]> git.proxmox.com Git - ovs.git/blob - vswitchd/bridge.c
netdev: Log a warning when netdev_set_config() fails.
[ovs.git] / vswitchd / bridge.c
1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
2 *
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:
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 *
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.
14 */
15
16 #include <config.h>
17 #include "bridge.h"
18 #include <errno.h>
19 #include <inttypes.h>
20 #include <stdlib.h>
21 #include "async-append.h"
22 #include "bfd.h"
23 #include "bitmap.h"
24 #include "cfm.h"
25 #include "coverage.h"
26 #include "daemon.h"
27 #include "dirs.h"
28 #include "dynamic-string.h"
29 #include "hash.h"
30 #include "hmap.h"
31 #include "hmapx.h"
32 #include "jsonrpc.h"
33 #include "lacp.h"
34 #include "list.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
37 #include "netdev.h"
38 #include "ofp-print.h"
39 #include "ofp-util.h"
40 #include "ofpbuf.h"
41 #include "ofproto/bond.h"
42 #include "ofproto/ofproto.h"
43 #include "poll-loop.h"
44 #include "sha1.h"
45 #include "shash.h"
46 #include "smap.h"
47 #include "socket-util.h"
48 #include "stream.h"
49 #include "stream-ssl.h"
50 #include "sset.h"
51 #include "system-stats.h"
52 #include "timeval.h"
53 #include "util.h"
54 #include "unixctl.h"
55 #include "vlandev.h"
56 #include "lib/vswitch-idl.h"
57 #include "xenserver.h"
58 #include "vlog.h"
59 #include "sflow_api.h"
60 #include "vlan-bitmap.h"
61
62 VLOG_DEFINE_THIS_MODULE(bridge);
63
64 COVERAGE_DEFINE(bridge_reconfigure);
65
66 /* Configuration of an uninstantiated iface. */
67 struct if_cfg {
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. */
72 };
73
74 /* OpenFlow port slated for removal from ofproto. */
75 struct ofpp_garbage {
76 struct list list_node; /* Node in bridge's ofpp_garbage. */
77 ofp_port_t ofp_port; /* Port to be deleted. */
78 };
79
80 struct iface {
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. */
86
87 /* These members are valid only after bridge_reconfigure() causes them to
88 * be initialized. */
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;
95 };
96
97 struct mirror {
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;
101 char *name;
102 const struct ovsrec_mirror *cfg;
103 };
104
105 struct port {
106 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
107 struct bridge *bridge;
108 char *name;
109
110 const struct ovsrec_port *cfg;
111
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. */
115 };
116
117 struct bridge {
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;
124
125 /* OpenFlow switch processing. */
126 struct ofproto *ofproto; /* OpenFlow switch. */
127
128 /* Bridge ports. */
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. */
132
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'. */
136
137 /* Port mirroring. */
138 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
139
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;
144 };
145
146 /* All bridges, indexed by name. */
147 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
148
149 /* OVSDB IDL used to obtain configuration. */
150 static struct ovsdb_idl *idl;
151
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.)
161 *
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;
168
169 /* Most recently processed IDL sequence number. */
170 static unsigned int idl_seqno;
171
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;
176
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;
187
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,
212 size_t n_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 *,
224 const char *name);
225 static bool port_is_bond_fake_iface(const struct port *);
226
227 static unixctl_cb_func qos_unixctl_show;
228
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 *);
238
239 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
240 static void run_system_stats(void);
241
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 *);
248
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 *);
272
273 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
274 *
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. */
279
280 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
281 static bool vlan_splinters_enabled_anywhere;
282
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);
290
291 static void
292 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
293 {
294 struct shash iface_hints;
295 static bool initialized = false;
296 int i;
297
298 if (initialized) {
299 return;
300 }
301
302 shash_init(&iface_hints);
303
304 if (cfg) {
305 for (i = 0; i < cfg->n_bridges; i++) {
306 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
307 int j;
308
309 for (j = 0; j < br_cfg->n_ports; j++) {
310 struct ovsrec_port *port_cfg = br_cfg->ports[j];
311 int k;
312
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;
316
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);
321
322 shash_add(&iface_hints, if_cfg->name, iface_hint);
323 }
324 }
325 }
326 }
327
328 ofproto_init(&iface_hints);
329
330 shash_destroy_free_data(&iface_hints);
331 initialized = true;
332 }
333 \f
334 /* Public functions. */
335
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(). */
339 void
340 bridge_init(const char *remote)
341 {
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);
347
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);
355
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);
359
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);
364
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_flap_count);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
382 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
383 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
384 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
385
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
387 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
388 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
389 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
390
391 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
392
393 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
394
395 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
396 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
397
398 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
401 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
402
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
406 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
407 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
408
409 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
410
411 /* Register unixctl commands. */
412 unixctl_command_register("qos/show", "interface", 1, 1,
413 qos_unixctl_show, NULL);
414 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
415 bridge_unixctl_dump_flows, NULL);
416 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
417 bridge_unixctl_reconnect, NULL);
418 lacp_init();
419 bond_init();
420 cfm_init();
421 stp_init();
422 }
423
424 void
425 bridge_exit(void)
426 {
427 struct bridge *br, *next_br;
428
429 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
430 bridge_destroy(br);
431 }
432 ovsdb_idl_destroy(idl);
433 }
434
435 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
436 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
437 * responsible for freeing '*managersp' (with free()).
438 *
439 * You may be asking yourself "why does ovs-vswitchd care?", because
440 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
441 * should not be and in fact is not directly involved in that. But
442 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
443 * it has to tell in-band control where the managers are to enable that.
444 * (Thus, only managers connected in-band are collected.)
445 */
446 static void
447 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
448 struct sockaddr_in **managersp, size_t *n_managersp)
449 {
450 struct sockaddr_in *managers = NULL;
451 size_t n_managers = 0;
452 struct sset targets;
453 size_t i;
454
455 /* Collect all of the potential targets from the "targets" columns of the
456 * rows pointed to by "manager_options", excluding any that are
457 * out-of-band. */
458 sset_init(&targets);
459 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
460 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
461
462 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
463 sset_find_and_delete(&targets, m->target);
464 } else {
465 sset_add(&targets, m->target);
466 }
467 }
468
469 /* Now extract the targets' IP addresses. */
470 if (!sset_is_empty(&targets)) {
471 const char *target;
472
473 managers = xmalloc(sset_count(&targets) * sizeof *managers);
474 SSET_FOR_EACH (target, &targets) {
475 struct sockaddr_in *sin = &managers[n_managers];
476
477 if (stream_parse_target_with_default_port(target,
478 OVSDB_OLD_PORT,
479 sin)) {
480 n_managers++;
481 }
482 }
483 }
484 sset_destroy(&targets);
485
486 *managersp = managers;
487 *n_managersp = n_managers;
488 }
489
490 static void
491 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
492 {
493 unsigned long int *splinter_vlans;
494 struct bridge *br;
495
496 COVERAGE_INC(bridge_reconfigure);
497
498 ovs_assert(!reconfiguring);
499 reconfiguring = true;
500
501 ofproto_set_flow_eviction_threshold(
502 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
503 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
504
505 ofproto_set_n_handler_threads(
506 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0));
507
508 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
509 "force-miss-model"));
510
511 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
512 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
513 * configuration otherwise.
514 *
515 * This is mostly an update to bridge data structures. Nothing is pushed
516 * down to ofproto or lower layers. */
517 add_del_bridges(ovs_cfg);
518 splinter_vlans = collect_splinter_vlans(ovs_cfg);
519 HMAP_FOR_EACH (br, node, &all_bridges) {
520 bridge_add_del_ports(br, splinter_vlans);
521 }
522 free(splinter_vlans);
523
524 /* Delete datapaths that are no longer configured, and create ones which
525 * don't exist but should. */
526 bridge_update_ofprotos();
527
528 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
529 HMAP_FOR_EACH (br, node, &all_bridges) {
530 bridge_refresh_ofp_port(br);
531 }
532
533 /* Clear database records for "if_cfg"s which haven't been instantiated. */
534 HMAP_FOR_EACH (br, node, &all_bridges) {
535 struct if_cfg *if_cfg;
536
537 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
538 iface_clear_db_record(if_cfg->cfg);
539 }
540 }
541
542 reconfigure_system_stats(ovs_cfg);
543 }
544
545 static bool
546 bridge_reconfigure_ofp(void)
547 {
548 long long int deadline;
549 struct bridge *br;
550
551 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
552
553 /* The kernel will reject any attempt to add a given port to a datapath if
554 * that port already belongs to a different datapath, so we must do all
555 * port deletions before any port additions. */
556 HMAP_FOR_EACH (br, node, &all_bridges) {
557 struct ofpp_garbage *garbage, *next;
558
559 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
560 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
561 * internal datastructures may not be consistent. Eventually, when
562 * port additions and deletions are cheaper, these calls should be
563 * removed. */
564 bridge_run_fast();
565 ofproto_port_del(br->ofproto, garbage->ofp_port);
566 list_remove(&garbage->list_node);
567 free(garbage);
568
569 if (time_msec() >= deadline) {
570 return false;
571 }
572 bridge_run_fast();
573 }
574 }
575
576 HMAP_FOR_EACH (br, node, &all_bridges) {
577 struct if_cfg *if_cfg, *next;
578
579 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
580 iface_create(br, if_cfg, OFPP_NONE);
581 if (time_msec() >= deadline) {
582 return false;
583 }
584 }
585 }
586
587 return true;
588 }
589
590 static bool
591 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
592 {
593 struct sockaddr_in *managers;
594 int sflow_bridge_number;
595 size_t n_managers;
596 struct bridge *br;
597 bool done;
598
599 ovs_assert(reconfiguring);
600 done = bridge_reconfigure_ofp();
601
602 /* Complete the configuration. */
603 sflow_bridge_number = 0;
604 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
605 HMAP_FOR_EACH (br, node, &all_bridges) {
606 struct port *port;
607
608 /* We need the datapath ID early to allow LACP ports to use it as the
609 * default system ID. */
610 bridge_configure_datapath_id(br);
611
612 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
613 struct iface *iface;
614
615 port_configure(port);
616
617 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
618 iface_configure_cfm(iface);
619 iface_configure_qos(iface, port->cfg->qos);
620 iface_set_mac(iface);
621 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
622 &iface->cfg->bfd);
623 }
624 }
625 bridge_configure_mirrors(br);
626 bridge_configure_forward_bpdu(br);
627 bridge_configure_mac_table(br);
628 bridge_configure_remotes(br, managers, n_managers);
629 bridge_configure_netflow(br);
630 bridge_configure_sflow(br, &sflow_bridge_number);
631 bridge_configure_ipfix(br);
632 bridge_configure_stp(br);
633 bridge_configure_tables(br);
634 bridge_configure_dp_desc(br);
635
636 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
637 /* XXX: Remove this warning message eventually. */
638 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
639 " moved to the Open_vSwitch table. Ignoring its"
640 " setting in the bridge table.");
641 }
642 }
643 free(managers);
644
645 return done;
646 }
647
648 /* Delete ofprotos which aren't configured or have the wrong type. Create
649 * ofprotos which don't exist but need to. */
650 static void
651 bridge_update_ofprotos(void)
652 {
653 struct bridge *br, *next;
654 struct sset names;
655 struct sset types;
656 const char *type;
657
658 /* Delete ofprotos with no bridge or with the wrong type. */
659 sset_init(&names);
660 sset_init(&types);
661 ofproto_enumerate_types(&types);
662 SSET_FOR_EACH (type, &types) {
663 const char *name;
664
665 ofproto_enumerate_names(type, &names);
666 SSET_FOR_EACH (name, &names) {
667 br = bridge_lookup(name);
668 if (!br || strcmp(type, br->type)) {
669 ofproto_delete(name, type);
670 }
671 }
672 }
673 sset_destroy(&names);
674 sset_destroy(&types);
675
676 /* Add ofprotos for bridges which don't have one yet. */
677 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
678 struct bridge *br2;
679 int error;
680
681 if (br->ofproto) {
682 continue;
683 }
684
685 /* Remove ports from any datapath with the same name as 'br'. If we
686 * don't do this, creating 'br''s ofproto will fail because a port with
687 * the same name as its local port already exists. */
688 HMAP_FOR_EACH (br2, node, &all_bridges) {
689 struct ofproto_port ofproto_port;
690
691 if (!br2->ofproto) {
692 continue;
693 }
694
695 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
696 &ofproto_port)) {
697 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
698 if (error) {
699 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
700 ovs_strerror(error));
701 }
702 ofproto_port_destroy(&ofproto_port);
703 }
704 }
705
706 error = ofproto_create(br->name, br->type, &br->ofproto);
707 if (error) {
708 VLOG_ERR("failed to create bridge %s: %s", br->name,
709 ovs_strerror(error));
710 bridge_destroy(br);
711 }
712 }
713 }
714
715 static void
716 port_configure(struct port *port)
717 {
718 const struct ovsrec_port *cfg = port->cfg;
719 struct bond_settings bond_settings;
720 struct lacp_settings lacp_settings;
721 struct ofproto_bundle_settings s;
722 struct iface *iface;
723
724 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
725 configure_splinter_port(port);
726 return;
727 }
728
729 /* Get name. */
730 s.name = port->name;
731
732 /* Get slaves. */
733 s.n_slaves = 0;
734 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
735 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
736 s.slaves[s.n_slaves++] = iface->ofp_port;
737 }
738
739 /* Get VLAN tag. */
740 s.vlan = -1;
741 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
742 s.vlan = *cfg->tag;
743 }
744
745 /* Get VLAN trunks. */
746 s.trunks = NULL;
747 if (cfg->n_trunks) {
748 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
749 }
750
751 /* Get VLAN mode. */
752 if (cfg->vlan_mode) {
753 if (!strcmp(cfg->vlan_mode, "access")) {
754 s.vlan_mode = PORT_VLAN_ACCESS;
755 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
756 s.vlan_mode = PORT_VLAN_TRUNK;
757 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
758 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
759 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
760 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
761 } else {
762 /* This "can't happen" because ovsdb-server should prevent it. */
763 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
764 s.vlan_mode = PORT_VLAN_TRUNK;
765 }
766 } else {
767 if (s.vlan >= 0) {
768 s.vlan_mode = PORT_VLAN_ACCESS;
769 if (cfg->n_trunks) {
770 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
771 port->name);
772 }
773 } else {
774 s.vlan_mode = PORT_VLAN_TRUNK;
775 }
776 }
777 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
778 false);
779
780 /* Get LACP settings. */
781 s.lacp = port_configure_lacp(port, &lacp_settings);
782 if (s.lacp) {
783 size_t i = 0;
784
785 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
786 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
787 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
788 }
789 } else {
790 s.lacp_slaves = NULL;
791 }
792
793 /* Get bond settings. */
794 if (s.n_slaves > 1) {
795 s.bond = &bond_settings;
796 port_configure_bond(port, &bond_settings);
797 } else {
798 s.bond = NULL;
799 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
800 netdev_set_miimon_interval(iface->netdev, 0);
801 }
802 }
803
804 /* Register. */
805 ofproto_bundle_register(port->bridge->ofproto, port, &s);
806
807 /* Clean up. */
808 free(s.slaves);
809 free(s.trunks);
810 free(s.lacp_slaves);
811 }
812
813 static void
814 bridge_configure_flow_miss_model(const char *opt)
815 {
816 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
817
818 if (opt) {
819 if (strcmp(opt, "with-facets")) {
820 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
821 } else if (strcmp(opt, "without-facets")) {
822 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
823 }
824 }
825
826 ofproto_set_flow_miss_model(model);
827 }
828
829 /* Pick local port hardware address and datapath ID for 'br'. */
830 static void
831 bridge_configure_datapath_id(struct bridge *br)
832 {
833 uint8_t ea[ETH_ADDR_LEN];
834 uint64_t dpid;
835 struct iface *local_iface;
836 struct iface *hw_addr_iface;
837 char *dpid_string;
838
839 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
840 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
841 if (local_iface) {
842 int error = netdev_set_etheraddr(local_iface->netdev, ea);
843 if (error) {
844 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
845 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
846 "Ethernet address: %s",
847 br->name, ovs_strerror(error));
848 }
849 }
850 memcpy(br->ea, ea, ETH_ADDR_LEN);
851
852 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
853 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
854 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
855 ofproto_set_datapath_id(br->ofproto, dpid);
856 }
857
858 dpid_string = xasprintf("%016"PRIx64, dpid);
859 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
860 free(dpid_string);
861 }
862
863 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
864 * 'br'. */
865 static uint32_t
866 bridge_get_allowed_versions(struct bridge *br)
867 {
868 if (!br->cfg->n_protocols)
869 return 0;
870
871 return ofputil_versions_from_strings(br->cfg->protocols,
872 br->cfg->n_protocols);
873 }
874
875 /* Set NetFlow configuration on 'br'. */
876 static void
877 bridge_configure_netflow(struct bridge *br)
878 {
879 struct ovsrec_netflow *cfg = br->cfg->netflow;
880 struct netflow_options opts;
881
882 if (!cfg) {
883 ofproto_set_netflow(br->ofproto, NULL);
884 return;
885 }
886
887 memset(&opts, 0, sizeof opts);
888
889 /* Get default NetFlow configuration from datapath.
890 * Apply overrides from 'cfg'. */
891 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
892 if (cfg->engine_type) {
893 opts.engine_type = *cfg->engine_type;
894 }
895 if (cfg->engine_id) {
896 opts.engine_id = *cfg->engine_id;
897 }
898
899 /* Configure active timeout interval. */
900 opts.active_timeout = cfg->active_timeout;
901 if (!opts.active_timeout) {
902 opts.active_timeout = -1;
903 } else if (opts.active_timeout < 0) {
904 VLOG_WARN("bridge %s: active timeout interval set to negative "
905 "value, using default instead (%d seconds)", br->name,
906 NF_ACTIVE_TIMEOUT_DEFAULT);
907 opts.active_timeout = -1;
908 }
909
910 /* Add engine ID to interface number to disambiguate bridgs? */
911 opts.add_id_to_iface = cfg->add_id_to_interface;
912 if (opts.add_id_to_iface) {
913 if (opts.engine_id > 0x7f) {
914 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
915 "another vswitch, choose an engine id less than 128",
916 br->name);
917 }
918 if (hmap_count(&br->ports) > 508) {
919 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
920 "another port when more than 508 ports are used",
921 br->name);
922 }
923 }
924
925 /* Collectors. */
926 sset_init(&opts.collectors);
927 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
928
929 /* Configure. */
930 if (ofproto_set_netflow(br->ofproto, &opts)) {
931 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
932 }
933 sset_destroy(&opts.collectors);
934 }
935
936 /* Set sFlow configuration on 'br'. */
937 static void
938 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
939 {
940 const struct ovsrec_sflow *cfg = br->cfg->sflow;
941 struct ovsrec_controller **controllers;
942 struct ofproto_sflow_options oso;
943 size_t n_controllers;
944 size_t i;
945
946 if (!cfg) {
947 ofproto_set_sflow(br->ofproto, NULL);
948 return;
949 }
950
951 memset(&oso, 0, sizeof oso);
952
953 sset_init(&oso.targets);
954 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
955
956 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
957 if (cfg->sampling) {
958 oso.sampling_rate = *cfg->sampling;
959 }
960
961 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
962 if (cfg->polling) {
963 oso.polling_interval = *cfg->polling;
964 }
965
966 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
967 if (cfg->header) {
968 oso.header_len = *cfg->header;
969 }
970
971 oso.sub_id = (*sflow_bridge_number)++;
972 oso.agent_device = cfg->agent;
973
974 oso.control_ip = NULL;
975 n_controllers = bridge_get_controllers(br, &controllers);
976 for (i = 0; i < n_controllers; i++) {
977 if (controllers[i]->local_ip) {
978 oso.control_ip = controllers[i]->local_ip;
979 break;
980 }
981 }
982 ofproto_set_sflow(br->ofproto, &oso);
983
984 sset_destroy(&oso.targets);
985 }
986
987 /* Returns whether a IPFIX row is valid. */
988 static bool
989 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
990 {
991 return ipfix && ipfix->n_targets > 0;
992 }
993
994 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
995 static bool
996 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
997 const struct bridge *br)
998 {
999 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1000 }
1001
1002 /* Set IPFIX configuration on 'br'. */
1003 static void
1004 bridge_configure_ipfix(struct bridge *br)
1005 {
1006 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1007 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1008 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1009 struct ofproto_ipfix_bridge_exporter_options be_opts;
1010 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1011 size_t n_fe_opts = 0;
1012
1013 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1014 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1015 n_fe_opts++;
1016 }
1017 }
1018
1019 if (!valid_be_cfg && n_fe_opts == 0) {
1020 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1021 return;
1022 }
1023
1024 if (valid_be_cfg) {
1025 memset(&be_opts, 0, sizeof be_opts);
1026
1027 sset_init(&be_opts.targets);
1028 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1029
1030 if (be_cfg->sampling) {
1031 be_opts.sampling_rate = *be_cfg->sampling;
1032 } else {
1033 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1034 }
1035 if (be_cfg->obs_domain_id) {
1036 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1037 }
1038 if (be_cfg->obs_point_id) {
1039 be_opts.obs_point_id = *be_cfg->obs_point_id;
1040 }
1041 if (be_cfg->cache_active_timeout) {
1042 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1043 }
1044 if (be_cfg->cache_max_flows) {
1045 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1046 }
1047 }
1048
1049 if (n_fe_opts > 0) {
1050 struct ofproto_ipfix_flow_exporter_options *opts;
1051 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1052 opts = fe_opts;
1053 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1054 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1055 opts->collector_set_id = fe_cfg->id;
1056 sset_init(&opts->targets);
1057 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1058 fe_cfg->ipfix->n_targets);
1059 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1060 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1061 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1062 ? *fe_cfg->ipfix->cache_max_flows : 0;
1063 opts++;
1064 }
1065 }
1066 }
1067
1068 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1069 n_fe_opts);
1070
1071 if (valid_be_cfg) {
1072 sset_destroy(&be_opts.targets);
1073 }
1074
1075 if (n_fe_opts > 0) {
1076 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1077 size_t i;
1078 for (i = 0; i < n_fe_opts; i++) {
1079 sset_destroy(&opts->targets);
1080 opts++;
1081 }
1082 free(fe_opts);
1083 }
1084 }
1085
1086 static void
1087 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1088 struct ofproto_port_stp_settings *port_s,
1089 int *port_num_counter, unsigned long *port_num_bitmap)
1090 {
1091 const char *config_str;
1092 struct iface *iface;
1093
1094 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1095 port_s->enable = false;
1096 return;
1097 } else {
1098 port_s->enable = true;
1099 }
1100
1101 /* STP over bonds is not supported. */
1102 if (!list_is_singleton(&port->ifaces)) {
1103 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1104 port->name);
1105 port_s->enable = false;
1106 return;
1107 }
1108
1109 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1110
1111 /* Internal ports shouldn't participate in spanning tree, so
1112 * skip them. */
1113 if (!strcmp(iface->type, "internal")) {
1114 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1115 port_s->enable = false;
1116 return;
1117 }
1118
1119 /* STP on mirror output ports is not supported. */
1120 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1121 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1122 port_s->enable = false;
1123 return;
1124 }
1125
1126 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1127 if (config_str) {
1128 unsigned long int port_num = strtoul(config_str, NULL, 0);
1129 int port_idx = port_num - 1;
1130
1131 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1132 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1133 port_s->enable = false;
1134 return;
1135 }
1136
1137 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1138 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1139 port->name, port_num);
1140 port_s->enable = false;
1141 return;
1142 }
1143 bitmap_set1(port_num_bitmap, port_idx);
1144 port_s->port_num = port_idx;
1145 } else {
1146 if (*port_num_counter >= STP_MAX_PORTS) {
1147 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1148 port_s->enable = false;
1149 return;
1150 }
1151
1152 port_s->port_num = (*port_num_counter)++;
1153 }
1154
1155 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1156 if (config_str) {
1157 port_s->path_cost = strtoul(config_str, NULL, 10);
1158 } else {
1159 enum netdev_features current;
1160 unsigned int mbps;
1161
1162 netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1163 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1164 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1165 }
1166
1167 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1168 if (config_str) {
1169 port_s->priority = strtoul(config_str, NULL, 0);
1170 } else {
1171 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1172 }
1173 }
1174
1175 /* Set spanning tree configuration on 'br'. */
1176 static void
1177 bridge_configure_stp(struct bridge *br)
1178 {
1179 if (!br->cfg->stp_enable) {
1180 ofproto_set_stp(br->ofproto, NULL);
1181 } else {
1182 struct ofproto_stp_settings br_s;
1183 const char *config_str;
1184 struct port *port;
1185 int port_num_counter;
1186 unsigned long *port_num_bitmap;
1187
1188 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1189 if (config_str) {
1190 uint8_t ea[ETH_ADDR_LEN];
1191
1192 if (eth_addr_from_string(config_str, ea)) {
1193 br_s.system_id = eth_addr_to_uint64(ea);
1194 } else {
1195 br_s.system_id = eth_addr_to_uint64(br->ea);
1196 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1197 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1198 }
1199 } else {
1200 br_s.system_id = eth_addr_to_uint64(br->ea);
1201 }
1202
1203 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1204 if (config_str) {
1205 br_s.priority = strtoul(config_str, NULL, 0);
1206 } else {
1207 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1208 }
1209
1210 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1211 if (config_str) {
1212 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1213 } else {
1214 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1215 }
1216
1217 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1218 if (config_str) {
1219 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1220 } else {
1221 br_s.max_age = STP_DEFAULT_MAX_AGE;
1222 }
1223
1224 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1225 if (config_str) {
1226 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1227 } else {
1228 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1229 }
1230
1231 /* Configure STP on the bridge. */
1232 if (ofproto_set_stp(br->ofproto, &br_s)) {
1233 VLOG_ERR("bridge %s: could not enable STP", br->name);
1234 return;
1235 }
1236
1237 /* Users must either set the port number with the "stp-port-num"
1238 * configuration on all ports or none. If manual configuration
1239 * is not done, then we allocate them sequentially. */
1240 port_num_counter = 0;
1241 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1242 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1243 struct ofproto_port_stp_settings port_s;
1244 struct iface *iface;
1245
1246 port_configure_stp(br->ofproto, port, &port_s,
1247 &port_num_counter, port_num_bitmap);
1248
1249 /* As bonds are not supported, just apply configuration to
1250 * all interfaces. */
1251 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1252 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1253 &port_s)) {
1254 VLOG_ERR("port %s: could not enable STP", port->name);
1255 continue;
1256 }
1257 }
1258 }
1259
1260 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1261 && port_num_counter) {
1262 VLOG_ERR("bridge %s: must manually configure all STP port "
1263 "IDs or none, disabling", br->name);
1264 ofproto_set_stp(br->ofproto, NULL);
1265 }
1266 bitmap_free(port_num_bitmap);
1267 }
1268 }
1269
1270 static bool
1271 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1272 {
1273 const struct port *port = port_lookup(br, name);
1274 return port && port_is_bond_fake_iface(port);
1275 }
1276
1277 static bool
1278 port_is_bond_fake_iface(const struct port *port)
1279 {
1280 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1281 }
1282
1283 static void
1284 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1285 {
1286 struct bridge *br, *next;
1287 struct shash new_br;
1288 size_t i;
1289
1290 /* Collect new bridges' names and types. */
1291 shash_init(&new_br);
1292 for (i = 0; i < cfg->n_bridges; i++) {
1293 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1294 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1295
1296 if (strchr(br_cfg->name, '/')) {
1297 /* Prevent remote ovsdb-server users from accessing arbitrary
1298 * directories, e.g. consider a bridge named "../../../etc/". */
1299 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1300 br_cfg->name);
1301 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1302 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1303 }
1304 }
1305
1306 /* Get rid of deleted bridges or those whose types have changed.
1307 * Update 'cfg' of bridges that still exist. */
1308 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1309 br->cfg = shash_find_data(&new_br, br->name);
1310 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1311 br->cfg->datapath_type))) {
1312 bridge_destroy(br);
1313 }
1314 }
1315
1316 /* Add new bridges. */
1317 for (i = 0; i < cfg->n_bridges; i++) {
1318 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1319 struct bridge *br = bridge_lookup(br_cfg->name);
1320 if (!br) {
1321 bridge_create(br_cfg);
1322 }
1323 }
1324
1325 shash_destroy(&new_br);
1326 }
1327
1328 static void
1329 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1330 {
1331 struct bridge *br = iface->port->bridge;
1332
1333 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1334 iface->ofp_port = ofp_port;
1335 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1336 hash_ofp_port(ofp_port));
1337 iface_set_ofport(iface->cfg, ofp_port);
1338 }
1339
1340 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1341 * Returns 0 if successful, otherwise a positive errno value. */
1342 static int
1343 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1344 struct netdev *netdev)
1345 {
1346 return netdev_set_config(netdev, &iface_cfg->options);
1347 }
1348
1349 /* This function determines whether 'ofproto_port', which is attached to
1350 * br->ofproto's datapath, is one that we want in 'br'.
1351 *
1352 * If it is, it returns true, after creating an iface (if necessary),
1353 * configuring the iface's netdev according to the iface's options, and setting
1354 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1355 *
1356 * If, on the other hand, 'port' should be removed, it returns false. The
1357 * caller should later detach the port from br->ofproto. */
1358 static bool
1359 bridge_refresh_one_ofp_port(struct bridge *br,
1360 const struct ofproto_port *ofproto_port)
1361 {
1362 const char *name = ofproto_port->name;
1363 const char *type = ofproto_port->type;
1364 ofp_port_t ofp_port = ofproto_port->ofp_port;
1365
1366 struct iface *iface = iface_lookup(br, name);
1367 if (iface) {
1368 /* Check that the name-to-number mapping is one-to-one. */
1369 if (iface->ofp_port != OFPP_NONE) {
1370 VLOG_WARN("bridge %s: interface %s reported twice",
1371 br->name, name);
1372 return false;
1373 } else if (iface_from_ofp_port(br, ofp_port)) {
1374 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1375 br->name, ofp_port);
1376 return false;
1377 }
1378
1379 /* There's a configured interface named 'name'. */
1380 if (strcmp(type, iface->type)
1381 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1382 /* It's the wrong type, or it's the right type but can't be
1383 * configured as the user requested, so we must destroy it. */
1384 return false;
1385 } else {
1386 /* It's the right type and configured correctly. Keep it. */
1387 iface_set_ofp_port(iface, ofp_port);
1388 return true;
1389 }
1390 } else if (bridge_has_bond_fake_iface(br, name)
1391 && !strcmp(type, "internal")) {
1392 /* It's a bond fake iface. Keep it. */
1393 return true;
1394 } else {
1395 /* There's no configured interface named 'name', but there might be an
1396 * interface of that name queued to be created.
1397 *
1398 * If there is, and it has the correct type, then try to configure it
1399 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1400 * delete it and later try to re-add it. */
1401 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1402 return (if_cfg
1403 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1404 && iface_create(br, if_cfg, ofp_port));
1405 }
1406 }
1407
1408 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1409 * consistent with the ofp_ports in "br->ofproto". */
1410 static void
1411 bridge_refresh_ofp_port(struct bridge *br)
1412 {
1413 struct ofproto_port_dump dump;
1414 struct ofproto_port ofproto_port;
1415 struct port *port, *port_next;
1416
1417 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1418 hmap_clear(&br->ifaces);
1419 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1420 struct iface *iface;
1421
1422 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1423 iface->ofp_port = OFPP_NONE;
1424 }
1425 }
1426
1427 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1428 * already exist in the datapath and promote them to full fledged "struct
1429 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1430 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1431 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1432 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1433 garbage->ofp_port = ofproto_port.ofp_port;
1434 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1435 }
1436 }
1437
1438 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1439 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1440 * later they can be added to ofproto. */
1441 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1442 struct iface *iface, *iface_next;
1443
1444 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1445 if (iface->ofp_port == OFPP_NONE) {
1446 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1447 iface_destroy(iface);
1448 }
1449 }
1450
1451 if (list_is_empty(&port->ifaces)) {
1452 port_destroy(port);
1453 }
1454 }
1455 }
1456
1457 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1458 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1459 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1460 * untouched.
1461 *
1462 * If successful, returns 0 and stores the network device in '*netdevp'. On
1463 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1464 static int
1465 iface_do_create(const struct bridge *br,
1466 const struct if_cfg *if_cfg,
1467 ofp_port_t *ofp_portp, struct netdev **netdevp)
1468 {
1469 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1470 const struct ovsrec_port *port_cfg = if_cfg->parent;
1471 struct netdev *netdev = NULL;
1472 int error;
1473
1474 if (netdev_is_reserved_name(iface_cfg->name)) {
1475 VLOG_WARN("could not create interface %s, name is reserved",
1476 iface_cfg->name);
1477 error = EINVAL;
1478 goto error;
1479 }
1480
1481 error = netdev_open(iface_cfg->name,
1482 iface_get_type(iface_cfg, br->cfg), &netdev);
1483 if (error) {
1484 VLOG_WARN("could not open network device %s (%s)",
1485 iface_cfg->name, ovs_strerror(error));
1486 goto error;
1487 }
1488
1489 error = iface_set_netdev_config(iface_cfg, netdev);
1490 if (error) {
1491 goto error;
1492 }
1493
1494 if (*ofp_portp == OFPP_NONE) {
1495 ofp_port_t ofp_port = if_cfg->ofport;
1496
1497 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1498 if (error) {
1499 goto error;
1500 }
1501 *ofp_portp = ofp_port;
1502
1503 VLOG_INFO("bridge %s: added interface %s on port %d",
1504 br->name, iface_cfg->name, *ofp_portp);
1505 } else {
1506 VLOG_DBG("bridge %s: interface %s is on port %d",
1507 br->name, iface_cfg->name, *ofp_portp);
1508 }
1509
1510 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1511 || iface_is_internal(iface_cfg, br->cfg)) {
1512 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1513 }
1514
1515 *netdevp = netdev;
1516 return 0;
1517
1518 error:
1519 *netdevp = NULL;
1520 netdev_close(netdev);
1521 return error;
1522 }
1523
1524 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1525 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1526 * automatically allocated for the iface. Takes ownership of and
1527 * deallocates 'if_cfg'.
1528 *
1529 * Return true if an iface is successfully created, false otherwise. */
1530 static bool
1531 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1532 {
1533 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1534 const struct ovsrec_port *port_cfg = if_cfg->parent;
1535
1536 struct netdev *netdev;
1537 struct iface *iface;
1538 struct port *port;
1539 int error;
1540 bool ok = true;
1541
1542 /* Do the bits that can fail up front.
1543 *
1544 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1545 * internal datastructures may not be consistent. Eventually, when port
1546 * additions and deletions are cheaper, these calls should be removed. */
1547 bridge_run_fast();
1548 ovs_assert(!iface_lookup(br, iface_cfg->name));
1549 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1550 bridge_run_fast();
1551 if (error) {
1552 iface_set_ofport(iface_cfg, OFPP_NONE);
1553 iface_clear_db_record(iface_cfg);
1554 ok = false;
1555 goto done;
1556 }
1557
1558 /* Get or create the port structure. */
1559 port = port_lookup(br, port_cfg->name);
1560 if (!port) {
1561 port = port_create(br, port_cfg);
1562 }
1563
1564 /* Create the iface structure. */
1565 iface = xzalloc(sizeof *iface);
1566 list_push_back(&port->ifaces, &iface->port_elem);
1567 hmap_insert(&br->iface_by_name, &iface->name_node,
1568 hash_string(iface_cfg->name, 0));
1569 iface->port = port;
1570 iface->name = xstrdup(iface_cfg->name);
1571 iface->ofp_port = OFPP_NONE;
1572 iface->netdev = netdev;
1573 iface->type = iface_get_type(iface_cfg, br->cfg);
1574 iface->cfg = iface_cfg;
1575
1576 iface_set_ofp_port(iface, ofp_port);
1577
1578 /* Populate initial status in database. */
1579 iface_refresh_stats(iface);
1580 iface_refresh_status(iface);
1581
1582 /* Add bond fake iface if necessary. */
1583 if (port_is_bond_fake_iface(port)) {
1584 struct ofproto_port ofproto_port;
1585
1586 if (ofproto_port_query_by_name(br->ofproto, port->name,
1587 &ofproto_port)) {
1588 struct netdev *netdev;
1589 int error;
1590
1591 error = netdev_open(port->name, "internal", &netdev);
1592 if (!error) {
1593 ofp_port_t fake_ofp_port = if_cfg->ofport;
1594
1595 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1596 netdev_close(netdev);
1597 } else {
1598 VLOG_WARN("could not open network device %s (%s)",
1599 port->name, ovs_strerror(error));
1600 }
1601 } else {
1602 /* Already exists, nothing to do. */
1603 ofproto_port_destroy(&ofproto_port);
1604 }
1605 }
1606
1607 done:
1608 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1609 free(if_cfg);
1610
1611 return ok;
1612 }
1613
1614 /* Set forward BPDU option. */
1615 static void
1616 bridge_configure_forward_bpdu(struct bridge *br)
1617 {
1618 ofproto_set_forward_bpdu(br->ofproto,
1619 smap_get_bool(&br->cfg->other_config,
1620 "forward-bpdu",
1621 false));
1622 }
1623
1624 /* Set MAC learning table configuration for 'br'. */
1625 static void
1626 bridge_configure_mac_table(struct bridge *br)
1627 {
1628 const char *idle_time_str;
1629 int idle_time;
1630
1631 const char *mac_table_size_str;
1632 int mac_table_size;
1633
1634 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1635 idle_time = (idle_time_str && atoi(idle_time_str)
1636 ? atoi(idle_time_str)
1637 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1638
1639 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1640 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1641 ? atoi(mac_table_size_str)
1642 : MAC_DEFAULT_MAX);
1643
1644 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1645 }
1646
1647 static void
1648 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1649 struct iface **hw_addr_iface)
1650 {
1651 struct hmapx mirror_output_ports;
1652 const char *hwaddr;
1653 struct port *port;
1654 bool found_addr = false;
1655 int error;
1656 int i;
1657
1658 *hw_addr_iface = NULL;
1659
1660 /* Did the user request a particular MAC? */
1661 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1662 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1663 if (eth_addr_is_multicast(ea)) {
1664 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1665 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1666 } else if (eth_addr_is_zero(ea)) {
1667 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1668 } else {
1669 return;
1670 }
1671 }
1672
1673 /* Mirror output ports don't participate in picking the local hardware
1674 * address. ofproto can't help us find out whether a given port is a
1675 * mirror output because we haven't configured mirrors yet, so we need to
1676 * accumulate them ourselves. */
1677 hmapx_init(&mirror_output_ports);
1678 for (i = 0; i < br->cfg->n_mirrors; i++) {
1679 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1680 if (m->output_port) {
1681 hmapx_add(&mirror_output_ports, m->output_port);
1682 }
1683 }
1684
1685 /* Otherwise choose the minimum non-local MAC address among all of the
1686 * interfaces. */
1687 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1688 uint8_t iface_ea[ETH_ADDR_LEN];
1689 struct iface *candidate;
1690 struct iface *iface;
1691
1692 /* Mirror output ports don't participate. */
1693 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1694 continue;
1695 }
1696
1697 /* Choose the MAC address to represent the port. */
1698 iface = NULL;
1699 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1700 /* Find the interface with this Ethernet address (if any) so that
1701 * we can provide the correct devname to the caller. */
1702 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1703 uint8_t candidate_ea[ETH_ADDR_LEN];
1704 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1705 && eth_addr_equals(iface_ea, candidate_ea)) {
1706 iface = candidate;
1707 }
1708 }
1709 } else {
1710 /* Choose the interface whose MAC address will represent the port.
1711 * The Linux kernel bonding code always chooses the MAC address of
1712 * the first slave added to a bond, and the Fedora networking
1713 * scripts always add slaves to a bond in alphabetical order, so
1714 * for compatibility we choose the interface with the name that is
1715 * first in alphabetical order. */
1716 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1717 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1718 iface = candidate;
1719 }
1720 }
1721
1722 /* The local port doesn't count (since we're trying to choose its
1723 * MAC address anyway). */
1724 if (iface->ofp_port == OFPP_LOCAL) {
1725 continue;
1726 }
1727
1728 /* Grab MAC. */
1729 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1730 if (error) {
1731 continue;
1732 }
1733 }
1734
1735 /* Compare against our current choice. */
1736 if (!eth_addr_is_multicast(iface_ea) &&
1737 !eth_addr_is_local(iface_ea) &&
1738 !eth_addr_is_reserved(iface_ea) &&
1739 !eth_addr_is_zero(iface_ea) &&
1740 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1741 {
1742 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1743 *hw_addr_iface = iface;
1744 found_addr = true;
1745 }
1746 }
1747
1748 if (!found_addr) {
1749 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1750 *hw_addr_iface = NULL;
1751 }
1752
1753 hmapx_destroy(&mirror_output_ports);
1754 }
1755
1756 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1757 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1758 * an interface on 'br', then that interface must be passed in as
1759 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1760 * 'hw_addr_iface' must be passed in as a null pointer. */
1761 static uint64_t
1762 bridge_pick_datapath_id(struct bridge *br,
1763 const uint8_t bridge_ea[ETH_ADDR_LEN],
1764 struct iface *hw_addr_iface)
1765 {
1766 /*
1767 * The procedure for choosing a bridge MAC address will, in the most
1768 * ordinary case, also choose a unique MAC that we can use as a datapath
1769 * ID. In some special cases, though, multiple bridges will end up with
1770 * the same MAC address. This is OK for the bridges, but it will confuse
1771 * the OpenFlow controller, because each datapath needs a unique datapath
1772 * ID.
1773 *
1774 * Datapath IDs must be unique. It is also very desirable that they be
1775 * stable from one run to the next, so that policy set on a datapath
1776 * "sticks".
1777 */
1778 const char *datapath_id;
1779 uint64_t dpid;
1780
1781 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1782 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1783 return dpid;
1784 }
1785
1786 if (!hw_addr_iface) {
1787 /*
1788 * A purely internal bridge, that is, one that has no non-virtual
1789 * network devices on it at all, is difficult because it has no
1790 * natural unique identifier at all.
1791 *
1792 * When the host is a XenServer, we handle this case by hashing the
1793 * host's UUID with the name of the bridge. Names of bridges are
1794 * persistent across XenServer reboots, although they can be reused if
1795 * an internal network is destroyed and then a new one is later
1796 * created, so this is fairly effective.
1797 *
1798 * When the host is not a XenServer, we punt by using a random MAC
1799 * address on each run.
1800 */
1801 const char *host_uuid = xenserver_get_host_uuid();
1802 if (host_uuid) {
1803 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1804 dpid = dpid_from_hash(combined, strlen(combined));
1805 free(combined);
1806 return dpid;
1807 }
1808 }
1809
1810 return eth_addr_to_uint64(bridge_ea);
1811 }
1812
1813 static uint64_t
1814 dpid_from_hash(const void *data, size_t n)
1815 {
1816 uint8_t hash[SHA1_DIGEST_SIZE];
1817
1818 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1819 sha1_bytes(data, n, hash);
1820 eth_addr_mark_random(hash);
1821 return eth_addr_to_uint64(hash);
1822 }
1823
1824 static void
1825 iface_refresh_status(struct iface *iface)
1826 {
1827 struct smap smap;
1828
1829 enum netdev_features current;
1830 int64_t bps;
1831 int mtu;
1832 int64_t mtu_64;
1833 uint8_t mac[ETH_ADDR_LEN];
1834 int64_t ifindex64;
1835 int error;
1836
1837 if (iface_is_synthetic(iface)) {
1838 return;
1839 }
1840
1841 smap_init(&smap);
1842
1843 if (!netdev_get_status(iface->netdev, &smap)) {
1844 ovsrec_interface_set_status(iface->cfg, &smap);
1845 } else {
1846 ovsrec_interface_set_status(iface->cfg, NULL);
1847 }
1848
1849 smap_destroy(&smap);
1850
1851 error = netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1852 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1853 if (bps) {
1854 ovsrec_interface_set_duplex(iface->cfg,
1855 netdev_features_is_full_duplex(current)
1856 ? "full" : "half");
1857 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1858 } else {
1859 ovsrec_interface_set_duplex(iface->cfg, NULL);
1860 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1861 }
1862
1863 error = netdev_get_mtu(iface->netdev, &mtu);
1864 if (!error) {
1865 mtu_64 = mtu;
1866 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1867 } else {
1868 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1869 }
1870
1871 error = netdev_get_etheraddr(iface->netdev, mac);
1872 if (!error) {
1873 char mac_string[32];
1874
1875 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1876 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1877 } else {
1878 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1879 }
1880
1881 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1882 * if there is no valid ifindex number. */
1883 ifindex64 = netdev_get_ifindex(iface->netdev);
1884 if (ifindex64 < 0) {
1885 ifindex64 = 0;
1886 }
1887 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1888 }
1889
1890 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1891 * synthetic. */
1892 static void
1893 iface_refresh_cfm_stats(struct iface *iface)
1894 {
1895 const struct ovsrec_interface *cfg = iface->cfg;
1896 struct ofproto_cfm_status status;
1897
1898 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1899 iface->ofp_port, &status)) {
1900 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1901 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1902 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1903 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1904 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1905 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1906 } else {
1907 const char *reasons[CFM_FAULT_N_REASONS];
1908 int64_t cfm_health = status.health;
1909 int64_t cfm_flap_count = status.flap_count;
1910 bool faulted = status.faults != 0;
1911 size_t i, j;
1912
1913 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1914
1915 j = 0;
1916 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1917 int reason = 1 << i;
1918 if (status.faults & reason) {
1919 reasons[j++] = cfm_fault_reason_to_str(reason);
1920 }
1921 }
1922 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1923
1924 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1925
1926 if (status.remote_opstate >= 0) {
1927 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1928 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1929 } else {
1930 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1931 }
1932
1933 ovsrec_interface_set_cfm_remote_mpids(cfg,
1934 (const int64_t *)status.rmps,
1935 status.n_rmps);
1936 if (cfm_health >= 0) {
1937 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1938 } else {
1939 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1940 }
1941
1942 free(status.rmps);
1943 }
1944 }
1945
1946 static void
1947 iface_refresh_stats(struct iface *iface)
1948 {
1949 #define IFACE_STATS \
1950 IFACE_STAT(rx_packets, "rx_packets") \
1951 IFACE_STAT(tx_packets, "tx_packets") \
1952 IFACE_STAT(rx_bytes, "rx_bytes") \
1953 IFACE_STAT(tx_bytes, "tx_bytes") \
1954 IFACE_STAT(rx_dropped, "rx_dropped") \
1955 IFACE_STAT(tx_dropped, "tx_dropped") \
1956 IFACE_STAT(rx_errors, "rx_errors") \
1957 IFACE_STAT(tx_errors, "tx_errors") \
1958 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1959 IFACE_STAT(rx_over_errors, "rx_over_err") \
1960 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1961 IFACE_STAT(collisions, "collisions")
1962
1963 #define IFACE_STAT(MEMBER, NAME) + 1
1964 enum { N_IFACE_STATS = IFACE_STATS };
1965 #undef IFACE_STAT
1966 int64_t values[N_IFACE_STATS];
1967 char *keys[N_IFACE_STATS];
1968 int n;
1969
1970 struct netdev_stats stats;
1971
1972 if (iface_is_synthetic(iface)) {
1973 return;
1974 }
1975
1976 /* Intentionally ignore return value, since errors will set 'stats' to
1977 * all-1s, and we will deal with that correctly below. */
1978 netdev_get_stats(iface->netdev, &stats);
1979
1980 /* Copy statistics into keys[] and values[]. */
1981 n = 0;
1982 #define IFACE_STAT(MEMBER, NAME) \
1983 if (stats.MEMBER != UINT64_MAX) { \
1984 keys[n] = NAME; \
1985 values[n] = stats.MEMBER; \
1986 n++; \
1987 }
1988 IFACE_STATS;
1989 #undef IFACE_STAT
1990 ovs_assert(n <= N_IFACE_STATS);
1991
1992 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1993 #undef IFACE_STATS
1994 }
1995
1996 static void
1997 br_refresh_stp_status(struct bridge *br)
1998 {
1999 struct smap smap = SMAP_INITIALIZER(&smap);
2000 struct ofproto *ofproto = br->ofproto;
2001 struct ofproto_stp_status status;
2002
2003 if (ofproto_get_stp_status(ofproto, &status)) {
2004 return;
2005 }
2006
2007 if (!status.enabled) {
2008 ovsrec_bridge_set_status(br->cfg, NULL);
2009 return;
2010 }
2011
2012 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2013 STP_ID_ARGS(status.bridge_id));
2014 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2015 STP_ID_ARGS(status.designated_root));
2016 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2017
2018 ovsrec_bridge_set_status(br->cfg, &smap);
2019 smap_destroy(&smap);
2020 }
2021
2022 static void
2023 port_refresh_stp_status(struct port *port)
2024 {
2025 struct ofproto *ofproto = port->bridge->ofproto;
2026 struct iface *iface;
2027 struct ofproto_port_stp_status status;
2028 struct smap smap;
2029
2030 if (port_is_synthetic(port)) {
2031 return;
2032 }
2033
2034 /* STP doesn't currently support bonds. */
2035 if (!list_is_singleton(&port->ifaces)) {
2036 ovsrec_port_set_status(port->cfg, NULL);
2037 return;
2038 }
2039
2040 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2041 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2042 return;
2043 }
2044
2045 if (!status.enabled) {
2046 ovsrec_port_set_status(port->cfg, NULL);
2047 return;
2048 }
2049
2050 /* Set Status column. */
2051 smap_init(&smap);
2052 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2053 smap_add(&smap, "stp_state", stp_state_name(status.state));
2054 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2055 smap_add(&smap, "stp_role", stp_role_name(status.role));
2056 ovsrec_port_set_status(port->cfg, &smap);
2057 smap_destroy(&smap);
2058 }
2059
2060 static void
2061 port_refresh_stp_stats(struct port *port)
2062 {
2063 struct ofproto *ofproto = port->bridge->ofproto;
2064 struct iface *iface;
2065 struct ofproto_port_stp_stats stats;
2066 char *keys[3];
2067 int64_t int_values[3];
2068
2069 if (port_is_synthetic(port)) {
2070 return;
2071 }
2072
2073 /* STP doesn't currently support bonds. */
2074 if (!list_is_singleton(&port->ifaces)) {
2075 return;
2076 }
2077
2078 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2079 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2080 return;
2081 }
2082
2083 if (!stats.enabled) {
2084 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2085 return;
2086 }
2087
2088 /* Set Statistics column. */
2089 keys[0] = "stp_tx_count";
2090 int_values[0] = stats.tx_count;
2091 keys[1] = "stp_rx_count";
2092 int_values[1] = stats.rx_count;
2093 keys[2] = "stp_error_count";
2094 int_values[2] = stats.error_count;
2095
2096 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2097 ARRAY_SIZE(int_values));
2098 }
2099
2100 static bool
2101 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2102 {
2103 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2104 }
2105
2106 static void
2107 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2108 {
2109 bool enable = enable_system_stats(cfg);
2110
2111 system_stats_enable(enable);
2112 if (!enable) {
2113 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2114 }
2115 }
2116
2117 static void
2118 run_system_stats(void)
2119 {
2120 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2121 struct smap *stats;
2122
2123 stats = system_stats_run();
2124 if (stats && cfg) {
2125 struct ovsdb_idl_txn *txn;
2126 struct ovsdb_datum datum;
2127
2128 txn = ovsdb_idl_txn_create(idl);
2129 ovsdb_datum_from_smap(&datum, stats);
2130 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2131 &datum);
2132 ovsdb_idl_txn_commit(txn);
2133 ovsdb_idl_txn_destroy(txn);
2134
2135 free(stats);
2136 }
2137 }
2138
2139 static inline const char *
2140 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2141 {
2142 switch (role) {
2143 case OFPCR12_ROLE_EQUAL:
2144 return "other";
2145 case OFPCR12_ROLE_MASTER:
2146 return "master";
2147 case OFPCR12_ROLE_SLAVE:
2148 return "slave";
2149 case OFPCR12_ROLE_NOCHANGE:
2150 default:
2151 return "*** INVALID ROLE ***";
2152 }
2153 }
2154
2155 static void
2156 refresh_controller_status(void)
2157 {
2158 struct bridge *br;
2159 struct shash info;
2160 const struct ovsrec_controller *cfg;
2161
2162 shash_init(&info);
2163
2164 /* Accumulate status for controllers on all bridges. */
2165 HMAP_FOR_EACH (br, node, &all_bridges) {
2166 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2167 }
2168
2169 /* Update each controller in the database with current status. */
2170 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2171 struct ofproto_controller_info *cinfo =
2172 shash_find_data(&info, cfg->target);
2173
2174 if (cinfo) {
2175 struct smap smap = SMAP_INITIALIZER(&smap);
2176 const char **values = cinfo->pairs.values;
2177 const char **keys = cinfo->pairs.keys;
2178 size_t i;
2179
2180 for (i = 0; i < cinfo->pairs.n; i++) {
2181 smap_add(&smap, keys[i], values[i]);
2182 }
2183
2184 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2185 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2186 cinfo->role));
2187 ovsrec_controller_set_status(cfg, &smap);
2188 smap_destroy(&smap);
2189 } else {
2190 ovsrec_controller_set_is_connected(cfg, false);
2191 ovsrec_controller_set_role(cfg, NULL);
2192 ovsrec_controller_set_status(cfg, NULL);
2193 }
2194 }
2195
2196 ofproto_free_ofproto_controller_info(&info);
2197 }
2198 \f
2199 /* "Instant" stats.
2200 *
2201 * Some information in the database must be kept as up-to-date as possible to
2202 * allow controllers to respond rapidly to network outages. We call these
2203 * statistics "instant" stats.
2204 *
2205 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2206 * assuming that they've changed. The only means we have to determine whether
2207 * they have changed are:
2208 *
2209 * - Try to commit changes to the database. If nothing changed, then
2210 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2211 * value.
2212 *
2213 * - instant_stats_run() is called late in the run loop, after anything that
2214 * might change any of the instant stats.
2215 *
2216 * We use these two facts together to avoid waking the process up every
2217 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2218 */
2219
2220 /* Minimum interval between writing updates to the instant stats to the
2221 * database. */
2222 #define INSTANT_INTERVAL_MSEC 100
2223
2224 /* Current instant stats database transaction, NULL if there is no ongoing
2225 * transaction. */
2226 static struct ovsdb_idl_txn *instant_txn;
2227
2228 /* Next time (in msec on monotonic clock) at which we will update the instant
2229 * stats. */
2230 static long long int instant_next_txn = LLONG_MIN;
2231
2232 /* True if the run loop has run since we last saw that the instant stats were
2233 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2234 * to refresh the instant stats. */
2235 static bool instant_stats_could_have_changed;
2236
2237 static void
2238 instant_stats_run(void)
2239 {
2240 enum ovsdb_idl_txn_status status;
2241
2242 instant_stats_could_have_changed = true;
2243
2244 if (!instant_txn) {
2245 struct bridge *br;
2246
2247 if (time_msec() < instant_next_txn) {
2248 return;
2249 }
2250 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2251
2252 instant_txn = ovsdb_idl_txn_create(idl);
2253 HMAP_FOR_EACH (br, node, &all_bridges) {
2254 struct iface *iface;
2255 struct port *port;
2256
2257 br_refresh_stp_status(br);
2258
2259 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2260 port_refresh_stp_status(port);
2261 }
2262
2263 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2264 enum netdev_flags flags;
2265 struct smap smap;
2266 const char *link_state;
2267 int64_t link_resets;
2268 int current, error;
2269
2270 if (iface_is_synthetic(iface)) {
2271 continue;
2272 }
2273
2274 current = ofproto_port_is_lacp_current(br->ofproto,
2275 iface->ofp_port);
2276 if (current >= 0) {
2277 bool bl = current;
2278 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2279 } else {
2280 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2281 }
2282
2283 error = netdev_get_flags(iface->netdev, &flags);
2284 if (!error) {
2285 const char *state = flags & NETDEV_UP ? "up" : "down";
2286 ovsrec_interface_set_admin_state(iface->cfg, state);
2287 } else {
2288 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2289 }
2290
2291 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2292 ovsrec_interface_set_link_state(iface->cfg, link_state);
2293
2294 link_resets = netdev_get_carrier_resets(iface->netdev);
2295 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2296
2297 iface_refresh_cfm_stats(iface);
2298
2299 smap_init(&smap);
2300 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2301 &smap);
2302 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2303 smap_destroy(&smap);
2304 }
2305 }
2306 }
2307
2308 status = ovsdb_idl_txn_commit(instant_txn);
2309 if (status != TXN_INCOMPLETE) {
2310 ovsdb_idl_txn_destroy(instant_txn);
2311 instant_txn = NULL;
2312 }
2313 if (status == TXN_UNCHANGED) {
2314 instant_stats_could_have_changed = false;
2315 }
2316 }
2317
2318 static void
2319 instant_stats_wait(void)
2320 {
2321 if (instant_txn) {
2322 ovsdb_idl_txn_wait(instant_txn);
2323 } else if (instant_stats_could_have_changed) {
2324 poll_timer_wait_until(instant_next_txn);
2325 }
2326 }
2327 \f
2328 /* Performs periodic activity required by bridges that needs to be done with
2329 * the least possible latency.
2330 *
2331 * It makes sense to call this function a couple of times per poll loop, to
2332 * provide a significant performance boost on some benchmarks with ofprotos
2333 * that use the ofproto-dpif implementation. */
2334 void
2335 bridge_run_fast(void)
2336 {
2337 struct sset types;
2338 const char *type;
2339 struct bridge *br;
2340
2341 sset_init(&types);
2342 ofproto_enumerate_types(&types);
2343 SSET_FOR_EACH (type, &types) {
2344 ofproto_type_run_fast(type);
2345 }
2346 sset_destroy(&types);
2347
2348 HMAP_FOR_EACH (br, node, &all_bridges) {
2349 ofproto_run_fast(br->ofproto);
2350 }
2351 }
2352
2353 void
2354 bridge_run(void)
2355 {
2356 static struct ovsrec_open_vswitch null_cfg;
2357 const struct ovsrec_open_vswitch *cfg;
2358 struct ovsdb_idl_txn *reconf_txn = NULL;
2359 struct sset types;
2360 const char *type;
2361
2362 bool vlan_splinters_changed;
2363 struct bridge *br;
2364
2365 ovsrec_open_vswitch_init(&null_cfg);
2366
2367 /* (Re)configure if necessary. */
2368 if (!reconfiguring) {
2369 ovsdb_idl_run(idl);
2370
2371 if (ovsdb_idl_is_lock_contended(idl)) {
2372 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2373 struct bridge *br, *next_br;
2374
2375 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2376 "disabling this process (pid %ld) until it goes away",
2377 (long int) getpid());
2378
2379 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2380 bridge_destroy(br);
2381 }
2382 /* Since we will not be running system_stats_run() in this process
2383 * with the current situation of multiple ovs-vswitchd daemons,
2384 * disable system stats collection. */
2385 system_stats_enable(false);
2386 return;
2387 } else if (!ovsdb_idl_has_lock(idl)) {
2388 return;
2389 }
2390 }
2391 cfg = ovsrec_open_vswitch_first(idl);
2392
2393 /* Initialize the ofproto library. This only needs to run once, but
2394 * it must be done after the configuration is set. If the
2395 * initialization has already occurred, bridge_init_ofproto()
2396 * returns immediately. */
2397 bridge_init_ofproto(cfg);
2398
2399 /* Once the value of flow-restore-wait is false, we no longer should
2400 * check its value from the database. */
2401 if (cfg && ofproto_get_flow_restore_wait()) {
2402 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2403 "flow-restore-wait", false));
2404 }
2405
2406 /* Let each datapath type do the work that it needs to do. */
2407 sset_init(&types);
2408 ofproto_enumerate_types(&types);
2409 SSET_FOR_EACH (type, &types) {
2410 ofproto_type_run(type);
2411 }
2412 sset_destroy(&types);
2413
2414 /* Let each bridge do the work that it needs to do. */
2415 HMAP_FOR_EACH (br, node, &all_bridges) {
2416 ofproto_run(br->ofproto);
2417 }
2418
2419 /* Re-configure SSL. We do this on every trip through the main loop,
2420 * instead of just when the database changes, because the contents of the
2421 * key and certificate files can change without the database changing.
2422 *
2423 * We do this before bridge_reconfigure() because that function might
2424 * initiate SSL connections and thus requires SSL to be configured. */
2425 if (cfg && cfg->ssl) {
2426 const struct ovsrec_ssl *ssl = cfg->ssl;
2427
2428 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2429 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2430 }
2431
2432 if (!reconfiguring) {
2433 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2434 * usage has changed. */
2435 vlan_splinters_changed = false;
2436 if (vlan_splinters_enabled_anywhere) {
2437 HMAP_FOR_EACH (br, node, &all_bridges) {
2438 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2439 vlan_splinters_changed = true;
2440 break;
2441 }
2442 }
2443 }
2444
2445 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2446 idl_seqno = ovsdb_idl_get_seqno(idl);
2447 if (cfg) {
2448 reconf_txn = ovsdb_idl_txn_create(idl);
2449 bridge_reconfigure(cfg);
2450 } else {
2451 /* We still need to reconfigure to avoid dangling pointers to
2452 * now-destroyed ovsrec structures inside bridge data. */
2453 bridge_reconfigure(&null_cfg);
2454 }
2455 }
2456 }
2457
2458 if (reconfiguring) {
2459 if (!reconf_txn) {
2460 reconf_txn = ovsdb_idl_txn_create(idl);
2461 }
2462
2463 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2464 reconfiguring = false;
2465
2466 if (cfg) {
2467 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2468 }
2469
2470 /* If we are completing our initial configuration for this run
2471 * of ovs-vswitchd, then keep the transaction around to monitor
2472 * it for completion. */
2473 if (!initial_config_done) {
2474 initial_config_done = true;
2475 daemonize_txn = reconf_txn;
2476 reconf_txn = NULL;
2477 }
2478 }
2479 }
2480
2481 if (reconf_txn) {
2482 ovsdb_idl_txn_commit(reconf_txn);
2483 ovsdb_idl_txn_destroy(reconf_txn);
2484 reconf_txn = NULL;
2485 }
2486
2487 if (daemonize_txn) {
2488 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2489 if (status != TXN_INCOMPLETE) {
2490 ovsdb_idl_txn_destroy(daemonize_txn);
2491 daemonize_txn = NULL;
2492
2493 /* ovs-vswitchd has completed initialization, so allow the
2494 * process that forked us to exit successfully. */
2495 daemonize_complete();
2496
2497 vlog_enable_async();
2498
2499 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2500 }
2501 }
2502
2503 /* Refresh interface and mirror stats if necessary. */
2504 if (time_msec() >= iface_stats_timer) {
2505 if (cfg) {
2506 struct ovsdb_idl_txn *txn;
2507
2508 txn = ovsdb_idl_txn_create(idl);
2509 HMAP_FOR_EACH (br, node, &all_bridges) {
2510 struct port *port;
2511 struct mirror *m;
2512
2513 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2514 struct iface *iface;
2515
2516 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2517 iface_refresh_stats(iface);
2518 iface_refresh_status(iface);
2519 }
2520
2521 port_refresh_stp_stats(port);
2522 }
2523
2524 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2525 mirror_refresh_stats(m);
2526 }
2527
2528 }
2529 refresh_controller_status();
2530 ovsdb_idl_txn_commit(txn);
2531 ovsdb_idl_txn_destroy(txn); /* XXX */
2532 }
2533
2534 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2535 }
2536
2537 run_system_stats();
2538 instant_stats_run();
2539 }
2540
2541 void
2542 bridge_wait(void)
2543 {
2544 struct sset types;
2545 const char *type;
2546
2547 ovsdb_idl_wait(idl);
2548 if (daemonize_txn) {
2549 ovsdb_idl_txn_wait(daemonize_txn);
2550 }
2551
2552 if (reconfiguring) {
2553 poll_immediate_wake();
2554 }
2555
2556 sset_init(&types);
2557 ofproto_enumerate_types(&types);
2558 SSET_FOR_EACH (type, &types) {
2559 ofproto_type_wait(type);
2560 }
2561 sset_destroy(&types);
2562
2563 if (!hmap_is_empty(&all_bridges)) {
2564 struct bridge *br;
2565
2566 HMAP_FOR_EACH (br, node, &all_bridges) {
2567 ofproto_wait(br->ofproto);
2568 }
2569 poll_timer_wait_until(iface_stats_timer);
2570 }
2571
2572 system_stats_wait();
2573 instant_stats_wait();
2574 }
2575
2576 /* Adds some memory usage statistics for bridges into 'usage', for use with
2577 * memory_report(). */
2578 void
2579 bridge_get_memory_usage(struct simap *usage)
2580 {
2581 struct bridge *br;
2582
2583 HMAP_FOR_EACH (br, node, &all_bridges) {
2584 ofproto_get_memory_usage(br->ofproto, usage);
2585 }
2586 }
2587 \f
2588 /* QoS unixctl user interface functions. */
2589
2590 struct qos_unixctl_show_cbdata {
2591 struct ds *ds;
2592 struct iface *iface;
2593 };
2594
2595 static void
2596 qos_unixctl_show_queue(unsigned int queue_id,
2597 const struct smap *details,
2598 struct iface *iface,
2599 struct ds *ds)
2600 {
2601 struct netdev_queue_stats stats;
2602 struct smap_node *node;
2603 int error;
2604
2605 ds_put_cstr(ds, "\n");
2606 if (queue_id) {
2607 ds_put_format(ds, "Queue %u:\n", queue_id);
2608 } else {
2609 ds_put_cstr(ds, "Default:\n");
2610 }
2611
2612 SMAP_FOR_EACH (node, details) {
2613 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2614 }
2615
2616 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2617 if (!error) {
2618 if (stats.tx_packets != UINT64_MAX) {
2619 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2620 }
2621
2622 if (stats.tx_bytes != UINT64_MAX) {
2623 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2624 }
2625
2626 if (stats.tx_errors != UINT64_MAX) {
2627 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2628 }
2629 } else {
2630 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2631 queue_id, ovs_strerror(error));
2632 }
2633 }
2634
2635 static void
2636 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2637 const char *argv[], void *aux OVS_UNUSED)
2638 {
2639 struct ds ds = DS_EMPTY_INITIALIZER;
2640 struct smap smap = SMAP_INITIALIZER(&smap);
2641 struct iface *iface;
2642 const char *type;
2643 struct smap_node *node;
2644
2645 iface = iface_find(argv[1]);
2646 if (!iface) {
2647 unixctl_command_reply_error(conn, "no such interface");
2648 return;
2649 }
2650
2651 netdev_get_qos(iface->netdev, &type, &smap);
2652
2653 if (*type != '\0') {
2654 struct netdev_queue_dump dump;
2655 struct smap details;
2656 unsigned int queue_id;
2657
2658 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2659
2660 SMAP_FOR_EACH (node, &smap) {
2661 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2662 }
2663
2664 smap_init(&details);
2665 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2666 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2667 }
2668 smap_destroy(&details);
2669
2670 unixctl_command_reply(conn, ds_cstr(&ds));
2671 } else {
2672 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2673 unixctl_command_reply_error(conn, ds_cstr(&ds));
2674 }
2675
2676 smap_destroy(&smap);
2677 ds_destroy(&ds);
2678 }
2679 \f
2680 /* Bridge reconfiguration functions. */
2681 static void
2682 bridge_create(const struct ovsrec_bridge *br_cfg)
2683 {
2684 struct bridge *br;
2685
2686 ovs_assert(!bridge_lookup(br_cfg->name));
2687 br = xzalloc(sizeof *br);
2688
2689 br->name = xstrdup(br_cfg->name);
2690 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2691 br->cfg = br_cfg;
2692
2693 /* Derive the default Ethernet address from the bridge's UUID. This should
2694 * be unique and it will be stable between ovs-vswitchd runs. */
2695 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2696 eth_addr_mark_random(br->default_ea);
2697
2698 hmap_init(&br->ports);
2699 hmap_init(&br->ifaces);
2700 hmap_init(&br->iface_by_name);
2701 hmap_init(&br->mirrors);
2702
2703 hmap_init(&br->if_cfg_todo);
2704 list_init(&br->ofpp_garbage);
2705
2706 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2707 }
2708
2709 static void
2710 bridge_destroy(struct bridge *br)
2711 {
2712 if (br) {
2713 struct mirror *mirror, *next_mirror;
2714 struct port *port, *next_port;
2715 struct if_cfg *if_cfg, *next_if_cfg;
2716 struct ofpp_garbage *garbage, *next_garbage;
2717
2718 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2719 port_destroy(port);
2720 }
2721 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2722 mirror_destroy(mirror);
2723 }
2724 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2725 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2726 free(if_cfg);
2727 }
2728 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2729 &br->ofpp_garbage) {
2730 list_remove(&garbage->list_node);
2731 free(garbage);
2732 }
2733
2734 hmap_remove(&all_bridges, &br->node);
2735 ofproto_destroy(br->ofproto);
2736 hmap_destroy(&br->ifaces);
2737 hmap_destroy(&br->ports);
2738 hmap_destroy(&br->iface_by_name);
2739 hmap_destroy(&br->mirrors);
2740 hmap_destroy(&br->if_cfg_todo);
2741 free(br->name);
2742 free(br->type);
2743 free(br);
2744 }
2745 }
2746
2747 static struct bridge *
2748 bridge_lookup(const char *name)
2749 {
2750 struct bridge *br;
2751
2752 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2753 if (!strcmp(br->name, name)) {
2754 return br;
2755 }
2756 }
2757 return NULL;
2758 }
2759
2760 /* Handle requests for a listing of all flows known by the OpenFlow
2761 * stack, including those normally hidden. */
2762 static void
2763 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2764 const char *argv[], void *aux OVS_UNUSED)
2765 {
2766 struct bridge *br;
2767 struct ds results;
2768
2769 br = bridge_lookup(argv[1]);
2770 if (!br) {
2771 unixctl_command_reply_error(conn, "Unknown bridge");
2772 return;
2773 }
2774
2775 ds_init(&results);
2776 ofproto_get_all_flows(br->ofproto, &results);
2777
2778 unixctl_command_reply(conn, ds_cstr(&results));
2779 ds_destroy(&results);
2780 }
2781
2782 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2783 * connections and reconnect. If BRIDGE is not specified, then all bridges
2784 * drop their controller connections and reconnect. */
2785 static void
2786 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2787 const char *argv[], void *aux OVS_UNUSED)
2788 {
2789 struct bridge *br;
2790 if (argc > 1) {
2791 br = bridge_lookup(argv[1]);
2792 if (!br) {
2793 unixctl_command_reply_error(conn, "Unknown bridge");
2794 return;
2795 }
2796 ofproto_reconnect_controllers(br->ofproto);
2797 } else {
2798 HMAP_FOR_EACH (br, node, &all_bridges) {
2799 ofproto_reconnect_controllers(br->ofproto);
2800 }
2801 }
2802 unixctl_command_reply(conn, NULL);
2803 }
2804
2805 static size_t
2806 bridge_get_controllers(const struct bridge *br,
2807 struct ovsrec_controller ***controllersp)
2808 {
2809 struct ovsrec_controller **controllers;
2810 size_t n_controllers;
2811
2812 controllers = br->cfg->controller;
2813 n_controllers = br->cfg->n_controller;
2814
2815 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2816 controllers = NULL;
2817 n_controllers = 0;
2818 }
2819
2820 if (controllersp) {
2821 *controllersp = controllers;
2822 }
2823 return n_controllers;
2824 }
2825
2826 static void
2827 bridge_queue_if_cfg(struct bridge *br,
2828 const struct ovsrec_interface *cfg,
2829 const struct ovsrec_port *parent)
2830 {
2831 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2832
2833 if_cfg->cfg = cfg;
2834 if_cfg->parent = parent;
2835 if_cfg->ofport = iface_pick_ofport(cfg);
2836 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2837 hash_string(if_cfg->cfg->name, 0));
2838 }
2839
2840 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2841 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2842 * 'br' needs to complete its configuration. */
2843 static void
2844 bridge_add_del_ports(struct bridge *br,
2845 const unsigned long int *splinter_vlans)
2846 {
2847 struct shash_node *port_node;
2848 struct port *port, *next;
2849 struct shash new_ports;
2850 size_t i;
2851
2852 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2853
2854 /* Collect new ports. */
2855 shash_init(&new_ports);
2856 for (i = 0; i < br->cfg->n_ports; i++) {
2857 const char *name = br->cfg->ports[i]->name;
2858 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2859 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2860 br->name, name);
2861 }
2862 }
2863 if (bridge_get_controllers(br, NULL)
2864 && !shash_find(&new_ports, br->name)) {
2865 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2866 br->name, br->name);
2867
2868 ovsrec_interface_init(&br->synth_local_iface);
2869 ovsrec_port_init(&br->synth_local_port);
2870
2871 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2872 br->synth_local_port.n_interfaces = 1;
2873 br->synth_local_port.name = br->name;
2874
2875 br->synth_local_iface.name = br->name;
2876 br->synth_local_iface.type = "internal";
2877
2878 br->synth_local_ifacep = &br->synth_local_iface;
2879
2880 shash_add(&new_ports, br->name, &br->synth_local_port);
2881 }
2882
2883 if (splinter_vlans) {
2884 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2885 }
2886
2887 /* Get rid of deleted ports.
2888 * Get rid of deleted interfaces on ports that still exist. */
2889 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2890 port->cfg = shash_find_data(&new_ports, port->name);
2891 if (!port->cfg) {
2892 port_destroy(port);
2893 } else {
2894 port_del_ifaces(port);
2895 }
2896 }
2897
2898 /* Update iface->cfg and iface->type in interfaces that still exist.
2899 * Add new interfaces to creation queue. */
2900 SHASH_FOR_EACH (port_node, &new_ports) {
2901 const struct ovsrec_port *port = port_node->data;
2902 size_t i;
2903
2904 for (i = 0; i < port->n_interfaces; i++) {
2905 const struct ovsrec_interface *cfg = port->interfaces[i];
2906 struct iface *iface = iface_lookup(br, cfg->name);
2907 const char *type = iface_get_type(cfg, br->cfg);
2908
2909 if (iface) {
2910 iface->cfg = cfg;
2911 iface->type = type;
2912 } else if (!strcmp(type, "null")) {
2913 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2914 " may be removed in February 2013. Please email"
2915 " dev@openvswitch.org with concerns.",
2916 cfg->name);
2917 } else {
2918 bridge_queue_if_cfg(br, cfg, port);
2919 }
2920 }
2921 }
2922
2923 shash_destroy(&new_ports);
2924 }
2925
2926 /* Initializes 'oc' appropriately as a management service controller for
2927 * 'br'.
2928 *
2929 * The caller must free oc->target when it is no longer needed. */
2930 static void
2931 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2932 struct ofproto_controller *oc)
2933 {
2934 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2935 oc->max_backoff = 0;
2936 oc->probe_interval = 60;
2937 oc->band = OFPROTO_OUT_OF_BAND;
2938 oc->rate_limit = 0;
2939 oc->burst_limit = 0;
2940 oc->enable_async_msgs = true;
2941 }
2942
2943 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2944 static void
2945 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2946 struct ofproto_controller *oc)
2947 {
2948 int dscp;
2949
2950 oc->target = c->target;
2951 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2952 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2953 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2954 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2955 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2956 oc->burst_limit = (c->controller_burst_limit
2957 ? *c->controller_burst_limit : 0);
2958 oc->enable_async_msgs = (!c->enable_async_messages
2959 || *c->enable_async_messages);
2960 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2961 if (dscp < 0 || dscp > 63) {
2962 dscp = DSCP_DEFAULT;
2963 }
2964 oc->dscp = dscp;
2965 }
2966
2967 /* Configures the IP stack for 'br''s local interface properly according to the
2968 * configuration in 'c'. */
2969 static void
2970 bridge_configure_local_iface_netdev(struct bridge *br,
2971 struct ovsrec_controller *c)
2972 {
2973 struct netdev *netdev;
2974 struct in_addr mask, gateway;
2975
2976 struct iface *local_iface;
2977 struct in_addr ip;
2978
2979 /* If there's no local interface or no IP address, give up. */
2980 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2981 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2982 return;
2983 }
2984
2985 /* Bring up the local interface. */
2986 netdev = local_iface->netdev;
2987 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2988
2989 /* Configure the IP address and netmask. */
2990 if (!c->local_netmask
2991 || !inet_aton(c->local_netmask, &mask)
2992 || !mask.s_addr) {
2993 mask.s_addr = guess_netmask(ip.s_addr);
2994 }
2995 if (!netdev_set_in4(netdev, ip, mask)) {
2996 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2997 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2998 }
2999
3000 /* Configure the default gateway. */
3001 if (c->local_gateway
3002 && inet_aton(c->local_gateway, &gateway)
3003 && gateway.s_addr) {
3004 if (!netdev_add_router(netdev, gateway)) {
3005 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3006 br->name, IP_ARGS(gateway.s_addr));
3007 }
3008 }
3009 }
3010
3011 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3012 * in either string are treated as equal to any number of slashes in the other,
3013 * e.g. "x///y" is equal to "x/y".
3014 *
3015 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3016 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3017 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3018 * 'b' against a prefix of 'a'.
3019 */
3020 static bool
3021 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3022 {
3023 const char *b_start = b;
3024 for (;;) {
3025 if (b - b_start >= b_stoplen) {
3026 return true;
3027 } else if (*a != *b) {
3028 return false;
3029 } else if (*a == '/') {
3030 a += strspn(a, "/");
3031 b += strspn(b, "/");
3032 } else if (*a == '\0') {
3033 return true;
3034 } else {
3035 a++;
3036 b++;
3037 }
3038 }
3039 }
3040
3041 static void
3042 bridge_configure_remotes(struct bridge *br,
3043 const struct sockaddr_in *managers, size_t n_managers)
3044 {
3045 bool disable_in_band;
3046
3047 struct ovsrec_controller **controllers;
3048 size_t n_controllers;
3049
3050 enum ofproto_fail_mode fail_mode;
3051
3052 struct ofproto_controller *ocs;
3053 size_t n_ocs;
3054 size_t i;
3055
3056 /* Check if we should disable in-band control on this bridge. */
3057 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3058 false);
3059
3060 /* Set OpenFlow queue ID for in-band control. */
3061 ofproto_set_in_band_queue(br->ofproto,
3062 smap_get_int(&br->cfg->other_config,
3063 "in-band-queue", -1));
3064
3065 if (disable_in_band) {
3066 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3067 } else {
3068 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3069 }
3070
3071 n_controllers = bridge_get_controllers(br, &controllers);
3072
3073 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3074 n_ocs = 0;
3075
3076 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3077 for (i = 0; i < n_controllers; i++) {
3078 struct ovsrec_controller *c = controllers[i];
3079
3080 if (!strncmp(c->target, "punix:", 6)
3081 || !strncmp(c->target, "unix:", 5)) {
3082 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3083 char *whitelist;
3084
3085 if (!strncmp(c->target, "unix:", 5)) {
3086 /* Connect to a listening socket */
3087 whitelist = xasprintf("unix:%s/", ovs_rundir());
3088 if (strchr(c->target, '/') &&
3089 !equal_pathnames(c->target, whitelist,
3090 strlen(whitelist))) {
3091 /* Absolute path specified, but not in ovs_rundir */
3092 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3093 "controller \"%s\" due to possibility for "
3094 "remote exploit. Instead, specify socket "
3095 "in whitelisted \"%s\" or connect to "
3096 "\"unix:%s/%s.mgmt\" (which is always "
3097 "available without special configuration).",
3098 br->name, c->target, whitelist,
3099 ovs_rundir(), br->name);
3100 free(whitelist);
3101 continue;
3102 }
3103 } else {
3104 whitelist = xasprintf("punix:%s/%s.controller",
3105 ovs_rundir(), br->name);
3106 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3107 /* Prevent remote ovsdb-server users from accessing
3108 * arbitrary Unix domain sockets and overwriting arbitrary
3109 * local files. */
3110 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3111 "controller \"%s\" due to possibility of "
3112 "overwriting local files. Instead, specify "
3113 "whitelisted \"%s\" or connect to "
3114 "\"unix:%s/%s.mgmt\" (which is always "
3115 "available without special configuration).",
3116 br->name, c->target, whitelist,
3117 ovs_rundir(), br->name);
3118 free(whitelist);
3119 continue;
3120 }
3121 }
3122
3123 free(whitelist);
3124 }
3125
3126 bridge_configure_local_iface_netdev(br, c);
3127 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3128 if (disable_in_band) {
3129 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3130 }
3131 n_ocs++;
3132 }
3133
3134 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3135 bridge_get_allowed_versions(br));
3136 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3137 free(ocs);
3138
3139 /* Set the fail-mode. */
3140 fail_mode = !br->cfg->fail_mode
3141 || !strcmp(br->cfg->fail_mode, "standalone")
3142 ? OFPROTO_FAIL_STANDALONE
3143 : OFPROTO_FAIL_SECURE;
3144 ofproto_set_fail_mode(br->ofproto, fail_mode);
3145
3146 /* Configure OpenFlow controller connection snooping. */
3147 if (!ofproto_has_snoops(br->ofproto)) {
3148 struct sset snoops;
3149
3150 sset_init(&snoops);
3151 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3152 ovs_rundir(), br->name));
3153 ofproto_set_snoops(br->ofproto, &snoops);
3154 sset_destroy(&snoops);
3155 }
3156 }
3157
3158 static void
3159 bridge_configure_tables(struct bridge *br)
3160 {
3161 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3162 int n_tables;
3163 int i, j, k;
3164
3165 n_tables = ofproto_get_n_tables(br->ofproto);
3166 j = 0;
3167 for (i = 0; i < n_tables; i++) {
3168 struct ofproto_table_settings s;
3169
3170 s.name = NULL;
3171 s.max_flows = UINT_MAX;
3172 s.groups = NULL;
3173 s.n_groups = 0;
3174 s.n_prefix_fields = 0;
3175 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3176
3177 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3178 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3179
3180 s.name = cfg->name;
3181 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3182 s.max_flows = *cfg->flow_limit;
3183 }
3184 if (cfg->overflow_policy
3185 && !strcmp(cfg->overflow_policy, "evict")) {
3186
3187 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3188 for (k = 0; k < cfg->n_groups; k++) {
3189 const char *string = cfg->groups[k];
3190 char *msg;
3191
3192 msg = mf_parse_subfield__(&s.groups[k], &string);
3193 if (msg) {
3194 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3195 "'groups' (%s)", br->name, i, msg);
3196 free(msg);
3197 } else if (*string) {
3198 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3199 "element '%s' contains trailing garbage",
3200 br->name, i, cfg->groups[k]);
3201 } else {
3202 s.n_groups++;
3203 }
3204 }
3205 }
3206 /* Prefix lookup fields. */
3207 s.n_prefix_fields = 0;
3208 for (k = 0; k < cfg->n_prefixes; k++) {
3209 const char *name = cfg->prefixes[k];
3210 const struct mf_field *mf = mf_from_name(name);
3211 if (!mf) {
3212 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3213 br->name, name);
3214 continue;
3215 }
3216 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3217 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3218 "%s", br->name, name);
3219 continue;
3220 }
3221 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3222 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3223 "field not used: %s", br->name, name);
3224 continue;
3225 }
3226 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3227 }
3228 if (s.n_prefix_fields > 0) {
3229 int k;
3230 struct ds ds = DS_EMPTY_INITIALIZER;
3231 for (k = 0; k < s.n_prefix_fields; k++) {
3232 if (k) {
3233 ds_put_char(&ds, ',');
3234 }
3235 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3236 }
3237 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3238 br->name, i, ds_cstr(&ds));
3239 ds_destroy(&ds);
3240 }
3241 }
3242
3243 ofproto_configure_table(br->ofproto, i, &s);
3244
3245 free(s.groups);
3246 }
3247 for (; j < br->cfg->n_flow_tables; j++) {
3248 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3249 "%"PRId64" not supported by this datapath", br->name,
3250 br->cfg->key_flow_tables[j]);
3251 }
3252 }
3253
3254 static void
3255 bridge_configure_dp_desc(struct bridge *br)
3256 {
3257 ofproto_set_dp_desc(br->ofproto,
3258 smap_get(&br->cfg->other_config, "dp-desc"));
3259 }
3260 \f
3261 /* Port functions. */
3262
3263 static struct port *
3264 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3265 {
3266 struct port *port;
3267
3268 port = xzalloc(sizeof *port);
3269 port->bridge = br;
3270 port->name = xstrdup(cfg->name);
3271 port->cfg = cfg;
3272 list_init(&port->ifaces);
3273
3274 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3275 return port;
3276 }
3277
3278 /* Deletes interfaces from 'port' that are no longer configured for it. */
3279 static void
3280 port_del_ifaces(struct port *port)
3281 {
3282 struct iface *iface, *next;
3283 struct sset new_ifaces;
3284 size_t i;
3285
3286 /* Collect list of new interfaces. */
3287 sset_init(&new_ifaces);
3288 for (i = 0; i < port->cfg->n_interfaces; i++) {
3289 const char *name = port->cfg->interfaces[i]->name;
3290 const char *type = port->cfg->interfaces[i]->type;
3291 if (strcmp(type, "null")) {
3292 sset_add(&new_ifaces, name);
3293 }
3294 }
3295
3296 /* Get rid of deleted interfaces. */
3297 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3298 if (!sset_contains(&new_ifaces, iface->name)) {
3299 iface_destroy(iface);
3300 }
3301 }
3302
3303 sset_destroy(&new_ifaces);
3304 }
3305
3306 static void
3307 port_destroy(struct port *port)
3308 {
3309 if (port) {
3310 struct bridge *br = port->bridge;
3311 struct iface *iface, *next;
3312
3313 if (br->ofproto) {
3314 ofproto_bundle_unregister(br->ofproto, port);
3315 }
3316
3317 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3318 iface_destroy(iface);
3319 }
3320
3321 hmap_remove(&br->ports, &port->hmap_node);
3322 free(port->name);
3323 free(port);
3324 }
3325 }
3326
3327 static struct port *
3328 port_lookup(const struct bridge *br, const char *name)
3329 {
3330 struct port *port;
3331
3332 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3333 &br->ports) {
3334 if (!strcmp(port->name, name)) {
3335 return port;
3336 }
3337 }
3338 return NULL;
3339 }
3340
3341 static bool
3342 enable_lacp(struct port *port, bool *activep)
3343 {
3344 if (!port->cfg->lacp) {
3345 /* XXX when LACP implementation has been sufficiently tested, enable by
3346 * default and make active on bonded ports. */
3347 return false;
3348 } else if (!strcmp(port->cfg->lacp, "off")) {
3349 return false;
3350 } else if (!strcmp(port->cfg->lacp, "active")) {
3351 *activep = true;
3352 return true;
3353 } else if (!strcmp(port->cfg->lacp, "passive")) {
3354 *activep = false;
3355 return true;
3356 } else {
3357 VLOG_WARN("port %s: unknown LACP mode %s",
3358 port->name, port->cfg->lacp);
3359 return false;
3360 }
3361 }
3362
3363 static struct lacp_settings *
3364 port_configure_lacp(struct port *port, struct lacp_settings *s)
3365 {
3366 const char *lacp_time, *system_id;
3367 int priority;
3368
3369 if (!enable_lacp(port, &s->active)) {
3370 return NULL;
3371 }
3372
3373 s->name = port->name;
3374
3375 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3376 if (system_id) {
3377 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3378 ETH_ADDR_SCAN_ARGS(s->id))) {
3379 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3380 " address.", port->name, system_id);
3381 return NULL;
3382 }
3383 } else {
3384 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3385 }
3386
3387 if (eth_addr_is_zero(s->id)) {
3388 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3389 return NULL;
3390 }
3391
3392 /* Prefer bondable links if unspecified. */
3393 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3394 0);
3395 s->priority = (priority > 0 && priority <= UINT16_MAX
3396 ? priority
3397 : UINT16_MAX - !list_is_short(&port->ifaces));
3398
3399 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3400 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3401
3402 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3403 "lacp-fallback-ab", false);
3404
3405 return s;
3406 }
3407
3408 static void
3409 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3410 {
3411 int priority, portid, key;
3412
3413 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3414 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3415 0);
3416 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3417
3418 if (portid <= 0 || portid > UINT16_MAX) {
3419 portid = ofp_to_u16(iface->ofp_port);
3420 }
3421
3422 if (priority <= 0 || priority > UINT16_MAX) {
3423 priority = UINT16_MAX;
3424 }
3425
3426 if (key < 0 || key > UINT16_MAX) {
3427 key = 0;
3428 }
3429
3430 s->name = iface->name;
3431 s->id = portid;
3432 s->priority = priority;
3433 s->key = key;
3434 }
3435
3436 static void
3437 port_configure_bond(struct port *port, struct bond_settings *s)
3438 {
3439 const char *detect_s;
3440 struct iface *iface;
3441 int miimon_interval;
3442
3443 s->name = port->name;
3444 s->balance = BM_AB;
3445 if (port->cfg->bond_mode) {
3446 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3447 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3448 port->name, port->cfg->bond_mode,
3449 bond_mode_to_string(s->balance));
3450 }
3451 } else {
3452 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3453
3454 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3455 * active-backup. At some point we should remove this warning. */
3456 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3457 " in previous versions, the default bond_mode was"
3458 " balance-slb", port->name,
3459 bond_mode_to_string(s->balance));
3460 }
3461 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3462 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3463 "please use another bond type or disable flood_vlans",
3464 port->name);
3465 }
3466
3467 miimon_interval = smap_get_int(&port->cfg->other_config,
3468 "bond-miimon-interval", 0);
3469 if (miimon_interval <= 0) {
3470 miimon_interval = 200;
3471 }
3472
3473 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3474 if (!detect_s || !strcmp(detect_s, "carrier")) {
3475 miimon_interval = 0;
3476 } else if (strcmp(detect_s, "miimon")) {
3477 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3478 "defaulting to carrier", port->name, detect_s);
3479 miimon_interval = 0;
3480 }
3481
3482 s->up_delay = MAX(0, port->cfg->bond_updelay);
3483 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3484 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3485 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3486 "bond-rebalance-interval", 10000);
3487 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3488 s->rebalance_interval = 1000;
3489 }
3490
3491 s->fake_iface = port->cfg->bond_fake_iface;
3492
3493 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3494 "lacp-fallback-ab", false);
3495
3496 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3497 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3498 }
3499 }
3500
3501 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3502 * instead of obtaining it from the database. */
3503 static bool
3504 port_is_synthetic(const struct port *port)
3505 {
3506 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3507 }
3508 \f
3509 /* Interface functions. */
3510
3511 static bool
3512 iface_is_internal(const struct ovsrec_interface *iface,
3513 const struct ovsrec_bridge *br)
3514 {
3515 /* The local port and "internal" ports are always "internal". */
3516 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3517 }
3518
3519 /* Returns the correct network device type for interface 'iface' in bridge
3520 * 'br'. */
3521 static const char *
3522 iface_get_type(const struct ovsrec_interface *iface,
3523 const struct ovsrec_bridge *br)
3524 {
3525 const char *type;
3526
3527 /* The local port always has type "internal". Other ports take
3528 * their type from the database and default to "system" if none is
3529 * specified. */
3530 if (iface_is_internal(iface, br)) {
3531 type = "internal";
3532 } else {
3533 type = iface->type[0] ? iface->type : "system";
3534 }
3535
3536 return ofproto_port_open_type(br->datapath_type, type);
3537 }
3538
3539 static void
3540 iface_destroy(struct iface *iface)
3541 {
3542 if (iface) {
3543 struct port *port = iface->port;
3544 struct bridge *br = port->bridge;
3545
3546 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3547 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3548 }
3549
3550 if (iface->ofp_port != OFPP_NONE) {
3551 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3552 }
3553
3554 list_remove(&iface->port_elem);
3555 hmap_remove(&br->iface_by_name, &iface->name_node);
3556
3557 netdev_close(iface->netdev);
3558
3559 free(iface->name);
3560 free(iface);
3561 }
3562 }
3563
3564 static struct iface *
3565 iface_lookup(const struct bridge *br, const char *name)
3566 {
3567 struct iface *iface;
3568
3569 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3570 &br->iface_by_name) {
3571 if (!strcmp(iface->name, name)) {
3572 return iface;
3573 }
3574 }
3575
3576 return NULL;
3577 }
3578
3579 static struct iface *
3580 iface_find(const char *name)
3581 {
3582 const struct bridge *br;
3583
3584 HMAP_FOR_EACH (br, node, &all_bridges) {
3585 struct iface *iface = iface_lookup(br, name);
3586
3587 if (iface) {
3588 return iface;
3589 }
3590 }
3591 return NULL;
3592 }
3593
3594 static struct if_cfg *
3595 if_cfg_lookup(const struct bridge *br, const char *name)
3596 {
3597 struct if_cfg *if_cfg;
3598
3599 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3600 &br->if_cfg_todo) {
3601 if (!strcmp(if_cfg->cfg->name, name)) {
3602 return if_cfg;
3603 }
3604 }
3605
3606 return NULL;
3607 }
3608
3609 static struct iface *
3610 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3611 {
3612 struct iface *iface;
3613
3614 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3615 &br->ifaces) {
3616 if (iface->ofp_port == ofp_port) {
3617 return iface;
3618 }
3619 }
3620 return NULL;
3621 }
3622
3623 /* Set Ethernet address of 'iface', if one is specified in the configuration
3624 * file. */
3625 static void
3626 iface_set_mac(struct iface *iface)
3627 {
3628 uint8_t ea[ETH_ADDR_LEN];
3629
3630 if (!strcmp(iface->type, "internal")
3631 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3632 if (iface->ofp_port == OFPP_LOCAL) {
3633 VLOG_ERR("interface %s: ignoring mac in Interface record "
3634 "(use Bridge record to set local port's mac)",
3635 iface->name);
3636 } else if (eth_addr_is_multicast(ea)) {
3637 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3638 iface->name);
3639 } else {
3640 int error = netdev_set_etheraddr(iface->netdev, ea);
3641 if (error) {
3642 VLOG_ERR("interface %s: setting MAC failed (%s)",
3643 iface->name, ovs_strerror(error));
3644 }
3645 }
3646 }
3647 }
3648
3649 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3650 static void
3651 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3652 {
3653 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3654 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3655 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3656 }
3657 }
3658
3659 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3660 * sets the "ofport" field to -1.
3661 *
3662 * This is appropriate when 'if_cfg''s interface cannot be created or is
3663 * otherwise invalid. */
3664 static void
3665 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3666 {
3667 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3668 ovsrec_interface_set_status(if_cfg, NULL);
3669 ovsrec_interface_set_admin_state(if_cfg, NULL);
3670 ovsrec_interface_set_duplex(if_cfg, NULL);
3671 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3672 ovsrec_interface_set_link_state(if_cfg, NULL);
3673 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3674 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3675 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3676 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3677 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3678 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3679 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3680 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3681 }
3682 }
3683
3684 static bool
3685 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3686 {
3687 union ovsdb_atom atom;
3688
3689 atom.integer = target;
3690 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3691 }
3692
3693 static void
3694 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3695 {
3696 struct ofpbuf queues_buf;
3697
3698 ofpbuf_init(&queues_buf, 0);
3699
3700 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3701 netdev_set_qos(iface->netdev, NULL, NULL);
3702 } else {
3703 const struct ovsdb_datum *queues;
3704 struct netdev_queue_dump dump;
3705 unsigned int queue_id;
3706 struct smap details;
3707 bool queue_zero;
3708 size_t i;
3709
3710 /* Configure top-level Qos for 'iface'. */
3711 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3712
3713 /* Deconfigure queues that were deleted. */
3714 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3715 OVSDB_TYPE_UUID);
3716 smap_init(&details);
3717 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3718 if (!queue_ids_include(queues, queue_id)) {
3719 netdev_delete_queue(iface->netdev, queue_id);
3720 }
3721 }
3722 smap_destroy(&details);
3723
3724 /* Configure queues for 'iface'. */
3725 queue_zero = false;
3726 for (i = 0; i < qos->n_queues; i++) {
3727 const struct ovsrec_queue *queue = qos->value_queues[i];
3728 unsigned int queue_id = qos->key_queues[i];
3729
3730 if (queue_id == 0) {
3731 queue_zero = true;
3732 }
3733
3734 if (queue->n_dscp == 1) {
3735 struct ofproto_port_queue *port_queue;
3736
3737 port_queue = ofpbuf_put_uninit(&queues_buf,
3738 sizeof *port_queue);
3739 port_queue->queue = queue_id;
3740 port_queue->dscp = queue->dscp[0];
3741 }
3742
3743 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3744 }
3745 if (!queue_zero) {
3746 struct smap details;
3747
3748 smap_init(&details);
3749 netdev_set_queue(iface->netdev, 0, &details);
3750 smap_destroy(&details);
3751 }
3752 }
3753
3754 if (iface->ofp_port != OFPP_NONE) {
3755 const struct ofproto_port_queue *port_queues = queues_buf.data;
3756 size_t n_queues = queues_buf.size / sizeof *port_queues;
3757
3758 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3759 port_queues, n_queues);
3760 }
3761
3762 netdev_set_policing(iface->netdev,
3763 iface->cfg->ingress_policing_rate,
3764 iface->cfg->ingress_policing_burst);
3765
3766 ofpbuf_uninit(&queues_buf);
3767 }
3768
3769 static void
3770 iface_configure_cfm(struct iface *iface)
3771 {
3772 const struct ovsrec_interface *cfg = iface->cfg;
3773 const char *opstate_str;
3774 const char *cfm_ccm_vlan;
3775 struct cfm_settings s;
3776 struct smap netdev_args;
3777
3778 if (!cfg->n_cfm_mpid) {
3779 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3780 return;
3781 }
3782
3783 s.check_tnl_key = false;
3784 smap_init(&netdev_args);
3785 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3786 const char *key = smap_get(&netdev_args, "key");
3787 const char *in_key = smap_get(&netdev_args, "in_key");
3788
3789 s.check_tnl_key = (key && !strcmp(key, "flow"))
3790 || (in_key && !strcmp(in_key, "flow"));
3791 }
3792 smap_destroy(&netdev_args);
3793
3794 s.mpid = *cfg->cfm_mpid;
3795 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3796 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3797 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3798
3799 if (s.interval <= 0) {
3800 s.interval = 1000;
3801 }
3802
3803 if (!cfm_ccm_vlan) {
3804 s.ccm_vlan = 0;
3805 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3806 s.ccm_vlan = CFM_RANDOM_VLAN;
3807 } else {
3808 s.ccm_vlan = atoi(cfm_ccm_vlan);
3809 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3810 s.ccm_vlan = 0;
3811 }
3812 }
3813
3814 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3815 false);
3816 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3817
3818 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3819 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3820
3821 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3822 }
3823
3824 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3825 * instead of obtaining it from the database. */
3826 static bool
3827 iface_is_synthetic(const struct iface *iface)
3828 {
3829 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3830 }
3831
3832 static ofp_port_t
3833 iface_pick_ofport(const struct ovsrec_interface *cfg)
3834 {
3835 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3836 : OFPP_NONE;
3837 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3838 : ofport;
3839 }
3840
3841 \f
3842 /* Port mirroring. */
3843
3844 static struct mirror *
3845 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3846 {
3847 struct mirror *m;
3848
3849 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3850 if (uuid_equals(uuid, &m->uuid)) {
3851 return m;
3852 }
3853 }
3854 return NULL;
3855 }
3856
3857 static void
3858 bridge_configure_mirrors(struct bridge *br)
3859 {
3860 const struct ovsdb_datum *mc;
3861 unsigned long *flood_vlans;
3862 struct mirror *m, *next;
3863 size_t i;
3864
3865 /* Get rid of deleted mirrors. */
3866 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3867 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3868 union ovsdb_atom atom;
3869
3870 atom.uuid = m->uuid;
3871 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3872 mirror_destroy(m);
3873 }
3874 }
3875
3876 /* Add new mirrors and reconfigure existing ones. */
3877 for (i = 0; i < br->cfg->n_mirrors; i++) {
3878 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3879 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3880 if (!m) {
3881 m = mirror_create(br, cfg);
3882 }
3883 m->cfg = cfg;
3884 if (!mirror_configure(m)) {
3885 mirror_destroy(m);
3886 }
3887 }
3888
3889 /* Update flooded vlans (for RSPAN). */
3890 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3891 br->cfg->n_flood_vlans);
3892 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3893 bitmap_free(flood_vlans);
3894 }
3895
3896 static struct mirror *
3897 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3898 {
3899 struct mirror *m;
3900
3901 m = xzalloc(sizeof *m);
3902 m->uuid = cfg->header_.uuid;
3903 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3904 m->bridge = br;
3905 m->name = xstrdup(cfg->name);
3906
3907 return m;
3908 }
3909
3910 static void
3911 mirror_destroy(struct mirror *m)
3912 {
3913 if (m) {
3914 struct bridge *br = m->bridge;
3915
3916 if (br->ofproto) {
3917 ofproto_mirror_unregister(br->ofproto, m);
3918 }
3919
3920 hmap_remove(&br->mirrors, &m->hmap_node);
3921 free(m->name);
3922 free(m);
3923 }
3924 }
3925
3926 static void
3927 mirror_collect_ports(struct mirror *m,
3928 struct ovsrec_port **in_ports, int n_in_ports,
3929 void ***out_portsp, size_t *n_out_portsp)
3930 {
3931 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3932 size_t n_out_ports = 0;
3933 size_t i;
3934
3935 for (i = 0; i < n_in_ports; i++) {
3936 const char *name = in_ports[i]->name;
3937 struct port *port = port_lookup(m->bridge, name);
3938 if (port) {
3939 out_ports[n_out_ports++] = port;
3940 } else {
3941 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3942 "port %s", m->bridge->name, m->name, name);
3943 }
3944 }
3945 *out_portsp = out_ports;
3946 *n_out_portsp = n_out_ports;
3947 }
3948
3949 static bool
3950 mirror_configure(struct mirror *m)
3951 {
3952 const struct ovsrec_mirror *cfg = m->cfg;
3953 struct ofproto_mirror_settings s;
3954
3955 /* Set name. */
3956 if (strcmp(cfg->name, m->name)) {
3957 free(m->name);
3958 m->name = xstrdup(cfg->name);
3959 }
3960 s.name = m->name;
3961
3962 /* Get output port or VLAN. */
3963 if (cfg->output_port) {
3964 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3965 if (!s.out_bundle) {
3966 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3967 m->bridge->name, m->name);
3968 return false;
3969 }
3970 s.out_vlan = UINT16_MAX;
3971
3972 if (cfg->output_vlan) {
3973 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3974 "output vlan; ignoring output vlan",
3975 m->bridge->name, m->name);
3976 }
3977 } else if (cfg->output_vlan) {
3978 /* The database should prevent invalid VLAN values. */
3979 s.out_bundle = NULL;
3980 s.out_vlan = *cfg->output_vlan;
3981 } else {
3982 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3983 m->bridge->name, m->name);
3984 return false;
3985 }
3986
3987 /* Get port selection. */
3988 if (cfg->select_all) {
3989 size_t n_ports = hmap_count(&m->bridge->ports);
3990 void **ports = xmalloc(n_ports * sizeof *ports);
3991 struct port *port;
3992 size_t i;
3993
3994 i = 0;
3995 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3996 ports[i++] = port;
3997 }
3998
3999 s.srcs = ports;
4000 s.n_srcs = n_ports;
4001
4002 s.dsts = ports;
4003 s.n_dsts = n_ports;
4004 } else {
4005 /* Get ports, dropping ports that don't exist.
4006 * The IDL ensures that there are no duplicates. */
4007 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4008 &s.srcs, &s.n_srcs);
4009 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4010 &s.dsts, &s.n_dsts);
4011 }
4012
4013 /* Get VLAN selection. */
4014 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4015
4016 /* Configure. */
4017 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4018
4019 /* Clean up. */
4020 if (s.srcs != s.dsts) {
4021 free(s.dsts);
4022 }
4023 free(s.srcs);
4024 free(s.src_vlans);
4025
4026 return true;
4027 }
4028 \f
4029 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4030 *
4031 * This is deprecated. It is only for compatibility with broken device drivers
4032 * in old versions of Linux that do not properly support VLANs when VLAN
4033 * devices are not used. When broken device drivers are no longer in
4034 * widespread use, we will delete these interfaces. */
4035
4036 static struct ovsrec_port **recs;
4037 static size_t n_recs, allocated_recs;
4038
4039 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4040 * splinters are reconfigured. */
4041 static void
4042 register_rec(struct ovsrec_port *rec)
4043 {
4044 if (n_recs >= allocated_recs) {
4045 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4046 }
4047 recs[n_recs++] = rec;
4048 }
4049
4050 /* Frees all of the ports registered with register_reg(). */
4051 static void
4052 free_registered_recs(void)
4053 {
4054 size_t i;
4055
4056 for (i = 0; i < n_recs; i++) {
4057 struct ovsrec_port *port = recs[i];
4058 size_t j;
4059
4060 for (j = 0; j < port->n_interfaces; j++) {
4061 struct ovsrec_interface *iface = port->interfaces[j];
4062 free(iface->name);
4063 free(iface);
4064 }
4065
4066 smap_destroy(&port->other_config);
4067 free(port->interfaces);
4068 free(port->name);
4069 free(port->tag);
4070 free(port);
4071 }
4072 n_recs = 0;
4073 }
4074
4075 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4076 * otherwise. */
4077 static bool
4078 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4079 {
4080 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4081 false);
4082 }
4083
4084 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4085 * splinters.
4086 *
4087 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4088 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4089 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4090 * with free().
4091 *
4092 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4093 * use, returns NULL.
4094 *
4095 * Updates 'vlan_splinters_enabled_anywhere'. */
4096 static unsigned long int *
4097 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4098 {
4099 unsigned long int *splinter_vlans;
4100 struct sset splinter_ifaces;
4101 const char *real_dev_name;
4102 struct shash *real_devs;
4103 struct shash_node *node;
4104 struct bridge *br;
4105 size_t i;
4106
4107 /* Free space allocated for synthesized ports and interfaces, since we're
4108 * in the process of reconstructing all of them. */
4109 free_registered_recs();
4110
4111 splinter_vlans = bitmap_allocate(4096);
4112 sset_init(&splinter_ifaces);
4113 vlan_splinters_enabled_anywhere = false;
4114 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4115 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4116 size_t j;
4117
4118 for (j = 0; j < br_cfg->n_ports; j++) {
4119 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4120 int k;
4121
4122 for (k = 0; k < port_cfg->n_interfaces; k++) {
4123 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4124
4125 if (vlan_splinters_is_enabled(iface_cfg)) {
4126 vlan_splinters_enabled_anywhere = true;
4127 sset_add(&splinter_ifaces, iface_cfg->name);
4128 vlan_bitmap_from_array__(port_cfg->trunks,
4129 port_cfg->n_trunks,
4130 splinter_vlans);
4131 }
4132 }
4133
4134 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4135 bitmap_set1(splinter_vlans, *port_cfg->tag);
4136 }
4137 }
4138 }
4139
4140 if (!vlan_splinters_enabled_anywhere) {
4141 free(splinter_vlans);
4142 sset_destroy(&splinter_ifaces);
4143 return NULL;
4144 }
4145
4146 HMAP_FOR_EACH (br, node, &all_bridges) {
4147 if (br->ofproto) {
4148 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4149 }
4150 }
4151
4152 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4153 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4154 * device to be created for it. */
4155 bitmap_set0(splinter_vlans, 0);
4156 bitmap_set0(splinter_vlans, 4095);
4157
4158 /* Delete all VLAN devices that we don't need. */
4159 vlandev_refresh();
4160 real_devs = vlandev_get_real_devs();
4161 SHASH_FOR_EACH (node, real_devs) {
4162 const struct vlan_real_dev *real_dev = node->data;
4163 const struct vlan_dev *vlan_dev;
4164 bool real_dev_has_splinters;
4165
4166 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4167 real_dev->name);
4168 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4169 if (!real_dev_has_splinters
4170 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4171 struct netdev *netdev;
4172
4173 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4174 if (!netdev_get_in4(netdev, NULL, NULL) ||
4175 !netdev_get_in6(netdev, NULL)) {
4176 /* It has an IP address configured, so we don't own
4177 * it. Don't delete it. */
4178 } else {
4179 vlandev_del(vlan_dev->name);
4180 }
4181 netdev_close(netdev);
4182 }
4183 }
4184
4185 }
4186 }
4187
4188 /* Add all VLAN devices that we need. */
4189 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4190 int vid;
4191
4192 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4193 if (!vlandev_get_name(real_dev_name, vid)) {
4194 vlandev_add(real_dev_name, vid);
4195 }
4196 }
4197 }
4198
4199 vlandev_refresh();
4200
4201 sset_destroy(&splinter_ifaces);
4202
4203 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4204 free(splinter_vlans);
4205 return NULL;
4206 }
4207 return splinter_vlans;
4208 }
4209
4210 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4211 * ofproto. */
4212 static void
4213 configure_splinter_port(struct port *port)
4214 {
4215 struct ofproto *ofproto = port->bridge->ofproto;
4216 ofp_port_t realdev_ofp_port;
4217 const char *realdev_name;
4218 struct iface *vlandev, *realdev;
4219
4220 ofproto_bundle_unregister(port->bridge->ofproto, port);
4221
4222 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4223 port_elem);
4224
4225 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4226 realdev = iface_lookup(port->bridge, realdev_name);
4227 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4228
4229 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4230 *port->cfg->tag);
4231 }
4232
4233 static struct ovsrec_port *
4234 synthesize_splinter_port(const char *real_dev_name,
4235 const char *vlan_dev_name, int vid)
4236 {
4237 struct ovsrec_interface *iface;
4238 struct ovsrec_port *port;
4239
4240 iface = xmalloc(sizeof *iface);
4241 ovsrec_interface_init(iface);
4242 iface->name = xstrdup(vlan_dev_name);
4243 iface->type = "system";
4244
4245 port = xmalloc(sizeof *port);
4246 ovsrec_port_init(port);
4247 port->interfaces = xmemdup(&iface, sizeof iface);
4248 port->n_interfaces = 1;
4249 port->name = xstrdup(vlan_dev_name);
4250 port->vlan_mode = "splinter";
4251 port->tag = xmalloc(sizeof *port->tag);
4252 *port->tag = vid;
4253
4254 smap_add(&port->other_config, "realdev", real_dev_name);
4255
4256 register_rec(port);
4257 return port;
4258 }
4259
4260 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4261 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4262 * 1-bit in the 'splinter_vlans' bitmap. */
4263 static void
4264 add_vlan_splinter_ports(struct bridge *br,
4265 const unsigned long int *splinter_vlans,
4266 struct shash *ports)
4267 {
4268 size_t i;
4269
4270 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4271 * we're modifying 'ports'. */
4272 for (i = 0; i < br->cfg->n_ports; i++) {
4273 const char *name = br->cfg->ports[i]->name;
4274 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4275 size_t j;
4276
4277 for (j = 0; j < port_cfg->n_interfaces; j++) {
4278 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4279
4280 if (vlan_splinters_is_enabled(iface_cfg)) {
4281 const char *real_dev_name;
4282 uint16_t vid;
4283
4284 real_dev_name = iface_cfg->name;
4285 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4286 const char *vlan_dev_name;
4287
4288 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4289 if (vlan_dev_name
4290 && !shash_find(ports, vlan_dev_name)) {
4291 shash_add(ports, vlan_dev_name,
4292 synthesize_splinter_port(
4293 real_dev_name, vlan_dev_name, vid));
4294 }
4295 }
4296 }
4297 }
4298 }
4299 }
4300
4301 static void
4302 mirror_refresh_stats(struct mirror *m)
4303 {
4304 struct ofproto *ofproto = m->bridge->ofproto;
4305 uint64_t tx_packets, tx_bytes;
4306 char *keys[2];
4307 int64_t values[2];
4308 size_t stat_cnt = 0;
4309
4310 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4311 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4312 return;
4313 }
4314
4315 if (tx_packets != UINT64_MAX) {
4316 keys[stat_cnt] = "tx_packets";
4317 values[stat_cnt] = tx_packets;
4318 stat_cnt++;
4319 }
4320 if (tx_bytes != UINT64_MAX) {
4321 keys[stat_cnt] = "tx_bytes";
4322 values[stat_cnt] = tx_bytes;
4323 stat_cnt++;
4324 }
4325
4326 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
4327 }