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 "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
64 VLOG_DEFINE_THIS_MODULE(bridge
);
66 COVERAGE_DEFINE(bridge_reconfigure
);
69 /* These members are always valid.
71 * They are immutable: they never change between iface_create() and
73 struct list port_elem
; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node
; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node
; /* In struct bridge's "ifaces" hmap. */
76 struct port
*port
; /* Containing port. */
77 char *name
; /* Host network device name. */
78 struct netdev
*netdev
; /* Network device. */
79 ofp_port_t ofp_port
; /* OpenFlow port number. */
81 /* These members are valid only within bridge_reconfigure(). */
82 const char *type
; /* Usually same as cfg->type. */
83 const struct ovsrec_interface
*cfg
;
87 struct uuid uuid
; /* UUID of this "mirror" record in database. */
88 struct hmap_node hmap_node
; /* In struct bridge's "mirrors" hmap. */
89 struct bridge
*bridge
;
91 const struct ovsrec_mirror
*cfg
;
95 struct hmap_node hmap_node
; /* Element in struct bridge's "ports" hmap. */
96 struct bridge
*bridge
;
99 const struct ovsrec_port
*cfg
;
101 /* An ordinary bridge port has 1 interface.
102 * A bridge port for bonding has at least 2 interfaces. */
103 struct list ifaces
; /* List of "struct iface"s. */
107 struct hmap_node node
; /* In 'all_bridges'. */
108 char *name
; /* User-specified arbitrary name. */
109 char *type
; /* Datapath type. */
110 uint8_t ea
[ETH_ADDR_LEN
]; /* Bridge Ethernet Address. */
111 uint8_t default_ea
[ETH_ADDR_LEN
]; /* Default MAC. */
112 const struct ovsrec_bridge
*cfg
;
114 /* OpenFlow switch processing. */
115 struct ofproto
*ofproto
; /* OpenFlow switch. */
118 struct hmap ports
; /* "struct port"s indexed by name. */
119 struct hmap ifaces
; /* "struct iface"s indexed by ofp_port. */
120 struct hmap iface_by_name
; /* "struct iface"s indexed by name. */
122 /* Port mirroring. */
123 struct hmap mirrors
; /* "struct mirror" indexed by UUID. */
125 /* Used during reconfiguration. */
126 struct shash wanted_ports
;
128 /* Synthetic local port if necessary. */
129 struct ovsrec_port synth_local_port
;
130 struct ovsrec_interface synth_local_iface
;
131 struct ovsrec_interface
*synth_local_ifacep
;
134 /* All bridges, indexed by name. */
135 static struct hmap all_bridges
= HMAP_INITIALIZER(&all_bridges
);
137 /* OVSDB IDL used to obtain configuration. */
138 static struct ovsdb_idl
*idl
;
140 /* We want to complete daemonization, fully detaching from our parent process,
141 * only after we have completed our initial configuration, committed our state
142 * to the database, and received confirmation back from the database server
143 * that it applied the commit. This allows our parent process to know that,
144 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
145 * to have already been filled in. (It is only "very likely" rather than
146 * certain because there is always a slim possibility that the transaction will
147 * fail or that some other client has added new bridges, ports, etc. while
148 * ovs-vswitchd was configuring using an old configuration.)
150 * We only need to do this once for our initial configuration at startup, so
151 * 'initial_config_done' tracks whether we've already done it. While we are
152 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
153 * itself and is otherwise NULL. */
154 static bool initial_config_done
;
155 static struct ovsdb_idl_txn
*daemonize_txn
;
157 /* Most recently processed IDL sequence number. */
158 static unsigned int idl_seqno
;
160 /* Track changes to port connectivity. */
161 static uint64_t connectivity_seqno
= LLONG_MIN
;
163 /* Each time this timer expires, the bridge fetches interface and mirror
164 * statistics and pushes them into the database. */
165 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
166 static long long int iface_stats_timer
= LLONG_MIN
;
168 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
169 * expensive. This can cause bridge_reconfigure() to take a long time during
170 * which no other work can be done. To deal with this problem, we limit port
171 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
172 * call to bridge_reconfigure(). If there is more work to do after the limit
173 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
174 * This allows the rest of the code to catch up on important things like
175 * forwarding packets. */
176 #define OFP_PORT_ACTION_WINDOW 10
178 static void add_del_bridges(const struct ovsrec_open_vswitch
*);
179 static void bridge_run__(void);
180 static void bridge_create(const struct ovsrec_bridge
*);
181 static void bridge_destroy(struct bridge
*);
182 static struct bridge
*bridge_lookup(const char *name
);
183 static unixctl_cb_func bridge_unixctl_dump_flows
;
184 static unixctl_cb_func bridge_unixctl_reconnect
;
185 static size_t bridge_get_controllers(const struct bridge
*br
,
186 struct ovsrec_controller
***controllersp
);
187 static void bridge_collect_wanted_ports(struct bridge
*,
188 const unsigned long *splinter_vlans
,
189 struct shash
*wanted_ports
);
190 static void bridge_delete_ofprotos(void);
191 static void bridge_delete_or_reconfigure_ports(struct bridge
*);
192 static void bridge_del_ports(struct bridge
*,
193 const struct shash
*wanted_ports
);
194 static void bridge_add_ports(struct bridge
*,
195 const struct shash
*wanted_ports
);
197 static void bridge_configure_datapath_id(struct bridge
*);
198 static void bridge_configure_netflow(struct bridge
*);
199 static void bridge_configure_forward_bpdu(struct bridge
*);
200 static void bridge_configure_mac_table(struct bridge
*);
201 static void bridge_configure_sflow(struct bridge
*, int *sflow_bridge_number
);
202 static void bridge_configure_ipfix(struct bridge
*);
203 static void bridge_configure_stp(struct bridge
*);
204 static void bridge_configure_tables(struct bridge
*);
205 static void bridge_configure_dp_desc(struct bridge
*);
206 static void bridge_configure_remotes(struct bridge
*,
207 const struct sockaddr_in
*managers
,
209 static void bridge_pick_local_hw_addr(struct bridge
*,
210 uint8_t ea
[ETH_ADDR_LEN
],
211 struct iface
**hw_addr_iface
);
212 static uint64_t bridge_pick_datapath_id(struct bridge
*,
213 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
214 struct iface
*hw_addr_iface
);
215 static uint64_t dpid_from_hash(const void *, size_t nbytes
);
216 static bool bridge_has_bond_fake_iface(const struct bridge
*,
218 static bool port_is_bond_fake_iface(const struct port
*);
220 static unixctl_cb_func qos_unixctl_show
;
222 static struct port
*port_create(struct bridge
*, const struct ovsrec_port
*);
223 static void port_del_ifaces(struct port
*);
224 static void port_destroy(struct port
*);
225 static struct port
*port_lookup(const struct bridge
*, const char *name
);
226 static void port_configure(struct port
*);
227 static struct lacp_settings
*port_configure_lacp(struct port
*,
228 struct lacp_settings
*);
229 static void port_configure_bond(struct port
*, struct bond_settings
*);
230 static bool port_is_synthetic(const struct port
*);
232 static void reconfigure_system_stats(const struct ovsrec_open_vswitch
*);
233 static void run_system_stats(void);
235 static void bridge_configure_mirrors(struct bridge
*);
236 static struct mirror
*mirror_create(struct bridge
*,
237 const struct ovsrec_mirror
*);
238 static void mirror_destroy(struct mirror
*);
239 static bool mirror_configure(struct mirror
*);
240 static void mirror_refresh_stats(struct mirror
*);
242 static void iface_configure_lacp(struct iface
*, struct lacp_slave_settings
*);
243 static bool iface_create(struct bridge
*, const struct ovsrec_interface
*,
244 const struct ovsrec_port
*);
245 static bool iface_is_internal(const struct ovsrec_interface
*iface
,
246 const struct ovsrec_bridge
*br
);
247 static const char *iface_get_type(const struct ovsrec_interface
*,
248 const struct ovsrec_bridge
*);
249 static void iface_destroy(struct iface
*);
250 static struct iface
*iface_lookup(const struct bridge
*, const char *name
);
251 static struct iface
*iface_find(const char *name
);
252 static struct iface
*iface_from_ofp_port(const struct bridge
*,
253 ofp_port_t ofp_port
);
254 static void iface_set_mac(struct iface
*, const uint8_t *);
255 static void iface_set_ofport(const struct ovsrec_interface
*, ofp_port_t ofport
);
256 static void iface_clear_db_record(const struct ovsrec_interface
*if_cfg
);
257 static void iface_configure_qos(struct iface
*, const struct ovsrec_qos
*);
258 static void iface_configure_cfm(struct iface
*);
259 static void iface_refresh_cfm_stats(struct iface
*);
260 static void iface_refresh_stats(struct iface
*);
261 static void iface_refresh_status(struct iface
*);
262 static bool iface_is_synthetic(const struct iface
*);
263 static ofp_port_t
iface_get_requested_ofp_port(
264 const struct ovsrec_interface
*);
265 static ofp_port_t
iface_pick_ofport(const struct ovsrec_interface
*);
267 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
269 * This is deprecated. It is only for compatibility with broken device drivers
270 * in old versions of Linux that do not properly support VLANs when VLAN
271 * devices are not used. When broken device drivers are no longer in
272 * widespread use, we will delete these interfaces. */
274 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
275 static bool vlan_splinters_enabled_anywhere
;
277 static bool vlan_splinters_is_enabled(const struct ovsrec_interface
*);
278 static unsigned long int *collect_splinter_vlans(
279 const struct ovsrec_open_vswitch
*);
280 static void configure_splinter_port(struct port
*);
281 static void add_vlan_splinter_ports(struct bridge
*,
282 const unsigned long int *splinter_vlans
,
283 struct shash
*ports
);
286 bridge_init_ofproto(const struct ovsrec_open_vswitch
*cfg
)
288 struct shash iface_hints
;
289 static bool initialized
= false;
296 shash_init(&iface_hints
);
299 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
300 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
303 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
304 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
307 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
308 struct ovsrec_interface
*if_cfg
= port_cfg
->interfaces
[k
];
309 struct iface_hint
*iface_hint
;
311 iface_hint
= xmalloc(sizeof *iface_hint
);
312 iface_hint
->br_name
= br_cfg
->name
;
313 iface_hint
->br_type
= br_cfg
->datapath_type
;
314 iface_hint
->ofp_port
= iface_pick_ofport(if_cfg
);
316 shash_add(&iface_hints
, if_cfg
->name
, iface_hint
);
322 ofproto_init(&iface_hints
);
324 shash_destroy_free_data(&iface_hints
);
328 /* Public functions. */
330 /* Initializes the bridge module, configuring it to obtain its configuration
331 * from an OVSDB server accessed over 'remote', which should be a string in a
332 * form acceptable to ovsdb_idl_create(). */
334 bridge_init(const char *remote
)
336 /* Create connection to database. */
337 idl
= ovsdb_idl_create(remote
, &ovsrec_idl_class
, true, true);
338 idl_seqno
= ovsdb_idl_get_seqno(idl
);
339 ovsdb_idl_set_lock(idl
, "ovs_vswitchd");
340 ovsdb_idl_verify_write_only(idl
);
342 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_cur_cfg
);
343 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_statistics
);
344 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_external_ids
);
345 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_ovs_version
);
346 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_db_version
);
347 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_type
);
348 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_version
);
350 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_id
);
351 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_status
);
352 ovsdb_idl_omit(idl
, &ovsrec_bridge_col_external_ids
);
354 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_status
);
355 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_statistics
);
356 ovsdb_idl_omit(idl
, &ovsrec_port_col_external_ids
);
358 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_admin_state
);
359 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_duplex
);
360 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_speed
);
361 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_state
);
362 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_resets
);
363 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mac_in_use
);
364 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ifindex
);
365 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mtu
);
366 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ofport
);
367 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_statistics
);
368 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_status
);
369 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault
);
370 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault_status
);
371 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_mpids
);
372 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_flap_count
);
373 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_health
);
374 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_opstate
);
375 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_bfd_status
);
376 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_lacp_current
);
377 ovsdb_idl_omit(idl
, &ovsrec_interface_col_external_ids
);
379 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_is_connected
);
380 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_role
);
381 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_status
);
382 ovsdb_idl_omit(idl
, &ovsrec_controller_col_external_ids
);
384 ovsdb_idl_omit(idl
, &ovsrec_qos_col_external_ids
);
386 ovsdb_idl_omit(idl
, &ovsrec_queue_col_external_ids
);
388 ovsdb_idl_omit(idl
, &ovsrec_mirror_col_external_ids
);
389 ovsdb_idl_omit_alert(idl
, &ovsrec_mirror_col_statistics
);
391 ovsdb_idl_omit(idl
, &ovsrec_netflow_col_external_ids
);
392 ovsdb_idl_omit(idl
, &ovsrec_sflow_col_external_ids
);
393 ovsdb_idl_omit(idl
, &ovsrec_ipfix_col_external_ids
);
394 ovsdb_idl_omit(idl
, &ovsrec_flow_sample_collector_set_col_external_ids
);
396 ovsdb_idl_omit(idl
, &ovsrec_manager_col_external_ids
);
397 ovsdb_idl_omit(idl
, &ovsrec_manager_col_inactivity_probe
);
398 ovsdb_idl_omit(idl
, &ovsrec_manager_col_is_connected
);
399 ovsdb_idl_omit(idl
, &ovsrec_manager_col_max_backoff
);
400 ovsdb_idl_omit(idl
, &ovsrec_manager_col_status
);
402 ovsdb_idl_omit(idl
, &ovsrec_ssl_col_external_ids
);
404 /* Register unixctl commands. */
405 unixctl_command_register("qos/show", "interface", 1, 1,
406 qos_unixctl_show
, NULL
);
407 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
408 bridge_unixctl_dump_flows
, NULL
);
409 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
410 bridge_unixctl_reconnect
, NULL
);
420 struct bridge
*br
, *next_br
;
422 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
425 ovsdb_idl_destroy(idl
);
428 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
429 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
430 * responsible for freeing '*managersp' (with free()).
432 * You may be asking yourself "why does ovs-vswitchd care?", because
433 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
434 * should not be and in fact is not directly involved in that. But
435 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
436 * it has to tell in-band control where the managers are to enable that.
437 * (Thus, only managers connected in-band are collected.)
440 collect_in_band_managers(const struct ovsrec_open_vswitch
*ovs_cfg
,
441 struct sockaddr_in
**managersp
, size_t *n_managersp
)
443 struct sockaddr_in
*managers
= NULL
;
444 size_t n_managers
= 0;
448 /* Collect all of the potential targets from the "targets" columns of the
449 * rows pointed to by "manager_options", excluding any that are
452 for (i
= 0; i
< ovs_cfg
->n_manager_options
; i
++) {
453 struct ovsrec_manager
*m
= ovs_cfg
->manager_options
[i
];
455 if (m
->connection_mode
&& !strcmp(m
->connection_mode
, "out-of-band")) {
456 sset_find_and_delete(&targets
, m
->target
);
458 sset_add(&targets
, m
->target
);
462 /* Now extract the targets' IP addresses. */
463 if (!sset_is_empty(&targets
)) {
466 managers
= xmalloc(sset_count(&targets
) * sizeof *managers
);
467 SSET_FOR_EACH (target
, &targets
) {
468 struct sockaddr_storage ss
;
470 if (stream_parse_target_with_default_port(target
, OVSDB_OLD_PORT
,
472 && ss
.ss_family
== AF_INET
) {
473 managers
[n_managers
++] = *(struct sockaddr_in
*) &ss
;
477 sset_destroy(&targets
);
479 *managersp
= managers
;
480 *n_managersp
= n_managers
;
484 bridge_reconfigure(const struct ovsrec_open_vswitch
*ovs_cfg
)
486 unsigned long int *splinter_vlans
;
487 struct sockaddr_in
*managers
;
488 struct bridge
*br
, *next
;
489 int sflow_bridge_number
;
492 COVERAGE_INC(bridge_reconfigure
);
494 ofproto_set_flow_limit(smap_get_int(&ovs_cfg
->other_config
, "flow-limit",
495 OFPROTO_FLOW_LIMIT_DEFAULT
));
498 smap_get_int(&ovs_cfg
->other_config
, "n-handler-threads", 0),
499 smap_get_int(&ovs_cfg
->other_config
, "n-revalidator-threads", 0));
501 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
502 * to 'ovs_cfg', with only very minimal configuration otherwise.
504 * This is mostly an update to bridge data structures. Nothing is pushed
505 * down to ofproto or lower layers. */
506 add_del_bridges(ovs_cfg
);
507 splinter_vlans
= collect_splinter_vlans(ovs_cfg
);
508 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
509 bridge_collect_wanted_ports(br
, splinter_vlans
, &br
->wanted_ports
);
510 bridge_del_ports(br
, &br
->wanted_ports
);
512 free(splinter_vlans
);
514 /* Start pushing configuration changes down to the ofproto layer:
516 * - Delete ofprotos that are no longer configured.
518 * - Delete ports that are no longer configured.
520 * - Reconfigure existing ports to their desired configurations, or
521 * delete them if not possible.
523 * We have to do all the deletions before we can do any additions, because
524 * the ports to be added might require resources that will be freed up by
525 * deletions (they might especially overlap in name). */
526 bridge_delete_ofprotos();
527 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
529 bridge_delete_or_reconfigure_ports(br
);
533 /* Finish pushing configuration changes to the ofproto layer:
535 * - Create ofprotos that are missing.
537 * - Add ports that are missing. */
538 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
542 error
= ofproto_create(br
->name
, br
->type
, &br
->ofproto
);
544 VLOG_ERR("failed to create bridge %s: %s", br
->name
,
545 ovs_strerror(error
));
546 shash_destroy(&br
->wanted_ports
);
551 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
552 bridge_add_ports(br
, &br
->wanted_ports
);
553 shash_destroy(&br
->wanted_ports
);
556 reconfigure_system_stats(ovs_cfg
);
558 /* Complete the configuration. */
559 sflow_bridge_number
= 0;
560 collect_in_band_managers(ovs_cfg
, &managers
, &n_managers
);
561 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
564 /* We need the datapath ID early to allow LACP ports to use it as the
565 * default system ID. */
566 bridge_configure_datapath_id(br
);
568 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
571 port_configure(port
);
573 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
574 iface_set_ofport(iface
->cfg
, iface
->ofp_port
);
575 iface_configure_cfm(iface
);
576 iface_configure_qos(iface
, port
->cfg
->qos
);
577 iface_set_mac(iface
, port
->cfg
->fake_bridge
? br
->ea
: NULL
);
578 ofproto_port_set_bfd(br
->ofproto
, iface
->ofp_port
,
582 bridge_configure_mirrors(br
);
583 bridge_configure_forward_bpdu(br
);
584 bridge_configure_mac_table(br
);
585 bridge_configure_remotes(br
, managers
, n_managers
);
586 bridge_configure_netflow(br
);
587 bridge_configure_sflow(br
, &sflow_bridge_number
);
588 bridge_configure_ipfix(br
);
589 bridge_configure_stp(br
);
590 bridge_configure_tables(br
);
591 bridge_configure_dp_desc(br
);
593 if (smap_get(&br
->cfg
->other_config
, "flow-eviction-threshold")) {
594 /* XXX: Remove this warning message eventually. */
595 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
596 " moved to the Open_vSwitch table. Ignoring its"
597 " setting in the bridge table.");
602 /* The ofproto-dpif provider does some final reconfiguration in its
603 * ->type_run() function. We have to call it before notifying the database
604 * client that reconfiguration is complete, otherwise there is a very
605 * narrow race window in which e.g. ofproto/trace will not recognize the
606 * new configuration (sometimes this causes unit test failures). */
610 /* Delete ofprotos which aren't configured or have the wrong type. Create
611 * ofprotos which don't exist but need to. */
613 bridge_delete_ofprotos(void)
620 /* Delete ofprotos with no bridge or with the wrong type. */
623 ofproto_enumerate_types(&types
);
624 SSET_FOR_EACH (type
, &types
) {
627 ofproto_enumerate_names(type
, &names
);
628 SSET_FOR_EACH (name
, &names
) {
629 br
= bridge_lookup(name
);
630 if (!br
|| strcmp(type
, br
->type
)) {
631 ofproto_delete(name
, type
);
635 sset_destroy(&names
);
636 sset_destroy(&types
);
640 add_ofp_port(ofp_port_t port
, ofp_port_t
*ports
, size_t *n
, size_t *allocated
)
642 if (*n
>= *allocated
) {
643 ports
= x2nrealloc(ports
, allocated
, sizeof *ports
);
645 ports
[(*n
)++] = port
;
650 bridge_delete_or_reconfigure_ports(struct bridge
*br
)
652 struct ofproto_port ofproto_port
;
653 struct ofproto_port_dump dump
;
655 /* List of "ofp_port"s to delete. We make a list instead of deleting them
656 * right away because ofproto implementations aren't necessarily able to
657 * iterate through a changing list of ports in an entirely robust way. */
665 OFPROTO_PORT_FOR_EACH (&ofproto_port
, &dump
, br
->ofproto
) {
666 ofp_port_t requested_ofp_port
;
669 iface
= iface_lookup(br
, ofproto_port
.name
);
671 /* No such iface is configured, so we should delete this
674 * As a corner case exception, keep the port if it's a bond fake
676 if (bridge_has_bond_fake_iface(br
, ofproto_port
.name
)
677 && !strcmp(ofproto_port
.type
, "internal")) {
683 if (strcmp(ofproto_port
.type
, iface
->type
)
684 || netdev_set_config(iface
->netdev
, &iface
->cfg
->options
)) {
685 /* The interface is the wrong type or can't be configured.
690 /* If the requested OpenFlow port for 'iface' changed, and it's not
691 * already the correct port, then we might want to temporarily delete
692 * this interface, so we can add it back again with the new OpenFlow
694 requested_ofp_port
= iface_get_requested_ofp_port(iface
->cfg
);
695 if (iface
->ofp_port
!= OFPP_LOCAL
&&
696 requested_ofp_port
!= OFPP_NONE
&&
697 requested_ofp_port
!= iface
->ofp_port
) {
698 ofp_port_t victim_request
;
699 struct iface
*victim
;
701 /* Check for an existing OpenFlow port currently occupying
702 * 'iface''s requested port number. If there isn't one, then
703 * delete this port. Otherwise we need to consider further. */
704 victim
= iface_from_ofp_port(br
, requested_ofp_port
);
709 /* 'victim' is a port currently using 'iface''s requested port
710 * number. Unless 'victim' specifically requested that port
711 * number, too, then we can delete both 'iface' and 'victim'
712 * temporarily. (We'll add both of them back again later with new
713 * OpenFlow port numbers.)
715 * If 'victim' did request port number 'requested_ofp_port', just
716 * like 'iface', then that's a configuration inconsistency that we
717 * can't resolve. We might as well let it keep its current port
719 victim_request
= iface_get_requested_ofp_port(victim
->cfg
);
720 if (victim_request
!= requested_ofp_port
) {
721 del
= add_ofp_port(victim
->ofp_port
, del
, &n
, &allocated
);
722 iface_destroy(victim
);
731 iface_destroy(iface
);
732 del
= add_ofp_port(ofproto_port
.ofp_port
, del
, &n
, &allocated
);
735 for (i
= 0; i
< n
; i
++) {
736 ofproto_port_del(br
->ofproto
, del
[i
]);
742 bridge_add_ports__(struct bridge
*br
, const struct shash
*wanted_ports
,
743 bool with_requested_port
)
745 struct shash_node
*port_node
;
747 SHASH_FOR_EACH (port_node
, wanted_ports
) {
748 const struct ovsrec_port
*port_cfg
= port_node
->data
;
751 for (i
= 0; i
< port_cfg
->n_interfaces
; i
++) {
752 const struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[i
];
753 ofp_port_t requested_ofp_port
;
755 requested_ofp_port
= iface_get_requested_ofp_port(iface_cfg
);
756 if ((requested_ofp_port
!= OFPP_NONE
) == with_requested_port
) {
757 struct iface
*iface
= iface_lookup(br
, iface_cfg
->name
);
760 iface_create(br
, iface_cfg
, port_cfg
);
768 bridge_add_ports(struct bridge
*br
, const struct shash
*wanted_ports
)
770 /* First add interfaces that request a particular port number. */
771 bridge_add_ports__(br
, wanted_ports
, true);
773 /* Then add interfaces that want automatic port number assignment.
774 * We add these afterward to avoid accidentally taking a specifically
775 * requested port number. */
776 bridge_add_ports__(br
, wanted_ports
, false);
780 port_configure(struct port
*port
)
782 const struct ovsrec_port
*cfg
= port
->cfg
;
783 struct bond_settings bond_settings
;
784 struct lacp_settings lacp_settings
;
785 struct ofproto_bundle_settings s
;
788 if (cfg
->vlan_mode
&& !strcmp(cfg
->vlan_mode
, "splinter")) {
789 configure_splinter_port(port
);
798 s
.slaves
= xmalloc(list_size(&port
->ifaces
) * sizeof *s
.slaves
);
799 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
800 s
.slaves
[s
.n_slaves
++] = iface
->ofp_port
;
805 if (cfg
->tag
&& *cfg
->tag
>= 0 && *cfg
->tag
<= 4095) {
809 /* Get VLAN trunks. */
812 s
.trunks
= vlan_bitmap_from_array(cfg
->trunks
, cfg
->n_trunks
);
816 if (cfg
->vlan_mode
) {
817 if (!strcmp(cfg
->vlan_mode
, "access")) {
818 s
.vlan_mode
= PORT_VLAN_ACCESS
;
819 } else if (!strcmp(cfg
->vlan_mode
, "trunk")) {
820 s
.vlan_mode
= PORT_VLAN_TRUNK
;
821 } else if (!strcmp(cfg
->vlan_mode
, "native-tagged")) {
822 s
.vlan_mode
= PORT_VLAN_NATIVE_TAGGED
;
823 } else if (!strcmp(cfg
->vlan_mode
, "native-untagged")) {
824 s
.vlan_mode
= PORT_VLAN_NATIVE_UNTAGGED
;
826 /* This "can't happen" because ovsdb-server should prevent it. */
827 VLOG_ERR("unknown VLAN mode %s", cfg
->vlan_mode
);
828 s
.vlan_mode
= PORT_VLAN_TRUNK
;
832 s
.vlan_mode
= PORT_VLAN_ACCESS
;
834 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
838 s
.vlan_mode
= PORT_VLAN_TRUNK
;
841 s
.use_priority_tags
= smap_get_bool(&cfg
->other_config
, "priority-tags",
844 /* Get LACP settings. */
845 s
.lacp
= port_configure_lacp(port
, &lacp_settings
);
849 s
.lacp_slaves
= xmalloc(s
.n_slaves
* sizeof *s
.lacp_slaves
);
850 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
851 iface_configure_lacp(iface
, &s
.lacp_slaves
[i
++]);
854 s
.lacp_slaves
= NULL
;
857 /* Get bond settings. */
858 if (s
.n_slaves
> 1) {
859 s
.bond
= &bond_settings
;
860 port_configure_bond(port
, &bond_settings
);
863 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
864 netdev_set_miimon_interval(iface
->netdev
, 0);
869 ofproto_bundle_register(port
->bridge
->ofproto
, port
, &s
);
877 /* Pick local port hardware address and datapath ID for 'br'. */
879 bridge_configure_datapath_id(struct bridge
*br
)
881 uint8_t ea
[ETH_ADDR_LEN
];
883 struct iface
*local_iface
;
884 struct iface
*hw_addr_iface
;
887 bridge_pick_local_hw_addr(br
, ea
, &hw_addr_iface
);
888 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
890 int error
= netdev_set_etheraddr(local_iface
->netdev
, ea
);
892 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
893 VLOG_ERR_RL(&rl
, "bridge %s: failed to set bridge "
894 "Ethernet address: %s",
895 br
->name
, ovs_strerror(error
));
898 memcpy(br
->ea
, ea
, ETH_ADDR_LEN
);
900 dpid
= bridge_pick_datapath_id(br
, ea
, hw_addr_iface
);
901 if (dpid
!= ofproto_get_datapath_id(br
->ofproto
)) {
902 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64
, br
->name
, dpid
);
903 ofproto_set_datapath_id(br
->ofproto
, dpid
);
906 dpid_string
= xasprintf("%016"PRIx64
, dpid
);
907 ovsrec_bridge_set_datapath_id(br
->cfg
, dpid_string
);
911 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
914 bridge_get_allowed_versions(struct bridge
*br
)
916 if (!br
->cfg
->n_protocols
)
919 return ofputil_versions_from_strings(br
->cfg
->protocols
,
920 br
->cfg
->n_protocols
);
923 /* Set NetFlow configuration on 'br'. */
925 bridge_configure_netflow(struct bridge
*br
)
927 struct ovsrec_netflow
*cfg
= br
->cfg
->netflow
;
928 struct netflow_options opts
;
931 ofproto_set_netflow(br
->ofproto
, NULL
);
935 memset(&opts
, 0, sizeof opts
);
937 /* Get default NetFlow configuration from datapath.
938 * Apply overrides from 'cfg'. */
939 ofproto_get_netflow_ids(br
->ofproto
, &opts
.engine_type
, &opts
.engine_id
);
940 if (cfg
->engine_type
) {
941 opts
.engine_type
= *cfg
->engine_type
;
943 if (cfg
->engine_id
) {
944 opts
.engine_id
= *cfg
->engine_id
;
947 /* Configure active timeout interval. */
948 opts
.active_timeout
= cfg
->active_timeout
;
949 if (!opts
.active_timeout
) {
950 opts
.active_timeout
= -1;
951 } else if (opts
.active_timeout
< 0) {
952 VLOG_WARN("bridge %s: active timeout interval set to negative "
953 "value, using default instead (%d seconds)", br
->name
,
954 NF_ACTIVE_TIMEOUT_DEFAULT
);
955 opts
.active_timeout
= -1;
958 /* Add engine ID to interface number to disambiguate bridgs? */
959 opts
.add_id_to_iface
= cfg
->add_id_to_interface
;
960 if (opts
.add_id_to_iface
) {
961 if (opts
.engine_id
> 0x7f) {
962 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
963 "another vswitch, choose an engine id less than 128",
966 if (hmap_count(&br
->ports
) > 508) {
967 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
968 "another port when more than 508 ports are used",
974 sset_init(&opts
.collectors
);
975 sset_add_array(&opts
.collectors
, cfg
->targets
, cfg
->n_targets
);
978 if (ofproto_set_netflow(br
->ofproto
, &opts
)) {
979 VLOG_ERR("bridge %s: problem setting netflow collectors", br
->name
);
981 sset_destroy(&opts
.collectors
);
984 /* Set sFlow configuration on 'br'. */
986 bridge_configure_sflow(struct bridge
*br
, int *sflow_bridge_number
)
988 const struct ovsrec_sflow
*cfg
= br
->cfg
->sflow
;
989 struct ovsrec_controller
**controllers
;
990 struct ofproto_sflow_options oso
;
991 size_t n_controllers
;
995 ofproto_set_sflow(br
->ofproto
, NULL
);
999 memset(&oso
, 0, sizeof oso
);
1001 sset_init(&oso
.targets
);
1002 sset_add_array(&oso
.targets
, cfg
->targets
, cfg
->n_targets
);
1004 oso
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1005 if (cfg
->sampling
) {
1006 oso
.sampling_rate
= *cfg
->sampling
;
1009 oso
.polling_interval
= SFL_DEFAULT_POLLING_INTERVAL
;
1011 oso
.polling_interval
= *cfg
->polling
;
1014 oso
.header_len
= SFL_DEFAULT_HEADER_SIZE
;
1016 oso
.header_len
= *cfg
->header
;
1019 oso
.sub_id
= (*sflow_bridge_number
)++;
1020 oso
.agent_device
= cfg
->agent
;
1022 oso
.control_ip
= NULL
;
1023 n_controllers
= bridge_get_controllers(br
, &controllers
);
1024 for (i
= 0; i
< n_controllers
; i
++) {
1025 if (controllers
[i
]->local_ip
) {
1026 oso
.control_ip
= controllers
[i
]->local_ip
;
1030 ofproto_set_sflow(br
->ofproto
, &oso
);
1032 sset_destroy(&oso
.targets
);
1035 /* Returns whether a IPFIX row is valid. */
1037 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix
*ipfix
)
1039 return ipfix
&& ipfix
->n_targets
> 0;
1042 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1044 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set
*fscs
,
1045 const struct bridge
*br
)
1047 return ovsrec_ipfix_is_valid(fscs
->ipfix
) && fscs
->bridge
== br
->cfg
;
1050 /* Set IPFIX configuration on 'br'. */
1052 bridge_configure_ipfix(struct bridge
*br
)
1054 const struct ovsrec_ipfix
*be_cfg
= br
->cfg
->ipfix
;
1055 bool valid_be_cfg
= ovsrec_ipfix_is_valid(be_cfg
);
1056 const struct ovsrec_flow_sample_collector_set
*fe_cfg
;
1057 struct ofproto_ipfix_bridge_exporter_options be_opts
;
1058 struct ofproto_ipfix_flow_exporter_options
*fe_opts
= NULL
;
1059 size_t n_fe_opts
= 0;
1061 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1062 if (ovsrec_fscs_is_valid(fe_cfg
, br
)) {
1067 if (!valid_be_cfg
&& n_fe_opts
== 0) {
1068 ofproto_set_ipfix(br
->ofproto
, NULL
, NULL
, 0);
1073 memset(&be_opts
, 0, sizeof be_opts
);
1075 sset_init(&be_opts
.targets
);
1076 sset_add_array(&be_opts
.targets
, be_cfg
->targets
, be_cfg
->n_targets
);
1078 if (be_cfg
->sampling
) {
1079 be_opts
.sampling_rate
= *be_cfg
->sampling
;
1081 be_opts
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1083 if (be_cfg
->obs_domain_id
) {
1084 be_opts
.obs_domain_id
= *be_cfg
->obs_domain_id
;
1086 if (be_cfg
->obs_point_id
) {
1087 be_opts
.obs_point_id
= *be_cfg
->obs_point_id
;
1089 if (be_cfg
->cache_active_timeout
) {
1090 be_opts
.cache_active_timeout
= *be_cfg
->cache_active_timeout
;
1092 if (be_cfg
->cache_max_flows
) {
1093 be_opts
.cache_max_flows
= *be_cfg
->cache_max_flows
;
1097 if (n_fe_opts
> 0) {
1098 struct ofproto_ipfix_flow_exporter_options
*opts
;
1099 fe_opts
= xcalloc(n_fe_opts
, sizeof *fe_opts
);
1101 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1102 if (ovsrec_fscs_is_valid(fe_cfg
, br
)) {
1103 opts
->collector_set_id
= fe_cfg
->id
;
1104 sset_init(&opts
->targets
);
1105 sset_add_array(&opts
->targets
, fe_cfg
->ipfix
->targets
,
1106 fe_cfg
->ipfix
->n_targets
);
1107 opts
->cache_active_timeout
= fe_cfg
->ipfix
->cache_active_timeout
1108 ? *fe_cfg
->ipfix
->cache_active_timeout
: 0;
1109 opts
->cache_max_flows
= fe_cfg
->ipfix
->cache_max_flows
1110 ? *fe_cfg
->ipfix
->cache_max_flows
: 0;
1116 ofproto_set_ipfix(br
->ofproto
, valid_be_cfg
? &be_opts
: NULL
, fe_opts
,
1120 sset_destroy(&be_opts
.targets
);
1123 if (n_fe_opts
> 0) {
1124 struct ofproto_ipfix_flow_exporter_options
*opts
= fe_opts
;
1126 for (i
= 0; i
< n_fe_opts
; i
++) {
1127 sset_destroy(&opts
->targets
);
1135 port_configure_stp(const struct ofproto
*ofproto
, struct port
*port
,
1136 struct ofproto_port_stp_settings
*port_s
,
1137 int *port_num_counter
, unsigned long *port_num_bitmap
)
1139 const char *config_str
;
1140 struct iface
*iface
;
1142 if (!smap_get_bool(&port
->cfg
->other_config
, "stp-enable", true)) {
1143 port_s
->enable
= false;
1146 port_s
->enable
= true;
1149 /* STP over bonds is not supported. */
1150 if (!list_is_singleton(&port
->ifaces
)) {
1151 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1153 port_s
->enable
= false;
1157 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
1159 /* Internal ports shouldn't participate in spanning tree, so
1161 if (!strcmp(iface
->type
, "internal")) {
1162 VLOG_DBG("port %s: disable STP on internal ports", port
->name
);
1163 port_s
->enable
= false;
1167 /* STP on mirror output ports is not supported. */
1168 if (ofproto_is_mirror_output_bundle(ofproto
, port
)) {
1169 VLOG_DBG("port %s: disable STP on mirror ports", port
->name
);
1170 port_s
->enable
= false;
1174 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-num");
1176 unsigned long int port_num
= strtoul(config_str
, NULL
, 0);
1177 int port_idx
= port_num
- 1;
1179 if (port_num
< 1 || port_num
> STP_MAX_PORTS
) {
1180 VLOG_ERR("port %s: invalid stp-port-num", port
->name
);
1181 port_s
->enable
= false;
1185 if (bitmap_is_set(port_num_bitmap
, port_idx
)) {
1186 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1187 port
->name
, port_num
);
1188 port_s
->enable
= false;
1191 bitmap_set1(port_num_bitmap
, port_idx
);
1192 port_s
->port_num
= port_idx
;
1194 if (*port_num_counter
>= STP_MAX_PORTS
) {
1195 VLOG_ERR("port %s: too many STP ports, disabling", port
->name
);
1196 port_s
->enable
= false;
1200 port_s
->port_num
= (*port_num_counter
)++;
1203 config_str
= smap_get(&port
->cfg
->other_config
, "stp-path-cost");
1205 port_s
->path_cost
= strtoul(config_str
, NULL
, 10);
1207 enum netdev_features current
;
1210 netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1211 mbps
= netdev_features_to_bps(current
, 100 * 1000 * 1000) / 1000000;
1212 port_s
->path_cost
= stp_convert_speed_to_cost(mbps
);
1215 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-priority");
1217 port_s
->priority
= strtoul(config_str
, NULL
, 0);
1219 port_s
->priority
= STP_DEFAULT_PORT_PRIORITY
;
1223 /* Set spanning tree configuration on 'br'. */
1225 bridge_configure_stp(struct bridge
*br
)
1227 if (!br
->cfg
->stp_enable
) {
1228 ofproto_set_stp(br
->ofproto
, NULL
);
1230 struct ofproto_stp_settings br_s
;
1231 const char *config_str
;
1233 int port_num_counter
;
1234 unsigned long *port_num_bitmap
;
1236 config_str
= smap_get(&br
->cfg
->other_config
, "stp-system-id");
1238 uint8_t ea
[ETH_ADDR_LEN
];
1240 if (eth_addr_from_string(config_str
, ea
)) {
1241 br_s
.system_id
= eth_addr_to_uint64(ea
);
1243 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1244 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1245 "to "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(br
->ea
));
1248 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1251 config_str
= smap_get(&br
->cfg
->other_config
, "stp-priority");
1253 br_s
.priority
= strtoul(config_str
, NULL
, 0);
1255 br_s
.priority
= STP_DEFAULT_BRIDGE_PRIORITY
;
1258 config_str
= smap_get(&br
->cfg
->other_config
, "stp-hello-time");
1260 br_s
.hello_time
= strtoul(config_str
, NULL
, 10) * 1000;
1262 br_s
.hello_time
= STP_DEFAULT_HELLO_TIME
;
1265 config_str
= smap_get(&br
->cfg
->other_config
, "stp-max-age");
1267 br_s
.max_age
= strtoul(config_str
, NULL
, 10) * 1000;
1269 br_s
.max_age
= STP_DEFAULT_MAX_AGE
;
1272 config_str
= smap_get(&br
->cfg
->other_config
, "stp-forward-delay");
1274 br_s
.fwd_delay
= strtoul(config_str
, NULL
, 10) * 1000;
1276 br_s
.fwd_delay
= STP_DEFAULT_FWD_DELAY
;
1279 /* Configure STP on the bridge. */
1280 if (ofproto_set_stp(br
->ofproto
, &br_s
)) {
1281 VLOG_ERR("bridge %s: could not enable STP", br
->name
);
1285 /* Users must either set the port number with the "stp-port-num"
1286 * configuration on all ports or none. If manual configuration
1287 * is not done, then we allocate them sequentially. */
1288 port_num_counter
= 0;
1289 port_num_bitmap
= bitmap_allocate(STP_MAX_PORTS
);
1290 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1291 struct ofproto_port_stp_settings port_s
;
1292 struct iface
*iface
;
1294 port_configure_stp(br
->ofproto
, port
, &port_s
,
1295 &port_num_counter
, port_num_bitmap
);
1297 /* As bonds are not supported, just apply configuration to
1298 * all interfaces. */
1299 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1300 if (ofproto_port_set_stp(br
->ofproto
, iface
->ofp_port
,
1302 VLOG_ERR("port %s: could not enable STP", port
->name
);
1308 if (bitmap_scan(port_num_bitmap
, 0, STP_MAX_PORTS
) != STP_MAX_PORTS
1309 && port_num_counter
) {
1310 VLOG_ERR("bridge %s: must manually configure all STP port "
1311 "IDs or none, disabling", br
->name
);
1312 ofproto_set_stp(br
->ofproto
, NULL
);
1314 bitmap_free(port_num_bitmap
);
1319 bridge_has_bond_fake_iface(const struct bridge
*br
, const char *name
)
1321 const struct port
*port
= port_lookup(br
, name
);
1322 return port
&& port_is_bond_fake_iface(port
);
1326 port_is_bond_fake_iface(const struct port
*port
)
1328 return port
->cfg
->bond_fake_iface
&& !list_is_short(&port
->ifaces
);
1332 add_del_bridges(const struct ovsrec_open_vswitch
*cfg
)
1334 struct bridge
*br
, *next
;
1335 struct shash new_br
;
1338 /* Collect new bridges' names and types. */
1339 shash_init(&new_br
);
1340 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1341 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1342 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1344 if (strchr(br_cfg
->name
, '/')) {
1345 /* Prevent remote ovsdb-server users from accessing arbitrary
1346 * directories, e.g. consider a bridge named "../../../etc/". */
1347 VLOG_WARN_RL(&rl
, "ignoring bridge with invalid name \"%s\"",
1349 } else if (!shash_add_once(&new_br
, br_cfg
->name
, br_cfg
)) {
1350 VLOG_WARN_RL(&rl
, "bridge %s specified twice", br_cfg
->name
);
1354 /* Get rid of deleted bridges or those whose types have changed.
1355 * Update 'cfg' of bridges that still exist. */
1356 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
1357 br
->cfg
= shash_find_data(&new_br
, br
->name
);
1358 if (!br
->cfg
|| strcmp(br
->type
, ofproto_normalize_type(
1359 br
->cfg
->datapath_type
))) {
1364 /* Add new bridges. */
1365 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1366 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1367 struct bridge
*br
= bridge_lookup(br_cfg
->name
);
1369 bridge_create(br_cfg
);
1373 shash_destroy(&new_br
);
1376 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1377 * Returns 0 if successful, otherwise a positive errno value. */
1379 iface_set_netdev_config(const struct ovsrec_interface
*iface_cfg
,
1380 struct netdev
*netdev
)
1382 return netdev_set_config(netdev
, &iface_cfg
->options
);
1385 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1386 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1388 * If successful, returns 0 and stores the network device in '*netdevp'. On
1389 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1391 iface_do_create(const struct bridge
*br
,
1392 const struct ovsrec_interface
*iface_cfg
,
1393 const struct ovsrec_port
*port_cfg
,
1394 ofp_port_t
*ofp_portp
, struct netdev
**netdevp
)
1396 struct netdev
*netdev
= NULL
;
1399 if (netdev_is_reserved_name(iface_cfg
->name
)) {
1400 VLOG_WARN("could not create interface %s, name is reserved",
1406 error
= netdev_open(iface_cfg
->name
,
1407 iface_get_type(iface_cfg
, br
->cfg
), &netdev
);
1409 VLOG_WARN("could not open network device %s (%s)",
1410 iface_cfg
->name
, ovs_strerror(error
));
1414 error
= iface_set_netdev_config(iface_cfg
, netdev
);
1419 *ofp_portp
= iface_pick_ofport(iface_cfg
);
1420 error
= ofproto_port_add(br
->ofproto
, netdev
, ofp_portp
);
1425 VLOG_INFO("bridge %s: added interface %s on port %d",
1426 br
->name
, iface_cfg
->name
, *ofp_portp
);
1428 if ((port_cfg
->vlan_mode
&& !strcmp(port_cfg
->vlan_mode
, "splinter"))
1429 || iface_is_internal(iface_cfg
, br
->cfg
)) {
1430 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
1438 netdev_close(netdev
);
1442 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1443 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1444 * automatically allocated for the iface. Takes ownership of and
1445 * deallocates 'if_cfg'.
1447 * Return true if an iface is successfully created, false otherwise. */
1449 iface_create(struct bridge
*br
, const struct ovsrec_interface
*iface_cfg
,
1450 const struct ovsrec_port
*port_cfg
)
1452 struct netdev
*netdev
;
1453 struct iface
*iface
;
1454 ofp_port_t ofp_port
;
1458 /* Do the bits that can fail up front. */
1459 ovs_assert(!iface_lookup(br
, iface_cfg
->name
));
1460 error
= iface_do_create(br
, iface_cfg
, port_cfg
, &ofp_port
, &netdev
);
1462 iface_clear_db_record(iface_cfg
);
1466 /* Get or create the port structure. */
1467 port
= port_lookup(br
, port_cfg
->name
);
1469 port
= port_create(br
, port_cfg
);
1472 /* Create the iface structure. */
1473 iface
= xzalloc(sizeof *iface
);
1474 list_push_back(&port
->ifaces
, &iface
->port_elem
);
1475 hmap_insert(&br
->iface_by_name
, &iface
->name_node
,
1476 hash_string(iface_cfg
->name
, 0));
1478 iface
->name
= xstrdup(iface_cfg
->name
);
1479 iface
->ofp_port
= ofp_port
;
1480 iface
->netdev
= netdev
;
1481 iface
->type
= iface_get_type(iface_cfg
, br
->cfg
);
1482 iface
->cfg
= iface_cfg
;
1483 hmap_insert(&br
->ifaces
, &iface
->ofp_port_node
,
1484 hash_ofp_port(ofp_port
));
1486 /* Populate initial status in database. */
1487 iface_refresh_stats(iface
);
1488 iface_refresh_status(iface
);
1490 /* Add bond fake iface if necessary. */
1491 if (port_is_bond_fake_iface(port
)) {
1492 struct ofproto_port ofproto_port
;
1494 if (ofproto_port_query_by_name(br
->ofproto
, port
->name
,
1496 struct netdev
*netdev
;
1499 error
= netdev_open(port
->name
, "internal", &netdev
);
1501 ofp_port_t fake_ofp_port
= OFPP_NONE
;
1502 ofproto_port_add(br
->ofproto
, netdev
, &fake_ofp_port
);
1503 netdev_close(netdev
);
1505 VLOG_WARN("could not open network device %s (%s)",
1506 port
->name
, ovs_strerror(error
));
1509 /* Already exists, nothing to do. */
1510 ofproto_port_destroy(&ofproto_port
);
1517 /* Set forward BPDU option. */
1519 bridge_configure_forward_bpdu(struct bridge
*br
)
1521 ofproto_set_forward_bpdu(br
->ofproto
,
1522 smap_get_bool(&br
->cfg
->other_config
,
1527 /* Set MAC learning table configuration for 'br'. */
1529 bridge_configure_mac_table(struct bridge
*br
)
1531 const char *idle_time_str
;
1534 const char *mac_table_size_str
;
1537 idle_time_str
= smap_get(&br
->cfg
->other_config
, "mac-aging-time");
1538 idle_time
= (idle_time_str
&& atoi(idle_time_str
)
1539 ? atoi(idle_time_str
)
1540 : MAC_ENTRY_DEFAULT_IDLE_TIME
);
1542 mac_table_size_str
= smap_get(&br
->cfg
->other_config
, "mac-table-size");
1543 mac_table_size
= (mac_table_size_str
&& atoi(mac_table_size_str
)
1544 ? atoi(mac_table_size_str
)
1547 ofproto_set_mac_table_config(br
->ofproto
, idle_time
, mac_table_size
);
1551 bridge_pick_local_hw_addr(struct bridge
*br
, uint8_t ea
[ETH_ADDR_LEN
],
1552 struct iface
**hw_addr_iface
)
1554 struct hmapx mirror_output_ports
;
1557 bool found_addr
= false;
1561 *hw_addr_iface
= NULL
;
1563 /* Did the user request a particular MAC? */
1564 hwaddr
= smap_get(&br
->cfg
->other_config
, "hwaddr");
1565 if (hwaddr
&& eth_addr_from_string(hwaddr
, ea
)) {
1566 if (eth_addr_is_multicast(ea
)) {
1567 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1568 "address "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(ea
));
1569 } else if (eth_addr_is_zero(ea
)) {
1570 VLOG_ERR("bridge %s: cannot set MAC address to zero", br
->name
);
1576 /* Mirror output ports don't participate in picking the local hardware
1577 * address. ofproto can't help us find out whether a given port is a
1578 * mirror output because we haven't configured mirrors yet, so we need to
1579 * accumulate them ourselves. */
1580 hmapx_init(&mirror_output_ports
);
1581 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
1582 struct ovsrec_mirror
*m
= br
->cfg
->mirrors
[i
];
1583 if (m
->output_port
) {
1584 hmapx_add(&mirror_output_ports
, m
->output_port
);
1588 /* Otherwise choose the minimum non-local MAC address among all of the
1590 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1591 uint8_t iface_ea
[ETH_ADDR_LEN
];
1592 struct iface
*candidate
;
1593 struct iface
*iface
;
1595 /* Mirror output ports don't participate. */
1596 if (hmapx_contains(&mirror_output_ports
, port
->cfg
)) {
1600 /* Choose the MAC address to represent the port. */
1602 if (port
->cfg
->mac
&& eth_addr_from_string(port
->cfg
->mac
, iface_ea
)) {
1603 /* Find the interface with this Ethernet address (if any) so that
1604 * we can provide the correct devname to the caller. */
1605 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1606 uint8_t candidate_ea
[ETH_ADDR_LEN
];
1607 if (!netdev_get_etheraddr(candidate
->netdev
, candidate_ea
)
1608 && eth_addr_equals(iface_ea
, candidate_ea
)) {
1613 /* Choose the interface whose MAC address will represent the port.
1614 * The Linux kernel bonding code always chooses the MAC address of
1615 * the first slave added to a bond, and the Fedora networking
1616 * scripts always add slaves to a bond in alphabetical order, so
1617 * for compatibility we choose the interface with the name that is
1618 * first in alphabetical order. */
1619 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1620 if (!iface
|| strcmp(candidate
->name
, iface
->name
) < 0) {
1625 /* The local port doesn't count (since we're trying to choose its
1626 * MAC address anyway). */
1627 if (iface
->ofp_port
== OFPP_LOCAL
) {
1632 error
= netdev_get_etheraddr(iface
->netdev
, iface_ea
);
1638 /* Compare against our current choice. */
1639 if (!eth_addr_is_multicast(iface_ea
) &&
1640 !eth_addr_is_local(iface_ea
) &&
1641 !eth_addr_is_reserved(iface_ea
) &&
1642 !eth_addr_is_zero(iface_ea
) &&
1643 (!found_addr
|| eth_addr_compare_3way(iface_ea
, ea
) < 0))
1645 memcpy(ea
, iface_ea
, ETH_ADDR_LEN
);
1646 *hw_addr_iface
= iface
;
1652 memcpy(ea
, br
->default_ea
, ETH_ADDR_LEN
);
1653 *hw_addr_iface
= NULL
;
1656 hmapx_destroy(&mirror_output_ports
);
1659 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1660 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1661 * an interface on 'br', then that interface must be passed in as
1662 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1663 * 'hw_addr_iface' must be passed in as a null pointer. */
1665 bridge_pick_datapath_id(struct bridge
*br
,
1666 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
1667 struct iface
*hw_addr_iface
)
1670 * The procedure for choosing a bridge MAC address will, in the most
1671 * ordinary case, also choose a unique MAC that we can use as a datapath
1672 * ID. In some special cases, though, multiple bridges will end up with
1673 * the same MAC address. This is OK for the bridges, but it will confuse
1674 * the OpenFlow controller, because each datapath needs a unique datapath
1677 * Datapath IDs must be unique. It is also very desirable that they be
1678 * stable from one run to the next, so that policy set on a datapath
1681 const char *datapath_id
;
1684 datapath_id
= smap_get(&br
->cfg
->other_config
, "datapath-id");
1685 if (datapath_id
&& dpid_from_string(datapath_id
, &dpid
)) {
1689 if (!hw_addr_iface
) {
1691 * A purely internal bridge, that is, one that has no non-virtual
1692 * network devices on it at all, is difficult because it has no
1693 * natural unique identifier at all.
1695 * When the host is a XenServer, we handle this case by hashing the
1696 * host's UUID with the name of the bridge. Names of bridges are
1697 * persistent across XenServer reboots, although they can be reused if
1698 * an internal network is destroyed and then a new one is later
1699 * created, so this is fairly effective.
1701 * When the host is not a XenServer, we punt by using a random MAC
1702 * address on each run.
1704 const char *host_uuid
= xenserver_get_host_uuid();
1706 char *combined
= xasprintf("%s,%s", host_uuid
, br
->name
);
1707 dpid
= dpid_from_hash(combined
, strlen(combined
));
1713 return eth_addr_to_uint64(bridge_ea
);
1717 dpid_from_hash(const void *data
, size_t n
)
1719 uint8_t hash
[SHA1_DIGEST_SIZE
];
1721 BUILD_ASSERT_DECL(sizeof hash
>= ETH_ADDR_LEN
);
1722 sha1_bytes(data
, n
, hash
);
1723 eth_addr_mark_random(hash
);
1724 return eth_addr_to_uint64(hash
);
1728 iface_refresh_status(struct iface
*iface
)
1732 enum netdev_features current
;
1736 uint8_t mac
[ETH_ADDR_LEN
];
1740 if (iface_is_synthetic(iface
)) {
1746 if (!netdev_get_status(iface
->netdev
, &smap
)) {
1747 ovsrec_interface_set_status(iface
->cfg
, &smap
);
1749 ovsrec_interface_set_status(iface
->cfg
, NULL
);
1752 smap_destroy(&smap
);
1754 error
= netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1755 bps
= !error
? netdev_features_to_bps(current
, 0) : 0;
1757 ovsrec_interface_set_duplex(iface
->cfg
,
1758 netdev_features_is_full_duplex(current
)
1760 ovsrec_interface_set_link_speed(iface
->cfg
, &bps
, 1);
1762 ovsrec_interface_set_duplex(iface
->cfg
, NULL
);
1763 ovsrec_interface_set_link_speed(iface
->cfg
, NULL
, 0);
1766 error
= netdev_get_mtu(iface
->netdev
, &mtu
);
1769 ovsrec_interface_set_mtu(iface
->cfg
, &mtu_64
, 1);
1771 ovsrec_interface_set_mtu(iface
->cfg
, NULL
, 0);
1774 error
= netdev_get_etheraddr(iface
->netdev
, mac
);
1776 char mac_string
[32];
1778 sprintf(mac_string
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
1779 ovsrec_interface_set_mac_in_use(iface
->cfg
, mac_string
);
1781 ovsrec_interface_set_mac_in_use(iface
->cfg
, NULL
);
1784 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1785 * if there is no valid ifindex number. */
1786 ifindex64
= netdev_get_ifindex(iface
->netdev
);
1787 if (ifindex64
< 0) {
1790 ovsrec_interface_set_ifindex(iface
->cfg
, &ifindex64
, 1);
1793 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1796 iface_refresh_cfm_stats(struct iface
*iface
)
1798 const struct ovsrec_interface
*cfg
= iface
->cfg
;
1799 struct ofproto_cfm_status status
;
1801 if (!ofproto_port_get_cfm_status(iface
->port
->bridge
->ofproto
,
1802 iface
->ofp_port
, &status
)) {
1803 ovsrec_interface_set_cfm_fault(cfg
, NULL
, 0);
1804 ovsrec_interface_set_cfm_fault_status(cfg
, NULL
, 0);
1805 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
1806 ovsrec_interface_set_cfm_flap_count(cfg
, NULL
, 0);
1807 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
1808 ovsrec_interface_set_cfm_remote_mpids(cfg
, NULL
, 0);
1810 const char *reasons
[CFM_FAULT_N_REASONS
];
1811 int64_t cfm_health
= status
.health
;
1812 int64_t cfm_flap_count
= status
.flap_count
;
1813 bool faulted
= status
.faults
!= 0;
1816 ovsrec_interface_set_cfm_fault(cfg
, &faulted
, 1);
1819 for (i
= 0; i
< CFM_FAULT_N_REASONS
; i
++) {
1820 int reason
= 1 << i
;
1821 if (status
.faults
& reason
) {
1822 reasons
[j
++] = cfm_fault_reason_to_str(reason
);
1825 ovsrec_interface_set_cfm_fault_status(cfg
, (char **) reasons
, j
);
1827 ovsrec_interface_set_cfm_flap_count(cfg
, &cfm_flap_count
, 1);
1829 if (status
.remote_opstate
>= 0) {
1830 const char *remote_opstate
= status
.remote_opstate
? "up" : "down";
1831 ovsrec_interface_set_cfm_remote_opstate(cfg
, remote_opstate
);
1833 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
1836 ovsrec_interface_set_cfm_remote_mpids(cfg
,
1837 (const int64_t *)status
.rmps
,
1839 if (cfm_health
>= 0) {
1840 ovsrec_interface_set_cfm_health(cfg
, &cfm_health
, 1);
1842 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
1850 iface_refresh_stats(struct iface
*iface
)
1852 #define IFACE_STATS \
1853 IFACE_STAT(rx_packets, "rx_packets") \
1854 IFACE_STAT(tx_packets, "tx_packets") \
1855 IFACE_STAT(rx_bytes, "rx_bytes") \
1856 IFACE_STAT(tx_bytes, "tx_bytes") \
1857 IFACE_STAT(rx_dropped, "rx_dropped") \
1858 IFACE_STAT(tx_dropped, "tx_dropped") \
1859 IFACE_STAT(rx_errors, "rx_errors") \
1860 IFACE_STAT(tx_errors, "tx_errors") \
1861 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1862 IFACE_STAT(rx_over_errors, "rx_over_err") \
1863 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1864 IFACE_STAT(collisions, "collisions")
1866 #define IFACE_STAT(MEMBER, NAME) + 1
1867 enum { N_IFACE_STATS
= IFACE_STATS
};
1869 int64_t values
[N_IFACE_STATS
];
1870 char *keys
[N_IFACE_STATS
];
1873 struct netdev_stats stats
;
1875 if (iface_is_synthetic(iface
)) {
1879 /* Intentionally ignore return value, since errors will set 'stats' to
1880 * all-1s, and we will deal with that correctly below. */
1881 netdev_get_stats(iface
->netdev
, &stats
);
1883 /* Copy statistics into keys[] and values[]. */
1885 #define IFACE_STAT(MEMBER, NAME) \
1886 if (stats.MEMBER != UINT64_MAX) { \
1888 values[n] = stats.MEMBER; \
1893 ovs_assert(n
<= N_IFACE_STATS
);
1895 ovsrec_interface_set_statistics(iface
->cfg
, keys
, values
, n
);
1900 br_refresh_stp_status(struct bridge
*br
)
1902 struct smap smap
= SMAP_INITIALIZER(&smap
);
1903 struct ofproto
*ofproto
= br
->ofproto
;
1904 struct ofproto_stp_status status
;
1906 if (ofproto_get_stp_status(ofproto
, &status
)) {
1910 if (!status
.enabled
) {
1911 ovsrec_bridge_set_status(br
->cfg
, NULL
);
1915 smap_add_format(&smap
, "stp_bridge_id", STP_ID_FMT
,
1916 STP_ID_ARGS(status
.bridge_id
));
1917 smap_add_format(&smap
, "stp_designated_root", STP_ID_FMT
,
1918 STP_ID_ARGS(status
.designated_root
));
1919 smap_add_format(&smap
, "stp_root_path_cost", "%d", status
.root_path_cost
);
1921 ovsrec_bridge_set_status(br
->cfg
, &smap
);
1922 smap_destroy(&smap
);
1926 port_refresh_stp_status(struct port
*port
)
1928 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
1929 struct iface
*iface
;
1930 struct ofproto_port_stp_status status
;
1933 if (port_is_synthetic(port
)) {
1937 /* STP doesn't currently support bonds. */
1938 if (!list_is_singleton(&port
->ifaces
)) {
1939 ovsrec_port_set_status(port
->cfg
, NULL
);
1943 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
1944 if (ofproto_port_get_stp_status(ofproto
, iface
->ofp_port
, &status
)) {
1948 if (!status
.enabled
) {
1949 ovsrec_port_set_status(port
->cfg
, NULL
);
1953 /* Set Status column. */
1955 smap_add_format(&smap
, "stp_port_id", STP_PORT_ID_FMT
, status
.port_id
);
1956 smap_add(&smap
, "stp_state", stp_state_name(status
.state
));
1957 smap_add_format(&smap
, "stp_sec_in_state", "%u", status
.sec_in_state
);
1958 smap_add(&smap
, "stp_role", stp_role_name(status
.role
));
1959 ovsrec_port_set_status(port
->cfg
, &smap
);
1960 smap_destroy(&smap
);
1964 port_refresh_stp_stats(struct port
*port
)
1966 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
1967 struct iface
*iface
;
1968 struct ofproto_port_stp_stats stats
;
1970 int64_t int_values
[3];
1972 if (port_is_synthetic(port
)) {
1976 /* STP doesn't currently support bonds. */
1977 if (!list_is_singleton(&port
->ifaces
)) {
1981 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
1982 if (ofproto_port_get_stp_stats(ofproto
, iface
->ofp_port
, &stats
)) {
1986 if (!stats
.enabled
) {
1987 ovsrec_port_set_statistics(port
->cfg
, NULL
, NULL
, 0);
1991 /* Set Statistics column. */
1992 keys
[0] = "stp_tx_count";
1993 int_values
[0] = stats
.tx_count
;
1994 keys
[1] = "stp_rx_count";
1995 int_values
[1] = stats
.rx_count
;
1996 keys
[2] = "stp_error_count";
1997 int_values
[2] = stats
.error_count
;
1999 ovsrec_port_set_statistics(port
->cfg
, keys
, int_values
,
2000 ARRAY_SIZE(int_values
));
2004 enable_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2006 return smap_get_bool(&cfg
->other_config
, "enable-statistics", false);
2010 reconfigure_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2012 bool enable
= enable_system_stats(cfg
);
2014 system_stats_enable(enable
);
2016 ovsrec_open_vswitch_set_statistics(cfg
, NULL
);
2021 run_system_stats(void)
2023 const struct ovsrec_open_vswitch
*cfg
= ovsrec_open_vswitch_first(idl
);
2026 stats
= system_stats_run();
2028 struct ovsdb_idl_txn
*txn
;
2029 struct ovsdb_datum datum
;
2031 txn
= ovsdb_idl_txn_create(idl
);
2032 ovsdb_datum_from_smap(&datum
, stats
);
2033 ovsdb_idl_txn_write(&cfg
->header_
, &ovsrec_open_vswitch_col_statistics
,
2035 ovsdb_idl_txn_commit(txn
);
2036 ovsdb_idl_txn_destroy(txn
);
2043 ofp12_controller_role_to_str(enum ofp12_controller_role role
)
2046 case OFPCR12_ROLE_EQUAL
:
2048 case OFPCR12_ROLE_MASTER
:
2050 case OFPCR12_ROLE_SLAVE
:
2052 case OFPCR12_ROLE_NOCHANGE
:
2054 return "*** INVALID ROLE ***";
2059 refresh_controller_status(void)
2063 const struct ovsrec_controller
*cfg
;
2067 /* Accumulate status for controllers on all bridges. */
2068 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2069 ofproto_get_ofproto_controller_info(br
->ofproto
, &info
);
2072 /* Update each controller in the database with current status. */
2073 OVSREC_CONTROLLER_FOR_EACH(cfg
, idl
) {
2074 struct ofproto_controller_info
*cinfo
=
2075 shash_find_data(&info
, cfg
->target
);
2078 struct smap smap
= SMAP_INITIALIZER(&smap
);
2079 const char **values
= cinfo
->pairs
.values
;
2080 const char **keys
= cinfo
->pairs
.keys
;
2083 for (i
= 0; i
< cinfo
->pairs
.n
; i
++) {
2084 smap_add(&smap
, keys
[i
], values
[i
]);
2087 ovsrec_controller_set_is_connected(cfg
, cinfo
->is_connected
);
2088 ovsrec_controller_set_role(cfg
, ofp12_controller_role_to_str(
2090 ovsrec_controller_set_status(cfg
, &smap
);
2091 smap_destroy(&smap
);
2093 ovsrec_controller_set_is_connected(cfg
, false);
2094 ovsrec_controller_set_role(cfg
, NULL
);
2095 ovsrec_controller_set_status(cfg
, NULL
);
2099 ofproto_free_ofproto_controller_info(&info
);
2104 * Some information in the database must be kept as up-to-date as possible to
2105 * allow controllers to respond rapidly to network outages. We call these
2106 * statistics "instant" stats.
2108 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2109 * assuming that they've changed. The only means we have to determine whether
2110 * they have changed are:
2112 * - Try to commit changes to the database. If nothing changed, then
2113 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2116 * - instant_stats_run() is called late in the run loop, after anything that
2117 * might change any of the instant stats.
2119 * We use these two facts together to avoid waking the process up every
2120 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2123 /* Minimum interval between writing updates to the instant stats to the
2125 #define INSTANT_INTERVAL_MSEC 100
2127 /* Current instant stats database transaction, NULL if there is no ongoing
2129 static struct ovsdb_idl_txn
*instant_txn
;
2131 /* Next time (in msec on monotonic clock) at which we will update the instant
2133 static long long int instant_next_txn
= LLONG_MIN
;
2135 /* True if the run loop has run since we last saw that the instant stats were
2136 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2137 * to refresh the instant stats. */
2138 static bool instant_stats_could_have_changed
;
2141 instant_stats_run(void)
2143 enum ovsdb_idl_txn_status status
;
2145 instant_stats_could_have_changed
= true;
2151 if (time_msec() < instant_next_txn
) {
2154 instant_next_txn
= time_msec() + INSTANT_INTERVAL_MSEC
;
2156 seq
= seq_read(connectivity_seq_get());
2157 if (seq
== connectivity_seqno
) {
2160 connectivity_seqno
= seq
;
2162 instant_txn
= ovsdb_idl_txn_create(idl
);
2163 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2164 struct iface
*iface
;
2167 br_refresh_stp_status(br
);
2169 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2170 port_refresh_stp_status(port
);
2173 HMAP_FOR_EACH (iface
, name_node
, &br
->iface_by_name
) {
2174 enum netdev_flags flags
;
2176 const char *link_state
;
2177 int64_t link_resets
;
2180 if (iface_is_synthetic(iface
)) {
2184 current
= ofproto_port_is_lacp_current(br
->ofproto
,
2188 ovsrec_interface_set_lacp_current(iface
->cfg
, &bl
, 1);
2190 ovsrec_interface_set_lacp_current(iface
->cfg
, NULL
, 0);
2193 error
= netdev_get_flags(iface
->netdev
, &flags
);
2195 const char *state
= flags
& NETDEV_UP
? "up" : "down";
2196 ovsrec_interface_set_admin_state(iface
->cfg
, state
);
2198 ovsrec_interface_set_admin_state(iface
->cfg
, NULL
);
2201 link_state
= netdev_get_carrier(iface
->netdev
) ? "up" : "down";
2202 ovsrec_interface_set_link_state(iface
->cfg
, link_state
);
2204 link_resets
= netdev_get_carrier_resets(iface
->netdev
);
2205 ovsrec_interface_set_link_resets(iface
->cfg
, &link_resets
, 1);
2207 iface_refresh_cfm_stats(iface
);
2210 ofproto_port_get_bfd_status(br
->ofproto
, iface
->ofp_port
,
2212 ovsrec_interface_set_bfd_status(iface
->cfg
, &smap
);
2213 smap_destroy(&smap
);
2218 status
= ovsdb_idl_txn_commit(instant_txn
);
2219 if (status
!= TXN_INCOMPLETE
) {
2220 ovsdb_idl_txn_destroy(instant_txn
);
2223 if (status
== TXN_UNCHANGED
) {
2224 instant_stats_could_have_changed
= false;
2229 instant_stats_wait(void)
2232 ovsdb_idl_txn_wait(instant_txn
);
2233 } else if (instant_stats_could_have_changed
) {
2234 poll_timer_wait_until(instant_next_txn
);
2245 /* Let each datapath type do the work that it needs to do. */
2247 ofproto_enumerate_types(&types
);
2248 SSET_FOR_EACH (type
, &types
) {
2249 ofproto_type_run(type
);
2251 sset_destroy(&types
);
2253 /* Let each bridge do the work that it needs to do. */
2254 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2255 ofproto_run(br
->ofproto
);
2262 static struct ovsrec_open_vswitch null_cfg
;
2263 const struct ovsrec_open_vswitch
*cfg
;
2265 bool vlan_splinters_changed
;
2268 ovsrec_open_vswitch_init(&null_cfg
);
2272 if (ovsdb_idl_is_lock_contended(idl
)) {
2273 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2274 struct bridge
*br
, *next_br
;
2276 VLOG_ERR_RL(&rl
, "another ovs-vswitchd process is running, "
2277 "disabling this process (pid %ld) until it goes away",
2278 (long int) getpid());
2280 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
2283 /* Since we will not be running system_stats_run() in this process
2284 * with the current situation of multiple ovs-vswitchd daemons,
2285 * disable system stats collection. */
2286 system_stats_enable(false);
2288 } else if (!ovsdb_idl_has_lock(idl
)) {
2291 cfg
= ovsrec_open_vswitch_first(idl
);
2293 /* Initialize the ofproto library. This only needs to run once, but
2294 * it must be done after the configuration is set. If the
2295 * initialization has already occurred, bridge_init_ofproto()
2296 * returns immediately. */
2297 bridge_init_ofproto(cfg
);
2299 /* Once the value of flow-restore-wait is false, we no longer should
2300 * check its value from the database. */
2301 if (cfg
&& ofproto_get_flow_restore_wait()) {
2302 ofproto_set_flow_restore_wait(smap_get_bool(&cfg
->other_config
,
2303 "flow-restore-wait", false));
2308 /* Re-configure SSL. We do this on every trip through the main loop,
2309 * instead of just when the database changes, because the contents of the
2310 * key and certificate files can change without the database changing.
2312 * We do this before bridge_reconfigure() because that function might
2313 * initiate SSL connections and thus requires SSL to be configured. */
2314 if (cfg
&& cfg
->ssl
) {
2315 const struct ovsrec_ssl
*ssl
= cfg
->ssl
;
2317 stream_ssl_set_key_and_cert(ssl
->private_key
, ssl
->certificate
);
2318 stream_ssl_set_ca_cert_file(ssl
->ca_cert
, ssl
->bootstrap_ca_cert
);
2321 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2322 * usage has changed. */
2323 vlan_splinters_changed
= false;
2324 if (vlan_splinters_enabled_anywhere
) {
2325 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2326 if (ofproto_has_vlan_usage_changed(br
->ofproto
)) {
2327 vlan_splinters_changed
= true;
2333 if (ovsdb_idl_get_seqno(idl
) != idl_seqno
|| vlan_splinters_changed
) {
2334 struct ovsdb_idl_txn
*txn
;
2336 idl_seqno
= ovsdb_idl_get_seqno(idl
);
2337 txn
= ovsdb_idl_txn_create(idl
);
2338 bridge_reconfigure(cfg
? cfg
: &null_cfg
);
2341 ovsrec_open_vswitch_set_cur_cfg(cfg
, cfg
->next_cfg
);
2344 /* If we are completing our initial configuration for this run
2345 * of ovs-vswitchd, then keep the transaction around to monitor
2346 * it for completion. */
2347 if (initial_config_done
) {
2348 ovsdb_idl_txn_commit(txn
);
2349 ovsdb_idl_txn_destroy(txn
);
2351 initial_config_done
= true;
2352 daemonize_txn
= txn
;
2356 if (daemonize_txn
) {
2357 enum ovsdb_idl_txn_status status
= ovsdb_idl_txn_commit(daemonize_txn
);
2358 if (status
!= TXN_INCOMPLETE
) {
2359 ovsdb_idl_txn_destroy(daemonize_txn
);
2360 daemonize_txn
= NULL
;
2362 /* ovs-vswitchd has completed initialization, so allow the
2363 * process that forked us to exit successfully. */
2364 daemonize_complete();
2366 vlog_enable_async();
2368 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name
, VERSION
);
2372 /* Refresh interface and mirror stats if necessary. */
2373 if (time_msec() >= iface_stats_timer
) {
2375 struct ovsdb_idl_txn
*txn
;
2377 txn
= ovsdb_idl_txn_create(idl
);
2378 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2382 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2383 struct iface
*iface
;
2385 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2386 iface_refresh_stats(iface
);
2387 iface_refresh_status(iface
);
2390 port_refresh_stp_stats(port
);
2393 HMAP_FOR_EACH (m
, hmap_node
, &br
->mirrors
) {
2394 mirror_refresh_stats(m
);
2398 refresh_controller_status();
2399 ovsdb_idl_txn_commit(txn
);
2400 ovsdb_idl_txn_destroy(txn
); /* XXX */
2403 iface_stats_timer
= time_msec() + IFACE_STATS_INTERVAL
;
2407 instant_stats_run();
2416 ovsdb_idl_wait(idl
);
2417 if (daemonize_txn
) {
2418 ovsdb_idl_txn_wait(daemonize_txn
);
2422 ofproto_enumerate_types(&types
);
2423 SSET_FOR_EACH (type
, &types
) {
2424 ofproto_type_wait(type
);
2426 sset_destroy(&types
);
2428 if (!hmap_is_empty(&all_bridges
)) {
2431 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2432 ofproto_wait(br
->ofproto
);
2434 poll_timer_wait_until(iface_stats_timer
);
2437 system_stats_wait();
2438 instant_stats_wait();
2441 /* Adds some memory usage statistics for bridges into 'usage', for use with
2442 * memory_report(). */
2444 bridge_get_memory_usage(struct simap
*usage
)
2451 ofproto_enumerate_types(&types
);
2452 SSET_FOR_EACH (type
, &types
) {
2453 ofproto_type_get_memory_usage(type
, usage
);
2455 sset_destroy(&types
);
2457 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2458 ofproto_get_memory_usage(br
->ofproto
, usage
);
2462 /* QoS unixctl user interface functions. */
2464 struct qos_unixctl_show_cbdata
{
2466 struct iface
*iface
;
2470 qos_unixctl_show_queue(unsigned int queue_id
,
2471 const struct smap
*details
,
2472 struct iface
*iface
,
2475 struct netdev_queue_stats stats
;
2476 struct smap_node
*node
;
2479 ds_put_cstr(ds
, "\n");
2481 ds_put_format(ds
, "Queue %u:\n", queue_id
);
2483 ds_put_cstr(ds
, "Default:\n");
2486 SMAP_FOR_EACH (node
, details
) {
2487 ds_put_format(ds
, "\t%s: %s\n", node
->key
, node
->value
);
2490 error
= netdev_get_queue_stats(iface
->netdev
, queue_id
, &stats
);
2492 if (stats
.tx_packets
!= UINT64_MAX
) {
2493 ds_put_format(ds
, "\ttx_packets: %"PRIu64
"\n", stats
.tx_packets
);
2496 if (stats
.tx_bytes
!= UINT64_MAX
) {
2497 ds_put_format(ds
, "\ttx_bytes: %"PRIu64
"\n", stats
.tx_bytes
);
2500 if (stats
.tx_errors
!= UINT64_MAX
) {
2501 ds_put_format(ds
, "\ttx_errors: %"PRIu64
"\n", stats
.tx_errors
);
2504 ds_put_format(ds
, "\tFailed to get statistics for queue %u: %s",
2505 queue_id
, ovs_strerror(error
));
2510 qos_unixctl_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
2511 const char *argv
[], void *aux OVS_UNUSED
)
2513 struct ds ds
= DS_EMPTY_INITIALIZER
;
2514 struct smap smap
= SMAP_INITIALIZER(&smap
);
2515 struct iface
*iface
;
2517 struct smap_node
*node
;
2519 iface
= iface_find(argv
[1]);
2521 unixctl_command_reply_error(conn
, "no such interface");
2525 netdev_get_qos(iface
->netdev
, &type
, &smap
);
2527 if (*type
!= '\0') {
2528 struct netdev_queue_dump dump
;
2529 struct smap details
;
2530 unsigned int queue_id
;
2532 ds_put_format(&ds
, "QoS: %s %s\n", iface
->name
, type
);
2534 SMAP_FOR_EACH (node
, &smap
) {
2535 ds_put_format(&ds
, "%s: %s\n", node
->key
, node
->value
);
2538 smap_init(&details
);
2539 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
2540 qos_unixctl_show_queue(queue_id
, &details
, iface
, &ds
);
2542 smap_destroy(&details
);
2544 unixctl_command_reply(conn
, ds_cstr(&ds
));
2546 ds_put_format(&ds
, "QoS not configured on %s\n", iface
->name
);
2547 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
2550 smap_destroy(&smap
);
2554 /* Bridge reconfiguration functions. */
2556 bridge_create(const struct ovsrec_bridge
*br_cfg
)
2560 ovs_assert(!bridge_lookup(br_cfg
->name
));
2561 br
= xzalloc(sizeof *br
);
2563 br
->name
= xstrdup(br_cfg
->name
);
2564 br
->type
= xstrdup(ofproto_normalize_type(br_cfg
->datapath_type
));
2567 /* Derive the default Ethernet address from the bridge's UUID. This should
2568 * be unique and it will be stable between ovs-vswitchd runs. */
2569 memcpy(br
->default_ea
, &br_cfg
->header_
.uuid
, ETH_ADDR_LEN
);
2570 eth_addr_mark_random(br
->default_ea
);
2572 hmap_init(&br
->ports
);
2573 hmap_init(&br
->ifaces
);
2574 hmap_init(&br
->iface_by_name
);
2575 hmap_init(&br
->mirrors
);
2577 hmap_insert(&all_bridges
, &br
->node
, hash_string(br
->name
, 0));
2581 bridge_destroy(struct bridge
*br
)
2584 struct mirror
*mirror
, *next_mirror
;
2585 struct port
*port
, *next_port
;
2587 HMAP_FOR_EACH_SAFE (port
, next_port
, hmap_node
, &br
->ports
) {
2590 HMAP_FOR_EACH_SAFE (mirror
, next_mirror
, hmap_node
, &br
->mirrors
) {
2591 mirror_destroy(mirror
);
2594 hmap_remove(&all_bridges
, &br
->node
);
2595 ofproto_destroy(br
->ofproto
);
2596 hmap_destroy(&br
->ifaces
);
2597 hmap_destroy(&br
->ports
);
2598 hmap_destroy(&br
->iface_by_name
);
2599 hmap_destroy(&br
->mirrors
);
2606 static struct bridge
*
2607 bridge_lookup(const char *name
)
2611 HMAP_FOR_EACH_WITH_HASH (br
, node
, hash_string(name
, 0), &all_bridges
) {
2612 if (!strcmp(br
->name
, name
)) {
2619 /* Handle requests for a listing of all flows known by the OpenFlow
2620 * stack, including those normally hidden. */
2622 bridge_unixctl_dump_flows(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
2623 const char *argv
[], void *aux OVS_UNUSED
)
2628 br
= bridge_lookup(argv
[1]);
2630 unixctl_command_reply_error(conn
, "Unknown bridge");
2635 ofproto_get_all_flows(br
->ofproto
, &results
);
2637 unixctl_command_reply(conn
, ds_cstr(&results
));
2638 ds_destroy(&results
);
2641 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2642 * connections and reconnect. If BRIDGE is not specified, then all bridges
2643 * drop their controller connections and reconnect. */
2645 bridge_unixctl_reconnect(struct unixctl_conn
*conn
, int argc
,
2646 const char *argv
[], void *aux OVS_UNUSED
)
2650 br
= bridge_lookup(argv
[1]);
2652 unixctl_command_reply_error(conn
, "Unknown bridge");
2655 ofproto_reconnect_controllers(br
->ofproto
);
2657 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2658 ofproto_reconnect_controllers(br
->ofproto
);
2661 unixctl_command_reply(conn
, NULL
);
2665 bridge_get_controllers(const struct bridge
*br
,
2666 struct ovsrec_controller
***controllersp
)
2668 struct ovsrec_controller
**controllers
;
2669 size_t n_controllers
;
2671 controllers
= br
->cfg
->controller
;
2672 n_controllers
= br
->cfg
->n_controller
;
2674 if (n_controllers
== 1 && !strcmp(controllers
[0]->target
, "none")) {
2680 *controllersp
= controllers
;
2682 return n_controllers
;
2686 bridge_collect_wanted_ports(struct bridge
*br
,
2687 const unsigned long int *splinter_vlans
,
2688 struct shash
*wanted_ports
)
2692 shash_init(wanted_ports
);
2694 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
2695 const char *name
= br
->cfg
->ports
[i
]->name
;
2696 if (!shash_add_once(wanted_ports
, name
, br
->cfg
->ports
[i
])) {
2697 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2701 if (bridge_get_controllers(br
, NULL
)
2702 && !shash_find(wanted_ports
, br
->name
)) {
2703 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2704 br
->name
, br
->name
);
2706 ovsrec_interface_init(&br
->synth_local_iface
);
2707 ovsrec_port_init(&br
->synth_local_port
);
2709 br
->synth_local_port
.interfaces
= &br
->synth_local_ifacep
;
2710 br
->synth_local_port
.n_interfaces
= 1;
2711 br
->synth_local_port
.name
= br
->name
;
2713 br
->synth_local_iface
.name
= br
->name
;
2714 br
->synth_local_iface
.type
= "internal";
2716 br
->synth_local_ifacep
= &br
->synth_local_iface
;
2718 shash_add(wanted_ports
, br
->name
, &br
->synth_local_port
);
2721 if (splinter_vlans
) {
2722 add_vlan_splinter_ports(br
, splinter_vlans
, wanted_ports
);
2726 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2727 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2728 * 'br' needs to complete its configuration. */
2730 bridge_del_ports(struct bridge
*br
, const struct shash
*wanted_ports
)
2732 struct shash_node
*port_node
;
2733 struct port
*port
, *next
;
2735 /* Get rid of deleted ports.
2736 * Get rid of deleted interfaces on ports that still exist. */
2737 HMAP_FOR_EACH_SAFE (port
, next
, hmap_node
, &br
->ports
) {
2738 port
->cfg
= shash_find_data(wanted_ports
, port
->name
);
2742 port_del_ifaces(port
);
2746 /* Update iface->cfg and iface->type in interfaces that still exist. */
2747 SHASH_FOR_EACH (port_node
, wanted_ports
) {
2748 const struct ovsrec_port
*port
= port_node
->data
;
2751 for (i
= 0; i
< port
->n_interfaces
; i
++) {
2752 const struct ovsrec_interface
*cfg
= port
->interfaces
[i
];
2753 struct iface
*iface
= iface_lookup(br
, cfg
->name
);
2754 const char *type
= iface_get_type(cfg
, br
->cfg
);
2759 } else if (!strcmp(type
, "null")) {
2760 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2761 " may be removed in February 2013. Please email"
2762 " dev@openvswitch.org with concerns.",
2765 /* We will add new interfaces later. */
2771 /* Initializes 'oc' appropriately as a management service controller for
2774 * The caller must free oc->target when it is no longer needed. */
2776 bridge_ofproto_controller_for_mgmt(const struct bridge
*br
,
2777 struct ofproto_controller
*oc
)
2779 oc
->target
= xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br
->name
);
2780 oc
->max_backoff
= 0;
2781 oc
->probe_interval
= 60;
2782 oc
->band
= OFPROTO_OUT_OF_BAND
;
2784 oc
->burst_limit
= 0;
2785 oc
->enable_async_msgs
= true;
2788 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2790 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller
*c
,
2791 struct ofproto_controller
*oc
)
2795 oc
->target
= c
->target
;
2796 oc
->max_backoff
= c
->max_backoff
? *c
->max_backoff
/ 1000 : 8;
2797 oc
->probe_interval
= c
->inactivity_probe
? *c
->inactivity_probe
/ 1000 : 5;
2798 oc
->band
= (!c
->connection_mode
|| !strcmp(c
->connection_mode
, "in-band")
2799 ? OFPROTO_IN_BAND
: OFPROTO_OUT_OF_BAND
);
2800 oc
->rate_limit
= c
->controller_rate_limit
? *c
->controller_rate_limit
: 0;
2801 oc
->burst_limit
= (c
->controller_burst_limit
2802 ? *c
->controller_burst_limit
: 0);
2803 oc
->enable_async_msgs
= (!c
->enable_async_messages
2804 || *c
->enable_async_messages
);
2805 dscp
= smap_get_int(&c
->other_config
, "dscp", DSCP_DEFAULT
);
2806 if (dscp
< 0 || dscp
> 63) {
2807 dscp
= DSCP_DEFAULT
;
2812 /* Configures the IP stack for 'br''s local interface properly according to the
2813 * configuration in 'c'. */
2815 bridge_configure_local_iface_netdev(struct bridge
*br
,
2816 struct ovsrec_controller
*c
)
2818 struct netdev
*netdev
;
2819 struct in_addr mask
, gateway
;
2821 struct iface
*local_iface
;
2824 /* If there's no local interface or no IP address, give up. */
2825 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
2826 if (!local_iface
|| !c
->local_ip
2827 || !inet_pton(AF_INET
, c
->local_ip
, &ip
)) {
2831 /* Bring up the local interface. */
2832 netdev
= local_iface
->netdev
;
2833 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
2835 /* Configure the IP address and netmask. */
2836 if (!c
->local_netmask
2837 || !inet_pton(AF_INET
, c
->local_netmask
, &mask
)
2839 mask
.s_addr
= guess_netmask(ip
.s_addr
);
2841 if (!netdev_set_in4(netdev
, ip
, mask
)) {
2842 VLOG_INFO("bridge %s: configured IP address "IP_FMT
", netmask "IP_FMT
,
2843 br
->name
, IP_ARGS(ip
.s_addr
), IP_ARGS(mask
.s_addr
));
2846 /* Configure the default gateway. */
2847 if (c
->local_gateway
2848 && inet_pton(AF_INET
, c
->local_gateway
, &gateway
)
2849 && gateway
.s_addr
) {
2850 if (!netdev_add_router(netdev
, gateway
)) {
2851 VLOG_INFO("bridge %s: configured gateway "IP_FMT
,
2852 br
->name
, IP_ARGS(gateway
.s_addr
));
2857 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2858 * in either string are treated as equal to any number of slashes in the other,
2859 * e.g. "x///y" is equal to "x/y".
2861 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2862 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2863 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2864 * 'b' against a prefix of 'a'.
2867 equal_pathnames(const char *a
, const char *b
, size_t b_stoplen
)
2869 const char *b_start
= b
;
2871 if (b
- b_start
>= b_stoplen
) {
2873 } else if (*a
!= *b
) {
2875 } else if (*a
== '/') {
2876 a
+= strspn(a
, "/");
2877 b
+= strspn(b
, "/");
2878 } else if (*a
== '\0') {
2888 bridge_configure_remotes(struct bridge
*br
,
2889 const struct sockaddr_in
*managers
, size_t n_managers
)
2891 bool disable_in_band
;
2893 struct ovsrec_controller
**controllers
;
2894 size_t n_controllers
;
2896 enum ofproto_fail_mode fail_mode
;
2898 struct ofproto_controller
*ocs
;
2902 /* Check if we should disable in-band control on this bridge. */
2903 disable_in_band
= smap_get_bool(&br
->cfg
->other_config
, "disable-in-band",
2906 /* Set OpenFlow queue ID for in-band control. */
2907 ofproto_set_in_band_queue(br
->ofproto
,
2908 smap_get_int(&br
->cfg
->other_config
,
2909 "in-band-queue", -1));
2911 if (disable_in_band
) {
2912 ofproto_set_extra_in_band_remotes(br
->ofproto
, NULL
, 0);
2914 ofproto_set_extra_in_band_remotes(br
->ofproto
, managers
, n_managers
);
2917 n_controllers
= bridge_get_controllers(br
, &controllers
);
2919 ocs
= xmalloc((n_controllers
+ 1) * sizeof *ocs
);
2922 bridge_ofproto_controller_for_mgmt(br
, &ocs
[n_ocs
++]);
2923 for (i
= 0; i
< n_controllers
; i
++) {
2924 struct ovsrec_controller
*c
= controllers
[i
];
2926 if (!strncmp(c
->target
, "punix:", 6)
2927 || !strncmp(c
->target
, "unix:", 5)) {
2928 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2931 if (!strncmp(c
->target
, "unix:", 5)) {
2932 /* Connect to a listening socket */
2933 whitelist
= xasprintf("unix:%s/", ovs_rundir());
2934 if (strchr(c
->target
, '/') &&
2935 !equal_pathnames(c
->target
, whitelist
,
2936 strlen(whitelist
))) {
2937 /* Absolute path specified, but not in ovs_rundir */
2938 VLOG_ERR_RL(&rl
, "bridge %s: Not connecting to socket "
2939 "controller \"%s\" due to possibility for "
2940 "remote exploit. Instead, specify socket "
2941 "in whitelisted \"%s\" or connect to "
2942 "\"unix:%s/%s.mgmt\" (which is always "
2943 "available without special configuration).",
2944 br
->name
, c
->target
, whitelist
,
2945 ovs_rundir(), br
->name
);
2950 whitelist
= xasprintf("punix:%s/%s.controller",
2951 ovs_rundir(), br
->name
);
2952 if (!equal_pathnames(c
->target
, whitelist
, SIZE_MAX
)) {
2953 /* Prevent remote ovsdb-server users from accessing
2954 * arbitrary Unix domain sockets and overwriting arbitrary
2956 VLOG_ERR_RL(&rl
, "bridge %s: Not adding Unix domain socket "
2957 "controller \"%s\" due to possibility of "
2958 "overwriting local files. Instead, specify "
2959 "whitelisted \"%s\" or connect to "
2960 "\"unix:%s/%s.mgmt\" (which is always "
2961 "available without special configuration).",
2962 br
->name
, c
->target
, whitelist
,
2963 ovs_rundir(), br
->name
);
2972 bridge_configure_local_iface_netdev(br
, c
);
2973 bridge_ofproto_controller_from_ovsrec(c
, &ocs
[n_ocs
]);
2974 if (disable_in_band
) {
2975 ocs
[n_ocs
].band
= OFPROTO_OUT_OF_BAND
;
2980 ofproto_set_controllers(br
->ofproto
, ocs
, n_ocs
,
2981 bridge_get_allowed_versions(br
));
2982 free(ocs
[0].target
); /* From bridge_ofproto_controller_for_mgmt(). */
2985 /* Set the fail-mode. */
2986 fail_mode
= !br
->cfg
->fail_mode
2987 || !strcmp(br
->cfg
->fail_mode
, "standalone")
2988 ? OFPROTO_FAIL_STANDALONE
2989 : OFPROTO_FAIL_SECURE
;
2990 ofproto_set_fail_mode(br
->ofproto
, fail_mode
);
2992 /* Configure OpenFlow controller connection snooping. */
2993 if (!ofproto_has_snoops(br
->ofproto
)) {
2997 sset_add_and_free(&snoops
, xasprintf("punix:%s/%s.snoop",
2998 ovs_rundir(), br
->name
));
2999 ofproto_set_snoops(br
->ofproto
, &snoops
);
3000 sset_destroy(&snoops
);
3005 bridge_configure_tables(struct bridge
*br
)
3007 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3011 n_tables
= ofproto_get_n_tables(br
->ofproto
);
3013 for (i
= 0; i
< n_tables
; i
++) {
3014 struct ofproto_table_settings s
;
3017 s
.max_flows
= UINT_MAX
;
3020 s
.n_prefix_fields
= 0;
3021 memset(s
.prefix_fields
, ~0, sizeof(s
.prefix_fields
));
3023 if (j
< br
->cfg
->n_flow_tables
&& i
== br
->cfg
->key_flow_tables
[j
]) {
3024 struct ovsrec_flow_table
*cfg
= br
->cfg
->value_flow_tables
[j
++];
3027 if (cfg
->n_flow_limit
&& *cfg
->flow_limit
< UINT_MAX
) {
3028 s
.max_flows
= *cfg
->flow_limit
;
3030 if (cfg
->overflow_policy
3031 && !strcmp(cfg
->overflow_policy
, "evict")) {
3033 s
.groups
= xmalloc(cfg
->n_groups
* sizeof *s
.groups
);
3034 for (k
= 0; k
< cfg
->n_groups
; k
++) {
3035 const char *string
= cfg
->groups
[k
];
3038 msg
= mf_parse_subfield__(&s
.groups
[k
], &string
);
3040 VLOG_WARN_RL(&rl
, "bridge %s table %d: error parsing "
3041 "'groups' (%s)", br
->name
, i
, msg
);
3043 } else if (*string
) {
3044 VLOG_WARN_RL(&rl
, "bridge %s table %d: 'groups' "
3045 "element '%s' contains trailing garbage",
3046 br
->name
, i
, cfg
->groups
[k
]);
3052 /* Prefix lookup fields. */
3053 s
.n_prefix_fields
= 0;
3054 for (k
= 0; k
< cfg
->n_prefixes
; k
++) {
3055 const char *name
= cfg
->prefixes
[k
];
3056 const struct mf_field
*mf
= mf_from_name(name
);
3058 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3062 if (mf
->flow_be32ofs
< 0 || mf
->n_bits
% 32) {
3063 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3064 "%s", br
->name
, name
);
3067 if (s
.n_prefix_fields
>= ARRAY_SIZE(s
.prefix_fields
)) {
3068 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3069 "field not used: %s", br
->name
, name
);
3072 s
.prefix_fields
[s
.n_prefix_fields
++] = mf
->id
;
3074 if (s
.n_prefix_fields
> 0) {
3076 struct ds ds
= DS_EMPTY_INITIALIZER
;
3077 for (k
= 0; k
< s
.n_prefix_fields
; k
++) {
3079 ds_put_char(&ds
, ',');
3081 ds_put_cstr(&ds
, mf_from_id(s
.prefix_fields
[k
])->name
);
3083 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3084 br
->name
, i
, ds_cstr(&ds
));
3089 ofproto_configure_table(br
->ofproto
, i
, &s
);
3093 for (; j
< br
->cfg
->n_flow_tables
; j
++) {
3094 VLOG_WARN_RL(&rl
, "bridge %s: ignoring configuration for flow table "
3095 "%"PRId64
" not supported by this datapath", br
->name
,
3096 br
->cfg
->key_flow_tables
[j
]);
3101 bridge_configure_dp_desc(struct bridge
*br
)
3103 ofproto_set_dp_desc(br
->ofproto
,
3104 smap_get(&br
->cfg
->other_config
, "dp-desc"));
3107 /* Port functions. */
3109 static void iface_destroy__(struct iface
*);
3111 static struct port
*
3112 port_create(struct bridge
*br
, const struct ovsrec_port
*cfg
)
3116 port
= xzalloc(sizeof *port
);
3118 port
->name
= xstrdup(cfg
->name
);
3120 list_init(&port
->ifaces
);
3122 hmap_insert(&br
->ports
, &port
->hmap_node
, hash_string(port
->name
, 0));
3126 /* Deletes interfaces from 'port' that are no longer configured for it. */
3128 port_del_ifaces(struct port
*port
)
3130 struct iface
*iface
, *next
;
3131 struct sset new_ifaces
;
3134 /* Collect list of new interfaces. */
3135 sset_init(&new_ifaces
);
3136 for (i
= 0; i
< port
->cfg
->n_interfaces
; i
++) {
3137 const char *name
= port
->cfg
->interfaces
[i
]->name
;
3138 const char *type
= port
->cfg
->interfaces
[i
]->type
;
3139 if (strcmp(type
, "null")) {
3140 sset_add(&new_ifaces
, name
);
3144 /* Get rid of deleted interfaces. */
3145 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
3146 if (!sset_contains(&new_ifaces
, iface
->name
)) {
3147 iface_destroy(iface
);
3151 sset_destroy(&new_ifaces
);
3155 port_destroy(struct port
*port
)
3158 struct bridge
*br
= port
->bridge
;
3159 struct iface
*iface
, *next
;
3162 ofproto_bundle_unregister(br
->ofproto
, port
);
3165 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
3166 iface_destroy__(iface
);
3169 hmap_remove(&br
->ports
, &port
->hmap_node
);
3175 static struct port
*
3176 port_lookup(const struct bridge
*br
, const char *name
)
3180 HMAP_FOR_EACH_WITH_HASH (port
, hmap_node
, hash_string(name
, 0),
3182 if (!strcmp(port
->name
, name
)) {
3190 enable_lacp(struct port
*port
, bool *activep
)
3192 if (!port
->cfg
->lacp
) {
3193 /* XXX when LACP implementation has been sufficiently tested, enable by
3194 * default and make active on bonded ports. */
3196 } else if (!strcmp(port
->cfg
->lacp
, "off")) {
3198 } else if (!strcmp(port
->cfg
->lacp
, "active")) {
3201 } else if (!strcmp(port
->cfg
->lacp
, "passive")) {
3205 VLOG_WARN("port %s: unknown LACP mode %s",
3206 port
->name
, port
->cfg
->lacp
);
3211 static struct lacp_settings
*
3212 port_configure_lacp(struct port
*port
, struct lacp_settings
*s
)
3214 const char *lacp_time
, *system_id
;
3217 if (!enable_lacp(port
, &s
->active
)) {
3221 s
->name
= port
->name
;
3223 system_id
= smap_get(&port
->cfg
->other_config
, "lacp-system-id");
3225 if (!ovs_scan(system_id
, ETH_ADDR_SCAN_FMT
,
3226 ETH_ADDR_SCAN_ARGS(s
->id
))) {
3227 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3228 " address.", port
->name
, system_id
);
3232 memcpy(s
->id
, port
->bridge
->ea
, ETH_ADDR_LEN
);
3235 if (eth_addr_is_zero(s
->id
)) {
3236 VLOG_WARN("port %s: Invalid zero LACP system ID.", port
->name
);
3240 /* Prefer bondable links if unspecified. */
3241 priority
= smap_get_int(&port
->cfg
->other_config
, "lacp-system-priority",
3243 s
->priority
= (priority
> 0 && priority
<= UINT16_MAX
3245 : UINT16_MAX
- !list_is_short(&port
->ifaces
));
3247 lacp_time
= smap_get(&port
->cfg
->other_config
, "lacp-time");
3248 s
->fast
= lacp_time
&& !strcasecmp(lacp_time
, "fast");
3250 s
->fallback_ab_cfg
= smap_get_bool(&port
->cfg
->other_config
,
3251 "lacp-fallback-ab", false);
3257 iface_configure_lacp(struct iface
*iface
, struct lacp_slave_settings
*s
)
3259 int priority
, portid
, key
;
3261 portid
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-id", 0);
3262 priority
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-priority",
3264 key
= smap_get_int(&iface
->cfg
->other_config
, "lacp-aggregation-key", 0);
3266 if (portid
<= 0 || portid
> UINT16_MAX
) {
3267 portid
= ofp_to_u16(iface
->ofp_port
);
3270 if (priority
<= 0 || priority
> UINT16_MAX
) {
3271 priority
= UINT16_MAX
;
3274 if (key
< 0 || key
> UINT16_MAX
) {
3278 s
->name
= iface
->name
;
3280 s
->priority
= priority
;
3285 port_configure_bond(struct port
*port
, struct bond_settings
*s
)
3287 const char *detect_s
;
3288 struct iface
*iface
;
3289 int miimon_interval
;
3291 s
->name
= port
->name
;
3293 if (port
->cfg
->bond_mode
) {
3294 if (!bond_mode_from_string(&s
->balance
, port
->cfg
->bond_mode
)) {
3295 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3296 port
->name
, port
->cfg
->bond_mode
,
3297 bond_mode_to_string(s
->balance
));
3300 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3302 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3303 * active-backup. At some point we should remove this warning. */
3304 VLOG_WARN_RL(&rl
, "port %s: Using the default bond_mode %s. Note that"
3305 " in previous versions, the default bond_mode was"
3306 " balance-slb", port
->name
,
3307 bond_mode_to_string(s
->balance
));
3309 if (s
->balance
== BM_SLB
&& port
->bridge
->cfg
->n_flood_vlans
) {
3310 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3311 "please use another bond type or disable flood_vlans",
3315 miimon_interval
= smap_get_int(&port
->cfg
->other_config
,
3316 "bond-miimon-interval", 0);
3317 if (miimon_interval
<= 0) {
3318 miimon_interval
= 200;
3321 detect_s
= smap_get(&port
->cfg
->other_config
, "bond-detect-mode");
3322 if (!detect_s
|| !strcmp(detect_s
, "carrier")) {
3323 miimon_interval
= 0;
3324 } else if (strcmp(detect_s
, "miimon")) {
3325 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3326 "defaulting to carrier", port
->name
, detect_s
);
3327 miimon_interval
= 0;
3330 s
->up_delay
= MAX(0, port
->cfg
->bond_updelay
);
3331 s
->down_delay
= MAX(0, port
->cfg
->bond_downdelay
);
3332 s
->basis
= smap_get_int(&port
->cfg
->other_config
, "bond-hash-basis", 0);
3333 s
->rebalance_interval
= smap_get_int(&port
->cfg
->other_config
,
3334 "bond-rebalance-interval", 10000);
3335 if (s
->rebalance_interval
&& s
->rebalance_interval
< 1000) {
3336 s
->rebalance_interval
= 1000;
3339 s
->fake_iface
= port
->cfg
->bond_fake_iface
;
3341 s
->lacp_fallback_ab_cfg
= smap_get_bool(&port
->cfg
->other_config
,
3342 "lacp-fallback-ab", false);
3344 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
3345 netdev_set_miimon_interval(iface
->netdev
, miimon_interval
);
3349 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3350 * instead of obtaining it from the database. */
3352 port_is_synthetic(const struct port
*port
)
3354 return ovsdb_idl_row_is_synthetic(&port
->cfg
->header_
);
3357 /* Interface functions. */
3360 iface_is_internal(const struct ovsrec_interface
*iface
,
3361 const struct ovsrec_bridge
*br
)
3363 /* The local port and "internal" ports are always "internal". */
3364 return !strcmp(iface
->type
, "internal") || !strcmp(iface
->name
, br
->name
);
3367 /* Returns the correct network device type for interface 'iface' in bridge
3370 iface_get_type(const struct ovsrec_interface
*iface
,
3371 const struct ovsrec_bridge
*br
)
3375 /* The local port always has type "internal". Other ports take
3376 * their type from the database and default to "system" if none is
3378 if (iface_is_internal(iface
, br
)) {
3381 type
= iface
->type
[0] ? iface
->type
: "system";
3384 return ofproto_port_open_type(br
->datapath_type
, type
);
3388 iface_destroy__(struct iface
*iface
)
3391 struct port
*port
= iface
->port
;
3392 struct bridge
*br
= port
->bridge
;
3394 if (br
->ofproto
&& iface
->ofp_port
!= OFPP_NONE
) {
3395 ofproto_port_unregister(br
->ofproto
, iface
->ofp_port
);
3398 if (iface
->ofp_port
!= OFPP_NONE
) {
3399 hmap_remove(&br
->ifaces
, &iface
->ofp_port_node
);
3402 list_remove(&iface
->port_elem
);
3403 hmap_remove(&br
->iface_by_name
, &iface
->name_node
);
3405 netdev_close(iface
->netdev
);
3413 iface_destroy(struct iface
*iface
)
3416 struct port
*port
= iface
->port
;
3418 iface_destroy__(iface
);
3419 if (list_is_empty(&port
->ifaces
)) {
3425 static struct iface
*
3426 iface_lookup(const struct bridge
*br
, const char *name
)
3428 struct iface
*iface
;
3430 HMAP_FOR_EACH_WITH_HASH (iface
, name_node
, hash_string(name
, 0),
3431 &br
->iface_by_name
) {
3432 if (!strcmp(iface
->name
, name
)) {
3440 static struct iface
*
3441 iface_find(const char *name
)
3443 const struct bridge
*br
;
3445 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3446 struct iface
*iface
= iface_lookup(br
, name
);
3455 static struct iface
*
3456 iface_from_ofp_port(const struct bridge
*br
, ofp_port_t ofp_port
)
3458 struct iface
*iface
;
3460 HMAP_FOR_EACH_IN_BUCKET (iface
, ofp_port_node
, hash_ofp_port(ofp_port
),
3462 if (iface
->ofp_port
== ofp_port
) {
3469 /* Set Ethernet address of 'iface', if one is specified in the configuration
3472 iface_set_mac(struct iface
*iface
, const uint8_t *mac
)
3474 uint8_t ea
[ETH_ADDR_LEN
];
3476 if (strcmp(iface
->type
, "internal")) {
3480 if (iface
->cfg
->mac
&& eth_addr_from_string(iface
->cfg
->mac
, ea
)) {
3485 if (iface
->ofp_port
== OFPP_LOCAL
) {
3486 VLOG_ERR("interface %s: ignoring mac in Interface record "
3487 "(use Bridge record to set local port's mac)",
3489 } else if (eth_addr_is_multicast(mac
)) {
3490 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3493 int error
= netdev_set_etheraddr(iface
->netdev
, mac
);
3495 VLOG_ERR("interface %s: setting MAC failed (%s)",
3496 iface
->name
, ovs_strerror(error
));
3502 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3504 iface_set_ofport(const struct ovsrec_interface
*if_cfg
, ofp_port_t ofport
)
3506 if (if_cfg
&& !ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
3507 int64_t port
= ofport
== OFPP_NONE
? -1 : ofp_to_u16(ofport
);
3508 ovsrec_interface_set_ofport(if_cfg
, &port
, 1);
3512 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3513 * sets the "ofport" field to -1.
3515 * This is appropriate when 'if_cfg''s interface cannot be created or is
3516 * otherwise invalid. */
3518 iface_clear_db_record(const struct ovsrec_interface
*if_cfg
)
3520 if (!ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
3521 iface_set_ofport(if_cfg
, OFPP_NONE
);
3522 ovsrec_interface_set_status(if_cfg
, NULL
);
3523 ovsrec_interface_set_admin_state(if_cfg
, NULL
);
3524 ovsrec_interface_set_duplex(if_cfg
, NULL
);
3525 ovsrec_interface_set_link_speed(if_cfg
, NULL
, 0);
3526 ovsrec_interface_set_link_state(if_cfg
, NULL
);
3527 ovsrec_interface_set_mac_in_use(if_cfg
, NULL
);
3528 ovsrec_interface_set_mtu(if_cfg
, NULL
, 0);
3529 ovsrec_interface_set_cfm_fault(if_cfg
, NULL
, 0);
3530 ovsrec_interface_set_cfm_fault_status(if_cfg
, NULL
, 0);
3531 ovsrec_interface_set_cfm_remote_mpids(if_cfg
, NULL
, 0);
3532 ovsrec_interface_set_lacp_current(if_cfg
, NULL
, 0);
3533 ovsrec_interface_set_statistics(if_cfg
, NULL
, NULL
, 0);
3534 ovsrec_interface_set_ifindex(if_cfg
, NULL
, 0);
3539 queue_ids_include(const struct ovsdb_datum
*queues
, int64_t target
)
3541 union ovsdb_atom atom
;
3543 atom
.integer
= target
;
3544 return ovsdb_datum_find_key(queues
, &atom
, OVSDB_TYPE_INTEGER
) != UINT_MAX
;
3548 iface_configure_qos(struct iface
*iface
, const struct ovsrec_qos
*qos
)
3550 struct ofpbuf queues_buf
;
3552 ofpbuf_init(&queues_buf
, 0);
3554 if (!qos
|| qos
->type
[0] == '\0' || qos
->n_queues
< 1) {
3555 netdev_set_qos(iface
->netdev
, NULL
, NULL
);
3557 const struct ovsdb_datum
*queues
;
3558 struct netdev_queue_dump dump
;
3559 unsigned int queue_id
;
3560 struct smap details
;
3564 /* Configure top-level Qos for 'iface'. */
3565 netdev_set_qos(iface
->netdev
, qos
->type
, &qos
->other_config
);
3567 /* Deconfigure queues that were deleted. */
3568 queues
= ovsrec_qos_get_queues(qos
, OVSDB_TYPE_INTEGER
,
3570 smap_init(&details
);
3571 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
3572 if (!queue_ids_include(queues
, queue_id
)) {
3573 netdev_delete_queue(iface
->netdev
, queue_id
);
3576 smap_destroy(&details
);
3578 /* Configure queues for 'iface'. */
3580 for (i
= 0; i
< qos
->n_queues
; i
++) {
3581 const struct ovsrec_queue
*queue
= qos
->value_queues
[i
];
3582 unsigned int queue_id
= qos
->key_queues
[i
];
3584 if (queue_id
== 0) {
3588 if (queue
->n_dscp
== 1) {
3589 struct ofproto_port_queue
*port_queue
;
3591 port_queue
= ofpbuf_put_uninit(&queues_buf
,
3592 sizeof *port_queue
);
3593 port_queue
->queue
= queue_id
;
3594 port_queue
->dscp
= queue
->dscp
[0];
3597 netdev_set_queue(iface
->netdev
, queue_id
, &queue
->other_config
);
3600 struct smap details
;
3602 smap_init(&details
);
3603 netdev_set_queue(iface
->netdev
, 0, &details
);
3604 smap_destroy(&details
);
3608 if (iface
->ofp_port
!= OFPP_NONE
) {
3609 const struct ofproto_port_queue
*port_queues
= queues_buf
.data
;
3610 size_t n_queues
= queues_buf
.size
/ sizeof *port_queues
;
3612 ofproto_port_set_queues(iface
->port
->bridge
->ofproto
, iface
->ofp_port
,
3613 port_queues
, n_queues
);
3616 netdev_set_policing(iface
->netdev
,
3617 iface
->cfg
->ingress_policing_rate
,
3618 iface
->cfg
->ingress_policing_burst
);
3620 ofpbuf_uninit(&queues_buf
);
3624 iface_configure_cfm(struct iface
*iface
)
3626 const struct ovsrec_interface
*cfg
= iface
->cfg
;
3627 const char *opstate_str
;
3628 const char *cfm_ccm_vlan
;
3629 struct cfm_settings s
;
3630 struct smap netdev_args
;
3632 if (!cfg
->n_cfm_mpid
) {
3633 ofproto_port_clear_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
);
3637 s
.check_tnl_key
= false;
3638 smap_init(&netdev_args
);
3639 if (!netdev_get_config(iface
->netdev
, &netdev_args
)) {
3640 const char *key
= smap_get(&netdev_args
, "key");
3641 const char *in_key
= smap_get(&netdev_args
, "in_key");
3643 s
.check_tnl_key
= (key
&& !strcmp(key
, "flow"))
3644 || (in_key
&& !strcmp(in_key
, "flow"));
3646 smap_destroy(&netdev_args
);
3648 s
.mpid
= *cfg
->cfm_mpid
;
3649 s
.interval
= smap_get_int(&iface
->cfg
->other_config
, "cfm_interval", 0);
3650 cfm_ccm_vlan
= smap_get(&iface
->cfg
->other_config
, "cfm_ccm_vlan");
3651 s
.ccm_pcp
= smap_get_int(&iface
->cfg
->other_config
, "cfm_ccm_pcp", 0);
3653 if (s
.interval
<= 0) {
3657 if (!cfm_ccm_vlan
) {
3659 } else if (!strcasecmp("random", cfm_ccm_vlan
)) {
3660 s
.ccm_vlan
= CFM_RANDOM_VLAN
;
3662 s
.ccm_vlan
= atoi(cfm_ccm_vlan
);
3663 if (s
.ccm_vlan
== CFM_RANDOM_VLAN
) {
3668 s
.extended
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_extended",
3670 s
.demand
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_demand", false);
3672 opstate_str
= smap_get(&iface
->cfg
->other_config
, "cfm_opstate");
3673 s
.opup
= !opstate_str
|| !strcasecmp("up", opstate_str
);
3675 ofproto_port_set_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
, &s
);
3678 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3679 * instead of obtaining it from the database. */
3681 iface_is_synthetic(const struct iface
*iface
)
3683 return ovsdb_idl_row_is_synthetic(&iface
->cfg
->header_
);
3687 iface_validate_ofport__(size_t n
, int64_t *ofport
)
3689 return (n
&& *ofport
>= 1 && *ofport
< ofp_to_u16(OFPP_MAX
)
3690 ? u16_to_ofp(*ofport
)
3695 iface_get_requested_ofp_port(const struct ovsrec_interface
*cfg
)
3697 return iface_validate_ofport__(cfg
->n_ofport_request
, cfg
->ofport_request
);
3701 iface_pick_ofport(const struct ovsrec_interface
*cfg
)
3703 ofp_port_t requested_ofport
= iface_get_requested_ofp_port(cfg
);
3704 return (requested_ofport
!= OFPP_NONE
3706 : iface_validate_ofport__(cfg
->n_ofport
, cfg
->ofport
));
3709 /* Port mirroring. */
3711 static struct mirror
*
3712 mirror_find_by_uuid(struct bridge
*br
, const struct uuid
*uuid
)
3716 HMAP_FOR_EACH_IN_BUCKET (m
, hmap_node
, uuid_hash(uuid
), &br
->mirrors
) {
3717 if (uuid_equals(uuid
, &m
->uuid
)) {
3725 bridge_configure_mirrors(struct bridge
*br
)
3727 const struct ovsdb_datum
*mc
;
3728 unsigned long *flood_vlans
;
3729 struct mirror
*m
, *next
;
3732 /* Get rid of deleted mirrors. */
3733 mc
= ovsrec_bridge_get_mirrors(br
->cfg
, OVSDB_TYPE_UUID
);
3734 HMAP_FOR_EACH_SAFE (m
, next
, hmap_node
, &br
->mirrors
) {
3735 union ovsdb_atom atom
;
3737 atom
.uuid
= m
->uuid
;
3738 if (ovsdb_datum_find_key(mc
, &atom
, OVSDB_TYPE_UUID
) == UINT_MAX
) {
3743 /* Add new mirrors and reconfigure existing ones. */
3744 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
3745 const struct ovsrec_mirror
*cfg
= br
->cfg
->mirrors
[i
];
3746 struct mirror
*m
= mirror_find_by_uuid(br
, &cfg
->header_
.uuid
);
3748 m
= mirror_create(br
, cfg
);
3751 if (!mirror_configure(m
)) {
3756 /* Update flooded vlans (for RSPAN). */
3757 flood_vlans
= vlan_bitmap_from_array(br
->cfg
->flood_vlans
,
3758 br
->cfg
->n_flood_vlans
);
3759 ofproto_set_flood_vlans(br
->ofproto
, flood_vlans
);
3760 bitmap_free(flood_vlans
);
3763 static struct mirror
*
3764 mirror_create(struct bridge
*br
, const struct ovsrec_mirror
*cfg
)
3768 m
= xzalloc(sizeof *m
);
3769 m
->uuid
= cfg
->header_
.uuid
;
3770 hmap_insert(&br
->mirrors
, &m
->hmap_node
, uuid_hash(&m
->uuid
));
3772 m
->name
= xstrdup(cfg
->name
);
3778 mirror_destroy(struct mirror
*m
)
3781 struct bridge
*br
= m
->bridge
;
3784 ofproto_mirror_unregister(br
->ofproto
, m
);
3787 hmap_remove(&br
->mirrors
, &m
->hmap_node
);
3794 mirror_collect_ports(struct mirror
*m
,
3795 struct ovsrec_port
**in_ports
, int n_in_ports
,
3796 void ***out_portsp
, size_t *n_out_portsp
)
3798 void **out_ports
= xmalloc(n_in_ports
* sizeof *out_ports
);
3799 size_t n_out_ports
= 0;
3802 for (i
= 0; i
< n_in_ports
; i
++) {
3803 const char *name
= in_ports
[i
]->name
;
3804 struct port
*port
= port_lookup(m
->bridge
, name
);
3806 out_ports
[n_out_ports
++] = port
;
3808 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3809 "port %s", m
->bridge
->name
, m
->name
, name
);
3812 *out_portsp
= out_ports
;
3813 *n_out_portsp
= n_out_ports
;
3817 mirror_configure(struct mirror
*m
)
3819 const struct ovsrec_mirror
*cfg
= m
->cfg
;
3820 struct ofproto_mirror_settings s
;
3823 if (strcmp(cfg
->name
, m
->name
)) {
3825 m
->name
= xstrdup(cfg
->name
);
3829 /* Get output port or VLAN. */
3830 if (cfg
->output_port
) {
3831 s
.out_bundle
= port_lookup(m
->bridge
, cfg
->output_port
->name
);
3832 if (!s
.out_bundle
) {
3833 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3834 m
->bridge
->name
, m
->name
);
3837 s
.out_vlan
= UINT16_MAX
;
3839 if (cfg
->output_vlan
) {
3840 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3841 "output vlan; ignoring output vlan",
3842 m
->bridge
->name
, m
->name
);
3844 } else if (cfg
->output_vlan
) {
3845 /* The database should prevent invalid VLAN values. */
3846 s
.out_bundle
= NULL
;
3847 s
.out_vlan
= *cfg
->output_vlan
;
3849 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3850 m
->bridge
->name
, m
->name
);
3854 /* Get port selection. */
3855 if (cfg
->select_all
) {
3856 size_t n_ports
= hmap_count(&m
->bridge
->ports
);
3857 void **ports
= xmalloc(n_ports
* sizeof *ports
);
3862 HMAP_FOR_EACH (port
, hmap_node
, &m
->bridge
->ports
) {
3872 /* Get ports, dropping ports that don't exist.
3873 * The IDL ensures that there are no duplicates. */
3874 mirror_collect_ports(m
, cfg
->select_src_port
, cfg
->n_select_src_port
,
3875 &s
.srcs
, &s
.n_srcs
);
3876 mirror_collect_ports(m
, cfg
->select_dst_port
, cfg
->n_select_dst_port
,
3877 &s
.dsts
, &s
.n_dsts
);
3880 /* Get VLAN selection. */
3881 s
.src_vlans
= vlan_bitmap_from_array(cfg
->select_vlan
, cfg
->n_select_vlan
);
3884 ofproto_mirror_register(m
->bridge
->ofproto
, m
, &s
);
3887 if (s
.srcs
!= s
.dsts
) {
3896 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3898 * This is deprecated. It is only for compatibility with broken device drivers
3899 * in old versions of Linux that do not properly support VLANs when VLAN
3900 * devices are not used. When broken device drivers are no longer in
3901 * widespread use, we will delete these interfaces. */
3903 static struct ovsrec_port
**recs
;
3904 static size_t n_recs
, allocated_recs
;
3906 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3907 * splinters are reconfigured. */
3909 register_rec(struct ovsrec_port
*rec
)
3911 if (n_recs
>= allocated_recs
) {
3912 recs
= x2nrealloc(recs
, &allocated_recs
, sizeof *recs
);
3914 recs
[n_recs
++] = rec
;
3917 /* Frees all of the ports registered with register_reg(). */
3919 free_registered_recs(void)
3923 for (i
= 0; i
< n_recs
; i
++) {
3924 struct ovsrec_port
*port
= recs
[i
];
3927 for (j
= 0; j
< port
->n_interfaces
; j
++) {
3928 struct ovsrec_interface
*iface
= port
->interfaces
[j
];
3933 smap_destroy(&port
->other_config
);
3934 free(port
->interfaces
);
3942 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3945 vlan_splinters_is_enabled(const struct ovsrec_interface
*iface_cfg
)
3947 return smap_get_bool(&iface_cfg
->other_config
, "enable-vlan-splinters",
3951 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3954 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3955 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3956 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3959 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3960 * use, returns NULL.
3962 * Updates 'vlan_splinters_enabled_anywhere'. */
3963 static unsigned long int *
3964 collect_splinter_vlans(const struct ovsrec_open_vswitch
*ovs_cfg
)
3966 unsigned long int *splinter_vlans
;
3967 struct sset splinter_ifaces
;
3968 const char *real_dev_name
;
3969 struct shash
*real_devs
;
3970 struct shash_node
*node
;
3974 /* Free space allocated for synthesized ports and interfaces, since we're
3975 * in the process of reconstructing all of them. */
3976 free_registered_recs();
3978 splinter_vlans
= bitmap_allocate(4096);
3979 sset_init(&splinter_ifaces
);
3980 vlan_splinters_enabled_anywhere
= false;
3981 for (i
= 0; i
< ovs_cfg
->n_bridges
; i
++) {
3982 struct ovsrec_bridge
*br_cfg
= ovs_cfg
->bridges
[i
];
3985 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
3986 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
3989 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
3990 struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[k
];
3992 if (vlan_splinters_is_enabled(iface_cfg
)) {
3993 vlan_splinters_enabled_anywhere
= true;
3994 sset_add(&splinter_ifaces
, iface_cfg
->name
);
3995 vlan_bitmap_from_array__(port_cfg
->trunks
,
4001 if (port_cfg
->tag
&& *port_cfg
->tag
> 0 && *port_cfg
->tag
< 4095) {
4002 bitmap_set1(splinter_vlans
, *port_cfg
->tag
);
4007 if (!vlan_splinters_enabled_anywhere
) {
4008 free(splinter_vlans
);
4009 sset_destroy(&splinter_ifaces
);
4013 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
4015 ofproto_get_vlan_usage(br
->ofproto
, splinter_vlans
);
4019 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4020 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4021 * device to be created for it. */
4022 bitmap_set0(splinter_vlans
, 0);
4023 bitmap_set0(splinter_vlans
, 4095);
4025 /* Delete all VLAN devices that we don't need. */
4027 real_devs
= vlandev_get_real_devs();
4028 SHASH_FOR_EACH (node
, real_devs
) {
4029 const struct vlan_real_dev
*real_dev
= node
->data
;
4030 const struct vlan_dev
*vlan_dev
;
4031 bool real_dev_has_splinters
;
4033 real_dev_has_splinters
= sset_contains(&splinter_ifaces
,
4035 HMAP_FOR_EACH (vlan_dev
, hmap_node
, &real_dev
->vlan_devs
) {
4036 if (!real_dev_has_splinters
4037 || !bitmap_is_set(splinter_vlans
, vlan_dev
->vid
)) {
4038 struct netdev
*netdev
;
4040 if (!netdev_open(vlan_dev
->name
, "system", &netdev
)) {
4041 if (!netdev_get_in4(netdev
, NULL
, NULL
) ||
4042 !netdev_get_in6(netdev
, NULL
)) {
4043 /* It has an IP address configured, so we don't own
4044 * it. Don't delete it. */
4046 vlandev_del(vlan_dev
->name
);
4048 netdev_close(netdev
);
4055 /* Add all VLAN devices that we need. */
4056 SSET_FOR_EACH (real_dev_name
, &splinter_ifaces
) {
4059 BITMAP_FOR_EACH_1 (vid
, 4096, splinter_vlans
) {
4060 if (!vlandev_get_name(real_dev_name
, vid
)) {
4061 vlandev_add(real_dev_name
, vid
);
4068 sset_destroy(&splinter_ifaces
);
4070 if (bitmap_scan(splinter_vlans
, 0, 4096) >= 4096) {
4071 free(splinter_vlans
);
4074 return splinter_vlans
;
4077 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4080 configure_splinter_port(struct port
*port
)
4082 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
4083 ofp_port_t realdev_ofp_port
;
4084 const char *realdev_name
;
4085 struct iface
*vlandev
, *realdev
;
4087 ofproto_bundle_unregister(port
->bridge
->ofproto
, port
);
4089 vlandev
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
,
4092 realdev_name
= smap_get(&port
->cfg
->other_config
, "realdev");
4093 realdev
= iface_lookup(port
->bridge
, realdev_name
);
4094 realdev_ofp_port
= realdev
? realdev
->ofp_port
: 0;
4096 ofproto_port_set_realdev(ofproto
, vlandev
->ofp_port
, realdev_ofp_port
,
4100 static struct ovsrec_port
*
4101 synthesize_splinter_port(const char *real_dev_name
,
4102 const char *vlan_dev_name
, int vid
)
4104 struct ovsrec_interface
*iface
;
4105 struct ovsrec_port
*port
;
4107 iface
= xmalloc(sizeof *iface
);
4108 ovsrec_interface_init(iface
);
4109 iface
->name
= xstrdup(vlan_dev_name
);
4110 iface
->type
= "system";
4112 port
= xmalloc(sizeof *port
);
4113 ovsrec_port_init(port
);
4114 port
->interfaces
= xmemdup(&iface
, sizeof iface
);
4115 port
->n_interfaces
= 1;
4116 port
->name
= xstrdup(vlan_dev_name
);
4117 port
->vlan_mode
= "splinter";
4118 port
->tag
= xmalloc(sizeof *port
->tag
);
4121 smap_add(&port
->other_config
, "realdev", real_dev_name
);
4127 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4128 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4129 * 1-bit in the 'splinter_vlans' bitmap. */
4131 add_vlan_splinter_ports(struct bridge
*br
,
4132 const unsigned long int *splinter_vlans
,
4133 struct shash
*ports
)
4137 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4138 * we're modifying 'ports'. */
4139 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
4140 const char *name
= br
->cfg
->ports
[i
]->name
;
4141 struct ovsrec_port
*port_cfg
= shash_find_data(ports
, name
);
4144 for (j
= 0; j
< port_cfg
->n_interfaces
; j
++) {
4145 struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[j
];
4147 if (vlan_splinters_is_enabled(iface_cfg
)) {
4148 const char *real_dev_name
;
4151 real_dev_name
= iface_cfg
->name
;
4152 BITMAP_FOR_EACH_1 (vid
, 4096, splinter_vlans
) {
4153 const char *vlan_dev_name
;
4155 vlan_dev_name
= vlandev_get_name(real_dev_name
, vid
);
4157 && !shash_find(ports
, vlan_dev_name
)) {
4158 shash_add(ports
, vlan_dev_name
,
4159 synthesize_splinter_port(
4160 real_dev_name
, vlan_dev_name
, vid
));
4169 mirror_refresh_stats(struct mirror
*m
)
4171 struct ofproto
*ofproto
= m
->bridge
->ofproto
;
4172 uint64_t tx_packets
, tx_bytes
;
4175 size_t stat_cnt
= 0;
4177 if (ofproto_mirror_get_stats(ofproto
, m
, &tx_packets
, &tx_bytes
)) {
4178 ovsrec_mirror_set_statistics(m
->cfg
, NULL
, NULL
, 0);
4182 if (tx_packets
!= UINT64_MAX
) {
4183 keys
[stat_cnt
] = "tx_packets";
4184 values
[stat_cnt
] = tx_packets
;
4187 if (tx_bytes
!= UINT64_MAX
) {
4188 keys
[stat_cnt
] = "tx_bytes";
4189 values
[stat_cnt
] = tx_bytes
;
4193 ovsrec_mirror_set_statistics(m
->cfg
, keys
, values
, stat_cnt
);