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.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
41 #include "ofproto/ofproto.h"
42 #include "poll-loop.h"
46 #include "socket-util.h"
48 #include "stream-ssl.h"
50 #include "system-stats.h"
55 #include "lib/vswitch-idl.h"
56 #include "xenserver.h"
58 #include "sflow_api.h"
59 #include "vlan-bitmap.h"
61 VLOG_DEFINE_THIS_MODULE(bridge);
63 COVERAGE_DEFINE(bridge_reconfigure);
65 /* Configuration of an uninstantiated iface. */
67 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
68 const struct ovsrec_interface *cfg; /* Interface record. */
69 const struct ovsrec_port *parent; /* Parent port record. */
70 uint16_t ofport; /* Requested OpenFlow port number. */
73 /* OpenFlow port slated for removal from ofproto. */
75 struct list list_node; /* Node in bridge's ofpp_garbage. */
76 uint16_t ofp_port; /* Port to be deleted. */
80 /* These members are always valid. */
81 struct list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
86 /* These members are valid only after bridge_reconfigure() causes them to
88 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
89 uint16_t ofp_port; /* OpenFlow port number, */
90 /* OFPP_NONE if unknown. */
91 struct netdev *netdev; /* Network device. */
92 const char *type; /* Usually same as cfg->type. */
93 const struct ovsrec_interface *cfg;
97 struct uuid uuid; /* UUID of this "mirror" record in database. */
98 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
99 struct bridge *bridge;
101 const struct ovsrec_mirror *cfg;
105 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
106 struct bridge *bridge;
109 const struct ovsrec_port *cfg;
111 /* An ordinary bridge port has 1 interface.
112 * A bridge port for bonding has at least 2 interfaces. */
113 struct list ifaces; /* List of "struct iface"s. */
117 struct hmap_node node; /* In 'all_bridges'. */
118 char *name; /* User-specified arbitrary name. */
119 char *type; /* Datapath type. */
120 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
121 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
122 const struct ovsrec_bridge *cfg;
124 /* OpenFlow switch processing. */
125 struct ofproto *ofproto; /* OpenFlow switch. */
128 struct hmap ports; /* "struct port"s indexed by name. */
129 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
130 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
132 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
133 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
134 Indexed on 'cfg->name'. */
136 /* Port mirroring. */
137 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
145 /* All bridges, indexed by name. */
146 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
148 /* OVSDB IDL used to obtain configuration. */
149 static struct ovsdb_idl *idl;
151 /* We want to complete daemonization, fully detaching from our parent process,
152 * only after we have completed our initial configuration, committed our state
153 * to the database, and received confirmation back from the database server
154 * that it applied the commit. This allows our parent process to know that,
155 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
156 * to have already been filled in. (It is only "very likely" rather than
157 * certain because there is always a slim possibility that the transaction will
158 * fail or that some other client has added new bridges, ports, etc. while
159 * ovs-vswitchd was configuring using an old configuration.)
161 * We only need to do this once for our initial configuration at startup, so
162 * 'initial_config_done' tracks whether we've already done it. While we are
163 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
164 * itself and is otherwise NULL. */
165 static bool initial_config_done;
166 static struct ovsdb_idl_txn *daemonize_txn;
168 /* Most recently processed IDL sequence number. */
169 static unsigned int idl_seqno;
171 /* Each time this timer expires, the bridge fetches interface and mirror
172 * statistics and pushes them into the database. */
173 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
174 static long long int iface_stats_timer = LLONG_MIN;
176 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
177 * expensive. This can cause bridge_reconfigure() to take a long time during
178 * which no other work can be done. To deal with this problem, we limit port
179 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
180 * call to bridge_reconfigure(). If there is more work to do after the limit
181 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
182 * This allows the rest of the code to catch up on important things like
183 * forwarding packets. */
184 #define OFP_PORT_ACTION_WINDOW 10
185 static bool reconfiguring = false;
187 static void add_del_bridges(const struct ovsrec_open_vswitch *);
188 static void bridge_update_ofprotos(void);
189 static void bridge_create(const struct ovsrec_bridge *);
190 static void bridge_destroy(struct bridge *);
191 static struct bridge *bridge_lookup(const char *name);
192 static unixctl_cb_func bridge_unixctl_dump_flows;
193 static unixctl_cb_func bridge_unixctl_reconnect;
194 static size_t bridge_get_controllers(const struct bridge *br,
195 struct ovsrec_controller ***controllersp);
196 static void bridge_add_del_ports(struct bridge *,
197 const unsigned long int *splinter_vlans);
198 static void bridge_refresh_ofp_port(struct bridge *);
199 static void bridge_configure_datapath_id(struct bridge *);
200 static void bridge_configure_flow_eviction_threshold(struct bridge *);
201 static void bridge_configure_netflow(struct bridge *);
202 static void bridge_configure_forward_bpdu(struct bridge *);
203 static void bridge_configure_mac_table(struct bridge *);
204 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
205 static void bridge_configure_ipfix(struct bridge *);
206 static void bridge_configure_stp(struct bridge *);
207 static void bridge_configure_tables(struct bridge *);
208 static void bridge_configure_dp_desc(struct bridge *);
209 static void bridge_configure_remotes(struct bridge *,
210 const struct sockaddr_in *managers,
212 static void bridge_pick_local_hw_addr(struct bridge *,
213 uint8_t ea[ETH_ADDR_LEN],
214 struct iface **hw_addr_iface);
215 static uint64_t bridge_pick_datapath_id(struct bridge *,
216 const uint8_t bridge_ea[ETH_ADDR_LEN],
217 struct iface *hw_addr_iface);
218 static void bridge_queue_if_cfg(struct bridge *,
219 const struct ovsrec_interface *,
220 const struct ovsrec_port *);
221 static uint64_t dpid_from_hash(const void *, size_t nbytes);
222 static bool bridge_has_bond_fake_iface(const struct bridge *,
224 static bool port_is_bond_fake_iface(const struct port *);
226 static unixctl_cb_func qos_unixctl_show;
228 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
229 static void port_del_ifaces(struct port *);
230 static void port_destroy(struct port *);
231 static struct port *port_lookup(const struct bridge *, const char *name);
232 static void port_configure(struct port *);
233 static struct lacp_settings *port_configure_lacp(struct port *,
234 struct lacp_settings *);
235 static void port_configure_bond(struct port *, struct bond_settings *);
236 static bool port_is_synthetic(const struct port *);
238 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
239 static void run_system_stats(void);
241 static void bridge_configure_mirrors(struct bridge *);
242 static struct mirror *mirror_create(struct bridge *,
243 const struct ovsrec_mirror *);
244 static void mirror_destroy(struct mirror *);
245 static bool mirror_configure(struct mirror *);
246 static void mirror_refresh_stats(struct mirror *);
248 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
249 static bool iface_create(struct bridge *, struct if_cfg *, uint16_t ofp_port);
250 static bool iface_is_internal(const struct ovsrec_interface *iface,
251 const struct ovsrec_bridge *br);
252 static const char *iface_get_type(const struct ovsrec_interface *,
253 const struct ovsrec_bridge *);
254 static void iface_destroy(struct iface *);
255 static struct iface *iface_lookup(const struct bridge *, const char *name);
256 static struct iface *iface_find(const char *name);
257 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
258 static struct iface *iface_from_ofp_port(const struct bridge *,
260 static void iface_set_mac(struct iface *);
261 static void iface_set_ofport(const struct ovsrec_interface *, uint16_t ofport);
262 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
263 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
264 static void iface_configure_cfm(struct iface *);
265 static void iface_refresh_cfm_stats(struct iface *);
266 static void iface_refresh_stats(struct iface *);
267 static void iface_refresh_status(struct iface *);
268 static bool iface_is_synthetic(const struct iface *);
269 static uint16_t iface_pick_ofport(const struct ovsrec_interface *);
271 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
273 * This is deprecated. It is only for compatibility with broken device drivers
274 * in old versions of Linux that do not properly support VLANs when VLAN
275 * devices are not used. When broken device drivers are no longer in
276 * widespread use, we will delete these interfaces. */
278 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
279 static bool vlan_splinters_enabled_anywhere;
281 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
282 static unsigned long int *collect_splinter_vlans(
283 const struct ovsrec_open_vswitch *);
284 static void configure_splinter_port(struct port *);
285 static void add_vlan_splinter_ports(struct bridge *,
286 const unsigned long int *splinter_vlans,
287 struct shash *ports);
290 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
292 struct shash iface_hints;
293 static bool initialized = false;
300 shash_init(&iface_hints);
303 for (i = 0; i < cfg->n_bridges; i++) {
304 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
307 for (j = 0; j < br_cfg->n_ports; j++) {
308 struct ovsrec_port *port_cfg = br_cfg->ports[j];
311 for (k = 0; k < port_cfg->n_interfaces; k++) {
312 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
313 struct iface_hint *iface_hint;
315 iface_hint = xmalloc(sizeof *iface_hint);
316 iface_hint->br_name = br_cfg->name;
317 iface_hint->br_type = br_cfg->datapath_type;
318 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
320 shash_add(&iface_hints, if_cfg->name, iface_hint);
326 ofproto_init(&iface_hints);
328 shash_destroy_free_data(&iface_hints);
332 /* Public functions. */
334 /* Initializes the bridge module, configuring it to obtain its configuration
335 * from an OVSDB server accessed over 'remote', which should be a string in a
336 * form acceptable to ovsdb_idl_create(). */
338 bridge_init(const char *remote)
340 /* Create connection to database. */
341 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
342 idl_seqno = ovsdb_idl_get_seqno(idl);
343 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
344 ovsdb_idl_verify_write_only(idl);
346 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
347 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
354 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
355 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
356 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
358 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
359 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
360 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
361 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
380 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
383 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
384 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
385 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
392 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
394 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
405 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
407 /* Register unixctl commands. */
408 unixctl_command_register("qos/show", "interface", 1, 1,
409 qos_unixctl_show, NULL);
410 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
411 bridge_unixctl_dump_flows, NULL);
412 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
413 bridge_unixctl_reconnect, NULL);
423 struct bridge *br, *next_br;
425 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
428 ovsdb_idl_destroy(idl);
431 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
432 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
433 * responsible for freeing '*managersp' (with free()).
435 * You may be asking yourself "why does ovs-vswitchd care?", because
436 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
437 * should not be and in fact is not directly involved in that. But
438 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
439 * it has to tell in-band control where the managers are to enable that.
440 * (Thus, only managers connected in-band are collected.)
443 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
444 struct sockaddr_in **managersp, size_t *n_managersp)
446 struct sockaddr_in *managers = NULL;
447 size_t n_managers = 0;
451 /* Collect all of the potential targets from the "targets" columns of the
452 * rows pointed to by "manager_options", excluding any that are
455 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
456 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
458 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
459 sset_find_and_delete(&targets, m->target);
461 sset_add(&targets, m->target);
465 /* Now extract the targets' IP addresses. */
466 if (!sset_is_empty(&targets)) {
469 managers = xmalloc(sset_count(&targets) * sizeof *managers);
470 SSET_FOR_EACH (target, &targets) {
471 struct sockaddr_in *sin = &managers[n_managers];
473 if (stream_parse_target_with_default_ports(target,
481 sset_destroy(&targets);
483 *managersp = managers;
484 *n_managersp = n_managers;
488 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
490 unsigned long int *splinter_vlans;
493 COVERAGE_INC(bridge_reconfigure);
495 ovs_assert(!reconfiguring);
496 reconfiguring = true;
498 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
499 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
500 * configuration otherwise.
502 * This is mostly an update to bridge data structures. Nothing is pushed
503 * down to ofproto or lower layers. */
504 add_del_bridges(ovs_cfg);
505 splinter_vlans = collect_splinter_vlans(ovs_cfg);
506 HMAP_FOR_EACH (br, node, &all_bridges) {
507 bridge_add_del_ports(br, splinter_vlans);
509 free(splinter_vlans);
511 /* Delete datapaths that are no longer configured, and create ones which
512 * don't exist but should. */
513 bridge_update_ofprotos();
515 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
516 HMAP_FOR_EACH (br, node, &all_bridges) {
517 bridge_refresh_ofp_port(br);
520 /* Clear database records for "if_cfg"s which haven't been instantiated. */
521 HMAP_FOR_EACH (br, node, &all_bridges) {
522 struct if_cfg *if_cfg;
524 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
525 iface_clear_db_record(if_cfg->cfg);
529 reconfigure_system_stats(ovs_cfg);
533 bridge_reconfigure_ofp(void)
535 long long int deadline;
539 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
541 /* The kernel will reject any attempt to add a given port to a datapath if
542 * that port already belongs to a different datapath, so we must do all
543 * port deletions before any port additions. */
544 HMAP_FOR_EACH (br, node, &all_bridges) {
545 struct ofpp_garbage *garbage, *next;
547 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
548 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
549 * internal datastructures may not be consistent. Eventually, when
550 * port additions and deletions are cheaper, these calls should be
553 ofproto_port_del(br->ofproto, garbage->ofp_port);
554 list_remove(&garbage->list_node);
558 if (time_msec() >= deadline) {
565 HMAP_FOR_EACH (br, node, &all_bridges) {
566 struct if_cfg *if_cfg, *next;
568 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
569 iface_create(br, if_cfg, OFPP_NONE);
571 if (time_msec() >= deadline) {
581 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
583 struct sockaddr_in *managers;
584 int sflow_bridge_number;
589 ovs_assert(reconfiguring);
590 done = bridge_reconfigure_ofp();
592 /* Complete the configuration. */
593 sflow_bridge_number = 0;
594 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
595 HMAP_FOR_EACH (br, node, &all_bridges) {
598 /* We need the datapath ID early to allow LACP ports to use it as the
599 * default system ID. */
600 bridge_configure_datapath_id(br);
602 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
605 port_configure(port);
607 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
608 iface_configure_cfm(iface);
609 iface_configure_qos(iface, port->cfg->qos);
610 iface_set_mac(iface);
611 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
615 bridge_configure_mirrors(br);
616 bridge_configure_flow_eviction_threshold(br);
617 bridge_configure_forward_bpdu(br);
618 bridge_configure_mac_table(br);
619 bridge_configure_remotes(br, managers, n_managers);
620 bridge_configure_netflow(br);
621 bridge_configure_sflow(br, &sflow_bridge_number);
622 bridge_configure_ipfix(br);
623 bridge_configure_stp(br);
624 bridge_configure_tables(br);
625 bridge_configure_dp_desc(br);
632 /* Delete ofprotos which aren't configured or have the wrong type. Create
633 * ofprotos which don't exist but need to. */
635 bridge_update_ofprotos(void)
637 struct bridge *br, *next;
642 /* Delete ofprotos with no bridge or with the wrong type. */
645 ofproto_enumerate_types(&types);
646 SSET_FOR_EACH (type, &types) {
649 ofproto_enumerate_names(type, &names);
650 SSET_FOR_EACH (name, &names) {
651 br = bridge_lookup(name);
652 if (!br || strcmp(type, br->type)) {
653 ofproto_delete(name, type);
657 sset_destroy(&names);
658 sset_destroy(&types);
660 /* Add ofprotos for bridges which don't have one yet. */
661 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
669 /* Remove ports from any datapath with the same name as 'br'. If we
670 * don't do this, creating 'br''s ofproto will fail because a port with
671 * the same name as its local port already exists. */
672 HMAP_FOR_EACH (br2, node, &all_bridges) {
673 struct ofproto_port ofproto_port;
679 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
681 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
683 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
686 ofproto_port_destroy(&ofproto_port);
690 error = ofproto_create(br->name, br->type, &br->ofproto);
692 VLOG_ERR("failed to create bridge %s: %s", br->name,
700 port_configure(struct port *port)
702 const struct ovsrec_port *cfg = port->cfg;
703 struct bond_settings bond_settings;
704 struct lacp_settings lacp_settings;
705 struct ofproto_bundle_settings s;
708 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
709 configure_splinter_port(port);
718 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
719 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
720 s.slaves[s.n_slaves++] = iface->ofp_port;
725 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
729 /* Get VLAN trunks. */
732 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
736 if (cfg->vlan_mode) {
737 if (!strcmp(cfg->vlan_mode, "access")) {
738 s.vlan_mode = PORT_VLAN_ACCESS;
739 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
740 s.vlan_mode = PORT_VLAN_TRUNK;
741 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
742 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
743 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
744 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
746 /* This "can't happen" because ovsdb-server should prevent it. */
747 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
748 s.vlan_mode = PORT_VLAN_TRUNK;
752 s.vlan_mode = PORT_VLAN_ACCESS;
754 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
758 s.vlan_mode = PORT_VLAN_TRUNK;
761 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
764 /* Get LACP settings. */
765 s.lacp = port_configure_lacp(port, &lacp_settings);
769 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
770 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
771 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
774 s.lacp_slaves = NULL;
777 /* Get bond settings. */
778 if (s.n_slaves > 1) {
779 s.bond = &bond_settings;
780 port_configure_bond(port, &bond_settings);
783 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
784 netdev_set_miimon_interval(iface->netdev, 0);
789 ofproto_bundle_register(port->bridge->ofproto, port, &s);
797 /* Pick local port hardware address and datapath ID for 'br'. */
799 bridge_configure_datapath_id(struct bridge *br)
801 uint8_t ea[ETH_ADDR_LEN];
803 struct iface *local_iface;
804 struct iface *hw_addr_iface;
807 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
808 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
810 int error = netdev_set_etheraddr(local_iface->netdev, ea);
812 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
813 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
814 "Ethernet address: %s",
815 br->name, strerror(error));
818 memcpy(br->ea, ea, ETH_ADDR_LEN);
820 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
821 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
822 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
823 ofproto_set_datapath_id(br->ofproto, dpid);
826 dpid_string = xasprintf("%016"PRIx64, dpid);
827 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
831 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
834 bridge_get_allowed_versions(struct bridge *br)
836 if (!br->cfg->n_protocols)
839 return ofputil_versions_from_strings(br->cfg->protocols,
840 br->cfg->n_protocols);
843 /* Set NetFlow configuration on 'br'. */
845 bridge_configure_netflow(struct bridge *br)
847 struct ovsrec_netflow *cfg = br->cfg->netflow;
848 struct netflow_options opts;
851 ofproto_set_netflow(br->ofproto, NULL);
855 memset(&opts, 0, sizeof opts);
857 /* Get default NetFlow configuration from datapath.
858 * Apply overrides from 'cfg'. */
859 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
860 if (cfg->engine_type) {
861 opts.engine_type = *cfg->engine_type;
863 if (cfg->engine_id) {
864 opts.engine_id = *cfg->engine_id;
867 /* Configure active timeout interval. */
868 opts.active_timeout = cfg->active_timeout;
869 if (!opts.active_timeout) {
870 opts.active_timeout = -1;
871 } else if (opts.active_timeout < 0) {
872 VLOG_WARN("bridge %s: active timeout interval set to negative "
873 "value, using default instead (%d seconds)", br->name,
874 NF_ACTIVE_TIMEOUT_DEFAULT);
875 opts.active_timeout = -1;
878 /* Add engine ID to interface number to disambiguate bridgs? */
879 opts.add_id_to_iface = cfg->add_id_to_interface;
880 if (opts.add_id_to_iface) {
881 if (opts.engine_id > 0x7f) {
882 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
883 "another vswitch, choose an engine id less than 128",
886 if (hmap_count(&br->ports) > 508) {
887 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
888 "another port when more than 508 ports are used",
894 sset_init(&opts.collectors);
895 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
898 if (ofproto_set_netflow(br->ofproto, &opts)) {
899 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
901 sset_destroy(&opts.collectors);
904 /* Set sFlow configuration on 'br'. */
906 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
908 const struct ovsrec_sflow *cfg = br->cfg->sflow;
909 struct ovsrec_controller **controllers;
910 struct ofproto_sflow_options oso;
911 size_t n_controllers;
915 ofproto_set_sflow(br->ofproto, NULL);
919 memset(&oso, 0, sizeof oso);
921 sset_init(&oso.targets);
922 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
924 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
926 oso.sampling_rate = *cfg->sampling;
929 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
931 oso.polling_interval = *cfg->polling;
934 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
936 oso.header_len = *cfg->header;
939 oso.sub_id = (*sflow_bridge_number)++;
940 oso.agent_device = cfg->agent;
942 oso.control_ip = NULL;
943 n_controllers = bridge_get_controllers(br, &controllers);
944 for (i = 0; i < n_controllers; i++) {
945 if (controllers[i]->local_ip) {
946 oso.control_ip = controllers[i]->local_ip;
950 ofproto_set_sflow(br->ofproto, &oso);
952 sset_destroy(&oso.targets);
955 /* Set IPFIX configuration on 'br'. */
957 bridge_configure_ipfix(struct bridge *br)
959 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
960 const struct ovsrec_flow_sample_collector_set *fe_cfg;
961 struct ofproto_ipfix_bridge_exporter_options be_opts;
962 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
963 size_t n_fe_opts = 0;
965 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
966 if (fe_cfg->bridge == br->cfg) {
971 if (!be_cfg && n_fe_opts == 0) {
972 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
977 memset(&be_opts, 0, sizeof be_opts);
979 sset_init(&be_opts.targets);
980 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
982 if (be_cfg->sampling) {
983 be_opts.sampling_rate = *be_cfg->sampling;
985 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
987 if (be_cfg->obs_domain_id) {
988 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
990 if (be_cfg->obs_point_id) {
991 be_opts.obs_point_id = *be_cfg->obs_point_id;
996 struct ofproto_ipfix_flow_exporter_options *opts;
997 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
999 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1000 if (fe_cfg->bridge == br->cfg) {
1001 opts->collector_set_id = fe_cfg->id;
1002 sset_init(&opts->targets);
1003 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1004 fe_cfg->ipfix->n_targets);
1010 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1014 sset_destroy(&be_opts.targets);
1017 if (n_fe_opts > 0) {
1018 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1020 for (i = 0; i < n_fe_opts; i++) {
1021 sset_destroy(&opts->targets);
1029 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1030 struct ofproto_port_stp_settings *port_s,
1031 int *port_num_counter, unsigned long *port_num_bitmap)
1033 const char *config_str;
1034 struct iface *iface;
1036 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1037 port_s->enable = false;
1040 port_s->enable = true;
1043 /* STP over bonds is not supported. */
1044 if (!list_is_singleton(&port->ifaces)) {
1045 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1047 port_s->enable = false;
1051 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1053 /* Internal ports shouldn't participate in spanning tree, so
1055 if (!strcmp(iface->type, "internal")) {
1056 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1057 port_s->enable = false;
1061 /* STP on mirror output ports is not supported. */
1062 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1063 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1064 port_s->enable = false;
1068 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1070 unsigned long int port_num = strtoul(config_str, NULL, 0);
1071 int port_idx = port_num - 1;
1073 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1074 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1075 port_s->enable = false;
1079 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1080 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1081 port->name, port_num);
1082 port_s->enable = false;
1085 bitmap_set1(port_num_bitmap, port_idx);
1086 port_s->port_num = port_idx;
1088 if (*port_num_counter >= STP_MAX_PORTS) {
1089 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1090 port_s->enable = false;
1094 port_s->port_num = (*port_num_counter)++;
1097 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1099 port_s->path_cost = strtoul(config_str, NULL, 10);
1101 enum netdev_features current;
1104 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1105 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1106 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1109 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1111 port_s->priority = strtoul(config_str, NULL, 0);
1113 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1117 /* Set spanning tree configuration on 'br'. */
1119 bridge_configure_stp(struct bridge *br)
1121 if (!br->cfg->stp_enable) {
1122 ofproto_set_stp(br->ofproto, NULL);
1124 struct ofproto_stp_settings br_s;
1125 const char *config_str;
1127 int port_num_counter;
1128 unsigned long *port_num_bitmap;
1130 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1132 uint8_t ea[ETH_ADDR_LEN];
1134 if (eth_addr_from_string(config_str, ea)) {
1135 br_s.system_id = eth_addr_to_uint64(ea);
1137 br_s.system_id = eth_addr_to_uint64(br->ea);
1138 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1139 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1142 br_s.system_id = eth_addr_to_uint64(br->ea);
1145 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1147 br_s.priority = strtoul(config_str, NULL, 0);
1149 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1152 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1154 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1156 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1159 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1161 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1163 br_s.max_age = STP_DEFAULT_MAX_AGE;
1166 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1168 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1170 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1173 /* Configure STP on the bridge. */
1174 if (ofproto_set_stp(br->ofproto, &br_s)) {
1175 VLOG_ERR("bridge %s: could not enable STP", br->name);
1179 /* Users must either set the port number with the "stp-port-num"
1180 * configuration on all ports or none. If manual configuration
1181 * is not done, then we allocate them sequentially. */
1182 port_num_counter = 0;
1183 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1184 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1185 struct ofproto_port_stp_settings port_s;
1186 struct iface *iface;
1188 port_configure_stp(br->ofproto, port, &port_s,
1189 &port_num_counter, port_num_bitmap);
1191 /* As bonds are not supported, just apply configuration to
1192 * all interfaces. */
1193 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1194 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1196 VLOG_ERR("port %s: could not enable STP", port->name);
1202 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1203 && port_num_counter) {
1204 VLOG_ERR("bridge %s: must manually configure all STP port "
1205 "IDs or none, disabling", br->name);
1206 ofproto_set_stp(br->ofproto, NULL);
1208 bitmap_free(port_num_bitmap);
1213 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1215 const struct port *port = port_lookup(br, name);
1216 return port && port_is_bond_fake_iface(port);
1220 port_is_bond_fake_iface(const struct port *port)
1222 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1226 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1228 struct bridge *br, *next;
1229 struct shash new_br;
1232 /* Collect new bridges' names and types. */
1233 shash_init(&new_br);
1234 for (i = 0; i < cfg->n_bridges; i++) {
1235 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1236 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1238 if (strchr(br_cfg->name, '/')) {
1239 /* Prevent remote ovsdb-server users from accessing arbitrary
1240 * directories, e.g. consider a bridge named "../../../etc/". */
1241 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1243 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1244 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1248 /* Get rid of deleted bridges or those whose types have changed.
1249 * Update 'cfg' of bridges that still exist. */
1250 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1251 br->cfg = shash_find_data(&new_br, br->name);
1252 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1253 br->cfg->datapath_type))) {
1258 /* Add new bridges. */
1259 for (i = 0; i < cfg->n_bridges; i++) {
1260 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1261 struct bridge *br = bridge_lookup(br_cfg->name);
1263 bridge_create(br_cfg);
1267 shash_destroy(&new_br);
1271 iface_set_ofp_port(struct iface *iface, uint16_t ofp_port)
1273 struct bridge *br = iface->port->bridge;
1275 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1276 iface->ofp_port = ofp_port;
1277 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1278 iface_set_ofport(iface->cfg, ofp_port);
1281 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1282 * Returns 0 if successful, otherwise a positive errno value. */
1284 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1285 struct netdev *netdev)
1289 error = netdev_set_config(netdev, &iface_cfg->options);
1291 VLOG_WARN("could not configure network device %s (%s)",
1292 iface_cfg->name, strerror(error));
1297 /* This function determines whether 'ofproto_port', which is attached to
1298 * br->ofproto's datapath, is one that we want in 'br'.
1300 * If it is, it returns true, after creating an iface (if necessary),
1301 * configuring the iface's netdev according to the iface's options, and setting
1302 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1304 * If, on the other hand, 'port' should be removed, it returns false. The
1305 * caller should later detach the port from br->ofproto. */
1307 bridge_refresh_one_ofp_port(struct bridge *br,
1308 const struct ofproto_port *ofproto_port)
1310 const char *name = ofproto_port->name;
1311 const char *type = ofproto_port->type;
1312 uint16_t ofp_port = ofproto_port->ofp_port;
1314 struct iface *iface = iface_lookup(br, name);
1316 /* Check that the name-to-number mapping is one-to-one. */
1317 if (iface->ofp_port != OFPP_NONE) {
1318 VLOG_WARN("bridge %s: interface %s reported twice",
1321 } else if (iface_from_ofp_port(br, ofp_port)) {
1322 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1323 br->name, ofp_port);
1327 /* There's a configured interface named 'name'. */
1328 if (strcmp(type, iface->type)
1329 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1330 /* It's the wrong type, or it's the right type but can't be
1331 * configured as the user requested, so we must destroy it. */
1334 /* It's the right type and configured correctly. Keep it. */
1335 iface_set_ofp_port(iface, ofp_port);
1338 } else if (bridge_has_bond_fake_iface(br, name)
1339 && !strcmp(type, "internal")) {
1340 /* It's a bond fake iface. Keep it. */
1343 /* There's no configured interface named 'name', but there might be an
1344 * interface of that name queued to be created.
1346 * If there is, and it has the correct type, then try to configure it
1347 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1348 * delete it and later try to re-add it. */
1349 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1351 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1352 && iface_create(br, if_cfg, ofp_port));
1356 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1357 * consistent with the ofp_ports in "br->ofproto". */
1359 bridge_refresh_ofp_port(struct bridge *br)
1361 struct ofproto_port_dump dump;
1362 struct ofproto_port ofproto_port;
1363 struct port *port, *port_next;
1365 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1366 hmap_clear(&br->ifaces);
1367 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1368 struct iface *iface;
1370 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1371 iface->ofp_port = OFPP_NONE;
1375 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1376 * already exist in the datapath and promote them to full fledged "struct
1377 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1378 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1379 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1380 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1381 garbage->ofp_port = ofproto_port.ofp_port;
1382 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1386 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1387 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1388 * later they can be added to ofproto. */
1389 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1390 struct iface *iface, *iface_next;
1392 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1393 if (iface->ofp_port == OFPP_NONE) {
1394 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1395 iface_destroy(iface);
1399 if (list_is_empty(&port->ifaces)) {
1405 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1406 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1407 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1410 * If successful, returns 0 and stores the network device in '*netdevp'. On
1411 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1413 iface_do_create(const struct bridge *br,
1414 const struct if_cfg *if_cfg,
1415 uint16_t *ofp_portp, struct netdev **netdevp)
1417 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1418 const struct ovsrec_port *port_cfg = if_cfg->parent;
1419 struct netdev *netdev = NULL;
1422 if (netdev_is_reserved_name(iface_cfg->name)) {
1423 VLOG_WARN("could not create interface %s, name is reserved",
1429 error = netdev_open(iface_cfg->name,
1430 iface_get_type(iface_cfg, br->cfg), &netdev);
1432 VLOG_WARN("could not open network device %s (%s)",
1433 iface_cfg->name, strerror(error));
1437 error = iface_set_netdev_config(iface_cfg, netdev);
1442 if (*ofp_portp == OFPP_NONE) {
1443 uint16_t ofp_port = if_cfg->ofport;
1445 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1449 *ofp_portp = ofp_port;
1451 VLOG_INFO("bridge %s: added interface %s on port %d",
1452 br->name, iface_cfg->name, *ofp_portp);
1454 VLOG_DBG("bridge %s: interface %s is on port %d",
1455 br->name, iface_cfg->name, *ofp_portp);
1458 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1459 || iface_is_internal(iface_cfg, br->cfg)) {
1460 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1468 netdev_close(netdev);
1472 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1473 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1474 * automatically allocated for the iface. Takes ownership of and
1475 * deallocates 'if_cfg'.
1477 * Return true if an iface is successfully created, false otherwise. */
1479 iface_create(struct bridge *br, struct if_cfg *if_cfg, uint16_t ofp_port)
1481 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1482 const struct ovsrec_port *port_cfg = if_cfg->parent;
1484 struct netdev *netdev;
1485 struct iface *iface;
1490 /* Do the bits that can fail up front.
1492 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1493 * internal datastructures may not be consistent. Eventually, when port
1494 * additions and deletions are cheaper, these calls should be removed. */
1496 ovs_assert(!iface_lookup(br, iface_cfg->name));
1497 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1500 iface_set_ofport(iface_cfg, OFPP_NONE);
1501 iface_clear_db_record(iface_cfg);
1506 /* Get or create the port structure. */
1507 port = port_lookup(br, port_cfg->name);
1509 port = port_create(br, port_cfg);
1512 /* Create the iface structure. */
1513 iface = xzalloc(sizeof *iface);
1514 list_push_back(&port->ifaces, &iface->port_elem);
1515 hmap_insert(&br->iface_by_name, &iface->name_node,
1516 hash_string(iface_cfg->name, 0));
1518 iface->name = xstrdup(iface_cfg->name);
1519 iface->ofp_port = OFPP_NONE;
1520 iface->netdev = netdev;
1521 iface->type = iface_get_type(iface_cfg, br->cfg);
1522 iface->cfg = iface_cfg;
1524 iface_set_ofp_port(iface, ofp_port);
1526 /* Populate initial status in database. */
1527 iface_refresh_stats(iface);
1528 iface_refresh_status(iface);
1530 /* Add bond fake iface if necessary. */
1531 if (port_is_bond_fake_iface(port)) {
1532 struct ofproto_port ofproto_port;
1534 if (ofproto_port_query_by_name(br->ofproto, port->name,
1536 struct netdev *netdev;
1539 error = netdev_open(port->name, "internal", &netdev);
1541 uint16_t fake_ofp_port = if_cfg->ofport;
1543 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1544 netdev_close(netdev);
1546 VLOG_WARN("could not open network device %s (%s)",
1547 port->name, strerror(error));
1550 /* Already exists, nothing to do. */
1551 ofproto_port_destroy(&ofproto_port);
1556 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1562 /* Set Flow eviction threshold */
1564 bridge_configure_flow_eviction_threshold(struct bridge *br)
1566 const char *threshold_str;
1569 threshold_str = smap_get(&br->cfg->other_config,
1570 "flow-eviction-threshold");
1571 if (threshold_str) {
1572 threshold = strtoul(threshold_str, NULL, 10);
1574 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1576 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1579 /* Set forward BPDU option. */
1581 bridge_configure_forward_bpdu(struct bridge *br)
1583 ofproto_set_forward_bpdu(br->ofproto,
1584 smap_get_bool(&br->cfg->other_config,
1589 /* Set MAC learning table configuration for 'br'. */
1591 bridge_configure_mac_table(struct bridge *br)
1593 const char *idle_time_str;
1596 const char *mac_table_size_str;
1599 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1600 idle_time = (idle_time_str && atoi(idle_time_str)
1601 ? atoi(idle_time_str)
1602 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1604 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1605 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1606 ? atoi(mac_table_size_str)
1609 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1613 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1614 struct iface **hw_addr_iface)
1616 struct hmapx mirror_output_ports;
1619 bool found_addr = false;
1623 *hw_addr_iface = NULL;
1625 /* Did the user request a particular MAC? */
1626 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1627 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1628 if (eth_addr_is_multicast(ea)) {
1629 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1630 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1631 } else if (eth_addr_is_zero(ea)) {
1632 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1638 /* Mirror output ports don't participate in picking the local hardware
1639 * address. ofproto can't help us find out whether a given port is a
1640 * mirror output because we haven't configured mirrors yet, so we need to
1641 * accumulate them ourselves. */
1642 hmapx_init(&mirror_output_ports);
1643 for (i = 0; i < br->cfg->n_mirrors; i++) {
1644 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1645 if (m->output_port) {
1646 hmapx_add(&mirror_output_ports, m->output_port);
1650 /* Otherwise choose the minimum non-local MAC address among all of the
1652 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1653 uint8_t iface_ea[ETH_ADDR_LEN];
1654 struct iface *candidate;
1655 struct iface *iface;
1657 /* Mirror output ports don't participate. */
1658 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1662 /* Choose the MAC address to represent the port. */
1664 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1665 /* Find the interface with this Ethernet address (if any) so that
1666 * we can provide the correct devname to the caller. */
1667 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1668 uint8_t candidate_ea[ETH_ADDR_LEN];
1669 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1670 && eth_addr_equals(iface_ea, candidate_ea)) {
1675 /* Choose the interface whose MAC address will represent the port.
1676 * The Linux kernel bonding code always chooses the MAC address of
1677 * the first slave added to a bond, and the Fedora networking
1678 * scripts always add slaves to a bond in alphabetical order, so
1679 * for compatibility we choose the interface with the name that is
1680 * first in alphabetical order. */
1681 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1682 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1687 /* The local port doesn't count (since we're trying to choose its
1688 * MAC address anyway). */
1689 if (iface->ofp_port == OFPP_LOCAL) {
1694 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1700 /* Compare against our current choice. */
1701 if (!eth_addr_is_multicast(iface_ea) &&
1702 !eth_addr_is_local(iface_ea) &&
1703 !eth_addr_is_reserved(iface_ea) &&
1704 !eth_addr_is_zero(iface_ea) &&
1705 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1707 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1708 *hw_addr_iface = iface;
1714 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1715 *hw_addr_iface = NULL;
1718 hmapx_destroy(&mirror_output_ports);
1721 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1722 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1723 * an interface on 'br', then that interface must be passed in as
1724 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1725 * 'hw_addr_iface' must be passed in as a null pointer. */
1727 bridge_pick_datapath_id(struct bridge *br,
1728 const uint8_t bridge_ea[ETH_ADDR_LEN],
1729 struct iface *hw_addr_iface)
1732 * The procedure for choosing a bridge MAC address will, in the most
1733 * ordinary case, also choose a unique MAC that we can use as a datapath
1734 * ID. In some special cases, though, multiple bridges will end up with
1735 * the same MAC address. This is OK for the bridges, but it will confuse
1736 * the OpenFlow controller, because each datapath needs a unique datapath
1739 * Datapath IDs must be unique. It is also very desirable that they be
1740 * stable from one run to the next, so that policy set on a datapath
1743 const char *datapath_id;
1746 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1747 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1751 if (!hw_addr_iface) {
1753 * A purely internal bridge, that is, one that has no non-virtual
1754 * network devices on it at all, is difficult because it has no
1755 * natural unique identifier at all.
1757 * When the host is a XenServer, we handle this case by hashing the
1758 * host's UUID with the name of the bridge. Names of bridges are
1759 * persistent across XenServer reboots, although they can be reused if
1760 * an internal network is destroyed and then a new one is later
1761 * created, so this is fairly effective.
1763 * When the host is not a XenServer, we punt by using a random MAC
1764 * address on each run.
1766 const char *host_uuid = xenserver_get_host_uuid();
1768 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1769 dpid = dpid_from_hash(combined, strlen(combined));
1775 return eth_addr_to_uint64(bridge_ea);
1779 dpid_from_hash(const void *data, size_t n)
1781 uint8_t hash[SHA1_DIGEST_SIZE];
1783 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1784 sha1_bytes(data, n, hash);
1785 eth_addr_mark_random(hash);
1786 return eth_addr_to_uint64(hash);
1790 iface_refresh_status(struct iface *iface)
1794 enum netdev_features current;
1798 uint8_t mac[ETH_ADDR_LEN];
1801 if (iface_is_synthetic(iface)) {
1807 if (!netdev_get_status(iface->netdev, &smap)) {
1808 ovsrec_interface_set_status(iface->cfg, &smap);
1810 ovsrec_interface_set_status(iface->cfg, NULL);
1813 smap_destroy(&smap);
1815 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1816 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1818 ovsrec_interface_set_duplex(iface->cfg,
1819 netdev_features_is_full_duplex(current)
1821 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1823 ovsrec_interface_set_duplex(iface->cfg, NULL);
1824 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1827 error = netdev_get_mtu(iface->netdev, &mtu);
1830 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1832 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1835 error = netdev_get_etheraddr(iface->netdev, mac);
1837 char mac_string[32];
1839 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1840 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1842 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1846 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1849 iface_refresh_cfm_stats(struct iface *iface)
1851 const struct ovsrec_interface *cfg = iface->cfg;
1852 struct ofproto_cfm_status status;
1854 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1855 iface->ofp_port, &status)) {
1856 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1857 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1858 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1859 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1860 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1862 const char *reasons[CFM_FAULT_N_REASONS];
1863 int64_t cfm_health = status.health;
1864 bool faulted = status.faults != 0;
1867 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1870 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1871 int reason = 1 << i;
1872 if (status.faults & reason) {
1873 reasons[j++] = cfm_fault_reason_to_str(reason);
1876 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1878 if (status.remote_opstate >= 0) {
1879 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1880 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1882 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1885 ovsrec_interface_set_cfm_remote_mpids(cfg,
1886 (const int64_t *)status.rmps,
1888 if (cfm_health >= 0) {
1889 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1891 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1897 iface_refresh_stats(struct iface *iface)
1899 #define IFACE_STATS \
1900 IFACE_STAT(rx_packets, "rx_packets") \
1901 IFACE_STAT(tx_packets, "tx_packets") \
1902 IFACE_STAT(rx_bytes, "rx_bytes") \
1903 IFACE_STAT(tx_bytes, "tx_bytes") \
1904 IFACE_STAT(rx_dropped, "rx_dropped") \
1905 IFACE_STAT(tx_dropped, "tx_dropped") \
1906 IFACE_STAT(rx_errors, "rx_errors") \
1907 IFACE_STAT(tx_errors, "tx_errors") \
1908 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1909 IFACE_STAT(rx_over_errors, "rx_over_err") \
1910 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1911 IFACE_STAT(collisions, "collisions")
1913 #define IFACE_STAT(MEMBER, NAME) + 1
1914 enum { N_IFACE_STATS = IFACE_STATS };
1916 int64_t values[N_IFACE_STATS];
1917 char *keys[N_IFACE_STATS];
1920 struct netdev_stats stats;
1922 if (iface_is_synthetic(iface)) {
1926 /* Intentionally ignore return value, since errors will set 'stats' to
1927 * all-1s, and we will deal with that correctly below. */
1928 netdev_get_stats(iface->netdev, &stats);
1930 /* Copy statistics into keys[] and values[]. */
1932 #define IFACE_STAT(MEMBER, NAME) \
1933 if (stats.MEMBER != UINT64_MAX) { \
1935 values[n] = stats.MEMBER; \
1940 ovs_assert(n <= N_IFACE_STATS);
1942 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1947 br_refresh_stp_status(struct bridge *br)
1949 struct smap smap = SMAP_INITIALIZER(&smap);
1950 struct ofproto *ofproto = br->ofproto;
1951 struct ofproto_stp_status status;
1953 if (ofproto_get_stp_status(ofproto, &status)) {
1957 if (!status.enabled) {
1958 ovsrec_bridge_set_status(br->cfg, NULL);
1962 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1963 STP_ID_ARGS(status.bridge_id));
1964 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1965 STP_ID_ARGS(status.designated_root));
1966 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1968 ovsrec_bridge_set_status(br->cfg, &smap);
1969 smap_destroy(&smap);
1973 port_refresh_stp_status(struct port *port)
1975 struct ofproto *ofproto = port->bridge->ofproto;
1976 struct iface *iface;
1977 struct ofproto_port_stp_status status;
1979 int64_t int_values[3];
1982 if (port_is_synthetic(port)) {
1986 /* STP doesn't currently support bonds. */
1987 if (!list_is_singleton(&port->ifaces)) {
1988 ovsrec_port_set_status(port->cfg, NULL);
1992 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1994 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1998 if (!status.enabled) {
1999 ovsrec_port_set_status(port->cfg, NULL);
2000 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2004 /* Set Status column. */
2006 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2007 smap_add(&smap, "stp_state", stp_state_name(status.state));
2008 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2009 smap_add(&smap, "stp_role", stp_role_name(status.role));
2010 ovsrec_port_set_status(port->cfg, &smap);
2011 smap_destroy(&smap);
2013 /* Set Statistics column. */
2014 keys[0] = "stp_tx_count";
2015 int_values[0] = status.tx_count;
2016 keys[1] = "stp_rx_count";
2017 int_values[1] = status.rx_count;
2018 keys[2] = "stp_error_count";
2019 int_values[2] = status.error_count;
2021 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2022 ARRAY_SIZE(int_values));
2026 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2028 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2032 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2034 bool enable = enable_system_stats(cfg);
2036 system_stats_enable(enable);
2038 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2043 run_system_stats(void)
2045 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2048 stats = system_stats_run();
2050 struct ovsdb_idl_txn *txn;
2051 struct ovsdb_datum datum;
2053 txn = ovsdb_idl_txn_create(idl);
2054 ovsdb_datum_from_smap(&datum, stats);
2055 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2057 ovsdb_idl_txn_commit(txn);
2058 ovsdb_idl_txn_destroy(txn);
2064 static inline const char *
2065 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2068 case OFPCR12_ROLE_EQUAL:
2070 case OFPCR12_ROLE_MASTER:
2072 case OFPCR12_ROLE_SLAVE:
2074 case OFPCR12_ROLE_NOCHANGE:
2076 return "*** INVALID ROLE ***";
2081 refresh_controller_status(void)
2085 const struct ovsrec_controller *cfg;
2089 /* Accumulate status for controllers on all bridges. */
2090 HMAP_FOR_EACH (br, node, &all_bridges) {
2091 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2094 /* Update each controller in the database with current status. */
2095 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2096 struct ofproto_controller_info *cinfo =
2097 shash_find_data(&info, cfg->target);
2100 struct smap smap = SMAP_INITIALIZER(&smap);
2101 const char **values = cinfo->pairs.values;
2102 const char **keys = cinfo->pairs.keys;
2105 for (i = 0; i < cinfo->pairs.n; i++) {
2106 smap_add(&smap, keys[i], values[i]);
2109 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2110 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2112 ovsrec_controller_set_status(cfg, &smap);
2113 smap_destroy(&smap);
2115 ovsrec_controller_set_is_connected(cfg, false);
2116 ovsrec_controller_set_role(cfg, NULL);
2117 ovsrec_controller_set_status(cfg, NULL);
2121 ofproto_free_ofproto_controller_info(&info);
2126 * Some information in the database must be kept as up-to-date as possible to
2127 * allow controllers to respond rapidly to network outages. We call these
2128 * statistics "instant" stats.
2130 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2131 * assuming that they've changed. The only means we have to determine whether
2132 * they have changed are:
2134 * - Try to commit changes to the database. If nothing changed, then
2135 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2138 * - instant_stats_run() is called late in the run loop, after anything that
2139 * might change any of the instant stats.
2141 * We use these two facts together to avoid waking the process up every
2142 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2145 /* Minimum interval between writing updates to the instant stats to the
2147 #define INSTANT_INTERVAL_MSEC 100
2149 /* Current instant stats database transaction, NULL if there is no ongoing
2151 static struct ovsdb_idl_txn *instant_txn;
2153 /* Next time (in msec on monotonic clock) at which we will update the instant
2155 static long long int instant_next_txn = LLONG_MIN;
2157 /* True if the run loop has run since we last saw that the instant stats were
2158 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2159 * to refresh the instant stats. */
2160 static bool instant_stats_could_have_changed;
2163 instant_stats_run(void)
2165 enum ovsdb_idl_txn_status status;
2167 instant_stats_could_have_changed = true;
2172 if (time_msec() < instant_next_txn) {
2175 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2177 instant_txn = ovsdb_idl_txn_create(idl);
2178 HMAP_FOR_EACH (br, node, &all_bridges) {
2179 struct iface *iface;
2182 br_refresh_stp_status(br);
2184 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2185 port_refresh_stp_status(port);
2188 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2189 enum netdev_flags flags;
2191 const char *link_state;
2192 int64_t link_resets;
2195 if (iface_is_synthetic(iface)) {
2199 current = ofproto_port_is_lacp_current(br->ofproto,
2203 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2205 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2208 error = netdev_get_flags(iface->netdev, &flags);
2210 const char *state = flags & NETDEV_UP ? "up" : "down";
2211 ovsrec_interface_set_admin_state(iface->cfg, state);
2213 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2216 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2217 ovsrec_interface_set_link_state(iface->cfg, link_state);
2219 link_resets = netdev_get_carrier_resets(iface->netdev);
2220 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2222 iface_refresh_cfm_stats(iface);
2225 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2227 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2228 smap_destroy(&smap);
2234 status = ovsdb_idl_txn_commit(instant_txn);
2235 if (status != TXN_INCOMPLETE) {
2236 ovsdb_idl_txn_destroy(instant_txn);
2239 if (status == TXN_UNCHANGED) {
2240 instant_stats_could_have_changed = false;
2245 instant_stats_wait(void)
2248 ovsdb_idl_txn_wait(instant_txn);
2249 } else if (instant_stats_could_have_changed) {
2250 poll_timer_wait_until(instant_next_txn);
2254 /* Performs periodic activity required by bridges that needs to be done with
2255 * the least possible latency.
2257 * It makes sense to call this function a couple of times per poll loop, to
2258 * provide a significant performance boost on some benchmarks with ofprotos
2259 * that use the ofproto-dpif implementation. */
2261 bridge_run_fast(void)
2268 ofproto_enumerate_types(&types);
2269 SSET_FOR_EACH (type, &types) {
2270 ofproto_type_run_fast(type);
2272 sset_destroy(&types);
2274 HMAP_FOR_EACH (br, node, &all_bridges) {
2275 ofproto_run_fast(br->ofproto);
2282 static struct ovsrec_open_vswitch null_cfg;
2283 const struct ovsrec_open_vswitch *cfg;
2284 struct ovsdb_idl_txn *reconf_txn = NULL;
2288 bool vlan_splinters_changed;
2291 ovsrec_open_vswitch_init(&null_cfg);
2293 /* (Re)configure if necessary. */
2294 if (!reconfiguring) {
2297 if (ovsdb_idl_is_lock_contended(idl)) {
2298 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2299 struct bridge *br, *next_br;
2301 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2302 "disabling this process (pid %ld) until it goes away",
2303 (long int) getpid());
2305 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2308 /* Since we will not be running system_stats_run() in this process
2309 * with the current situation of multiple ovs-vswitchd daemons,
2310 * disable system stats collection. */
2311 system_stats_enable(false);
2313 } else if (!ovsdb_idl_has_lock(idl)) {
2317 cfg = ovsrec_open_vswitch_first(idl);
2319 /* Initialize the ofproto library. This only needs to run once, but
2320 * it must be done after the configuration is set. If the
2321 * initialization has already occurred, bridge_init_ofproto()
2322 * returns immediately. */
2323 bridge_init_ofproto(cfg);
2325 /* Once the value of flow-restore-wait is false, we no longer should
2326 * check its value from the database. */
2327 if (cfg && ofproto_get_flow_restore_wait()) {
2328 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2329 "flow-restore-wait", false));
2332 /* Let each datapath type do the work that it needs to do. */
2334 ofproto_enumerate_types(&types);
2335 SSET_FOR_EACH (type, &types) {
2336 ofproto_type_run(type);
2338 sset_destroy(&types);
2340 /* Let each bridge do the work that it needs to do. */
2341 HMAP_FOR_EACH (br, node, &all_bridges) {
2342 ofproto_run(br->ofproto);
2345 /* Re-configure SSL. We do this on every trip through the main loop,
2346 * instead of just when the database changes, because the contents of the
2347 * key and certificate files can change without the database changing.
2349 * We do this before bridge_reconfigure() because that function might
2350 * initiate SSL connections and thus requires SSL to be configured. */
2351 if (cfg && cfg->ssl) {
2352 const struct ovsrec_ssl *ssl = cfg->ssl;
2354 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2355 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2358 if (!reconfiguring) {
2359 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2360 * usage has changed. */
2361 vlan_splinters_changed = false;
2362 if (vlan_splinters_enabled_anywhere) {
2363 HMAP_FOR_EACH (br, node, &all_bridges) {
2364 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2365 vlan_splinters_changed = true;
2371 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2372 idl_seqno = ovsdb_idl_get_seqno(idl);
2374 reconf_txn = ovsdb_idl_txn_create(idl);
2375 bridge_reconfigure(cfg);
2377 /* We still need to reconfigure to avoid dangling pointers to
2378 * now-destroyed ovsrec structures inside bridge data. */
2379 bridge_reconfigure(&null_cfg);
2384 if (reconfiguring) {
2386 reconf_txn = ovsdb_idl_txn_create(idl);
2389 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2390 reconfiguring = false;
2393 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2396 /* If we are completing our initial configuration for this run
2397 * of ovs-vswitchd, then keep the transaction around to monitor
2398 * it for completion. */
2399 if (!initial_config_done) {
2400 initial_config_done = true;
2401 daemonize_txn = reconf_txn;
2408 ovsdb_idl_txn_commit(reconf_txn);
2409 ovsdb_idl_txn_destroy(reconf_txn);
2413 if (daemonize_txn) {
2414 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2415 if (status != TXN_INCOMPLETE) {
2416 ovsdb_idl_txn_destroy(daemonize_txn);
2417 daemonize_txn = NULL;
2419 /* ovs-vswitchd has completed initialization, so allow the
2420 * process that forked us to exit successfully. */
2421 daemonize_complete();
2423 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2427 /* Refresh interface and mirror stats if necessary. */
2428 if (time_msec() >= iface_stats_timer) {
2430 struct ovsdb_idl_txn *txn;
2432 txn = ovsdb_idl_txn_create(idl);
2433 HMAP_FOR_EACH (br, node, &all_bridges) {
2437 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2438 struct iface *iface;
2440 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2441 iface_refresh_stats(iface);
2442 iface_refresh_status(iface);
2446 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2447 mirror_refresh_stats(m);
2451 refresh_controller_status();
2452 ovsdb_idl_txn_commit(txn);
2453 ovsdb_idl_txn_destroy(txn); /* XXX */
2456 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2460 instant_stats_run();
2469 ovsdb_idl_wait(idl);
2470 if (daemonize_txn) {
2471 ovsdb_idl_txn_wait(daemonize_txn);
2474 if (reconfiguring) {
2475 poll_immediate_wake();
2479 ofproto_enumerate_types(&types);
2480 SSET_FOR_EACH (type, &types) {
2481 ofproto_type_wait(type);
2483 sset_destroy(&types);
2485 if (!hmap_is_empty(&all_bridges)) {
2488 HMAP_FOR_EACH (br, node, &all_bridges) {
2489 ofproto_wait(br->ofproto);
2491 poll_timer_wait_until(iface_stats_timer);
2494 system_stats_wait();
2495 instant_stats_wait();
2498 /* Adds some memory usage statistics for bridges into 'usage', for use with
2499 * memory_report(). */
2501 bridge_get_memory_usage(struct simap *usage)
2505 HMAP_FOR_EACH (br, node, &all_bridges) {
2506 ofproto_get_memory_usage(br->ofproto, usage);
2510 /* QoS unixctl user interface functions. */
2512 struct qos_unixctl_show_cbdata {
2514 struct iface *iface;
2518 qos_unixctl_show_cb(unsigned int queue_id,
2519 const struct smap *details,
2522 struct qos_unixctl_show_cbdata *data = aux;
2523 struct ds *ds = data->ds;
2524 struct iface *iface = data->iface;
2525 struct netdev_queue_stats stats;
2526 struct smap_node *node;
2529 ds_put_cstr(ds, "\n");
2531 ds_put_format(ds, "Queue %u:\n", queue_id);
2533 ds_put_cstr(ds, "Default:\n");
2536 SMAP_FOR_EACH (node, details) {
2537 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2540 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2542 if (stats.tx_packets != UINT64_MAX) {
2543 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2546 if (stats.tx_bytes != UINT64_MAX) {
2547 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2550 if (stats.tx_errors != UINT64_MAX) {
2551 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2554 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2555 queue_id, strerror(error));
2560 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2561 const char *argv[], void *aux OVS_UNUSED)
2563 struct ds ds = DS_EMPTY_INITIALIZER;
2564 struct smap smap = SMAP_INITIALIZER(&smap);
2565 struct iface *iface;
2567 struct smap_node *node;
2568 struct qos_unixctl_show_cbdata data;
2571 iface = iface_find(argv[1]);
2573 unixctl_command_reply_error(conn, "no such interface");
2577 netdev_get_qos(iface->netdev, &type, &smap);
2579 if (*type != '\0') {
2580 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2582 SMAP_FOR_EACH (node, &smap) {
2583 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2588 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2591 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2593 unixctl_command_reply(conn, ds_cstr(&ds));
2595 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2596 unixctl_command_reply_error(conn, ds_cstr(&ds));
2599 smap_destroy(&smap);
2603 /* Bridge reconfiguration functions. */
2605 bridge_create(const struct ovsrec_bridge *br_cfg)
2609 ovs_assert(!bridge_lookup(br_cfg->name));
2610 br = xzalloc(sizeof *br);
2612 br->name = xstrdup(br_cfg->name);
2613 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2616 /* Derive the default Ethernet address from the bridge's UUID. This should
2617 * be unique and it will be stable between ovs-vswitchd runs. */
2618 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2619 eth_addr_mark_random(br->default_ea);
2621 hmap_init(&br->ports);
2622 hmap_init(&br->ifaces);
2623 hmap_init(&br->iface_by_name);
2624 hmap_init(&br->mirrors);
2626 hmap_init(&br->if_cfg_todo);
2627 list_init(&br->ofpp_garbage);
2629 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2633 bridge_destroy(struct bridge *br)
2636 struct mirror *mirror, *next_mirror;
2637 struct port *port, *next_port;
2638 struct if_cfg *if_cfg, *next_if_cfg;
2639 struct ofpp_garbage *garbage, *next_garbage;
2641 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2644 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2645 mirror_destroy(mirror);
2647 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2648 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2651 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2652 &br->ofpp_garbage) {
2653 list_remove(&garbage->list_node);
2657 hmap_remove(&all_bridges, &br->node);
2658 ofproto_destroy(br->ofproto);
2659 hmap_destroy(&br->ifaces);
2660 hmap_destroy(&br->ports);
2661 hmap_destroy(&br->iface_by_name);
2662 hmap_destroy(&br->mirrors);
2663 hmap_destroy(&br->if_cfg_todo);
2670 static struct bridge *
2671 bridge_lookup(const char *name)
2675 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2676 if (!strcmp(br->name, name)) {
2683 /* Handle requests for a listing of all flows known by the OpenFlow
2684 * stack, including those normally hidden. */
2686 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2687 const char *argv[], void *aux OVS_UNUSED)
2692 br = bridge_lookup(argv[1]);
2694 unixctl_command_reply_error(conn, "Unknown bridge");
2699 ofproto_get_all_flows(br->ofproto, &results);
2701 unixctl_command_reply(conn, ds_cstr(&results));
2702 ds_destroy(&results);
2705 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2706 * connections and reconnect. If BRIDGE is not specified, then all bridges
2707 * drop their controller connections and reconnect. */
2709 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2710 const char *argv[], void *aux OVS_UNUSED)
2714 br = bridge_lookup(argv[1]);
2716 unixctl_command_reply_error(conn, "Unknown bridge");
2719 ofproto_reconnect_controllers(br->ofproto);
2721 HMAP_FOR_EACH (br, node, &all_bridges) {
2722 ofproto_reconnect_controllers(br->ofproto);
2725 unixctl_command_reply(conn, NULL);
2729 bridge_get_controllers(const struct bridge *br,
2730 struct ovsrec_controller ***controllersp)
2732 struct ovsrec_controller **controllers;
2733 size_t n_controllers;
2735 controllers = br->cfg->controller;
2736 n_controllers = br->cfg->n_controller;
2738 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2744 *controllersp = controllers;
2746 return n_controllers;
2750 bridge_queue_if_cfg(struct bridge *br,
2751 const struct ovsrec_interface *cfg,
2752 const struct ovsrec_port *parent)
2754 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2757 if_cfg->parent = parent;
2758 if_cfg->ofport = iface_pick_ofport(cfg);
2759 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2760 hash_string(if_cfg->cfg->name, 0));
2763 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2764 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2765 * 'br' needs to complete its configuration. */
2767 bridge_add_del_ports(struct bridge *br,
2768 const unsigned long int *splinter_vlans)
2770 struct shash_node *port_node;
2771 struct port *port, *next;
2772 struct shash new_ports;
2775 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2777 /* Collect new ports. */
2778 shash_init(&new_ports);
2779 for (i = 0; i < br->cfg->n_ports; i++) {
2780 const char *name = br->cfg->ports[i]->name;
2781 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2782 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2786 if (bridge_get_controllers(br, NULL)
2787 && !shash_find(&new_ports, br->name)) {
2788 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2789 br->name, br->name);
2791 ovsrec_interface_init(&br->synth_local_iface);
2792 ovsrec_port_init(&br->synth_local_port);
2794 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2795 br->synth_local_port.n_interfaces = 1;
2796 br->synth_local_port.name = br->name;
2798 br->synth_local_iface.name = br->name;
2799 br->synth_local_iface.type = "internal";
2801 br->synth_local_ifacep = &br->synth_local_iface;
2803 shash_add(&new_ports, br->name, &br->synth_local_port);
2806 if (splinter_vlans) {
2807 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2810 /* Get rid of deleted ports.
2811 * Get rid of deleted interfaces on ports that still exist. */
2812 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2813 port->cfg = shash_find_data(&new_ports, port->name);
2817 port_del_ifaces(port);
2821 /* Update iface->cfg and iface->type in interfaces that still exist.
2822 * Add new interfaces to creation queue. */
2823 SHASH_FOR_EACH (port_node, &new_ports) {
2824 const struct ovsrec_port *port = port_node->data;
2827 for (i = 0; i < port->n_interfaces; i++) {
2828 const struct ovsrec_interface *cfg = port->interfaces[i];
2829 struct iface *iface = iface_lookup(br, cfg->name);
2830 const char *type = iface_get_type(cfg, br->cfg);
2835 } else if (!strcmp(type, "null")) {
2836 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2837 " may be removed in February 2013. Please email"
2838 " dev@openvswitch.org with concerns.",
2841 bridge_queue_if_cfg(br, cfg, port);
2846 shash_destroy(&new_ports);
2849 /* Initializes 'oc' appropriately as a management service controller for
2852 * The caller must free oc->target when it is no longer needed. */
2854 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2855 struct ofproto_controller *oc)
2857 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2858 oc->max_backoff = 0;
2859 oc->probe_interval = 60;
2860 oc->band = OFPROTO_OUT_OF_BAND;
2862 oc->burst_limit = 0;
2863 oc->enable_async_msgs = true;
2866 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2868 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2869 struct ofproto_controller *oc)
2873 oc->target = c->target;
2874 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2875 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2876 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2877 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2878 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2879 oc->burst_limit = (c->controller_burst_limit
2880 ? *c->controller_burst_limit : 0);
2881 oc->enable_async_msgs = (!c->enable_async_messages
2882 || *c->enable_async_messages);
2883 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2884 if (dscp < 0 || dscp > 63) {
2885 dscp = DSCP_DEFAULT;
2890 /* Configures the IP stack for 'br''s local interface properly according to the
2891 * configuration in 'c'. */
2893 bridge_configure_local_iface_netdev(struct bridge *br,
2894 struct ovsrec_controller *c)
2896 struct netdev *netdev;
2897 struct in_addr mask, gateway;
2899 struct iface *local_iface;
2902 /* If there's no local interface or no IP address, give up. */
2903 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2904 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2908 /* Bring up the local interface. */
2909 netdev = local_iface->netdev;
2910 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2912 /* Configure the IP address and netmask. */
2913 if (!c->local_netmask
2914 || !inet_aton(c->local_netmask, &mask)
2916 mask.s_addr = guess_netmask(ip.s_addr);
2918 if (!netdev_set_in4(netdev, ip, mask)) {
2919 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2920 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2923 /* Configure the default gateway. */
2924 if (c->local_gateway
2925 && inet_aton(c->local_gateway, &gateway)
2926 && gateway.s_addr) {
2927 if (!netdev_add_router(netdev, gateway)) {
2928 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2929 br->name, IP_ARGS(gateway.s_addr));
2934 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2935 * in either string are treated as equal to any number of slashes in the other,
2936 * e.g. "x///y" is equal to "x/y".
2938 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2939 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2940 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2941 * 'b' against a prefix of 'a'.
2944 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2946 const char *b_start = b;
2948 if (b - b_start >= b_stoplen) {
2950 } else if (*a != *b) {
2952 } else if (*a == '/') {
2953 a += strspn(a, "/");
2954 b += strspn(b, "/");
2955 } else if (*a == '\0') {
2965 bridge_configure_remotes(struct bridge *br,
2966 const struct sockaddr_in *managers, size_t n_managers)
2968 bool disable_in_band;
2970 struct ovsrec_controller **controllers;
2971 size_t n_controllers;
2973 enum ofproto_fail_mode fail_mode;
2975 struct ofproto_controller *ocs;
2979 /* Check if we should disable in-band control on this bridge. */
2980 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2983 /* Set OpenFlow queue ID for in-band control. */
2984 ofproto_set_in_band_queue(br->ofproto,
2985 smap_get_int(&br->cfg->other_config,
2986 "in-band-queue", -1));
2988 if (disable_in_band) {
2989 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2991 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2994 n_controllers = bridge_get_controllers(br, &controllers);
2996 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2999 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3000 for (i = 0; i < n_controllers; i++) {
3001 struct ovsrec_controller *c = controllers[i];
3003 if (!strncmp(c->target, "punix:", 6)
3004 || !strncmp(c->target, "unix:", 5)) {
3005 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3008 if (!strncmp(c->target, "unix:", 5)) {
3009 /* Connect to a listening socket */
3010 whitelist = xasprintf("unix:%s/", ovs_rundir());
3011 if (strchr(c->target, '/') &&
3012 !equal_pathnames(c->target, whitelist,
3013 strlen(whitelist))) {
3014 /* Absolute path specified, but not in ovs_rundir */
3015 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3016 "controller \"%s\" due to possibility for "
3017 "remote exploit. Instead, specify socket "
3018 "in whitelisted \"%s\" or connect to "
3019 "\"unix:%s/%s.mgmt\" (which is always "
3020 "available without special configuration).",
3021 br->name, c->target, whitelist,
3022 ovs_rundir(), br->name);
3027 whitelist = xasprintf("punix:%s/%s.controller",
3028 ovs_rundir(), br->name);
3029 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3030 /* Prevent remote ovsdb-server users from accessing
3031 * arbitrary Unix domain sockets and overwriting arbitrary
3033 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3034 "controller \"%s\" due to possibility of "
3035 "overwriting local files. Instead, specify "
3036 "whitelisted \"%s\" or connect to "
3037 "\"unix:%s/%s.mgmt\" (which is always "
3038 "available without special configuration).",
3039 br->name, c->target, whitelist,
3040 ovs_rundir(), br->name);
3049 bridge_configure_local_iface_netdev(br, c);
3050 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3051 if (disable_in_band) {
3052 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3057 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3058 bridge_get_allowed_versions(br));
3059 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3062 /* Set the fail-mode. */
3063 fail_mode = !br->cfg->fail_mode
3064 || !strcmp(br->cfg->fail_mode, "standalone")
3065 ? OFPROTO_FAIL_STANDALONE
3066 : OFPROTO_FAIL_SECURE;
3067 ofproto_set_fail_mode(br->ofproto, fail_mode);
3069 /* Configure OpenFlow controller connection snooping. */
3070 if (!ofproto_has_snoops(br->ofproto)) {
3074 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3075 ovs_rundir(), br->name));
3076 ofproto_set_snoops(br->ofproto, &snoops);
3077 sset_destroy(&snoops);
3082 bridge_configure_tables(struct bridge *br)
3084 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3088 n_tables = ofproto_get_n_tables(br->ofproto);
3090 for (i = 0; i < n_tables; i++) {
3091 struct ofproto_table_settings s;
3094 s.max_flows = UINT_MAX;
3098 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3099 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3102 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3103 s.max_flows = *cfg->flow_limit;
3105 if (cfg->overflow_policy
3106 && !strcmp(cfg->overflow_policy, "evict")) {
3109 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3110 for (k = 0; k < cfg->n_groups; k++) {
3111 const char *string = cfg->groups[k];
3114 msg = mf_parse_subfield__(&s.groups[k], &string);
3116 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3117 "'groups' (%s)", br->name, i, msg);
3119 } else if (*string) {
3120 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3121 "element '%s' contains trailing garbage",
3122 br->name, i, cfg->groups[k]);
3130 ofproto_configure_table(br->ofproto, i, &s);
3134 for (; j < br->cfg->n_flow_tables; j++) {
3135 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3136 "%"PRId64" not supported by this datapath", br->name,
3137 br->cfg->key_flow_tables[j]);
3142 bridge_configure_dp_desc(struct bridge *br)
3144 ofproto_set_dp_desc(br->ofproto,
3145 smap_get(&br->cfg->other_config, "dp-desc"));
3148 /* Port functions. */
3150 static struct port *
3151 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3155 port = xzalloc(sizeof *port);
3157 port->name = xstrdup(cfg->name);
3159 list_init(&port->ifaces);
3161 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3165 /* Deletes interfaces from 'port' that are no longer configured for it. */
3167 port_del_ifaces(struct port *port)
3169 struct iface *iface, *next;
3170 struct sset new_ifaces;
3173 /* Collect list of new interfaces. */
3174 sset_init(&new_ifaces);
3175 for (i = 0; i < port->cfg->n_interfaces; i++) {
3176 const char *name = port->cfg->interfaces[i]->name;
3177 const char *type = port->cfg->interfaces[i]->type;
3178 if (strcmp(type, "null")) {
3179 sset_add(&new_ifaces, name);
3183 /* Get rid of deleted interfaces. */
3184 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3185 if (!sset_contains(&new_ifaces, iface->name)) {
3186 iface_destroy(iface);
3190 sset_destroy(&new_ifaces);
3194 port_destroy(struct port *port)
3197 struct bridge *br = port->bridge;
3198 struct iface *iface, *next;
3201 ofproto_bundle_unregister(br->ofproto, port);
3204 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3205 iface_destroy(iface);
3208 hmap_remove(&br->ports, &port->hmap_node);
3214 static struct port *
3215 port_lookup(const struct bridge *br, const char *name)
3219 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3221 if (!strcmp(port->name, name)) {
3229 enable_lacp(struct port *port, bool *activep)
3231 if (!port->cfg->lacp) {
3232 /* XXX when LACP implementation has been sufficiently tested, enable by
3233 * default and make active on bonded ports. */
3235 } else if (!strcmp(port->cfg->lacp, "off")) {
3237 } else if (!strcmp(port->cfg->lacp, "active")) {
3240 } else if (!strcmp(port->cfg->lacp, "passive")) {
3244 VLOG_WARN("port %s: unknown LACP mode %s",
3245 port->name, port->cfg->lacp);
3250 static struct lacp_settings *
3251 port_configure_lacp(struct port *port, struct lacp_settings *s)
3253 const char *lacp_time, *system_id;
3256 if (!enable_lacp(port, &s->active)) {
3260 s->name = port->name;
3262 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3264 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3265 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3266 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3267 " address.", port->name, system_id);
3271 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3274 if (eth_addr_is_zero(s->id)) {
3275 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3279 /* Prefer bondable links if unspecified. */
3280 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3282 s->priority = (priority > 0 && priority <= UINT16_MAX
3284 : UINT16_MAX - !list_is_short(&port->ifaces));
3286 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3287 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3292 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3294 int priority, portid, key;
3296 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3297 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3299 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3301 if (portid <= 0 || portid > UINT16_MAX) {
3302 portid = iface->ofp_port;
3305 if (priority <= 0 || priority > UINT16_MAX) {
3306 priority = UINT16_MAX;
3309 if (key < 0 || key > UINT16_MAX) {
3313 s->name = iface->name;
3315 s->priority = priority;
3320 port_configure_bond(struct port *port, struct bond_settings *s)
3322 const char *detect_s;
3323 struct iface *iface;
3324 int miimon_interval;
3326 s->name = port->name;
3328 if (port->cfg->bond_mode) {
3329 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3330 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3331 port->name, port->cfg->bond_mode,
3332 bond_mode_to_string(s->balance));
3335 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3337 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3338 * active-backup. At some point we should remove this warning. */
3339 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3340 " in previous versions, the default bond_mode was"
3341 " balance-slb", port->name,
3342 bond_mode_to_string(s->balance));
3344 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3345 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3346 "please use another bond type or disable flood_vlans",
3350 miimon_interval = smap_get_int(&port->cfg->other_config,
3351 "bond-miimon-interval", 0);
3352 if (miimon_interval <= 0) {
3353 miimon_interval = 200;
3356 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3357 if (!detect_s || !strcmp(detect_s, "carrier")) {
3358 miimon_interval = 0;
3359 } else if (strcmp(detect_s, "miimon")) {
3360 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3361 "defaulting to carrier", port->name, detect_s);
3362 miimon_interval = 0;
3365 s->up_delay = MAX(0, port->cfg->bond_updelay);
3366 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3367 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3368 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3369 "bond-rebalance-interval", 10000);
3370 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3371 s->rebalance_interval = 1000;
3374 s->fake_iface = port->cfg->bond_fake_iface;
3376 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3377 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3381 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3382 * instead of obtaining it from the database. */
3384 port_is_synthetic(const struct port *port)
3386 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3389 /* Interface functions. */
3392 iface_is_internal(const struct ovsrec_interface *iface,
3393 const struct ovsrec_bridge *br)
3395 /* The local port and "internal" ports are always "internal". */
3396 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3399 /* Returns the correct network device type for interface 'iface' in bridge
3402 iface_get_type(const struct ovsrec_interface *iface,
3403 const struct ovsrec_bridge *br)
3407 /* The local port always has type "internal". Other ports take
3408 * their type from the database and default to "system" if none is
3410 if (iface_is_internal(iface, br)) {
3413 type = iface->type[0] ? iface->type : "system";
3416 return ofproto_port_open_type(br->datapath_type, type);
3420 iface_destroy(struct iface *iface)
3423 struct port *port = iface->port;
3424 struct bridge *br = port->bridge;
3426 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3427 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3430 if (iface->ofp_port != OFPP_NONE) {
3431 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3434 list_remove(&iface->port_elem);
3435 hmap_remove(&br->iface_by_name, &iface->name_node);
3437 netdev_close(iface->netdev);
3444 static struct iface *
3445 iface_lookup(const struct bridge *br, const char *name)
3447 struct iface *iface;
3449 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3450 &br->iface_by_name) {
3451 if (!strcmp(iface->name, name)) {
3459 static struct iface *
3460 iface_find(const char *name)
3462 const struct bridge *br;
3464 HMAP_FOR_EACH (br, node, &all_bridges) {
3465 struct iface *iface = iface_lookup(br, name);
3474 static struct if_cfg *
3475 if_cfg_lookup(const struct bridge *br, const char *name)
3477 struct if_cfg *if_cfg;
3479 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3481 if (!strcmp(if_cfg->cfg->name, name)) {
3489 static struct iface *
3490 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3492 struct iface *iface;
3494 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3495 hash_int(ofp_port, 0), &br->ifaces) {
3496 if (iface->ofp_port == ofp_port) {
3503 /* Set Ethernet address of 'iface', if one is specified in the configuration
3506 iface_set_mac(struct iface *iface)
3508 uint8_t ea[ETH_ADDR_LEN];
3510 if (!strcmp(iface->type, "internal")
3511 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3512 if (iface->ofp_port == OFPP_LOCAL) {
3513 VLOG_ERR("interface %s: ignoring mac in Interface record "
3514 "(use Bridge record to set local port's mac)",
3516 } else if (eth_addr_is_multicast(ea)) {
3517 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3520 int error = netdev_set_etheraddr(iface->netdev, ea);
3522 VLOG_ERR("interface %s: setting MAC failed (%s)",
3523 iface->name, strerror(error));
3529 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3531 iface_set_ofport(const struct ovsrec_interface *if_cfg, uint16_t ofport)
3534 port_ = (ofport == OFPP_NONE) ? -1 : ofport;
3535 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3536 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3540 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3541 * sets the "ofport" field to -1.
3543 * This is appropriate when 'if_cfg''s interface cannot be created or is
3544 * otherwise invalid. */
3546 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3548 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3549 ovsrec_interface_set_status(if_cfg, NULL);
3550 ovsrec_interface_set_admin_state(if_cfg, NULL);
3551 ovsrec_interface_set_duplex(if_cfg, NULL);
3552 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3553 ovsrec_interface_set_link_state(if_cfg, NULL);
3554 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3555 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3556 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3557 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3558 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3559 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3560 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3564 struct iface_delete_queues_cbdata {
3565 struct netdev *netdev;
3566 const struct ovsdb_datum *queues;
3570 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3572 union ovsdb_atom atom;
3574 atom.integer = target;
3575 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3579 iface_delete_queues(unsigned int queue_id,
3580 const struct smap *details OVS_UNUSED, void *cbdata_)
3582 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3584 if (!queue_ids_include(cbdata->queues, queue_id)) {
3585 netdev_delete_queue(cbdata->netdev, queue_id);
3590 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3592 struct ofpbuf queues_buf;
3594 ofpbuf_init(&queues_buf, 0);
3596 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3597 netdev_set_qos(iface->netdev, NULL, NULL);
3599 struct iface_delete_queues_cbdata cbdata;
3603 /* Configure top-level Qos for 'iface'. */
3604 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3606 /* Deconfigure queues that were deleted. */
3607 cbdata.netdev = iface->netdev;
3608 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3610 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3612 /* Configure queues for 'iface'. */
3614 for (i = 0; i < qos->n_queues; i++) {
3615 const struct ovsrec_queue *queue = qos->value_queues[i];
3616 unsigned int queue_id = qos->key_queues[i];
3618 if (queue_id == 0) {
3622 if (queue->n_dscp == 1) {
3623 struct ofproto_port_queue *port_queue;
3625 port_queue = ofpbuf_put_uninit(&queues_buf,
3626 sizeof *port_queue);
3627 port_queue->queue = queue_id;
3628 port_queue->dscp = queue->dscp[0];
3631 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3634 struct smap details;
3636 smap_init(&details);
3637 netdev_set_queue(iface->netdev, 0, &details);
3638 smap_destroy(&details);
3642 if (iface->ofp_port != OFPP_NONE) {
3643 const struct ofproto_port_queue *port_queues = queues_buf.data;
3644 size_t n_queues = queues_buf.size / sizeof *port_queues;
3646 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3647 port_queues, n_queues);
3650 netdev_set_policing(iface->netdev,
3651 iface->cfg->ingress_policing_rate,
3652 iface->cfg->ingress_policing_burst);
3654 ofpbuf_uninit(&queues_buf);
3658 iface_configure_cfm(struct iface *iface)
3660 const struct ovsrec_interface *cfg = iface->cfg;
3661 const char *opstate_str;
3662 const char *cfm_ccm_vlan;
3663 struct cfm_settings s;
3664 struct smap netdev_args;
3666 if (!cfg->n_cfm_mpid) {
3667 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3671 s.check_tnl_key = false;
3672 smap_init(&netdev_args);
3673 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3674 const char *key = smap_get(&netdev_args, "key");
3675 const char *in_key = smap_get(&netdev_args, "in_key");
3677 s.check_tnl_key = (key && !strcmp(key, "flow"))
3678 || (in_key && !strcmp(in_key, "flow"));
3680 smap_destroy(&netdev_args);
3682 s.mpid = *cfg->cfm_mpid;
3683 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3684 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3685 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3687 if (s.interval <= 0) {
3691 if (!cfm_ccm_vlan) {
3693 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3694 s.ccm_vlan = CFM_RANDOM_VLAN;
3696 s.ccm_vlan = atoi(cfm_ccm_vlan);
3697 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3702 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3704 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3706 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3707 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3709 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3712 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3713 * instead of obtaining it from the database. */
3715 iface_is_synthetic(const struct iface *iface)
3717 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3721 iface_pick_ofport(const struct ovsrec_interface *cfg)
3723 uint16_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3724 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3728 /* Port mirroring. */
3730 static struct mirror *
3731 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3735 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3736 if (uuid_equals(uuid, &m->uuid)) {
3744 bridge_configure_mirrors(struct bridge *br)
3746 const struct ovsdb_datum *mc;
3747 unsigned long *flood_vlans;
3748 struct mirror *m, *next;
3751 /* Get rid of deleted mirrors. */
3752 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3753 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3754 union ovsdb_atom atom;
3756 atom.uuid = m->uuid;
3757 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3762 /* Add new mirrors and reconfigure existing ones. */
3763 for (i = 0; i < br->cfg->n_mirrors; i++) {
3764 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3765 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3767 m = mirror_create(br, cfg);
3770 if (!mirror_configure(m)) {
3775 /* Update flooded vlans (for RSPAN). */
3776 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3777 br->cfg->n_flood_vlans);
3778 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3779 bitmap_free(flood_vlans);
3782 static struct mirror *
3783 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3787 m = xzalloc(sizeof *m);
3788 m->uuid = cfg->header_.uuid;
3789 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3791 m->name = xstrdup(cfg->name);
3797 mirror_destroy(struct mirror *m)
3800 struct bridge *br = m->bridge;
3803 ofproto_mirror_unregister(br->ofproto, m);
3806 hmap_remove(&br->mirrors, &m->hmap_node);
3813 mirror_collect_ports(struct mirror *m,
3814 struct ovsrec_port **in_ports, int n_in_ports,
3815 void ***out_portsp, size_t *n_out_portsp)
3817 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3818 size_t n_out_ports = 0;
3821 for (i = 0; i < n_in_ports; i++) {
3822 const char *name = in_ports[i]->name;
3823 struct port *port = port_lookup(m->bridge, name);
3825 out_ports[n_out_ports++] = port;
3827 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3828 "port %s", m->bridge->name, m->name, name);
3831 *out_portsp = out_ports;
3832 *n_out_portsp = n_out_ports;
3836 mirror_configure(struct mirror *m)
3838 const struct ovsrec_mirror *cfg = m->cfg;
3839 struct ofproto_mirror_settings s;
3842 if (strcmp(cfg->name, m->name)) {
3844 m->name = xstrdup(cfg->name);
3848 /* Get output port or VLAN. */
3849 if (cfg->output_port) {
3850 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3851 if (!s.out_bundle) {
3852 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3853 m->bridge->name, m->name);
3856 s.out_vlan = UINT16_MAX;
3858 if (cfg->output_vlan) {
3859 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3860 "output vlan; ignoring output vlan",
3861 m->bridge->name, m->name);
3863 } else if (cfg->output_vlan) {
3864 /* The database should prevent invalid VLAN values. */
3865 s.out_bundle = NULL;
3866 s.out_vlan = *cfg->output_vlan;
3868 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3869 m->bridge->name, m->name);
3873 /* Get port selection. */
3874 if (cfg->select_all) {
3875 size_t n_ports = hmap_count(&m->bridge->ports);
3876 void **ports = xmalloc(n_ports * sizeof *ports);
3881 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3891 /* Get ports, dropping ports that don't exist.
3892 * The IDL ensures that there are no duplicates. */
3893 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3894 &s.srcs, &s.n_srcs);
3895 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3896 &s.dsts, &s.n_dsts);
3899 /* Get VLAN selection. */
3900 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3903 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3906 if (s.srcs != s.dsts) {
3915 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3917 * This is deprecated. It is only for compatibility with broken device drivers
3918 * in old versions of Linux that do not properly support VLANs when VLAN
3919 * devices are not used. When broken device drivers are no longer in
3920 * widespread use, we will delete these interfaces. */
3922 static struct ovsrec_port **recs;
3923 static size_t n_recs, allocated_recs;
3925 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3926 * splinters are reconfigured. */
3928 register_rec(struct ovsrec_port *rec)
3930 if (n_recs >= allocated_recs) {
3931 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3933 recs[n_recs++] = rec;
3936 /* Frees all of the ports registered with register_reg(). */
3938 free_registered_recs(void)
3942 for (i = 0; i < n_recs; i++) {
3943 struct ovsrec_port *port = recs[i];
3946 for (j = 0; j < port->n_interfaces; j++) {
3947 struct ovsrec_interface *iface = port->interfaces[j];
3952 smap_destroy(&port->other_config);
3953 free(port->interfaces);
3961 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3964 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3966 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3970 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3973 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3974 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3975 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3978 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3979 * use, returns NULL.
3981 * Updates 'vlan_splinters_enabled_anywhere'. */
3982 static unsigned long int *
3983 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3985 unsigned long int *splinter_vlans;
3986 struct sset splinter_ifaces;
3987 const char *real_dev_name;
3988 struct shash *real_devs;
3989 struct shash_node *node;
3993 /* Free space allocated for synthesized ports and interfaces, since we're
3994 * in the process of reconstructing all of them. */
3995 free_registered_recs();
3997 splinter_vlans = bitmap_allocate(4096);
3998 sset_init(&splinter_ifaces);
3999 vlan_splinters_enabled_anywhere = false;
4000 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4001 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4004 for (j = 0; j < br_cfg->n_ports; j++) {
4005 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4008 for (k = 0; k < port_cfg->n_interfaces; k++) {
4009 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4011 if (vlan_splinters_is_enabled(iface_cfg)) {
4012 vlan_splinters_enabled_anywhere = true;
4013 sset_add(&splinter_ifaces, iface_cfg->name);
4014 vlan_bitmap_from_array__(port_cfg->trunks,
4020 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4021 bitmap_set1(splinter_vlans, *port_cfg->tag);
4026 if (!vlan_splinters_enabled_anywhere) {
4027 free(splinter_vlans);
4028 sset_destroy(&splinter_ifaces);
4032 HMAP_FOR_EACH (br, node, &all_bridges) {
4034 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4038 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4039 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4040 * device to be created for it. */
4041 bitmap_set0(splinter_vlans, 0);
4042 bitmap_set0(splinter_vlans, 4095);
4044 /* Delete all VLAN devices that we don't need. */
4046 real_devs = vlandev_get_real_devs();
4047 SHASH_FOR_EACH (node, real_devs) {
4048 const struct vlan_real_dev *real_dev = node->data;
4049 const struct vlan_dev *vlan_dev;
4050 bool real_dev_has_splinters;
4052 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4054 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4055 if (!real_dev_has_splinters
4056 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4057 struct netdev *netdev;
4059 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4060 if (!netdev_get_in4(netdev, NULL, NULL) ||
4061 !netdev_get_in6(netdev, NULL)) {
4062 vlandev_del(vlan_dev->name);
4064 /* It has an IP address configured, so we don't own
4065 * it. Don't delete it. */
4067 netdev_close(netdev);
4074 /* Add all VLAN devices that we need. */
4075 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4078 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4079 if (!vlandev_get_name(real_dev_name, vid)) {
4080 vlandev_add(real_dev_name, vid);
4087 sset_destroy(&splinter_ifaces);
4089 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4090 free(splinter_vlans);
4093 return splinter_vlans;
4096 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4099 configure_splinter_port(struct port *port)
4101 struct ofproto *ofproto = port->bridge->ofproto;
4102 uint16_t realdev_ofp_port;
4103 const char *realdev_name;
4104 struct iface *vlandev, *realdev;
4106 ofproto_bundle_unregister(port->bridge->ofproto, port);
4108 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4111 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4112 realdev = iface_lookup(port->bridge, realdev_name);
4113 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4115 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4119 static struct ovsrec_port *
4120 synthesize_splinter_port(const char *real_dev_name,
4121 const char *vlan_dev_name, int vid)
4123 struct ovsrec_interface *iface;
4124 struct ovsrec_port *port;
4126 iface = xmalloc(sizeof *iface);
4127 ovsrec_interface_init(iface);
4128 iface->name = xstrdup(vlan_dev_name);
4129 iface->type = "system";
4131 port = xmalloc(sizeof *port);
4132 ovsrec_port_init(port);
4133 port->interfaces = xmemdup(&iface, sizeof iface);
4134 port->n_interfaces = 1;
4135 port->name = xstrdup(vlan_dev_name);
4136 port->vlan_mode = "splinter";
4137 port->tag = xmalloc(sizeof *port->tag);
4140 smap_add(&port->other_config, "realdev", real_dev_name);
4146 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4147 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4148 * 1-bit in the 'splinter_vlans' bitmap. */
4150 add_vlan_splinter_ports(struct bridge *br,
4151 const unsigned long int *splinter_vlans,
4152 struct shash *ports)
4156 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4157 * we're modifying 'ports'. */
4158 for (i = 0; i < br->cfg->n_ports; i++) {
4159 const char *name = br->cfg->ports[i]->name;
4160 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4163 for (j = 0; j < port_cfg->n_interfaces; j++) {
4164 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4166 if (vlan_splinters_is_enabled(iface_cfg)) {
4167 const char *real_dev_name;
4170 real_dev_name = iface_cfg->name;
4171 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4172 const char *vlan_dev_name;
4174 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4176 && !shash_find(ports, vlan_dev_name)) {
4177 shash_add(ports, vlan_dev_name,
4178 synthesize_splinter_port(
4179 real_dev_name, vlan_dev_name, vid));
4188 mirror_refresh_stats(struct mirror *m)
4190 struct ofproto *ofproto = m->bridge->ofproto;
4191 uint64_t tx_packets, tx_bytes;
4194 size_t stat_cnt = 0;
4196 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4197 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4201 if (tx_packets != UINT64_MAX) {
4202 keys[stat_cnt] = "tx_packets";
4203 values[stat_cnt] = tx_packets;
4206 if (tx_bytes != UINT64_MAX) {
4207 keys[stat_cnt] = "tx_bytes";
4208 values[stat_cnt] = tx_bytes;
4212 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);