 free(str_values[i]);
1678 /* Set Statistics column. */
1679 keys[0] = "stp_tx_count";
1680 int_values[0] = status.tx_count;
1681 keys[1] = "stp_rx_count";
1682 int_values[1] = status.rx_count;
1683 keys[2] = "stp_error_count";
1684 int_values[2] = status.error_count;
1686 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1687 ARRAY_SIZE(int_values));
1691 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1695 /* Use other-config:enable-system-stats by preference. */
1696 enable = get_ovsrec_key_value(cfg->key_other_config,
1697 cfg->value_other_config,
1698 cfg->n_other_config,
1699 "enable-statistics");
1701 return !strcmp(enable, "true");
1704 /* Disable by default. */
1709 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1711 struct ovsdb_datum datum;
1715 if (enable_system_stats(cfg)) {
1716 get_system_stats(&stats);
1719 ovsdb_datum_from_shash(&datum, &stats);
1720 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1724 static inline const char *
1725 nx_role_to_str(enum nx_role role)
1730 case NX_ROLE_MASTER:
1735 return "*** INVALID ROLE ***";
1740 refresh_controller_status(void)
1744 const struct ovsrec_controller *cfg;
1748 /* Accumulate status for controllers on all bridges. */
1749 HMAP_FOR_EACH (br, node, &all_bridges) {
1750 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1753 /* Update each controller in the database with current status. */
1754 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1755 struct ofproto_controller_info *cinfo =
1756 shash_find_data(&info, cfg->target);
1759 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1760 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1761 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1762 (char **) cinfo->pairs.values,
1765 ovsrec_controller_set_is_connected(cfg, false);
1766 ovsrec_controller_set_role(cfg, NULL);
1767 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1771 ofproto_free_ofproto_controller_info(&info);
1775 refresh_cfm_stats(void)
1777 static struct ovsdb_idl_txn *txn = NULL;
1782 txn = ovsdb_idl_txn_create(idl);
1784 HMAP_FOR_EACH (br, node, &all_bridges) {
1785 struct iface *iface;
1787 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1788 iface_refresh_cfm_stats(iface);
1793 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1794 ovsdb_idl_txn_destroy(txn);
1799 /* Performs periodic activity required by bridges that needs to be done with
1800 * the least possible latency.
1802 * It makes sense to call this function a couple of times per poll loop, to
1803 * provide a significant performance boost on some benchmarks with ofprotos
1804 * that use the ofproto-dpif implementation. */
1806 bridge_run_fast(void)
1810 HMAP_FOR_EACH (br, node, &all_bridges) {
1811 ofproto_run_fast(br->ofproto);
1818 const struct ovsrec_open_vswitch *cfg;
1820 bool vlan_splinters_changed;
1821 bool database_changed;
1824 /* (Re)configure if necessary. */
1825 database_changed = ovsdb_idl_run(idl);
1826 if (ovsdb_idl_is_lock_contended(idl)) {
1827 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1828 struct bridge *br, *next_br;
1830 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1831 "disabling this process until it goes away");
1833 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1837 } else if (!ovsdb_idl_has_lock(idl)) {
1840 cfg = ovsrec_open_vswitch_first(idl);
1842 /* Let each bridge do the work that it needs to do. */
1843 HMAP_FOR_EACH (br, node, &all_bridges) {
1844 ofproto_run(br->ofproto);
1847 /* Re-configure SSL. We do this on every trip through the main loop,
1848 * instead of just when the database changes, because the contents of the
1849 * key and certificate files can change without the database changing.
1851 * We do this before bridge_reconfigure() because that function might
1852 * initiate SSL connections and thus requires SSL to be configured. */
1853 if (cfg && cfg->ssl) {
1854 const struct ovsrec_ssl *ssl = cfg->ssl;
1856 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1857 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1860 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1862 vlan_splinters_changed = false;
1863 if (vlan_splinters_enabled_anywhere) {
1864 HMAP_FOR_EACH (br, node, &all_bridges) {
1865 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1866 vlan_splinters_changed = true;
1872 if (database_changed || vlan_splinters_changed) {
1874 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1876 bridge_reconfigure(cfg);
1878 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1879 ovsdb_idl_txn_commit(txn);
1880 ovsdb_idl_txn_destroy(txn); /* XXX */
1882 /* We still need to reconfigure to avoid dangling pointers to
1883 * now-destroyed ovsrec structures inside bridge data. */
1884 static const struct ovsrec_open_vswitch null_cfg;
1886 bridge_reconfigure(&null_cfg);
1890 /* Refresh system and interface stats if necessary. */
1891 if (time_msec() >= stats_timer) {
1893 struct ovsdb_idl_txn *txn;
1895 txn = ovsdb_idl_txn_create(idl);
1896 HMAP_FOR_EACH (br, node, &all_bridges) {
1900 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1901 struct iface *iface;
1903 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1904 iface_refresh_stats(iface);
1905 iface_refresh_status(iface);
1909 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1910 mirror_refresh_stats(m);
1914 refresh_system_stats(cfg);
1915 refresh_controller_status();
1916 ovsdb_idl_txn_commit(txn);
1917 ovsdb_idl_txn_destroy(txn); /* XXX */
1920 stats_timer = time_msec() + STATS_INTERVAL;
1923 if (time_msec() >= db_limiter) {
1924 struct ovsdb_idl_txn *txn;
1926 txn = ovsdb_idl_txn_create(idl);
1927 HMAP_FOR_EACH (br, node, &all_bridges) {
1928 struct iface *iface;
1931 br_refresh_stp_status(br);
1933 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1934 port_refresh_stp_status(port);
1937 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1938 const char *link_state;
1939 int64_t link_resets;
1942 if (iface_is_synthetic(iface)) {
1946 current = ofproto_port_is_lacp_current(br->ofproto,
1950 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1952 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1955 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1956 ovsrec_interface_set_link_state(iface->cfg, link_state);
1958 link_resets = netdev_get_carrier_resets(iface->netdev);
1959 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1963 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
1964 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1966 ovsdb_idl_txn_destroy(txn);
1969 refresh_cfm_stats();
1975 ovsdb_idl_wait(idl);
1976 if (!hmap_is_empty(&all_bridges)) {
1979 HMAP_FOR_EACH (br, node, &all_bridges) {
1980 ofproto_wait(br->ofproto);
1982 poll_timer_wait_until(stats_timer);
1984 if (db_limiter > time_msec()) {
1985 poll_timer_wait_until(db_limiter);
1990 /* QoS unixctl user interface functions. */
1992 struct qos_unixctl_show_cbdata {
1994 struct iface *iface;
1998 qos_unixctl_show_cb(unsigned int queue_id,
1999 const struct shash *details,
2002 struct qos_unixctl_show_cbdata *data = aux;
2003 struct ds *ds = data->ds;
2004 struct iface *iface = data->iface;
2005 struct netdev_queue_stats stats;
2006 struct shash_node *node;
2009 ds_put_cstr(ds, "\n");
2011 ds_put_format(ds, "Queue %u:\n", queue_id);
2013 ds_put_cstr(ds, "Default:\n");
2016 SHASH_FOR_EACH (node, details) {
2017 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2020 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2022 if (stats.tx_packets != UINT64_MAX) {
2023 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2026 if (stats.tx_bytes != UINT64_MAX) {
2027 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2030 if (stats.tx_errors != UINT64_MAX) {
2031 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2034 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2035 queue_id, strerror(error));
2040 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2041 const char *argv[], void *aux OVS_UNUSED)
2043 struct ds ds = DS_EMPTY_INITIALIZER;
2044 struct shash sh = SHASH_INITIALIZER(&sh);
2045 struct iface *iface;
2047 struct shash_node *node;
2048 struct qos_unixctl_show_cbdata data;
2051 iface = iface_find(argv[1]);
2053 unixctl_command_reply(conn, 501, "no such interface");
2057 netdev_get_qos(iface->netdev, &type, &sh);
2059 if (*type != '\0') {
2060 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2062 SHASH_FOR_EACH (node, &sh) {
2063 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2068 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2071 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2073 unixctl_command_reply(conn, 200, ds_cstr(&ds));
2075 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2076 unixctl_command_reply(conn, 501, ds_cstr(&ds));
2079 shash_destroy_free_data(&sh);
2083 /* Bridge reconfiguration functions. */
2085 bridge_create(const struct ovsrec_bridge *br_cfg)
2089 assert(!bridge_lookup(br_cfg->name));
2090 br = xzalloc(sizeof *br);
2092 br->name = xstrdup(br_cfg->name);
2093 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2096 /* Derive the default Ethernet address from the bridge's UUID. This should
2097 * be unique and it will be stable between ovs-vswitchd runs. */
2098 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2099 eth_addr_mark_random(br->default_ea);
2101 hmap_init(&br->ports);
2102 hmap_init(&br->ifaces);
2103 hmap_init(&br->iface_by_name);
2104 hmap_init(&br->mirrors);
2106 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2110 bridge_destroy(struct bridge *br)
2113 struct mirror *mirror, *next_mirror;
2114 struct port *port, *next_port;
2116 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2119 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2120 mirror_destroy(mirror);
2122 hmap_remove(&all_bridges, &br->node);
2123 ofproto_destroy(br->ofproto);
2124 hmap_destroy(&br->ifaces);
2125 hmap_destroy(&br->ports);
2126 hmap_destroy(&br->iface_by_name);
2127 hmap_destroy(&br->mirrors);
2134 static struct bridge *
2135 bridge_lookup(const char *name)
2139 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2140 if (!strcmp(br->name, name)) {
2147 /* Handle requests for a listing of all flows known by the OpenFlow
2148 * stack, including those normally hidden. */
2150 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2151 const char *argv[], void *aux OVS_UNUSED)
2156 br = bridge_lookup(argv[1]);
2158 unixctl_command_reply(conn, 501, "Unknown bridge");
2163 ofproto_get_all_flows(br->ofproto, &results);
2165 unixctl_command_reply(conn, 200, ds_cstr(&results));
2166 ds_destroy(&results);
2169 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2170 * connections and reconnect. If BRIDGE is not specified, then all bridges
2171 * drop their controller connections and reconnect. */
2173 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2174 const char *argv[], void *aux OVS_UNUSED)
2178 br = bridge_lookup(argv[1]);
2180 unixctl_command_reply(conn, 501, "Unknown bridge");
2183 ofproto_reconnect_controllers(br->ofproto);
2185 HMAP_FOR_EACH (br, node, &all_bridges) {
2186 ofproto_reconnect_controllers(br->ofproto);
2189 unixctl_command_reply(conn, 200, NULL);
2193 bridge_get_controllers(const struct bridge *br,
2194 struct ovsrec_controller ***controllersp)
2196 struct ovsrec_controller **controllers;
2197 size_t n_controllers;
2199 controllers = br->cfg->controller;
2200 n_controllers = br->cfg->n_controller;
2202 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2208 *controllersp = controllers;
2210 return n_controllers;
2213 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2214 * those configured in 'br->cfg'. */
2216 bridge_add_del_ports(struct bridge *br,
2217 const unsigned long int *splinter_vlans)
2219 struct port *port, *next;
2220 struct shash_node *node;
2221 struct shash new_ports;
2224 /* Collect new ports. */
2225 shash_init(&new_ports);
2226 for (i = 0; i < br->cfg->n_ports; i++) {
2227 const char *name = br->cfg->ports[i]->name;
2228 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2229 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2233 if (bridge_get_controllers(br, NULL)
2234 && !shash_find(&new_ports, br->name)) {
2235 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2236 br->name, br->name);
2238 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2239 br->synth_local_port.n_interfaces = 1;
2240 br->synth_local_port.name = br->name;
2242 br->synth_local_iface.name = br->name;
2243 br->synth_local_iface.type = "internal";
2245 br->synth_local_ifacep = &br->synth_local_iface;
2247 shash_add(&new_ports, br->name, &br->synth_local_port);
2250 if (splinter_vlans) {
2251 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2254 /* Get rid of deleted ports.
2255 * Get rid of deleted interfaces on ports that still exist. */
2256 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2257 port->cfg = shash_find_data(&new_ports, port->name);
2261 port_del_ifaces(port);
2265 /* Create new ports.
2266 * Add new interfaces to existing ports. */
2267 SHASH_FOR_EACH (node, &new_ports) {
2268 struct port *port = port_lookup(br, node->name);
2270 struct ovsrec_port *cfg = node->data;
2271 port = port_create(br, cfg);
2273 port_add_ifaces(port);
2274 if (list_is_empty(&port->ifaces)) {
2275 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2276 br->name, port->name);
2280 shash_destroy(&new_ports);
2283 /* Initializes 'oc' appropriately as a management service controller for
2286 * The caller must free oc->target when it is no longer needed. */
2288 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2289 struct ofproto_controller *oc)
2291 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2292 oc->max_backoff = 0;
2293 oc->probe_interval = 60;
2294 oc->band = OFPROTO_OUT_OF_BAND;
2296 oc->burst_limit = 0;
2299 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2301 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2302 struct ofproto_controller *oc)
2304 oc->target = c->target;
2305 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2306 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2307 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2308 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2309 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2310 oc->burst_limit = (c->controller_burst_limit
2311 ? *c->controller_burst_limit : 0);
2314 /* Configures the IP stack for 'br''s local interface properly according to the
2315 * configuration in 'c'. */
2317 bridge_configure_local_iface_netdev(struct bridge *br,
2318 struct ovsrec_controller *c)
2320 struct netdev *netdev;
2321 struct in_addr mask, gateway;
2323 struct iface *local_iface;
2326 /* If there's no local interface or no IP address, give up. */
2327 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2328 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2332 /* Bring up the local interface. */
2333 netdev = local_iface->netdev;
2334 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2336 /* Configure the IP address and netmask. */
2337 if (!c->local_netmask
2338 || !inet_aton(c->local_netmask, &mask)
2340 mask.s_addr = guess_netmask(ip.s_addr);
2342 if (!netdev_set_in4(netdev, ip, mask)) {
2343 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2344 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2347 /* Configure the default gateway. */
2348 if (c->local_gateway
2349 && inet_aton(c->local_gateway, &gateway)
2350 && gateway.s_addr) {
2351 if (!netdev_add_router(netdev, gateway)) {
2352 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2353 br->name, IP_ARGS(&gateway.s_addr));
2358 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2359 * in either string are treated as equal to any number of slashes in the other,
2360 * e.g. "x///y" is equal to "x/y". */
2362 equal_pathnames(const char *a, const char *b)
2366 a += strspn(a, "/");
2367 b += strspn(b, "/");
2368 } else if (*a == '\0') {
2379 bridge_configure_remotes(struct bridge *br,
2380 const struct sockaddr_in *managers, size_t n_managers)
2382 const char *disable_ib_str, *queue_id_str;
2383 bool disable_in_band = false;
2386 struct ovsrec_controller **controllers;
2387 size_t n_controllers;
2389 enum ofproto_fail_mode fail_mode;
2391 struct ofproto_controller *ocs;
2395 /* Check if we should disable in-band control on this bridge. */
2396 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2397 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2398 disable_in_band = true;
2401 /* Set OpenFlow queue ID for in-band control. */
2402 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2403 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2404 ofproto_set_in_band_queue(br->ofproto, queue_id);
2406 if (disable_in_band) {
2407 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2409 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2412 n_controllers = bridge_get_controllers(br, &controllers);
2414 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2417 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2418 for (i = 0; i < n_controllers; i++) {
2419 struct ovsrec_controller *c = controllers[i];
2421 if (!strncmp(c->target, "punix:", 6)
2422 || !strncmp(c->target, "unix:", 5)) {
2423 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2426 whitelist = xasprintf("unix:%s/%s.controller",
2427 ovs_rundir(), br->name);
2428 if (!equal_pathnames(c->target, whitelist)) {
2429 /* Prevent remote ovsdb-server users from accessing arbitrary
2430 * Unix domain sockets and overwriting arbitrary local
2432 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2433 "controller \"%s\" due to possibility for remote "
2434 "exploit. Instead, specify whitelisted \"%s\" or "
2435 "connect to \"unix:%s/%s.mgmt\" (which is always "
2436 "available without special configuration).",
2437 br->name, c->target, whitelist,
2438 ovs_rundir(), br->name);
2446 bridge_configure_local_iface_netdev(br, c);
2447 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2448 if (disable_in_band) {
2449 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2454 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2455 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2458 /* Set the fail-mode. */
2459 fail_mode = !br->cfg->fail_mode
2460 || !strcmp(br->cfg->fail_mode, "standalone")
2461 ? OFPROTO_FAIL_STANDALONE
2462 : OFPROTO_FAIL_SECURE;
2463 ofproto_set_fail_mode(br->ofproto, fail_mode);
2465 /* Configure OpenFlow controller connection snooping. */
2466 if (!ofproto_has_snoops(br->ofproto)) {
2470 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2471 ovs_rundir(), br->name));
2472 ofproto_set_snoops(br->ofproto, &snoops);
2473 sset_destroy(&snoops);
2477 /* Port functions. */
2479 static struct port *
2480 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2484 port = xzalloc(sizeof *port);
2486 port->name = xstrdup(cfg->name);
2488 list_init(&port->ifaces);
2490 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2492 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2498 get_port_other_config(const struct ovsrec_port *port, const char *key,
2499 const char *default_value)
2503 value = get_ovsrec_key_value(port->key_other_config,
2504 port->value_other_config,
2505 port->n_other_config, key);
2506 return value ? value : default_value;
2510 get_interface_other_config(const struct ovsrec_interface *iface,
2511 const char *key, const char *default_value)
2515 value = get_ovsrec_key_value(iface->key_other_config,
2516 iface->value_other_config,
2517 iface->n_other_config, key);
2518 return value ? value : default_value;
2521 /* Deletes interfaces from 'port' that are no longer configured for it. */
2523 port_del_ifaces(struct port *port)
2525 struct iface *iface, *next;
2526 struct sset new_ifaces;
2529 /* Collect list of new interfaces. */
2530 sset_init(&new_ifaces);
2531 for (i = 0; i < port->cfg->n_interfaces; i++) {
2532 const char *name = port->cfg->interfaces[i]->name;
2533 const char *type = port->cfg->interfaces[i]->name;
2534 if (strcmp(type, "null")) {
2535 sset_add(&new_ifaces, name);
2539 /* Get rid of deleted interfaces. */
2540 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2541 if (!sset_contains(&new_ifaces, iface->name)) {
2542 iface_destroy(iface);
2546 sset_destroy(&new_ifaces);
2549 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2552 port_add_ifaces(struct port *port)
2554 struct shash new_ifaces;
2555 struct shash_node *node;
2558 /* Collect new ifaces. */
2559 shash_init(&new_ifaces);
2560 for (i = 0; i < port->cfg->n_interfaces; i++) {
2561 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2562 if (strcmp(cfg->type, "null")
2563 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2564 VLOG_WARN("port %s: %s specified twice as port interface",
2565 port->name, cfg->name);
2566 iface_clear_db_record(cfg);
2570 /* Create new interfaces.
2571 * Update interface types and 'cfg' members. */
2572 SHASH_FOR_EACH (node, &new_ifaces) {
2573 const struct ovsrec_interface *cfg = node->data;
2574 const char *iface_name = node->name;
2575 struct iface *iface;
2577 iface = iface_lookup(port->bridge, iface_name);
2579 iface = iface_create(port, cfg);
2584 /* Determine interface type. The local port always has type
2585 * "internal". Other ports take their type from the database and
2586 * default to "system" if none is specified. */
2587 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2588 : cfg->type[0] ? cfg->type
2591 shash_destroy(&new_ifaces);
2595 port_destroy(struct port *port)
2598 struct bridge *br = port->bridge;
2599 struct iface *iface, *next;
2602 ofproto_bundle_unregister(br->ofproto, port);
2605 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2606 iface_destroy(iface);
2609 hmap_remove(&br->ports, &port->hmap_node);
2611 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2618 static struct port *
2619 port_lookup(const struct bridge *br, const char *name)
2623 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2625 if (!strcmp(port->name, name)) {
2633 enable_lacp(struct port *port, bool *activep)
2635 if (!port->cfg->lacp) {
2636 /* XXX when LACP implementation has been sufficiently tested, enable by
2637 * default and make active on bonded ports. */
2639 } else if (!strcmp(port->cfg->lacp, "off")) {
2641 } else if (!strcmp(port->cfg->lacp, "active")) {
2644 } else if (!strcmp(port->cfg->lacp, "passive")) {
2648 VLOG_WARN("port %s: unknown LACP mode %s",
2649 port->name, port->cfg->lacp);
2654 static struct lacp_settings *
2655 port_configure_lacp(struct port *port, struct lacp_settings *s)
2657 const char *lacp_time, *system_id;
2658 long long int custom_time;
2661 if (!enable_lacp(port, &s->active)) {
2665 s->name = port->name;
2667 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2669 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2670 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2671 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2672 " address.", port->name, system_id);
2676 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2679 if (eth_addr_is_zero(s->id)) {
2680 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2684 /* Prefer bondable links if unspecified. */
2685 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2687 s->priority = (priority > 0 && priority <= UINT16_MAX
2689 : UINT16_MAX - !list_is_short(&port->ifaces));
2691 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2695 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2696 custom_time = atoi(lacp_time);
2697 if (!strcmp(lacp_time, "fast")) {
2698 s->lacp_time = LACP_TIME_FAST;
2699 } else if (!strcmp(lacp_time, "slow")) {
2700 s->lacp_time = LACP_TIME_SLOW;
2701 } else if (custom_time > 0) {
2702 s->lacp_time = LACP_TIME_CUSTOM;
2703 s->custom_time = custom_time;
2705 s->lacp_time = LACP_TIME_SLOW;
2712 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2714 int priority, portid, key;
2716 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2717 priority = atoi(get_interface_other_config(iface->cfg,
2718 "lacp-port-priority", "0"));
2719 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2722 if (portid <= 0 || portid > UINT16_MAX) {
2723 portid = iface->ofp_port;
2726 if (priority <= 0 || priority > UINT16_MAX) {
2727 priority = UINT16_MAX;
2730 if (key < 0 || key > UINT16_MAX) {
2734 s->name = iface->name;
2736 s->priority = priority;
2741 port_configure_bond(struct port *port, struct bond_settings *s,
2742 uint32_t *bond_stable_ids)
2744 const char *detect_s;
2745 struct iface *iface;
2746 int miimon_interval;
2749 s->name = port->name;
2751 if (port->cfg->bond_mode) {
2752 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2753 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2754 port->name, port->cfg->bond_mode,
2755 bond_mode_to_string(s->balance));
2758 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2760 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2761 * active-backup. At some point we should remove this warning. */
2762 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2763 " in previous versions, the default bond_mode was"
2764 " balance-slb", port->name,
2765 bond_mode_to_string(s->balance));
2767 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2768 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2769 "please use another bond type or disable flood_vlans",
2773 miimon_interval = atoi(get_port_other_config(port->cfg,
2774 "bond-miimon-interval", "0"));
2775 if (miimon_interval <= 0) {
2776 miimon_interval = 200;
2779 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2780 if (!strcmp(detect_s, "carrier")) {
2781 miimon_interval = 0;
2782 } else if (strcmp(detect_s, "miimon")) {
2783 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2784 "defaulting to carrier", port->name, detect_s);
2785 miimon_interval = 0;
2788 s->up_delay = MAX(0, port->cfg->bond_updelay);
2789 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2790 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2791 s->rebalance_interval = atoi(
2792 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2793 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2794 s->rebalance_interval = 1000;
2797 s->fake_iface = port->cfg->bond_fake_iface;
2800 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2801 long long stable_id;
2803 stable_id = atoll(get_interface_other_config(iface->cfg,
2804 "bond-stable-id", "0"));
2805 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2806 stable_id = iface->ofp_port;
2808 bond_stable_ids[i++] = stable_id;
2810 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2814 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2815 * instead of obtaining it from the database. */
2817 port_is_synthetic(const struct port *port)
2819 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2822 /* Interface functions. */
2824 static struct iface *
2825 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2827 struct bridge *br = port->bridge;
2828 struct iface *iface;
2829 char *name = if_cfg->name;
2831 iface = xzalloc(sizeof *iface);
2833 iface->name = xstrdup(name);
2834 iface->ofp_port = -1;
2835 iface->tag = tag_create_random();
2836 iface->netdev = NULL;
2837 iface->cfg = if_cfg;
2839 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2841 list_push_back(&port->ifaces, &iface->port_elem);
2843 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2849 iface_destroy(struct iface *iface)
2852 struct port *port = iface->port;
2853 struct bridge *br = port->bridge;
2855 if (br->ofproto && iface->ofp_port >= 0) {
2856 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2859 if (iface->ofp_port >= 0) {
2860 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2863 list_remove(&iface->port_elem);
2864 hmap_remove(&br->iface_by_name, &iface->name_node);
2866 netdev_close(iface->netdev);
2873 static struct iface *
2874 iface_lookup(const struct bridge *br, const char *name)
2876 struct iface *iface;
2878 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2879 &br->iface_by_name) {
2880 if (!strcmp(iface->name, name)) {
2888 static struct iface *
2889 iface_find(const char *name)
2891 const struct bridge *br;
2893 HMAP_FOR_EACH (br, node, &all_bridges) {
2894 struct iface *iface = iface_lookup(br, name);
2903 static struct iface *
2904 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2906 struct iface *iface;
2908 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2909 hash_int(ofp_port, 0), &br->ifaces) {
2910 if (iface->ofp_port == ofp_port) {
2917 /* Set Ethernet address of 'iface', if one is specified in the configuration
2920 iface_set_mac(struct iface *iface)
2922 uint8_t ea[ETH_ADDR_LEN];
2924 if (!strcmp(iface->type, "internal")
2925 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2926 if (iface->ofp_port == OFPP_LOCAL) {
2927 VLOG_ERR("interface %s: ignoring mac in Interface record "
2928 "(use Bridge record to set local port's mac)",
2930 } else if (eth_addr_is_multicast(ea)) {
2931 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2934 int error = netdev_set_etheraddr(iface->netdev, ea);
2936 VLOG_ERR("interface %s: setting MAC failed (%s)",
2937 iface->name, strerror(error));
2943 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2945 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2947 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2948 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2952 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
2953 * sets the "ofport" field to -1.
2955 * This is appropriate when 'if_cfg''s interface cannot be created or is
2956 * otherwise invalid. */
2958 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
2960 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2961 iface_set_ofport(if_cfg, -1);
2962 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
2963 ovsrec_interface_set_admin_state(if_cfg, NULL);
2964 ovsrec_interface_set_duplex(if_cfg, NULL);
2965 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
2966 ovsrec_interface_set_link_state(if_cfg, NULL);
2967 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
2968 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
2969 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
2970 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
2971 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
2975 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2977 * The value strings in '*shash' are taken directly from values[], not copied,
2978 * so the caller should not modify or free them. */
2980 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2981 struct shash *shash)
2986 for (i = 0; i < n; i++) {
2987 shash_add(shash, keys[i], values[i]);
2991 /* Creates 'keys' and 'values' arrays from 'shash'.
2993 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2994 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2995 * are populated with with strings taken directly from 'shash' and thus have
2996 * the same ownership of the key-value pairs in shash.
2999 shash_to_ovs_idl_map(struct shash *shash,
3000 char ***keys, char ***values, size_t *n)
3004 struct shash_node *sn;
3006 count = shash_count(shash);
3008 k = xmalloc(count * sizeof *k);
3009 v = xmalloc(count * sizeof *v);
3012 SHASH_FOR_EACH(sn, shash) {
3023 struct iface_delete_queues_cbdata {
3024 struct netdev *netdev;
3025 const struct ovsdb_datum *queues;
3029 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3031 union ovsdb_atom atom;
3033 atom.integer = target;
3034 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3038 iface_delete_queues(unsigned int queue_id,
3039 const struct shash *details OVS_UNUSED, void *cbdata_)
3041 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3043 if (!queue_ids_include(cbdata->queues, queue_id)) {
3044 netdev_delete_queue(cbdata->netdev, queue_id);
3049 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3051 struct ofpbuf queues_buf;
3053 ofpbuf_init(&queues_buf, 0);
3055 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3056 netdev_set_qos(iface->netdev, NULL, NULL);
3058 struct iface_delete_queues_cbdata cbdata;
3059 struct shash details;
3063 /* Configure top-level Qos for 'iface'. */
3064 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3065 qos->n_other_config, &details);
3066 netdev_set_qos(iface->netdev, qos->type, &details);
3067 shash_destroy(&details);
3069 /* Deconfigure queues that were deleted. */
3070 cbdata.netdev = iface->netdev;
3071 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3073 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3075 /* Configure queues for 'iface'. */
3077 for (i = 0; i < qos->n_queues; i++) {
3078 const struct ovsrec_queue *queue = qos->value_queues[i];
3079 unsigned int queue_id = qos->key_queues[i];
3081 if (queue_id == 0) {
3085 if (queue->n_dscp == 1) {
3086 struct ofproto_port_queue *port_queue;
3088 port_queue = ofpbuf_put_uninit(&queues_buf,
3089 sizeof *port_queue);
3090 port_queue->queue = queue_id;
3091 port_queue->dscp = queue->dscp[0];
3094 shash_from_ovs_idl_map(queue->key_other_config,
3095 queue->value_other_config,
3096 queue->n_other_config, &details);
3097 netdev_set_queue(iface->netdev, queue_id, &details);
3098 shash_destroy(&details);
3101 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3102 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
3103 "queue (queue 0). Packets not directed to a "
3104 "correctly configured queue may be dropped.",
3109 if (iface->ofp_port >= 0) {
3110 const struct ofproto_port_queue *port_queues = queues_buf.data;
3111 size_t n_queues = queues_buf.size / sizeof *port_queues;
3113 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3114 port_queues, n_queues);
3117 netdev_set_policing(iface->netdev,
3118 iface->cfg->ingress_policing_rate,
3119 iface->cfg->ingress_policing_burst);
3121 ofpbuf_uninit(&queues_buf);
3125 iface_configure_cfm(struct iface *iface)
3127 const struct ovsrec_interface *cfg = iface->cfg;
3128 const char *extended_str, *opstate_str;
3129 struct cfm_settings s;
3131 if (!cfg->n_cfm_mpid) {
3132 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3136 s.mpid = *cfg->cfm_mpid;
3137 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3139 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3141 if (s.interval <= 0) {
3145 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3147 s.extended = !strcasecmp("true", extended_str);
3149 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3150 s.opup = !strcasecmp("up", opstate_str);
3152 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3155 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3156 * instead of obtaining it from the database. */
3158 iface_is_synthetic(const struct iface *iface)
3160 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3163 /* Port mirroring. */
3165 static struct mirror *
3166 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3170 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3171 if (uuid_equals(uuid, &m->uuid)) {
3179 bridge_configure_mirrors(struct bridge *br)
3181 const struct ovsdb_datum *mc;
3182 unsigned long *flood_vlans;
3183 struct mirror *m, *next;
3186 /* Get rid of deleted mirrors. */
3187 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3188 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3189 union ovsdb_atom atom;
3191 atom.uuid = m->uuid;
3192 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3197 /* Add new mirrors and reconfigure existing ones. */
3198 for (i = 0; i < br->cfg->n_mirrors; i++) {
3199 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3200 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3202 m = mirror_create(br, cfg);
3205 if (!mirror_configure(m)) {
3210 /* Update flooded vlans (for RSPAN). */
3211 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3212 br->cfg->n_flood_vlans);
3213 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3214 bitmap_free(flood_vlans);
3217 static struct mirror *
3218 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3222 m = xzalloc(sizeof *m);
3223 m->uuid = cfg->header_.uuid;
3224 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3226 m->name = xstrdup(cfg->name);
3232 mirror_destroy(struct mirror *m)
3235 struct bridge *br = m->bridge;
3238 ofproto_mirror_unregister(br->ofproto, m);
3241 hmap_remove(&br->mirrors, &m->hmap_node);
3248 mirror_collect_ports(struct mirror *m,
3249 struct ovsrec_port **in_ports, int n_in_ports,
3250 void ***out_portsp, size_t *n_out_portsp)
3252 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3253 size_t n_out_ports = 0;
3256 for (i = 0; i < n_in_ports; i++) {
3257 const char *name = in_ports[i]->name;
3258 struct port *port = port_lookup(m->bridge, name);
3260 out_ports[n_out_ports++] = port;
3262 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3263 "port %s", m->bridge->name, m->name, name);
3266 *out_portsp = out_ports;
3267 *n_out_portsp = n_out_ports;
3271 mirror_configure(struct mirror *m)
3273 const struct ovsrec_mirror *cfg = m->cfg;
3274 struct ofproto_mirror_settings s;
3277 if (strcmp(cfg->name, m->name)) {
3279 m->name = xstrdup(cfg->name);
3283 /* Get output port or VLAN. */
3284 if (cfg->output_port) {
3285 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3286 if (!s.out_bundle) {
3287 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3288 m->bridge->name, m->name);
3291 s.out_vlan = UINT16_MAX;
3293 if (cfg->output_vlan) {
3294 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3295 "output vlan; ignoring output vlan",
3296 m->bridge->name, m->name);
3298 } else if (cfg->output_vlan) {
3299 /* The database should prevent invalid VLAN values. */
3300 s.out_bundle = NULL;
3301 s.out_vlan = *cfg->output_vlan;
3303 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3304 m->bridge->name, m->name);
3308 /* Get port selection. */
3309 if (cfg->select_all) {
3310 size_t n_ports = hmap_count(&m->bridge->ports);
3311 void **ports = xmalloc(n_ports * sizeof *ports);
3316 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3326 /* Get ports, dropping ports that don't exist.
3327 * The IDL ensures that there are no duplicates. */
3328 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3329 &s.srcs, &s.n_srcs);
3330 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3331 &s.dsts, &s.n_dsts);
3334 /* Get VLAN selection. */
3335 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3338 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3341 if (s.srcs != s.dsts) {
3350 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3352 * This is deprecated. It is only for compatibility with broken device drivers
3353 * in old versions of Linux that do not properly support VLANs when VLAN
3354 * devices are not used. When broken device drivers are no longer in
3355 * widespread use, we will delete these interfaces. */
3357 static void **blocks;
3358 static size_t n_blocks, allocated_blocks;
3360 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3361 * VLAN splinters are reconfigured. */
3363 register_block(void *block)
3365 if (n_blocks >= allocated_blocks) {
3366 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3368 blocks[n_blocks++] = block;
3371 /* Frees all of the blocks registered with register_block(). */
3373 free_registered_blocks(void)
3377 for (i = 0; i < n_blocks; i++) {
3383 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3386 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3390 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3391 return !strcmp(value, "true");
3394 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3397 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3398 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3399 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3402 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3403 * use, returns NULL.
3405 * Updates 'vlan_splinters_enabled_anywhere'. */
3406 static unsigned long int *
3407 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3409 unsigned long int *splinter_vlans;
3410 struct sset splinter_ifaces;
3411 const char *real_dev_name;
3412 struct shash *real_devs;
3413 struct shash_node *node;
3417 /* Free space allocated for synthesized ports and interfaces, since we're
3418 * in the process of reconstructing all of them. */
3419 free_registered_blocks();
3421 splinter_vlans = bitmap_allocate(4096);
3422 sset_init(&splinter_ifaces);
3423 vlan_splinters_enabled_anywhere = false;
3424 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3425 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3428 for (j = 0; j < br_cfg->n_ports; j++) {
3429 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3432 for (k = 0; k < port_cfg->n_interfaces; k++) {
3433 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3435 if (vlan_splinters_is_enabled(iface_cfg)) {
3436 vlan_splinters_enabled_anywhere = true;
3437 sset_add(&splinter_ifaces, iface_cfg->name);
3438 vlan_bitmap_from_array__(port_cfg->trunks,
3444 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3445 bitmap_set1(splinter_vlans, *port_cfg->tag);
3450 if (!vlan_splinters_enabled_anywhere) {
3451 free(splinter_vlans);
3452 sset_destroy(&splinter_ifaces);
3456 HMAP_FOR_EACH (br, node, &all_bridges) {
3458 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3462 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3463 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3464 * device to be created for it. */
3465 bitmap_set0(splinter_vlans, 0);
3466 bitmap_set0(splinter_vlans, 4095);
3468 /* Delete all VLAN devices that we don't need. */
3470 real_devs = vlandev_get_real_devs();
3471 SHASH_FOR_EACH (node, real_devs) {
3472 const struct vlan_real_dev *real_dev = node->data;
3473 const struct vlan_dev *vlan_dev;
3474 bool real_dev_has_splinters;
3476 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3478 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3479 if (!real_dev_has_splinters
3480 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3481 struct netdev *netdev;
3483 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3484 if (!netdev_get_in4(netdev, NULL, NULL) ||
3485 !netdev_get_in6(netdev, NULL)) {
3486 vlandev_del(vlan_dev->name);
3488 /* It has an IP address configured, so we don't own
3489 * it. Don't delete it. */
3491 netdev_close(netdev);
3498 /* Add all VLAN devices that we need. */
3499 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3502 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3503 if (!vlandev_get_name(real_dev_name, vid)) {
3504 vlandev_add(real_dev_name, vid);
3511 sset_destroy(&splinter_ifaces);
3513 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3514 free(splinter_vlans);
3517 return splinter_vlans;
3520 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3523 configure_splinter_port(struct port *port)
3525 struct ofproto *ofproto = port->bridge->ofproto;
3526 uint16_t realdev_ofp_port;
3527 const char *realdev_name;
3528 struct iface *vlandev, *realdev;
3530 ofproto_bundle_unregister(port->bridge->ofproto, port);
3532 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3535 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3536 realdev = iface_lookup(port->bridge, realdev_name);
3537 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3539 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3543 static struct ovsrec_port *
3544 synthesize_splinter_port(const char *real_dev_name,
3545 const char *vlan_dev_name, int vid)
3547 struct ovsrec_interface *iface;
3548 struct ovsrec_port *port;
3550 iface = xzalloc(sizeof *iface);
3551 iface->name = xstrdup(vlan_dev_name);
3552 iface->type = "system";
3554 port = xzalloc(sizeof *port);
3555 port->interfaces = xmemdup(&iface, sizeof iface);
3556 port->n_interfaces = 1;
3557 port->name = xstrdup(vlan_dev_name);
3558 port->vlan_mode = "splinter";
3559 port->tag = xmalloc(sizeof *port->tag);
3561 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3562 port->key_other_config[0] = "realdev";
3563 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3564 port->value_other_config[0] = xstrdup(real_dev_name);
3565 port->n_other_config = 1;
3567 register_block(iface);
3568 register_block(iface->name);
3569 register_block(port);
3570 register_block(port->interfaces);
3571 register_block(port->name);
3572 register_block(port->tag);
3573 register_block(port->key_other_config);
3574 register_block(port->value_other_config);
3575 register_block(port->value_other_config[0]);
3580 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3581 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3582 * 1-bit in the 'splinter_vlans' bitmap. */
3584 add_vlan_splinter_ports(struct bridge *br,
3585 const unsigned long int *splinter_vlans,
3586 struct shash *ports)
3590 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3591 * we're modifying 'ports'. */
3592 for (i = 0; i < br->cfg->n_ports; i++) {
3593 const char *name = br->cfg->ports[i]->name;
3594 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3597 for (j = 0; j < port_cfg->n_interfaces; j++) {
3598 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3600 if (vlan_splinters_is_enabled(iface_cfg)) {
3601 const char *real_dev_name;
3604 real_dev_name = iface_cfg->name;
3605 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3606 const char *vlan_dev_name;
3608 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3610 && !shash_find(ports, vlan_dev_name)) {
3611 shash_add(ports, vlan_dev_name,
3612 synthesize_splinter_port(
3613 real_dev_name, vlan_dev_name, vid));
3622 mirror_refresh_stats(struct mirror *m)
3624 struct ofproto *ofproto = m->bridge->ofproto;
3625 uint64_t tx_packets, tx_bytes;
3628 size_t stat_cnt = 0;
3630 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3631 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3635 if (tx_packets != UINT64_MAX) {
3636 keys[stat_cnt] = "tx_packets";
3637 values[stat_cnt] = tx_packets;
3640 if (tx_bytes != UINT64_MAX) {
3641 keys[stat_cnt] = "tx_bytes";
3642 values[stat_cnt] = tx_bytes;
3646 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);