1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/ofproto.h"
43 #include "poll-loop.h"
47 #include "socket-util.h"
49 #include "stream-ssl.h"
51 #include "system-stats.h"
56 #include "lib/vswitch-idl.h"
57 #include "xenserver.h"
59 #include "sflow_api.h"
60 #include "vlan-bitmap.h"
62 VLOG_DEFINE_THIS_MODULE(bridge
);
64 COVERAGE_DEFINE(bridge_reconfigure
);
66 /* Configuration of an uninstantiated iface. */
68 struct hmap_node hmap_node
; /* Node in bridge's if_cfg_todo. */
69 const struct ovsrec_interface
*cfg
; /* Interface record. */
70 const struct ovsrec_port
*parent
; /* Parent port record. */
71 ofp_port_t ofport
; /* Requested OpenFlow port number. */
74 /* OpenFlow port slated for removal from ofproto. */
76 struct list list_node
; /* Node in bridge's ofpp_garbage. */
77 ofp_port_t ofp_port
; /* Port to be deleted. */
81 /* These members are always valid. */
82 struct list port_elem
; /* Element in struct port's "ifaces" list. */
83 struct hmap_node name_node
; /* In struct bridge's "iface_by_name" hmap. */
84 struct port
*port
; /* Containing port. */
85 char *name
; /* Host network device name. */
87 /* These members are valid only after bridge_reconfigure() causes them to
89 struct hmap_node ofp_port_node
; /* In struct bridge's "ifaces" hmap. */
90 ofp_port_t ofp_port
; /* OpenFlow port number, */
91 /* OFPP_NONE if unknown. */
92 struct netdev
*netdev
; /* Network device. */
93 const char *type
; /* Usually same as cfg->type. */
94 const struct ovsrec_interface
*cfg
;
98 struct uuid uuid
; /* UUID of this "mirror" record in database. */
99 struct hmap_node hmap_node
; /* In struct bridge's "mirrors" hmap. */
100 struct bridge
*bridge
;
102 const struct ovsrec_mirror
*cfg
;
106 struct hmap_node hmap_node
; /* Element in struct bridge's "ports" hmap. */
107 struct bridge
*bridge
;
110 const struct ovsrec_port
*cfg
;
112 /* An ordinary bridge port has 1 interface.
113 * A bridge port for bonding has at least 2 interfaces. */
114 struct list ifaces
; /* List of "struct iface"s. */
118 struct hmap_node node
; /* In 'all_bridges'. */
119 char *name
; /* User-specified arbitrary name. */
120 char *type
; /* Datapath type. */
121 uint8_t ea
[ETH_ADDR_LEN
]; /* Bridge Ethernet Address. */
122 uint8_t default_ea
[ETH_ADDR_LEN
]; /* Default MAC. */
123 const struct ovsrec_bridge
*cfg
;
125 /* OpenFlow switch processing. */
126 struct ofproto
*ofproto
; /* OpenFlow switch. */
129 struct hmap ports
; /* "struct port"s indexed by name. */
130 struct hmap ifaces
; /* "struct iface"s indexed by ofp_port. */
131 struct hmap iface_by_name
; /* "struct iface"s indexed by name. */
133 struct list ofpp_garbage
; /* "struct ofpp_garbage" slated for removal. */
134 struct hmap if_cfg_todo
; /* "struct if_cfg"s slated for creation.
135 Indexed on 'cfg->name'. */
137 /* Port mirroring. */
138 struct hmap mirrors
; /* "struct mirror" indexed by UUID. */
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port
;
142 struct ovsrec_interface synth_local_iface
;
143 struct ovsrec_interface
*synth_local_ifacep
;
146 /* All bridges, indexed by name. */
147 static struct hmap all_bridges
= HMAP_INITIALIZER(&all_bridges
);
149 /* OVSDB IDL used to obtain configuration. */
150 static struct ovsdb_idl
*idl
;
152 /* We want to complete daemonization, fully detaching from our parent process,
153 * only after we have completed our initial configuration, committed our state
154 * to the database, and received confirmation back from the database server
155 * that it applied the commit. This allows our parent process to know that,
156 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
157 * to have already been filled in. (It is only "very likely" rather than
158 * certain because there is always a slim possibility that the transaction will
159 * fail or that some other client has added new bridges, ports, etc. while
160 * ovs-vswitchd was configuring using an old configuration.)
162 * We only need to do this once for our initial configuration at startup, so
163 * 'initial_config_done' tracks whether we've already done it. While we are
164 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
165 * itself and is otherwise NULL. */
166 static bool initial_config_done
;
167 static struct ovsdb_idl_txn
*daemonize_txn
;
169 /* Most recently processed IDL sequence number. */
170 static unsigned int idl_seqno
;
172 /* Each time this timer expires, the bridge fetches interface and mirror
173 * statistics and pushes them into the database. */
174 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
175 static long long int iface_stats_timer
= LLONG_MIN
;
177 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
178 * expensive. This can cause bridge_reconfigure() to take a long time during
179 * which no other work can be done. To deal with this problem, we limit port
180 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
181 * call to bridge_reconfigure(). If there is more work to do after the limit
182 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
183 * This allows the rest of the code to catch up on important things like
184 * forwarding packets. */
185 #define OFP_PORT_ACTION_WINDOW 10
186 static bool reconfiguring
= false;
188 static void add_del_bridges(const struct ovsrec_open_vswitch
*);
189 static void bridge_update_ofprotos(void);
190 static void bridge_create(const struct ovsrec_bridge
*);
191 static void bridge_destroy(struct bridge
*);
192 static struct bridge
*bridge_lookup(const char *name
);
193 static unixctl_cb_func bridge_unixctl_dump_flows
;
194 static unixctl_cb_func bridge_unixctl_reconnect
;
195 static size_t bridge_get_controllers(const struct bridge
*br
,
196 struct ovsrec_controller
***controllersp
);
197 static void bridge_add_del_ports(struct bridge
*,
198 const unsigned long int *splinter_vlans
);
199 static void bridge_refresh_ofp_port(struct bridge
*);
200 static void bridge_configure_flow_miss_model(const char *opt
);
201 static void bridge_configure_datapath_id(struct bridge
*);
202 static void bridge_configure_netflow(struct bridge
*);
203 static void bridge_configure_forward_bpdu(struct bridge
*);
204 static void bridge_configure_mac_table(struct bridge
*);
205 static void bridge_configure_sflow(struct bridge
*, int *sflow_bridge_number
);
206 static void bridge_configure_ipfix(struct bridge
*);
207 static void bridge_configure_stp(struct bridge
*);
208 static void bridge_configure_tables(struct bridge
*);
209 static void bridge_configure_dp_desc(struct bridge
*);
210 static void bridge_configure_remotes(struct bridge
*,
211 const struct sockaddr_in
*managers
,
213 static void bridge_pick_local_hw_addr(struct bridge
*,
214 uint8_t ea
[ETH_ADDR_LEN
],
215 struct iface
**hw_addr_iface
);
216 static uint64_t bridge_pick_datapath_id(struct bridge
*,
217 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
218 struct iface
*hw_addr_iface
);
219 static void bridge_queue_if_cfg(struct bridge
*,
220 const struct ovsrec_interface
*,
221 const struct ovsrec_port
*);
222 static uint64_t dpid_from_hash(const void *, size_t nbytes
);
223 static bool bridge_has_bond_fake_iface(const struct bridge
*,
225 static bool port_is_bond_fake_iface(const struct port
*);
227 static unixctl_cb_func qos_unixctl_show
;
229 static struct port
*port_create(struct bridge
*, const struct ovsrec_port
*);
230 static void port_del_ifaces(struct port
*);
231 static void port_destroy(struct port
*);
232 static struct port
*port_lookup(const struct bridge
*, const char *name
);
233 static void port_configure(struct port
*);
234 static struct lacp_settings
*port_configure_lacp(struct port
*,
235 struct lacp_settings
*);
236 static void port_configure_bond(struct port
*, struct bond_settings
*);
237 static bool port_is_synthetic(const struct port
*);
239 static void reconfigure_system_stats(const struct ovsrec_open_vswitch
*);
240 static void run_system_stats(void);
242 static void bridge_configure_mirrors(struct bridge
*);
243 static struct mirror
*mirror_create(struct bridge
*,
244 const struct ovsrec_mirror
*);
245 static void mirror_destroy(struct mirror
*);
246 static bool mirror_configure(struct mirror
*);
247 static void mirror_refresh_stats(struct mirror
*);
249 static void iface_configure_lacp(struct iface
*, struct lacp_slave_settings
*);
250 static bool iface_create(struct bridge
*, struct if_cfg
*,
251 ofp_port_t ofp_port
);
252 static bool iface_is_internal(const struct ovsrec_interface
*iface
,
253 const struct ovsrec_bridge
*br
);
254 static const char *iface_get_type(const struct ovsrec_interface
*,
255 const struct ovsrec_bridge
*);
256 static void iface_destroy(struct iface
*);
257 static struct iface
*iface_lookup(const struct bridge
*, const char *name
);
258 static struct iface
*iface_find(const char *name
);
259 static struct if_cfg
*if_cfg_lookup(const struct bridge
*, const char *name
);
260 static struct iface
*iface_from_ofp_port(const struct bridge
*,
261 ofp_port_t ofp_port
);
262 static void iface_set_mac(struct iface
*);
263 static void iface_set_ofport(const struct ovsrec_interface
*, ofp_port_t ofport
);
264 static void iface_clear_db_record(const struct ovsrec_interface
*if_cfg
);
265 static void iface_configure_qos(struct iface
*, const struct ovsrec_qos
*);
266 static void iface_configure_cfm(struct iface
*);
267 static void iface_refresh_cfm_stats(struct iface
*);
268 static void iface_refresh_stats(struct iface
*);
269 static void iface_refresh_status(struct iface
*);
270 static bool iface_is_synthetic(const struct iface
*);
271 static ofp_port_t
iface_pick_ofport(const struct ovsrec_interface
*);
273 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
275 * This is deprecated. It is only for compatibility with broken device drivers
276 * in old versions of Linux that do not properly support VLANs when VLAN
277 * devices are not used. When broken device drivers are no longer in
278 * widespread use, we will delete these interfaces. */
280 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
281 static bool vlan_splinters_enabled_anywhere
;
283 static bool vlan_splinters_is_enabled(const struct ovsrec_interface
*);
284 static unsigned long int *collect_splinter_vlans(
285 const struct ovsrec_open_vswitch
*);
286 static void configure_splinter_port(struct port
*);
287 static void add_vlan_splinter_ports(struct bridge
*,
288 const unsigned long int *splinter_vlans
,
289 struct shash
*ports
);
292 bridge_init_ofproto(const struct ovsrec_open_vswitch
*cfg
)
294 struct shash iface_hints
;
295 static bool initialized
= false;
302 shash_init(&iface_hints
);
305 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
306 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
309 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
310 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
313 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
314 struct ovsrec_interface
*if_cfg
= port_cfg
->interfaces
[k
];
315 struct iface_hint
*iface_hint
;
317 iface_hint
= xmalloc(sizeof *iface_hint
);
318 iface_hint
->br_name
= br_cfg
->name
;
319 iface_hint
->br_type
= br_cfg
->datapath_type
;
320 iface_hint
->ofp_port
= iface_pick_ofport(if_cfg
);
322 shash_add(&iface_hints
, if_cfg
->name
, iface_hint
);
328 ofproto_init(&iface_hints
);
330 shash_destroy_free_data(&iface_hints
);
334 /* Public functions. */
336 /* Initializes the bridge module, configuring it to obtain its configuration
337 * from an OVSDB server accessed over 'remote', which should be a string in a
338 * form acceptable to ovsdb_idl_create(). */
340 bridge_init(const char *remote
)
342 /* Create connection to database. */
343 idl
= ovsdb_idl_create(remote
, &ovsrec_idl_class
, true, true);
344 idl_seqno
= ovsdb_idl_get_seqno(idl
);
345 ovsdb_idl_set_lock(idl
, "ovs_vswitchd");
346 ovsdb_idl_verify_write_only(idl
);
348 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_cur_cfg
);
349 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_statistics
);
350 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_external_ids
);
351 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_ovs_version
);
352 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_db_version
);
353 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_type
);
354 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_version
);
356 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_id
);
357 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_status
);
358 ovsdb_idl_omit(idl
, &ovsrec_bridge_col_external_ids
);
360 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_status
);
361 ovsdb_idl_omit_alert(idl
, &ovsrec_port_col_statistics
);
362 ovsdb_idl_omit(idl
, &ovsrec_port_col_external_ids
);
363 ovsdb_idl_omit(idl
, &ovsrec_port_col_fake_bridge
);
365 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_admin_state
);
366 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_duplex
);
367 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_speed
);
368 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_state
);
369 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_resets
);
370 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mac_in_use
);
371 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ifindex
);
372 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mtu
);
373 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ofport
);
374 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_statistics
);
375 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_status
);
376 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault
);
377 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault_status
);
378 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_mpids
);
379 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_health
);
380 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_opstate
);
381 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_bfd_status
);
382 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_lacp_current
);
383 ovsdb_idl_omit(idl
, &ovsrec_interface_col_external_ids
);
385 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_is_connected
);
386 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_role
);
387 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_status
);
388 ovsdb_idl_omit(idl
, &ovsrec_controller_col_external_ids
);
390 ovsdb_idl_omit(idl
, &ovsrec_qos_col_external_ids
);
392 ovsdb_idl_omit(idl
, &ovsrec_queue_col_external_ids
);
394 ovsdb_idl_omit(idl
, &ovsrec_mirror_col_external_ids
);
395 ovsdb_idl_omit_alert(idl
, &ovsrec_mirror_col_statistics
);
397 ovsdb_idl_omit(idl
, &ovsrec_netflow_col_external_ids
);
398 ovsdb_idl_omit(idl
, &ovsrec_sflow_col_external_ids
);
399 ovsdb_idl_omit(idl
, &ovsrec_ipfix_col_external_ids
);
400 ovsdb_idl_omit(idl
, &ovsrec_flow_sample_collector_set_col_external_ids
);
402 ovsdb_idl_omit(idl
, &ovsrec_manager_col_external_ids
);
403 ovsdb_idl_omit(idl
, &ovsrec_manager_col_inactivity_probe
);
404 ovsdb_idl_omit(idl
, &ovsrec_manager_col_is_connected
);
405 ovsdb_idl_omit(idl
, &ovsrec_manager_col_max_backoff
);
406 ovsdb_idl_omit(idl
, &ovsrec_manager_col_status
);
408 ovsdb_idl_omit(idl
, &ovsrec_ssl_col_external_ids
);
410 /* Register unixctl commands. */
411 unixctl_command_register("qos/show", "interface", 1, 1,
412 qos_unixctl_show
, NULL
);
413 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
414 bridge_unixctl_dump_flows
, NULL
);
415 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
416 bridge_unixctl_reconnect
, NULL
);
426 struct bridge
*br
, *next_br
;
428 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
431 ovsdb_idl_destroy(idl
);
434 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
435 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
436 * responsible for freeing '*managersp' (with free()).
438 * You may be asking yourself "why does ovs-vswitchd care?", because
439 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
440 * should not be and in fact is not directly involved in that. But
441 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
442 * it has to tell in-band control where the managers are to enable that.
443 * (Thus, only managers connected in-band are collected.)
446 collect_in_band_managers(const struct ovsrec_open_vswitch
*ovs_cfg
,
447 struct sockaddr_in
**managersp
, size_t *n_managersp
)
449 struct sockaddr_in
*managers
= NULL
;
450 size_t n_managers
= 0;
454 /* Collect all of the potential targets from the "targets" columns of the
455 * rows pointed to by "manager_options", excluding any that are
458 for (i
= 0; i
< ovs_cfg
->n_manager_options
; i
++) {
459 struct ovsrec_manager
*m
= ovs_cfg
->manager_options
[i
];
461 if (m
->connection_mode
&& !strcmp(m
->connection_mode
, "out-of-band")) {
462 sset_find_and_delete(&targets
, m
->target
);
464 sset_add(&targets
, m
->target
);
468 /* Now extract the targets' IP addresses. */
469 if (!sset_is_empty(&targets
)) {
472 managers
= xmalloc(sset_count(&targets
) * sizeof *managers
);
473 SSET_FOR_EACH (target
, &targets
) {
474 struct sockaddr_in
*sin
= &managers
[n_managers
];
476 if (stream_parse_target_with_default_port(target
,
483 sset_destroy(&targets
);
485 *managersp
= managers
;
486 *n_managersp
= n_managers
;
490 bridge_reconfigure(const struct ovsrec_open_vswitch
*ovs_cfg
)
492 unsigned long int *splinter_vlans
;
495 COVERAGE_INC(bridge_reconfigure
);
497 ovs_assert(!reconfiguring
);
498 reconfiguring
= true;
500 ofproto_set_flow_eviction_threshold(
501 smap_get_int(&ovs_cfg
->other_config
, "flow-eviction-threshold",
502 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT
));
504 ofproto_set_n_handler_threads(
505 smap_get_int(&ovs_cfg
->other_config
, "n-handler-threads", 0));
507 bridge_configure_flow_miss_model(smap_get(&ovs_cfg
->other_config
,
508 "force-miss-model"));
510 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
511 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
512 * configuration otherwise.
514 * This is mostly an update to bridge data structures. Nothing is pushed
515 * down to ofproto or lower layers. */
516 add_del_bridges(ovs_cfg
);
517 splinter_vlans
= collect_splinter_vlans(ovs_cfg
);
518 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
519 bridge_add_del_ports(br
, splinter_vlans
);
521 free(splinter_vlans
);
523 /* Delete datapaths that are no longer configured, and create ones which
524 * don't exist but should. */
525 bridge_update_ofprotos();
527 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
528 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
529 bridge_refresh_ofp_port(br
);
532 /* Clear database records for "if_cfg"s which haven't been instantiated. */
533 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
534 struct if_cfg
*if_cfg
;
536 HMAP_FOR_EACH (if_cfg
, hmap_node
, &br
->if_cfg_todo
) {
537 iface_clear_db_record(if_cfg
->cfg
);
541 reconfigure_system_stats(ovs_cfg
);
545 bridge_reconfigure_ofp(void)
547 long long int deadline
;
550 deadline
= time_msec() + OFP_PORT_ACTION_WINDOW
;
552 /* The kernel will reject any attempt to add a given port to a datapath if
553 * that port already belongs to a different datapath, so we must do all
554 * port deletions before any port additions. */
555 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
556 struct ofpp_garbage
*garbage
, *next
;
558 LIST_FOR_EACH_SAFE (garbage
, next
, list_node
, &br
->ofpp_garbage
) {
559 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
560 * internal datastructures may not be consistent. Eventually, when
561 * port additions and deletions are cheaper, these calls should be
564 ofproto_port_del(br
->ofproto
, garbage
->ofp_port
);
565 list_remove(&garbage
->list_node
);
568 if (time_msec() >= deadline
) {
575 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
576 struct if_cfg
*if_cfg
, *next
;
578 HMAP_FOR_EACH_SAFE (if_cfg
, next
, hmap_node
, &br
->if_cfg_todo
) {
579 iface_create(br
, if_cfg
, OFPP_NONE
);
580 if (time_msec() >= deadline
) {
590 bridge_reconfigure_continue(const struct ovsrec_open_vswitch
*ovs_cfg
)
592 struct sockaddr_in
*managers
;
593 int sflow_bridge_number
;
598 ovs_assert(reconfiguring
);
599 done
= bridge_reconfigure_ofp();
601 /* Complete the configuration. */
602 sflow_bridge_number
= 0;
603 collect_in_band_managers(ovs_cfg
, &managers
, &n_managers
);
604 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
607 /* We need the datapath ID early to allow LACP ports to use it as the
608 * default system ID. */
609 bridge_configure_datapath_id(br
);
611 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
614 port_configure(port
);
616 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
617 iface_configure_cfm(iface
);
618 iface_configure_qos(iface
, port
->cfg
->qos
);
619 iface_set_mac(iface
);
620 ofproto_port_set_bfd(br
->ofproto
, iface
->ofp_port
,
624 bridge_configure_mirrors(br
);
625 bridge_configure_forward_bpdu(br
);
626 bridge_configure_mac_table(br
);
627 bridge_configure_remotes(br
, managers
, n_managers
);
628 bridge_configure_netflow(br
);
629 bridge_configure_sflow(br
, &sflow_bridge_number
);
630 bridge_configure_ipfix(br
);
631 bridge_configure_stp(br
);
632 bridge_configure_tables(br
);
633 bridge_configure_dp_desc(br
);
635 if (smap_get(&br
->cfg
->other_config
, "flow-eviction-threshold")) {
636 /* XXX: Remove this warning message eventually. */
637 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
638 " moved to the Open_vSwitch table. Ignoring its"
639 " setting in the bridge table.");
647 /* Delete ofprotos which aren't configured or have the wrong type. Create
648 * ofprotos which don't exist but need to. */
650 bridge_update_ofprotos(void)
652 struct bridge
*br
, *next
;
657 /* Delete ofprotos with no bridge or with the wrong type. */
660 ofproto_enumerate_types(&types
);
661 SSET_FOR_EACH (type
, &types
) {
664 ofproto_enumerate_names(type
, &names
);
665 SSET_FOR_EACH (name
, &names
) {
666 br
= bridge_lookup(name
);
667 if (!br
|| strcmp(type
, br
->type
)) {
668 ofproto_delete(name
, type
);
672 sset_destroy(&names
);
673 sset_destroy(&types
);
675 /* Add ofprotos for bridges which don't have one yet. */
676 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
684 /* Remove ports from any datapath with the same name as 'br'. If we
685 * don't do this, creating 'br''s ofproto will fail because a port with
686 * the same name as its local port already exists. */
687 HMAP_FOR_EACH (br2
, node
, &all_bridges
) {
688 struct ofproto_port ofproto_port
;
694 if (!ofproto_port_query_by_name(br2
->ofproto
, br
->name
,
696 error
= ofproto_port_del(br2
->ofproto
, ofproto_port
.ofp_port
);
698 VLOG_ERR("failed to delete port %s: %s", ofproto_port
.name
,
699 ovs_strerror(error
));
701 ofproto_port_destroy(&ofproto_port
);
705 error
= ofproto_create(br
->name
, br
->type
, &br
->ofproto
);
707 VLOG_ERR("failed to create bridge %s: %s", br
->name
,
708 ovs_strerror(error
));
715 port_configure(struct port
*port
)
717 const struct ovsrec_port
*cfg
= port
->cfg
;
718 struct bond_settings bond_settings
;
719 struct lacp_settings lacp_settings
;
720 struct ofproto_bundle_settings s
;
723 if (cfg
->vlan_mode
&& !strcmp(cfg
->vlan_mode
, "splinter")) {
724 configure_splinter_port(port
);
733 s
.slaves
= xmalloc(list_size(&port
->ifaces
) * sizeof *s
.slaves
);
734 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
735 s
.slaves
[s
.n_slaves
++] = iface
->ofp_port
;
740 if (cfg
->tag
&& *cfg
->tag
>= 0 && *cfg
->tag
<= 4095) {
744 /* Get VLAN trunks. */
747 s
.trunks
= vlan_bitmap_from_array(cfg
->trunks
, cfg
->n_trunks
);
751 if (cfg
->vlan_mode
) {
752 if (!strcmp(cfg
->vlan_mode
, "access")) {
753 s
.vlan_mode
= PORT_VLAN_ACCESS
;
754 } else if (!strcmp(cfg
->vlan_mode
, "trunk")) {
755 s
.vlan_mode
= PORT_VLAN_TRUNK
;
756 } else if (!strcmp(cfg
->vlan_mode
, "native-tagged")) {
757 s
.vlan_mode
= PORT_VLAN_NATIVE_TAGGED
;
758 } else if (!strcmp(cfg
->vlan_mode
, "native-untagged")) {
759 s
.vlan_mode
= PORT_VLAN_NATIVE_UNTAGGED
;
761 /* This "can't happen" because ovsdb-server should prevent it. */
762 VLOG_ERR("unknown VLAN mode %s", cfg
->vlan_mode
);
763 s
.vlan_mode
= PORT_VLAN_TRUNK
;
767 s
.vlan_mode
= PORT_VLAN_ACCESS
;
769 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
773 s
.vlan_mode
= PORT_VLAN_TRUNK
;
776 s
.use_priority_tags
= smap_get_bool(&cfg
->other_config
, "priority-tags",
779 /* Get LACP settings. */
780 s
.lacp
= port_configure_lacp(port
, &lacp_settings
);
784 s
.lacp_slaves
= xmalloc(s
.n_slaves
* sizeof *s
.lacp_slaves
);
785 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
786 iface_configure_lacp(iface
, &s
.lacp_slaves
[i
++]);
789 s
.lacp_slaves
= NULL
;
792 /* Get bond settings. */
793 if (s
.n_slaves
> 1) {
794 s
.bond
= &bond_settings
;
795 port_configure_bond(port
, &bond_settings
);
798 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
799 netdev_set_miimon_interval(iface
->netdev
, 0);
804 ofproto_bundle_register(port
->bridge
->ofproto
, port
, &s
);
813 bridge_configure_flow_miss_model(const char *opt
)
815 enum ofproto_flow_miss_model model
= OFPROTO_HANDLE_MISS_AUTO
;
818 if (strcmp(opt
, "with-facets")) {
819 model
= OFPROTO_HANDLE_MISS_WITH_FACETS
;
820 } else if (strcmp(opt
, "without-facets")) {
821 model
= OFPROTO_HANDLE_MISS_WITHOUT_FACETS
;
825 ofproto_set_flow_miss_model(model
);
828 /* Pick local port hardware address and datapath ID for 'br'. */
830 bridge_configure_datapath_id(struct bridge
*br
)
832 uint8_t ea
[ETH_ADDR_LEN
];
834 struct iface
*local_iface
;
835 struct iface
*hw_addr_iface
;
838 bridge_pick_local_hw_addr(br
, ea
, &hw_addr_iface
);
839 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
841 int error
= netdev_set_etheraddr(local_iface
->netdev
, ea
);
843 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
844 VLOG_ERR_RL(&rl
, "bridge %s: failed to set bridge "
845 "Ethernet address: %s",
846 br
->name
, ovs_strerror(error
));
849 memcpy(br
->ea
, ea
, ETH_ADDR_LEN
);
851 dpid
= bridge_pick_datapath_id(br
, ea
, hw_addr_iface
);
852 if (dpid
!= ofproto_get_datapath_id(br
->ofproto
)) {
853 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64
, br
->name
, dpid
);
854 ofproto_set_datapath_id(br
->ofproto
, dpid
);
857 dpid_string
= xasprintf("%016"PRIx64
, dpid
);
858 ovsrec_bridge_set_datapath_id(br
->cfg
, dpid_string
);
862 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
865 bridge_get_allowed_versions(struct bridge
*br
)
867 if (!br
->cfg
->n_protocols
)
870 return ofputil_versions_from_strings(br
->cfg
->protocols
,
871 br
->cfg
->n_protocols
);
874 /* Set NetFlow configuration on 'br'. */
876 bridge_configure_netflow(struct bridge
*br
)
878 struct ovsrec_netflow
*cfg
= br
->cfg
->netflow
;
879 struct netflow_options opts
;
882 ofproto_set_netflow(br
->ofproto
, NULL
);
886 memset(&opts
, 0, sizeof opts
);
888 /* Get default NetFlow configuration from datapath.
889 * Apply overrides from 'cfg'. */
890 ofproto_get_netflow_ids(br
->ofproto
, &opts
.engine_type
, &opts
.engine_id
);
891 if (cfg
->engine_type
) {
892 opts
.engine_type
= *cfg
->engine_type
;
894 if (cfg
->engine_id
) {
895 opts
.engine_id
= *cfg
->engine_id
;
898 /* Configure active timeout interval. */
899 opts
.active_timeout
= cfg
->active_timeout
;
900 if (!opts
.active_timeout
) {
901 opts
.active_timeout
= -1;
902 } else if (opts
.active_timeout
< 0) {
903 VLOG_WARN("bridge %s: active timeout interval set to negative "
904 "value, using default instead (%d seconds)", br
->name
,
905 NF_ACTIVE_TIMEOUT_DEFAULT
);
906 opts
.active_timeout
= -1;
909 /* Add engine ID to interface number to disambiguate bridgs? */
910 opts
.add_id_to_iface
= cfg
->add_id_to_interface
;
911 if (opts
.add_id_to_iface
) {
912 if (opts
.engine_id
> 0x7f) {
913 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
914 "another vswitch, choose an engine id less than 128",
917 if (hmap_count(&br
->ports
) > 508) {
918 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
919 "another port when more than 508 ports are used",
925 sset_init(&opts
.collectors
);
926 sset_add_array(&opts
.collectors
, cfg
->targets
, cfg
->n_targets
);
929 if (ofproto_set_netflow(br
->ofproto
, &opts
)) {
930 VLOG_ERR("bridge %s: problem setting netflow collectors", br
->name
);
932 sset_destroy(&opts
.collectors
);
935 /* Set sFlow configuration on 'br'. */
937 bridge_configure_sflow(struct bridge
*br
, int *sflow_bridge_number
)
939 const struct ovsrec_sflow
*cfg
= br
->cfg
->sflow
;
940 struct ovsrec_controller
**controllers
;
941 struct ofproto_sflow_options oso
;
942 size_t n_controllers
;
946 ofproto_set_sflow(br
->ofproto
, NULL
);
950 memset(&oso
, 0, sizeof oso
);
952 sset_init(&oso
.targets
);
953 sset_add_array(&oso
.targets
, cfg
->targets
, cfg
->n_targets
);
955 oso
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
957 oso
.sampling_rate
= *cfg
->sampling
;
960 oso
.polling_interval
= SFL_DEFAULT_POLLING_INTERVAL
;
962 oso
.polling_interval
= *cfg
->polling
;
965 oso
.header_len
= SFL_DEFAULT_HEADER_SIZE
;
967 oso
.header_len
= *cfg
->header
;
970 oso
.sub_id
= (*sflow_bridge_number
)++;
971 oso
.agent_device
= cfg
->agent
;
973 oso
.control_ip
= NULL
;
974 n_controllers
= bridge_get_controllers(br
, &controllers
);
975 for (i
= 0; i
< n_controllers
; i
++) {
976 if (controllers
[i
]->local_ip
) {
977 oso
.control_ip
= controllers
[i
]->local_ip
;
981 ofproto_set_sflow(br
->ofproto
, &oso
);
983 sset_destroy(&oso
.targets
);
986 /* Set IPFIX configuration on 'br'. */
988 bridge_configure_ipfix(struct bridge
*br
)
990 const struct ovsrec_ipfix
*be_cfg
= br
->cfg
->ipfix
;
991 const struct ovsrec_flow_sample_collector_set
*fe_cfg
;
992 struct ofproto_ipfix_bridge_exporter_options be_opts
;
993 struct ofproto_ipfix_flow_exporter_options
*fe_opts
= NULL
;
994 size_t n_fe_opts
= 0;
996 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
997 if (fe_cfg
->bridge
== br
->cfg
) {
1002 if (!be_cfg
&& n_fe_opts
== 0) {
1003 ofproto_set_ipfix(br
->ofproto
, NULL
, NULL
, 0);
1008 memset(&be_opts
, 0, sizeof be_opts
);
1010 sset_init(&be_opts
.targets
);
1011 sset_add_array(&be_opts
.targets
, be_cfg
->targets
, be_cfg
->n_targets
);
1013 if (be_cfg
->sampling
) {
1014 be_opts
.sampling_rate
= *be_cfg
->sampling
;
1016 be_opts
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
1018 if (be_cfg
->obs_domain_id
) {
1019 be_opts
.obs_domain_id
= *be_cfg
->obs_domain_id
;
1021 if (be_cfg
->obs_point_id
) {
1022 be_opts
.obs_point_id
= *be_cfg
->obs_point_id
;
1024 if (be_cfg
->cache_active_timeout
) {
1025 be_opts
.cache_active_timeout
= *be_cfg
->cache_active_timeout
;
1027 if (be_cfg
->cache_max_flows
) {
1028 be_opts
.cache_max_flows
= *be_cfg
->cache_max_flows
;
1032 if (n_fe_opts
> 0) {
1033 struct ofproto_ipfix_flow_exporter_options
*opts
;
1034 fe_opts
= xcalloc(n_fe_opts
, sizeof *fe_opts
);
1036 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg
, idl
) {
1037 if (fe_cfg
->bridge
== br
->cfg
) {
1038 opts
->collector_set_id
= fe_cfg
->id
;
1039 sset_init(&opts
->targets
);
1040 sset_add_array(&opts
->targets
, fe_cfg
->ipfix
->targets
,
1041 fe_cfg
->ipfix
->n_targets
);
1042 opts
->cache_active_timeout
= fe_cfg
->ipfix
->cache_active_timeout
1043 ? *fe_cfg
->ipfix
->cache_active_timeout
: 0;
1044 opts
->cache_max_flows
= fe_cfg
->ipfix
->cache_max_flows
1045 ? *fe_cfg
->ipfix
->cache_max_flows
: 0;
1051 ofproto_set_ipfix(br
->ofproto
, be_cfg
? &be_opts
: NULL
, fe_opts
,
1055 sset_destroy(&be_opts
.targets
);
1058 if (n_fe_opts
> 0) {
1059 struct ofproto_ipfix_flow_exporter_options
*opts
= fe_opts
;
1061 for (i
= 0; i
< n_fe_opts
; i
++) {
1062 sset_destroy(&opts
->targets
);
1070 port_configure_stp(const struct ofproto
*ofproto
, struct port
*port
,
1071 struct ofproto_port_stp_settings
*port_s
,
1072 int *port_num_counter
, unsigned long *port_num_bitmap
)
1074 const char *config_str
;
1075 struct iface
*iface
;
1077 if (!smap_get_bool(&port
->cfg
->other_config
, "stp-enable", true)) {
1078 port_s
->enable
= false;
1081 port_s
->enable
= true;
1084 /* STP over bonds is not supported. */
1085 if (!list_is_singleton(&port
->ifaces
)) {
1086 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1088 port_s
->enable
= false;
1092 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
1094 /* Internal ports shouldn't participate in spanning tree, so
1096 if (!strcmp(iface
->type
, "internal")) {
1097 VLOG_DBG("port %s: disable STP on internal ports", port
->name
);
1098 port_s
->enable
= false;
1102 /* STP on mirror output ports is not supported. */
1103 if (ofproto_is_mirror_output_bundle(ofproto
, port
)) {
1104 VLOG_DBG("port %s: disable STP on mirror ports", port
->name
);
1105 port_s
->enable
= false;
1109 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-num");
1111 unsigned long int port_num
= strtoul(config_str
, NULL
, 0);
1112 int port_idx
= port_num
- 1;
1114 if (port_num
< 1 || port_num
> STP_MAX_PORTS
) {
1115 VLOG_ERR("port %s: invalid stp-port-num", port
->name
);
1116 port_s
->enable
= false;
1120 if (bitmap_is_set(port_num_bitmap
, port_idx
)) {
1121 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1122 port
->name
, port_num
);
1123 port_s
->enable
= false;
1126 bitmap_set1(port_num_bitmap
, port_idx
);
1127 port_s
->port_num
= port_idx
;
1129 if (*port_num_counter
>= STP_MAX_PORTS
) {
1130 VLOG_ERR("port %s: too many STP ports, disabling", port
->name
);
1131 port_s
->enable
= false;
1135 port_s
->port_num
= (*port_num_counter
)++;
1138 config_str
= smap_get(&port
->cfg
->other_config
, "stp-path-cost");
1140 port_s
->path_cost
= strtoul(config_str
, NULL
, 10);
1142 enum netdev_features current
;
1145 netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1146 mbps
= netdev_features_to_bps(current
, 100 * 1000 * 1000) / 1000000;
1147 port_s
->path_cost
= stp_convert_speed_to_cost(mbps
);
1150 config_str
= smap_get(&port
->cfg
->other_config
, "stp-port-priority");
1152 port_s
->priority
= strtoul(config_str
, NULL
, 0);
1154 port_s
->priority
= STP_DEFAULT_PORT_PRIORITY
;
1158 /* Set spanning tree configuration on 'br'. */
1160 bridge_configure_stp(struct bridge
*br
)
1162 if (!br
->cfg
->stp_enable
) {
1163 ofproto_set_stp(br
->ofproto
, NULL
);
1165 struct ofproto_stp_settings br_s
;
1166 const char *config_str
;
1168 int port_num_counter
;
1169 unsigned long *port_num_bitmap
;
1171 config_str
= smap_get(&br
->cfg
->other_config
, "stp-system-id");
1173 uint8_t ea
[ETH_ADDR_LEN
];
1175 if (eth_addr_from_string(config_str
, ea
)) {
1176 br_s
.system_id
= eth_addr_to_uint64(ea
);
1178 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1179 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1180 "to "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(br
->ea
));
1183 br_s
.system_id
= eth_addr_to_uint64(br
->ea
);
1186 config_str
= smap_get(&br
->cfg
->other_config
, "stp-priority");
1188 br_s
.priority
= strtoul(config_str
, NULL
, 0);
1190 br_s
.priority
= STP_DEFAULT_BRIDGE_PRIORITY
;
1193 config_str
= smap_get(&br
->cfg
->other_config
, "stp-hello-time");
1195 br_s
.hello_time
= strtoul(config_str
, NULL
, 10) * 1000;
1197 br_s
.hello_time
= STP_DEFAULT_HELLO_TIME
;
1200 config_str
= smap_get(&br
->cfg
->other_config
, "stp-max-age");
1202 br_s
.max_age
= strtoul(config_str
, NULL
, 10) * 1000;
1204 br_s
.max_age
= STP_DEFAULT_MAX_AGE
;
1207 config_str
= smap_get(&br
->cfg
->other_config
, "stp-forward-delay");
1209 br_s
.fwd_delay
= strtoul(config_str
, NULL
, 10) * 1000;
1211 br_s
.fwd_delay
= STP_DEFAULT_FWD_DELAY
;
1214 /* Configure STP on the bridge. */
1215 if (ofproto_set_stp(br
->ofproto
, &br_s
)) {
1216 VLOG_ERR("bridge %s: could not enable STP", br
->name
);
1220 /* Users must either set the port number with the "stp-port-num"
1221 * configuration on all ports or none. If manual configuration
1222 * is not done, then we allocate them sequentially. */
1223 port_num_counter
= 0;
1224 port_num_bitmap
= bitmap_allocate(STP_MAX_PORTS
);
1225 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1226 struct ofproto_port_stp_settings port_s
;
1227 struct iface
*iface
;
1229 port_configure_stp(br
->ofproto
, port
, &port_s
,
1230 &port_num_counter
, port_num_bitmap
);
1232 /* As bonds are not supported, just apply configuration to
1233 * all interfaces. */
1234 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1235 if (ofproto_port_set_stp(br
->ofproto
, iface
->ofp_port
,
1237 VLOG_ERR("port %s: could not enable STP", port
->name
);
1243 if (bitmap_scan(port_num_bitmap
, 0, STP_MAX_PORTS
) != STP_MAX_PORTS
1244 && port_num_counter
) {
1245 VLOG_ERR("bridge %s: must manually configure all STP port "
1246 "IDs or none, disabling", br
->name
);
1247 ofproto_set_stp(br
->ofproto
, NULL
);
1249 bitmap_free(port_num_bitmap
);
1254 bridge_has_bond_fake_iface(const struct bridge
*br
, const char *name
)
1256 const struct port
*port
= port_lookup(br
, name
);
1257 return port
&& port_is_bond_fake_iface(port
);
1261 port_is_bond_fake_iface(const struct port
*port
)
1263 return port
->cfg
->bond_fake_iface
&& !list_is_short(&port
->ifaces
);
1267 add_del_bridges(const struct ovsrec_open_vswitch
*cfg
)
1269 struct bridge
*br
, *next
;
1270 struct shash new_br
;
1273 /* Collect new bridges' names and types. */
1274 shash_init(&new_br
);
1275 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1276 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1277 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1279 if (strchr(br_cfg
->name
, '/')) {
1280 /* Prevent remote ovsdb-server users from accessing arbitrary
1281 * directories, e.g. consider a bridge named "../../../etc/". */
1282 VLOG_WARN_RL(&rl
, "ignoring bridge with invalid name \"%s\"",
1284 } else if (!shash_add_once(&new_br
, br_cfg
->name
, br_cfg
)) {
1285 VLOG_WARN_RL(&rl
, "bridge %s specified twice", br_cfg
->name
);
1289 /* Get rid of deleted bridges or those whose types have changed.
1290 * Update 'cfg' of bridges that still exist. */
1291 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
1292 br
->cfg
= shash_find_data(&new_br
, br
->name
);
1293 if (!br
->cfg
|| strcmp(br
->type
, ofproto_normalize_type(
1294 br
->cfg
->datapath_type
))) {
1299 /* Add new bridges. */
1300 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
1301 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
1302 struct bridge
*br
= bridge_lookup(br_cfg
->name
);
1304 bridge_create(br_cfg
);
1308 shash_destroy(&new_br
);
1312 iface_set_ofp_port(struct iface
*iface
, ofp_port_t ofp_port
)
1314 struct bridge
*br
= iface
->port
->bridge
;
1316 ovs_assert(iface
->ofp_port
== OFPP_NONE
&& ofp_port
!= OFPP_NONE
);
1317 iface
->ofp_port
= ofp_port
;
1318 hmap_insert(&br
->ifaces
, &iface
->ofp_port_node
,
1319 hash_ofp_port(ofp_port
));
1320 iface_set_ofport(iface
->cfg
, ofp_port
);
1323 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1324 * Returns 0 if successful, otherwise a positive errno value. */
1326 iface_set_netdev_config(const struct ovsrec_interface
*iface_cfg
,
1327 struct netdev
*netdev
)
1331 error
= netdev_set_config(netdev
, &iface_cfg
->options
);
1333 VLOG_WARN("could not configure network device %s (%s)",
1334 iface_cfg
->name
, ovs_strerror(error
));
1339 /* This function determines whether 'ofproto_port', which is attached to
1340 * br->ofproto's datapath, is one that we want in 'br'.
1342 * If it is, it returns true, after creating an iface (if necessary),
1343 * configuring the iface's netdev according to the iface's options, and setting
1344 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1346 * If, on the other hand, 'port' should be removed, it returns false. The
1347 * caller should later detach the port from br->ofproto. */
1349 bridge_refresh_one_ofp_port(struct bridge
*br
,
1350 const struct ofproto_port
*ofproto_port
)
1352 const char *name
= ofproto_port
->name
;
1353 const char *type
= ofproto_port
->type
;
1354 ofp_port_t ofp_port
= ofproto_port
->ofp_port
;
1356 struct iface
*iface
= iface_lookup(br
, name
);
1358 /* Check that the name-to-number mapping is one-to-one. */
1359 if (iface
->ofp_port
!= OFPP_NONE
) {
1360 VLOG_WARN("bridge %s: interface %s reported twice",
1363 } else if (iface_from_ofp_port(br
, ofp_port
)) {
1364 VLOG_WARN("bridge %s: interface %"PRIu16
" reported twice",
1365 br
->name
, ofp_port
);
1369 /* There's a configured interface named 'name'. */
1370 if (strcmp(type
, iface
->type
)
1371 || iface_set_netdev_config(iface
->cfg
, iface
->netdev
)) {
1372 /* It's the wrong type, or it's the right type but can't be
1373 * configured as the user requested, so we must destroy it. */
1376 /* It's the right type and configured correctly. Keep it. */
1377 iface_set_ofp_port(iface
, ofp_port
);
1380 } else if (bridge_has_bond_fake_iface(br
, name
)
1381 && !strcmp(type
, "internal")) {
1382 /* It's a bond fake iface. Keep it. */
1385 /* There's no configured interface named 'name', but there might be an
1386 * interface of that name queued to be created.
1388 * If there is, and it has the correct type, then try to configure it
1389 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1390 * delete it and later try to re-add it. */
1391 struct if_cfg
*if_cfg
= if_cfg_lookup(br
, name
);
1393 && !strcmp(type
, iface_get_type(if_cfg
->cfg
, br
->cfg
))
1394 && iface_create(br
, if_cfg
, ofp_port
));
1398 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1399 * consistent with the ofp_ports in "br->ofproto". */
1401 bridge_refresh_ofp_port(struct bridge
*br
)
1403 struct ofproto_port_dump dump
;
1404 struct ofproto_port ofproto_port
;
1405 struct port
*port
, *port_next
;
1407 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1408 hmap_clear(&br
->ifaces
);
1409 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1410 struct iface
*iface
;
1412 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1413 iface
->ofp_port
= OFPP_NONE
;
1417 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1418 * already exist in the datapath and promote them to full fledged "struct
1419 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1420 OFPROTO_PORT_FOR_EACH (&ofproto_port
, &dump
, br
->ofproto
) {
1421 if (!bridge_refresh_one_ofp_port(br
, &ofproto_port
)) {
1422 struct ofpp_garbage
*garbage
= xmalloc(sizeof *garbage
);
1423 garbage
->ofp_port
= ofproto_port
.ofp_port
;
1424 list_push_front(&br
->ofpp_garbage
, &garbage
->list_node
);
1428 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1429 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1430 * later they can be added to ofproto. */
1431 HMAP_FOR_EACH_SAFE (port
, port_next
, hmap_node
, &br
->ports
) {
1432 struct iface
*iface
, *iface_next
;
1434 LIST_FOR_EACH_SAFE (iface
, iface_next
, port_elem
, &port
->ifaces
) {
1435 if (iface
->ofp_port
== OFPP_NONE
) {
1436 bridge_queue_if_cfg(br
, iface
->cfg
, port
->cfg
);
1437 iface_destroy(iface
);
1441 if (list_is_empty(&port
->ifaces
)) {
1447 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1448 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1449 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1452 * If successful, returns 0 and stores the network device in '*netdevp'. On
1453 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1455 iface_do_create(const struct bridge
*br
,
1456 const struct if_cfg
*if_cfg
,
1457 ofp_port_t
*ofp_portp
, struct netdev
**netdevp
)
1459 const struct ovsrec_interface
*iface_cfg
= if_cfg
->cfg
;
1460 const struct ovsrec_port
*port_cfg
= if_cfg
->parent
;
1461 struct netdev
*netdev
= NULL
;
1464 if (netdev_is_reserved_name(iface_cfg
->name
)) {
1465 VLOG_WARN("could not create interface %s, name is reserved",
1471 error
= netdev_open(iface_cfg
->name
,
1472 iface_get_type(iface_cfg
, br
->cfg
), &netdev
);
1474 VLOG_WARN("could not open network device %s (%s)",
1475 iface_cfg
->name
, ovs_strerror(error
));
1479 error
= iface_set_netdev_config(iface_cfg
, netdev
);
1484 if (*ofp_portp
== OFPP_NONE
) {
1485 ofp_port_t ofp_port
= if_cfg
->ofport
;
1487 error
= ofproto_port_add(br
->ofproto
, netdev
, &ofp_port
);
1491 *ofp_portp
= ofp_port
;
1493 VLOG_INFO("bridge %s: added interface %s on port %d",
1494 br
->name
, iface_cfg
->name
, *ofp_portp
);
1496 VLOG_DBG("bridge %s: interface %s is on port %d",
1497 br
->name
, iface_cfg
->name
, *ofp_portp
);
1500 if ((port_cfg
->vlan_mode
&& !strcmp(port_cfg
->vlan_mode
, "splinter"))
1501 || iface_is_internal(iface_cfg
, br
->cfg
)) {
1502 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
1510 netdev_close(netdev
);
1514 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1515 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1516 * automatically allocated for the iface. Takes ownership of and
1517 * deallocates 'if_cfg'.
1519 * Return true if an iface is successfully created, false otherwise. */
1521 iface_create(struct bridge
*br
, struct if_cfg
*if_cfg
, ofp_port_t ofp_port
)
1523 const struct ovsrec_interface
*iface_cfg
= if_cfg
->cfg
;
1524 const struct ovsrec_port
*port_cfg
= if_cfg
->parent
;
1526 struct netdev
*netdev
;
1527 struct iface
*iface
;
1532 /* Do the bits that can fail up front.
1534 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1535 * internal datastructures may not be consistent. Eventually, when port
1536 * additions and deletions are cheaper, these calls should be removed. */
1538 ovs_assert(!iface_lookup(br
, iface_cfg
->name
));
1539 error
= iface_do_create(br
, if_cfg
, &ofp_port
, &netdev
);
1542 iface_set_ofport(iface_cfg
, OFPP_NONE
);
1543 iface_clear_db_record(iface_cfg
);
1548 /* Get or create the port structure. */
1549 port
= port_lookup(br
, port_cfg
->name
);
1551 port
= port_create(br
, port_cfg
);
1554 /* Create the iface structure. */
1555 iface
= xzalloc(sizeof *iface
);
1556 list_push_back(&port
->ifaces
, &iface
->port_elem
);
1557 hmap_insert(&br
->iface_by_name
, &iface
->name_node
,
1558 hash_string(iface_cfg
->name
, 0));
1560 iface
->name
= xstrdup(iface_cfg
->name
);
1561 iface
->ofp_port
= OFPP_NONE
;
1562 iface
->netdev
= netdev
;
1563 iface
->type
= iface_get_type(iface_cfg
, br
->cfg
);
1564 iface
->cfg
= iface_cfg
;
1566 iface_set_ofp_port(iface
, ofp_port
);
1568 /* Populate initial status in database. */
1569 iface_refresh_stats(iface
);
1570 iface_refresh_status(iface
);
1572 /* Add bond fake iface if necessary. */
1573 if (port_is_bond_fake_iface(port
)) {
1574 struct ofproto_port ofproto_port
;
1576 if (ofproto_port_query_by_name(br
->ofproto
, port
->name
,
1578 struct netdev
*netdev
;
1581 error
= netdev_open(port
->name
, "internal", &netdev
);
1583 ofp_port_t fake_ofp_port
= if_cfg
->ofport
;
1585 ofproto_port_add(br
->ofproto
, netdev
, &fake_ofp_port
);
1586 netdev_close(netdev
);
1588 VLOG_WARN("could not open network device %s (%s)",
1589 port
->name
, ovs_strerror(error
));
1592 /* Already exists, nothing to do. */
1593 ofproto_port_destroy(&ofproto_port
);
1598 hmap_remove(&br
->if_cfg_todo
, &if_cfg
->hmap_node
);
1604 /* Set forward BPDU option. */
1606 bridge_configure_forward_bpdu(struct bridge
*br
)
1608 ofproto_set_forward_bpdu(br
->ofproto
,
1609 smap_get_bool(&br
->cfg
->other_config
,
1614 /* Set MAC learning table configuration for 'br'. */
1616 bridge_configure_mac_table(struct bridge
*br
)
1618 const char *idle_time_str
;
1621 const char *mac_table_size_str
;
1624 idle_time_str
= smap_get(&br
->cfg
->other_config
, "mac-aging-time");
1625 idle_time
= (idle_time_str
&& atoi(idle_time_str
)
1626 ? atoi(idle_time_str
)
1627 : MAC_ENTRY_DEFAULT_IDLE_TIME
);
1629 mac_table_size_str
= smap_get(&br
->cfg
->other_config
, "mac-table-size");
1630 mac_table_size
= (mac_table_size_str
&& atoi(mac_table_size_str
)
1631 ? atoi(mac_table_size_str
)
1634 ofproto_set_mac_table_config(br
->ofproto
, idle_time
, mac_table_size
);
1638 bridge_pick_local_hw_addr(struct bridge
*br
, uint8_t ea
[ETH_ADDR_LEN
],
1639 struct iface
**hw_addr_iface
)
1641 struct hmapx mirror_output_ports
;
1644 bool found_addr
= false;
1648 *hw_addr_iface
= NULL
;
1650 /* Did the user request a particular MAC? */
1651 hwaddr
= smap_get(&br
->cfg
->other_config
, "hwaddr");
1652 if (hwaddr
&& eth_addr_from_string(hwaddr
, ea
)) {
1653 if (eth_addr_is_multicast(ea
)) {
1654 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1655 "address "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(ea
));
1656 } else if (eth_addr_is_zero(ea
)) {
1657 VLOG_ERR("bridge %s: cannot set MAC address to zero", br
->name
);
1663 /* Mirror output ports don't participate in picking the local hardware
1664 * address. ofproto can't help us find out whether a given port is a
1665 * mirror output because we haven't configured mirrors yet, so we need to
1666 * accumulate them ourselves. */
1667 hmapx_init(&mirror_output_ports
);
1668 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
1669 struct ovsrec_mirror
*m
= br
->cfg
->mirrors
[i
];
1670 if (m
->output_port
) {
1671 hmapx_add(&mirror_output_ports
, m
->output_port
);
1675 /* Otherwise choose the minimum non-local MAC address among all of the
1677 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1678 uint8_t iface_ea
[ETH_ADDR_LEN
];
1679 struct iface
*candidate
;
1680 struct iface
*iface
;
1682 /* Mirror output ports don't participate. */
1683 if (hmapx_contains(&mirror_output_ports
, port
->cfg
)) {
1687 /* Choose the MAC address to represent the port. */
1689 if (port
->cfg
->mac
&& eth_addr_from_string(port
->cfg
->mac
, iface_ea
)) {
1690 /* Find the interface with this Ethernet address (if any) so that
1691 * we can provide the correct devname to the caller. */
1692 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1693 uint8_t candidate_ea
[ETH_ADDR_LEN
];
1694 if (!netdev_get_etheraddr(candidate
->netdev
, candidate_ea
)
1695 && eth_addr_equals(iface_ea
, candidate_ea
)) {
1700 /* Choose the interface whose MAC address will represent the port.
1701 * The Linux kernel bonding code always chooses the MAC address of
1702 * the first slave added to a bond, and the Fedora networking
1703 * scripts always add slaves to a bond in alphabetical order, so
1704 * for compatibility we choose the interface with the name that is
1705 * first in alphabetical order. */
1706 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1707 if (!iface
|| strcmp(candidate
->name
, iface
->name
) < 0) {
1712 /* The local port doesn't count (since we're trying to choose its
1713 * MAC address anyway). */
1714 if (iface
->ofp_port
== OFPP_LOCAL
) {
1719 error
= netdev_get_etheraddr(iface
->netdev
, iface_ea
);
1725 /* Compare against our current choice. */
1726 if (!eth_addr_is_multicast(iface_ea
) &&
1727 !eth_addr_is_local(iface_ea
) &&
1728 !eth_addr_is_reserved(iface_ea
) &&
1729 !eth_addr_is_zero(iface_ea
) &&
1730 (!found_addr
|| eth_addr_compare_3way(iface_ea
, ea
) < 0))
1732 memcpy(ea
, iface_ea
, ETH_ADDR_LEN
);
1733 *hw_addr_iface
= iface
;
1739 memcpy(ea
, br
->default_ea
, ETH_ADDR_LEN
);
1740 *hw_addr_iface
= NULL
;
1743 hmapx_destroy(&mirror_output_ports
);
1746 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1747 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1748 * an interface on 'br', then that interface must be passed in as
1749 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1750 * 'hw_addr_iface' must be passed in as a null pointer. */
1752 bridge_pick_datapath_id(struct bridge
*br
,
1753 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
1754 struct iface
*hw_addr_iface
)
1757 * The procedure for choosing a bridge MAC address will, in the most
1758 * ordinary case, also choose a unique MAC that we can use as a datapath
1759 * ID. In some special cases, though, multiple bridges will end up with
1760 * the same MAC address. This is OK for the bridges, but it will confuse
1761 * the OpenFlow controller, because each datapath needs a unique datapath
1764 * Datapath IDs must be unique. It is also very desirable that they be
1765 * stable from one run to the next, so that policy set on a datapath
1768 const char *datapath_id
;
1771 datapath_id
= smap_get(&br
->cfg
->other_config
, "datapath-id");
1772 if (datapath_id
&& dpid_from_string(datapath_id
, &dpid
)) {
1776 if (!hw_addr_iface
) {
1778 * A purely internal bridge, that is, one that has no non-virtual
1779 * network devices on it at all, is difficult because it has no
1780 * natural unique identifier at all.
1782 * When the host is a XenServer, we handle this case by hashing the
1783 * host's UUID with the name of the bridge. Names of bridges are
1784 * persistent across XenServer reboots, although they can be reused if
1785 * an internal network is destroyed and then a new one is later
1786 * created, so this is fairly effective.
1788 * When the host is not a XenServer, we punt by using a random MAC
1789 * address on each run.
1791 const char *host_uuid
= xenserver_get_host_uuid();
1793 char *combined
= xasprintf("%s,%s", host_uuid
, br
->name
);
1794 dpid
= dpid_from_hash(combined
, strlen(combined
));
1800 return eth_addr_to_uint64(bridge_ea
);
1804 dpid_from_hash(const void *data
, size_t n
)
1806 uint8_t hash
[SHA1_DIGEST_SIZE
];
1808 BUILD_ASSERT_DECL(sizeof hash
>= ETH_ADDR_LEN
);
1809 sha1_bytes(data
, n
, hash
);
1810 eth_addr_mark_random(hash
);
1811 return eth_addr_to_uint64(hash
);
1815 iface_refresh_status(struct iface
*iface
)
1819 enum netdev_features current
;
1823 uint8_t mac
[ETH_ADDR_LEN
];
1827 if (iface_is_synthetic(iface
)) {
1833 if (!netdev_get_status(iface
->netdev
, &smap
)) {
1834 ovsrec_interface_set_status(iface
->cfg
, &smap
);
1836 ovsrec_interface_set_status(iface
->cfg
, NULL
);
1839 smap_destroy(&smap
);
1841 error
= netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1842 bps
= !error
? netdev_features_to_bps(current
, 0) : 0;
1844 ovsrec_interface_set_duplex(iface
->cfg
,
1845 netdev_features_is_full_duplex(current
)
1847 ovsrec_interface_set_link_speed(iface
->cfg
, &bps
, 1);
1849 ovsrec_interface_set_duplex(iface
->cfg
, NULL
);
1850 ovsrec_interface_set_link_speed(iface
->cfg
, NULL
, 0);
1853 error
= netdev_get_mtu(iface
->netdev
, &mtu
);
1856 ovsrec_interface_set_mtu(iface
->cfg
, &mtu_64
, 1);
1858 ovsrec_interface_set_mtu(iface
->cfg
, NULL
, 0);
1861 error
= netdev_get_etheraddr(iface
->netdev
, mac
);
1863 char mac_string
[32];
1865 sprintf(mac_string
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
1866 ovsrec_interface_set_mac_in_use(iface
->cfg
, mac_string
);
1868 ovsrec_interface_set_mac_in_use(iface
->cfg
, NULL
);
1871 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1872 * if there is no valid ifindex number. */
1873 ifindex64
= netdev_get_ifindex(iface
->netdev
);
1874 if (ifindex64
< 0) {
1877 ovsrec_interface_set_ifindex(iface
->cfg
, &ifindex64
, 1);
1880 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1883 iface_refresh_cfm_stats(struct iface
*iface
)
1885 const struct ovsrec_interface
*cfg
= iface
->cfg
;
1886 struct ofproto_cfm_status status
;
1888 if (!ofproto_port_get_cfm_status(iface
->port
->bridge
->ofproto
,
1889 iface
->ofp_port
, &status
)) {
1890 ovsrec_interface_set_cfm_fault(cfg
, NULL
, 0);
1891 ovsrec_interface_set_cfm_fault_status(cfg
, NULL
, 0);
1892 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
1893 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
1894 ovsrec_interface_set_cfm_remote_mpids(cfg
, NULL
, 0);
1896 const char *reasons
[CFM_FAULT_N_REASONS
];
1897 int64_t cfm_health
= status
.health
;
1898 bool faulted
= status
.faults
!= 0;
1901 ovsrec_interface_set_cfm_fault(cfg
, &faulted
, 1);
1904 for (i
= 0; i
< CFM_FAULT_N_REASONS
; i
++) {
1905 int reason
= 1 << i
;
1906 if (status
.faults
& reason
) {
1907 reasons
[j
++] = cfm_fault_reason_to_str(reason
);
1910 ovsrec_interface_set_cfm_fault_status(cfg
, (char **) reasons
, j
);
1912 if (status
.remote_opstate
>= 0) {
1913 const char *remote_opstate
= status
.remote_opstate
? "up" : "down";
1914 ovsrec_interface_set_cfm_remote_opstate(cfg
, remote_opstate
);
1916 ovsrec_interface_set_cfm_remote_opstate(cfg
, NULL
);
1919 ovsrec_interface_set_cfm_remote_mpids(cfg
,
1920 (const int64_t *)status
.rmps
,
1922 if (cfm_health
>= 0) {
1923 ovsrec_interface_set_cfm_health(cfg
, &cfm_health
, 1);
1925 ovsrec_interface_set_cfm_health(cfg
, NULL
, 0);
1933 iface_refresh_stats(struct iface
*iface
)
1935 #define IFACE_STATS \
1936 IFACE_STAT(rx_packets, "rx_packets") \
1937 IFACE_STAT(tx_packets, "tx_packets") \
1938 IFACE_STAT(rx_bytes, "rx_bytes") \
1939 IFACE_STAT(tx_bytes, "tx_bytes") \
1940 IFACE_STAT(rx_dropped, "rx_dropped") \
1941 IFACE_STAT(tx_dropped, "tx_dropped") \
1942 IFACE_STAT(rx_errors, "rx_errors") \
1943 IFACE_STAT(tx_errors, "tx_errors") \
1944 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1945 IFACE_STAT(rx_over_errors, "rx_over_err") \
1946 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1947 IFACE_STAT(collisions, "collisions")
1949 #define IFACE_STAT(MEMBER, NAME) + 1
1950 enum { N_IFACE_STATS
= IFACE_STATS
};
1952 int64_t values
[N_IFACE_STATS
];
1953 char *keys
[N_IFACE_STATS
];
1956 struct netdev_stats stats
;
1958 if (iface_is_synthetic(iface
)) {
1962 /* Intentionally ignore return value, since errors will set 'stats' to
1963 * all-1s, and we will deal with that correctly below. */
1964 netdev_get_stats(iface
->netdev
, &stats
);
1966 /* Copy statistics into keys[] and values[]. */
1968 #define IFACE_STAT(MEMBER, NAME) \
1969 if (stats.MEMBER != UINT64_MAX) { \
1971 values[n] = stats.MEMBER; \
1976 ovs_assert(n
<= N_IFACE_STATS
);
1978 ovsrec_interface_set_statistics(iface
->cfg
, keys
, values
, n
);
1983 br_refresh_stp_status(struct bridge
*br
)
1985 struct smap smap
= SMAP_INITIALIZER(&smap
);
1986 struct ofproto
*ofproto
= br
->ofproto
;
1987 struct ofproto_stp_status status
;
1989 if (ofproto_get_stp_status(ofproto
, &status
)) {
1993 if (!status
.enabled
) {
1994 ovsrec_bridge_set_status(br
->cfg
, NULL
);
1998 smap_add_format(&smap
, "stp_bridge_id", STP_ID_FMT
,
1999 STP_ID_ARGS(status
.bridge_id
));
2000 smap_add_format(&smap
, "stp_designated_root", STP_ID_FMT
,
2001 STP_ID_ARGS(status
.designated_root
));
2002 smap_add_format(&smap
, "stp_root_path_cost", "%d", status
.root_path_cost
);
2004 ovsrec_bridge_set_status(br
->cfg
, &smap
);
2005 smap_destroy(&smap
);
2009 port_refresh_stp_status(struct port
*port
)
2011 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
2012 struct iface
*iface
;
2013 struct ofproto_port_stp_status status
;
2015 int64_t int_values
[3];
2018 if (port_is_synthetic(port
)) {
2022 /* STP doesn't currently support bonds. */
2023 if (!list_is_singleton(&port
->ifaces
)) {
2024 ovsrec_port_set_status(port
->cfg
, NULL
);
2028 iface
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
, port_elem
);
2030 if (ofproto_port_get_stp_status(ofproto
, iface
->ofp_port
, &status
)) {
2034 if (!status
.enabled
) {
2035 ovsrec_port_set_status(port
->cfg
, NULL
);
2036 ovsrec_port_set_statistics(port
->cfg
, NULL
, NULL
, 0);
2040 /* Set Status column. */
2042 smap_add_format(&smap
, "stp_port_id", STP_PORT_ID_FMT
, status
.port_id
);
2043 smap_add(&smap
, "stp_state", stp_state_name(status
.state
));
2044 smap_add_format(&smap
, "stp_sec_in_state", "%u", status
.sec_in_state
);
2045 smap_add(&smap
, "stp_role", stp_role_name(status
.role
));
2046 ovsrec_port_set_status(port
->cfg
, &smap
);
2047 smap_destroy(&smap
);
2049 /* Set Statistics column. */
2050 keys
[0] = "stp_tx_count";
2051 int_values
[0] = status
.tx_count
;
2052 keys
[1] = "stp_rx_count";
2053 int_values
[1] = status
.rx_count
;
2054 keys
[2] = "stp_error_count";
2055 int_values
[2] = status
.error_count
;
2057 ovsrec_port_set_statistics(port
->cfg
, keys
, int_values
,
2058 ARRAY_SIZE(int_values
));
2062 enable_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2064 return smap_get_bool(&cfg
->other_config
, "enable-statistics", false);
2068 reconfigure_system_stats(const struct ovsrec_open_vswitch
*cfg
)
2070 bool enable
= enable_system_stats(cfg
);
2072 system_stats_enable(enable
);
2074 ovsrec_open_vswitch_set_statistics(cfg
, NULL
);
2079 run_system_stats(void)
2081 const struct ovsrec_open_vswitch
*cfg
= ovsrec_open_vswitch_first(idl
);
2084 stats
= system_stats_run();
2086 struct ovsdb_idl_txn
*txn
;
2087 struct ovsdb_datum datum
;
2089 txn
= ovsdb_idl_txn_create(idl
);
2090 ovsdb_datum_from_smap(&datum
, stats
);
2091 ovsdb_idl_txn_write(&cfg
->header_
, &ovsrec_open_vswitch_col_statistics
,
2093 ovsdb_idl_txn_commit(txn
);
2094 ovsdb_idl_txn_destroy(txn
);
2100 static inline const char *
2101 ofp12_controller_role_to_str(enum ofp12_controller_role role
)
2104 case OFPCR12_ROLE_EQUAL
:
2106 case OFPCR12_ROLE_MASTER
:
2108 case OFPCR12_ROLE_SLAVE
:
2110 case OFPCR12_ROLE_NOCHANGE
:
2112 return "*** INVALID ROLE ***";
2117 refresh_controller_status(void)
2121 const struct ovsrec_controller
*cfg
;
2125 /* Accumulate status for controllers on all bridges. */
2126 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2127 ofproto_get_ofproto_controller_info(br
->ofproto
, &info
);
2130 /* Update each controller in the database with current status. */
2131 OVSREC_CONTROLLER_FOR_EACH(cfg
, idl
) {
2132 struct ofproto_controller_info
*cinfo
=
2133 shash_find_data(&info
, cfg
->target
);
2136 struct smap smap
= SMAP_INITIALIZER(&smap
);
2137 const char **values
= cinfo
->pairs
.values
;
2138 const char **keys
= cinfo
->pairs
.keys
;
2141 for (i
= 0; i
< cinfo
->pairs
.n
; i
++) {
2142 smap_add(&smap
, keys
[i
], values
[i
]);
2145 ovsrec_controller_set_is_connected(cfg
, cinfo
->is_connected
);
2146 ovsrec_controller_set_role(cfg
, ofp12_controller_role_to_str(
2148 ovsrec_controller_set_status(cfg
, &smap
);
2149 smap_destroy(&smap
);
2151 ovsrec_controller_set_is_connected(cfg
, false);
2152 ovsrec_controller_set_role(cfg
, NULL
);
2153 ovsrec_controller_set_status(cfg
, NULL
);
2157 ofproto_free_ofproto_controller_info(&info
);
2162 * Some information in the database must be kept as up-to-date as possible to
2163 * allow controllers to respond rapidly to network outages. We call these
2164 * statistics "instant" stats.
2166 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2167 * assuming that they've changed. The only means we have to determine whether
2168 * they have changed are:
2170 * - Try to commit changes to the database. If nothing changed, then
2171 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2174 * - instant_stats_run() is called late in the run loop, after anything that
2175 * might change any of the instant stats.
2177 * We use these two facts together to avoid waking the process up every
2178 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2181 /* Minimum interval between writing updates to the instant stats to the
2183 #define INSTANT_INTERVAL_MSEC 100
2185 /* Current instant stats database transaction, NULL if there is no ongoing
2187 static struct ovsdb_idl_txn
*instant_txn
;
2189 /* Next time (in msec on monotonic clock) at which we will update the instant
2191 static long long int instant_next_txn
= LLONG_MIN
;
2193 /* True if the run loop has run since we last saw that the instant stats were
2194 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2195 * to refresh the instant stats. */
2196 static bool instant_stats_could_have_changed
;
2199 instant_stats_run(void)
2201 enum ovsdb_idl_txn_status status
;
2203 instant_stats_could_have_changed
= true;
2208 if (time_msec() < instant_next_txn
) {
2211 instant_next_txn
= time_msec() + INSTANT_INTERVAL_MSEC
;
2213 instant_txn
= ovsdb_idl_txn_create(idl
);
2214 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2215 struct iface
*iface
;
2218 br_refresh_stp_status(br
);
2220 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2221 port_refresh_stp_status(port
);
2224 HMAP_FOR_EACH (iface
, name_node
, &br
->iface_by_name
) {
2225 enum netdev_flags flags
;
2227 const char *link_state
;
2228 int64_t link_resets
;
2231 if (iface_is_synthetic(iface
)) {
2235 current
= ofproto_port_is_lacp_current(br
->ofproto
,
2239 ovsrec_interface_set_lacp_current(iface
->cfg
, &bl
, 1);
2241 ovsrec_interface_set_lacp_current(iface
->cfg
, NULL
, 0);
2244 error
= netdev_get_flags(iface
->netdev
, &flags
);
2246 const char *state
= flags
& NETDEV_UP
? "up" : "down";
2247 ovsrec_interface_set_admin_state(iface
->cfg
, state
);
2249 ovsrec_interface_set_admin_state(iface
->cfg
, NULL
);
2252 link_state
= netdev_get_carrier(iface
->netdev
) ? "up" : "down";
2253 ovsrec_interface_set_link_state(iface
->cfg
, link_state
);
2255 link_resets
= netdev_get_carrier_resets(iface
->netdev
);
2256 ovsrec_interface_set_link_resets(iface
->cfg
, &link_resets
, 1);
2258 iface_refresh_cfm_stats(iface
);
2261 ofproto_port_get_bfd_status(br
->ofproto
, iface
->ofp_port
,
2263 ovsrec_interface_set_bfd_status(iface
->cfg
, &smap
);
2264 smap_destroy(&smap
);
2269 status
= ovsdb_idl_txn_commit(instant_txn
);
2270 if (status
!= TXN_INCOMPLETE
) {
2271 ovsdb_idl_txn_destroy(instant_txn
);
2274 if (status
== TXN_UNCHANGED
) {
2275 instant_stats_could_have_changed
= false;
2280 instant_stats_wait(void)
2283 ovsdb_idl_txn_wait(instant_txn
);
2284 } else if (instant_stats_could_have_changed
) {
2285 poll_timer_wait_until(instant_next_txn
);
2289 /* Performs periodic activity required by bridges that needs to be done with
2290 * the least possible latency.
2292 * It makes sense to call this function a couple of times per poll loop, to
2293 * provide a significant performance boost on some benchmarks with ofprotos
2294 * that use the ofproto-dpif implementation. */
2296 bridge_run_fast(void)
2303 ofproto_enumerate_types(&types
);
2304 SSET_FOR_EACH (type
, &types
) {
2305 ofproto_type_run_fast(type
);
2307 sset_destroy(&types
);
2309 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2310 ofproto_run_fast(br
->ofproto
);
2317 static struct ovsrec_open_vswitch null_cfg
;
2318 const struct ovsrec_open_vswitch
*cfg
;
2319 struct ovsdb_idl_txn
*reconf_txn
= NULL
;
2323 bool vlan_splinters_changed
;
2326 ovsrec_open_vswitch_init(&null_cfg
);
2328 /* (Re)configure if necessary. */
2329 if (!reconfiguring
) {
2332 if (ovsdb_idl_is_lock_contended(idl
)) {
2333 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2334 struct bridge
*br
, *next_br
;
2336 VLOG_ERR_RL(&rl
, "another ovs-vswitchd process is running, "
2337 "disabling this process (pid %ld) until it goes away",
2338 (long int) getpid());
2340 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
2343 /* Since we will not be running system_stats_run() in this process
2344 * with the current situation of multiple ovs-vswitchd daemons,
2345 * disable system stats collection. */
2346 system_stats_enable(false);
2348 } else if (!ovsdb_idl_has_lock(idl
)) {
2352 cfg
= ovsrec_open_vswitch_first(idl
);
2354 /* Initialize the ofproto library. This only needs to run once, but
2355 * it must be done after the configuration is set. If the
2356 * initialization has already occurred, bridge_init_ofproto()
2357 * returns immediately. */
2358 bridge_init_ofproto(cfg
);
2360 /* Once the value of flow-restore-wait is false, we no longer should
2361 * check its value from the database. */
2362 if (cfg
&& ofproto_get_flow_restore_wait()) {
2363 ofproto_set_flow_restore_wait(smap_get_bool(&cfg
->other_config
,
2364 "flow-restore-wait", false));
2367 /* Let each datapath type do the work that it needs to do. */
2369 ofproto_enumerate_types(&types
);
2370 SSET_FOR_EACH (type
, &types
) {
2371 ofproto_type_run(type
);
2373 sset_destroy(&types
);
2375 /* Let each bridge do the work that it needs to do. */
2376 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2377 ofproto_run(br
->ofproto
);
2380 /* Re-configure SSL. We do this on every trip through the main loop,
2381 * instead of just when the database changes, because the contents of the
2382 * key and certificate files can change without the database changing.
2384 * We do this before bridge_reconfigure() because that function might
2385 * initiate SSL connections and thus requires SSL to be configured. */
2386 if (cfg
&& cfg
->ssl
) {
2387 const struct ovsrec_ssl
*ssl
= cfg
->ssl
;
2389 stream_ssl_set_key_and_cert(ssl
->private_key
, ssl
->certificate
);
2390 stream_ssl_set_ca_cert_file(ssl
->ca_cert
, ssl
->bootstrap_ca_cert
);
2393 if (!reconfiguring
) {
2394 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2395 * usage has changed. */
2396 vlan_splinters_changed
= false;
2397 if (vlan_splinters_enabled_anywhere
) {
2398 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2399 if (ofproto_has_vlan_usage_changed(br
->ofproto
)) {
2400 vlan_splinters_changed
= true;
2406 if (ovsdb_idl_get_seqno(idl
) != idl_seqno
|| vlan_splinters_changed
) {
2407 idl_seqno
= ovsdb_idl_get_seqno(idl
);
2409 reconf_txn
= ovsdb_idl_txn_create(idl
);
2410 bridge_reconfigure(cfg
);
2412 /* We still need to reconfigure to avoid dangling pointers to
2413 * now-destroyed ovsrec structures inside bridge data. */
2414 bridge_reconfigure(&null_cfg
);
2419 if (reconfiguring
) {
2421 reconf_txn
= ovsdb_idl_txn_create(idl
);
2424 if (bridge_reconfigure_continue(cfg
? cfg
: &null_cfg
)) {
2425 reconfiguring
= false;
2428 ovsrec_open_vswitch_set_cur_cfg(cfg
, cfg
->next_cfg
);
2431 /* If we are completing our initial configuration for this run
2432 * of ovs-vswitchd, then keep the transaction around to monitor
2433 * it for completion. */
2434 if (!initial_config_done
) {
2435 initial_config_done
= true;
2436 daemonize_txn
= reconf_txn
;
2443 ovsdb_idl_txn_commit(reconf_txn
);
2444 ovsdb_idl_txn_destroy(reconf_txn
);
2448 if (daemonize_txn
) {
2449 enum ovsdb_idl_txn_status status
= ovsdb_idl_txn_commit(daemonize_txn
);
2450 if (status
!= TXN_INCOMPLETE
) {
2451 ovsdb_idl_txn_destroy(daemonize_txn
);
2452 daemonize_txn
= NULL
;
2454 /* ovs-vswitchd has completed initialization, so allow the
2455 * process that forked us to exit successfully. */
2456 daemonize_complete();
2458 vlog_enable_async();
2460 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name
, VERSION
);
2464 /* Refresh interface and mirror stats if necessary. */
2465 if (time_msec() >= iface_stats_timer
) {
2467 struct ovsdb_idl_txn
*txn
;
2469 txn
= ovsdb_idl_txn_create(idl
);
2470 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2474 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
2475 struct iface
*iface
;
2477 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2478 iface_refresh_stats(iface
);
2479 iface_refresh_status(iface
);
2483 HMAP_FOR_EACH (m
, hmap_node
, &br
->mirrors
) {
2484 mirror_refresh_stats(m
);
2488 refresh_controller_status();
2489 ovsdb_idl_txn_commit(txn
);
2490 ovsdb_idl_txn_destroy(txn
); /* XXX */
2493 iface_stats_timer
= time_msec() + IFACE_STATS_INTERVAL
;
2497 instant_stats_run();
2506 ovsdb_idl_wait(idl
);
2507 if (daemonize_txn
) {
2508 ovsdb_idl_txn_wait(daemonize_txn
);
2511 if (reconfiguring
) {
2512 poll_immediate_wake();
2516 ofproto_enumerate_types(&types
);
2517 SSET_FOR_EACH (type
, &types
) {
2518 ofproto_type_wait(type
);
2520 sset_destroy(&types
);
2522 if (!hmap_is_empty(&all_bridges
)) {
2525 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2526 ofproto_wait(br
->ofproto
);
2528 poll_timer_wait_until(iface_stats_timer
);
2531 system_stats_wait();
2532 instant_stats_wait();
2535 /* Adds some memory usage statistics for bridges into 'usage', for use with
2536 * memory_report(). */
2538 bridge_get_memory_usage(struct simap
*usage
)
2542 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2543 ofproto_get_memory_usage(br
->ofproto
, usage
);
2547 /* QoS unixctl user interface functions. */
2549 struct qos_unixctl_show_cbdata
{
2551 struct iface
*iface
;
2555 qos_unixctl_show_queue(unsigned int queue_id
,
2556 const struct smap
*details
,
2557 struct iface
*iface
,
2560 struct netdev_queue_stats stats
;
2561 struct smap_node
*node
;
2564 ds_put_cstr(ds
, "\n");
2566 ds_put_format(ds
, "Queue %u:\n", queue_id
);
2568 ds_put_cstr(ds
, "Default:\n");
2571 SMAP_FOR_EACH (node
, details
) {
2572 ds_put_format(ds
, "\t%s: %s\n", node
->key
, node
->value
);
2575 error
= netdev_get_queue_stats(iface
->netdev
, queue_id
, &stats
);
2577 if (stats
.tx_packets
!= UINT64_MAX
) {
2578 ds_put_format(ds
, "\ttx_packets: %"PRIu64
"\n", stats
.tx_packets
);
2581 if (stats
.tx_bytes
!= UINT64_MAX
) {
2582 ds_put_format(ds
, "\ttx_bytes: %"PRIu64
"\n", stats
.tx_bytes
);
2585 if (stats
.tx_errors
!= UINT64_MAX
) {
2586 ds_put_format(ds
, "\ttx_errors: %"PRIu64
"\n", stats
.tx_errors
);
2589 ds_put_format(ds
, "\tFailed to get statistics for queue %u: %s",
2590 queue_id
, ovs_strerror(error
));
2595 qos_unixctl_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
2596 const char *argv
[], void *aux OVS_UNUSED
)
2598 struct ds ds
= DS_EMPTY_INITIALIZER
;
2599 struct smap smap
= SMAP_INITIALIZER(&smap
);
2600 struct iface
*iface
;
2602 struct smap_node
*node
;
2604 iface
= iface_find(argv
[1]);
2606 unixctl_command_reply_error(conn
, "no such interface");
2610 netdev_get_qos(iface
->netdev
, &type
, &smap
);
2612 if (*type
!= '\0') {
2613 struct netdev_queue_dump dump
;
2614 struct smap details
;
2615 unsigned int queue_id
;
2617 ds_put_format(&ds
, "QoS: %s %s\n", iface
->name
, type
);
2619 SMAP_FOR_EACH (node
, &smap
) {
2620 ds_put_format(&ds
, "%s: %s\n", node
->key
, node
->value
);
2623 smap_init(&details
);
2624 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
2625 qos_unixctl_show_queue(queue_id
, &details
, iface
, &ds
);
2627 smap_destroy(&details
);
2629 unixctl_command_reply(conn
, ds_cstr(&ds
));
2631 ds_put_format(&ds
, "QoS not configured on %s\n", iface
->name
);
2632 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
2635 smap_destroy(&smap
);
2639 /* Bridge reconfiguration functions. */
2641 bridge_create(const struct ovsrec_bridge
*br_cfg
)
2645 ovs_assert(!bridge_lookup(br_cfg
->name
));
2646 br
= xzalloc(sizeof *br
);
2648 br
->name
= xstrdup(br_cfg
->name
);
2649 br
->type
= xstrdup(ofproto_normalize_type(br_cfg
->datapath_type
));
2652 /* Derive the default Ethernet address from the bridge's UUID. This should
2653 * be unique and it will be stable between ovs-vswitchd runs. */
2654 memcpy(br
->default_ea
, &br_cfg
->header_
.uuid
, ETH_ADDR_LEN
);
2655 eth_addr_mark_random(br
->default_ea
);
2657 hmap_init(&br
->ports
);
2658 hmap_init(&br
->ifaces
);
2659 hmap_init(&br
->iface_by_name
);
2660 hmap_init(&br
->mirrors
);
2662 hmap_init(&br
->if_cfg_todo
);
2663 list_init(&br
->ofpp_garbage
);
2665 hmap_insert(&all_bridges
, &br
->node
, hash_string(br
->name
, 0));
2669 bridge_destroy(struct bridge
*br
)
2672 struct mirror
*mirror
, *next_mirror
;
2673 struct port
*port
, *next_port
;
2674 struct if_cfg
*if_cfg
, *next_if_cfg
;
2675 struct ofpp_garbage
*garbage
, *next_garbage
;
2677 HMAP_FOR_EACH_SAFE (port
, next_port
, hmap_node
, &br
->ports
) {
2680 HMAP_FOR_EACH_SAFE (mirror
, next_mirror
, hmap_node
, &br
->mirrors
) {
2681 mirror_destroy(mirror
);
2683 HMAP_FOR_EACH_SAFE (if_cfg
, next_if_cfg
, hmap_node
, &br
->if_cfg_todo
) {
2684 hmap_remove(&br
->if_cfg_todo
, &if_cfg
->hmap_node
);
2687 LIST_FOR_EACH_SAFE (garbage
, next_garbage
, list_node
,
2688 &br
->ofpp_garbage
) {
2689 list_remove(&garbage
->list_node
);
2693 hmap_remove(&all_bridges
, &br
->node
);
2694 ofproto_destroy(br
->ofproto
);
2695 hmap_destroy(&br
->ifaces
);
2696 hmap_destroy(&br
->ports
);
2697 hmap_destroy(&br
->iface_by_name
);
2698 hmap_destroy(&br
->mirrors
);
2699 hmap_destroy(&br
->if_cfg_todo
);
2706 static struct bridge
*
2707 bridge_lookup(const char *name
)
2711 HMAP_FOR_EACH_WITH_HASH (br
, node
, hash_string(name
, 0), &all_bridges
) {
2712 if (!strcmp(br
->name
, name
)) {
2719 /* Handle requests for a listing of all flows known by the OpenFlow
2720 * stack, including those normally hidden. */
2722 bridge_unixctl_dump_flows(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
2723 const char *argv
[], void *aux OVS_UNUSED
)
2728 br
= bridge_lookup(argv
[1]);
2730 unixctl_command_reply_error(conn
, "Unknown bridge");
2735 ofproto_get_all_flows(br
->ofproto
, &results
);
2737 unixctl_command_reply(conn
, ds_cstr(&results
));
2738 ds_destroy(&results
);
2741 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2742 * connections and reconnect. If BRIDGE is not specified, then all bridges
2743 * drop their controller connections and reconnect. */
2745 bridge_unixctl_reconnect(struct unixctl_conn
*conn
, int argc
,
2746 const char *argv
[], void *aux OVS_UNUSED
)
2750 br
= bridge_lookup(argv
[1]);
2752 unixctl_command_reply_error(conn
, "Unknown bridge");
2755 ofproto_reconnect_controllers(br
->ofproto
);
2757 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2758 ofproto_reconnect_controllers(br
->ofproto
);
2761 unixctl_command_reply(conn
, NULL
);
2765 bridge_get_controllers(const struct bridge
*br
,
2766 struct ovsrec_controller
***controllersp
)
2768 struct ovsrec_controller
**controllers
;
2769 size_t n_controllers
;
2771 controllers
= br
->cfg
->controller
;
2772 n_controllers
= br
->cfg
->n_controller
;
2774 if (n_controllers
== 1 && !strcmp(controllers
[0]->target
, "none")) {
2780 *controllersp
= controllers
;
2782 return n_controllers
;
2786 bridge_queue_if_cfg(struct bridge
*br
,
2787 const struct ovsrec_interface
*cfg
,
2788 const struct ovsrec_port
*parent
)
2790 struct if_cfg
*if_cfg
= xmalloc(sizeof *if_cfg
);
2793 if_cfg
->parent
= parent
;
2794 if_cfg
->ofport
= iface_pick_ofport(cfg
);
2795 hmap_insert(&br
->if_cfg_todo
, &if_cfg
->hmap_node
,
2796 hash_string(if_cfg
->cfg
->name
, 0));
2799 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2800 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2801 * 'br' needs to complete its configuration. */
2803 bridge_add_del_ports(struct bridge
*br
,
2804 const unsigned long int *splinter_vlans
)
2806 struct shash_node
*port_node
;
2807 struct port
*port
, *next
;
2808 struct shash new_ports
;
2811 ovs_assert(hmap_is_empty(&br
->if_cfg_todo
));
2813 /* Collect new ports. */
2814 shash_init(&new_ports
);
2815 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
2816 const char *name
= br
->cfg
->ports
[i
]->name
;
2817 if (!shash_add_once(&new_ports
, name
, br
->cfg
->ports
[i
])) {
2818 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2822 if (bridge_get_controllers(br
, NULL
)
2823 && !shash_find(&new_ports
, br
->name
)) {
2824 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2825 br
->name
, br
->name
);
2827 ovsrec_interface_init(&br
->synth_local_iface
);
2828 ovsrec_port_init(&br
->synth_local_port
);
2830 br
->synth_local_port
.interfaces
= &br
->synth_local_ifacep
;
2831 br
->synth_local_port
.n_interfaces
= 1;
2832 br
->synth_local_port
.name
= br
->name
;
2834 br
->synth_local_iface
.name
= br
->name
;
2835 br
->synth_local_iface
.type
= "internal";
2837 br
->synth_local_ifacep
= &br
->synth_local_iface
;
2839 shash_add(&new_ports
, br
->name
, &br
->synth_local_port
);
2842 if (splinter_vlans
) {
2843 add_vlan_splinter_ports(br
, splinter_vlans
, &new_ports
);
2846 /* Get rid of deleted ports.
2847 * Get rid of deleted interfaces on ports that still exist. */
2848 HMAP_FOR_EACH_SAFE (port
, next
, hmap_node
, &br
->ports
) {
2849 port
->cfg
= shash_find_data(&new_ports
, port
->name
);
2853 port_del_ifaces(port
);
2857 /* Update iface->cfg and iface->type in interfaces that still exist.
2858 * Add new interfaces to creation queue. */
2859 SHASH_FOR_EACH (port_node
, &new_ports
) {
2860 const struct ovsrec_port
*port
= port_node
->data
;
2863 for (i
= 0; i
< port
->n_interfaces
; i
++) {
2864 const struct ovsrec_interface
*cfg
= port
->interfaces
[i
];
2865 struct iface
*iface
= iface_lookup(br
, cfg
->name
);
2866 const char *type
= iface_get_type(cfg
, br
->cfg
);
2871 } else if (!strcmp(type
, "null")) {
2872 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2873 " may be removed in February 2013. Please email"
2874 " dev@openvswitch.org with concerns.",
2877 bridge_queue_if_cfg(br
, cfg
, port
);
2882 shash_destroy(&new_ports
);
2885 /* Initializes 'oc' appropriately as a management service controller for
2888 * The caller must free oc->target when it is no longer needed. */
2890 bridge_ofproto_controller_for_mgmt(const struct bridge
*br
,
2891 struct ofproto_controller
*oc
)
2893 oc
->target
= xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br
->name
);
2894 oc
->max_backoff
= 0;
2895 oc
->probe_interval
= 60;
2896 oc
->band
= OFPROTO_OUT_OF_BAND
;
2898 oc
->burst_limit
= 0;
2899 oc
->enable_async_msgs
= true;
2902 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2904 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller
*c
,
2905 struct ofproto_controller
*oc
)
2909 oc
->target
= c
->target
;
2910 oc
->max_backoff
= c
->max_backoff
? *c
->max_backoff
/ 1000 : 8;
2911 oc
->probe_interval
= c
->inactivity_probe
? *c
->inactivity_probe
/ 1000 : 5;
2912 oc
->band
= (!c
->connection_mode
|| !strcmp(c
->connection_mode
, "in-band")
2913 ? OFPROTO_IN_BAND
: OFPROTO_OUT_OF_BAND
);
2914 oc
->rate_limit
= c
->controller_rate_limit
? *c
->controller_rate_limit
: 0;
2915 oc
->burst_limit
= (c
->controller_burst_limit
2916 ? *c
->controller_burst_limit
: 0);
2917 oc
->enable_async_msgs
= (!c
->enable_async_messages
2918 || *c
->enable_async_messages
);
2919 dscp
= smap_get_int(&c
->other_config
, "dscp", DSCP_DEFAULT
);
2920 if (dscp
< 0 || dscp
> 63) {
2921 dscp
= DSCP_DEFAULT
;
2926 /* Configures the IP stack for 'br''s local interface properly according to the
2927 * configuration in 'c'. */
2929 bridge_configure_local_iface_netdev(struct bridge
*br
,
2930 struct ovsrec_controller
*c
)
2932 struct netdev
*netdev
;
2933 struct in_addr mask
, gateway
;
2935 struct iface
*local_iface
;
2938 /* If there's no local interface or no IP address, give up. */
2939 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
2940 if (!local_iface
|| !c
->local_ip
|| !inet_aton(c
->local_ip
, &ip
)) {
2944 /* Bring up the local interface. */
2945 netdev
= local_iface
->netdev
;
2946 netdev_turn_flags_on(netdev
, NETDEV_UP
, NULL
);
2948 /* Configure the IP address and netmask. */
2949 if (!c
->local_netmask
2950 || !inet_aton(c
->local_netmask
, &mask
)
2952 mask
.s_addr
= guess_netmask(ip
.s_addr
);
2954 if (!netdev_set_in4(netdev
, ip
, mask
)) {
2955 VLOG_INFO("bridge %s: configured IP address "IP_FMT
", netmask "IP_FMT
,
2956 br
->name
, IP_ARGS(ip
.s_addr
), IP_ARGS(mask
.s_addr
));
2959 /* Configure the default gateway. */
2960 if (c
->local_gateway
2961 && inet_aton(c
->local_gateway
, &gateway
)
2962 && gateway
.s_addr
) {
2963 if (!netdev_add_router(netdev
, gateway
)) {
2964 VLOG_INFO("bridge %s: configured gateway "IP_FMT
,
2965 br
->name
, IP_ARGS(gateway
.s_addr
));
2970 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2971 * in either string are treated as equal to any number of slashes in the other,
2972 * e.g. "x///y" is equal to "x/y".
2974 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2975 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2976 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2977 * 'b' against a prefix of 'a'.
2980 equal_pathnames(const char *a
, const char *b
, size_t b_stoplen
)
2982 const char *b_start
= b
;
2984 if (b
- b_start
>= b_stoplen
) {
2986 } else if (*a
!= *b
) {
2988 } else if (*a
== '/') {
2989 a
+= strspn(a
, "/");
2990 b
+= strspn(b
, "/");
2991 } else if (*a
== '\0') {
3001 bridge_configure_remotes(struct bridge
*br
,
3002 const struct sockaddr_in
*managers
, size_t n_managers
)
3004 bool disable_in_band
;
3006 struct ovsrec_controller
**controllers
;
3007 size_t n_controllers
;
3009 enum ofproto_fail_mode fail_mode
;
3011 struct ofproto_controller
*ocs
;
3015 /* Check if we should disable in-band control on this bridge. */
3016 disable_in_band
= smap_get_bool(&br
->cfg
->other_config
, "disable-in-band",
3019 /* Set OpenFlow queue ID for in-band control. */
3020 ofproto_set_in_band_queue(br
->ofproto
,
3021 smap_get_int(&br
->cfg
->other_config
,
3022 "in-band-queue", -1));
3024 if (disable_in_band
) {
3025 ofproto_set_extra_in_band_remotes(br
->ofproto
, NULL
, 0);
3027 ofproto_set_extra_in_band_remotes(br
->ofproto
, managers
, n_managers
);
3030 n_controllers
= bridge_get_controllers(br
, &controllers
);
3032 ocs
= xmalloc((n_controllers
+ 1) * sizeof *ocs
);
3035 bridge_ofproto_controller_for_mgmt(br
, &ocs
[n_ocs
++]);
3036 for (i
= 0; i
< n_controllers
; i
++) {
3037 struct ovsrec_controller
*c
= controllers
[i
];
3039 if (!strncmp(c
->target
, "punix:", 6)
3040 || !strncmp(c
->target
, "unix:", 5)) {
3041 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3044 if (!strncmp(c
->target
, "unix:", 5)) {
3045 /* Connect to a listening socket */
3046 whitelist
= xasprintf("unix:%s/", ovs_rundir());
3047 if (strchr(c
->target
, '/') &&
3048 !equal_pathnames(c
->target
, whitelist
,
3049 strlen(whitelist
))) {
3050 /* Absolute path specified, but not in ovs_rundir */
3051 VLOG_ERR_RL(&rl
, "bridge %s: Not connecting to socket "
3052 "controller \"%s\" due to possibility for "
3053 "remote exploit. Instead, specify socket "
3054 "in whitelisted \"%s\" or connect to "
3055 "\"unix:%s/%s.mgmt\" (which is always "
3056 "available without special configuration).",
3057 br
->name
, c
->target
, whitelist
,
3058 ovs_rundir(), br
->name
);
3063 whitelist
= xasprintf("punix:%s/%s.controller",
3064 ovs_rundir(), br
->name
);
3065 if (!equal_pathnames(c
->target
, whitelist
, SIZE_MAX
)) {
3066 /* Prevent remote ovsdb-server users from accessing
3067 * arbitrary Unix domain sockets and overwriting arbitrary
3069 VLOG_ERR_RL(&rl
, "bridge %s: Not adding Unix domain socket "
3070 "controller \"%s\" due to possibility of "
3071 "overwriting local files. Instead, specify "
3072 "whitelisted \"%s\" or connect to "
3073 "\"unix:%s/%s.mgmt\" (which is always "
3074 "available without special configuration).",
3075 br
->name
, c
->target
, whitelist
,
3076 ovs_rundir(), br
->name
);
3085 bridge_configure_local_iface_netdev(br
, c
);
3086 bridge_ofproto_controller_from_ovsrec(c
, &ocs
[n_ocs
]);
3087 if (disable_in_band
) {
3088 ocs
[n_ocs
].band
= OFPROTO_OUT_OF_BAND
;
3093 ofproto_set_controllers(br
->ofproto
, ocs
, n_ocs
,
3094 bridge_get_allowed_versions(br
));
3095 free(ocs
[0].target
); /* From bridge_ofproto_controller_for_mgmt(). */
3098 /* Set the fail-mode. */
3099 fail_mode
= !br
->cfg
->fail_mode
3100 || !strcmp(br
->cfg
->fail_mode
, "standalone")
3101 ? OFPROTO_FAIL_STANDALONE
3102 : OFPROTO_FAIL_SECURE
;
3103 ofproto_set_fail_mode(br
->ofproto
, fail_mode
);
3105 /* Configure OpenFlow controller connection snooping. */
3106 if (!ofproto_has_snoops(br
->ofproto
)) {
3110 sset_add_and_free(&snoops
, xasprintf("punix:%s/%s.snoop",
3111 ovs_rundir(), br
->name
));
3112 ofproto_set_snoops(br
->ofproto
, &snoops
);
3113 sset_destroy(&snoops
);
3118 bridge_configure_tables(struct bridge
*br
)
3120 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3124 n_tables
= ofproto_get_n_tables(br
->ofproto
);
3126 for (i
= 0; i
< n_tables
; i
++) {
3127 struct ofproto_table_settings s
;
3130 s
.max_flows
= UINT_MAX
;
3134 if (j
< br
->cfg
->n_flow_tables
&& i
== br
->cfg
->key_flow_tables
[j
]) {
3135 struct ovsrec_flow_table
*cfg
= br
->cfg
->value_flow_tables
[j
++];
3138 if (cfg
->n_flow_limit
&& *cfg
->flow_limit
< UINT_MAX
) {
3139 s
.max_flows
= *cfg
->flow_limit
;
3141 if (cfg
->overflow_policy
3142 && !strcmp(cfg
->overflow_policy
, "evict")) {
3145 s
.groups
= xmalloc(cfg
->n_groups
* sizeof *s
.groups
);
3146 for (k
= 0; k
< cfg
->n_groups
; k
++) {
3147 const char *string
= cfg
->groups
[k
];
3150 msg
= mf_parse_subfield__(&s
.groups
[k
], &string
);
3152 VLOG_WARN_RL(&rl
, "bridge %s table %d: error parsing "
3153 "'groups' (%s)", br
->name
, i
, msg
);
3155 } else if (*string
) {
3156 VLOG_WARN_RL(&rl
, "bridge %s table %d: 'groups' "
3157 "element '%s' contains trailing garbage",
3158 br
->name
, i
, cfg
->groups
[k
]);
3166 ofproto_configure_table(br
->ofproto
, i
, &s
);
3170 for (; j
< br
->cfg
->n_flow_tables
; j
++) {
3171 VLOG_WARN_RL(&rl
, "bridge %s: ignoring configuration for flow table "
3172 "%"PRId64
" not supported by this datapath", br
->name
,
3173 br
->cfg
->key_flow_tables
[j
]);
3178 bridge_configure_dp_desc(struct bridge
*br
)
3180 ofproto_set_dp_desc(br
->ofproto
,
3181 smap_get(&br
->cfg
->other_config
, "dp-desc"));
3184 /* Port functions. */
3186 static struct port
*
3187 port_create(struct bridge
*br
, const struct ovsrec_port
*cfg
)
3191 port
= xzalloc(sizeof *port
);
3193 port
->name
= xstrdup(cfg
->name
);
3195 list_init(&port
->ifaces
);
3197 hmap_insert(&br
->ports
, &port
->hmap_node
, hash_string(port
->name
, 0));
3201 /* Deletes interfaces from 'port' that are no longer configured for it. */
3203 port_del_ifaces(struct port
*port
)
3205 struct iface
*iface
, *next
;
3206 struct sset new_ifaces
;
3209 /* Collect list of new interfaces. */
3210 sset_init(&new_ifaces
);
3211 for (i
= 0; i
< port
->cfg
->n_interfaces
; i
++) {
3212 const char *name
= port
->cfg
->interfaces
[i
]->name
;
3213 const char *type
= port
->cfg
->interfaces
[i
]->type
;
3214 if (strcmp(type
, "null")) {
3215 sset_add(&new_ifaces
, name
);
3219 /* Get rid of deleted interfaces. */
3220 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
3221 if (!sset_contains(&new_ifaces
, iface
->name
)) {
3222 iface_destroy(iface
);
3226 sset_destroy(&new_ifaces
);
3230 port_destroy(struct port
*port
)
3233 struct bridge
*br
= port
->bridge
;
3234 struct iface
*iface
, *next
;
3237 ofproto_bundle_unregister(br
->ofproto
, port
);
3240 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
3241 iface_destroy(iface
);
3244 hmap_remove(&br
->ports
, &port
->hmap_node
);
3250 static struct port
*
3251 port_lookup(const struct bridge
*br
, const char *name
)
3255 HMAP_FOR_EACH_WITH_HASH (port
, hmap_node
, hash_string(name
, 0),
3257 if (!strcmp(port
->name
, name
)) {
3265 enable_lacp(struct port
*port
, bool *activep
)
3267 if (!port
->cfg
->lacp
) {
3268 /* XXX when LACP implementation has been sufficiently tested, enable by
3269 * default and make active on bonded ports. */
3271 } else if (!strcmp(port
->cfg
->lacp
, "off")) {
3273 } else if (!strcmp(port
->cfg
->lacp
, "active")) {
3276 } else if (!strcmp(port
->cfg
->lacp
, "passive")) {
3280 VLOG_WARN("port %s: unknown LACP mode %s",
3281 port
->name
, port
->cfg
->lacp
);
3286 static struct lacp_settings
*
3287 port_configure_lacp(struct port
*port
, struct lacp_settings
*s
)
3289 const char *lacp_time
, *system_id
;
3292 if (!enable_lacp(port
, &s
->active
)) {
3296 s
->name
= port
->name
;
3298 system_id
= smap_get(&port
->cfg
->other_config
, "lacp-system-id");
3300 if (sscanf(system_id
, ETH_ADDR_SCAN_FMT
,
3301 ETH_ADDR_SCAN_ARGS(s
->id
)) != ETH_ADDR_SCAN_COUNT
) {
3302 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3303 " address.", port
->name
, system_id
);
3307 memcpy(s
->id
, port
->bridge
->ea
, ETH_ADDR_LEN
);
3310 if (eth_addr_is_zero(s
->id
)) {
3311 VLOG_WARN("port %s: Invalid zero LACP system ID.", port
->name
);
3315 /* Prefer bondable links if unspecified. */
3316 priority
= smap_get_int(&port
->cfg
->other_config
, "lacp-system-priority",
3318 s
->priority
= (priority
> 0 && priority
<= UINT16_MAX
3320 : UINT16_MAX
- !list_is_short(&port
->ifaces
));
3322 lacp_time
= smap_get(&port
->cfg
->other_config
, "lacp-time");
3323 s
->fast
= lacp_time
&& !strcasecmp(lacp_time
, "fast");
3328 iface_configure_lacp(struct iface
*iface
, struct lacp_slave_settings
*s
)
3330 int priority
, portid
, key
;
3332 portid
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-id", 0);
3333 priority
= smap_get_int(&iface
->cfg
->other_config
, "lacp-port-priority",
3335 key
= smap_get_int(&iface
->cfg
->other_config
, "lacp-aggregation-key", 0);
3337 if (portid
<= 0 || portid
> UINT16_MAX
) {
3338 portid
= ofp_to_u16(iface
->ofp_port
);
3341 if (priority
<= 0 || priority
> UINT16_MAX
) {
3342 priority
= UINT16_MAX
;
3345 if (key
< 0 || key
> UINT16_MAX
) {
3349 s
->name
= iface
->name
;
3351 s
->priority
= priority
;
3356 port_configure_bond(struct port
*port
, struct bond_settings
*s
)
3358 const char *detect_s
;
3359 struct iface
*iface
;
3360 int miimon_interval
;
3362 s
->name
= port
->name
;
3364 if (port
->cfg
->bond_mode
) {
3365 if (!bond_mode_from_string(&s
->balance
, port
->cfg
->bond_mode
)) {
3366 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3367 port
->name
, port
->cfg
->bond_mode
,
3368 bond_mode_to_string(s
->balance
));
3371 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3373 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3374 * active-backup. At some point we should remove this warning. */
3375 VLOG_WARN_RL(&rl
, "port %s: Using the default bond_mode %s. Note that"
3376 " in previous versions, the default bond_mode was"
3377 " balance-slb", port
->name
,
3378 bond_mode_to_string(s
->balance
));
3380 if (s
->balance
== BM_SLB
&& port
->bridge
->cfg
->n_flood_vlans
) {
3381 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3382 "please use another bond type or disable flood_vlans",
3386 miimon_interval
= smap_get_int(&port
->cfg
->other_config
,
3387 "bond-miimon-interval", 0);
3388 if (miimon_interval
<= 0) {
3389 miimon_interval
= 200;
3392 detect_s
= smap_get(&port
->cfg
->other_config
, "bond-detect-mode");
3393 if (!detect_s
|| !strcmp(detect_s
, "carrier")) {
3394 miimon_interval
= 0;
3395 } else if (strcmp(detect_s
, "miimon")) {
3396 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3397 "defaulting to carrier", port
->name
, detect_s
);
3398 miimon_interval
= 0;
3401 s
->up_delay
= MAX(0, port
->cfg
->bond_updelay
);
3402 s
->down_delay
= MAX(0, port
->cfg
->bond_downdelay
);
3403 s
->basis
= smap_get_int(&port
->cfg
->other_config
, "bond-hash-basis", 0);
3404 s
->rebalance_interval
= smap_get_int(&port
->cfg
->other_config
,
3405 "bond-rebalance-interval", 10000);
3406 if (s
->rebalance_interval
&& s
->rebalance_interval
< 1000) {
3407 s
->rebalance_interval
= 1000;
3410 s
->fake_iface
= port
->cfg
->bond_fake_iface
;
3412 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
3413 netdev_set_miimon_interval(iface
->netdev
, miimon_interval
);
3417 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3418 * instead of obtaining it from the database. */
3420 port_is_synthetic(const struct port
*port
)
3422 return ovsdb_idl_row_is_synthetic(&port
->cfg
->header_
);
3425 /* Interface functions. */
3428 iface_is_internal(const struct ovsrec_interface
*iface
,
3429 const struct ovsrec_bridge
*br
)
3431 /* The local port and "internal" ports are always "internal". */
3432 return !strcmp(iface
->type
, "internal") || !strcmp(iface
->name
, br
->name
);
3435 /* Returns the correct network device type for interface 'iface' in bridge
3438 iface_get_type(const struct ovsrec_interface
*iface
,
3439 const struct ovsrec_bridge
*br
)
3443 /* The local port always has type "internal". Other ports take
3444 * their type from the database and default to "system" if none is
3446 if (iface_is_internal(iface
, br
)) {
3449 type
= iface
->type
[0] ? iface
->type
: "system";
3452 return ofproto_port_open_type(br
->datapath_type
, type
);
3456 iface_destroy(struct iface
*iface
)
3459 struct port
*port
= iface
->port
;
3460 struct bridge
*br
= port
->bridge
;
3462 if (br
->ofproto
&& iface
->ofp_port
!= OFPP_NONE
) {
3463 ofproto_port_unregister(br
->ofproto
, iface
->ofp_port
);
3466 if (iface
->ofp_port
!= OFPP_NONE
) {
3467 hmap_remove(&br
->ifaces
, &iface
->ofp_port_node
);
3470 list_remove(&iface
->port_elem
);
3471 hmap_remove(&br
->iface_by_name
, &iface
->name_node
);
3473 netdev_close(iface
->netdev
);
3480 static struct iface
*
3481 iface_lookup(const struct bridge
*br
, const char *name
)
3483 struct iface
*iface
;
3485 HMAP_FOR_EACH_WITH_HASH (iface
, name_node
, hash_string(name
, 0),
3486 &br
->iface_by_name
) {
3487 if (!strcmp(iface
->name
, name
)) {
3495 static struct iface
*
3496 iface_find(const char *name
)
3498 const struct bridge
*br
;
3500 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
3501 struct iface
*iface
= iface_lookup(br
, name
);
3510 static struct if_cfg
*
3511 if_cfg_lookup(const struct bridge
*br
, const char *name
)
3513 struct if_cfg
*if_cfg
;
3515 HMAP_FOR_EACH_WITH_HASH (if_cfg
, hmap_node
, hash_string(name
, 0),
3517 if (!strcmp(if_cfg
->cfg
->name
, name
)) {
3525 static struct iface
*
3526 iface_from_ofp_port(const struct bridge
*br
, ofp_port_t ofp_port
)
3528 struct iface
*iface
;
3530 HMAP_FOR_EACH_IN_BUCKET (iface
, ofp_port_node
, hash_ofp_port(ofp_port
),
3532 if (iface
->ofp_port
== ofp_port
) {
3539 /* Set Ethernet address of 'iface', if one is specified in the configuration
3542 iface_set_mac(struct iface
*iface
)
3544 uint8_t ea
[ETH_ADDR_LEN
];
3546 if (!strcmp(iface
->type
, "internal")
3547 && iface
->cfg
->mac
&& eth_addr_from_string(iface
->cfg
->mac
, ea
)) {
3548 if (iface
->ofp_port
== OFPP_LOCAL
) {
3549 VLOG_ERR("interface %s: ignoring mac in Interface record "
3550 "(use Bridge record to set local port's mac)",
3552 } else if (eth_addr_is_multicast(ea
)) {
3553 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3556 int error
= netdev_set_etheraddr(iface
->netdev
, ea
);
3558 VLOG_ERR("interface %s: setting MAC failed (%s)",
3559 iface
->name
, ovs_strerror(error
));
3565 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3567 iface_set_ofport(const struct ovsrec_interface
*if_cfg
, ofp_port_t ofport
)
3570 port_
= (ofport
== OFPP_NONE
) ? -1 : ofp_to_u16(ofport
);
3571 if (if_cfg
&& !ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
3572 ovsrec_interface_set_ofport(if_cfg
, &port_
, 1);
3576 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3577 * sets the "ofport" field to -1.
3579 * This is appropriate when 'if_cfg''s interface cannot be created or is
3580 * otherwise invalid. */
3582 iface_clear_db_record(const struct ovsrec_interface
*if_cfg
)
3584 if (!ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
3585 ovsrec_interface_set_status(if_cfg
, NULL
);
3586 ovsrec_interface_set_admin_state(if_cfg
, NULL
);
3587 ovsrec_interface_set_duplex(if_cfg
, NULL
);
3588 ovsrec_interface_set_link_speed(if_cfg
, NULL
, 0);
3589 ovsrec_interface_set_link_state(if_cfg
, NULL
);
3590 ovsrec_interface_set_mac_in_use(if_cfg
, NULL
);
3591 ovsrec_interface_set_mtu(if_cfg
, NULL
, 0);
3592 ovsrec_interface_set_cfm_fault(if_cfg
, NULL
, 0);
3593 ovsrec_interface_set_cfm_fault_status(if_cfg
, NULL
, 0);
3594 ovsrec_interface_set_cfm_remote_mpids(if_cfg
, NULL
, 0);
3595 ovsrec_interface_set_lacp_current(if_cfg
, NULL
, 0);
3596 ovsrec_interface_set_statistics(if_cfg
, NULL
, NULL
, 0);
3597 ovsrec_interface_set_ifindex(if_cfg
, NULL
, 0);
3602 queue_ids_include(const struct ovsdb_datum
*queues
, int64_t target
)
3604 union ovsdb_atom atom
;
3606 atom
.integer
= target
;
3607 return ovsdb_datum_find_key(queues
, &atom
, OVSDB_TYPE_INTEGER
) != UINT_MAX
;
3611 iface_configure_qos(struct iface
*iface
, const struct ovsrec_qos
*qos
)
3613 struct ofpbuf queues_buf
;
3615 ofpbuf_init(&queues_buf
, 0);
3617 if (!qos
|| qos
->type
[0] == '\0' || qos
->n_queues
< 1) {
3618 netdev_set_qos(iface
->netdev
, NULL
, NULL
);
3620 const struct ovsdb_datum
*queues
;
3621 struct netdev_queue_dump dump
;
3622 unsigned int queue_id
;
3623 struct smap details
;
3627 /* Configure top-level Qos for 'iface'. */
3628 netdev_set_qos(iface
->netdev
, qos
->type
, &qos
->other_config
);
3630 /* Deconfigure queues that were deleted. */
3631 queues
= ovsrec_qos_get_queues(qos
, OVSDB_TYPE_INTEGER
,
3633 smap_init(&details
);
3634 NETDEV_QUEUE_FOR_EACH (&queue_id
, &details
, &dump
, iface
->netdev
) {
3635 if (!queue_ids_include(queues
, queue_id
)) {
3636 netdev_delete_queue(iface
->netdev
, queue_id
);
3639 smap_destroy(&details
);
3641 /* Configure queues for 'iface'. */
3643 for (i
= 0; i
< qos
->n_queues
; i
++) {
3644 const struct ovsrec_queue
*queue
= qos
->value_queues
[i
];
3645 unsigned int queue_id
= qos
->key_queues
[i
];
3647 if (queue_id
== 0) {
3651 if (queue
->n_dscp
== 1) {
3652 struct ofproto_port_queue
*port_queue
;
3654 port_queue
= ofpbuf_put_uninit(&queues_buf
,
3655 sizeof *port_queue
);
3656 port_queue
->queue
= queue_id
;
3657 port_queue
->dscp
= queue
->dscp
[0];
3660 netdev_set_queue(iface
->netdev
, queue_id
, &queue
->other_config
);
3663 struct smap details
;
3665 smap_init(&details
);
3666 netdev_set_queue(iface
->netdev
, 0, &details
);
3667 smap_destroy(&details
);
3671 if (iface
->ofp_port
!= OFPP_NONE
) {
3672 const struct ofproto_port_queue
*port_queues
= queues_buf
.data
;
3673 size_t n_queues
= queues_buf
.size
/ sizeof *port_queues
;
3675 ofproto_port_set_queues(iface
->port
->bridge
->ofproto
, iface
->ofp_port
,
3676 port_queues
, n_queues
);
3679 netdev_set_policing(iface
->netdev
,
3680 iface
->cfg
->ingress_policing_rate
,
3681 iface
->cfg
->ingress_policing_burst
);
3683 ofpbuf_uninit(&queues_buf
);
3687 iface_configure_cfm(struct iface
*iface
)
3689 const struct ovsrec_interface
*cfg
= iface
->cfg
;
3690 const char *opstate_str
;
3691 const char *cfm_ccm_vlan
;
3692 struct cfm_settings s
;
3693 struct smap netdev_args
;
3695 if (!cfg
->n_cfm_mpid
) {
3696 ofproto_port_clear_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
);
3700 s
.check_tnl_key
= false;
3701 smap_init(&netdev_args
);
3702 if (!netdev_get_config(iface
->netdev
, &netdev_args
)) {
3703 const char *key
= smap_get(&netdev_args
, "key");
3704 const char *in_key
= smap_get(&netdev_args
, "in_key");
3706 s
.check_tnl_key
= (key
&& !strcmp(key
, "flow"))
3707 || (in_key
&& !strcmp(in_key
, "flow"));
3709 smap_destroy(&netdev_args
);
3711 s
.mpid
= *cfg
->cfm_mpid
;
3712 s
.interval
= smap_get_int(&iface
->cfg
->other_config
, "cfm_interval", 0);
3713 cfm_ccm_vlan
= smap_get(&iface
->cfg
->other_config
, "cfm_ccm_vlan");
3714 s
.ccm_pcp
= smap_get_int(&iface
->cfg
->other_config
, "cfm_ccm_pcp", 0);
3716 if (s
.interval
<= 0) {
3720 if (!cfm_ccm_vlan
) {
3722 } else if (!strcasecmp("random", cfm_ccm_vlan
)) {
3723 s
.ccm_vlan
= CFM_RANDOM_VLAN
;
3725 s
.ccm_vlan
= atoi(cfm_ccm_vlan
);
3726 if (s
.ccm_vlan
== CFM_RANDOM_VLAN
) {
3731 s
.extended
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_extended",
3733 s
.demand
= smap_get_bool(&iface
->cfg
->other_config
, "cfm_demand", false);
3735 opstate_str
= smap_get(&iface
->cfg
->other_config
, "cfm_opstate");
3736 s
.opup
= !opstate_str
|| !strcasecmp("up", opstate_str
);
3738 ofproto_port_set_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
, &s
);
3741 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3742 * instead of obtaining it from the database. */
3744 iface_is_synthetic(const struct iface
*iface
)
3746 return ovsdb_idl_row_is_synthetic(&iface
->cfg
->header_
);
3750 iface_pick_ofport(const struct ovsrec_interface
*cfg
)
3752 ofp_port_t ofport
= cfg
->n_ofport
? u16_to_ofp(*cfg
->ofport
)
3754 return cfg
->n_ofport_request
? u16_to_ofp(*cfg
->ofport_request
)
3759 /* Port mirroring. */
3761 static struct mirror
*
3762 mirror_find_by_uuid(struct bridge
*br
, const struct uuid
*uuid
)
3766 HMAP_FOR_EACH_IN_BUCKET (m
, hmap_node
, uuid_hash(uuid
), &br
->mirrors
) {
3767 if (uuid_equals(uuid
, &m
->uuid
)) {
3775 bridge_configure_mirrors(struct bridge
*br
)
3777 const struct ovsdb_datum
*mc
;
3778 unsigned long *flood_vlans
;
3779 struct mirror
*m
, *next
;
3782 /* Get rid of deleted mirrors. */
3783 mc
= ovsrec_bridge_get_mirrors(br
->cfg
, OVSDB_TYPE_UUID
);
3784 HMAP_FOR_EACH_SAFE (m
, next
, hmap_node
, &br
->mirrors
) {
3785 union ovsdb_atom atom
;
3787 atom
.uuid
= m
->uuid
;
3788 if (ovsdb_datum_find_key(mc
, &atom
, OVSDB_TYPE_UUID
) == UINT_MAX
) {
3793 /* Add new mirrors and reconfigure existing ones. */
3794 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
3795 const struct ovsrec_mirror
*cfg
= br
->cfg
->mirrors
[i
];
3796 struct mirror
*m
= mirror_find_by_uuid(br
, &cfg
->header_
.uuid
);
3798 m
= mirror_create(br
, cfg
);
3801 if (!mirror_configure(m
)) {
3806 /* Update flooded vlans (for RSPAN). */
3807 flood_vlans
= vlan_bitmap_from_array(br
->cfg
->flood_vlans
,
3808 br
->cfg
->n_flood_vlans
);
3809 ofproto_set_flood_vlans(br
->ofproto
, flood_vlans
);
3810 bitmap_free(flood_vlans
);
3813 static struct mirror
*
3814 mirror_create(struct bridge
*br
, const struct ovsrec_mirror
*cfg
)
3818 m
= xzalloc(sizeof *m
);
3819 m
->uuid
= cfg
->header_
.uuid
;
3820 hmap_insert(&br
->mirrors
, &m
->hmap_node
, uuid_hash(&m
->uuid
));
3822 m
->name
= xstrdup(cfg
->name
);
3828 mirror_destroy(struct mirror
*m
)
3831 struct bridge
*br
= m
->bridge
;
3834 ofproto_mirror_unregister(br
->ofproto
, m
);
3837 hmap_remove(&br
->mirrors
, &m
->hmap_node
);
3844 mirror_collect_ports(struct mirror
*m
,
3845 struct ovsrec_port
**in_ports
, int n_in_ports
,
3846 void ***out_portsp
, size_t *n_out_portsp
)
3848 void **out_ports
= xmalloc(n_in_ports
* sizeof *out_ports
);
3849 size_t n_out_ports
= 0;
3852 for (i
= 0; i
< n_in_ports
; i
++) {
3853 const char *name
= in_ports
[i
]->name
;
3854 struct port
*port
= port_lookup(m
->bridge
, name
);
3856 out_ports
[n_out_ports
++] = port
;
3858 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3859 "port %s", m
->bridge
->name
, m
->name
, name
);
3862 *out_portsp
= out_ports
;
3863 *n_out_portsp
= n_out_ports
;
3867 mirror_configure(struct mirror
*m
)
3869 const struct ovsrec_mirror
*cfg
= m
->cfg
;
3870 struct ofproto_mirror_settings s
;
3873 if (strcmp(cfg
->name
, m
->name
)) {
3875 m
->name
= xstrdup(cfg
->name
);
3879 /* Get output port or VLAN. */
3880 if (cfg
->output_port
) {
3881 s
.out_bundle
= port_lookup(m
->bridge
, cfg
->output_port
->name
);
3882 if (!s
.out_bundle
) {
3883 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3884 m
->bridge
->name
, m
->name
);
3887 s
.out_vlan
= UINT16_MAX
;
3889 if (cfg
->output_vlan
) {
3890 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3891 "output vlan; ignoring output vlan",
3892 m
->bridge
->name
, m
->name
);
3894 } else if (cfg
->output_vlan
) {
3895 /* The database should prevent invalid VLAN values. */
3896 s
.out_bundle
= NULL
;
3897 s
.out_vlan
= *cfg
->output_vlan
;
3899 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3900 m
->bridge
->name
, m
->name
);
3904 /* Get port selection. */
3905 if (cfg
->select_all
) {
3906 size_t n_ports
= hmap_count(&m
->bridge
->ports
);
3907 void **ports
= xmalloc(n_ports
* sizeof *ports
);
3912 HMAP_FOR_EACH (port
, hmap_node
, &m
->bridge
->ports
) {
3922 /* Get ports, dropping ports that don't exist.
3923 * The IDL ensures that there are no duplicates. */
3924 mirror_collect_ports(m
, cfg
->select_src_port
, cfg
->n_select_src_port
,
3925 &s
.srcs
, &s
.n_srcs
);
3926 mirror_collect_ports(m
, cfg
->select_dst_port
, cfg
->n_select_dst_port
,
3927 &s
.dsts
, &s
.n_dsts
);
3930 /* Get VLAN selection. */
3931 s
.src_vlans
= vlan_bitmap_from_array(cfg
->select_vlan
, cfg
->n_select_vlan
);
3934 ofproto_mirror_register(m
->bridge
->ofproto
, m
, &s
);
3937 if (s
.srcs
!= s
.dsts
) {
3946 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3948 * This is deprecated. It is only for compatibility with broken device drivers
3949 * in old versions of Linux that do not properly support VLANs when VLAN
3950 * devices are not used. When broken device drivers are no longer in
3951 * widespread use, we will delete these interfaces. */
3953 static struct ovsrec_port
**recs
;
3954 static size_t n_recs
, allocated_recs
;
3956 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3957 * splinters are reconfigured. */
3959 register_rec(struct ovsrec_port
*rec
)
3961 if (n_recs
>= allocated_recs
) {
3962 recs
= x2nrealloc(recs
, &allocated_recs
, sizeof *recs
);
3964 recs
[n_recs
++] = rec
;
3967 /* Frees all of the ports registered with register_reg(). */
3969 free_registered_recs(void)
3973 for (i
= 0; i
< n_recs
; i
++) {
3974 struct ovsrec_port
*port
= recs
[i
];
3977 for (j
= 0; j
< port
->n_interfaces
; j
++) {
3978 struct ovsrec_interface
*iface
= port
->interfaces
[j
];
3983 smap_destroy(&port
->other_config
);
3984 free(port
->interfaces
);
3992 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3995 vlan_splinters_is_enabled(const struct ovsrec_interface
*iface_cfg
)
3997 return smap_get_bool(&iface_cfg
->other_config
, "enable-vlan-splinters",
4001 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4004 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4005 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4006 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4009 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4010 * use, returns NULL.
4012 * Updates 'vlan_splinters_enabled_anywhere'. */
4013 static unsigned long int *
4014 collect_splinter_vlans(const struct ovsrec_open_vswitch
*ovs_cfg
)
4016 unsigned long int *splinter_vlans
;
4017 struct sset splinter_ifaces
;
4018 const char *real_dev_name
;
4019 struct shash
*real_devs
;
4020 struct shash_node
*node
;
4024 /* Free space allocated for synthesized ports and interfaces, since we're
4025 * in the process of reconstructing all of them. */
4026 free_registered_recs();
4028 splinter_vlans
= bitmap_allocate(4096);
4029 sset_init(&splinter_ifaces
);
4030 vlan_splinters_enabled_anywhere
= false;
4031 for (i
= 0; i
< ovs_cfg
->n_bridges
; i
++) {
4032 struct ovsrec_bridge
*br_cfg
= ovs_cfg
->bridges
[i
];
4035 for (j
= 0; j
< br_cfg
->n_ports
; j
++) {
4036 struct ovsrec_port
*port_cfg
= br_cfg
->ports
[j
];
4039 for (k
= 0; k
< port_cfg
->n_interfaces
; k
++) {
4040 struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[k
];
4042 if (vlan_splinters_is_enabled(iface_cfg
)) {
4043 vlan_splinters_enabled_anywhere
= true;
4044 sset_add(&splinter_ifaces
, iface_cfg
->name
);
4045 vlan_bitmap_from_array__(port_cfg
->trunks
,
4051 if (port_cfg
->tag
&& *port_cfg
->tag
> 0 && *port_cfg
->tag
< 4095) {
4052 bitmap_set1(splinter_vlans
, *port_cfg
->tag
);
4057 if (!vlan_splinters_enabled_anywhere
) {
4058 free(splinter_vlans
);
4059 sset_destroy(&splinter_ifaces
);
4063 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
4065 ofproto_get_vlan_usage(br
->ofproto
, splinter_vlans
);
4069 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4070 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4071 * device to be created for it. */
4072 bitmap_set0(splinter_vlans
, 0);
4073 bitmap_set0(splinter_vlans
, 4095);
4075 /* Delete all VLAN devices that we don't need. */
4077 real_devs
= vlandev_get_real_devs();
4078 SHASH_FOR_EACH (node
, real_devs
) {
4079 const struct vlan_real_dev
*real_dev
= node
->data
;
4080 const struct vlan_dev
*vlan_dev
;
4081 bool real_dev_has_splinters
;
4083 real_dev_has_splinters
= sset_contains(&splinter_ifaces
,
4085 HMAP_FOR_EACH (vlan_dev
, hmap_node
, &real_dev
->vlan_devs
) {
4086 if (!real_dev_has_splinters
4087 || !bitmap_is_set(splinter_vlans
, vlan_dev
->vid
)) {
4088 struct netdev
*netdev
;
4090 if (!netdev_open(vlan_dev
->name
, "system", &netdev
)) {
4091 if (!netdev_get_in4(netdev
, NULL
, NULL
) ||
4092 !netdev_get_in6(netdev
, NULL
)) {
4093 /* It has an IP address configured, so we don't own
4094 * it. Don't delete it. */
4096 vlandev_del(vlan_dev
->name
);
4098 netdev_close(netdev
);
4105 /* Add all VLAN devices that we need. */
4106 SSET_FOR_EACH (real_dev_name
, &splinter_ifaces
) {
4109 BITMAP_FOR_EACH_1 (vid
, 4096, splinter_vlans
) {
4110 if (!vlandev_get_name(real_dev_name
, vid
)) {
4111 vlandev_add(real_dev_name
, vid
);
4118 sset_destroy(&splinter_ifaces
);
4120 if (bitmap_scan(splinter_vlans
, 0, 4096) >= 4096) {
4121 free(splinter_vlans
);
4124 return splinter_vlans
;
4127 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4130 configure_splinter_port(struct port
*port
)
4132 struct ofproto
*ofproto
= port
->bridge
->ofproto
;
4133 ofp_port_t realdev_ofp_port
;
4134 const char *realdev_name
;
4135 struct iface
*vlandev
, *realdev
;
4137 ofproto_bundle_unregister(port
->bridge
->ofproto
, port
);
4139 vlandev
= CONTAINER_OF(list_front(&port
->ifaces
), struct iface
,
4142 realdev_name
= smap_get(&port
->cfg
->other_config
, "realdev");
4143 realdev
= iface_lookup(port
->bridge
, realdev_name
);
4144 realdev_ofp_port
= realdev
? realdev
->ofp_port
: 0;
4146 ofproto_port_set_realdev(ofproto
, vlandev
->ofp_port
, realdev_ofp_port
,
4150 static struct ovsrec_port
*
4151 synthesize_splinter_port(const char *real_dev_name
,
4152 const char *vlan_dev_name
, int vid
)
4154 struct ovsrec_interface
*iface
;
4155 struct ovsrec_port
*port
;
4157 iface
= xmalloc(sizeof *iface
);
4158 ovsrec_interface_init(iface
);
4159 iface
->name
= xstrdup(vlan_dev_name
);
4160 iface
->type
= "system";
4162 port
= xmalloc(sizeof *port
);
4163 ovsrec_port_init(port
);
4164 port
->interfaces
= xmemdup(&iface
, sizeof iface
);
4165 port
->n_interfaces
= 1;
4166 port
->name
= xstrdup(vlan_dev_name
);
4167 port
->vlan_mode
= "splinter";
4168 port
->tag
= xmalloc(sizeof *port
->tag
);
4171 smap_add(&port
->other_config
, "realdev", real_dev_name
);
4177 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4178 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4179 * 1-bit in the 'splinter_vlans' bitmap. */
4181 add_vlan_splinter_ports(struct bridge
*br
,
4182 const unsigned long int *splinter_vlans
,
4183 struct shash
*ports
)
4187 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4188 * we're modifying 'ports'. */
4189 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
4190 const char *name
= br
->cfg
->ports
[i
]->name
;
4191 struct ovsrec_port
*port_cfg
= shash_find_data(ports
, name
);
4194 for (j
= 0; j
< port_cfg
->n_interfaces
; j
++) {
4195 struct ovsrec_interface
*iface_cfg
= port_cfg
->interfaces
[j
];
4197 if (vlan_splinters_is_enabled(iface_cfg
)) {
4198 const char *real_dev_name
;
4201 real_dev_name
= iface_cfg
->name
;
4202 BITMAP_FOR_EACH_1 (vid
, 4096, splinter_vlans
) {
4203 const char *vlan_dev_name
;
4205 vlan_dev_name
= vlandev_get_name(real_dev_name
, vid
);
4207 && !shash_find(ports
, vlan_dev_name
)) {
4208 shash_add(ports
, vlan_dev_name
,
4209 synthesize_splinter_port(
4210 real_dev_name
, vlan_dev_name
, vid
));
4219 mirror_refresh_stats(struct mirror
*m
)
4221 struct ofproto
*ofproto
= m
->bridge
->ofproto
;
4222 uint64_t tx_packets
, tx_bytes
;
4225 size_t stat_cnt
= 0;
4227 if (ofproto_mirror_get_stats(ofproto
, m
, &tx_packets
, &tx_bytes
)) {
4228 ovsrec_mirror_set_statistics(m
->cfg
, NULL
, NULL
, 0);
4232 if (tx_packets
!= UINT64_MAX
) {
4233 keys
[stat_cnt
] = "tx_packets";
4234 values
[stat_cnt
] = tx_packets
;
4237 if (tx_bytes
!= UINT64_MAX
) {
4238 keys
[stat_cnt
] = "tx_bytes";
4239 values
[stat_cnt
] = tx_bytes
;
4243 ovsrec_mirror_set_statistics(m
->cfg
, keys
, values
, stat_cnt
);