]> git.proxmox.com Git - mirror_ovs.git/blob - vswitchd/bridge.c
idl: New helpers for accessing string maps.
[mirror_ovs.git] / vswitchd / bridge.c
1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks
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 <assert.h>
19 #include <errno.h>
20 #include <inttypes.h>
21 #include <stdlib.h>
22 #include "bitmap.h"
23 #include "bond.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 "ofpbuf.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
42 #include "sha1.h"
43 #include "shash.h"
44 #include "socket-util.h"
45 #include "stream.h"
46 #include "stream-ssl.h"
47 #include "sset.h"
48 #include "system-stats.h"
49 #include "timeval.h"
50 #include "util.h"
51 #include "unixctl.h"
52 #include "vlandev.h"
53 #include "vswitch-idl.h"
54 #include "xenserver.h"
55 #include "vlog.h"
56 #include "sflow_api.h"
57 #include "vlan-bitmap.h"
58
59 VLOG_DEFINE_THIS_MODULE(bridge);
60
61 COVERAGE_DEFINE(bridge_reconfigure);
62
63 struct iface {
64 /* These members are always valid. */
65 struct list port_elem; /* Element in struct port's "ifaces" list. */
66 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
67 struct port *port; /* Containing port. */
68 char *name; /* Host network device name. */
69 tag_type tag; /* Tag associated with this interface. */
70
71 /* These members are valid only after bridge_reconfigure() causes them to
72 * be initialized. */
73 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
74 int ofp_port; /* OpenFlow port number, -1 if unknown. */
75 struct netdev *netdev; /* Network device. */
76 const char *type; /* Usually same as cfg->type. */
77 const struct ovsrec_interface *cfg;
78 };
79
80 struct mirror {
81 struct uuid uuid; /* UUID of this "mirror" record in database. */
82 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
83 struct bridge *bridge;
84 char *name;
85 const struct ovsrec_mirror *cfg;
86 };
87
88 struct port {
89 struct bridge *bridge;
90 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
91 char *name;
92
93 const struct ovsrec_port *cfg;
94
95 /* An ordinary bridge port has 1 interface.
96 * A bridge port for bonding has at least 2 interfaces. */
97 struct list ifaces; /* List of "struct iface"s. */
98 };
99
100 struct bridge {
101 struct hmap_node node; /* In 'all_bridges'. */
102 char *name; /* User-specified arbitrary name. */
103 char *type; /* Datapath type. */
104 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
105 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
106 const struct ovsrec_bridge *cfg;
107
108 /* OpenFlow switch processing. */
109 struct ofproto *ofproto; /* OpenFlow switch. */
110
111 /* Bridge ports. */
112 struct hmap ports; /* "struct port"s indexed by name. */
113 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
114 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
115
116 /* Port mirroring. */
117 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
118
119 /* Synthetic local port if necessary. */
120 struct ovsrec_port synth_local_port;
121 struct ovsrec_interface synth_local_iface;
122 struct ovsrec_interface *synth_local_ifacep;
123 };
124
125 /* All bridges, indexed by name. */
126 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
127
128 /* OVSDB IDL used to obtain configuration. */
129 static struct ovsdb_idl *idl;
130
131 /* Each time this timer expires, the bridge fetches systems and interface
132 * statistics and pushes them into the database. */
133 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
134 static long long int stats_timer = LLONG_MIN;
135
136 /* Stores the time after which rate limited statistics may be written to the
137 * database. Only updated when changes to the database require rate limiting.
138 */
139 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
140 static long long int db_limiter = LLONG_MIN;
141
142 static void add_del_bridges(const struct ovsrec_open_vswitch *);
143 static void bridge_del_ofprotos(void);
144 static bool bridge_add_ofprotos(struct bridge *);
145 static void bridge_create(const struct ovsrec_bridge *);
146 static void bridge_destroy(struct bridge *);
147 static struct bridge *bridge_lookup(const char *name);
148 static unixctl_cb_func bridge_unixctl_dump_flows;
149 static unixctl_cb_func bridge_unixctl_reconnect;
150 static size_t bridge_get_controllers(const struct bridge *br,
151 struct ovsrec_controller ***controllersp);
152 static void bridge_add_del_ports(struct bridge *,
153 const unsigned long int *splinter_vlans);
154 static void bridge_add_ofproto_ports(struct bridge *);
155 static void bridge_del_ofproto_ports(struct bridge *);
156 static void bridge_refresh_ofp_port(struct bridge *);
157 static void bridge_configure_datapath_id(struct bridge *);
158 static void bridge_configure_flow_eviction_threshold(struct bridge *);
159 static void bridge_configure_netflow(struct bridge *);
160 static void bridge_configure_forward_bpdu(struct bridge *);
161 static void bridge_configure_mac_idle_time(struct bridge *);
162 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
163 static void bridge_configure_stp(struct bridge *);
164 static void bridge_configure_tables(struct bridge *);
165 static void bridge_configure_remotes(struct bridge *,
166 const struct sockaddr_in *managers,
167 size_t n_managers);
168 static void bridge_pick_local_hw_addr(struct bridge *,
169 uint8_t ea[ETH_ADDR_LEN],
170 struct iface **hw_addr_iface);
171 static uint64_t bridge_pick_datapath_id(struct bridge *,
172 const uint8_t bridge_ea[ETH_ADDR_LEN],
173 struct iface *hw_addr_iface);
174 static uint64_t dpid_from_hash(const void *, size_t nbytes);
175 static bool bridge_has_bond_fake_iface(const struct bridge *,
176 const char *name);
177 static bool port_is_bond_fake_iface(const struct port *);
178
179 static unixctl_cb_func qos_unixctl_show;
180
181 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
182 static void port_add_ifaces(struct port *);
183 static void port_del_ifaces(struct port *);
184 static void port_destroy(struct port *);
185 static struct port *port_lookup(const struct bridge *, const char *name);
186 static void port_configure(struct port *);
187 static struct lacp_settings *port_configure_lacp(struct port *,
188 struct lacp_settings *);
189 static void port_configure_bond(struct port *, struct bond_settings *,
190 uint32_t *bond_stable_ids);
191 static bool port_is_synthetic(const struct port *);
192
193 static void bridge_configure_mirrors(struct bridge *);
194 static struct mirror *mirror_create(struct bridge *,
195 const struct ovsrec_mirror *);
196 static void mirror_destroy(struct mirror *);
197 static bool mirror_configure(struct mirror *);
198 static void mirror_refresh_stats(struct mirror *);
199
200 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
201 static struct iface *iface_create(struct port *port,
202 const struct ovsrec_interface *if_cfg);
203 static void iface_destroy(struct iface *);
204 static struct iface *iface_lookup(const struct bridge *, const char *name);
205 static struct iface *iface_find(const char *name);
206 static struct iface *iface_from_ofp_port(const struct bridge *,
207 uint16_t ofp_port);
208 static void iface_set_mac(struct iface *);
209 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
210 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
211 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
212 static void iface_configure_cfm(struct iface *);
213 static void iface_refresh_cfm_stats(struct iface *);
214 static void iface_refresh_stats(struct iface *);
215 static void iface_refresh_status(struct iface *);
216 static bool iface_is_synthetic(const struct iface *);
217
218 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
219 struct shash *);
220 static void shash_to_ovs_idl_map(struct shash *,
221 char ***keys, char ***values, size_t *n);
222
223 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
224 *
225 * This is deprecated. It is only for compatibility with broken device drivers
226 * in old versions of Linux that do not properly support VLANs when VLAN
227 * devices are not used. When broken device drivers are no longer in
228 * widespread use, we will delete these interfaces. */
229
230 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
231 static bool vlan_splinters_enabled_anywhere;
232
233 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
234 static unsigned long int *collect_splinter_vlans(
235 const struct ovsrec_open_vswitch *);
236 static void configure_splinter_port(struct port *);
237 static void add_vlan_splinter_ports(struct bridge *,
238 const unsigned long int *splinter_vlans,
239 struct shash *ports);
240 \f
241 /* Public functions. */
242
243 /* Initializes the bridge module, configuring it to obtain its configuration
244 * from an OVSDB server accessed over 'remote', which should be a string in a
245 * form acceptable to ovsdb_idl_create(). */
246 void
247 bridge_init(const char *remote)
248 {
249 /* Create connection to database. */
250 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
251 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
252
253 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
254 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
255 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
257 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
258 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
259 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
260
261 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
262 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
263 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
264
265 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
266 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
267 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
268 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
269
270 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
271 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
272 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
273 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
274 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
275 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
276 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
277 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
283 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
284
285 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
286 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
287 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
288 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
289
290 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
291
292 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
293
294 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
295 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
296
297 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
298
299 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
300
301 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
302 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
303 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
304 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
305 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
306
307 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
308
309 /* Register unixctl commands. */
310 unixctl_command_register("qos/show", "interface", 1, 1,
311 qos_unixctl_show, NULL);
312 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
313 bridge_unixctl_dump_flows, NULL);
314 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
315 bridge_unixctl_reconnect, NULL);
316 lacp_init();
317 bond_init();
318 cfm_init();
319 stp_init();
320 }
321
322 void
323 bridge_exit(void)
324 {
325 struct bridge *br, *next_br;
326
327 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
328 bridge_destroy(br);
329 }
330 ovsdb_idl_destroy(idl);
331 }
332
333 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
334 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
335 * responsible for freeing '*managersp' (with free()).
336 *
337 * You may be asking yourself "why does ovs-vswitchd care?", because
338 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
339 * should not be and in fact is not directly involved in that. But
340 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
341 * it has to tell in-band control where the managers are to enable that.
342 * (Thus, only managers connected in-band are collected.)
343 */
344 static void
345 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
346 struct sockaddr_in **managersp, size_t *n_managersp)
347 {
348 struct sockaddr_in *managers = NULL;
349 size_t n_managers = 0;
350 struct sset targets;
351 size_t i;
352
353 /* Collect all of the potential targets from the "targets" columns of the
354 * rows pointed to by "manager_options", excluding any that are
355 * out-of-band. */
356 sset_init(&targets);
357 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
358 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
359
360 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
361 sset_find_and_delete(&targets, m->target);
362 } else {
363 sset_add(&targets, m->target);
364 }
365 }
366
367 /* Now extract the targets' IP addresses. */
368 if (!sset_is_empty(&targets)) {
369 const char *target;
370
371 managers = xmalloc(sset_count(&targets) * sizeof *managers);
372 SSET_FOR_EACH (target, &targets) {
373 struct sockaddr_in *sin = &managers[n_managers];
374
375 if (stream_parse_target_with_default_ports(target,
376 JSONRPC_TCP_PORT,
377 JSONRPC_SSL_PORT,
378 sin)) {
379 n_managers++;
380 }
381 }
382 }
383 sset_destroy(&targets);
384
385 *managersp = managers;
386 *n_managersp = n_managers;
387 }
388
389 static void
390 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
391 {
392 unsigned long int *splinter_vlans;
393 struct sockaddr_in *managers;
394 struct bridge *br, *next;
395 int sflow_bridge_number;
396 size_t n_managers;
397
398 COVERAGE_INC(bridge_reconfigure);
399
400 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
401 * iface"s according to 'ovs_cfg', with only very minimal configuration
402 * otherwise.
403 *
404 * This is mostly an update to bridge data structures. Very little is
405 * pushed down to ofproto or lower layers. */
406 add_del_bridges(ovs_cfg);
407 splinter_vlans = collect_splinter_vlans(ovs_cfg);
408 HMAP_FOR_EACH (br, node, &all_bridges) {
409 bridge_add_del_ports(br, splinter_vlans);
410 }
411 free(splinter_vlans);
412
413 /* Delete all datapaths and datapath ports that are no longer configured.
414 *
415 * The kernel will reject any attempt to add a given port to a datapath if
416 * that port already belongs to a different datapath, so we must do all
417 * port deletions before any port additions. A datapath always has a
418 * "local port" so we must delete not-configured datapaths too. */
419 bridge_del_ofprotos();
420 HMAP_FOR_EACH (br, node, &all_bridges) {
421 if (br->ofproto) {
422 bridge_del_ofproto_ports(br);
423 }
424 }
425
426 /* Create datapaths and datapath ports that are missing.
427 *
428 * After this is done, we have our final set of bridges, ports, and
429 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
430 * has at least one iface, every "struct iface" has a valid ofp_port and
431 * netdev. */
432 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
433 if (!br->ofproto) {
434 if (bridge_add_ofprotos(br)) {
435 bridge_del_ofproto_ports(br);
436 } else {
437 bridge_destroy(br);
438 }
439 }
440 }
441 HMAP_FOR_EACH (br, node, &all_bridges) {
442 bridge_refresh_ofp_port(br);
443 bridge_add_ofproto_ports(br);
444 }
445
446 /* Complete the configuration. */
447 sflow_bridge_number = 0;
448 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
449 HMAP_FOR_EACH (br, node, &all_bridges) {
450 struct port *port;
451
452 /* We need the datapath ID early to allow LACP ports to use it as the
453 * default system ID. */
454 bridge_configure_datapath_id(br);
455
456 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
457 struct iface *iface;
458
459 port_configure(port);
460
461 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
462 iface_configure_cfm(iface);
463 iface_configure_qos(iface, port->cfg->qos);
464 iface_set_mac(iface);
465 }
466 }
467 bridge_configure_mirrors(br);
468 bridge_configure_flow_eviction_threshold(br);
469 bridge_configure_forward_bpdu(br);
470 bridge_configure_mac_idle_time(br);
471 bridge_configure_remotes(br, managers, n_managers);
472 bridge_configure_netflow(br);
473 bridge_configure_sflow(br, &sflow_bridge_number);
474 bridge_configure_stp(br);
475 bridge_configure_tables(br);
476 }
477 free(managers);
478
479 /* ovs-vswitchd has completed initialization, so allow the process that
480 * forked us to exit successfully. */
481 daemonize_complete();
482 }
483
484 /* Iterate over all ofprotos and delete any of them that do not have a
485 * configured bridge or that are the wrong type. */
486 static void
487 bridge_del_ofprotos(void)
488 {
489 struct sset names;
490 struct sset types;
491 const char *type;
492
493 sset_init(&names);
494 sset_init(&types);
495 ofproto_enumerate_types(&types);
496 SSET_FOR_EACH (type, &types) {
497 const char *name;
498
499 ofproto_enumerate_names(type, &names);
500 SSET_FOR_EACH (name, &names) {
501 struct bridge *br = bridge_lookup(name);
502 if (!br || strcmp(type, br->type)) {
503 ofproto_delete(name, type);
504 }
505 }
506 }
507 sset_destroy(&names);
508 sset_destroy(&types);
509 }
510
511 static bool
512 bridge_add_ofprotos(struct bridge *br)
513 {
514 int error = ofproto_create(br->name, br->type, &br->ofproto);
515 if (error) {
516 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
517 return false;
518 }
519 return true;
520 }
521
522 static void
523 port_configure(struct port *port)
524 {
525 const struct ovsrec_port *cfg = port->cfg;
526 struct bond_settings bond_settings;
527 struct lacp_settings lacp_settings;
528 struct ofproto_bundle_settings s;
529 struct iface *iface;
530
531 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
532 configure_splinter_port(port);
533 return;
534 }
535
536 /* Get name. */
537 s.name = port->name;
538
539 /* Get slaves. */
540 s.n_slaves = 0;
541 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
542 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
543 s.slaves[s.n_slaves++] = iface->ofp_port;
544 }
545
546 /* Get VLAN tag. */
547 s.vlan = -1;
548 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
549 s.vlan = *cfg->tag;
550 }
551
552 /* Get VLAN trunks. */
553 s.trunks = NULL;
554 if (cfg->n_trunks) {
555 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
556 }
557
558 /* Get VLAN mode. */
559 if (cfg->vlan_mode) {
560 if (!strcmp(cfg->vlan_mode, "access")) {
561 s.vlan_mode = PORT_VLAN_ACCESS;
562 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
563 s.vlan_mode = PORT_VLAN_TRUNK;
564 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
565 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
566 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
567 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
568 } else {
569 /* This "can't happen" because ovsdb-server should prevent it. */
570 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
571 s.vlan_mode = PORT_VLAN_TRUNK;
572 }
573 } else {
574 if (s.vlan >= 0) {
575 s.vlan_mode = PORT_VLAN_ACCESS;
576 if (cfg->n_trunks) {
577 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
578 port->name);
579 }
580 } else {
581 s.vlan_mode = PORT_VLAN_TRUNK;
582 }
583 }
584 s.use_priority_tags = !strcmp("true", ovsrec_port_get_other_config_value(
585 cfg, "priority-tags", ""));
586
587 /* Get LACP settings. */
588 s.lacp = port_configure_lacp(port, &lacp_settings);
589 if (s.lacp) {
590 size_t i = 0;
591
592 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
593 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
594 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
595 }
596 } else {
597 s.lacp_slaves = NULL;
598 }
599
600 /* Get bond settings. */
601 if (s.n_slaves > 1) {
602 s.bond = &bond_settings;
603 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
604 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
605 } else {
606 s.bond = NULL;
607 s.bond_stable_ids = NULL;
608
609 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
610 netdev_set_miimon_interval(iface->netdev, 0);
611 }
612 }
613
614 /* Register. */
615 ofproto_bundle_register(port->bridge->ofproto, port, &s);
616
617 /* Clean up. */
618 free(s.slaves);
619 free(s.trunks);
620 free(s.lacp_slaves);
621 free(s.bond_stable_ids);
622 }
623
624 /* Pick local port hardware address and datapath ID for 'br'. */
625 static void
626 bridge_configure_datapath_id(struct bridge *br)
627 {
628 uint8_t ea[ETH_ADDR_LEN];
629 uint64_t dpid;
630 struct iface *local_iface;
631 struct iface *hw_addr_iface;
632 char *dpid_string;
633
634 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
635 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
636 if (local_iface) {
637 int error = netdev_set_etheraddr(local_iface->netdev, ea);
638 if (error) {
639 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
640 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
641 "Ethernet address: %s",
642 br->name, strerror(error));
643 }
644 }
645 memcpy(br->ea, ea, ETH_ADDR_LEN);
646
647 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
648 ofproto_set_datapath_id(br->ofproto, dpid);
649
650 dpid_string = xasprintf("%016"PRIx64, dpid);
651 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
652 free(dpid_string);
653 }
654
655 /* Set NetFlow configuration on 'br'. */
656 static void
657 bridge_configure_netflow(struct bridge *br)
658 {
659 struct ovsrec_netflow *cfg = br->cfg->netflow;
660 struct netflow_options opts;
661
662 if (!cfg) {
663 ofproto_set_netflow(br->ofproto, NULL);
664 return;
665 }
666
667 memset(&opts, 0, sizeof opts);
668
669 /* Get default NetFlow configuration from datapath.
670 * Apply overrides from 'cfg'. */
671 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
672 if (cfg->engine_type) {
673 opts.engine_type = *cfg->engine_type;
674 }
675 if (cfg->engine_id) {
676 opts.engine_id = *cfg->engine_id;
677 }
678
679 /* Configure active timeout interval. */
680 opts.active_timeout = cfg->active_timeout;
681 if (!opts.active_timeout) {
682 opts.active_timeout = -1;
683 } else if (opts.active_timeout < 0) {
684 VLOG_WARN("bridge %s: active timeout interval set to negative "
685 "value, using default instead (%d seconds)", br->name,
686 NF_ACTIVE_TIMEOUT_DEFAULT);
687 opts.active_timeout = -1;
688 }
689
690 /* Add engine ID to interface number to disambiguate bridgs? */
691 opts.add_id_to_iface = cfg->add_id_to_interface;
692 if (opts.add_id_to_iface) {
693 if (opts.engine_id > 0x7f) {
694 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
695 "another vswitch, choose an engine id less than 128",
696 br->name);
697 }
698 if (hmap_count(&br->ports) > 508) {
699 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
700 "another port when more than 508 ports are used",
701 br->name);
702 }
703 }
704
705 /* Collectors. */
706 sset_init(&opts.collectors);
707 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
708
709 /* Configure. */
710 if (ofproto_set_netflow(br->ofproto, &opts)) {
711 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
712 }
713 sset_destroy(&opts.collectors);
714 }
715
716 /* Set sFlow configuration on 'br'. */
717 static void
718 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
719 {
720 const struct ovsrec_sflow *cfg = br->cfg->sflow;
721 struct ovsrec_controller **controllers;
722 struct ofproto_sflow_options oso;
723 size_t n_controllers;
724 size_t i;
725
726 if (!cfg) {
727 ofproto_set_sflow(br->ofproto, NULL);
728 return;
729 }
730
731 memset(&oso, 0, sizeof oso);
732
733 sset_init(&oso.targets);
734 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
735
736 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
737 if (cfg->sampling) {
738 oso.sampling_rate = *cfg->sampling;
739 }
740
741 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
742 if (cfg->polling) {
743 oso.polling_interval = *cfg->polling;
744 }
745
746 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
747 if (cfg->header) {
748 oso.header_len = *cfg->header;
749 }
750
751 oso.sub_id = (*sflow_bridge_number)++;
752 oso.agent_device = cfg->agent;
753
754 oso.control_ip = NULL;
755 n_controllers = bridge_get_controllers(br, &controllers);
756 for (i = 0; i < n_controllers; i++) {
757 if (controllers[i]->local_ip) {
758 oso.control_ip = controllers[i]->local_ip;
759 break;
760 }
761 }
762 ofproto_set_sflow(br->ofproto, &oso);
763
764 sset_destroy(&oso.targets);
765 }
766
767 static void
768 port_configure_stp(const struct ofproto *ofproto, struct port *port,
769 struct ofproto_port_stp_settings *port_s,
770 int *port_num_counter, unsigned long *port_num_bitmap)
771 {
772 const char *config_str;
773 struct iface *iface;
774
775 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-enable",
776 NULL);
777 if (config_str && !strcmp(config_str, "false")) {
778 port_s->enable = false;
779 return;
780 } else {
781 port_s->enable = true;
782 }
783
784 /* STP over bonds is not supported. */
785 if (!list_is_singleton(&port->ifaces)) {
786 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
787 port->name);
788 port_s->enable = false;
789 return;
790 }
791
792 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
793
794 /* Internal ports shouldn't participate in spanning tree, so
795 * skip them. */
796 if (!strcmp(iface->type, "internal")) {
797 VLOG_DBG("port %s: disable STP on internal ports", port->name);
798 port_s->enable = false;
799 return;
800 }
801
802 /* STP on mirror output ports is not supported. */
803 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
804 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
805 port_s->enable = false;
806 return;
807 }
808
809 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-port-num",
810 NULL);
811 if (config_str) {
812 unsigned long int port_num = strtoul(config_str, NULL, 0);
813 int port_idx = port_num - 1;
814
815 if (port_num < 1 || port_num > STP_MAX_PORTS) {
816 VLOG_ERR("port %s: invalid stp-port-num", port->name);
817 port_s->enable = false;
818 return;
819 }
820
821 if (bitmap_is_set(port_num_bitmap, port_idx)) {
822 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
823 port->name, port_num);
824 port_s->enable = false;
825 return;
826 }
827 bitmap_set1(port_num_bitmap, port_idx);
828 port_s->port_num = port_idx;
829 } else {
830 if (*port_num_counter > STP_MAX_PORTS) {
831 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
832 port_s->enable = false;
833 return;
834 }
835
836 port_s->port_num = (*port_num_counter)++;
837 }
838
839 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-path-cost",
840 NULL);
841 if (config_str) {
842 port_s->path_cost = strtoul(config_str, NULL, 10);
843 } else {
844 enum netdev_features current;
845
846 if (netdev_get_features(iface->netdev, &current, NULL, NULL, NULL)) {
847 /* Couldn't get speed, so assume 100Mb/s. */
848 port_s->path_cost = 19;
849 } else {
850 unsigned int mbps;
851
852 mbps = netdev_features_to_bps(current) / 1000000;
853 port_s->path_cost = stp_convert_speed_to_cost(mbps);
854 }
855 }
856
857 config_str = ovsrec_port_get_other_config_value(port->cfg,
858 "stp-port-priority",
859 NULL);
860 if (config_str) {
861 port_s->priority = strtoul(config_str, NULL, 0);
862 } else {
863 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
864 }
865 }
866
867 /* Set spanning tree configuration on 'br'. */
868 static void
869 bridge_configure_stp(struct bridge *br)
870 {
871 if (!br->cfg->stp_enable) {
872 ofproto_set_stp(br->ofproto, NULL);
873 } else {
874 struct ofproto_stp_settings br_s;
875 const char *config_str;
876 struct port *port;
877 int port_num_counter;
878 unsigned long *port_num_bitmap;
879
880 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
881 "stp-system-id",
882 NULL);
883 if (config_str) {
884 uint8_t ea[ETH_ADDR_LEN];
885
886 if (eth_addr_from_string(config_str, ea)) {
887 br_s.system_id = eth_addr_to_uint64(ea);
888 } else {
889 br_s.system_id = eth_addr_to_uint64(br->ea);
890 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
891 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
892 }
893 } else {
894 br_s.system_id = eth_addr_to_uint64(br->ea);
895 }
896
897 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
898 "stp-priority",
899 NULL);
900 if (config_str) {
901 br_s.priority = strtoul(config_str, NULL, 0);
902 } else {
903 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
904 }
905
906 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
907 "stp-hello-time",
908 NULL);
909 if (config_str) {
910 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
911 } else {
912 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
913 }
914
915 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
916 "stp-max-age",
917 NULL);
918 if (config_str) {
919 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
920 } else {
921 br_s.max_age = STP_DEFAULT_MAX_AGE;
922 }
923
924 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
925 "stp-forward-delay",
926 NULL);
927 if (config_str) {
928 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
929 } else {
930 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
931 }
932
933 /* Configure STP on the bridge. */
934 if (ofproto_set_stp(br->ofproto, &br_s)) {
935 VLOG_ERR("bridge %s: could not enable STP", br->name);
936 return;
937 }
938
939 /* Users must either set the port number with the "stp-port-num"
940 * configuration on all ports or none. If manual configuration
941 * is not done, then we allocate them sequentially. */
942 port_num_counter = 0;
943 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
944 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
945 struct ofproto_port_stp_settings port_s;
946 struct iface *iface;
947
948 port_configure_stp(br->ofproto, port, &port_s,
949 &port_num_counter, port_num_bitmap);
950
951 /* As bonds are not supported, just apply configuration to
952 * all interfaces. */
953 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
954 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
955 &port_s)) {
956 VLOG_ERR("port %s: could not enable STP", port->name);
957 continue;
958 }
959 }
960 }
961
962 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
963 && port_num_counter) {
964 VLOG_ERR("bridge %s: must manually configure all STP port "
965 "IDs or none, disabling", br->name);
966 ofproto_set_stp(br->ofproto, NULL);
967 }
968 bitmap_free(port_num_bitmap);
969 }
970 }
971
972 static bool
973 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
974 {
975 const struct port *port = port_lookup(br, name);
976 return port && port_is_bond_fake_iface(port);
977 }
978
979 static bool
980 port_is_bond_fake_iface(const struct port *port)
981 {
982 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
983 }
984
985 static void
986 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
987 {
988 struct bridge *br, *next;
989 struct shash new_br;
990 size_t i;
991
992 /* Collect new bridges' names and types. */
993 shash_init(&new_br);
994 for (i = 0; i < cfg->n_bridges; i++) {
995 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
996 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
997
998 if (strchr(br_cfg->name, '/')) {
999 /* Prevent remote ovsdb-server users from accessing arbitrary
1000 * directories, e.g. consider a bridge named "../../../etc/". */
1001 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1002 br_cfg->name);
1003 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1004 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1005 }
1006 }
1007
1008 /* Get rid of deleted bridges or those whose types have changed.
1009 * Update 'cfg' of bridges that still exist. */
1010 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1011 br->cfg = shash_find_data(&new_br, br->name);
1012 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1013 br->cfg->datapath_type))) {
1014 bridge_destroy(br);
1015 }
1016 }
1017
1018 /* Add new bridges. */
1019 for (i = 0; i < cfg->n_bridges; i++) {
1020 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1021 struct bridge *br = bridge_lookup(br_cfg->name);
1022 if (!br) {
1023 bridge_create(br_cfg);
1024 }
1025 }
1026
1027 shash_destroy(&new_br);
1028 }
1029
1030 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1031 * iface".
1032 *
1033 * The kernel will reject any attempt to add a given port to a datapath if that
1034 * port already belongs to a different datapath, so we must do all port
1035 * deletions before any port additions. */
1036 static void
1037 bridge_del_ofproto_ports(struct bridge *br)
1038 {
1039 struct ofproto_port_dump dump;
1040 struct ofproto_port ofproto_port;
1041
1042 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1043 const char *name = ofproto_port.name;
1044 struct iface *iface;
1045 const char *type;
1046 int error;
1047
1048 /* Ignore the local port. We can't change it anyhow. */
1049 if (!strcmp(name, br->name)) {
1050 continue;
1051 }
1052
1053 /* Get the type that 'ofproto_port' should have (ordinarily the
1054 * type of its corresponding iface) or NULL if it should be
1055 * deleted. */
1056 iface = iface_lookup(br, name);
1057 type = (iface ? iface->type
1058 : bridge_has_bond_fake_iface(br, name) ? "internal"
1059 : NULL);
1060
1061 /* If it's the wrong type then delete the ofproto port. */
1062 if (type
1063 && !strcmp(ofproto_port.type, type)
1064 && (!iface || !iface->netdev
1065 || !strcmp(netdev_get_type(iface->netdev), type))) {
1066 continue;
1067 }
1068 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1069 if (error) {
1070 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1071 br->name, name, strerror(error));
1072 }
1073 if (iface) {
1074 netdev_close(iface->netdev);
1075 iface->netdev = NULL;
1076 }
1077 }
1078 }
1079
1080 static void
1081 iface_set_ofp_port(struct iface *iface, int ofp_port)
1082 {
1083 struct bridge *br = iface->port->bridge;
1084
1085 assert(iface->ofp_port < 0 && ofp_port >= 0);
1086 iface->ofp_port = ofp_port;
1087 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1088 iface_set_ofport(iface->cfg, ofp_port);
1089 }
1090
1091 static void
1092 bridge_refresh_ofp_port(struct bridge *br)
1093 {
1094 struct ofproto_port_dump dump;
1095 struct ofproto_port ofproto_port;
1096 struct port *port;
1097
1098 /* Clear all the "ofp_port"es. */
1099 hmap_clear(&br->ifaces);
1100 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1101 struct iface *iface;
1102
1103 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1104 iface->ofp_port = -1;
1105 }
1106 }
1107
1108 /* Obtain the correct "ofp_port"s from ofproto. */
1109 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1110 struct iface *iface = iface_lookup(br, ofproto_port.name);
1111 if (iface) {
1112 if (iface->ofp_port >= 0) {
1113 VLOG_WARN("bridge %s: interface %s reported twice",
1114 br->name, ofproto_port.name);
1115 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1116 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1117 br->name, ofproto_port.ofp_port);
1118 } else {
1119 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1120 }
1121 }
1122 }
1123 }
1124
1125 /* Add an ofproto port for any "struct iface" that doesn't have one.
1126 * Delete any "struct iface" for which this fails.
1127 * Delete any "struct port" that thereby ends up with no ifaces. */
1128 static void
1129 bridge_add_ofproto_ports(struct bridge *br)
1130 {
1131 struct port *port, *next_port;
1132
1133 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1134 struct iface *iface, *next_iface;
1135 struct ofproto_port ofproto_port;
1136
1137 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1138 int error;
1139
1140 /* Open the netdev. */
1141 if (!iface->netdev) {
1142 error = netdev_open(iface->name, iface->type, &iface->netdev);
1143 if (error) {
1144 VLOG_WARN("could not open network device %s (%s)",
1145 iface->name, strerror(error));
1146 }
1147
1148 if (iface->netdev
1149 && port->cfg->vlan_mode
1150 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1151 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1152 }
1153 } else {
1154 error = 0;
1155 }
1156
1157 /* Configure the netdev. */
1158 if (iface->netdev) {
1159 struct shash args;
1160
1161 shash_init(&args);
1162 shash_from_ovs_idl_map(iface->cfg->key_options,
1163 iface->cfg->value_options,
1164 iface->cfg->n_options, &args);
1165 error = netdev_set_config(iface->netdev, &args);
1166 shash_destroy(&args);
1167
1168 if (error) {
1169 VLOG_WARN("could not configure network device %s (%s)",
1170 iface->name, strerror(error));
1171 netdev_close(iface->netdev);
1172 iface->netdev = NULL;
1173 }
1174 }
1175
1176 /* Add the port, if necessary. */
1177 if (iface->netdev && iface->ofp_port < 0) {
1178 uint16_t ofp_port;
1179 int error;
1180
1181 error = ofproto_port_add(br->ofproto, iface->netdev,
1182 &ofp_port);
1183 if (!error) {
1184 iface_set_ofp_port(iface, ofp_port);
1185 } else {
1186 netdev_close(iface->netdev);
1187 iface->netdev = NULL;
1188 }
1189 }
1190
1191 /* Populate stats columns in new Interface rows. */
1192 if (iface->netdev && !iface->cfg->mtu) {
1193 iface_refresh_stats(iface);
1194 iface_refresh_status(iface);
1195 }
1196
1197 /* Delete the iface if we failed. */
1198 if (iface->netdev && iface->ofp_port >= 0) {
1199 VLOG_DBG("bridge %s: interface %s is on port %d",
1200 br->name, iface->name, iface->ofp_port);
1201 } else {
1202 if (iface->netdev) {
1203 VLOG_ERR("bridge %s: missing %s interface, dropping",
1204 br->name, iface->name);
1205 } else {
1206 /* We already reported a related error, don't bother
1207 * duplicating it. */
1208 }
1209 iface_clear_db_record(iface->cfg);
1210 iface_destroy(iface);
1211 }
1212 }
1213 if (list_is_empty(&port->ifaces)) {
1214 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1215 port_destroy(port);
1216 continue;
1217 }
1218
1219 /* Add bond fake iface if necessary. */
1220 if (port_is_bond_fake_iface(port)) {
1221 if (ofproto_port_query_by_name(br->ofproto, port->name,
1222 &ofproto_port)) {
1223 struct netdev *netdev;
1224 int error;
1225
1226 error = netdev_open(port->name, "internal", &netdev);
1227 if (!error) {
1228 ofproto_port_add(br->ofproto, netdev, NULL);
1229 netdev_close(netdev);
1230 } else {
1231 VLOG_WARN("could not open network device %s (%s)",
1232 port->name, strerror(error));
1233 }
1234 } else {
1235 /* Already exists, nothing to do. */
1236 ofproto_port_destroy(&ofproto_port);
1237 }
1238 }
1239 }
1240 }
1241
1242 /* Set Flow eviction threshold */
1243 static void
1244 bridge_configure_flow_eviction_threshold(struct bridge *br)
1245 {
1246 const char *threshold_str;
1247 unsigned threshold;
1248
1249 threshold_str =
1250 ovsrec_bridge_get_other_config_value(br->cfg,
1251 "flow-eviction-threshold",
1252 NULL);
1253 if (threshold_str) {
1254 threshold = strtoul(threshold_str, NULL, 10);
1255 } else {
1256 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1257 }
1258 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1259 }
1260
1261 /* Set forward BPDU option. */
1262 static void
1263 bridge_configure_forward_bpdu(struct bridge *br)
1264 {
1265 const char *forward_bpdu_str;
1266 bool forward_bpdu = false;
1267
1268 forward_bpdu_str = ovsrec_bridge_get_other_config_value(br->cfg,
1269 "forward-bpdu",
1270 NULL);
1271 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1272 forward_bpdu = true;
1273 }
1274 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1275 }
1276
1277 /* Set MAC aging time for 'br'. */
1278 static void
1279 bridge_configure_mac_idle_time(struct bridge *br)
1280 {
1281 const char *idle_time_str;
1282 int idle_time;
1283
1284 idle_time_str = ovsrec_bridge_get_other_config_value(br->cfg,
1285 "mac-aging-time",
1286 NULL);
1287 idle_time = (idle_time_str && atoi(idle_time_str)
1288 ? atoi(idle_time_str)
1289 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1290 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1291 }
1292
1293 static void
1294 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1295 struct iface **hw_addr_iface)
1296 {
1297 struct hmapx mirror_output_ports;
1298 const char *hwaddr;
1299 struct port *port;
1300 bool found_addr = false;
1301 int error;
1302 int i;
1303
1304 *hw_addr_iface = NULL;
1305
1306 /* Did the user request a particular MAC? */
1307 hwaddr = ovsrec_bridge_get_other_config_value(br->cfg, "hwaddr", NULL);
1308 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1309 if (eth_addr_is_multicast(ea)) {
1310 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1311 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1312 } else if (eth_addr_is_zero(ea)) {
1313 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1314 } else {
1315 return;
1316 }
1317 }
1318
1319 /* Mirror output ports don't participate in picking the local hardware
1320 * address. ofproto can't help us find out whether a given port is a
1321 * mirror output because we haven't configured mirrors yet, so we need to
1322 * accumulate them ourselves. */
1323 hmapx_init(&mirror_output_ports);
1324 for (i = 0; i < br->cfg->n_mirrors; i++) {
1325 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1326 if (m->output_port) {
1327 hmapx_add(&mirror_output_ports, m->output_port);
1328 }
1329 }
1330
1331 /* Otherwise choose the minimum non-local MAC address among all of the
1332 * interfaces. */
1333 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1334 uint8_t iface_ea[ETH_ADDR_LEN];
1335 struct iface *candidate;
1336 struct iface *iface;
1337
1338 /* Mirror output ports don't participate. */
1339 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1340 continue;
1341 }
1342
1343 /* Choose the MAC address to represent the port. */
1344 iface = NULL;
1345 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1346 /* Find the interface with this Ethernet address (if any) so that
1347 * we can provide the correct devname to the caller. */
1348 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1349 uint8_t candidate_ea[ETH_ADDR_LEN];
1350 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1351 && eth_addr_equals(iface_ea, candidate_ea)) {
1352 iface = candidate;
1353 }
1354 }
1355 } else {
1356 /* Choose the interface whose MAC address will represent the port.
1357 * The Linux kernel bonding code always chooses the MAC address of
1358 * the first slave added to a bond, and the Fedora networking
1359 * scripts always add slaves to a bond in alphabetical order, so
1360 * for compatibility we choose the interface with the name that is
1361 * first in alphabetical order. */
1362 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1363 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1364 iface = candidate;
1365 }
1366 }
1367
1368 /* The local port doesn't count (since we're trying to choose its
1369 * MAC address anyway). */
1370 if (iface->ofp_port == OFPP_LOCAL) {
1371 continue;
1372 }
1373
1374 /* Grab MAC. */
1375 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1376 if (error) {
1377 continue;
1378 }
1379 }
1380
1381 /* Compare against our current choice. */
1382 if (!eth_addr_is_multicast(iface_ea) &&
1383 !eth_addr_is_local(iface_ea) &&
1384 !eth_addr_is_reserved(iface_ea) &&
1385 !eth_addr_is_zero(iface_ea) &&
1386 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1387 {
1388 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1389 *hw_addr_iface = iface;
1390 found_addr = true;
1391 }
1392 }
1393 if (found_addr) {
1394 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1395 br->name, ETH_ADDR_ARGS(ea));
1396 } else {
1397 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1398 *hw_addr_iface = NULL;
1399 VLOG_WARN("bridge %s: using default bridge Ethernet "
1400 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1401 }
1402
1403 hmapx_destroy(&mirror_output_ports);
1404 }
1405
1406 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1407 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1408 * an interface on 'br', then that interface must be passed in as
1409 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1410 * 'hw_addr_iface' must be passed in as a null pointer. */
1411 static uint64_t
1412 bridge_pick_datapath_id(struct bridge *br,
1413 const uint8_t bridge_ea[ETH_ADDR_LEN],
1414 struct iface *hw_addr_iface)
1415 {
1416 /*
1417 * The procedure for choosing a bridge MAC address will, in the most
1418 * ordinary case, also choose a unique MAC that we can use as a datapath
1419 * ID. In some special cases, though, multiple bridges will end up with
1420 * the same MAC address. This is OK for the bridges, but it will confuse
1421 * the OpenFlow controller, because each datapath needs a unique datapath
1422 * ID.
1423 *
1424 * Datapath IDs must be unique. It is also very desirable that they be
1425 * stable from one run to the next, so that policy set on a datapath
1426 * "sticks".
1427 */
1428 const char *datapath_id;
1429 uint64_t dpid;
1430
1431 datapath_id = ovsrec_bridge_get_other_config_value(br->cfg, "datapath-id",
1432 NULL);
1433 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1434 return dpid;
1435 }
1436
1437 if (!hw_addr_iface) {
1438 /*
1439 * A purely internal bridge, that is, one that has no non-virtual
1440 * network devices on it at all, is difficult because it has no
1441 * natural unique identifier at all.
1442 *
1443 * When the host is a XenServer, we handle this case by hashing the
1444 * host's UUID with the name of the bridge. Names of bridges are
1445 * persistent across XenServer reboots, although they can be reused if
1446 * an internal network is destroyed and then a new one is later
1447 * created, so this is fairly effective.
1448 *
1449 * When the host is not a XenServer, we punt by using a random MAC
1450 * address on each run.
1451 */
1452 const char *host_uuid = xenserver_get_host_uuid();
1453 if (host_uuid) {
1454 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1455 dpid = dpid_from_hash(combined, strlen(combined));
1456 free(combined);
1457 return dpid;
1458 }
1459 }
1460
1461 return eth_addr_to_uint64(bridge_ea);
1462 }
1463
1464 static uint64_t
1465 dpid_from_hash(const void *data, size_t n)
1466 {
1467 uint8_t hash[SHA1_DIGEST_SIZE];
1468
1469 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1470 sha1_bytes(data, n, hash);
1471 eth_addr_mark_random(hash);
1472 return eth_addr_to_uint64(hash);
1473 }
1474
1475 static void
1476 iface_refresh_status(struct iface *iface)
1477 {
1478 struct shash sh;
1479
1480 enum netdev_features current;
1481 enum netdev_flags flags;
1482 int64_t bps;
1483 int mtu;
1484 int64_t mtu_64;
1485 int error;
1486
1487 if (iface_is_synthetic(iface)) {
1488 return;
1489 }
1490
1491 shash_init(&sh);
1492
1493 if (!netdev_get_drv_info(iface->netdev, &sh)) {
1494 size_t n;
1495 char **keys, **values;
1496
1497 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1498 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1499
1500 free(keys);
1501 free(values);
1502 } else {
1503 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1504 }
1505
1506 shash_destroy_free_data(&sh);
1507
1508 error = netdev_get_flags(iface->netdev, &flags);
1509 if (!error) {
1510 ovsrec_interface_set_admin_state(iface->cfg,
1511 flags & NETDEV_UP ? "up" : "down");
1512 }
1513 else {
1514 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1515 }
1516
1517 error = netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1518 if (!error) {
1519 ovsrec_interface_set_duplex(iface->cfg,
1520 netdev_features_is_full_duplex(current)
1521 ? "full" : "half");
1522 /* warning: uint64_t -> int64_t conversion */
1523 bps = netdev_features_to_bps(current);
1524 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1525 }
1526 else {
1527 ovsrec_interface_set_duplex(iface->cfg, NULL);
1528 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1529 }
1530
1531 error = netdev_get_mtu(iface->netdev, &mtu);
1532 if (!error) {
1533 mtu_64 = mtu;
1534 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1535 }
1536 else {
1537 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1538 }
1539 }
1540
1541 /* Writes 'iface''s CFM statistics to the database. */
1542 static void
1543 iface_refresh_cfm_stats(struct iface *iface)
1544 {
1545 const struct ovsrec_interface *cfg = iface->cfg;
1546 int fault, error;
1547 const uint64_t *rmps;
1548 size_t n_rmps;
1549
1550 if (iface_is_synthetic(iface)) {
1551 return;
1552 }
1553
1554 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1555 iface->ofp_port);
1556 if (fault >= 0) {
1557 const char *reasons[CFM_FAULT_N_REASONS];
1558 bool fault_bool = fault;
1559 size_t i, j;
1560
1561 j = 0;
1562 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1563 int reason = 1 << i;
1564 if (fault & reason) {
1565 reasons[j++] = cfm_fault_reason_to_str(reason);
1566 }
1567 }
1568
1569 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1570 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1571 } else {
1572 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1573 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1574 }
1575
1576 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1577 iface->ofp_port, &rmps, &n_rmps);
1578 if (error >= 0) {
1579 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1580 n_rmps);
1581 } else {
1582 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1583 }
1584 }
1585
1586 static void
1587 iface_refresh_stats(struct iface *iface)
1588 {
1589 #define IFACE_STATS \
1590 IFACE_STAT(rx_packets, "rx_packets") \
1591 IFACE_STAT(tx_packets, "tx_packets") \
1592 IFACE_STAT(rx_bytes, "rx_bytes") \
1593 IFACE_STAT(tx_bytes, "tx_bytes") \
1594 IFACE_STAT(rx_dropped, "rx_dropped") \
1595 IFACE_STAT(tx_dropped, "tx_dropped") \
1596 IFACE_STAT(rx_errors, "rx_errors") \
1597 IFACE_STAT(tx_errors, "tx_errors") \
1598 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1599 IFACE_STAT(rx_over_errors, "rx_over_err") \
1600 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1601 IFACE_STAT(collisions, "collisions")
1602
1603 #define IFACE_STAT(MEMBER, NAME) NAME,
1604 static char *keys[] = { IFACE_STATS };
1605 #undef IFACE_STAT
1606 int64_t values[ARRAY_SIZE(keys)];
1607 int i;
1608
1609 struct netdev_stats stats;
1610
1611 if (iface_is_synthetic(iface)) {
1612 return;
1613 }
1614
1615 /* Intentionally ignore return value, since errors will set 'stats' to
1616 * all-1s, and we will deal with that correctly below. */
1617 netdev_get_stats(iface->netdev, &stats);
1618
1619 /* Copy statistics into values[] array. */
1620 i = 0;
1621 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1622 IFACE_STATS;
1623 #undef IFACE_STAT
1624 assert(i == ARRAY_SIZE(keys));
1625
1626 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1627 ARRAY_SIZE(keys));
1628 #undef IFACE_STATS
1629 }
1630
1631 static void
1632 br_refresh_stp_status(struct bridge *br)
1633 {
1634 struct ofproto *ofproto = br->ofproto;
1635 struct ofproto_stp_status status;
1636 char *keys[3], *values[3];
1637 size_t i;
1638
1639 if (ofproto_get_stp_status(ofproto, &status)) {
1640 return;
1641 }
1642
1643 if (!status.enabled) {
1644 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1645 return;
1646 }
1647
1648 keys[0] = "stp_bridge_id",
1649 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1650 keys[1] = "stp_designated_root",
1651 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1652 keys[2] = "stp_root_path_cost",
1653 values[2] = xasprintf("%d", status.root_path_cost);
1654
1655 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1656
1657 for (i = 0; i < ARRAY_SIZE(values); i++) {
1658 free(values[i]);
1659 }
1660 }
1661
1662 static void
1663 port_refresh_stp_status(struct port *port)
1664 {
1665 struct ofproto *ofproto = port->bridge->ofproto;
1666 struct iface *iface;
1667 struct ofproto_port_stp_status status;
1668 char *keys[4];
1669 char *str_values[4];
1670 int64_t int_values[3];
1671 size_t i;
1672
1673 if (port_is_synthetic(port)) {
1674 return;
1675 }
1676
1677 /* STP doesn't currently support bonds. */
1678 if (!list_is_singleton(&port->ifaces)) {
1679 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1680 return;
1681 }
1682
1683 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1684
1685 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1686 return;
1687 }
1688
1689 if (!status.enabled) {
1690 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1691 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1692 return;
1693 }
1694
1695 /* Set Status column. */
1696 keys[0] = "stp_port_id";
1697 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1698 keys[1] = "stp_state";
1699 str_values[1] = xstrdup(stp_state_name(status.state));
1700 keys[2] = "stp_sec_in_state";
1701 str_values[2] = xasprintf("%u", status.sec_in_state);
1702 keys[3] = "stp_role";
1703 str_values[3] = xstrdup(stp_role_name(status.role));
1704
1705 ovsrec_port_set_status(port->cfg, keys, str_values,
1706 ARRAY_SIZE(str_values));
1707
1708 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1709 free(str_values[i]);
1710 }
1711
1712 /* Set Statistics column. */
1713 keys[0] = "stp_tx_count";
1714 int_values[0] = status.tx_count;
1715 keys[1] = "stp_rx_count";
1716 int_values[1] = status.rx_count;
1717 keys[2] = "stp_error_count";
1718 int_values[2] = status.error_count;
1719
1720 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1721 ARRAY_SIZE(int_values));
1722 }
1723
1724 static bool
1725 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1726 {
1727 const char *enable;
1728
1729 /* Use other-config:enable-system-stats by preference. */
1730 enable = ovsrec_open_vswitch_get_other_config_value(cfg,
1731 "enable-statistics",
1732 NULL);
1733 if (enable) {
1734 return !strcmp(enable, "true");
1735 }
1736
1737 /* Disable by default. */
1738 return false;
1739 }
1740
1741 static void
1742 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1743 {
1744 struct ovsdb_datum datum;
1745 struct shash stats;
1746
1747 shash_init(&stats);
1748 if (enable_system_stats(cfg)) {
1749 get_system_stats(&stats);
1750 }
1751
1752 ovsdb_datum_from_shash(&datum, &stats);
1753 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1754 &datum);
1755 }
1756
1757 static inline const char *
1758 nx_role_to_str(enum nx_role role)
1759 {
1760 switch (role) {
1761 case NX_ROLE_OTHER:
1762 return "other";
1763 case NX_ROLE_MASTER:
1764 return "master";
1765 case NX_ROLE_SLAVE:
1766 return "slave";
1767 default:
1768 return "*** INVALID ROLE ***";
1769 }
1770 }
1771
1772 static void
1773 refresh_controller_status(void)
1774 {
1775 struct bridge *br;
1776 struct shash info;
1777 const struct ovsrec_controller *cfg;
1778
1779 shash_init(&info);
1780
1781 /* Accumulate status for controllers on all bridges. */
1782 HMAP_FOR_EACH (br, node, &all_bridges) {
1783 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1784 }
1785
1786 /* Update each controller in the database with current status. */
1787 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1788 struct ofproto_controller_info *cinfo =
1789 shash_find_data(&info, cfg->target);
1790
1791 if (cinfo) {
1792 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1793 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1794 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1795 (char **) cinfo->pairs.values,
1796 cinfo->pairs.n);
1797 } else {
1798 ovsrec_controller_set_is_connected(cfg, false);
1799 ovsrec_controller_set_role(cfg, NULL);
1800 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1801 }
1802 }
1803
1804 ofproto_free_ofproto_controller_info(&info);
1805 }
1806
1807 static void
1808 refresh_cfm_stats(void)
1809 {
1810 static struct ovsdb_idl_txn *txn = NULL;
1811
1812 if (!txn) {
1813 struct bridge *br;
1814
1815 txn = ovsdb_idl_txn_create(idl);
1816
1817 HMAP_FOR_EACH (br, node, &all_bridges) {
1818 struct iface *iface;
1819
1820 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1821 iface_refresh_cfm_stats(iface);
1822 }
1823 }
1824 }
1825
1826 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1827 ovsdb_idl_txn_destroy(txn);
1828 txn = NULL;
1829 }
1830 }
1831
1832 /* Performs periodic activity required by bridges that needs to be done with
1833 * the least possible latency.
1834 *
1835 * It makes sense to call this function a couple of times per poll loop, to
1836 * provide a significant performance boost on some benchmarks with ofprotos
1837 * that use the ofproto-dpif implementation. */
1838 void
1839 bridge_run_fast(void)
1840 {
1841 struct bridge *br;
1842
1843 HMAP_FOR_EACH (br, node, &all_bridges) {
1844 ofproto_run_fast(br->ofproto);
1845 }
1846 }
1847
1848 void
1849 bridge_run(void)
1850 {
1851 const struct ovsrec_open_vswitch *cfg;
1852
1853 bool vlan_splinters_changed;
1854 bool database_changed;
1855 struct bridge *br;
1856
1857 /* (Re)configure if necessary. */
1858 database_changed = ovsdb_idl_run(idl);
1859 if (ovsdb_idl_is_lock_contended(idl)) {
1860 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1861 struct bridge *br, *next_br;
1862
1863 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1864 "disabling this process until it goes away");
1865
1866 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1867 bridge_destroy(br);
1868 }
1869 return;
1870 } else if (!ovsdb_idl_has_lock(idl)) {
1871 return;
1872 }
1873 cfg = ovsrec_open_vswitch_first(idl);
1874
1875 /* Let each bridge do the work that it needs to do. */
1876 HMAP_FOR_EACH (br, node, &all_bridges) {
1877 ofproto_run(br->ofproto);
1878 }
1879
1880 /* Re-configure SSL. We do this on every trip through the main loop,
1881 * instead of just when the database changes, because the contents of the
1882 * key and certificate files can change without the database changing.
1883 *
1884 * We do this before bridge_reconfigure() because that function might
1885 * initiate SSL connections and thus requires SSL to be configured. */
1886 if (cfg && cfg->ssl) {
1887 const struct ovsrec_ssl *ssl = cfg->ssl;
1888
1889 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1890 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1891 }
1892
1893 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1894 * has changed. */
1895 vlan_splinters_changed = false;
1896 if (vlan_splinters_enabled_anywhere) {
1897 HMAP_FOR_EACH (br, node, &all_bridges) {
1898 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1899 vlan_splinters_changed = true;
1900 break;
1901 }
1902 }
1903 }
1904
1905 if (database_changed || vlan_splinters_changed) {
1906 if (cfg) {
1907 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1908
1909 bridge_reconfigure(cfg);
1910
1911 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1912 ovsdb_idl_txn_commit(txn);
1913 ovsdb_idl_txn_destroy(txn); /* XXX */
1914 } else {
1915 /* We still need to reconfigure to avoid dangling pointers to
1916 * now-destroyed ovsrec structures inside bridge data. */
1917 static const struct ovsrec_open_vswitch null_cfg;
1918
1919 bridge_reconfigure(&null_cfg);
1920 }
1921 }
1922
1923 /* Refresh system and interface stats if necessary. */
1924 if (time_msec() >= stats_timer) {
1925 if (cfg) {
1926 struct ovsdb_idl_txn *txn;
1927
1928 txn = ovsdb_idl_txn_create(idl);
1929 HMAP_FOR_EACH (br, node, &all_bridges) {
1930 struct port *port;
1931 struct mirror *m;
1932
1933 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1934 struct iface *iface;
1935
1936 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1937 iface_refresh_stats(iface);
1938 iface_refresh_status(iface);
1939 }
1940 }
1941
1942 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1943 mirror_refresh_stats(m);
1944 }
1945
1946 }
1947 refresh_system_stats(cfg);
1948 refresh_controller_status();
1949 ovsdb_idl_txn_commit(txn);
1950 ovsdb_idl_txn_destroy(txn); /* XXX */
1951 }
1952
1953 stats_timer = time_msec() + STATS_INTERVAL;
1954 }
1955
1956 if (time_msec() >= db_limiter) {
1957 struct ovsdb_idl_txn *txn;
1958
1959 txn = ovsdb_idl_txn_create(idl);
1960 HMAP_FOR_EACH (br, node, &all_bridges) {
1961 struct iface *iface;
1962 struct port *port;
1963
1964 br_refresh_stp_status(br);
1965
1966 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1967 port_refresh_stp_status(port);
1968 }
1969
1970 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1971 const char *link_state;
1972 int64_t link_resets;
1973 int current;
1974
1975 if (iface_is_synthetic(iface)) {
1976 continue;
1977 }
1978
1979 current = ofproto_port_is_lacp_current(br->ofproto,
1980 iface->ofp_port);
1981 if (current >= 0) {
1982 bool bl = current;
1983 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1984 } else {
1985 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1986 }
1987
1988 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1989 ovsrec_interface_set_link_state(iface->cfg, link_state);
1990
1991 link_resets = netdev_get_carrier_resets(iface->netdev);
1992 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1993 }
1994 }
1995
1996 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
1997 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1998 }
1999 ovsdb_idl_txn_destroy(txn);
2000 }
2001
2002 refresh_cfm_stats();
2003 }
2004
2005 void
2006 bridge_wait(void)
2007 {
2008 ovsdb_idl_wait(idl);
2009 if (!hmap_is_empty(&all_bridges)) {
2010 struct bridge *br;
2011
2012 HMAP_FOR_EACH (br, node, &all_bridges) {
2013 ofproto_wait(br->ofproto);
2014 }
2015 poll_timer_wait_until(stats_timer);
2016
2017 if (db_limiter > time_msec()) {
2018 poll_timer_wait_until(db_limiter);
2019 }
2020 }
2021 }
2022 \f
2023 /* QoS unixctl user interface functions. */
2024
2025 struct qos_unixctl_show_cbdata {
2026 struct ds *ds;
2027 struct iface *iface;
2028 };
2029
2030 static void
2031 qos_unixctl_show_cb(unsigned int queue_id,
2032 const struct shash *details,
2033 void *aux)
2034 {
2035 struct qos_unixctl_show_cbdata *data = aux;
2036 struct ds *ds = data->ds;
2037 struct iface *iface = data->iface;
2038 struct netdev_queue_stats stats;
2039 struct shash_node *node;
2040 int error;
2041
2042 ds_put_cstr(ds, "\n");
2043 if (queue_id) {
2044 ds_put_format(ds, "Queue %u:\n", queue_id);
2045 } else {
2046 ds_put_cstr(ds, "Default:\n");
2047 }
2048
2049 SHASH_FOR_EACH (node, details) {
2050 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2051 }
2052
2053 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2054 if (!error) {
2055 if (stats.tx_packets != UINT64_MAX) {
2056 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2057 }
2058
2059 if (stats.tx_bytes != UINT64_MAX) {
2060 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2061 }
2062
2063 if (stats.tx_errors != UINT64_MAX) {
2064 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2065 }
2066 } else {
2067 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2068 queue_id, strerror(error));
2069 }
2070 }
2071
2072 static void
2073 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2074 const char *argv[], void *aux OVS_UNUSED)
2075 {
2076 struct ds ds = DS_EMPTY_INITIALIZER;
2077 struct shash sh = SHASH_INITIALIZER(&sh);
2078 struct iface *iface;
2079 const char *type;
2080 struct shash_node *node;
2081 struct qos_unixctl_show_cbdata data;
2082 int error;
2083
2084 iface = iface_find(argv[1]);
2085 if (!iface) {
2086 unixctl_command_reply_error(conn, "no such interface");
2087 return;
2088 }
2089
2090 netdev_get_qos(iface->netdev, &type, &sh);
2091
2092 if (*type != '\0') {
2093 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2094
2095 SHASH_FOR_EACH (node, &sh) {
2096 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2097 }
2098
2099 data.ds = &ds;
2100 data.iface = iface;
2101 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2102
2103 if (error) {
2104 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2105 }
2106 unixctl_command_reply(conn, ds_cstr(&ds));
2107 } else {
2108 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2109 unixctl_command_reply_error(conn, ds_cstr(&ds));
2110 }
2111
2112 shash_destroy_free_data(&sh);
2113 ds_destroy(&ds);
2114 }
2115 \f
2116 /* Bridge reconfiguration functions. */
2117 static void
2118 bridge_create(const struct ovsrec_bridge *br_cfg)
2119 {
2120 struct bridge *br;
2121
2122 assert(!bridge_lookup(br_cfg->name));
2123 br = xzalloc(sizeof *br);
2124
2125 br->name = xstrdup(br_cfg->name);
2126 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2127 br->cfg = br_cfg;
2128
2129 /* Derive the default Ethernet address from the bridge's UUID. This should
2130 * be unique and it will be stable between ovs-vswitchd runs. */
2131 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2132 eth_addr_mark_random(br->default_ea);
2133
2134 hmap_init(&br->ports);
2135 hmap_init(&br->ifaces);
2136 hmap_init(&br->iface_by_name);
2137 hmap_init(&br->mirrors);
2138
2139 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2140 }
2141
2142 static void
2143 bridge_destroy(struct bridge *br)
2144 {
2145 if (br) {
2146 struct mirror *mirror, *next_mirror;
2147 struct port *port, *next_port;
2148
2149 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2150 port_destroy(port);
2151 }
2152 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2153 mirror_destroy(mirror);
2154 }
2155 hmap_remove(&all_bridges, &br->node);
2156 ofproto_destroy(br->ofproto);
2157 hmap_destroy(&br->ifaces);
2158 hmap_destroy(&br->ports);
2159 hmap_destroy(&br->iface_by_name);
2160 hmap_destroy(&br->mirrors);
2161 free(br->name);
2162 free(br->type);
2163 free(br);
2164 }
2165 }
2166
2167 static struct bridge *
2168 bridge_lookup(const char *name)
2169 {
2170 struct bridge *br;
2171
2172 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2173 if (!strcmp(br->name, name)) {
2174 return br;
2175 }
2176 }
2177 return NULL;
2178 }
2179
2180 /* Handle requests for a listing of all flows known by the OpenFlow
2181 * stack, including those normally hidden. */
2182 static void
2183 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2184 const char *argv[], void *aux OVS_UNUSED)
2185 {
2186 struct bridge *br;
2187 struct ds results;
2188
2189 br = bridge_lookup(argv[1]);
2190 if (!br) {
2191 unixctl_command_reply_error(conn, "Unknown bridge");
2192 return;
2193 }
2194
2195 ds_init(&results);
2196 ofproto_get_all_flows(br->ofproto, &results);
2197
2198 unixctl_command_reply(conn, ds_cstr(&results));
2199 ds_destroy(&results);
2200 }
2201
2202 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2203 * connections and reconnect. If BRIDGE is not specified, then all bridges
2204 * drop their controller connections and reconnect. */
2205 static void
2206 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2207 const char *argv[], void *aux OVS_UNUSED)
2208 {
2209 struct bridge *br;
2210 if (argc > 1) {
2211 br = bridge_lookup(argv[1]);
2212 if (!br) {
2213 unixctl_command_reply_error(conn, "Unknown bridge");
2214 return;
2215 }
2216 ofproto_reconnect_controllers(br->ofproto);
2217 } else {
2218 HMAP_FOR_EACH (br, node, &all_bridges) {
2219 ofproto_reconnect_controllers(br->ofproto);
2220 }
2221 }
2222 unixctl_command_reply(conn, NULL);
2223 }
2224
2225 static size_t
2226 bridge_get_controllers(const struct bridge *br,
2227 struct ovsrec_controller ***controllersp)
2228 {
2229 struct ovsrec_controller **controllers;
2230 size_t n_controllers;
2231
2232 controllers = br->cfg->controller;
2233 n_controllers = br->cfg->n_controller;
2234
2235 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2236 controllers = NULL;
2237 n_controllers = 0;
2238 }
2239
2240 if (controllersp) {
2241 *controllersp = controllers;
2242 }
2243 return n_controllers;
2244 }
2245
2246 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2247 * those configured in 'br->cfg'. */
2248 static void
2249 bridge_add_del_ports(struct bridge *br,
2250 const unsigned long int *splinter_vlans)
2251 {
2252 struct port *port, *next;
2253 struct shash_node *node;
2254 struct shash new_ports;
2255 size_t i;
2256
2257 /* Collect new ports. */
2258 shash_init(&new_ports);
2259 for (i = 0; i < br->cfg->n_ports; i++) {
2260 const char *name = br->cfg->ports[i]->name;
2261 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2262 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2263 br->name, name);
2264 }
2265 }
2266 if (bridge_get_controllers(br, NULL)
2267 && !shash_find(&new_ports, br->name)) {
2268 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2269 br->name, br->name);
2270
2271 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2272 br->synth_local_port.n_interfaces = 1;
2273 br->synth_local_port.name = br->name;
2274
2275 br->synth_local_iface.name = br->name;
2276 br->synth_local_iface.type = "internal";
2277
2278 br->synth_local_ifacep = &br->synth_local_iface;
2279
2280 shash_add(&new_ports, br->name, &br->synth_local_port);
2281 }
2282
2283 if (splinter_vlans) {
2284 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2285 }
2286
2287 /* Get rid of deleted ports.
2288 * Get rid of deleted interfaces on ports that still exist. */
2289 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2290 port->cfg = shash_find_data(&new_ports, port->name);
2291 if (!port->cfg) {
2292 port_destroy(port);
2293 } else {
2294 port_del_ifaces(port);
2295 }
2296 }
2297
2298 /* Create new ports.
2299 * Add new interfaces to existing ports. */
2300 SHASH_FOR_EACH (node, &new_ports) {
2301 struct port *port = port_lookup(br, node->name);
2302 if (!port) {
2303 struct ovsrec_port *cfg = node->data;
2304 port = port_create(br, cfg);
2305 }
2306 port_add_ifaces(port);
2307 if (list_is_empty(&port->ifaces)) {
2308 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2309 br->name, port->name);
2310 port_destroy(port);
2311 }
2312 }
2313 shash_destroy(&new_ports);
2314 }
2315
2316 /* Initializes 'oc' appropriately as a management service controller for
2317 * 'br'.
2318 *
2319 * The caller must free oc->target when it is no longer needed. */
2320 static void
2321 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2322 struct ofproto_controller *oc)
2323 {
2324 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2325 oc->max_backoff = 0;
2326 oc->probe_interval = 60;
2327 oc->band = OFPROTO_OUT_OF_BAND;
2328 oc->rate_limit = 0;
2329 oc->burst_limit = 0;
2330 oc->enable_async_msgs = true;
2331 }
2332
2333 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2334 static void
2335 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2336 struct ofproto_controller *oc)
2337 {
2338 oc->target = c->target;
2339 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2340 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2341 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2342 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2343 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2344 oc->burst_limit = (c->controller_burst_limit
2345 ? *c->controller_burst_limit : 0);
2346 oc->enable_async_msgs = (!c->enable_async_messages
2347 || *c->enable_async_messages);
2348 }
2349
2350 /* Configures the IP stack for 'br''s local interface properly according to the
2351 * configuration in 'c'. */
2352 static void
2353 bridge_configure_local_iface_netdev(struct bridge *br,
2354 struct ovsrec_controller *c)
2355 {
2356 struct netdev *netdev;
2357 struct in_addr mask, gateway;
2358
2359 struct iface *local_iface;
2360 struct in_addr ip;
2361
2362 /* If there's no local interface or no IP address, give up. */
2363 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2364 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2365 return;
2366 }
2367
2368 /* Bring up the local interface. */
2369 netdev = local_iface->netdev;
2370 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2371
2372 /* Configure the IP address and netmask. */
2373 if (!c->local_netmask
2374 || !inet_aton(c->local_netmask, &mask)
2375 || !mask.s_addr) {
2376 mask.s_addr = guess_netmask(ip.s_addr);
2377 }
2378 if (!netdev_set_in4(netdev, ip, mask)) {
2379 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2380 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2381 }
2382
2383 /* Configure the default gateway. */
2384 if (c->local_gateway
2385 && inet_aton(c->local_gateway, &gateway)
2386 && gateway.s_addr) {
2387 if (!netdev_add_router(netdev, gateway)) {
2388 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2389 br->name, IP_ARGS(&gateway.s_addr));
2390 }
2391 }
2392 }
2393
2394 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2395 * in either string are treated as equal to any number of slashes in the other,
2396 * e.g. "x///y" is equal to "x/y". */
2397 static bool
2398 equal_pathnames(const char *a, const char *b)
2399 {
2400 while (*a == *b) {
2401 if (*a == '/') {
2402 a += strspn(a, "/");
2403 b += strspn(b, "/");
2404 } else if (*a == '\0') {
2405 return true;
2406 } else {
2407 a++;
2408 b++;
2409 }
2410 }
2411 return false;
2412 }
2413
2414 static void
2415 bridge_configure_remotes(struct bridge *br,
2416 const struct sockaddr_in *managers, size_t n_managers)
2417 {
2418 const char *disable_ib_str, *queue_id_str;
2419 bool disable_in_band = false;
2420 int queue_id;
2421
2422 struct ovsrec_controller **controllers;
2423 size_t n_controllers;
2424
2425 enum ofproto_fail_mode fail_mode;
2426
2427 struct ofproto_controller *ocs;
2428 size_t n_ocs;
2429 size_t i;
2430
2431 /* Check if we should disable in-band control on this bridge. */
2432 disable_ib_str = ovsrec_bridge_get_other_config_value(br->cfg,
2433 "disable-in-band",
2434 NULL);
2435 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2436 disable_in_band = true;
2437 }
2438
2439 /* Set OpenFlow queue ID for in-band control. */
2440 queue_id_str = ovsrec_bridge_get_other_config_value(br->cfg,
2441 "in-band-queue",
2442 NULL);
2443 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2444 ofproto_set_in_band_queue(br->ofproto, queue_id);
2445
2446 if (disable_in_band) {
2447 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2448 } else {
2449 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2450 }
2451
2452 n_controllers = bridge_get_controllers(br, &controllers);
2453
2454 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2455 n_ocs = 0;
2456
2457 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2458 for (i = 0; i < n_controllers; i++) {
2459 struct ovsrec_controller *c = controllers[i];
2460
2461 if (!strncmp(c->target, "punix:", 6)
2462 || !strncmp(c->target, "unix:", 5)) {
2463 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2464 char *whitelist;
2465
2466 whitelist = xasprintf("unix:%s/%s.controller",
2467 ovs_rundir(), br->name);
2468 if (!equal_pathnames(c->target, whitelist)) {
2469 /* Prevent remote ovsdb-server users from accessing arbitrary
2470 * Unix domain sockets and overwriting arbitrary local
2471 * files. */
2472 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2473 "controller \"%s\" due to possibility for remote "
2474 "exploit. Instead, specify whitelisted \"%s\" or "
2475 "connect to \"unix:%s/%s.mgmt\" (which is always "
2476 "available without special configuration).",
2477 br->name, c->target, whitelist,
2478 ovs_rundir(), br->name);
2479 free(whitelist);
2480 continue;
2481 }
2482
2483 free(whitelist);
2484 }
2485
2486 bridge_configure_local_iface_netdev(br, c);
2487 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2488 if (disable_in_band) {
2489 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2490 }
2491 n_ocs++;
2492 }
2493
2494 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2495 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2496 free(ocs);
2497
2498 /* Set the fail-mode. */
2499 fail_mode = !br->cfg->fail_mode
2500 || !strcmp(br->cfg->fail_mode, "standalone")
2501 ? OFPROTO_FAIL_STANDALONE
2502 : OFPROTO_FAIL_SECURE;
2503 ofproto_set_fail_mode(br->ofproto, fail_mode);
2504
2505 /* Configure OpenFlow controller connection snooping. */
2506 if (!ofproto_has_snoops(br->ofproto)) {
2507 struct sset snoops;
2508
2509 sset_init(&snoops);
2510 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2511 ovs_rundir(), br->name));
2512 ofproto_set_snoops(br->ofproto, &snoops);
2513 sset_destroy(&snoops);
2514 }
2515 }
2516
2517 static void
2518 bridge_configure_tables(struct bridge *br)
2519 {
2520 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2521 int n_tables;
2522 int i, j;
2523
2524 n_tables = ofproto_get_n_tables(br->ofproto);
2525 j = 0;
2526 for (i = 0; i < n_tables; i++) {
2527 struct ofproto_table_settings s;
2528
2529 s.name = NULL;
2530 s.max_flows = UINT_MAX;
2531 s.groups = NULL;
2532 s.n_groups = 0;
2533
2534 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2535 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2536
2537 s.name = cfg->name;
2538 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2539 s.max_flows = *cfg->flow_limit;
2540 }
2541 if (cfg->overflow_policy
2542 && !strcmp(cfg->overflow_policy, "evict")) {
2543 size_t k;
2544
2545 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2546 for (k = 0; k < cfg->n_groups; k++) {
2547 const char *string = cfg->groups[k];
2548 char *msg;
2549
2550 msg = mf_parse_subfield__(&s.groups[k], &string);
2551 if (msg) {
2552 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2553 "'groups' (%s)", br->name, i, msg);
2554 free(msg);
2555 } else if (*string) {
2556 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2557 "element '%s' contains trailing garbage",
2558 br->name, i, cfg->groups[k]);
2559 } else {
2560 s.n_groups++;
2561 }
2562 }
2563 }
2564 }
2565
2566 ofproto_configure_table(br->ofproto, i, &s);
2567
2568 free(s.groups);
2569 }
2570 for (; j < br->cfg->n_flow_tables; j++) {
2571 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2572 "%"PRId64" not supported by this datapath", br->name,
2573 br->cfg->key_flow_tables[j]);
2574 }
2575 }
2576 \f
2577 /* Port functions. */
2578
2579 static struct port *
2580 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2581 {
2582 struct port *port;
2583
2584 port = xzalloc(sizeof *port);
2585 port->bridge = br;
2586 port->name = xstrdup(cfg->name);
2587 port->cfg = cfg;
2588 list_init(&port->ifaces);
2589
2590 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2591
2592 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2593
2594 return port;
2595 }
2596
2597 /* Deletes interfaces from 'port' that are no longer configured for it. */
2598 static void
2599 port_del_ifaces(struct port *port)
2600 {
2601 struct iface *iface, *next;
2602 struct sset new_ifaces;
2603 size_t i;
2604
2605 /* Collect list of new interfaces. */
2606 sset_init(&new_ifaces);
2607 for (i = 0; i < port->cfg->n_interfaces; i++) {
2608 const char *name = port->cfg->interfaces[i]->name;
2609 const char *type = port->cfg->interfaces[i]->name;
2610 if (strcmp(type, "null")) {
2611 sset_add(&new_ifaces, name);
2612 }
2613 }
2614
2615 /* Get rid of deleted interfaces. */
2616 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2617 if (!sset_contains(&new_ifaces, iface->name)) {
2618 iface_destroy(iface);
2619 }
2620 }
2621
2622 sset_destroy(&new_ifaces);
2623 }
2624
2625 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2626 * existing ones. */
2627 static void
2628 port_add_ifaces(struct port *port)
2629 {
2630 struct shash new_ifaces;
2631 struct shash_node *node;
2632 size_t i;
2633
2634 /* Collect new ifaces. */
2635 shash_init(&new_ifaces);
2636 for (i = 0; i < port->cfg->n_interfaces; i++) {
2637 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2638 if (strcmp(cfg->type, "null")
2639 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2640 VLOG_WARN("port %s: %s specified twice as port interface",
2641 port->name, cfg->name);
2642 iface_clear_db_record(cfg);
2643 }
2644 }
2645
2646 /* Create new interfaces.
2647 * Update interface types and 'cfg' members. */
2648 SHASH_FOR_EACH (node, &new_ifaces) {
2649 const struct ovsrec_interface *cfg = node->data;
2650 const char *iface_name = node->name;
2651 struct iface *iface;
2652
2653 iface = iface_lookup(port->bridge, iface_name);
2654 if (!iface) {
2655 iface = iface_create(port, cfg);
2656 } else {
2657 iface->cfg = cfg;
2658 }
2659
2660 /* Determine interface type. The local port always has type
2661 * "internal". Other ports take their type from the database and
2662 * default to "system" if none is specified. */
2663 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2664 : cfg->type[0] ? cfg->type
2665 : "system");
2666 }
2667 shash_destroy(&new_ifaces);
2668 }
2669
2670 static void
2671 port_destroy(struct port *port)
2672 {
2673 if (port) {
2674 struct bridge *br = port->bridge;
2675 struct iface *iface, *next;
2676
2677 if (br->ofproto) {
2678 ofproto_bundle_unregister(br->ofproto, port);
2679 }
2680
2681 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2682 iface_destroy(iface);
2683 }
2684
2685 hmap_remove(&br->ports, &port->hmap_node);
2686
2687 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2688
2689 free(port->name);
2690 free(port);
2691 }
2692 }
2693
2694 static struct port *
2695 port_lookup(const struct bridge *br, const char *name)
2696 {
2697 struct port *port;
2698
2699 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2700 &br->ports) {
2701 if (!strcmp(port->name, name)) {
2702 return port;
2703 }
2704 }
2705 return NULL;
2706 }
2707
2708 static bool
2709 enable_lacp(struct port *port, bool *activep)
2710 {
2711 if (!port->cfg->lacp) {
2712 /* XXX when LACP implementation has been sufficiently tested, enable by
2713 * default and make active on bonded ports. */
2714 return false;
2715 } else if (!strcmp(port->cfg->lacp, "off")) {
2716 return false;
2717 } else if (!strcmp(port->cfg->lacp, "active")) {
2718 *activep = true;
2719 return true;
2720 } else if (!strcmp(port->cfg->lacp, "passive")) {
2721 *activep = false;
2722 return true;
2723 } else {
2724 VLOG_WARN("port %s: unknown LACP mode %s",
2725 port->name, port->cfg->lacp);
2726 return false;
2727 }
2728 }
2729
2730 static struct lacp_settings *
2731 port_configure_lacp(struct port *port, struct lacp_settings *s)
2732 {
2733 const char *lacp_time, *system_id;
2734 long long int custom_time;
2735 int priority;
2736
2737 if (!enable_lacp(port, &s->active)) {
2738 return NULL;
2739 }
2740
2741 s->name = port->name;
2742
2743 system_id = ovsrec_port_get_other_config_value(port->cfg, "lacp-system-id",
2744 NULL);
2745 if (system_id) {
2746 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2747 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2748 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2749 " address.", port->name, system_id);
2750 return NULL;
2751 }
2752 } else {
2753 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2754 }
2755
2756 if (eth_addr_is_zero(s->id)) {
2757 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2758 return NULL;
2759 }
2760
2761 /* Prefer bondable links if unspecified. */
2762 priority = atoi(ovsrec_port_get_other_config_value(port->cfg,
2763 "lacp-system-priority",
2764 "0"));
2765 s->priority = (priority > 0 && priority <= UINT16_MAX
2766 ? priority
2767 : UINT16_MAX - !list_is_short(&port->ifaces));
2768
2769 s->heartbeat = !strcmp(ovsrec_port_get_other_config_value(port->cfg,
2770 "lacp-heartbeat",
2771 "false"),
2772 "true");
2773
2774 lacp_time = ovsrec_port_get_other_config_value(port->cfg, "lacp-time",
2775 "slow");
2776 custom_time = atoi(lacp_time);
2777 if (!strcmp(lacp_time, "fast")) {
2778 s->lacp_time = LACP_TIME_FAST;
2779 } else if (!strcmp(lacp_time, "slow")) {
2780 s->lacp_time = LACP_TIME_SLOW;
2781 } else if (custom_time > 0) {
2782 s->lacp_time = LACP_TIME_CUSTOM;
2783 s->custom_time = custom_time;
2784 } else {
2785 s->lacp_time = LACP_TIME_SLOW;
2786 }
2787
2788 return s;
2789 }
2790
2791 static void
2792 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2793 {
2794 int priority, portid, key;
2795
2796 portid = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2797 "lacp-port-id",
2798 "0"));
2799 priority =
2800 atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2801 "lacp-port-priority",
2802 "0"));
2803 key = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2804 "lacp-aggregation-key",
2805 "0"));
2806
2807 if (portid <= 0 || portid > UINT16_MAX) {
2808 portid = iface->ofp_port;
2809 }
2810
2811 if (priority <= 0 || priority > UINT16_MAX) {
2812 priority = UINT16_MAX;
2813 }
2814
2815 if (key < 0 || key > UINT16_MAX) {
2816 key = 0;
2817 }
2818
2819 s->name = iface->name;
2820 s->id = portid;
2821 s->priority = priority;
2822 s->key = key;
2823 }
2824
2825 static void
2826 port_configure_bond(struct port *port, struct bond_settings *s,
2827 uint32_t *bond_stable_ids)
2828 {
2829 const char *detect_s;
2830 struct iface *iface;
2831 int miimon_interval;
2832 size_t i;
2833
2834 s->name = port->name;
2835 s->balance = BM_AB;
2836 if (port->cfg->bond_mode) {
2837 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2838 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2839 port->name, port->cfg->bond_mode,
2840 bond_mode_to_string(s->balance));
2841 }
2842 } else {
2843 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2844
2845 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2846 * active-backup. At some point we should remove this warning. */
2847 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2848 " in previous versions, the default bond_mode was"
2849 " balance-slb", port->name,
2850 bond_mode_to_string(s->balance));
2851 }
2852 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2853 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2854 "please use another bond type or disable flood_vlans",
2855 port->name);
2856 }
2857
2858 miimon_interval =
2859 atoi(ovsrec_port_get_other_config_value(port->cfg,
2860 "bond-miimon-interval", "0"));
2861 if (miimon_interval <= 0) {
2862 miimon_interval = 200;
2863 }
2864
2865 detect_s = ovsrec_port_get_other_config_value(port->cfg,
2866 "bond-detect-mode",
2867 "carrier");
2868 if (!strcmp(detect_s, "carrier")) {
2869 miimon_interval = 0;
2870 } else if (strcmp(detect_s, "miimon")) {
2871 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2872 "defaulting to carrier", port->name, detect_s);
2873 miimon_interval = 0;
2874 }
2875
2876 s->up_delay = MAX(0, port->cfg->bond_updelay);
2877 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2878 s->basis = atoi(ovsrec_port_get_other_config_value(port->cfg,
2879 "bond-hash-basis",
2880 "0"));
2881 s->rebalance_interval = atoi(
2882 ovsrec_port_get_other_config_value(port->cfg,
2883 "bond-rebalance-interval",
2884 "10000"));
2885 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2886 s->rebalance_interval = 1000;
2887 }
2888
2889 s->fake_iface = port->cfg->bond_fake_iface;
2890
2891 i = 0;
2892 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2893 long long stable_id;
2894
2895 stable_id =
2896 atoll(ovsrec_interface_get_other_config_value(iface->cfg,
2897 "bond-stable-id",
2898 "0"));
2899 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2900 stable_id = iface->ofp_port;
2901 }
2902 bond_stable_ids[i++] = stable_id;
2903
2904 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2905 }
2906 }
2907
2908 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2909 * instead of obtaining it from the database. */
2910 static bool
2911 port_is_synthetic(const struct port *port)
2912 {
2913 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2914 }
2915 \f
2916 /* Interface functions. */
2917
2918 static struct iface *
2919 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2920 {
2921 struct bridge *br = port->bridge;
2922 struct iface *iface;
2923 char *name = if_cfg->name;
2924
2925 iface = xzalloc(sizeof *iface);
2926 iface->port = port;
2927 iface->name = xstrdup(name);
2928 iface->ofp_port = -1;
2929 iface->tag = tag_create_random();
2930 iface->netdev = NULL;
2931 iface->cfg = if_cfg;
2932
2933 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2934
2935 list_push_back(&port->ifaces, &iface->port_elem);
2936
2937 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2938
2939 return iface;
2940 }
2941
2942 static void
2943 iface_destroy(struct iface *iface)
2944 {
2945 if (iface) {
2946 struct port *port = iface->port;
2947 struct bridge *br = port->bridge;
2948
2949 if (br->ofproto && iface->ofp_port >= 0) {
2950 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2951 }
2952
2953 if (iface->ofp_port >= 0) {
2954 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2955 }
2956
2957 list_remove(&iface->port_elem);
2958 hmap_remove(&br->iface_by_name, &iface->name_node);
2959
2960 netdev_close(iface->netdev);
2961
2962 free(iface->name);
2963 free(iface);
2964 }
2965 }
2966
2967 static struct iface *
2968 iface_lookup(const struct bridge *br, const char *name)
2969 {
2970 struct iface *iface;
2971
2972 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2973 &br->iface_by_name) {
2974 if (!strcmp(iface->name, name)) {
2975 return iface;
2976 }
2977 }
2978
2979 return NULL;
2980 }
2981
2982 static struct iface *
2983 iface_find(const char *name)
2984 {
2985 const struct bridge *br;
2986
2987 HMAP_FOR_EACH (br, node, &all_bridges) {
2988 struct iface *iface = iface_lookup(br, name);
2989
2990 if (iface) {
2991 return iface;
2992 }
2993 }
2994 return NULL;
2995 }
2996
2997 static struct iface *
2998 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2999 {
3000 struct iface *iface;
3001
3002 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3003 hash_int(ofp_port, 0), &br->ifaces) {
3004 if (iface->ofp_port == ofp_port) {
3005 return iface;
3006 }
3007 }
3008 return NULL;
3009 }
3010
3011 /* Set Ethernet address of 'iface', if one is specified in the configuration
3012 * file. */
3013 static void
3014 iface_set_mac(struct iface *iface)
3015 {
3016 uint8_t ea[ETH_ADDR_LEN];
3017
3018 if (!strcmp(iface->type, "internal")
3019 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3020 if (iface->ofp_port == OFPP_LOCAL) {
3021 VLOG_ERR("interface %s: ignoring mac in Interface record "
3022 "(use Bridge record to set local port's mac)",
3023 iface->name);
3024 } else if (eth_addr_is_multicast(ea)) {
3025 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3026 iface->name);
3027 } else {
3028 int error = netdev_set_etheraddr(iface->netdev, ea);
3029 if (error) {
3030 VLOG_ERR("interface %s: setting MAC failed (%s)",
3031 iface->name, strerror(error));
3032 }
3033 }
3034 }
3035 }
3036
3037 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3038 static void
3039 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3040 {
3041 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3042 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3043 }
3044 }
3045
3046 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3047 * sets the "ofport" field to -1.
3048 *
3049 * This is appropriate when 'if_cfg''s interface cannot be created or is
3050 * otherwise invalid. */
3051 static void
3052 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3053 {
3054 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3055 iface_set_ofport(if_cfg, -1);
3056 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3057 ovsrec_interface_set_admin_state(if_cfg, NULL);
3058 ovsrec_interface_set_duplex(if_cfg, NULL);
3059 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3060 ovsrec_interface_set_link_state(if_cfg, NULL);
3061 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3062 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3063 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3064 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3065 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3066 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3067 }
3068 }
3069
3070 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3071 *
3072 * The value strings in '*shash' are taken directly from values[], not copied,
3073 * so the caller should not modify or free them. */
3074 static void
3075 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3076 struct shash *shash)
3077 {
3078 size_t i;
3079
3080 shash_init(shash);
3081 for (i = 0; i < n; i++) {
3082 shash_add(shash, keys[i], values[i]);
3083 }
3084 }
3085
3086 /* Creates 'keys' and 'values' arrays from 'shash'.
3087 *
3088 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3089 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3090 * are populated with with strings taken directly from 'shash' and thus have
3091 * the same ownership of the key-value pairs in shash.
3092 */
3093 static void
3094 shash_to_ovs_idl_map(struct shash *shash,
3095 char ***keys, char ***values, size_t *n)
3096 {
3097 size_t i, count;
3098 char **k, **v;
3099 struct shash_node *sn;
3100
3101 count = shash_count(shash);
3102
3103 k = xmalloc(count * sizeof *k);
3104 v = xmalloc(count * sizeof *v);
3105
3106 i = 0;
3107 SHASH_FOR_EACH(sn, shash) {
3108 k[i] = sn->name;
3109 v[i] = sn->data;
3110 i++;
3111 }
3112
3113 *n = count;
3114 *keys = k;
3115 *values = v;
3116 }
3117
3118 struct iface_delete_queues_cbdata {
3119 struct netdev *netdev;
3120 const struct ovsdb_datum *queues;
3121 };
3122
3123 static bool
3124 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3125 {
3126 union ovsdb_atom atom;
3127
3128 atom.integer = target;
3129 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3130 }
3131
3132 static void
3133 iface_delete_queues(unsigned int queue_id,
3134 const struct shash *details OVS_UNUSED, void *cbdata_)
3135 {
3136 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3137
3138 if (!queue_ids_include(cbdata->queues, queue_id)) {
3139 netdev_delete_queue(cbdata->netdev, queue_id);
3140 }
3141 }
3142
3143 static void
3144 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3145 {
3146 struct ofpbuf queues_buf;
3147
3148 ofpbuf_init(&queues_buf, 0);
3149
3150 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3151 netdev_set_qos(iface->netdev, NULL, NULL);
3152 } else {
3153 struct iface_delete_queues_cbdata cbdata;
3154 struct shash details;
3155 bool queue_zero;
3156 size_t i;
3157
3158 /* Configure top-level Qos for 'iface'. */
3159 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3160 qos->n_other_config, &details);
3161 netdev_set_qos(iface->netdev, qos->type, &details);
3162 shash_destroy(&details);
3163
3164 /* Deconfigure queues that were deleted. */
3165 cbdata.netdev = iface->netdev;
3166 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3167 OVSDB_TYPE_UUID);
3168 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3169
3170 /* Configure queues for 'iface'. */
3171 queue_zero = false;
3172 for (i = 0; i < qos->n_queues; i++) {
3173 const struct ovsrec_queue *queue = qos->value_queues[i];
3174 unsigned int queue_id = qos->key_queues[i];
3175
3176 if (queue_id == 0) {
3177 queue_zero = true;
3178 }
3179
3180 if (queue->n_dscp == 1) {
3181 struct ofproto_port_queue *port_queue;
3182
3183 port_queue = ofpbuf_put_uninit(&queues_buf,
3184 sizeof *port_queue);
3185 port_queue->queue = queue_id;
3186 port_queue->dscp = queue->dscp[0];
3187 }
3188
3189 shash_from_ovs_idl_map(queue->key_other_config,
3190 queue->value_other_config,
3191 queue->n_other_config, &details);
3192 netdev_set_queue(iface->netdev, queue_id, &details);
3193 shash_destroy(&details);
3194 }
3195 if (!queue_zero) {
3196 shash_init(&details);
3197 netdev_set_queue(iface->netdev, 0, &details);
3198 shash_destroy(&details);
3199 }
3200 }
3201
3202 if (iface->ofp_port >= 0) {
3203 const struct ofproto_port_queue *port_queues = queues_buf.data;
3204 size_t n_queues = queues_buf.size / sizeof *port_queues;
3205
3206 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3207 port_queues, n_queues);
3208 }
3209
3210 netdev_set_policing(iface->netdev,
3211 iface->cfg->ingress_policing_rate,
3212 iface->cfg->ingress_policing_burst);
3213
3214 ofpbuf_uninit(&queues_buf);
3215 }
3216
3217 static void
3218 iface_configure_cfm(struct iface *iface)
3219 {
3220 const struct ovsrec_interface *cfg = iface->cfg;
3221 const char *extended_str, *opstate_str;
3222 const char *cfm_ccm_vlan;
3223 struct cfm_settings s;
3224
3225 if (!cfg->n_cfm_mpid) {
3226 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3227 return;
3228 }
3229
3230 s.mpid = *cfg->cfm_mpid;
3231 s.interval = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3232 "cfm_interval",
3233 "0"));
3234 cfm_ccm_vlan = ovsrec_interface_get_other_config_value(iface->cfg,
3235 "cfm_ccm_vlan",
3236 "0");
3237 s.ccm_pcp = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3238 "cfm_ccm_pcp",
3239 "0"));
3240 if (s.interval <= 0) {
3241 s.interval = 1000;
3242 }
3243
3244 if (!strcasecmp("random", cfm_ccm_vlan)) {
3245 s.ccm_vlan = CFM_RANDOM_VLAN;
3246 } else {
3247 s.ccm_vlan = atoi(cfm_ccm_vlan);
3248 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3249 s.ccm_vlan = 0;
3250 }
3251 }
3252
3253 extended_str = ovsrec_interface_get_other_config_value(iface->cfg,
3254 "cfm_extended",
3255 "false");
3256 s.extended = !strcasecmp("true", extended_str);
3257
3258 opstate_str = ovsrec_interface_get_other_config_value(iface->cfg,
3259 "cfm_opstate",
3260 "up");
3261 s.opup = !strcasecmp("up", opstate_str);
3262
3263 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3264 }
3265
3266 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3267 * instead of obtaining it from the database. */
3268 static bool
3269 iface_is_synthetic(const struct iface *iface)
3270 {
3271 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3272 }
3273 \f
3274 /* Port mirroring. */
3275
3276 static struct mirror *
3277 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3278 {
3279 struct mirror *m;
3280
3281 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3282 if (uuid_equals(uuid, &m->uuid)) {
3283 return m;
3284 }
3285 }
3286 return NULL;
3287 }
3288
3289 static void
3290 bridge_configure_mirrors(struct bridge *br)
3291 {
3292 const struct ovsdb_datum *mc;
3293 unsigned long *flood_vlans;
3294 struct mirror *m, *next;
3295 size_t i;
3296
3297 /* Get rid of deleted mirrors. */
3298 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3299 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3300 union ovsdb_atom atom;
3301
3302 atom.uuid = m->uuid;
3303 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3304 mirror_destroy(m);
3305 }
3306 }
3307
3308 /* Add new mirrors and reconfigure existing ones. */
3309 for (i = 0; i < br->cfg->n_mirrors; i++) {
3310 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3311 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3312 if (!m) {
3313 m = mirror_create(br, cfg);
3314 }
3315 m->cfg = cfg;
3316 if (!mirror_configure(m)) {
3317 mirror_destroy(m);
3318 }
3319 }
3320
3321 /* Update flooded vlans (for RSPAN). */
3322 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3323 br->cfg->n_flood_vlans);
3324 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3325 bitmap_free(flood_vlans);
3326 }
3327
3328 static struct mirror *
3329 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3330 {
3331 struct mirror *m;
3332
3333 m = xzalloc(sizeof *m);
3334 m->uuid = cfg->header_.uuid;
3335 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3336 m->bridge = br;
3337 m->name = xstrdup(cfg->name);
3338
3339 return m;
3340 }
3341
3342 static void
3343 mirror_destroy(struct mirror *m)
3344 {
3345 if (m) {
3346 struct bridge *br = m->bridge;
3347
3348 if (br->ofproto) {
3349 ofproto_mirror_unregister(br->ofproto, m);
3350 }
3351
3352 hmap_remove(&br->mirrors, &m->hmap_node);
3353 free(m->name);
3354 free(m);
3355 }
3356 }
3357
3358 static void
3359 mirror_collect_ports(struct mirror *m,
3360 struct ovsrec_port **in_ports, int n_in_ports,
3361 void ***out_portsp, size_t *n_out_portsp)
3362 {
3363 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3364 size_t n_out_ports = 0;
3365 size_t i;
3366
3367 for (i = 0; i < n_in_ports; i++) {
3368 const char *name = in_ports[i]->name;
3369 struct port *port = port_lookup(m->bridge, name);
3370 if (port) {
3371 out_ports[n_out_ports++] = port;
3372 } else {
3373 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3374 "port %s", m->bridge->name, m->name, name);
3375 }
3376 }
3377 *out_portsp = out_ports;
3378 *n_out_portsp = n_out_ports;
3379 }
3380
3381 static bool
3382 mirror_configure(struct mirror *m)
3383 {
3384 const struct ovsrec_mirror *cfg = m->cfg;
3385 struct ofproto_mirror_settings s;
3386
3387 /* Set name. */
3388 if (strcmp(cfg->name, m->name)) {
3389 free(m->name);
3390 m->name = xstrdup(cfg->name);
3391 }
3392 s.name = m->name;
3393
3394 /* Get output port or VLAN. */
3395 if (cfg->output_port) {
3396 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3397 if (!s.out_bundle) {
3398 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3399 m->bridge->name, m->name);
3400 return false;
3401 }
3402 s.out_vlan = UINT16_MAX;
3403
3404 if (cfg->output_vlan) {
3405 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3406 "output vlan; ignoring output vlan",
3407 m->bridge->name, m->name);
3408 }
3409 } else if (cfg->output_vlan) {
3410 /* The database should prevent invalid VLAN values. */
3411 s.out_bundle = NULL;
3412 s.out_vlan = *cfg->output_vlan;
3413 } else {
3414 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3415 m->bridge->name, m->name);
3416 return false;
3417 }
3418
3419 /* Get port selection. */
3420 if (cfg->select_all) {
3421 size_t n_ports = hmap_count(&m->bridge->ports);
3422 void **ports = xmalloc(n_ports * sizeof *ports);
3423 struct port *port;
3424 size_t i;
3425
3426 i = 0;
3427 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3428 ports[i++] = port;
3429 }
3430
3431 s.srcs = ports;
3432 s.n_srcs = n_ports;
3433
3434 s.dsts = ports;
3435 s.n_dsts = n_ports;
3436 } else {
3437 /* Get ports, dropping ports that don't exist.
3438 * The IDL ensures that there are no duplicates. */
3439 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3440 &s.srcs, &s.n_srcs);
3441 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3442 &s.dsts, &s.n_dsts);
3443 }
3444
3445 /* Get VLAN selection. */
3446 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3447
3448 /* Configure. */
3449 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3450
3451 /* Clean up. */
3452 if (s.srcs != s.dsts) {
3453 free(s.dsts);
3454 }
3455 free(s.srcs);
3456 free(s.src_vlans);
3457
3458 return true;
3459 }
3460 \f
3461 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3462 *
3463 * This is deprecated. It is only for compatibility with broken device drivers
3464 * in old versions of Linux that do not properly support VLANs when VLAN
3465 * devices are not used. When broken device drivers are no longer in
3466 * widespread use, we will delete these interfaces. */
3467
3468 static void **blocks;
3469 static size_t n_blocks, allocated_blocks;
3470
3471 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3472 * VLAN splinters are reconfigured. */
3473 static void
3474 register_block(void *block)
3475 {
3476 if (n_blocks >= allocated_blocks) {
3477 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3478 }
3479 blocks[n_blocks++] = block;
3480 }
3481
3482 /* Frees all of the blocks registered with register_block(). */
3483 static void
3484 free_registered_blocks(void)
3485 {
3486 size_t i;
3487
3488 for (i = 0; i < n_blocks; i++) {
3489 free(blocks[i]);
3490 }
3491 n_blocks = 0;
3492 }
3493
3494 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3495 * otherwise. */
3496 static bool
3497 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3498 {
3499 const char *value;
3500
3501 value = ovsrec_interface_get_other_config_value(iface_cfg,
3502 "enable-vlan-splinters",
3503 "");
3504 return !strcmp(value, "true");
3505 }
3506
3507 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3508 * splinters.
3509 *
3510 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3511 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3512 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3513 * with free().
3514 *
3515 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3516 * use, returns NULL.
3517 *
3518 * Updates 'vlan_splinters_enabled_anywhere'. */
3519 static unsigned long int *
3520 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3521 {
3522 unsigned long int *splinter_vlans;
3523 struct sset splinter_ifaces;
3524 const char *real_dev_name;
3525 struct shash *real_devs;
3526 struct shash_node *node;
3527 struct bridge *br;
3528 size_t i;
3529
3530 /* Free space allocated for synthesized ports and interfaces, since we're
3531 * in the process of reconstructing all of them. */
3532 free_registered_blocks();
3533
3534 splinter_vlans = bitmap_allocate(4096);
3535 sset_init(&splinter_ifaces);
3536 vlan_splinters_enabled_anywhere = false;
3537 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3538 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3539 size_t j;
3540
3541 for (j = 0; j < br_cfg->n_ports; j++) {
3542 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3543 int k;
3544
3545 for (k = 0; k < port_cfg->n_interfaces; k++) {
3546 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3547
3548 if (vlan_splinters_is_enabled(iface_cfg)) {
3549 vlan_splinters_enabled_anywhere = true;
3550 sset_add(&splinter_ifaces, iface_cfg->name);
3551 vlan_bitmap_from_array__(port_cfg->trunks,
3552 port_cfg->n_trunks,
3553 splinter_vlans);
3554 }
3555 }
3556
3557 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3558 bitmap_set1(splinter_vlans, *port_cfg->tag);
3559 }
3560 }
3561 }
3562
3563 if (!vlan_splinters_enabled_anywhere) {
3564 free(splinter_vlans);
3565 sset_destroy(&splinter_ifaces);
3566 return NULL;
3567 }
3568
3569 HMAP_FOR_EACH (br, node, &all_bridges) {
3570 if (br->ofproto) {
3571 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3572 }
3573 }
3574
3575 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3576 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3577 * device to be created for it. */
3578 bitmap_set0(splinter_vlans, 0);
3579 bitmap_set0(splinter_vlans, 4095);
3580
3581 /* Delete all VLAN devices that we don't need. */
3582 vlandev_refresh();
3583 real_devs = vlandev_get_real_devs();
3584 SHASH_FOR_EACH (node, real_devs) {
3585 const struct vlan_real_dev *real_dev = node->data;
3586 const struct vlan_dev *vlan_dev;
3587 bool real_dev_has_splinters;
3588
3589 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3590 real_dev->name);
3591 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3592 if (!real_dev_has_splinters
3593 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3594 struct netdev *netdev;
3595
3596 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3597 if (!netdev_get_in4(netdev, NULL, NULL) ||
3598 !netdev_get_in6(netdev, NULL)) {
3599 vlandev_del(vlan_dev->name);
3600 } else {
3601 /* It has an IP address configured, so we don't own
3602 * it. Don't delete it. */
3603 }
3604 netdev_close(netdev);
3605 }
3606 }
3607
3608 }
3609 }
3610
3611 /* Add all VLAN devices that we need. */
3612 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3613 int vid;
3614
3615 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3616 if (!vlandev_get_name(real_dev_name, vid)) {
3617 vlandev_add(real_dev_name, vid);
3618 }
3619 }
3620 }
3621
3622 vlandev_refresh();
3623
3624 sset_destroy(&splinter_ifaces);
3625
3626 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3627 free(splinter_vlans);
3628 return NULL;
3629 }
3630 return splinter_vlans;
3631 }
3632
3633 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3634 * ofproto. */
3635 static void
3636 configure_splinter_port(struct port *port)
3637 {
3638 struct ofproto *ofproto = port->bridge->ofproto;
3639 uint16_t realdev_ofp_port;
3640 const char *realdev_name;
3641 struct iface *vlandev, *realdev;
3642
3643 ofproto_bundle_unregister(port->bridge->ofproto, port);
3644
3645 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3646 port_elem);
3647
3648 realdev_name = ovsrec_port_get_other_config_value(port->cfg,
3649 "realdev", NULL);
3650 realdev = iface_lookup(port->bridge, realdev_name);
3651 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3652
3653 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3654 *port->cfg->tag);
3655 }
3656
3657 static struct ovsrec_port *
3658 synthesize_splinter_port(const char *real_dev_name,
3659 const char *vlan_dev_name, int vid)
3660 {
3661 struct ovsrec_interface *iface;
3662 struct ovsrec_port *port;
3663
3664 iface = xzalloc(sizeof *iface);
3665 iface->name = xstrdup(vlan_dev_name);
3666 iface->type = "system";
3667
3668 port = xzalloc(sizeof *port);
3669 port->interfaces = xmemdup(&iface, sizeof iface);
3670 port->n_interfaces = 1;
3671 port->name = xstrdup(vlan_dev_name);
3672 port->vlan_mode = "splinter";
3673 port->tag = xmalloc(sizeof *port->tag);
3674 *port->tag = vid;
3675 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3676 port->key_other_config[0] = "realdev";
3677 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3678 port->value_other_config[0] = xstrdup(real_dev_name);
3679 port->n_other_config = 1;
3680
3681 register_block(iface);
3682 register_block(iface->name);
3683 register_block(port);
3684 register_block(port->interfaces);
3685 register_block(port->name);
3686 register_block(port->tag);
3687 register_block(port->key_other_config);
3688 register_block(port->value_other_config);
3689 register_block(port->value_other_config[0]);
3690
3691 return port;
3692 }
3693
3694 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3695 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3696 * 1-bit in the 'splinter_vlans' bitmap. */
3697 static void
3698 add_vlan_splinter_ports(struct bridge *br,
3699 const unsigned long int *splinter_vlans,
3700 struct shash *ports)
3701 {
3702 size_t i;
3703
3704 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3705 * we're modifying 'ports'. */
3706 for (i = 0; i < br->cfg->n_ports; i++) {
3707 const char *name = br->cfg->ports[i]->name;
3708 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3709 size_t j;
3710
3711 for (j = 0; j < port_cfg->n_interfaces; j++) {
3712 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3713
3714 if (vlan_splinters_is_enabled(iface_cfg)) {
3715 const char *real_dev_name;
3716 uint16_t vid;
3717
3718 real_dev_name = iface_cfg->name;
3719 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3720 const char *vlan_dev_name;
3721
3722 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3723 if (vlan_dev_name
3724 && !shash_find(ports, vlan_dev_name)) {
3725 shash_add(ports, vlan_dev_name,
3726 synthesize_splinter_port(
3727 real_dev_name, vlan_dev_name, vid));
3728 }
3729 }
3730 }
3731 }
3732 }
3733 }
3734
3735 static void
3736 mirror_refresh_stats(struct mirror *m)
3737 {
3738 struct ofproto *ofproto = m->bridge->ofproto;
3739 uint64_t tx_packets, tx_bytes;
3740 char *keys[2];
3741 int64_t values[2];
3742 size_t stat_cnt = 0;
3743
3744 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3745 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3746 return;
3747 }
3748
3749 if (tx_packets != UINT64_MAX) {
3750 keys[stat_cnt] = "tx_packets";
3751 values[stat_cnt] = tx_packets;
3752 stat_cnt++;
3753 }
3754 if (tx_bytes != UINT64_MAX) {
3755 keys[stat_cnt] = "tx_bytes";
3756 values[stat_cnt] = tx_bytes;
3757 stat_cnt++;
3758 }
3759
3760 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
3761 }