1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 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.
22 #include "async-append.h"
26 #include "connectivity.h"
34 #include "if-notifier.h"
37 #include "lib/netdev-dpdk.h"
38 #include "mac-learning.h"
39 #include "mcast-snooping.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "openvswitch/dynamic-string.h"
45 #include "openvswitch/list.h"
46 #include "openvswitch/meta-flow.h"
47 #include "openvswitch/ofp-print.h"
48 #include "openvswitch/ofp-util.h"
49 #include "openvswitch/ofpbuf.h"
50 #include "openvswitch/vlog.h"
54 #include "poll-loop.h"
56 #include "sflow_api.h"
60 #include "socket-util.h"
62 #include "stream-ssl.h"
64 #include "system-stats.h"
68 #include "lib/vswitch-idl.h"
69 #include "xenserver.h"
70 #include "vlan-bitmap.h"
72 VLOG_DEFINE_THIS_MODULE(bridge
);
74 COVERAGE_DEFINE(bridge_reconfigure
);
77 /* These members are always valid.
79 * They are immutable: they never change between iface_create() and
81 struct ovs_list port_elem
; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node
; /* In struct bridge's "iface_by_name" hmap. */
83 struct hmap_node ofp_port_node
; /* In struct bridge's "ifaces" hmap. */
84 struct port
*port
; /* Containing port. */
85 char *name
; /* Host network device name. */
86 struct netdev
*netdev
; /* Network device. */
87 ofp_port_t ofp_port
; /* OpenFlow port number. */
90 /* These members are valid only within bridge_reconfigure(). */
91 const char *type
; /* Usually same as cfg->type. */
92 const struct ovsrec_interface
*cfg
;
96 struct uuid uuid
; /* UUID of this "mirror" record in database. */
97 struct hmap_node hmap_node
; /* In struct bridge's "mirrors" hmap. */
98 struct bridge
*bridge
;
100 const struct ovsrec_mirror
*cfg
;
104 struct hmap_node hmap_node
; /* Element in struct bridge's "ports" hmap. */
105 struct bridge
*bridge
;
108 const struct ovsrec_port
*cfg
;
110 /* An ordinary bridge port has 1 interface.
111 * A bridge port for bonding has at least 2 interfaces. */
112 struct ovs_list ifaces
; /* List of "struct iface"s. */
116 struct hmap_node node
; /* In 'all_bridges'. */
117 char *name
; /* User-specified arbitrary name. */
118 char *type
; /* Datapath type. */
119 struct eth_addr ea
; /* Bridge Ethernet Address. */
120 struct eth_addr default_ea
; /* Default MAC. */
121 const struct ovsrec_bridge
*cfg
;
123 /* OpenFlow switch processing. */
124 struct ofproto
*ofproto
; /* OpenFlow switch. */
127 struct hmap ports
; /* "struct port"s indexed by name. */
128 struct hmap ifaces
; /* "struct iface"s indexed by ofp_port. */
129 struct hmap iface_by_name
; /* "struct iface"s indexed by name. */
131 /* Port mirroring. */
132 struct hmap mirrors
; /* "struct mirror" indexed by UUID. */
135 struct hmap mappings
; /* "struct" indexed by UUID */
137 /* Used during reconfiguration. */
138 struct shash wanted_ports
;
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port
;
142 struct ovsrec_interface synth_local_iface
;
143 struct ovsrec_interface
*synth_local_ifacep
;
147 struct hmap_node hmap_node
; /* In struct bridge's "mappings" hmap. */
148 struct bridge
*bridge
;
154 /* All bridges, indexed by name. */
155 static struct hmap all_bridges
= HMAP_INITIALIZER(&all_bridges
);
157 /* OVSDB IDL used to obtain configuration. */
158 static struct ovsdb_idl
*idl
;
160 /* We want to complete daemonization, fully detaching from our parent process,
161 * only after we have completed our initial configuration, committed our state
162 * to the database, and received confirmation back from the database server
163 * that it applied the commit. This allows our parent process to know that,
164 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
165 * to have already been filled in. (It is only "very likely" rather than
166 * certain because there is always a slim possibility that the transaction will
167 * fail or that some other client has added new bridges, ports, etc. while
168 * ovs-vswitchd was configuring using an old configuration.)
170 * We only need to do this once for our initial configuration at startup, so
171 * 'initial_config_done' tracks whether we've already done it. While we are
172 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
173 * itself and is otherwise NULL. */
174 static bool initial_config_done
;
175 static struct ovsdb_idl_txn
*daemonize_txn
;
177 /* Most recently processed IDL sequence number. */
178 static unsigned int idl_seqno
;
180 /* Track changes to port connectivity. */
181 static uint64_t connectivity_seqno
= LLONG_MIN
;
183 /* Status update to database.
185 * Some information in the database must be kept as up-to-date as possible to
186 * allow controllers to respond rapidly to network outages. Those status are
187 * updated via the 'status_txn'.
189 * We use the global connectivity sequence number to detect the status change.
190 * Also, to prevent the status update from sending too much to the database,
191 * we check the return status of each update transaction and do not start new
192 * update if the previous transaction status is 'TXN_INCOMPLETE'.
194 * 'statux_txn' is NULL if there is no ongoing status update.
196 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
197 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
198 * which will cause the main thread wake up soon and retry the status update.
200 static struct ovsdb_idl_txn
*status_txn
;
201 static bool status_txn_try_again
;
203 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
204 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
205 #define STATUS_CHECK_AGAIN_MSEC 100
207 /* Statistics update to database. */
208 static struct ovsdb_idl_txn
*stats_txn
;
210 /* Each time this timer expires, the bridge fetches interface and mirror
211 * statistics and pushes them into the database. */
212 static int stats_timer_interval
;
213 static long long int stats_timer
= LLONG_MIN
;
215 /* Each time this timer expires, the bridge fetches the list of port/VLAN
216 * membership that has been modified by the AA.
218 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
219 static long long int aa_refresh_timer
= LLONG_MIN
;
221 /* Whenever system interfaces are added, removed or change state, the bridge
222 * will be reconfigured.
224 static struct if_notifier
*ifnotifier
;
225 static bool ifaces_changed
= false;
227 static void add_del_bridges(const struct ovsrec_open_vswitch
*);
228 static void bridge_run__(void);
229 static void bridge_create(const struct ovsrec_bridge
*);
230 static void bridge_destroy(struct bridge
*, bool del
);
231 static struct bridge
*bridge_lookup(const char *name
);
232 static unixctl_cb_func bridge_unixctl_dump_flows
;
233 static unixctl_cb_func bridge_unixctl_reconnect
;
234 static size_t bridge_get_controllers(const struct bridge
*br
,
235 struct ovsrec_controller
***controllersp
);
236 static void bridge_collect_wanted_ports(struct bridge
*,
237 struct shash
*wanted_ports
);
238 static void bridge_delete_ofprotos(void);
239 static void bridge_delete_or_reconfigure_ports(struct bridge
*);
240 static void bridge_del_ports(struct bridge
*,
241 const struct shash
*wanted_ports
);
242 static void bridge_add_ports(struct bridge
*,
243 const struct shash
*wanted_ports
);
245 static void bridge_configure_datapath_id(struct bridge
*);
246 static void bridge_configure_netflow(struct bridge
*);
247 static void bridge_configure_forward_bpdu(struct bridge
*);
248 static void bridge_configure_mac_table(struct bridge
*);
249 static void bridge_configure_mcast_snooping(struct bridge
*);
250 static void bridge_configure_sflow(struct bridge
*, int *sflow_bridge_number
);
251 static void bridge_configure_ipfix(struct bridge
*);
252 static void bridge_configure_spanning_tree(struct bridge
*);
253 static void bridge_configure_tables(struct bridge
*);
254 static void bridge_configure_dp_desc(struct bridge
*);
255 static void bridge_configure_aa(struct bridge
*);
256 static void bridge_aa_refresh_queued(struct bridge
*);
257 static bool bridge_aa_need_refresh(struct bridge
*);
258 static void bridge_configure_remotes(struct bridge
*,
259 const struct sockaddr_in
*managers
,
261 static void bridge_pick_local_hw_addr(struct bridge
*, struct eth_addr
*ea
,
262 struct iface
**hw_addr_iface
);
263 static uint64_t bridge_pick_datapath_id(struct bridge
*,
264 const struct eth_addr bridge_ea
,
265 struct iface
*hw_addr_iface
);
266 static uint64_t dpid_from_hash(const void *, size_t nbytes
);
267 static bool bridge_has_bond_fake_iface(const struct bridge
*,
269 static bool port_is_bond_fake_iface(const struct port
*);
271 static unixctl_cb_func qos_unixctl_show_types
;
272 static unixctl_cb_func qos_unixctl_show
;
274 static struct port
*port_create(struct bridge
*, const struct ovsrec_port
*);
275 static void port_del_ifaces(struct port
*);
276 static void port_destroy(struct port
*);
277 static struct port
*port_lookup(const struct bridge
*, const char *name
);
278 static void port_configure(struct port
*);
279 static struct lacp_settings
*port_configure_lacp(struct port
*,
280 struct lacp_settings
*);
281 static void port_configure_bond(struct port
*, struct bond_settings
*);
282 static bool port_is_synthetic(const struct port
*);
284 static void reconfigure_system_stats(const struct ovsrec_open_vswitch
*);
285 static void run_system_stats(void);
287 static void bridge_configure_mirrors(struct bridge
*);
288 static struct mirror
*mirror_create(struct bridge
*,
289 const struct ovsrec_mirror
*);
290 static void mirror_destroy(struct mirror
*);
291 static bool mirror_configure(struct mirror
*);
292 static void mirror_refresh_stats(struct mirror
*);
294 static void iface_configure_lacp(struct iface
*, struct lacp_slave_settings
*);
295 static bool iface_create(struct bridge
*, const struct ovsrec_interface
*,
296 const struct ovsrec_port
*);
297 static bool iface_is_internal(const struct ovsrec_interface
*iface
,
298 const struct ovsrec_bridge
*br
);
299 static const char *iface_get_type(const struct ovsrec_interface
*,
300 const struct ovsrec_bridge
*);
301 static void iface_destroy(struct iface
*);
302 static void iface_destroy__(struct iface
*);
303 static struct iface
*iface_lookup(const struct bridge
*, const char *name
);
304 static struct iface
*iface_find(const char *name
);
305 static struct iface
*iface_from_ofp_port(const struct bridge
*,
306 ofp_port_t ofp_port
);
307 static void iface_set_mac(const struct bridge
*, const struct port
*, struct iface
*);
308 static void iface_set_ofport(const struct ovsrec_interface
*, ofp_port_t ofport
);
309 static void iface_clear_db_record(const struct ovsrec_interface
*if_cfg
, char *errp
);
310 static void iface_configure_qos(struct iface
*, const struct ovsrec_qos
*);
311 static void iface_configure_cfm(struct iface
*);
312 static void iface_refresh_cfm_stats(struct iface
*);
313 static void iface_refresh_stats(struct iface
*);
314 static void iface_refresh_netdev_status(struct iface
*);
315 static void iface_refresh_ofproto_status(struct iface
*);
316 static bool iface_is_synthetic(const struct iface
*);
317 static ofp_port_t
iface_get_requested_ofp_port(
318 const struct ovsrec_interface
*);
319 static ofp_port_t
iface_pick_ofport(const struct ovsrec_interface
*);
322 static void discover_types(const struct ovsrec_open_vswitch
*cfg
);
325 bridge_init_ofproto(const struct ovsrec_open_vswitch
*cfg
)
327 struct shash iface_hints
;
328 static bool initialized
= false;
335 shash_init(&iface_hints
);
338 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
339 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
342 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
343 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
346 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
347 struct ovsrec_interface
*if_cfg
= port_cfg
->interfaces
[k
];
348 struct iface_hint
*iface_hint
;
350 iface_hint
= xmalloc(sizeof *iface_hint
);
351 iface_hint
->br_name
= br_cfg
->name
;
352 iface_hint
->br_type
= br_cfg
->datapath_type
;
353 iface_hint
->ofp_port
= iface_pick_ofport(if_cfg
);
355 shash_add(&iface_hints
, if_cfg
->name
, iface_hint
);
361 ofproto_init(&iface_hints
);
363 shash_destroy_free_data(&iface_hints
);
368 if_change_cb(void *aux OVS_UNUSED
)
370 ifaces_changed
= true;
373 /* Public functions. */
375 /* Initializes the bridge module, configuring it to obtain its configuration
376 * from an OVSDB server accessed over 'remote', which should be a string in a
377 * form acceptable to ovsdb_idl_create(). */
379 bridge_init(const char *remote
)
381 /* Create connection to database. */
382 idl
= ovsdb_idl_create(remote
, &ovsrec_idl_class
, true, true);
383 idl_seqno
= ovsdb_idl_get_seqno(idl
);
384 ovsdb_idl_set_lock(idl
, "ovs_vswitchd");
385 ovsdb_idl_verify_write_only(idl
);
387 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_cur_cfg
);
388 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_statistics
);
389 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_datapath_types
);
390 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_iface_types
);
391 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_external_ids
);
392 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_ovs_version
);
393 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_db_version
);
394 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_type
);
395 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_version
);
397 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_id
);
398 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_version
);
399 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_status
);
400 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_rstp_status
);
401 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_stp_enable
);
402 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_rstp_enable
);
403 ovsdb_idl_omit(idl
, &ovsrec_bridge_col_external_ids
);
405 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_status
);
406 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_rstp_status
);
407 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_rstp_statistics
);
408 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_statistics
);
409 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_bond_active_slave
);
410 ovsdb_idl_omit(idl
, &ovsrec_port_col_external_ids
);
411 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_trunks
);
412 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_vlan_mode
);
413 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_admin_state
);
414 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_duplex
);
415 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_speed
);
416 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_state
);
417 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_resets
);
418 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mac_in_use
);
419 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ifindex
);
420 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mtu
);
421 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ofport
);
422 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_statistics
);
423 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_status
);
424 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault
);
425 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault_status
);
426 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_mpids
);
427 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_flap_count
);
428 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_health
);
429 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_opstate
);
430 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_bfd_status
);
431 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_lacp_current
);
432 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_error
);
433 ovsdb_idl_omit(idl
, &ovsrec_interface_col_external_ids
);
435 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_is_connected
);
436 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_role
);
437 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_status
);
438 ovsdb_idl_omit(idl
, &ovsrec_controller_col_external_ids
);
440 ovsdb_idl_omit(idl
, &ovsrec_qos_col_external_ids
);
442 ovsdb_idl_omit(idl
, &ovsrec_queue_col_external_ids
);
444 ovsdb_idl_omit(idl
, &ovsrec_mirror_col_external_ids
);
445 ovsdb_idl_omit_alert(idl
, &ovsrec_mirror_col_statistics
);
447 ovsdb_idl_omit(idl
, &ovsrec_netflow_col_external_ids
);
448 ovsdb_idl_omit(idl
, &ovsrec_sflow_col_external_ids
);
449 ovsdb_idl_omit(idl
, &ovsrec_ipfix_col_external_ids
);
450 ovsdb_idl_omit(idl
, &ovsrec_flow_sample_collector_set_col_external_ids
);
452 ovsdb_idl_omit(idl
, &ovsrec_manager_col_external_ids
);
453 ovsdb_idl_omit(idl
, &ovsrec_manager_col_inactivity_probe
);
454 ovsdb_idl_omit(idl
, &ovsrec_manager_col_is_connected
);
455 ovsdb_idl_omit(idl
, &ovsrec_manager_col_max_backoff
);
456 ovsdb_idl_omit(idl
, &ovsrec_manager_col_status
);
458 ovsdb_idl_omit(idl
, &ovsrec_ssl_col_external_ids
);
460 /* Register unixctl commands. */
461 unixctl_command_register("qos/show-types", "interface", 1, 1,
462 qos_unixctl_show_types
, NULL
);
463 unixctl_command_register("qos/show", "interface", 1, 1,
464 qos_unixctl_show
, NULL
);
465 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
466 bridge_unixctl_dump_flows
, NULL
);
467 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
468 bridge_unixctl_reconnect
, NULL
);
477 ifnotifier
= if_notifier_create(if_change_cb
, NULL
);
483 struct bridge
*br
, *next_br
;
485 if_notifier_destroy(ifnotifier
);
486 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
487 bridge_destroy(br
, false);
489 ovsdb_idl_destroy(idl
);
492 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
493 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
494 * responsible for freeing '*managersp' (with free()).
496 * You may be asking yourself "why does ovs-vswitchd care?", because
497 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
498 * should not be and in fact is not directly involved in that. But
499 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
500 * it has to tell in-band control where the managers are to enable that.
501 * (Thus, only managers connected in-band and with non-loopback addresses
505 collect_in_band_managers(const struct ovsrec_open_vswitch
*ovs_cfg
,
506 struct sockaddr_in
**managersp
, size_t *n_managersp
)
508 struct sockaddr_in
*managers
= NULL
;
509 size_t n_managers
= 0;
513 /* Collect all of the potential targets from the "targets" columns of the
514 * rows pointed to by "manager_options", excluding any that are
517 for (i
= 0; i
< ovs_cfg
->n_manager_options
; i
++) {
518 struct ovsrec_manager
*m
= ovs_cfg
->manager_options
[i
];
520 if (m
->connection_mode
&& !strcmp(m
->connection_mode
, "out-of-band")) {
521 sset_find_and_delete(&targets
, m
->target
);
523 sset_add(&targets
, m
->target
);
527 /* Now extract the targets' IP addresses. */
528 if (!sset_is_empty(&targets
)) {
531 managers
= xmalloc(sset_count(&targets
) * sizeof *managers
);
532 SSET_FOR_EACH (target
, &targets
) {
534 struct sockaddr_storage ss
;
535 struct sockaddr_in in
;
538 /* Ignore loopback. */
539 if (stream_parse_target_with_default_port(target
, OVSDB_PORT
,
541 && sa
.ss
.ss_family
== AF_INET
542 && sa
.in
.sin_addr
.s_addr
!= htonl(INADDR_LOOPBACK
)) {
543 managers
[n_managers
++] = sa
.in
;
547 sset_destroy(&targets
);
549 *managersp
= managers
;
550 *n_managersp
= n_managers
;
554 bridge_reconfigure(const struct ovsrec_open_vswitch
*ovs_cfg
)
556 struct sockaddr_in
*managers
;
557 struct bridge
*br
, *next
;
558 int sflow_bridge_number
;
561 COVERAGE_INC(bridge_reconfigure
);
563 ofproto_set_flow_limit(smap_get_int(&ovs_cfg
->other_config
, "flow-limit",
564 OFPROTO_FLOW_LIMIT_DEFAULT
));
565 ofproto_set_max_idle(smap_get_int(&ovs_cfg
->other_config
, "max-idle",
566 OFPROTO_MAX_IDLE_DEFAULT
));
567 ofproto_set_cpu_mask(smap_get(&ovs_cfg
->other_config
, "pmd-cpu-mask"));
570 smap_get_int(&ovs_cfg
->other_config
, "n-handler-threads", 0),
571 smap_get_int(&ovs_cfg
->other_config
, "n-revalidator-threads", 0));
573 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
574 * to 'ovs_cfg', with only very minimal configuration otherwise.
576 * This is mostly an update to bridge data structures. Nothing is pushed
577 * down to ofproto or lower layers. */
578 add_del_bridges(ovs_cfg
);
579 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
580 bridge_collect_wanted_ports(br
, &br
->wanted_ports
);
581 bridge_del_ports(br
, &br
->wanted_ports
);
584 /* Start pushing configuration changes down to the ofproto layer:
586 * - Delete ofprotos that are no longer configured.
588 * - Delete ports that are no longer configured.
590 * - Reconfigure existing ports to their desired configurations, or
591 * delete them if not possible.
593 * We have to do all the deletions before we can do any additions, because
594 * the ports to be added might require resources that will be freed up by
595 * deletions (they might especially overlap in name). */
596 bridge_delete_ofprotos();
597 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
599 bridge_delete_or_reconfigure_ports(br
);
603 /* Finish pushing configuration changes to the ofproto layer:
605 * - Create ofprotos that are missing.
607 * - Add ports that are missing. */
608 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
612 error
= ofproto_create(br
->name
, br
->type
, &br
->ofproto
);
614 VLOG_ERR("failed to create bridge %s: %s", br
->name
,
615 ovs_strerror(error
));
616 shash_destroy(&br
->wanted_ports
);
617 bridge_destroy(br
, true);
619 /* Trigger storing datapath version. */
620 seq_change(connectivity_seq_get());
624 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
625 bridge_add_ports(br
, &br
->wanted_ports
);
626 shash_destroy(&br
->wanted_ports
);
629 reconfigure_system_stats(ovs_cfg
);
631 /* Complete the configuration. */
632 sflow_bridge_number
= 0;
633 collect_in_band_managers(ovs_cfg
, &managers
, &n_managers
);
634 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
637 /* We need the datapath ID early to allow LACP ports to use it as the
638 * default system ID. */
639 bridge_configure_datapath_id(br
);
641 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
644 port_configure(port
);
646 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
647 iface_set_ofport(iface
->cfg
, iface
->ofp_port
);
648 /* Clear eventual previous errors */
649 ovsrec_interface_set_error(iface
->cfg
, NULL
);
650 iface_configure_cfm(iface
);
651 iface_configure_qos(iface
, port
->cfg
->qos
);
652 iface_set_mac(br
, port
, iface
);
653 ofproto_port_set_bfd(br
->ofproto
, iface
->ofp_port
,
655 ofproto_port_set_lldp(br
->ofproto
, iface
->ofp_port
,
659 bridge_configure_mirrors(br
);
660 bridge_configure_forward_bpdu(br
);
661 bridge_configure_mac_table(br
);
662 bridge_configure_mcast_snooping(br
);
663 bridge_configure_remotes(br
, managers
, n_managers
);
664 bridge_configure_netflow(br
);
665 bridge_configure_sflow(br
, &sflow_bridge_number
);
666 bridge_configure_ipfix(br
);
667 bridge_configure_spanning_tree(br
);
668 bridge_configure_tables(br
);
669 bridge_configure_dp_desc(br
);
670 bridge_configure_aa(br
);
674 /* The ofproto-dpif provider does some final reconfiguration in its
675 * ->type_run() function. We have to call it before notifying the database
676 * client that reconfiguration is complete, otherwise there is a very
677 * narrow race window in which e.g. ofproto/trace will not recognize the
678 * new configuration (sometimes this causes unit test failures). */
682 /* Delete ofprotos which aren't configured or have the wrong type. Create
683 * ofprotos which don't exist but need to. */
685 bridge_delete_ofprotos(void)
692 /* Delete ofprotos with no bridge or with the wrong type. */
695 ofproto_enumerate_types(&types
);
696 SSET_FOR_EACH (type
, &types
) {
699 ofproto_enumerate_names(type
, &names
);
700 SSET_FOR_EACH (name
, &names
) {
701 br
= bridge_lookup(name
);
702 if (!br
|| strcmp(type
, br
->type
)) {
703 ofproto_delete(name
, type
);
707 sset_destroy(&names
);
708 sset_destroy(&types
);
712 add_ofp_port(ofp_port_t port
, ofp_port_t
*ports
, size_t *n
, size_t *allocated
)
714 if (*n
>= *allocated
) {
715 ports
= x2nrealloc(ports
, allocated
, sizeof *ports
);
717 ports
[(*n
)++] = port
;
722 bridge_delete_or_reconfigure_ports(struct bridge
*br
)
724 struct ofproto_port ofproto_port
;
725 struct ofproto_port_dump dump
;
727 struct sset ofproto_ports
;
728 struct port
*port
, *port_next
;
730 /* List of "ofp_port"s to delete. We make a list instead of deleting them
731 * right away because ofproto implementations aren't necessarily able to
732 * iterate through a changing list of ports in an entirely robust way. */
739 sset_init(&ofproto_ports
);
741 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
742 * that are not configured in the database. (This commonly happens when
743 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
745 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
746 * that have the wrong OpenFlow port number (and arrange to add them back
747 * with the correct OpenFlow port number). */
748 OFPROTO_PORT_FOR_EACH (&ofproto_port
, &dump
, br
->ofproto
) {
749 ofp_port_t requested_ofp_port
;
752 sset_add(&ofproto_ports
, ofproto_port
.name
);
754 iface
= iface_lookup(br
, ofproto_port
.name
);
756 /* No such iface is configured, so we should delete this
759 * As a corner case exception, keep the port if it's a bond fake
761 if (bridge_has_bond_fake_iface(br
, ofproto_port
.name
)
762 && !strcmp(ofproto_port
.type
, "internal")) {
768 if (strcmp(ofproto_port
.type
, iface
->type
)
769 || netdev_set_config(iface
->netdev
, &iface
->cfg
->options
, NULL
)) {
770 /* The interface is the wrong type or can't be configured.
775 /* If the requested OpenFlow port for 'iface' changed, and it's not
776 * already the correct port, then we might want to temporarily delete
777 * this interface, so we can add it back again with the new OpenFlow
779 requested_ofp_port
= iface_get_requested_ofp_port(iface
->cfg
);
780 if (iface
->ofp_port
!= OFPP_LOCAL
&&
781 requested_ofp_port
!= OFPP_NONE
&&
782 requested_ofp_port
!= iface
->ofp_port
) {
783 ofp_port_t victim_request
;
784 struct iface
*victim
;
786 /* Check for an existing OpenFlow port currently occupying
787 * 'iface''s requested port number. If there isn't one, then
788 * delete this port. Otherwise we need to consider further. */
789 victim
= iface_from_ofp_port(br
, requested_ofp_port
);
794 /* 'victim' is a port currently using 'iface''s requested port
795 * number. Unless 'victim' specifically requested that port
796 * number, too, then we can delete both 'iface' and 'victim'
797 * temporarily. (We'll add both of them back again later with new
798 * OpenFlow port numbers.)
800 * If 'victim' did request port number 'requested_ofp_port', just
801 * like 'iface', then that's a configuration inconsistency that we
802 * can't resolve. We might as well let it keep its current port
804 victim_request
= iface_get_requested_ofp_port(victim
->cfg
);
805 if (victim_request
!= requested_ofp_port
) {
806 del
= add_ofp_port(victim
->ofp_port
, del
, &n
, &allocated
);
807 iface_destroy(victim
);
816 iface_destroy(iface
);
817 del
= add_ofp_port(ofproto_port
.ofp_port
, del
, &n
, &allocated
);
819 for (i
= 0; i
< n
; i
++) {
820 ofproto_port_del(br
->ofproto
, del
[i
]);
824 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
825 * that we didn't see when we iterated through the datapath, i.e. ports
826 * that disappeared underneath use. This is an unusual situation, but it
827 * can happen in some cases:
829 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
830 * idea but could happen).
832 * - The port represented a device that disappeared, e.g. a tuntap
833 * device destroyed via "tunctl -d", a physical Ethernet device
834 * whose module was just unloaded via "rmmod", or a virtual NIC for a
835 * VM whose VM was just terminated. */
836 HMAP_FOR_EACH_SAFE (port
, port_next
, hmap_node
, &br
->ports
) {
837 struct iface
*iface
, *iface_next
;
839 LIST_FOR_EACH_SAFE (iface
, iface_next
, port_elem
, &port
->ifaces
) {
840 if (!sset_contains(&ofproto_ports
, iface
->name
)) {
841 iface_destroy__(iface
);
845 if (ovs_list_is_empty(&port
->ifaces
)) {
849 sset_destroy(&ofproto_ports
);
853 bridge_add_ports__(struct bridge
*br
, const struct shash
*wanted_ports
,
854 bool with_requested_port
)
856 struct shash_node
*port_node
;
858 SHASH_FOR_EACH (port_node
, wanted_ports
) {
859 const struct ovsrec_port
*port_cfg
= port_node
->data
;
862 for (i
= 0; i
< port_cfg
->n_interfaces
; i
++) {
863 const struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[i
];
864 ofp_port_t requested_ofp_port
;
866 requested_ofp_port
= iface_get_requested_ofp_port(iface_cfg
);
867 if ((requested_ofp_port
!= OFPP_NONE
) == with_requested_port
) {
868 struct iface
*iface
= iface_lookup(br
, iface_cfg
->name
);
871 iface_create(br
, iface_cfg
, port_cfg
);
879 bridge_add_ports(struct bridge
*br
, const struct shash
*wanted_ports
)
881 /* First add interfaces that request a particular port number. */
882 bridge_add_ports__(br
, wanted_ports
, true);
884 /* Then add interfaces that want automatic port number assignment.
885 * We add these afterward to avoid accidentally taking a specifically
886 * requested port number. */
887 bridge_add_ports__(br
, wanted_ports
, false);
891 port_configure(struct port
*port
)
893 const struct ovsrec_port
*cfg
= port
->cfg
;
894 struct bond_settings bond_settings
;
895 struct lacp_settings lacp_settings
;
896 struct ofproto_bundle_settings s
;
904 s
.slaves
= xmalloc(ovs_list_size(&port
->ifaces
) * sizeof *s
.slaves
);
905 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
906 s
.slaves
[s
.n_slaves
++] = iface
->ofp_port
;
911 if (cfg
->tag
&& *cfg
->tag
>= 0 && *cfg
->tag
<= 4095) {
915 /* Get VLAN trunks. */
918 s
.trunks
= vlan_bitmap_from_array(cfg
->trunks
, cfg
->n_trunks
);
922 if (cfg
->vlan_mode
) {
923 if (!strcmp(cfg
->vlan_mode
, "access")) {
924 s
.vlan_mode
= PORT_VLAN_ACCESS
;
925 } else if (!strcmp(cfg
->vlan_mode
, "trunk")) {
926 s
.vlan_mode
= PORT_VLAN_TRUNK
;
927 } else if (!strcmp(cfg
->vlan_mode
, "native-tagged")) {
928 s
.vlan_mode
= PORT_VLAN_NATIVE_TAGGED
;
929 } else if (!strcmp(cfg
->vlan_mode
, "native-untagged")) {
930 s
.vlan_mode
= PORT_VLAN_NATIVE_UNTAGGED
;
932 /* This "can't happen" because ovsdb-server should prevent it. */
933 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
934 "back to trunk mode", port
->name
, cfg
->vlan_mode
);
935 s
.vlan_mode
= PORT_VLAN_TRUNK
;
939 s
.vlan_mode
= PORT_VLAN_ACCESS
;
941 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
945 s
.vlan_mode
= PORT_VLAN_TRUNK
;
948 s
.use_priority_tags
= smap_get_bool(&cfg
->other_config
, "priority-tags",
951 /* Get LACP settings. */
952 s
.lacp
= port_configure_lacp(port
, &lacp_settings
);
956 s
.lacp_slaves
= xmalloc(s
.n_slaves
* sizeof *s
.lacp_slaves
);
957 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
958 iface_configure_lacp(iface
, &s
.lacp_slaves
[i
++]);
961 s
.lacp_slaves
= NULL
;
964 /* Get bond settings. */
965 if (s
.n_slaves
> 1) {
966 s
.bond
= &bond_settings
;
967 port_configure_bond(port
, &bond_settings
);
970 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
971 netdev_set_miimon_interval(iface
->netdev
, 0);
976 ofproto_bundle_register(port
->bridge
->ofproto
, port
, &s
);
984 /* Pick local port hardware address and datapath ID for 'br'. */
986 bridge_configure_datapath_id(struct bridge
*br
)
990 struct iface
*local_iface
;
991 struct iface
*hw_addr_iface
;
994 bridge_pick_local_hw_addr(br
, &ea
, &hw_addr_iface
);
995 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
997 int error
= netdev_set_etheraddr(local_iface
->netdev
, ea
);
999 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1000 VLOG_ERR_RL(&rl
, "bridge %s: failed to set bridge "
1001 "Ethernet address: %s",
1002 br
->name
, ovs_strerror(error
));
1007 dpid
= bridge_pick_datapath_id(br
, ea
, hw_addr_iface
);
1008 if (dpid
!= ofproto_get_datapath_id(br
->ofproto
)) {
1009 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64
, br
->name
, dpid
);
1010 ofproto_set_datapath_id(br
->ofproto
, dpid
);
1013 dpid_string
= xasprintf("%016"PRIx64
, dpid
);
1014 ovsrec_bridge_set_datapath_id(br
->cfg
, dpid_string
);
1018 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1021 bridge_get_allowed_versions(struct bridge
*br
)
1023 if (!br
->cfg
->n_protocols
) {
1027 return ofputil_versions_from_strings(br
->cfg
->protocols
,
1028 br
->cfg
->n_protocols
);
1031 /* Set NetFlow configuration on 'br'. */
1033 bridge_configure_netflow(struct bridge
*br
)
1035 struct ovsrec_netflow
*cfg
= br
->cfg
->netflow
;
1036 struct netflow_options opts
;
1039 ofproto_set_netflow(br
->ofproto
, NULL
);
1043 memset(&opts
, 0, sizeof opts
);
1045 /* Get default NetFlow configuration from datapath.
1046 * Apply overrides from 'cfg'. */
1047 ofproto_get_netflow_ids(br
->ofproto
, &opts
.engine_type
, &opts
.engine_id
);
1048 if (cfg
->engine_type
) {
1049 opts
.engine_type
= *cfg
->engine_type
;
1051 if (cfg
->engine_id
) {
1052 opts
.engine_id
= *cfg
->engine_id
;
1055 /* Configure active timeout interval. */
1056 opts
.active_timeout
= cfg
->active_timeout
;
1057 if (!opts
.active_timeout
) {
1058 opts
.active_timeout
= -1;
1059 } else if (opts
.active_timeout
< 0) {
1060 VLOG_WARN("bridge %s: active timeout interval set to negative "
1061 "value, using default instead (%d seconds)", br
->name
,
1062 NF_ACTIVE_TIMEOUT_DEFAULT
);
1063 opts
.active_timeout
= -1;
1066 /* Add engine ID to interface number to disambiguate bridgs? */
1067 opts
.add_id_to_iface
= cfg
->add_id_to_interface
;
1068 if (opts
.add_id_to_iface
) {
1069 if (opts
.engine_id
> 0x7f) {
1070 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1071 "another vswitch, choose an engine id less than 128",
1074 if (hmap_count(&br
->ports
) > 508) {
1075 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1076 "another port when more than 508 ports are used",
1082 sset_init(&opts
.collectors
);
1083 sset_add_array(&opts
.collectors
, cfg
->targets
, cfg
->n_targets
);
1086 if (ofproto_set_netflow(br
->ofproto
, &opts
)) {
1087 VLOG_ERR("bridge %s: problem setting netflow collectors", br
->name
);
1089 sset_destroy(&opts
.collectors
);
1092 /* Set sFlow configuration on 'br'. */
1094 bridge_configure_sflow(struct bridge
*br
, int *sflow_bridge_number
)
1096 const struct ovsrec_sflow
*cfg
= br
->cfg
->sflow
;
1097 struct ovsrec_controller
**controllers
;
1098 struct ofproto_sflow_options oso
;
1099 size_t n_controllers
;
1103 ofproto_set_sflow(br
->ofproto
, NULL
);
1107 memset(&oso
, 0, sizeof oso
);
1109 sset_init(&oso
.targets
);
1110 sset_add_array(&oso
.targets
, cfg
->targets
, cfg
->n_targets
);
1112 oso
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1113 if (cfg
->sampling
) {
1114 oso
.sampling_rate
= *cfg
->sampling
;
1117 oso
.polling_interval
= SFL_DEFAULT_POLLING_INTERVAL
;
1119 oso
.polling_interval
= *cfg
->polling
;
1122 oso
.header_len
= SFL_DEFAULT_HEADER_SIZE
;
1124 oso
.header_len
= *cfg
->header
;
1127 oso
.sub_id
= (*sflow_bridge_number
)++;
1128 oso
.agent_device
= cfg
->agent
;
1130 oso
.control_ip
= NULL
;
1131 n_controllers
= bridge_get_controllers(br
, &controllers
);
1132 for (i
= 0; i
< n_controllers
; i
++) {
1133 if (controllers
[i
]->local_ip
) {
1134 oso
.control_ip
= controllers
[i
]->local_ip
;
1138 ofproto_set_sflow(br
->ofproto
, &oso
);
1140 sset_destroy(&oso
.targets
);
1143 /* Returns whether a IPFIX row is valid. */
1145 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix
*ipfix
)
1147 return ipfix
&& ipfix
->n_targets
> 0;
1150 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1152 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set
*fscs
,
1153 const struct bridge
*br
)
1155 return ovsrec_ipfix_is_valid(fscs
->ipfix
) && fscs
->bridge
== br
->cfg
;
1158 /* Set IPFIX configuration on 'br'. */
1160 bridge_configure_ipfix(struct bridge
*br
)
1162 const struct ovsrec_ipfix
*be_cfg
= br
->cfg
->ipfix
;
1163 bool valid_be_cfg
= ovsrec_ipfix_is_valid(be_cfg
);
1164 const struct ovsrec_flow_sample_collector_set
*fe_cfg
;
1165 struct ofproto_ipfix_bridge_exporter_options be_opts
;
1166 struct ofproto_ipfix_flow_exporter_options
*fe_opts
= NULL
;
1167 size_t n_fe_opts
= 0;
1169 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1170 if (ovsrec_fscs_is_valid(fe_cfg
, br
)) {
1175 if (!valid_be_cfg
&& n_fe_opts
== 0) {
1176 ofproto_set_ipfix(br
->ofproto
, NULL
, NULL
, 0);
1181 memset(&be_opts
, 0, sizeof be_opts
);
1183 sset_init(&be_opts
.targets
);
1184 sset_add_array(&be_opts
.targets
, be_cfg
->targets
, be_cfg
->n_targets
);
1186 if (be_cfg
->sampling
) {
1187 be_opts
.sampling_rate
= *be_cfg
->sampling
;
1189 be_opts
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1191 if (be_cfg
->obs_domain_id
) {
1192 be_opts
.obs_domain_id
= *be_cfg
->obs_domain_id
;
1194 if (be_cfg
->obs_point_id
) {
1195 be_opts
.obs_point_id
= *be_cfg
->obs_point_id
;
1197 if (be_cfg
->cache_active_timeout
) {
1198 be_opts
.cache_active_timeout
= *be_cfg
->cache_active_timeout
;
1200 if (be_cfg
->cache_max_flows
) {
1201 be_opts
.cache_max_flows
= *be_cfg
->cache_max_flows
;
1204 be_opts
.enable_tunnel_sampling
= smap_get_bool(&be_cfg
->other_config
,
1205 "enable-tunnel-sampling", true);
1207 be_opts
.enable_input_sampling
= !smap_get_bool(&be_cfg
->other_config
,
1208 "enable-input-sampling", false);
1210 be_opts
.enable_output_sampling
= !smap_get_bool(&be_cfg
->other_config
,
1211 "enable-output-sampling", false);
1214 if (n_fe_opts
> 0) {
1215 struct ofproto_ipfix_flow_exporter_options
*opts
;
1216 fe_opts
= xcalloc(n_fe_opts
, sizeof *fe_opts
);
1218 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1219 if (ovsrec_fscs_is_valid(fe_cfg
, br
)) {
1220 opts
->collector_set_id
= fe_cfg
->id
;
1221 sset_init(&opts
->targets
);
1222 sset_add_array(&opts
->targets
, fe_cfg
->ipfix
->targets
,
1223 fe_cfg
->ipfix
->n_targets
);
1224 opts
->cache_active_timeout
= fe_cfg
->ipfix
->cache_active_timeout
1225 ? *fe_cfg
->ipfix
->cache_active_timeout
: 0;
1226 opts
->cache_max_flows
= fe_cfg
->ipfix
->cache_max_flows
1227 ? *fe_cfg
->ipfix
->cache_max_flows
: 0;
1228 opts
->enable_tunnel_sampling
= smap_get_bool(
1229 &fe_cfg
->ipfix
->other_config
,
1230 "enable-tunnel-sampling", true);
1236 ofproto_set_ipfix(br
->ofproto
, valid_be_cfg
? &be_opts
: NULL
, fe_opts
,
1240 sset_destroy(&be_opts
.targets
);
1243 if (n_fe_opts
> 0) {
1244 struct ofproto_ipfix_flow_exporter_options
*opts
= fe_opts
;
1246 for (i
= 0; i
< n_fe_opts
; i
++) {
1247 sset_destroy(&opts
->targets
);
1255 port_configure_stp(const struct ofproto
*ofproto
, struct port
*port
,
1256 struct ofproto_port_stp_settings
*port_s
,
1257 int *port_num_counter
, unsigned long *port_num_bitmap
)
1259 const char *config_str
;
1260 struct iface
*iface
;
1262 if (!smap_get_bool(&port
->cfg
->other_config
, "stp-enable", true)) {
1263 port_s
->enable
= false;
1266 port_s
->enable
= true;
1269 /* STP over bonds is not supported. */
1270 if (!ovs_list_is_singleton(&port
->ifaces
)) {
1271 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1273 port_s
->enable
= false;
1277 iface
= CONTAINER_OF(ovs_list_front(&port
->ifaces
), struct iface
, port_elem
);
1279 /* Internal ports shouldn't participate in spanning tree, so
1281 if (!strcmp(iface
->type
, "internal")) {
1282 VLOG_DBG("port %s: disable STP on internal ports", port
->name
);
1283 port_s
->enable
= false;
1287 /* STP on mirror output ports is not supported. */
1288 if (ofproto_is_mirror_output_bundle(ofproto
, port
)) {
1289 VLOG_DBG("port %s: disable STP on mirror ports", port
->name
);
1290 port_s
->enable
= false;
1294 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-num");
1296 unsigned long int port_num
= strtoul(config_str
, NULL
, 0);
1297 int port_idx
= port_num
- 1;
1299 if (port_num
< 1 || port_num
> STP_MAX_PORTS
) {
1300 VLOG_ERR("port %s: invalid stp-port-num", port
->name
);
1301 port_s
->enable
= false;
1305 if (bitmap_is_set(port_num_bitmap
, port_idx
)) {
1306 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1307 port
->name
, port_num
);
1308 port_s
->enable
= false;
1311 bitmap_set1(port_num_bitmap
, port_idx
);
1312 port_s
->port_num
= port_idx
;
1314 if (*port_num_counter
>= STP_MAX_PORTS
) {
1315 VLOG_ERR("port %s: too many STP ports, disabling", port
->name
);
1316 port_s
->enable
= false;
1320 port_s
->port_num
= (*port_num_counter
)++;
1323 config_str
= smap_get(&port
->cfg
->other_config
, "stp-path-cost");
1325 port_s
->path_cost
= strtoul(config_str
, NULL
, 10);
1327 enum netdev_features current
;
1330 netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1331 mbps
= netdev_features_to_bps(current
, 100 * 1000 * 1000) / 1000000;
1332 port_s
->path_cost
= stp_convert_speed_to_cost(mbps
);
1335 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-priority");
1337 port_s
->priority
= strtoul(config_str
, NULL
, 0);
1339 port_s
->priority
= STP_DEFAULT_PORT_PRIORITY
;
1344 port_configure_rstp(const struct ofproto
*ofproto
, struct port
*port
,
1345 struct ofproto_port_rstp_settings
*port_s
, int *port_num_counter
)
1347 const char *config_str
;
1348 struct iface
*iface
;
1350 if (!smap_get_bool(&port
->cfg
->other_config
, "rstp-enable", true)) {
1351 port_s
->enable
= false;
1354 port_s
->enable
= true;
1357 /* RSTP over bonds is not supported. */
1358 if (!ovs_list_is_singleton(&port
->ifaces
)) {
1359 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1361 port_s
->enable
= false;
1365 iface
= CONTAINER_OF(ovs_list_front(&port
->ifaces
), struct iface
, port_elem
);
1367 /* Internal ports shouldn't participate in spanning tree, so
1369 if (!strcmp(iface
->type
, "internal")) {
1370 VLOG_DBG("port %s: disable RSTP on internal ports", port
->name
);
1371 port_s
->enable
= false;
1375 /* RSTP on mirror output ports is not supported. */
1376 if (ofproto_is_mirror_output_bundle(ofproto
, port
)) {
1377 VLOG_DBG("port %s: disable RSTP on mirror ports", port
->name
);
1378 port_s
->enable
= false;
1382 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-port-num");
1384 unsigned long int port_num
= strtoul(config_str
, NULL
, 0);
1385 if (port_num
< 1 || port_num
> RSTP_MAX_PORTS
) {
1386 VLOG_ERR("port %s: invalid rstp-port-num", port
->name
);
1387 port_s
->enable
= false;
1390 port_s
->port_num
= port_num
;
1392 if (*port_num_counter
>= RSTP_MAX_PORTS
) {
1393 VLOG_ERR("port %s: too many RSTP ports, disabling", port
->name
);
1394 port_s
->enable
= false;
1397 /* If rstp-port-num is not specified, use 0.
1398 * rstp_port_set_port_number() will look for the first free one. */
1399 port_s
->port_num
= 0;
1402 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-path-cost");
1404 port_s
->path_cost
= strtoul(config_str
, NULL
, 10);
1406 enum netdev_features current
;
1409 netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1410 mbps
= netdev_features_to_bps(current
, 100 * 1000 * 1000) / 1000000;
1411 port_s
->path_cost
= rstp_convert_speed_to_cost(mbps
);
1414 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-port-priority");
1416 port_s
->priority
= strtoul(config_str
, NULL
, 0);
1418 port_s
->priority
= RSTP_DEFAULT_PORT_PRIORITY
;
1421 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-admin-p2p-mac");
1423 port_s
->admin_p2p_mac_state
= strtoul(config_str
, NULL
, 0);
1425 port_s
->admin_p2p_mac_state
= RSTP_ADMIN_P2P_MAC_FORCE_TRUE
;
1428 port_s
->admin_port_state
= smap_get_bool(&port
->cfg
->other_config
,
1429 "rstp-admin-port-state", true);
1431 port_s
->admin_edge_port
= smap_get_bool(&port
->cfg
->other_config
,
1432 "rstp-port-admin-edge", false);
1433 port_s
->auto_edge
= smap_get_bool(&port
->cfg
->other_config
,
1434 "rstp-port-auto-edge", true);
1435 port_s
->mcheck
= smap_get_bool(&port
->cfg
->other_config
,
1436 "rstp-port-mcheck", false);
1439 /* Set spanning tree configuration on 'br'. */
1441 bridge_configure_stp(struct bridge
*br
, bool enable_stp
)
1444 ofproto_set_stp(br
->ofproto
, NULL
);
1446 struct ofproto_stp_settings br_s
;
1447 const char *config_str
;
1449 int port_num_counter
;
1450 unsigned long *port_num_bitmap
;
1452 config_str
= smap_get(&br
->cfg
->other_config
, "stp-system-id");
1456 if (eth_addr_from_string(config_str
, &ea
)) {
1457 br_s
.system_id
= eth_addr_to_uint64(ea
);
1459 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1460 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1461 "to "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(br
->ea
));
1464 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1467 config_str
= smap_get(&br
->cfg
->other_config
, "stp-priority");
1469 br_s
.priority
= strtoul(config_str
, NULL
, 0);
1471 br_s
.priority
= STP_DEFAULT_BRIDGE_PRIORITY
;
1474 config_str
= smap_get(&br
->cfg
->other_config
, "stp-hello-time");
1476 br_s
.hello_time
= strtoul(config_str
, NULL
, 10) * 1000;
1478 br_s
.hello_time
= STP_DEFAULT_HELLO_TIME
;
1481 config_str
= smap_get(&br
->cfg
->other_config
, "stp-max-age");
1483 br_s
.max_age
= strtoul(config_str
, NULL
, 10) * 1000;
1485 br_s
.max_age
= STP_DEFAULT_MAX_AGE
;
1488 config_str
= smap_get(&br
->cfg
->other_config
, "stp-forward-delay");
1490 br_s
.fwd_delay
= strtoul(config_str
, NULL
, 10) * 1000;
1492 br_s
.fwd_delay
= STP_DEFAULT_FWD_DELAY
;
1495 /* Configure STP on the bridge. */
1496 if (ofproto_set_stp(br
->ofproto
, &br_s
)) {
1497 VLOG_ERR("bridge %s: could not enable STP", br
->name
);
1501 /* Users must either set the port number with the "stp-port-num"
1502 * configuration on all ports or none. If manual configuration
1503 * is not done, then we allocate them sequentially. */
1504 port_num_counter
= 0;
1505 port_num_bitmap
= bitmap_allocate(STP_MAX_PORTS
);
1506 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1507 struct ofproto_port_stp_settings port_s
;
1508 struct iface
*iface
;
1510 port_configure_stp(br
->ofproto
, port
, &port_s
,
1511 &port_num_counter
, port_num_bitmap
);
1513 /* As bonds are not supported, just apply configuration to
1514 * all interfaces. */
1515 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1516 if (ofproto_port_set_stp(br
->ofproto
, iface
->ofp_port
,
1518 VLOG_ERR("port %s: could not enable STP", port
->name
);
1524 if (bitmap_scan(port_num_bitmap
, 1, 0, STP_MAX_PORTS
) != STP_MAX_PORTS
1525 && port_num_counter
) {
1526 VLOG_ERR("bridge %s: must manually configure all STP port "
1527 "IDs or none, disabling", br
->name
);
1528 ofproto_set_stp(br
->ofproto
, NULL
);
1530 bitmap_free(port_num_bitmap
);
1535 bridge_configure_rstp(struct bridge
*br
, bool enable_rstp
)
1538 ofproto_set_rstp(br
->ofproto
, NULL
);
1540 struct ofproto_rstp_settings br_s
;
1541 const char *config_str
;
1543 int port_num_counter
;
1545 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-address");
1549 if (eth_addr_from_string(config_str
, &ea
)) {
1550 br_s
.address
= eth_addr_to_uint64(ea
);
1553 br_s
.address
= eth_addr_to_uint64(br
->ea
);
1554 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1555 "to "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(br
->ea
));
1559 br_s
.address
= eth_addr_to_uint64(br
->ea
);
1562 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-priority");
1564 br_s
.priority
= strtoul(config_str
, NULL
, 0);
1566 br_s
.priority
= RSTP_DEFAULT_PRIORITY
;
1569 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-ageing-time");
1571 br_s
.ageing_time
= strtoul(config_str
, NULL
, 0);
1573 br_s
.ageing_time
= RSTP_DEFAULT_AGEING_TIME
;
1576 config_str
= smap_get(&br
->cfg
->other_config
,
1577 "rstp-force-protocol-version");
1579 br_s
.force_protocol_version
= strtoul(config_str
, NULL
, 0);
1581 br_s
.force_protocol_version
= FPV_DEFAULT
;
1584 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-max-age");
1586 br_s
.bridge_max_age
= strtoul(config_str
, NULL
, 10);
1588 br_s
.bridge_max_age
= RSTP_DEFAULT_BRIDGE_MAX_AGE
;
1591 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-forward-delay");
1593 br_s
.bridge_forward_delay
= strtoul(config_str
, NULL
, 10);
1595 br_s
.bridge_forward_delay
= RSTP_DEFAULT_BRIDGE_FORWARD_DELAY
;
1598 config_str
= smap_get(&br
->cfg
->other_config
,
1599 "rstp-transmit-hold-count");
1601 br_s
.transmit_hold_count
= strtoul(config_str
, NULL
, 10);
1603 br_s
.transmit_hold_count
= RSTP_DEFAULT_TRANSMIT_HOLD_COUNT
;
1606 /* Configure RSTP on the bridge. */
1607 if (ofproto_set_rstp(br
->ofproto
, &br_s
)) {
1608 VLOG_ERR("bridge %s: could not enable RSTP", br
->name
);
1612 port_num_counter
= 0;
1613 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1614 struct ofproto_port_rstp_settings port_s
;
1615 struct iface
*iface
;
1617 port_configure_rstp(br
->ofproto
, port
, &port_s
,
1620 /* As bonds are not supported, just apply configuration to
1621 * all interfaces. */
1622 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1623 if (ofproto_port_set_rstp(br
->ofproto
, iface
->ofp_port
,
1625 VLOG_ERR("port %s: could not enable RSTP", port
->name
);
1634 bridge_configure_spanning_tree(struct bridge
*br
)
1636 bool enable_rstp
= br
->cfg
->rstp_enable
;
1637 bool enable_stp
= br
->cfg
->stp_enable
;
1639 if (enable_rstp
&& enable_stp
) {
1640 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1641 "configured; enabling RSTP", br
->name
);
1645 bridge_configure_stp(br
, enable_stp
);
1646 bridge_configure_rstp(br
, enable_rstp
);
1650 bridge_has_bond_fake_iface(const struct bridge
*br
, const char *name
)
1652 const struct port
*port
= port_lookup(br
, name
);
1653 return port
&& port_is_bond_fake_iface(port
);
1657 port_is_bond_fake_iface(const struct port
*port
)
1659 return port
->cfg
->bond_fake_iface
&& !ovs_list_is_short(&port
->ifaces
);
1663 add_del_bridges(const struct ovsrec_open_vswitch
*cfg
)
1665 struct bridge
*br
, *next
;
1666 struct shash_node
*node
;
1667 struct shash new_br
;
1670 /* Collect new bridges' names and types. */
1671 shash_init(&new_br
);
1672 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1673 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1674 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1676 if (strchr(br_cfg
->name
, '/') || strchr(br_cfg
->name
, '\\')) {
1677 /* Prevent remote ovsdb-server users from accessing arbitrary
1678 * directories, e.g. consider a bridge named "../../../etc/".
1680 * Prohibiting "\" is only necessary on Windows but it's no great
1681 * loss elsewhere. */
1682 VLOG_WARN_RL(&rl
, "ignoring bridge with invalid name \"%s\"",
1684 } else if (!shash_add_once(&new_br
, br_cfg
->name
, br_cfg
)) {
1685 VLOG_WARN_RL(&rl
, "bridge %s specified twice", br_cfg
->name
);
1689 /* Get rid of deleted bridges or those whose types have changed.
1690 * Update 'cfg' of bridges that still exist. */
1691 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
1692 br
->cfg
= shash_find_data(&new_br
, br
->name
);
1693 if (!br
->cfg
|| strcmp(br
->type
, ofproto_normalize_type(
1694 br
->cfg
->datapath_type
))) {
1695 bridge_destroy(br
, true);
1699 /* Add new bridges. */
1700 SHASH_FOR_EACH(node
, &new_br
) {
1701 const struct ovsrec_bridge
*br_cfg
= node
->data
;
1702 struct bridge
*br
= bridge_lookup(br_cfg
->name
);
1704 bridge_create(br_cfg
);
1708 shash_destroy(&new_br
);
1711 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1712 * Returns 0 if successful, otherwise a positive errno value. */
1714 iface_set_netdev_config(const struct ovsrec_interface
*iface_cfg
,
1715 struct netdev
*netdev
, char **errp
)
1717 return netdev_set_config(netdev
, &iface_cfg
->options
, errp
);
1720 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1721 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1723 * If successful, returns 0 and stores the network device in '*netdevp'. On
1724 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1726 iface_do_create(const struct bridge
*br
,
1727 const struct ovsrec_interface
*iface_cfg
,
1728 ofp_port_t
*ofp_portp
, struct netdev
**netdevp
,
1731 struct netdev
*netdev
= NULL
;
1734 if (netdev_is_reserved_name(iface_cfg
->name
)) {
1735 VLOG_WARN("could not create interface %s, name is reserved",
1741 error
= netdev_open(iface_cfg
->name
,
1742 iface_get_type(iface_cfg
, br
->cfg
), &netdev
);
1744 VLOG_WARN_BUF(errp
, "could not open network device %s (%s)",
1745 iface_cfg
->name
, ovs_strerror(error
));
1749 error
= iface_set_netdev_config(iface_cfg
, netdev
, errp
);
1754 *ofp_portp
= iface_pick_ofport(iface_cfg
);
1755 error
= ofproto_port_add(br
->ofproto
, netdev
, ofp_portp
);
1760 VLOG_INFO("bridge %s: added interface %s on port %d",
1761 br
->name
, iface_cfg
->name
, *ofp_portp
);
1768 netdev_close(netdev
);
1772 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1773 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1774 * automatically allocated for the iface. Takes ownership of and
1775 * deallocates 'if_cfg'.
1777 * Return true if an iface is successfully created, false otherwise. */
1779 iface_create(struct bridge
*br
, const struct ovsrec_interface
*iface_cfg
,
1780 const struct ovsrec_port
*port_cfg
)
1782 struct netdev
*netdev
;
1783 struct iface
*iface
;
1784 ofp_port_t ofp_port
;
1789 /* Do the bits that can fail up front. */
1790 ovs_assert(!iface_lookup(br
, iface_cfg
->name
));
1791 error
= iface_do_create(br
, iface_cfg
, &ofp_port
, &netdev
, &errp
);
1793 iface_clear_db_record(iface_cfg
, errp
);
1798 /* Get or create the port structure. */
1799 port
= port_lookup(br
, port_cfg
->name
);
1801 port
= port_create(br
, port_cfg
);
1804 /* Create the iface structure. */
1805 iface
= xzalloc(sizeof *iface
);
1806 ovs_list_push_back(&port
->ifaces
, &iface
->port_elem
);
1807 hmap_insert(&br
->iface_by_name
, &iface
->name_node
,
1808 hash_string(iface_cfg
->name
, 0));
1810 iface
->name
= xstrdup(iface_cfg
->name
);
1811 iface
->ofp_port
= ofp_port
;
1812 iface
->netdev
= netdev
;
1813 iface
->type
= iface_get_type(iface_cfg
, br
->cfg
);
1814 iface
->cfg
= iface_cfg
;
1815 hmap_insert(&br
->ifaces
, &iface
->ofp_port_node
,
1816 hash_ofp_port(ofp_port
));
1818 /* Populate initial status in database. */
1819 iface_refresh_stats(iface
);
1820 iface_refresh_netdev_status(iface
);
1822 /* Add bond fake iface if necessary. */
1823 if (port_is_bond_fake_iface(port
)) {
1824 struct ofproto_port ofproto_port
;
1826 if (ofproto_port_query_by_name(br
->ofproto
, port
->name
,
1828 struct netdev
*netdev
;
1831 error
= netdev_open(port
->name
, "internal", &netdev
);
1833 ofp_port_t fake_ofp_port
= OFPP_NONE
;
1834 ofproto_port_add(br
->ofproto
, netdev
, &fake_ofp_port
);
1835 netdev_close(netdev
);
1837 VLOG_WARN("could not open network device %s (%s)",
1838 port
->name
, ovs_strerror(error
));
1841 /* Already exists, nothing to do. */
1842 ofproto_port_destroy(&ofproto_port
);
1849 /* Set forward BPDU option. */
1851 bridge_configure_forward_bpdu(struct bridge
*br
)
1853 ofproto_set_forward_bpdu(br
->ofproto
,
1854 smap_get_bool(&br
->cfg
->other_config
,
1859 /* Set MAC learning table configuration for 'br'. */
1861 bridge_configure_mac_table(struct bridge
*br
)
1863 const char *idle_time_str
;
1866 const char *mac_table_size_str
;
1869 idle_time_str
= smap_get(&br
->cfg
->other_config
, "mac-aging-time");
1870 idle_time
= (idle_time_str
&& atoi(idle_time_str
)
1871 ? atoi(idle_time_str
)
1872 : MAC_ENTRY_DEFAULT_IDLE_TIME
);
1874 mac_table_size_str
= smap_get(&br
->cfg
->other_config
, "mac-table-size");
1875 mac_table_size
= (mac_table_size_str
&& atoi(mac_table_size_str
)
1876 ? atoi(mac_table_size_str
)
1879 ofproto_set_mac_table_config(br
->ofproto
, idle_time
, mac_table_size
);
1882 /* Set multicast snooping table configuration for 'br'. */
1884 bridge_configure_mcast_snooping(struct bridge
*br
)
1886 if (!br
->cfg
->mcast_snooping_enable
) {
1887 ofproto_set_mcast_snooping(br
->ofproto
, NULL
);
1890 struct ofproto_mcast_snooping_settings br_s
;
1891 const char *idle_time_str
;
1892 const char *max_entries_str
;
1894 idle_time_str
= smap_get(&br
->cfg
->other_config
,
1895 "mcast-snooping-aging-time");
1896 br_s
.idle_time
= (idle_time_str
&& atoi(idle_time_str
)
1897 ? atoi(idle_time_str
)
1898 : MCAST_ENTRY_DEFAULT_IDLE_TIME
);
1900 max_entries_str
= smap_get(&br
->cfg
->other_config
,
1901 "mcast-snooping-table-size");
1902 br_s
.max_entries
= (max_entries_str
&& atoi(max_entries_str
)
1903 ? atoi(max_entries_str
)
1904 : MCAST_DEFAULT_MAX_ENTRIES
);
1906 br_s
.flood_unreg
= !smap_get_bool(&br
->cfg
->other_config
,
1907 "mcast-snooping-disable-flood-unregistered",
1910 /* Configure multicast snooping on the bridge */
1911 if (ofproto_set_mcast_snooping(br
->ofproto
, &br_s
)) {
1912 VLOG_ERR("bridge %s: could not enable multicast snooping",
1917 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1918 struct ofproto_mcast_snooping_port_settings port_s
;
1919 port_s
.flood
= smap_get_bool(&port
->cfg
->other_config
,
1920 "mcast-snooping-flood", false);
1921 port_s
.flood_reports
= smap_get_bool(&port
->cfg
->other_config
,
1922 "mcast-snooping-flood-reports", false);
1923 if (ofproto_port_set_mcast_snooping(br
->ofproto
, port
, &port_s
)) {
1924 VLOG_ERR("port %s: could not configure mcast snooping",
1932 find_local_hw_addr(const struct bridge
*br
, struct eth_addr
*ea
,
1933 const struct port
*fake_br
, struct iface
**hw_addr_iface
)
1935 struct hmapx mirror_output_ports
;
1937 bool found_addr
= false;
1941 /* Mirror output ports don't participate in picking the local hardware
1942 * address. ofproto can't help us find out whether a given port is a
1943 * mirror output because we haven't configured mirrors yet, so we need to
1944 * accumulate them ourselves. */
1945 hmapx_init(&mirror_output_ports
);
1946 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
1947 struct ovsrec_mirror
*m
= br
->cfg
->mirrors
[i
];
1948 if (m
->output_port
) {
1949 hmapx_add(&mirror_output_ports
, m
->output_port
);
1953 /* Otherwise choose the minimum non-local MAC address among all of the
1955 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1956 struct eth_addr iface_ea
;
1957 struct iface
*candidate
;
1958 struct iface
*iface
;
1960 /* Mirror output ports don't participate. */
1961 if (hmapx_contains(&mirror_output_ports
, port
->cfg
)) {
1965 /* Choose the MAC address to represent the port. */
1967 if (port
->cfg
->mac
&& eth_addr_from_string(port
->cfg
->mac
,
1969 /* Find the interface with this Ethernet address (if any) so that
1970 * we can provide the correct devname to the caller. */
1971 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1972 struct eth_addr candidate_ea
;
1973 if (!netdev_get_etheraddr(candidate
->netdev
, &candidate_ea
)
1974 && eth_addr_equals(iface_ea
, candidate_ea
)) {
1979 /* Choose the interface whose MAC address will represent the port.
1980 * The Linux kernel bonding code always chooses the MAC address of
1981 * the first slave added to a bond, and the Fedora networking
1982 * scripts always add slaves to a bond in alphabetical order, so
1983 * for compatibility we choose the interface with the name that is
1984 * first in alphabetical order. */
1985 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1986 if (!iface
|| strcmp(candidate
->name
, iface
->name
) < 0) {
1991 /* The local port doesn't count (since we're trying to choose its
1992 * MAC address anyway). */
1993 if (iface
->ofp_port
== OFPP_LOCAL
) {
1997 /* For fake bridges we only choose from ports with the same tag */
1998 if (fake_br
&& fake_br
->cfg
&& fake_br
->cfg
->tag
) {
1999 if (!port
->cfg
->tag
) {
2002 if (*port
->cfg
->tag
!= *fake_br
->cfg
->tag
) {
2008 error
= netdev_get_etheraddr(iface
->netdev
, &iface_ea
);
2014 /* Compare against our current choice. */
2015 if (!eth_addr_is_multicast(iface_ea
) &&
2016 !eth_addr_is_local(iface_ea
) &&
2017 !eth_addr_is_reserved(iface_ea
) &&
2018 !eth_addr_is_zero(iface_ea
) &&
2019 (!found_addr
|| eth_addr_compare_3way(iface_ea
, *ea
) < 0))
2022 *hw_addr_iface
= iface
;
2028 *ea
= br
->default_ea
;
2029 *hw_addr_iface
= NULL
;
2032 hmapx_destroy(&mirror_output_ports
);
2036 bridge_pick_local_hw_addr(struct bridge
*br
, struct eth_addr
*ea
,
2037 struct iface
**hw_addr_iface
)
2040 *hw_addr_iface
= NULL
;
2042 /* Did the user request a particular MAC? */
2043 hwaddr
= smap_get(&br
->cfg
->other_config
, "hwaddr");
2044 if (hwaddr
&& eth_addr_from_string(hwaddr
, ea
)) {
2045 if (eth_addr_is_multicast(*ea
)) {
2046 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2047 "address "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(*ea
));
2048 } else if (eth_addr_is_zero(*ea
)) {
2049 VLOG_ERR("bridge %s: cannot set MAC address to zero", br
->name
);
2055 /* Find a local hw address */
2056 find_local_hw_addr(br
, ea
, NULL
, hw_addr_iface
);
2059 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2060 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2061 * an interface on 'br', then that interface must be passed in as
2062 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2063 * 'hw_addr_iface' must be passed in as a null pointer. */
2065 bridge_pick_datapath_id(struct bridge
*br
,
2066 const struct eth_addr bridge_ea
,
2067 struct iface
*hw_addr_iface
)
2070 * The procedure for choosing a bridge MAC address will, in the most
2071 * ordinary case, also choose a unique MAC that we can use as a datapath
2072 * ID. In some special cases, though, multiple bridges will end up with
2073 * the same MAC address. This is OK for the bridges, but it will confuse
2074 * the OpenFlow controller, because each datapath needs a unique datapath
2077 * Datapath IDs must be unique. It is also very desirable that they be
2078 * stable from one run to the next, so that policy set on a datapath
2081 const char *datapath_id
;
2084 datapath_id
= smap_get(&br
->cfg
->other_config
, "datapath-id");
2085 if (datapath_id
&& dpid_from_string(datapath_id
, &dpid
)) {
2089 if (!hw_addr_iface
) {
2091 * A purely internal bridge, that is, one that has no non-virtual
2092 * network devices on it at all, is difficult because it has no
2093 * natural unique identifier at all.
2095 * When the host is a XenServer, we handle this case by hashing the
2096 * host's UUID with the name of the bridge. Names of bridges are
2097 * persistent across XenServer reboots, although they can be reused if
2098 * an internal network is destroyed and then a new one is later
2099 * created, so this is fairly effective.
2101 * When the host is not a XenServer, we punt by using a random MAC
2102 * address on each run.
2104 const char *host_uuid
= xenserver_get_host_uuid();
2106 char *combined
= xasprintf("%s,%s", host_uuid
, br
->name
);
2107 dpid
= dpid_from_hash(combined
, strlen(combined
));
2113 return eth_addr_to_uint64(bridge_ea
);
2117 dpid_from_hash(const void *data
, size_t n
)
2120 uint8_t bytes
[SHA1_DIGEST_SIZE
];
2124 sha1_bytes(data
, n
, hash
.bytes
);
2125 eth_addr_mark_random(&hash
.ea
);
2126 return eth_addr_to_uint64(hash
.ea
);
2130 iface_refresh_netdev_status(struct iface
*iface
)
2134 enum netdev_features current
;
2135 enum netdev_flags flags
;
2136 const char *link_state
;
2137 struct eth_addr mac
;
2138 int64_t bps
, mtu_64
, ifindex64
, link_resets
;
2141 if (iface_is_synthetic(iface
)) {
2145 if (iface
->change_seq
== netdev_get_change_seq(iface
->netdev
)
2146 && !status_txn_try_again
) {
2150 iface
->change_seq
= netdev_get_change_seq(iface
->netdev
);
2154 if (!netdev_get_status(iface
->netdev
, &smap
)) {
2155 ovsrec_interface_set_status(iface
->cfg
, &smap
);
2157 ovsrec_interface_set_status(iface
->cfg
, NULL
);
2160 smap_destroy(&smap
);
2162 error
= netdev_get_flags(iface
->netdev
, &flags
);
2164 const char *state
= flags
& NETDEV_UP
? "up" : "down";
2166 ovsrec_interface_set_admin_state(iface
->cfg
, state
);
2168 ovsrec_interface_set_admin_state(iface
->cfg
, NULL
);
2171 link_state
= netdev_get_carrier(iface
->netdev
) ? "up" : "down";
2172 ovsrec_interface_set_link_state(iface
->cfg
, link_state
);
2174 link_resets
= netdev_get_carrier_resets(iface
->netdev
);
2175 ovsrec_interface_set_link_resets(iface
->cfg
, &link_resets
, 1);
2177 error
= netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
2178 bps
= !error
? netdev_features_to_bps(current
, 0) : 0;
2180 ovsrec_interface_set_duplex(iface
->cfg
,
2181 netdev_features_is_full_duplex(current
)
2183 ovsrec_interface_set_link_speed(iface
->cfg
, &bps
, 1);
2185 ovsrec_interface_set_duplex(iface
->cfg
, NULL
);
2186 ovsrec_interface_set_link_speed(iface
->cfg
, NULL
, 0);
2189 error
= netdev_get_mtu(iface
->netdev
, &mtu
);
2192 ovsrec_interface_set_mtu(iface
->cfg
, &mtu_64
, 1);
2194 ovsrec_interface_set_mtu(iface
->cfg
, NULL
, 0);
2197 error
= netdev_get_etheraddr(iface
->netdev
, &mac
);
2199 char mac_string
[ETH_ADDR_STRLEN
+ 1];
2201 snprintf(mac_string
, sizeof mac_string
,
2202 ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
2203 ovsrec_interface_set_mac_in_use(iface
->cfg
, mac_string
);
2205 ovsrec_interface_set_mac_in_use(iface
->cfg
, NULL
);
2208 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2209 * if there is no valid ifindex number. */
2210 ifindex64
= netdev_get_ifindex(iface
->netdev
);
2211 if (ifindex64
< 0) {
2214 ovsrec_interface_set_ifindex(iface
->cfg
, &ifindex64
, 1);
2218 iface_refresh_ofproto_status(struct iface
*iface
)
2222 if (iface_is_synthetic(iface
)) {
2226 current
= ofproto_port_is_lacp_current(iface
->port
->bridge
->ofproto
,
2230 ovsrec_interface_set_lacp_current(iface
->cfg
, &bl
, 1);
2232 ovsrec_interface_set_lacp_current(iface
->cfg
, NULL
, 0);
2235 if (ofproto_port_cfm_status_changed(iface
->port
->bridge
->ofproto
,
2237 || status_txn_try_again
) {
2238 iface_refresh_cfm_stats(iface
);
2241 if (ofproto_port_bfd_status_changed(iface
->port
->bridge
->ofproto
,
2243 || status_txn_try_again
) {
2247 ofproto_port_get_bfd_status(iface
->port
->bridge
->ofproto
,
2248 iface
->ofp_port
, &smap
);
2249 ovsrec_interface_set_bfd_status(iface
->cfg
, &smap
);
2250 smap_destroy(&smap
);
2254 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2257 iface_refresh_cfm_stats(struct iface
*iface
)
2259 const struct ovsrec_interface
*cfg
= iface
->cfg
;
2260 struct cfm_status status
;
2263 error
= ofproto_port_get_cfm_status(iface
->port
->bridge
->ofproto
,
2264 iface
->ofp_port
, &status
);
2266 ovsrec_interface_set_cfm_fault(cfg
, NULL
, 0);
2267 ovsrec_interface_set_cfm_fault_status(cfg
, NULL
, 0);
2268 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
2269 ovsrec_interface_set_cfm_flap_count(cfg
, NULL
, 0);
2270 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
2271 ovsrec_interface_set_cfm_remote_mpids(cfg
, NULL
, 0);
2273 const char *reasons
[CFM_FAULT_N_REASONS
];
2274 int64_t cfm_health
= status
.health
;
2275 int64_t cfm_flap_count
= status
.flap_count
;
2276 bool faulted
= status
.faults
!= 0;
2279 ovsrec_interface_set_cfm_fault(cfg
, &faulted
, 1);
2282 for (i
= 0; i
< CFM_FAULT_N_REASONS
; i
++) {
2283 int reason
= 1 << i
;
2284 if (status
.faults
& reason
) {
2285 reasons
[j
++] = cfm_fault_reason_to_str(reason
);
2288 ovsrec_interface_set_cfm_fault_status(cfg
, reasons
, j
);
2290 ovsrec_interface_set_cfm_flap_count(cfg
, &cfm_flap_count
, 1);
2292 if (status
.remote_opstate
>= 0) {
2293 const char *remote_opstate
= status
.remote_opstate
? "up" : "down";
2294 ovsrec_interface_set_cfm_remote_opstate(cfg
, remote_opstate
);
2296 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
2299 ovsrec_interface_set_cfm_remote_mpids(cfg
,
2300 (const int64_t *)status
.rmps
,
2302 if (cfm_health
>= 0) {
2303 ovsrec_interface_set_cfm_health(cfg
, &cfm_health
, 1);
2305 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
2313 iface_refresh_stats(struct iface
*iface
)
2315 #define IFACE_STATS \
2316 IFACE_STAT(rx_packets, "rx_packets") \
2317 IFACE_STAT(tx_packets, "tx_packets") \
2318 IFACE_STAT(rx_bytes, "rx_bytes") \
2319 IFACE_STAT(tx_bytes, "tx_bytes") \
2320 IFACE_STAT(rx_dropped, "rx_dropped") \
2321 IFACE_STAT(tx_dropped, "tx_dropped") \
2322 IFACE_STAT(rx_errors, "rx_errors") \
2323 IFACE_STAT(tx_errors, "tx_errors") \
2324 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2325 IFACE_STAT(rx_over_errors, "rx_over_err") \
2326 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2327 IFACE_STAT(collisions, "collisions") \
2328 IFACE_STAT(rx_1_to_64_packets, "rx_1_to_64_packets") \
2329 IFACE_STAT(rx_65_to_127_packets, "rx_65_to_127_packets") \
2330 IFACE_STAT(rx_128_to_255_packets, "rx_128_to_255_packets") \
2331 IFACE_STAT(rx_256_to_511_packets, "rx_256_to_511_packets") \
2332 IFACE_STAT(rx_512_to_1023_packets, "rx_512_to_1023_packets") \
2333 IFACE_STAT(rx_1024_to_1522_packets, "rx_1024_to_1518_packets") \
2334 IFACE_STAT(rx_1523_to_max_packets, "rx_1523_to_max_packets") \
2335 IFACE_STAT(tx_1_to_64_packets, "tx_1_to_64_packets") \
2336 IFACE_STAT(tx_65_to_127_packets, "tx_65_to_127_packets") \
2337 IFACE_STAT(tx_128_to_255_packets, "tx_128_to_255_packets") \
2338 IFACE_STAT(tx_256_to_511_packets, "tx_256_to_511_packets") \
2339 IFACE_STAT(tx_512_to_1023_packets, "tx_512_to_1023_packets") \
2340 IFACE_STAT(tx_1024_to_1522_packets, "tx_1024_to_1518_packets") \
2341 IFACE_STAT(tx_1523_to_max_packets, "tx_1523_to_max_packets") \
2342 IFACE_STAT(tx_multicast_packets, "tx_multicast_packets") \
2343 IFACE_STAT(rx_broadcast_packets, "rx_broadcast_packets") \
2344 IFACE_STAT(tx_broadcast_packets, "tx_broadcast_packets") \
2345 IFACE_STAT(rx_undersized_errors, "rx_undersized_errors") \
2346 IFACE_STAT(rx_oversize_errors, "rx_oversize_errors") \
2347 IFACE_STAT(rx_fragmented_errors, "rx_fragmented_errors") \
2348 IFACE_STAT(rx_jabber_errors, "rx_jabber_errors")
2350 #define IFACE_STAT(MEMBER, NAME) + 1
2351 enum { N_IFACE_STATS
= IFACE_STATS
};
2353 int64_t values
[N_IFACE_STATS
];
2354 const char *keys
[N_IFACE_STATS
];
2357 struct netdev_stats stats
;
2359 if (iface_is_synthetic(iface
)) {
2363 /* Intentionally ignore return value, since errors will set 'stats' to
2364 * all-1s, and we will deal with that correctly below. */
2365 netdev_get_stats(iface
->netdev
, &stats
);
2367 /* Copy statistics into keys[] and values[]. */
2369 #define IFACE_STAT(MEMBER, NAME) \
2370 if (stats.MEMBER != UINT64_MAX) { \
2372 values[n] = stats.MEMBER; \
2377 ovs_assert(n
<= N_IFACE_STATS
);
2379 ovsrec_interface_set_statistics(iface
->cfg
, keys
, values
, n
);
2384 br_refresh_datapath_info(struct bridge
*br
)
2386 const char *version
;
2388 version
= (br
->ofproto
&& br
->ofproto
->ofproto_class
->get_datapath_version
2389 ? br
->ofproto
->ofproto_class
->get_datapath_version(br
->ofproto
)
2392 ovsrec_bridge_set_datapath_version(br
->cfg
,
2393 version
? version
: "<unknown>");
2397 br_refresh_stp_status(struct bridge
*br
)
2399 struct smap smap
= SMAP_INITIALIZER(&smap
);
2400 struct ofproto
*ofproto
= br
->ofproto
;
2401 struct ofproto_stp_status status
;
2403 if (ofproto_get_stp_status(ofproto
, &status
)) {
2407 if (!status
.enabled
) {
2408 ovsrec_bridge_set_status(br
->cfg
, NULL
);
2412 smap_add_format(&smap
, "stp_bridge_id", STP_ID_FMT
,
2413 STP_ID_ARGS(status
.bridge_id
));
2414 smap_add_format(&smap
, "stp_designated_root", STP_ID_FMT
,
2415 STP_ID_ARGS(status
.designated_root
));
2416 smap_add_format(&smap
, "stp_root_path_cost", "%d", status
.root_path_cost
);
2418 ovsrec_bridge_set_status(br
->cfg
, &smap
);
2419 smap_destroy(&smap
);
2423 port_refresh_stp_status(struct port
*port
)
2425 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2426 struct iface
*iface
;
2427 struct ofproto_port_stp_status status
;
2430 if (port_is_synthetic(port
)) {
2434 /* STP doesn't currently support bonds. */
2435 if (!ovs_list_is_singleton(&port
->ifaces
)) {
2436 ovsrec_port_set_status(port
->cfg
, NULL
);
2440 iface
= CONTAINER_OF(ovs_list_front(&port
->ifaces
), struct iface
, port_elem
);
2441 if (ofproto_port_get_stp_status(ofproto
, iface
->ofp_port
, &status
)) {
2445 if (!status
.enabled
) {
2446 ovsrec_port_set_status(port
->cfg
, NULL
);
2450 /* Set Status column. */
2452 smap_add_format(&smap
, "stp_port_id", STP_PORT_ID_FMT
, status
.port_id
);
2453 smap_add(&smap
, "stp_state", stp_state_name(status
.state
));
2454 smap_add_format(&smap
, "stp_sec_in_state", "%u", status
.sec_in_state
);
2455 smap_add(&smap
, "stp_role", stp_role_name(status
.role
));
2456 ovsrec_port_set_status(port
->cfg
, &smap
);
2457 smap_destroy(&smap
);
2461 port_refresh_stp_stats(struct port
*port
)
2463 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2464 struct iface
*iface
;
2465 struct ofproto_port_stp_stats stats
;
2466 const char *keys
[3];
2467 int64_t int_values
[3];
2469 if (port_is_synthetic(port
)) {
2473 /* STP doesn't currently support bonds. */
2474 if (!ovs_list_is_singleton(&port
->ifaces
)) {
2478 iface
= CONTAINER_OF(ovs_list_front(&port
->ifaces
), struct iface
, port_elem
);
2479 if (ofproto_port_get_stp_stats(ofproto
, iface
->ofp_port
, &stats
)) {
2483 if (!stats
.enabled
) {
2484 ovsrec_port_set_statistics(port
->cfg
, NULL
, NULL
, 0);
2488 /* Set Statistics column. */
2489 keys
[0] = "stp_tx_count";
2490 int_values
[0] = stats
.tx_count
;
2491 keys
[1] = "stp_rx_count";
2492 int_values
[1] = stats
.rx_count
;
2493 keys
[2] = "stp_error_count";
2494 int_values
[2] = stats
.error_count
;
2496 ovsrec_port_set_statistics(port
->cfg
, keys
, int_values
,
2497 ARRAY_SIZE(int_values
));
2501 br_refresh_rstp_status(struct bridge
*br
)
2503 struct smap smap
= SMAP_INITIALIZER(&smap
);
2504 struct ofproto
*ofproto
= br
->ofproto
;
2505 struct ofproto_rstp_status status
;
2507 if (ofproto_get_rstp_status(ofproto
, &status
)) {
2510 if (!status
.enabled
) {
2511 ovsrec_bridge_set_rstp_status(br
->cfg
, NULL
);
2514 smap_add_format(&smap
, "rstp_bridge_id", RSTP_ID_FMT
,
2515 RSTP_ID_ARGS(status
.bridge_id
));
2516 smap_add_format(&smap
, "rstp_root_path_cost", "%"PRIu32
,
2517 status
.root_path_cost
);
2518 smap_add_format(&smap
, "rstp_root_id", RSTP_ID_FMT
,
2519 RSTP_ID_ARGS(status
.root_id
));
2520 smap_add_format(&smap
, "rstp_designated_id", RSTP_ID_FMT
,
2521 RSTP_ID_ARGS(status
.designated_id
));
2522 smap_add_format(&smap
, "rstp_designated_port_id", RSTP_PORT_ID_FMT
,
2523 status
.designated_port_id
);
2524 smap_add_format(&smap
, "rstp_bridge_port_id", RSTP_PORT_ID_FMT
,
2525 status
.bridge_port_id
);
2526 ovsrec_bridge_set_rstp_status(br
->cfg
, &smap
);
2527 smap_destroy(&smap
);
2531 port_refresh_rstp_status(struct port
*port
)
2533 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2534 struct iface
*iface
;
2535 struct ofproto_port_rstp_status status
;
2536 const char *keys
[4];
2537 int64_t int_values
[4];
2540 if (port_is_synthetic(port
)) {
2544 /* RSTP doesn't currently support bonds. */
2545 if (!ovs_list_is_singleton(&port
->ifaces
)) {
2546 ovsrec_port_set_rstp_status(port
->cfg
, NULL
);
2550 iface
= CONTAINER_OF(ovs_list_front(&port
->ifaces
), struct iface
, port_elem
);
2551 if (ofproto_port_get_rstp_status(ofproto
, iface
->ofp_port
, &status
)) {
2555 if (!status
.enabled
) {
2556 ovsrec_port_set_rstp_status(port
->cfg
, NULL
);
2557 ovsrec_port_set_rstp_statistics(port
->cfg
, NULL
, NULL
, 0);
2560 /* Set Status column. */
2563 smap_add_format(&smap
, "rstp_port_id", RSTP_PORT_ID_FMT
,
2565 smap_add_format(&smap
, "rstp_port_role", "%s",
2566 rstp_port_role_name(status
.role
));
2567 smap_add_format(&smap
, "rstp_port_state", "%s",
2568 rstp_state_name(status
.state
));
2569 smap_add_format(&smap
, "rstp_designated_bridge_id", RSTP_ID_FMT
,
2570 RSTP_ID_ARGS(status
.designated_bridge_id
));
2571 smap_add_format(&smap
, "rstp_designated_port_id", RSTP_PORT_ID_FMT
,
2572 status
.designated_port_id
);
2573 smap_add_format(&smap
, "rstp_designated_path_cost", "%"PRIu32
,
2574 status
.designated_path_cost
);
2576 ovsrec_port_set_rstp_status(port
->cfg
, &smap
);
2577 smap_destroy(&smap
);
2579 /* Set Statistics column. */
2580 keys
[0] = "rstp_tx_count";
2581 int_values
[0] = status
.tx_count
;
2582 keys
[1] = "rstp_rx_count";
2583 int_values
[1] = status
.rx_count
;
2584 keys
[2] = "rstp_uptime";
2585 int_values
[2] = status
.uptime
;
2586 keys
[3] = "rstp_error_count";
2587 int_values
[3] = status
.error_count
;
2588 ovsrec_port_set_rstp_statistics(port
->cfg
, keys
, int_values
,
2589 ARRAY_SIZE(int_values
));
2593 port_refresh_bond_status(struct port
*port
, bool force_update
)
2595 struct eth_addr mac
;
2597 /* Return if port is not a bond */
2598 if (ovs_list_is_singleton(&port
->ifaces
)) {
2602 if (bond_get_changed_active_slave(port
->name
, &mac
, force_update
)) {
2606 ds_put_format(&mac_s
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
2607 ovsrec_port_set_bond_active_slave(port
->cfg
, ds_cstr(&mac_s
));
2613 enable_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2615 return smap_get_bool(&cfg
->other_config
, "enable-statistics", false);
2619 reconfigure_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2621 bool enable
= enable_system_stats(cfg
);
2623 system_stats_enable(enable
);
2625 ovsrec_open_vswitch_set_statistics(cfg
, NULL
);
2630 run_system_stats(void)
2632 const struct ovsrec_open_vswitch
*cfg
= ovsrec_open_vswitch_first(idl
);
2635 stats
= system_stats_run();
2637 struct ovsdb_idl_txn
*txn
;
2638 struct ovsdb_datum datum
;
2640 txn
= ovsdb_idl_txn_create(idl
);
2641 ovsdb_datum_from_smap(&datum
, stats
);
2642 ovsdb_idl_txn_write(&cfg
->header_
, &ovsrec_open_vswitch_col_statistics
,
2644 ovsdb_idl_txn_commit(txn
);
2645 ovsdb_idl_txn_destroy(txn
);
2652 ofp12_controller_role_to_str(enum ofp12_controller_role role
)
2655 case OFPCR12_ROLE_EQUAL
:
2657 case OFPCR12_ROLE_MASTER
:
2659 case OFPCR12_ROLE_SLAVE
:
2661 case OFPCR12_ROLE_NOCHANGE
:
2663 return "*** INVALID ROLE ***";
2668 refresh_controller_status(void)
2672 const struct ovsrec_controller
*cfg
;
2676 /* Accumulate status for controllers on all bridges. */
2677 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2678 ofproto_get_ofproto_controller_info(br
->ofproto
, &info
);
2681 /* Update each controller in the database with current status. */
2682 OVSREC_CONTROLLER_FOR_EACH(cfg
, idl
) {
2683 struct ofproto_controller_info
*cinfo
=
2684 shash_find_data(&info
, cfg
->target
);
2687 ovsrec_controller_set_is_connected(cfg
, cinfo
->is_connected
);
2688 ovsrec_controller_set_role(cfg
, ofp12_controller_role_to_str(
2690 ovsrec_controller_set_status(cfg
, &cinfo
->pairs
);
2692 ovsrec_controller_set_is_connected(cfg
, false);
2693 ovsrec_controller_set_role(cfg
, NULL
);
2694 ovsrec_controller_set_status(cfg
, NULL
);
2698 ofproto_free_ofproto_controller_info(&info
);
2701 /* Update interface and mirror statistics if necessary. */
2703 run_stats_update(void)
2705 const struct ovsrec_open_vswitch
*cfg
= ovsrec_open_vswitch_first(idl
);
2712 /* Statistics update interval should always be greater than or equal to
2714 stats_interval
= MAX(smap_get_int(&cfg
->other_config
,
2715 "stats-update-interval",
2717 if (stats_timer_interval
!= stats_interval
) {
2718 stats_timer_interval
= stats_interval
;
2719 stats_timer
= LLONG_MIN
;
2722 if (time_msec() >= stats_timer
) {
2723 enum ovsdb_idl_txn_status status
;
2725 /* Rate limit the update. Do not start a new update if the
2726 * previous one is not done. */
2730 stats_txn
= ovsdb_idl_txn_create(idl
);
2731 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2735 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2736 struct iface
*iface
;
2738 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2739 iface_refresh_stats(iface
);
2741 port_refresh_stp_stats(port
);
2743 HMAP_FOR_EACH (m
, hmap_node
, &br
->mirrors
) {
2744 mirror_refresh_stats(m
);
2747 refresh_controller_status();
2750 status
= ovsdb_idl_txn_commit(stats_txn
);
2751 if (status
!= TXN_INCOMPLETE
) {
2752 stats_timer
= time_msec() + stats_timer_interval
;
2753 ovsdb_idl_txn_destroy(stats_txn
);
2760 stats_update_wait(void)
2762 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2763 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2765 ovsdb_idl_txn_wait(stats_txn
);
2767 poll_timer_wait_until(stats_timer
);
2771 /* Update bridge/port/interface status if necessary. */
2773 run_status_update(void)
2778 /* Rate limit the update. Do not start a new update if the
2779 * previous one is not done. */
2780 seq
= seq_read(connectivity_seq_get());
2781 if (seq
!= connectivity_seqno
|| status_txn_try_again
) {
2784 connectivity_seqno
= seq
;
2785 status_txn
= ovsdb_idl_txn_create(idl
);
2786 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2789 br_refresh_stp_status(br
);
2790 br_refresh_rstp_status(br
);
2791 br_refresh_datapath_info(br
);
2792 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2793 struct iface
*iface
;
2795 port_refresh_stp_status(port
);
2796 port_refresh_rstp_status(port
);
2797 port_refresh_bond_status(port
, status_txn_try_again
);
2798 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2799 iface_refresh_netdev_status(iface
);
2800 iface_refresh_ofproto_status(iface
);
2807 /* Commit the transaction and get the status. If the transaction finishes,
2808 * then destroy the transaction. Otherwise, keep it so that we can check
2809 * progress the next time that this function is called. */
2811 enum ovsdb_idl_txn_status status
;
2813 status
= ovsdb_idl_txn_commit(status_txn
);
2814 if (status
!= TXN_INCOMPLETE
) {
2815 ovsdb_idl_txn_destroy(status_txn
);
2818 /* Sets the 'status_txn_try_again' if the transaction fails. */
2819 if (status
== TXN_SUCCESS
|| status
== TXN_UNCHANGED
) {
2820 status_txn_try_again
= false;
2822 status_txn_try_again
= true;
2827 /* Refresh AA port status if necessary. */
2828 if (time_msec() >= aa_refresh_timer
) {
2831 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2832 if (bridge_aa_need_refresh(br
)) {
2833 struct ovsdb_idl_txn
*txn
;
2835 txn
= ovsdb_idl_txn_create(idl
);
2836 bridge_aa_refresh_queued(br
);
2837 ovsdb_idl_txn_commit(txn
);
2838 ovsdb_idl_txn_destroy(txn
);
2842 aa_refresh_timer
= time_msec() + AA_REFRESH_INTERVAL
;
2847 status_update_wait(void)
2849 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2850 * transaction to complete. If the status update to database needs to be
2851 * run again (transaction fails), registers a timeout in
2852 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2853 * sequence number. */
2855 ovsdb_idl_txn_wait(status_txn
);
2856 } else if (status_txn_try_again
) {
2857 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC
);
2859 seq_wait(connectivity_seq_get(), connectivity_seqno
);
2870 /* Let each datapath type do the work that it needs to do. */
2872 ofproto_enumerate_types(&types
);
2873 SSET_FOR_EACH (type
, &types
) {
2874 ofproto_type_run(type
);
2876 sset_destroy(&types
);
2878 /* Let each bridge do the work that it needs to do. */
2879 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2880 ofproto_run(br
->ofproto
);
2887 static struct ovsrec_open_vswitch null_cfg
;
2888 const struct ovsrec_open_vswitch
*cfg
;
2890 ovsrec_open_vswitch_init(&null_cfg
);
2896 if (ovsdb_idl_is_lock_contended(idl
)) {
2897 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2898 struct bridge
*br
, *next_br
;
2900 VLOG_ERR_RL(&rl
, "another ovs-vswitchd process is running, "
2901 "disabling this process (pid %ld) until it goes away",
2902 (long int) getpid());
2904 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
2905 bridge_destroy(br
, false);
2907 /* Since we will not be running system_stats_run() in this process
2908 * with the current situation of multiple ovs-vswitchd daemons,
2909 * disable system stats collection. */
2910 system_stats_enable(false);
2912 } else if (!ovsdb_idl_has_lock(idl
)
2913 || !ovsdb_idl_has_ever_connected(idl
)) {
2914 /* Returns if not holding the lock or not done retrieving db
2918 cfg
= ovsrec_open_vswitch_first(idl
);
2921 dpdk_init(&cfg
->other_config
);
2924 /* Initialize the ofproto library. This only needs to run once, but
2925 * it must be done after the configuration is set. If the
2926 * initialization has already occurred, bridge_init_ofproto()
2927 * returns immediately. */
2928 bridge_init_ofproto(cfg
);
2930 /* Once the value of flow-restore-wait is false, we no longer should
2931 * check its value from the database. */
2932 if (cfg
&& ofproto_get_flow_restore_wait()) {
2933 ofproto_set_flow_restore_wait(smap_get_bool(&cfg
->other_config
,
2934 "flow-restore-wait", false));
2939 /* Re-configure SSL. We do this on every trip through the main loop,
2940 * instead of just when the database changes, because the contents of the
2941 * key and certificate files can change without the database changing.
2943 * We do this before bridge_reconfigure() because that function might
2944 * initiate SSL connections and thus requires SSL to be configured. */
2945 if (cfg
&& cfg
->ssl
) {
2946 const struct ovsrec_ssl
*ssl
= cfg
->ssl
;
2948 stream_ssl_set_key_and_cert(ssl
->private_key
, ssl
->certificate
);
2949 stream_ssl_set_ca_cert_file(ssl
->ca_cert
, ssl
->bootstrap_ca_cert
);
2952 if (ovsdb_idl_get_seqno(idl
) != idl_seqno
|| ifaces_changed
) {
2953 struct ovsdb_idl_txn
*txn
;
2955 ifaces_changed
= false;
2957 idl_seqno
= ovsdb_idl_get_seqno(idl
);
2958 txn
= ovsdb_idl_txn_create(idl
);
2959 bridge_reconfigure(cfg
? cfg
: &null_cfg
);
2962 ovsrec_open_vswitch_set_cur_cfg(cfg
, cfg
->next_cfg
);
2963 discover_types(cfg
);
2966 /* If we are completing our initial configuration for this run
2967 * of ovs-vswitchd, then keep the transaction around to monitor
2968 * it for completion. */
2969 if (initial_config_done
) {
2970 /* Always sets the 'status_txn_try_again' to check again,
2971 * in case that this transaction fails. */
2972 status_txn_try_again
= true;
2973 ovsdb_idl_txn_commit(txn
);
2974 ovsdb_idl_txn_destroy(txn
);
2976 initial_config_done
= true;
2977 daemonize_txn
= txn
;
2981 if (daemonize_txn
) {
2982 enum ovsdb_idl_txn_status status
= ovsdb_idl_txn_commit(daemonize_txn
);
2983 if (status
!= TXN_INCOMPLETE
) {
2984 ovsdb_idl_txn_destroy(daemonize_txn
);
2985 daemonize_txn
= NULL
;
2987 /* ovs-vswitchd has completed initialization, so allow the
2988 * process that forked us to exit successfully. */
2989 daemonize_complete();
2991 vlog_enable_async();
2993 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name
, VERSION
);
2998 run_status_update();
3008 ovsdb_idl_wait(idl
);
3009 if (daemonize_txn
) {
3010 ovsdb_idl_txn_wait(daemonize_txn
);
3014 if (ifaces_changed
) {
3015 poll_immediate_wake();
3019 ofproto_enumerate_types(&types
);
3020 SSET_FOR_EACH (type
, &types
) {
3021 ofproto_type_wait(type
);
3023 sset_destroy(&types
);
3025 if (!hmap_is_empty(&all_bridges
)) {
3028 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3029 ofproto_wait(br
->ofproto
);
3031 stats_update_wait();
3032 status_update_wait();
3035 system_stats_wait();
3038 /* Adds some memory usage statistics for bridges into 'usage', for use with
3039 * memory_report(). */
3041 bridge_get_memory_usage(struct simap
*usage
)
3048 ofproto_enumerate_types(&types
);
3049 SSET_FOR_EACH (type
, &types
) {
3050 ofproto_type_get_memory_usage(type
, usage
);
3052 sset_destroy(&types
);
3054 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3055 ofproto_get_memory_usage(br
->ofproto
, usage
);
3059 /* QoS unixctl user interface functions. */
3061 struct qos_unixctl_show_cbdata
{
3063 struct iface
*iface
;
3067 qos_unixctl_show_queue(unsigned int queue_id
,
3068 const struct smap
*details
,
3069 struct iface
*iface
,
3072 struct netdev_queue_stats stats
;
3073 struct smap_node
*node
;
3076 ds_put_cstr(ds
, "\n");
3078 ds_put_format(ds
, "Queue %u:\n", queue_id
);
3080 ds_put_cstr(ds
, "Default:\n");
3083 SMAP_FOR_EACH (node
, details
) {
3084 ds_put_format(ds
, "\t%s: %s\n", node
->key
, node
->value
);
3087 error
= netdev_get_queue_stats(iface
->netdev
, queue_id
, &stats
);
3089 if (stats
.tx_packets
!= UINT64_MAX
) {
3090 ds_put_format(ds
, "\ttx_packets: %"PRIu64
"\n", stats
.tx_packets
);
3093 if (stats
.tx_bytes
!= UINT64_MAX
) {
3094 ds_put_format(ds
, "\ttx_bytes: %"PRIu64
"\n", stats
.tx_bytes
);
3097 if (stats
.tx_errors
!= UINT64_MAX
) {
3098 ds_put_format(ds
, "\ttx_errors: %"PRIu64
"\n", stats
.tx_errors
);
3101 ds_put_format(ds
, "\tFailed to get statistics for queue %u: %s",
3102 queue_id
, ovs_strerror(error
));
3107 qos_unixctl_show_types(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
3108 const char *argv
[], void *aux OVS_UNUSED
)
3110 struct ds ds
= DS_EMPTY_INITIALIZER
;
3111 struct sset types
= SSET_INITIALIZER(&types
);
3112 struct iface
*iface
;
3113 const char * types_name
;
3116 iface
= iface_find(argv
[1]);
3118 unixctl_command_reply_error(conn
, "no such interface");
3122 error
= netdev_get_qos_types(iface
->netdev
, &types
);
3124 if (!sset_is_empty(&types
)) {
3125 SSET_FOR_EACH (types_name
, &types
) {
3126 ds_put_format(&ds
, "QoS type: %s\n", types_name
);
3128 unixctl_command_reply(conn
, ds_cstr(&ds
));
3130 ds_put_format(&ds
, "No QoS types supported for interface: %s\n",
3132 unixctl_command_reply(conn
, ds_cstr(&ds
));
3135 ds_put_format(&ds
, "%s: failed to retrieve supported QoS types (%s)",
3136 iface
->name
, ovs_strerror(error
));
3137 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
3140 sset_destroy(&types
);
3145 qos_unixctl_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
3146 const char *argv
[], void *aux OVS_UNUSED
)
3148 struct ds ds
= DS_EMPTY_INITIALIZER
;
3149 struct smap smap
= SMAP_INITIALIZER(&smap
);
3150 struct iface
*iface
;
3152 struct smap_node
*node
;
3155 iface
= iface_find(argv
[1]);
3157 unixctl_command_reply_error(conn
, "no such interface");
3161 error
= netdev_get_qos(iface
->netdev
, &type
, &smap
);
3163 if (*type
!= '\0') {
3164 struct netdev_queue_dump dump
;
3165 struct smap details
;
3166 unsigned int queue_id
;
3168 ds_put_format(&ds
, "QoS: %s %s\n", iface
->name
, type
);
3170 SMAP_FOR_EACH (node
, &smap
) {
3171 ds_put_format(&ds
, "%s: %s\n", node
->key
, node
->value
);
3174 smap_init(&details
);
3175 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
3176 qos_unixctl_show_queue(queue_id
, &details
, iface
, &ds
);
3178 smap_destroy(&details
);
3180 unixctl_command_reply(conn
, ds_cstr(&ds
));
3182 ds_put_format(&ds
, "QoS not configured on %s\n", iface
->name
);
3183 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
3186 ds_put_format(&ds
, "%s: failed to retrieve QOS configuration (%s)\n",
3187 iface
->name
, ovs_strerror(error
));
3188 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
3191 smap_destroy(&smap
);
3195 /* Bridge reconfiguration functions. */
3197 bridge_create(const struct ovsrec_bridge
*br_cfg
)
3201 ovs_assert(!bridge_lookup(br_cfg
->name
));
3202 br
= xzalloc(sizeof *br
);
3204 br
->name
= xstrdup(br_cfg
->name
);
3205 br
->type
= xstrdup(ofproto_normalize_type(br_cfg
->datapath_type
));
3208 /* Derive the default Ethernet address from the bridge's UUID. This should
3209 * be unique and it will be stable between ovs-vswitchd runs. */
3210 memcpy(&br
->default_ea
, &br_cfg
->header_
.uuid
, ETH_ADDR_LEN
);
3211 eth_addr_mark_random(&br
->default_ea
);
3213 hmap_init(&br
->ports
);
3214 hmap_init(&br
->ifaces
);
3215 hmap_init(&br
->iface_by_name
);
3216 hmap_init(&br
->mirrors
);
3218 hmap_init(&br
->mappings
);
3219 hmap_insert(&all_bridges
, &br
->node
, hash_string(br
->name
, 0));
3223 bridge_destroy(struct bridge
*br
, bool del
)
3226 struct mirror
*mirror
, *next_mirror
;
3227 struct port
*port
, *next_port
;
3229 HMAP_FOR_EACH_SAFE (port
, next_port
, hmap_node
, &br
->ports
) {
3232 HMAP_FOR_EACH_SAFE (mirror
, next_mirror
, hmap_node
, &br
->mirrors
) {
3233 mirror_destroy(mirror
);
3236 hmap_remove(&all_bridges
, &br
->node
);
3237 ofproto_destroy(br
->ofproto
, del
);
3238 hmap_destroy(&br
->ifaces
);
3239 hmap_destroy(&br
->ports
);
3240 hmap_destroy(&br
->iface_by_name
);
3241 hmap_destroy(&br
->mirrors
);
3242 hmap_destroy(&br
->mappings
);
3249 static struct bridge
*
3250 bridge_lookup(const char *name
)
3254 HMAP_FOR_EACH_WITH_HASH (br
, node
, hash_string(name
, 0), &all_bridges
) {
3255 if (!strcmp(br
->name
, name
)) {
3262 /* Handle requests for a listing of all flows known by the OpenFlow
3263 * stack, including those normally hidden. */
3265 bridge_unixctl_dump_flows(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
3266 const char *argv
[], void *aux OVS_UNUSED
)
3271 br
= bridge_lookup(argv
[1]);
3273 unixctl_command_reply_error(conn
, "Unknown bridge");
3278 ofproto_get_all_flows(br
->ofproto
, &results
);
3280 unixctl_command_reply(conn
, ds_cstr(&results
));
3281 ds_destroy(&results
);
3284 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3285 * connections and reconnect. If BRIDGE is not specified, then all bridges
3286 * drop their controller connections and reconnect. */
3288 bridge_unixctl_reconnect(struct unixctl_conn
*conn
, int argc
,
3289 const char *argv
[], void *aux OVS_UNUSED
)
3293 br
= bridge_lookup(argv
[1]);
3295 unixctl_command_reply_error(conn
, "Unknown bridge");
3298 ofproto_reconnect_controllers(br
->ofproto
);
3300 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3301 ofproto_reconnect_controllers(br
->ofproto
);
3304 unixctl_command_reply(conn
, NULL
);
3308 bridge_get_controllers(const struct bridge
*br
,
3309 struct ovsrec_controller
***controllersp
)
3311 struct ovsrec_controller
**controllers
;
3312 size_t n_controllers
;
3314 controllers
= br
->cfg
->controller
;
3315 n_controllers
= br
->cfg
->n_controller
;
3317 if (n_controllers
== 1 && !strcmp(controllers
[0]->target
, "none")) {
3323 *controllersp
= controllers
;
3325 return n_controllers
;
3329 bridge_collect_wanted_ports(struct bridge
*br
,
3330 struct shash
*wanted_ports
)
3334 shash_init(wanted_ports
);
3336 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
3337 const char *name
= br
->cfg
->ports
[i
]->name
;
3338 if (!shash_add_once(wanted_ports
, name
, br
->cfg
->ports
[i
])) {
3339 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3343 if (bridge_get_controllers(br
, NULL
)
3344 && !shash_find(wanted_ports
, br
->name
)) {
3345 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3346 br
->name
, br
->name
);
3348 ovsrec_interface_init(&br
->synth_local_iface
);
3349 ovsrec_port_init(&br
->synth_local_port
);
3351 br
->synth_local_port
.interfaces
= &br
->synth_local_ifacep
;
3352 br
->synth_local_port
.n_interfaces
= 1;
3353 br
->synth_local_port
.name
= br
->name
;
3355 br
->synth_local_iface
.name
= br
->name
;
3356 br
->synth_local_iface
.type
= "internal";
3358 br
->synth_local_ifacep
= &br
->synth_local_iface
;
3360 shash_add(wanted_ports
, br
->name
, &br
->synth_local_port
);
3364 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3365 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3366 * 'br' needs to complete its configuration. */
3368 bridge_del_ports(struct bridge
*br
, const struct shash
*wanted_ports
)
3370 struct shash_node
*port_node
;
3371 struct port
*port
, *next
;
3373 /* Get rid of deleted ports.
3374 * Get rid of deleted interfaces on ports that still exist. */
3375 HMAP_FOR_EACH_SAFE (port
, next
, hmap_node
, &br
->ports
) {
3376 port
->cfg
= shash_find_data(wanted_ports
, port
->name
);
3380 port_del_ifaces(port
);
3384 /* Update iface->cfg and iface->type in interfaces that still exist. */
3385 SHASH_FOR_EACH (port_node
, wanted_ports
) {
3386 const struct ovsrec_port
*port
= port_node
->data
;
3389 for (i
= 0; i
< port
->n_interfaces
; i
++) {
3390 const struct ovsrec_interface
*cfg
= port
->interfaces
[i
];
3391 struct iface
*iface
= iface_lookup(br
, cfg
->name
);
3392 const char *type
= iface_get_type(cfg
, br
->cfg
);
3397 } else if (!strcmp(type
, "null")) {
3398 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3399 " may be removed in February 2013. Please email"
3400 " dev@openvswitch.org with concerns.",
3403 /* We will add new interfaces later. */
3409 /* Initializes 'oc' appropriately as a management service controller for
3412 * The caller must free oc->target when it is no longer needed. */
3414 bridge_ofproto_controller_for_mgmt(const struct bridge
*br
,
3415 struct ofproto_controller
*oc
)
3417 oc
->target
= xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br
->name
);
3418 oc
->max_backoff
= 0;
3419 oc
->probe_interval
= 60;
3420 oc
->band
= OFPROTO_OUT_OF_BAND
;
3422 oc
->burst_limit
= 0;
3423 oc
->enable_async_msgs
= true;
3427 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3429 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller
*c
,
3430 struct ofproto_controller
*oc
)
3434 oc
->target
= c
->target
;
3435 oc
->max_backoff
= c
->max_backoff
? *c
->max_backoff
/ 1000 : 8;
3436 oc
->probe_interval
= c
->inactivity_probe
? *c
->inactivity_probe
/ 1000 : 5;
3437 oc
->band
= (!c
->connection_mode
|| !strcmp(c
->connection_mode
, "in-band")
3438 ? OFPROTO_IN_BAND
: OFPROTO_OUT_OF_BAND
);
3439 oc
->rate_limit
= c
->controller_rate_limit
? *c
->controller_rate_limit
: 0;
3440 oc
->burst_limit
= (c
->controller_burst_limit
3441 ? *c
->controller_burst_limit
: 0);
3442 oc
->enable_async_msgs
= (!c
->enable_async_messages
3443 || *c
->enable_async_messages
);
3444 dscp
= smap_get_int(&c
->other_config
, "dscp", DSCP_DEFAULT
);
3445 if (dscp
< 0 || dscp
> 63) {
3446 dscp
= DSCP_DEFAULT
;
3451 /* Configures the IP stack for 'br''s local interface properly according to the
3452 * configuration in 'c'. */
3454 bridge_configure_local_iface_netdev(struct bridge
*br
,
3455 struct ovsrec_controller
*c
)
3457 struct netdev
*netdev
;
3458 struct in_addr mask
, gateway
;
3460 struct iface
*local_iface
;
3463 /* If there's no local interface or no IP address, give up. */
3464 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
3465 if (!local_iface
|| !c
->local_ip
|| !ip_parse(c
->local_ip
, &ip
.s_addr
)) {
3469 /* Bring up the local interface. */
3470 netdev
= local_iface
->netdev
;
3471 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
3473 /* Configure the IP address and netmask. */
3474 if (!c
->local_netmask
3475 || !ip_parse(c
->local_netmask
, &mask
.s_addr
)
3477 mask
.s_addr
= guess_netmask(ip
.s_addr
);
3479 if (!netdev_set_in4(netdev
, ip
, mask
)) {
3480 VLOG_INFO("bridge %s: configured IP address "IP_FMT
", netmask "IP_FMT
,
3481 br
->name
, IP_ARGS(ip
.s_addr
), IP_ARGS(mask
.s_addr
));
3484 /* Configure the default gateway. */
3485 if (c
->local_gateway
3486 && ip_parse(c
->local_gateway
, &gateway
.s_addr
)
3487 && gateway
.s_addr
) {
3488 if (!netdev_add_router(netdev
, gateway
)) {
3489 VLOG_INFO("bridge %s: configured gateway "IP_FMT
,
3490 br
->name
, IP_ARGS(gateway
.s_addr
));
3495 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3496 * in either string are treated as equal to any number of slashes in the other,
3497 * e.g. "x///y" is equal to "x/y".
3499 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3500 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3501 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3502 * 'b' against a prefix of 'a'.
3505 equal_pathnames(const char *a
, const char *b
, size_t b_stoplen
)
3507 const char *b_start
= b
;
3509 if (b
- b_start
>= b_stoplen
) {
3511 } else if (*a
!= *b
) {
3513 } else if (*a
== '/') {
3514 a
+= strspn(a
, "/");
3515 b
+= strspn(b
, "/");
3516 } else if (*a
== '\0') {
3526 bridge_configure_remotes(struct bridge
*br
,
3527 const struct sockaddr_in
*managers
, size_t n_managers
)
3529 bool disable_in_band
;
3531 struct ovsrec_controller
**controllers
;
3532 size_t n_controllers
;
3534 enum ofproto_fail_mode fail_mode
;
3536 struct ofproto_controller
*ocs
;
3540 /* Check if we should disable in-band control on this bridge. */
3541 disable_in_band
= smap_get_bool(&br
->cfg
->other_config
, "disable-in-band",
3544 /* Set OpenFlow queue ID for in-band control. */
3545 ofproto_set_in_band_queue(br
->ofproto
,
3546 smap_get_int(&br
->cfg
->other_config
,
3547 "in-band-queue", -1));
3549 if (disable_in_band
) {
3550 ofproto_set_extra_in_band_remotes(br
->ofproto
, NULL
, 0);
3552 ofproto_set_extra_in_band_remotes(br
->ofproto
, managers
, n_managers
);
3555 n_controllers
= bridge_get_controllers(br
, &controllers
);
3557 ocs
= xmalloc((n_controllers
+ 1) * sizeof *ocs
);
3560 bridge_ofproto_controller_for_mgmt(br
, &ocs
[n_ocs
++]);
3561 for (i
= 0; i
< n_controllers
; i
++) {
3562 struct ovsrec_controller
*c
= controllers
[i
];
3564 if (!strncmp(c
->target
, "punix:", 6)
3565 || !strncmp(c
->target
, "unix:", 5)) {
3566 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3569 if (!strncmp(c
->target
, "unix:", 5)) {
3570 /* Connect to a listening socket */
3571 whitelist
= xasprintf("unix:%s/", ovs_rundir());
3572 if (strchr(c
->target
, '/') &&
3573 !equal_pathnames(c
->target
, whitelist
,
3574 strlen(whitelist
))) {
3575 /* Absolute path specified, but not in ovs_rundir */
3576 VLOG_ERR_RL(&rl
, "bridge %s: Not connecting to socket "
3577 "controller \"%s\" due to possibility for "
3578 "remote exploit. Instead, specify socket "
3579 "in whitelisted \"%s\" or connect to "
3580 "\"unix:%s/%s.mgmt\" (which is always "
3581 "available without special configuration).",
3582 br
->name
, c
->target
, whitelist
,
3583 ovs_rundir(), br
->name
);
3588 whitelist
= xasprintf("punix:%s/%s.",
3589 ovs_rundir(), br
->name
);
3590 if (!equal_pathnames(c
->target
, whitelist
, strlen(whitelist
))
3591 || strchr(c
->target
+ strlen(whitelist
), '/')) {
3592 /* Prevent remote ovsdb-server users from accessing
3593 * arbitrary Unix domain sockets and overwriting arbitrary
3595 VLOG_ERR_RL(&rl
, "bridge %s: Not adding Unix domain socket "
3596 "controller \"%s\" due to possibility of "
3597 "overwriting local files. Instead, specify "
3598 "path in whitelisted format \"%s*\" or "
3599 "connect to \"unix:%s/%s.mgmt\" (which is "
3600 "always available without special "
3602 br
->name
, c
->target
, whitelist
,
3603 ovs_rundir(), br
->name
);
3612 bridge_configure_local_iface_netdev(br
, c
);
3613 bridge_ofproto_controller_from_ovsrec(c
, &ocs
[n_ocs
]);
3614 if (disable_in_band
) {
3615 ocs
[n_ocs
].band
= OFPROTO_OUT_OF_BAND
;
3620 ofproto_set_controllers(br
->ofproto
, ocs
, n_ocs
,
3621 bridge_get_allowed_versions(br
));
3622 free(ocs
[0].target
); /* From bridge_ofproto_controller_for_mgmt(). */
3625 /* Set the fail-mode. */
3626 fail_mode
= !br
->cfg
->fail_mode
3627 || !strcmp(br
->cfg
->fail_mode
, "standalone")
3628 ? OFPROTO_FAIL_STANDALONE
3629 : OFPROTO_FAIL_SECURE
;
3630 ofproto_set_fail_mode(br
->ofproto
, fail_mode
);
3632 /* Configure OpenFlow controller connection snooping. */
3633 if (!ofproto_has_snoops(br
->ofproto
)) {
3637 sset_add_and_free(&snoops
, xasprintf("punix:%s/%s.snoop",
3638 ovs_rundir(), br
->name
));
3639 ofproto_set_snoops(br
->ofproto
, &snoops
);
3640 sset_destroy(&snoops
);
3645 bridge_configure_tables(struct bridge
*br
)
3647 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3651 n_tables
= ofproto_get_n_tables(br
->ofproto
);
3653 for (i
= 0; i
< n_tables
; i
++) {
3654 struct ofproto_table_settings s
;
3655 bool use_default_prefixes
= true;
3658 s
.max_flows
= UINT_MAX
;
3660 s
.enable_eviction
= false;
3662 s
.n_prefix_fields
= 0;
3663 memset(s
.prefix_fields
, ~0, sizeof(s
.prefix_fields
));
3665 if (j
< br
->cfg
->n_flow_tables
&& i
== br
->cfg
->key_flow_tables
[j
]) {
3666 struct ovsrec_flow_table
*cfg
= br
->cfg
->value_flow_tables
[j
++];
3669 if (cfg
->n_flow_limit
&& *cfg
->flow_limit
< UINT_MAX
) {
3670 s
.max_flows
= *cfg
->flow_limit
;
3673 s
.enable_eviction
= (cfg
->overflow_policy
3674 && !strcmp(cfg
->overflow_policy
, "evict"));
3675 if (cfg
->n_groups
) {
3676 s
.groups
= xmalloc(cfg
->n_groups
* sizeof *s
.groups
);
3677 for (k
= 0; k
< cfg
->n_groups
; k
++) {
3678 const char *string
= cfg
->groups
[k
];
3681 msg
= mf_parse_subfield__(&s
.groups
[k
], &string
);
3683 VLOG_WARN_RL(&rl
, "bridge %s table %d: error parsing "
3684 "'groups' (%s)", br
->name
, i
, msg
);
3686 } else if (*string
) {
3687 VLOG_WARN_RL(&rl
, "bridge %s table %d: 'groups' "
3688 "element '%s' contains trailing garbage",
3689 br
->name
, i
, cfg
->groups
[k
]);
3696 /* Prefix lookup fields. */
3697 s
.n_prefix_fields
= 0;
3698 for (k
= 0; k
< cfg
->n_prefixes
; k
++) {
3699 const char *name
= cfg
->prefixes
[k
];
3700 const struct mf_field
*mf
;
3702 if (strcmp(name
, "none") == 0) {
3703 use_default_prefixes
= false;
3704 s
.n_prefix_fields
= 0;
3707 mf
= mf_from_name(name
);
3709 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3713 if (mf
->flow_be32ofs
< 0 || mf
->n_bits
% 32) {
3714 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3715 "%s", br
->name
, name
);
3718 if (s
.n_prefix_fields
>= ARRAY_SIZE(s
.prefix_fields
)) {
3719 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3720 "field not used: %s", br
->name
, name
);
3723 use_default_prefixes
= false;
3724 s
.prefix_fields
[s
.n_prefix_fields
++] = mf
->id
;
3727 if (use_default_prefixes
) {
3728 /* Use default values. */
3729 s
.n_prefix_fields
= ARRAY_SIZE(default_prefix_fields
);
3730 memcpy(s
.prefix_fields
, default_prefix_fields
,
3731 sizeof default_prefix_fields
);
3734 struct ds ds
= DS_EMPTY_INITIALIZER
;
3735 for (k
= 0; k
< s
.n_prefix_fields
; k
++) {
3737 ds_put_char(&ds
, ',');
3739 ds_put_cstr(&ds
, mf_from_id(s
.prefix_fields
[k
])->name
);
3741 if (s
.n_prefix_fields
== 0) {
3742 ds_put_cstr(&ds
, "none");
3744 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3745 br
->name
, i
, ds_cstr(&ds
));
3749 ofproto_configure_table(br
->ofproto
, i
, &s
);
3753 for (; j
< br
->cfg
->n_flow_tables
; j
++) {
3754 VLOG_WARN_RL(&rl
, "bridge %s: ignoring configuration for flow table "
3755 "%"PRId64
" not supported by this datapath", br
->name
,
3756 br
->cfg
->key_flow_tables
[j
]);
3761 bridge_configure_dp_desc(struct bridge
*br
)
3763 ofproto_set_dp_desc(br
->ofproto
,
3764 smap_get(&br
->cfg
->other_config
, "dp-desc"));
3767 static struct aa_mapping
*
3768 bridge_aa_mapping_find(struct bridge
*br
, const int64_t isid
)
3770 struct aa_mapping
*m
;
3772 HMAP_FOR_EACH_IN_BUCKET (m
,
3774 hash_bytes(&isid
, sizeof isid
, 0),
3776 if (isid
== m
->isid
) {
3783 static struct aa_mapping
*
3784 bridge_aa_mapping_create(struct bridge
*br
,
3788 struct aa_mapping
*m
;
3790 m
= xzalloc(sizeof *m
);
3794 m
->br_name
= xstrdup(br
->name
);
3795 hmap_insert(&br
->mappings
,
3797 hash_bytes(&isid
, sizeof isid
, 0));
3803 bridge_aa_mapping_destroy(struct aa_mapping
*m
)
3806 struct bridge
*br
= m
->bridge
;
3809 ofproto_aa_mapping_unregister(br
->ofproto
, m
);
3812 hmap_remove(&br
->mappings
, &m
->hmap_node
);
3821 bridge_aa_mapping_configure(struct aa_mapping
*m
)
3823 struct aa_mapping_settings s
;
3829 ofproto_aa_mapping_register(m
->bridge
->ofproto
, m
, &s
);
3835 bridge_configure_aa(struct bridge
*br
)
3837 const struct ovsdb_datum
*mc
;
3838 struct ovsrec_autoattach
*auto_attach
= br
->cfg
->auto_attach
;
3839 struct aa_settings aa_s
;
3840 struct aa_mapping
*m
, *next
;
3844 ofproto_set_aa(br
->ofproto
, NULL
, NULL
);
3848 memset(&aa_s
, 0, sizeof aa_s
);
3849 aa_s
.system_description
= auto_attach
->system_description
;
3850 aa_s
.system_name
= auto_attach
->system_name
;
3851 ofproto_set_aa(br
->ofproto
, NULL
, &aa_s
);
3853 mc
= ovsrec_autoattach_get_mappings(auto_attach
,
3855 OVSDB_TYPE_INTEGER
);
3856 HMAP_FOR_EACH_SAFE (m
, next
, hmap_node
, &br
->mappings
) {
3857 union ovsdb_atom atom
;
3859 atom
.integer
= m
->isid
;
3860 if (ovsdb_datum_find_key(mc
, &atom
, OVSDB_TYPE_UUID
) == UINT_MAX
) {
3861 VLOG_INFO("Deleting isid=%"PRIu32
", vlan=%"PRIu16
,
3863 bridge_aa_mapping_destroy(m
);
3867 /* Add new mappings and reconfigure existing ones. */
3868 for (i
= 0; i
< auto_attach
->n_mappings
; ++i
) {
3869 struct aa_mapping
*m
=
3870 bridge_aa_mapping_find(br
, auto_attach
->key_mappings
[i
]);
3873 VLOG_INFO("Adding isid=%"PRId64
", vlan=%"PRId64
,
3874 auto_attach
->key_mappings
[i
],
3875 auto_attach
->value_mappings
[i
]);
3876 m
= bridge_aa_mapping_create(br
,
3877 auto_attach
->key_mappings
[i
],
3878 auto_attach
->value_mappings
[i
]);
3880 if (!bridge_aa_mapping_configure(m
)) {
3881 bridge_aa_mapping_destroy(m
);
3888 bridge_aa_need_refresh(struct bridge
*br
)
3890 return ofproto_aa_vlan_get_queue_size(br
->ofproto
) > 0;
3894 bridge_aa_update_trunks(struct port
*port
, struct bridge_aa_vlan
*m
)
3896 int64_t *trunks
= NULL
;
3898 bool found
= false, reconfigure
= false;
3900 for (i
= 0; i
< port
->cfg
->n_trunks
; i
++) {
3901 if (port
->cfg
->trunks
[i
] == m
->vlan
) {
3908 case BRIDGE_AA_VLAN_OPER_ADD
:
3910 trunks
= xmalloc(sizeof *trunks
* (port
->cfg
->n_trunks
+ 1));
3912 for (i
= 0; i
< port
->cfg
->n_trunks
; i
++) {
3913 trunks
[i
] = port
->cfg
->trunks
[i
];
3915 trunks
[i
++] = m
->vlan
;
3921 case BRIDGE_AA_VLAN_OPER_REMOVE
:
3925 trunks
= xmalloc(sizeof *trunks
* (port
->cfg
->n_trunks
- 1));
3927 for (i
= 0; i
< port
->cfg
->n_trunks
; i
++) {
3928 if (port
->cfg
->trunks
[i
] != m
->vlan
) {
3929 trunks
[j
++] = port
->cfg
->trunks
[i
];
3938 case BRIDGE_AA_VLAN_OPER_UNDEF
:
3940 VLOG_WARN("unrecognized operation %u", m
->oper
);
3945 /* VLAN switching under trunk mode cause the trunk port to switch all
3946 * VLANs, see ovs-vswitchd.conf.db
3949 static char *vlan_mode_access
= "access";
3950 ovsrec_port_set_vlan_mode(port
->cfg
, vlan_mode_access
);
3954 static char *vlan_mode_trunk
= "trunk";
3955 ovsrec_port_set_vlan_mode(port
->cfg
, vlan_mode_trunk
);
3958 ovsrec_port_set_trunks(port
->cfg
, trunks
, i
);
3960 /* Force reconfigure of the port. */
3961 port_configure(port
);
3966 bridge_aa_refresh_queued(struct bridge
*br
)
3968 struct ovs_list
*list
= xmalloc(sizeof *list
);
3969 struct bridge_aa_vlan
*node
, *next
;
3971 ovs_list_init(list
);
3972 ofproto_aa_vlan_get_queued(br
->ofproto
, list
);
3974 LIST_FOR_EACH_SAFE (node
, next
, list_node
, list
) {
3977 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node
->port_name
, node
->vlan
,
3980 port
= port_lookup(br
, node
->port_name
);
3982 bridge_aa_update_trunks(port
, node
);
3985 ovs_list_remove(&node
->list_node
);
3986 free(node
->port_name
);
3994 /* Port functions. */
3996 static struct port
*
3997 port_create(struct bridge
*br
, const struct ovsrec_port
*cfg
)
4001 port
= xzalloc(sizeof *port
);
4003 port
->name
= xstrdup(cfg
->name
);
4005 ovs_list_init(&port
->ifaces
);
4007 hmap_insert(&br
->ports
, &port
->hmap_node
, hash_string(port
->name
, 0));
4011 /* Deletes interfaces from 'port' that are no longer configured for it. */
4013 port_del_ifaces(struct port
*port
)
4015 struct iface
*iface
, *next
;
4016 struct sset new_ifaces
;
4019 /* Collect list of new interfaces. */
4020 sset_init(&new_ifaces
);
4021 for (i
= 0; i
< port
->cfg
->n_interfaces
; i
++) {
4022 const char *name
= port
->cfg
->interfaces
[i
]->name
;
4023 const char *type
= port
->cfg
->interfaces
[i
]->type
;
4024 if (strcmp(type
, "null")) {
4025 sset_add(&new_ifaces
, name
);
4029 /* Get rid of deleted interfaces. */
4030 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
4031 if (!sset_contains(&new_ifaces
, iface
->name
)) {
4032 iface_destroy(iface
);
4036 sset_destroy(&new_ifaces
);
4040 port_destroy(struct port
*port
)
4043 struct bridge
*br
= port
->bridge
;
4044 struct iface
*iface
, *next
;
4047 ofproto_bundle_unregister(br
->ofproto
, port
);
4050 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
4051 iface_destroy__(iface
);
4054 hmap_remove(&br
->ports
, &port
->hmap_node
);
4060 static struct port
*
4061 port_lookup(const struct bridge
*br
, const char *name
)
4065 HMAP_FOR_EACH_WITH_HASH (port
, hmap_node
, hash_string(name
, 0),
4067 if (!strcmp(port
->name
, name
)) {
4075 enable_lacp(struct port
*port
, bool *activep
)
4077 if (!port
->cfg
->lacp
) {
4078 /* XXX when LACP implementation has been sufficiently tested, enable by
4079 * default and make active on bonded ports. */
4081 } else if (!strcmp(port
->cfg
->lacp
, "off")) {
4083 } else if (!strcmp(port
->cfg
->lacp
, "active")) {
4086 } else if (!strcmp(port
->cfg
->lacp
, "passive")) {
4090 VLOG_WARN("port %s: unknown LACP mode %s",
4091 port
->name
, port
->cfg
->lacp
);
4096 static struct lacp_settings
*
4097 port_configure_lacp(struct port
*port
, struct lacp_settings
*s
)
4099 const char *lacp_time
, *system_id
;
4102 if (!enable_lacp(port
, &s
->active
)) {
4106 s
->name
= port
->name
;
4108 system_id
= smap_get(&port
->cfg
->other_config
, "lacp-system-id");
4110 if (!ovs_scan(system_id
, ETH_ADDR_SCAN_FMT
,
4111 ETH_ADDR_SCAN_ARGS(s
->id
))) {
4112 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4113 " address.", port
->name
, system_id
);
4117 s
->id
= port
->bridge
->ea
;
4120 if (eth_addr_is_zero(s
->id
)) {
4121 VLOG_WARN("port %s: Invalid zero LACP system ID.", port
->name
);
4125 /* Prefer bondable links if unspecified. */
4126 priority
= smap_get_int(&port
->cfg
->other_config
, "lacp-system-priority",
4128 s
->priority
= (priority
> 0 && priority
<= UINT16_MAX
4130 : UINT16_MAX
- !ovs_list_is_short(&port
->ifaces
));
4132 lacp_time
= smap_get(&port
->cfg
->other_config
, "lacp-time");
4133 s
->fast
= lacp_time
&& !strcasecmp(lacp_time
, "fast");
4135 s
->fallback_ab_cfg
= smap_get_bool(&port
->cfg
->other_config
,
4136 "lacp-fallback-ab", false);
4142 iface_configure_lacp(struct iface
*iface
, struct lacp_slave_settings
*s
)
4144 int priority
, portid
, key
;
4146 portid
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-id", 0);
4147 priority
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-priority",
4149 key
= smap_get_int(&iface
->cfg
->other_config
, "lacp-aggregation-key", 0);
4151 if (portid
<= 0 || portid
> UINT16_MAX
) {
4152 portid
= ofp_to_u16(iface
->ofp_port
);
4155 if (priority
<= 0 || priority
> UINT16_MAX
) {
4156 priority
= UINT16_MAX
;
4159 if (key
< 0 || key
> UINT16_MAX
) {
4163 s
->name
= iface
->name
;
4165 s
->priority
= priority
;
4170 port_configure_bond(struct port
*port
, struct bond_settings
*s
)
4172 const char *detect_s
;
4173 struct iface
*iface
;
4175 int miimon_interval
;
4177 s
->name
= port
->name
;
4179 if (port
->cfg
->bond_mode
) {
4180 if (!bond_mode_from_string(&s
->balance
, port
->cfg
->bond_mode
)) {
4181 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4182 port
->name
, port
->cfg
->bond_mode
,
4183 bond_mode_to_string(s
->balance
));
4186 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
4188 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4189 * active-backup. At some point we should remove this warning. */
4190 VLOG_WARN_RL(&rl
, "port %s: Using the default bond_mode %s. Note that"
4191 " in previous versions, the default bond_mode was"
4192 " balance-slb", port
->name
,
4193 bond_mode_to_string(s
->balance
));
4195 if (s
->balance
== BM_SLB
&& port
->bridge
->cfg
->n_flood_vlans
) {
4196 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4197 "please use another bond type or disable flood_vlans",
4201 miimon_interval
= smap_get_int(&port
->cfg
->other_config
,
4202 "bond-miimon-interval", 0);
4203 if (miimon_interval
<= 0) {
4204 miimon_interval
= 200;
4207 detect_s
= smap_get(&port
->cfg
->other_config
, "bond-detect-mode");
4208 if (!detect_s
|| !strcmp(detect_s
, "carrier")) {
4209 miimon_interval
= 0;
4210 } else if (strcmp(detect_s
, "miimon")) {
4211 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4212 "defaulting to carrier", port
->name
, detect_s
);
4213 miimon_interval
= 0;
4216 s
->up_delay
= MAX(0, port
->cfg
->bond_updelay
);
4217 s
->down_delay
= MAX(0, port
->cfg
->bond_downdelay
);
4218 s
->basis
= smap_get_int(&port
->cfg
->other_config
, "bond-hash-basis", 0);
4219 s
->rebalance_interval
= smap_get_int(&port
->cfg
->other_config
,
4220 "bond-rebalance-interval", 10000);
4221 if (s
->rebalance_interval
&& s
->rebalance_interval
< 1000) {
4222 s
->rebalance_interval
= 1000;
4225 s
->lacp_fallback_ab_cfg
= smap_get_bool(&port
->cfg
->other_config
,
4226 "lacp-fallback-ab", false);
4228 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
4229 netdev_set_miimon_interval(iface
->netdev
, miimon_interval
);
4232 mac_s
= port
->cfg
->bond_active_slave
;
4233 if (!mac_s
|| !ovs_scan(mac_s
, ETH_ADDR_SCAN_FMT
,
4234 ETH_ADDR_SCAN_ARGS(s
->active_slave_mac
))) {
4235 /* OVSDB did not store the last active interface */
4236 s
->active_slave_mac
= eth_addr_zero
;
4240 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4241 * instead of obtaining it from the database. */
4243 port_is_synthetic(const struct port
*port
)
4245 return ovsdb_idl_row_is_synthetic(&port
->cfg
->header_
);
4248 /* Interface functions. */
4251 iface_is_internal(const struct ovsrec_interface
*iface
,
4252 const struct ovsrec_bridge
*br
)
4254 /* The local port and "internal" ports are always "internal". */
4255 return !strcmp(iface
->type
, "internal") || !strcmp(iface
->name
, br
->name
);
4258 /* Returns the correct network device type for interface 'iface' in bridge
4261 iface_get_type(const struct ovsrec_interface
*iface
,
4262 const struct ovsrec_bridge
*br
)
4266 /* The local port always has type "internal". Other ports take
4267 * their type from the database and default to "system" if none is
4269 if (iface_is_internal(iface
, br
)) {
4272 type
= iface
->type
[0] ? iface
->type
: "system";
4275 return ofproto_port_open_type(br
->datapath_type
, type
);
4279 iface_destroy__(struct iface
*iface
)
4282 struct port
*port
= iface
->port
;
4283 struct bridge
*br
= port
->bridge
;
4285 if (br
->ofproto
&& iface
->ofp_port
!= OFPP_NONE
) {
4286 ofproto_port_unregister(br
->ofproto
, iface
->ofp_port
);
4289 if (iface
->ofp_port
!= OFPP_NONE
) {
4290 hmap_remove(&br
->ifaces
, &iface
->ofp_port_node
);
4293 ovs_list_remove(&iface
->port_elem
);
4294 hmap_remove(&br
->iface_by_name
, &iface
->name_node
);
4296 /* The user is changing configuration here, so netdev_remove needs to be
4297 * used as opposed to netdev_close */
4298 netdev_remove(iface
->netdev
);
4306 iface_destroy(struct iface
*iface
)
4309 struct port
*port
= iface
->port
;
4311 iface_destroy__(iface
);
4312 if (ovs_list_is_empty(&port
->ifaces
)) {
4318 static struct iface
*
4319 iface_lookup(const struct bridge
*br
, const char *name
)
4321 struct iface
*iface
;
4323 HMAP_FOR_EACH_WITH_HASH (iface
, name_node
, hash_string(name
, 0),
4324 &br
->iface_by_name
) {
4325 if (!strcmp(iface
->name
, name
)) {
4333 static struct iface
*
4334 iface_find(const char *name
)
4336 const struct bridge
*br
;
4338 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
4339 struct iface
*iface
= iface_lookup(br
, name
);
4348 static struct iface
*
4349 iface_from_ofp_port(const struct bridge
*br
, ofp_port_t ofp_port
)
4351 struct iface
*iface
;
4353 HMAP_FOR_EACH_IN_BUCKET (iface
, ofp_port_node
, hash_ofp_port(ofp_port
),
4355 if (iface
->ofp_port
== ofp_port
) {
4362 /* Set Ethernet address of 'iface', if one is specified in the configuration
4365 iface_set_mac(const struct bridge
*br
, const struct port
*port
, struct iface
*iface
)
4367 struct eth_addr ea
, *mac
= NULL
;
4368 struct iface
*hw_addr_iface
;
4370 if (strcmp(iface
->type
, "internal")) {
4374 if (iface
->cfg
->mac
&& eth_addr_from_string(iface
->cfg
->mac
, &ea
)) {
4376 } else if (port
->cfg
->fake_bridge
) {
4377 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4378 find_local_hw_addr(br
, &ea
, port
, &hw_addr_iface
);
4383 if (iface
->ofp_port
== OFPP_LOCAL
) {
4384 VLOG_ERR("interface %s: ignoring mac in Interface record "
4385 "(use Bridge record to set local port's mac)",
4387 } else if (eth_addr_is_multicast(*mac
)) {
4388 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4391 int error
= netdev_set_etheraddr(iface
->netdev
, *mac
);
4393 VLOG_ERR("interface %s: setting MAC failed (%s)",
4394 iface
->name
, ovs_strerror(error
));
4400 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4402 iface_set_ofport(const struct ovsrec_interface
*if_cfg
, ofp_port_t ofport
)
4404 if (if_cfg
&& !ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
4405 int64_t port
= ofport
== OFPP_NONE
? -1 : ofp_to_u16(ofport
);
4406 ovsrec_interface_set_ofport(if_cfg
, &port
, 1);
4410 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4411 * sets the "ofport" field to -1.
4413 * This is appropriate when 'if_cfg''s interface cannot be created or is
4414 * otherwise invalid. */
4416 iface_clear_db_record(const struct ovsrec_interface
*if_cfg
, char *errp
)
4418 if (!ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
4419 iface_set_ofport(if_cfg
, OFPP_NONE
);
4420 ovsrec_interface_set_error(if_cfg
, errp
);
4421 ovsrec_interface_set_status(if_cfg
, NULL
);
4422 ovsrec_interface_set_admin_state(if_cfg
, NULL
);
4423 ovsrec_interface_set_duplex(if_cfg
, NULL
);
4424 ovsrec_interface_set_link_speed(if_cfg
, NULL
, 0);
4425 ovsrec_interface_set_link_state(if_cfg
, NULL
);
4426 ovsrec_interface_set_mac_in_use(if_cfg
, NULL
);
4427 ovsrec_interface_set_mtu(if_cfg
, NULL
, 0);
4428 ovsrec_interface_set_cfm_fault(if_cfg
, NULL
, 0);
4429 ovsrec_interface_set_cfm_fault_status(if_cfg
, NULL
, 0);
4430 ovsrec_interface_set_cfm_remote_mpids(if_cfg
, NULL
, 0);
4431 ovsrec_interface_set_lacp_current(if_cfg
, NULL
, 0);
4432 ovsrec_interface_set_statistics(if_cfg
, NULL
, NULL
, 0);
4433 ovsrec_interface_set_ifindex(if_cfg
, NULL
, 0);
4438 queue_ids_include(const struct ovsdb_datum
*queues
, int64_t target
)
4440 union ovsdb_atom atom
;
4442 atom
.integer
= target
;
4443 return ovsdb_datum_find_key(queues
, &atom
, OVSDB_TYPE_INTEGER
) != UINT_MAX
;
4447 iface_configure_qos(struct iface
*iface
, const struct ovsrec_qos
*qos
)
4449 struct ofpbuf queues_buf
;
4451 ofpbuf_init(&queues_buf
, 0);
4453 if (!qos
|| qos
->type
[0] == '\0') {
4454 netdev_set_qos(iface
->netdev
, NULL
, NULL
);
4456 const struct ovsdb_datum
*queues
;
4457 struct netdev_queue_dump dump
;
4458 unsigned int queue_id
;
4459 struct smap details
;
4463 /* Configure top-level Qos for 'iface'. */
4464 netdev_set_qos(iface
->netdev
, qos
->type
, &qos
->other_config
);
4466 /* Deconfigure queues that were deleted. */
4467 queues
= ovsrec_qos_get_queues(qos
, OVSDB_TYPE_INTEGER
,
4469 smap_init(&details
);
4470 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
4471 if (!queue_ids_include(queues
, queue_id
)) {
4472 netdev_delete_queue(iface
->netdev
, queue_id
);
4475 smap_destroy(&details
);
4477 /* Configure queues for 'iface'. */
4479 for (i
= 0; i
< qos
->n_queues
; i
++) {
4480 const struct ovsrec_queue
*queue
= qos
->value_queues
[i
];
4481 unsigned int queue_id
= qos
->key_queues
[i
];
4483 if (queue_id
== 0) {
4487 if (queue
->n_dscp
== 1) {
4488 struct ofproto_port_queue
*port_queue
;
4490 port_queue
= ofpbuf_put_uninit(&queues_buf
,
4491 sizeof *port_queue
);
4492 port_queue
->queue
= queue_id
;
4493 port_queue
->dscp
= queue
->dscp
[0];
4496 netdev_set_queue(iface
->netdev
, queue_id
, &queue
->other_config
);
4499 struct smap details
;
4501 smap_init(&details
);
4502 netdev_set_queue(iface
->netdev
, 0, &details
);
4503 smap_destroy(&details
);
4507 if (iface
->ofp_port
!= OFPP_NONE
) {
4508 const struct ofproto_port_queue
*port_queues
= queues_buf
.data
;
4509 size_t n_queues
= queues_buf
.size
/ sizeof *port_queues
;
4511 ofproto_port_set_queues(iface
->port
->bridge
->ofproto
, iface
->ofp_port
,
4512 port_queues
, n_queues
);
4515 netdev_set_policing(iface
->netdev
,
4516 MIN(UINT32_MAX
, iface
->cfg
->ingress_policing_rate
),
4517 MIN(UINT32_MAX
, iface
->cfg
->ingress_policing_burst
));
4519 ofpbuf_uninit(&queues_buf
);
4523 iface_configure_cfm(struct iface
*iface
)
4525 const struct ovsrec_interface
*cfg
= iface
->cfg
;
4526 const char *opstate_str
;
4527 const char *cfm_ccm_vlan
;
4528 struct cfm_settings s
;
4529 struct smap netdev_args
;
4531 if (!cfg
->n_cfm_mpid
) {
4532 ofproto_port_clear_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
);
4536 s
.check_tnl_key
= false;
4537 smap_init(&netdev_args
);
4538 if (!netdev_get_config(iface
->netdev
, &netdev_args
)) {
4539 const char *key
= smap_get(&netdev_args
, "key");
4540 const char *in_key
= smap_get(&netdev_args
, "in_key");
4542 s
.check_tnl_key
= (key
&& !strcmp(key
, "flow"))
4543 || (in_key
&& !strcmp(in_key
, "flow"));
4545 smap_destroy(&netdev_args
);
4547 s
.mpid
= *cfg
->cfm_mpid
;
4548 s
.interval
= smap_get_int(&iface
->cfg
->other_config
, "cfm_interval", 0);
4549 cfm_ccm_vlan
= smap_get(&iface
->cfg
->other_config
, "cfm_ccm_vlan");
4550 s
.ccm_pcp
= smap_get_int(&iface
->cfg
->other_config
, "cfm_ccm_pcp", 0);
4552 if (s
.interval
<= 0) {
4556 if (!cfm_ccm_vlan
) {
4558 } else if (!strcasecmp("random", cfm_ccm_vlan
)) {
4559 s
.ccm_vlan
= CFM_RANDOM_VLAN
;
4561 s
.ccm_vlan
= atoi(cfm_ccm_vlan
);
4562 if (s
.ccm_vlan
== CFM_RANDOM_VLAN
) {
4567 s
.extended
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_extended",
4569 s
.demand
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_demand", false);
4571 opstate_str
= smap_get(&iface
->cfg
->other_config
, "cfm_opstate");
4572 s
.opup
= !opstate_str
|| !strcasecmp("up", opstate_str
);
4574 ofproto_port_set_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
, &s
);
4577 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4578 * instead of obtaining it from the database. */
4580 iface_is_synthetic(const struct iface
*iface
)
4582 return ovsdb_idl_row_is_synthetic(&iface
->cfg
->header_
);
4586 iface_validate_ofport__(size_t n
, int64_t *ofport
)
4588 return (n
&& *ofport
>= 1 && *ofport
< ofp_to_u16(OFPP_MAX
)
4589 ? u16_to_ofp(*ofport
)
4594 iface_get_requested_ofp_port(const struct ovsrec_interface
*cfg
)
4596 return iface_validate_ofport__(cfg
->n_ofport_request
, cfg
->ofport_request
);
4600 iface_pick_ofport(const struct ovsrec_interface
*cfg
)
4602 ofp_port_t requested_ofport
= iface_get_requested_ofp_port(cfg
);
4603 return (requested_ofport
!= OFPP_NONE
4605 : iface_validate_ofport__(cfg
->n_ofport
, cfg
->ofport
));
4608 /* Port mirroring. */
4610 static struct mirror
*
4611 mirror_find_by_uuid(struct bridge
*br
, const struct uuid
*uuid
)
4615 HMAP_FOR_EACH_IN_BUCKET (m
, hmap_node
, uuid_hash(uuid
), &br
->mirrors
) {
4616 if (uuid_equals(uuid
, &m
->uuid
)) {
4624 bridge_configure_mirrors(struct bridge
*br
)
4626 const struct ovsdb_datum
*mc
;
4627 unsigned long *flood_vlans
;
4628 struct mirror
*m
, *next
;
4631 /* Get rid of deleted mirrors. */
4632 mc
= ovsrec_bridge_get_mirrors(br
->cfg
, OVSDB_TYPE_UUID
);
4633 HMAP_FOR_EACH_SAFE (m
, next
, hmap_node
, &br
->mirrors
) {
4634 union ovsdb_atom atom
;
4636 atom
.uuid
= m
->uuid
;
4637 if (ovsdb_datum_find_key(mc
, &atom
, OVSDB_TYPE_UUID
) == UINT_MAX
) {
4642 /* Add new mirrors and reconfigure existing ones. */
4643 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
4644 const struct ovsrec_mirror
*cfg
= br
->cfg
->mirrors
[i
];
4645 struct mirror
*m
= mirror_find_by_uuid(br
, &cfg
->header_
.uuid
);
4647 m
= mirror_create(br
, cfg
);
4650 if (!mirror_configure(m
)) {
4655 /* Update flooded vlans (for RSPAN). */
4656 flood_vlans
= vlan_bitmap_from_array(br
->cfg
->flood_vlans
,
4657 br
->cfg
->n_flood_vlans
);
4658 ofproto_set_flood_vlans(br
->ofproto
, flood_vlans
);
4659 bitmap_free(flood_vlans
);
4662 static struct mirror
*
4663 mirror_create(struct bridge
*br
, const struct ovsrec_mirror
*cfg
)
4667 m
= xzalloc(sizeof *m
);
4668 m
->uuid
= cfg
->header_
.uuid
;
4669 hmap_insert(&br
->mirrors
, &m
->hmap_node
, uuid_hash(&m
->uuid
));
4671 m
->name
= xstrdup(cfg
->name
);
4677 mirror_destroy(struct mirror
*m
)
4680 struct bridge
*br
= m
->bridge
;
4683 ofproto_mirror_unregister(br
->ofproto
, m
);
4686 hmap_remove(&br
->mirrors
, &m
->hmap_node
);
4693 mirror_collect_ports(struct mirror
*m
,
4694 struct ovsrec_port
**in_ports
, int n_in_ports
,
4695 void ***out_portsp
, size_t *n_out_portsp
)
4697 void **out_ports
= xmalloc(n_in_ports
* sizeof *out_ports
);
4698 size_t n_out_ports
= 0;
4701 for (i
= 0; i
< n_in_ports
; i
++) {
4702 const char *name
= in_ports
[i
]->name
;
4703 struct port
*port
= port_lookup(m
->bridge
, name
);
4705 out_ports
[n_out_ports
++] = port
;
4707 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4708 "port %s", m
->bridge
->name
, m
->name
, name
);
4711 *out_portsp
= out_ports
;
4712 *n_out_portsp
= n_out_ports
;
4716 mirror_configure(struct mirror
*m
)
4718 const struct ovsrec_mirror
*cfg
= m
->cfg
;
4719 struct ofproto_mirror_settings s
;
4722 if (strcmp(cfg
->name
, m
->name
)) {
4724 m
->name
= xstrdup(cfg
->name
);
4728 /* Get output port or VLAN. */
4729 if (cfg
->output_port
) {
4730 s
.out_bundle
= port_lookup(m
->bridge
, cfg
->output_port
->name
);
4731 if (!s
.out_bundle
) {
4732 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4733 m
->bridge
->name
, m
->name
);
4736 s
.out_vlan
= UINT16_MAX
;
4738 if (cfg
->output_vlan
) {
4739 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4740 "output vlan; ignoring output vlan",
4741 m
->bridge
->name
, m
->name
);
4743 } else if (cfg
->output_vlan
) {
4744 /* The database should prevent invalid VLAN values. */
4745 s
.out_bundle
= NULL
;
4746 s
.out_vlan
= *cfg
->output_vlan
;
4748 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4749 m
->bridge
->name
, m
->name
);
4753 /* Get port selection. */
4754 if (cfg
->select_all
) {
4755 size_t n_ports
= hmap_count(&m
->bridge
->ports
);
4756 void **ports
= xmalloc(n_ports
* sizeof *ports
);
4761 HMAP_FOR_EACH (port
, hmap_node
, &m
->bridge
->ports
) {
4771 /* Get ports, dropping ports that don't exist.
4772 * The IDL ensures that there are no duplicates. */
4773 mirror_collect_ports(m
, cfg
->select_src_port
, cfg
->n_select_src_port
,
4774 &s
.srcs
, &s
.n_srcs
);
4775 mirror_collect_ports(m
, cfg
->select_dst_port
, cfg
->n_select_dst_port
,
4776 &s
.dsts
, &s
.n_dsts
);
4779 /* Get VLAN selection. */
4780 s
.src_vlans
= vlan_bitmap_from_array(cfg
->select_vlan
, cfg
->n_select_vlan
);
4783 ofproto_mirror_register(m
->bridge
->ofproto
, m
, &s
);
4786 if (s
.srcs
!= s
.dsts
) {
4797 mirror_refresh_stats(struct mirror
*m
)
4799 struct ofproto
*ofproto
= m
->bridge
->ofproto
;
4800 uint64_t tx_packets
, tx_bytes
;
4801 const char *keys
[2];
4803 size_t stat_cnt
= 0;
4805 if (ofproto_mirror_get_stats(ofproto
, m
, &tx_packets
, &tx_bytes
)) {
4806 ovsrec_mirror_set_statistics(m
->cfg
, NULL
, NULL
, 0);
4810 if (tx_packets
!= UINT64_MAX
) {
4811 keys
[stat_cnt
] = "tx_packets";
4812 values
[stat_cnt
] = tx_packets
;
4815 if (tx_bytes
!= UINT64_MAX
) {
4816 keys
[stat_cnt
] = "tx_bytes";
4817 values
[stat_cnt
] = tx_bytes
;
4821 ovsrec_mirror_set_statistics(m
->cfg
, keys
, values
, stat_cnt
);
4825 * Add registered netdev and dpif types to ovsdb to allow external
4826 * applications to query the capabilities of the Open vSwitch instance
4827 * running on the node.
4830 discover_types(const struct ovsrec_open_vswitch
*cfg
)
4834 /* Datapath types. */
4836 dp_enumerate_types(&types
);
4837 const char **datapath_types
= sset_array(&types
);
4838 ovsrec_open_vswitch_set_datapath_types(cfg
, datapath_types
,
4839 sset_count(&types
));
4840 free(datapath_types
);
4841 sset_destroy(&types
);
4845 netdev_enumerate_types(&types
);
4846 const char **iface_types
= sset_array(&types
);
4847 ovsrec_open_vswitch_set_iface_types(cfg
, iface_types
, sset_count(&types
));
4849 sset_destroy(&types
);