1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "mcast-snooping.h"
38 #include "meta-flow.h"
41 #include "ofp-print.h"
44 #include "ofproto/bond.h"
45 #include "ofproto/ofproto.h"
47 #include "poll-loop.h"
52 #include "socket-util.h"
54 #include "stream-ssl.h"
56 #include "system-stats.h"
61 #include "lib/vswitch-idl.h"
62 #include "xenserver.h"
64 #include "sflow_api.h"
65 #include "vlan-bitmap.h"
68 VLOG_DEFINE_THIS_MODULE(bridge
);
70 COVERAGE_DEFINE(bridge_reconfigure
);
73 /* These members are always valid.
75 * They are immutable: they never change between iface_create() and
77 struct list port_elem
; /* Element in struct port's "ifaces" list. */
78 struct hmap_node name_node
; /* In struct bridge's "iface_by_name" hmap. */
79 struct hmap_node ofp_port_node
; /* In struct bridge's "ifaces" hmap. */
80 struct port
*port
; /* Containing port. */
81 char *name
; /* Host network device name. */
82 struct netdev
*netdev
; /* Network device. */
83 ofp_port_t ofp_port
; /* OpenFlow port number. */
86 /* These members are valid only within bridge_reconfigure(). */
87 const char *type
; /* Usually same as cfg->type. */
88 const struct ovsrec_interface
*cfg
;
92 struct uuid uuid
; /* UUID of this "mirror" record in database. */
93 struct hmap_node hmap_node
; /* In struct bridge's "mirrors" hmap. */
94 struct bridge
*bridge
;
96 const struct ovsrec_mirror
*cfg
;
100 struct hmap_node hmap_node
; /* Element in struct bridge's "ports" hmap. */
101 struct bridge
*bridge
;
104 const struct ovsrec_port
*cfg
;
106 /* An ordinary bridge port has 1 interface.
107 * A bridge port for bonding has at least 2 interfaces. */
108 struct list ifaces
; /* List of "struct iface"s. */
112 struct hmap_node node
; /* In 'all_bridges'. */
113 char *name
; /* User-specified arbitrary name. */
114 char *type
; /* Datapath type. */
115 uint8_t ea
[ETH_ADDR_LEN
]; /* Bridge Ethernet Address. */
116 uint8_t default_ea
[ETH_ADDR_LEN
]; /* Default MAC. */
117 const struct ovsrec_bridge
*cfg
;
119 /* OpenFlow switch processing. */
120 struct ofproto
*ofproto
; /* OpenFlow switch. */
123 struct hmap ports
; /* "struct port"s indexed by name. */
124 struct hmap ifaces
; /* "struct iface"s indexed by ofp_port. */
125 struct hmap iface_by_name
; /* "struct iface"s indexed by name. */
127 /* Port mirroring. */
128 struct hmap mirrors
; /* "struct mirror" indexed by UUID. */
130 /* Used during reconfiguration. */
131 struct shash wanted_ports
;
133 /* Synthetic local port if necessary. */
134 struct ovsrec_port synth_local_port
;
135 struct ovsrec_interface synth_local_iface
;
136 struct ovsrec_interface
*synth_local_ifacep
;
139 /* All bridges, indexed by name. */
140 static struct hmap all_bridges
= HMAP_INITIALIZER(&all_bridges
);
142 /* OVSDB IDL used to obtain configuration. */
143 static struct ovsdb_idl
*idl
;
145 /* We want to complete daemonization, fully detaching from our parent process,
146 * only after we have completed our initial configuration, committed our state
147 * to the database, and received confirmation back from the database server
148 * that it applied the commit. This allows our parent process to know that,
149 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
150 * to have already been filled in. (It is only "very likely" rather than
151 * certain because there is always a slim possibility that the transaction will
152 * fail or that some other client has added new bridges, ports, etc. while
153 * ovs-vswitchd was configuring using an old configuration.)
155 * We only need to do this once for our initial configuration at startup, so
156 * 'initial_config_done' tracks whether we've already done it. While we are
157 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
158 * itself and is otherwise NULL. */
159 static bool initial_config_done
;
160 static struct ovsdb_idl_txn
*daemonize_txn
;
162 /* Most recently processed IDL sequence number. */
163 static unsigned int idl_seqno
;
165 /* Track changes to port connectivity. */
166 static uint64_t connectivity_seqno
= LLONG_MIN
;
168 /* Status update to database.
170 * Some information in the database must be kept as up-to-date as possible to
171 * allow controllers to respond rapidly to network outages. Those status are
172 * updated via the 'status_txn'.
174 * We use the global connectivity sequence number to detect the status change.
175 * Also, to prevent the status update from sending too much to the database,
176 * we check the return status of each update transaction and do not start new
177 * update if the previous transaction status is 'TXN_INCOMPLETE'.
179 * 'statux_txn' is NULL if there is no ongoing status update.
181 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
182 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
183 * which will cause the main thread wake up soon and retry the status update.
185 static struct ovsdb_idl_txn
*status_txn
;
186 static bool status_txn_try_again
;
188 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
189 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
190 #define STATUS_CHECK_AGAIN_MSEC 100
192 /* Each time this timer expires, the bridge fetches interface and mirror
193 * statistics and pushes them into the database. */
194 static int stats_timer_interval
;
195 static long long int stats_timer
= LLONG_MIN
;
197 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
198 * expensive. This can cause bridge_reconfigure() to take a long time during
199 * which no other work can be done. To deal with this problem, we limit port
200 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
201 * call to bridge_reconfigure(). If there is more work to do after the limit
202 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
203 * This allows the rest of the code to catch up on important things like
204 * forwarding packets. */
205 #define OFP_PORT_ACTION_WINDOW 10
207 static void add_del_bridges(const struct ovsrec_open_vswitch
*);
208 static void bridge_run__(void);
209 static void bridge_create(const struct ovsrec_bridge
*);
210 static void bridge_destroy(struct bridge
*);
211 static struct bridge
*bridge_lookup(const char *name
);
212 static unixctl_cb_func bridge_unixctl_dump_flows
;
213 static unixctl_cb_func bridge_unixctl_reconnect
;
214 static size_t bridge_get_controllers(const struct bridge
*br
,
215 struct ovsrec_controller
***controllersp
);
216 static void bridge_collect_wanted_ports(struct bridge
*,
217 const unsigned long *splinter_vlans
,
218 struct shash
*wanted_ports
);
219 static void bridge_delete_ofprotos(void);
220 static void bridge_delete_or_reconfigure_ports(struct bridge
*);
221 static void bridge_del_ports(struct bridge
*,
222 const struct shash
*wanted_ports
);
223 static void bridge_add_ports(struct bridge
*,
224 const struct shash
*wanted_ports
);
226 static void bridge_configure_datapath_id(struct bridge
*);
227 static void bridge_configure_netflow(struct bridge
*);
228 static void bridge_configure_forward_bpdu(struct bridge
*);
229 static void bridge_configure_mac_table(struct bridge
*);
230 static void bridge_configure_mcast_snooping(struct bridge
*);
231 static void bridge_configure_sflow(struct bridge
*, int *sflow_bridge_number
);
232 static void bridge_configure_ipfix(struct bridge
*);
233 static void bridge_configure_stp(struct bridge
*);
234 static void bridge_configure_rstp(struct bridge
*);
235 static void bridge_configure_tables(struct bridge
*);
236 static void bridge_configure_dp_desc(struct bridge
*);
237 static void bridge_configure_remotes(struct bridge
*,
238 const struct sockaddr_in
*managers
,
240 static void bridge_pick_local_hw_addr(struct bridge
*,
241 uint8_t ea
[ETH_ADDR_LEN
],
242 struct iface
**hw_addr_iface
);
243 static uint64_t bridge_pick_datapath_id(struct bridge
*,
244 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
245 struct iface
*hw_addr_iface
);
246 static uint64_t dpid_from_hash(const void *, size_t nbytes
);
247 static bool bridge_has_bond_fake_iface(const struct bridge
*,
249 static bool port_is_bond_fake_iface(const struct port
*);
251 static unixctl_cb_func qos_unixctl_show
;
253 static struct port
*port_create(struct bridge
*, const struct ovsrec_port
*);
254 static void port_del_ifaces(struct port
*);
255 static void port_destroy(struct port
*);
256 static struct port
*port_lookup(const struct bridge
*, const char *name
);
257 static void port_configure(struct port
*);
258 static struct lacp_settings
*port_configure_lacp(struct port
*,
259 struct lacp_settings
*);
260 static void port_configure_bond(struct port
*, struct bond_settings
*);
261 static bool port_is_synthetic(const struct port
*);
263 static void reconfigure_system_stats(const struct ovsrec_open_vswitch
*);
264 static void run_system_stats(void);
266 static void bridge_configure_mirrors(struct bridge
*);
267 static struct mirror
*mirror_create(struct bridge
*,
268 const struct ovsrec_mirror
*);
269 static void mirror_destroy(struct mirror
*);
270 static bool mirror_configure(struct mirror
*);
271 static void mirror_refresh_stats(struct mirror
*);
273 static void iface_configure_lacp(struct iface
*, struct lacp_slave_settings
*);
274 static bool iface_create(struct bridge
*, const struct ovsrec_interface
*,
275 const struct ovsrec_port
*);
276 static bool iface_is_internal(const struct ovsrec_interface
*iface
,
277 const struct ovsrec_bridge
*br
);
278 static const char *iface_get_type(const struct ovsrec_interface
*,
279 const struct ovsrec_bridge
*);
280 static void iface_destroy(struct iface
*);
281 static void iface_destroy__(struct iface
*);
282 static struct iface
*iface_lookup(const struct bridge
*, const char *name
);
283 static struct iface
*iface_find(const char *name
);
284 static struct iface
*iface_from_ofp_port(const struct bridge
*,
285 ofp_port_t ofp_port
);
286 static void iface_set_mac(const struct bridge
*, const struct port
*, struct iface
*);
287 static void iface_set_ofport(const struct ovsrec_interface
*, ofp_port_t ofport
);
288 static void iface_clear_db_record(const struct ovsrec_interface
*if_cfg
, char *errp
);
289 static void iface_configure_qos(struct iface
*, const struct ovsrec_qos
*);
290 static void iface_configure_cfm(struct iface
*);
291 static void iface_refresh_cfm_stats(struct iface
*);
292 static void iface_refresh_stats(struct iface
*);
293 static void iface_refresh_netdev_status(struct iface
*);
294 static void iface_refresh_ofproto_status(struct iface
*);
295 static bool iface_is_synthetic(const struct iface
*);
296 static ofp_port_t
iface_get_requested_ofp_port(
297 const struct ovsrec_interface
*);
298 static ofp_port_t
iface_pick_ofport(const struct ovsrec_interface
*);
300 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
302 * This is deprecated. It is only for compatibility with broken device drivers
303 * in old versions of Linux that do not properly support VLANs when VLAN
304 * devices are not used. When broken device drivers are no longer in
305 * widespread use, we will delete these interfaces. */
307 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
308 static bool vlan_splinters_enabled_anywhere
;
310 static bool vlan_splinters_is_enabled(const struct ovsrec_interface
*);
311 static unsigned long int *collect_splinter_vlans(
312 const struct ovsrec_open_vswitch
*);
313 static void configure_splinter_port(struct port
*);
314 static void add_vlan_splinter_ports(struct bridge
*,
315 const unsigned long int *splinter_vlans
,
316 struct shash
*ports
);
319 bridge_init_ofproto(const struct ovsrec_open_vswitch
*cfg
)
321 struct shash iface_hints
;
322 static bool initialized
= false;
329 shash_init(&iface_hints
);
332 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
333 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
336 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
337 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
340 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
341 struct ovsrec_interface
*if_cfg
= port_cfg
->interfaces
[k
];
342 struct iface_hint
*iface_hint
;
344 iface_hint
= xmalloc(sizeof *iface_hint
);
345 iface_hint
->br_name
= br_cfg
->name
;
346 iface_hint
->br_type
= br_cfg
->datapath_type
;
347 iface_hint
->ofp_port
= iface_pick_ofport(if_cfg
);
349 shash_add(&iface_hints
, if_cfg
->name
, iface_hint
);
355 ofproto_init(&iface_hints
);
357 shash_destroy_free_data(&iface_hints
);
361 /* Public functions. */
363 /* Initializes the bridge module, configuring it to obtain its configuration
364 * from an OVSDB server accessed over 'remote', which should be a string in a
365 * form acceptable to ovsdb_idl_create(). */
367 bridge_init(const char *remote
)
369 /* Create connection to database. */
370 idl
= ovsdb_idl_create(remote
, &ovsrec_idl_class
, true, true);
371 idl_seqno
= ovsdb_idl_get_seqno(idl
);
372 ovsdb_idl_set_lock(idl
, "ovs_vswitchd");
373 ovsdb_idl_verify_write_only(idl
);
375 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_cur_cfg
);
376 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_statistics
);
377 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_external_ids
);
378 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_ovs_version
);
379 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_db_version
);
380 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_type
);
381 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_version
);
383 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_id
);
384 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_version
);
385 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_status
);
386 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_rstp_status
);
387 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_stp_enable
);
388 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_rstp_enable
);
389 ovsdb_idl_omit(idl
, &ovsrec_bridge_col_external_ids
);
391 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_status
);
392 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_rstp_status
);
393 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_rstp_statistics
);
394 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_statistics
);
395 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_bond_active_slave
);
396 ovsdb_idl_omit(idl
, &ovsrec_port_col_external_ids
);
398 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_admin_state
);
399 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_duplex
);
400 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_speed
);
401 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_state
);
402 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_resets
);
403 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mac_in_use
);
404 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ifindex
);
405 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mtu
);
406 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ofport
);
407 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_statistics
);
408 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_status
);
409 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault
);
410 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault_status
);
411 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_mpids
);
412 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_flap_count
);
413 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_health
);
414 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_opstate
);
415 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_bfd_status
);
416 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_lacp_current
);
417 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_error
);
418 ovsdb_idl_omit(idl
, &ovsrec_interface_col_external_ids
);
420 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_is_connected
);
421 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_role
);
422 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_status
);
423 ovsdb_idl_omit(idl
, &ovsrec_controller_col_external_ids
);
425 ovsdb_idl_omit(idl
, &ovsrec_qos_col_external_ids
);
427 ovsdb_idl_omit(idl
, &ovsrec_queue_col_external_ids
);
429 ovsdb_idl_omit(idl
, &ovsrec_mirror_col_external_ids
);
430 ovsdb_idl_omit_alert(idl
, &ovsrec_mirror_col_statistics
);
432 ovsdb_idl_omit(idl
, &ovsrec_netflow_col_external_ids
);
433 ovsdb_idl_omit(idl
, &ovsrec_sflow_col_external_ids
);
434 ovsdb_idl_omit(idl
, &ovsrec_ipfix_col_external_ids
);
435 ovsdb_idl_omit(idl
, &ovsrec_flow_sample_collector_set_col_external_ids
);
437 ovsdb_idl_omit(idl
, &ovsrec_manager_col_external_ids
);
438 ovsdb_idl_omit(idl
, &ovsrec_manager_col_inactivity_probe
);
439 ovsdb_idl_omit(idl
, &ovsrec_manager_col_is_connected
);
440 ovsdb_idl_omit(idl
, &ovsrec_manager_col_max_backoff
);
441 ovsdb_idl_omit(idl
, &ovsrec_manager_col_status
);
443 ovsdb_idl_omit(idl
, &ovsrec_ssl_col_external_ids
);
445 /* Register unixctl commands. */
446 unixctl_command_register("qos/show", "interface", 1, 1,
447 qos_unixctl_show
, NULL
);
448 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
449 bridge_unixctl_dump_flows
, NULL
);
450 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
451 bridge_unixctl_reconnect
, NULL
);
463 struct bridge
*br
, *next_br
;
465 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
468 ovsdb_idl_destroy(idl
);
471 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
472 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
473 * responsible for freeing '*managersp' (with free()).
475 * You may be asking yourself "why does ovs-vswitchd care?", because
476 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
477 * should not be and in fact is not directly involved in that. But
478 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
479 * it has to tell in-band control where the managers are to enable that.
480 * (Thus, only managers connected in-band are collected.)
483 collect_in_band_managers(const struct ovsrec_open_vswitch
*ovs_cfg
,
484 struct sockaddr_in
**managersp
, size_t *n_managersp
)
486 struct sockaddr_in
*managers
= NULL
;
487 size_t n_managers
= 0;
491 /* Collect all of the potential targets from the "targets" columns of the
492 * rows pointed to by "manager_options", excluding any that are
495 for (i
= 0; i
< ovs_cfg
->n_manager_options
; i
++) {
496 struct ovsrec_manager
*m
= ovs_cfg
->manager_options
[i
];
498 if (m
->connection_mode
&& !strcmp(m
->connection_mode
, "out-of-band")) {
499 sset_find_and_delete(&targets
, m
->target
);
501 sset_add(&targets
, m
->target
);
505 /* Now extract the targets' IP addresses. */
506 if (!sset_is_empty(&targets
)) {
509 managers
= xmalloc(sset_count(&targets
) * sizeof *managers
);
510 SSET_FOR_EACH (target
, &targets
) {
512 struct sockaddr_storage ss
;
513 struct sockaddr_in in
;
516 if (stream_parse_target_with_default_port(target
, OVSDB_OLD_PORT
,
518 && sa
.ss
.ss_family
== AF_INET
) {
519 managers
[n_managers
++] = sa
.in
;
523 sset_destroy(&targets
);
525 *managersp
= managers
;
526 *n_managersp
= n_managers
;
530 bridge_reconfigure(const struct ovsrec_open_vswitch
*ovs_cfg
)
532 unsigned long int *splinter_vlans
;
533 struct sockaddr_in
*managers
;
534 struct bridge
*br
, *next
;
535 int sflow_bridge_number
;
538 COVERAGE_INC(bridge_reconfigure
);
540 ofproto_set_flow_limit(smap_get_int(&ovs_cfg
->other_config
, "flow-limit",
541 OFPROTO_FLOW_LIMIT_DEFAULT
));
542 ofproto_set_max_idle(smap_get_int(&ovs_cfg
->other_config
, "max-idle",
543 OFPROTO_MAX_IDLE_DEFAULT
));
544 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg
->other_config
,
546 ofproto_set_cpu_mask(smap_get(&ovs_cfg
->other_config
, "pmd-cpu-mask"));
549 smap_get_int(&ovs_cfg
->other_config
, "n-handler-threads", 0),
550 smap_get_int(&ovs_cfg
->other_config
, "n-revalidator-threads", 0));
552 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
553 * to 'ovs_cfg', with only very minimal configuration otherwise.
555 * This is mostly an update to bridge data structures. Nothing is pushed
556 * down to ofproto or lower layers. */
557 add_del_bridges(ovs_cfg
);
558 splinter_vlans
= collect_splinter_vlans(ovs_cfg
);
559 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
560 bridge_collect_wanted_ports(br
, splinter_vlans
, &br
->wanted_ports
);
561 bridge_del_ports(br
, &br
->wanted_ports
);
563 free(splinter_vlans
);
565 /* Start pushing configuration changes down to the ofproto layer:
567 * - Delete ofprotos that are no longer configured.
569 * - Delete ports that are no longer configured.
571 * - Reconfigure existing ports to their desired configurations, or
572 * delete them if not possible.
574 * We have to do all the deletions before we can do any additions, because
575 * the ports to be added might require resources that will be freed up by
576 * deletions (they might especially overlap in name). */
577 bridge_delete_ofprotos();
578 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
580 bridge_delete_or_reconfigure_ports(br
);
584 /* Finish pushing configuration changes to the ofproto layer:
586 * - Create ofprotos that are missing.
588 * - Add ports that are missing. */
589 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
593 error
= ofproto_create(br
->name
, br
->type
, &br
->ofproto
);
595 VLOG_ERR("failed to create bridge %s: %s", br
->name
,
596 ovs_strerror(error
));
597 shash_destroy(&br
->wanted_ports
);
600 /* Trigger storing datapath version. */
601 seq_change(connectivity_seq_get());
605 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
606 bridge_add_ports(br
, &br
->wanted_ports
);
607 shash_destroy(&br
->wanted_ports
);
610 reconfigure_system_stats(ovs_cfg
);
612 /* Complete the configuration. */
613 sflow_bridge_number
= 0;
614 collect_in_band_managers(ovs_cfg
, &managers
, &n_managers
);
615 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
618 /* We need the datapath ID early to allow LACP ports to use it as the
619 * default system ID. */
620 bridge_configure_datapath_id(br
);
622 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
625 port_configure(port
);
627 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
628 iface_set_ofport(iface
->cfg
, iface
->ofp_port
);
629 /* Clear eventual previous errors */
630 ovsrec_interface_set_error(iface
->cfg
, NULL
);
631 iface_configure_cfm(iface
);
632 iface_configure_qos(iface
, port
->cfg
->qos
);
633 iface_set_mac(br
, port
, iface
);
634 ofproto_port_set_bfd(br
->ofproto
, iface
->ofp_port
,
638 bridge_configure_mirrors(br
);
639 bridge_configure_forward_bpdu(br
);
640 bridge_configure_mac_table(br
);
641 bridge_configure_mcast_snooping(br
);
642 bridge_configure_remotes(br
, managers
, n_managers
);
643 bridge_configure_netflow(br
);
644 bridge_configure_sflow(br
, &sflow_bridge_number
);
645 bridge_configure_ipfix(br
);
646 bridge_configure_stp(br
);
647 bridge_configure_rstp(br
);
648 bridge_configure_tables(br
);
649 bridge_configure_dp_desc(br
);
653 /* The ofproto-dpif provider does some final reconfiguration in its
654 * ->type_run() function. We have to call it before notifying the database
655 * client that reconfiguration is complete, otherwise there is a very
656 * narrow race window in which e.g. ofproto/trace will not recognize the
657 * new configuration (sometimes this causes unit test failures). */
661 /* Delete ofprotos which aren't configured or have the wrong type. Create
662 * ofprotos which don't exist but need to. */
664 bridge_delete_ofprotos(void)
671 /* Delete ofprotos with no bridge or with the wrong type. */
674 ofproto_enumerate_types(&types
);
675 SSET_FOR_EACH (type
, &types
) {
678 ofproto_enumerate_names(type
, &names
);
679 SSET_FOR_EACH (name
, &names
) {
680 br
= bridge_lookup(name
);
681 if (!br
|| strcmp(type
, br
->type
)) {
682 ofproto_delete(name
, type
);
686 sset_destroy(&names
);
687 sset_destroy(&types
);
691 add_ofp_port(ofp_port_t port
, ofp_port_t
*ports
, size_t *n
, size_t *allocated
)
693 if (*n
>= *allocated
) {
694 ports
= x2nrealloc(ports
, allocated
, sizeof *ports
);
696 ports
[(*n
)++] = port
;
701 bridge_delete_or_reconfigure_ports(struct bridge
*br
)
703 struct ofproto_port ofproto_port
;
704 struct ofproto_port_dump dump
;
706 struct sset ofproto_ports
;
707 struct port
*port
, *port_next
;
709 /* List of "ofp_port"s to delete. We make a list instead of deleting them
710 * right away because ofproto implementations aren't necessarily able to
711 * iterate through a changing list of ports in an entirely robust way. */
718 sset_init(&ofproto_ports
);
720 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
721 * that are not configured in the database. (This commonly happens when
722 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
724 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
725 * that have the wrong OpenFlow port number (and arrange to add them back
726 * with the correct OpenFlow port number). */
727 OFPROTO_PORT_FOR_EACH (&ofproto_port
, &dump
, br
->ofproto
) {
728 ofp_port_t requested_ofp_port
;
731 sset_add(&ofproto_ports
, ofproto_port
.name
);
733 iface
= iface_lookup(br
, ofproto_port
.name
);
735 /* No such iface is configured, so we should delete this
738 * As a corner case exception, keep the port if it's a bond fake
740 if (bridge_has_bond_fake_iface(br
, ofproto_port
.name
)
741 && !strcmp(ofproto_port
.type
, "internal")) {
747 if (strcmp(ofproto_port
.type
, iface
->type
)
748 || netdev_set_config(iface
->netdev
, &iface
->cfg
->options
, NULL
)) {
749 /* The interface is the wrong type or can't be configured.
754 /* If the requested OpenFlow port for 'iface' changed, and it's not
755 * already the correct port, then we might want to temporarily delete
756 * this interface, so we can add it back again with the new OpenFlow
758 requested_ofp_port
= iface_get_requested_ofp_port(iface
->cfg
);
759 if (iface
->ofp_port
!= OFPP_LOCAL
&&
760 requested_ofp_port
!= OFPP_NONE
&&
761 requested_ofp_port
!= iface
->ofp_port
) {
762 ofp_port_t victim_request
;
763 struct iface
*victim
;
765 /* Check for an existing OpenFlow port currently occupying
766 * 'iface''s requested port number. If there isn't one, then
767 * delete this port. Otherwise we need to consider further. */
768 victim
= iface_from_ofp_port(br
, requested_ofp_port
);
773 /* 'victim' is a port currently using 'iface''s requested port
774 * number. Unless 'victim' specifically requested that port
775 * number, too, then we can delete both 'iface' and 'victim'
776 * temporarily. (We'll add both of them back again later with new
777 * OpenFlow port numbers.)
779 * If 'victim' did request port number 'requested_ofp_port', just
780 * like 'iface', then that's a configuration inconsistency that we
781 * can't resolve. We might as well let it keep its current port
783 victim_request
= iface_get_requested_ofp_port(victim
->cfg
);
784 if (victim_request
!= requested_ofp_port
) {
785 del
= add_ofp_port(victim
->ofp_port
, del
, &n
, &allocated
);
786 iface_destroy(victim
);
795 iface_destroy(iface
);
796 del
= add_ofp_port(ofproto_port
.ofp_port
, del
, &n
, &allocated
);
798 for (i
= 0; i
< n
; i
++) {
799 ofproto_port_del(br
->ofproto
, del
[i
]);
803 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
804 * that we didn't see when we iterated through the datapath, i.e. ports
805 * that disappeared underneath use. This is an unusual situation, but it
806 * can happen in some cases:
808 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
809 * idea but could happen).
811 * - The port represented a device that disappeared, e.g. a tuntap
812 * device destroyed via "tunctl -d", a physical Ethernet device
813 * whose module was just unloaded via "rmmod", or a virtual NIC for a
814 * VM whose VM was just terminated. */
815 HMAP_FOR_EACH_SAFE (port
, port_next
, hmap_node
, &br
->ports
) {
816 struct iface
*iface
, *iface_next
;
818 LIST_FOR_EACH_SAFE (iface
, iface_next
, port_elem
, &port
->ifaces
) {
819 if (!sset_contains(&ofproto_ports
, iface
->name
)) {
820 iface_destroy__(iface
);
824 if (list_is_empty(&port
->ifaces
)) {
828 sset_destroy(&ofproto_ports
);
832 bridge_add_ports__(struct bridge
*br
, const struct shash
*wanted_ports
,
833 bool with_requested_port
)
835 struct shash_node
*port_node
;
837 SHASH_FOR_EACH (port_node
, wanted_ports
) {
838 const struct ovsrec_port
*port_cfg
= port_node
->data
;
841 for (i
= 0; i
< port_cfg
->n_interfaces
; i
++) {
842 const struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[i
];
843 ofp_port_t requested_ofp_port
;
845 requested_ofp_port
= iface_get_requested_ofp_port(iface_cfg
);
846 if ((requested_ofp_port
!= OFPP_NONE
) == with_requested_port
) {
847 struct iface
*iface
= iface_lookup(br
, iface_cfg
->name
);
850 iface_create(br
, iface_cfg
, port_cfg
);
858 bridge_add_ports(struct bridge
*br
, const struct shash
*wanted_ports
)
860 /* First add interfaces that request a particular port number. */
861 bridge_add_ports__(br
, wanted_ports
, true);
863 /* Then add interfaces that want automatic port number assignment.
864 * We add these afterward to avoid accidentally taking a specifically
865 * requested port number. */
866 bridge_add_ports__(br
, wanted_ports
, false);
870 port_configure(struct port
*port
)
872 const struct ovsrec_port
*cfg
= port
->cfg
;
873 struct bond_settings bond_settings
;
874 struct lacp_settings lacp_settings
;
875 struct ofproto_bundle_settings s
;
878 if (cfg
->vlan_mode
&& !strcmp(cfg
->vlan_mode
, "splinter")) {
879 configure_splinter_port(port
);
888 s
.slaves
= xmalloc(list_size(&port
->ifaces
) * sizeof *s
.slaves
);
889 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
890 s
.slaves
[s
.n_slaves
++] = iface
->ofp_port
;
895 if (cfg
->tag
&& *cfg
->tag
>= 0 && *cfg
->tag
<= 4095) {
899 /* Get VLAN trunks. */
902 s
.trunks
= vlan_bitmap_from_array(cfg
->trunks
, cfg
->n_trunks
);
906 if (cfg
->vlan_mode
) {
907 if (!strcmp(cfg
->vlan_mode
, "access")) {
908 s
.vlan_mode
= PORT_VLAN_ACCESS
;
909 } else if (!strcmp(cfg
->vlan_mode
, "trunk")) {
910 s
.vlan_mode
= PORT_VLAN_TRUNK
;
911 } else if (!strcmp(cfg
->vlan_mode
, "native-tagged")) {
912 s
.vlan_mode
= PORT_VLAN_NATIVE_TAGGED
;
913 } else if (!strcmp(cfg
->vlan_mode
, "native-untagged")) {
914 s
.vlan_mode
= PORT_VLAN_NATIVE_UNTAGGED
;
916 /* This "can't happen" because ovsdb-server should prevent it. */
917 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
918 "back to trunk mode", port
->name
, cfg
->vlan_mode
);
919 s
.vlan_mode
= PORT_VLAN_TRUNK
;
923 s
.vlan_mode
= PORT_VLAN_ACCESS
;
925 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
929 s
.vlan_mode
= PORT_VLAN_TRUNK
;
932 s
.use_priority_tags
= smap_get_bool(&cfg
->other_config
, "priority-tags",
935 /* Get LACP settings. */
936 s
.lacp
= port_configure_lacp(port
, &lacp_settings
);
940 s
.lacp_slaves
= xmalloc(s
.n_slaves
* sizeof *s
.lacp_slaves
);
941 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
942 iface_configure_lacp(iface
, &s
.lacp_slaves
[i
++]);
945 s
.lacp_slaves
= NULL
;
948 /* Get bond settings. */
949 if (s
.n_slaves
> 1) {
950 s
.bond
= &bond_settings
;
951 port_configure_bond(port
, &bond_settings
);
954 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
955 netdev_set_miimon_interval(iface
->netdev
, 0);
960 ofproto_bundle_register(port
->bridge
->ofproto
, port
, &s
);
968 /* Pick local port hardware address and datapath ID for 'br'. */
970 bridge_configure_datapath_id(struct bridge
*br
)
972 uint8_t ea
[ETH_ADDR_LEN
];
974 struct iface
*local_iface
;
975 struct iface
*hw_addr_iface
;
978 bridge_pick_local_hw_addr(br
, ea
, &hw_addr_iface
);
979 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
981 int error
= netdev_set_etheraddr(local_iface
->netdev
, ea
);
983 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
984 VLOG_ERR_RL(&rl
, "bridge %s: failed to set bridge "
985 "Ethernet address: %s",
986 br
->name
, ovs_strerror(error
));
989 memcpy(br
->ea
, ea
, ETH_ADDR_LEN
);
991 dpid
= bridge_pick_datapath_id(br
, ea
, hw_addr_iface
);
992 if (dpid
!= ofproto_get_datapath_id(br
->ofproto
)) {
993 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64
, br
->name
, dpid
);
994 ofproto_set_datapath_id(br
->ofproto
, dpid
);
997 dpid_string
= xasprintf("%016"PRIx64
, dpid
);
998 ovsrec_bridge_set_datapath_id(br
->cfg
, dpid_string
);
1002 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1005 bridge_get_allowed_versions(struct bridge
*br
)
1007 if (!br
->cfg
->n_protocols
)
1010 return ofputil_versions_from_strings(br
->cfg
->protocols
,
1011 br
->cfg
->n_protocols
);
1014 /* Set NetFlow configuration on 'br'. */
1016 bridge_configure_netflow(struct bridge
*br
)
1018 struct ovsrec_netflow
*cfg
= br
->cfg
->netflow
;
1019 struct netflow_options opts
;
1022 ofproto_set_netflow(br
->ofproto
, NULL
);
1026 memset(&opts
, 0, sizeof opts
);
1028 /* Get default NetFlow configuration from datapath.
1029 * Apply overrides from 'cfg'. */
1030 ofproto_get_netflow_ids(br
->ofproto
, &opts
.engine_type
, &opts
.engine_id
);
1031 if (cfg
->engine_type
) {
1032 opts
.engine_type
= *cfg
->engine_type
;
1034 if (cfg
->engine_id
) {
1035 opts
.engine_id
= *cfg
->engine_id
;
1038 /* Configure active timeout interval. */
1039 opts
.active_timeout
= cfg
->active_timeout
;
1040 if (!opts
.active_timeout
) {
1041 opts
.active_timeout
= -1;
1042 } else if (opts
.active_timeout
< 0) {
1043 VLOG_WARN("bridge %s: active timeout interval set to negative "
1044 "value, using default instead (%d seconds)", br
->name
,
1045 NF_ACTIVE_TIMEOUT_DEFAULT
);
1046 opts
.active_timeout
= -1;
1049 /* Add engine ID to interface number to disambiguate bridgs? */
1050 opts
.add_id_to_iface
= cfg
->add_id_to_interface
;
1051 if (opts
.add_id_to_iface
) {
1052 if (opts
.engine_id
> 0x7f) {
1053 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1054 "another vswitch, choose an engine id less than 128",
1057 if (hmap_count(&br
->ports
) > 508) {
1058 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1059 "another port when more than 508 ports are used",
1065 sset_init(&opts
.collectors
);
1066 sset_add_array(&opts
.collectors
, cfg
->targets
, cfg
->n_targets
);
1069 if (ofproto_set_netflow(br
->ofproto
, &opts
)) {
1070 VLOG_ERR("bridge %s: problem setting netflow collectors", br
->name
);
1072 sset_destroy(&opts
.collectors
);
1075 /* Set sFlow configuration on 'br'. */
1077 bridge_configure_sflow(struct bridge
*br
, int *sflow_bridge_number
)
1079 const struct ovsrec_sflow
*cfg
= br
->cfg
->sflow
;
1080 struct ovsrec_controller
**controllers
;
1081 struct ofproto_sflow_options oso
;
1082 size_t n_controllers
;
1086 ofproto_set_sflow(br
->ofproto
, NULL
);
1090 memset(&oso
, 0, sizeof oso
);
1092 sset_init(&oso
.targets
);
1093 sset_add_array(&oso
.targets
, cfg
->targets
, cfg
->n_targets
);
1095 oso
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1096 if (cfg
->sampling
) {
1097 oso
.sampling_rate
= *cfg
->sampling
;
1100 oso
.polling_interval
= SFL_DEFAULT_POLLING_INTERVAL
;
1102 oso
.polling_interval
= *cfg
->polling
;
1105 oso
.header_len
= SFL_DEFAULT_HEADER_SIZE
;
1107 oso
.header_len
= *cfg
->header
;
1110 oso
.sub_id
= (*sflow_bridge_number
)++;
1111 oso
.agent_device
= cfg
->agent
;
1113 oso
.control_ip
= NULL
;
1114 n_controllers
= bridge_get_controllers(br
, &controllers
);
1115 for (i
= 0; i
< n_controllers
; i
++) {
1116 if (controllers
[i
]->local_ip
) {
1117 oso
.control_ip
= controllers
[i
]->local_ip
;
1121 ofproto_set_sflow(br
->ofproto
, &oso
);
1123 sset_destroy(&oso
.targets
);
1126 /* Returns whether a IPFIX row is valid. */
1128 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix
*ipfix
)
1130 return ipfix
&& ipfix
->n_targets
> 0;
1133 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1135 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set
*fscs
,
1136 const struct bridge
*br
)
1138 return ovsrec_ipfix_is_valid(fscs
->ipfix
) && fscs
->bridge
== br
->cfg
;
1141 /* Set IPFIX configuration on 'br'. */
1143 bridge_configure_ipfix(struct bridge
*br
)
1145 const struct ovsrec_ipfix
*be_cfg
= br
->cfg
->ipfix
;
1146 bool valid_be_cfg
= ovsrec_ipfix_is_valid(be_cfg
);
1147 const struct ovsrec_flow_sample_collector_set
*fe_cfg
;
1148 struct ofproto_ipfix_bridge_exporter_options be_opts
;
1149 struct ofproto_ipfix_flow_exporter_options
*fe_opts
= NULL
;
1150 size_t n_fe_opts
= 0;
1152 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1153 if (ovsrec_fscs_is_valid(fe_cfg
, br
)) {
1158 if (!valid_be_cfg
&& n_fe_opts
== 0) {
1159 ofproto_set_ipfix(br
->ofproto
, NULL
, NULL
, 0);
1164 memset(&be_opts
, 0, sizeof be_opts
);
1166 sset_init(&be_opts
.targets
);
1167 sset_add_array(&be_opts
.targets
, be_cfg
->targets
, be_cfg
->n_targets
);
1169 if (be_cfg
->sampling
) {
1170 be_opts
.sampling_rate
= *be_cfg
->sampling
;
1172 be_opts
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1174 if (be_cfg
->obs_domain_id
) {
1175 be_opts
.obs_domain_id
= *be_cfg
->obs_domain_id
;
1177 if (be_cfg
->obs_point_id
) {
1178 be_opts
.obs_point_id
= *be_cfg
->obs_point_id
;
1180 if (be_cfg
->cache_active_timeout
) {
1181 be_opts
.cache_active_timeout
= *be_cfg
->cache_active_timeout
;
1183 if (be_cfg
->cache_max_flows
) {
1184 be_opts
.cache_max_flows
= *be_cfg
->cache_max_flows
;
1187 be_opts
.enable_tunnel_sampling
= smap_get_bool(&be_cfg
->other_config
,
1188 "enable-tunnel-sampling", true);
1190 be_opts
.enable_input_sampling
= !smap_get_bool(&be_cfg
->other_config
,
1191 "enable-input-sampling", false);
1193 be_opts
.enable_output_sampling
= !smap_get_bool(&be_cfg
->other_config
,
1194 "enable-output-sampling", false);
1197 if (n_fe_opts
> 0) {
1198 struct ofproto_ipfix_flow_exporter_options
*opts
;
1199 fe_opts
= xcalloc(n_fe_opts
, sizeof *fe_opts
);
1201 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1202 if (ovsrec_fscs_is_valid(fe_cfg
, br
)) {
1203 opts
->collector_set_id
= fe_cfg
->id
;
1204 sset_init(&opts
->targets
);
1205 sset_add_array(&opts
->targets
, fe_cfg
->ipfix
->targets
,
1206 fe_cfg
->ipfix
->n_targets
);
1207 opts
->cache_active_timeout
= fe_cfg
->ipfix
->cache_active_timeout
1208 ? *fe_cfg
->ipfix
->cache_active_timeout
: 0;
1209 opts
->cache_max_flows
= fe_cfg
->ipfix
->cache_max_flows
1210 ? *fe_cfg
->ipfix
->cache_max_flows
: 0;
1216 ofproto_set_ipfix(br
->ofproto
, valid_be_cfg
? &be_opts
: NULL
, fe_opts
,
1220 sset_destroy(&be_opts
.targets
);
1223 if (n_fe_opts
> 0) {
1224 struct ofproto_ipfix_flow_exporter_options
*opts
= fe_opts
;
1226 for (i
= 0; i
< n_fe_opts
; i
++) {
1227 sset_destroy(&opts
->targets
);
1235 port_configure_stp(const struct ofproto
*ofproto
, struct port
*port
,
1236 struct ofproto_port_stp_settings
*port_s
,
1237 int *port_num_counter
, unsigned long *port_num_bitmap
)
1239 const char *config_str
;
1240 struct iface
*iface
;
1242 if (!smap_get_bool(&port
->cfg
->other_config
, "stp-enable", true)) {
1243 port_s
->enable
= false;
1246 port_s
->enable
= true;
1249 /* STP over bonds is not supported. */
1250 if (!list_is_singleton(&port
->ifaces
)) {
1251 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1253 port_s
->enable
= false;
1257 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
1259 /* Internal ports shouldn't participate in spanning tree, so
1261 if (!strcmp(iface
->type
, "internal")) {
1262 VLOG_DBG("port %s: disable STP on internal ports", port
->name
);
1263 port_s
->enable
= false;
1267 /* STP on mirror output ports is not supported. */
1268 if (ofproto_is_mirror_output_bundle(ofproto
, port
)) {
1269 VLOG_DBG("port %s: disable STP on mirror ports", port
->name
);
1270 port_s
->enable
= false;
1274 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-num");
1276 unsigned long int port_num
= strtoul(config_str
, NULL
, 0);
1277 int port_idx
= port_num
- 1;
1279 if (port_num
< 1 || port_num
> STP_MAX_PORTS
) {
1280 VLOG_ERR("port %s: invalid stp-port-num", port
->name
);
1281 port_s
->enable
= false;
1285 if (bitmap_is_set(port_num_bitmap
, port_idx
)) {
1286 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1287 port
->name
, port_num
);
1288 port_s
->enable
= false;
1291 bitmap_set1(port_num_bitmap
, port_idx
);
1292 port_s
->port_num
= port_idx
;
1294 if (*port_num_counter
>= STP_MAX_PORTS
) {
1295 VLOG_ERR("port %s: too many STP ports, disabling", port
->name
);
1296 port_s
->enable
= false;
1300 port_s
->port_num
= (*port_num_counter
)++;
1303 config_str
= smap_get(&port
->cfg
->other_config
, "stp-path-cost");
1305 port_s
->path_cost
= strtoul(config_str
, NULL
, 10);
1307 enum netdev_features current
;
1310 netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1311 mbps
= netdev_features_to_bps(current
, 100 * 1000 * 1000) / 1000000;
1312 port_s
->path_cost
= stp_convert_speed_to_cost(mbps
);
1315 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-priority");
1317 port_s
->priority
= strtoul(config_str
, NULL
, 0);
1319 port_s
->priority
= STP_DEFAULT_PORT_PRIORITY
;
1324 port_configure_rstp(const struct ofproto
*ofproto
, struct port
*port
,
1325 struct ofproto_port_rstp_settings
*port_s
, int *port_num_counter
)
1327 const char *config_str
;
1328 struct iface
*iface
;
1330 if (!smap_get_bool(&port
->cfg
->other_config
, "rstp-enable", true)) {
1331 port_s
->enable
= false;
1334 port_s
->enable
= true;
1337 /* RSTP over bonds is not supported. */
1338 if (!list_is_singleton(&port
->ifaces
)) {
1339 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1341 port_s
->enable
= false;
1345 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
1347 /* Internal ports shouldn't participate in spanning tree, so
1349 if (!strcmp(iface
->type
, "internal")) {
1350 VLOG_DBG("port %s: disable RSTP on internal ports", port
->name
);
1351 port_s
->enable
= false;
1355 /* RSTP on mirror output ports is not supported. */
1356 if (ofproto_is_mirror_output_bundle(ofproto
, port
)) {
1357 VLOG_DBG("port %s: disable RSTP on mirror ports", port
->name
);
1358 port_s
->enable
= false;
1362 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-port-num");
1364 unsigned long int port_num
= strtoul(config_str
, NULL
, 0);
1365 if (port_num
< 1 || port_num
> RSTP_MAX_PORTS
) {
1366 VLOG_ERR("port %s: invalid rstp-port-num", port
->name
);
1367 port_s
->enable
= false;
1370 port_s
->port_num
= port_num
;
1372 if (*port_num_counter
>= RSTP_MAX_PORTS
) {
1373 VLOG_ERR("port %s: too many RSTP ports, disabling", port
->name
);
1374 port_s
->enable
= false;
1377 /* If rstp-port-num is not specified, use 0.
1378 * rstp_port_set_port_number() will look for the first free one. */
1379 port_s
->port_num
= 0;
1382 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-path-cost");
1384 port_s
->path_cost
= strtoul(config_str
, NULL
, 10);
1386 enum netdev_features current
;
1389 netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1390 mbps
= netdev_features_to_bps(current
, 100 * 1000 * 1000) / 1000000;
1391 port_s
->path_cost
= rstp_convert_speed_to_cost(mbps
);
1394 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-port-priority");
1396 port_s
->priority
= strtoul(config_str
, NULL
, 0);
1398 port_s
->priority
= RSTP_DEFAULT_PORT_PRIORITY
;
1401 config_str
= smap_get(&port
->cfg
->other_config
, "rstp-admin-p2p-mac");
1403 port_s
->admin_p2p_mac_state
= strtoul(config_str
, NULL
, 0);
1405 port_s
->admin_p2p_mac_state
= RSTP_ADMIN_P2P_MAC_FORCE_TRUE
;
1408 port_s
->admin_port_state
= smap_get_bool(&port
->cfg
->other_config
,
1409 "rstp-admin-port-state", true);
1411 port_s
->admin_edge_port
= smap_get_bool(&port
->cfg
->other_config
,
1412 "rstp-port-admin-edge", false);
1413 port_s
->auto_edge
= smap_get_bool(&port
->cfg
->other_config
,
1414 "rstp-port-auto-edge", true);
1415 port_s
->mcheck
= smap_get_bool(&port
->cfg
->other_config
,
1416 "rstp-port-mcheck", false);
1419 /* Set spanning tree configuration on 'br'. */
1421 bridge_configure_stp(struct bridge
*br
)
1423 struct ofproto_rstp_status rstp_status
;
1425 ofproto_get_rstp_status(br
->ofproto
, &rstp_status
);
1426 if (!br
->cfg
->stp_enable
) {
1427 ofproto_set_stp(br
->ofproto
, NULL
);
1428 } else if (rstp_status
.enabled
) {
1429 /* Do not activate STP if RSTP is enabled. */
1430 VLOG_ERR("STP cannot be enabled if RSTP is running.");
1431 ofproto_set_stp(br
->ofproto
, NULL
);
1432 ovsrec_bridge_set_stp_enable(br
->cfg
, false);
1434 struct ofproto_stp_settings br_s
;
1435 const char *config_str
;
1437 int port_num_counter
;
1438 unsigned long *port_num_bitmap
;
1440 config_str
= smap_get(&br
->cfg
->other_config
, "stp-system-id");
1442 uint8_t ea
[ETH_ADDR_LEN
];
1444 if (eth_addr_from_string(config_str
, ea
)) {
1445 br_s
.system_id
= eth_addr_to_uint64(ea
);
1447 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1448 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1449 "to "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(br
->ea
));
1452 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1455 config_str
= smap_get(&br
->cfg
->other_config
, "stp-priority");
1457 br_s
.priority
= strtoul(config_str
, NULL
, 0);
1459 br_s
.priority
= STP_DEFAULT_BRIDGE_PRIORITY
;
1462 config_str
= smap_get(&br
->cfg
->other_config
, "stp-hello-time");
1464 br_s
.hello_time
= strtoul(config_str
, NULL
, 10) * 1000;
1466 br_s
.hello_time
= STP_DEFAULT_HELLO_TIME
;
1469 config_str
= smap_get(&br
->cfg
->other_config
, "stp-max-age");
1471 br_s
.max_age
= strtoul(config_str
, NULL
, 10) * 1000;
1473 br_s
.max_age
= STP_DEFAULT_MAX_AGE
;
1476 config_str
= smap_get(&br
->cfg
->other_config
, "stp-forward-delay");
1478 br_s
.fwd_delay
= strtoul(config_str
, NULL
, 10) * 1000;
1480 br_s
.fwd_delay
= STP_DEFAULT_FWD_DELAY
;
1483 /* Configure STP on the bridge. */
1484 if (ofproto_set_stp(br
->ofproto
, &br_s
)) {
1485 VLOG_ERR("bridge %s: could not enable STP", br
->name
);
1489 /* Users must either set the port number with the "stp-port-num"
1490 * configuration on all ports or none. If manual configuration
1491 * is not done, then we allocate them sequentially. */
1492 port_num_counter
= 0;
1493 port_num_bitmap
= bitmap_allocate(STP_MAX_PORTS
);
1494 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1495 struct ofproto_port_stp_settings port_s
;
1496 struct iface
*iface
;
1498 port_configure_stp(br
->ofproto
, port
, &port_s
,
1499 &port_num_counter
, port_num_bitmap
);
1501 /* As bonds are not supported, just apply configuration to
1502 * all interfaces. */
1503 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1504 if (ofproto_port_set_stp(br
->ofproto
, iface
->ofp_port
,
1506 VLOG_ERR("port %s: could not enable STP", port
->name
);
1512 if (bitmap_scan(port_num_bitmap
, 1, 0, STP_MAX_PORTS
) != STP_MAX_PORTS
1513 && port_num_counter
) {
1514 VLOG_ERR("bridge %s: must manually configure all STP port "
1515 "IDs or none, disabling", br
->name
);
1516 ofproto_set_stp(br
->ofproto
, NULL
);
1518 bitmap_free(port_num_bitmap
);
1523 bridge_configure_rstp(struct bridge
*br
)
1525 struct ofproto_stp_status stp_status
;
1527 ofproto_get_stp_status(br
->ofproto
, &stp_status
);
1528 if (!br
->cfg
->rstp_enable
) {
1529 ofproto_set_rstp(br
->ofproto
, NULL
);
1530 } else if (stp_status
.enabled
) {
1531 /* Do not activate RSTP if STP is enabled. */
1532 VLOG_ERR("RSTP cannot be enabled if STP is running.");
1533 ofproto_set_rstp(br
->ofproto
, NULL
);
1534 ovsrec_bridge_set_rstp_enable(br
->cfg
, false);
1536 struct ofproto_rstp_settings br_s
;
1537 const char *config_str
;
1539 int port_num_counter
;
1541 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-address");
1543 uint8_t ea
[ETH_ADDR_LEN
];
1545 if (eth_addr_from_string(config_str
, ea
)) {
1546 br_s
.address
= eth_addr_to_uint64(ea
);
1549 br_s
.address
= eth_addr_to_uint64(br
->ea
);
1550 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1551 "to "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(br
->ea
));
1555 br_s
.address
= eth_addr_to_uint64(br
->ea
);
1558 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-priority");
1560 br_s
.priority
= strtoul(config_str
, NULL
, 0);
1562 br_s
.priority
= RSTP_DEFAULT_PRIORITY
;
1565 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-ageing-time");
1567 br_s
.ageing_time
= strtoul(config_str
, NULL
, 0);
1569 br_s
.ageing_time
= RSTP_DEFAULT_AGEING_TIME
;
1572 config_str
= smap_get(&br
->cfg
->other_config
,
1573 "rstp-force-protocol-version");
1575 br_s
.force_protocol_version
= strtoul(config_str
, NULL
, 0);
1577 br_s
.force_protocol_version
= FPV_DEFAULT
;
1580 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-max-age");
1582 br_s
.bridge_max_age
= strtoul(config_str
, NULL
, 10);
1584 br_s
.bridge_max_age
= RSTP_DEFAULT_BRIDGE_MAX_AGE
;
1587 config_str
= smap_get(&br
->cfg
->other_config
, "rstp-forward-delay");
1589 br_s
.bridge_forward_delay
= strtoul(config_str
, NULL
, 10);
1591 br_s
.bridge_forward_delay
= RSTP_DEFAULT_BRIDGE_FORWARD_DELAY
;
1594 config_str
= smap_get(&br
->cfg
->other_config
,
1595 "rstp-transmit-hold-count");
1597 br_s
.transmit_hold_count
= strtoul(config_str
, NULL
, 10);
1599 br_s
.transmit_hold_count
= RSTP_DEFAULT_TRANSMIT_HOLD_COUNT
;
1602 /* Configure RSTP on the bridge. */
1603 if (ofproto_set_rstp(br
->ofproto
, &br_s
)) {
1604 VLOG_ERR("bridge %s: could not enable RSTP", br
->name
);
1608 port_num_counter
= 0;
1609 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1610 struct ofproto_port_rstp_settings port_s
;
1611 struct iface
*iface
;
1613 port_configure_rstp(br
->ofproto
, port
, &port_s
,
1616 /* As bonds are not supported, just apply configuration to
1617 * all interfaces. */
1618 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1619 if (ofproto_port_set_rstp(br
->ofproto
, iface
->ofp_port
,
1621 VLOG_ERR("port %s: could not enable RSTP", port
->name
);
1630 bridge_has_bond_fake_iface(const struct bridge
*br
, const char *name
)
1632 const struct port
*port
= port_lookup(br
, name
);
1633 return port
&& port_is_bond_fake_iface(port
);
1637 port_is_bond_fake_iface(const struct port
*port
)
1639 return port
->cfg
->bond_fake_iface
&& !list_is_short(&port
->ifaces
);
1643 add_del_bridges(const struct ovsrec_open_vswitch
*cfg
)
1645 struct bridge
*br
, *next
;
1646 struct shash new_br
;
1649 /* Collect new bridges' names and types. */
1650 shash_init(&new_br
);
1651 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1652 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1653 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1655 if (strchr(br_cfg
->name
, '/')) {
1656 /* Prevent remote ovsdb-server users from accessing arbitrary
1657 * directories, e.g. consider a bridge named "../../../etc/". */
1658 VLOG_WARN_RL(&rl
, "ignoring bridge with invalid name \"%s\"",
1660 } else if (!shash_add_once(&new_br
, br_cfg
->name
, br_cfg
)) {
1661 VLOG_WARN_RL(&rl
, "bridge %s specified twice", br_cfg
->name
);
1665 /* Get rid of deleted bridges or those whose types have changed.
1666 * Update 'cfg' of bridges that still exist. */
1667 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
1668 br
->cfg
= shash_find_data(&new_br
, br
->name
);
1669 if (!br
->cfg
|| strcmp(br
->type
, ofproto_normalize_type(
1670 br
->cfg
->datapath_type
))) {
1675 /* Add new bridges. */
1676 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1677 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1678 struct bridge
*br
= bridge_lookup(br_cfg
->name
);
1680 bridge_create(br_cfg
);
1684 shash_destroy(&new_br
);
1687 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1688 * Returns 0 if successful, otherwise a positive errno value. */
1690 iface_set_netdev_config(const struct ovsrec_interface
*iface_cfg
,
1691 struct netdev
*netdev
, char **errp
)
1693 return netdev_set_config(netdev
, &iface_cfg
->options
, errp
);
1696 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1697 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1699 * If successful, returns 0 and stores the network device in '*netdevp'. On
1700 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1702 iface_do_create(const struct bridge
*br
,
1703 const struct ovsrec_interface
*iface_cfg
,
1704 const struct ovsrec_port
*port_cfg
,
1705 ofp_port_t
*ofp_portp
, struct netdev
**netdevp
,
1708 struct netdev
*netdev
= NULL
;
1711 if (netdev_is_reserved_name(iface_cfg
->name
)) {
1712 VLOG_WARN("could not create interface %s, name is reserved",
1718 error
= netdev_open(iface_cfg
->name
,
1719 iface_get_type(iface_cfg
, br
->cfg
), &netdev
);
1721 VLOG_WARN_BUF(errp
, "could not open network device %s (%s)",
1722 iface_cfg
->name
, ovs_strerror(error
));
1726 error
= iface_set_netdev_config(iface_cfg
, netdev
, errp
);
1731 *ofp_portp
= iface_pick_ofport(iface_cfg
);
1732 error
= ofproto_port_add(br
->ofproto
, netdev
, ofp_portp
);
1737 VLOG_INFO("bridge %s: added interface %s on port %d",
1738 br
->name
, iface_cfg
->name
, *ofp_portp
);
1740 if (port_cfg
->vlan_mode
&& !strcmp(port_cfg
->vlan_mode
, "splinter")) {
1741 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
1749 netdev_close(netdev
);
1753 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1754 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1755 * automatically allocated for the iface. Takes ownership of and
1756 * deallocates 'if_cfg'.
1758 * Return true if an iface is successfully created, false otherwise. */
1760 iface_create(struct bridge
*br
, const struct ovsrec_interface
*iface_cfg
,
1761 const struct ovsrec_port
*port_cfg
)
1763 struct netdev
*netdev
;
1764 struct iface
*iface
;
1765 ofp_port_t ofp_port
;
1770 /* Do the bits that can fail up front. */
1771 ovs_assert(!iface_lookup(br
, iface_cfg
->name
));
1772 error
= iface_do_create(br
, iface_cfg
, port_cfg
, &ofp_port
, &netdev
, &errp
);
1774 iface_clear_db_record(iface_cfg
, errp
);
1779 /* Get or create the port structure. */
1780 port
= port_lookup(br
, port_cfg
->name
);
1782 port
= port_create(br
, port_cfg
);
1785 /* Create the iface structure. */
1786 iface
= xzalloc(sizeof *iface
);
1787 list_push_back(&port
->ifaces
, &iface
->port_elem
);
1788 hmap_insert(&br
->iface_by_name
, &iface
->name_node
,
1789 hash_string(iface_cfg
->name
, 0));
1791 iface
->name
= xstrdup(iface_cfg
->name
);
1792 iface
->ofp_port
= ofp_port
;
1793 iface
->netdev
= netdev
;
1794 iface
->type
= iface_get_type(iface_cfg
, br
->cfg
);
1795 iface
->cfg
= iface_cfg
;
1796 hmap_insert(&br
->ifaces
, &iface
->ofp_port_node
,
1797 hash_ofp_port(ofp_port
));
1799 /* Populate initial status in database. */
1800 iface_refresh_stats(iface
);
1801 iface_refresh_netdev_status(iface
);
1803 /* Add bond fake iface if necessary. */
1804 if (port_is_bond_fake_iface(port
)) {
1805 struct ofproto_port ofproto_port
;
1807 if (ofproto_port_query_by_name(br
->ofproto
, port
->name
,
1809 struct netdev
*netdev
;
1812 error
= netdev_open(port
->name
, "internal", &netdev
);
1814 ofp_port_t fake_ofp_port
= OFPP_NONE
;
1815 ofproto_port_add(br
->ofproto
, netdev
, &fake_ofp_port
);
1816 netdev_close(netdev
);
1818 VLOG_WARN("could not open network device %s (%s)",
1819 port
->name
, ovs_strerror(error
));
1822 /* Already exists, nothing to do. */
1823 ofproto_port_destroy(&ofproto_port
);
1830 /* Set forward BPDU option. */
1832 bridge_configure_forward_bpdu(struct bridge
*br
)
1834 ofproto_set_forward_bpdu(br
->ofproto
,
1835 smap_get_bool(&br
->cfg
->other_config
,
1840 /* Set MAC learning table configuration for 'br'. */
1842 bridge_configure_mac_table(struct bridge
*br
)
1844 const char *idle_time_str
;
1847 const char *mac_table_size_str
;
1850 idle_time_str
= smap_get(&br
->cfg
->other_config
, "mac-aging-time");
1851 idle_time
= (idle_time_str
&& atoi(idle_time_str
)
1852 ? atoi(idle_time_str
)
1853 : MAC_ENTRY_DEFAULT_IDLE_TIME
);
1855 mac_table_size_str
= smap_get(&br
->cfg
->other_config
, "mac-table-size");
1856 mac_table_size
= (mac_table_size_str
&& atoi(mac_table_size_str
)
1857 ? atoi(mac_table_size_str
)
1860 ofproto_set_mac_table_config(br
->ofproto
, idle_time
, mac_table_size
);
1863 /* Set multicast snooping table configuration for 'br'. */
1865 bridge_configure_mcast_snooping(struct bridge
*br
)
1867 if (!br
->cfg
->mcast_snooping_enable
) {
1868 ofproto_set_mcast_snooping(br
->ofproto
, NULL
);
1871 struct ofproto_mcast_snooping_settings br_s
;
1872 const char *idle_time_str
;
1873 const char *max_entries_str
;
1875 idle_time_str
= smap_get(&br
->cfg
->other_config
,
1876 "mcast-snooping-aging-time");
1877 br_s
.idle_time
= (idle_time_str
&& atoi(idle_time_str
)
1878 ? atoi(idle_time_str
)
1879 : MCAST_ENTRY_DEFAULT_IDLE_TIME
);
1881 max_entries_str
= smap_get(&br
->cfg
->other_config
,
1882 "mcast-snooping-table-size");
1883 br_s
.max_entries
= (max_entries_str
&& atoi(max_entries_str
)
1884 ? atoi(max_entries_str
)
1885 : MCAST_DEFAULT_MAX_ENTRIES
);
1887 br_s
.flood_unreg
= !smap_get_bool(&br
->cfg
->other_config
,
1888 "mcast-snooping-disable-flood-unregistered",
1891 /* Configure multicast snooping on the bridge */
1892 if (ofproto_set_mcast_snooping(br
->ofproto
, &br_s
)) {
1893 VLOG_ERR("bridge %s: could not enable multicast snooping",
1898 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1899 bool flood
= smap_get_bool(&port
->cfg
->other_config
,
1900 "mcast-snooping-flood", false);
1901 if (ofproto_port_set_mcast_snooping(br
->ofproto
, port
, flood
)) {
1902 VLOG_ERR("port %s: could not configure mcast snooping",
1910 find_local_hw_addr(const struct bridge
*br
, uint8_t ea
[ETH_ADDR_LEN
],
1911 const struct port
*fake_br
, struct iface
**hw_addr_iface
)
1913 struct hmapx mirror_output_ports
;
1915 bool found_addr
= false;
1919 /* Mirror output ports don't participate in picking the local hardware
1920 * address. ofproto can't help us find out whether a given port is a
1921 * mirror output because we haven't configured mirrors yet, so we need to
1922 * accumulate them ourselves. */
1923 hmapx_init(&mirror_output_ports
);
1924 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
1925 struct ovsrec_mirror
*m
= br
->cfg
->mirrors
[i
];
1926 if (m
->output_port
) {
1927 hmapx_add(&mirror_output_ports
, m
->output_port
);
1931 /* Otherwise choose the minimum non-local MAC address among all of the
1933 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1934 uint8_t iface_ea
[ETH_ADDR_LEN
];
1935 struct iface
*candidate
;
1936 struct iface
*iface
;
1938 /* Mirror output ports don't participate. */
1939 if (hmapx_contains(&mirror_output_ports
, port
->cfg
)) {
1943 /* Choose the MAC address to represent the port. */
1945 if (port
->cfg
->mac
&& eth_addr_from_string(port
->cfg
->mac
, iface_ea
)) {
1946 /* Find the interface with this Ethernet address (if any) so that
1947 * we can provide the correct devname to the caller. */
1948 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1949 uint8_t candidate_ea
[ETH_ADDR_LEN
];
1950 if (!netdev_get_etheraddr(candidate
->netdev
, candidate_ea
)
1951 && eth_addr_equals(iface_ea
, candidate_ea
)) {
1956 /* Choose the interface whose MAC address will represent the port.
1957 * The Linux kernel bonding code always chooses the MAC address of
1958 * the first slave added to a bond, and the Fedora networking
1959 * scripts always add slaves to a bond in alphabetical order, so
1960 * for compatibility we choose the interface with the name that is
1961 * first in alphabetical order. */
1962 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1963 if (!iface
|| strcmp(candidate
->name
, iface
->name
) < 0) {
1968 /* The local port doesn't count (since we're trying to choose its
1969 * MAC address anyway). */
1970 if (iface
->ofp_port
== OFPP_LOCAL
) {
1974 /* For fake bridges we only choose from ports with the same tag */
1975 if (fake_br
&& fake_br
->cfg
&& fake_br
->cfg
->tag
) {
1976 if (!port
->cfg
->tag
) {
1979 if (*port
->cfg
->tag
!= *fake_br
->cfg
->tag
) {
1985 error
= netdev_get_etheraddr(iface
->netdev
, iface_ea
);
1991 /* Compare against our current choice. */
1992 if (!eth_addr_is_multicast(iface_ea
) &&
1993 !eth_addr_is_local(iface_ea
) &&
1994 !eth_addr_is_reserved(iface_ea
) &&
1995 !eth_addr_is_zero(iface_ea
) &&
1996 (!found_addr
|| eth_addr_compare_3way(iface_ea
, ea
) < 0))
1998 memcpy(ea
, iface_ea
, ETH_ADDR_LEN
);
1999 *hw_addr_iface
= iface
;
2005 memcpy(ea
, br
->default_ea
, ETH_ADDR_LEN
);
2006 *hw_addr_iface
= NULL
;
2009 hmapx_destroy(&mirror_output_ports
);
2013 bridge_pick_local_hw_addr(struct bridge
*br
, uint8_t ea
[ETH_ADDR_LEN
],
2014 struct iface
**hw_addr_iface
)
2017 *hw_addr_iface
= NULL
;
2019 /* Did the user request a particular MAC? */
2020 hwaddr
= smap_get(&br
->cfg
->other_config
, "hwaddr");
2021 if (hwaddr
&& eth_addr_from_string(hwaddr
, ea
)) {
2022 if (eth_addr_is_multicast(ea
)) {
2023 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2024 "address "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(ea
));
2025 } else if (eth_addr_is_zero(ea
)) {
2026 VLOG_ERR("bridge %s: cannot set MAC address to zero", br
->name
);
2032 /* Find a local hw address */
2033 find_local_hw_addr(br
, ea
, NULL
, hw_addr_iface
);
2036 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2037 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2038 * an interface on 'br', then that interface must be passed in as
2039 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2040 * 'hw_addr_iface' must be passed in as a null pointer. */
2042 bridge_pick_datapath_id(struct bridge
*br
,
2043 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
2044 struct iface
*hw_addr_iface
)
2047 * The procedure for choosing a bridge MAC address will, in the most
2048 * ordinary case, also choose a unique MAC that we can use as a datapath
2049 * ID. In some special cases, though, multiple bridges will end up with
2050 * the same MAC address. This is OK for the bridges, but it will confuse
2051 * the OpenFlow controller, because each datapath needs a unique datapath
2054 * Datapath IDs must be unique. It is also very desirable that they be
2055 * stable from one run to the next, so that policy set on a datapath
2058 const char *datapath_id
;
2061 datapath_id
= smap_get(&br
->cfg
->other_config
, "datapath-id");
2062 if (datapath_id
&& dpid_from_string(datapath_id
, &dpid
)) {
2066 if (!hw_addr_iface
) {
2068 * A purely internal bridge, that is, one that has no non-virtual
2069 * network devices on it at all, is difficult because it has no
2070 * natural unique identifier at all.
2072 * When the host is a XenServer, we handle this case by hashing the
2073 * host's UUID with the name of the bridge. Names of bridges are
2074 * persistent across XenServer reboots, although they can be reused if
2075 * an internal network is destroyed and then a new one is later
2076 * created, so this is fairly effective.
2078 * When the host is not a XenServer, we punt by using a random MAC
2079 * address on each run.
2081 const char *host_uuid
= xenserver_get_host_uuid();
2083 char *combined
= xasprintf("%s,%s", host_uuid
, br
->name
);
2084 dpid
= dpid_from_hash(combined
, strlen(combined
));
2090 return eth_addr_to_uint64(bridge_ea
);
2094 dpid_from_hash(const void *data
, size_t n
)
2096 uint8_t hash
[SHA1_DIGEST_SIZE
];
2098 BUILD_ASSERT_DECL(sizeof hash
>= ETH_ADDR_LEN
);
2099 sha1_bytes(data
, n
, hash
);
2100 eth_addr_mark_random(hash
);
2101 return eth_addr_to_uint64(hash
);
2105 iface_refresh_netdev_status(struct iface
*iface
)
2109 enum netdev_features current
;
2110 enum netdev_flags flags
;
2111 const char *link_state
;
2112 uint8_t mac
[ETH_ADDR_LEN
];
2113 int64_t bps
, mtu_64
, ifindex64
, link_resets
;
2116 if (iface_is_synthetic(iface
)) {
2120 if (iface
->change_seq
== netdev_get_change_seq(iface
->netdev
)
2121 && !status_txn_try_again
) {
2125 iface
->change_seq
= netdev_get_change_seq(iface
->netdev
);
2129 if (!netdev_get_status(iface
->netdev
, &smap
)) {
2130 ovsrec_interface_set_status(iface
->cfg
, &smap
);
2132 ovsrec_interface_set_status(iface
->cfg
, NULL
);
2135 smap_destroy(&smap
);
2137 error
= netdev_get_flags(iface
->netdev
, &flags
);
2139 const char *state
= flags
& NETDEV_UP
? "up" : "down";
2141 ovsrec_interface_set_admin_state(iface
->cfg
, state
);
2143 ovsrec_interface_set_admin_state(iface
->cfg
, NULL
);
2146 link_state
= netdev_get_carrier(iface
->netdev
) ? "up" : "down";
2147 ovsrec_interface_set_link_state(iface
->cfg
, link_state
);
2149 link_resets
= netdev_get_carrier_resets(iface
->netdev
);
2150 ovsrec_interface_set_link_resets(iface
->cfg
, &link_resets
, 1);
2152 error
= netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
2153 bps
= !error
? netdev_features_to_bps(current
, 0) : 0;
2155 ovsrec_interface_set_duplex(iface
->cfg
,
2156 netdev_features_is_full_duplex(current
)
2158 ovsrec_interface_set_link_speed(iface
->cfg
, &bps
, 1);
2160 ovsrec_interface_set_duplex(iface
->cfg
, NULL
);
2161 ovsrec_interface_set_link_speed(iface
->cfg
, NULL
, 0);
2164 error
= netdev_get_mtu(iface
->netdev
, &mtu
);
2167 ovsrec_interface_set_mtu(iface
->cfg
, &mtu_64
, 1);
2169 ovsrec_interface_set_mtu(iface
->cfg
, NULL
, 0);
2172 error
= netdev_get_etheraddr(iface
->netdev
, mac
);
2174 char mac_string
[32];
2176 sprintf(mac_string
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
2177 ovsrec_interface_set_mac_in_use(iface
->cfg
, mac_string
);
2179 ovsrec_interface_set_mac_in_use(iface
->cfg
, NULL
);
2182 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2183 * if there is no valid ifindex number. */
2184 ifindex64
= netdev_get_ifindex(iface
->netdev
);
2185 if (ifindex64
< 0) {
2188 ovsrec_interface_set_ifindex(iface
->cfg
, &ifindex64
, 1);
2192 iface_refresh_ofproto_status(struct iface
*iface
)
2196 if (iface_is_synthetic(iface
)) {
2200 current
= ofproto_port_is_lacp_current(iface
->port
->bridge
->ofproto
,
2204 ovsrec_interface_set_lacp_current(iface
->cfg
, &bl
, 1);
2206 ovsrec_interface_set_lacp_current(iface
->cfg
, NULL
, 0);
2209 if (ofproto_port_cfm_status_changed(iface
->port
->bridge
->ofproto
,
2211 || status_txn_try_again
) {
2212 iface_refresh_cfm_stats(iface
);
2215 if (ofproto_port_bfd_status_changed(iface
->port
->bridge
->ofproto
,
2217 || status_txn_try_again
) {
2221 ofproto_port_get_bfd_status(iface
->port
->bridge
->ofproto
,
2222 iface
->ofp_port
, &smap
);
2223 ovsrec_interface_set_bfd_status(iface
->cfg
, &smap
);
2224 smap_destroy(&smap
);
2228 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2231 iface_refresh_cfm_stats(struct iface
*iface
)
2233 const struct ovsrec_interface
*cfg
= iface
->cfg
;
2234 struct cfm_status status
;
2237 error
= ofproto_port_get_cfm_status(iface
->port
->bridge
->ofproto
,
2238 iface
->ofp_port
, &status
);
2240 ovsrec_interface_set_cfm_fault(cfg
, NULL
, 0);
2241 ovsrec_interface_set_cfm_fault_status(cfg
, NULL
, 0);
2242 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
2243 ovsrec_interface_set_cfm_flap_count(cfg
, NULL
, 0);
2244 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
2245 ovsrec_interface_set_cfm_remote_mpids(cfg
, NULL
, 0);
2247 const char *reasons
[CFM_FAULT_N_REASONS
];
2248 int64_t cfm_health
= status
.health
;
2249 int64_t cfm_flap_count
= status
.flap_count
;
2250 bool faulted
= status
.faults
!= 0;
2253 ovsrec_interface_set_cfm_fault(cfg
, &faulted
, 1);
2256 for (i
= 0; i
< CFM_FAULT_N_REASONS
; i
++) {
2257 int reason
= 1 << i
;
2258 if (status
.faults
& reason
) {
2259 reasons
[j
++] = cfm_fault_reason_to_str(reason
);
2262 ovsrec_interface_set_cfm_fault_status(cfg
, (char **) reasons
, j
);
2264 ovsrec_interface_set_cfm_flap_count(cfg
, &cfm_flap_count
, 1);
2266 if (status
.remote_opstate
>= 0) {
2267 const char *remote_opstate
= status
.remote_opstate
? "up" : "down";
2268 ovsrec_interface_set_cfm_remote_opstate(cfg
, remote_opstate
);
2270 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
2273 ovsrec_interface_set_cfm_remote_mpids(cfg
,
2274 (const int64_t *)status
.rmps
,
2276 if (cfm_health
>= 0) {
2277 ovsrec_interface_set_cfm_health(cfg
, &cfm_health
, 1);
2279 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
2287 iface_refresh_stats(struct iface
*iface
)
2289 #define IFACE_STATS \
2290 IFACE_STAT(rx_packets, "rx_packets") \
2291 IFACE_STAT(tx_packets, "tx_packets") \
2292 IFACE_STAT(rx_bytes, "rx_bytes") \
2293 IFACE_STAT(tx_bytes, "tx_bytes") \
2294 IFACE_STAT(rx_dropped, "rx_dropped") \
2295 IFACE_STAT(tx_dropped, "tx_dropped") \
2296 IFACE_STAT(rx_errors, "rx_errors") \
2297 IFACE_STAT(tx_errors, "tx_errors") \
2298 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2299 IFACE_STAT(rx_over_errors, "rx_over_err") \
2300 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2301 IFACE_STAT(collisions, "collisions")
2303 #define IFACE_STAT(MEMBER, NAME) + 1
2304 enum { N_IFACE_STATS
= IFACE_STATS
};
2306 int64_t values
[N_IFACE_STATS
];
2307 char *keys
[N_IFACE_STATS
];
2310 struct netdev_stats stats
;
2312 if (iface_is_synthetic(iface
)) {
2316 /* Intentionally ignore return value, since errors will set 'stats' to
2317 * all-1s, and we will deal with that correctly below. */
2318 netdev_get_stats(iface
->netdev
, &stats
);
2320 /* Copy statistics into keys[] and values[]. */
2322 #define IFACE_STAT(MEMBER, NAME) \
2323 if (stats.MEMBER != UINT64_MAX) { \
2325 values[n] = stats.MEMBER; \
2330 ovs_assert(n
<= N_IFACE_STATS
);
2332 ovsrec_interface_set_statistics(iface
->cfg
, keys
, values
, n
);
2337 br_refresh_datapath_info(struct bridge
*br
)
2339 const char *version
;
2341 version
= (br
->ofproto
&& br
->ofproto
->ofproto_class
->get_datapath_version
2342 ? br
->ofproto
->ofproto_class
->get_datapath_version(br
->ofproto
)
2345 ovsrec_bridge_set_datapath_version(br
->cfg
,
2346 version
? version
: "<unknown>");
2350 br_refresh_stp_status(struct bridge
*br
)
2352 struct smap smap
= SMAP_INITIALIZER(&smap
);
2353 struct ofproto
*ofproto
= br
->ofproto
;
2354 struct ofproto_stp_status status
;
2356 if (ofproto_get_stp_status(ofproto
, &status
)) {
2360 if (!status
.enabled
) {
2361 ovsrec_bridge_set_status(br
->cfg
, NULL
);
2365 smap_add_format(&smap
, "stp_bridge_id", STP_ID_FMT
,
2366 STP_ID_ARGS(status
.bridge_id
));
2367 smap_add_format(&smap
, "stp_designated_root", STP_ID_FMT
,
2368 STP_ID_ARGS(status
.designated_root
));
2369 smap_add_format(&smap
, "stp_root_path_cost", "%d", status
.root_path_cost
);
2371 ovsrec_bridge_set_status(br
->cfg
, &smap
);
2372 smap_destroy(&smap
);
2376 port_refresh_stp_status(struct port
*port
)
2378 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2379 struct iface
*iface
;
2380 struct ofproto_port_stp_status status
;
2383 if (port_is_synthetic(port
)) {
2387 /* STP doesn't currently support bonds. */
2388 if (!list_is_singleton(&port
->ifaces
)) {
2389 ovsrec_port_set_status(port
->cfg
, NULL
);
2393 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
2394 if (ofproto_port_get_stp_status(ofproto
, iface
->ofp_port
, &status
)) {
2398 if (!status
.enabled
) {
2399 ovsrec_port_set_status(port
->cfg
, NULL
);
2403 /* Set Status column. */
2405 smap_add_format(&smap
, "stp_port_id", STP_PORT_ID_FMT
, status
.port_id
);
2406 smap_add(&smap
, "stp_state", stp_state_name(status
.state
));
2407 smap_add_format(&smap
, "stp_sec_in_state", "%u", status
.sec_in_state
);
2408 smap_add(&smap
, "stp_role", stp_role_name(status
.role
));
2409 ovsrec_port_set_status(port
->cfg
, &smap
);
2410 smap_destroy(&smap
);
2414 port_refresh_stp_stats(struct port
*port
)
2416 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2417 struct iface
*iface
;
2418 struct ofproto_port_stp_stats stats
;
2420 int64_t int_values
[3];
2422 if (port_is_synthetic(port
)) {
2426 /* STP doesn't currently support bonds. */
2427 if (!list_is_singleton(&port
->ifaces
)) {
2431 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
2432 if (ofproto_port_get_stp_stats(ofproto
, iface
->ofp_port
, &stats
)) {
2436 if (!stats
.enabled
) {
2437 ovsrec_port_set_statistics(port
->cfg
, NULL
, NULL
, 0);
2441 /* Set Statistics column. */
2442 keys
[0] = "stp_tx_count";
2443 int_values
[0] = stats
.tx_count
;
2444 keys
[1] = "stp_rx_count";
2445 int_values
[1] = stats
.rx_count
;
2446 keys
[2] = "stp_error_count";
2447 int_values
[2] = stats
.error_count
;
2449 ovsrec_port_set_statistics(port
->cfg
, keys
, int_values
,
2450 ARRAY_SIZE(int_values
));
2454 br_refresh_rstp_status(struct bridge
*br
)
2456 struct smap smap
= SMAP_INITIALIZER(&smap
);
2457 struct ofproto
*ofproto
= br
->ofproto
;
2458 struct ofproto_rstp_status status
;
2460 if (ofproto_get_rstp_status(ofproto
, &status
)) {
2463 if (!status
.enabled
) {
2464 ovsrec_bridge_set_rstp_status(br
->cfg
, NULL
);
2467 smap_add_format(&smap
, "rstp_bridge_id", RSTP_ID_FMT
,
2468 RSTP_ID_ARGS(status
.bridge_id
));
2469 smap_add_format(&smap
, "rstp_root_path_cost", "%"PRIu32
,
2470 status
.root_path_cost
);
2471 smap_add_format(&smap
, "rstp_root_id", RSTP_ID_FMT
,
2472 RSTP_ID_ARGS(status
.root_id
));
2473 smap_add_format(&smap
, "rstp_designated_id", RSTP_ID_FMT
,
2474 RSTP_ID_ARGS(status
.designated_id
));
2475 smap_add_format(&smap
, "rstp_designated_port_id", RSTP_PORT_ID_FMT
,
2476 status
.designated_port_id
);
2477 smap_add_format(&smap
, "rstp_bridge_port_id", RSTP_PORT_ID_FMT
,
2478 status
.bridge_port_id
);
2479 ovsrec_bridge_set_rstp_status(br
->cfg
, &smap
);
2480 smap_destroy(&smap
);
2484 port_refresh_rstp_status(struct port
*port
)
2486 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2487 struct iface
*iface
;
2488 struct ofproto_port_rstp_status status
;
2490 int64_t int_values
[3];
2493 if (port_is_synthetic(port
)) {
2497 /* RSTP doesn't currently support bonds. */
2498 if (!list_is_singleton(&port
->ifaces
)) {
2499 ovsrec_port_set_rstp_status(port
->cfg
, NULL
);
2503 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
2504 if (ofproto_port_get_rstp_status(ofproto
, iface
->ofp_port
, &status
)) {
2508 if (!status
.enabled
) {
2509 ovsrec_port_set_rstp_status(port
->cfg
, NULL
);
2510 ovsrec_port_set_rstp_statistics(port
->cfg
, NULL
, NULL
, 0);
2513 /* Set Status column. */
2516 smap_add_format(&smap
, "rstp_port_id", RSTP_PORT_ID_FMT
,
2518 smap_add_format(&smap
, "rstp_port_role", "%s",
2519 rstp_port_role_name(status
.role
));
2520 smap_add_format(&smap
, "rstp_port_state", "%s",
2521 rstp_state_name(status
.state
));
2522 smap_add_format(&smap
, "rstp_designated_bridge_id", RSTP_ID_FMT
,
2523 RSTP_ID_ARGS(status
.designated_bridge_id
));
2524 smap_add_format(&smap
, "rstp_designated_port_id", RSTP_PORT_ID_FMT
,
2525 status
.designated_port_id
);
2526 smap_add_format(&smap
, "rstp_designated_path_cost", "%"PRIu32
,
2527 status
.designated_path_cost
);
2529 ovsrec_port_set_rstp_status(port
->cfg
, &smap
);
2530 smap_destroy(&smap
);
2532 /* Set Statistics column. */
2533 keys
[0] = "rstp_tx_count";
2534 int_values
[0] = status
.tx_count
;
2535 keys
[1] = "rstp_rx_count";
2536 int_values
[1] = status
.rx_count
;
2537 keys
[2] = "rstp_uptime";
2538 int_values
[2] = status
.uptime
;
2539 ovsrec_port_set_rstp_statistics(port
->cfg
, keys
, int_values
,
2540 ARRAY_SIZE(int_values
));
2544 port_refresh_bond_status(struct port
*port
, bool force_update
)
2546 uint8_t mac
[ETH_ADDR_LEN
];
2548 /* Return if port is not a bond */
2549 if (list_is_singleton(&port
->ifaces
)) {
2553 if (bond_get_changed_active_slave(port
->name
, mac
, force_update
)) {
2557 ds_put_format(&mac_s
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
2558 ovsrec_port_set_bond_active_slave(port
->cfg
, ds_cstr(&mac_s
));
2564 enable_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2566 return smap_get_bool(&cfg
->other_config
, "enable-statistics", false);
2570 reconfigure_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2572 bool enable
= enable_system_stats(cfg
);
2574 system_stats_enable(enable
);
2576 ovsrec_open_vswitch_set_statistics(cfg
, NULL
);
2581 run_system_stats(void)
2583 const struct ovsrec_open_vswitch
*cfg
= ovsrec_open_vswitch_first(idl
);
2586 stats
= system_stats_run();
2588 struct ovsdb_idl_txn
*txn
;
2589 struct ovsdb_datum datum
;
2591 txn
= ovsdb_idl_txn_create(idl
);
2592 ovsdb_datum_from_smap(&datum
, stats
);
2593 ovsdb_idl_txn_write(&cfg
->header_
, &ovsrec_open_vswitch_col_statistics
,
2595 ovsdb_idl_txn_commit(txn
);
2596 ovsdb_idl_txn_destroy(txn
);
2603 ofp12_controller_role_to_str(enum ofp12_controller_role role
)
2606 case OFPCR12_ROLE_EQUAL
:
2608 case OFPCR12_ROLE_MASTER
:
2610 case OFPCR12_ROLE_SLAVE
:
2612 case OFPCR12_ROLE_NOCHANGE
:
2614 return "*** INVALID ROLE ***";
2619 refresh_controller_status(void)
2623 const struct ovsrec_controller
*cfg
;
2627 /* Accumulate status for controllers on all bridges. */
2628 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2629 ofproto_get_ofproto_controller_info(br
->ofproto
, &info
);
2632 /* Update each controller in the database with current status. */
2633 OVSREC_CONTROLLER_FOR_EACH(cfg
, idl
) {
2634 struct ofproto_controller_info
*cinfo
=
2635 shash_find_data(&info
, cfg
->target
);
2638 ovsrec_controller_set_is_connected(cfg
, cinfo
->is_connected
);
2639 ovsrec_controller_set_role(cfg
, ofp12_controller_role_to_str(
2641 ovsrec_controller_set_status(cfg
, &cinfo
->pairs
);
2643 ovsrec_controller_set_is_connected(cfg
, false);
2644 ovsrec_controller_set_role(cfg
, NULL
);
2645 ovsrec_controller_set_status(cfg
, NULL
);
2649 ofproto_free_ofproto_controller_info(&info
);
2652 /* Update interface and mirror statistics if necessary. */
2654 run_stats_update(void)
2656 static struct ovsdb_idl_txn
*stats_txn
;
2657 const struct ovsrec_open_vswitch
*cfg
= ovsrec_open_vswitch_first(idl
);
2664 /* Statistics update interval should always be greater than or equal to
2666 stats_interval
= MAX(smap_get_int(&cfg
->other_config
,
2667 "stats-update-interval",
2669 if (stats_timer_interval
!= stats_interval
) {
2670 stats_timer_interval
= stats_interval
;
2671 stats_timer
= LLONG_MIN
;
2674 if (time_msec() >= stats_timer
) {
2675 enum ovsdb_idl_txn_status status
;
2677 /* Rate limit the update. Do not start a new update if the
2678 * previous one is not done. */
2682 stats_txn
= ovsdb_idl_txn_create(idl
);
2683 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2687 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2688 struct iface
*iface
;
2690 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2691 iface_refresh_stats(iface
);
2693 port_refresh_stp_stats(port
);
2695 HMAP_FOR_EACH (m
, hmap_node
, &br
->mirrors
) {
2696 mirror_refresh_stats(m
);
2699 refresh_controller_status();
2702 status
= ovsdb_idl_txn_commit(stats_txn
);
2703 if (status
!= TXN_INCOMPLETE
) {
2704 stats_timer
= time_msec() + stats_timer_interval
;
2705 ovsdb_idl_txn_destroy(stats_txn
);
2711 /* Update bridge/port/interface status if necessary. */
2713 run_status_update(void)
2718 /* Rate limit the update. Do not start a new update if the
2719 * previous one is not done. */
2720 seq
= seq_read(connectivity_seq_get());
2721 if (seq
!= connectivity_seqno
|| status_txn_try_again
) {
2724 connectivity_seqno
= seq
;
2725 status_txn
= ovsdb_idl_txn_create(idl
);
2726 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2729 br_refresh_stp_status(br
);
2730 br_refresh_rstp_status(br
);
2731 br_refresh_datapath_info(br
);
2732 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2733 struct iface
*iface
;
2735 port_refresh_stp_status(port
);
2736 port_refresh_rstp_status(port
);
2737 port_refresh_bond_status(port
, status_txn_try_again
);
2738 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2739 iface_refresh_netdev_status(iface
);
2740 iface_refresh_ofproto_status(iface
);
2747 /* Commit the transaction and get the status. If the transaction finishes,
2748 * then destroy the transaction. Otherwise, keep it so that we can check
2749 * progress the next time that this function is called. */
2751 enum ovsdb_idl_txn_status status
;
2753 status
= ovsdb_idl_txn_commit(status_txn
);
2754 if (status
!= TXN_INCOMPLETE
) {
2755 ovsdb_idl_txn_destroy(status_txn
);
2758 /* Sets the 'status_txn_try_again' if the transaction fails. */
2759 if (status
== TXN_SUCCESS
|| status
== TXN_UNCHANGED
) {
2760 status_txn_try_again
= false;
2762 status_txn_try_again
= true;
2769 status_update_wait(void)
2771 /* This prevents the process from constantly waking up on
2772 * connectivity seq, when there is no connection to ovsdb. */
2773 if (!ovsdb_idl_has_lock(idl
)) {
2777 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2778 * transaction to complete. If the status update to database needs to be
2779 * run again (transaction fails), registers a timeout in
2780 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2781 * sequence number. */
2783 ovsdb_idl_txn_wait(status_txn
);
2784 } else if (status_txn_try_again
) {
2785 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC
);
2787 seq_wait(connectivity_seq_get(), connectivity_seqno
);
2798 /* Let each datapath type do the work that it needs to do. */
2800 ofproto_enumerate_types(&types
);
2801 SSET_FOR_EACH (type
, &types
) {
2802 ofproto_type_run(type
);
2804 sset_destroy(&types
);
2806 /* Let each bridge do the work that it needs to do. */
2807 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2808 ofproto_run(br
->ofproto
);
2815 static struct ovsrec_open_vswitch null_cfg
;
2816 const struct ovsrec_open_vswitch
*cfg
;
2818 bool vlan_splinters_changed
;
2820 ovsrec_open_vswitch_init(&null_cfg
);
2824 if (ovsdb_idl_is_lock_contended(idl
)) {
2825 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2826 struct bridge
*br
, *next_br
;
2828 VLOG_ERR_RL(&rl
, "another ovs-vswitchd process is running, "
2829 "disabling this process (pid %ld) until it goes away",
2830 (long int) getpid());
2832 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
2835 /* Since we will not be running system_stats_run() in this process
2836 * with the current situation of multiple ovs-vswitchd daemons,
2837 * disable system stats collection. */
2838 system_stats_enable(false);
2840 } else if (!ovsdb_idl_has_lock(idl
)) {
2843 cfg
= ovsrec_open_vswitch_first(idl
);
2845 /* Initialize the ofproto library. This only needs to run once, but
2846 * it must be done after the configuration is set. If the
2847 * initialization has already occurred, bridge_init_ofproto()
2848 * returns immediately. */
2849 bridge_init_ofproto(cfg
);
2851 /* Once the value of flow-restore-wait is false, we no longer should
2852 * check its value from the database. */
2853 if (cfg
&& ofproto_get_flow_restore_wait()) {
2854 ofproto_set_flow_restore_wait(smap_get_bool(&cfg
->other_config
,
2855 "flow-restore-wait", false));
2860 /* Re-configure SSL. We do this on every trip through the main loop,
2861 * instead of just when the database changes, because the contents of the
2862 * key and certificate files can change without the database changing.
2864 * We do this before bridge_reconfigure() because that function might
2865 * initiate SSL connections and thus requires SSL to be configured. */
2866 if (cfg
&& cfg
->ssl
) {
2867 const struct ovsrec_ssl
*ssl
= cfg
->ssl
;
2869 stream_ssl_set_key_and_cert(ssl
->private_key
, ssl
->certificate
);
2870 stream_ssl_set_ca_cert_file(ssl
->ca_cert
, ssl
->bootstrap_ca_cert
);
2873 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2874 * usage has changed. */
2875 vlan_splinters_changed
= false;
2876 if (vlan_splinters_enabled_anywhere
) {
2879 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2880 if (ofproto_has_vlan_usage_changed(br
->ofproto
)) {
2881 vlan_splinters_changed
= true;
2887 if (ovsdb_idl_get_seqno(idl
) != idl_seqno
|| vlan_splinters_changed
) {
2888 struct ovsdb_idl_txn
*txn
;
2890 idl_seqno
= ovsdb_idl_get_seqno(idl
);
2891 txn
= ovsdb_idl_txn_create(idl
);
2892 bridge_reconfigure(cfg
? cfg
: &null_cfg
);
2895 ovsrec_open_vswitch_set_cur_cfg(cfg
, cfg
->next_cfg
);
2898 /* If we are completing our initial configuration for this run
2899 * of ovs-vswitchd, then keep the transaction around to monitor
2900 * it for completion. */
2901 if (initial_config_done
) {
2902 /* Always sets the 'status_txn_try_again' to check again,
2903 * in case that this transaction fails. */
2904 status_txn_try_again
= true;
2905 ovsdb_idl_txn_commit(txn
);
2906 ovsdb_idl_txn_destroy(txn
);
2908 initial_config_done
= true;
2909 daemonize_txn
= txn
;
2913 if (daemonize_txn
) {
2914 enum ovsdb_idl_txn_status status
= ovsdb_idl_txn_commit(daemonize_txn
);
2915 if (status
!= TXN_INCOMPLETE
) {
2916 ovsdb_idl_txn_destroy(daemonize_txn
);
2917 daemonize_txn
= NULL
;
2919 /* ovs-vswitchd has completed initialization, so allow the
2920 * process that forked us to exit successfully. */
2921 daemonize_complete();
2923 vlog_enable_async();
2925 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name
, VERSION
);
2930 run_status_update();
2940 ovsdb_idl_wait(idl
);
2941 if (daemonize_txn
) {
2942 ovsdb_idl_txn_wait(daemonize_txn
);
2946 ofproto_enumerate_types(&types
);
2947 SSET_FOR_EACH (type
, &types
) {
2948 ofproto_type_wait(type
);
2950 sset_destroy(&types
);
2952 if (!hmap_is_empty(&all_bridges
)) {
2955 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2956 ofproto_wait(br
->ofproto
);
2959 poll_timer_wait_until(stats_timer
);
2962 status_update_wait();
2963 system_stats_wait();
2966 /* Adds some memory usage statistics for bridges into 'usage', for use with
2967 * memory_report(). */
2969 bridge_get_memory_usage(struct simap
*usage
)
2976 ofproto_enumerate_types(&types
);
2977 SSET_FOR_EACH (type
, &types
) {
2978 ofproto_type_get_memory_usage(type
, usage
);
2980 sset_destroy(&types
);
2982 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2983 ofproto_get_memory_usage(br
->ofproto
, usage
);
2987 /* QoS unixctl user interface functions. */
2989 struct qos_unixctl_show_cbdata
{
2991 struct iface
*iface
;
2995 qos_unixctl_show_queue(unsigned int queue_id
,
2996 const struct smap
*details
,
2997 struct iface
*iface
,
3000 struct netdev_queue_stats stats
;
3001 struct smap_node
*node
;
3004 ds_put_cstr(ds
, "\n");
3006 ds_put_format(ds
, "Queue %u:\n", queue_id
);
3008 ds_put_cstr(ds
, "Default:\n");
3011 SMAP_FOR_EACH (node
, details
) {
3012 ds_put_format(ds
, "\t%s: %s\n", node
->key
, node
->value
);
3015 error
= netdev_get_queue_stats(iface
->netdev
, queue_id
, &stats
);
3017 if (stats
.tx_packets
!= UINT64_MAX
) {
3018 ds_put_format(ds
, "\ttx_packets: %"PRIu64
"\n", stats
.tx_packets
);
3021 if (stats
.tx_bytes
!= UINT64_MAX
) {
3022 ds_put_format(ds
, "\ttx_bytes: %"PRIu64
"\n", stats
.tx_bytes
);
3025 if (stats
.tx_errors
!= UINT64_MAX
) {
3026 ds_put_format(ds
, "\ttx_errors: %"PRIu64
"\n", stats
.tx_errors
);
3029 ds_put_format(ds
, "\tFailed to get statistics for queue %u: %s",
3030 queue_id
, ovs_strerror(error
));
3035 qos_unixctl_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
3036 const char *argv
[], void *aux OVS_UNUSED
)
3038 struct ds ds
= DS_EMPTY_INITIALIZER
;
3039 struct smap smap
= SMAP_INITIALIZER(&smap
);
3040 struct iface
*iface
;
3042 struct smap_node
*node
;
3044 iface
= iface_find(argv
[1]);
3046 unixctl_command_reply_error(conn
, "no such interface");
3050 netdev_get_qos(iface
->netdev
, &type
, &smap
);
3052 if (*type
!= '\0') {
3053 struct netdev_queue_dump dump
;
3054 struct smap details
;
3055 unsigned int queue_id
;
3057 ds_put_format(&ds
, "QoS: %s %s\n", iface
->name
, type
);
3059 SMAP_FOR_EACH (node
, &smap
) {
3060 ds_put_format(&ds
, "%s: %s\n", node
->key
, node
->value
);
3063 smap_init(&details
);
3064 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
3065 qos_unixctl_show_queue(queue_id
, &details
, iface
, &ds
);
3067 smap_destroy(&details
);
3069 unixctl_command_reply(conn
, ds_cstr(&ds
));
3071 ds_put_format(&ds
, "QoS not configured on %s\n", iface
->name
);
3072 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
3075 smap_destroy(&smap
);
3079 /* Bridge reconfiguration functions. */
3081 bridge_create(const struct ovsrec_bridge
*br_cfg
)
3085 ovs_assert(!bridge_lookup(br_cfg
->name
));
3086 br
= xzalloc(sizeof *br
);
3088 br
->name
= xstrdup(br_cfg
->name
);
3089 br
->type
= xstrdup(ofproto_normalize_type(br_cfg
->datapath_type
));
3092 /* Derive the default Ethernet address from the bridge's UUID. This should
3093 * be unique and it will be stable between ovs-vswitchd runs. */
3094 memcpy(br
->default_ea
, &br_cfg
->header_
.uuid
, ETH_ADDR_LEN
);
3095 eth_addr_mark_random(br
->default_ea
);
3097 hmap_init(&br
->ports
);
3098 hmap_init(&br
->ifaces
);
3099 hmap_init(&br
->iface_by_name
);
3100 hmap_init(&br
->mirrors
);
3102 hmap_insert(&all_bridges
, &br
->node
, hash_string(br
->name
, 0));
3106 bridge_destroy(struct bridge
*br
)
3109 struct mirror
*mirror
, *next_mirror
;
3110 struct port
*port
, *next_port
;
3112 HMAP_FOR_EACH_SAFE (port
, next_port
, hmap_node
, &br
->ports
) {
3115 HMAP_FOR_EACH_SAFE (mirror
, next_mirror
, hmap_node
, &br
->mirrors
) {
3116 mirror_destroy(mirror
);
3119 hmap_remove(&all_bridges
, &br
->node
);
3120 ofproto_destroy(br
->ofproto
);
3121 hmap_destroy(&br
->ifaces
);
3122 hmap_destroy(&br
->ports
);
3123 hmap_destroy(&br
->iface_by_name
);
3124 hmap_destroy(&br
->mirrors
);
3131 static struct bridge
*
3132 bridge_lookup(const char *name
)
3136 HMAP_FOR_EACH_WITH_HASH (br
, node
, hash_string(name
, 0), &all_bridges
) {
3137 if (!strcmp(br
->name
, name
)) {
3144 /* Handle requests for a listing of all flows known by the OpenFlow
3145 * stack, including those normally hidden. */
3147 bridge_unixctl_dump_flows(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
3148 const char *argv
[], void *aux OVS_UNUSED
)
3153 br
= bridge_lookup(argv
[1]);
3155 unixctl_command_reply_error(conn
, "Unknown bridge");
3160 ofproto_get_all_flows(br
->ofproto
, &results
);
3162 unixctl_command_reply(conn
, ds_cstr(&results
));
3163 ds_destroy(&results
);
3166 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3167 * connections and reconnect. If BRIDGE is not specified, then all bridges
3168 * drop their controller connections and reconnect. */
3170 bridge_unixctl_reconnect(struct unixctl_conn
*conn
, int argc
,
3171 const char *argv
[], void *aux OVS_UNUSED
)
3175 br
= bridge_lookup(argv
[1]);
3177 unixctl_command_reply_error(conn
, "Unknown bridge");
3180 ofproto_reconnect_controllers(br
->ofproto
);
3182 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3183 ofproto_reconnect_controllers(br
->ofproto
);
3186 unixctl_command_reply(conn
, NULL
);
3190 bridge_get_controllers(const struct bridge
*br
,
3191 struct ovsrec_controller
***controllersp
)
3193 struct ovsrec_controller
**controllers
;
3194 size_t n_controllers
;
3196 controllers
= br
->cfg
->controller
;
3197 n_controllers
= br
->cfg
->n_controller
;
3199 if (n_controllers
== 1 && !strcmp(controllers
[0]->target
, "none")) {
3205 *controllersp
= controllers
;
3207 return n_controllers
;
3211 bridge_collect_wanted_ports(struct bridge
*br
,
3212 const unsigned long int *splinter_vlans
,
3213 struct shash
*wanted_ports
)
3217 shash_init(wanted_ports
);
3219 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
3220 const char *name
= br
->cfg
->ports
[i
]->name
;
3221 if (!shash_add_once(wanted_ports
, name
, br
->cfg
->ports
[i
])) {
3222 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3226 if (bridge_get_controllers(br
, NULL
)
3227 && !shash_find(wanted_ports
, br
->name
)) {
3228 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3229 br
->name
, br
->name
);
3231 ovsrec_interface_init(&br
->synth_local_iface
);
3232 ovsrec_port_init(&br
->synth_local_port
);
3234 br
->synth_local_port
.interfaces
= &br
->synth_local_ifacep
;
3235 br
->synth_local_port
.n_interfaces
= 1;
3236 br
->synth_local_port
.name
= br
->name
;
3238 br
->synth_local_iface
.name
= br
->name
;
3239 br
->synth_local_iface
.type
= "internal";
3241 br
->synth_local_ifacep
= &br
->synth_local_iface
;
3243 shash_add(wanted_ports
, br
->name
, &br
->synth_local_port
);
3246 if (splinter_vlans
) {
3247 add_vlan_splinter_ports(br
, splinter_vlans
, wanted_ports
);
3251 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3252 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3253 * 'br' needs to complete its configuration. */
3255 bridge_del_ports(struct bridge
*br
, const struct shash
*wanted_ports
)
3257 struct shash_node
*port_node
;
3258 struct port
*port
, *next
;
3260 /* Get rid of deleted ports.
3261 * Get rid of deleted interfaces on ports that still exist. */
3262 HMAP_FOR_EACH_SAFE (port
, next
, hmap_node
, &br
->ports
) {
3263 port
->cfg
= shash_find_data(wanted_ports
, port
->name
);
3267 port_del_ifaces(port
);
3271 /* Update iface->cfg and iface->type in interfaces that still exist. */
3272 SHASH_FOR_EACH (port_node
, wanted_ports
) {
3273 const struct ovsrec_port
*port
= port_node
->data
;
3276 for (i
= 0; i
< port
->n_interfaces
; i
++) {
3277 const struct ovsrec_interface
*cfg
= port
->interfaces
[i
];
3278 struct iface
*iface
= iface_lookup(br
, cfg
->name
);
3279 const char *type
= iface_get_type(cfg
, br
->cfg
);
3284 } else if (!strcmp(type
, "null")) {
3285 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3286 " may be removed in February 2013. Please email"
3287 " dev@openvswitch.org with concerns.",
3290 /* We will add new interfaces later. */
3296 /* Initializes 'oc' appropriately as a management service controller for
3299 * The caller must free oc->target when it is no longer needed. */
3301 bridge_ofproto_controller_for_mgmt(const struct bridge
*br
,
3302 struct ofproto_controller
*oc
)
3304 oc
->target
= xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br
->name
);
3305 oc
->max_backoff
= 0;
3306 oc
->probe_interval
= 60;
3307 oc
->band
= OFPROTO_OUT_OF_BAND
;
3309 oc
->burst_limit
= 0;
3310 oc
->enable_async_msgs
= true;
3314 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3316 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller
*c
,
3317 struct ofproto_controller
*oc
)
3321 oc
->target
= c
->target
;
3322 oc
->max_backoff
= c
->max_backoff
? *c
->max_backoff
/ 1000 : 8;
3323 oc
->probe_interval
= c
->inactivity_probe
? *c
->inactivity_probe
/ 1000 : 5;
3324 oc
->band
= (!c
->connection_mode
|| !strcmp(c
->connection_mode
, "in-band")
3325 ? OFPROTO_IN_BAND
: OFPROTO_OUT_OF_BAND
);
3326 oc
->rate_limit
= c
->controller_rate_limit
? *c
->controller_rate_limit
: 0;
3327 oc
->burst_limit
= (c
->controller_burst_limit
3328 ? *c
->controller_burst_limit
: 0);
3329 oc
->enable_async_msgs
= (!c
->enable_async_messages
3330 || *c
->enable_async_messages
);
3331 dscp
= smap_get_int(&c
->other_config
, "dscp", DSCP_DEFAULT
);
3332 if (dscp
< 0 || dscp
> 63) {
3333 dscp
= DSCP_DEFAULT
;
3338 /* Configures the IP stack for 'br''s local interface properly according to the
3339 * configuration in 'c'. */
3341 bridge_configure_local_iface_netdev(struct bridge
*br
,
3342 struct ovsrec_controller
*c
)
3344 struct netdev
*netdev
;
3345 struct in_addr mask
, gateway
;
3347 struct iface
*local_iface
;
3350 /* If there's no local interface or no IP address, give up. */
3351 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
3352 if (!local_iface
|| !c
->local_ip
3353 || !inet_pton(AF_INET
, c
->local_ip
, &ip
)) {
3357 /* Bring up the local interface. */
3358 netdev
= local_iface
->netdev
;
3359 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
3361 /* Configure the IP address and netmask. */
3362 if (!c
->local_netmask
3363 || !inet_pton(AF_INET
, c
->local_netmask
, &mask
)
3365 mask
.s_addr
= guess_netmask(ip
.s_addr
);
3367 if (!netdev_set_in4(netdev
, ip
, mask
)) {
3368 VLOG_INFO("bridge %s: configured IP address "IP_FMT
", netmask "IP_FMT
,
3369 br
->name
, IP_ARGS(ip
.s_addr
), IP_ARGS(mask
.s_addr
));
3372 /* Configure the default gateway. */
3373 if (c
->local_gateway
3374 && inet_pton(AF_INET
, c
->local_gateway
, &gateway
)
3375 && gateway
.s_addr
) {
3376 if (!netdev_add_router(netdev
, gateway
)) {
3377 VLOG_INFO("bridge %s: configured gateway "IP_FMT
,
3378 br
->name
, IP_ARGS(gateway
.s_addr
));
3383 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3384 * in either string are treated as equal to any number of slashes in the other,
3385 * e.g. "x///y" is equal to "x/y".
3387 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3388 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3389 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3390 * 'b' against a prefix of 'a'.
3393 equal_pathnames(const char *a
, const char *b
, size_t b_stoplen
)
3395 const char *b_start
= b
;
3397 if (b
- b_start
>= b_stoplen
) {
3399 } else if (*a
!= *b
) {
3401 } else if (*a
== '/') {
3402 a
+= strspn(a
, "/");
3403 b
+= strspn(b
, "/");
3404 } else if (*a
== '\0') {
3414 bridge_configure_remotes(struct bridge
*br
,
3415 const struct sockaddr_in
*managers
, size_t n_managers
)
3417 bool disable_in_band
;
3419 struct ovsrec_controller
**controllers
;
3420 size_t n_controllers
;
3422 enum ofproto_fail_mode fail_mode
;
3424 struct ofproto_controller
*ocs
;
3428 /* Check if we should disable in-band control on this bridge. */
3429 disable_in_band
= smap_get_bool(&br
->cfg
->other_config
, "disable-in-band",
3432 /* Set OpenFlow queue ID for in-band control. */
3433 ofproto_set_in_band_queue(br
->ofproto
,
3434 smap_get_int(&br
->cfg
->other_config
,
3435 "in-band-queue", -1));
3437 if (disable_in_band
) {
3438 ofproto_set_extra_in_band_remotes(br
->ofproto
, NULL
, 0);
3440 ofproto_set_extra_in_band_remotes(br
->ofproto
, managers
, n_managers
);
3443 n_controllers
= bridge_get_controllers(br
, &controllers
);
3445 ocs
= xmalloc((n_controllers
+ 1) * sizeof *ocs
);
3448 bridge_ofproto_controller_for_mgmt(br
, &ocs
[n_ocs
++]);
3449 for (i
= 0; i
< n_controllers
; i
++) {
3450 struct ovsrec_controller
*c
= controllers
[i
];
3452 if (!strncmp(c
->target
, "punix:", 6)
3453 || !strncmp(c
->target
, "unix:", 5)) {
3454 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3457 if (!strncmp(c
->target
, "unix:", 5)) {
3458 /* Connect to a listening socket */
3459 whitelist
= xasprintf("unix:%s/", ovs_rundir());
3460 if (strchr(c
->target
, '/') &&
3461 !equal_pathnames(c
->target
, whitelist
,
3462 strlen(whitelist
))) {
3463 /* Absolute path specified, but not in ovs_rundir */
3464 VLOG_ERR_RL(&rl
, "bridge %s: Not connecting to socket "
3465 "controller \"%s\" due to possibility for "
3466 "remote exploit. Instead, specify socket "
3467 "in whitelisted \"%s\" or connect to "
3468 "\"unix:%s/%s.mgmt\" (which is always "
3469 "available without special configuration).",
3470 br
->name
, c
->target
, whitelist
,
3471 ovs_rundir(), br
->name
);
3476 whitelist
= xasprintf("punix:%s/%s.controller",
3477 ovs_rundir(), br
->name
);
3478 if (!equal_pathnames(c
->target
, whitelist
, SIZE_MAX
)) {
3479 /* Prevent remote ovsdb-server users from accessing
3480 * arbitrary Unix domain sockets and overwriting arbitrary
3482 VLOG_ERR_RL(&rl
, "bridge %s: Not adding Unix domain socket "
3483 "controller \"%s\" due to possibility of "
3484 "overwriting local files. Instead, specify "
3485 "whitelisted \"%s\" or connect to "
3486 "\"unix:%s/%s.mgmt\" (which is always "
3487 "available without special configuration).",
3488 br
->name
, c
->target
, whitelist
,
3489 ovs_rundir(), br
->name
);
3498 bridge_configure_local_iface_netdev(br
, c
);
3499 bridge_ofproto_controller_from_ovsrec(c
, &ocs
[n_ocs
]);
3500 if (disable_in_band
) {
3501 ocs
[n_ocs
].band
= OFPROTO_OUT_OF_BAND
;
3506 ofproto_set_controllers(br
->ofproto
, ocs
, n_ocs
,
3507 bridge_get_allowed_versions(br
));
3508 free(ocs
[0].target
); /* From bridge_ofproto_controller_for_mgmt(). */
3511 /* Set the fail-mode. */
3512 fail_mode
= !br
->cfg
->fail_mode
3513 || !strcmp(br
->cfg
->fail_mode
, "standalone")
3514 ? OFPROTO_FAIL_STANDALONE
3515 : OFPROTO_FAIL_SECURE
;
3516 ofproto_set_fail_mode(br
->ofproto
, fail_mode
);
3518 /* Configure OpenFlow controller connection snooping. */
3519 if (!ofproto_has_snoops(br
->ofproto
)) {
3523 sset_add_and_free(&snoops
, xasprintf("punix:%s/%s.snoop",
3524 ovs_rundir(), br
->name
));
3525 ofproto_set_snoops(br
->ofproto
, &snoops
);
3526 sset_destroy(&snoops
);
3531 bridge_configure_tables(struct bridge
*br
)
3533 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3537 n_tables
= ofproto_get_n_tables(br
->ofproto
);
3539 for (i
= 0; i
< n_tables
; i
++) {
3540 struct ofproto_table_settings s
;
3541 bool use_default_prefixes
= true;
3544 s
.max_flows
= UINT_MAX
;
3547 s
.n_prefix_fields
= 0;
3548 memset(s
.prefix_fields
, ~0, sizeof(s
.prefix_fields
));
3550 if (j
< br
->cfg
->n_flow_tables
&& i
== br
->cfg
->key_flow_tables
[j
]) {
3551 struct ovsrec_flow_table
*cfg
= br
->cfg
->value_flow_tables
[j
++];
3554 if (cfg
->n_flow_limit
&& *cfg
->flow_limit
< UINT_MAX
) {
3555 s
.max_flows
= *cfg
->flow_limit
;
3557 if (cfg
->overflow_policy
3558 && !strcmp(cfg
->overflow_policy
, "evict")) {
3560 s
.groups
= xmalloc(cfg
->n_groups
* sizeof *s
.groups
);
3561 for (k
= 0; k
< cfg
->n_groups
; k
++) {
3562 const char *string
= cfg
->groups
[k
];
3565 msg
= mf_parse_subfield__(&s
.groups
[k
], &string
);
3567 VLOG_WARN_RL(&rl
, "bridge %s table %d: error parsing "
3568 "'groups' (%s)", br
->name
, i
, msg
);
3570 } else if (*string
) {
3571 VLOG_WARN_RL(&rl
, "bridge %s table %d: 'groups' "
3572 "element '%s' contains trailing garbage",
3573 br
->name
, i
, cfg
->groups
[k
]);
3579 /* Prefix lookup fields. */
3580 s
.n_prefix_fields
= 0;
3581 for (k
= 0; k
< cfg
->n_prefixes
; k
++) {
3582 const char *name
= cfg
->prefixes
[k
];
3583 const struct mf_field
*mf
;
3585 if (strcmp(name
, "none") == 0) {
3586 use_default_prefixes
= false;
3587 s
.n_prefix_fields
= 0;
3590 mf
= mf_from_name(name
);
3592 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3596 if (mf
->flow_be32ofs
< 0 || mf
->n_bits
% 32) {
3597 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3598 "%s", br
->name
, name
);
3601 if (s
.n_prefix_fields
>= ARRAY_SIZE(s
.prefix_fields
)) {
3602 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3603 "field not used: %s", br
->name
, name
);
3606 use_default_prefixes
= false;
3607 s
.prefix_fields
[s
.n_prefix_fields
++] = mf
->id
;
3610 if (use_default_prefixes
) {
3611 /* Use default values. */
3612 s
.n_prefix_fields
= ARRAY_SIZE(default_prefix_fields
);
3613 memcpy(s
.prefix_fields
, default_prefix_fields
,
3614 sizeof default_prefix_fields
);
3617 struct ds ds
= DS_EMPTY_INITIALIZER
;
3618 for (k
= 0; k
< s
.n_prefix_fields
; k
++) {
3620 ds_put_char(&ds
, ',');
3622 ds_put_cstr(&ds
, mf_from_id(s
.prefix_fields
[k
])->name
);
3624 if (s
.n_prefix_fields
== 0) {
3625 ds_put_cstr(&ds
, "none");
3627 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3628 br
->name
, i
, ds_cstr(&ds
));
3632 ofproto_configure_table(br
->ofproto
, i
, &s
);
3636 for (; j
< br
->cfg
->n_flow_tables
; j
++) {
3637 VLOG_WARN_RL(&rl
, "bridge %s: ignoring configuration for flow table "
3638 "%"PRId64
" not supported by this datapath", br
->name
,
3639 br
->cfg
->key_flow_tables
[j
]);
3644 bridge_configure_dp_desc(struct bridge
*br
)
3646 ofproto_set_dp_desc(br
->ofproto
,
3647 smap_get(&br
->cfg
->other_config
, "dp-desc"));
3650 /* Port functions. */
3652 static struct port
*
3653 port_create(struct bridge
*br
, const struct ovsrec_port
*cfg
)
3657 port
= xzalloc(sizeof *port
);
3659 port
->name
= xstrdup(cfg
->name
);
3661 list_init(&port
->ifaces
);
3663 hmap_insert(&br
->ports
, &port
->hmap_node
, hash_string(port
->name
, 0));
3667 /* Deletes interfaces from 'port' that are no longer configured for it. */
3669 port_del_ifaces(struct port
*port
)
3671 struct iface
*iface
, *next
;
3672 struct sset new_ifaces
;
3675 /* Collect list of new interfaces. */
3676 sset_init(&new_ifaces
);
3677 for (i
= 0; i
< port
->cfg
->n_interfaces
; i
++) {
3678 const char *name
= port
->cfg
->interfaces
[i
]->name
;
3679 const char *type
= port
->cfg
->interfaces
[i
]->type
;
3680 if (strcmp(type
, "null")) {
3681 sset_add(&new_ifaces
, name
);
3685 /* Get rid of deleted interfaces. */
3686 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
3687 if (!sset_contains(&new_ifaces
, iface
->name
)) {
3688 iface_destroy(iface
);
3692 sset_destroy(&new_ifaces
);
3696 port_destroy(struct port
*port
)
3699 struct bridge
*br
= port
->bridge
;
3700 struct iface
*iface
, *next
;
3703 ofproto_bundle_unregister(br
->ofproto
, port
);
3706 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
3707 iface_destroy__(iface
);
3710 hmap_remove(&br
->ports
, &port
->hmap_node
);
3716 static struct port
*
3717 port_lookup(const struct bridge
*br
, const char *name
)
3721 HMAP_FOR_EACH_WITH_HASH (port
, hmap_node
, hash_string(name
, 0),
3723 if (!strcmp(port
->name
, name
)) {
3731 enable_lacp(struct port
*port
, bool *activep
)
3733 if (!port
->cfg
->lacp
) {
3734 /* XXX when LACP implementation has been sufficiently tested, enable by
3735 * default and make active on bonded ports. */
3737 } else if (!strcmp(port
->cfg
->lacp
, "off")) {
3739 } else if (!strcmp(port
->cfg
->lacp
, "active")) {
3742 } else if (!strcmp(port
->cfg
->lacp
, "passive")) {
3746 VLOG_WARN("port %s: unknown LACP mode %s",
3747 port
->name
, port
->cfg
->lacp
);
3752 static struct lacp_settings
*
3753 port_configure_lacp(struct port
*port
, struct lacp_settings
*s
)
3755 const char *lacp_time
, *system_id
;
3758 if (!enable_lacp(port
, &s
->active
)) {
3762 s
->name
= port
->name
;
3764 system_id
= smap_get(&port
->cfg
->other_config
, "lacp-system-id");
3766 if (!ovs_scan(system_id
, ETH_ADDR_SCAN_FMT
,
3767 ETH_ADDR_SCAN_ARGS(s
->id
))) {
3768 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3769 " address.", port
->name
, system_id
);
3773 memcpy(s
->id
, port
->bridge
->ea
, ETH_ADDR_LEN
);
3776 if (eth_addr_is_zero(s
->id
)) {
3777 VLOG_WARN("port %s: Invalid zero LACP system ID.", port
->name
);
3781 /* Prefer bondable links if unspecified. */
3782 priority
= smap_get_int(&port
->cfg
->other_config
, "lacp-system-priority",
3784 s
->priority
= (priority
> 0 && priority
<= UINT16_MAX
3786 : UINT16_MAX
- !list_is_short(&port
->ifaces
));
3788 lacp_time
= smap_get(&port
->cfg
->other_config
, "lacp-time");
3789 s
->fast
= lacp_time
&& !strcasecmp(lacp_time
, "fast");
3791 s
->fallback_ab_cfg
= smap_get_bool(&port
->cfg
->other_config
,
3792 "lacp-fallback-ab", false);
3798 iface_configure_lacp(struct iface
*iface
, struct lacp_slave_settings
*s
)
3800 int priority
, portid
, key
;
3802 portid
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-id", 0);
3803 priority
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-priority",
3805 key
= smap_get_int(&iface
->cfg
->other_config
, "lacp-aggregation-key", 0);
3807 if (portid
<= 0 || portid
> UINT16_MAX
) {
3808 portid
= ofp_to_u16(iface
->ofp_port
);
3811 if (priority
<= 0 || priority
> UINT16_MAX
) {
3812 priority
= UINT16_MAX
;
3815 if (key
< 0 || key
> UINT16_MAX
) {
3819 s
->name
= iface
->name
;
3821 s
->priority
= priority
;
3826 port_configure_bond(struct port
*port
, struct bond_settings
*s
)
3828 const char *detect_s
;
3829 struct iface
*iface
;
3831 int miimon_interval
;
3833 s
->name
= port
->name
;
3835 if (port
->cfg
->bond_mode
) {
3836 if (!bond_mode_from_string(&s
->balance
, port
->cfg
->bond_mode
)) {
3837 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3838 port
->name
, port
->cfg
->bond_mode
,
3839 bond_mode_to_string(s
->balance
));
3842 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3844 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3845 * active-backup. At some point we should remove this warning. */
3846 VLOG_WARN_RL(&rl
, "port %s: Using the default bond_mode %s. Note that"
3847 " in previous versions, the default bond_mode was"
3848 " balance-slb", port
->name
,
3849 bond_mode_to_string(s
->balance
));
3851 if (s
->balance
== BM_SLB
&& port
->bridge
->cfg
->n_flood_vlans
) {
3852 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3853 "please use another bond type or disable flood_vlans",
3857 miimon_interval
= smap_get_int(&port
->cfg
->other_config
,
3858 "bond-miimon-interval", 0);
3859 if (miimon_interval
<= 0) {
3860 miimon_interval
= 200;
3863 detect_s
= smap_get(&port
->cfg
->other_config
, "bond-detect-mode");
3864 if (!detect_s
|| !strcmp(detect_s
, "carrier")) {
3865 miimon_interval
= 0;
3866 } else if (strcmp(detect_s
, "miimon")) {
3867 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3868 "defaulting to carrier", port
->name
, detect_s
);
3869 miimon_interval
= 0;
3872 s
->up_delay
= MAX(0, port
->cfg
->bond_updelay
);
3873 s
->down_delay
= MAX(0, port
->cfg
->bond_downdelay
);
3874 s
->basis
= smap_get_int(&port
->cfg
->other_config
, "bond-hash-basis", 0);
3875 s
->rebalance_interval
= smap_get_int(&port
->cfg
->other_config
,
3876 "bond-rebalance-interval", 10000);
3877 if (s
->rebalance_interval
&& s
->rebalance_interval
< 1000) {
3878 s
->rebalance_interval
= 1000;
3881 s
->lacp_fallback_ab_cfg
= smap_get_bool(&port
->cfg
->other_config
,
3882 "lacp-fallback-ab", false);
3884 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
3885 netdev_set_miimon_interval(iface
->netdev
, miimon_interval
);
3888 mac_s
= port
->cfg
->bond_active_slave
;
3889 if (!mac_s
|| !ovs_scan(mac_s
, ETH_ADDR_SCAN_FMT
,
3890 ETH_ADDR_SCAN_ARGS(s
->active_slave_mac
))) {
3891 /* OVSDB did not store the last active interface */
3892 memset(s
->active_slave_mac
, 0, sizeof(s
->active_slave_mac
));
3896 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3897 * instead of obtaining it from the database. */
3899 port_is_synthetic(const struct port
*port
)
3901 return ovsdb_idl_row_is_synthetic(&port
->cfg
->header_
);
3904 /* Interface functions. */
3907 iface_is_internal(const struct ovsrec_interface
*iface
,
3908 const struct ovsrec_bridge
*br
)
3910 /* The local port and "internal" ports are always "internal". */
3911 return !strcmp(iface
->type
, "internal") || !strcmp(iface
->name
, br
->name
);
3914 /* Returns the correct network device type for interface 'iface' in bridge
3917 iface_get_type(const struct ovsrec_interface
*iface
,
3918 const struct ovsrec_bridge
*br
)
3922 /* The local port always has type "internal". Other ports take
3923 * their type from the database and default to "system" if none is
3925 if (iface_is_internal(iface
, br
)) {
3928 type
= iface
->type
[0] ? iface
->type
: "system";
3931 return ofproto_port_open_type(br
->datapath_type
, type
);
3935 iface_destroy__(struct iface
*iface
)
3938 struct port
*port
= iface
->port
;
3939 struct bridge
*br
= port
->bridge
;
3941 if (br
->ofproto
&& iface
->ofp_port
!= OFPP_NONE
) {
3942 ofproto_port_unregister(br
->ofproto
, iface
->ofp_port
);
3945 if (iface
->ofp_port
!= OFPP_NONE
) {
3946 hmap_remove(&br
->ifaces
, &iface
->ofp_port_node
);
3949 list_remove(&iface
->port_elem
);
3950 hmap_remove(&br
->iface_by_name
, &iface
->name_node
);
3952 /* The user is changing configuration here, so netdev_remove needs to be
3953 * used as opposed to netdev_close */
3954 netdev_remove(iface
->netdev
);
3962 iface_destroy(struct iface
*iface
)
3965 struct port
*port
= iface
->port
;
3967 iface_destroy__(iface
);
3968 if (list_is_empty(&port
->ifaces
)) {
3974 static struct iface
*
3975 iface_lookup(const struct bridge
*br
, const char *name
)
3977 struct iface
*iface
;
3979 HMAP_FOR_EACH_WITH_HASH (iface
, name_node
, hash_string(name
, 0),
3980 &br
->iface_by_name
) {
3981 if (!strcmp(iface
->name
, name
)) {
3989 static struct iface
*
3990 iface_find(const char *name
)
3992 const struct bridge
*br
;
3994 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3995 struct iface
*iface
= iface_lookup(br
, name
);
4004 static struct iface
*
4005 iface_from_ofp_port(const struct bridge
*br
, ofp_port_t ofp_port
)
4007 struct iface
*iface
;
4009 HMAP_FOR_EACH_IN_BUCKET (iface
, ofp_port_node
, hash_ofp_port(ofp_port
),
4011 if (iface
->ofp_port
== ofp_port
) {
4018 /* Set Ethernet address of 'iface', if one is specified in the configuration
4021 iface_set_mac(const struct bridge
*br
, const struct port
*port
, struct iface
*iface
)
4023 uint8_t ea
[ETH_ADDR_LEN
], *mac
= NULL
;
4024 struct iface
*hw_addr_iface
;
4026 if (strcmp(iface
->type
, "internal")) {
4030 if (iface
->cfg
->mac
&& eth_addr_from_string(iface
->cfg
->mac
, ea
)) {
4032 } else if (port
->cfg
->fake_bridge
) {
4033 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4034 find_local_hw_addr(br
, ea
, port
, &hw_addr_iface
);
4039 if (iface
->ofp_port
== OFPP_LOCAL
) {
4040 VLOG_ERR("interface %s: ignoring mac in Interface record "
4041 "(use Bridge record to set local port's mac)",
4043 } else if (eth_addr_is_multicast(mac
)) {
4044 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4047 int error
= netdev_set_etheraddr(iface
->netdev
, mac
);
4049 VLOG_ERR("interface %s: setting MAC failed (%s)",
4050 iface
->name
, ovs_strerror(error
));
4056 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4058 iface_set_ofport(const struct ovsrec_interface
*if_cfg
, ofp_port_t ofport
)
4060 if (if_cfg
&& !ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
4061 int64_t port
= ofport
== OFPP_NONE
? -1 : ofp_to_u16(ofport
);
4062 ovsrec_interface_set_ofport(if_cfg
, &port
, 1);
4066 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4067 * sets the "ofport" field to -1.
4069 * This is appropriate when 'if_cfg''s interface cannot be created or is
4070 * otherwise invalid. */
4072 iface_clear_db_record(const struct ovsrec_interface
*if_cfg
, char *errp
)
4074 if (!ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
4075 iface_set_ofport(if_cfg
, OFPP_NONE
);
4076 ovsrec_interface_set_error(if_cfg
, errp
);
4077 ovsrec_interface_set_status(if_cfg
, NULL
);
4078 ovsrec_interface_set_admin_state(if_cfg
, NULL
);
4079 ovsrec_interface_set_duplex(if_cfg
, NULL
);
4080 ovsrec_interface_set_link_speed(if_cfg
, NULL
, 0);
4081 ovsrec_interface_set_link_state(if_cfg
, NULL
);
4082 ovsrec_interface_set_mac_in_use(if_cfg
, NULL
);
4083 ovsrec_interface_set_mtu(if_cfg
, NULL
, 0);
4084 ovsrec_interface_set_cfm_fault(if_cfg
, NULL
, 0);
4085 ovsrec_interface_set_cfm_fault_status(if_cfg
, NULL
, 0);
4086 ovsrec_interface_set_cfm_remote_mpids(if_cfg
, NULL
, 0);
4087 ovsrec_interface_set_lacp_current(if_cfg
, NULL
, 0);
4088 ovsrec_interface_set_statistics(if_cfg
, NULL
, NULL
, 0);
4089 ovsrec_interface_set_ifindex(if_cfg
, NULL
, 0);
4094 queue_ids_include(const struct ovsdb_datum
*queues
, int64_t target
)
4096 union ovsdb_atom atom
;
4098 atom
.integer
= target
;
4099 return ovsdb_datum_find_key(queues
, &atom
, OVSDB_TYPE_INTEGER
) != UINT_MAX
;
4103 iface_configure_qos(struct iface
*iface
, const struct ovsrec_qos
*qos
)
4105 struct ofpbuf queues_buf
;
4107 ofpbuf_init(&queues_buf
, 0);
4109 if (!qos
|| qos
->type
[0] == '\0' || qos
->n_queues
< 1) {
4110 netdev_set_qos(iface
->netdev
, NULL
, NULL
);
4112 const struct ovsdb_datum
*queues
;
4113 struct netdev_queue_dump dump
;
4114 unsigned int queue_id
;
4115 struct smap details
;
4119 /* Configure top-level Qos for 'iface'. */
4120 netdev_set_qos(iface
->netdev
, qos
->type
, &qos
->other_config
);
4122 /* Deconfigure queues that were deleted. */
4123 queues
= ovsrec_qos_get_queues(qos
, OVSDB_TYPE_INTEGER
,
4125 smap_init(&details
);
4126 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
4127 if (!queue_ids_include(queues
, queue_id
)) {
4128 netdev_delete_queue(iface
->netdev
, queue_id
);
4131 smap_destroy(&details
);
4133 /* Configure queues for 'iface'. */
4135 for (i
= 0; i
< qos
->n_queues
; i
++) {
4136 const struct ovsrec_queue
*queue
= qos
->value_queues
[i
];
4137 unsigned int queue_id
= qos
->key_queues
[i
];
4139 if (queue_id
== 0) {
4143 if (queue
->n_dscp
== 1) {
4144 struct ofproto_port_queue
*port_queue
;
4146 port_queue
= ofpbuf_put_uninit(&queues_buf
,
4147 sizeof *port_queue
);
4148 port_queue
->queue
= queue_id
;
4149 port_queue
->dscp
= queue
->dscp
[0];
4152 netdev_set_queue(iface
->netdev
, queue_id
, &queue
->other_config
);
4155 struct smap details
;
4157 smap_init(&details
);
4158 netdev_set_queue(iface
->netdev
, 0, &details
);
4159 smap_destroy(&details
);
4163 if (iface
->ofp_port
!= OFPP_NONE
) {
4164 const struct ofproto_port_queue
*port_queues
= ofpbuf_data(&queues_buf
);
4165 size_t n_queues
= ofpbuf_size(&queues_buf
) / sizeof *port_queues
;
4167 ofproto_port_set_queues(iface
->port
->bridge
->ofproto
, iface
->ofp_port
,
4168 port_queues
, n_queues
);
4171 netdev_set_policing(iface
->netdev
,
4172 iface
->cfg
->ingress_policing_rate
,
4173 iface
->cfg
->ingress_policing_burst
);
4175 ofpbuf_uninit(&queues_buf
);
4179 iface_configure_cfm(struct iface
*iface
)
4181 const struct ovsrec_interface
*cfg
= iface
->cfg
;
4182 const char *opstate_str
;
4183 const char *cfm_ccm_vlan
;
4184 struct cfm_settings s
;
4185 struct smap netdev_args
;
4187 if (!cfg
->n_cfm_mpid
) {
4188 ofproto_port_clear_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
);
4192 s
.check_tnl_key
= false;
4193 smap_init(&netdev_args
);
4194 if (!netdev_get_config(iface
->netdev
, &netdev_args
)) {
4195 const char *key
= smap_get(&netdev_args
, "key");
4196 const char *in_key
= smap_get(&netdev_args
, "in_key");
4198 s
.check_tnl_key
= (key
&& !strcmp(key
, "flow"))
4199 || (in_key
&& !strcmp(in_key
, "flow"));
4201 smap_destroy(&netdev_args
);
4203 s
.mpid
= *cfg
->cfm_mpid
;
4204 s
.interval
= smap_get_int(&iface
->cfg
->other_config
, "cfm_interval", 0);
4205 cfm_ccm_vlan
= smap_get(&iface
->cfg
->other_config
, "cfm_ccm_vlan");
4206 s
.ccm_pcp
= smap_get_int(&iface
->cfg
->other_config
, "cfm_ccm_pcp", 0);
4208 if (s
.interval
<= 0) {
4212 if (!cfm_ccm_vlan
) {
4214 } else if (!strcasecmp("random", cfm_ccm_vlan
)) {
4215 s
.ccm_vlan
= CFM_RANDOM_VLAN
;
4217 s
.ccm_vlan
= atoi(cfm_ccm_vlan
);
4218 if (s
.ccm_vlan
== CFM_RANDOM_VLAN
) {
4223 s
.extended
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_extended",
4225 s
.demand
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_demand", false);
4227 opstate_str
= smap_get(&iface
->cfg
->other_config
, "cfm_opstate");
4228 s
.opup
= !opstate_str
|| !strcasecmp("up", opstate_str
);
4230 ofproto_port_set_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
, &s
);
4233 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4234 * instead of obtaining it from the database. */
4236 iface_is_synthetic(const struct iface
*iface
)
4238 return ovsdb_idl_row_is_synthetic(&iface
->cfg
->header_
);
4242 iface_validate_ofport__(size_t n
, int64_t *ofport
)
4244 return (n
&& *ofport
>= 1 && *ofport
< ofp_to_u16(OFPP_MAX
)
4245 ? u16_to_ofp(*ofport
)
4250 iface_get_requested_ofp_port(const struct ovsrec_interface
*cfg
)
4252 return iface_validate_ofport__(cfg
->n_ofport_request
, cfg
->ofport_request
);
4256 iface_pick_ofport(const struct ovsrec_interface
*cfg
)
4258 ofp_port_t requested_ofport
= iface_get_requested_ofp_port(cfg
);
4259 return (requested_ofport
!= OFPP_NONE
4261 : iface_validate_ofport__(cfg
->n_ofport
, cfg
->ofport
));
4264 /* Port mirroring. */
4266 static struct mirror
*
4267 mirror_find_by_uuid(struct bridge
*br
, const struct uuid
*uuid
)
4271 HMAP_FOR_EACH_IN_BUCKET (m
, hmap_node
, uuid_hash(uuid
), &br
->mirrors
) {
4272 if (uuid_equals(uuid
, &m
->uuid
)) {
4280 bridge_configure_mirrors(struct bridge
*br
)
4282 const struct ovsdb_datum
*mc
;
4283 unsigned long *flood_vlans
;
4284 struct mirror
*m
, *next
;
4287 /* Get rid of deleted mirrors. */
4288 mc
= ovsrec_bridge_get_mirrors(br
->cfg
, OVSDB_TYPE_UUID
);
4289 HMAP_FOR_EACH_SAFE (m
, next
, hmap_node
, &br
->mirrors
) {
4290 union ovsdb_atom atom
;
4292 atom
.uuid
= m
->uuid
;
4293 if (ovsdb_datum_find_key(mc
, &atom
, OVSDB_TYPE_UUID
) == UINT_MAX
) {
4298 /* Add new mirrors and reconfigure existing ones. */
4299 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
4300 const struct ovsrec_mirror
*cfg
= br
->cfg
->mirrors
[i
];
4301 struct mirror
*m
= mirror_find_by_uuid(br
, &cfg
->header_
.uuid
);
4303 m
= mirror_create(br
, cfg
);
4306 if (!mirror_configure(m
)) {
4311 /* Update flooded vlans (for RSPAN). */
4312 flood_vlans
= vlan_bitmap_from_array(br
->cfg
->flood_vlans
,
4313 br
->cfg
->n_flood_vlans
);
4314 ofproto_set_flood_vlans(br
->ofproto
, flood_vlans
);
4315 bitmap_free(flood_vlans
);
4318 static struct mirror
*
4319 mirror_create(struct bridge
*br
, const struct ovsrec_mirror
*cfg
)
4323 m
= xzalloc(sizeof *m
);
4324 m
->uuid
= cfg
->header_
.uuid
;
4325 hmap_insert(&br
->mirrors
, &m
->hmap_node
, uuid_hash(&m
->uuid
));
4327 m
->name
= xstrdup(cfg
->name
);
4333 mirror_destroy(struct mirror
*m
)
4336 struct bridge
*br
= m
->bridge
;
4339 ofproto_mirror_unregister(br
->ofproto
, m
);
4342 hmap_remove(&br
->mirrors
, &m
->hmap_node
);
4349 mirror_collect_ports(struct mirror
*m
,
4350 struct ovsrec_port
**in_ports
, int n_in_ports
,
4351 void ***out_portsp
, size_t *n_out_portsp
)
4353 void **out_ports
= xmalloc(n_in_ports
* sizeof *out_ports
);
4354 size_t n_out_ports
= 0;
4357 for (i
= 0; i
< n_in_ports
; i
++) {
4358 const char *name
= in_ports
[i
]->name
;
4359 struct port
*port
= port_lookup(m
->bridge
, name
);
4361 out_ports
[n_out_ports
++] = port
;
4363 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4364 "port %s", m
->bridge
->name
, m
->name
, name
);
4367 *out_portsp
= out_ports
;
4368 *n_out_portsp
= n_out_ports
;
4372 mirror_configure(struct mirror
*m
)
4374 const struct ovsrec_mirror
*cfg
= m
->cfg
;
4375 struct ofproto_mirror_settings s
;
4378 if (strcmp(cfg
->name
, m
->name
)) {
4380 m
->name
= xstrdup(cfg
->name
);
4384 /* Get output port or VLAN. */
4385 if (cfg
->output_port
) {
4386 s
.out_bundle
= port_lookup(m
->bridge
, cfg
->output_port
->name
);
4387 if (!s
.out_bundle
) {
4388 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4389 m
->bridge
->name
, m
->name
);
4392 s
.out_vlan
= UINT16_MAX
;
4394 if (cfg
->output_vlan
) {
4395 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4396 "output vlan; ignoring output vlan",
4397 m
->bridge
->name
, m
->name
);
4399 } else if (cfg
->output_vlan
) {
4400 /* The database should prevent invalid VLAN values. */
4401 s
.out_bundle
= NULL
;
4402 s
.out_vlan
= *cfg
->output_vlan
;
4404 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4405 m
->bridge
->name
, m
->name
);
4409 /* Get port selection. */
4410 if (cfg
->select_all
) {
4411 size_t n_ports
= hmap_count(&m
->bridge
->ports
);
4412 void **ports
= xmalloc(n_ports
* sizeof *ports
);
4417 HMAP_FOR_EACH (port
, hmap_node
, &m
->bridge
->ports
) {
4427 /* Get ports, dropping ports that don't exist.
4428 * The IDL ensures that there are no duplicates. */
4429 mirror_collect_ports(m
, cfg
->select_src_port
, cfg
->n_select_src_port
,
4430 &s
.srcs
, &s
.n_srcs
);
4431 mirror_collect_ports(m
, cfg
->select_dst_port
, cfg
->n_select_dst_port
,
4432 &s
.dsts
, &s
.n_dsts
);
4435 /* Get VLAN selection. */
4436 s
.src_vlans
= vlan_bitmap_from_array(cfg
->select_vlan
, cfg
->n_select_vlan
);
4439 ofproto_mirror_register(m
->bridge
->ofproto
, m
, &s
);
4442 if (s
.srcs
!= s
.dsts
) {
4451 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4453 * This is deprecated. It is only for compatibility with broken device drivers
4454 * in old versions of Linux that do not properly support VLANs when VLAN
4455 * devices are not used. When broken device drivers are no longer in
4456 * widespread use, we will delete these interfaces. */
4458 static struct ovsrec_port
**recs
;
4459 static size_t n_recs
, allocated_recs
;
4461 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4462 * splinters are reconfigured. */
4464 register_rec(struct ovsrec_port
*rec
)
4466 if (n_recs
>= allocated_recs
) {
4467 recs
= x2nrealloc(recs
, &allocated_recs
, sizeof *recs
);
4469 recs
[n_recs
++] = rec
;
4472 /* Frees all of the ports registered with register_reg(). */
4474 free_registered_recs(void)
4478 for (i
= 0; i
< n_recs
; i
++) {
4479 struct ovsrec_port
*port
= recs
[i
];
4482 for (j
= 0; j
< port
->n_interfaces
; j
++) {
4483 struct ovsrec_interface
*iface
= port
->interfaces
[j
];
4488 smap_destroy(&port
->other_config
);
4489 free(port
->interfaces
);
4497 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4500 vlan_splinters_is_enabled(const struct ovsrec_interface
*iface_cfg
)
4502 return smap_get_bool(&iface_cfg
->other_config
, "enable-vlan-splinters",
4506 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4509 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4510 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4511 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4514 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4515 * use, returns NULL.
4517 * Updates 'vlan_splinters_enabled_anywhere'. */
4518 static unsigned long int *
4519 collect_splinter_vlans(const struct ovsrec_open_vswitch
*ovs_cfg
)
4521 unsigned long int *splinter_vlans
;
4522 struct sset splinter_ifaces
;
4523 const char *real_dev_name
;
4524 struct shash
*real_devs
;
4525 struct shash_node
*node
;
4529 /* Free space allocated for synthesized ports and interfaces, since we're
4530 * in the process of reconstructing all of them. */
4531 free_registered_recs();
4533 splinter_vlans
= bitmap_allocate(4096);
4534 sset_init(&splinter_ifaces
);
4535 vlan_splinters_enabled_anywhere
= false;
4536 for (i
= 0; i
< ovs_cfg
->n_bridges
; i
++) {
4537 struct ovsrec_bridge
*br_cfg
= ovs_cfg
->bridges
[i
];
4540 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
4541 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
4544 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
4545 struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[k
];
4547 if (vlan_splinters_is_enabled(iface_cfg
)) {
4548 vlan_splinters_enabled_anywhere
= true;
4549 sset_add(&splinter_ifaces
, iface_cfg
->name
);
4550 vlan_bitmap_from_array__(port_cfg
->trunks
,
4556 if (port_cfg
->tag
&& *port_cfg
->tag
> 0 && *port_cfg
->tag
< 4095) {
4557 bitmap_set1(splinter_vlans
, *port_cfg
->tag
);
4562 if (!vlan_splinters_enabled_anywhere
) {
4563 free(splinter_vlans
);
4564 sset_destroy(&splinter_ifaces
);
4568 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
4570 ofproto_get_vlan_usage(br
->ofproto
, splinter_vlans
);
4574 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4575 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4576 * device to be created for it. */
4577 bitmap_set0(splinter_vlans
, 0);
4578 bitmap_set0(splinter_vlans
, 4095);
4580 /* Delete all VLAN devices that we don't need. */
4582 real_devs
= vlandev_get_real_devs();
4583 SHASH_FOR_EACH (node
, real_devs
) {
4584 const struct vlan_real_dev
*real_dev
= node
->data
;
4585 const struct vlan_dev
*vlan_dev
;
4586 bool real_dev_has_splinters
;
4588 real_dev_has_splinters
= sset_contains(&splinter_ifaces
,
4590 HMAP_FOR_EACH (vlan_dev
, hmap_node
, &real_dev
->vlan_devs
) {
4591 if (!real_dev_has_splinters
4592 || !bitmap_is_set(splinter_vlans
, vlan_dev
->vid
)) {
4593 struct netdev
*netdev
;
4595 if (!netdev_open(vlan_dev
->name
, "system", &netdev
)) {
4596 if (!netdev_get_in4(netdev
, NULL
, NULL
) ||
4597 !netdev_get_in6(netdev
, NULL
)) {
4598 /* It has an IP address configured, so we don't own
4599 * it. Don't delete it. */
4601 vlandev_del(vlan_dev
->name
);
4603 netdev_close(netdev
);
4610 /* Add all VLAN devices that we need. */
4611 SSET_FOR_EACH (real_dev_name
, &splinter_ifaces
) {
4614 BITMAP_FOR_EACH_1 (vid
, 4096, splinter_vlans
) {
4615 if (!vlandev_get_name(real_dev_name
, vid
)) {
4616 vlandev_add(real_dev_name
, vid
);
4623 sset_destroy(&splinter_ifaces
);
4625 if (bitmap_scan(splinter_vlans
, 1, 0, 4096) >= 4096) {
4626 free(splinter_vlans
);
4629 return splinter_vlans
;
4632 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4635 configure_splinter_port(struct port
*port
)
4637 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
4638 ofp_port_t realdev_ofp_port
;
4639 const char *realdev_name
;
4640 struct iface
*vlandev
, *realdev
;
4642 ofproto_bundle_unregister(port
->bridge
->ofproto
, port
);
4644 vlandev
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
,
4647 realdev_name
= smap_get(&port
->cfg
->other_config
, "realdev");
4648 realdev
= iface_lookup(port
->bridge
, realdev_name
);
4649 realdev_ofp_port
= realdev
? realdev
->ofp_port
: 0;
4651 ofproto_port_set_realdev(ofproto
, vlandev
->ofp_port
, realdev_ofp_port
,
4655 static struct ovsrec_port
*
4656 synthesize_splinter_port(const char *real_dev_name
,
4657 const char *vlan_dev_name
, int vid
)
4659 struct ovsrec_interface
*iface
;
4660 struct ovsrec_port
*port
;
4662 iface
= xmalloc(sizeof *iface
);
4663 ovsrec_interface_init(iface
);
4664 iface
->name
= xstrdup(vlan_dev_name
);
4665 iface
->type
= "system";
4667 port
= xmalloc(sizeof *port
);
4668 ovsrec_port_init(port
);
4669 port
->interfaces
= xmemdup(&iface
, sizeof iface
);
4670 port
->n_interfaces
= 1;
4671 port
->name
= xstrdup(vlan_dev_name
);
4672 port
->vlan_mode
= "splinter";
4673 port
->tag
= xmalloc(sizeof *port
->tag
);
4676 smap_add(&port
->other_config
, "realdev", real_dev_name
);
4682 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4683 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4684 * 1-bit in the 'splinter_vlans' bitmap. */
4686 add_vlan_splinter_ports(struct bridge
*br
,
4687 const unsigned long int *splinter_vlans
,
4688 struct shash
*ports
)
4692 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4693 * we're modifying 'ports'. */
4694 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
4695 const char *name
= br
->cfg
->ports
[i
]->name
;
4696 struct ovsrec_port
*port_cfg
= shash_find_data(ports
, name
);
4699 for (j
= 0; j
< port_cfg
->n_interfaces
; j
++) {
4700 struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[j
];
4702 if (vlan_splinters_is_enabled(iface_cfg
)) {
4703 const char *real_dev_name
;
4706 real_dev_name
= iface_cfg
->name
;
4707 BITMAP_FOR_EACH_1 (vid
, 4096, splinter_vlans
) {
4708 const char *vlan_dev_name
;
4710 vlan_dev_name
= vlandev_get_name(real_dev_name
, vid
);
4712 && !shash_find(ports
, vlan_dev_name
)) {
4713 shash_add(ports
, vlan_dev_name
,
4714 synthesize_splinter_port(
4715 real_dev_name
, vlan_dev_name
, vid
));
4724 mirror_refresh_stats(struct mirror
*m
)
4726 struct ofproto
*ofproto
= m
->bridge
->ofproto
;
4727 uint64_t tx_packets
, tx_bytes
;
4730 size_t stat_cnt
= 0;
4732 if (ofproto_mirror_get_stats(ofproto
, m
, &tx_packets
, &tx_bytes
)) {
4733 ovsrec_mirror_set_statistics(m
->cfg
, NULL
, NULL
, 0);
4737 if (tx_packets
!= UINT64_MAX
) {
4738 keys
[stat_cnt
] = "tx_packets";
4739 values
[stat_cnt
] = tx_packets
;
4742 if (tx_bytes
!= UINT64_MAX
) {
4743 keys
[stat_cnt
] = "tx_bytes";
4744 values
[stat_cnt
] = tx_bytes
;
4748 ovsrec_mirror_set_statistics(m
->cfg
, keys
, values
, stat_cnt
);