1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
71 /* OpenFlow port slated for removal from ofproto. */
73 struct list list_node; /* Node in bridge's ofpp_garbage. */
74 uint16_t ofp_port; /* Port to be deleted. */
78 /* These members are always valid. */
79 struct list port_elem; /* Element in struct port's "ifaces" list. */
80 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
84 /* These members are valid only after bridge_reconfigure() causes them to
86 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
87 int ofp_port; /* OpenFlow port number, -1 if unknown. */
88 struct netdev *netdev; /* Network device. */
89 const char *type; /* Usually same as cfg->type. */
90 const struct ovsrec_interface *cfg;
94 struct uuid uuid; /* UUID of this "mirror" record in database. */
95 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
96 struct bridge *bridge;
98 const struct ovsrec_mirror *cfg;
102 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
103 struct bridge *bridge;
106 const struct ovsrec_port *cfg;
108 /* An ordinary bridge port has 1 interface.
109 * A bridge port for bonding has at least 2 interfaces. */
110 struct list ifaces; /* List of "struct iface"s. */
114 struct hmap_node node; /* In 'all_bridges'. */
115 char *name; /* User-specified arbitrary name. */
116 char *type; /* Datapath type. */
117 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
118 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
119 const struct ovsrec_bridge *cfg;
121 /* OpenFlow switch processing. */
122 struct ofproto *ofproto; /* OpenFlow switch. */
125 struct hmap ports; /* "struct port"s indexed by name. */
126 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
127 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
129 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
130 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
131 Indexed on 'cfg->name'. */
133 /* Port mirroring. */
134 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
136 /* Synthetic local port if necessary. */
137 struct ovsrec_port synth_local_port;
138 struct ovsrec_interface synth_local_iface;
139 struct ovsrec_interface *synth_local_ifacep;
142 /* All bridges, indexed by name. */
143 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
145 /* OVSDB IDL used to obtain configuration. */
146 static struct ovsdb_idl *idl;
148 /* Most recently processed IDL sequence number. */
149 static unsigned int idl_seqno;
151 /* Each time this timer expires, the bridge fetches interface and mirror
152 * statistics and pushes them into the database. */
153 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
154 static long long int iface_stats_timer = LLONG_MIN;
156 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
157 * expensive. This can cause bridge_reconfigure() to take a long time during
158 * which no other work can be done. To deal with this problem, we limit port
159 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
160 * call to bridge_reconfigure(). If there is more work to do after the limit
161 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
162 * This allows the rest of the code to catch up on important things like
163 * forwarding packets. */
164 #define OFP_PORT_ACTION_WINDOW 10
165 static bool reconfiguring = false;
167 static void add_del_bridges(const struct ovsrec_open_vswitch *);
168 static void bridge_update_ofprotos(void);
169 static void bridge_create(const struct ovsrec_bridge *);
170 static void bridge_destroy(struct bridge *);
171 static struct bridge *bridge_lookup(const char *name);
172 static unixctl_cb_func bridge_unixctl_dump_flows;
173 static unixctl_cb_func bridge_unixctl_reconnect;
174 static size_t bridge_get_controllers(const struct bridge *br,
175 struct ovsrec_controller ***controllersp);
176 static void bridge_add_del_ports(struct bridge *,
177 const unsigned long int *splinter_vlans);
178 static void bridge_refresh_ofp_port(struct bridge *);
179 static void bridge_configure_datapath_id(struct bridge *);
180 static void bridge_configure_flow_eviction_threshold(struct bridge *);
181 static void bridge_configure_netflow(struct bridge *);
182 static void bridge_configure_forward_bpdu(struct bridge *);
183 static void bridge_configure_mac_idle_time(struct bridge *);
184 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
185 static void bridge_configure_stp(struct bridge *);
186 static void bridge_configure_tables(struct bridge *);
187 static void bridge_configure_remotes(struct bridge *,
188 const struct sockaddr_in *managers,
190 static void bridge_pick_local_hw_addr(struct bridge *,
191 uint8_t ea[ETH_ADDR_LEN],
192 struct iface **hw_addr_iface);
193 static uint64_t bridge_pick_datapath_id(struct bridge *,
194 const uint8_t bridge_ea[ETH_ADDR_LEN],
195 struct iface *hw_addr_iface);
196 static void bridge_queue_if_cfg(struct bridge *,
197 const struct ovsrec_interface *,
198 const struct ovsrec_port *);
199 static uint64_t dpid_from_hash(const void *, size_t nbytes);
200 static bool bridge_has_bond_fake_iface(const struct bridge *,
202 static bool port_is_bond_fake_iface(const struct port *);
204 static unixctl_cb_func qos_unixctl_show;
206 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
207 static void port_del_ifaces(struct port *);
208 static void port_destroy(struct port *);
209 static struct port *port_lookup(const struct bridge *, const char *name);
210 static void port_configure(struct port *);
211 static struct lacp_settings *port_configure_lacp(struct port *,
212 struct lacp_settings *);
213 static void port_configure_bond(struct port *, struct bond_settings *,
214 uint32_t *bond_stable_ids);
215 static bool port_is_synthetic(const struct port *);
217 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
218 static void run_system_stats(void);
220 static void bridge_configure_mirrors(struct bridge *);
221 static struct mirror *mirror_create(struct bridge *,
222 const struct ovsrec_mirror *);
223 static void mirror_destroy(struct mirror *);
224 static bool mirror_configure(struct mirror *);
225 static void mirror_refresh_stats(struct mirror *);
227 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
228 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
229 static const char *iface_get_type(const struct ovsrec_interface *,
230 const struct ovsrec_bridge *);
231 static void iface_destroy(struct iface *);
232 static struct iface *iface_lookup(const struct bridge *, const char *name);
233 static struct iface *iface_find(const char *name);
234 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
235 static struct iface *iface_from_ofp_port(const struct bridge *,
237 static void iface_set_mac(struct iface *);
238 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
239 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
240 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
241 static void iface_configure_cfm(struct iface *);
242 static void iface_refresh_cfm_stats(struct iface *);
243 static void iface_refresh_stats(struct iface *);
244 static void iface_refresh_status(struct iface *);
245 static bool iface_is_synthetic(const struct iface *);
247 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
249 * This is deprecated. It is only for compatibility with broken device drivers
250 * in old versions of Linux that do not properly support VLANs when VLAN
251 * devices are not used. When broken device drivers are no longer in
252 * widespread use, we will delete these interfaces. */
254 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
255 static bool vlan_splinters_enabled_anywhere;
257 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
258 static unsigned long int *collect_splinter_vlans(
259 const struct ovsrec_open_vswitch *);
260 static void configure_splinter_port(struct port *);
261 static void add_vlan_splinter_ports(struct bridge *,
262 const unsigned long int *splinter_vlans,
263 struct shash *ports);
265 /* Public functions. */
267 /* Initializes the bridge module, configuring it to obtain its configuration
268 * from an OVSDB server accessed over 'remote', which should be a string in a
269 * form acceptable to ovsdb_idl_create(). */
271 bridge_init(const char *remote)
273 /* Create connection to database. */
274 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
275 idl_seqno = ovsdb_idl_get_seqno(idl);
276 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
277 ovsdb_idl_verify_write_only(idl);
279 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
280 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
281 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
282 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
283 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
284 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
287 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
288 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
289 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
291 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
292 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
293 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
294 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
296 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
297 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
311 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
313 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
314 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
315 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
316 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
318 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
320 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
322 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
323 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
325 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
327 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
329 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
335 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
337 /* Register unixctl commands. */
338 unixctl_command_register("qos/show", "interface", 1, 1,
339 qos_unixctl_show, NULL);
340 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
341 bridge_unixctl_dump_flows, NULL);
342 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
343 bridge_unixctl_reconnect, NULL);
353 struct bridge *br, *next_br;
355 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
358 ovsdb_idl_destroy(idl);
361 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
362 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
363 * responsible for freeing '*managersp' (with free()).
365 * You may be asking yourself "why does ovs-vswitchd care?", because
366 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
367 * should not be and in fact is not directly involved in that. But
368 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
369 * it has to tell in-band control where the managers are to enable that.
370 * (Thus, only managers connected in-band are collected.)
373 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
374 struct sockaddr_in **managersp, size_t *n_managersp)
376 struct sockaddr_in *managers = NULL;
377 size_t n_managers = 0;
381 /* Collect all of the potential targets from the "targets" columns of the
382 * rows pointed to by "manager_options", excluding any that are
385 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
386 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
388 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
389 sset_find_and_delete(&targets, m->target);
391 sset_add(&targets, m->target);
395 /* Now extract the targets' IP addresses. */
396 if (!sset_is_empty(&targets)) {
399 managers = xmalloc(sset_count(&targets) * sizeof *managers);
400 SSET_FOR_EACH (target, &targets) {
401 struct sockaddr_in *sin = &managers[n_managers];
403 if (stream_parse_target_with_default_ports(target,
411 sset_destroy(&targets);
413 *managersp = managers;
414 *n_managersp = n_managers;
418 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
420 unsigned long int *splinter_vlans;
423 COVERAGE_INC(bridge_reconfigure);
425 assert(!reconfiguring);
426 reconfiguring = true;
428 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
429 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
430 * configuration otherwise.
432 * This is mostly an update to bridge data structures. Nothing is pushed
433 * down to ofproto or lower layers. */
434 add_del_bridges(ovs_cfg);
435 splinter_vlans = collect_splinter_vlans(ovs_cfg);
436 HMAP_FOR_EACH (br, node, &all_bridges) {
437 bridge_add_del_ports(br, splinter_vlans);
439 free(splinter_vlans);
441 /* Delete datapaths that are no longer configured, and create ones which
442 * don't exist but should. */
443 bridge_update_ofprotos();
445 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
446 HMAP_FOR_EACH (br, node, &all_bridges) {
447 bridge_refresh_ofp_port(br);
450 /* Clear database records for "if_cfg"s which haven't been instantiated. */
451 HMAP_FOR_EACH (br, node, &all_bridges) {
452 struct if_cfg *if_cfg;
454 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
455 iface_clear_db_record(if_cfg->cfg);
459 reconfigure_system_stats(ovs_cfg);
463 bridge_reconfigure_ofp(void)
465 long long int deadline;
469 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
471 /* The kernel will reject any attempt to add a given port to a datapath if
472 * that port already belongs to a different datapath, so we must do all
473 * port deletions before any port additions. */
474 HMAP_FOR_EACH (br, node, &all_bridges) {
475 struct ofpp_garbage *garbage, *next;
477 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
478 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
479 * internal datastructures may not be consistent. Eventually, when
480 * port additions and deletions are cheaper, these calls should be
483 ofproto_port_del(br->ofproto, garbage->ofp_port);
484 list_remove(&garbage->list_node);
488 if (time_msec() >= deadline) {
495 HMAP_FOR_EACH (br, node, &all_bridges) {
496 struct if_cfg *if_cfg, *next;
498 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
499 iface_create(br, if_cfg, -1);
501 if (time_msec() >= deadline) {
511 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
513 struct sockaddr_in *managers;
514 int sflow_bridge_number;
519 assert(reconfiguring);
520 done = bridge_reconfigure_ofp();
522 /* Complete the configuration. */
523 sflow_bridge_number = 0;
524 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
525 HMAP_FOR_EACH (br, node, &all_bridges) {
528 /* We need the datapath ID early to allow LACP ports to use it as the
529 * default system ID. */
530 bridge_configure_datapath_id(br);
532 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
535 port_configure(port);
537 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
538 iface_configure_cfm(iface);
539 iface_configure_qos(iface, port->cfg->qos);
540 iface_set_mac(iface);
543 bridge_configure_mirrors(br);
544 bridge_configure_flow_eviction_threshold(br);
545 bridge_configure_forward_bpdu(br);
546 bridge_configure_mac_idle_time(br);
547 bridge_configure_remotes(br, managers, n_managers);
548 bridge_configure_netflow(br);
549 bridge_configure_sflow(br, &sflow_bridge_number);
550 bridge_configure_stp(br);
551 bridge_configure_tables(br);
556 /* ovs-vswitchd has completed initialization, so allow the process that
557 * forked us to exit successfully. */
558 daemonize_complete();
559 reconfiguring = false;
561 VLOG_INFO("%s (Open vSwitch) %s", program_name, VERSION);
567 /* Delete ofprotos which aren't configured or have the wrong type. Create
568 * ofprotos which don't exist but need to. */
570 bridge_update_ofprotos(void)
572 struct bridge *br, *next;
577 /* Delete ofprotos with no bridge or with the wrong type. */
580 ofproto_enumerate_types(&types);
581 SSET_FOR_EACH (type, &types) {
584 ofproto_enumerate_names(type, &names);
585 SSET_FOR_EACH (name, &names) {
586 br = bridge_lookup(name);
587 if (!br || strcmp(type, br->type)) {
588 ofproto_delete(name, type);
592 sset_destroy(&names);
593 sset_destroy(&types);
595 /* Add ofprotos for bridges which don't have one yet. */
596 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
604 /* Remove ports from any datapath with the same name as 'br'. If we
605 * don't do this, creating 'br''s ofproto will fail because a port with
606 * the same name as its local port already exists. */
607 HMAP_FOR_EACH (br2, node, &all_bridges) {
608 struct ofproto_port ofproto_port;
614 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
616 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
618 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
621 ofproto_port_destroy(&ofproto_port);
625 error = ofproto_create(br->name, br->type, &br->ofproto);
627 VLOG_ERR("failed to create bridge %s: %s", br->name,
635 port_configure(struct port *port)
637 const struct ovsrec_port *cfg = port->cfg;
638 struct bond_settings bond_settings;
639 struct lacp_settings lacp_settings;
640 struct ofproto_bundle_settings s;
643 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
644 configure_splinter_port(port);
653 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
654 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
655 s.slaves[s.n_slaves++] = iface->ofp_port;
660 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
664 /* Get VLAN trunks. */
667 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
671 if (cfg->vlan_mode) {
672 if (!strcmp(cfg->vlan_mode, "access")) {
673 s.vlan_mode = PORT_VLAN_ACCESS;
674 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
675 s.vlan_mode = PORT_VLAN_TRUNK;
676 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
677 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
678 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
679 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
681 /* This "can't happen" because ovsdb-server should prevent it. */
682 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
683 s.vlan_mode = PORT_VLAN_TRUNK;
687 s.vlan_mode = PORT_VLAN_ACCESS;
689 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
693 s.vlan_mode = PORT_VLAN_TRUNK;
696 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
699 /* Get LACP settings. */
700 s.lacp = port_configure_lacp(port, &lacp_settings);
704 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
705 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
706 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
709 s.lacp_slaves = NULL;
712 /* Get bond settings. */
713 if (s.n_slaves > 1) {
714 s.bond = &bond_settings;
715 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
716 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
719 s.bond_stable_ids = NULL;
721 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
722 netdev_set_miimon_interval(iface->netdev, 0);
727 ofproto_bundle_register(port->bridge->ofproto, port, &s);
733 free(s.bond_stable_ids);
736 /* Pick local port hardware address and datapath ID for 'br'. */
738 bridge_configure_datapath_id(struct bridge *br)
740 uint8_t ea[ETH_ADDR_LEN];
742 struct iface *local_iface;
743 struct iface *hw_addr_iface;
746 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
747 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
749 int error = netdev_set_etheraddr(local_iface->netdev, ea);
751 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
752 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
753 "Ethernet address: %s",
754 br->name, strerror(error));
757 memcpy(br->ea, ea, ETH_ADDR_LEN);
759 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
760 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
761 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
762 ofproto_set_datapath_id(br->ofproto, dpid);
765 dpid_string = xasprintf("%016"PRIx64, dpid);
766 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
770 /* Set NetFlow configuration on 'br'. */
772 bridge_configure_netflow(struct bridge *br)
774 struct ovsrec_netflow *cfg = br->cfg->netflow;
775 struct netflow_options opts;
778 ofproto_set_netflow(br->ofproto, NULL);
782 memset(&opts, 0, sizeof opts);
784 /* Get default NetFlow configuration from datapath.
785 * Apply overrides from 'cfg'. */
786 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
787 if (cfg->engine_type) {
788 opts.engine_type = *cfg->engine_type;
790 if (cfg->engine_id) {
791 opts.engine_id = *cfg->engine_id;
794 /* Configure active timeout interval. */
795 opts.active_timeout = cfg->active_timeout;
796 if (!opts.active_timeout) {
797 opts.active_timeout = -1;
798 } else if (opts.active_timeout < 0) {
799 VLOG_WARN("bridge %s: active timeout interval set to negative "
800 "value, using default instead (%d seconds)", br->name,
801 NF_ACTIVE_TIMEOUT_DEFAULT);
802 opts.active_timeout = -1;
805 /* Add engine ID to interface number to disambiguate bridgs? */
806 opts.add_id_to_iface = cfg->add_id_to_interface;
807 if (opts.add_id_to_iface) {
808 if (opts.engine_id > 0x7f) {
809 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
810 "another vswitch, choose an engine id less than 128",
813 if (hmap_count(&br->ports) > 508) {
814 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
815 "another port when more than 508 ports are used",
821 sset_init(&opts.collectors);
822 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
825 if (ofproto_set_netflow(br->ofproto, &opts)) {
826 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
828 sset_destroy(&opts.collectors);
831 /* Set sFlow configuration on 'br'. */
833 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
835 const struct ovsrec_sflow *cfg = br->cfg->sflow;
836 struct ovsrec_controller **controllers;
837 struct ofproto_sflow_options oso;
838 size_t n_controllers;
842 ofproto_set_sflow(br->ofproto, NULL);
846 memset(&oso, 0, sizeof oso);
848 sset_init(&oso.targets);
849 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
851 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
853 oso.sampling_rate = *cfg->sampling;
856 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
858 oso.polling_interval = *cfg->polling;
861 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
863 oso.header_len = *cfg->header;
866 oso.sub_id = (*sflow_bridge_number)++;
867 oso.agent_device = cfg->agent;
869 oso.control_ip = NULL;
870 n_controllers = bridge_get_controllers(br, &controllers);
871 for (i = 0; i < n_controllers; i++) {
872 if (controllers[i]->local_ip) {
873 oso.control_ip = controllers[i]->local_ip;
877 ofproto_set_sflow(br->ofproto, &oso);
879 sset_destroy(&oso.targets);
883 port_configure_stp(const struct ofproto *ofproto, struct port *port,
884 struct ofproto_port_stp_settings *port_s,
885 int *port_num_counter, unsigned long *port_num_bitmap)
887 const char *config_str;
890 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
891 port_s->enable = false;
894 port_s->enable = true;
897 /* STP over bonds is not supported. */
898 if (!list_is_singleton(&port->ifaces)) {
899 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
901 port_s->enable = false;
905 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
907 /* Internal ports shouldn't participate in spanning tree, so
909 if (!strcmp(iface->type, "internal")) {
910 VLOG_DBG("port %s: disable STP on internal ports", port->name);
911 port_s->enable = false;
915 /* STP on mirror output ports is not supported. */
916 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
917 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
918 port_s->enable = false;
922 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
924 unsigned long int port_num = strtoul(config_str, NULL, 0);
925 int port_idx = port_num - 1;
927 if (port_num < 1 || port_num > STP_MAX_PORTS) {
928 VLOG_ERR("port %s: invalid stp-port-num", port->name);
929 port_s->enable = false;
933 if (bitmap_is_set(port_num_bitmap, port_idx)) {
934 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
935 port->name, port_num);
936 port_s->enable = false;
939 bitmap_set1(port_num_bitmap, port_idx);
940 port_s->port_num = port_idx;
942 if (*port_num_counter >= STP_MAX_PORTS) {
943 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
944 port_s->enable = false;
948 port_s->port_num = (*port_num_counter)++;
951 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
953 port_s->path_cost = strtoul(config_str, NULL, 10);
955 enum netdev_features current;
957 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
958 /* Couldn't get speed, so assume 100Mb/s. */
959 port_s->path_cost = 19;
963 mbps = netdev_features_to_bps(current) / 1000000;
964 port_s->path_cost = stp_convert_speed_to_cost(mbps);
968 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
970 port_s->priority = strtoul(config_str, NULL, 0);
972 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
976 /* Set spanning tree configuration on 'br'. */
978 bridge_configure_stp(struct bridge *br)
980 if (!br->cfg->stp_enable) {
981 ofproto_set_stp(br->ofproto, NULL);
983 struct ofproto_stp_settings br_s;
984 const char *config_str;
986 int port_num_counter;
987 unsigned long *port_num_bitmap;
989 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
991 uint8_t ea[ETH_ADDR_LEN];
993 if (eth_addr_from_string(config_str, ea)) {
994 br_s.system_id = eth_addr_to_uint64(ea);
996 br_s.system_id = eth_addr_to_uint64(br->ea);
997 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
998 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1001 br_s.system_id = eth_addr_to_uint64(br->ea);
1004 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1006 br_s.priority = strtoul(config_str, NULL, 0);
1008 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1011 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1013 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1015 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1018 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1020 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1022 br_s.max_age = STP_DEFAULT_MAX_AGE;
1025 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1027 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1029 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1032 /* Configure STP on the bridge. */
1033 if (ofproto_set_stp(br->ofproto, &br_s)) {
1034 VLOG_ERR("bridge %s: could not enable STP", br->name);
1038 /* Users must either set the port number with the "stp-port-num"
1039 * configuration on all ports or none. If manual configuration
1040 * is not done, then we allocate them sequentially. */
1041 port_num_counter = 0;
1042 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1043 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1044 struct ofproto_port_stp_settings port_s;
1045 struct iface *iface;
1047 port_configure_stp(br->ofproto, port, &port_s,
1048 &port_num_counter, port_num_bitmap);
1050 /* As bonds are not supported, just apply configuration to
1051 * all interfaces. */
1052 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1053 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1055 VLOG_ERR("port %s: could not enable STP", port->name);
1061 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1062 && port_num_counter) {
1063 VLOG_ERR("bridge %s: must manually configure all STP port "
1064 "IDs or none, disabling", br->name);
1065 ofproto_set_stp(br->ofproto, NULL);
1067 bitmap_free(port_num_bitmap);
1072 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1074 const struct port *port = port_lookup(br, name);
1075 return port && port_is_bond_fake_iface(port);
1079 port_is_bond_fake_iface(const struct port *port)
1081 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1085 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1087 struct bridge *br, *next;
1088 struct shash new_br;
1091 /* Collect new bridges' names and types. */
1092 shash_init(&new_br);
1093 for (i = 0; i < cfg->n_bridges; i++) {
1094 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1095 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1097 if (strchr(br_cfg->name, '/')) {
1098 /* Prevent remote ovsdb-server users from accessing arbitrary
1099 * directories, e.g. consider a bridge named "../../../etc/". */
1100 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1102 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1103 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1107 /* Get rid of deleted bridges or those whose types have changed.
1108 * Update 'cfg' of bridges that still exist. */
1109 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1110 br->cfg = shash_find_data(&new_br, br->name);
1111 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1112 br->cfg->datapath_type))) {
1117 /* Add new bridges. */
1118 for (i = 0; i < cfg->n_bridges; i++) {
1119 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1120 struct bridge *br = bridge_lookup(br_cfg->name);
1122 bridge_create(br_cfg);
1126 shash_destroy(&new_br);
1130 iface_set_ofp_port(struct iface *iface, int ofp_port)
1132 struct bridge *br = iface->port->bridge;
1134 assert(iface->ofp_port < 0 && ofp_port >= 0);
1135 iface->ofp_port = ofp_port;
1136 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1137 iface_set_ofport(iface->cfg, ofp_port);
1140 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1141 * Returns 0 if successful, otherwise a positive errno value. */
1143 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1144 struct netdev *netdev)
1148 error = netdev_set_config(netdev, &iface_cfg->options);
1150 VLOG_WARN("could not configure network device %s (%s)",
1151 iface_cfg->name, strerror(error));
1156 /* This function determines whether 'ofproto_port', which is attached to
1157 * br->ofproto's datapath, is one that we want in 'br'.
1159 * If it is, it returns true, after creating an iface (if necessary),
1160 * configuring the iface's netdev according to the iface's options, and setting
1161 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1163 * If, on the other hand, 'port' should be removed, it returns false. The
1164 * caller should later detach the port from br->ofproto. */
1166 bridge_refresh_one_ofp_port(struct bridge *br,
1167 const struct ofproto_port *ofproto_port)
1169 const char *name = ofproto_port->name;
1170 const char *type = ofproto_port->type;
1171 uint16_t ofp_port = ofproto_port->ofp_port;
1173 struct iface *iface = iface_lookup(br, name);
1175 /* Check that the name-to-number mapping is one-to-one. */
1176 if (iface->ofp_port >= 0) {
1177 VLOG_WARN("bridge %s: interface %s reported twice",
1180 } else if (iface_from_ofp_port(br, ofp_port)) {
1181 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1182 br->name, ofp_port);
1186 /* There's a configured interface named 'name'. */
1187 if (strcmp(type, iface->type)
1188 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1189 /* It's the wrong type, or it's the right type but can't be
1190 * configured as the user requested, so we must destroy it. */
1193 /* It's the right type and configured correctly. keep it. */
1194 iface_set_ofp_port(iface, ofp_port);
1197 } else if (bridge_has_bond_fake_iface(br, name)
1198 && !strcmp(type, "internal")) {
1199 /* It's a bond fake iface. Keep it. */
1202 /* There's no configured interface named 'name', but there might be an
1203 * interface of that name queued to be created.
1205 * If there is, and it has the correct type, then try to configure it
1206 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1207 * delete it and later try to re-add it. */
1208 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1210 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1211 && iface_create(br, if_cfg, ofp_port));
1215 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1216 * consistent with the ofp_ports in "br->ofproto". */
1218 bridge_refresh_ofp_port(struct bridge *br)
1220 struct ofproto_port_dump dump;
1221 struct ofproto_port ofproto_port;
1222 struct port *port, *port_next;
1224 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1225 hmap_clear(&br->ifaces);
1226 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1227 struct iface *iface;
1229 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1230 iface->ofp_port = -1;
1234 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1235 * already exist in the datapath and promote them to full fledged "struct
1236 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1237 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1238 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1239 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1240 garbage->ofp_port = ofproto_port.ofp_port;
1241 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1245 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1246 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1247 * later they can be added to ofproto. */
1248 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1249 struct iface *iface, *iface_next;
1251 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1252 if (iface->ofp_port < 0) {
1253 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1254 iface_destroy(iface);
1258 if (list_is_empty(&port->ifaces)) {
1264 /* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp'
1265 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1266 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1269 * If successful, returns 0 and stores the network device in '*netdevp'. On
1270 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1272 iface_do_create(const struct bridge *br,
1273 const struct ovsrec_interface *iface_cfg,
1274 const struct ovsrec_port *port_cfg,
1275 int *ofp_portp, struct netdev **netdevp)
1277 struct netdev *netdev;
1280 error = netdev_open(iface_cfg->name,
1281 iface_get_type(iface_cfg, br->cfg), &netdev);
1283 VLOG_WARN("could not open network device %s (%s)",
1284 iface_cfg->name, strerror(error));
1288 error = iface_set_netdev_config(iface_cfg, netdev);
1293 if (*ofp_portp < 0) {
1296 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1300 *ofp_portp = ofp_port;
1302 VLOG_INFO("bridge %s: added interface %s on port %d",
1303 br->name, iface_cfg->name, *ofp_portp);
1305 VLOG_DBG("bridge %s: interface %s is on port %d",
1306 br->name, iface_cfg->name, *ofp_portp);
1309 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1310 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1318 netdev_close(netdev);
1322 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1323 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1324 * automatically allocated for the iface. Takes ownership of and
1325 * deallocates 'if_cfg'.
1327 * Return true if an iface is successfully created, false otherwise. */
1329 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1331 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1332 const struct ovsrec_port *port_cfg = if_cfg->parent;
1334 struct netdev *netdev;
1335 struct iface *iface;
1339 /* Get rid of 'if_cfg' itself. We already copied out the interesting
1341 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1344 /* Do the bits that can fail up front.
1346 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1347 * internal datastructures may not be consistent. Eventually, when port
1348 * additions and deletions are cheaper, these calls should be removed. */
1350 assert(!iface_lookup(br, iface_cfg->name));
1351 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1354 iface_clear_db_record(iface_cfg);
1358 /* Get or create the port structure. */
1359 port = port_lookup(br, port_cfg->name);
1361 port = port_create(br, port_cfg);
1364 /* Create the iface structure. */
1365 iface = xzalloc(sizeof *iface);
1366 list_push_back(&port->ifaces, &iface->port_elem);
1367 hmap_insert(&br->iface_by_name, &iface->name_node,
1368 hash_string(iface_cfg->name, 0));
1370 iface->name = xstrdup(iface_cfg->name);
1371 iface->ofp_port = -1;
1372 iface->netdev = netdev;
1373 iface->type = iface_get_type(iface_cfg, br->cfg);
1374 iface->cfg = iface_cfg;
1376 iface_set_ofp_port(iface, ofp_port);
1378 /* Populate initial status in database. */
1379 iface_refresh_stats(iface);
1380 iface_refresh_status(iface);
1382 /* Add bond fake iface if necessary. */
1383 if (port_is_bond_fake_iface(port)) {
1384 struct ofproto_port ofproto_port;
1386 if (ofproto_port_query_by_name(br->ofproto, port->name,
1388 struct netdev *netdev;
1391 error = netdev_open(port->name, "internal", &netdev);
1393 ofproto_port_add(br->ofproto, netdev, NULL);
1394 netdev_close(netdev);
1396 VLOG_WARN("could not open network device %s (%s)",
1397 port->name, strerror(error));
1400 /* Already exists, nothing to do. */
1401 ofproto_port_destroy(&ofproto_port);
1408 /* Set Flow eviction threshold */
1410 bridge_configure_flow_eviction_threshold(struct bridge *br)
1412 const char *threshold_str;
1415 threshold_str = smap_get(&br->cfg->other_config,
1416 "flow-eviction-threshold");
1417 if (threshold_str) {
1418 threshold = strtoul(threshold_str, NULL, 10);
1420 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1422 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1425 /* Set forward BPDU option. */
1427 bridge_configure_forward_bpdu(struct bridge *br)
1429 ofproto_set_forward_bpdu(br->ofproto,
1430 smap_get_bool(&br->cfg->other_config,
1435 /* Set MAC aging time for 'br'. */
1437 bridge_configure_mac_idle_time(struct bridge *br)
1439 const char *idle_time_str;
1442 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1443 idle_time = (idle_time_str && atoi(idle_time_str)
1444 ? atoi(idle_time_str)
1445 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1446 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1450 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1451 struct iface **hw_addr_iface)
1453 struct hmapx mirror_output_ports;
1456 bool found_addr = false;
1460 *hw_addr_iface = NULL;
1462 /* Did the user request a particular MAC? */
1463 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1464 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1465 if (eth_addr_is_multicast(ea)) {
1466 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1467 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1468 } else if (eth_addr_is_zero(ea)) {
1469 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1475 /* Mirror output ports don't participate in picking the local hardware
1476 * address. ofproto can't help us find out whether a given port is a
1477 * mirror output because we haven't configured mirrors yet, so we need to
1478 * accumulate them ourselves. */
1479 hmapx_init(&mirror_output_ports);
1480 for (i = 0; i < br->cfg->n_mirrors; i++) {
1481 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1482 if (m->output_port) {
1483 hmapx_add(&mirror_output_ports, m->output_port);
1487 /* Otherwise choose the minimum non-local MAC address among all of the
1489 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1490 uint8_t iface_ea[ETH_ADDR_LEN];
1491 struct iface *candidate;
1492 struct iface *iface;
1494 /* Mirror output ports don't participate. */
1495 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1499 /* Choose the MAC address to represent the port. */
1501 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1502 /* Find the interface with this Ethernet address (if any) so that
1503 * we can provide the correct devname to the caller. */
1504 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1505 uint8_t candidate_ea[ETH_ADDR_LEN];
1506 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1507 && eth_addr_equals(iface_ea, candidate_ea)) {
1512 /* Choose the interface whose MAC address will represent the port.
1513 * The Linux kernel bonding code always chooses the MAC address of
1514 * the first slave added to a bond, and the Fedora networking
1515 * scripts always add slaves to a bond in alphabetical order, so
1516 * for compatibility we choose the interface with the name that is
1517 * first in alphabetical order. */
1518 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1519 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1524 /* The local port doesn't count (since we're trying to choose its
1525 * MAC address anyway). */
1526 if (iface->ofp_port == OFPP_LOCAL) {
1531 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1537 /* Compare against our current choice. */
1538 if (!eth_addr_is_multicast(iface_ea) &&
1539 !eth_addr_is_local(iface_ea) &&
1540 !eth_addr_is_reserved(iface_ea) &&
1541 !eth_addr_is_zero(iface_ea) &&
1542 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1544 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1545 *hw_addr_iface = iface;
1550 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1551 br->name, ETH_ADDR_ARGS(ea));
1553 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1554 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1555 *hw_addr_iface = NULL;
1556 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1557 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1560 hmapx_destroy(&mirror_output_ports);
1563 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1564 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1565 * an interface on 'br', then that interface must be passed in as
1566 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1567 * 'hw_addr_iface' must be passed in as a null pointer. */
1569 bridge_pick_datapath_id(struct bridge *br,
1570 const uint8_t bridge_ea[ETH_ADDR_LEN],
1571 struct iface *hw_addr_iface)
1574 * The procedure for choosing a bridge MAC address will, in the most
1575 * ordinary case, also choose a unique MAC that we can use as a datapath
1576 * ID. In some special cases, though, multiple bridges will end up with
1577 * the same MAC address. This is OK for the bridges, but it will confuse
1578 * the OpenFlow controller, because each datapath needs a unique datapath
1581 * Datapath IDs must be unique. It is also very desirable that they be
1582 * stable from one run to the next, so that policy set on a datapath
1585 const char *datapath_id;
1588 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1589 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1593 if (!hw_addr_iface) {
1595 * A purely internal bridge, that is, one that has no non-virtual
1596 * network devices on it at all, is difficult because it has no
1597 * natural unique identifier at all.
1599 * When the host is a XenServer, we handle this case by hashing the
1600 * host's UUID with the name of the bridge. Names of bridges are
1601 * persistent across XenServer reboots, although they can be reused if
1602 * an internal network is destroyed and then a new one is later
1603 * created, so this is fairly effective.
1605 * When the host is not a XenServer, we punt by using a random MAC
1606 * address on each run.
1608 const char *host_uuid = xenserver_get_host_uuid();
1610 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1611 dpid = dpid_from_hash(combined, strlen(combined));
1617 return eth_addr_to_uint64(bridge_ea);
1621 dpid_from_hash(const void *data, size_t n)
1623 uint8_t hash[SHA1_DIGEST_SIZE];
1625 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1626 sha1_bytes(data, n, hash);
1627 eth_addr_mark_random(hash);
1628 return eth_addr_to_uint64(hash);
1632 iface_refresh_status(struct iface *iface)
1636 enum netdev_features current;
1642 if (iface_is_synthetic(iface)) {
1648 if (!netdev_get_drv_info(iface->netdev, &smap)) {
1649 ovsrec_interface_set_status(iface->cfg, &smap);
1651 ovsrec_interface_set_status(iface->cfg, NULL);
1654 smap_destroy(&smap);
1656 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1658 ovsrec_interface_set_duplex(iface->cfg,
1659 netdev_features_is_full_duplex(current)
1661 /* warning: uint64_t -> int64_t conversion */
1662 bps = netdev_features_to_bps(current);
1663 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1666 ovsrec_interface_set_duplex(iface->cfg, NULL);
1667 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1670 error = netdev_get_mtu(iface->netdev, &mtu);
1673 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1676 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1680 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1683 iface_refresh_cfm_stats(struct iface *iface)
1685 const struct ovsrec_interface *cfg = iface->cfg;
1686 int fault, opup, error;
1687 const uint64_t *rmps;
1691 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1694 const char *reasons[CFM_FAULT_N_REASONS];
1695 bool fault_bool = fault;
1699 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1700 int reason = 1 << i;
1701 if (fault & reason) {
1702 reasons[j++] = cfm_fault_reason_to_str(reason);
1706 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1707 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1709 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1710 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1713 opup = ofproto_port_get_cfm_opup(iface->port->bridge->ofproto,
1716 ovsrec_interface_set_cfm_remote_opstate(cfg, opup ? "up" : "down");
1718 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1721 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1722 iface->ofp_port, &rmps, &n_rmps);
1724 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1727 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1730 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1733 int64_t cfm_health = health;
1734 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1736 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1741 iface_refresh_stats(struct iface *iface)
1743 #define IFACE_STATS \
1744 IFACE_STAT(rx_packets, "rx_packets") \
1745 IFACE_STAT(tx_packets, "tx_packets") \
1746 IFACE_STAT(rx_bytes, "rx_bytes") \
1747 IFACE_STAT(tx_bytes, "tx_bytes") \
1748 IFACE_STAT(rx_dropped, "rx_dropped") \
1749 IFACE_STAT(tx_dropped, "tx_dropped") \
1750 IFACE_STAT(rx_errors, "rx_errors") \
1751 IFACE_STAT(tx_errors, "tx_errors") \
1752 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1753 IFACE_STAT(rx_over_errors, "rx_over_err") \
1754 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1755 IFACE_STAT(collisions, "collisions")
1757 #define IFACE_STAT(MEMBER, NAME) NAME,
1758 static char *keys[] = { IFACE_STATS };
1760 int64_t values[ARRAY_SIZE(keys)];
1763 struct netdev_stats stats;
1765 if (iface_is_synthetic(iface)) {
1769 /* Intentionally ignore return value, since errors will set 'stats' to
1770 * all-1s, and we will deal with that correctly below. */
1771 netdev_get_stats(iface->netdev, &stats);
1773 /* Copy statistics into values[] array. */
1775 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1778 assert(i == ARRAY_SIZE(keys));
1780 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1786 br_refresh_stp_status(struct bridge *br)
1788 struct smap smap = SMAP_INITIALIZER(&smap);
1789 struct ofproto *ofproto = br->ofproto;
1790 struct ofproto_stp_status status;
1792 if (ofproto_get_stp_status(ofproto, &status)) {
1796 if (!status.enabled) {
1797 ovsrec_bridge_set_status(br->cfg, NULL);
1801 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1802 STP_ID_ARGS(status.bridge_id));
1803 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1804 STP_ID_ARGS(status.designated_root));
1805 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1807 ovsrec_bridge_set_status(br->cfg, &smap);
1808 smap_destroy(&smap);
1812 port_refresh_stp_status(struct port *port)
1814 struct ofproto *ofproto = port->bridge->ofproto;
1815 struct iface *iface;
1816 struct ofproto_port_stp_status status;
1818 int64_t int_values[3];
1821 if (port_is_synthetic(port)) {
1825 /* STP doesn't currently support bonds. */
1826 if (!list_is_singleton(&port->ifaces)) {
1827 ovsrec_port_set_status(port->cfg, NULL);
1831 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1833 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1837 if (!status.enabled) {
1838 ovsrec_port_set_status(port->cfg, NULL);
1839 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1843 /* Set Status column. */
1845 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1846 smap_add(&smap, "stp_state", stp_state_name(status.state));
1847 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1848 smap_add(&smap, "stp_role", stp_role_name(status.role));
1849 ovsrec_port_set_status(port->cfg, &smap);
1850 smap_destroy(&smap);
1852 /* Set Statistics column. */
1853 keys[0] = "stp_tx_count";
1854 int_values[0] = status.tx_count;
1855 keys[1] = "stp_rx_count";
1856 int_values[1] = status.rx_count;
1857 keys[2] = "stp_error_count";
1858 int_values[2] = status.error_count;
1860 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1861 ARRAY_SIZE(int_values));
1865 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1867 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1871 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1873 bool enable = enable_system_stats(cfg);
1875 system_stats_enable(enable);
1877 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1882 run_system_stats(void)
1884 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1887 stats = system_stats_run();
1889 struct ovsdb_idl_txn *txn;
1890 struct ovsdb_datum datum;
1892 txn = ovsdb_idl_txn_create(idl);
1893 ovsdb_datum_from_smap(&datum, stats);
1894 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1896 ovsdb_idl_txn_commit(txn);
1897 ovsdb_idl_txn_destroy(txn);
1903 static inline const char *
1904 nx_role_to_str(enum nx_role role)
1909 case NX_ROLE_MASTER:
1914 return "*** INVALID ROLE ***";
1919 refresh_controller_status(void)
1923 const struct ovsrec_controller *cfg;
1927 /* Accumulate status for controllers on all bridges. */
1928 HMAP_FOR_EACH (br, node, &all_bridges) {
1929 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1932 /* Update each controller in the database with current status. */
1933 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1934 struct ofproto_controller_info *cinfo =
1935 shash_find_data(&info, cfg->target);
1938 struct smap smap = SMAP_INITIALIZER(&smap);
1939 const char **values = cinfo->pairs.values;
1940 const char **keys = cinfo->pairs.keys;
1943 for (i = 0; i < cinfo->pairs.n; i++) {
1944 smap_add(&smap, keys[i], values[i]);
1947 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1948 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1949 ovsrec_controller_set_status(cfg, &smap);
1950 smap_destroy(&smap);
1952 ovsrec_controller_set_is_connected(cfg, false);
1953 ovsrec_controller_set_role(cfg, NULL);
1954 ovsrec_controller_set_status(cfg, NULL);
1958 ofproto_free_ofproto_controller_info(&info);
1962 refresh_instant_stats(void)
1964 static struct ovsdb_idl_txn *txn = NULL;
1969 txn = ovsdb_idl_txn_create(idl);
1971 HMAP_FOR_EACH (br, node, &all_bridges) {
1972 struct iface *iface;
1975 br_refresh_stp_status(br);
1977 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1978 port_refresh_stp_status(port);
1981 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1982 enum netdev_flags flags;
1983 const char *link_state;
1984 int64_t link_resets;
1987 if (iface_is_synthetic(iface)) {
1991 current = ofproto_port_is_lacp_current(br->ofproto,
1995 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1997 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2000 error = netdev_get_flags(iface->netdev, &flags);
2002 const char *state = flags & NETDEV_UP ? "up" : "down";
2003 ovsrec_interface_set_admin_state(iface->cfg, state);
2005 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2008 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2009 ovsrec_interface_set_link_state(iface->cfg, link_state);
2011 link_resets = netdev_get_carrier_resets(iface->netdev);
2012 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2014 iface_refresh_cfm_stats(iface);
2019 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
2020 ovsdb_idl_txn_destroy(txn);
2025 /* Performs periodic activity required by bridges that needs to be done with
2026 * the least possible latency.
2028 * It makes sense to call this function a couple of times per poll loop, to
2029 * provide a significant performance boost on some benchmarks with ofprotos
2030 * that use the ofproto-dpif implementation. */
2032 bridge_run_fast(void)
2036 HMAP_FOR_EACH (br, node, &all_bridges) {
2037 ofproto_run_fast(br->ofproto);
2044 static const struct ovsrec_open_vswitch null_cfg;
2045 const struct ovsrec_open_vswitch *cfg;
2046 struct ovsdb_idl_txn *reconf_txn = NULL;
2048 bool vlan_splinters_changed;
2051 ovsrec_open_vswitch_init((struct ovsrec_open_vswitch *) &null_cfg);
2053 /* (Re)configure if necessary. */
2054 if (!reconfiguring) {
2057 if (ovsdb_idl_is_lock_contended(idl)) {
2058 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2059 struct bridge *br, *next_br;
2061 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2062 "disabling this process until it goes away");
2064 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2068 } else if (!ovsdb_idl_has_lock(idl)) {
2072 cfg = ovsrec_open_vswitch_first(idl);
2074 /* Let each bridge do the work that it needs to do. */
2075 HMAP_FOR_EACH (br, node, &all_bridges) {
2076 ofproto_run(br->ofproto);
2079 /* Re-configure SSL. We do this on every trip through the main loop,
2080 * instead of just when the database changes, because the contents of the
2081 * key and certificate files can change without the database changing.
2083 * We do this before bridge_reconfigure() because that function might
2084 * initiate SSL connections and thus requires SSL to be configured. */
2085 if (cfg && cfg->ssl) {
2086 const struct ovsrec_ssl *ssl = cfg->ssl;
2088 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2089 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2092 if (!reconfiguring) {
2093 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2094 * usage has changed. */
2095 vlan_splinters_changed = false;
2096 if (vlan_splinters_enabled_anywhere) {
2097 HMAP_FOR_EACH (br, node, &all_bridges) {
2098 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2099 vlan_splinters_changed = true;
2105 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2106 idl_seqno = ovsdb_idl_get_seqno(idl);
2108 reconf_txn = ovsdb_idl_txn_create(idl);
2109 bridge_reconfigure(cfg);
2111 /* We still need to reconfigure to avoid dangling pointers to
2112 * now-destroyed ovsrec structures inside bridge data. */
2113 bridge_reconfigure(&null_cfg);
2118 if (reconfiguring) {
2121 reconf_txn = ovsdb_idl_txn_create(idl);
2123 if (bridge_reconfigure_continue(cfg)) {
2124 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2127 bridge_reconfigure_continue(&null_cfg);
2132 ovsdb_idl_txn_commit(reconf_txn);
2133 ovsdb_idl_txn_destroy(reconf_txn);
2137 /* Refresh interface and mirror stats if necessary. */
2138 if (time_msec() >= iface_stats_timer) {
2140 struct ovsdb_idl_txn *txn;
2142 txn = ovsdb_idl_txn_create(idl);
2143 HMAP_FOR_EACH (br, node, &all_bridges) {
2147 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2148 struct iface *iface;
2150 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2151 iface_refresh_stats(iface);
2152 iface_refresh_status(iface);
2156 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2157 mirror_refresh_stats(m);
2161 refresh_controller_status();
2162 ovsdb_idl_txn_commit(txn);
2163 ovsdb_idl_txn_destroy(txn); /* XXX */
2166 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2170 refresh_instant_stats();
2176 ovsdb_idl_wait(idl);
2178 if (reconfiguring) {
2179 poll_immediate_wake();
2182 if (!hmap_is_empty(&all_bridges)) {
2185 HMAP_FOR_EACH (br, node, &all_bridges) {
2186 ofproto_wait(br->ofproto);
2188 poll_timer_wait_until(iface_stats_timer);
2191 system_stats_wait();
2194 /* Adds some memory usage statistics for bridges into 'usage', for use with
2195 * memory_report(). */
2197 bridge_get_memory_usage(struct simap *usage)
2201 HMAP_FOR_EACH (br, node, &all_bridges) {
2202 ofproto_get_memory_usage(br->ofproto, usage);
2206 /* QoS unixctl user interface functions. */
2208 struct qos_unixctl_show_cbdata {
2210 struct iface *iface;
2214 qos_unixctl_show_cb(unsigned int queue_id,
2215 const struct smap *details,
2218 struct qos_unixctl_show_cbdata *data = aux;
2219 struct ds *ds = data->ds;
2220 struct iface *iface = data->iface;
2221 struct netdev_queue_stats stats;
2222 struct smap_node *node;
2225 ds_put_cstr(ds, "\n");
2227 ds_put_format(ds, "Queue %u:\n", queue_id);
2229 ds_put_cstr(ds, "Default:\n");
2232 SMAP_FOR_EACH (node, details) {
2233 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2236 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2238 if (stats.tx_packets != UINT64_MAX) {
2239 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2242 if (stats.tx_bytes != UINT64_MAX) {
2243 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2246 if (stats.tx_errors != UINT64_MAX) {
2247 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2250 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2251 queue_id, strerror(error));
2256 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2257 const char *argv[], void *aux OVS_UNUSED)
2259 struct ds ds = DS_EMPTY_INITIALIZER;
2260 struct smap smap = SMAP_INITIALIZER(&smap);
2261 struct iface *iface;
2263 struct smap_node *node;
2264 struct qos_unixctl_show_cbdata data;
2267 iface = iface_find(argv[1]);
2269 unixctl_command_reply_error(conn, "no such interface");
2273 netdev_get_qos(iface->netdev, &type, &smap);
2275 if (*type != '\0') {
2276 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2278 SMAP_FOR_EACH (node, &smap) {
2279 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2284 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2287 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2289 unixctl_command_reply(conn, ds_cstr(&ds));
2291 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2292 unixctl_command_reply_error(conn, ds_cstr(&ds));
2295 smap_destroy(&smap);
2299 /* Bridge reconfiguration functions. */
2301 bridge_create(const struct ovsrec_bridge *br_cfg)
2305 assert(!bridge_lookup(br_cfg->name));
2306 br = xzalloc(sizeof *br);
2308 br->name = xstrdup(br_cfg->name);
2309 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2312 /* Derive the default Ethernet address from the bridge's UUID. This should
2313 * be unique and it will be stable between ovs-vswitchd runs. */
2314 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2315 eth_addr_mark_random(br->default_ea);
2317 hmap_init(&br->ports);
2318 hmap_init(&br->ifaces);
2319 hmap_init(&br->iface_by_name);
2320 hmap_init(&br->mirrors);
2322 hmap_init(&br->if_cfg_todo);
2323 list_init(&br->ofpp_garbage);
2325 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2329 bridge_destroy(struct bridge *br)
2332 struct mirror *mirror, *next_mirror;
2333 struct port *port, *next_port;
2334 struct if_cfg *if_cfg, *next_if_cfg;
2335 struct ofpp_garbage *garbage, *next_garbage;
2337 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2340 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2341 mirror_destroy(mirror);
2343 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2344 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2347 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2348 &br->ofpp_garbage) {
2349 list_remove(&garbage->list_node);
2353 hmap_remove(&all_bridges, &br->node);
2354 ofproto_destroy(br->ofproto);
2355 hmap_destroy(&br->ifaces);
2356 hmap_destroy(&br->ports);
2357 hmap_destroy(&br->iface_by_name);
2358 hmap_destroy(&br->mirrors);
2359 hmap_destroy(&br->if_cfg_todo);
2366 static struct bridge *
2367 bridge_lookup(const char *name)
2371 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2372 if (!strcmp(br->name, name)) {
2379 /* Handle requests for a listing of all flows known by the OpenFlow
2380 * stack, including those normally hidden. */
2382 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2383 const char *argv[], void *aux OVS_UNUSED)
2388 br = bridge_lookup(argv[1]);
2390 unixctl_command_reply_error(conn, "Unknown bridge");
2395 ofproto_get_all_flows(br->ofproto, &results);
2397 unixctl_command_reply(conn, ds_cstr(&results));
2398 ds_destroy(&results);
2401 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2402 * connections and reconnect. If BRIDGE is not specified, then all bridges
2403 * drop their controller connections and reconnect. */
2405 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2406 const char *argv[], void *aux OVS_UNUSED)
2410 br = bridge_lookup(argv[1]);
2412 unixctl_command_reply_error(conn, "Unknown bridge");
2415 ofproto_reconnect_controllers(br->ofproto);
2417 HMAP_FOR_EACH (br, node, &all_bridges) {
2418 ofproto_reconnect_controllers(br->ofproto);
2421 unixctl_command_reply(conn, NULL);
2425 bridge_get_controllers(const struct bridge *br,
2426 struct ovsrec_controller ***controllersp)
2428 struct ovsrec_controller **controllers;
2429 size_t n_controllers;
2431 controllers = br->cfg->controller;
2432 n_controllers = br->cfg->n_controller;
2434 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2440 *controllersp = controllers;
2442 return n_controllers;
2446 bridge_queue_if_cfg(struct bridge *br,
2447 const struct ovsrec_interface *cfg,
2448 const struct ovsrec_port *parent)
2450 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2453 if_cfg->parent = parent;
2454 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2455 hash_string(if_cfg->cfg->name, 0));
2458 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2459 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2460 * 'br' needs to complete its configuration. */
2462 bridge_add_del_ports(struct bridge *br,
2463 const unsigned long int *splinter_vlans)
2465 struct shash_node *port_node;
2466 struct port *port, *next;
2467 struct shash new_ports;
2470 assert(hmap_is_empty(&br->if_cfg_todo));
2472 /* Collect new ports. */
2473 shash_init(&new_ports);
2474 for (i = 0; i < br->cfg->n_ports; i++) {
2475 const char *name = br->cfg->ports[i]->name;
2476 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2477 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2481 if (bridge_get_controllers(br, NULL)
2482 && !shash_find(&new_ports, br->name)) {
2483 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2484 br->name, br->name);
2486 ovsrec_interface_init(&br->synth_local_iface);
2487 ovsrec_port_init(&br->synth_local_port);
2489 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2490 br->synth_local_port.n_interfaces = 1;
2491 br->synth_local_port.name = br->name;
2493 br->synth_local_iface.name = br->name;
2494 br->synth_local_iface.type = "internal";
2496 br->synth_local_ifacep = &br->synth_local_iface;
2498 shash_add(&new_ports, br->name, &br->synth_local_port);
2501 if (splinter_vlans) {
2502 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2505 /* Get rid of deleted ports.
2506 * Get rid of deleted interfaces on ports that still exist. */
2507 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2508 port->cfg = shash_find_data(&new_ports, port->name);
2512 port_del_ifaces(port);
2516 /* Update iface->cfg and iface->type in interfaces that still exist.
2517 * Add new interfaces to creation queue. */
2518 SHASH_FOR_EACH (port_node, &new_ports) {
2519 const struct ovsrec_port *port = port_node->data;
2522 for (i = 0; i < port->n_interfaces; i++) {
2523 const struct ovsrec_interface *cfg = port->interfaces[i];
2524 struct iface *iface = iface_lookup(br, cfg->name);
2525 const char *type = iface_get_type(cfg, br->cfg);
2530 } else if (!strcmp(type, "null")) {
2531 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2532 " may be removed in February 2013. Please email"
2533 " dev@openvswitch.org with concerns.",
2536 bridge_queue_if_cfg(br, cfg, port);
2541 shash_destroy(&new_ports);
2544 /* Initializes 'oc' appropriately as a management service controller for
2547 * The caller must free oc->target when it is no longer needed. */
2549 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2550 struct ofproto_controller *oc)
2552 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2553 oc->max_backoff = 0;
2554 oc->probe_interval = 60;
2555 oc->band = OFPROTO_OUT_OF_BAND;
2557 oc->burst_limit = 0;
2558 oc->enable_async_msgs = true;
2561 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2563 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2564 struct ofproto_controller *oc)
2568 oc->target = c->target;
2569 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2570 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2571 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2572 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2573 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2574 oc->burst_limit = (c->controller_burst_limit
2575 ? *c->controller_burst_limit : 0);
2576 oc->enable_async_msgs = (!c->enable_async_messages
2577 || *c->enable_async_messages);
2578 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2579 if (dscp < 0 || dscp > 63) {
2580 dscp = DSCP_DEFAULT;
2585 /* Configures the IP stack for 'br''s local interface properly according to the
2586 * configuration in 'c'. */
2588 bridge_configure_local_iface_netdev(struct bridge *br,
2589 struct ovsrec_controller *c)
2591 struct netdev *netdev;
2592 struct in_addr mask, gateway;
2594 struct iface *local_iface;
2597 /* If there's no local interface or no IP address, give up. */
2598 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2599 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2603 /* Bring up the local interface. */
2604 netdev = local_iface->netdev;
2605 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2607 /* Configure the IP address and netmask. */
2608 if (!c->local_netmask
2609 || !inet_aton(c->local_netmask, &mask)
2611 mask.s_addr = guess_netmask(ip.s_addr);
2613 if (!netdev_set_in4(netdev, ip, mask)) {
2614 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2615 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2618 /* Configure the default gateway. */
2619 if (c->local_gateway
2620 && inet_aton(c->local_gateway, &gateway)
2621 && gateway.s_addr) {
2622 if (!netdev_add_router(netdev, gateway)) {
2623 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2624 br->name, IP_ARGS(&gateway.s_addr));
2629 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2630 * in either string are treated as equal to any number of slashes in the other,
2631 * e.g. "x///y" is equal to "x/y". */
2633 equal_pathnames(const char *a, const char *b)
2637 a += strspn(a, "/");
2638 b += strspn(b, "/");
2639 } else if (*a == '\0') {
2650 bridge_configure_remotes(struct bridge *br,
2651 const struct sockaddr_in *managers, size_t n_managers)
2653 bool disable_in_band;
2655 struct ovsrec_controller **controllers;
2656 size_t n_controllers;
2658 enum ofproto_fail_mode fail_mode;
2660 struct ofproto_controller *ocs;
2664 /* Check if we should disable in-band control on this bridge. */
2665 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2668 /* Set OpenFlow queue ID for in-band control. */
2669 ofproto_set_in_band_queue(br->ofproto,
2670 smap_get_int(&br->cfg->other_config,
2671 "in-band-queue", -1));
2673 if (disable_in_band) {
2674 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2676 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2679 n_controllers = bridge_get_controllers(br, &controllers);
2681 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2684 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2685 for (i = 0; i < n_controllers; i++) {
2686 struct ovsrec_controller *c = controllers[i];
2688 if (!strncmp(c->target, "punix:", 6)
2689 || !strncmp(c->target, "unix:", 5)) {
2690 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2693 whitelist = xasprintf("unix:%s/%s.controller",
2694 ovs_rundir(), br->name);
2695 if (!equal_pathnames(c->target, whitelist)) {
2696 /* Prevent remote ovsdb-server users from accessing arbitrary
2697 * Unix domain sockets and overwriting arbitrary local
2699 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2700 "controller \"%s\" due to possibility for remote "
2701 "exploit. Instead, specify whitelisted \"%s\" or "
2702 "connect to \"unix:%s/%s.mgmt\" (which is always "
2703 "available without special configuration).",
2704 br->name, c->target, whitelist,
2705 ovs_rundir(), br->name);
2713 bridge_configure_local_iface_netdev(br, c);
2714 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2715 if (disable_in_band) {
2716 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2721 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2722 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2725 /* Set the fail-mode. */
2726 fail_mode = !br->cfg->fail_mode
2727 || !strcmp(br->cfg->fail_mode, "standalone")
2728 ? OFPROTO_FAIL_STANDALONE
2729 : OFPROTO_FAIL_SECURE;
2730 ofproto_set_fail_mode(br->ofproto, fail_mode);
2732 /* Configure OpenFlow controller connection snooping. */
2733 if (!ofproto_has_snoops(br->ofproto)) {
2737 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2738 ovs_rundir(), br->name));
2739 ofproto_set_snoops(br->ofproto, &snoops);
2740 sset_destroy(&snoops);
2745 bridge_configure_tables(struct bridge *br)
2747 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2751 n_tables = ofproto_get_n_tables(br->ofproto);
2753 for (i = 0; i < n_tables; i++) {
2754 struct ofproto_table_settings s;
2757 s.max_flows = UINT_MAX;
2761 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2762 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2765 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2766 s.max_flows = *cfg->flow_limit;
2768 if (cfg->overflow_policy
2769 && !strcmp(cfg->overflow_policy, "evict")) {
2772 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2773 for (k = 0; k < cfg->n_groups; k++) {
2774 const char *string = cfg->groups[k];
2777 msg = mf_parse_subfield__(&s.groups[k], &string);
2779 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2780 "'groups' (%s)", br->name, i, msg);
2782 } else if (*string) {
2783 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2784 "element '%s' contains trailing garbage",
2785 br->name, i, cfg->groups[k]);
2793 ofproto_configure_table(br->ofproto, i, &s);
2797 for (; j < br->cfg->n_flow_tables; j++) {
2798 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2799 "%"PRId64" not supported by this datapath", br->name,
2800 br->cfg->key_flow_tables[j]);
2804 /* Port functions. */
2806 static struct port *
2807 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2811 port = xzalloc(sizeof *port);
2813 port->name = xstrdup(cfg->name);
2815 list_init(&port->ifaces);
2817 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2821 /* Deletes interfaces from 'port' that are no longer configured for it. */
2823 port_del_ifaces(struct port *port)
2825 struct iface *iface, *next;
2826 struct sset new_ifaces;
2829 /* Collect list of new interfaces. */
2830 sset_init(&new_ifaces);
2831 for (i = 0; i < port->cfg->n_interfaces; i++) {
2832 const char *name = port->cfg->interfaces[i]->name;
2833 const char *type = port->cfg->interfaces[i]->type;
2834 if (strcmp(type, "null")) {
2835 sset_add(&new_ifaces, name);
2839 /* Get rid of deleted interfaces. */
2840 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2841 if (!sset_contains(&new_ifaces, iface->name)) {
2842 iface_destroy(iface);
2846 sset_destroy(&new_ifaces);
2850 port_destroy(struct port *port)
2853 struct bridge *br = port->bridge;
2854 struct iface *iface, *next;
2857 ofproto_bundle_unregister(br->ofproto, port);
2860 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2861 iface_destroy(iface);
2864 hmap_remove(&br->ports, &port->hmap_node);
2870 static struct port *
2871 port_lookup(const struct bridge *br, const char *name)
2875 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2877 if (!strcmp(port->name, name)) {
2885 enable_lacp(struct port *port, bool *activep)
2887 if (!port->cfg->lacp) {
2888 /* XXX when LACP implementation has been sufficiently tested, enable by
2889 * default and make active on bonded ports. */
2891 } else if (!strcmp(port->cfg->lacp, "off")) {
2893 } else if (!strcmp(port->cfg->lacp, "active")) {
2896 } else if (!strcmp(port->cfg->lacp, "passive")) {
2900 VLOG_WARN("port %s: unknown LACP mode %s",
2901 port->name, port->cfg->lacp);
2906 static struct lacp_settings *
2907 port_configure_lacp(struct port *port, struct lacp_settings *s)
2909 const char *lacp_time, *system_id;
2912 if (!enable_lacp(port, &s->active)) {
2916 s->name = port->name;
2918 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
2920 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2921 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2922 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2923 " address.", port->name, system_id);
2927 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2930 if (eth_addr_is_zero(s->id)) {
2931 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2935 /* Prefer bondable links if unspecified. */
2936 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
2938 s->priority = (priority > 0 && priority <= UINT16_MAX
2940 : UINT16_MAX - !list_is_short(&port->ifaces));
2942 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
2943 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
2948 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2950 int priority, portid, key;
2952 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
2953 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
2955 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
2957 if (portid <= 0 || portid > UINT16_MAX) {
2958 portid = iface->ofp_port;
2961 if (priority <= 0 || priority > UINT16_MAX) {
2962 priority = UINT16_MAX;
2965 if (key < 0 || key > UINT16_MAX) {
2969 s->name = iface->name;
2971 s->priority = priority;
2976 port_configure_bond(struct port *port, struct bond_settings *s,
2977 uint32_t *bond_stable_ids)
2979 const char *detect_s;
2980 struct iface *iface;
2981 int miimon_interval;
2984 s->name = port->name;
2986 if (port->cfg->bond_mode) {
2987 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2988 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2989 port->name, port->cfg->bond_mode,
2990 bond_mode_to_string(s->balance));
2993 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2995 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2996 * active-backup. At some point we should remove this warning. */
2997 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2998 " in previous versions, the default bond_mode was"
2999 " balance-slb", port->name,
3000 bond_mode_to_string(s->balance));
3002 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3003 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3004 "please use another bond type or disable flood_vlans",
3008 miimon_interval = smap_get_int(&port->cfg->other_config,
3009 "bond-miimon-interval", 0);
3010 if (miimon_interval <= 0) {
3011 miimon_interval = 200;
3014 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3015 if (!detect_s || !strcmp(detect_s, "carrier")) {
3016 miimon_interval = 0;
3017 } else if (strcmp(detect_s, "miimon")) {
3018 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3019 "defaulting to carrier", port->name, detect_s);
3020 miimon_interval = 0;
3023 s->up_delay = MAX(0, port->cfg->bond_updelay);
3024 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3025 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3026 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3027 "bond-rebalance-interval", 10000);
3028 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3029 s->rebalance_interval = 1000;
3032 s->fake_iface = port->cfg->bond_fake_iface;
3035 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3036 long long stable_id;
3038 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3040 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3041 stable_id = iface->ofp_port;
3043 bond_stable_ids[i++] = stable_id;
3045 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3049 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3050 * instead of obtaining it from the database. */
3052 port_is_synthetic(const struct port *port)
3054 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3057 /* Interface functions. */
3059 /* Returns the correct network device type for interface 'iface' in bridge
3062 iface_get_type(const struct ovsrec_interface *iface,
3063 const struct ovsrec_bridge *br)
3065 /* The local port always has type "internal". Other ports take their type
3066 * from the database and default to "system" if none is specified. */
3067 return (!strcmp(iface->name, br->name) ? "internal"
3068 : iface->type[0] ? iface->type
3073 iface_destroy(struct iface *iface)
3076 struct port *port = iface->port;
3077 struct bridge *br = port->bridge;
3079 if (br->ofproto && iface->ofp_port >= 0) {
3080 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3083 if (iface->ofp_port >= 0) {
3084 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3087 list_remove(&iface->port_elem);
3088 hmap_remove(&br->iface_by_name, &iface->name_node);
3090 netdev_close(iface->netdev);
3097 static struct iface *
3098 iface_lookup(const struct bridge *br, const char *name)
3100 struct iface *iface;
3102 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3103 &br->iface_by_name) {
3104 if (!strcmp(iface->name, name)) {
3112 static struct iface *
3113 iface_find(const char *name)
3115 const struct bridge *br;
3117 HMAP_FOR_EACH (br, node, &all_bridges) {
3118 struct iface *iface = iface_lookup(br, name);
3127 static struct if_cfg *
3128 if_cfg_lookup(const struct bridge *br, const char *name)
3130 struct if_cfg *if_cfg;
3132 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3134 if (!strcmp(if_cfg->cfg->name, name)) {
3142 static struct iface *
3143 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3145 struct iface *iface;
3147 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3148 hash_int(ofp_port, 0), &br->ifaces) {
3149 if (iface->ofp_port == ofp_port) {
3156 /* Set Ethernet address of 'iface', if one is specified in the configuration
3159 iface_set_mac(struct iface *iface)
3161 uint8_t ea[ETH_ADDR_LEN];
3163 if (!strcmp(iface->type, "internal")
3164 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3165 if (iface->ofp_port == OFPP_LOCAL) {
3166 VLOG_ERR("interface %s: ignoring mac in Interface record "
3167 "(use Bridge record to set local port's mac)",
3169 } else if (eth_addr_is_multicast(ea)) {
3170 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3173 int error = netdev_set_etheraddr(iface->netdev, ea);
3175 VLOG_ERR("interface %s: setting MAC failed (%s)",
3176 iface->name, strerror(error));
3182 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3184 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3186 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3187 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3191 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3192 * sets the "ofport" field to -1.
3194 * This is appropriate when 'if_cfg''s interface cannot be created or is
3195 * otherwise invalid. */
3197 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3199 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3200 iface_set_ofport(if_cfg, -1);
3201 ovsrec_interface_set_status(if_cfg, NULL);
3202 ovsrec_interface_set_admin_state(if_cfg, NULL);
3203 ovsrec_interface_set_duplex(if_cfg, NULL);
3204 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3205 ovsrec_interface_set_link_state(if_cfg, NULL);
3206 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3207 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3208 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3209 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3210 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3211 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3215 struct iface_delete_queues_cbdata {
3216 struct netdev *netdev;
3217 const struct ovsdb_datum *queues;
3221 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3223 union ovsdb_atom atom;
3225 atom.integer = target;
3226 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3230 iface_delete_queues(unsigned int queue_id,
3231 const struct smap *details OVS_UNUSED, void *cbdata_)
3233 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3235 if (!queue_ids_include(cbdata->queues, queue_id)) {
3236 netdev_delete_queue(cbdata->netdev, queue_id);
3241 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3243 struct ofpbuf queues_buf;
3245 ofpbuf_init(&queues_buf, 0);
3247 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3248 netdev_set_qos(iface->netdev, NULL, NULL);
3250 struct iface_delete_queues_cbdata cbdata;
3254 /* Configure top-level Qos for 'iface'. */
3255 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3257 /* Deconfigure queues that were deleted. */
3258 cbdata.netdev = iface->netdev;
3259 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3261 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3263 /* Configure queues for 'iface'. */
3265 for (i = 0; i < qos->n_queues; i++) {
3266 const struct ovsrec_queue *queue = qos->value_queues[i];
3267 unsigned int queue_id = qos->key_queues[i];
3269 if (queue_id == 0) {
3273 if (queue->n_dscp == 1) {
3274 struct ofproto_port_queue *port_queue;
3276 port_queue = ofpbuf_put_uninit(&queues_buf,
3277 sizeof *port_queue);
3278 port_queue->queue = queue_id;
3279 port_queue->dscp = queue->dscp[0];
3282 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3285 struct smap details;
3287 smap_init(&details);
3288 netdev_set_queue(iface->netdev, 0, &details);
3289 smap_destroy(&details);
3293 if (iface->ofp_port >= 0) {
3294 const struct ofproto_port_queue *port_queues = queues_buf.data;
3295 size_t n_queues = queues_buf.size / sizeof *port_queues;
3297 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3298 port_queues, n_queues);
3301 netdev_set_policing(iface->netdev,
3302 iface->cfg->ingress_policing_rate,
3303 iface->cfg->ingress_policing_burst);
3305 ofpbuf_uninit(&queues_buf);
3309 iface_configure_cfm(struct iface *iface)
3311 const struct ovsrec_interface *cfg = iface->cfg;
3312 const char *opstate_str;
3313 const char *cfm_ccm_vlan;
3314 struct cfm_settings s;
3316 if (!cfg->n_cfm_mpid) {
3317 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3321 s.mpid = *cfg->cfm_mpid;
3322 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3323 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3324 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3326 if (s.interval <= 0) {
3330 if (!cfm_ccm_vlan) {
3332 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3333 s.ccm_vlan = CFM_RANDOM_VLAN;
3335 s.ccm_vlan = atoi(cfm_ccm_vlan);
3336 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3341 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3344 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3345 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3347 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3350 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3351 * instead of obtaining it from the database. */
3353 iface_is_synthetic(const struct iface *iface)
3355 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3359 /* Port mirroring. */
3361 static struct mirror *
3362 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3366 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3367 if (uuid_equals(uuid, &m->uuid)) {
3375 bridge_configure_mirrors(struct bridge *br)
3377 const struct ovsdb_datum *mc;
3378 unsigned long *flood_vlans;
3379 struct mirror *m, *next;
3382 /* Get rid of deleted mirrors. */
3383 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3384 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3385 union ovsdb_atom atom;
3387 atom.uuid = m->uuid;
3388 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3393 /* Add new mirrors and reconfigure existing ones. */
3394 for (i = 0; i < br->cfg->n_mirrors; i++) {
3395 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3396 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3398 m = mirror_create(br, cfg);
3401 if (!mirror_configure(m)) {
3406 /* Update flooded vlans (for RSPAN). */
3407 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3408 br->cfg->n_flood_vlans);
3409 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3410 bitmap_free(flood_vlans);
3413 static struct mirror *
3414 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3418 m = xzalloc(sizeof *m);
3419 m->uuid = cfg->header_.uuid;
3420 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3422 m->name = xstrdup(cfg->name);
3428 mirror_destroy(struct mirror *m)
3431 struct bridge *br = m->bridge;
3434 ofproto_mirror_unregister(br->ofproto, m);
3437 hmap_remove(&br->mirrors, &m->hmap_node);
3444 mirror_collect_ports(struct mirror *m,
3445 struct ovsrec_port **in_ports, int n_in_ports,
3446 void ***out_portsp, size_t *n_out_portsp)
3448 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3449 size_t n_out_ports = 0;
3452 for (i = 0; i < n_in_ports; i++) {
3453 const char *name = in_ports[i]->name;
3454 struct port *port = port_lookup(m->bridge, name);
3456 out_ports[n_out_ports++] = port;
3458 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3459 "port %s", m->bridge->name, m->name, name);
3462 *out_portsp = out_ports;
3463 *n_out_portsp = n_out_ports;
3467 mirror_configure(struct mirror *m)
3469 const struct ovsrec_mirror *cfg = m->cfg;
3470 struct ofproto_mirror_settings s;
3473 if (strcmp(cfg->name, m->name)) {
3475 m->name = xstrdup(cfg->name);
3479 /* Get output port or VLAN. */
3480 if (cfg->output_port) {
3481 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3482 if (!s.out_bundle) {
3483 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3484 m->bridge->name, m->name);
3487 s.out_vlan = UINT16_MAX;
3489 if (cfg->output_vlan) {
3490 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3491 "output vlan; ignoring output vlan",
3492 m->bridge->name, m->name);
3494 } else if (cfg->output_vlan) {
3495 /* The database should prevent invalid VLAN values. */
3496 s.out_bundle = NULL;
3497 s.out_vlan = *cfg->output_vlan;
3499 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3500 m->bridge->name, m->name);
3504 /* Get port selection. */
3505 if (cfg->select_all) {
3506 size_t n_ports = hmap_count(&m->bridge->ports);
3507 void **ports = xmalloc(n_ports * sizeof *ports);
3512 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3522 /* Get ports, dropping ports that don't exist.
3523 * The IDL ensures that there are no duplicates. */
3524 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3525 &s.srcs, &s.n_srcs);
3526 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3527 &s.dsts, &s.n_dsts);
3530 /* Get VLAN selection. */
3531 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3534 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3537 if (s.srcs != s.dsts) {
3546 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3548 * This is deprecated. It is only for compatibility with broken device drivers
3549 * in old versions of Linux that do not properly support VLANs when VLAN
3550 * devices are not used. When broken device drivers are no longer in
3551 * widespread use, we will delete these interfaces. */
3553 static struct ovsrec_port **recs;
3554 static size_t n_recs, allocated_recs;
3556 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3557 * splinters are reconfigured. */
3559 register_rec(struct ovsrec_port *rec)
3561 if (n_recs >= allocated_recs) {
3562 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3564 recs[n_recs++] = rec;
3567 /* Frees all of the ports registered with register_reg(). */
3569 free_registered_recs(void)
3573 for (i = 0; i < n_recs; i++) {
3574 struct ovsrec_port *port = recs[i];
3577 for (j = 0; j < port->n_interfaces; j++) {
3578 struct ovsrec_interface *iface = port->interfaces[j];
3583 smap_destroy(&port->other_config);
3584 free(port->interfaces);
3592 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3595 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3597 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3601 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3604 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3605 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3606 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3609 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3610 * use, returns NULL.
3612 * Updates 'vlan_splinters_enabled_anywhere'. */
3613 static unsigned long int *
3614 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3616 unsigned long int *splinter_vlans;
3617 struct sset splinter_ifaces;
3618 const char *real_dev_name;
3619 struct shash *real_devs;
3620 struct shash_node *node;
3624 /* Free space allocated for synthesized ports and interfaces, since we're
3625 * in the process of reconstructing all of them. */
3626 free_registered_recs();
3628 splinter_vlans = bitmap_allocate(4096);
3629 sset_init(&splinter_ifaces);
3630 vlan_splinters_enabled_anywhere = false;
3631 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3632 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3635 for (j = 0; j < br_cfg->n_ports; j++) {
3636 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3639 for (k = 0; k < port_cfg->n_interfaces; k++) {
3640 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3642 if (vlan_splinters_is_enabled(iface_cfg)) {
3643 vlan_splinters_enabled_anywhere = true;
3644 sset_add(&splinter_ifaces, iface_cfg->name);
3645 vlan_bitmap_from_array__(port_cfg->trunks,
3651 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3652 bitmap_set1(splinter_vlans, *port_cfg->tag);
3657 if (!vlan_splinters_enabled_anywhere) {
3658 free(splinter_vlans);
3659 sset_destroy(&splinter_ifaces);
3663 HMAP_FOR_EACH (br, node, &all_bridges) {
3665 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3669 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3670 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3671 * device to be created for it. */
3672 bitmap_set0(splinter_vlans, 0);
3673 bitmap_set0(splinter_vlans, 4095);
3675 /* Delete all VLAN devices that we don't need. */
3677 real_devs = vlandev_get_real_devs();
3678 SHASH_FOR_EACH (node, real_devs) {
3679 const struct vlan_real_dev *real_dev = node->data;
3680 const struct vlan_dev *vlan_dev;
3681 bool real_dev_has_splinters;
3683 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3685 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3686 if (!real_dev_has_splinters
3687 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3688 struct netdev *netdev;
3690 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3691 if (!netdev_get_in4(netdev, NULL, NULL) ||
3692 !netdev_get_in6(netdev, NULL)) {
3693 vlandev_del(vlan_dev->name);
3695 /* It has an IP address configured, so we don't own
3696 * it. Don't delete it. */
3698 netdev_close(netdev);
3705 /* Add all VLAN devices that we need. */
3706 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3709 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3710 if (!vlandev_get_name(real_dev_name, vid)) {
3711 vlandev_add(real_dev_name, vid);
3718 sset_destroy(&splinter_ifaces);
3720 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3721 free(splinter_vlans);
3724 return splinter_vlans;
3727 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3730 configure_splinter_port(struct port *port)
3732 struct ofproto *ofproto = port->bridge->ofproto;
3733 uint16_t realdev_ofp_port;
3734 const char *realdev_name;
3735 struct iface *vlandev, *realdev;
3737 ofproto_bundle_unregister(port->bridge->ofproto, port);
3739 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3742 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3743 realdev = iface_lookup(port->bridge, realdev_name);
3744 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3746 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3750 static struct ovsrec_port *
3751 synthesize_splinter_port(const char *real_dev_name,
3752 const char *vlan_dev_name, int vid)
3754 struct ovsrec_interface *iface;
3755 struct ovsrec_port *port;
3757 iface = xmalloc(sizeof *iface);
3758 ovsrec_interface_init(iface);
3759 iface->name = xstrdup(vlan_dev_name);
3760 iface->type = "system";
3762 port = xmalloc(sizeof *port);
3763 ovsrec_port_init(port);
3764 port->interfaces = xmemdup(&iface, sizeof iface);
3765 port->n_interfaces = 1;
3766 port->name = xstrdup(vlan_dev_name);
3767 port->vlan_mode = "splinter";
3768 port->tag = xmalloc(sizeof *port->tag);
3771 smap_add(&port->other_config, "realdev", real_dev_name);
3777 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3778 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3779 * 1-bit in the 'splinter_vlans' bitmap. */
3781 add_vlan_splinter_ports(struct bridge *br,
3782 const unsigned long int *splinter_vlans,
3783 struct shash *ports)
3787 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3788 * we're modifying 'ports'. */
3789 for (i = 0; i < br->cfg->n_ports; i++) {
3790 const char *name = br->cfg->ports[i]->name;
3791 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3794 for (j = 0; j < port_cfg->n_interfaces; j++) {
3795 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3797 if (vlan_splinters_is_enabled(iface_cfg)) {
3798 const char *real_dev_name;
3801 real_dev_name = iface_cfg->name;
3802 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3803 const char *vlan_dev_name;
3805 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3807 && !shash_find(ports, vlan_dev_name)) {
3808 shash_add(ports, vlan_dev_name,
3809 synthesize_splinter_port(
3810 real_dev_name, vlan_dev_name, vid));
3819 mirror_refresh_stats(struct mirror *m)
3821 struct ofproto *ofproto = m->bridge->ofproto;
3822 uint64_t tx_packets, tx_bytes;
3825 size_t stat_cnt = 0;
3827 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3828 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3832 if (tx_packets != UINT64_MAX) {
3833 keys[stat_cnt] = "tx_packets";
3834 values[stat_cnt] = tx_packets;
3837 if (tx_bytes != UINT64_MAX) {
3838 keys[stat_cnt] = "tx_bytes";
3839 values[stat_cnt] = tx_bytes;
3843 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);