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.
21 #include "async-append.h"
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
41 #include "ofproto/bond.h"
42 #include "ofproto/ofproto.h"
43 #include "poll-loop.h"
47 #include "socket-util.h"
49 #include "stream-ssl.h"
51 #include "system-stats.h"
56 #include "lib/vswitch-idl.h"
57 #include "xenserver.h"
59 #include "sflow_api.h"
60 #include "vlan-bitmap.h"
62 VLOG_DEFINE_THIS_MODULE(bridge);
64 COVERAGE_DEFINE(bridge_reconfigure);
67 /* These members are always valid.
69 * They are immutable: they never change between iface_create() and
71 struct list port_elem; /* Element in struct port's "ifaces" list. */
72 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
73 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
74 struct port *port; /* Containing port. */
75 char *name; /* Host network device name. */
76 struct netdev *netdev; /* Network device. */
77 ofp_port_t ofp_port; /* OpenFlow port number. */
79 /* These members are valid only within bridge_reconfigure(). */
80 const char *type; /* Usually same as cfg->type. */
81 const struct ovsrec_interface *cfg;
85 struct uuid uuid; /* UUID of this "mirror" record in database. */
86 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
87 struct bridge *bridge;
89 const struct ovsrec_mirror *cfg;
93 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
94 struct bridge *bridge;
97 const struct ovsrec_port *cfg;
99 /* An ordinary bridge port has 1 interface.
100 * A bridge port for bonding has at least 2 interfaces. */
101 struct list ifaces; /* List of "struct iface"s. */
105 struct hmap_node node; /* In 'all_bridges'. */
106 char *name; /* User-specified arbitrary name. */
107 char *type; /* Datapath type. */
108 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
109 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
110 const struct ovsrec_bridge *cfg;
112 /* OpenFlow switch processing. */
113 struct ofproto *ofproto; /* OpenFlow switch. */
116 struct hmap ports; /* "struct port"s indexed by name. */
117 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
118 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
120 /* Port mirroring. */
121 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
123 /* Used during reconfiguration. */
124 struct shash wanted_ports;
126 /* Synthetic local port if necessary. */
127 struct ovsrec_port synth_local_port;
128 struct ovsrec_interface synth_local_iface;
129 struct ovsrec_interface *synth_local_ifacep;
132 /* All bridges, indexed by name. */
133 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
135 /* OVSDB IDL used to obtain configuration. */
136 static struct ovsdb_idl *idl;
138 /* We want to complete daemonization, fully detaching from our parent process,
139 * only after we have completed our initial configuration, committed our state
140 * to the database, and received confirmation back from the database server
141 * that it applied the commit. This allows our parent process to know that,
142 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
143 * to have already been filled in. (It is only "very likely" rather than
144 * certain because there is always a slim possibility that the transaction will
145 * fail or that some other client has added new bridges, ports, etc. while
146 * ovs-vswitchd was configuring using an old configuration.)
148 * We only need to do this once for our initial configuration at startup, so
149 * 'initial_config_done' tracks whether we've already done it. While we are
150 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
151 * itself and is otherwise NULL. */
152 static bool initial_config_done;
153 static struct ovsdb_idl_txn *daemonize_txn;
155 /* Most recently processed IDL sequence number. */
156 static unsigned int idl_seqno;
158 /* Each time this timer expires, the bridge fetches interface and mirror
159 * statistics and pushes them into the database. */
160 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
161 static long long int iface_stats_timer = LLONG_MIN;
163 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
164 * expensive. This can cause bridge_reconfigure() to take a long time during
165 * which no other work can be done. To deal with this problem, we limit port
166 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
167 * call to bridge_reconfigure(). If there is more work to do after the limit
168 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
169 * This allows the rest of the code to catch up on important things like
170 * forwarding packets. */
171 #define OFP_PORT_ACTION_WINDOW 10
173 static void add_del_bridges(const struct ovsrec_open_vswitch *);
174 static void bridge_create(const struct ovsrec_bridge *);
175 static void bridge_destroy(struct bridge *);
176 static struct bridge *bridge_lookup(const char *name);
177 static unixctl_cb_func bridge_unixctl_dump_flows;
178 static unixctl_cb_func bridge_unixctl_reconnect;
179 static size_t bridge_get_controllers(const struct bridge *br,
180 struct ovsrec_controller ***controllersp);
181 static void bridge_collect_wanted_ports(struct bridge *,
182 const unsigned long *splinter_vlans,
183 struct shash *wanted_ports);
184 static void bridge_delete_ofprotos(void);
185 static void bridge_delete_or_reconfigure_ports(struct bridge *);
186 static void bridge_del_ports(struct bridge *,
187 const struct shash *wanted_ports);
188 static void bridge_add_ports(struct bridge *,
189 const struct shash *wanted_ports);
191 static void bridge_configure_flow_miss_model(const char *opt);
192 static void bridge_configure_datapath_id(struct bridge *);
193 static void bridge_configure_netflow(struct bridge *);
194 static void bridge_configure_forward_bpdu(struct bridge *);
195 static void bridge_configure_mac_table(struct bridge *);
196 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
197 static void bridge_configure_ipfix(struct bridge *);
198 static void bridge_configure_stp(struct bridge *);
199 static void bridge_configure_tables(struct bridge *);
200 static void bridge_configure_dp_desc(struct bridge *);
201 static void bridge_configure_remotes(struct bridge *,
202 const struct sockaddr_in *managers,
204 static void bridge_pick_local_hw_addr(struct bridge *,
205 uint8_t ea[ETH_ADDR_LEN],
206 struct iface **hw_addr_iface);
207 static uint64_t bridge_pick_datapath_id(struct bridge *,
208 const uint8_t bridge_ea[ETH_ADDR_LEN],
209 struct iface *hw_addr_iface);
210 static uint64_t dpid_from_hash(const void *, size_t nbytes);
211 static bool bridge_has_bond_fake_iface(const struct bridge *,
213 static bool port_is_bond_fake_iface(const struct port *);
215 static unixctl_cb_func qos_unixctl_show;
217 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
218 static void port_del_ifaces(struct port *);
219 static void port_destroy(struct port *);
220 static struct port *port_lookup(const struct bridge *, const char *name);
221 static void port_configure(struct port *);
222 static struct lacp_settings *port_configure_lacp(struct port *,
223 struct lacp_settings *);
224 static void port_configure_bond(struct port *, struct bond_settings *);
225 static bool port_is_synthetic(const struct port *);
227 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
228 static void run_system_stats(void);
230 static void bridge_configure_mirrors(struct bridge *);
231 static struct mirror *mirror_create(struct bridge *,
232 const struct ovsrec_mirror *);
233 static void mirror_destroy(struct mirror *);
234 static bool mirror_configure(struct mirror *);
235 static void mirror_refresh_stats(struct mirror *);
237 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
238 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
239 const struct ovsrec_port *);
240 static bool iface_is_internal(const struct ovsrec_interface *iface,
241 const struct ovsrec_bridge *br);
242 static const char *iface_get_type(const struct ovsrec_interface *,
243 const struct ovsrec_bridge *);
244 static void iface_destroy(struct iface *);
245 static struct iface *iface_lookup(const struct bridge *, const char *name);
246 static struct iface *iface_find(const char *name);
247 static struct iface *iface_from_ofp_port(const struct bridge *,
248 ofp_port_t ofp_port);
249 static void iface_set_mac(struct iface *);
250 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
251 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
252 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
253 static void iface_configure_cfm(struct iface *);
254 static void iface_refresh_cfm_stats(struct iface *);
255 static void iface_refresh_stats(struct iface *);
256 static void iface_refresh_status(struct iface *);
257 static bool iface_is_synthetic(const struct iface *);
258 static ofp_port_t iface_get_requested_ofp_port(
259 const struct ovsrec_interface *);
260 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
262 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
264 * This is deprecated. It is only for compatibility with broken device drivers
265 * in old versions of Linux that do not properly support VLANs when VLAN
266 * devices are not used. When broken device drivers are no longer in
267 * widespread use, we will delete these interfaces. */
269 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
270 static bool vlan_splinters_enabled_anywhere;
272 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
273 static unsigned long int *collect_splinter_vlans(
274 const struct ovsrec_open_vswitch *);
275 static void configure_splinter_port(struct port *);
276 static void add_vlan_splinter_ports(struct bridge *,
277 const unsigned long int *splinter_vlans,
278 struct shash *ports);
281 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
283 struct shash iface_hints;
284 static bool initialized = false;
291 shash_init(&iface_hints);
294 for (i = 0; i < cfg->n_bridges; i++) {
295 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
298 for (j = 0; j < br_cfg->n_ports; j++) {
299 struct ovsrec_port *port_cfg = br_cfg->ports[j];
302 for (k = 0; k < port_cfg->n_interfaces; k++) {
303 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
304 struct iface_hint *iface_hint;
306 iface_hint = xmalloc(sizeof *iface_hint);
307 iface_hint->br_name = br_cfg->name;
308 iface_hint->br_type = br_cfg->datapath_type;
309 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
311 shash_add(&iface_hints, if_cfg->name, iface_hint);
317 ofproto_init(&iface_hints);
319 shash_destroy_free_data(&iface_hints);
323 /* Public functions. */
325 /* Initializes the bridge module, configuring it to obtain its configuration
326 * from an OVSDB server accessed over 'remote', which should be a string in a
327 * form acceptable to ovsdb_idl_create(). */
329 bridge_init(const char *remote)
331 /* Create connection to database. */
332 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
333 idl_seqno = ovsdb_idl_get_seqno(idl);
334 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
335 ovsdb_idl_verify_write_only(idl);
337 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
338 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
339 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
340 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
341 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
342 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
343 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
345 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
346 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
347 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
349 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
350 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
351 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
352 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
354 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
355 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
356 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
357 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
358 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
359 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
373 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
375 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
376 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
377 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
378 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
380 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
382 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
384 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
385 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
387 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
394 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
395 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
396 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
398 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
400 /* Register unixctl commands. */
401 unixctl_command_register("qos/show", "interface", 1, 1,
402 qos_unixctl_show, NULL);
403 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
404 bridge_unixctl_dump_flows, NULL);
405 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
406 bridge_unixctl_reconnect, NULL);
416 struct bridge *br, *next_br;
418 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
421 ovsdb_idl_destroy(idl);
424 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
425 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
426 * responsible for freeing '*managersp' (with free()).
428 * You may be asking yourself "why does ovs-vswitchd care?", because
429 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
430 * should not be and in fact is not directly involved in that. But
431 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
432 * it has to tell in-band control where the managers are to enable that.
433 * (Thus, only managers connected in-band are collected.)
436 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
437 struct sockaddr_in **managersp, size_t *n_managersp)
439 struct sockaddr_in *managers = NULL;
440 size_t n_managers = 0;
444 /* Collect all of the potential targets from the "targets" columns of the
445 * rows pointed to by "manager_options", excluding any that are
448 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
449 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
451 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
452 sset_find_and_delete(&targets, m->target);
454 sset_add(&targets, m->target);
458 /* Now extract the targets' IP addresses. */
459 if (!sset_is_empty(&targets)) {
462 managers = xmalloc(sset_count(&targets) * sizeof *managers);
463 SSET_FOR_EACH (target, &targets) {
464 struct sockaddr_in *sin = &managers[n_managers];
466 if (stream_parse_target_with_default_port(target,
473 sset_destroy(&targets);
475 *managersp = managers;
476 *n_managersp = n_managers;
480 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
482 unsigned long int *splinter_vlans;
483 struct sockaddr_in *managers;
484 struct bridge *br, *next;
485 int sflow_bridge_number;
488 COVERAGE_INC(bridge_reconfigure);
490 ofproto_set_flow_eviction_threshold(
491 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
492 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
494 ofproto_set_n_handler_threads(
495 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0));
497 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
498 "force-miss-model"));
500 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
501 * to 'ovs_cfg', with only very minimal configuration otherwise.
503 * This is mostly an update to bridge data structures. Nothing is pushed
504 * down to ofproto or lower layers. */
505 add_del_bridges(ovs_cfg);
506 splinter_vlans = collect_splinter_vlans(ovs_cfg);
507 HMAP_FOR_EACH (br, node, &all_bridges) {
508 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
509 bridge_del_ports(br, &br->wanted_ports);
511 free(splinter_vlans);
513 /* Start pushing configuration changes down to the ofproto layer:
515 * - Delete ofprotos that are no longer configured.
517 * - Delete ports that are no longer configured.
519 * - Reconfigure existing ports to their desired configurations, or
520 * delete them if not possible.
522 * We have to do all the deletions before we can do any additions, because
523 * the ports to be added might require resources that will be freed up by
524 * deletions (they might especially overlap in name). */
525 bridge_delete_ofprotos();
526 HMAP_FOR_EACH (br, node, &all_bridges) {
528 bridge_delete_or_reconfigure_ports(br);
532 /* Finish pushing configuration changes to the ofproto layer:
534 * - Create ofprotos that are missing.
536 * - Add ports that are missing. */
537 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
541 error = ofproto_create(br->name, br->type, &br->ofproto);
543 VLOG_ERR("failed to create bridge %s: %s", br->name,
544 ovs_strerror(error));
545 shash_destroy(&br->wanted_ports);
550 HMAP_FOR_EACH (br, node, &all_bridges) {
551 bridge_add_ports(br, &br->wanted_ports);
552 shash_destroy(&br->wanted_ports);
555 reconfigure_system_stats(ovs_cfg);
557 /* Complete the configuration. */
558 sflow_bridge_number = 0;
559 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
560 HMAP_FOR_EACH (br, node, &all_bridges) {
563 /* We need the datapath ID early to allow LACP ports to use it as the
564 * default system ID. */
565 bridge_configure_datapath_id(br);
567 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
570 port_configure(port);
572 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
573 iface_configure_cfm(iface);
574 iface_configure_qos(iface, port->cfg->qos);
575 iface_set_mac(iface);
576 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
580 bridge_configure_mirrors(br);
581 bridge_configure_forward_bpdu(br);
582 bridge_configure_mac_table(br);
583 bridge_configure_remotes(br, managers, n_managers);
584 bridge_configure_netflow(br);
585 bridge_configure_sflow(br, &sflow_bridge_number);
586 bridge_configure_ipfix(br);
587 bridge_configure_stp(br);
588 bridge_configure_tables(br);
589 bridge_configure_dp_desc(br);
591 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
592 /* XXX: Remove this warning message eventually. */
593 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
594 " moved to the Open_vSwitch table. Ignoring its"
595 " setting in the bridge table.");
601 /* Delete ofprotos which aren't configured or have the wrong type. Create
602 * ofprotos which don't exist but need to. */
604 bridge_delete_ofprotos(void)
611 /* Delete ofprotos with no bridge or with the wrong type. */
614 ofproto_enumerate_types(&types);
615 SSET_FOR_EACH (type, &types) {
618 ofproto_enumerate_names(type, &names);
619 SSET_FOR_EACH (name, &names) {
620 br = bridge_lookup(name);
621 if (!br || strcmp(type, br->type)) {
622 ofproto_delete(name, type);
626 sset_destroy(&names);
627 sset_destroy(&types);
631 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
633 if (*n >= *allocated) {
634 ports = x2nrealloc(ports, allocated, sizeof *ports);
636 ports[(*n)++] = port;
641 bridge_delete_or_reconfigure_ports(struct bridge *br)
643 struct ofproto_port ofproto_port;
644 struct ofproto_port_dump dump;
646 /* List of "ofp_port"s to delete. We make a list instead of deleting them
647 * right away because ofproto implementations aren't necessarily able to
648 * iterate through a changing list of ports in an entirely robust way. */
656 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
657 ofp_port_t requested_ofp_port;
660 iface = iface_lookup(br, ofproto_port.name);
662 /* No such iface is configured, so we should delete this
665 * As a corner case exception, keep the port if it's a bond fake
667 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
668 && !strcmp(ofproto_port.type, "internal")) {
674 if (strcmp(ofproto_port.type, iface->type)
675 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
676 /* The interface is the wrong type or can't be configured.
681 /* If the requested OpenFlow port for 'iface' changed, and it's not
682 * already the correct port, then we might want to temporarily delete
683 * this interface, so we can add it back again with the new OpenFlow
685 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
686 if (iface->ofp_port != OFPP_LOCAL &&
687 requested_ofp_port != OFPP_NONE &&
688 requested_ofp_port != iface->ofp_port) {
689 ofp_port_t victim_request;
690 struct iface *victim;
692 /* Check for an existing OpenFlow port currently occupying
693 * 'iface''s requested port number. If there isn't one, then
694 * delete this port. Otherwise we need to consider further. */
695 victim = iface_from_ofp_port(br, requested_ofp_port);
700 /* 'victim' is a port currently using 'iface''s requested port
701 * number. Unless 'victim' specifically requested that port
702 * number, too, then we can delete both 'iface' and 'victim'
703 * temporarily. (We'll add both of them back again later with new
704 * OpenFlow port numbers.)
706 * If 'victim' did request port number 'requested_ofp_port', just
707 * like 'iface', then that's a configuration inconsistency that we
708 * can't resolve. We might as well let it keep its current port
710 victim_request = iface_get_requested_ofp_port(victim->cfg);
711 if (victim_request != requested_ofp_port) {
712 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
713 iface_destroy(victim);
722 iface_destroy(iface);
723 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
726 for (i = 0; i < n; i++) {
727 ofproto_port_del(br->ofproto, del[i]);
733 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
734 bool with_requested_port)
736 struct shash_node *port_node;
738 SHASH_FOR_EACH (port_node, wanted_ports) {
739 const struct ovsrec_port *port_cfg = port_node->data;
742 for (i = 0; i < port_cfg->n_interfaces; i++) {
743 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
744 ofp_port_t requested_ofp_port;
746 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
747 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
748 struct iface *iface = iface_lookup(br, iface_cfg->name);
751 iface_create(br, iface_cfg, port_cfg);
759 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
761 /* First add interfaces that request a particular port number. */
762 bridge_add_ports__(br, wanted_ports, true);
764 /* Then add interfaces that want automatic port number assignment.
765 * We add these afterward to avoid accidentally taking a specifically
766 * requested port number. */
767 bridge_add_ports__(br, wanted_ports, false);
771 port_configure(struct port *port)
773 const struct ovsrec_port *cfg = port->cfg;
774 struct bond_settings bond_settings;
775 struct lacp_settings lacp_settings;
776 struct ofproto_bundle_settings s;
779 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
780 configure_splinter_port(port);
789 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
790 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
791 s.slaves[s.n_slaves++] = iface->ofp_port;
796 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
800 /* Get VLAN trunks. */
803 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
807 if (cfg->vlan_mode) {
808 if (!strcmp(cfg->vlan_mode, "access")) {
809 s.vlan_mode = PORT_VLAN_ACCESS;
810 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
811 s.vlan_mode = PORT_VLAN_TRUNK;
812 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
813 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
814 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
815 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
817 /* This "can't happen" because ovsdb-server should prevent it. */
818 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
819 s.vlan_mode = PORT_VLAN_TRUNK;
823 s.vlan_mode = PORT_VLAN_ACCESS;
825 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
829 s.vlan_mode = PORT_VLAN_TRUNK;
832 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
835 /* Get LACP settings. */
836 s.lacp = port_configure_lacp(port, &lacp_settings);
840 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
841 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
842 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
845 s.lacp_slaves = NULL;
848 /* Get bond settings. */
849 if (s.n_slaves > 1) {
850 s.bond = &bond_settings;
851 port_configure_bond(port, &bond_settings);
854 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
855 netdev_set_miimon_interval(iface->netdev, 0);
860 ofproto_bundle_register(port->bridge->ofproto, port, &s);
869 bridge_configure_flow_miss_model(const char *opt)
871 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
874 if (strcmp(opt, "with-facets")) {
875 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
876 } else if (strcmp(opt, "without-facets")) {
877 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
881 ofproto_set_flow_miss_model(model);
884 /* Pick local port hardware address and datapath ID for 'br'. */
886 bridge_configure_datapath_id(struct bridge *br)
888 uint8_t ea[ETH_ADDR_LEN];
890 struct iface *local_iface;
891 struct iface *hw_addr_iface;
894 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
895 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
897 int error = netdev_set_etheraddr(local_iface->netdev, ea);
899 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
900 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
901 "Ethernet address: %s",
902 br->name, ovs_strerror(error));
905 memcpy(br->ea, ea, ETH_ADDR_LEN);
907 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
908 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
909 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
910 ofproto_set_datapath_id(br->ofproto, dpid);
913 dpid_string = xasprintf("%016"PRIx64, dpid);
914 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
918 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
921 bridge_get_allowed_versions(struct bridge *br)
923 if (!br->cfg->n_protocols)
926 return ofputil_versions_from_strings(br->cfg->protocols,
927 br->cfg->n_protocols);
930 /* Set NetFlow configuration on 'br'. */
932 bridge_configure_netflow(struct bridge *br)
934 struct ovsrec_netflow *cfg = br->cfg->netflow;
935 struct netflow_options opts;
938 ofproto_set_netflow(br->ofproto, NULL);
942 memset(&opts, 0, sizeof opts);
944 /* Get default NetFlow configuration from datapath.
945 * Apply overrides from 'cfg'. */
946 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
947 if (cfg->engine_type) {
948 opts.engine_type = *cfg->engine_type;
950 if (cfg->engine_id) {
951 opts.engine_id = *cfg->engine_id;
954 /* Configure active timeout interval. */
955 opts.active_timeout = cfg->active_timeout;
956 if (!opts.active_timeout) {
957 opts.active_timeout = -1;
958 } else if (opts.active_timeout < 0) {
959 VLOG_WARN("bridge %s: active timeout interval set to negative "
960 "value, using default instead (%d seconds)", br->name,
961 NF_ACTIVE_TIMEOUT_DEFAULT);
962 opts.active_timeout = -1;
965 /* Add engine ID to interface number to disambiguate bridgs? */
966 opts.add_id_to_iface = cfg->add_id_to_interface;
967 if (opts.add_id_to_iface) {
968 if (opts.engine_id > 0x7f) {
969 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
970 "another vswitch, choose an engine id less than 128",
973 if (hmap_count(&br->ports) > 508) {
974 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
975 "another port when more than 508 ports are used",
981 sset_init(&opts.collectors);
982 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
985 if (ofproto_set_netflow(br->ofproto, &opts)) {
986 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
988 sset_destroy(&opts.collectors);
991 /* Set sFlow configuration on 'br'. */
993 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
995 const struct ovsrec_sflow *cfg = br->cfg->sflow;
996 struct ovsrec_controller **controllers;
997 struct ofproto_sflow_options oso;
998 size_t n_controllers;
1002 ofproto_set_sflow(br->ofproto, NULL);
1006 memset(&oso, 0, sizeof oso);
1008 sset_init(&oso.targets);
1009 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1011 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1012 if (cfg->sampling) {
1013 oso.sampling_rate = *cfg->sampling;
1016 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1018 oso.polling_interval = *cfg->polling;
1021 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1023 oso.header_len = *cfg->header;
1026 oso.sub_id = (*sflow_bridge_number)++;
1027 oso.agent_device = cfg->agent;
1029 oso.control_ip = NULL;
1030 n_controllers = bridge_get_controllers(br, &controllers);
1031 for (i = 0; i < n_controllers; i++) {
1032 if (controllers[i]->local_ip) {
1033 oso.control_ip = controllers[i]->local_ip;
1037 ofproto_set_sflow(br->ofproto, &oso);
1039 sset_destroy(&oso.targets);
1042 /* Returns whether a IPFIX row is valid. */
1044 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1046 return ipfix && ipfix->n_targets > 0;
1049 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1051 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1052 const struct bridge *br)
1054 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1057 /* Set IPFIX configuration on 'br'. */
1059 bridge_configure_ipfix(struct bridge *br)
1061 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1062 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1063 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1064 struct ofproto_ipfix_bridge_exporter_options be_opts;
1065 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1066 size_t n_fe_opts = 0;
1068 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1069 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1074 if (!valid_be_cfg && n_fe_opts == 0) {
1075 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1080 memset(&be_opts, 0, sizeof be_opts);
1082 sset_init(&be_opts.targets);
1083 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1085 if (be_cfg->sampling) {
1086 be_opts.sampling_rate = *be_cfg->sampling;
1088 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1090 if (be_cfg->obs_domain_id) {
1091 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1093 if (be_cfg->obs_point_id) {
1094 be_opts.obs_point_id = *be_cfg->obs_point_id;
1096 if (be_cfg->cache_active_timeout) {
1097 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1099 if (be_cfg->cache_max_flows) {
1100 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1104 if (n_fe_opts > 0) {
1105 struct ofproto_ipfix_flow_exporter_options *opts;
1106 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1108 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1109 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1110 opts->collector_set_id = fe_cfg->id;
1111 sset_init(&opts->targets);
1112 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1113 fe_cfg->ipfix->n_targets);
1114 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1115 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1116 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1117 ? *fe_cfg->ipfix->cache_max_flows : 0;
1123 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1127 sset_destroy(&be_opts.targets);
1130 if (n_fe_opts > 0) {
1131 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1133 for (i = 0; i < n_fe_opts; i++) {
1134 sset_destroy(&opts->targets);
1142 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1143 struct ofproto_port_stp_settings *port_s,
1144 int *port_num_counter, unsigned long *port_num_bitmap)
1146 const char *config_str;
1147 struct iface *iface;
1149 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1150 port_s->enable = false;
1153 port_s->enable = true;
1156 /* STP over bonds is not supported. */
1157 if (!list_is_singleton(&port->ifaces)) {
1158 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1160 port_s->enable = false;
1164 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1166 /* Internal ports shouldn't participate in spanning tree, so
1168 if (!strcmp(iface->type, "internal")) {
1169 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1170 port_s->enable = false;
1174 /* STP on mirror output ports is not supported. */
1175 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1176 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1177 port_s->enable = false;
1181 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1183 unsigned long int port_num = strtoul(config_str, NULL, 0);
1184 int port_idx = port_num - 1;
1186 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1187 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1188 port_s->enable = false;
1192 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1193 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1194 port->name, port_num);
1195 port_s->enable = false;
1198 bitmap_set1(port_num_bitmap, port_idx);
1199 port_s->port_num = port_idx;
1201 if (*port_num_counter >= STP_MAX_PORTS) {
1202 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1203 port_s->enable = false;
1207 port_s->port_num = (*port_num_counter)++;
1210 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1212 port_s->path_cost = strtoul(config_str, NULL, 10);
1214 enum netdev_features current;
1217 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1218 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1219 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1222 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1224 port_s->priority = strtoul(config_str, NULL, 0);
1226 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1230 /* Set spanning tree configuration on 'br'. */
1232 bridge_configure_stp(struct bridge *br)
1234 if (!br->cfg->stp_enable) {
1235 ofproto_set_stp(br->ofproto, NULL);
1237 struct ofproto_stp_settings br_s;
1238 const char *config_str;
1240 int port_num_counter;
1241 unsigned long *port_num_bitmap;
1243 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1245 uint8_t ea[ETH_ADDR_LEN];
1247 if (eth_addr_from_string(config_str, ea)) {
1248 br_s.system_id = eth_addr_to_uint64(ea);
1250 br_s.system_id = eth_addr_to_uint64(br->ea);
1251 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1252 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1255 br_s.system_id = eth_addr_to_uint64(br->ea);
1258 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1260 br_s.priority = strtoul(config_str, NULL, 0);
1262 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1265 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1267 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1269 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1272 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1274 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1276 br_s.max_age = STP_DEFAULT_MAX_AGE;
1279 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1281 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1283 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1286 /* Configure STP on the bridge. */
1287 if (ofproto_set_stp(br->ofproto, &br_s)) {
1288 VLOG_ERR("bridge %s: could not enable STP", br->name);
1292 /* Users must either set the port number with the "stp-port-num"
1293 * configuration on all ports or none. If manual configuration
1294 * is not done, then we allocate them sequentially. */
1295 port_num_counter = 0;
1296 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1297 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1298 struct ofproto_port_stp_settings port_s;
1299 struct iface *iface;
1301 port_configure_stp(br->ofproto, port, &port_s,
1302 &port_num_counter, port_num_bitmap);
1304 /* As bonds are not supported, just apply configuration to
1305 * all interfaces. */
1306 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1307 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1309 VLOG_ERR("port %s: could not enable STP", port->name);
1315 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1316 && port_num_counter) {
1317 VLOG_ERR("bridge %s: must manually configure all STP port "
1318 "IDs or none, disabling", br->name);
1319 ofproto_set_stp(br->ofproto, NULL);
1321 bitmap_free(port_num_bitmap);
1326 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1328 const struct port *port = port_lookup(br, name);
1329 return port && port_is_bond_fake_iface(port);
1333 port_is_bond_fake_iface(const struct port *port)
1335 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1339 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1341 struct bridge *br, *next;
1342 struct shash new_br;
1345 /* Collect new bridges' names and types. */
1346 shash_init(&new_br);
1347 for (i = 0; i < cfg->n_bridges; i++) {
1348 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1349 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1351 if (strchr(br_cfg->name, '/')) {
1352 /* Prevent remote ovsdb-server users from accessing arbitrary
1353 * directories, e.g. consider a bridge named "../../../etc/". */
1354 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1356 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1357 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1361 /* Get rid of deleted bridges or those whose types have changed.
1362 * Update 'cfg' of bridges that still exist. */
1363 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1364 br->cfg = shash_find_data(&new_br, br->name);
1365 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1366 br->cfg->datapath_type))) {
1371 /* Add new bridges. */
1372 for (i = 0; i < cfg->n_bridges; i++) {
1373 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1374 struct bridge *br = bridge_lookup(br_cfg->name);
1376 bridge_create(br_cfg);
1380 shash_destroy(&new_br);
1383 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1384 * Returns 0 if successful, otherwise a positive errno value. */
1386 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1387 struct netdev *netdev)
1389 return netdev_set_config(netdev, &iface_cfg->options);
1392 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1393 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1394 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1397 * If successful, returns 0 and stores the network device in '*netdevp'. On
1398 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1400 iface_do_create(const struct bridge *br,
1401 const struct ovsrec_interface *iface_cfg,
1402 const struct ovsrec_port *port_cfg,
1403 ofp_port_t *ofp_portp, struct netdev **netdevp)
1405 struct netdev *netdev = NULL;
1408 if (netdev_is_reserved_name(iface_cfg->name)) {
1409 VLOG_WARN("could not create interface %s, name is reserved",
1415 error = netdev_open(iface_cfg->name,
1416 iface_get_type(iface_cfg, br->cfg), &netdev);
1418 VLOG_WARN("could not open network device %s (%s)",
1419 iface_cfg->name, ovs_strerror(error));
1423 error = iface_set_netdev_config(iface_cfg, netdev);
1428 *ofp_portp = iface_pick_ofport(iface_cfg);
1429 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1434 VLOG_INFO("bridge %s: added interface %s on port %d",
1435 br->name, iface_cfg->name, *ofp_portp);
1437 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1438 || iface_is_internal(iface_cfg, br->cfg)) {
1439 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1447 netdev_close(netdev);
1451 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1452 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1453 * automatically allocated for the iface. Takes ownership of and
1454 * deallocates 'if_cfg'.
1456 * Return true if an iface is successfully created, false otherwise. */
1458 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1459 const struct ovsrec_port *port_cfg)
1461 struct netdev *netdev;
1462 struct iface *iface;
1463 ofp_port_t ofp_port;
1467 /* Do the bits that can fail up front.
1469 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1470 * internal datastructures may not be consistent. Eventually, when port
1471 * additions and deletions are cheaper, these calls should be removed. */
1473 ovs_assert(!iface_lookup(br, iface_cfg->name));
1474 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1477 iface_set_ofport(iface_cfg, OFPP_NONE);
1478 iface_clear_db_record(iface_cfg);
1482 /* Get or create the port structure. */
1483 port = port_lookup(br, port_cfg->name);
1485 port = port_create(br, port_cfg);
1488 /* Create the iface structure. */
1489 iface = xzalloc(sizeof *iface);
1490 list_push_back(&port->ifaces, &iface->port_elem);
1491 hmap_insert(&br->iface_by_name, &iface->name_node,
1492 hash_string(iface_cfg->name, 0));
1494 iface->name = xstrdup(iface_cfg->name);
1495 iface->ofp_port = ofp_port;
1496 iface->netdev = netdev;
1497 iface->type = iface_get_type(iface_cfg, br->cfg);
1498 iface->cfg = iface_cfg;
1499 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1500 hash_ofp_port(ofp_port));
1502 iface_set_ofport(iface->cfg, ofp_port);
1504 /* Populate initial status in database. */
1505 iface_refresh_stats(iface);
1506 iface_refresh_status(iface);
1508 /* Add bond fake iface if necessary. */
1509 if (port_is_bond_fake_iface(port)) {
1510 struct ofproto_port ofproto_port;
1512 if (ofproto_port_query_by_name(br->ofproto, port->name,
1514 struct netdev *netdev;
1517 error = netdev_open(port->name, "internal", &netdev);
1519 ofp_port_t fake_ofp_port = OFPP_NONE;
1520 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1521 netdev_close(netdev);
1523 VLOG_WARN("could not open network device %s (%s)",
1524 port->name, ovs_strerror(error));
1527 /* Already exists, nothing to do. */
1528 ofproto_port_destroy(&ofproto_port);
1535 /* Set forward BPDU option. */
1537 bridge_configure_forward_bpdu(struct bridge *br)
1539 ofproto_set_forward_bpdu(br->ofproto,
1540 smap_get_bool(&br->cfg->other_config,
1545 /* Set MAC learning table configuration for 'br'. */
1547 bridge_configure_mac_table(struct bridge *br)
1549 const char *idle_time_str;
1552 const char *mac_table_size_str;
1555 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1556 idle_time = (idle_time_str && atoi(idle_time_str)
1557 ? atoi(idle_time_str)
1558 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1560 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1561 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1562 ? atoi(mac_table_size_str)
1565 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1569 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1570 struct iface **hw_addr_iface)
1572 struct hmapx mirror_output_ports;
1575 bool found_addr = false;
1579 *hw_addr_iface = NULL;
1581 /* Did the user request a particular MAC? */
1582 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1583 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1584 if (eth_addr_is_multicast(ea)) {
1585 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1586 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1587 } else if (eth_addr_is_zero(ea)) {
1588 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1594 /* Mirror output ports don't participate in picking the local hardware
1595 * address. ofproto can't help us find out whether a given port is a
1596 * mirror output because we haven't configured mirrors yet, so we need to
1597 * accumulate them ourselves. */
1598 hmapx_init(&mirror_output_ports);
1599 for (i = 0; i < br->cfg->n_mirrors; i++) {
1600 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1601 if (m->output_port) {
1602 hmapx_add(&mirror_output_ports, m->output_port);
1606 /* Otherwise choose the minimum non-local MAC address among all of the
1608 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1609 uint8_t iface_ea[ETH_ADDR_LEN];
1610 struct iface *candidate;
1611 struct iface *iface;
1613 /* Mirror output ports don't participate. */
1614 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1618 /* Choose the MAC address to represent the port. */
1620 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1621 /* Find the interface with this Ethernet address (if any) so that
1622 * we can provide the correct devname to the caller. */
1623 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1624 uint8_t candidate_ea[ETH_ADDR_LEN];
1625 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1626 && eth_addr_equals(iface_ea, candidate_ea)) {
1631 /* Choose the interface whose MAC address will represent the port.
1632 * The Linux kernel bonding code always chooses the MAC address of
1633 * the first slave added to a bond, and the Fedora networking
1634 * scripts always add slaves to a bond in alphabetical order, so
1635 * for compatibility we choose the interface with the name that is
1636 * first in alphabetical order. */
1637 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1638 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1643 /* The local port doesn't count (since we're trying to choose its
1644 * MAC address anyway). */
1645 if (iface->ofp_port == OFPP_LOCAL) {
1650 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1656 /* Compare against our current choice. */
1657 if (!eth_addr_is_multicast(iface_ea) &&
1658 !eth_addr_is_local(iface_ea) &&
1659 !eth_addr_is_reserved(iface_ea) &&
1660 !eth_addr_is_zero(iface_ea) &&
1661 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1663 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1664 *hw_addr_iface = iface;
1670 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1671 *hw_addr_iface = NULL;
1674 hmapx_destroy(&mirror_output_ports);
1677 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1678 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1679 * an interface on 'br', then that interface must be passed in as
1680 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1681 * 'hw_addr_iface' must be passed in as a null pointer. */
1683 bridge_pick_datapath_id(struct bridge *br,
1684 const uint8_t bridge_ea[ETH_ADDR_LEN],
1685 struct iface *hw_addr_iface)
1688 * The procedure for choosing a bridge MAC address will, in the most
1689 * ordinary case, also choose a unique MAC that we can use as a datapath
1690 * ID. In some special cases, though, multiple bridges will end up with
1691 * the same MAC address. This is OK for the bridges, but it will confuse
1692 * the OpenFlow controller, because each datapath needs a unique datapath
1695 * Datapath IDs must be unique. It is also very desirable that they be
1696 * stable from one run to the next, so that policy set on a datapath
1699 const char *datapath_id;
1702 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1703 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1707 if (!hw_addr_iface) {
1709 * A purely internal bridge, that is, one that has no non-virtual
1710 * network devices on it at all, is difficult because it has no
1711 * natural unique identifier at all.
1713 * When the host is a XenServer, we handle this case by hashing the
1714 * host's UUID with the name of the bridge. Names of bridges are
1715 * persistent across XenServer reboots, although they can be reused if
1716 * an internal network is destroyed and then a new one is later
1717 * created, so this is fairly effective.
1719 * When the host is not a XenServer, we punt by using a random MAC
1720 * address on each run.
1722 const char *host_uuid = xenserver_get_host_uuid();
1724 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1725 dpid = dpid_from_hash(combined, strlen(combined));
1731 return eth_addr_to_uint64(bridge_ea);
1735 dpid_from_hash(const void *data, size_t n)
1737 uint8_t hash[SHA1_DIGEST_SIZE];
1739 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1740 sha1_bytes(data, n, hash);
1741 eth_addr_mark_random(hash);
1742 return eth_addr_to_uint64(hash);
1746 iface_refresh_status(struct iface *iface)
1750 enum netdev_features current;
1754 uint8_t mac[ETH_ADDR_LEN];
1758 if (iface_is_synthetic(iface)) {
1764 if (!netdev_get_status(iface->netdev, &smap)) {
1765 ovsrec_interface_set_status(iface->cfg, &smap);
1767 ovsrec_interface_set_status(iface->cfg, NULL);
1770 smap_destroy(&smap);
1772 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1773 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1775 ovsrec_interface_set_duplex(iface->cfg,
1776 netdev_features_is_full_duplex(current)
1778 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1780 ovsrec_interface_set_duplex(iface->cfg, NULL);
1781 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1784 error = netdev_get_mtu(iface->netdev, &mtu);
1787 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1789 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1792 error = netdev_get_etheraddr(iface->netdev, mac);
1794 char mac_string[32];
1796 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1797 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1799 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1802 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1803 * if there is no valid ifindex number. */
1804 ifindex64 = netdev_get_ifindex(iface->netdev);
1805 if (ifindex64 < 0) {
1808 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1811 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1814 iface_refresh_cfm_stats(struct iface *iface)
1816 const struct ovsrec_interface *cfg = iface->cfg;
1817 struct ofproto_cfm_status status;
1819 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1820 iface->ofp_port, &status)) {
1821 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1822 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1823 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1824 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1825 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1826 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1828 const char *reasons[CFM_FAULT_N_REASONS];
1829 int64_t cfm_health = status.health;
1830 int64_t cfm_flap_count = status.flap_count;
1831 bool faulted = status.faults != 0;
1834 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1837 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1838 int reason = 1 << i;
1839 if (status.faults & reason) {
1840 reasons[j++] = cfm_fault_reason_to_str(reason);
1843 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1845 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1847 if (status.remote_opstate >= 0) {
1848 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1849 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1851 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1854 ovsrec_interface_set_cfm_remote_mpids(cfg,
1855 (const int64_t *)status.rmps,
1857 if (cfm_health >= 0) {
1858 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1860 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1868 iface_refresh_stats(struct iface *iface)
1870 #define IFACE_STATS \
1871 IFACE_STAT(rx_packets, "rx_packets") \
1872 IFACE_STAT(tx_packets, "tx_packets") \
1873 IFACE_STAT(rx_bytes, "rx_bytes") \
1874 IFACE_STAT(tx_bytes, "tx_bytes") \
1875 IFACE_STAT(rx_dropped, "rx_dropped") \
1876 IFACE_STAT(tx_dropped, "tx_dropped") \
1877 IFACE_STAT(rx_errors, "rx_errors") \
1878 IFACE_STAT(tx_errors, "tx_errors") \
1879 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1880 IFACE_STAT(rx_over_errors, "rx_over_err") \
1881 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1882 IFACE_STAT(collisions, "collisions")
1884 #define IFACE_STAT(MEMBER, NAME) + 1
1885 enum { N_IFACE_STATS = IFACE_STATS };
1887 int64_t values[N_IFACE_STATS];
1888 char *keys[N_IFACE_STATS];
1891 struct netdev_stats stats;
1893 if (iface_is_synthetic(iface)) {
1897 /* Intentionally ignore return value, since errors will set 'stats' to
1898 * all-1s, and we will deal with that correctly below. */
1899 netdev_get_stats(iface->netdev, &stats);
1901 /* Copy statistics into keys[] and values[]. */
1903 #define IFACE_STAT(MEMBER, NAME) \
1904 if (stats.MEMBER != UINT64_MAX) { \
1906 values[n] = stats.MEMBER; \
1911 ovs_assert(n <= N_IFACE_STATS);
1913 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1918 br_refresh_stp_status(struct bridge *br)
1920 struct smap smap = SMAP_INITIALIZER(&smap);
1921 struct ofproto *ofproto = br->ofproto;
1922 struct ofproto_stp_status status;
1924 if (ofproto_get_stp_status(ofproto, &status)) {
1928 if (!status.enabled) {
1929 ovsrec_bridge_set_status(br->cfg, NULL);
1933 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1934 STP_ID_ARGS(status.bridge_id));
1935 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1936 STP_ID_ARGS(status.designated_root));
1937 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1939 ovsrec_bridge_set_status(br->cfg, &smap);
1940 smap_destroy(&smap);
1944 port_refresh_stp_status(struct port *port)
1946 struct ofproto *ofproto = port->bridge->ofproto;
1947 struct iface *iface;
1948 struct ofproto_port_stp_status status;
1951 if (port_is_synthetic(port)) {
1955 /* STP doesn't currently support bonds. */
1956 if (!list_is_singleton(&port->ifaces)) {
1957 ovsrec_port_set_status(port->cfg, NULL);
1961 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1962 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1966 if (!status.enabled) {
1967 ovsrec_port_set_status(port->cfg, NULL);
1971 /* Set Status column. */
1973 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1974 smap_add(&smap, "stp_state", stp_state_name(status.state));
1975 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1976 smap_add(&smap, "stp_role", stp_role_name(status.role));
1977 ovsrec_port_set_status(port->cfg, &smap);
1978 smap_destroy(&smap);
1982 port_refresh_stp_stats(struct port *port)
1984 struct ofproto *ofproto = port->bridge->ofproto;
1985 struct iface *iface;
1986 struct ofproto_port_stp_stats stats;
1988 int64_t int_values[3];
1990 if (port_is_synthetic(port)) {
1994 /* STP doesn't currently support bonds. */
1995 if (!list_is_singleton(&port->ifaces)) {
1999 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2000 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2004 if (!stats.enabled) {
2005 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2009 /* Set Statistics column. */
2010 keys[0] = "stp_tx_count";
2011 int_values[0] = stats.tx_count;
2012 keys[1] = "stp_rx_count";
2013 int_values[1] = stats.rx_count;
2014 keys[2] = "stp_error_count";
2015 int_values[2] = stats.error_count;
2017 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2018 ARRAY_SIZE(int_values));
2022 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2024 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2028 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2030 bool enable = enable_system_stats(cfg);
2032 system_stats_enable(enable);
2034 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2039 run_system_stats(void)
2041 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2044 stats = system_stats_run();
2046 struct ovsdb_idl_txn *txn;
2047 struct ovsdb_datum datum;
2049 txn = ovsdb_idl_txn_create(idl);
2050 ovsdb_datum_from_smap(&datum, stats);
2051 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2053 ovsdb_idl_txn_commit(txn);
2054 ovsdb_idl_txn_destroy(txn);
2060 static inline const char *
2061 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2064 case OFPCR12_ROLE_EQUAL:
2066 case OFPCR12_ROLE_MASTER:
2068 case OFPCR12_ROLE_SLAVE:
2070 case OFPCR12_ROLE_NOCHANGE:
2072 return "*** INVALID ROLE ***";
2077 refresh_controller_status(void)
2081 const struct ovsrec_controller *cfg;
2085 /* Accumulate status for controllers on all bridges. */
2086 HMAP_FOR_EACH (br, node, &all_bridges) {
2087 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2090 /* Update each controller in the database with current status. */
2091 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2092 struct ofproto_controller_info *cinfo =
2093 shash_find_data(&info, cfg->target);
2096 struct smap smap = SMAP_INITIALIZER(&smap);
2097 const char **values = cinfo->pairs.values;
2098 const char **keys = cinfo->pairs.keys;
2101 for (i = 0; i < cinfo->pairs.n; i++) {
2102 smap_add(&smap, keys[i], values[i]);
2105 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2106 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2108 ovsrec_controller_set_status(cfg, &smap);
2109 smap_destroy(&smap);
2111 ovsrec_controller_set_is_connected(cfg, false);
2112 ovsrec_controller_set_role(cfg, NULL);
2113 ovsrec_controller_set_status(cfg, NULL);
2117 ofproto_free_ofproto_controller_info(&info);
2122 * Some information in the database must be kept as up-to-date as possible to
2123 * allow controllers to respond rapidly to network outages. We call these
2124 * statistics "instant" stats.
2126 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2127 * assuming that they've changed. The only means we have to determine whether
2128 * they have changed are:
2130 * - Try to commit changes to the database. If nothing changed, then
2131 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2134 * - instant_stats_run() is called late in the run loop, after anything that
2135 * might change any of the instant stats.
2137 * We use these two facts together to avoid waking the process up every
2138 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2141 /* Minimum interval between writing updates to the instant stats to the
2143 #define INSTANT_INTERVAL_MSEC 100
2145 /* Current instant stats database transaction, NULL if there is no ongoing
2147 static struct ovsdb_idl_txn *instant_txn;
2149 /* Next time (in msec on monotonic clock) at which we will update the instant
2151 static long long int instant_next_txn = LLONG_MIN;
2153 /* True if the run loop has run since we last saw that the instant stats were
2154 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2155 * to refresh the instant stats. */
2156 static bool instant_stats_could_have_changed;
2159 instant_stats_run(void)
2161 enum ovsdb_idl_txn_status status;
2163 instant_stats_could_have_changed = true;
2168 if (time_msec() < instant_next_txn) {
2171 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2173 instant_txn = ovsdb_idl_txn_create(idl);
2174 HMAP_FOR_EACH (br, node, &all_bridges) {
2175 struct iface *iface;
2178 br_refresh_stp_status(br);
2180 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2181 port_refresh_stp_status(port);
2184 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2185 enum netdev_flags flags;
2187 const char *link_state;
2188 int64_t link_resets;
2191 if (iface_is_synthetic(iface)) {
2195 current = ofproto_port_is_lacp_current(br->ofproto,
2199 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2201 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2204 error = netdev_get_flags(iface->netdev, &flags);
2206 const char *state = flags & NETDEV_UP ? "up" : "down";
2207 ovsrec_interface_set_admin_state(iface->cfg, state);
2209 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2212 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2213 ovsrec_interface_set_link_state(iface->cfg, link_state);
2215 link_resets = netdev_get_carrier_resets(iface->netdev);
2216 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2218 iface_refresh_cfm_stats(iface);
2221 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2223 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2224 smap_destroy(&smap);
2229 status = ovsdb_idl_txn_commit(instant_txn);
2230 if (status != TXN_INCOMPLETE) {
2231 ovsdb_idl_txn_destroy(instant_txn);
2234 if (status == TXN_UNCHANGED) {
2235 instant_stats_could_have_changed = false;
2240 instant_stats_wait(void)
2243 ovsdb_idl_txn_wait(instant_txn);
2244 } else if (instant_stats_could_have_changed) {
2245 poll_timer_wait_until(instant_next_txn);
2249 /* Performs periodic activity required by bridges that needs to be done with
2250 * the least possible latency.
2252 * It makes sense to call this function a couple of times per poll loop, to
2253 * provide a significant performance boost on some benchmarks with ofprotos
2254 * that use the ofproto-dpif implementation. */
2256 bridge_run_fast(void)
2263 ofproto_enumerate_types(&types);
2264 SSET_FOR_EACH (type, &types) {
2265 ofproto_type_run_fast(type);
2267 sset_destroy(&types);
2269 HMAP_FOR_EACH (br, node, &all_bridges) {
2270 ofproto_run_fast(br->ofproto);
2277 static struct ovsrec_open_vswitch null_cfg;
2278 const struct ovsrec_open_vswitch *cfg;
2282 bool vlan_splinters_changed;
2285 ovsrec_open_vswitch_init(&null_cfg);
2289 if (ovsdb_idl_is_lock_contended(idl)) {
2290 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2291 struct bridge *br, *next_br;
2293 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2294 "disabling this process (pid %ld) until it goes away",
2295 (long int) getpid());
2297 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2300 /* Since we will not be running system_stats_run() in this process
2301 * with the current situation of multiple ovs-vswitchd daemons,
2302 * disable system stats collection. */
2303 system_stats_enable(false);
2305 } else if (!ovsdb_idl_has_lock(idl)) {
2308 cfg = ovsrec_open_vswitch_first(idl);
2310 /* Initialize the ofproto library. This only needs to run once, but
2311 * it must be done after the configuration is set. If the
2312 * initialization has already occurred, bridge_init_ofproto()
2313 * returns immediately. */
2314 bridge_init_ofproto(cfg);
2316 /* Once the value of flow-restore-wait is false, we no longer should
2317 * check its value from the database. */
2318 if (cfg && ofproto_get_flow_restore_wait()) {
2319 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2320 "flow-restore-wait", false));
2323 /* Let each datapath type do the work that it needs to do. */
2325 ofproto_enumerate_types(&types);
2326 SSET_FOR_EACH (type, &types) {
2327 ofproto_type_run(type);
2329 sset_destroy(&types);
2331 /* Let each bridge do the work that it needs to do. */
2332 HMAP_FOR_EACH (br, node, &all_bridges) {
2333 ofproto_run(br->ofproto);
2336 /* Re-configure SSL. We do this on every trip through the main loop,
2337 * instead of just when the database changes, because the contents of the
2338 * key and certificate files can change without the database changing.
2340 * We do this before bridge_reconfigure() because that function might
2341 * initiate SSL connections and thus requires SSL to be configured. */
2342 if (cfg && cfg->ssl) {
2343 const struct ovsrec_ssl *ssl = cfg->ssl;
2345 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2346 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2349 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2350 * usage has changed. */
2351 vlan_splinters_changed = false;
2352 if (vlan_splinters_enabled_anywhere) {
2353 HMAP_FOR_EACH (br, node, &all_bridges) {
2354 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2355 vlan_splinters_changed = true;
2361 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2362 struct ovsdb_idl_txn *txn;
2364 idl_seqno = ovsdb_idl_get_seqno(idl);
2365 txn = ovsdb_idl_txn_create(idl);
2366 bridge_reconfigure(cfg ? cfg : &null_cfg);
2369 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2372 /* If we are completing our initial configuration for this run
2373 * of ovs-vswitchd, then keep the transaction around to monitor
2374 * it for completion. */
2375 if (initial_config_done) {
2376 ovsdb_idl_txn_commit(txn);
2377 ovsdb_idl_txn_destroy(txn);
2379 initial_config_done = true;
2380 daemonize_txn = txn;
2384 if (daemonize_txn) {
2385 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2386 if (status != TXN_INCOMPLETE) {
2387 ovsdb_idl_txn_destroy(daemonize_txn);
2388 daemonize_txn = NULL;
2390 /* ovs-vswitchd has completed initialization, so allow the
2391 * process that forked us to exit successfully. */
2392 daemonize_complete();
2394 vlog_enable_async();
2396 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2400 /* Refresh interface and mirror stats if necessary. */
2401 if (time_msec() >= iface_stats_timer) {
2403 struct ovsdb_idl_txn *txn;
2405 txn = ovsdb_idl_txn_create(idl);
2406 HMAP_FOR_EACH (br, node, &all_bridges) {
2410 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2411 struct iface *iface;
2413 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2414 iface_refresh_stats(iface);
2415 iface_refresh_status(iface);
2418 port_refresh_stp_stats(port);
2421 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2422 mirror_refresh_stats(m);
2426 refresh_controller_status();
2427 ovsdb_idl_txn_commit(txn);
2428 ovsdb_idl_txn_destroy(txn); /* XXX */
2431 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2435 instant_stats_run();
2444 ovsdb_idl_wait(idl);
2445 if (daemonize_txn) {
2446 ovsdb_idl_txn_wait(daemonize_txn);
2450 ofproto_enumerate_types(&types);
2451 SSET_FOR_EACH (type, &types) {
2452 ofproto_type_wait(type);
2454 sset_destroy(&types);
2456 if (!hmap_is_empty(&all_bridges)) {
2459 HMAP_FOR_EACH (br, node, &all_bridges) {
2460 ofproto_wait(br->ofproto);
2462 poll_timer_wait_until(iface_stats_timer);
2465 system_stats_wait();
2466 instant_stats_wait();
2469 /* Adds some memory usage statistics for bridges into 'usage', for use with
2470 * memory_report(). */
2472 bridge_get_memory_usage(struct simap *usage)
2476 HMAP_FOR_EACH (br, node, &all_bridges) {
2477 ofproto_get_memory_usage(br->ofproto, usage);
2481 /* QoS unixctl user interface functions. */
2483 struct qos_unixctl_show_cbdata {
2485 struct iface *iface;
2489 qos_unixctl_show_queue(unsigned int queue_id,
2490 const struct smap *details,
2491 struct iface *iface,
2494 struct netdev_queue_stats stats;
2495 struct smap_node *node;
2498 ds_put_cstr(ds, "\n");
2500 ds_put_format(ds, "Queue %u:\n", queue_id);
2502 ds_put_cstr(ds, "Default:\n");
2505 SMAP_FOR_EACH (node, details) {
2506 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2509 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2511 if (stats.tx_packets != UINT64_MAX) {
2512 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2515 if (stats.tx_bytes != UINT64_MAX) {
2516 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2519 if (stats.tx_errors != UINT64_MAX) {
2520 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2523 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2524 queue_id, ovs_strerror(error));
2529 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2530 const char *argv[], void *aux OVS_UNUSED)
2532 struct ds ds = DS_EMPTY_INITIALIZER;
2533 struct smap smap = SMAP_INITIALIZER(&smap);
2534 struct iface *iface;
2536 struct smap_node *node;
2538 iface = iface_find(argv[1]);
2540 unixctl_command_reply_error(conn, "no such interface");
2544 netdev_get_qos(iface->netdev, &type, &smap);
2546 if (*type != '\0') {
2547 struct netdev_queue_dump dump;
2548 struct smap details;
2549 unsigned int queue_id;
2551 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2553 SMAP_FOR_EACH (node, &smap) {
2554 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2557 smap_init(&details);
2558 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2559 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2561 smap_destroy(&details);
2563 unixctl_command_reply(conn, ds_cstr(&ds));
2565 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2566 unixctl_command_reply_error(conn, ds_cstr(&ds));
2569 smap_destroy(&smap);
2573 /* Bridge reconfiguration functions. */
2575 bridge_create(const struct ovsrec_bridge *br_cfg)
2579 ovs_assert(!bridge_lookup(br_cfg->name));
2580 br = xzalloc(sizeof *br);
2582 br->name = xstrdup(br_cfg->name);
2583 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2586 /* Derive the default Ethernet address from the bridge's UUID. This should
2587 * be unique and it will be stable between ovs-vswitchd runs. */
2588 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2589 eth_addr_mark_random(br->default_ea);
2591 hmap_init(&br->ports);
2592 hmap_init(&br->ifaces);
2593 hmap_init(&br->iface_by_name);
2594 hmap_init(&br->mirrors);
2596 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2600 bridge_destroy(struct bridge *br)
2603 struct mirror *mirror, *next_mirror;
2604 struct port *port, *next_port;
2606 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2609 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2610 mirror_destroy(mirror);
2613 hmap_remove(&all_bridges, &br->node);
2614 ofproto_destroy(br->ofproto);
2615 hmap_destroy(&br->ifaces);
2616 hmap_destroy(&br->ports);
2617 hmap_destroy(&br->iface_by_name);
2618 hmap_destroy(&br->mirrors);
2625 static struct bridge *
2626 bridge_lookup(const char *name)
2630 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2631 if (!strcmp(br->name, name)) {
2638 /* Handle requests for a listing of all flows known by the OpenFlow
2639 * stack, including those normally hidden. */
2641 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2642 const char *argv[], void *aux OVS_UNUSED)
2647 br = bridge_lookup(argv[1]);
2649 unixctl_command_reply_error(conn, "Unknown bridge");
2654 ofproto_get_all_flows(br->ofproto, &results);
2656 unixctl_command_reply(conn, ds_cstr(&results));
2657 ds_destroy(&results);
2660 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2661 * connections and reconnect. If BRIDGE is not specified, then all bridges
2662 * drop their controller connections and reconnect. */
2664 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2665 const char *argv[], void *aux OVS_UNUSED)
2669 br = bridge_lookup(argv[1]);
2671 unixctl_command_reply_error(conn, "Unknown bridge");
2674 ofproto_reconnect_controllers(br->ofproto);
2676 HMAP_FOR_EACH (br, node, &all_bridges) {
2677 ofproto_reconnect_controllers(br->ofproto);
2680 unixctl_command_reply(conn, NULL);
2684 bridge_get_controllers(const struct bridge *br,
2685 struct ovsrec_controller ***controllersp)
2687 struct ovsrec_controller **controllers;
2688 size_t n_controllers;
2690 controllers = br->cfg->controller;
2691 n_controllers = br->cfg->n_controller;
2693 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2699 *controllersp = controllers;
2701 return n_controllers;
2705 bridge_collect_wanted_ports(struct bridge *br,
2706 const unsigned long int *splinter_vlans,
2707 struct shash *wanted_ports)
2711 shash_init(wanted_ports);
2713 for (i = 0; i < br->cfg->n_ports; i++) {
2714 const char *name = br->cfg->ports[i]->name;
2715 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2716 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2720 if (bridge_get_controllers(br, NULL)
2721 && !shash_find(wanted_ports, br->name)) {
2722 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2723 br->name, br->name);
2725 ovsrec_interface_init(&br->synth_local_iface);
2726 ovsrec_port_init(&br->synth_local_port);
2728 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2729 br->synth_local_port.n_interfaces = 1;
2730 br->synth_local_port.name = br->name;
2732 br->synth_local_iface.name = br->name;
2733 br->synth_local_iface.type = "internal";
2735 br->synth_local_ifacep = &br->synth_local_iface;
2737 shash_add(wanted_ports, br->name, &br->synth_local_port);
2740 if (splinter_vlans) {
2741 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2745 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2746 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2747 * 'br' needs to complete its configuration. */
2749 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2751 struct shash_node *port_node;
2752 struct port *port, *next;
2754 /* Get rid of deleted ports.
2755 * Get rid of deleted interfaces on ports that still exist. */
2756 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2757 port->cfg = shash_find_data(wanted_ports, port->name);
2761 port_del_ifaces(port);
2765 /* Update iface->cfg and iface->type in interfaces that still exist. */
2766 SHASH_FOR_EACH (port_node, wanted_ports) {
2767 const struct ovsrec_port *port = port_node->data;
2770 for (i = 0; i < port->n_interfaces; i++) {
2771 const struct ovsrec_interface *cfg = port->interfaces[i];
2772 struct iface *iface = iface_lookup(br, cfg->name);
2773 const char *type = iface_get_type(cfg, br->cfg);
2778 } else if (!strcmp(type, "null")) {
2779 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2780 " may be removed in February 2013. Please email"
2781 " dev@openvswitch.org with concerns.",
2784 /* We will add new interfaces later. */
2790 /* Initializes 'oc' appropriately as a management service controller for
2793 * The caller must free oc->target when it is no longer needed. */
2795 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2796 struct ofproto_controller *oc)
2798 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2799 oc->max_backoff = 0;
2800 oc->probe_interval = 60;
2801 oc->band = OFPROTO_OUT_OF_BAND;
2803 oc->burst_limit = 0;
2804 oc->enable_async_msgs = true;
2807 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2809 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2810 struct ofproto_controller *oc)
2814 oc->target = c->target;
2815 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2816 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2817 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2818 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2819 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2820 oc->burst_limit = (c->controller_burst_limit
2821 ? *c->controller_burst_limit : 0);
2822 oc->enable_async_msgs = (!c->enable_async_messages
2823 || *c->enable_async_messages);
2824 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2825 if (dscp < 0 || dscp > 63) {
2826 dscp = DSCP_DEFAULT;
2831 /* Configures the IP stack for 'br''s local interface properly according to the
2832 * configuration in 'c'. */
2834 bridge_configure_local_iface_netdev(struct bridge *br,
2835 struct ovsrec_controller *c)
2837 struct netdev *netdev;
2838 struct in_addr mask, gateway;
2840 struct iface *local_iface;
2843 /* If there's no local interface or no IP address, give up. */
2844 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2845 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2849 /* Bring up the local interface. */
2850 netdev = local_iface->netdev;
2851 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2853 /* Configure the IP address and netmask. */
2854 if (!c->local_netmask
2855 || !inet_aton(c->local_netmask, &mask)
2857 mask.s_addr = guess_netmask(ip.s_addr);
2859 if (!netdev_set_in4(netdev, ip, mask)) {
2860 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2861 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2864 /* Configure the default gateway. */
2865 if (c->local_gateway
2866 && inet_aton(c->local_gateway, &gateway)
2867 && gateway.s_addr) {
2868 if (!netdev_add_router(netdev, gateway)) {
2869 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2870 br->name, IP_ARGS(gateway.s_addr));
2875 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2876 * in either string are treated as equal to any number of slashes in the other,
2877 * e.g. "x///y" is equal to "x/y".
2879 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2880 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2881 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2882 * 'b' against a prefix of 'a'.
2885 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2887 const char *b_start = b;
2889 if (b - b_start >= b_stoplen) {
2891 } else if (*a != *b) {
2893 } else if (*a == '/') {
2894 a += strspn(a, "/");
2895 b += strspn(b, "/");
2896 } else if (*a == '\0') {
2906 bridge_configure_remotes(struct bridge *br,
2907 const struct sockaddr_in *managers, size_t n_managers)
2909 bool disable_in_band;
2911 struct ovsrec_controller **controllers;
2912 size_t n_controllers;
2914 enum ofproto_fail_mode fail_mode;
2916 struct ofproto_controller *ocs;
2920 /* Check if we should disable in-band control on this bridge. */
2921 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2924 /* Set OpenFlow queue ID for in-band control. */
2925 ofproto_set_in_band_queue(br->ofproto,
2926 smap_get_int(&br->cfg->other_config,
2927 "in-band-queue", -1));
2929 if (disable_in_band) {
2930 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2932 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2935 n_controllers = bridge_get_controllers(br, &controllers);
2937 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2940 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2941 for (i = 0; i < n_controllers; i++) {
2942 struct ovsrec_controller *c = controllers[i];
2944 if (!strncmp(c->target, "punix:", 6)
2945 || !strncmp(c->target, "unix:", 5)) {
2946 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2949 if (!strncmp(c->target, "unix:", 5)) {
2950 /* Connect to a listening socket */
2951 whitelist = xasprintf("unix:%s/", ovs_rundir());
2952 if (strchr(c->target, '/') &&
2953 !equal_pathnames(c->target, whitelist,
2954 strlen(whitelist))) {
2955 /* Absolute path specified, but not in ovs_rundir */
2956 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2957 "controller \"%s\" due to possibility for "
2958 "remote exploit. Instead, specify socket "
2959 "in whitelisted \"%s\" or connect to "
2960 "\"unix:%s/%s.mgmt\" (which is always "
2961 "available without special configuration).",
2962 br->name, c->target, whitelist,
2963 ovs_rundir(), br->name);
2968 whitelist = xasprintf("punix:%s/%s.controller",
2969 ovs_rundir(), br->name);
2970 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2971 /* Prevent remote ovsdb-server users from accessing
2972 * arbitrary Unix domain sockets and overwriting arbitrary
2974 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2975 "controller \"%s\" due to possibility of "
2976 "overwriting local files. Instead, specify "
2977 "whitelisted \"%s\" or connect to "
2978 "\"unix:%s/%s.mgmt\" (which is always "
2979 "available without special configuration).",
2980 br->name, c->target, whitelist,
2981 ovs_rundir(), br->name);
2990 bridge_configure_local_iface_netdev(br, c);
2991 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2992 if (disable_in_band) {
2993 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2998 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
2999 bridge_get_allowed_versions(br));
3000 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3003 /* Set the fail-mode. */
3004 fail_mode = !br->cfg->fail_mode
3005 || !strcmp(br->cfg->fail_mode, "standalone")
3006 ? OFPROTO_FAIL_STANDALONE
3007 : OFPROTO_FAIL_SECURE;
3008 ofproto_set_fail_mode(br->ofproto, fail_mode);
3010 /* Configure OpenFlow controller connection snooping. */
3011 if (!ofproto_has_snoops(br->ofproto)) {
3015 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3016 ovs_rundir(), br->name));
3017 ofproto_set_snoops(br->ofproto, &snoops);
3018 sset_destroy(&snoops);
3023 bridge_configure_tables(struct bridge *br)
3025 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3029 n_tables = ofproto_get_n_tables(br->ofproto);
3031 for (i = 0; i < n_tables; i++) {
3032 struct ofproto_table_settings s;
3035 s.max_flows = UINT_MAX;
3038 s.n_prefix_fields = 0;
3039 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3041 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3042 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3045 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3046 s.max_flows = *cfg->flow_limit;
3048 if (cfg->overflow_policy
3049 && !strcmp(cfg->overflow_policy, "evict")) {
3051 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3052 for (k = 0; k < cfg->n_groups; k++) {
3053 const char *string = cfg->groups[k];
3056 msg = mf_parse_subfield__(&s.groups[k], &string);
3058 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3059 "'groups' (%s)", br->name, i, msg);
3061 } else if (*string) {
3062 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3063 "element '%s' contains trailing garbage",
3064 br->name, i, cfg->groups[k]);
3070 /* Prefix lookup fields. */
3071 s.n_prefix_fields = 0;
3072 for (k = 0; k < cfg->n_prefixes; k++) {
3073 const char *name = cfg->prefixes[k];
3074 const struct mf_field *mf = mf_from_name(name);
3076 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3080 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3081 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3082 "%s", br->name, name);
3085 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3086 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3087 "field not used: %s", br->name, name);
3090 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3092 if (s.n_prefix_fields > 0) {
3094 struct ds ds = DS_EMPTY_INITIALIZER;
3095 for (k = 0; k < s.n_prefix_fields; k++) {
3097 ds_put_char(&ds, ',');
3099 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3101 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3102 br->name, i, ds_cstr(&ds));
3107 ofproto_configure_table(br->ofproto, i, &s);
3111 for (; j < br->cfg->n_flow_tables; j++) {
3112 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3113 "%"PRId64" not supported by this datapath", br->name,
3114 br->cfg->key_flow_tables[j]);
3119 bridge_configure_dp_desc(struct bridge *br)
3121 ofproto_set_dp_desc(br->ofproto,
3122 smap_get(&br->cfg->other_config, "dp-desc"));
3125 /* Port functions. */
3127 static void iface_destroy__(struct iface *);
3129 static struct port *
3130 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3134 port = xzalloc(sizeof *port);
3136 port->name = xstrdup(cfg->name);
3138 list_init(&port->ifaces);
3140 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3144 /* Deletes interfaces from 'port' that are no longer configured for it. */
3146 port_del_ifaces(struct port *port)
3148 struct iface *iface, *next;
3149 struct sset new_ifaces;
3152 /* Collect list of new interfaces. */
3153 sset_init(&new_ifaces);
3154 for (i = 0; i < port->cfg->n_interfaces; i++) {
3155 const char *name = port->cfg->interfaces[i]->name;
3156 const char *type = port->cfg->interfaces[i]->type;
3157 if (strcmp(type, "null")) {
3158 sset_add(&new_ifaces, name);
3162 /* Get rid of deleted interfaces. */
3163 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3164 if (!sset_contains(&new_ifaces, iface->name)) {
3165 iface_destroy(iface);
3169 sset_destroy(&new_ifaces);
3173 port_destroy(struct port *port)
3176 struct bridge *br = port->bridge;
3177 struct iface *iface, *next;
3180 ofproto_bundle_unregister(br->ofproto, port);
3183 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3184 iface_destroy__(iface);
3187 hmap_remove(&br->ports, &port->hmap_node);
3193 static struct port *
3194 port_lookup(const struct bridge *br, const char *name)
3198 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3200 if (!strcmp(port->name, name)) {
3208 enable_lacp(struct port *port, bool *activep)
3210 if (!port->cfg->lacp) {
3211 /* XXX when LACP implementation has been sufficiently tested, enable by
3212 * default and make active on bonded ports. */
3214 } else if (!strcmp(port->cfg->lacp, "off")) {
3216 } else if (!strcmp(port->cfg->lacp, "active")) {
3219 } else if (!strcmp(port->cfg->lacp, "passive")) {
3223 VLOG_WARN("port %s: unknown LACP mode %s",
3224 port->name, port->cfg->lacp);
3229 static struct lacp_settings *
3230 port_configure_lacp(struct port *port, struct lacp_settings *s)
3232 const char *lacp_time, *system_id;
3235 if (!enable_lacp(port, &s->active)) {
3239 s->name = port->name;
3241 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3243 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3244 ETH_ADDR_SCAN_ARGS(s->id))) {
3245 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3246 " address.", port->name, system_id);
3250 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3253 if (eth_addr_is_zero(s->id)) {
3254 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3258 /* Prefer bondable links if unspecified. */
3259 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3261 s->priority = (priority > 0 && priority <= UINT16_MAX
3263 : UINT16_MAX - !list_is_short(&port->ifaces));
3265 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3266 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3268 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3269 "lacp-fallback-ab", false);
3275 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3277 int priority, portid, key;
3279 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3280 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3282 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3284 if (portid <= 0 || portid > UINT16_MAX) {
3285 portid = ofp_to_u16(iface->ofp_port);
3288 if (priority <= 0 || priority > UINT16_MAX) {
3289 priority = UINT16_MAX;
3292 if (key < 0 || key > UINT16_MAX) {
3296 s->name = iface->name;
3298 s->priority = priority;
3303 port_configure_bond(struct port *port, struct bond_settings *s)
3305 const char *detect_s;
3306 struct iface *iface;
3307 int miimon_interval;
3309 s->name = port->name;
3311 if (port->cfg->bond_mode) {
3312 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3313 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3314 port->name, port->cfg->bond_mode,
3315 bond_mode_to_string(s->balance));
3318 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3320 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3321 * active-backup. At some point we should remove this warning. */
3322 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3323 " in previous versions, the default bond_mode was"
3324 " balance-slb", port->name,
3325 bond_mode_to_string(s->balance));
3327 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3328 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3329 "please use another bond type or disable flood_vlans",
3333 miimon_interval = smap_get_int(&port->cfg->other_config,
3334 "bond-miimon-interval", 0);
3335 if (miimon_interval <= 0) {
3336 miimon_interval = 200;
3339 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3340 if (!detect_s || !strcmp(detect_s, "carrier")) {
3341 miimon_interval = 0;
3342 } else if (strcmp(detect_s, "miimon")) {
3343 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3344 "defaulting to carrier", port->name, detect_s);
3345 miimon_interval = 0;
3348 s->up_delay = MAX(0, port->cfg->bond_updelay);
3349 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3350 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3351 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3352 "bond-rebalance-interval", 10000);
3353 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3354 s->rebalance_interval = 1000;
3357 s->fake_iface = port->cfg->bond_fake_iface;
3359 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3360 "lacp-fallback-ab", false);
3362 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3363 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3367 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3368 * instead of obtaining it from the database. */
3370 port_is_synthetic(const struct port *port)
3372 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3375 /* Interface functions. */
3378 iface_is_internal(const struct ovsrec_interface *iface,
3379 const struct ovsrec_bridge *br)
3381 /* The local port and "internal" ports are always "internal". */
3382 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3385 /* Returns the correct network device type for interface 'iface' in bridge
3388 iface_get_type(const struct ovsrec_interface *iface,
3389 const struct ovsrec_bridge *br)
3393 /* The local port always has type "internal". Other ports take
3394 * their type from the database and default to "system" if none is
3396 if (iface_is_internal(iface, br)) {
3399 type = iface->type[0] ? iface->type : "system";
3402 return ofproto_port_open_type(br->datapath_type, type);
3406 iface_destroy__(struct iface *iface)
3409 struct port *port = iface->port;
3410 struct bridge *br = port->bridge;
3412 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3413 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3416 if (iface->ofp_port != OFPP_NONE) {
3417 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3420 list_remove(&iface->port_elem);
3421 hmap_remove(&br->iface_by_name, &iface->name_node);
3423 netdev_close(iface->netdev);
3431 iface_destroy(struct iface *iface)
3434 struct port *port = iface->port;
3436 iface_destroy__(iface);
3437 if (list_is_empty(&port->ifaces)) {
3443 static struct iface *
3444 iface_lookup(const struct bridge *br, const char *name)
3446 struct iface *iface;
3448 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3449 &br->iface_by_name) {
3450 if (!strcmp(iface->name, name)) {
3458 static struct iface *
3459 iface_find(const char *name)
3461 const struct bridge *br;
3463 HMAP_FOR_EACH (br, node, &all_bridges) {
3464 struct iface *iface = iface_lookup(br, name);
3473 static struct iface *
3474 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3476 struct iface *iface;
3478 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3480 if (iface->ofp_port == ofp_port) {
3487 /* Set Ethernet address of 'iface', if one is specified in the configuration
3490 iface_set_mac(struct iface *iface)
3492 uint8_t ea[ETH_ADDR_LEN];
3494 if (!strcmp(iface->type, "internal")
3495 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3496 if (iface->ofp_port == OFPP_LOCAL) {
3497 VLOG_ERR("interface %s: ignoring mac in Interface record "
3498 "(use Bridge record to set local port's mac)",
3500 } else if (eth_addr_is_multicast(ea)) {
3501 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3504 int error = netdev_set_etheraddr(iface->netdev, ea);
3506 VLOG_ERR("interface %s: setting MAC failed (%s)",
3507 iface->name, ovs_strerror(error));
3513 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3515 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3517 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3518 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3519 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3523 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3524 * sets the "ofport" field to -1.
3526 * This is appropriate when 'if_cfg''s interface cannot be created or is
3527 * otherwise invalid. */
3529 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3531 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3532 ovsrec_interface_set_status(if_cfg, NULL);
3533 ovsrec_interface_set_admin_state(if_cfg, NULL);
3534 ovsrec_interface_set_duplex(if_cfg, NULL);
3535 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3536 ovsrec_interface_set_link_state(if_cfg, NULL);
3537 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3538 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3539 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3540 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3541 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3542 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3543 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3544 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3549 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3551 union ovsdb_atom atom;
3553 atom.integer = target;
3554 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3558 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3560 struct ofpbuf queues_buf;
3562 ofpbuf_init(&queues_buf, 0);
3564 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3565 netdev_set_qos(iface->netdev, NULL, NULL);
3567 const struct ovsdb_datum *queues;
3568 struct netdev_queue_dump dump;
3569 unsigned int queue_id;
3570 struct smap details;
3574 /* Configure top-level Qos for 'iface'. */
3575 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3577 /* Deconfigure queues that were deleted. */
3578 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3580 smap_init(&details);
3581 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3582 if (!queue_ids_include(queues, queue_id)) {
3583 netdev_delete_queue(iface->netdev, queue_id);
3586 smap_destroy(&details);
3588 /* Configure queues for 'iface'. */
3590 for (i = 0; i < qos->n_queues; i++) {
3591 const struct ovsrec_queue *queue = qos->value_queues[i];
3592 unsigned int queue_id = qos->key_queues[i];
3594 if (queue_id == 0) {
3598 if (queue->n_dscp == 1) {
3599 struct ofproto_port_queue *port_queue;
3601 port_queue = ofpbuf_put_uninit(&queues_buf,
3602 sizeof *port_queue);
3603 port_queue->queue = queue_id;
3604 port_queue->dscp = queue->dscp[0];
3607 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3610 struct smap details;
3612 smap_init(&details);
3613 netdev_set_queue(iface->netdev, 0, &details);
3614 smap_destroy(&details);
3618 if (iface->ofp_port != OFPP_NONE) {
3619 const struct ofproto_port_queue *port_queues = queues_buf.data;
3620 size_t n_queues = queues_buf.size / sizeof *port_queues;
3622 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3623 port_queues, n_queues);
3626 netdev_set_policing(iface->netdev,
3627 iface->cfg->ingress_policing_rate,
3628 iface->cfg->ingress_policing_burst);
3630 ofpbuf_uninit(&queues_buf);
3634 iface_configure_cfm(struct iface *iface)
3636 const struct ovsrec_interface *cfg = iface->cfg;
3637 const char *opstate_str;
3638 const char *cfm_ccm_vlan;
3639 struct cfm_settings s;
3640 struct smap netdev_args;
3642 if (!cfg->n_cfm_mpid) {
3643 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3647 s.check_tnl_key = false;
3648 smap_init(&netdev_args);
3649 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3650 const char *key = smap_get(&netdev_args, "key");
3651 const char *in_key = smap_get(&netdev_args, "in_key");
3653 s.check_tnl_key = (key && !strcmp(key, "flow"))
3654 || (in_key && !strcmp(in_key, "flow"));
3656 smap_destroy(&netdev_args);
3658 s.mpid = *cfg->cfm_mpid;
3659 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3660 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3661 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3663 if (s.interval <= 0) {
3667 if (!cfm_ccm_vlan) {
3669 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3670 s.ccm_vlan = CFM_RANDOM_VLAN;
3672 s.ccm_vlan = atoi(cfm_ccm_vlan);
3673 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3678 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3680 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3682 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3683 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3685 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3688 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3689 * instead of obtaining it from the database. */
3691 iface_is_synthetic(const struct iface *iface)
3693 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3697 iface_validate_ofport__(size_t n, int64_t *ofport)
3699 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3700 ? u16_to_ofp(*ofport)
3705 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3707 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3711 iface_pick_ofport(const struct ovsrec_interface *cfg)
3713 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3714 return (requested_ofport != OFPP_NONE
3716 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3719 /* Port mirroring. */
3721 static struct mirror *
3722 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3726 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3727 if (uuid_equals(uuid, &m->uuid)) {
3735 bridge_configure_mirrors(struct bridge *br)
3737 const struct ovsdb_datum *mc;
3738 unsigned long *flood_vlans;
3739 struct mirror *m, *next;
3742 /* Get rid of deleted mirrors. */
3743 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3744 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3745 union ovsdb_atom atom;
3747 atom.uuid = m->uuid;
3748 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3753 /* Add new mirrors and reconfigure existing ones. */
3754 for (i = 0; i < br->cfg->n_mirrors; i++) {
3755 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3756 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3758 m = mirror_create(br, cfg);
3761 if (!mirror_configure(m)) {
3766 /* Update flooded vlans (for RSPAN). */
3767 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3768 br->cfg->n_flood_vlans);
3769 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3770 bitmap_free(flood_vlans);
3773 static struct mirror *
3774 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3778 m = xzalloc(sizeof *m);
3779 m->uuid = cfg->header_.uuid;
3780 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3782 m->name = xstrdup(cfg->name);
3788 mirror_destroy(struct mirror *m)
3791 struct bridge *br = m->bridge;
3794 ofproto_mirror_unregister(br->ofproto, m);
3797 hmap_remove(&br->mirrors, &m->hmap_node);
3804 mirror_collect_ports(struct mirror *m,
3805 struct ovsrec_port **in_ports, int n_in_ports,
3806 void ***out_portsp, size_t *n_out_portsp)
3808 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3809 size_t n_out_ports = 0;
3812 for (i = 0; i < n_in_ports; i++) {
3813 const char *name = in_ports[i]->name;
3814 struct port *port = port_lookup(m->bridge, name);
3816 out_ports[n_out_ports++] = port;
3818 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3819 "port %s", m->bridge->name, m->name, name);
3822 *out_portsp = out_ports;
3823 *n_out_portsp = n_out_ports;
3827 mirror_configure(struct mirror *m)
3829 const struct ovsrec_mirror *cfg = m->cfg;
3830 struct ofproto_mirror_settings s;
3833 if (strcmp(cfg->name, m->name)) {
3835 m->name = xstrdup(cfg->name);
3839 /* Get output port or VLAN. */
3840 if (cfg->output_port) {
3841 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3842 if (!s.out_bundle) {
3843 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3844 m->bridge->name, m->name);
3847 s.out_vlan = UINT16_MAX;
3849 if (cfg->output_vlan) {
3850 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3851 "output vlan; ignoring output vlan",
3852 m->bridge->name, m->name);
3854 } else if (cfg->output_vlan) {
3855 /* The database should prevent invalid VLAN values. */
3856 s.out_bundle = NULL;
3857 s.out_vlan = *cfg->output_vlan;
3859 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3860 m->bridge->name, m->name);
3864 /* Get port selection. */
3865 if (cfg->select_all) {
3866 size_t n_ports = hmap_count(&m->bridge->ports);
3867 void **ports = xmalloc(n_ports * sizeof *ports);
3872 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3882 /* Get ports, dropping ports that don't exist.
3883 * The IDL ensures that there are no duplicates. */
3884 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3885 &s.srcs, &s.n_srcs);
3886 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3887 &s.dsts, &s.n_dsts);
3890 /* Get VLAN selection. */
3891 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3894 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3897 if (s.srcs != s.dsts) {
3906 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3908 * This is deprecated. It is only for compatibility with broken device drivers
3909 * in old versions of Linux that do not properly support VLANs when VLAN
3910 * devices are not used. When broken device drivers are no longer in
3911 * widespread use, we will delete these interfaces. */
3913 static struct ovsrec_port **recs;
3914 static size_t n_recs, allocated_recs;
3916 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3917 * splinters are reconfigured. */
3919 register_rec(struct ovsrec_port *rec)
3921 if (n_recs >= allocated_recs) {
3922 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3924 recs[n_recs++] = rec;
3927 /* Frees all of the ports registered with register_reg(). */
3929 free_registered_recs(void)
3933 for (i = 0; i < n_recs; i++) {
3934 struct ovsrec_port *port = recs[i];
3937 for (j = 0; j < port->n_interfaces; j++) {
3938 struct ovsrec_interface *iface = port->interfaces[j];
3943 smap_destroy(&port->other_config);
3944 free(port->interfaces);
3952 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3955 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3957 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3961 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3964 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3965 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3966 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3969 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3970 * use, returns NULL.
3972 * Updates 'vlan_splinters_enabled_anywhere'. */
3973 static unsigned long int *
3974 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3976 unsigned long int *splinter_vlans;
3977 struct sset splinter_ifaces;
3978 const char *real_dev_name;
3979 struct shash *real_devs;
3980 struct shash_node *node;
3984 /* Free space allocated for synthesized ports and interfaces, since we're
3985 * in the process of reconstructing all of them. */
3986 free_registered_recs();
3988 splinter_vlans = bitmap_allocate(4096);
3989 sset_init(&splinter_ifaces);
3990 vlan_splinters_enabled_anywhere = false;
3991 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3992 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3995 for (j = 0; j < br_cfg->n_ports; j++) {
3996 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3999 for (k = 0; k < port_cfg->n_interfaces; k++) {
4000 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4002 if (vlan_splinters_is_enabled(iface_cfg)) {
4003 vlan_splinters_enabled_anywhere = true;
4004 sset_add(&splinter_ifaces, iface_cfg->name);
4005 vlan_bitmap_from_array__(port_cfg->trunks,
4011 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4012 bitmap_set1(splinter_vlans, *port_cfg->tag);
4017 if (!vlan_splinters_enabled_anywhere) {
4018 free(splinter_vlans);
4019 sset_destroy(&splinter_ifaces);
4023 HMAP_FOR_EACH (br, node, &all_bridges) {
4025 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4029 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4030 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4031 * device to be created for it. */
4032 bitmap_set0(splinter_vlans, 0);
4033 bitmap_set0(splinter_vlans, 4095);
4035 /* Delete all VLAN devices that we don't need. */
4037 real_devs = vlandev_get_real_devs();
4038 SHASH_FOR_EACH (node, real_devs) {
4039 const struct vlan_real_dev *real_dev = node->data;
4040 const struct vlan_dev *vlan_dev;
4041 bool real_dev_has_splinters;
4043 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4045 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4046 if (!real_dev_has_splinters
4047 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4048 struct netdev *netdev;
4050 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4051 if (!netdev_get_in4(netdev, NULL, NULL) ||
4052 !netdev_get_in6(netdev, NULL)) {
4053 /* It has an IP address configured, so we don't own
4054 * it. Don't delete it. */
4056 vlandev_del(vlan_dev->name);
4058 netdev_close(netdev);
4065 /* Add all VLAN devices that we need. */
4066 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4069 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4070 if (!vlandev_get_name(real_dev_name, vid)) {
4071 vlandev_add(real_dev_name, vid);
4078 sset_destroy(&splinter_ifaces);
4080 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4081 free(splinter_vlans);
4084 return splinter_vlans;
4087 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4090 configure_splinter_port(struct port *port)
4092 struct ofproto *ofproto = port->bridge->ofproto;
4093 ofp_port_t realdev_ofp_port;
4094 const char *realdev_name;
4095 struct iface *vlandev, *realdev;
4097 ofproto_bundle_unregister(port->bridge->ofproto, port);
4099 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4102 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4103 realdev = iface_lookup(port->bridge, realdev_name);
4104 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4106 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4110 static struct ovsrec_port *
4111 synthesize_splinter_port(const char *real_dev_name,
4112 const char *vlan_dev_name, int vid)
4114 struct ovsrec_interface *iface;
4115 struct ovsrec_port *port;
4117 iface = xmalloc(sizeof *iface);
4118 ovsrec_interface_init(iface);
4119 iface->name = xstrdup(vlan_dev_name);
4120 iface->type = "system";
4122 port = xmalloc(sizeof *port);
4123 ovsrec_port_init(port);
4124 port->interfaces = xmemdup(&iface, sizeof iface);
4125 port->n_interfaces = 1;
4126 port->name = xstrdup(vlan_dev_name);
4127 port->vlan_mode = "splinter";
4128 port->tag = xmalloc(sizeof *port->tag);
4131 smap_add(&port->other_config, "realdev", real_dev_name);
4137 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4138 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4139 * 1-bit in the 'splinter_vlans' bitmap. */
4141 add_vlan_splinter_ports(struct bridge *br,
4142 const unsigned long int *splinter_vlans,
4143 struct shash *ports)
4147 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4148 * we're modifying 'ports'. */
4149 for (i = 0; i < br->cfg->n_ports; i++) {
4150 const char *name = br->cfg->ports[i]->name;
4151 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4154 for (j = 0; j < port_cfg->n_interfaces; j++) {
4155 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4157 if (vlan_splinters_is_enabled(iface_cfg)) {
4158 const char *real_dev_name;
4161 real_dev_name = iface_cfg->name;
4162 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4163 const char *vlan_dev_name;
4165 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4167 && !shash_find(ports, vlan_dev_name)) {
4168 shash_add(ports, vlan_dev_name,
4169 synthesize_splinter_port(
4170 real_dev_name, vlan_dev_name, vid));
4179 mirror_refresh_stats(struct mirror *m)
4181 struct ofproto *ofproto = m->bridge->ofproto;
4182 uint64_t tx_packets, tx_bytes;
4185 size_t stat_cnt = 0;
4187 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4188 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4192 if (tx_packets != UINT64_MAX) {
4193 keys[stat_cnt] = "tx_packets";
4194 values[stat_cnt] = tx_packets;
4197 if (tx_bytes != UINT64_MAX) {
4198 keys[stat_cnt] = "tx_bytes";
4199 values[stat_cnt] = tx_bytes;
4203 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);