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"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
44 #include "socket-util.h"
46 #include "stream-ssl.h"
48 #include "system-stats.h"
53 #include "vswitchd/vswitch-idl.h"
54 #include "xenserver.h"
56 #include "sflow_api.h"
57 #include "vlan-bitmap.h"
59 VLOG_DEFINE_THIS_MODULE(bridge);
61 COVERAGE_DEFINE(bridge_reconfigure);
64 /* These members are always valid. */
65 struct list port_elem; /* Element in struct port's "ifaces" list. */
66 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
67 struct port *port; /* Containing port. */
68 char *name; /* Host network device name. */
69 tag_type tag; /* Tag associated with this interface. */
71 /* These members are valid only after bridge_reconfigure() causes them to
73 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
74 int ofp_port; /* OpenFlow port number, -1 if unknown. */
75 struct netdev *netdev; /* Network device. */
76 const char *type; /* Usually same as cfg->type. */
77 const struct ovsrec_interface *cfg;
78 bool need_refresh; /* Refresh iface after create. */
82 struct uuid uuid; /* UUID of this "mirror" record in database. */
83 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
84 struct bridge *bridge;
86 const struct ovsrec_mirror *cfg;
90 struct bridge *bridge;
91 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
94 const struct ovsrec_port *cfg;
96 /* An ordinary bridge port has 1 interface.
97 * A bridge port for bonding has at least 2 interfaces. */
98 struct list ifaces; /* List of "struct iface"s. */
102 struct hmap_node node; /* In 'all_bridges'. */
103 char *name; /* User-specified arbitrary name. */
104 char *type; /* Datapath type. */
105 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
106 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
107 const struct ovsrec_bridge *cfg;
109 /* OpenFlow switch processing. */
110 struct ofproto *ofproto; /* OpenFlow switch. */
113 struct hmap ports; /* "struct port"s indexed by name. */
114 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
115 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
117 /* Port mirroring. */
118 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
120 /* Synthetic local port if necessary. */
121 struct ovsrec_port synth_local_port;
122 struct ovsrec_interface synth_local_iface;
123 struct ovsrec_interface *synth_local_ifacep;
126 /* All bridges, indexed by name. */
127 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
129 /* OVSDB IDL used to obtain configuration. */
130 static struct ovsdb_idl *idl;
132 /* Each time this timer expires, the bridge fetches systems and interface
133 * statistics and pushes them into the database. */
134 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
135 static long long int stats_timer = LLONG_MIN;
137 /* Stores the time after which rate limited statistics may be written to the
138 * database. Only updated when changes to the database require rate limiting.
140 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
141 static long long int db_limiter = LLONG_MIN;
143 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
144 * expensive. This can cause bridge_reconfigure() to take a long time during
145 * which no other work can be done. To deal with this problem, we limit port
146 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
147 * call to bridge_reconfigure(). If there is more work to do after the limit
148 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
149 * This allows the rest of the code to catch up on important things like
150 * forwarding packets. */
151 #define OFP_PORT_ACTION_WINDOW 10
152 static bool need_reconfigure = false;
154 static void add_del_bridges(const struct ovsrec_open_vswitch *);
155 static void bridge_del_ofprotos(void);
156 static bool bridge_add_ofprotos(struct bridge *);
157 static void bridge_create(const struct ovsrec_bridge *);
158 static void bridge_destroy(struct bridge *);
159 static struct bridge *bridge_lookup(const char *name);
160 static unixctl_cb_func bridge_unixctl_dump_flows;
161 static unixctl_cb_func bridge_unixctl_reconnect;
162 static size_t bridge_get_controllers(const struct bridge *br,
163 struct ovsrec_controller ***controllersp);
164 static void bridge_add_del_ports(struct bridge *,
165 const unsigned long int *splinter_vlans);
166 static void bridge_add_ofproto_ports(struct bridge *,
167 long long int *port_action_timer);
168 static void bridge_del_ofproto_ports(struct bridge *,
169 long long int *port_action_timer);
170 static void bridge_refresh_ofp_port(struct bridge *);
171 static void bridge_configure_datapath_id(struct bridge *);
172 static void bridge_configure_flow_eviction_threshold(struct bridge *);
173 static void bridge_configure_netflow(struct bridge *);
174 static void bridge_configure_forward_bpdu(struct bridge *);
175 static void bridge_configure_mac_idle_time(struct bridge *);
176 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
177 static void bridge_configure_stp(struct bridge *);
178 static void bridge_configure_tables(struct bridge *);
179 static void bridge_configure_remotes(struct bridge *,
180 const struct sockaddr_in *managers,
182 static void bridge_pick_local_hw_addr(struct bridge *,
183 uint8_t ea[ETH_ADDR_LEN],
184 struct iface **hw_addr_iface);
185 static uint64_t bridge_pick_datapath_id(struct bridge *,
186 const uint8_t bridge_ea[ETH_ADDR_LEN],
187 struct iface *hw_addr_iface);
188 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
190 static const char *get_port_other_config(const struct ovsrec_port *,
192 const char *default_value);
193 static uint64_t dpid_from_hash(const void *, size_t nbytes);
194 static bool bridge_has_bond_fake_iface(const struct bridge *,
196 static bool port_is_bond_fake_iface(const struct port *);
198 static unixctl_cb_func qos_unixctl_show;
200 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
201 static void port_add_ifaces(struct port *);
202 static void port_del_ifaces(struct port *);
203 static void port_destroy(struct port *);
204 static struct port *port_lookup(const struct bridge *, const char *name);
205 static void port_configure(struct port *);
206 static struct lacp_settings *port_configure_lacp(struct port *,
207 struct lacp_settings *);
208 static void port_configure_bond(struct port *, struct bond_settings *,
209 uint32_t *bond_stable_ids);
210 static bool port_is_synthetic(const struct port *);
212 static void bridge_configure_mirrors(struct bridge *);
213 static struct mirror *mirror_create(struct bridge *,
214 const struct ovsrec_mirror *);
215 static void mirror_destroy(struct mirror *);
216 static bool mirror_configure(struct mirror *);
217 static void mirror_refresh_stats(struct mirror *);
219 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
220 static struct iface *iface_create(struct port *port,
221 const struct ovsrec_interface *if_cfg);
222 static void iface_destroy(struct iface *);
223 static struct iface *iface_lookup(const struct bridge *, const char *name);
224 static struct iface *iface_find(const char *name);
225 static struct iface *iface_from_ofp_port(const struct bridge *,
227 static void iface_set_mac(struct iface *);
228 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
229 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
230 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
231 static void iface_configure_cfm(struct iface *);
232 static void iface_refresh_cfm_stats(struct iface *);
233 static void iface_refresh_stats(struct iface *);
234 static void iface_refresh_status(struct iface *);
235 static bool iface_is_synthetic(const struct iface *);
236 static const char *get_interface_other_config(const struct ovsrec_interface *,
238 const char *default_value);
240 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
242 static void shash_to_ovs_idl_map(struct shash *,
243 char ***keys, char ***values, size_t *n);
245 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
247 * This is deprecated. It is only for compatibility with broken device drivers
248 * in old versions of Linux that do not properly support VLANs when VLAN
249 * devices are not used. When broken device drivers are no longer in
250 * widespread use, we will delete these interfaces. */
252 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
253 static bool vlan_splinters_enabled_anywhere;
255 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
256 static unsigned long int *collect_splinter_vlans(
257 const struct ovsrec_open_vswitch *);
258 static void configure_splinter_port(struct port *);
259 static void add_vlan_splinter_ports(struct bridge *,
260 const unsigned long int *splinter_vlans,
261 struct shash *ports);
263 /* Public functions. */
265 /* Initializes the bridge module, configuring it to obtain its configuration
266 * from an OVSDB server accessed over 'remote', which should be a string in a
267 * form acceptable to ovsdb_idl_create(). */
269 bridge_init(const char *remote)
271 /* Create connection to database. */
272 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
273 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
275 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
276 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
277 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
278 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
279 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
280 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
281 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
283 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
284 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
285 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
287 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
288 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
289 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
290 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
292 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
293 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
294 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
295 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
296 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
297 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
305 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
307 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
308 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
309 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
310 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
312 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
314 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
316 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
317 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
319 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
321 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
323 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
324 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
325 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
326 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
327 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
329 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
331 /* Register unixctl commands. */
332 unixctl_command_register("qos/show", "interface", 1, 1,
333 qos_unixctl_show, NULL);
334 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
335 bridge_unixctl_dump_flows, NULL);
336 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
337 bridge_unixctl_reconnect, NULL);
347 struct bridge *br, *next_br;
349 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
352 ovsdb_idl_destroy(idl);
355 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
356 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
357 * responsible for freeing '*managersp' (with free()).
359 * You may be asking yourself "why does ovs-vswitchd care?", because
360 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
361 * should not be and in fact is not directly involved in that. But
362 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
363 * it has to tell in-band control where the managers are to enable that.
364 * (Thus, only managers connected in-band are collected.)
367 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
368 struct sockaddr_in **managersp, size_t *n_managersp)
370 struct sockaddr_in *managers = NULL;
371 size_t n_managers = 0;
375 /* Collect all of the potential targets from the "targets" columns of the
376 * rows pointed to by "manager_options", excluding any that are
379 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
380 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
382 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
383 sset_find_and_delete(&targets, m->target);
385 sset_add(&targets, m->target);
389 /* Now extract the targets' IP addresses. */
390 if (!sset_is_empty(&targets)) {
393 managers = xmalloc(sset_count(&targets) * sizeof *managers);
394 SSET_FOR_EACH (target, &targets) {
395 struct sockaddr_in *sin = &managers[n_managers];
397 if (stream_parse_target_with_default_ports(target,
405 sset_destroy(&targets);
407 *managersp = managers;
408 *n_managersp = n_managers;
412 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
414 unsigned long int *splinter_vlans;
415 long long int port_action_timer;
416 struct sockaddr_in *managers;
417 struct bridge *br, *next;
418 int sflow_bridge_number;
421 COVERAGE_INC(bridge_reconfigure);
423 port_action_timer = LLONG_MAX;
424 need_reconfigure = false;
426 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
427 * iface"s according to 'ovs_cfg', with only very minimal configuration
430 * This is mostly an update to bridge data structures. Very little is
431 * pushed down to ofproto or lower layers. */
432 add_del_bridges(ovs_cfg);
433 splinter_vlans = collect_splinter_vlans(ovs_cfg);
434 HMAP_FOR_EACH (br, node, &all_bridges) {
435 bridge_add_del_ports(br, splinter_vlans);
437 free(splinter_vlans);
439 /* Delete all datapaths and datapath ports that are no longer configured.
441 * The kernel will reject any attempt to add a given port to a datapath if
442 * that port already belongs to a different datapath, so we must do all
443 * port deletions before any port additions. A datapath always has a
444 * "local port" so we must delete not-configured datapaths too. */
445 bridge_del_ofprotos();
446 HMAP_FOR_EACH (br, node, &all_bridges) {
448 bridge_del_ofproto_ports(br, &port_action_timer);
452 /* Create datapaths and datapath ports that are missing.
454 * After this is done, we have our final set of bridges, ports, and
455 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
456 * has at least one iface, every "struct iface" has a valid ofp_port and
458 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
460 if (bridge_add_ofprotos(br)) {
461 bridge_del_ofproto_ports(br, &port_action_timer);
467 HMAP_FOR_EACH (br, node, &all_bridges) {
468 bridge_refresh_ofp_port(br);
469 bridge_add_ofproto_ports(br, &port_action_timer);
472 /* Complete the configuration. */
473 sflow_bridge_number = 0;
474 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
475 HMAP_FOR_EACH (br, node, &all_bridges) {
478 /* We need the datapath ID early to allow LACP ports to use it as the
479 * default system ID. */
480 bridge_configure_datapath_id(br);
482 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
485 port_configure(port);
487 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
488 iface_configure_cfm(iface);
489 iface_configure_qos(iface, port->cfg->qos);
490 iface_set_mac(iface);
493 bridge_configure_mirrors(br);
494 bridge_configure_flow_eviction_threshold(br);
495 bridge_configure_forward_bpdu(br);
496 bridge_configure_mac_idle_time(br);
497 bridge_configure_remotes(br, managers, n_managers);
498 bridge_configure_netflow(br);
499 bridge_configure_sflow(br, &sflow_bridge_number);
500 bridge_configure_stp(br);
501 bridge_configure_tables(br);
505 if (!need_reconfigure) {
506 /* ovs-vswitchd has completed initialization, so allow the process that
507 * forked us to exit successfully. */
508 daemonize_complete();
512 /* Iterate over all ofprotos and delete any of them that do not have a
513 * configured bridge or that are the wrong type. */
515 bridge_del_ofprotos(void)
523 ofproto_enumerate_types(&types);
524 SSET_FOR_EACH (type, &types) {
527 ofproto_enumerate_names(type, &names);
528 SSET_FOR_EACH (name, &names) {
529 struct bridge *br = bridge_lookup(name);
530 if (!br || strcmp(type, br->type)) {
531 ofproto_delete(name, type);
535 sset_destroy(&names);
536 sset_destroy(&types);
540 bridge_add_ofprotos(struct bridge *br)
542 int error = ofproto_create(br->name, br->type, &br->ofproto);
544 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
551 port_configure(struct port *port)
553 const struct ovsrec_port *cfg = port->cfg;
554 struct bond_settings bond_settings;
555 struct lacp_settings lacp_settings;
556 struct ofproto_bundle_settings s;
559 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
560 configure_splinter_port(port);
569 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
570 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
571 s.slaves[s.n_slaves++] = iface->ofp_port;
576 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
580 /* Get VLAN trunks. */
583 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
587 if (cfg->vlan_mode) {
588 if (!strcmp(cfg->vlan_mode, "access")) {
589 s.vlan_mode = PORT_VLAN_ACCESS;
590 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
591 s.vlan_mode = PORT_VLAN_TRUNK;
592 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
593 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
594 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
595 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
597 /* This "can't happen" because ovsdb-server should prevent it. */
598 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
599 s.vlan_mode = PORT_VLAN_TRUNK;
603 s.vlan_mode = PORT_VLAN_ACCESS;
605 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
609 s.vlan_mode = PORT_VLAN_TRUNK;
612 s.use_priority_tags = !strcmp("true", get_port_other_config(
613 cfg, "priority-tags", ""));
615 /* Get LACP settings. */
616 s.lacp = port_configure_lacp(port, &lacp_settings);
620 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
621 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
622 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
625 s.lacp_slaves = NULL;
628 /* Get bond settings. */
629 if (s.n_slaves > 1) {
630 s.bond = &bond_settings;
631 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
632 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
635 s.bond_stable_ids = NULL;
637 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
638 netdev_set_miimon_interval(iface->netdev, 0);
643 ofproto_bundle_register(port->bridge->ofproto, port, &s);
649 free(s.bond_stable_ids);
652 /* Pick local port hardware address and datapath ID for 'br'. */
654 bridge_configure_datapath_id(struct bridge *br)
656 uint8_t ea[ETH_ADDR_LEN];
658 struct iface *local_iface;
659 struct iface *hw_addr_iface;
662 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
663 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
665 int error = netdev_set_etheraddr(local_iface->netdev, ea);
667 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
668 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
669 "Ethernet address: %s",
670 br->name, strerror(error));
673 memcpy(br->ea, ea, ETH_ADDR_LEN);
675 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
676 ofproto_set_datapath_id(br->ofproto, dpid);
678 dpid_string = xasprintf("%016"PRIx64, dpid);
679 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
683 /* Set NetFlow configuration on 'br'. */
685 bridge_configure_netflow(struct bridge *br)
687 struct ovsrec_netflow *cfg = br->cfg->netflow;
688 struct netflow_options opts;
691 ofproto_set_netflow(br->ofproto, NULL);
695 memset(&opts, 0, sizeof opts);
697 /* Get default NetFlow configuration from datapath.
698 * Apply overrides from 'cfg'. */
699 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
700 if (cfg->engine_type) {
701 opts.engine_type = *cfg->engine_type;
703 if (cfg->engine_id) {
704 opts.engine_id = *cfg->engine_id;
707 /* Configure active timeout interval. */
708 opts.active_timeout = cfg->active_timeout;
709 if (!opts.active_timeout) {
710 opts.active_timeout = -1;
711 } else if (opts.active_timeout < 0) {
712 VLOG_WARN("bridge %s: active timeout interval set to negative "
713 "value, using default instead (%d seconds)", br->name,
714 NF_ACTIVE_TIMEOUT_DEFAULT);
715 opts.active_timeout = -1;
718 /* Add engine ID to interface number to disambiguate bridgs? */
719 opts.add_id_to_iface = cfg->add_id_to_interface;
720 if (opts.add_id_to_iface) {
721 if (opts.engine_id > 0x7f) {
722 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
723 "another vswitch, choose an engine id less than 128",
726 if (hmap_count(&br->ports) > 508) {
727 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
728 "another port when more than 508 ports are used",
734 sset_init(&opts.collectors);
735 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
738 if (ofproto_set_netflow(br->ofproto, &opts)) {
739 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
741 sset_destroy(&opts.collectors);
744 /* Set sFlow configuration on 'br'. */
746 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
748 const struct ovsrec_sflow *cfg = br->cfg->sflow;
749 struct ovsrec_controller **controllers;
750 struct ofproto_sflow_options oso;
751 size_t n_controllers;
755 ofproto_set_sflow(br->ofproto, NULL);
759 memset(&oso, 0, sizeof oso);
761 sset_init(&oso.targets);
762 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
764 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
766 oso.sampling_rate = *cfg->sampling;
769 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
771 oso.polling_interval = *cfg->polling;
774 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
776 oso.header_len = *cfg->header;
779 oso.sub_id = (*sflow_bridge_number)++;
780 oso.agent_device = cfg->agent;
782 oso.control_ip = NULL;
783 n_controllers = bridge_get_controllers(br, &controllers);
784 for (i = 0; i < n_controllers; i++) {
785 if (controllers[i]->local_ip) {
786 oso.control_ip = controllers[i]->local_ip;
790 ofproto_set_sflow(br->ofproto, &oso);
792 sset_destroy(&oso.targets);
796 port_configure_stp(const struct ofproto *ofproto, struct port *port,
797 struct ofproto_port_stp_settings *port_s,
798 int *port_num_counter, unsigned long *port_num_bitmap)
800 const char *config_str;
803 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
804 if (config_str && !strcmp(config_str, "false")) {
805 port_s->enable = false;
808 port_s->enable = true;
811 /* STP over bonds is not supported. */
812 if (!list_is_singleton(&port->ifaces)) {
813 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
815 port_s->enable = false;
819 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
821 /* Internal ports shouldn't participate in spanning tree, so
823 if (!strcmp(iface->type, "internal")) {
824 VLOG_DBG("port %s: disable STP on internal ports", port->name);
825 port_s->enable = false;
829 /* STP on mirror output ports is not supported. */
830 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
831 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
832 port_s->enable = false;
836 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
838 unsigned long int port_num = strtoul(config_str, NULL, 0);
839 int port_idx = port_num - 1;
841 if (port_num < 1 || port_num > STP_MAX_PORTS) {
842 VLOG_ERR("port %s: invalid stp-port-num", port->name);
843 port_s->enable = false;
847 if (bitmap_is_set(port_num_bitmap, port_idx)) {
848 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
849 port->name, port_num);
850 port_s->enable = false;
853 bitmap_set1(port_num_bitmap, port_idx);
854 port_s->port_num = port_idx;
856 if (*port_num_counter > STP_MAX_PORTS) {
857 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
858 port_s->enable = false;
862 port_s->port_num = (*port_num_counter)++;
865 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
867 port_s->path_cost = strtoul(config_str, NULL, 10);
871 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
872 /* Couldn't get speed, so assume 100Mb/s. */
873 port_s->path_cost = 19;
877 mbps = netdev_features_to_bps(current) / 1000000;
878 port_s->path_cost = stp_convert_speed_to_cost(mbps);
882 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
884 port_s->priority = strtoul(config_str, NULL, 0);
886 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
890 /* Set spanning tree configuration on 'br'. */
892 bridge_configure_stp(struct bridge *br)
894 if (!br->cfg->stp_enable) {
895 ofproto_set_stp(br->ofproto, NULL);
897 struct ofproto_stp_settings br_s;
898 const char *config_str;
900 int port_num_counter;
901 unsigned long *port_num_bitmap;
903 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
905 uint8_t ea[ETH_ADDR_LEN];
907 if (eth_addr_from_string(config_str, ea)) {
908 br_s.system_id = eth_addr_to_uint64(ea);
910 br_s.system_id = eth_addr_to_uint64(br->ea);
911 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
912 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
915 br_s.system_id = eth_addr_to_uint64(br->ea);
918 config_str = bridge_get_other_config(br->cfg, "stp-priority");
920 br_s.priority = strtoul(config_str, NULL, 0);
922 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
925 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
927 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
929 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
932 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
934 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
936 br_s.max_age = STP_DEFAULT_MAX_AGE;
939 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
941 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
943 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
946 /* Configure STP on the bridge. */
947 if (ofproto_set_stp(br->ofproto, &br_s)) {
948 VLOG_ERR("bridge %s: could not enable STP", br->name);
952 /* Users must either set the port number with the "stp-port-num"
953 * configuration on all ports or none. If manual configuration
954 * is not done, then we allocate them sequentially. */
955 port_num_counter = 0;
956 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
957 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
958 struct ofproto_port_stp_settings port_s;
961 port_configure_stp(br->ofproto, port, &port_s,
962 &port_num_counter, port_num_bitmap);
964 /* As bonds are not supported, just apply configuration to
966 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
967 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
969 VLOG_ERR("port %s: could not enable STP", port->name);
975 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
976 && port_num_counter) {
977 VLOG_ERR("bridge %s: must manually configure all STP port "
978 "IDs or none, disabling", br->name);
979 ofproto_set_stp(br->ofproto, NULL);
981 bitmap_free(port_num_bitmap);
986 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
988 const struct port *port = port_lookup(br, name);
989 return port && port_is_bond_fake_iface(port);
993 port_is_bond_fake_iface(const struct port *port)
995 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
999 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1001 struct bridge *br, *next;
1002 struct shash new_br;
1005 /* Collect new bridges' names and types. */
1006 shash_init(&new_br);
1007 for (i = 0; i < cfg->n_bridges; i++) {
1008 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1009 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1011 if (strchr(br_cfg->name, '/')) {
1012 /* Prevent remote ovsdb-server users from accessing arbitrary
1013 * directories, e.g. consider a bridge named "../../../etc/". */
1014 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1016 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1017 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1021 /* Get rid of deleted bridges or those whose types have changed.
1022 * Update 'cfg' of bridges that still exist. */
1023 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1024 br->cfg = shash_find_data(&new_br, br->name);
1025 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1026 br->cfg->datapath_type))) {
1031 /* Add new bridges. */
1032 for (i = 0; i < cfg->n_bridges; i++) {
1033 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1034 struct bridge *br = bridge_lookup(br_cfg->name);
1036 bridge_create(br_cfg);
1040 shash_destroy(&new_br);
1044 may_port_action(long long int *port_action_timer)
1046 if (need_reconfigure) {
1051 if (*port_action_timer == LLONG_MAX) {
1052 *port_action_timer = time_msec() + OFP_PORT_ACTION_WINDOW;
1053 } else if (time_msec() > *port_action_timer) {
1054 need_reconfigure = true;
1060 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1063 * The kernel will reject any attempt to add a given port to a datapath if that
1064 * port already belongs to a different datapath, so we must do all port
1065 * deletions before any port additions. */
1067 bridge_del_ofproto_ports(struct bridge *br,
1068 long long int *port_action_timer)
1070 struct ofproto_port_dump dump;
1071 struct ofproto_port ofproto_port;
1073 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1074 const char *name = ofproto_port.name;
1075 struct iface *iface;
1079 /* Ignore the local port. We can't change it anyhow. */
1080 if (!strcmp(name, br->name)) {
1084 /* Get the type that 'ofproto_port' should have (ordinarily the
1085 * type of its corresponding iface) or NULL if it should be
1087 iface = iface_lookup(br, name);
1088 type = (iface ? iface->type
1089 : bridge_has_bond_fake_iface(br, name) ? "internal"
1092 /* If it's the wrong type then delete the ofproto port. */
1094 && !strcmp(ofproto_port.type, type)
1095 && (!iface || !iface->netdev
1096 || !strcmp(netdev_get_type(iface->netdev), type))) {
1100 if (may_port_action(port_action_timer)) {
1101 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1103 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1104 br->name, name, strerror(error));
1106 VLOG_INFO("bridge %s: removed interface %s (%d)", br->name,
1107 name, ofproto_port.ofp_port);
1111 netdev_close(iface->netdev);
1112 iface->netdev = NULL;
1118 iface_set_ofp_port(struct iface *iface, int ofp_port)
1120 struct bridge *br = iface->port->bridge;
1122 assert(iface->ofp_port < 0 && ofp_port >= 0);
1123 iface->ofp_port = ofp_port;
1124 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1125 iface_set_ofport(iface->cfg, ofp_port);
1129 bridge_refresh_ofp_port(struct bridge *br)
1131 struct ofproto_port_dump dump;
1132 struct ofproto_port ofproto_port;
1135 /* Clear all the "ofp_port"es. */
1136 hmap_clear(&br->ifaces);
1137 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1138 struct iface *iface;
1140 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1141 iface->ofp_port = -1;
1145 /* Obtain the correct "ofp_port"s from ofproto. */
1146 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1147 struct iface *iface = iface_lookup(br, ofproto_port.name);
1149 if (iface->ofp_port >= 0) {
1150 VLOG_WARN("bridge %s: interface %s reported twice",
1151 br->name, ofproto_port.name);
1152 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1153 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1154 br->name, ofproto_port.ofp_port);
1156 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1162 /* Add an ofproto port for any "struct iface" that doesn't have one.
1163 * Delete any "struct iface" for which this fails.
1164 * Delete any "struct port" that thereby ends up with no ifaces. */
1166 bridge_add_ofproto_ports(struct bridge *br,
1167 long long int *port_action_timer)
1169 struct port *port, *next_port;
1171 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1172 struct iface *iface, *next_iface;
1173 struct ofproto_port ofproto_port;
1175 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1178 if (iface->ofp_port < 0 && !may_port_action(port_action_timer)) {
1179 iface_clear_db_record(iface->cfg);
1180 iface_destroy(iface);
1184 /* Open the netdev. */
1185 if (!iface->netdev) {
1186 error = netdev_open(iface->name, iface->type, &iface->netdev);
1188 VLOG_WARN("could not open network device %s (%s)",
1189 iface->name, strerror(error));
1193 && port->cfg->vlan_mode
1194 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1195 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1201 /* Configure the netdev. */
1202 if (iface->netdev) {
1206 shash_from_ovs_idl_map(iface->cfg->key_options,
1207 iface->cfg->value_options,
1208 iface->cfg->n_options, &args);
1209 error = netdev_set_config(iface->netdev, &args);
1210 shash_destroy(&args);
1213 VLOG_WARN("could not configure network device %s (%s)",
1214 iface->name, strerror(error));
1215 netdev_close(iface->netdev);
1216 iface->netdev = NULL;
1220 /* Add the port, if necessary. */
1221 if (iface->netdev && iface->ofp_port < 0) {
1225 error = ofproto_port_add(br->ofproto, iface->netdev,
1228 VLOG_INFO("bridge %s: added interface %s (%d)", br->name,
1229 iface->name, ofp_port);
1230 iface_set_ofp_port(iface, ofp_port);
1232 netdev_close(iface->netdev);
1233 iface->netdev = NULL;
1237 /* Populate stats columns in new Interface rows. */
1238 if (iface->netdev && iface->need_refresh) {
1239 iface_refresh_stats(iface);
1240 iface_refresh_status(iface);
1241 iface->need_refresh = false;
1244 /* Delete the iface if we failed. */
1245 if (iface->netdev && iface->ofp_port >= 0) {
1246 VLOG_DBG("bridge %s: interface %s is on port %d",
1247 br->name, iface->name, iface->ofp_port);
1249 if (iface->netdev) {
1250 VLOG_ERR("bridge %s: missing %s interface, dropping",
1251 br->name, iface->name);
1253 /* We already reported a related error, don't bother
1254 * duplicating it. */
1256 if (!ofproto_port_query_by_name(br->ofproto, port->name,
1258 VLOG_INFO("bridge %s: removed interface %s (%d)",
1259 br->name, port->name, ofproto_port.ofp_port);
1260 ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1261 ofproto_port_destroy(&ofproto_port);
1263 iface_clear_db_record(iface->cfg);
1264 iface_destroy(iface);
1268 if (list_is_empty(&port->ifaces)) {
1269 if (!need_reconfigure) {
1270 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1276 /* Add bond fake iface if necessary. */
1277 if (port_is_bond_fake_iface(port)) {
1278 if (ofproto_port_query_by_name(br->ofproto, port->name,
1280 struct netdev *netdev;
1283 error = netdev_open(port->name, "internal", &netdev);
1285 /* There are unlikely to be a great number of fake
1286 * interfaces so we don't bother rate limiting their
1288 ofproto_port_add(br->ofproto, netdev, NULL);
1289 netdev_close(netdev);
1291 VLOG_WARN("could not open network device %s (%s)",
1292 port->name, strerror(error));
1295 /* Already exists, nothing to do. */
1296 ofproto_port_destroy(&ofproto_port);
1303 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1307 for (i = 0; i < n; i++) {
1308 if (!strcmp(keys[i], key)) {
1316 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1318 return get_ovsrec_key_value(br_cfg->key_other_config,
1319 br_cfg->value_other_config,
1320 br_cfg->n_other_config, key);
1323 /* Set Flow eviction threshold */
1325 bridge_configure_flow_eviction_threshold(struct bridge *br)
1327 const char *threshold_str;
1330 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1331 if (threshold_str) {
1332 threshold = strtoul(threshold_str, NULL, 10);
1334 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1336 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1339 /* Set forward BPDU option. */
1341 bridge_configure_forward_bpdu(struct bridge *br)
1343 const char *forward_bpdu_str;
1344 bool forward_bpdu = false;
1346 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1347 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1348 forward_bpdu = true;
1350 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1353 /* Set MAC aging time for 'br'. */
1355 bridge_configure_mac_idle_time(struct bridge *br)
1357 const char *idle_time_str;
1360 idle_time_str = bridge_get_other_config(br->cfg, "mac-aging-time");
1361 idle_time = (idle_time_str && atoi(idle_time_str)
1362 ? atoi(idle_time_str)
1363 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1364 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1368 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1369 struct iface **hw_addr_iface)
1371 struct hmapx mirror_output_ports;
1374 bool found_addr = false;
1378 *hw_addr_iface = NULL;
1380 /* Did the user request a particular MAC? */
1381 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1382 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1383 if (eth_addr_is_multicast(ea)) {
1384 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1385 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1386 } else if (eth_addr_is_zero(ea)) {
1387 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1393 /* Mirror output ports don't participate in picking the local hardware
1394 * address. ofproto can't help us find out whether a given port is a
1395 * mirror output because we haven't configured mirrors yet, so we need to
1396 * accumulate them ourselves. */
1397 hmapx_init(&mirror_output_ports);
1398 for (i = 0; i < br->cfg->n_mirrors; i++) {
1399 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1400 if (m->output_port) {
1401 hmapx_add(&mirror_output_ports, m->output_port);
1405 /* Otherwise choose the minimum non-local MAC address among all of the
1407 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1408 uint8_t iface_ea[ETH_ADDR_LEN];
1409 struct iface *candidate;
1410 struct iface *iface;
1412 /* Mirror output ports don't participate. */
1413 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1417 /* Choose the MAC address to represent the port. */
1419 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1420 /* Find the interface with this Ethernet address (if any) so that
1421 * we can provide the correct devname to the caller. */
1422 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1423 uint8_t candidate_ea[ETH_ADDR_LEN];
1424 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1425 && eth_addr_equals(iface_ea, candidate_ea)) {
1430 /* Choose the interface whose MAC address will represent the port.
1431 * The Linux kernel bonding code always chooses the MAC address of
1432 * the first slave added to a bond, and the Fedora networking
1433 * scripts always add slaves to a bond in alphabetical order, so
1434 * for compatibility we choose the interface with the name that is
1435 * first in alphabetical order. */
1436 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1437 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1442 /* The local port doesn't count (since we're trying to choose its
1443 * MAC address anyway). */
1444 if (iface->ofp_port == OFPP_LOCAL) {
1449 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1455 /* Compare against our current choice. */
1456 if (!eth_addr_is_multicast(iface_ea) &&
1457 !eth_addr_is_local(iface_ea) &&
1458 !eth_addr_is_reserved(iface_ea) &&
1459 !eth_addr_is_zero(iface_ea) &&
1460 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1462 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1463 *hw_addr_iface = iface;
1468 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1469 br->name, ETH_ADDR_ARGS(ea));
1471 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1472 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1473 *hw_addr_iface = NULL;
1474 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1475 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1478 hmapx_destroy(&mirror_output_ports);
1481 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1482 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1483 * an interface on 'br', then that interface must be passed in as
1484 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1485 * 'hw_addr_iface' must be passed in as a null pointer. */
1487 bridge_pick_datapath_id(struct bridge *br,
1488 const uint8_t bridge_ea[ETH_ADDR_LEN],
1489 struct iface *hw_addr_iface)
1492 * The procedure for choosing a bridge MAC address will, in the most
1493 * ordinary case, also choose a unique MAC that we can use as a datapath
1494 * ID. In some special cases, though, multiple bridges will end up with
1495 * the same MAC address. This is OK for the bridges, but it will confuse
1496 * the OpenFlow controller, because each datapath needs a unique datapath
1499 * Datapath IDs must be unique. It is also very desirable that they be
1500 * stable from one run to the next, so that policy set on a datapath
1503 const char *datapath_id;
1506 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1507 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1511 if (!hw_addr_iface) {
1513 * A purely internal bridge, that is, one that has no non-virtual
1514 * network devices on it at all, is difficult because it has no
1515 * natural unique identifier at all.
1517 * When the host is a XenServer, we handle this case by hashing the
1518 * host's UUID with the name of the bridge. Names of bridges are
1519 * persistent across XenServer reboots, although they can be reused if
1520 * an internal network is destroyed and then a new one is later
1521 * created, so this is fairly effective.
1523 * When the host is not a XenServer, we punt by using a random MAC
1524 * address on each run.
1526 const char *host_uuid = xenserver_get_host_uuid();
1528 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1529 dpid = dpid_from_hash(combined, strlen(combined));
1535 return eth_addr_to_uint64(bridge_ea);
1539 dpid_from_hash(const void *data, size_t n)
1541 uint8_t hash[SHA1_DIGEST_SIZE];
1543 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1544 sha1_bytes(data, n, hash);
1545 eth_addr_mark_random(hash);
1546 return eth_addr_to_uint64(hash);
1550 iface_refresh_status(struct iface *iface)
1554 enum netdev_flags flags;
1561 if (iface_is_synthetic(iface)) {
1567 if (!netdev_get_status(iface->netdev, &sh)) {
1569 char **keys, **values;
1571 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1572 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1577 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1580 shash_destroy_free_data(&sh);
1582 error = netdev_get_flags(iface->netdev, &flags);
1584 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1587 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1590 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1592 ovsrec_interface_set_duplex(iface->cfg,
1593 netdev_features_is_full_duplex(current)
1595 /* warning: uint64_t -> int64_t conversion */
1596 bps = netdev_features_to_bps(current);
1597 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1600 ovsrec_interface_set_duplex(iface->cfg, NULL);
1601 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1604 error = netdev_get_mtu(iface->netdev, &mtu);
1607 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1610 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1614 /* Writes 'iface''s CFM statistics to the database. */
1616 iface_refresh_cfm_stats(struct iface *iface)
1618 const struct ovsrec_interface *cfg = iface->cfg;
1620 const uint64_t *rmps;
1623 if (iface_is_synthetic(iface)) {
1627 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1630 const char *reasons[CFM_FAULT_N_REASONS];
1631 bool fault_bool = fault;
1635 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1636 int reason = 1 << i;
1637 if (fault & reason) {
1638 reasons[j++] = cfm_fault_reason_to_str(reason);
1642 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1643 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1645 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1646 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1649 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1650 iface->ofp_port, &rmps, &n_rmps);
1652 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1655 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1660 iface_refresh_stats(struct iface *iface)
1662 #define IFACE_STATS \
1663 IFACE_STAT(rx_packets, "rx_packets") \
1664 IFACE_STAT(tx_packets, "tx_packets") \
1665 IFACE_STAT(rx_bytes, "rx_bytes") \
1666 IFACE_STAT(tx_bytes, "tx_bytes") \
1667 IFACE_STAT(rx_dropped, "rx_dropped") \
1668 IFACE_STAT(tx_dropped, "tx_dropped") \
1669 IFACE_STAT(rx_errors, "rx_errors") \
1670 IFACE_STAT(tx_errors, "tx_errors") \
1671 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1672 IFACE_STAT(rx_over_errors, "rx_over_err") \
1673 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1674 IFACE_STAT(collisions, "collisions")
1676 #define IFACE_STAT(MEMBER, NAME) NAME,
1677 static char *keys[] = { IFACE_STATS };
1679 int64_t values[ARRAY_SIZE(keys)];
1682 struct netdev_stats stats;
1684 if (iface_is_synthetic(iface)) {
1688 /* Intentionally ignore return value, since errors will set 'stats' to
1689 * all-1s, and we will deal with that correctly below. */
1690 netdev_get_stats(iface->netdev, &stats);
1692 /* Copy statistics into values[] array. */
1694 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1697 assert(i == ARRAY_SIZE(keys));
1699 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1704 br_refresh_stp_status(struct bridge *br)
1706 struct ofproto *ofproto = br->ofproto;
1707 struct ofproto_stp_status status;
1708 char *keys[3], *values[3];
1711 if (ofproto_get_stp_status(ofproto, &status)) {
1715 if (!status.enabled) {
1716 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1720 keys[0] = "stp_bridge_id",
1721 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1722 keys[1] = "stp_designated_root",
1723 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1724 keys[2] = "stp_root_path_cost",
1725 values[2] = xasprintf("%d", status.root_path_cost);
1727 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1729 for (i = 0; i < ARRAY_SIZE(values); i++) {
1735 port_refresh_stp_status(struct port *port)
1737 struct ofproto *ofproto = port->bridge->ofproto;
1738 struct iface *iface;
1739 struct ofproto_port_stp_status status;
1741 char *str_values[4];
1742 int64_t int_values[3];
1745 if (port_is_synthetic(port)) {
1749 /* STP doesn't currently support bonds. */
1750 if (!list_is_singleton(&port->ifaces)) {
1751 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1755 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1757 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1761 if (!status.enabled) {
1762 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1763 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1767 /* Set Status column. */
1768 keys[0] = "stp_port_id";
1769 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1770 keys[1] = "stp_state";
1771 str_values[1] = xstrdup(stp_state_name(status.state));
1772 keys[2] = "stp_sec_in_state";
1773 str_values[2] = xasprintf("%u", status.sec_in_state);
1774 keys[3] = "stp_role";
1775 str_values[3] = xstrdup(stp_role_name(status.role));
1777 ovsrec_port_set_status(port->cfg, keys, str_values,
1778 ARRAY_SIZE(str_values));
1780 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1781 free(str_values[i]);
1784 /* Set Statistics column. */
1785 keys[0] = "stp_tx_count";
1786 int_values[0] = status.tx_count;
1787 keys[1] = "stp_rx_count";
1788 int_values[1] = status.rx_count;
1789 keys[2] = "stp_error_count";
1790 int_values[2] = status.error_count;
1792 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1793 ARRAY_SIZE(int_values));
1797 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1801 /* Use other-config:enable-system-stats by preference. */
1802 enable = get_ovsrec_key_value(cfg->key_other_config,
1803 cfg->value_other_config,
1804 cfg->n_other_config,
1805 "enable-statistics");
1807 return !strcmp(enable, "true");
1810 /* Disable by default. */
1815 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1817 struct ovsdb_datum datum;
1821 if (enable_system_stats(cfg)) {
1822 get_system_stats(&stats);
1825 ovsdb_datum_from_shash(&datum, &stats);
1826 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1830 static inline const char *
1831 nx_role_to_str(enum nx_role role)
1836 case NX_ROLE_MASTER:
1841 return "*** INVALID ROLE ***";
1846 refresh_controller_status(void)
1850 const struct ovsrec_controller *cfg;
1854 /* Accumulate status for controllers on all bridges. */
1855 HMAP_FOR_EACH (br, node, &all_bridges) {
1856 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1859 /* Update each controller in the database with current status. */
1860 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1861 struct ofproto_controller_info *cinfo =
1862 shash_find_data(&info, cfg->target);
1865 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1866 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1867 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1868 (char **) cinfo->pairs.values,
1871 ovsrec_controller_set_is_connected(cfg, false);
1872 ovsrec_controller_set_role(cfg, NULL);
1873 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1877 ofproto_free_ofproto_controller_info(&info);
1881 refresh_cfm_stats(void)
1883 static struct ovsdb_idl_txn *txn = NULL;
1888 txn = ovsdb_idl_txn_create(idl);
1890 HMAP_FOR_EACH (br, node, &all_bridges) {
1891 struct iface *iface;
1893 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1894 iface_refresh_cfm_stats(iface);
1899 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1900 ovsdb_idl_txn_destroy(txn);
1905 /* Performs periodic activity required by bridges that needs to be done with
1906 * the least possible latency.
1908 * It makes sense to call this function a couple of times per poll loop, to
1909 * provide a significant performance boost on some benchmarks with ofprotos
1910 * that use the ofproto-dpif implementation. */
1912 bridge_run_fast(void)
1916 HMAP_FOR_EACH (br, node, &all_bridges) {
1917 ofproto_run_fast(br->ofproto);
1924 const struct ovsrec_open_vswitch *cfg;
1926 bool vlan_splinters_changed;
1927 bool database_changed;
1930 /* (Re)configure if necessary. */
1931 database_changed = ovsdb_idl_run(idl);
1932 if (ovsdb_idl_is_lock_contended(idl)) {
1933 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1934 struct bridge *br, *next_br;
1936 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1937 "disabling this process until it goes away");
1939 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1943 } else if (!ovsdb_idl_has_lock(idl)) {
1946 cfg = ovsrec_open_vswitch_first(idl);
1948 /* Let each bridge do the work that it needs to do. */
1949 HMAP_FOR_EACH (br, node, &all_bridges) {
1950 ofproto_run(br->ofproto);
1953 /* Re-configure SSL. We do this on every trip through the main loop,
1954 * instead of just when the database changes, because the contents of the
1955 * key and certificate files can change without the database changing.
1957 * We do this before bridge_reconfigure() because that function might
1958 * initiate SSL connections and thus requires SSL to be configured. */
1959 if (cfg && cfg->ssl) {
1960 const struct ovsrec_ssl *ssl = cfg->ssl;
1962 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1963 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1966 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1968 vlan_splinters_changed = false;
1969 if (vlan_splinters_enabled_anywhere) {
1970 HMAP_FOR_EACH (br, node, &all_bridges) {
1971 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1972 vlan_splinters_changed = true;
1978 if (need_reconfigure || database_changed || vlan_splinters_changed) {
1980 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1982 bridge_reconfigure(cfg);
1984 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1985 ovsdb_idl_txn_commit(txn);
1986 ovsdb_idl_txn_destroy(txn); /* XXX */
1988 /* We still need to reconfigure to avoid dangling pointers to
1989 * now-destroyed ovsrec structures inside bridge data. */
1990 static const struct ovsrec_open_vswitch null_cfg;
1992 bridge_reconfigure(&null_cfg);
1996 /* Refresh system and interface stats if necessary. */
1997 if (time_msec() >= stats_timer) {
1999 struct ovsdb_idl_txn *txn;
2001 txn = ovsdb_idl_txn_create(idl);
2002 HMAP_FOR_EACH (br, node, &all_bridges) {
2006 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2007 struct iface *iface;
2009 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2010 iface_refresh_stats(iface);
2011 iface_refresh_status(iface);
2015 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2016 mirror_refresh_stats(m);
2020 refresh_system_stats(cfg);
2021 refresh_controller_status();
2022 ovsdb_idl_txn_commit(txn);
2023 ovsdb_idl_txn_destroy(txn); /* XXX */
2026 stats_timer = time_msec() + STATS_INTERVAL;
2029 if (time_msec() >= db_limiter) {
2030 struct ovsdb_idl_txn *txn;
2032 txn = ovsdb_idl_txn_create(idl);
2033 HMAP_FOR_EACH (br, node, &all_bridges) {
2034 struct iface *iface;
2037 br_refresh_stp_status(br);
2039 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2040 port_refresh_stp_status(port);
2043 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2044 const char *link_state;
2045 int64_t link_resets;
2048 if (iface_is_synthetic(iface)) {
2052 current = ofproto_port_is_lacp_current(br->ofproto,
2056 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2058 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2061 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2062 ovsrec_interface_set_link_state(iface->cfg, link_state);
2064 link_resets = netdev_get_carrier_resets(iface->netdev);
2065 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2069 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2070 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2072 ovsdb_idl_txn_destroy(txn);
2075 refresh_cfm_stats();
2081 ovsdb_idl_wait(idl);
2083 if (need_reconfigure) {
2084 poll_immediate_wake();
2087 if (!hmap_is_empty(&all_bridges)) {
2090 HMAP_FOR_EACH (br, node, &all_bridges) {
2091 ofproto_wait(br->ofproto);
2093 poll_timer_wait_until(stats_timer);
2095 if (db_limiter > time_msec()) {
2096 poll_timer_wait_until(db_limiter);
2101 /* QoS unixctl user interface functions. */
2103 struct qos_unixctl_show_cbdata {
2105 struct iface *iface;
2109 qos_unixctl_show_cb(unsigned int queue_id,
2110 const struct shash *details,
2113 struct qos_unixctl_show_cbdata *data = aux;
2114 struct ds *ds = data->ds;
2115 struct iface *iface = data->iface;
2116 struct netdev_queue_stats stats;
2117 struct shash_node *node;
2120 ds_put_cstr(ds, "\n");
2122 ds_put_format(ds, "Queue %u:\n", queue_id);
2124 ds_put_cstr(ds, "Default:\n");
2127 SHASH_FOR_EACH (node, details) {
2128 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2131 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2133 if (stats.tx_packets != UINT64_MAX) {
2134 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2137 if (stats.tx_bytes != UINT64_MAX) {
2138 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2141 if (stats.tx_errors != UINT64_MAX) {
2142 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2145 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2146 queue_id, strerror(error));
2151 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2152 const char *argv[], void *aux OVS_UNUSED)
2154 struct ds ds = DS_EMPTY_INITIALIZER;
2155 struct shash sh = SHASH_INITIALIZER(&sh);
2156 struct iface *iface;
2158 struct shash_node *node;
2159 struct qos_unixctl_show_cbdata data;
2162 iface = iface_find(argv[1]);
2164 unixctl_command_reply_error(conn, "no such interface");
2168 netdev_get_qos(iface->netdev, &type, &sh);
2170 if (*type != '\0') {
2171 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2173 SHASH_FOR_EACH (node, &sh) {
2174 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2179 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2182 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2184 unixctl_command_reply(conn, ds_cstr(&ds));
2186 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2187 unixctl_command_reply_error(conn, ds_cstr(&ds));
2190 shash_destroy_free_data(&sh);
2194 /* Bridge reconfiguration functions. */
2196 bridge_create(const struct ovsrec_bridge *br_cfg)
2200 assert(!bridge_lookup(br_cfg->name));
2201 br = xzalloc(sizeof *br);
2203 br->name = xstrdup(br_cfg->name);
2204 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2207 /* Derive the default Ethernet address from the bridge's UUID. This should
2208 * be unique and it will be stable between ovs-vswitchd runs. */
2209 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2210 eth_addr_mark_random(br->default_ea);
2212 hmap_init(&br->ports);
2213 hmap_init(&br->ifaces);
2214 hmap_init(&br->iface_by_name);
2215 hmap_init(&br->mirrors);
2217 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2221 bridge_destroy(struct bridge *br)
2224 struct mirror *mirror, *next_mirror;
2225 struct port *port, *next_port;
2227 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2230 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2231 mirror_destroy(mirror);
2233 hmap_remove(&all_bridges, &br->node);
2234 ofproto_destroy(br->ofproto);
2235 hmap_destroy(&br->ifaces);
2236 hmap_destroy(&br->ports);
2237 hmap_destroy(&br->iface_by_name);
2238 hmap_destroy(&br->mirrors);
2245 static struct bridge *
2246 bridge_lookup(const char *name)
2250 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2251 if (!strcmp(br->name, name)) {
2258 /* Handle requests for a listing of all flows known by the OpenFlow
2259 * stack, including those normally hidden. */
2261 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2262 const char *argv[], void *aux OVS_UNUSED)
2267 br = bridge_lookup(argv[1]);
2269 unixctl_command_reply_error(conn, "Unknown bridge");
2274 ofproto_get_all_flows(br->ofproto, &results);
2276 unixctl_command_reply(conn, ds_cstr(&results));
2277 ds_destroy(&results);
2280 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2281 * connections and reconnect. If BRIDGE is not specified, then all bridges
2282 * drop their controller connections and reconnect. */
2284 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2285 const char *argv[], void *aux OVS_UNUSED)
2289 br = bridge_lookup(argv[1]);
2291 unixctl_command_reply_error(conn, "Unknown bridge");
2294 ofproto_reconnect_controllers(br->ofproto);
2296 HMAP_FOR_EACH (br, node, &all_bridges) {
2297 ofproto_reconnect_controllers(br->ofproto);
2300 unixctl_command_reply(conn, NULL);
2304 bridge_get_controllers(const struct bridge *br,
2305 struct ovsrec_controller ***controllersp)
2307 struct ovsrec_controller **controllers;
2308 size_t n_controllers;
2310 controllers = br->cfg->controller;
2311 n_controllers = br->cfg->n_controller;
2313 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2319 *controllersp = controllers;
2321 return n_controllers;
2324 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2325 * those configured in 'br->cfg'. */
2327 bridge_add_del_ports(struct bridge *br,
2328 const unsigned long int *splinter_vlans)
2330 struct port *port, *next;
2331 struct shash_node *node;
2332 struct shash new_ports;
2335 /* Collect new ports. */
2336 shash_init(&new_ports);
2337 for (i = 0; i < br->cfg->n_ports; i++) {
2338 const char *name = br->cfg->ports[i]->name;
2339 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2340 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2344 if (bridge_get_controllers(br, NULL)
2345 && !shash_find(&new_ports, br->name)) {
2346 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2347 br->name, br->name);
2349 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2350 br->synth_local_port.n_interfaces = 1;
2351 br->synth_local_port.name = br->name;
2353 br->synth_local_iface.name = br->name;
2354 br->synth_local_iface.type = "internal";
2356 br->synth_local_ifacep = &br->synth_local_iface;
2358 shash_add(&new_ports, br->name, &br->synth_local_port);
2361 if (splinter_vlans) {
2362 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2365 /* Get rid of deleted ports.
2366 * Get rid of deleted interfaces on ports that still exist. */
2367 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2368 port->cfg = shash_find_data(&new_ports, port->name);
2372 port_del_ifaces(port);
2376 /* Create new ports.
2377 * Add new interfaces to existing ports. */
2378 SHASH_FOR_EACH (node, &new_ports) {
2379 struct port *port = port_lookup(br, node->name);
2381 struct ovsrec_port *cfg = node->data;
2382 port = port_create(br, cfg);
2384 port_add_ifaces(port);
2385 if (list_is_empty(&port->ifaces)) {
2386 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2387 br->name, port->name);
2391 shash_destroy(&new_ports);
2394 /* Initializes 'oc' appropriately as a management service controller for
2397 * The caller must free oc->target when it is no longer needed. */
2399 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2400 struct ofproto_controller *oc)
2402 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2403 oc->max_backoff = 0;
2404 oc->probe_interval = 60;
2405 oc->band = OFPROTO_OUT_OF_BAND;
2407 oc->burst_limit = 0;
2408 oc->enable_async_msgs = true;
2411 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2413 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2414 struct ofproto_controller *oc)
2416 oc->target = c->target;
2417 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2418 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2419 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2420 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2421 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2422 oc->burst_limit = (c->controller_burst_limit
2423 ? *c->controller_burst_limit : 0);
2424 oc->enable_async_msgs = (!c->enable_async_messages
2425 || *c->enable_async_messages);
2428 /* Configures the IP stack for 'br''s local interface properly according to the
2429 * configuration in 'c'. */
2431 bridge_configure_local_iface_netdev(struct bridge *br,
2432 struct ovsrec_controller *c)
2434 struct netdev *netdev;
2435 struct in_addr mask, gateway;
2437 struct iface *local_iface;
2440 /* If there's no local interface or no IP address, give up. */
2441 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2442 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2446 /* Bring up the local interface. */
2447 netdev = local_iface->netdev;
2448 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2450 /* Configure the IP address and netmask. */
2451 if (!c->local_netmask
2452 || !inet_aton(c->local_netmask, &mask)
2454 mask.s_addr = guess_netmask(ip.s_addr);
2456 if (!netdev_set_in4(netdev, ip, mask)) {
2457 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2458 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2461 /* Configure the default gateway. */
2462 if (c->local_gateway
2463 && inet_aton(c->local_gateway, &gateway)
2464 && gateway.s_addr) {
2465 if (!netdev_add_router(netdev, gateway)) {
2466 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2467 br->name, IP_ARGS(&gateway.s_addr));
2472 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2473 * in either string are treated as equal to any number of slashes in the other,
2474 * e.g. "x///y" is equal to "x/y". */
2476 equal_pathnames(const char *a, const char *b)
2480 a += strspn(a, "/");
2481 b += strspn(b, "/");
2482 } else if (*a == '\0') {
2493 bridge_configure_remotes(struct bridge *br,
2494 const struct sockaddr_in *managers, size_t n_managers)
2496 const char *disable_ib_str, *queue_id_str;
2497 bool disable_in_band = false;
2500 struct ovsrec_controller **controllers;
2501 size_t n_controllers;
2503 enum ofproto_fail_mode fail_mode;
2505 struct ofproto_controller *ocs;
2509 /* Check if we should disable in-band control on this bridge. */
2510 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2511 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2512 disable_in_band = true;
2515 /* Set OpenFlow queue ID for in-band control. */
2516 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2517 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2518 ofproto_set_in_band_queue(br->ofproto, queue_id);
2520 if (disable_in_band) {
2521 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2523 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2526 n_controllers = bridge_get_controllers(br, &controllers);
2528 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2531 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2532 for (i = 0; i < n_controllers; i++) {
2533 struct ovsrec_controller *c = controllers[i];
2535 if (!strncmp(c->target, "punix:", 6)
2536 || !strncmp(c->target, "unix:", 5)) {
2537 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2540 whitelist = xasprintf("unix:%s/%s.controller",
2541 ovs_rundir(), br->name);
2542 if (!equal_pathnames(c->target, whitelist)) {
2543 /* Prevent remote ovsdb-server users from accessing arbitrary
2544 * Unix domain sockets and overwriting arbitrary local
2546 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2547 "controller \"%s\" due to possibility for remote "
2548 "exploit. Instead, specify whitelisted \"%s\" or "
2549 "connect to \"unix:%s/%s.mgmt\" (which is always "
2550 "available without special configuration).",
2551 br->name, c->target, whitelist,
2552 ovs_rundir(), br->name);
2560 bridge_configure_local_iface_netdev(br, c);
2561 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2562 if (disable_in_band) {
2563 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2568 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2569 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2572 /* Set the fail-mode. */
2573 fail_mode = !br->cfg->fail_mode
2574 || !strcmp(br->cfg->fail_mode, "standalone")
2575 ? OFPROTO_FAIL_STANDALONE
2576 : OFPROTO_FAIL_SECURE;
2577 ofproto_set_fail_mode(br->ofproto, fail_mode);
2579 /* Configure OpenFlow controller connection snooping. */
2580 if (!ofproto_has_snoops(br->ofproto)) {
2584 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2585 ovs_rundir(), br->name));
2586 ofproto_set_snoops(br->ofproto, &snoops);
2587 sset_destroy(&snoops);
2592 bridge_configure_tables(struct bridge *br)
2594 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2598 n_tables = ofproto_get_n_tables(br->ofproto);
2600 for (i = 0; i < n_tables; i++) {
2601 struct ofproto_table_settings s;
2604 s.max_flows = UINT_MAX;
2608 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2609 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2612 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2613 s.max_flows = *cfg->flow_limit;
2615 if (cfg->overflow_policy
2616 && !strcmp(cfg->overflow_policy, "evict")) {
2619 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2620 for (k = 0; k < cfg->n_groups; k++) {
2621 const char *string = cfg->groups[k];
2624 msg = mf_parse_subfield__(&s.groups[k], &string);
2626 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2627 "'groups' (%s)", br->name, i, msg);
2629 } else if (*string) {
2630 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2631 "element '%s' contains trailing garbage",
2632 br->name, i, cfg->groups[k]);
2640 ofproto_configure_table(br->ofproto, i, &s);
2644 for (; j < br->cfg->n_flow_tables; j++) {
2645 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2646 "%"PRId64" not supported by this datapath", br->name,
2647 br->cfg->key_flow_tables[j]);
2651 /* Port functions. */
2653 static struct port *
2654 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2658 port = xzalloc(sizeof *port);
2660 port->name = xstrdup(cfg->name);
2662 list_init(&port->ifaces);
2664 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2669 get_port_other_config(const struct ovsrec_port *port, const char *key,
2670 const char *default_value)
2674 value = get_ovsrec_key_value(port->key_other_config,
2675 port->value_other_config,
2676 port->n_other_config, key);
2677 return value ? value : default_value;
2681 get_interface_other_config(const struct ovsrec_interface *iface,
2682 const char *key, const char *default_value)
2686 value = get_ovsrec_key_value(iface->key_other_config,
2687 iface->value_other_config,
2688 iface->n_other_config, key);
2689 return value ? value : default_value;
2692 /* Deletes interfaces from 'port' that are no longer configured for it. */
2694 port_del_ifaces(struct port *port)
2696 struct iface *iface, *next;
2697 struct sset new_ifaces;
2700 /* Collect list of new interfaces. */
2701 sset_init(&new_ifaces);
2702 for (i = 0; i < port->cfg->n_interfaces; i++) {
2703 const char *name = port->cfg->interfaces[i]->name;
2704 const char *type = port->cfg->interfaces[i]->name;
2705 if (strcmp(type, "null")) {
2706 sset_add(&new_ifaces, name);
2710 /* Get rid of deleted interfaces. */
2711 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2712 if (!sset_contains(&new_ifaces, iface->name)) {
2713 iface_destroy(iface);
2717 sset_destroy(&new_ifaces);
2720 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2723 port_add_ifaces(struct port *port)
2725 struct shash new_ifaces;
2726 struct shash_node *node;
2729 /* Collect new ifaces. */
2730 shash_init(&new_ifaces);
2731 for (i = 0; i < port->cfg->n_interfaces; i++) {
2732 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2733 if (strcmp(cfg->type, "null")
2734 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2735 VLOG_WARN("port %s: %s specified twice as port interface",
2736 port->name, cfg->name);
2737 iface_clear_db_record(cfg);
2741 /* Create new interfaces.
2742 * Update interface types and 'cfg' members. */
2743 SHASH_FOR_EACH (node, &new_ifaces) {
2744 const struct ovsrec_interface *cfg = node->data;
2745 const char *iface_name = node->name;
2746 struct iface *iface;
2748 iface = iface_lookup(port->bridge, iface_name);
2750 iface = iface_create(port, cfg);
2755 /* Determine interface type. The local port always has type
2756 * "internal". Other ports take their type from the database and
2757 * default to "system" if none is specified. */
2758 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2759 : cfg->type[0] ? cfg->type
2762 shash_destroy(&new_ifaces);
2766 port_destroy(struct port *port)
2769 struct bridge *br = port->bridge;
2770 struct iface *iface, *next;
2773 ofproto_bundle_unregister(br->ofproto, port);
2776 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2777 iface_destroy(iface);
2780 hmap_remove(&br->ports, &port->hmap_node);
2786 static struct port *
2787 port_lookup(const struct bridge *br, const char *name)
2791 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2793 if (!strcmp(port->name, name)) {
2801 enable_lacp(struct port *port, bool *activep)
2803 if (!port->cfg->lacp) {
2804 /* XXX when LACP implementation has been sufficiently tested, enable by
2805 * default and make active on bonded ports. */
2807 } else if (!strcmp(port->cfg->lacp, "off")) {
2809 } else if (!strcmp(port->cfg->lacp, "active")) {
2812 } else if (!strcmp(port->cfg->lacp, "passive")) {
2816 VLOG_WARN("port %s: unknown LACP mode %s",
2817 port->name, port->cfg->lacp);
2822 static struct lacp_settings *
2823 port_configure_lacp(struct port *port, struct lacp_settings *s)
2825 const char *lacp_time, *system_id;
2826 long long int custom_time;
2829 if (!enable_lacp(port, &s->active)) {
2833 s->name = port->name;
2835 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2837 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2838 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2839 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2840 " address.", port->name, system_id);
2844 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2847 if (eth_addr_is_zero(s->id)) {
2848 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2852 /* Prefer bondable links if unspecified. */
2853 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2855 s->priority = (priority > 0 && priority <= UINT16_MAX
2857 : UINT16_MAX - !list_is_short(&port->ifaces));
2859 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2863 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2864 custom_time = atoi(lacp_time);
2865 if (!strcmp(lacp_time, "fast")) {
2866 s->lacp_time = LACP_TIME_FAST;
2867 } else if (!strcmp(lacp_time, "slow")) {
2868 s->lacp_time = LACP_TIME_SLOW;
2869 } else if (custom_time > 0) {
2870 s->lacp_time = LACP_TIME_CUSTOM;
2871 s->custom_time = custom_time;
2873 s->lacp_time = LACP_TIME_SLOW;
2880 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2882 int priority, portid, key;
2884 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2885 priority = atoi(get_interface_other_config(iface->cfg,
2886 "lacp-port-priority", "0"));
2887 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2890 if (portid <= 0 || portid > UINT16_MAX) {
2891 portid = iface->ofp_port;
2894 if (priority <= 0 || priority > UINT16_MAX) {
2895 priority = UINT16_MAX;
2898 if (key < 0 || key > UINT16_MAX) {
2902 s->name = iface->name;
2904 s->priority = priority;
2909 port_configure_bond(struct port *port, struct bond_settings *s,
2910 uint32_t *bond_stable_ids)
2912 const char *detect_s;
2913 struct iface *iface;
2914 int miimon_interval;
2917 s->name = port->name;
2919 if (port->cfg->bond_mode) {
2920 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2921 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2922 port->name, port->cfg->bond_mode,
2923 bond_mode_to_string(s->balance));
2926 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2928 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2929 * active-backup. At some point we should remove this warning. */
2930 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2931 " in previous versions, the default bond_mode was"
2932 " balance-slb", port->name,
2933 bond_mode_to_string(s->balance));
2935 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2936 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2937 "please use another bond type or disable flood_vlans",
2941 miimon_interval = atoi(get_port_other_config(port->cfg,
2942 "bond-miimon-interval", "0"));
2943 if (miimon_interval <= 0) {
2944 miimon_interval = 200;
2947 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2948 if (!strcmp(detect_s, "carrier")) {
2949 miimon_interval = 0;
2950 } else if (strcmp(detect_s, "miimon")) {
2951 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2952 "defaulting to carrier", port->name, detect_s);
2953 miimon_interval = 0;
2956 s->up_delay = MAX(0, port->cfg->bond_updelay);
2957 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2958 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2959 s->rebalance_interval = atoi(
2960 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2961 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2962 s->rebalance_interval = 1000;
2965 s->fake_iface = port->cfg->bond_fake_iface;
2968 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2969 long long stable_id;
2971 stable_id = atoll(get_interface_other_config(iface->cfg,
2972 "bond-stable-id", "0"));
2973 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2974 stable_id = iface->ofp_port;
2976 bond_stable_ids[i++] = stable_id;
2978 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2982 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2983 * instead of obtaining it from the database. */
2985 port_is_synthetic(const struct port *port)
2987 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2990 /* Interface functions. */
2992 static struct iface *
2993 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2995 struct bridge *br = port->bridge;
2996 struct iface *iface;
2997 char *name = if_cfg->name;
2999 iface = xzalloc(sizeof *iface);
3001 iface->name = xstrdup(name);
3002 iface->ofp_port = -1;
3003 iface->tag = tag_create_random();
3004 iface->netdev = NULL;
3005 iface->cfg = if_cfg;
3006 iface->need_refresh = true;
3008 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
3010 list_push_back(&port->ifaces, &iface->port_elem);
3012 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3018 iface_destroy(struct iface *iface)
3021 struct port *port = iface->port;
3022 struct bridge *br = port->bridge;
3024 if (br->ofproto && iface->ofp_port >= 0) {
3025 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3028 if (iface->ofp_port >= 0) {
3029 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3032 list_remove(&iface->port_elem);
3033 hmap_remove(&br->iface_by_name, &iface->name_node);
3035 netdev_close(iface->netdev);
3042 static struct iface *
3043 iface_lookup(const struct bridge *br, const char *name)
3045 struct iface *iface;
3047 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3048 &br->iface_by_name) {
3049 if (!strcmp(iface->name, name)) {
3057 static struct iface *
3058 iface_find(const char *name)
3060 const struct bridge *br;
3062 HMAP_FOR_EACH (br, node, &all_bridges) {
3063 struct iface *iface = iface_lookup(br, name);
3072 static struct iface *
3073 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3075 struct iface *iface;
3077 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3078 hash_int(ofp_port, 0), &br->ifaces) {
3079 if (iface->ofp_port == ofp_port) {
3086 /* Set Ethernet address of 'iface', if one is specified in the configuration
3089 iface_set_mac(struct iface *iface)
3091 uint8_t ea[ETH_ADDR_LEN];
3093 if (!strcmp(iface->type, "internal")
3094 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3095 if (iface->ofp_port == OFPP_LOCAL) {
3096 VLOG_ERR("interface %s: ignoring mac in Interface record "
3097 "(use Bridge record to set local port's mac)",
3099 } else if (eth_addr_is_multicast(ea)) {
3100 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3103 int error = netdev_set_etheraddr(iface->netdev, ea);
3105 VLOG_ERR("interface %s: setting MAC failed (%s)",
3106 iface->name, strerror(error));
3112 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3114 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3116 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3117 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3121 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3122 * sets the "ofport" field to -1.
3124 * This is appropriate when 'if_cfg''s interface cannot be created or is
3125 * otherwise invalid. */
3127 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3129 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3130 iface_set_ofport(if_cfg, -1);
3131 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3132 ovsrec_interface_set_admin_state(if_cfg, NULL);
3133 ovsrec_interface_set_duplex(if_cfg, NULL);
3134 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3135 ovsrec_interface_set_link_state(if_cfg, NULL);
3136 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3137 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3138 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3139 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3140 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3141 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3145 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3147 * The value strings in '*shash' are taken directly from values[], not copied,
3148 * so the caller should not modify or free them. */
3150 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3151 struct shash *shash)
3156 for (i = 0; i < n; i++) {
3157 shash_add(shash, keys[i], values[i]);
3161 /* Creates 'keys' and 'values' arrays from 'shash'.
3163 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3164 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3165 * are populated with with strings taken directly from 'shash' and thus have
3166 * the same ownership of the key-value pairs in shash.
3169 shash_to_ovs_idl_map(struct shash *shash,
3170 char ***keys, char ***values, size_t *n)
3174 struct shash_node *sn;
3176 count = shash_count(shash);
3178 k = xmalloc(count * sizeof *k);
3179 v = xmalloc(count * sizeof *v);
3182 SHASH_FOR_EACH(sn, shash) {
3193 struct iface_delete_queues_cbdata {
3194 struct netdev *netdev;
3195 const struct ovsdb_datum *queues;
3199 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3201 union ovsdb_atom atom;
3203 atom.integer = target;
3204 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3208 iface_delete_queues(unsigned int queue_id,
3209 const struct shash *details OVS_UNUSED, void *cbdata_)
3211 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3213 if (!queue_ids_include(cbdata->queues, queue_id)) {
3214 netdev_delete_queue(cbdata->netdev, queue_id);
3219 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3221 struct ofpbuf queues_buf;
3223 ofpbuf_init(&queues_buf, 0);
3225 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3226 netdev_set_qos(iface->netdev, NULL, NULL);
3228 struct iface_delete_queues_cbdata cbdata;
3229 struct shash details;
3233 /* Configure top-level Qos for 'iface'. */
3234 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3235 qos->n_other_config, &details);
3236 netdev_set_qos(iface->netdev, qos->type, &details);
3237 shash_destroy(&details);
3239 /* Deconfigure queues that were deleted. */
3240 cbdata.netdev = iface->netdev;
3241 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3243 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3245 /* Configure queues for 'iface'. */
3247 for (i = 0; i < qos->n_queues; i++) {
3248 const struct ovsrec_queue *queue = qos->value_queues[i];
3249 unsigned int queue_id = qos->key_queues[i];
3251 if (queue_id == 0) {
3255 if (queue->n_dscp == 1) {
3256 struct ofproto_port_queue *port_queue;
3258 port_queue = ofpbuf_put_uninit(&queues_buf,
3259 sizeof *port_queue);
3260 port_queue->queue = queue_id;
3261 port_queue->dscp = queue->dscp[0];
3264 shash_from_ovs_idl_map(queue->key_other_config,
3265 queue->value_other_config,
3266 queue->n_other_config, &details);
3267 netdev_set_queue(iface->netdev, queue_id, &details);
3268 shash_destroy(&details);
3271 shash_init(&details);
3272 netdev_set_queue(iface->netdev, 0, &details);
3273 shash_destroy(&details);
3277 if (iface->ofp_port >= 0) {
3278 const struct ofproto_port_queue *port_queues = queues_buf.data;
3279 size_t n_queues = queues_buf.size / sizeof *port_queues;
3281 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3282 port_queues, n_queues);
3285 netdev_set_policing(iface->netdev,
3286 iface->cfg->ingress_policing_rate,
3287 iface->cfg->ingress_policing_burst);
3289 ofpbuf_uninit(&queues_buf);
3293 iface_configure_cfm(struct iface *iface)
3295 const struct ovsrec_interface *cfg = iface->cfg;
3296 const char *extended_str, *opstate_str;
3297 struct cfm_settings s;
3299 if (!cfg->n_cfm_mpid) {
3300 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3304 s.mpid = *cfg->cfm_mpid;
3305 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3307 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3309 s.ccm_pcp = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_pcp",
3311 if (s.interval <= 0) {
3315 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3317 s.extended = !strcasecmp("true", extended_str);
3319 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3320 s.opup = !strcasecmp("up", opstate_str);
3322 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3325 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3326 * instead of obtaining it from the database. */
3328 iface_is_synthetic(const struct iface *iface)
3330 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3333 /* Port mirroring. */
3335 static struct mirror *
3336 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3340 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3341 if (uuid_equals(uuid, &m->uuid)) {
3349 bridge_configure_mirrors(struct bridge *br)
3351 const struct ovsdb_datum *mc;
3352 unsigned long *flood_vlans;
3353 struct mirror *m, *next;
3356 /* Get rid of deleted mirrors. */
3357 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3358 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3359 union ovsdb_atom atom;
3361 atom.uuid = m->uuid;
3362 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3367 /* Add new mirrors and reconfigure existing ones. */
3368 for (i = 0; i < br->cfg->n_mirrors; i++) {
3369 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3370 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3372 m = mirror_create(br, cfg);
3375 if (!mirror_configure(m)) {
3380 /* Update flooded vlans (for RSPAN). */
3381 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3382 br->cfg->n_flood_vlans);
3383 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3384 bitmap_free(flood_vlans);
3387 static struct mirror *
3388 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3392 m = xzalloc(sizeof *m);
3393 m->uuid = cfg->header_.uuid;
3394 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3396 m->name = xstrdup(cfg->name);
3402 mirror_destroy(struct mirror *m)
3405 struct bridge *br = m->bridge;
3408 ofproto_mirror_unregister(br->ofproto, m);
3411 hmap_remove(&br->mirrors, &m->hmap_node);
3418 mirror_collect_ports(struct mirror *m,
3419 struct ovsrec_port **in_ports, int n_in_ports,
3420 void ***out_portsp, size_t *n_out_portsp)
3422 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3423 size_t n_out_ports = 0;
3426 for (i = 0; i < n_in_ports; i++) {
3427 const char *name = in_ports[i]->name;
3428 struct port *port = port_lookup(m->bridge, name);
3430 out_ports[n_out_ports++] = port;
3432 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3433 "port %s", m->bridge->name, m->name, name);
3436 *out_portsp = out_ports;
3437 *n_out_portsp = n_out_ports;
3441 mirror_configure(struct mirror *m)
3443 const struct ovsrec_mirror *cfg = m->cfg;
3444 struct ofproto_mirror_settings s;
3447 if (strcmp(cfg->name, m->name)) {
3449 m->name = xstrdup(cfg->name);
3453 /* Get output port or VLAN. */
3454 if (cfg->output_port) {
3455 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3456 if (!s.out_bundle) {
3457 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3458 m->bridge->name, m->name);
3461 s.out_vlan = UINT16_MAX;
3463 if (cfg->output_vlan) {
3464 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3465 "output vlan; ignoring output vlan",
3466 m->bridge->name, m->name);
3468 } else if (cfg->output_vlan) {
3469 /* The database should prevent invalid VLAN values. */
3470 s.out_bundle = NULL;
3471 s.out_vlan = *cfg->output_vlan;
3473 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3474 m->bridge->name, m->name);
3478 /* Get port selection. */
3479 if (cfg->select_all) {
3480 size_t n_ports = hmap_count(&m->bridge->ports);
3481 void **ports = xmalloc(n_ports * sizeof *ports);
3486 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3496 /* Get ports, dropping ports that don't exist.
3497 * The IDL ensures that there are no duplicates. */
3498 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3499 &s.srcs, &s.n_srcs);
3500 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3501 &s.dsts, &s.n_dsts);
3504 /* Get VLAN selection. */
3505 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3508 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3511 if (s.srcs != s.dsts) {
3520 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3522 * This is deprecated. It is only for compatibility with broken device drivers
3523 * in old versions of Linux that do not properly support VLANs when VLAN
3524 * devices are not used. When broken device drivers are no longer in
3525 * widespread use, we will delete these interfaces. */
3527 static void **blocks;
3528 static size_t n_blocks, allocated_blocks;
3530 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3531 * VLAN splinters are reconfigured. */
3533 register_block(void *block)
3535 if (n_blocks >= allocated_blocks) {
3536 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3538 blocks[n_blocks++] = block;
3541 /* Frees all of the blocks registered with register_block(). */
3543 free_registered_blocks(void)
3547 for (i = 0; i < n_blocks; i++) {
3553 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3556 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3560 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3561 return !strcmp(value, "true");
3564 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3567 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3568 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3569 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3572 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3573 * use, returns NULL.
3575 * Updates 'vlan_splinters_enabled_anywhere'. */
3576 static unsigned long int *
3577 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3579 unsigned long int *splinter_vlans;
3580 struct sset splinter_ifaces;
3581 const char *real_dev_name;
3582 struct shash *real_devs;
3583 struct shash_node *node;
3587 /* Free space allocated for synthesized ports and interfaces, since we're
3588 * in the process of reconstructing all of them. */
3589 free_registered_blocks();
3591 splinter_vlans = bitmap_allocate(4096);
3592 sset_init(&splinter_ifaces);
3593 vlan_splinters_enabled_anywhere = false;
3594 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3595 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3598 for (j = 0; j < br_cfg->n_ports; j++) {
3599 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3602 for (k = 0; k < port_cfg->n_interfaces; k++) {
3603 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3605 if (vlan_splinters_is_enabled(iface_cfg)) {
3606 vlan_splinters_enabled_anywhere = true;
3607 sset_add(&splinter_ifaces, iface_cfg->name);
3608 vlan_bitmap_from_array__(port_cfg->trunks,
3614 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3615 bitmap_set1(splinter_vlans, *port_cfg->tag);
3620 if (!vlan_splinters_enabled_anywhere) {
3621 free(splinter_vlans);
3622 sset_destroy(&splinter_ifaces);
3626 HMAP_FOR_EACH (br, node, &all_bridges) {
3628 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3632 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3633 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3634 * device to be created for it. */
3635 bitmap_set0(splinter_vlans, 0);
3636 bitmap_set0(splinter_vlans, 4095);
3638 /* Delete all VLAN devices that we don't need. */
3640 real_devs = vlandev_get_real_devs();
3641 SHASH_FOR_EACH (node, real_devs) {
3642 const struct vlan_real_dev *real_dev = node->data;
3643 const struct vlan_dev *vlan_dev;
3644 bool real_dev_has_splinters;
3646 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3648 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3649 if (!real_dev_has_splinters
3650 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3651 struct netdev *netdev;
3653 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3654 if (!netdev_get_in4(netdev, NULL, NULL) ||
3655 !netdev_get_in6(netdev, NULL)) {
3656 vlandev_del(vlan_dev->name);
3658 /* It has an IP address configured, so we don't own
3659 * it. Don't delete it. */
3661 netdev_close(netdev);
3668 /* Add all VLAN devices that we need. */
3669 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3672 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3673 if (!vlandev_get_name(real_dev_name, vid)) {
3674 vlandev_add(real_dev_name, vid);
3681 sset_destroy(&splinter_ifaces);
3683 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3684 free(splinter_vlans);
3687 return splinter_vlans;
3690 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3693 configure_splinter_port(struct port *port)
3695 struct ofproto *ofproto = port->bridge->ofproto;
3696 uint16_t realdev_ofp_port;
3697 const char *realdev_name;
3698 struct iface *vlandev, *realdev;
3700 ofproto_bundle_unregister(port->bridge->ofproto, port);
3702 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3705 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3706 realdev = iface_lookup(port->bridge, realdev_name);
3707 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3709 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3713 static struct ovsrec_port *
3714 synthesize_splinter_port(const char *real_dev_name,
3715 const char *vlan_dev_name, int vid)
3717 struct ovsrec_interface *iface;
3718 struct ovsrec_port *port;
3720 iface = xzalloc(sizeof *iface);
3721 iface->name = xstrdup(vlan_dev_name);
3722 iface->type = "system";
3724 port = xzalloc(sizeof *port);
3725 port->interfaces = xmemdup(&iface, sizeof iface);
3726 port->n_interfaces = 1;
3727 port->name = xstrdup(vlan_dev_name);
3728 port->vlan_mode = "splinter";
3729 port->tag = xmalloc(sizeof *port->tag);
3731 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3732 port->key_other_config[0] = "realdev";
3733 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3734 port->value_other_config[0] = xstrdup(real_dev_name);
3735 port->n_other_config = 1;
3737 register_block(iface);
3738 register_block(iface->name);
3739 register_block(port);
3740 register_block(port->interfaces);
3741 register_block(port->name);
3742 register_block(port->tag);
3743 register_block(port->key_other_config);
3744 register_block(port->value_other_config);
3745 register_block(port->value_other_config[0]);
3750 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3751 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3752 * 1-bit in the 'splinter_vlans' bitmap. */
3754 add_vlan_splinter_ports(struct bridge *br,
3755 const unsigned long int *splinter_vlans,
3756 struct shash *ports)
3760 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3761 * we're modifying 'ports'. */
3762 for (i = 0; i < br->cfg->n_ports; i++) {
3763 const char *name = br->cfg->ports[i]->name;
3764 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3767 for (j = 0; j < port_cfg->n_interfaces; j++) {
3768 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3770 if (vlan_splinters_is_enabled(iface_cfg)) {
3771 const char *real_dev_name;
3774 real_dev_name = iface_cfg->name;
3775 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3776 const char *vlan_dev_name;
3778 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3780 && !shash_find(ports, vlan_dev_name)) {
3781 shash_add(ports, vlan_dev_name,
3782 synthesize_splinter_port(
3783 real_dev_name, vlan_dev_name, vid));
3792 mirror_refresh_stats(struct mirror *m)
3794 struct ofproto *ofproto = m->bridge->ofproto;
3795 uint64_t tx_packets, tx_bytes;
3798 size_t stat_cnt = 0;
3800 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3801 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3805 if (tx_packets != UINT64_MAX) {
3806 keys[stat_cnt] = "tx_packets";
3807 values[stat_cnt] = tx_packets;
3810 if (tx_bytes != UINT64_MAX) {
3811 keys[stat_cnt] = "tx_bytes";
3812 values[stat_cnt] = tx_bytes;
3816 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);