1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 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.
27 #include "dynamic-string.h"
34 #include "mac-learning.h"
35 #include "meta-flow.h"
37 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
69 int64_t ofport; /* Requested OpenFlow port number. */
72 /* OpenFlow port slated for removal from ofproto. */
74 struct list list_node; /* Node in bridge's ofpp_garbage. */
75 uint16_t ofp_port; /* Port to be deleted. */
79 /* These members are always valid. */
80 struct list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct port *port; /* Containing port. */
83 char *name; /* Host network device name. */
85 /* These members are valid only after bridge_reconfigure() causes them to
87 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
88 int ofp_port; /* OpenFlow port number, -1 if unknown. */
89 struct netdev *netdev; /* Network device. */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
131 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
132 Indexed on 'cfg->name'. */
134 /* Port mirroring. */
135 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
137 /* Synthetic local port if necessary. */
138 struct ovsrec_port synth_local_port;
139 struct ovsrec_interface synth_local_iface;
140 struct ovsrec_interface *synth_local_ifacep;
143 /* All bridges, indexed by name. */
144 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
146 /* OVSDB IDL used to obtain configuration. */
147 static struct ovsdb_idl *idl;
149 /* Most recently processed IDL sequence number. */
150 static unsigned int idl_seqno;
152 /* Each time this timer expires, the bridge fetches interface and mirror
153 * statistics and pushes them into the database. */
154 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
155 static long long int iface_stats_timer = LLONG_MIN;
157 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
158 * expensive. This can cause bridge_reconfigure() to take a long time during
159 * which no other work can be done. To deal with this problem, we limit port
160 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
161 * call to bridge_reconfigure(). If there is more work to do after the limit
162 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
163 * This allows the rest of the code to catch up on important things like
164 * forwarding packets. */
165 #define OFP_PORT_ACTION_WINDOW 10
166 static bool reconfiguring = false;
168 static void add_del_bridges(const struct ovsrec_open_vswitch *);
169 static void bridge_update_ofprotos(void);
170 static void bridge_create(const struct ovsrec_bridge *);
171 static void bridge_destroy(struct bridge *);
172 static struct bridge *bridge_lookup(const char *name);
173 static unixctl_cb_func bridge_unixctl_dump_flows;
174 static unixctl_cb_func bridge_unixctl_reconnect;
175 static size_t bridge_get_controllers(const struct bridge *br,
176 struct ovsrec_controller ***controllersp);
177 static void bridge_add_del_ports(struct bridge *,
178 const unsigned long int *splinter_vlans);
179 static void bridge_refresh_ofp_port(struct bridge *);
180 static void bridge_configure_datapath_id(struct bridge *);
181 static void bridge_configure_flow_eviction_threshold(struct bridge *);
182 static void bridge_configure_netflow(struct bridge *);
183 static void bridge_configure_forward_bpdu(struct bridge *);
184 static void bridge_configure_mac_table(struct bridge *);
185 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
186 static void bridge_configure_stp(struct bridge *);
187 static void bridge_configure_tables(struct bridge *);
188 static void bridge_configure_dp_desc(struct bridge *);
189 static void bridge_configure_remotes(struct bridge *,
190 const struct sockaddr_in *managers,
192 static void bridge_pick_local_hw_addr(struct bridge *,
193 uint8_t ea[ETH_ADDR_LEN],
194 struct iface **hw_addr_iface);
195 static uint64_t bridge_pick_datapath_id(struct bridge *,
196 const uint8_t bridge_ea[ETH_ADDR_LEN],
197 struct iface *hw_addr_iface);
198 static void bridge_queue_if_cfg(struct bridge *,
199 const struct ovsrec_interface *,
200 const struct ovsrec_port *);
201 static uint64_t dpid_from_hash(const void *, size_t nbytes);
202 static bool bridge_has_bond_fake_iface(const struct bridge *,
204 static bool port_is_bond_fake_iface(const struct port *);
206 static unixctl_cb_func qos_unixctl_show;
208 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
209 static void port_del_ifaces(struct port *);
210 static void port_destroy(struct port *);
211 static struct port *port_lookup(const struct bridge *, const char *name);
212 static void port_configure(struct port *);
213 static struct lacp_settings *port_configure_lacp(struct port *,
214 struct lacp_settings *);
215 static void port_configure_bond(struct port *, struct bond_settings *,
216 uint32_t *bond_stable_ids);
217 static bool port_is_synthetic(const struct port *);
219 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
220 static void run_system_stats(void);
222 static void bridge_configure_mirrors(struct bridge *);
223 static struct mirror *mirror_create(struct bridge *,
224 const struct ovsrec_mirror *);
225 static void mirror_destroy(struct mirror *);
226 static bool mirror_configure(struct mirror *);
227 static void mirror_refresh_stats(struct mirror *);
229 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
230 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
231 static bool iface_is_internal(const struct ovsrec_interface *iface,
232 const struct ovsrec_bridge *br);
233 static const char *iface_get_type(const struct ovsrec_interface *,
234 const struct ovsrec_bridge *);
235 static void iface_destroy(struct iface *);
236 static struct iface *iface_lookup(const struct bridge *, const char *name);
237 static struct iface *iface_find(const char *name);
238 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
239 static struct iface *iface_from_ofp_port(const struct bridge *,
241 static void iface_set_mac(struct iface *);
242 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
243 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
244 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
245 static void iface_configure_cfm(struct iface *);
246 static void iface_refresh_cfm_stats(struct iface *);
247 static void iface_refresh_stats(struct iface *);
248 static void iface_refresh_status(struct iface *);
249 static bool iface_is_synthetic(const struct iface *);
250 static int64_t iface_pick_ofport(const struct ovsrec_interface *);
252 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
254 * This is deprecated. It is only for compatibility with broken device drivers
255 * in old versions of Linux that do not properly support VLANs when VLAN
256 * devices are not used. When broken device drivers are no longer in
257 * widespread use, we will delete these interfaces. */
259 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
260 static bool vlan_splinters_enabled_anywhere;
262 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
263 static unsigned long int *collect_splinter_vlans(
264 const struct ovsrec_open_vswitch *);
265 static void configure_splinter_port(struct port *);
266 static void add_vlan_splinter_ports(struct bridge *,
267 const unsigned long int *splinter_vlans,
268 struct shash *ports);
271 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
273 struct shash iface_hints;
274 static bool initialized = false;
281 shash_init(&iface_hints);
284 for (i = 0; i < cfg->n_bridges; i++) {
285 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
288 for (j = 0; j < br_cfg->n_ports; j++) {
289 struct ovsrec_port *port_cfg = br_cfg->ports[j];
292 for (k = 0; k < port_cfg->n_interfaces; k++) {
293 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
294 struct iface_hint *iface_hint;
296 iface_hint = xmalloc(sizeof *iface_hint);
297 iface_hint->br_name = br_cfg->name;
298 iface_hint->br_type = br_cfg->datapath_type;
299 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
301 shash_add(&iface_hints, if_cfg->name, iface_hint);
307 ofproto_init(&iface_hints);
309 shash_destroy_free_data(&iface_hints);
313 /* Public functions. */
315 /* Initializes the bridge module, configuring it to obtain its configuration
316 * from an OVSDB server accessed over 'remote', which should be a string in a
317 * form acceptable to ovsdb_idl_create(). */
319 bridge_init(const char *remote)
321 /* Create connection to database. */
322 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
323 idl_seqno = ovsdb_idl_get_seqno(idl);
324 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
325 ovsdb_idl_verify_write_only(idl);
327 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
328 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
329 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
331 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
332 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
333 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
335 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
336 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
337 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
339 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
340 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
341 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
342 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
344 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
345 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
346 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
347 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
348 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
349 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
350 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
351 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
352 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
353 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
354 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
355 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
356 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
357 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
358 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
359 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
360 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
362 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
363 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
364 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
365 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
367 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
369 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
371 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
372 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
374 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
376 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
378 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
379 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
380 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
381 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
382 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
384 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
386 /* Register unixctl commands. */
387 unixctl_command_register("qos/show", "interface", 1, 1,
388 qos_unixctl_show, NULL);
389 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
390 bridge_unixctl_dump_flows, NULL);
391 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
392 bridge_unixctl_reconnect, NULL);
402 struct bridge *br, *next_br;
404 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
407 ovsdb_idl_destroy(idl);
410 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
411 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
412 * responsible for freeing '*managersp' (with free()).
414 * You may be asking yourself "why does ovs-vswitchd care?", because
415 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
416 * should not be and in fact is not directly involved in that. But
417 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
418 * it has to tell in-band control where the managers are to enable that.
419 * (Thus, only managers connected in-band are collected.)
422 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
423 struct sockaddr_in **managersp, size_t *n_managersp)
425 struct sockaddr_in *managers = NULL;
426 size_t n_managers = 0;
430 /* Collect all of the potential targets from the "targets" columns of the
431 * rows pointed to by "manager_options", excluding any that are
434 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
435 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
437 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
438 sset_find_and_delete(&targets, m->target);
440 sset_add(&targets, m->target);
444 /* Now extract the targets' IP addresses. */
445 if (!sset_is_empty(&targets)) {
448 managers = xmalloc(sset_count(&targets) * sizeof *managers);
449 SSET_FOR_EACH (target, &targets) {
450 struct sockaddr_in *sin = &managers[n_managers];
452 if (stream_parse_target_with_default_ports(target,
460 sset_destroy(&targets);
462 *managersp = managers;
463 *n_managersp = n_managers;
467 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
469 unsigned long int *splinter_vlans;
472 COVERAGE_INC(bridge_reconfigure);
474 ovs_assert(!reconfiguring);
475 reconfiguring = true;
477 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
478 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
479 * configuration otherwise.
481 * This is mostly an update to bridge data structures. Nothing is pushed
482 * down to ofproto or lower layers. */
483 add_del_bridges(ovs_cfg);
484 splinter_vlans = collect_splinter_vlans(ovs_cfg);
485 HMAP_FOR_EACH (br, node, &all_bridges) {
486 bridge_add_del_ports(br, splinter_vlans);
488 free(splinter_vlans);
490 /* Delete datapaths that are no longer configured, and create ones which
491 * don't exist but should. */
492 bridge_update_ofprotos();
494 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
495 HMAP_FOR_EACH (br, node, &all_bridges) {
496 bridge_refresh_ofp_port(br);
499 /* Clear database records for "if_cfg"s which haven't been instantiated. */
500 HMAP_FOR_EACH (br, node, &all_bridges) {
501 struct if_cfg *if_cfg;
503 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
504 iface_clear_db_record(if_cfg->cfg);
508 reconfigure_system_stats(ovs_cfg);
512 bridge_reconfigure_ofp(void)
514 long long int deadline;
518 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
520 /* The kernel will reject any attempt to add a given port to a datapath if
521 * that port already belongs to a different datapath, so we must do all
522 * port deletions before any port additions. */
523 HMAP_FOR_EACH (br, node, &all_bridges) {
524 struct ofpp_garbage *garbage, *next;
526 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
527 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
528 * internal datastructures may not be consistent. Eventually, when
529 * port additions and deletions are cheaper, these calls should be
532 ofproto_port_del(br->ofproto, garbage->ofp_port);
533 list_remove(&garbage->list_node);
537 if (time_msec() >= deadline) {
544 HMAP_FOR_EACH (br, node, &all_bridges) {
545 struct if_cfg *if_cfg, *next;
547 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
548 iface_create(br, if_cfg, -1);
550 if (time_msec() >= deadline) {
560 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
562 struct sockaddr_in *managers;
563 int sflow_bridge_number;
568 ovs_assert(reconfiguring);
569 done = bridge_reconfigure_ofp();
571 /* Complete the configuration. */
572 sflow_bridge_number = 0;
573 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
574 HMAP_FOR_EACH (br, node, &all_bridges) {
577 /* We need the datapath ID early to allow LACP ports to use it as the
578 * default system ID. */
579 bridge_configure_datapath_id(br);
581 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
584 port_configure(port);
586 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
587 iface_configure_cfm(iface);
588 iface_configure_qos(iface, port->cfg->qos);
589 iface_set_mac(iface);
592 bridge_configure_mirrors(br);
593 bridge_configure_flow_eviction_threshold(br);
594 bridge_configure_forward_bpdu(br);
595 bridge_configure_mac_table(br);
596 bridge_configure_remotes(br, managers, n_managers);
597 bridge_configure_netflow(br);
598 bridge_configure_sflow(br, &sflow_bridge_number);
599 bridge_configure_stp(br);
600 bridge_configure_tables(br);
601 bridge_configure_dp_desc(br);
606 /* ovs-vswitchd has completed initialization, so allow the process that
607 * forked us to exit successfully. */
608 daemonize_complete();
609 reconfiguring = false;
611 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
617 /* Delete ofprotos which aren't configured or have the wrong type. Create
618 * ofprotos which don't exist but need to. */
620 bridge_update_ofprotos(void)
622 struct bridge *br, *next;
627 /* Delete ofprotos with no bridge or with the wrong type. */
630 ofproto_enumerate_types(&types);
631 SSET_FOR_EACH (type, &types) {
634 ofproto_enumerate_names(type, &names);
635 SSET_FOR_EACH (name, &names) {
636 br = bridge_lookup(name);
637 if (!br || strcmp(type, br->type)) {
638 ofproto_delete(name, type);
642 sset_destroy(&names);
643 sset_destroy(&types);
645 /* Add ofprotos for bridges which don't have one yet. */
646 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
654 /* Remove ports from any datapath with the same name as 'br'. If we
655 * don't do this, creating 'br''s ofproto will fail because a port with
656 * the same name as its local port already exists. */
657 HMAP_FOR_EACH (br2, node, &all_bridges) {
658 struct ofproto_port ofproto_port;
664 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
666 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
668 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
671 ofproto_port_destroy(&ofproto_port);
675 error = ofproto_create(br->name, br->type, &br->ofproto);
677 VLOG_ERR("failed to create bridge %s: %s", br->name,
685 port_configure(struct port *port)
687 const struct ovsrec_port *cfg = port->cfg;
688 struct bond_settings bond_settings;
689 struct lacp_settings lacp_settings;
690 struct ofproto_bundle_settings s;
693 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
694 configure_splinter_port(port);
703 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
704 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
705 s.slaves[s.n_slaves++] = iface->ofp_port;
710 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
714 /* Get VLAN trunks. */
717 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
721 if (cfg->vlan_mode) {
722 if (!strcmp(cfg->vlan_mode, "access")) {
723 s.vlan_mode = PORT_VLAN_ACCESS;
724 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
725 s.vlan_mode = PORT_VLAN_TRUNK;
726 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
727 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
728 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
729 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
731 /* This "can't happen" because ovsdb-server should prevent it. */
732 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
733 s.vlan_mode = PORT_VLAN_TRUNK;
737 s.vlan_mode = PORT_VLAN_ACCESS;
739 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
743 s.vlan_mode = PORT_VLAN_TRUNK;
746 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
749 /* Get LACP settings. */
750 s.lacp = port_configure_lacp(port, &lacp_settings);
754 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
755 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
756 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
759 s.lacp_slaves = NULL;
762 /* Get bond settings. */
763 if (s.n_slaves > 1) {
764 s.bond = &bond_settings;
765 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
766 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
769 s.bond_stable_ids = NULL;
771 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
772 netdev_set_miimon_interval(iface->netdev, 0);
777 ofproto_bundle_register(port->bridge->ofproto, port, &s);
783 free(s.bond_stable_ids);
786 /* Pick local port hardware address and datapath ID for 'br'. */
788 bridge_configure_datapath_id(struct bridge *br)
790 uint8_t ea[ETH_ADDR_LEN];
792 struct iface *local_iface;
793 struct iface *hw_addr_iface;
796 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
797 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
799 int error = netdev_set_etheraddr(local_iface->netdev, ea);
801 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
802 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
803 "Ethernet address: %s",
804 br->name, strerror(error));
807 memcpy(br->ea, ea, ETH_ADDR_LEN);
809 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
810 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
811 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
812 ofproto_set_datapath_id(br->ofproto, dpid);
815 dpid_string = xasprintf("%016"PRIx64, dpid);
816 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
820 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
823 bridge_get_allowed_versions(struct bridge *br)
825 if (!br->cfg->n_protocols)
828 return ofputil_versions_from_strings(br->cfg->protocols,
829 br->cfg->n_protocols);
832 /* Set NetFlow configuration on 'br'. */
834 bridge_configure_netflow(struct bridge *br)
836 struct ovsrec_netflow *cfg = br->cfg->netflow;
837 struct netflow_options opts;
840 ofproto_set_netflow(br->ofproto, NULL);
844 memset(&opts, 0, sizeof opts);
846 /* Get default NetFlow configuration from datapath.
847 * Apply overrides from 'cfg'. */
848 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
849 if (cfg->engine_type) {
850 opts.engine_type = *cfg->engine_type;
852 if (cfg->engine_id) {
853 opts.engine_id = *cfg->engine_id;
856 /* Configure active timeout interval. */
857 opts.active_timeout = cfg->active_timeout;
858 if (!opts.active_timeout) {
859 opts.active_timeout = -1;
860 } else if (opts.active_timeout < 0) {
861 VLOG_WARN("bridge %s: active timeout interval set to negative "
862 "value, using default instead (%d seconds)", br->name,
863 NF_ACTIVE_TIMEOUT_DEFAULT);
864 opts.active_timeout = -1;
867 /* Add engine ID to interface number to disambiguate bridgs? */
868 opts.add_id_to_iface = cfg->add_id_to_interface;
869 if (opts.add_id_to_iface) {
870 if (opts.engine_id > 0x7f) {
871 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
872 "another vswitch, choose an engine id less than 128",
875 if (hmap_count(&br->ports) > 508) {
876 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
877 "another port when more than 508 ports are used",
883 sset_init(&opts.collectors);
884 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
887 if (ofproto_set_netflow(br->ofproto, &opts)) {
888 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
890 sset_destroy(&opts.collectors);
893 /* Set sFlow configuration on 'br'. */
895 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
897 const struct ovsrec_sflow *cfg = br->cfg->sflow;
898 struct ovsrec_controller **controllers;
899 struct ofproto_sflow_options oso;
900 size_t n_controllers;
904 ofproto_set_sflow(br->ofproto, NULL);
908 memset(&oso, 0, sizeof oso);
910 sset_init(&oso.targets);
911 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
913 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
915 oso.sampling_rate = *cfg->sampling;
918 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
920 oso.polling_interval = *cfg->polling;
923 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
925 oso.header_len = *cfg->header;
928 oso.sub_id = (*sflow_bridge_number)++;
929 oso.agent_device = cfg->agent;
931 oso.control_ip = NULL;
932 n_controllers = bridge_get_controllers(br, &controllers);
933 for (i = 0; i < n_controllers; i++) {
934 if (controllers[i]->local_ip) {
935 oso.control_ip = controllers[i]->local_ip;
939 ofproto_set_sflow(br->ofproto, &oso);
941 sset_destroy(&oso.targets);
945 port_configure_stp(const struct ofproto *ofproto, struct port *port,
946 struct ofproto_port_stp_settings *port_s,
947 int *port_num_counter, unsigned long *port_num_bitmap)
949 const char *config_str;
952 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
953 port_s->enable = false;
956 port_s->enable = true;
959 /* STP over bonds is not supported. */
960 if (!list_is_singleton(&port->ifaces)) {
961 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
963 port_s->enable = false;
967 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
969 /* Internal ports shouldn't participate in spanning tree, so
971 if (!strcmp(iface->type, "internal")) {
972 VLOG_DBG("port %s: disable STP on internal ports", port->name);
973 port_s->enable = false;
977 /* STP on mirror output ports is not supported. */
978 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
979 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
980 port_s->enable = false;
984 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
986 unsigned long int port_num = strtoul(config_str, NULL, 0);
987 int port_idx = port_num - 1;
989 if (port_num < 1 || port_num > STP_MAX_PORTS) {
990 VLOG_ERR("port %s: invalid stp-port-num", port->name);
991 port_s->enable = false;
995 if (bitmap_is_set(port_num_bitmap, port_idx)) {
996 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
997 port->name, port_num);
998 port_s->enable = false;
1001 bitmap_set1(port_num_bitmap, port_idx);
1002 port_s->port_num = port_idx;
1004 if (*port_num_counter >= STP_MAX_PORTS) {
1005 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1006 port_s->enable = false;
1010 port_s->port_num = (*port_num_counter)++;
1013 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1015 port_s->path_cost = strtoul(config_str, NULL, 10);
1017 enum netdev_features current;
1020 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1021 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1022 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1025 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1027 port_s->priority = strtoul(config_str, NULL, 0);
1029 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1033 /* Set spanning tree configuration on 'br'. */
1035 bridge_configure_stp(struct bridge *br)
1037 if (!br->cfg->stp_enable) {
1038 ofproto_set_stp(br->ofproto, NULL);
1040 struct ofproto_stp_settings br_s;
1041 const char *config_str;
1043 int port_num_counter;
1044 unsigned long *port_num_bitmap;
1046 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1048 uint8_t ea[ETH_ADDR_LEN];
1050 if (eth_addr_from_string(config_str, ea)) {
1051 br_s.system_id = eth_addr_to_uint64(ea);
1053 br_s.system_id = eth_addr_to_uint64(br->ea);
1054 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1055 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1058 br_s.system_id = eth_addr_to_uint64(br->ea);
1061 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1063 br_s.priority = strtoul(config_str, NULL, 0);
1065 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1068 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1070 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1072 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1075 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1077 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1079 br_s.max_age = STP_DEFAULT_MAX_AGE;
1082 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1084 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1086 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1089 /* Configure STP on the bridge. */
1090 if (ofproto_set_stp(br->ofproto, &br_s)) {
1091 VLOG_ERR("bridge %s: could not enable STP", br->name);
1095 /* Users must either set the port number with the "stp-port-num"
1096 * configuration on all ports or none. If manual configuration
1097 * is not done, then we allocate them sequentially. */
1098 port_num_counter = 0;
1099 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1100 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1101 struct ofproto_port_stp_settings port_s;
1102 struct iface *iface;
1104 port_configure_stp(br->ofproto, port, &port_s,
1105 &port_num_counter, port_num_bitmap);
1107 /* As bonds are not supported, just apply configuration to
1108 * all interfaces. */
1109 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1110 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1112 VLOG_ERR("port %s: could not enable STP", port->name);
1118 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1119 && port_num_counter) {
1120 VLOG_ERR("bridge %s: must manually configure all STP port "
1121 "IDs or none, disabling", br->name);
1122 ofproto_set_stp(br->ofproto, NULL);
1124 bitmap_free(port_num_bitmap);
1129 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1131 const struct port *port = port_lookup(br, name);
1132 return port && port_is_bond_fake_iface(port);
1136 port_is_bond_fake_iface(const struct port *port)
1138 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1142 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1144 struct bridge *br, *next;
1145 struct shash new_br;
1148 /* Collect new bridges' names and types. */
1149 shash_init(&new_br);
1150 for (i = 0; i < cfg->n_bridges; i++) {
1151 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1152 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1154 if (strchr(br_cfg->name, '/')) {
1155 /* Prevent remote ovsdb-server users from accessing arbitrary
1156 * directories, e.g. consider a bridge named "../../../etc/". */
1157 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1159 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1160 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1164 /* Get rid of deleted bridges or those whose types have changed.
1165 * Update 'cfg' of bridges that still exist. */
1166 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1167 br->cfg = shash_find_data(&new_br, br->name);
1168 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1169 br->cfg->datapath_type))) {
1174 /* Add new bridges. */
1175 for (i = 0; i < cfg->n_bridges; i++) {
1176 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1177 struct bridge *br = bridge_lookup(br_cfg->name);
1179 bridge_create(br_cfg);
1183 shash_destroy(&new_br);
1187 iface_set_ofp_port(struct iface *iface, int ofp_port)
1189 struct bridge *br = iface->port->bridge;
1191 ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
1192 iface->ofp_port = ofp_port;
1193 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1194 iface_set_ofport(iface->cfg, ofp_port);
1197 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1198 * Returns 0 if successful, otherwise a positive errno value. */
1200 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1201 struct netdev *netdev)
1205 error = netdev_set_config(netdev, &iface_cfg->options);
1207 VLOG_WARN("could not configure network device %s (%s)",
1208 iface_cfg->name, strerror(error));
1213 /* This function determines whether 'ofproto_port', which is attached to
1214 * br->ofproto's datapath, is one that we want in 'br'.
1216 * If it is, it returns true, after creating an iface (if necessary),
1217 * configuring the iface's netdev according to the iface's options, and setting
1218 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1220 * If, on the other hand, 'port' should be removed, it returns false. The
1221 * caller should later detach the port from br->ofproto. */
1223 bridge_refresh_one_ofp_port(struct bridge *br,
1224 const struct ofproto_port *ofproto_port)
1226 const char *name = ofproto_port->name;
1227 const char *type = ofproto_port->type;
1228 uint16_t ofp_port = ofproto_port->ofp_port;
1230 struct iface *iface = iface_lookup(br, name);
1232 /* Check that the name-to-number mapping is one-to-one. */
1233 if (iface->ofp_port >= 0) {
1234 VLOG_WARN("bridge %s: interface %s reported twice",
1237 } else if (iface_from_ofp_port(br, ofp_port)) {
1238 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1239 br->name, ofp_port);
1243 /* There's a configured interface named 'name'. */
1244 if (strcmp(type, iface->type)
1245 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1246 /* It's the wrong type, or it's the right type but can't be
1247 * configured as the user requested, so we must destroy it. */
1250 /* It's the right type and configured correctly. Keep it. */
1251 iface_set_ofp_port(iface, ofp_port);
1254 } else if (bridge_has_bond_fake_iface(br, name)
1255 && !strcmp(type, "internal")) {
1256 /* It's a bond fake iface. Keep it. */
1259 /* There's no configured interface named 'name', but there might be an
1260 * interface of that name queued to be created.
1262 * If there is, and it has the correct type, then try to configure it
1263 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1264 * delete it and later try to re-add it. */
1265 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1267 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1268 && iface_create(br, if_cfg, ofp_port));
1272 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1273 * consistent with the ofp_ports in "br->ofproto". */
1275 bridge_refresh_ofp_port(struct bridge *br)
1277 struct ofproto_port_dump dump;
1278 struct ofproto_port ofproto_port;
1279 struct port *port, *port_next;
1281 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1282 hmap_clear(&br->ifaces);
1283 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1284 struct iface *iface;
1286 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1287 iface->ofp_port = -1;
1291 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1292 * already exist in the datapath and promote them to full fledged "struct
1293 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1294 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1295 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1296 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1297 garbage->ofp_port = ofproto_port.ofp_port;
1298 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1302 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1303 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1304 * later they can be added to ofproto. */
1305 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1306 struct iface *iface, *iface_next;
1308 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1309 if (iface->ofp_port < 0) {
1310 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1311 iface_destroy(iface);
1315 if (list_is_empty(&port->ifaces)) {
1321 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1322 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1323 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1326 * If successful, returns 0 and stores the network device in '*netdevp'. On
1327 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1329 iface_do_create(const struct bridge *br,
1330 const struct if_cfg *if_cfg,
1331 int *ofp_portp, struct netdev **netdevp)
1333 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1334 const struct ovsrec_port *port_cfg = if_cfg->parent;
1335 struct netdev *netdev;
1338 error = netdev_open(iface_cfg->name,
1339 iface_get_type(iface_cfg, br->cfg), &netdev);
1341 VLOG_WARN("could not open network device %s (%s)",
1342 iface_cfg->name, strerror(error));
1346 error = iface_set_netdev_config(iface_cfg, netdev);
1351 if (*ofp_portp < 0) {
1352 uint16_t ofp_port = if_cfg->ofport;
1354 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1358 *ofp_portp = ofp_port;
1360 VLOG_INFO("bridge %s: added interface %s on port %d",
1361 br->name, iface_cfg->name, *ofp_portp);
1363 VLOG_DBG("bridge %s: interface %s is on port %d",
1364 br->name, iface_cfg->name, *ofp_portp);
1367 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1368 || iface_is_internal(iface_cfg, br->cfg)) {
1369 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1377 netdev_close(netdev);
1381 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1382 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1383 * automatically allocated for the iface. Takes ownership of and
1384 * deallocates 'if_cfg'.
1386 * Return true if an iface is successfully created, false otherwise. */
1388 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1390 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1391 const struct ovsrec_port *port_cfg = if_cfg->parent;
1393 struct netdev *netdev;
1394 struct iface *iface;
1399 /* Do the bits that can fail up front.
1401 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1402 * internal datastructures may not be consistent. Eventually, when port
1403 * additions and deletions are cheaper, these calls should be removed. */
1405 ovs_assert(!iface_lookup(br, iface_cfg->name));
1406 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1409 iface_set_ofport(iface_cfg, -1);
1410 iface_clear_db_record(iface_cfg);
1415 /* Get or create the port structure. */
1416 port = port_lookup(br, port_cfg->name);
1418 port = port_create(br, port_cfg);
1421 /* Create the iface structure. */
1422 iface = xzalloc(sizeof *iface);
1423 list_push_back(&port->ifaces, &iface->port_elem);
1424 hmap_insert(&br->iface_by_name, &iface->name_node,
1425 hash_string(iface_cfg->name, 0));
1427 iface->name = xstrdup(iface_cfg->name);
1428 iface->ofp_port = -1;
1429 iface->netdev = netdev;
1430 iface->type = iface_get_type(iface_cfg, br->cfg);
1431 iface->cfg = iface_cfg;
1433 iface_set_ofp_port(iface, ofp_port);
1435 /* Populate initial status in database. */
1436 iface_refresh_stats(iface);
1437 iface_refresh_status(iface);
1439 /* Add bond fake iface if necessary. */
1440 if (port_is_bond_fake_iface(port)) {
1441 struct ofproto_port ofproto_port;
1443 if (ofproto_port_query_by_name(br->ofproto, port->name,
1445 struct netdev *netdev;
1448 error = netdev_open(port->name, "internal", &netdev);
1450 uint16_t fake_ofp_port = if_cfg->ofport;
1452 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1453 netdev_close(netdev);
1455 VLOG_WARN("could not open network device %s (%s)",
1456 port->name, strerror(error));
1459 /* Already exists, nothing to do. */
1460 ofproto_port_destroy(&ofproto_port);
1465 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1471 /* Set Flow eviction threshold */
1473 bridge_configure_flow_eviction_threshold(struct bridge *br)
1475 const char *threshold_str;
1478 threshold_str = smap_get(&br->cfg->other_config,
1479 "flow-eviction-threshold");
1480 if (threshold_str) {
1481 threshold = strtoul(threshold_str, NULL, 10);
1483 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1485 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1488 /* Set forward BPDU option. */
1490 bridge_configure_forward_bpdu(struct bridge *br)
1492 ofproto_set_forward_bpdu(br->ofproto,
1493 smap_get_bool(&br->cfg->other_config,
1498 /* Set MAC learning table configuration for 'br'. */
1500 bridge_configure_mac_table(struct bridge *br)
1502 const char *idle_time_str;
1505 const char *mac_table_size_str;
1508 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1509 idle_time = (idle_time_str && atoi(idle_time_str)
1510 ? atoi(idle_time_str)
1511 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1513 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1514 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1515 ? atoi(mac_table_size_str)
1518 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1522 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1523 struct iface **hw_addr_iface)
1525 struct hmapx mirror_output_ports;
1528 bool found_addr = false;
1532 *hw_addr_iface = NULL;
1534 /* Did the user request a particular MAC? */
1535 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1536 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1537 if (eth_addr_is_multicast(ea)) {
1538 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1539 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1540 } else if (eth_addr_is_zero(ea)) {
1541 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1547 /* Mirror output ports don't participate in picking the local hardware
1548 * address. ofproto can't help us find out whether a given port is a
1549 * mirror output because we haven't configured mirrors yet, so we need to
1550 * accumulate them ourselves. */
1551 hmapx_init(&mirror_output_ports);
1552 for (i = 0; i < br->cfg->n_mirrors; i++) {
1553 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1554 if (m->output_port) {
1555 hmapx_add(&mirror_output_ports, m->output_port);
1559 /* Otherwise choose the minimum non-local MAC address among all of the
1561 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1562 uint8_t iface_ea[ETH_ADDR_LEN];
1563 struct iface *candidate;
1564 struct iface *iface;
1566 /* Mirror output ports don't participate. */
1567 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1571 /* Choose the MAC address to represent the port. */
1573 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1574 /* Find the interface with this Ethernet address (if any) so that
1575 * we can provide the correct devname to the caller. */
1576 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1577 uint8_t candidate_ea[ETH_ADDR_LEN];
1578 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1579 && eth_addr_equals(iface_ea, candidate_ea)) {
1584 /* Choose the interface whose MAC address will represent the port.
1585 * The Linux kernel bonding code always chooses the MAC address of
1586 * the first slave added to a bond, and the Fedora networking
1587 * scripts always add slaves to a bond in alphabetical order, so
1588 * for compatibility we choose the interface with the name that is
1589 * first in alphabetical order. */
1590 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1591 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1596 /* The local port doesn't count (since we're trying to choose its
1597 * MAC address anyway). */
1598 if (iface->ofp_port == OFPP_LOCAL) {
1603 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1609 /* Compare against our current choice. */
1610 if (!eth_addr_is_multicast(iface_ea) &&
1611 !eth_addr_is_local(iface_ea) &&
1612 !eth_addr_is_reserved(iface_ea) &&
1613 !eth_addr_is_zero(iface_ea) &&
1614 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1616 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1617 *hw_addr_iface = iface;
1623 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1624 *hw_addr_iface = NULL;
1627 hmapx_destroy(&mirror_output_ports);
1630 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1631 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1632 * an interface on 'br', then that interface must be passed in as
1633 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1634 * 'hw_addr_iface' must be passed in as a null pointer. */
1636 bridge_pick_datapath_id(struct bridge *br,
1637 const uint8_t bridge_ea[ETH_ADDR_LEN],
1638 struct iface *hw_addr_iface)
1641 * The procedure for choosing a bridge MAC address will, in the most
1642 * ordinary case, also choose a unique MAC that we can use as a datapath
1643 * ID. In some special cases, though, multiple bridges will end up with
1644 * the same MAC address. This is OK for the bridges, but it will confuse
1645 * the OpenFlow controller, because each datapath needs a unique datapath
1648 * Datapath IDs must be unique. It is also very desirable that they be
1649 * stable from one run to the next, so that policy set on a datapath
1652 const char *datapath_id;
1655 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1656 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1660 if (!hw_addr_iface) {
1662 * A purely internal bridge, that is, one that has no non-virtual
1663 * network devices on it at all, is difficult because it has no
1664 * natural unique identifier at all.
1666 * When the host is a XenServer, we handle this case by hashing the
1667 * host's UUID with the name of the bridge. Names of bridges are
1668 * persistent across XenServer reboots, although they can be reused if
1669 * an internal network is destroyed and then a new one is later
1670 * created, so this is fairly effective.
1672 * When the host is not a XenServer, we punt by using a random MAC
1673 * address on each run.
1675 const char *host_uuid = xenserver_get_host_uuid();
1677 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1678 dpid = dpid_from_hash(combined, strlen(combined));
1684 return eth_addr_to_uint64(bridge_ea);
1688 dpid_from_hash(const void *data, size_t n)
1690 uint8_t hash[SHA1_DIGEST_SIZE];
1692 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1693 sha1_bytes(data, n, hash);
1694 eth_addr_mark_random(hash);
1695 return eth_addr_to_uint64(hash);
1699 iface_refresh_status(struct iface *iface)
1703 enum netdev_features current;
1707 uint8_t mac[ETH_ADDR_LEN];
1710 if (iface_is_synthetic(iface)) {
1716 if (!netdev_get_status(iface->netdev, &smap)) {
1717 ovsrec_interface_set_status(iface->cfg, &smap);
1719 ovsrec_interface_set_status(iface->cfg, NULL);
1722 smap_destroy(&smap);
1724 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1725 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1727 ovsrec_interface_set_duplex(iface->cfg,
1728 netdev_features_is_full_duplex(current)
1730 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1733 ovsrec_interface_set_duplex(iface->cfg, NULL);
1734 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1737 error = netdev_get_mtu(iface->netdev, &mtu);
1740 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1743 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1746 error = netdev_get_etheraddr(iface->netdev, mac);
1748 char mac_string[32];
1750 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1751 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1753 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1757 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1760 iface_refresh_cfm_stats(struct iface *iface)
1762 const struct ovsrec_interface *cfg = iface->cfg;
1763 int fault, opup, error;
1764 const uint64_t *rmps;
1768 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1771 const char *reasons[CFM_FAULT_N_REASONS];
1772 bool fault_bool = fault;
1776 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1777 int reason = 1 << i;
1778 if (fault & reason) {
1779 reasons[j++] = cfm_fault_reason_to_str(reason);
1783 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1784 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1786 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1787 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1790 opup = ofproto_port_get_cfm_opup(iface->port->bridge->ofproto,
1793 ovsrec_interface_set_cfm_remote_opstate(cfg, opup ? "up" : "down");
1795 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1798 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1799 iface->ofp_port, &rmps, &n_rmps);
1801 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1804 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1807 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1810 int64_t cfm_health = health;
1811 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1813 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1818 iface_refresh_stats(struct iface *iface)
1820 #define IFACE_STATS \
1821 IFACE_STAT(rx_packets, "rx_packets") \
1822 IFACE_STAT(tx_packets, "tx_packets") \
1823 IFACE_STAT(rx_bytes, "rx_bytes") \
1824 IFACE_STAT(tx_bytes, "tx_bytes") \
1825 IFACE_STAT(rx_dropped, "rx_dropped") \
1826 IFACE_STAT(tx_dropped, "tx_dropped") \
1827 IFACE_STAT(rx_errors, "rx_errors") \
1828 IFACE_STAT(tx_errors, "tx_errors") \
1829 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1830 IFACE_STAT(rx_over_errors, "rx_over_err") \
1831 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1832 IFACE_STAT(collisions, "collisions")
1834 #define IFACE_STAT(MEMBER, NAME) NAME,
1835 static char *keys[] = { IFACE_STATS };
1837 int64_t values[ARRAY_SIZE(keys)];
1840 struct netdev_stats stats;
1842 if (iface_is_synthetic(iface)) {
1846 /* Intentionally ignore return value, since errors will set 'stats' to
1847 * all-1s, and we will deal with that correctly below. */
1848 netdev_get_stats(iface->netdev, &stats);
1850 /* Copy statistics into values[] array. */
1852 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1855 ovs_assert(i == ARRAY_SIZE(keys));
1857 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1863 br_refresh_stp_status(struct bridge *br)
1865 struct smap smap = SMAP_INITIALIZER(&smap);
1866 struct ofproto *ofproto = br->ofproto;
1867 struct ofproto_stp_status status;
1869 if (ofproto_get_stp_status(ofproto, &status)) {
1873 if (!status.enabled) {
1874 ovsrec_bridge_set_status(br->cfg, NULL);
1878 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1879 STP_ID_ARGS(status.bridge_id));
1880 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1881 STP_ID_ARGS(status.designated_root));
1882 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1884 ovsrec_bridge_set_status(br->cfg, &smap);
1885 smap_destroy(&smap);
1889 port_refresh_stp_status(struct port *port)
1891 struct ofproto *ofproto = port->bridge->ofproto;
1892 struct iface *iface;
1893 struct ofproto_port_stp_status status;
1895 int64_t int_values[3];
1898 if (port_is_synthetic(port)) {
1902 /* STP doesn't currently support bonds. */
1903 if (!list_is_singleton(&port->ifaces)) {
1904 ovsrec_port_set_status(port->cfg, NULL);
1908 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1910 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1914 if (!status.enabled) {
1915 ovsrec_port_set_status(port->cfg, NULL);
1916 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1920 /* Set Status column. */
1922 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1923 smap_add(&smap, "stp_state", stp_state_name(status.state));
1924 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1925 smap_add(&smap, "stp_role", stp_role_name(status.role));
1926 ovsrec_port_set_status(port->cfg, &smap);
1927 smap_destroy(&smap);
1929 /* Set Statistics column. */
1930 keys[0] = "stp_tx_count";
1931 int_values[0] = status.tx_count;
1932 keys[1] = "stp_rx_count";
1933 int_values[1] = status.rx_count;
1934 keys[2] = "stp_error_count";
1935 int_values[2] = status.error_count;
1937 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1938 ARRAY_SIZE(int_values));
1942 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1944 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1948 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1950 bool enable = enable_system_stats(cfg);
1952 system_stats_enable(enable);
1954 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1959 run_system_stats(void)
1961 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1964 stats = system_stats_run();
1966 struct ovsdb_idl_txn *txn;
1967 struct ovsdb_datum datum;
1969 txn = ovsdb_idl_txn_create(idl);
1970 ovsdb_datum_from_smap(&datum, stats);
1971 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1973 ovsdb_idl_txn_commit(txn);
1974 ovsdb_idl_txn_destroy(txn);
1980 static inline const char *
1981 nx_role_to_str(enum nx_role role)
1986 case NX_ROLE_MASTER:
1991 return "*** INVALID ROLE ***";
1996 refresh_controller_status(void)
2000 const struct ovsrec_controller *cfg;
2004 /* Accumulate status for controllers on all bridges. */
2005 HMAP_FOR_EACH (br, node, &all_bridges) {
2006 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2009 /* Update each controller in the database with current status. */
2010 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2011 struct ofproto_controller_info *cinfo =
2012 shash_find_data(&info, cfg->target);
2015 struct smap smap = SMAP_INITIALIZER(&smap);
2016 const char **values = cinfo->pairs.values;
2017 const char **keys = cinfo->pairs.keys;
2020 for (i = 0; i < cinfo->pairs.n; i++) {
2021 smap_add(&smap, keys[i], values[i]);
2024 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2025 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
2026 ovsrec_controller_set_status(cfg, &smap);
2027 smap_destroy(&smap);
2029 ovsrec_controller_set_is_connected(cfg, false);
2030 ovsrec_controller_set_role(cfg, NULL);
2031 ovsrec_controller_set_status(cfg, NULL);
2035 ofproto_free_ofproto_controller_info(&info);
2040 * Some information in the database must be kept as up-to-date as possible to
2041 * allow controllers to respond rapidly to network outages. We call these
2042 * statistics "instant" stats.
2044 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2045 * assuming that they've changed. The only means we have to determine whether
2046 * they have changed are:
2048 * - Try to commit changes to the database. If nothing changed, then
2049 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2052 * - instant_stats_run() is called late in the run loop, after anything that
2053 * might change any of the instant stats.
2055 * We use these two facts together to avoid waking the process up every
2056 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2059 /* Minimum interval between writing updates to the instant stats to the
2061 #define INSTANT_INTERVAL_MSEC 100
2063 /* Current instant stats database transaction, NULL if there is no ongoing
2065 static struct ovsdb_idl_txn *instant_txn;
2067 /* Next time (in msec on monotonic clock) at which we will update the instant
2069 static long long int instant_next_txn = LLONG_MIN;
2071 /* True if the run loop has run since we last saw that the instant stats were
2072 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2073 * to refresh the instant stats. */
2074 static bool instant_stats_could_have_changed;
2077 instant_stats_run(void)
2079 enum ovsdb_idl_txn_status status;
2081 instant_stats_could_have_changed = true;
2086 if (time_msec() < instant_next_txn) {
2089 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2091 instant_txn = ovsdb_idl_txn_create(idl);
2092 HMAP_FOR_EACH (br, node, &all_bridges) {
2093 struct iface *iface;
2096 br_refresh_stp_status(br);
2098 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2099 port_refresh_stp_status(port);
2102 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2103 enum netdev_flags flags;
2104 const char *link_state;
2105 int64_t link_resets;
2108 if (iface_is_synthetic(iface)) {
2112 current = ofproto_port_is_lacp_current(br->ofproto,
2116 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2118 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2121 error = netdev_get_flags(iface->netdev, &flags);
2123 const char *state = flags & NETDEV_UP ? "up" : "down";
2124 ovsrec_interface_set_admin_state(iface->cfg, state);
2126 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2129 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2130 ovsrec_interface_set_link_state(iface->cfg, link_state);
2132 link_resets = netdev_get_carrier_resets(iface->netdev);
2133 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2135 iface_refresh_cfm_stats(iface);
2140 status = ovsdb_idl_txn_commit(instant_txn);
2141 if (status != TXN_INCOMPLETE) {
2142 ovsdb_idl_txn_destroy(instant_txn);
2145 if (status == TXN_UNCHANGED) {
2146 instant_stats_could_have_changed = false;
2151 instant_stats_wait(void)
2154 ovsdb_idl_txn_wait(instant_txn);
2155 } else if (instant_stats_could_have_changed) {
2156 poll_timer_wait_until(instant_next_txn);
2160 /* Performs periodic activity required by bridges that needs to be done with
2161 * the least possible latency.
2163 * It makes sense to call this function a couple of times per poll loop, to
2164 * provide a significant performance boost on some benchmarks with ofprotos
2165 * that use the ofproto-dpif implementation. */
2167 bridge_run_fast(void)
2174 ofproto_enumerate_types(&types);
2175 SSET_FOR_EACH (type, &types) {
2176 ofproto_type_run_fast(type);
2178 sset_destroy(&types);
2180 HMAP_FOR_EACH (br, node, &all_bridges) {
2181 ofproto_run_fast(br->ofproto);
2188 static struct ovsrec_open_vswitch null_cfg;
2189 const struct ovsrec_open_vswitch *cfg;
2190 struct ovsdb_idl_txn *reconf_txn = NULL;
2194 bool vlan_splinters_changed;
2197 ovsrec_open_vswitch_init(&null_cfg);
2199 /* (Re)configure if necessary. */
2200 if (!reconfiguring) {
2203 if (ovsdb_idl_is_lock_contended(idl)) {
2204 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2205 struct bridge *br, *next_br;
2207 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2208 "disabling this process until it goes away");
2210 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2214 } else if (!ovsdb_idl_has_lock(idl)) {
2218 cfg = ovsrec_open_vswitch_first(idl);
2220 /* Initialize the ofproto library. This only needs to run once, but
2221 * it must be done after the configuration is set. If the
2222 * initialization has already occurred, bridge_init_ofproto()
2223 * returns immediately. */
2224 bridge_init_ofproto(cfg);
2226 /* Let each datapath type do the work that it needs to do. */
2228 ofproto_enumerate_types(&types);
2229 SSET_FOR_EACH (type, &types) {
2230 ofproto_type_run(type);
2232 sset_destroy(&types);
2234 /* Let each bridge do the work that it needs to do. */
2235 HMAP_FOR_EACH (br, node, &all_bridges) {
2236 ofproto_run(br->ofproto);
2239 /* Re-configure SSL. We do this on every trip through the main loop,
2240 * instead of just when the database changes, because the contents of the
2241 * key and certificate files can change without the database changing.
2243 * We do this before bridge_reconfigure() because that function might
2244 * initiate SSL connections and thus requires SSL to be configured. */
2245 if (cfg && cfg->ssl) {
2246 const struct ovsrec_ssl *ssl = cfg->ssl;
2248 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2249 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2252 if (!reconfiguring) {
2253 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2254 * usage has changed. */
2255 vlan_splinters_changed = false;
2256 if (vlan_splinters_enabled_anywhere) {
2257 HMAP_FOR_EACH (br, node, &all_bridges) {
2258 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2259 vlan_splinters_changed = true;
2265 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2266 idl_seqno = ovsdb_idl_get_seqno(idl);
2268 reconf_txn = ovsdb_idl_txn_create(idl);
2269 bridge_reconfigure(cfg);
2271 /* We still need to reconfigure to avoid dangling pointers to
2272 * now-destroyed ovsrec structures inside bridge data. */
2273 bridge_reconfigure(&null_cfg);
2278 if (reconfiguring) {
2281 reconf_txn = ovsdb_idl_txn_create(idl);
2283 if (bridge_reconfigure_continue(cfg)) {
2284 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2287 bridge_reconfigure_continue(&null_cfg);
2292 ovsdb_idl_txn_commit(reconf_txn);
2293 ovsdb_idl_txn_destroy(reconf_txn);
2297 /* Refresh interface and mirror stats if necessary. */
2298 if (time_msec() >= iface_stats_timer) {
2300 struct ovsdb_idl_txn *txn;
2302 txn = ovsdb_idl_txn_create(idl);
2303 HMAP_FOR_EACH (br, node, &all_bridges) {
2307 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2308 struct iface *iface;
2310 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2311 iface_refresh_stats(iface);
2312 iface_refresh_status(iface);
2316 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2317 mirror_refresh_stats(m);
2321 refresh_controller_status();
2322 ovsdb_idl_txn_commit(txn);
2323 ovsdb_idl_txn_destroy(txn); /* XXX */
2326 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2330 instant_stats_run();
2339 ovsdb_idl_wait(idl);
2341 if (reconfiguring) {
2342 poll_immediate_wake();
2346 ofproto_enumerate_types(&types);
2347 SSET_FOR_EACH (type, &types) {
2348 ofproto_type_wait(type);
2350 sset_destroy(&types);
2352 if (!hmap_is_empty(&all_bridges)) {
2355 HMAP_FOR_EACH (br, node, &all_bridges) {
2356 ofproto_wait(br->ofproto);
2358 poll_timer_wait_until(iface_stats_timer);
2361 system_stats_wait();
2362 instant_stats_wait();
2365 /* Adds some memory usage statistics for bridges into 'usage', for use with
2366 * memory_report(). */
2368 bridge_get_memory_usage(struct simap *usage)
2372 HMAP_FOR_EACH (br, node, &all_bridges) {
2373 ofproto_get_memory_usage(br->ofproto, usage);
2377 /* QoS unixctl user interface functions. */
2379 struct qos_unixctl_show_cbdata {
2381 struct iface *iface;
2385 qos_unixctl_show_cb(unsigned int queue_id,
2386 const struct smap *details,
2389 struct qos_unixctl_show_cbdata *data = aux;
2390 struct ds *ds = data->ds;
2391 struct iface *iface = data->iface;
2392 struct netdev_queue_stats stats;
2393 struct smap_node *node;
2396 ds_put_cstr(ds, "\n");
2398 ds_put_format(ds, "Queue %u:\n", queue_id);
2400 ds_put_cstr(ds, "Default:\n");
2403 SMAP_FOR_EACH (node, details) {
2404 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2407 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2409 if (stats.tx_packets != UINT64_MAX) {
2410 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2413 if (stats.tx_bytes != UINT64_MAX) {
2414 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2417 if (stats.tx_errors != UINT64_MAX) {
2418 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2421 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2422 queue_id, strerror(error));
2427 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2428 const char *argv[], void *aux OVS_UNUSED)
2430 struct ds ds = DS_EMPTY_INITIALIZER;
2431 struct smap smap = SMAP_INITIALIZER(&smap);
2432 struct iface *iface;
2434 struct smap_node *node;
2435 struct qos_unixctl_show_cbdata data;
2438 iface = iface_find(argv[1]);
2440 unixctl_command_reply_error(conn, "no such interface");
2444 netdev_get_qos(iface->netdev, &type, &smap);
2446 if (*type != '\0') {
2447 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2449 SMAP_FOR_EACH (node, &smap) {
2450 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2455 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2458 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2460 unixctl_command_reply(conn, ds_cstr(&ds));
2462 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2463 unixctl_command_reply_error(conn, ds_cstr(&ds));
2466 smap_destroy(&smap);
2470 /* Bridge reconfiguration functions. */
2472 bridge_create(const struct ovsrec_bridge *br_cfg)
2476 ovs_assert(!bridge_lookup(br_cfg->name));
2477 br = xzalloc(sizeof *br);
2479 br->name = xstrdup(br_cfg->name);
2480 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2483 /* Derive the default Ethernet address from the bridge's UUID. This should
2484 * be unique and it will be stable between ovs-vswitchd runs. */
2485 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2486 eth_addr_mark_random(br->default_ea);
2488 hmap_init(&br->ports);
2489 hmap_init(&br->ifaces);
2490 hmap_init(&br->iface_by_name);
2491 hmap_init(&br->mirrors);
2493 hmap_init(&br->if_cfg_todo);
2494 list_init(&br->ofpp_garbage);
2496 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2500 bridge_destroy(struct bridge *br)
2503 struct mirror *mirror, *next_mirror;
2504 struct port *port, *next_port;
2505 struct if_cfg *if_cfg, *next_if_cfg;
2506 struct ofpp_garbage *garbage, *next_garbage;
2508 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2511 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2512 mirror_destroy(mirror);
2514 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2515 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2518 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2519 &br->ofpp_garbage) {
2520 list_remove(&garbage->list_node);
2524 hmap_remove(&all_bridges, &br->node);
2525 ofproto_destroy(br->ofproto);
2526 hmap_destroy(&br->ifaces);
2527 hmap_destroy(&br->ports);
2528 hmap_destroy(&br->iface_by_name);
2529 hmap_destroy(&br->mirrors);
2530 hmap_destroy(&br->if_cfg_todo);
2537 static struct bridge *
2538 bridge_lookup(const char *name)
2542 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2543 if (!strcmp(br->name, name)) {
2550 /* Handle requests for a listing of all flows known by the OpenFlow
2551 * stack, including those normally hidden. */
2553 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2554 const char *argv[], void *aux OVS_UNUSED)
2559 br = bridge_lookup(argv[1]);
2561 unixctl_command_reply_error(conn, "Unknown bridge");
2566 ofproto_get_all_flows(br->ofproto, &results);
2568 unixctl_command_reply(conn, ds_cstr(&results));
2569 ds_destroy(&results);
2572 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2573 * connections and reconnect. If BRIDGE is not specified, then all bridges
2574 * drop their controller connections and reconnect. */
2576 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2577 const char *argv[], void *aux OVS_UNUSED)
2581 br = bridge_lookup(argv[1]);
2583 unixctl_command_reply_error(conn, "Unknown bridge");
2586 ofproto_reconnect_controllers(br->ofproto);
2588 HMAP_FOR_EACH (br, node, &all_bridges) {
2589 ofproto_reconnect_controllers(br->ofproto);
2592 unixctl_command_reply(conn, NULL);
2596 bridge_get_controllers(const struct bridge *br,
2597 struct ovsrec_controller ***controllersp)
2599 struct ovsrec_controller **controllers;
2600 size_t n_controllers;
2602 controllers = br->cfg->controller;
2603 n_controllers = br->cfg->n_controller;
2605 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2611 *controllersp = controllers;
2613 return n_controllers;
2617 bridge_queue_if_cfg(struct bridge *br,
2618 const struct ovsrec_interface *cfg,
2619 const struct ovsrec_port *parent)
2621 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2624 if_cfg->parent = parent;
2625 if_cfg->ofport = iface_pick_ofport(cfg);
2626 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2627 hash_string(if_cfg->cfg->name, 0));
2630 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2631 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2632 * 'br' needs to complete its configuration. */
2634 bridge_add_del_ports(struct bridge *br,
2635 const unsigned long int *splinter_vlans)
2637 struct shash_node *port_node;
2638 struct port *port, *next;
2639 struct shash new_ports;
2642 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2644 /* Collect new ports. */
2645 shash_init(&new_ports);
2646 for (i = 0; i < br->cfg->n_ports; i++) {
2647 const char *name = br->cfg->ports[i]->name;
2648 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2649 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2653 if (bridge_get_controllers(br, NULL)
2654 && !shash_find(&new_ports, br->name)) {
2655 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2656 br->name, br->name);
2658 ovsrec_interface_init(&br->synth_local_iface);
2659 ovsrec_port_init(&br->synth_local_port);
2661 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2662 br->synth_local_port.n_interfaces = 1;
2663 br->synth_local_port.name = br->name;
2665 br->synth_local_iface.name = br->name;
2666 br->synth_local_iface.type = "internal";
2668 br->synth_local_ifacep = &br->synth_local_iface;
2670 shash_add(&new_ports, br->name, &br->synth_local_port);
2673 if (splinter_vlans) {
2674 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2677 /* Get rid of deleted ports.
2678 * Get rid of deleted interfaces on ports that still exist. */
2679 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2680 port->cfg = shash_find_data(&new_ports, port->name);
2684 port_del_ifaces(port);
2688 /* Update iface->cfg and iface->type in interfaces that still exist.
2689 * Add new interfaces to creation queue. */
2690 SHASH_FOR_EACH (port_node, &new_ports) {
2691 const struct ovsrec_port *port = port_node->data;
2694 for (i = 0; i < port->n_interfaces; i++) {
2695 const struct ovsrec_interface *cfg = port->interfaces[i];
2696 struct iface *iface = iface_lookup(br, cfg->name);
2697 const char *type = iface_get_type(cfg, br->cfg);
2702 } else if (!strcmp(type, "null")) {
2703 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2704 " may be removed in February 2013. Please email"
2705 " dev@openvswitch.org with concerns.",
2708 bridge_queue_if_cfg(br, cfg, port);
2713 shash_destroy(&new_ports);
2716 /* Initializes 'oc' appropriately as a management service controller for
2719 * The caller must free oc->target when it is no longer needed. */
2721 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2722 struct ofproto_controller *oc)
2724 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2725 oc->max_backoff = 0;
2726 oc->probe_interval = 60;
2727 oc->band = OFPROTO_OUT_OF_BAND;
2729 oc->burst_limit = 0;
2730 oc->enable_async_msgs = true;
2733 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2735 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2736 struct ofproto_controller *oc)
2740 oc->target = c->target;
2741 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2742 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2743 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2744 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2745 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2746 oc->burst_limit = (c->controller_burst_limit
2747 ? *c->controller_burst_limit : 0);
2748 oc->enable_async_msgs = (!c->enable_async_messages
2749 || *c->enable_async_messages);
2750 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2751 if (dscp < 0 || dscp > 63) {
2752 dscp = DSCP_DEFAULT;
2757 /* Configures the IP stack for 'br''s local interface properly according to the
2758 * configuration in 'c'. */
2760 bridge_configure_local_iface_netdev(struct bridge *br,
2761 struct ovsrec_controller *c)
2763 struct netdev *netdev;
2764 struct in_addr mask, gateway;
2766 struct iface *local_iface;
2769 /* If there's no local interface or no IP address, give up. */
2770 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2771 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2775 /* Bring up the local interface. */
2776 netdev = local_iface->netdev;
2777 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2779 /* Configure the IP address and netmask. */
2780 if (!c->local_netmask
2781 || !inet_aton(c->local_netmask, &mask)
2783 mask.s_addr = guess_netmask(ip.s_addr);
2785 if (!netdev_set_in4(netdev, ip, mask)) {
2786 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2787 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2790 /* Configure the default gateway. */
2791 if (c->local_gateway
2792 && inet_aton(c->local_gateway, &gateway)
2793 && gateway.s_addr) {
2794 if (!netdev_add_router(netdev, gateway)) {
2795 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2796 br->name, IP_ARGS(gateway.s_addr));
2801 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2802 * in either string are treated as equal to any number of slashes in the other,
2803 * e.g. "x///y" is equal to "x/y".
2805 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2806 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2807 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2808 * 'b' against a prefix of 'a'.
2811 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2813 const char *b_start = b;
2815 if (b - b_start >= b_stoplen) {
2817 } else if (*a != *b) {
2819 } else if (*a == '/') {
2820 a += strspn(a, "/");
2821 b += strspn(b, "/");
2822 } else if (*a == '\0') {
2832 bridge_configure_remotes(struct bridge *br,
2833 const struct sockaddr_in *managers, size_t n_managers)
2835 bool disable_in_band;
2837 struct ovsrec_controller **controllers;
2838 size_t n_controllers;
2840 enum ofproto_fail_mode fail_mode;
2842 struct ofproto_controller *ocs;
2846 /* Check if we should disable in-band control on this bridge. */
2847 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2850 /* Set OpenFlow queue ID for in-band control. */
2851 ofproto_set_in_band_queue(br->ofproto,
2852 smap_get_int(&br->cfg->other_config,
2853 "in-band-queue", -1));
2855 if (disable_in_band) {
2856 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2858 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2861 n_controllers = bridge_get_controllers(br, &controllers);
2863 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2866 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2867 for (i = 0; i < n_controllers; i++) {
2868 struct ovsrec_controller *c = controllers[i];
2870 if (!strncmp(c->target, "punix:", 6)
2871 || !strncmp(c->target, "unix:", 5)) {
2872 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2875 if (!strncmp(c->target, "unix:", 5)) {
2876 /* Connect to a listening socket */
2877 whitelist = xasprintf("unix:%s/", ovs_rundir());
2878 if (strchr(c->target, '/') &&
2879 !equal_pathnames(c->target, whitelist,
2880 strlen(whitelist))) {
2881 /* Absolute path specified, but not in ovs_rundir */
2882 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2883 "controller \"%s\" due to possibility for "
2884 "remote exploit. Instead, specify socket "
2885 "in whitelisted \"%s\" or connect to "
2886 "\"unix:%s/%s.mgmt\" (which is always "
2887 "available without special configuration).",
2888 br->name, c->target, whitelist,
2889 ovs_rundir(), br->name);
2894 whitelist = xasprintf("punix:%s/%s.controller",
2895 ovs_rundir(), br->name);
2896 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2897 /* Prevent remote ovsdb-server users from accessing
2898 * arbitrary Unix domain sockets and overwriting arbitrary
2900 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2901 "controller \"%s\" due to possibility of "
2902 "overwriting local files. Instead, specify "
2903 "whitelisted \"%s\" or connect to "
2904 "\"unix:%s/%s.mgmt\" (which is always "
2905 "available without special configuration).",
2906 br->name, c->target, whitelist,
2907 ovs_rundir(), br->name);
2916 bridge_configure_local_iface_netdev(br, c);
2917 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2918 if (disable_in_band) {
2919 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2924 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
2925 bridge_get_allowed_versions(br));
2926 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2929 /* Set the fail-mode. */
2930 fail_mode = !br->cfg->fail_mode
2931 || !strcmp(br->cfg->fail_mode, "standalone")
2932 ? OFPROTO_FAIL_STANDALONE
2933 : OFPROTO_FAIL_SECURE;
2934 ofproto_set_fail_mode(br->ofproto, fail_mode);
2936 /* Configure OpenFlow controller connection snooping. */
2937 if (!ofproto_has_snoops(br->ofproto)) {
2941 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2942 ovs_rundir(), br->name));
2943 ofproto_set_snoops(br->ofproto, &snoops);
2944 sset_destroy(&snoops);
2949 bridge_configure_tables(struct bridge *br)
2951 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2955 n_tables = ofproto_get_n_tables(br->ofproto);
2957 for (i = 0; i < n_tables; i++) {
2958 struct ofproto_table_settings s;
2961 s.max_flows = UINT_MAX;
2965 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2966 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2969 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2970 s.max_flows = *cfg->flow_limit;
2972 if (cfg->overflow_policy
2973 && !strcmp(cfg->overflow_policy, "evict")) {
2976 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2977 for (k = 0; k < cfg->n_groups; k++) {
2978 const char *string = cfg->groups[k];
2981 msg = mf_parse_subfield__(&s.groups[k], &string);
2983 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2984 "'groups' (%s)", br->name, i, msg);
2986 } else if (*string) {
2987 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2988 "element '%s' contains trailing garbage",
2989 br->name, i, cfg->groups[k]);
2997 ofproto_configure_table(br->ofproto, i, &s);
3001 for (; j < br->cfg->n_flow_tables; j++) {
3002 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3003 "%"PRId64" not supported by this datapath", br->name,
3004 br->cfg->key_flow_tables[j]);
3009 bridge_configure_dp_desc(struct bridge *br)
3011 ofproto_set_dp_desc(br->ofproto,
3012 smap_get(&br->cfg->other_config, "dp-desc"));
3015 /* Port functions. */
3017 static struct port *
3018 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3022 port = xzalloc(sizeof *port);
3024 port->name = xstrdup(cfg->name);
3026 list_init(&port->ifaces);
3028 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3032 /* Deletes interfaces from 'port' that are no longer configured for it. */
3034 port_del_ifaces(struct port *port)
3036 struct iface *iface, *next;
3037 struct sset new_ifaces;
3040 /* Collect list of new interfaces. */
3041 sset_init(&new_ifaces);
3042 for (i = 0; i < port->cfg->n_interfaces; i++) {
3043 const char *name = port->cfg->interfaces[i]->name;
3044 const char *type = port->cfg->interfaces[i]->type;
3045 if (strcmp(type, "null")) {
3046 sset_add(&new_ifaces, name);
3050 /* Get rid of deleted interfaces. */
3051 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3052 if (!sset_contains(&new_ifaces, iface->name)) {
3053 iface_destroy(iface);
3057 sset_destroy(&new_ifaces);
3061 port_destroy(struct port *port)
3064 struct bridge *br = port->bridge;
3065 struct iface *iface, *next;
3068 ofproto_bundle_unregister(br->ofproto, port);
3071 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3072 iface_destroy(iface);
3075 hmap_remove(&br->ports, &port->hmap_node);
3081 static struct port *
3082 port_lookup(const struct bridge *br, const char *name)
3086 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3088 if (!strcmp(port->name, name)) {
3096 enable_lacp(struct port *port, bool *activep)
3098 if (!port->cfg->lacp) {
3099 /* XXX when LACP implementation has been sufficiently tested, enable by
3100 * default and make active on bonded ports. */
3102 } else if (!strcmp(port->cfg->lacp, "off")) {
3104 } else if (!strcmp(port->cfg->lacp, "active")) {
3107 } else if (!strcmp(port->cfg->lacp, "passive")) {
3111 VLOG_WARN("port %s: unknown LACP mode %s",
3112 port->name, port->cfg->lacp);
3117 static struct lacp_settings *
3118 port_configure_lacp(struct port *port, struct lacp_settings *s)
3120 const char *lacp_time, *system_id;
3123 if (!enable_lacp(port, &s->active)) {
3127 s->name = port->name;
3129 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3131 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3132 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3133 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3134 " address.", port->name, system_id);
3138 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3141 if (eth_addr_is_zero(s->id)) {
3142 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3146 /* Prefer bondable links if unspecified. */
3147 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3149 s->priority = (priority > 0 && priority <= UINT16_MAX
3151 : UINT16_MAX - !list_is_short(&port->ifaces));
3153 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3154 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3159 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3161 int priority, portid, key;
3163 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3164 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3166 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3168 if (portid <= 0 || portid > UINT16_MAX) {
3169 portid = iface->ofp_port;
3172 if (priority <= 0 || priority > UINT16_MAX) {
3173 priority = UINT16_MAX;
3176 if (key < 0 || key > UINT16_MAX) {
3180 s->name = iface->name;
3182 s->priority = priority;
3187 port_configure_bond(struct port *port, struct bond_settings *s,
3188 uint32_t *bond_stable_ids)
3190 const char *detect_s;
3191 struct iface *iface;
3192 int miimon_interval;
3195 s->name = port->name;
3197 if (port->cfg->bond_mode) {
3198 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3199 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3200 port->name, port->cfg->bond_mode,
3201 bond_mode_to_string(s->balance));
3204 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3206 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3207 * active-backup. At some point we should remove this warning. */
3208 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3209 " in previous versions, the default bond_mode was"
3210 " balance-slb", port->name,
3211 bond_mode_to_string(s->balance));
3213 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3214 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3215 "please use another bond type or disable flood_vlans",
3219 miimon_interval = smap_get_int(&port->cfg->other_config,
3220 "bond-miimon-interval", 0);
3221 if (miimon_interval <= 0) {
3222 miimon_interval = 200;
3225 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3226 if (!detect_s || !strcmp(detect_s, "carrier")) {
3227 miimon_interval = 0;
3228 } else if (strcmp(detect_s, "miimon")) {
3229 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3230 "defaulting to carrier", port->name, detect_s);
3231 miimon_interval = 0;
3234 s->up_delay = MAX(0, port->cfg->bond_updelay);
3235 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3236 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3237 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3238 "bond-rebalance-interval", 10000);
3239 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3240 s->rebalance_interval = 1000;
3243 s->fake_iface = port->cfg->bond_fake_iface;
3246 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3247 long long stable_id;
3249 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3251 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3252 stable_id = iface->ofp_port;
3254 bond_stable_ids[i++] = stable_id;
3256 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3260 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3261 * instead of obtaining it from the database. */
3263 port_is_synthetic(const struct port *port)
3265 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3268 /* Interface functions. */
3271 iface_is_internal(const struct ovsrec_interface *iface,
3272 const struct ovsrec_bridge *br)
3274 /* The local port and "internal" ports are always "internal". */
3275 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3278 /* Returns the correct network device type for interface 'iface' in bridge
3281 iface_get_type(const struct ovsrec_interface *iface,
3282 const struct ovsrec_bridge *br)
3286 /* The local port always has type "internal". Other ports take
3287 * their type from the database and default to "system" if none is
3289 if (iface_is_internal(iface, br)) {
3292 type = iface->type[0] ? iface->type : "system";
3295 return ofproto_port_open_type(br->datapath_type, type);
3299 iface_destroy(struct iface *iface)
3302 struct port *port = iface->port;
3303 struct bridge *br = port->bridge;
3305 if (br->ofproto && iface->ofp_port >= 0) {
3306 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3309 if (iface->ofp_port >= 0) {
3310 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3313 list_remove(&iface->port_elem);
3314 hmap_remove(&br->iface_by_name, &iface->name_node);
3316 netdev_close(iface->netdev);
3323 static struct iface *
3324 iface_lookup(const struct bridge *br, const char *name)
3326 struct iface *iface;
3328 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3329 &br->iface_by_name) {
3330 if (!strcmp(iface->name, name)) {
3338 static struct iface *
3339 iface_find(const char *name)
3341 const struct bridge *br;
3343 HMAP_FOR_EACH (br, node, &all_bridges) {
3344 struct iface *iface = iface_lookup(br, name);
3353 static struct if_cfg *
3354 if_cfg_lookup(const struct bridge *br, const char *name)
3356 struct if_cfg *if_cfg;
3358 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3360 if (!strcmp(if_cfg->cfg->name, name)) {
3368 static struct iface *
3369 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3371 struct iface *iface;
3373 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3374 hash_int(ofp_port, 0), &br->ifaces) {
3375 if (iface->ofp_port == ofp_port) {
3382 /* Set Ethernet address of 'iface', if one is specified in the configuration
3385 iface_set_mac(struct iface *iface)
3387 uint8_t ea[ETH_ADDR_LEN];
3389 if (!strcmp(iface->type, "internal")
3390 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3391 if (iface->ofp_port == OFPP_LOCAL) {
3392 VLOG_ERR("interface %s: ignoring mac in Interface record "
3393 "(use Bridge record to set local port's mac)",
3395 } else if (eth_addr_is_multicast(ea)) {
3396 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3399 int error = netdev_set_etheraddr(iface->netdev, ea);
3401 VLOG_ERR("interface %s: setting MAC failed (%s)",
3402 iface->name, strerror(error));
3408 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3410 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3412 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3413 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3417 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3418 * sets the "ofport" field to -1.
3420 * This is appropriate when 'if_cfg''s interface cannot be created or is
3421 * otherwise invalid. */
3423 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3425 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3426 ovsrec_interface_set_status(if_cfg, NULL);
3427 ovsrec_interface_set_admin_state(if_cfg, NULL);
3428 ovsrec_interface_set_duplex(if_cfg, NULL);
3429 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3430 ovsrec_interface_set_link_state(if_cfg, NULL);
3431 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3432 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3433 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3434 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3435 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3436 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3437 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3441 struct iface_delete_queues_cbdata {
3442 struct netdev *netdev;
3443 const struct ovsdb_datum *queues;
3447 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3449 union ovsdb_atom atom;
3451 atom.integer = target;
3452 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3456 iface_delete_queues(unsigned int queue_id,
3457 const struct smap *details OVS_UNUSED, void *cbdata_)
3459 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3461 if (!queue_ids_include(cbdata->queues, queue_id)) {
3462 netdev_delete_queue(cbdata->netdev, queue_id);
3467 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3469 struct ofpbuf queues_buf;
3471 ofpbuf_init(&queues_buf, 0);
3473 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3474 netdev_set_qos(iface->netdev, NULL, NULL);
3476 struct iface_delete_queues_cbdata cbdata;
3480 /* Configure top-level Qos for 'iface'. */
3481 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3483 /* Deconfigure queues that were deleted. */
3484 cbdata.netdev = iface->netdev;
3485 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3487 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3489 /* Configure queues for 'iface'. */
3491 for (i = 0; i < qos->n_queues; i++) {
3492 const struct ovsrec_queue *queue = qos->value_queues[i];
3493 unsigned int queue_id = qos->key_queues[i];
3495 if (queue_id == 0) {
3499 if (queue->n_dscp == 1) {
3500 struct ofproto_port_queue *port_queue;
3502 port_queue = ofpbuf_put_uninit(&queues_buf,
3503 sizeof *port_queue);
3504 port_queue->queue = queue_id;
3505 port_queue->dscp = queue->dscp[0];
3508 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3511 struct smap details;
3513 smap_init(&details);
3514 netdev_set_queue(iface->netdev, 0, &details);
3515 smap_destroy(&details);
3519 if (iface->ofp_port >= 0) {
3520 const struct ofproto_port_queue *port_queues = queues_buf.data;
3521 size_t n_queues = queues_buf.size / sizeof *port_queues;
3523 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3524 port_queues, n_queues);
3527 netdev_set_policing(iface->netdev,
3528 iface->cfg->ingress_policing_rate,
3529 iface->cfg->ingress_policing_burst);
3531 ofpbuf_uninit(&queues_buf);
3535 iface_configure_cfm(struct iface *iface)
3537 const struct ovsrec_interface *cfg = iface->cfg;
3538 const char *opstate_str;
3539 const char *cfm_ccm_vlan;
3540 struct cfm_settings s;
3541 struct smap netdev_args;
3543 if (!cfg->n_cfm_mpid) {
3544 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3548 s.check_tnl_key = false;
3549 smap_init(&netdev_args);
3550 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3551 const char *key = smap_get(&netdev_args, "key");
3552 const char *in_key = smap_get(&netdev_args, "in_key");
3554 s.check_tnl_key = (key && !strcmp(key, "flow"))
3555 || (in_key && !strcmp(in_key, "flow"));
3557 smap_destroy(&netdev_args);
3559 s.mpid = *cfg->cfm_mpid;
3560 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3561 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3562 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3564 if (s.interval <= 0) {
3568 if (!cfm_ccm_vlan) {
3570 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3571 s.ccm_vlan = CFM_RANDOM_VLAN;
3573 s.ccm_vlan = atoi(cfm_ccm_vlan);
3574 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3579 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3582 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3583 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3585 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3588 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3589 * instead of obtaining it from the database. */
3591 iface_is_synthetic(const struct iface *iface)
3593 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3597 iface_pick_ofport(const struct ovsrec_interface *cfg)
3599 int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3600 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3604 /* Port mirroring. */
3606 static struct mirror *
3607 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3611 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3612 if (uuid_equals(uuid, &m->uuid)) {
3620 bridge_configure_mirrors(struct bridge *br)
3622 const struct ovsdb_datum *mc;
3623 unsigned long *flood_vlans;
3624 struct mirror *m, *next;
3627 /* Get rid of deleted mirrors. */
3628 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3629 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3630 union ovsdb_atom atom;
3632 atom.uuid = m->uuid;
3633 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3638 /* Add new mirrors and reconfigure existing ones. */
3639 for (i = 0; i < br->cfg->n_mirrors; i++) {
3640 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3641 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3643 m = mirror_create(br, cfg);
3646 if (!mirror_configure(m)) {
3651 /* Update flooded vlans (for RSPAN). */
3652 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3653 br->cfg->n_flood_vlans);
3654 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3655 bitmap_free(flood_vlans);
3658 static struct mirror *
3659 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3663 m = xzalloc(sizeof *m);
3664 m->uuid = cfg->header_.uuid;
3665 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3667 m->name = xstrdup(cfg->name);
3673 mirror_destroy(struct mirror *m)
3676 struct bridge *br = m->bridge;
3679 ofproto_mirror_unregister(br->ofproto, m);
3682 hmap_remove(&br->mirrors, &m->hmap_node);
3689 mirror_collect_ports(struct mirror *m,
3690 struct ovsrec_port **in_ports, int n_in_ports,
3691 void ***out_portsp, size_t *n_out_portsp)
3693 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3694 size_t n_out_ports = 0;
3697 for (i = 0; i < n_in_ports; i++) {
3698 const char *name = in_ports[i]->name;
3699 struct port *port = port_lookup(m->bridge, name);
3701 out_ports[n_out_ports++] = port;
3703 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3704 "port %s", m->bridge->name, m->name, name);
3707 *out_portsp = out_ports;
3708 *n_out_portsp = n_out_ports;
3712 mirror_configure(struct mirror *m)
3714 const struct ovsrec_mirror *cfg = m->cfg;
3715 struct ofproto_mirror_settings s;
3718 if (strcmp(cfg->name, m->name)) {
3720 m->name = xstrdup(cfg->name);
3724 /* Get output port or VLAN. */
3725 if (cfg->output_port) {
3726 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3727 if (!s.out_bundle) {
3728 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3729 m->bridge->name, m->name);
3732 s.out_vlan = UINT16_MAX;
3734 if (cfg->output_vlan) {
3735 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3736 "output vlan; ignoring output vlan",
3737 m->bridge->name, m->name);
3739 } else if (cfg->output_vlan) {
3740 /* The database should prevent invalid VLAN values. */
3741 s.out_bundle = NULL;
3742 s.out_vlan = *cfg->output_vlan;
3744 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3745 m->bridge->name, m->name);
3749 /* Get port selection. */
3750 if (cfg->select_all) {
3751 size_t n_ports = hmap_count(&m->bridge->ports);
3752 void **ports = xmalloc(n_ports * sizeof *ports);
3757 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3767 /* Get ports, dropping ports that don't exist.
3768 * The IDL ensures that there are no duplicates. */
3769 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3770 &s.srcs, &s.n_srcs);
3771 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3772 &s.dsts, &s.n_dsts);
3775 /* Get VLAN selection. */
3776 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3779 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3782 if (s.srcs != s.dsts) {
3791 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3793 * This is deprecated. It is only for compatibility with broken device drivers
3794 * in old versions of Linux that do not properly support VLANs when VLAN
3795 * devices are not used. When broken device drivers are no longer in
3796 * widespread use, we will delete these interfaces. */
3798 static struct ovsrec_port **recs;
3799 static size_t n_recs, allocated_recs;
3801 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3802 * splinters are reconfigured. */
3804 register_rec(struct ovsrec_port *rec)
3806 if (n_recs >= allocated_recs) {
3807 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3809 recs[n_recs++] = rec;
3812 /* Frees all of the ports registered with register_reg(). */
3814 free_registered_recs(void)
3818 for (i = 0; i < n_recs; i++) {
3819 struct ovsrec_port *port = recs[i];
3822 for (j = 0; j < port->n_interfaces; j++) {
3823 struct ovsrec_interface *iface = port->interfaces[j];
3828 smap_destroy(&port->other_config);
3829 free(port->interfaces);
3837 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3840 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3842 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3846 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3849 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3850 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3851 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3854 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3855 * use, returns NULL.
3857 * Updates 'vlan_splinters_enabled_anywhere'. */
3858 static unsigned long int *
3859 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3861 unsigned long int *splinter_vlans;
3862 struct sset splinter_ifaces;
3863 const char *real_dev_name;
3864 struct shash *real_devs;
3865 struct shash_node *node;
3869 /* Free space allocated for synthesized ports and interfaces, since we're
3870 * in the process of reconstructing all of them. */
3871 free_registered_recs();
3873 splinter_vlans = bitmap_allocate(4096);
3874 sset_init(&splinter_ifaces);
3875 vlan_splinters_enabled_anywhere = false;
3876 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3877 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3880 for (j = 0; j < br_cfg->n_ports; j++) {
3881 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3884 for (k = 0; k < port_cfg->n_interfaces; k++) {
3885 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3887 if (vlan_splinters_is_enabled(iface_cfg)) {
3888 vlan_splinters_enabled_anywhere = true;
3889 sset_add(&splinter_ifaces, iface_cfg->name);
3890 vlan_bitmap_from_array__(port_cfg->trunks,
3896 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3897 bitmap_set1(splinter_vlans, *port_cfg->tag);
3902 if (!vlan_splinters_enabled_anywhere) {
3903 free(splinter_vlans);
3904 sset_destroy(&splinter_ifaces);
3908 HMAP_FOR_EACH (br, node, &all_bridges) {
3910 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3914 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3915 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3916 * device to be created for it. */
3917 bitmap_set0(splinter_vlans, 0);
3918 bitmap_set0(splinter_vlans, 4095);
3920 /* Delete all VLAN devices that we don't need. */
3922 real_devs = vlandev_get_real_devs();
3923 SHASH_FOR_EACH (node, real_devs) {
3924 const struct vlan_real_dev *real_dev = node->data;
3925 const struct vlan_dev *vlan_dev;
3926 bool real_dev_has_splinters;
3928 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3930 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3931 if (!real_dev_has_splinters
3932 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3933 struct netdev *netdev;
3935 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3936 if (!netdev_get_in4(netdev, NULL, NULL) ||
3937 !netdev_get_in6(netdev, NULL)) {
3938 vlandev_del(vlan_dev->name);
3940 /* It has an IP address configured, so we don't own
3941 * it. Don't delete it. */
3943 netdev_close(netdev);
3950 /* Add all VLAN devices that we need. */
3951 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3954 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3955 if (!vlandev_get_name(real_dev_name, vid)) {
3956 vlandev_add(real_dev_name, vid);
3963 sset_destroy(&splinter_ifaces);
3965 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3966 free(splinter_vlans);
3969 return splinter_vlans;
3972 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3975 configure_splinter_port(struct port *port)
3977 struct ofproto *ofproto = port->bridge->ofproto;
3978 uint16_t realdev_ofp_port;
3979 const char *realdev_name;
3980 struct iface *vlandev, *realdev;
3982 ofproto_bundle_unregister(port->bridge->ofproto, port);
3984 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3987 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3988 realdev = iface_lookup(port->bridge, realdev_name);
3989 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3991 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3995 static struct ovsrec_port *
3996 synthesize_splinter_port(const char *real_dev_name,
3997 const char *vlan_dev_name, int vid)
3999 struct ovsrec_interface *iface;
4000 struct ovsrec_port *port;
4002 iface = xmalloc(sizeof *iface);
4003 ovsrec_interface_init(iface);
4004 iface->name = xstrdup(vlan_dev_name);
4005 iface->type = "system";
4007 port = xmalloc(sizeof *port);
4008 ovsrec_port_init(port);
4009 port->interfaces = xmemdup(&iface, sizeof iface);
4010 port->n_interfaces = 1;
4011 port->name = xstrdup(vlan_dev_name);
4012 port->vlan_mode = "splinter";
4013 port->tag = xmalloc(sizeof *port->tag);
4016 smap_add(&port->other_config, "realdev", real_dev_name);
4022 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4023 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4024 * 1-bit in the 'splinter_vlans' bitmap. */
4026 add_vlan_splinter_ports(struct bridge *br,
4027 const unsigned long int *splinter_vlans,
4028 struct shash *ports)
4032 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4033 * we're modifying 'ports'. */
4034 for (i = 0; i < br->cfg->n_ports; i++) {
4035 const char *name = br->cfg->ports[i]->name;
4036 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4039 for (j = 0; j < port_cfg->n_interfaces; j++) {
4040 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4042 if (vlan_splinters_is_enabled(iface_cfg)) {
4043 const char *real_dev_name;
4046 real_dev_name = iface_cfg->name;
4047 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4048 const char *vlan_dev_name;
4050 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4052 && !shash_find(ports, vlan_dev_name)) {
4053 shash_add(ports, vlan_dev_name,
4054 synthesize_splinter_port(
4055 real_dev_name, vlan_dev_name, vid));
4064 mirror_refresh_stats(struct mirror *m)
4066 struct ofproto *ofproto = m->bridge->ofproto;
4067 uint64_t tx_packets, tx_bytes;
4070 size_t stat_cnt = 0;
4072 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4073 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4077 if (tx_packets != UINT64_MAX) {
4078 keys[stat_cnt] = "tx_packets";
4079 values[stat_cnt] = tx_packets;
4082 if (tx_bytes != UINT64_MAX) {
4083 keys[stat_cnt] = "tx_bytes";
4084 values[stat_cnt] = tx_bytes;
4088 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);