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_pick_ofport(const struct ovsrec_interface *);
260 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
262 * This is deprecated. It is only for compatibility with broken device drivers
263 * in old versions of Linux that do not properly support VLANs when VLAN
264 * devices are not used. When broken device drivers are no longer in
265 * widespread use, we will delete these interfaces. */
267 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
268 static bool vlan_splinters_enabled_anywhere;
270 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
271 static unsigned long int *collect_splinter_vlans(
272 const struct ovsrec_open_vswitch *);
273 static void configure_splinter_port(struct port *);
274 static void add_vlan_splinter_ports(struct bridge *,
275 const unsigned long int *splinter_vlans,
276 struct shash *ports);
279 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
281 struct shash iface_hints;
282 static bool initialized = false;
289 shash_init(&iface_hints);
292 for (i = 0; i < cfg->n_bridges; i++) {
293 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
296 for (j = 0; j < br_cfg->n_ports; j++) {
297 struct ovsrec_port *port_cfg = br_cfg->ports[j];
300 for (k = 0; k < port_cfg->n_interfaces; k++) {
301 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
302 struct iface_hint *iface_hint;
304 iface_hint = xmalloc(sizeof *iface_hint);
305 iface_hint->br_name = br_cfg->name;
306 iface_hint->br_type = br_cfg->datapath_type;
307 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
309 shash_add(&iface_hints, if_cfg->name, iface_hint);
315 ofproto_init(&iface_hints);
317 shash_destroy_free_data(&iface_hints);
321 /* Public functions. */
323 /* Initializes the bridge module, configuring it to obtain its configuration
324 * from an OVSDB server accessed over 'remote', which should be a string in a
325 * form acceptable to ovsdb_idl_create(). */
327 bridge_init(const char *remote)
329 /* Create connection to database. */
330 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
331 idl_seqno = ovsdb_idl_get_seqno(idl);
332 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
333 ovsdb_idl_verify_write_only(idl);
335 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
336 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
337 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
338 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
339 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
340 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
341 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
343 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
344 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
345 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
347 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
348 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
349 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
350 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
352 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
353 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
354 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
355 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
356 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
357 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
358 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
359 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
371 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
373 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
374 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
375 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
376 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
378 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
380 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
382 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
383 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
385 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
386 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
392 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
393 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
394 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
396 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
398 /* Register unixctl commands. */
399 unixctl_command_register("qos/show", "interface", 1, 1,
400 qos_unixctl_show, NULL);
401 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
402 bridge_unixctl_dump_flows, NULL);
403 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
404 bridge_unixctl_reconnect, NULL);
414 struct bridge *br, *next_br;
416 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
419 ovsdb_idl_destroy(idl);
422 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
423 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
424 * responsible for freeing '*managersp' (with free()).
426 * You may be asking yourself "why does ovs-vswitchd care?", because
427 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
428 * should not be and in fact is not directly involved in that. But
429 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
430 * it has to tell in-band control where the managers are to enable that.
431 * (Thus, only managers connected in-band are collected.)
434 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
435 struct sockaddr_in **managersp, size_t *n_managersp)
437 struct sockaddr_in *managers = NULL;
438 size_t n_managers = 0;
442 /* Collect all of the potential targets from the "targets" columns of the
443 * rows pointed to by "manager_options", excluding any that are
446 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
447 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
449 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
450 sset_find_and_delete(&targets, m->target);
452 sset_add(&targets, m->target);
456 /* Now extract the targets' IP addresses. */
457 if (!sset_is_empty(&targets)) {
460 managers = xmalloc(sset_count(&targets) * sizeof *managers);
461 SSET_FOR_EACH (target, &targets) {
462 struct sockaddr_in *sin = &managers[n_managers];
464 if (stream_parse_target_with_default_port(target,
471 sset_destroy(&targets);
473 *managersp = managers;
474 *n_managersp = n_managers;
478 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
480 unsigned long int *splinter_vlans;
481 struct sockaddr_in *managers;
482 struct bridge *br, *next;
483 int sflow_bridge_number;
486 COVERAGE_INC(bridge_reconfigure);
488 ofproto_set_flow_eviction_threshold(
489 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
490 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
492 ofproto_set_n_handler_threads(
493 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0));
495 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
496 "force-miss-model"));
498 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
499 * to 'ovs_cfg', with only very minimal configuration otherwise.
501 * This is mostly an update to bridge data structures. Nothing is pushed
502 * down to ofproto or lower layers. */
503 add_del_bridges(ovs_cfg);
504 splinter_vlans = collect_splinter_vlans(ovs_cfg);
505 HMAP_FOR_EACH (br, node, &all_bridges) {
506 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
507 bridge_del_ports(br, &br->wanted_ports);
509 free(splinter_vlans);
511 /* Start pushing configuration changes down to the ofproto layer:
513 * - Delete ofprotos that are no longer configured.
515 * - Delete ports that are no longer configured.
517 * - Reconfigure existing ports to their desired configurations, or
518 * delete them if not possible.
520 * We have to do all the deletions before we can do any additions, because
521 * the ports to be added might require resources that will be freed up by
522 * deletions (they might especially overlap in name). */
523 bridge_delete_ofprotos();
524 HMAP_FOR_EACH (br, node, &all_bridges) {
526 bridge_delete_or_reconfigure_ports(br);
530 /* Finish pushing configuration changes to the ofproto layer:
532 * - Create ofprotos that are missing.
534 * - Add ports that are missing. */
535 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
539 error = ofproto_create(br->name, br->type, &br->ofproto);
541 VLOG_ERR("failed to create bridge %s: %s", br->name,
542 ovs_strerror(error));
543 shash_destroy(&br->wanted_ports);
548 HMAP_FOR_EACH (br, node, &all_bridges) {
549 bridge_add_ports(br, &br->wanted_ports);
550 shash_destroy(&br->wanted_ports);
553 reconfigure_system_stats(ovs_cfg);
555 /* Complete the configuration. */
556 sflow_bridge_number = 0;
557 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
558 HMAP_FOR_EACH (br, node, &all_bridges) {
561 /* We need the datapath ID early to allow LACP ports to use it as the
562 * default system ID. */
563 bridge_configure_datapath_id(br);
565 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
568 port_configure(port);
570 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
571 iface_configure_cfm(iface);
572 iface_configure_qos(iface, port->cfg->qos);
573 iface_set_mac(iface);
574 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
578 bridge_configure_mirrors(br);
579 bridge_configure_forward_bpdu(br);
580 bridge_configure_mac_table(br);
581 bridge_configure_remotes(br, managers, n_managers);
582 bridge_configure_netflow(br);
583 bridge_configure_sflow(br, &sflow_bridge_number);
584 bridge_configure_ipfix(br);
585 bridge_configure_stp(br);
586 bridge_configure_tables(br);
587 bridge_configure_dp_desc(br);
589 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
590 /* XXX: Remove this warning message eventually. */
591 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
592 " moved to the Open_vSwitch table. Ignoring its"
593 " setting in the bridge table.");
599 /* Delete ofprotos which aren't configured or have the wrong type. Create
600 * ofprotos which don't exist but need to. */
602 bridge_delete_ofprotos(void)
609 /* Delete ofprotos with no bridge or with the wrong type. */
612 ofproto_enumerate_types(&types);
613 SSET_FOR_EACH (type, &types) {
616 ofproto_enumerate_names(type, &names);
617 SSET_FOR_EACH (name, &names) {
618 br = bridge_lookup(name);
619 if (!br || strcmp(type, br->type)) {
620 ofproto_delete(name, type);
624 sset_destroy(&names);
625 sset_destroy(&types);
629 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
631 if (*n >= *allocated) {
632 ports = x2nrealloc(ports, allocated, sizeof *ports);
634 ports[(*n)++] = port;
639 bridge_delete_or_reconfigure_ports(struct bridge *br)
641 struct ofproto_port ofproto_port;
642 struct ofproto_port_dump dump;
644 /* List of "ofp_port"s to delete. We make a list instead of deleting them
645 * right away because ofproto implementations aren't necessarily able to
646 * iterate through a changing list of ports in an entirely robust way. */
654 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
657 iface = iface_lookup(br, ofproto_port.name);
659 /* No such iface is configured, so we should delete this
662 * As a corner case exception, keep the port if it's a bond fake
664 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
665 && !strcmp(ofproto_port.type, "internal")) {
671 if (strcmp(ofproto_port.type, iface->type)
672 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
673 /* The interface is the wrong type or can't be configured.
682 iface_destroy(iface);
683 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
686 for (i = 0; i < n; i++) {
687 ofproto_port_del(br->ofproto, del[i]);
693 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
695 struct shash_node *port_node;
697 SHASH_FOR_EACH (port_node, wanted_ports) {
698 const struct ovsrec_port *port_cfg = port_node->data;
701 for (i = 0; i < port_cfg->n_interfaces; i++) {
702 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
703 struct iface *iface = iface_lookup(br, iface_cfg->name);
706 iface_create(br, iface_cfg, port_cfg);
713 port_configure(struct port *port)
715 const struct ovsrec_port *cfg = port->cfg;
716 struct bond_settings bond_settings;
717 struct lacp_settings lacp_settings;
718 struct ofproto_bundle_settings s;
721 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
722 configure_splinter_port(port);
731 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
732 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
733 s.slaves[s.n_slaves++] = iface->ofp_port;
738 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
742 /* Get VLAN trunks. */
745 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
749 if (cfg->vlan_mode) {
750 if (!strcmp(cfg->vlan_mode, "access")) {
751 s.vlan_mode = PORT_VLAN_ACCESS;
752 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
753 s.vlan_mode = PORT_VLAN_TRUNK;
754 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
755 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
756 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
757 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
759 /* This "can't happen" because ovsdb-server should prevent it. */
760 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
761 s.vlan_mode = PORT_VLAN_TRUNK;
765 s.vlan_mode = PORT_VLAN_ACCESS;
767 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
771 s.vlan_mode = PORT_VLAN_TRUNK;
774 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
777 /* Get LACP settings. */
778 s.lacp = port_configure_lacp(port, &lacp_settings);
782 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
783 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
784 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
787 s.lacp_slaves = NULL;
790 /* Get bond settings. */
791 if (s.n_slaves > 1) {
792 s.bond = &bond_settings;
793 port_configure_bond(port, &bond_settings);
796 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
797 netdev_set_miimon_interval(iface->netdev, 0);
802 ofproto_bundle_register(port->bridge->ofproto, port, &s);
811 bridge_configure_flow_miss_model(const char *opt)
813 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
816 if (strcmp(opt, "with-facets")) {
817 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
818 } else if (strcmp(opt, "without-facets")) {
819 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
823 ofproto_set_flow_miss_model(model);
826 /* Pick local port hardware address and datapath ID for 'br'. */
828 bridge_configure_datapath_id(struct bridge *br)
830 uint8_t ea[ETH_ADDR_LEN];
832 struct iface *local_iface;
833 struct iface *hw_addr_iface;
836 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
837 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
839 int error = netdev_set_etheraddr(local_iface->netdev, ea);
841 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
842 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
843 "Ethernet address: %s",
844 br->name, ovs_strerror(error));
847 memcpy(br->ea, ea, ETH_ADDR_LEN);
849 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
850 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
851 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
852 ofproto_set_datapath_id(br->ofproto, dpid);
855 dpid_string = xasprintf("%016"PRIx64, dpid);
856 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
860 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
863 bridge_get_allowed_versions(struct bridge *br)
865 if (!br->cfg->n_protocols)
868 return ofputil_versions_from_strings(br->cfg->protocols,
869 br->cfg->n_protocols);
872 /* Set NetFlow configuration on 'br'. */
874 bridge_configure_netflow(struct bridge *br)
876 struct ovsrec_netflow *cfg = br->cfg->netflow;
877 struct netflow_options opts;
880 ofproto_set_netflow(br->ofproto, NULL);
884 memset(&opts, 0, sizeof opts);
886 /* Get default NetFlow configuration from datapath.
887 * Apply overrides from 'cfg'. */
888 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
889 if (cfg->engine_type) {
890 opts.engine_type = *cfg->engine_type;
892 if (cfg->engine_id) {
893 opts.engine_id = *cfg->engine_id;
896 /* Configure active timeout interval. */
897 opts.active_timeout = cfg->active_timeout;
898 if (!opts.active_timeout) {
899 opts.active_timeout = -1;
900 } else if (opts.active_timeout < 0) {
901 VLOG_WARN("bridge %s: active timeout interval set to negative "
902 "value, using default instead (%d seconds)", br->name,
903 NF_ACTIVE_TIMEOUT_DEFAULT);
904 opts.active_timeout = -1;
907 /* Add engine ID to interface number to disambiguate bridgs? */
908 opts.add_id_to_iface = cfg->add_id_to_interface;
909 if (opts.add_id_to_iface) {
910 if (opts.engine_id > 0x7f) {
911 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
912 "another vswitch, choose an engine id less than 128",
915 if (hmap_count(&br->ports) > 508) {
916 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
917 "another port when more than 508 ports are used",
923 sset_init(&opts.collectors);
924 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
927 if (ofproto_set_netflow(br->ofproto, &opts)) {
928 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
930 sset_destroy(&opts.collectors);
933 /* Set sFlow configuration on 'br'. */
935 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
937 const struct ovsrec_sflow *cfg = br->cfg->sflow;
938 struct ovsrec_controller **controllers;
939 struct ofproto_sflow_options oso;
940 size_t n_controllers;
944 ofproto_set_sflow(br->ofproto, NULL);
948 memset(&oso, 0, sizeof oso);
950 sset_init(&oso.targets);
951 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
953 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
955 oso.sampling_rate = *cfg->sampling;
958 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
960 oso.polling_interval = *cfg->polling;
963 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
965 oso.header_len = *cfg->header;
968 oso.sub_id = (*sflow_bridge_number)++;
969 oso.agent_device = cfg->agent;
971 oso.control_ip = NULL;
972 n_controllers = bridge_get_controllers(br, &controllers);
973 for (i = 0; i < n_controllers; i++) {
974 if (controllers[i]->local_ip) {
975 oso.control_ip = controllers[i]->local_ip;
979 ofproto_set_sflow(br->ofproto, &oso);
981 sset_destroy(&oso.targets);
984 /* Returns whether a IPFIX row is valid. */
986 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
988 return ipfix && ipfix->n_targets > 0;
991 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
993 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
994 const struct bridge *br)
996 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
999 /* Set IPFIX configuration on 'br'. */
1001 bridge_configure_ipfix(struct bridge *br)
1003 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1004 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1005 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1006 struct ofproto_ipfix_bridge_exporter_options be_opts;
1007 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1008 size_t n_fe_opts = 0;
1010 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1011 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1016 if (!valid_be_cfg && n_fe_opts == 0) {
1017 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1022 memset(&be_opts, 0, sizeof be_opts);
1024 sset_init(&be_opts.targets);
1025 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1027 if (be_cfg->sampling) {
1028 be_opts.sampling_rate = *be_cfg->sampling;
1030 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1032 if (be_cfg->obs_domain_id) {
1033 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1035 if (be_cfg->obs_point_id) {
1036 be_opts.obs_point_id = *be_cfg->obs_point_id;
1038 if (be_cfg->cache_active_timeout) {
1039 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1041 if (be_cfg->cache_max_flows) {
1042 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1046 if (n_fe_opts > 0) {
1047 struct ofproto_ipfix_flow_exporter_options *opts;
1048 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1050 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1051 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1052 opts->collector_set_id = fe_cfg->id;
1053 sset_init(&opts->targets);
1054 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1055 fe_cfg->ipfix->n_targets);
1056 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1057 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1058 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1059 ? *fe_cfg->ipfix->cache_max_flows : 0;
1065 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1069 sset_destroy(&be_opts.targets);
1072 if (n_fe_opts > 0) {
1073 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1075 for (i = 0; i < n_fe_opts; i++) {
1076 sset_destroy(&opts->targets);
1084 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1085 struct ofproto_port_stp_settings *port_s,
1086 int *port_num_counter, unsigned long *port_num_bitmap)
1088 const char *config_str;
1089 struct iface *iface;
1091 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1092 port_s->enable = false;
1095 port_s->enable = true;
1098 /* STP over bonds is not supported. */
1099 if (!list_is_singleton(&port->ifaces)) {
1100 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1102 port_s->enable = false;
1106 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1108 /* Internal ports shouldn't participate in spanning tree, so
1110 if (!strcmp(iface->type, "internal")) {
1111 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1112 port_s->enable = false;
1116 /* STP on mirror output ports is not supported. */
1117 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1118 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1119 port_s->enable = false;
1123 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1125 unsigned long int port_num = strtoul(config_str, NULL, 0);
1126 int port_idx = port_num - 1;
1128 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1129 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1130 port_s->enable = false;
1134 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1135 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1136 port->name, port_num);
1137 port_s->enable = false;
1140 bitmap_set1(port_num_bitmap, port_idx);
1141 port_s->port_num = port_idx;
1143 if (*port_num_counter >= STP_MAX_PORTS) {
1144 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1145 port_s->enable = false;
1149 port_s->port_num = (*port_num_counter)++;
1152 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1154 port_s->path_cost = strtoul(config_str, NULL, 10);
1156 enum netdev_features current;
1159 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1160 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1161 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1164 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1166 port_s->priority = strtoul(config_str, NULL, 0);
1168 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1172 /* Set spanning tree configuration on 'br'. */
1174 bridge_configure_stp(struct bridge *br)
1176 if (!br->cfg->stp_enable) {
1177 ofproto_set_stp(br->ofproto, NULL);
1179 struct ofproto_stp_settings br_s;
1180 const char *config_str;
1182 int port_num_counter;
1183 unsigned long *port_num_bitmap;
1185 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1187 uint8_t ea[ETH_ADDR_LEN];
1189 if (eth_addr_from_string(config_str, ea)) {
1190 br_s.system_id = eth_addr_to_uint64(ea);
1192 br_s.system_id = eth_addr_to_uint64(br->ea);
1193 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1194 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1197 br_s.system_id = eth_addr_to_uint64(br->ea);
1200 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1202 br_s.priority = strtoul(config_str, NULL, 0);
1204 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1207 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1209 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1211 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1214 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1216 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1218 br_s.max_age = STP_DEFAULT_MAX_AGE;
1221 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1223 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1225 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1228 /* Configure STP on the bridge. */
1229 if (ofproto_set_stp(br->ofproto, &br_s)) {
1230 VLOG_ERR("bridge %s: could not enable STP", br->name);
1234 /* Users must either set the port number with the "stp-port-num"
1235 * configuration on all ports or none. If manual configuration
1236 * is not done, then we allocate them sequentially. */
1237 port_num_counter = 0;
1238 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1239 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1240 struct ofproto_port_stp_settings port_s;
1241 struct iface *iface;
1243 port_configure_stp(br->ofproto, port, &port_s,
1244 &port_num_counter, port_num_bitmap);
1246 /* As bonds are not supported, just apply configuration to
1247 * all interfaces. */
1248 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1249 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1251 VLOG_ERR("port %s: could not enable STP", port->name);
1257 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1258 && port_num_counter) {
1259 VLOG_ERR("bridge %s: must manually configure all STP port "
1260 "IDs or none, disabling", br->name);
1261 ofproto_set_stp(br->ofproto, NULL);
1263 bitmap_free(port_num_bitmap);
1268 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1270 const struct port *port = port_lookup(br, name);
1271 return port && port_is_bond_fake_iface(port);
1275 port_is_bond_fake_iface(const struct port *port)
1277 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1281 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1283 struct bridge *br, *next;
1284 struct shash new_br;
1287 /* Collect new bridges' names and types. */
1288 shash_init(&new_br);
1289 for (i = 0; i < cfg->n_bridges; i++) {
1290 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1291 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1293 if (strchr(br_cfg->name, '/')) {
1294 /* Prevent remote ovsdb-server users from accessing arbitrary
1295 * directories, e.g. consider a bridge named "../../../etc/". */
1296 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1298 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1299 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1303 /* Get rid of deleted bridges or those whose types have changed.
1304 * Update 'cfg' of bridges that still exist. */
1305 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1306 br->cfg = shash_find_data(&new_br, br->name);
1307 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1308 br->cfg->datapath_type))) {
1313 /* Add new bridges. */
1314 for (i = 0; i < cfg->n_bridges; i++) {
1315 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1316 struct bridge *br = bridge_lookup(br_cfg->name);
1318 bridge_create(br_cfg);
1322 shash_destroy(&new_br);
1325 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1326 * Returns 0 if successful, otherwise a positive errno value. */
1328 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1329 struct netdev *netdev)
1331 return netdev_set_config(netdev, &iface_cfg->options);
1334 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1335 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1336 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1339 * If successful, returns 0 and stores the network device in '*netdevp'. On
1340 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1342 iface_do_create(const struct bridge *br,
1343 const struct ovsrec_interface *iface_cfg,
1344 const struct ovsrec_port *port_cfg,
1345 ofp_port_t *ofp_portp, struct netdev **netdevp)
1347 struct netdev *netdev = NULL;
1350 if (netdev_is_reserved_name(iface_cfg->name)) {
1351 VLOG_WARN("could not create interface %s, name is reserved",
1357 error = netdev_open(iface_cfg->name,
1358 iface_get_type(iface_cfg, br->cfg), &netdev);
1360 VLOG_WARN("could not open network device %s (%s)",
1361 iface_cfg->name, ovs_strerror(error));
1365 error = iface_set_netdev_config(iface_cfg, netdev);
1370 *ofp_portp = iface_pick_ofport(iface_cfg);
1371 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1376 VLOG_INFO("bridge %s: added interface %s on port %d",
1377 br->name, iface_cfg->name, *ofp_portp);
1379 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1380 || iface_is_internal(iface_cfg, br->cfg)) {
1381 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1389 netdev_close(netdev);
1393 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1394 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1395 * automatically allocated for the iface. Takes ownership of and
1396 * deallocates 'if_cfg'.
1398 * Return true if an iface is successfully created, false otherwise. */
1400 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1401 const struct ovsrec_port *port_cfg)
1403 struct netdev *netdev;
1404 struct iface *iface;
1405 ofp_port_t ofp_port;
1409 /* Do the bits that can fail up front.
1411 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1412 * internal datastructures may not be consistent. Eventually, when port
1413 * additions and deletions are cheaper, these calls should be removed. */
1415 ovs_assert(!iface_lookup(br, iface_cfg->name));
1416 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1419 iface_set_ofport(iface_cfg, OFPP_NONE);
1420 iface_clear_db_record(iface_cfg);
1424 /* Get or create the port structure. */
1425 port = port_lookup(br, port_cfg->name);
1427 port = port_create(br, port_cfg);
1430 /* Create the iface structure. */
1431 iface = xzalloc(sizeof *iface);
1432 list_push_back(&port->ifaces, &iface->port_elem);
1433 hmap_insert(&br->iface_by_name, &iface->name_node,
1434 hash_string(iface_cfg->name, 0));
1436 iface->name = xstrdup(iface_cfg->name);
1437 iface->ofp_port = ofp_port;
1438 iface->netdev = netdev;
1439 iface->type = iface_get_type(iface_cfg, br->cfg);
1440 iface->cfg = iface_cfg;
1441 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1442 hash_ofp_port(ofp_port));
1444 iface_set_ofport(iface->cfg, ofp_port);
1446 /* Populate initial status in database. */
1447 iface_refresh_stats(iface);
1448 iface_refresh_status(iface);
1450 /* Add bond fake iface if necessary. */
1451 if (port_is_bond_fake_iface(port)) {
1452 struct ofproto_port ofproto_port;
1454 if (ofproto_port_query_by_name(br->ofproto, port->name,
1456 struct netdev *netdev;
1459 error = netdev_open(port->name, "internal", &netdev);
1461 ofp_port_t fake_ofp_port = iface_pick_ofport(iface_cfg);
1462 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1463 netdev_close(netdev);
1465 VLOG_WARN("could not open network device %s (%s)",
1466 port->name, ovs_strerror(error));
1469 /* Already exists, nothing to do. */
1470 ofproto_port_destroy(&ofproto_port);
1477 /* Set forward BPDU option. */
1479 bridge_configure_forward_bpdu(struct bridge *br)
1481 ofproto_set_forward_bpdu(br->ofproto,
1482 smap_get_bool(&br->cfg->other_config,
1487 /* Set MAC learning table configuration for 'br'. */
1489 bridge_configure_mac_table(struct bridge *br)
1491 const char *idle_time_str;
1494 const char *mac_table_size_str;
1497 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1498 idle_time = (idle_time_str && atoi(idle_time_str)
1499 ? atoi(idle_time_str)
1500 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1502 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1503 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1504 ? atoi(mac_table_size_str)
1507 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1511 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1512 struct iface **hw_addr_iface)
1514 struct hmapx mirror_output_ports;
1517 bool found_addr = false;
1521 *hw_addr_iface = NULL;
1523 /* Did the user request a particular MAC? */
1524 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1525 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1526 if (eth_addr_is_multicast(ea)) {
1527 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1528 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1529 } else if (eth_addr_is_zero(ea)) {
1530 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1536 /* Mirror output ports don't participate in picking the local hardware
1537 * address. ofproto can't help us find out whether a given port is a
1538 * mirror output because we haven't configured mirrors yet, so we need to
1539 * accumulate them ourselves. */
1540 hmapx_init(&mirror_output_ports);
1541 for (i = 0; i < br->cfg->n_mirrors; i++) {
1542 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1543 if (m->output_port) {
1544 hmapx_add(&mirror_output_ports, m->output_port);
1548 /* Otherwise choose the minimum non-local MAC address among all of the
1550 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1551 uint8_t iface_ea[ETH_ADDR_LEN];
1552 struct iface *candidate;
1553 struct iface *iface;
1555 /* Mirror output ports don't participate. */
1556 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1560 /* Choose the MAC address to represent the port. */
1562 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1563 /* Find the interface with this Ethernet address (if any) so that
1564 * we can provide the correct devname to the caller. */
1565 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1566 uint8_t candidate_ea[ETH_ADDR_LEN];
1567 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1568 && eth_addr_equals(iface_ea, candidate_ea)) {
1573 /* Choose the interface whose MAC address will represent the port.
1574 * The Linux kernel bonding code always chooses the MAC address of
1575 * the first slave added to a bond, and the Fedora networking
1576 * scripts always add slaves to a bond in alphabetical order, so
1577 * for compatibility we choose the interface with the name that is
1578 * first in alphabetical order. */
1579 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1580 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1585 /* The local port doesn't count (since we're trying to choose its
1586 * MAC address anyway). */
1587 if (iface->ofp_port == OFPP_LOCAL) {
1592 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1598 /* Compare against our current choice. */
1599 if (!eth_addr_is_multicast(iface_ea) &&
1600 !eth_addr_is_local(iface_ea) &&
1601 !eth_addr_is_reserved(iface_ea) &&
1602 !eth_addr_is_zero(iface_ea) &&
1603 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1605 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1606 *hw_addr_iface = iface;
1612 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1613 *hw_addr_iface = NULL;
1616 hmapx_destroy(&mirror_output_ports);
1619 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1620 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1621 * an interface on 'br', then that interface must be passed in as
1622 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1623 * 'hw_addr_iface' must be passed in as a null pointer. */
1625 bridge_pick_datapath_id(struct bridge *br,
1626 const uint8_t bridge_ea[ETH_ADDR_LEN],
1627 struct iface *hw_addr_iface)
1630 * The procedure for choosing a bridge MAC address will, in the most
1631 * ordinary case, also choose a unique MAC that we can use as a datapath
1632 * ID. In some special cases, though, multiple bridges will end up with
1633 * the same MAC address. This is OK for the bridges, but it will confuse
1634 * the OpenFlow controller, because each datapath needs a unique datapath
1637 * Datapath IDs must be unique. It is also very desirable that they be
1638 * stable from one run to the next, so that policy set on a datapath
1641 const char *datapath_id;
1644 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1645 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1649 if (!hw_addr_iface) {
1651 * A purely internal bridge, that is, one that has no non-virtual
1652 * network devices on it at all, is difficult because it has no
1653 * natural unique identifier at all.
1655 * When the host is a XenServer, we handle this case by hashing the
1656 * host's UUID with the name of the bridge. Names of bridges are
1657 * persistent across XenServer reboots, although they can be reused if
1658 * an internal network is destroyed and then a new one is later
1659 * created, so this is fairly effective.
1661 * When the host is not a XenServer, we punt by using a random MAC
1662 * address on each run.
1664 const char *host_uuid = xenserver_get_host_uuid();
1666 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1667 dpid = dpid_from_hash(combined, strlen(combined));
1673 return eth_addr_to_uint64(bridge_ea);
1677 dpid_from_hash(const void *data, size_t n)
1679 uint8_t hash[SHA1_DIGEST_SIZE];
1681 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1682 sha1_bytes(data, n, hash);
1683 eth_addr_mark_random(hash);
1684 return eth_addr_to_uint64(hash);
1688 iface_refresh_status(struct iface *iface)
1692 enum netdev_features current;
1696 uint8_t mac[ETH_ADDR_LEN];
1700 if (iface_is_synthetic(iface)) {
1706 if (!netdev_get_status(iface->netdev, &smap)) {
1707 ovsrec_interface_set_status(iface->cfg, &smap);
1709 ovsrec_interface_set_status(iface->cfg, NULL);
1712 smap_destroy(&smap);
1714 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1715 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1717 ovsrec_interface_set_duplex(iface->cfg,
1718 netdev_features_is_full_duplex(current)
1720 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1722 ovsrec_interface_set_duplex(iface->cfg, NULL);
1723 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1726 error = netdev_get_mtu(iface->netdev, &mtu);
1729 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1731 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1734 error = netdev_get_etheraddr(iface->netdev, mac);
1736 char mac_string[32];
1738 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1739 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1741 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1744 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1745 * if there is no valid ifindex number. */
1746 ifindex64 = netdev_get_ifindex(iface->netdev);
1747 if (ifindex64 < 0) {
1750 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1753 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1756 iface_refresh_cfm_stats(struct iface *iface)
1758 const struct ovsrec_interface *cfg = iface->cfg;
1759 struct ofproto_cfm_status status;
1761 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1762 iface->ofp_port, &status)) {
1763 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1764 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1765 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1766 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1767 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1768 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1770 const char *reasons[CFM_FAULT_N_REASONS];
1771 int64_t cfm_health = status.health;
1772 int64_t cfm_flap_count = status.flap_count;
1773 bool faulted = status.faults != 0;
1776 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1779 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1780 int reason = 1 << i;
1781 if (status.faults & reason) {
1782 reasons[j++] = cfm_fault_reason_to_str(reason);
1785 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1787 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1789 if (status.remote_opstate >= 0) {
1790 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1791 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1793 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1796 ovsrec_interface_set_cfm_remote_mpids(cfg,
1797 (const int64_t *)status.rmps,
1799 if (cfm_health >= 0) {
1800 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1802 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1810 iface_refresh_stats(struct iface *iface)
1812 #define IFACE_STATS \
1813 IFACE_STAT(rx_packets, "rx_packets") \
1814 IFACE_STAT(tx_packets, "tx_packets") \
1815 IFACE_STAT(rx_bytes, "rx_bytes") \
1816 IFACE_STAT(tx_bytes, "tx_bytes") \
1817 IFACE_STAT(rx_dropped, "rx_dropped") \
1818 IFACE_STAT(tx_dropped, "tx_dropped") \
1819 IFACE_STAT(rx_errors, "rx_errors") \
1820 IFACE_STAT(tx_errors, "tx_errors") \
1821 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1822 IFACE_STAT(rx_over_errors, "rx_over_err") \
1823 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1824 IFACE_STAT(collisions, "collisions")
1826 #define IFACE_STAT(MEMBER, NAME) + 1
1827 enum { N_IFACE_STATS = IFACE_STATS };
1829 int64_t values[N_IFACE_STATS];
1830 char *keys[N_IFACE_STATS];
1833 struct netdev_stats stats;
1835 if (iface_is_synthetic(iface)) {
1839 /* Intentionally ignore return value, since errors will set 'stats' to
1840 * all-1s, and we will deal with that correctly below. */
1841 netdev_get_stats(iface->netdev, &stats);
1843 /* Copy statistics into keys[] and values[]. */
1845 #define IFACE_STAT(MEMBER, NAME) \
1846 if (stats.MEMBER != UINT64_MAX) { \
1848 values[n] = stats.MEMBER; \
1853 ovs_assert(n <= N_IFACE_STATS);
1855 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1860 br_refresh_stp_status(struct bridge *br)
1862 struct smap smap = SMAP_INITIALIZER(&smap);
1863 struct ofproto *ofproto = br->ofproto;
1864 struct ofproto_stp_status status;
1866 if (ofproto_get_stp_status(ofproto, &status)) {
1870 if (!status.enabled) {
1871 ovsrec_bridge_set_status(br->cfg, NULL);
1875 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1876 STP_ID_ARGS(status.bridge_id));
1877 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1878 STP_ID_ARGS(status.designated_root));
1879 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1881 ovsrec_bridge_set_status(br->cfg, &smap);
1882 smap_destroy(&smap);
1886 port_refresh_stp_status(struct port *port)
1888 struct ofproto *ofproto = port->bridge->ofproto;
1889 struct iface *iface;
1890 struct ofproto_port_stp_status status;
1893 if (port_is_synthetic(port)) {
1897 /* STP doesn't currently support bonds. */
1898 if (!list_is_singleton(&port->ifaces)) {
1899 ovsrec_port_set_status(port->cfg, NULL);
1903 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1904 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1908 if (!status.enabled) {
1909 ovsrec_port_set_status(port->cfg, NULL);
1913 /* Set Status column. */
1915 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1916 smap_add(&smap, "stp_state", stp_state_name(status.state));
1917 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1918 smap_add(&smap, "stp_role", stp_role_name(status.role));
1919 ovsrec_port_set_status(port->cfg, &smap);
1920 smap_destroy(&smap);
1924 port_refresh_stp_stats(struct port *port)
1926 struct ofproto *ofproto = port->bridge->ofproto;
1927 struct iface *iface;
1928 struct ofproto_port_stp_stats stats;
1930 int64_t int_values[3];
1932 if (port_is_synthetic(port)) {
1936 /* STP doesn't currently support bonds. */
1937 if (!list_is_singleton(&port->ifaces)) {
1941 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1942 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
1946 if (!stats.enabled) {
1947 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1951 /* Set Statistics column. */
1952 keys[0] = "stp_tx_count";
1953 int_values[0] = stats.tx_count;
1954 keys[1] = "stp_rx_count";
1955 int_values[1] = stats.rx_count;
1956 keys[2] = "stp_error_count";
1957 int_values[2] = stats.error_count;
1959 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1960 ARRAY_SIZE(int_values));
1964 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1966 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1970 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1972 bool enable = enable_system_stats(cfg);
1974 system_stats_enable(enable);
1976 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1981 run_system_stats(void)
1983 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1986 stats = system_stats_run();
1988 struct ovsdb_idl_txn *txn;
1989 struct ovsdb_datum datum;
1991 txn = ovsdb_idl_txn_create(idl);
1992 ovsdb_datum_from_smap(&datum, stats);
1993 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1995 ovsdb_idl_txn_commit(txn);
1996 ovsdb_idl_txn_destroy(txn);
2002 static inline const char *
2003 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2006 case OFPCR12_ROLE_EQUAL:
2008 case OFPCR12_ROLE_MASTER:
2010 case OFPCR12_ROLE_SLAVE:
2012 case OFPCR12_ROLE_NOCHANGE:
2014 return "*** INVALID ROLE ***";
2019 refresh_controller_status(void)
2023 const struct ovsrec_controller *cfg;
2027 /* Accumulate status for controllers on all bridges. */
2028 HMAP_FOR_EACH (br, node, &all_bridges) {
2029 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2032 /* Update each controller in the database with current status. */
2033 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2034 struct ofproto_controller_info *cinfo =
2035 shash_find_data(&info, cfg->target);
2038 struct smap smap = SMAP_INITIALIZER(&smap);
2039 const char **values = cinfo->pairs.values;
2040 const char **keys = cinfo->pairs.keys;
2043 for (i = 0; i < cinfo->pairs.n; i++) {
2044 smap_add(&smap, keys[i], values[i]);
2047 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2048 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2050 ovsrec_controller_set_status(cfg, &smap);
2051 smap_destroy(&smap);
2053 ovsrec_controller_set_is_connected(cfg, false);
2054 ovsrec_controller_set_role(cfg, NULL);
2055 ovsrec_controller_set_status(cfg, NULL);
2059 ofproto_free_ofproto_controller_info(&info);
2064 * Some information in the database must be kept as up-to-date as possible to
2065 * allow controllers to respond rapidly to network outages. We call these
2066 * statistics "instant" stats.
2068 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2069 * assuming that they've changed. The only means we have to determine whether
2070 * they have changed are:
2072 * - Try to commit changes to the database. If nothing changed, then
2073 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2076 * - instant_stats_run() is called late in the run loop, after anything that
2077 * might change any of the instant stats.
2079 * We use these two facts together to avoid waking the process up every
2080 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2083 /* Minimum interval between writing updates to the instant stats to the
2085 #define INSTANT_INTERVAL_MSEC 100
2087 /* Current instant stats database transaction, NULL if there is no ongoing
2089 static struct ovsdb_idl_txn *instant_txn;
2091 /* Next time (in msec on monotonic clock) at which we will update the instant
2093 static long long int instant_next_txn = LLONG_MIN;
2095 /* True if the run loop has run since we last saw that the instant stats were
2096 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2097 * to refresh the instant stats. */
2098 static bool instant_stats_could_have_changed;
2101 instant_stats_run(void)
2103 enum ovsdb_idl_txn_status status;
2105 instant_stats_could_have_changed = true;
2110 if (time_msec() < instant_next_txn) {
2113 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2115 instant_txn = ovsdb_idl_txn_create(idl);
2116 HMAP_FOR_EACH (br, node, &all_bridges) {
2117 struct iface *iface;
2120 br_refresh_stp_status(br);
2122 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2123 port_refresh_stp_status(port);
2126 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2127 enum netdev_flags flags;
2129 const char *link_state;
2130 int64_t link_resets;
2133 if (iface_is_synthetic(iface)) {
2137 current = ofproto_port_is_lacp_current(br->ofproto,
2141 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2143 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2146 error = netdev_get_flags(iface->netdev, &flags);
2148 const char *state = flags & NETDEV_UP ? "up" : "down";
2149 ovsrec_interface_set_admin_state(iface->cfg, state);
2151 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2154 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2155 ovsrec_interface_set_link_state(iface->cfg, link_state);
2157 link_resets = netdev_get_carrier_resets(iface->netdev);
2158 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2160 iface_refresh_cfm_stats(iface);
2163 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2165 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2166 smap_destroy(&smap);
2171 status = ovsdb_idl_txn_commit(instant_txn);
2172 if (status != TXN_INCOMPLETE) {
2173 ovsdb_idl_txn_destroy(instant_txn);
2176 if (status == TXN_UNCHANGED) {
2177 instant_stats_could_have_changed = false;
2182 instant_stats_wait(void)
2185 ovsdb_idl_txn_wait(instant_txn);
2186 } else if (instant_stats_could_have_changed) {
2187 poll_timer_wait_until(instant_next_txn);
2191 /* Performs periodic activity required by bridges that needs to be done with
2192 * the least possible latency.
2194 * It makes sense to call this function a couple of times per poll loop, to
2195 * provide a significant performance boost on some benchmarks with ofprotos
2196 * that use the ofproto-dpif implementation. */
2198 bridge_run_fast(void)
2205 ofproto_enumerate_types(&types);
2206 SSET_FOR_EACH (type, &types) {
2207 ofproto_type_run_fast(type);
2209 sset_destroy(&types);
2211 HMAP_FOR_EACH (br, node, &all_bridges) {
2212 ofproto_run_fast(br->ofproto);
2219 static struct ovsrec_open_vswitch null_cfg;
2220 const struct ovsrec_open_vswitch *cfg;
2224 bool vlan_splinters_changed;
2227 ovsrec_open_vswitch_init(&null_cfg);
2231 if (ovsdb_idl_is_lock_contended(idl)) {
2232 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2233 struct bridge *br, *next_br;
2235 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2236 "disabling this process (pid %ld) until it goes away",
2237 (long int) getpid());
2239 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2242 /* Since we will not be running system_stats_run() in this process
2243 * with the current situation of multiple ovs-vswitchd daemons,
2244 * disable system stats collection. */
2245 system_stats_enable(false);
2247 } else if (!ovsdb_idl_has_lock(idl)) {
2250 cfg = ovsrec_open_vswitch_first(idl);
2252 /* Initialize the ofproto library. This only needs to run once, but
2253 * it must be done after the configuration is set. If the
2254 * initialization has already occurred, bridge_init_ofproto()
2255 * returns immediately. */
2256 bridge_init_ofproto(cfg);
2258 /* Once the value of flow-restore-wait is false, we no longer should
2259 * check its value from the database. */
2260 if (cfg && ofproto_get_flow_restore_wait()) {
2261 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2262 "flow-restore-wait", false));
2265 /* Let each datapath type do the work that it needs to do. */
2267 ofproto_enumerate_types(&types);
2268 SSET_FOR_EACH (type, &types) {
2269 ofproto_type_run(type);
2271 sset_destroy(&types);
2273 /* Let each bridge do the work that it needs to do. */
2274 HMAP_FOR_EACH (br, node, &all_bridges) {
2275 ofproto_run(br->ofproto);
2278 /* Re-configure SSL. We do this on every trip through the main loop,
2279 * instead of just when the database changes, because the contents of the
2280 * key and certificate files can change without the database changing.
2282 * We do this before bridge_reconfigure() because that function might
2283 * initiate SSL connections and thus requires SSL to be configured. */
2284 if (cfg && cfg->ssl) {
2285 const struct ovsrec_ssl *ssl = cfg->ssl;
2287 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2288 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2291 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2292 * usage has changed. */
2293 vlan_splinters_changed = false;
2294 if (vlan_splinters_enabled_anywhere) {
2295 HMAP_FOR_EACH (br, node, &all_bridges) {
2296 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2297 vlan_splinters_changed = true;
2303 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2304 struct ovsdb_idl_txn *txn;
2306 idl_seqno = ovsdb_idl_get_seqno(idl);
2307 txn = ovsdb_idl_txn_create(idl);
2308 bridge_reconfigure(cfg ? cfg : &null_cfg);
2311 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2314 /* If we are completing our initial configuration for this run
2315 * of ovs-vswitchd, then keep the transaction around to monitor
2316 * it for completion. */
2317 if (initial_config_done) {
2318 ovsdb_idl_txn_commit(txn);
2319 ovsdb_idl_txn_destroy(txn);
2321 initial_config_done = true;
2322 daemonize_txn = txn;
2326 if (daemonize_txn) {
2327 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2328 if (status != TXN_INCOMPLETE) {
2329 ovsdb_idl_txn_destroy(daemonize_txn);
2330 daemonize_txn = NULL;
2332 /* ovs-vswitchd has completed initialization, so allow the
2333 * process that forked us to exit successfully. */
2334 daemonize_complete();
2336 vlog_enable_async();
2338 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2342 /* Refresh interface and mirror stats if necessary. */
2343 if (time_msec() >= iface_stats_timer) {
2345 struct ovsdb_idl_txn *txn;
2347 txn = ovsdb_idl_txn_create(idl);
2348 HMAP_FOR_EACH (br, node, &all_bridges) {
2352 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2353 struct iface *iface;
2355 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2356 iface_refresh_stats(iface);
2357 iface_refresh_status(iface);
2360 port_refresh_stp_stats(port);
2363 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2364 mirror_refresh_stats(m);
2368 refresh_controller_status();
2369 ovsdb_idl_txn_commit(txn);
2370 ovsdb_idl_txn_destroy(txn); /* XXX */
2373 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2377 instant_stats_run();
2386 ovsdb_idl_wait(idl);
2387 if (daemonize_txn) {
2388 ovsdb_idl_txn_wait(daemonize_txn);
2392 ofproto_enumerate_types(&types);
2393 SSET_FOR_EACH (type, &types) {
2394 ofproto_type_wait(type);
2396 sset_destroy(&types);
2398 if (!hmap_is_empty(&all_bridges)) {
2401 HMAP_FOR_EACH (br, node, &all_bridges) {
2402 ofproto_wait(br->ofproto);
2404 poll_timer_wait_until(iface_stats_timer);
2407 system_stats_wait();
2408 instant_stats_wait();
2411 /* Adds some memory usage statistics for bridges into 'usage', for use with
2412 * memory_report(). */
2414 bridge_get_memory_usage(struct simap *usage)
2418 HMAP_FOR_EACH (br, node, &all_bridges) {
2419 ofproto_get_memory_usage(br->ofproto, usage);
2423 /* QoS unixctl user interface functions. */
2425 struct qos_unixctl_show_cbdata {
2427 struct iface *iface;
2431 qos_unixctl_show_queue(unsigned int queue_id,
2432 const struct smap *details,
2433 struct iface *iface,
2436 struct netdev_queue_stats stats;
2437 struct smap_node *node;
2440 ds_put_cstr(ds, "\n");
2442 ds_put_format(ds, "Queue %u:\n", queue_id);
2444 ds_put_cstr(ds, "Default:\n");
2447 SMAP_FOR_EACH (node, details) {
2448 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2451 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2453 if (stats.tx_packets != UINT64_MAX) {
2454 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2457 if (stats.tx_bytes != UINT64_MAX) {
2458 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2461 if (stats.tx_errors != UINT64_MAX) {
2462 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2465 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2466 queue_id, ovs_strerror(error));
2471 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2472 const char *argv[], void *aux OVS_UNUSED)
2474 struct ds ds = DS_EMPTY_INITIALIZER;
2475 struct smap smap = SMAP_INITIALIZER(&smap);
2476 struct iface *iface;
2478 struct smap_node *node;
2480 iface = iface_find(argv[1]);
2482 unixctl_command_reply_error(conn, "no such interface");
2486 netdev_get_qos(iface->netdev, &type, &smap);
2488 if (*type != '\0') {
2489 struct netdev_queue_dump dump;
2490 struct smap details;
2491 unsigned int queue_id;
2493 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2495 SMAP_FOR_EACH (node, &smap) {
2496 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2499 smap_init(&details);
2500 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2501 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2503 smap_destroy(&details);
2505 unixctl_command_reply(conn, ds_cstr(&ds));
2507 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2508 unixctl_command_reply_error(conn, ds_cstr(&ds));
2511 smap_destroy(&smap);
2515 /* Bridge reconfiguration functions. */
2517 bridge_create(const struct ovsrec_bridge *br_cfg)
2521 ovs_assert(!bridge_lookup(br_cfg->name));
2522 br = xzalloc(sizeof *br);
2524 br->name = xstrdup(br_cfg->name);
2525 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2528 /* Derive the default Ethernet address from the bridge's UUID. This should
2529 * be unique and it will be stable between ovs-vswitchd runs. */
2530 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2531 eth_addr_mark_random(br->default_ea);
2533 hmap_init(&br->ports);
2534 hmap_init(&br->ifaces);
2535 hmap_init(&br->iface_by_name);
2536 hmap_init(&br->mirrors);
2538 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2542 bridge_destroy(struct bridge *br)
2545 struct mirror *mirror, *next_mirror;
2546 struct port *port, *next_port;
2548 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2551 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2552 mirror_destroy(mirror);
2555 hmap_remove(&all_bridges, &br->node);
2556 ofproto_destroy(br->ofproto);
2557 hmap_destroy(&br->ifaces);
2558 hmap_destroy(&br->ports);
2559 hmap_destroy(&br->iface_by_name);
2560 hmap_destroy(&br->mirrors);
2567 static struct bridge *
2568 bridge_lookup(const char *name)
2572 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2573 if (!strcmp(br->name, name)) {
2580 /* Handle requests for a listing of all flows known by the OpenFlow
2581 * stack, including those normally hidden. */
2583 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2584 const char *argv[], void *aux OVS_UNUSED)
2589 br = bridge_lookup(argv[1]);
2591 unixctl_command_reply_error(conn, "Unknown bridge");
2596 ofproto_get_all_flows(br->ofproto, &results);
2598 unixctl_command_reply(conn, ds_cstr(&results));
2599 ds_destroy(&results);
2602 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2603 * connections and reconnect. If BRIDGE is not specified, then all bridges
2604 * drop their controller connections and reconnect. */
2606 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2607 const char *argv[], void *aux OVS_UNUSED)
2611 br = bridge_lookup(argv[1]);
2613 unixctl_command_reply_error(conn, "Unknown bridge");
2616 ofproto_reconnect_controllers(br->ofproto);
2618 HMAP_FOR_EACH (br, node, &all_bridges) {
2619 ofproto_reconnect_controllers(br->ofproto);
2622 unixctl_command_reply(conn, NULL);
2626 bridge_get_controllers(const struct bridge *br,
2627 struct ovsrec_controller ***controllersp)
2629 struct ovsrec_controller **controllers;
2630 size_t n_controllers;
2632 controllers = br->cfg->controller;
2633 n_controllers = br->cfg->n_controller;
2635 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2641 *controllersp = controllers;
2643 return n_controllers;
2647 bridge_collect_wanted_ports(struct bridge *br,
2648 const unsigned long int *splinter_vlans,
2649 struct shash *wanted_ports)
2653 shash_init(wanted_ports);
2655 for (i = 0; i < br->cfg->n_ports; i++) {
2656 const char *name = br->cfg->ports[i]->name;
2657 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2658 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2662 if (bridge_get_controllers(br, NULL)
2663 && !shash_find(wanted_ports, br->name)) {
2664 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2665 br->name, br->name);
2667 ovsrec_interface_init(&br->synth_local_iface);
2668 ovsrec_port_init(&br->synth_local_port);
2670 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2671 br->synth_local_port.n_interfaces = 1;
2672 br->synth_local_port.name = br->name;
2674 br->synth_local_iface.name = br->name;
2675 br->synth_local_iface.type = "internal";
2677 br->synth_local_ifacep = &br->synth_local_iface;
2679 shash_add(wanted_ports, br->name, &br->synth_local_port);
2682 if (splinter_vlans) {
2683 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2687 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2688 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2689 * 'br' needs to complete its configuration. */
2691 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2693 struct shash_node *port_node;
2694 struct port *port, *next;
2696 /* Get rid of deleted ports.
2697 * Get rid of deleted interfaces on ports that still exist. */
2698 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2699 port->cfg = shash_find_data(wanted_ports, port->name);
2703 port_del_ifaces(port);
2707 /* Update iface->cfg and iface->type in interfaces that still exist. */
2708 SHASH_FOR_EACH (port_node, wanted_ports) {
2709 const struct ovsrec_port *port = port_node->data;
2712 for (i = 0; i < port->n_interfaces; i++) {
2713 const struct ovsrec_interface *cfg = port->interfaces[i];
2714 struct iface *iface = iface_lookup(br, cfg->name);
2715 const char *type = iface_get_type(cfg, br->cfg);
2720 } else if (!strcmp(type, "null")) {
2721 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2722 " may be removed in February 2013. Please email"
2723 " dev@openvswitch.org with concerns.",
2726 /* We will add new interfaces later. */
2732 /* Initializes 'oc' appropriately as a management service controller for
2735 * The caller must free oc->target when it is no longer needed. */
2737 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2738 struct ofproto_controller *oc)
2740 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2741 oc->max_backoff = 0;
2742 oc->probe_interval = 60;
2743 oc->band = OFPROTO_OUT_OF_BAND;
2745 oc->burst_limit = 0;
2746 oc->enable_async_msgs = true;
2749 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2751 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2752 struct ofproto_controller *oc)
2756 oc->target = c->target;
2757 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2758 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2759 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2760 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2761 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2762 oc->burst_limit = (c->controller_burst_limit
2763 ? *c->controller_burst_limit : 0);
2764 oc->enable_async_msgs = (!c->enable_async_messages
2765 || *c->enable_async_messages);
2766 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2767 if (dscp < 0 || dscp > 63) {
2768 dscp = DSCP_DEFAULT;
2773 /* Configures the IP stack for 'br''s local interface properly according to the
2774 * configuration in 'c'. */
2776 bridge_configure_local_iface_netdev(struct bridge *br,
2777 struct ovsrec_controller *c)
2779 struct netdev *netdev;
2780 struct in_addr mask, gateway;
2782 struct iface *local_iface;
2785 /* If there's no local interface or no IP address, give up. */
2786 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2787 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2791 /* Bring up the local interface. */
2792 netdev = local_iface->netdev;
2793 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2795 /* Configure the IP address and netmask. */
2796 if (!c->local_netmask
2797 || !inet_aton(c->local_netmask, &mask)
2799 mask.s_addr = guess_netmask(ip.s_addr);
2801 if (!netdev_set_in4(netdev, ip, mask)) {
2802 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2803 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2806 /* Configure the default gateway. */
2807 if (c->local_gateway
2808 && inet_aton(c->local_gateway, &gateway)
2809 && gateway.s_addr) {
2810 if (!netdev_add_router(netdev, gateway)) {
2811 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2812 br->name, IP_ARGS(gateway.s_addr));
2817 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2818 * in either string are treated as equal to any number of slashes in the other,
2819 * e.g. "x///y" is equal to "x/y".
2821 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2822 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2823 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2824 * 'b' against a prefix of 'a'.
2827 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2829 const char *b_start = b;
2831 if (b - b_start >= b_stoplen) {
2833 } else if (*a != *b) {
2835 } else if (*a == '/') {
2836 a += strspn(a, "/");
2837 b += strspn(b, "/");
2838 } else if (*a == '\0') {
2848 bridge_configure_remotes(struct bridge *br,
2849 const struct sockaddr_in *managers, size_t n_managers)
2851 bool disable_in_band;
2853 struct ovsrec_controller **controllers;
2854 size_t n_controllers;
2856 enum ofproto_fail_mode fail_mode;
2858 struct ofproto_controller *ocs;
2862 /* Check if we should disable in-band control on this bridge. */
2863 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2866 /* Set OpenFlow queue ID for in-band control. */
2867 ofproto_set_in_band_queue(br->ofproto,
2868 smap_get_int(&br->cfg->other_config,
2869 "in-band-queue", -1));
2871 if (disable_in_band) {
2872 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2874 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2877 n_controllers = bridge_get_controllers(br, &controllers);
2879 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2882 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2883 for (i = 0; i < n_controllers; i++) {
2884 struct ovsrec_controller *c = controllers[i];
2886 if (!strncmp(c->target, "punix:", 6)
2887 || !strncmp(c->target, "unix:", 5)) {
2888 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2891 if (!strncmp(c->target, "unix:", 5)) {
2892 /* Connect to a listening socket */
2893 whitelist = xasprintf("unix:%s/", ovs_rundir());
2894 if (strchr(c->target, '/') &&
2895 !equal_pathnames(c->target, whitelist,
2896 strlen(whitelist))) {
2897 /* Absolute path specified, but not in ovs_rundir */
2898 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2899 "controller \"%s\" due to possibility for "
2900 "remote exploit. Instead, specify socket "
2901 "in whitelisted \"%s\" or connect to "
2902 "\"unix:%s/%s.mgmt\" (which is always "
2903 "available without special configuration).",
2904 br->name, c->target, whitelist,
2905 ovs_rundir(), br->name);
2910 whitelist = xasprintf("punix:%s/%s.controller",
2911 ovs_rundir(), br->name);
2912 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2913 /* Prevent remote ovsdb-server users from accessing
2914 * arbitrary Unix domain sockets and overwriting arbitrary
2916 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2917 "controller \"%s\" due to possibility of "
2918 "overwriting local files. Instead, specify "
2919 "whitelisted \"%s\" or connect to "
2920 "\"unix:%s/%s.mgmt\" (which is always "
2921 "available without special configuration).",
2922 br->name, c->target, whitelist,
2923 ovs_rundir(), br->name);
2932 bridge_configure_local_iface_netdev(br, c);
2933 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2934 if (disable_in_band) {
2935 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2940 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
2941 bridge_get_allowed_versions(br));
2942 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2945 /* Set the fail-mode. */
2946 fail_mode = !br->cfg->fail_mode
2947 || !strcmp(br->cfg->fail_mode, "standalone")
2948 ? OFPROTO_FAIL_STANDALONE
2949 : OFPROTO_FAIL_SECURE;
2950 ofproto_set_fail_mode(br->ofproto, fail_mode);
2952 /* Configure OpenFlow controller connection snooping. */
2953 if (!ofproto_has_snoops(br->ofproto)) {
2957 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2958 ovs_rundir(), br->name));
2959 ofproto_set_snoops(br->ofproto, &snoops);
2960 sset_destroy(&snoops);
2965 bridge_configure_tables(struct bridge *br)
2967 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2971 n_tables = ofproto_get_n_tables(br->ofproto);
2973 for (i = 0; i < n_tables; i++) {
2974 struct ofproto_table_settings s;
2977 s.max_flows = UINT_MAX;
2980 s.n_prefix_fields = 0;
2981 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
2983 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2984 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2987 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2988 s.max_flows = *cfg->flow_limit;
2990 if (cfg->overflow_policy
2991 && !strcmp(cfg->overflow_policy, "evict")) {
2993 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2994 for (k = 0; k < cfg->n_groups; k++) {
2995 const char *string = cfg->groups[k];
2998 msg = mf_parse_subfield__(&s.groups[k], &string);
3000 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3001 "'groups' (%s)", br->name, i, msg);
3003 } else if (*string) {
3004 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3005 "element '%s' contains trailing garbage",
3006 br->name, i, cfg->groups[k]);
3012 /* Prefix lookup fields. */
3013 s.n_prefix_fields = 0;
3014 for (k = 0; k < cfg->n_prefixes; k++) {
3015 const char *name = cfg->prefixes[k];
3016 const struct mf_field *mf = mf_from_name(name);
3018 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3022 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3023 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3024 "%s", br->name, name);
3027 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3028 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3029 "field not used: %s", br->name, name);
3032 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3034 if (s.n_prefix_fields > 0) {
3036 struct ds ds = DS_EMPTY_INITIALIZER;
3037 for (k = 0; k < s.n_prefix_fields; k++) {
3039 ds_put_char(&ds, ',');
3041 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3043 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3044 br->name, i, ds_cstr(&ds));
3049 ofproto_configure_table(br->ofproto, i, &s);
3053 for (; j < br->cfg->n_flow_tables; j++) {
3054 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3055 "%"PRId64" not supported by this datapath", br->name,
3056 br->cfg->key_flow_tables[j]);
3061 bridge_configure_dp_desc(struct bridge *br)
3063 ofproto_set_dp_desc(br->ofproto,
3064 smap_get(&br->cfg->other_config, "dp-desc"));
3067 /* Port functions. */
3069 static void iface_destroy__(struct iface *);
3071 static struct port *
3072 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3076 port = xzalloc(sizeof *port);
3078 port->name = xstrdup(cfg->name);
3080 list_init(&port->ifaces);
3082 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3086 /* Deletes interfaces from 'port' that are no longer configured for it. */
3088 port_del_ifaces(struct port *port)
3090 struct iface *iface, *next;
3091 struct sset new_ifaces;
3094 /* Collect list of new interfaces. */
3095 sset_init(&new_ifaces);
3096 for (i = 0; i < port->cfg->n_interfaces; i++) {
3097 const char *name = port->cfg->interfaces[i]->name;
3098 const char *type = port->cfg->interfaces[i]->type;
3099 if (strcmp(type, "null")) {
3100 sset_add(&new_ifaces, name);
3104 /* Get rid of deleted interfaces. */
3105 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3106 if (!sset_contains(&new_ifaces, iface->name)) {
3107 iface_destroy(iface);
3111 sset_destroy(&new_ifaces);
3115 port_destroy(struct port *port)
3118 struct bridge *br = port->bridge;
3119 struct iface *iface, *next;
3122 ofproto_bundle_unregister(br->ofproto, port);
3125 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3126 iface_destroy__(iface);
3129 hmap_remove(&br->ports, &port->hmap_node);
3135 static struct port *
3136 port_lookup(const struct bridge *br, const char *name)
3140 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3142 if (!strcmp(port->name, name)) {
3150 enable_lacp(struct port *port, bool *activep)
3152 if (!port->cfg->lacp) {
3153 /* XXX when LACP implementation has been sufficiently tested, enable by
3154 * default and make active on bonded ports. */
3156 } else if (!strcmp(port->cfg->lacp, "off")) {
3158 } else if (!strcmp(port->cfg->lacp, "active")) {
3161 } else if (!strcmp(port->cfg->lacp, "passive")) {
3165 VLOG_WARN("port %s: unknown LACP mode %s",
3166 port->name, port->cfg->lacp);
3171 static struct lacp_settings *
3172 port_configure_lacp(struct port *port, struct lacp_settings *s)
3174 const char *lacp_time, *system_id;
3177 if (!enable_lacp(port, &s->active)) {
3181 s->name = port->name;
3183 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3185 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3186 ETH_ADDR_SCAN_ARGS(s->id))) {
3187 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3188 " address.", port->name, system_id);
3192 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3195 if (eth_addr_is_zero(s->id)) {
3196 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3200 /* Prefer bondable links if unspecified. */
3201 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3203 s->priority = (priority > 0 && priority <= UINT16_MAX
3205 : UINT16_MAX - !list_is_short(&port->ifaces));
3207 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3208 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3210 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3211 "lacp-fallback-ab", false);
3217 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3219 int priority, portid, key;
3221 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3222 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3224 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3226 if (portid <= 0 || portid > UINT16_MAX) {
3227 portid = ofp_to_u16(iface->ofp_port);
3230 if (priority <= 0 || priority > UINT16_MAX) {
3231 priority = UINT16_MAX;
3234 if (key < 0 || key > UINT16_MAX) {
3238 s->name = iface->name;
3240 s->priority = priority;
3245 port_configure_bond(struct port *port, struct bond_settings *s)
3247 const char *detect_s;
3248 struct iface *iface;
3249 int miimon_interval;
3251 s->name = port->name;
3253 if (port->cfg->bond_mode) {
3254 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3255 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3256 port->name, port->cfg->bond_mode,
3257 bond_mode_to_string(s->balance));
3260 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3262 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3263 * active-backup. At some point we should remove this warning. */
3264 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3265 " in previous versions, the default bond_mode was"
3266 " balance-slb", port->name,
3267 bond_mode_to_string(s->balance));
3269 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3270 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3271 "please use another bond type or disable flood_vlans",
3275 miimon_interval = smap_get_int(&port->cfg->other_config,
3276 "bond-miimon-interval", 0);
3277 if (miimon_interval <= 0) {
3278 miimon_interval = 200;
3281 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3282 if (!detect_s || !strcmp(detect_s, "carrier")) {
3283 miimon_interval = 0;
3284 } else if (strcmp(detect_s, "miimon")) {
3285 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3286 "defaulting to carrier", port->name, detect_s);
3287 miimon_interval = 0;
3290 s->up_delay = MAX(0, port->cfg->bond_updelay);
3291 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3292 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3293 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3294 "bond-rebalance-interval", 10000);
3295 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3296 s->rebalance_interval = 1000;
3299 s->fake_iface = port->cfg->bond_fake_iface;
3301 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3302 "lacp-fallback-ab", false);
3304 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3305 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3309 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3310 * instead of obtaining it from the database. */
3312 port_is_synthetic(const struct port *port)
3314 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3317 /* Interface functions. */
3320 iface_is_internal(const struct ovsrec_interface *iface,
3321 const struct ovsrec_bridge *br)
3323 /* The local port and "internal" ports are always "internal". */
3324 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3327 /* Returns the correct network device type for interface 'iface' in bridge
3330 iface_get_type(const struct ovsrec_interface *iface,
3331 const struct ovsrec_bridge *br)
3335 /* The local port always has type "internal". Other ports take
3336 * their type from the database and default to "system" if none is
3338 if (iface_is_internal(iface, br)) {
3341 type = iface->type[0] ? iface->type : "system";
3344 return ofproto_port_open_type(br->datapath_type, type);
3348 iface_destroy__(struct iface *iface)
3351 struct port *port = iface->port;
3352 struct bridge *br = port->bridge;
3354 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3355 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3358 if (iface->ofp_port != OFPP_NONE) {
3359 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3362 list_remove(&iface->port_elem);
3363 hmap_remove(&br->iface_by_name, &iface->name_node);
3365 netdev_close(iface->netdev);
3373 iface_destroy(struct iface *iface)
3376 struct port *port = iface->port;
3378 iface_destroy__(iface);
3379 if (list_is_empty(&port->ifaces)) {
3385 static struct iface *
3386 iface_lookup(const struct bridge *br, const char *name)
3388 struct iface *iface;
3390 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3391 &br->iface_by_name) {
3392 if (!strcmp(iface->name, name)) {
3400 static struct iface *
3401 iface_find(const char *name)
3403 const struct bridge *br;
3405 HMAP_FOR_EACH (br, node, &all_bridges) {
3406 struct iface *iface = iface_lookup(br, name);
3415 static struct iface *
3416 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3418 struct iface *iface;
3420 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3422 if (iface->ofp_port == ofp_port) {
3429 /* Set Ethernet address of 'iface', if one is specified in the configuration
3432 iface_set_mac(struct iface *iface)
3434 uint8_t ea[ETH_ADDR_LEN];
3436 if (!strcmp(iface->type, "internal")
3437 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3438 if (iface->ofp_port == OFPP_LOCAL) {
3439 VLOG_ERR("interface %s: ignoring mac in Interface record "
3440 "(use Bridge record to set local port's mac)",
3442 } else if (eth_addr_is_multicast(ea)) {
3443 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3446 int error = netdev_set_etheraddr(iface->netdev, ea);
3448 VLOG_ERR("interface %s: setting MAC failed (%s)",
3449 iface->name, ovs_strerror(error));
3455 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3457 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3459 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3460 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3461 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3465 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3466 * sets the "ofport" field to -1.
3468 * This is appropriate when 'if_cfg''s interface cannot be created or is
3469 * otherwise invalid. */
3471 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3473 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3474 ovsrec_interface_set_status(if_cfg, NULL);
3475 ovsrec_interface_set_admin_state(if_cfg, NULL);
3476 ovsrec_interface_set_duplex(if_cfg, NULL);
3477 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3478 ovsrec_interface_set_link_state(if_cfg, NULL);
3479 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3480 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3481 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3482 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3483 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3484 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3485 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3486 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3491 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3493 union ovsdb_atom atom;
3495 atom.integer = target;
3496 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3500 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3502 struct ofpbuf queues_buf;
3504 ofpbuf_init(&queues_buf, 0);
3506 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3507 netdev_set_qos(iface->netdev, NULL, NULL);
3509 const struct ovsdb_datum *queues;
3510 struct netdev_queue_dump dump;
3511 unsigned int queue_id;
3512 struct smap details;
3516 /* Configure top-level Qos for 'iface'. */
3517 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3519 /* Deconfigure queues that were deleted. */
3520 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3522 smap_init(&details);
3523 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3524 if (!queue_ids_include(queues, queue_id)) {
3525 netdev_delete_queue(iface->netdev, queue_id);
3528 smap_destroy(&details);
3530 /* Configure queues for 'iface'. */
3532 for (i = 0; i < qos->n_queues; i++) {
3533 const struct ovsrec_queue *queue = qos->value_queues[i];
3534 unsigned int queue_id = qos->key_queues[i];
3536 if (queue_id == 0) {
3540 if (queue->n_dscp == 1) {
3541 struct ofproto_port_queue *port_queue;
3543 port_queue = ofpbuf_put_uninit(&queues_buf,
3544 sizeof *port_queue);
3545 port_queue->queue = queue_id;
3546 port_queue->dscp = queue->dscp[0];
3549 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3552 struct smap details;
3554 smap_init(&details);
3555 netdev_set_queue(iface->netdev, 0, &details);
3556 smap_destroy(&details);
3560 if (iface->ofp_port != OFPP_NONE) {
3561 const struct ofproto_port_queue *port_queues = queues_buf.data;
3562 size_t n_queues = queues_buf.size / sizeof *port_queues;
3564 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3565 port_queues, n_queues);
3568 netdev_set_policing(iface->netdev,
3569 iface->cfg->ingress_policing_rate,
3570 iface->cfg->ingress_policing_burst);
3572 ofpbuf_uninit(&queues_buf);
3576 iface_configure_cfm(struct iface *iface)
3578 const struct ovsrec_interface *cfg = iface->cfg;
3579 const char *opstate_str;
3580 const char *cfm_ccm_vlan;
3581 struct cfm_settings s;
3582 struct smap netdev_args;
3584 if (!cfg->n_cfm_mpid) {
3585 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3589 s.check_tnl_key = false;
3590 smap_init(&netdev_args);
3591 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3592 const char *key = smap_get(&netdev_args, "key");
3593 const char *in_key = smap_get(&netdev_args, "in_key");
3595 s.check_tnl_key = (key && !strcmp(key, "flow"))
3596 || (in_key && !strcmp(in_key, "flow"));
3598 smap_destroy(&netdev_args);
3600 s.mpid = *cfg->cfm_mpid;
3601 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3602 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3603 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3605 if (s.interval <= 0) {
3609 if (!cfm_ccm_vlan) {
3611 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3612 s.ccm_vlan = CFM_RANDOM_VLAN;
3614 s.ccm_vlan = atoi(cfm_ccm_vlan);
3615 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3620 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3622 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3624 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3625 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3627 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3630 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3631 * instead of obtaining it from the database. */
3633 iface_is_synthetic(const struct iface *iface)
3635 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3639 iface_pick_ofport(const struct ovsrec_interface *cfg)
3641 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3643 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3648 /* Port mirroring. */
3650 static struct mirror *
3651 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3655 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3656 if (uuid_equals(uuid, &m->uuid)) {
3664 bridge_configure_mirrors(struct bridge *br)
3666 const struct ovsdb_datum *mc;
3667 unsigned long *flood_vlans;
3668 struct mirror *m, *next;
3671 /* Get rid of deleted mirrors. */
3672 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3673 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3674 union ovsdb_atom atom;
3676 atom.uuid = m->uuid;
3677 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3682 /* Add new mirrors and reconfigure existing ones. */
3683 for (i = 0; i < br->cfg->n_mirrors; i++) {
3684 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3685 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3687 m = mirror_create(br, cfg);
3690 if (!mirror_configure(m)) {
3695 /* Update flooded vlans (for RSPAN). */
3696 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3697 br->cfg->n_flood_vlans);
3698 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3699 bitmap_free(flood_vlans);
3702 static struct mirror *
3703 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3707 m = xzalloc(sizeof *m);
3708 m->uuid = cfg->header_.uuid;
3709 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3711 m->name = xstrdup(cfg->name);
3717 mirror_destroy(struct mirror *m)
3720 struct bridge *br = m->bridge;
3723 ofproto_mirror_unregister(br->ofproto, m);
3726 hmap_remove(&br->mirrors, &m->hmap_node);
3733 mirror_collect_ports(struct mirror *m,
3734 struct ovsrec_port **in_ports, int n_in_ports,
3735 void ***out_portsp, size_t *n_out_portsp)
3737 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3738 size_t n_out_ports = 0;
3741 for (i = 0; i < n_in_ports; i++) {
3742 const char *name = in_ports[i]->name;
3743 struct port *port = port_lookup(m->bridge, name);
3745 out_ports[n_out_ports++] = port;
3747 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3748 "port %s", m->bridge->name, m->name, name);
3751 *out_portsp = out_ports;
3752 *n_out_portsp = n_out_ports;
3756 mirror_configure(struct mirror *m)
3758 const struct ovsrec_mirror *cfg = m->cfg;
3759 struct ofproto_mirror_settings s;
3762 if (strcmp(cfg->name, m->name)) {
3764 m->name = xstrdup(cfg->name);
3768 /* Get output port or VLAN. */
3769 if (cfg->output_port) {
3770 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3771 if (!s.out_bundle) {
3772 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3773 m->bridge->name, m->name);
3776 s.out_vlan = UINT16_MAX;
3778 if (cfg->output_vlan) {
3779 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3780 "output vlan; ignoring output vlan",
3781 m->bridge->name, m->name);
3783 } else if (cfg->output_vlan) {
3784 /* The database should prevent invalid VLAN values. */
3785 s.out_bundle = NULL;
3786 s.out_vlan = *cfg->output_vlan;
3788 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3789 m->bridge->name, m->name);
3793 /* Get port selection. */
3794 if (cfg->select_all) {
3795 size_t n_ports = hmap_count(&m->bridge->ports);
3796 void **ports = xmalloc(n_ports * sizeof *ports);
3801 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3811 /* Get ports, dropping ports that don't exist.
3812 * The IDL ensures that there are no duplicates. */
3813 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3814 &s.srcs, &s.n_srcs);
3815 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3816 &s.dsts, &s.n_dsts);
3819 /* Get VLAN selection. */
3820 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3823 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3826 if (s.srcs != s.dsts) {
3835 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3837 * This is deprecated. It is only for compatibility with broken device drivers
3838 * in old versions of Linux that do not properly support VLANs when VLAN
3839 * devices are not used. When broken device drivers are no longer in
3840 * widespread use, we will delete these interfaces. */
3842 static struct ovsrec_port **recs;
3843 static size_t n_recs, allocated_recs;
3845 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3846 * splinters are reconfigured. */
3848 register_rec(struct ovsrec_port *rec)
3850 if (n_recs >= allocated_recs) {
3851 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3853 recs[n_recs++] = rec;
3856 /* Frees all of the ports registered with register_reg(). */
3858 free_registered_recs(void)
3862 for (i = 0; i < n_recs; i++) {
3863 struct ovsrec_port *port = recs[i];
3866 for (j = 0; j < port->n_interfaces; j++) {
3867 struct ovsrec_interface *iface = port->interfaces[j];
3872 smap_destroy(&port->other_config);
3873 free(port->interfaces);
3881 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3884 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3886 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3890 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3893 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3894 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3895 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3898 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3899 * use, returns NULL.
3901 * Updates 'vlan_splinters_enabled_anywhere'. */
3902 static unsigned long int *
3903 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3905 unsigned long int *splinter_vlans;
3906 struct sset splinter_ifaces;
3907 const char *real_dev_name;
3908 struct shash *real_devs;
3909 struct shash_node *node;
3913 /* Free space allocated for synthesized ports and interfaces, since we're
3914 * in the process of reconstructing all of them. */
3915 free_registered_recs();
3917 splinter_vlans = bitmap_allocate(4096);
3918 sset_init(&splinter_ifaces);
3919 vlan_splinters_enabled_anywhere = false;
3920 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3921 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3924 for (j = 0; j < br_cfg->n_ports; j++) {
3925 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3928 for (k = 0; k < port_cfg->n_interfaces; k++) {
3929 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3931 if (vlan_splinters_is_enabled(iface_cfg)) {
3932 vlan_splinters_enabled_anywhere = true;
3933 sset_add(&splinter_ifaces, iface_cfg->name);
3934 vlan_bitmap_from_array__(port_cfg->trunks,
3940 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3941 bitmap_set1(splinter_vlans, *port_cfg->tag);
3946 if (!vlan_splinters_enabled_anywhere) {
3947 free(splinter_vlans);
3948 sset_destroy(&splinter_ifaces);
3952 HMAP_FOR_EACH (br, node, &all_bridges) {
3954 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3958 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3959 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3960 * device to be created for it. */
3961 bitmap_set0(splinter_vlans, 0);
3962 bitmap_set0(splinter_vlans, 4095);
3964 /* Delete all VLAN devices that we don't need. */
3966 real_devs = vlandev_get_real_devs();
3967 SHASH_FOR_EACH (node, real_devs) {
3968 const struct vlan_real_dev *real_dev = node->data;
3969 const struct vlan_dev *vlan_dev;
3970 bool real_dev_has_splinters;
3972 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3974 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3975 if (!real_dev_has_splinters
3976 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3977 struct netdev *netdev;
3979 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3980 if (!netdev_get_in4(netdev, NULL, NULL) ||
3981 !netdev_get_in6(netdev, NULL)) {
3982 /* It has an IP address configured, so we don't own
3983 * it. Don't delete it. */
3985 vlandev_del(vlan_dev->name);
3987 netdev_close(netdev);
3994 /* Add all VLAN devices that we need. */
3995 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3998 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3999 if (!vlandev_get_name(real_dev_name, vid)) {
4000 vlandev_add(real_dev_name, vid);
4007 sset_destroy(&splinter_ifaces);
4009 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4010 free(splinter_vlans);
4013 return splinter_vlans;
4016 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4019 configure_splinter_port(struct port *port)
4021 struct ofproto *ofproto = port->bridge->ofproto;
4022 ofp_port_t realdev_ofp_port;
4023 const char *realdev_name;
4024 struct iface *vlandev, *realdev;
4026 ofproto_bundle_unregister(port->bridge->ofproto, port);
4028 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4031 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4032 realdev = iface_lookup(port->bridge, realdev_name);
4033 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4035 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4039 static struct ovsrec_port *
4040 synthesize_splinter_port(const char *real_dev_name,
4041 const char *vlan_dev_name, int vid)
4043 struct ovsrec_interface *iface;
4044 struct ovsrec_port *port;
4046 iface = xmalloc(sizeof *iface);
4047 ovsrec_interface_init(iface);
4048 iface->name = xstrdup(vlan_dev_name);
4049 iface->type = "system";
4051 port = xmalloc(sizeof *port);
4052 ovsrec_port_init(port);
4053 port->interfaces = xmemdup(&iface, sizeof iface);
4054 port->n_interfaces = 1;
4055 port->name = xstrdup(vlan_dev_name);
4056 port->vlan_mode = "splinter";
4057 port->tag = xmalloc(sizeof *port->tag);
4060 smap_add(&port->other_config, "realdev", real_dev_name);
4066 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4067 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4068 * 1-bit in the 'splinter_vlans' bitmap. */
4070 add_vlan_splinter_ports(struct bridge *br,
4071 const unsigned long int *splinter_vlans,
4072 struct shash *ports)
4076 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4077 * we're modifying 'ports'. */
4078 for (i = 0; i < br->cfg->n_ports; i++) {
4079 const char *name = br->cfg->ports[i]->name;
4080 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4083 for (j = 0; j < port_cfg->n_interfaces; j++) {
4084 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4086 if (vlan_splinters_is_enabled(iface_cfg)) {
4087 const char *real_dev_name;
4090 real_dev_name = iface_cfg->name;
4091 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4092 const char *vlan_dev_name;
4094 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4096 && !shash_find(ports, vlan_dev_name)) {
4097 shash_add(ports, vlan_dev_name,
4098 synthesize_splinter_port(
4099 real_dev_name, vlan_dev_name, vid));
4108 mirror_refresh_stats(struct mirror *m)
4110 struct ofproto *ofproto = m->bridge->ofproto;
4111 uint64_t tx_packets, tx_bytes;
4114 size_t stat_cnt = 0;
4116 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4117 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4121 if (tx_packets != UINT64_MAX) {
4122 keys[stat_cnt] = "tx_packets";
4123 values[stat_cnt] = tx_packets;
4126 if (tx_bytes != UINT64_MAX) {
4127 keys[stat_cnt] = "tx_bytes";
4128 values[stat_cnt] = tx_bytes;
4132 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);