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 int64_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 int ofp_port; /* OpenFlow port number, -1 if unknown. */
90 struct netdev *netdev; /* Network device. */
91 const char *type; /* Usually same as cfg->type. */
92 const struct ovsrec_interface *cfg;
96 struct uuid uuid; /* UUID of this "mirror" record in database. */
97 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
98 struct bridge *bridge;
100 const struct ovsrec_mirror *cfg;
104 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
105 struct bridge *bridge;
108 const struct ovsrec_port *cfg;
110 /* An ordinary bridge port has 1 interface.
111 * A bridge port for bonding has at least 2 interfaces. */
112 struct list ifaces; /* List of "struct iface"s. */
116 struct hmap_node node; /* In 'all_bridges'. */
117 char *name; /* User-specified arbitrary name. */
118 char *type; /* Datapath type. */
119 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
120 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
121 const struct ovsrec_bridge *cfg;
123 /* OpenFlow switch processing. */
124 struct ofproto *ofproto; /* OpenFlow switch. */
127 struct hmap ports; /* "struct port"s indexed by name. */
128 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
129 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
131 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
132 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
133 Indexed on 'cfg->name'. */
135 /* Port mirroring. */
136 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
138 /* Synthetic local port if necessary. */
139 struct ovsrec_port synth_local_port;
140 struct ovsrec_interface synth_local_iface;
141 struct ovsrec_interface *synth_local_ifacep;
144 /* All bridges, indexed by name. */
145 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
147 /* OVSDB IDL used to obtain configuration. */
148 static struct ovsdb_idl *idl;
150 /* We want to complete daemonization, fully detaching from our parent process,
151 * only after we have completed our initial configuration, committed our state
152 * to the database, and received confirmation back from the database server
153 * that it applied the commit. This allows our parent process to know that,
154 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
155 * to have already been filled in. (It is only "very likely" rather than
156 * certain because there is always a slim possibility that the transaction will
157 * fail or that some other client has added new bridges, ports, etc. while
158 * ovs-vswitchd was configuring using an old configuration.)
160 * We only need to do this once for our initial configuration at startup, so
161 * 'initial_config_done' tracks whether we've already done it. While we are
162 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
163 * itself and is otherwise NULL. */
164 static bool initial_config_done;
165 static struct ovsdb_idl_txn *daemonize_txn;
167 /* Most recently processed IDL sequence number. */
168 static unsigned int idl_seqno;
170 /* Each time this timer expires, the bridge fetches interface and mirror
171 * statistics and pushes them into the database. */
172 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
173 static long long int iface_stats_timer = LLONG_MIN;
175 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
176 * expensive. This can cause bridge_reconfigure() to take a long time during
177 * which no other work can be done. To deal with this problem, we limit port
178 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
179 * call to bridge_reconfigure(). If there is more work to do after the limit
180 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
181 * This allows the rest of the code to catch up on important things like
182 * forwarding packets. */
183 #define OFP_PORT_ACTION_WINDOW 10
184 static bool reconfiguring = false;
186 static void add_del_bridges(const struct ovsrec_open_vswitch *);
187 static void bridge_update_ofprotos(void);
188 static void bridge_create(const struct ovsrec_bridge *);
189 static void bridge_destroy(struct bridge *);
190 static struct bridge *bridge_lookup(const char *name);
191 static unixctl_cb_func bridge_unixctl_dump_flows;
192 static unixctl_cb_func bridge_unixctl_reconnect;
193 static size_t bridge_get_controllers(const struct bridge *br,
194 struct ovsrec_controller ***controllersp);
195 static void bridge_add_del_ports(struct bridge *,
196 const unsigned long int *splinter_vlans);
197 static void bridge_refresh_ofp_port(struct bridge *);
198 static void bridge_configure_datapath_id(struct bridge *);
199 static void bridge_configure_flow_eviction_threshold(struct bridge *);
200 static void bridge_configure_netflow(struct bridge *);
201 static void bridge_configure_forward_bpdu(struct bridge *);
202 static void bridge_configure_mac_table(struct bridge *);
203 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
204 static void bridge_configure_ipfix(struct bridge *);
205 static void bridge_configure_stp(struct bridge *);
206 static void bridge_configure_tables(struct bridge *);
207 static void bridge_configure_dp_desc(struct bridge *);
208 static void bridge_configure_remotes(struct bridge *,
209 const struct sockaddr_in *managers,
211 static void bridge_pick_local_hw_addr(struct bridge *,
212 uint8_t ea[ETH_ADDR_LEN],
213 struct iface **hw_addr_iface);
214 static uint64_t bridge_pick_datapath_id(struct bridge *,
215 const uint8_t bridge_ea[ETH_ADDR_LEN],
216 struct iface *hw_addr_iface);
217 static void bridge_queue_if_cfg(struct bridge *,
218 const struct ovsrec_interface *,
219 const struct ovsrec_port *);
220 static uint64_t dpid_from_hash(const void *, size_t nbytes);
221 static bool bridge_has_bond_fake_iface(const struct bridge *,
223 static bool port_is_bond_fake_iface(const struct port *);
225 static unixctl_cb_func qos_unixctl_show;
227 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
228 static void port_del_ifaces(struct port *);
229 static void port_destroy(struct port *);
230 static struct port *port_lookup(const struct bridge *, const char *name);
231 static void port_configure(struct port *);
232 static struct lacp_settings *port_configure_lacp(struct port *,
233 struct lacp_settings *);
234 static void port_configure_bond(struct port *, struct bond_settings *);
235 static bool port_is_synthetic(const struct port *);
237 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
238 static void run_system_stats(void);
240 static void bridge_configure_mirrors(struct bridge *);
241 static struct mirror *mirror_create(struct bridge *,
242 const struct ovsrec_mirror *);
243 static void mirror_destroy(struct mirror *);
244 static bool mirror_configure(struct mirror *);
245 static void mirror_refresh_stats(struct mirror *);
247 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
248 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
249 static bool iface_is_internal(const struct ovsrec_interface *iface,
250 const struct ovsrec_bridge *br);
251 static const char *iface_get_type(const struct ovsrec_interface *,
252 const struct ovsrec_bridge *);
253 static void iface_destroy(struct iface *);
254 static struct iface *iface_lookup(const struct bridge *, const char *name);
255 static struct iface *iface_find(const char *name);
256 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
257 static struct iface *iface_from_ofp_port(const struct bridge *,
259 static void iface_set_mac(struct iface *);
260 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
261 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
262 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
263 static void iface_configure_cfm(struct iface *);
264 static void iface_refresh_cfm_stats(struct iface *);
265 static void iface_refresh_stats(struct iface *);
266 static void iface_refresh_status(struct iface *);
267 static bool iface_is_synthetic(const struct iface *);
268 static int64_t iface_pick_ofport(const struct ovsrec_interface *);
270 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
272 * This is deprecated. It is only for compatibility with broken device drivers
273 * in old versions of Linux that do not properly support VLANs when VLAN
274 * devices are not used. When broken device drivers are no longer in
275 * widespread use, we will delete these interfaces. */
277 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
278 static bool vlan_splinters_enabled_anywhere;
280 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
281 static unsigned long int *collect_splinter_vlans(
282 const struct ovsrec_open_vswitch *);
283 static void configure_splinter_port(struct port *);
284 static void add_vlan_splinter_ports(struct bridge *,
285 const unsigned long int *splinter_vlans,
286 struct shash *ports);
289 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
291 struct shash iface_hints;
292 static bool initialized = false;
299 shash_init(&iface_hints);
302 for (i = 0; i < cfg->n_bridges; i++) {
303 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
306 for (j = 0; j < br_cfg->n_ports; j++) {
307 struct ovsrec_port *port_cfg = br_cfg->ports[j];
310 for (k = 0; k < port_cfg->n_interfaces; k++) {
311 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
312 struct iface_hint *iface_hint;
314 iface_hint = xmalloc(sizeof *iface_hint);
315 iface_hint->br_name = br_cfg->name;
316 iface_hint->br_type = br_cfg->datapath_type;
317 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
319 shash_add(&iface_hints, if_cfg->name, iface_hint);
325 ofproto_init(&iface_hints);
327 shash_destroy_free_data(&iface_hints);
331 /* Public functions. */
333 /* Initializes the bridge module, configuring it to obtain its configuration
334 * from an OVSDB server accessed over 'remote', which should be a string in a
335 * form acceptable to ovsdb_idl_create(). */
337 bridge_init(const char *remote)
339 /* Create connection to database. */
340 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
341 idl_seqno = ovsdb_idl_get_seqno(idl);
342 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
343 ovsdb_idl_verify_write_only(idl);
345 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
346 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
353 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
354 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
355 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
357 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
358 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
359 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
360 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
379 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
383 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
384 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
386 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
391 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
393 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
404 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
406 /* Register unixctl commands. */
407 unixctl_command_register("qos/show", "interface", 1, 1,
408 qos_unixctl_show, NULL);
409 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
410 bridge_unixctl_dump_flows, NULL);
411 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
412 bridge_unixctl_reconnect, NULL);
422 struct bridge *br, *next_br;
424 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
427 ovsdb_idl_destroy(idl);
430 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
431 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
432 * responsible for freeing '*managersp' (with free()).
434 * You may be asking yourself "why does ovs-vswitchd care?", because
435 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
436 * should not be and in fact is not directly involved in that. But
437 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
438 * it has to tell in-band control where the managers are to enable that.
439 * (Thus, only managers connected in-band are collected.)
442 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
443 struct sockaddr_in **managersp, size_t *n_managersp)
445 struct sockaddr_in *managers = NULL;
446 size_t n_managers = 0;
450 /* Collect all of the potential targets from the "targets" columns of the
451 * rows pointed to by "manager_options", excluding any that are
454 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
455 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
457 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
458 sset_find_and_delete(&targets, m->target);
460 sset_add(&targets, m->target);
464 /* Now extract the targets' IP addresses. */
465 if (!sset_is_empty(&targets)) {
468 managers = xmalloc(sset_count(&targets) * sizeof *managers);
469 SSET_FOR_EACH (target, &targets) {
470 struct sockaddr_in *sin = &managers[n_managers];
472 if (stream_parse_target_with_default_ports(target,
480 sset_destroy(&targets);
482 *managersp = managers;
483 *n_managersp = n_managers;
487 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
489 unsigned long int *splinter_vlans;
492 COVERAGE_INC(bridge_reconfigure);
494 ovs_assert(!reconfiguring);
495 reconfiguring = true;
497 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
498 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
499 * configuration otherwise.
501 * This is mostly an update to bridge data structures. Nothing is pushed
502 * down to ofproto or lower layers. */
503 add_del_bridges(ovs_cfg);
504 splinter_vlans = collect_splinter_vlans(ovs_cfg);
505 HMAP_FOR_EACH (br, node, &all_bridges) {
506 bridge_add_del_ports(br, splinter_vlans);
508 free(splinter_vlans);
510 /* Delete datapaths that are no longer configured, and create ones which
511 * don't exist but should. */
512 bridge_update_ofprotos();
514 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
515 HMAP_FOR_EACH (br, node, &all_bridges) {
516 bridge_refresh_ofp_port(br);
519 /* Clear database records for "if_cfg"s which haven't been instantiated. */
520 HMAP_FOR_EACH (br, node, &all_bridges) {
521 struct if_cfg *if_cfg;
523 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
524 iface_clear_db_record(if_cfg->cfg);
528 reconfigure_system_stats(ovs_cfg);
532 bridge_reconfigure_ofp(void)
534 long long int deadline;
538 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
540 /* The kernel will reject any attempt to add a given port to a datapath if
541 * that port already belongs to a different datapath, so we must do all
542 * port deletions before any port additions. */
543 HMAP_FOR_EACH (br, node, &all_bridges) {
544 struct ofpp_garbage *garbage, *next;
546 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
547 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
548 * internal datastructures may not be consistent. Eventually, when
549 * port additions and deletions are cheaper, these calls should be
552 ofproto_port_del(br->ofproto, garbage->ofp_port);
553 list_remove(&garbage->list_node);
557 if (time_msec() >= deadline) {
564 HMAP_FOR_EACH (br, node, &all_bridges) {
565 struct if_cfg *if_cfg, *next;
567 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
568 iface_create(br, if_cfg, -1);
570 if (time_msec() >= deadline) {
580 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
582 struct sockaddr_in *managers;
583 int sflow_bridge_number;
588 ovs_assert(reconfiguring);
589 done = bridge_reconfigure_ofp();
591 /* Complete the configuration. */
592 sflow_bridge_number = 0;
593 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
594 HMAP_FOR_EACH (br, node, &all_bridges) {
597 /* We need the datapath ID early to allow LACP ports to use it as the
598 * default system ID. */
599 bridge_configure_datapath_id(br);
601 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
604 port_configure(port);
606 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
607 iface_configure_cfm(iface);
608 iface_configure_qos(iface, port->cfg->qos);
609 iface_set_mac(iface);
610 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
614 bridge_configure_mirrors(br);
615 bridge_configure_flow_eviction_threshold(br);
616 bridge_configure_forward_bpdu(br);
617 bridge_configure_mac_table(br);
618 bridge_configure_remotes(br, managers, n_managers);
619 bridge_configure_netflow(br);
620 bridge_configure_sflow(br, &sflow_bridge_number);
621 bridge_configure_ipfix(br);
622 bridge_configure_stp(br);
623 bridge_configure_tables(br);
624 bridge_configure_dp_desc(br);
631 /* Delete ofprotos which aren't configured or have the wrong type. Create
632 * ofprotos which don't exist but need to. */
634 bridge_update_ofprotos(void)
636 struct bridge *br, *next;
641 /* Delete ofprotos with no bridge or with the wrong type. */
644 ofproto_enumerate_types(&types);
645 SSET_FOR_EACH (type, &types) {
648 ofproto_enumerate_names(type, &names);
649 SSET_FOR_EACH (name, &names) {
650 br = bridge_lookup(name);
651 if (!br || strcmp(type, br->type)) {
652 ofproto_delete(name, type);
656 sset_destroy(&names);
657 sset_destroy(&types);
659 /* Add ofprotos for bridges which don't have one yet. */
660 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
668 /* Remove ports from any datapath with the same name as 'br'. If we
669 * don't do this, creating 'br''s ofproto will fail because a port with
670 * the same name as its local port already exists. */
671 HMAP_FOR_EACH (br2, node, &all_bridges) {
672 struct ofproto_port ofproto_port;
678 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
680 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
682 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
685 ofproto_port_destroy(&ofproto_port);
689 error = ofproto_create(br->name, br->type, &br->ofproto);
691 VLOG_ERR("failed to create bridge %s: %s", br->name,
699 port_configure(struct port *port)
701 const struct ovsrec_port *cfg = port->cfg;
702 struct bond_settings bond_settings;
703 struct lacp_settings lacp_settings;
704 struct ofproto_bundle_settings s;
707 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
708 configure_splinter_port(port);
717 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
718 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
719 s.slaves[s.n_slaves++] = iface->ofp_port;
724 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
728 /* Get VLAN trunks. */
731 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
735 if (cfg->vlan_mode) {
736 if (!strcmp(cfg->vlan_mode, "access")) {
737 s.vlan_mode = PORT_VLAN_ACCESS;
738 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
739 s.vlan_mode = PORT_VLAN_TRUNK;
740 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
741 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
742 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
743 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
745 /* This "can't happen" because ovsdb-server should prevent it. */
746 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
747 s.vlan_mode = PORT_VLAN_TRUNK;
751 s.vlan_mode = PORT_VLAN_ACCESS;
753 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
757 s.vlan_mode = PORT_VLAN_TRUNK;
760 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
763 /* Get LACP settings. */
764 s.lacp = port_configure_lacp(port, &lacp_settings);
768 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
769 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
770 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
773 s.lacp_slaves = NULL;
776 /* Get bond settings. */
777 if (s.n_slaves > 1) {
778 s.bond = &bond_settings;
779 port_configure_bond(port, &bond_settings);
782 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
783 netdev_set_miimon_interval(iface->netdev, 0);
788 ofproto_bundle_register(port->bridge->ofproto, port, &s);
796 /* Pick local port hardware address and datapath ID for 'br'. */
798 bridge_configure_datapath_id(struct bridge *br)
800 uint8_t ea[ETH_ADDR_LEN];
802 struct iface *local_iface;
803 struct iface *hw_addr_iface;
806 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
807 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
809 int error = netdev_set_etheraddr(local_iface->netdev, ea);
811 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
812 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
813 "Ethernet address: %s",
814 br->name, strerror(error));
817 memcpy(br->ea, ea, ETH_ADDR_LEN);
819 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
820 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
821 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
822 ofproto_set_datapath_id(br->ofproto, dpid);
825 dpid_string = xasprintf("%016"PRIx64, dpid);
826 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
830 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
833 bridge_get_allowed_versions(struct bridge *br)
835 if (!br->cfg->n_protocols)
838 return ofputil_versions_from_strings(br->cfg->protocols,
839 br->cfg->n_protocols);
842 /* Set NetFlow configuration on 'br'. */
844 bridge_configure_netflow(struct bridge *br)
846 struct ovsrec_netflow *cfg = br->cfg->netflow;
847 struct netflow_options opts;
850 ofproto_set_netflow(br->ofproto, NULL);
854 memset(&opts, 0, sizeof opts);
856 /* Get default NetFlow configuration from datapath.
857 * Apply overrides from 'cfg'. */
858 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
859 if (cfg->engine_type) {
860 opts.engine_type = *cfg->engine_type;
862 if (cfg->engine_id) {
863 opts.engine_id = *cfg->engine_id;
866 /* Configure active timeout interval. */
867 opts.active_timeout = cfg->active_timeout;
868 if (!opts.active_timeout) {
869 opts.active_timeout = -1;
870 } else if (opts.active_timeout < 0) {
871 VLOG_WARN("bridge %s: active timeout interval set to negative "
872 "value, using default instead (%d seconds)", br->name,
873 NF_ACTIVE_TIMEOUT_DEFAULT);
874 opts.active_timeout = -1;
877 /* Add engine ID to interface number to disambiguate bridgs? */
878 opts.add_id_to_iface = cfg->add_id_to_interface;
879 if (opts.add_id_to_iface) {
880 if (opts.engine_id > 0x7f) {
881 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
882 "another vswitch, choose an engine id less than 128",
885 if (hmap_count(&br->ports) > 508) {
886 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
887 "another port when more than 508 ports are used",
893 sset_init(&opts.collectors);
894 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
897 if (ofproto_set_netflow(br->ofproto, &opts)) {
898 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
900 sset_destroy(&opts.collectors);
903 /* Set sFlow configuration on 'br'. */
905 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
907 const struct ovsrec_sflow *cfg = br->cfg->sflow;
908 struct ovsrec_controller **controllers;
909 struct ofproto_sflow_options oso;
910 size_t n_controllers;
914 ofproto_set_sflow(br->ofproto, NULL);
918 memset(&oso, 0, sizeof oso);
920 sset_init(&oso.targets);
921 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
923 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
925 oso.sampling_rate = *cfg->sampling;
928 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
930 oso.polling_interval = *cfg->polling;
933 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
935 oso.header_len = *cfg->header;
938 oso.sub_id = (*sflow_bridge_number)++;
939 oso.agent_device = cfg->agent;
941 oso.control_ip = NULL;
942 n_controllers = bridge_get_controllers(br, &controllers);
943 for (i = 0; i < n_controllers; i++) {
944 if (controllers[i]->local_ip) {
945 oso.control_ip = controllers[i]->local_ip;
949 ofproto_set_sflow(br->ofproto, &oso);
951 sset_destroy(&oso.targets);
954 /* Set IPFIX configuration on 'br'. */
956 bridge_configure_ipfix(struct bridge *br)
958 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
959 const struct ovsrec_flow_sample_collector_set *fe_cfg;
960 struct ofproto_ipfix_bridge_exporter_options be_opts;
961 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
962 size_t n_fe_opts = 0;
964 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
965 if (fe_cfg->bridge == br->cfg) {
970 if (!be_cfg && n_fe_opts == 0) {
971 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
976 memset(&be_opts, 0, sizeof be_opts);
978 sset_init(&be_opts.targets);
979 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
981 if (be_cfg->sampling) {
982 be_opts.sampling_rate = *be_cfg->sampling;
984 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
986 if (be_cfg->obs_domain_id) {
987 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
989 if (be_cfg->obs_point_id) {
990 be_opts.obs_point_id = *be_cfg->obs_point_id;
995 struct ofproto_ipfix_flow_exporter_options *opts;
996 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
998 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
999 if (fe_cfg->bridge == br->cfg) {
1000 opts->collector_set_id = fe_cfg->id;
1001 sset_init(&opts->targets);
1002 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1003 fe_cfg->ipfix->n_targets);
1009 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1013 sset_destroy(&be_opts.targets);
1016 if (n_fe_opts > 0) {
1017 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1019 for (i = 0; i < n_fe_opts; i++) {
1020 sset_destroy(&opts->targets);
1028 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1029 struct ofproto_port_stp_settings *port_s,
1030 int *port_num_counter, unsigned long *port_num_bitmap)
1032 const char *config_str;
1033 struct iface *iface;
1035 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1036 port_s->enable = false;
1039 port_s->enable = true;
1042 /* STP over bonds is not supported. */
1043 if (!list_is_singleton(&port->ifaces)) {
1044 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1046 port_s->enable = false;
1050 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1052 /* Internal ports shouldn't participate in spanning tree, so
1054 if (!strcmp(iface->type, "internal")) {
1055 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1056 port_s->enable = false;
1060 /* STP on mirror output ports is not supported. */
1061 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1062 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1063 port_s->enable = false;
1067 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1069 unsigned long int port_num = strtoul(config_str, NULL, 0);
1070 int port_idx = port_num - 1;
1072 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1073 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1074 port_s->enable = false;
1078 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1079 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1080 port->name, port_num);
1081 port_s->enable = false;
1084 bitmap_set1(port_num_bitmap, port_idx);
1085 port_s->port_num = port_idx;
1087 if (*port_num_counter >= STP_MAX_PORTS) {
1088 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1089 port_s->enable = false;
1093 port_s->port_num = (*port_num_counter)++;
1096 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1098 port_s->path_cost = strtoul(config_str, NULL, 10);
1100 enum netdev_features current;
1103 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1104 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1105 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1108 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1110 port_s->priority = strtoul(config_str, NULL, 0);
1112 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1116 /* Set spanning tree configuration on 'br'. */
1118 bridge_configure_stp(struct bridge *br)
1120 if (!br->cfg->stp_enable) {
1121 ofproto_set_stp(br->ofproto, NULL);
1123 struct ofproto_stp_settings br_s;
1124 const char *config_str;
1126 int port_num_counter;
1127 unsigned long *port_num_bitmap;
1129 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1131 uint8_t ea[ETH_ADDR_LEN];
1133 if (eth_addr_from_string(config_str, ea)) {
1134 br_s.system_id = eth_addr_to_uint64(ea);
1136 br_s.system_id = eth_addr_to_uint64(br->ea);
1137 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1138 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1141 br_s.system_id = eth_addr_to_uint64(br->ea);
1144 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1146 br_s.priority = strtoul(config_str, NULL, 0);
1148 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1151 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1153 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1155 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1158 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1160 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1162 br_s.max_age = STP_DEFAULT_MAX_AGE;
1165 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1167 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1169 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1172 /* Configure STP on the bridge. */
1173 if (ofproto_set_stp(br->ofproto, &br_s)) {
1174 VLOG_ERR("bridge %s: could not enable STP", br->name);
1178 /* Users must either set the port number with the "stp-port-num"
1179 * configuration on all ports or none. If manual configuration
1180 * is not done, then we allocate them sequentially. */
1181 port_num_counter = 0;
1182 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1183 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1184 struct ofproto_port_stp_settings port_s;
1185 struct iface *iface;
1187 port_configure_stp(br->ofproto, port, &port_s,
1188 &port_num_counter, port_num_bitmap);
1190 /* As bonds are not supported, just apply configuration to
1191 * all interfaces. */
1192 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1193 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1195 VLOG_ERR("port %s: could not enable STP", port->name);
1201 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1202 && port_num_counter) {
1203 VLOG_ERR("bridge %s: must manually configure all STP port "
1204 "IDs or none, disabling", br->name);
1205 ofproto_set_stp(br->ofproto, NULL);
1207 bitmap_free(port_num_bitmap);
1212 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1214 const struct port *port = port_lookup(br, name);
1215 return port && port_is_bond_fake_iface(port);
1219 port_is_bond_fake_iface(const struct port *port)
1221 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1225 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1227 struct bridge *br, *next;
1228 struct shash new_br;
1231 /* Collect new bridges' names and types. */
1232 shash_init(&new_br);
1233 for (i = 0; i < cfg->n_bridges; i++) {
1234 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1235 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1237 if (strchr(br_cfg->name, '/')) {
1238 /* Prevent remote ovsdb-server users from accessing arbitrary
1239 * directories, e.g. consider a bridge named "../../../etc/". */
1240 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1242 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1243 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1247 /* Get rid of deleted bridges or those whose types have changed.
1248 * Update 'cfg' of bridges that still exist. */
1249 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1250 br->cfg = shash_find_data(&new_br, br->name);
1251 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1252 br->cfg->datapath_type))) {
1257 /* Add new bridges. */
1258 for (i = 0; i < cfg->n_bridges; i++) {
1259 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1260 struct bridge *br = bridge_lookup(br_cfg->name);
1262 bridge_create(br_cfg);
1266 shash_destroy(&new_br);
1270 iface_set_ofp_port(struct iface *iface, int ofp_port)
1272 struct bridge *br = iface->port->bridge;
1274 ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
1275 iface->ofp_port = ofp_port;
1276 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1277 iface_set_ofport(iface->cfg, ofp_port);
1280 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1281 * Returns 0 if successful, otherwise a positive errno value. */
1283 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1284 struct netdev *netdev)
1288 error = netdev_set_config(netdev, &iface_cfg->options);
1290 VLOG_WARN("could not configure network device %s (%s)",
1291 iface_cfg->name, strerror(error));
1296 /* This function determines whether 'ofproto_port', which is attached to
1297 * br->ofproto's datapath, is one that we want in 'br'.
1299 * If it is, it returns true, after creating an iface (if necessary),
1300 * configuring the iface's netdev according to the iface's options, and setting
1301 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1303 * If, on the other hand, 'port' should be removed, it returns false. The
1304 * caller should later detach the port from br->ofproto. */
1306 bridge_refresh_one_ofp_port(struct bridge *br,
1307 const struct ofproto_port *ofproto_port)
1309 const char *name = ofproto_port->name;
1310 const char *type = ofproto_port->type;
1311 uint16_t ofp_port = ofproto_port->ofp_port;
1313 struct iface *iface = iface_lookup(br, name);
1315 /* Check that the name-to-number mapping is one-to-one. */
1316 if (iface->ofp_port >= 0) {
1317 VLOG_WARN("bridge %s: interface %s reported twice",
1320 } else if (iface_from_ofp_port(br, ofp_port)) {
1321 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1322 br->name, ofp_port);
1326 /* There's a configured interface named 'name'. */
1327 if (strcmp(type, iface->type)
1328 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1329 /* It's the wrong type, or it's the right type but can't be
1330 * configured as the user requested, so we must destroy it. */
1333 /* It's the right type and configured correctly. Keep it. */
1334 iface_set_ofp_port(iface, ofp_port);
1337 } else if (bridge_has_bond_fake_iface(br, name)
1338 && !strcmp(type, "internal")) {
1339 /* It's a bond fake iface. Keep it. */
1342 /* There's no configured interface named 'name', but there might be an
1343 * interface of that name queued to be created.
1345 * If there is, and it has the correct type, then try to configure it
1346 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1347 * delete it and later try to re-add it. */
1348 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1350 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1351 && iface_create(br, if_cfg, ofp_port));
1355 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1356 * consistent with the ofp_ports in "br->ofproto". */
1358 bridge_refresh_ofp_port(struct bridge *br)
1360 struct ofproto_port_dump dump;
1361 struct ofproto_port ofproto_port;
1362 struct port *port, *port_next;
1364 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1365 hmap_clear(&br->ifaces);
1366 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1367 struct iface *iface;
1369 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1370 iface->ofp_port = -1;
1374 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1375 * already exist in the datapath and promote them to full fledged "struct
1376 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1377 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1378 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1379 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1380 garbage->ofp_port = ofproto_port.ofp_port;
1381 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1385 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1386 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1387 * later they can be added to ofproto. */
1388 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1389 struct iface *iface, *iface_next;
1391 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1392 if (iface->ofp_port < 0) {
1393 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1394 iface_destroy(iface);
1398 if (list_is_empty(&port->ifaces)) {
1404 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1405 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1406 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1409 * If successful, returns 0 and stores the network device in '*netdevp'. On
1410 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1412 iface_do_create(const struct bridge *br,
1413 const struct if_cfg *if_cfg,
1414 int *ofp_portp, struct netdev **netdevp)
1416 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1417 const struct ovsrec_port *port_cfg = if_cfg->parent;
1418 struct netdev *netdev = NULL;
1421 if (netdev_is_reserved_name(iface_cfg->name)) {
1422 VLOG_WARN("could not create interface %s, name is reserved",
1428 error = netdev_open(iface_cfg->name,
1429 iface_get_type(iface_cfg, br->cfg), &netdev);
1431 VLOG_WARN("could not open network device %s (%s)",
1432 iface_cfg->name, strerror(error));
1436 error = iface_set_netdev_config(iface_cfg, netdev);
1441 if (*ofp_portp < 0) {
1442 uint16_t ofp_port = if_cfg->ofport;
1444 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1448 *ofp_portp = ofp_port;
1450 VLOG_INFO("bridge %s: added interface %s on port %d",
1451 br->name, iface_cfg->name, *ofp_portp);
1453 VLOG_DBG("bridge %s: interface %s is on port %d",
1454 br->name, iface_cfg->name, *ofp_portp);
1457 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1458 || iface_is_internal(iface_cfg, br->cfg)) {
1459 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1467 netdev_close(netdev);
1471 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1472 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1473 * automatically allocated for the iface. Takes ownership of and
1474 * deallocates 'if_cfg'.
1476 * Return true if an iface is successfully created, false otherwise. */
1478 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1480 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1481 const struct ovsrec_port *port_cfg = if_cfg->parent;
1483 struct netdev *netdev;
1484 struct iface *iface;
1489 /* Do the bits that can fail up front.
1491 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1492 * internal datastructures may not be consistent. Eventually, when port
1493 * additions and deletions are cheaper, these calls should be removed. */
1495 ovs_assert(!iface_lookup(br, iface_cfg->name));
1496 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1499 iface_set_ofport(iface_cfg, -1);
1500 iface_clear_db_record(iface_cfg);
1505 /* Get or create the port structure. */
1506 port = port_lookup(br, port_cfg->name);
1508 port = port_create(br, port_cfg);
1511 /* Create the iface structure. */
1512 iface = xzalloc(sizeof *iface);
1513 list_push_back(&port->ifaces, &iface->port_elem);
1514 hmap_insert(&br->iface_by_name, &iface->name_node,
1515 hash_string(iface_cfg->name, 0));
1517 iface->name = xstrdup(iface_cfg->name);
1518 iface->ofp_port = -1;
1519 iface->netdev = netdev;
1520 iface->type = iface_get_type(iface_cfg, br->cfg);
1521 iface->cfg = iface_cfg;
1523 iface_set_ofp_port(iface, ofp_port);
1525 /* Populate initial status in database. */
1526 iface_refresh_stats(iface);
1527 iface_refresh_status(iface);
1529 /* Add bond fake iface if necessary. */
1530 if (port_is_bond_fake_iface(port)) {
1531 struct ofproto_port ofproto_port;
1533 if (ofproto_port_query_by_name(br->ofproto, port->name,
1535 struct netdev *netdev;
1538 error = netdev_open(port->name, "internal", &netdev);
1540 uint16_t fake_ofp_port = if_cfg->ofport;
1542 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1543 netdev_close(netdev);
1545 VLOG_WARN("could not open network device %s (%s)",
1546 port->name, strerror(error));
1549 /* Already exists, nothing to do. */
1550 ofproto_port_destroy(&ofproto_port);
1555 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1561 /* Set Flow eviction threshold */
1563 bridge_configure_flow_eviction_threshold(struct bridge *br)
1565 const char *threshold_str;
1568 threshold_str = smap_get(&br->cfg->other_config,
1569 "flow-eviction-threshold");
1570 if (threshold_str) {
1571 threshold = strtoul(threshold_str, NULL, 10);
1573 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1575 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1578 /* Set forward BPDU option. */
1580 bridge_configure_forward_bpdu(struct bridge *br)
1582 ofproto_set_forward_bpdu(br->ofproto,
1583 smap_get_bool(&br->cfg->other_config,
1588 /* Set MAC learning table configuration for 'br'. */
1590 bridge_configure_mac_table(struct bridge *br)
1592 const char *idle_time_str;
1595 const char *mac_table_size_str;
1598 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1599 idle_time = (idle_time_str && atoi(idle_time_str)
1600 ? atoi(idle_time_str)
1601 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1603 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1604 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1605 ? atoi(mac_table_size_str)
1608 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1612 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1613 struct iface **hw_addr_iface)
1615 struct hmapx mirror_output_ports;
1618 bool found_addr = false;
1622 *hw_addr_iface = NULL;
1624 /* Did the user request a particular MAC? */
1625 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1626 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1627 if (eth_addr_is_multicast(ea)) {
1628 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1629 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1630 } else if (eth_addr_is_zero(ea)) {
1631 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1637 /* Mirror output ports don't participate in picking the local hardware
1638 * address. ofproto can't help us find out whether a given port is a
1639 * mirror output because we haven't configured mirrors yet, so we need to
1640 * accumulate them ourselves. */
1641 hmapx_init(&mirror_output_ports);
1642 for (i = 0; i < br->cfg->n_mirrors; i++) {
1643 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1644 if (m->output_port) {
1645 hmapx_add(&mirror_output_ports, m->output_port);
1649 /* Otherwise choose the minimum non-local MAC address among all of the
1651 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1652 uint8_t iface_ea[ETH_ADDR_LEN];
1653 struct iface *candidate;
1654 struct iface *iface;
1656 /* Mirror output ports don't participate. */
1657 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1661 /* Choose the MAC address to represent the port. */
1663 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1664 /* Find the interface with this Ethernet address (if any) so that
1665 * we can provide the correct devname to the caller. */
1666 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1667 uint8_t candidate_ea[ETH_ADDR_LEN];
1668 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1669 && eth_addr_equals(iface_ea, candidate_ea)) {
1674 /* Choose the interface whose MAC address will represent the port.
1675 * The Linux kernel bonding code always chooses the MAC address of
1676 * the first slave added to a bond, and the Fedora networking
1677 * scripts always add slaves to a bond in alphabetical order, so
1678 * for compatibility we choose the interface with the name that is
1679 * first in alphabetical order. */
1680 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1681 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1686 /* The local port doesn't count (since we're trying to choose its
1687 * MAC address anyway). */
1688 if (iface->ofp_port == OFPP_LOCAL) {
1693 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1699 /* Compare against our current choice. */
1700 if (!eth_addr_is_multicast(iface_ea) &&
1701 !eth_addr_is_local(iface_ea) &&
1702 !eth_addr_is_reserved(iface_ea) &&
1703 !eth_addr_is_zero(iface_ea) &&
1704 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1706 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1707 *hw_addr_iface = iface;
1713 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1714 *hw_addr_iface = NULL;
1717 hmapx_destroy(&mirror_output_ports);
1720 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1721 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1722 * an interface on 'br', then that interface must be passed in as
1723 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1724 * 'hw_addr_iface' must be passed in as a null pointer. */
1726 bridge_pick_datapath_id(struct bridge *br,
1727 const uint8_t bridge_ea[ETH_ADDR_LEN],
1728 struct iface *hw_addr_iface)
1731 * The procedure for choosing a bridge MAC address will, in the most
1732 * ordinary case, also choose a unique MAC that we can use as a datapath
1733 * ID. In some special cases, though, multiple bridges will end up with
1734 * the same MAC address. This is OK for the bridges, but it will confuse
1735 * the OpenFlow controller, because each datapath needs a unique datapath
1738 * Datapath IDs must be unique. It is also very desirable that they be
1739 * stable from one run to the next, so that policy set on a datapath
1742 const char *datapath_id;
1745 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1746 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1750 if (!hw_addr_iface) {
1752 * A purely internal bridge, that is, one that has no non-virtual
1753 * network devices on it at all, is difficult because it has no
1754 * natural unique identifier at all.
1756 * When the host is a XenServer, we handle this case by hashing the
1757 * host's UUID with the name of the bridge. Names of bridges are
1758 * persistent across XenServer reboots, although they can be reused if
1759 * an internal network is destroyed and then a new one is later
1760 * created, so this is fairly effective.
1762 * When the host is not a XenServer, we punt by using a random MAC
1763 * address on each run.
1765 const char *host_uuid = xenserver_get_host_uuid();
1767 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1768 dpid = dpid_from_hash(combined, strlen(combined));
1774 return eth_addr_to_uint64(bridge_ea);
1778 dpid_from_hash(const void *data, size_t n)
1780 uint8_t hash[SHA1_DIGEST_SIZE];
1782 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1783 sha1_bytes(data, n, hash);
1784 eth_addr_mark_random(hash);
1785 return eth_addr_to_uint64(hash);
1789 iface_refresh_status(struct iface *iface)
1793 enum netdev_features current;
1797 uint8_t mac[ETH_ADDR_LEN];
1800 if (iface_is_synthetic(iface)) {
1806 if (!netdev_get_status(iface->netdev, &smap)) {
1807 ovsrec_interface_set_status(iface->cfg, &smap);
1809 ovsrec_interface_set_status(iface->cfg, NULL);
1812 smap_destroy(&smap);
1814 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1815 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1817 ovsrec_interface_set_duplex(iface->cfg,
1818 netdev_features_is_full_duplex(current)
1820 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1822 ovsrec_interface_set_duplex(iface->cfg, NULL);
1823 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1826 error = netdev_get_mtu(iface->netdev, &mtu);
1829 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1831 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1834 error = netdev_get_etheraddr(iface->netdev, mac);
1836 char mac_string[32];
1838 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1839 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1841 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1845 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1848 iface_refresh_cfm_stats(struct iface *iface)
1850 const struct ovsrec_interface *cfg = iface->cfg;
1851 struct ofproto_cfm_status status;
1853 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1854 iface->ofp_port, &status)) {
1855 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1856 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1857 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1858 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1859 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1861 const char *reasons[CFM_FAULT_N_REASONS];
1862 int64_t cfm_health = status.health;
1863 bool faulted = status.faults != 0;
1866 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1869 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1870 int reason = 1 << i;
1871 if (status.faults & reason) {
1872 reasons[j++] = cfm_fault_reason_to_str(reason);
1875 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1877 if (status.remote_opstate >= 0) {
1878 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1879 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1881 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1884 ovsrec_interface_set_cfm_remote_mpids(cfg,
1885 (const int64_t *)status.rmps,
1887 if (cfm_health >= 0) {
1888 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1890 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1896 iface_refresh_stats(struct iface *iface)
1898 #define IFACE_STATS \
1899 IFACE_STAT(rx_packets, "rx_packets") \
1900 IFACE_STAT(tx_packets, "tx_packets") \
1901 IFACE_STAT(rx_bytes, "rx_bytes") \
1902 IFACE_STAT(tx_bytes, "tx_bytes") \
1903 IFACE_STAT(rx_dropped, "rx_dropped") \
1904 IFACE_STAT(tx_dropped, "tx_dropped") \
1905 IFACE_STAT(rx_errors, "rx_errors") \
1906 IFACE_STAT(tx_errors, "tx_errors") \
1907 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1908 IFACE_STAT(rx_over_errors, "rx_over_err") \
1909 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1910 IFACE_STAT(collisions, "collisions")
1912 #define IFACE_STAT(MEMBER, NAME) + 1
1913 enum { N_IFACE_STATS = IFACE_STATS };
1915 int64_t values[N_IFACE_STATS];
1916 char *keys[N_IFACE_STATS];
1919 struct netdev_stats stats;
1921 if (iface_is_synthetic(iface)) {
1925 /* Intentionally ignore return value, since errors will set 'stats' to
1926 * all-1s, and we will deal with that correctly below. */
1927 netdev_get_stats(iface->netdev, &stats);
1929 /* Copy statistics into keys[] and values[]. */
1931 #define IFACE_STAT(MEMBER, NAME) \
1932 if (stats.MEMBER != UINT64_MAX) { \
1934 values[n] = stats.MEMBER; \
1939 ovs_assert(n <= N_IFACE_STATS);
1941 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1946 br_refresh_stp_status(struct bridge *br)
1948 struct smap smap = SMAP_INITIALIZER(&smap);
1949 struct ofproto *ofproto = br->ofproto;
1950 struct ofproto_stp_status status;
1952 if (ofproto_get_stp_status(ofproto, &status)) {
1956 if (!status.enabled) {
1957 ovsrec_bridge_set_status(br->cfg, NULL);
1961 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1962 STP_ID_ARGS(status.bridge_id));
1963 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1964 STP_ID_ARGS(status.designated_root));
1965 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1967 ovsrec_bridge_set_status(br->cfg, &smap);
1968 smap_destroy(&smap);
1972 port_refresh_stp_status(struct port *port)
1974 struct ofproto *ofproto = port->bridge->ofproto;
1975 struct iface *iface;
1976 struct ofproto_port_stp_status status;
1978 int64_t int_values[3];
1981 if (port_is_synthetic(port)) {
1985 /* STP doesn't currently support bonds. */
1986 if (!list_is_singleton(&port->ifaces)) {
1987 ovsrec_port_set_status(port->cfg, NULL);
1991 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1993 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1997 if (!status.enabled) {
1998 ovsrec_port_set_status(port->cfg, NULL);
1999 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2003 /* Set Status column. */
2005 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2006 smap_add(&smap, "stp_state", stp_state_name(status.state));
2007 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2008 smap_add(&smap, "stp_role", stp_role_name(status.role));
2009 ovsrec_port_set_status(port->cfg, &smap);
2010 smap_destroy(&smap);
2012 /* Set Statistics column. */
2013 keys[0] = "stp_tx_count";
2014 int_values[0] = status.tx_count;
2015 keys[1] = "stp_rx_count";
2016 int_values[1] = status.rx_count;
2017 keys[2] = "stp_error_count";
2018 int_values[2] = status.error_count;
2020 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2021 ARRAY_SIZE(int_values));
2025 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2027 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2031 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2033 bool enable = enable_system_stats(cfg);
2035 system_stats_enable(enable);
2037 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2042 run_system_stats(void)
2044 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2047 stats = system_stats_run();
2049 struct ovsdb_idl_txn *txn;
2050 struct ovsdb_datum datum;
2052 txn = ovsdb_idl_txn_create(idl);
2053 ovsdb_datum_from_smap(&datum, stats);
2054 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2056 ovsdb_idl_txn_commit(txn);
2057 ovsdb_idl_txn_destroy(txn);
2063 static inline const char *
2064 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2067 case OFPCR12_ROLE_EQUAL:
2069 case OFPCR12_ROLE_MASTER:
2071 case OFPCR12_ROLE_SLAVE:
2073 case OFPCR12_ROLE_NOCHANGE:
2075 return "*** INVALID ROLE ***";
2080 refresh_controller_status(void)
2084 const struct ovsrec_controller *cfg;
2088 /* Accumulate status for controllers on all bridges. */
2089 HMAP_FOR_EACH (br, node, &all_bridges) {
2090 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2093 /* Update each controller in the database with current status. */
2094 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2095 struct ofproto_controller_info *cinfo =
2096 shash_find_data(&info, cfg->target);
2099 struct smap smap = SMAP_INITIALIZER(&smap);
2100 const char **values = cinfo->pairs.values;
2101 const char **keys = cinfo->pairs.keys;
2104 for (i = 0; i < cinfo->pairs.n; i++) {
2105 smap_add(&smap, keys[i], values[i]);
2108 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2109 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2111 ovsrec_controller_set_status(cfg, &smap);
2112 smap_destroy(&smap);
2114 ovsrec_controller_set_is_connected(cfg, false);
2115 ovsrec_controller_set_role(cfg, NULL);
2116 ovsrec_controller_set_status(cfg, NULL);
2120 ofproto_free_ofproto_controller_info(&info);
2125 * Some information in the database must be kept as up-to-date as possible to
2126 * allow controllers to respond rapidly to network outages. We call these
2127 * statistics "instant" stats.
2129 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2130 * assuming that they've changed. The only means we have to determine whether
2131 * they have changed are:
2133 * - Try to commit changes to the database. If nothing changed, then
2134 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2137 * - instant_stats_run() is called late in the run loop, after anything that
2138 * might change any of the instant stats.
2140 * We use these two facts together to avoid waking the process up every
2141 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2144 /* Minimum interval between writing updates to the instant stats to the
2146 #define INSTANT_INTERVAL_MSEC 100
2148 /* Current instant stats database transaction, NULL if there is no ongoing
2150 static struct ovsdb_idl_txn *instant_txn;
2152 /* Next time (in msec on monotonic clock) at which we will update the instant
2154 static long long int instant_next_txn = LLONG_MIN;
2156 /* True if the run loop has run since we last saw that the instant stats were
2157 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2158 * to refresh the instant stats. */
2159 static bool instant_stats_could_have_changed;
2162 instant_stats_run(void)
2164 enum ovsdb_idl_txn_status status;
2166 instant_stats_could_have_changed = true;
2171 if (time_msec() < instant_next_txn) {
2174 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2176 instant_txn = ovsdb_idl_txn_create(idl);
2177 HMAP_FOR_EACH (br, node, &all_bridges) {
2178 struct iface *iface;
2181 br_refresh_stp_status(br);
2183 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2184 port_refresh_stp_status(port);
2187 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2188 enum netdev_flags flags;
2190 const char *link_state;
2191 int64_t link_resets;
2194 if (iface_is_synthetic(iface)) {
2198 current = ofproto_port_is_lacp_current(br->ofproto,
2202 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2204 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2207 error = netdev_get_flags(iface->netdev, &flags);
2209 const char *state = flags & NETDEV_UP ? "up" : "down";
2210 ovsrec_interface_set_admin_state(iface->cfg, state);
2212 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2215 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2216 ovsrec_interface_set_link_state(iface->cfg, link_state);
2218 link_resets = netdev_get_carrier_resets(iface->netdev);
2219 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2221 iface_refresh_cfm_stats(iface);
2224 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2226 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2227 smap_destroy(&smap);
2233 status = ovsdb_idl_txn_commit(instant_txn);
2234 if (status != TXN_INCOMPLETE) {
2235 ovsdb_idl_txn_destroy(instant_txn);
2238 if (status == TXN_UNCHANGED) {
2239 instant_stats_could_have_changed = false;
2244 instant_stats_wait(void)
2247 ovsdb_idl_txn_wait(instant_txn);
2248 } else if (instant_stats_could_have_changed) {
2249 poll_timer_wait_until(instant_next_txn);
2253 /* Performs periodic activity required by bridges that needs to be done with
2254 * the least possible latency.
2256 * It makes sense to call this function a couple of times per poll loop, to
2257 * provide a significant performance boost on some benchmarks with ofprotos
2258 * that use the ofproto-dpif implementation. */
2260 bridge_run_fast(void)
2267 ofproto_enumerate_types(&types);
2268 SSET_FOR_EACH (type, &types) {
2269 ofproto_type_run_fast(type);
2271 sset_destroy(&types);
2273 HMAP_FOR_EACH (br, node, &all_bridges) {
2274 ofproto_run_fast(br->ofproto);
2281 static struct ovsrec_open_vswitch null_cfg;
2282 const struct ovsrec_open_vswitch *cfg;
2283 struct ovsdb_idl_txn *reconf_txn = NULL;
2287 bool vlan_splinters_changed;
2290 ovsrec_open_vswitch_init(&null_cfg);
2292 /* (Re)configure if necessary. */
2293 if (!reconfiguring) {
2296 if (ovsdb_idl_is_lock_contended(idl)) {
2297 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2298 struct bridge *br, *next_br;
2300 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2301 "disabling this process (pid %ld) until it goes away",
2302 (long int) getpid());
2304 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2307 /* Since we will not be running system_stats_run() in this process
2308 * with the current situation of multiple ovs-vswitchd daemons,
2309 * disable system stats collection. */
2310 system_stats_enable(false);
2312 } else if (!ovsdb_idl_has_lock(idl)) {
2316 cfg = ovsrec_open_vswitch_first(idl);
2318 /* Initialize the ofproto library. This only needs to run once, but
2319 * it must be done after the configuration is set. If the
2320 * initialization has already occurred, bridge_init_ofproto()
2321 * returns immediately. */
2322 bridge_init_ofproto(cfg);
2324 /* Let each datapath type do the work that it needs to do. */
2326 ofproto_enumerate_types(&types);
2327 SSET_FOR_EACH (type, &types) {
2328 ofproto_type_run(type);
2330 sset_destroy(&types);
2332 /* Let each bridge do the work that it needs to do. */
2333 HMAP_FOR_EACH (br, node, &all_bridges) {
2334 ofproto_run(br->ofproto);
2337 /* Re-configure SSL. We do this on every trip through the main loop,
2338 * instead of just when the database changes, because the contents of the
2339 * key and certificate files can change without the database changing.
2341 * We do this before bridge_reconfigure() because that function might
2342 * initiate SSL connections and thus requires SSL to be configured. */
2343 if (cfg && cfg->ssl) {
2344 const struct ovsrec_ssl *ssl = cfg->ssl;
2346 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2347 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2350 if (!reconfiguring) {
2351 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2352 * usage has changed. */
2353 vlan_splinters_changed = false;
2354 if (vlan_splinters_enabled_anywhere) {
2355 HMAP_FOR_EACH (br, node, &all_bridges) {
2356 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2357 vlan_splinters_changed = true;
2363 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2364 idl_seqno = ovsdb_idl_get_seqno(idl);
2366 reconf_txn = ovsdb_idl_txn_create(idl);
2367 bridge_reconfigure(cfg);
2369 /* We still need to reconfigure to avoid dangling pointers to
2370 * now-destroyed ovsrec structures inside bridge data. */
2371 bridge_reconfigure(&null_cfg);
2376 if (reconfiguring) {
2378 reconf_txn = ovsdb_idl_txn_create(idl);
2381 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2382 reconfiguring = false;
2385 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2388 /* If we are completing our initial configuration for this run
2389 * of ovs-vswitchd, then keep the transaction around to monitor
2390 * it for completion. */
2391 if (!initial_config_done) {
2392 initial_config_done = true;
2393 daemonize_txn = reconf_txn;
2400 ovsdb_idl_txn_commit(reconf_txn);
2401 ovsdb_idl_txn_destroy(reconf_txn);
2405 if (daemonize_txn) {
2406 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2407 if (status != TXN_INCOMPLETE) {
2408 ovsdb_idl_txn_destroy(daemonize_txn);
2409 daemonize_txn = NULL;
2411 /* ovs-vswitchd has completed initialization, so allow the
2412 * process that forked us to exit successfully. */
2413 daemonize_complete();
2415 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2419 /* Refresh interface and mirror stats if necessary. */
2420 if (time_msec() >= iface_stats_timer) {
2422 struct ovsdb_idl_txn *txn;
2424 txn = ovsdb_idl_txn_create(idl);
2425 HMAP_FOR_EACH (br, node, &all_bridges) {
2429 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2430 struct iface *iface;
2432 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2433 iface_refresh_stats(iface);
2434 iface_refresh_status(iface);
2438 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2439 mirror_refresh_stats(m);
2443 refresh_controller_status();
2444 ovsdb_idl_txn_commit(txn);
2445 ovsdb_idl_txn_destroy(txn); /* XXX */
2448 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2452 instant_stats_run();
2461 ovsdb_idl_wait(idl);
2462 if (daemonize_txn) {
2463 ovsdb_idl_txn_wait(daemonize_txn);
2466 if (reconfiguring) {
2467 poll_immediate_wake();
2471 ofproto_enumerate_types(&types);
2472 SSET_FOR_EACH (type, &types) {
2473 ofproto_type_wait(type);
2475 sset_destroy(&types);
2477 if (!hmap_is_empty(&all_bridges)) {
2480 HMAP_FOR_EACH (br, node, &all_bridges) {
2481 ofproto_wait(br->ofproto);
2483 poll_timer_wait_until(iface_stats_timer);
2486 system_stats_wait();
2487 instant_stats_wait();
2490 /* Adds some memory usage statistics for bridges into 'usage', for use with
2491 * memory_report(). */
2493 bridge_get_memory_usage(struct simap *usage)
2497 HMAP_FOR_EACH (br, node, &all_bridges) {
2498 ofproto_get_memory_usage(br->ofproto, usage);
2502 /* QoS unixctl user interface functions. */
2504 struct qos_unixctl_show_cbdata {
2506 struct iface *iface;
2510 qos_unixctl_show_cb(unsigned int queue_id,
2511 const struct smap *details,
2514 struct qos_unixctl_show_cbdata *data = aux;
2515 struct ds *ds = data->ds;
2516 struct iface *iface = data->iface;
2517 struct netdev_queue_stats stats;
2518 struct smap_node *node;
2521 ds_put_cstr(ds, "\n");
2523 ds_put_format(ds, "Queue %u:\n", queue_id);
2525 ds_put_cstr(ds, "Default:\n");
2528 SMAP_FOR_EACH (node, details) {
2529 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2532 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2534 if (stats.tx_packets != UINT64_MAX) {
2535 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2538 if (stats.tx_bytes != UINT64_MAX) {
2539 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2542 if (stats.tx_errors != UINT64_MAX) {
2543 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2546 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2547 queue_id, strerror(error));
2552 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2553 const char *argv[], void *aux OVS_UNUSED)
2555 struct ds ds = DS_EMPTY_INITIALIZER;
2556 struct smap smap = SMAP_INITIALIZER(&smap);
2557 struct iface *iface;
2559 struct smap_node *node;
2560 struct qos_unixctl_show_cbdata data;
2563 iface = iface_find(argv[1]);
2565 unixctl_command_reply_error(conn, "no such interface");
2569 netdev_get_qos(iface->netdev, &type, &smap);
2571 if (*type != '\0') {
2572 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2574 SMAP_FOR_EACH (node, &smap) {
2575 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2580 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2583 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2585 unixctl_command_reply(conn, ds_cstr(&ds));
2587 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2588 unixctl_command_reply_error(conn, ds_cstr(&ds));
2591 smap_destroy(&smap);
2595 /* Bridge reconfiguration functions. */
2597 bridge_create(const struct ovsrec_bridge *br_cfg)
2601 ovs_assert(!bridge_lookup(br_cfg->name));
2602 br = xzalloc(sizeof *br);
2604 br->name = xstrdup(br_cfg->name);
2605 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2608 /* Derive the default Ethernet address from the bridge's UUID. This should
2609 * be unique and it will be stable between ovs-vswitchd runs. */
2610 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2611 eth_addr_mark_random(br->default_ea);
2613 hmap_init(&br->ports);
2614 hmap_init(&br->ifaces);
2615 hmap_init(&br->iface_by_name);
2616 hmap_init(&br->mirrors);
2618 hmap_init(&br->if_cfg_todo);
2619 list_init(&br->ofpp_garbage);
2621 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2625 bridge_destroy(struct bridge *br)
2628 struct mirror *mirror, *next_mirror;
2629 struct port *port, *next_port;
2630 struct if_cfg *if_cfg, *next_if_cfg;
2631 struct ofpp_garbage *garbage, *next_garbage;
2633 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2636 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2637 mirror_destroy(mirror);
2639 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2640 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2643 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2644 &br->ofpp_garbage) {
2645 list_remove(&garbage->list_node);
2649 hmap_remove(&all_bridges, &br->node);
2650 ofproto_destroy(br->ofproto);
2651 hmap_destroy(&br->ifaces);
2652 hmap_destroy(&br->ports);
2653 hmap_destroy(&br->iface_by_name);
2654 hmap_destroy(&br->mirrors);
2655 hmap_destroy(&br->if_cfg_todo);
2662 static struct bridge *
2663 bridge_lookup(const char *name)
2667 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2668 if (!strcmp(br->name, name)) {
2675 /* Handle requests for a listing of all flows known by the OpenFlow
2676 * stack, including those normally hidden. */
2678 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2679 const char *argv[], void *aux OVS_UNUSED)
2684 br = bridge_lookup(argv[1]);
2686 unixctl_command_reply_error(conn, "Unknown bridge");
2691 ofproto_get_all_flows(br->ofproto, &results);
2693 unixctl_command_reply(conn, ds_cstr(&results));
2694 ds_destroy(&results);
2697 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2698 * connections and reconnect. If BRIDGE is not specified, then all bridges
2699 * drop their controller connections and reconnect. */
2701 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2702 const char *argv[], void *aux OVS_UNUSED)
2706 br = bridge_lookup(argv[1]);
2708 unixctl_command_reply_error(conn, "Unknown bridge");
2711 ofproto_reconnect_controllers(br->ofproto);
2713 HMAP_FOR_EACH (br, node, &all_bridges) {
2714 ofproto_reconnect_controllers(br->ofproto);
2717 unixctl_command_reply(conn, NULL);
2721 bridge_get_controllers(const struct bridge *br,
2722 struct ovsrec_controller ***controllersp)
2724 struct ovsrec_controller **controllers;
2725 size_t n_controllers;
2727 controllers = br->cfg->controller;
2728 n_controllers = br->cfg->n_controller;
2730 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2736 *controllersp = controllers;
2738 return n_controllers;
2742 bridge_queue_if_cfg(struct bridge *br,
2743 const struct ovsrec_interface *cfg,
2744 const struct ovsrec_port *parent)
2746 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2749 if_cfg->parent = parent;
2750 if_cfg->ofport = iface_pick_ofport(cfg);
2751 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2752 hash_string(if_cfg->cfg->name, 0));
2755 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2756 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2757 * 'br' needs to complete its configuration. */
2759 bridge_add_del_ports(struct bridge *br,
2760 const unsigned long int *splinter_vlans)
2762 struct shash_node *port_node;
2763 struct port *port, *next;
2764 struct shash new_ports;
2767 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2769 /* Collect new ports. */
2770 shash_init(&new_ports);
2771 for (i = 0; i < br->cfg->n_ports; i++) {
2772 const char *name = br->cfg->ports[i]->name;
2773 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2774 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2778 if (bridge_get_controllers(br, NULL)
2779 && !shash_find(&new_ports, br->name)) {
2780 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2781 br->name, br->name);
2783 ovsrec_interface_init(&br->synth_local_iface);
2784 ovsrec_port_init(&br->synth_local_port);
2786 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2787 br->synth_local_port.n_interfaces = 1;
2788 br->synth_local_port.name = br->name;
2790 br->synth_local_iface.name = br->name;
2791 br->synth_local_iface.type = "internal";
2793 br->synth_local_ifacep = &br->synth_local_iface;
2795 shash_add(&new_ports, br->name, &br->synth_local_port);
2798 if (splinter_vlans) {
2799 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2802 /* Get rid of deleted ports.
2803 * Get rid of deleted interfaces on ports that still exist. */
2804 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2805 port->cfg = shash_find_data(&new_ports, port->name);
2809 port_del_ifaces(port);
2813 /* Update iface->cfg and iface->type in interfaces that still exist.
2814 * Add new interfaces to creation queue. */
2815 SHASH_FOR_EACH (port_node, &new_ports) {
2816 const struct ovsrec_port *port = port_node->data;
2819 for (i = 0; i < port->n_interfaces; i++) {
2820 const struct ovsrec_interface *cfg = port->interfaces[i];
2821 struct iface *iface = iface_lookup(br, cfg->name);
2822 const char *type = iface_get_type(cfg, br->cfg);
2827 } else if (!strcmp(type, "null")) {
2828 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2829 " may be removed in February 2013. Please email"
2830 " dev@openvswitch.org with concerns.",
2833 bridge_queue_if_cfg(br, cfg, port);
2838 shash_destroy(&new_ports);
2841 /* Initializes 'oc' appropriately as a management service controller for
2844 * The caller must free oc->target when it is no longer needed. */
2846 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2847 struct ofproto_controller *oc)
2849 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2850 oc->max_backoff = 0;
2851 oc->probe_interval = 60;
2852 oc->band = OFPROTO_OUT_OF_BAND;
2854 oc->burst_limit = 0;
2855 oc->enable_async_msgs = true;
2858 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2860 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2861 struct ofproto_controller *oc)
2865 oc->target = c->target;
2866 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2867 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2868 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2869 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2870 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2871 oc->burst_limit = (c->controller_burst_limit
2872 ? *c->controller_burst_limit : 0);
2873 oc->enable_async_msgs = (!c->enable_async_messages
2874 || *c->enable_async_messages);
2875 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2876 if (dscp < 0 || dscp > 63) {
2877 dscp = DSCP_DEFAULT;
2882 /* Configures the IP stack for 'br''s local interface properly according to the
2883 * configuration in 'c'. */
2885 bridge_configure_local_iface_netdev(struct bridge *br,
2886 struct ovsrec_controller *c)
2888 struct netdev *netdev;
2889 struct in_addr mask, gateway;
2891 struct iface *local_iface;
2894 /* If there's no local interface or no IP address, give up. */
2895 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2896 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2900 /* Bring up the local interface. */
2901 netdev = local_iface->netdev;
2902 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2904 /* Configure the IP address and netmask. */
2905 if (!c->local_netmask
2906 || !inet_aton(c->local_netmask, &mask)
2908 mask.s_addr = guess_netmask(ip.s_addr);
2910 if (!netdev_set_in4(netdev, ip, mask)) {
2911 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2912 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2915 /* Configure the default gateway. */
2916 if (c->local_gateway
2917 && inet_aton(c->local_gateway, &gateway)
2918 && gateway.s_addr) {
2919 if (!netdev_add_router(netdev, gateway)) {
2920 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2921 br->name, IP_ARGS(gateway.s_addr));
2926 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2927 * in either string are treated as equal to any number of slashes in the other,
2928 * e.g. "x///y" is equal to "x/y".
2930 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2931 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2932 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2933 * 'b' against a prefix of 'a'.
2936 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2938 const char *b_start = b;
2940 if (b - b_start >= b_stoplen) {
2942 } else if (*a != *b) {
2944 } else if (*a == '/') {
2945 a += strspn(a, "/");
2946 b += strspn(b, "/");
2947 } else if (*a == '\0') {
2957 bridge_configure_remotes(struct bridge *br,
2958 const struct sockaddr_in *managers, size_t n_managers)
2960 bool disable_in_band;
2962 struct ovsrec_controller **controllers;
2963 size_t n_controllers;
2965 enum ofproto_fail_mode fail_mode;
2967 struct ofproto_controller *ocs;
2971 /* Check if we should disable in-band control on this bridge. */
2972 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2975 /* Set OpenFlow queue ID for in-band control. */
2976 ofproto_set_in_band_queue(br->ofproto,
2977 smap_get_int(&br->cfg->other_config,
2978 "in-band-queue", -1));
2980 if (disable_in_band) {
2981 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2983 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2986 n_controllers = bridge_get_controllers(br, &controllers);
2988 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2991 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2992 for (i = 0; i < n_controllers; i++) {
2993 struct ovsrec_controller *c = controllers[i];
2995 if (!strncmp(c->target, "punix:", 6)
2996 || !strncmp(c->target, "unix:", 5)) {
2997 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3000 if (!strncmp(c->target, "unix:", 5)) {
3001 /* Connect to a listening socket */
3002 whitelist = xasprintf("unix:%s/", ovs_rundir());
3003 if (strchr(c->target, '/') &&
3004 !equal_pathnames(c->target, whitelist,
3005 strlen(whitelist))) {
3006 /* Absolute path specified, but not in ovs_rundir */
3007 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3008 "controller \"%s\" due to possibility for "
3009 "remote exploit. Instead, specify socket "
3010 "in whitelisted \"%s\" or connect to "
3011 "\"unix:%s/%s.mgmt\" (which is always "
3012 "available without special configuration).",
3013 br->name, c->target, whitelist,
3014 ovs_rundir(), br->name);
3019 whitelist = xasprintf("punix:%s/%s.controller",
3020 ovs_rundir(), br->name);
3021 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3022 /* Prevent remote ovsdb-server users from accessing
3023 * arbitrary Unix domain sockets and overwriting arbitrary
3025 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3026 "controller \"%s\" due to possibility of "
3027 "overwriting local files. Instead, specify "
3028 "whitelisted \"%s\" or connect to "
3029 "\"unix:%s/%s.mgmt\" (which is always "
3030 "available without special configuration).",
3031 br->name, c->target, whitelist,
3032 ovs_rundir(), br->name);
3041 bridge_configure_local_iface_netdev(br, c);
3042 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3043 if (disable_in_band) {
3044 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3049 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3050 bridge_get_allowed_versions(br));
3051 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3054 /* Set the fail-mode. */
3055 fail_mode = !br->cfg->fail_mode
3056 || !strcmp(br->cfg->fail_mode, "standalone")
3057 ? OFPROTO_FAIL_STANDALONE
3058 : OFPROTO_FAIL_SECURE;
3059 ofproto_set_fail_mode(br->ofproto, fail_mode);
3061 /* Configure OpenFlow controller connection snooping. */
3062 if (!ofproto_has_snoops(br->ofproto)) {
3066 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3067 ovs_rundir(), br->name));
3068 ofproto_set_snoops(br->ofproto, &snoops);
3069 sset_destroy(&snoops);
3074 bridge_configure_tables(struct bridge *br)
3076 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3080 n_tables = ofproto_get_n_tables(br->ofproto);
3082 for (i = 0; i < n_tables; i++) {
3083 struct ofproto_table_settings s;
3086 s.max_flows = UINT_MAX;
3090 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3091 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3094 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3095 s.max_flows = *cfg->flow_limit;
3097 if (cfg->overflow_policy
3098 && !strcmp(cfg->overflow_policy, "evict")) {
3101 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3102 for (k = 0; k < cfg->n_groups; k++) {
3103 const char *string = cfg->groups[k];
3106 msg = mf_parse_subfield__(&s.groups[k], &string);
3108 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3109 "'groups' (%s)", br->name, i, msg);
3111 } else if (*string) {
3112 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3113 "element '%s' contains trailing garbage",
3114 br->name, i, cfg->groups[k]);
3122 ofproto_configure_table(br->ofproto, i, &s);
3126 for (; j < br->cfg->n_flow_tables; j++) {
3127 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3128 "%"PRId64" not supported by this datapath", br->name,
3129 br->cfg->key_flow_tables[j]);
3134 bridge_configure_dp_desc(struct bridge *br)
3136 ofproto_set_dp_desc(br->ofproto,
3137 smap_get(&br->cfg->other_config, "dp-desc"));
3140 /* Port functions. */
3142 static struct port *
3143 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3147 port = xzalloc(sizeof *port);
3149 port->name = xstrdup(cfg->name);
3151 list_init(&port->ifaces);
3153 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3157 /* Deletes interfaces from 'port' that are no longer configured for it. */
3159 port_del_ifaces(struct port *port)
3161 struct iface *iface, *next;
3162 struct sset new_ifaces;
3165 /* Collect list of new interfaces. */
3166 sset_init(&new_ifaces);
3167 for (i = 0; i < port->cfg->n_interfaces; i++) {
3168 const char *name = port->cfg->interfaces[i]->name;
3169 const char *type = port->cfg->interfaces[i]->type;
3170 if (strcmp(type, "null")) {
3171 sset_add(&new_ifaces, name);
3175 /* Get rid of deleted interfaces. */
3176 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3177 if (!sset_contains(&new_ifaces, iface->name)) {
3178 iface_destroy(iface);
3182 sset_destroy(&new_ifaces);
3186 port_destroy(struct port *port)
3189 struct bridge *br = port->bridge;
3190 struct iface *iface, *next;
3193 ofproto_bundle_unregister(br->ofproto, port);
3196 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3197 iface_destroy(iface);
3200 hmap_remove(&br->ports, &port->hmap_node);
3206 static struct port *
3207 port_lookup(const struct bridge *br, const char *name)
3211 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3213 if (!strcmp(port->name, name)) {
3221 enable_lacp(struct port *port, bool *activep)
3223 if (!port->cfg->lacp) {
3224 /* XXX when LACP implementation has been sufficiently tested, enable by
3225 * default and make active on bonded ports. */
3227 } else if (!strcmp(port->cfg->lacp, "off")) {
3229 } else if (!strcmp(port->cfg->lacp, "active")) {
3232 } else if (!strcmp(port->cfg->lacp, "passive")) {
3236 VLOG_WARN("port %s: unknown LACP mode %s",
3237 port->name, port->cfg->lacp);
3242 static struct lacp_settings *
3243 port_configure_lacp(struct port *port, struct lacp_settings *s)
3245 const char *lacp_time, *system_id;
3248 if (!enable_lacp(port, &s->active)) {
3252 s->name = port->name;
3254 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3256 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3257 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3258 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3259 " address.", port->name, system_id);
3263 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3266 if (eth_addr_is_zero(s->id)) {
3267 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3271 /* Prefer bondable links if unspecified. */
3272 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3274 s->priority = (priority > 0 && priority <= UINT16_MAX
3276 : UINT16_MAX - !list_is_short(&port->ifaces));
3278 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3279 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3284 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3286 int priority, portid, key;
3288 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3289 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3291 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3293 if (portid <= 0 || portid > UINT16_MAX) {
3294 portid = iface->ofp_port;
3297 if (priority <= 0 || priority > UINT16_MAX) {
3298 priority = UINT16_MAX;
3301 if (key < 0 || key > UINT16_MAX) {
3305 s->name = iface->name;
3307 s->priority = priority;
3312 port_configure_bond(struct port *port, struct bond_settings *s)
3314 const char *detect_s;
3315 struct iface *iface;
3316 int miimon_interval;
3318 s->name = port->name;
3320 if (port->cfg->bond_mode) {
3321 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3322 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3323 port->name, port->cfg->bond_mode,
3324 bond_mode_to_string(s->balance));
3327 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3329 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3330 * active-backup. At some point we should remove this warning. */
3331 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3332 " in previous versions, the default bond_mode was"
3333 " balance-slb", port->name,
3334 bond_mode_to_string(s->balance));
3336 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3337 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3338 "please use another bond type or disable flood_vlans",
3342 miimon_interval = smap_get_int(&port->cfg->other_config,
3343 "bond-miimon-interval", 0);
3344 if (miimon_interval <= 0) {
3345 miimon_interval = 200;
3348 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3349 if (!detect_s || !strcmp(detect_s, "carrier")) {
3350 miimon_interval = 0;
3351 } else if (strcmp(detect_s, "miimon")) {
3352 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3353 "defaulting to carrier", port->name, detect_s);
3354 miimon_interval = 0;
3357 s->up_delay = MAX(0, port->cfg->bond_updelay);
3358 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3359 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3360 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3361 "bond-rebalance-interval", 10000);
3362 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3363 s->rebalance_interval = 1000;
3366 s->fake_iface = port->cfg->bond_fake_iface;
3368 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3369 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3373 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3374 * instead of obtaining it from the database. */
3376 port_is_synthetic(const struct port *port)
3378 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3381 /* Interface functions. */
3384 iface_is_internal(const struct ovsrec_interface *iface,
3385 const struct ovsrec_bridge *br)
3387 /* The local port and "internal" ports are always "internal". */
3388 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3391 /* Returns the correct network device type for interface 'iface' in bridge
3394 iface_get_type(const struct ovsrec_interface *iface,
3395 const struct ovsrec_bridge *br)
3399 /* The local port always has type "internal". Other ports take
3400 * their type from the database and default to "system" if none is
3402 if (iface_is_internal(iface, br)) {
3405 type = iface->type[0] ? iface->type : "system";
3408 return ofproto_port_open_type(br->datapath_type, type);
3412 iface_destroy(struct iface *iface)
3415 struct port *port = iface->port;
3416 struct bridge *br = port->bridge;
3418 if (br->ofproto && iface->ofp_port >= 0) {
3419 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3422 if (iface->ofp_port >= 0) {
3423 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3426 list_remove(&iface->port_elem);
3427 hmap_remove(&br->iface_by_name, &iface->name_node);
3429 netdev_close(iface->netdev);
3436 static struct iface *
3437 iface_lookup(const struct bridge *br, const char *name)
3439 struct iface *iface;
3441 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3442 &br->iface_by_name) {
3443 if (!strcmp(iface->name, name)) {
3451 static struct iface *
3452 iface_find(const char *name)
3454 const struct bridge *br;
3456 HMAP_FOR_EACH (br, node, &all_bridges) {
3457 struct iface *iface = iface_lookup(br, name);
3466 static struct if_cfg *
3467 if_cfg_lookup(const struct bridge *br, const char *name)
3469 struct if_cfg *if_cfg;
3471 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3473 if (!strcmp(if_cfg->cfg->name, name)) {
3481 static struct iface *
3482 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3484 struct iface *iface;
3486 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3487 hash_int(ofp_port, 0), &br->ifaces) {
3488 if (iface->ofp_port == ofp_port) {
3495 /* Set Ethernet address of 'iface', if one is specified in the configuration
3498 iface_set_mac(struct iface *iface)
3500 uint8_t ea[ETH_ADDR_LEN];
3502 if (!strcmp(iface->type, "internal")
3503 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3504 if (iface->ofp_port == OFPP_LOCAL) {
3505 VLOG_ERR("interface %s: ignoring mac in Interface record "
3506 "(use Bridge record to set local port's mac)",
3508 } else if (eth_addr_is_multicast(ea)) {
3509 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3512 int error = netdev_set_etheraddr(iface->netdev, ea);
3514 VLOG_ERR("interface %s: setting MAC failed (%s)",
3515 iface->name, strerror(error));
3521 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3523 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3525 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3526 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3530 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3531 * sets the "ofport" field to -1.
3533 * This is appropriate when 'if_cfg''s interface cannot be created or is
3534 * otherwise invalid. */
3536 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3538 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3539 ovsrec_interface_set_status(if_cfg, NULL);
3540 ovsrec_interface_set_admin_state(if_cfg, NULL);
3541 ovsrec_interface_set_duplex(if_cfg, NULL);
3542 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3543 ovsrec_interface_set_link_state(if_cfg, NULL);
3544 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3545 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3546 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3547 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3548 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3549 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3550 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3554 struct iface_delete_queues_cbdata {
3555 struct netdev *netdev;
3556 const struct ovsdb_datum *queues;
3560 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3562 union ovsdb_atom atom;
3564 atom.integer = target;
3565 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3569 iface_delete_queues(unsigned int queue_id,
3570 const struct smap *details OVS_UNUSED, void *cbdata_)
3572 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3574 if (!queue_ids_include(cbdata->queues, queue_id)) {
3575 netdev_delete_queue(cbdata->netdev, queue_id);
3580 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3582 struct ofpbuf queues_buf;
3584 ofpbuf_init(&queues_buf, 0);
3586 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3587 netdev_set_qos(iface->netdev, NULL, NULL);
3589 struct iface_delete_queues_cbdata cbdata;
3593 /* Configure top-level Qos for 'iface'. */
3594 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3596 /* Deconfigure queues that were deleted. */
3597 cbdata.netdev = iface->netdev;
3598 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3600 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3602 /* Configure queues for 'iface'. */
3604 for (i = 0; i < qos->n_queues; i++) {
3605 const struct ovsrec_queue *queue = qos->value_queues[i];
3606 unsigned int queue_id = qos->key_queues[i];
3608 if (queue_id == 0) {
3612 if (queue->n_dscp == 1) {
3613 struct ofproto_port_queue *port_queue;
3615 port_queue = ofpbuf_put_uninit(&queues_buf,
3616 sizeof *port_queue);
3617 port_queue->queue = queue_id;
3618 port_queue->dscp = queue->dscp[0];
3621 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3624 struct smap details;
3626 smap_init(&details);
3627 netdev_set_queue(iface->netdev, 0, &details);
3628 smap_destroy(&details);
3632 if (iface->ofp_port >= 0) {
3633 const struct ofproto_port_queue *port_queues = queues_buf.data;
3634 size_t n_queues = queues_buf.size / sizeof *port_queues;
3636 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3637 port_queues, n_queues);
3640 netdev_set_policing(iface->netdev,
3641 iface->cfg->ingress_policing_rate,
3642 iface->cfg->ingress_policing_burst);
3644 ofpbuf_uninit(&queues_buf);
3648 iface_configure_cfm(struct iface *iface)
3650 const struct ovsrec_interface *cfg = iface->cfg;
3651 const char *opstate_str;
3652 const char *cfm_ccm_vlan;
3653 struct cfm_settings s;
3654 struct smap netdev_args;
3656 if (!cfg->n_cfm_mpid) {
3657 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3661 s.check_tnl_key = false;
3662 smap_init(&netdev_args);
3663 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3664 const char *key = smap_get(&netdev_args, "key");
3665 const char *in_key = smap_get(&netdev_args, "in_key");
3667 s.check_tnl_key = (key && !strcmp(key, "flow"))
3668 || (in_key && !strcmp(in_key, "flow"));
3670 smap_destroy(&netdev_args);
3672 s.mpid = *cfg->cfm_mpid;
3673 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3674 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3675 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3677 if (s.interval <= 0) {
3681 if (!cfm_ccm_vlan) {
3683 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3684 s.ccm_vlan = CFM_RANDOM_VLAN;
3686 s.ccm_vlan = atoi(cfm_ccm_vlan);
3687 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3692 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3695 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3696 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3698 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3701 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3702 * instead of obtaining it from the database. */
3704 iface_is_synthetic(const struct iface *iface)
3706 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3710 iface_pick_ofport(const struct ovsrec_interface *cfg)
3712 int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3713 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3717 /* Port mirroring. */
3719 static struct mirror *
3720 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3724 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3725 if (uuid_equals(uuid, &m->uuid)) {
3733 bridge_configure_mirrors(struct bridge *br)
3735 const struct ovsdb_datum *mc;
3736 unsigned long *flood_vlans;
3737 struct mirror *m, *next;
3740 /* Get rid of deleted mirrors. */
3741 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3742 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3743 union ovsdb_atom atom;
3745 atom.uuid = m->uuid;
3746 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3751 /* Add new mirrors and reconfigure existing ones. */
3752 for (i = 0; i < br->cfg->n_mirrors; i++) {
3753 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3754 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3756 m = mirror_create(br, cfg);
3759 if (!mirror_configure(m)) {
3764 /* Update flooded vlans (for RSPAN). */
3765 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3766 br->cfg->n_flood_vlans);
3767 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3768 bitmap_free(flood_vlans);
3771 static struct mirror *
3772 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3776 m = xzalloc(sizeof *m);
3777 m->uuid = cfg->header_.uuid;
3778 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3780 m->name = xstrdup(cfg->name);
3786 mirror_destroy(struct mirror *m)
3789 struct bridge *br = m->bridge;
3792 ofproto_mirror_unregister(br->ofproto, m);
3795 hmap_remove(&br->mirrors, &m->hmap_node);
3802 mirror_collect_ports(struct mirror *m,
3803 struct ovsrec_port **in_ports, int n_in_ports,
3804 void ***out_portsp, size_t *n_out_portsp)
3806 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3807 size_t n_out_ports = 0;
3810 for (i = 0; i < n_in_ports; i++) {
3811 const char *name = in_ports[i]->name;
3812 struct port *port = port_lookup(m->bridge, name);
3814 out_ports[n_out_ports++] = port;
3816 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3817 "port %s", m->bridge->name, m->name, name);
3820 *out_portsp = out_ports;
3821 *n_out_portsp = n_out_ports;
3825 mirror_configure(struct mirror *m)
3827 const struct ovsrec_mirror *cfg = m->cfg;
3828 struct ofproto_mirror_settings s;
3831 if (strcmp(cfg->name, m->name)) {
3833 m->name = xstrdup(cfg->name);
3837 /* Get output port or VLAN. */
3838 if (cfg->output_port) {
3839 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3840 if (!s.out_bundle) {
3841 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3842 m->bridge->name, m->name);
3845 s.out_vlan = UINT16_MAX;
3847 if (cfg->output_vlan) {
3848 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3849 "output vlan; ignoring output vlan",
3850 m->bridge->name, m->name);
3852 } else if (cfg->output_vlan) {
3853 /* The database should prevent invalid VLAN values. */
3854 s.out_bundle = NULL;
3855 s.out_vlan = *cfg->output_vlan;
3857 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3858 m->bridge->name, m->name);
3862 /* Get port selection. */
3863 if (cfg->select_all) {
3864 size_t n_ports = hmap_count(&m->bridge->ports);
3865 void **ports = xmalloc(n_ports * sizeof *ports);
3870 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3880 /* Get ports, dropping ports that don't exist.
3881 * The IDL ensures that there are no duplicates. */
3882 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3883 &s.srcs, &s.n_srcs);
3884 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3885 &s.dsts, &s.n_dsts);
3888 /* Get VLAN selection. */
3889 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3892 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3895 if (s.srcs != s.dsts) {
3904 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3906 * This is deprecated. It is only for compatibility with broken device drivers
3907 * in old versions of Linux that do not properly support VLANs when VLAN
3908 * devices are not used. When broken device drivers are no longer in
3909 * widespread use, we will delete these interfaces. */
3911 static struct ovsrec_port **recs;
3912 static size_t n_recs, allocated_recs;
3914 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3915 * splinters are reconfigured. */
3917 register_rec(struct ovsrec_port *rec)
3919 if (n_recs >= allocated_recs) {
3920 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3922 recs[n_recs++] = rec;
3925 /* Frees all of the ports registered with register_reg(). */
3927 free_registered_recs(void)
3931 for (i = 0; i < n_recs; i++) {
3932 struct ovsrec_port *port = recs[i];
3935 for (j = 0; j < port->n_interfaces; j++) {
3936 struct ovsrec_interface *iface = port->interfaces[j];
3941 smap_destroy(&port->other_config);
3942 free(port->interfaces);
3950 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3953 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3955 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3959 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3962 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3963 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3964 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3967 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3968 * use, returns NULL.
3970 * Updates 'vlan_splinters_enabled_anywhere'. */
3971 static unsigned long int *
3972 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3974 unsigned long int *splinter_vlans;
3975 struct sset splinter_ifaces;
3976 const char *real_dev_name;
3977 struct shash *real_devs;
3978 struct shash_node *node;
3982 /* Free space allocated for synthesized ports and interfaces, since we're
3983 * in the process of reconstructing all of them. */
3984 free_registered_recs();
3986 splinter_vlans = bitmap_allocate(4096);
3987 sset_init(&splinter_ifaces);
3988 vlan_splinters_enabled_anywhere = false;
3989 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3990 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3993 for (j = 0; j < br_cfg->n_ports; j++) {
3994 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3997 for (k = 0; k < port_cfg->n_interfaces; k++) {
3998 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4000 if (vlan_splinters_is_enabled(iface_cfg)) {
4001 vlan_splinters_enabled_anywhere = true;
4002 sset_add(&splinter_ifaces, iface_cfg->name);
4003 vlan_bitmap_from_array__(port_cfg->trunks,
4009 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4010 bitmap_set1(splinter_vlans, *port_cfg->tag);
4015 if (!vlan_splinters_enabled_anywhere) {
4016 free(splinter_vlans);
4017 sset_destroy(&splinter_ifaces);
4021 HMAP_FOR_EACH (br, node, &all_bridges) {
4023 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4027 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4028 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4029 * device to be created for it. */
4030 bitmap_set0(splinter_vlans, 0);
4031 bitmap_set0(splinter_vlans, 4095);
4033 /* Delete all VLAN devices that we don't need. */
4035 real_devs = vlandev_get_real_devs();
4036 SHASH_FOR_EACH (node, real_devs) {
4037 const struct vlan_real_dev *real_dev = node->data;
4038 const struct vlan_dev *vlan_dev;
4039 bool real_dev_has_splinters;
4041 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4043 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4044 if (!real_dev_has_splinters
4045 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4046 struct netdev *netdev;
4048 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4049 if (!netdev_get_in4(netdev, NULL, NULL) ||
4050 !netdev_get_in6(netdev, NULL)) {
4051 vlandev_del(vlan_dev->name);
4053 /* It has an IP address configured, so we don't own
4054 * it. Don't delete it. */
4056 netdev_close(netdev);
4063 /* Add all VLAN devices that we need. */
4064 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4067 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4068 if (!vlandev_get_name(real_dev_name, vid)) {
4069 vlandev_add(real_dev_name, vid);
4076 sset_destroy(&splinter_ifaces);
4078 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4079 free(splinter_vlans);
4082 return splinter_vlans;
4085 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4088 configure_splinter_port(struct port *port)
4090 struct ofproto *ofproto = port->bridge->ofproto;
4091 uint16_t realdev_ofp_port;
4092 const char *realdev_name;
4093 struct iface *vlandev, *realdev;
4095 ofproto_bundle_unregister(port->bridge->ofproto, port);
4097 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4100 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4101 realdev = iface_lookup(port->bridge, realdev_name);
4102 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4104 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4108 static struct ovsrec_port *
4109 synthesize_splinter_port(const char *real_dev_name,
4110 const char *vlan_dev_name, int vid)
4112 struct ovsrec_interface *iface;
4113 struct ovsrec_port *port;
4115 iface = xmalloc(sizeof *iface);
4116 ovsrec_interface_init(iface);
4117 iface->name = xstrdup(vlan_dev_name);
4118 iface->type = "system";
4120 port = xmalloc(sizeof *port);
4121 ovsrec_port_init(port);
4122 port->interfaces = xmemdup(&iface, sizeof iface);
4123 port->n_interfaces = 1;
4124 port->name = xstrdup(vlan_dev_name);
4125 port->vlan_mode = "splinter";
4126 port->tag = xmalloc(sizeof *port->tag);
4129 smap_add(&port->other_config, "realdev", real_dev_name);
4135 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4136 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4137 * 1-bit in the 'splinter_vlans' bitmap. */
4139 add_vlan_splinter_ports(struct bridge *br,
4140 const unsigned long int *splinter_vlans,
4141 struct shash *ports)
4145 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4146 * we're modifying 'ports'. */
4147 for (i = 0; i < br->cfg->n_ports; i++) {
4148 const char *name = br->cfg->ports[i]->name;
4149 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4152 for (j = 0; j < port_cfg->n_interfaces; j++) {
4153 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4155 if (vlan_splinters_is_enabled(iface_cfg)) {
4156 const char *real_dev_name;
4159 real_dev_name = iface_cfg->name;
4160 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4161 const char *vlan_dev_name;
4163 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4165 && !shash_find(ports, vlan_dev_name)) {
4166 shash_add(ports, vlan_dev_name,
4167 synthesize_splinter_port(
4168 real_dev_name, vlan_dev_name, vid));
4177 mirror_refresh_stats(struct mirror *m)
4179 struct ofproto *ofproto = m->bridge->ofproto;
4180 uint64_t tx_packets, tx_bytes;
4183 size_t stat_cnt = 0;
4185 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4186 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4190 if (tx_packets != UINT64_MAX) {
4191 keys[stat_cnt] = "tx_packets";
4192 values[stat_cnt] = tx_packets;
4195 if (tx_bytes != UINT64_MAX) {
4196 keys[stat_cnt] = "tx_bytes";
4197 values[stat_cnt] = tx_bytes;
4201 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);