]> git.proxmox.com Git - ovs.git/blob - vswitchd/bridge.c
bridge: Fix reversed string parsing in bridge_configure_flow_miss_model().
[ovs.git] / vswitchd / bridge.c
1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
2 *
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 *
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
14 */
15
16 #include <config.h>
17 #include "bridge.h"
18 #include <errno.h>
19 #include <inttypes.h>
20 #include <stdlib.h>
21 #include "async-append.h"
22 #include "bfd.h"
23 #include "bitmap.h"
24 #include "cfm.h"
25 #include "connectivity.h"
26 #include "coverage.h"
27 #include "daemon.h"
28 #include "dirs.h"
29 #include "dynamic-string.h"
30 #include "hash.h"
31 #include "hmap.h"
32 #include "hmapx.h"
33 #include "jsonrpc.h"
34 #include "lacp.h"
35 #include "list.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
38 #include "netdev.h"
39 #include "ofp-print.h"
40 #include "ofp-util.h"
41 #include "ofpbuf.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
45 #include "seq.h"
46 #include "sha1.h"
47 #include "shash.h"
48 #include "smap.h"
49 #include "socket-util.h"
50 #include "stream.h"
51 #include "stream-ssl.h"
52 #include "sset.h"
53 #include "system-stats.h"
54 #include "timeval.h"
55 #include "util.h"
56 #include "unixctl.h"
57 #include "vlandev.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
60 #include "vlog.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
63
64 VLOG_DEFINE_THIS_MODULE(bridge);
65
66 COVERAGE_DEFINE(bridge_reconfigure);
67
68 struct iface {
69 /* These members are always valid.
70 *
71 * They are immutable: they never change between iface_create() and
72 * iface_destroy(). */
73 struct list port_elem; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
76 struct port *port; /* Containing port. */
77 char *name; /* Host network device name. */
78 struct netdev *netdev; /* Network device. */
79 ofp_port_t ofp_port; /* OpenFlow port number. */
80
81 /* These members are valid only within bridge_reconfigure(). */
82 const char *type; /* Usually same as cfg->type. */
83 const struct ovsrec_interface *cfg;
84 };
85
86 struct mirror {
87 struct uuid uuid; /* UUID of this "mirror" record in database. */
88 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
89 struct bridge *bridge;
90 char *name;
91 const struct ovsrec_mirror *cfg;
92 };
93
94 struct port {
95 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
96 struct bridge *bridge;
97 char *name;
98
99 const struct ovsrec_port *cfg;
100
101 /* An ordinary bridge port has 1 interface.
102 * A bridge port for bonding has at least 2 interfaces. */
103 struct list ifaces; /* List of "struct iface"s. */
104 };
105
106 struct bridge {
107 struct hmap_node node; /* In 'all_bridges'. */
108 char *name; /* User-specified arbitrary name. */
109 char *type; /* Datapath type. */
110 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
111 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
112 const struct ovsrec_bridge *cfg;
113
114 /* OpenFlow switch processing. */
115 struct ofproto *ofproto; /* OpenFlow switch. */
116
117 /* Bridge ports. */
118 struct hmap ports; /* "struct port"s indexed by name. */
119 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
120 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
121
122 /* Port mirroring. */
123 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
124
125 /* Used during reconfiguration. */
126 struct shash wanted_ports;
127
128 /* Synthetic local port if necessary. */
129 struct ovsrec_port synth_local_port;
130 struct ovsrec_interface synth_local_iface;
131 struct ovsrec_interface *synth_local_ifacep;
132 };
133
134 /* All bridges, indexed by name. */
135 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
136
137 /* OVSDB IDL used to obtain configuration. */
138 static struct ovsdb_idl *idl;
139
140 /* We want to complete daemonization, fully detaching from our parent process,
141 * only after we have completed our initial configuration, committed our state
142 * to the database, and received confirmation back from the database server
143 * that it applied the commit. This allows our parent process to know that,
144 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
145 * to have already been filled in. (It is only "very likely" rather than
146 * certain because there is always a slim possibility that the transaction will
147 * fail or that some other client has added new bridges, ports, etc. while
148 * ovs-vswitchd was configuring using an old configuration.)
149 *
150 * We only need to do this once for our initial configuration at startup, so
151 * 'initial_config_done' tracks whether we've already done it. While we are
152 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
153 * itself and is otherwise NULL. */
154 static bool initial_config_done;
155 static struct ovsdb_idl_txn *daemonize_txn;
156
157 /* Most recently processed IDL sequence number. */
158 static unsigned int idl_seqno;
159
160 /* Track changes to port connectivity. */
161 static uint64_t connectivity_seqno = LLONG_MIN;
162
163 /* Each time this timer expires, the bridge fetches interface and mirror
164 * statistics and pushes them into the database. */
165 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
166 static long long int iface_stats_timer = LLONG_MIN;
167
168 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
169 * expensive. This can cause bridge_reconfigure() to take a long time during
170 * which no other work can be done. To deal with this problem, we limit port
171 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
172 * call to bridge_reconfigure(). If there is more work to do after the limit
173 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
174 * This allows the rest of the code to catch up on important things like
175 * forwarding packets. */
176 #define OFP_PORT_ACTION_WINDOW 10
177
178 static void add_del_bridges(const struct ovsrec_open_vswitch *);
179 static void bridge_run__(void);
180 static void bridge_create(const struct ovsrec_bridge *);
181 static void bridge_destroy(struct bridge *);
182 static struct bridge *bridge_lookup(const char *name);
183 static unixctl_cb_func bridge_unixctl_dump_flows;
184 static unixctl_cb_func bridge_unixctl_reconnect;
185 static size_t bridge_get_controllers(const struct bridge *br,
186 struct ovsrec_controller ***controllersp);
187 static void bridge_collect_wanted_ports(struct bridge *,
188 const unsigned long *splinter_vlans,
189 struct shash *wanted_ports);
190 static void bridge_delete_ofprotos(void);
191 static void bridge_delete_or_reconfigure_ports(struct bridge *);
192 static void bridge_del_ports(struct bridge *,
193 const struct shash *wanted_ports);
194 static void bridge_add_ports(struct bridge *,
195 const struct shash *wanted_ports);
196
197 static void bridge_configure_flow_miss_model(const char *opt);
198 static void bridge_configure_datapath_id(struct bridge *);
199 static void bridge_configure_netflow(struct bridge *);
200 static void bridge_configure_forward_bpdu(struct bridge *);
201 static void bridge_configure_mac_table(struct bridge *);
202 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
203 static void bridge_configure_ipfix(struct bridge *);
204 static void bridge_configure_stp(struct bridge *);
205 static void bridge_configure_tables(struct bridge *);
206 static void bridge_configure_dp_desc(struct bridge *);
207 static void bridge_configure_remotes(struct bridge *,
208 const struct sockaddr_in *managers,
209 size_t n_managers);
210 static void bridge_pick_local_hw_addr(struct bridge *,
211 uint8_t ea[ETH_ADDR_LEN],
212 struct iface **hw_addr_iface);
213 static uint64_t bridge_pick_datapath_id(struct bridge *,
214 const uint8_t bridge_ea[ETH_ADDR_LEN],
215 struct iface *hw_addr_iface);
216 static uint64_t dpid_from_hash(const void *, size_t nbytes);
217 static bool bridge_has_bond_fake_iface(const struct bridge *,
218 const char *name);
219 static bool port_is_bond_fake_iface(const struct port *);
220
221 static unixctl_cb_func qos_unixctl_show;
222
223 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
224 static void port_del_ifaces(struct port *);
225 static void port_destroy(struct port *);
226 static struct port *port_lookup(const struct bridge *, const char *name);
227 static void port_configure(struct port *);
228 static struct lacp_settings *port_configure_lacp(struct port *,
229 struct lacp_settings *);
230 static void port_configure_bond(struct port *, struct bond_settings *);
231 static bool port_is_synthetic(const struct port *);
232
233 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
234 static void run_system_stats(void);
235
236 static void bridge_configure_mirrors(struct bridge *);
237 static struct mirror *mirror_create(struct bridge *,
238 const struct ovsrec_mirror *);
239 static void mirror_destroy(struct mirror *);
240 static bool mirror_configure(struct mirror *);
241 static void mirror_refresh_stats(struct mirror *);
242
243 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
244 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
245 const struct ovsrec_port *);
246 static bool iface_is_internal(const struct ovsrec_interface *iface,
247 const struct ovsrec_bridge *br);
248 static const char *iface_get_type(const struct ovsrec_interface *,
249 const struct ovsrec_bridge *);
250 static void iface_destroy(struct iface *);
251 static struct iface *iface_lookup(const struct bridge *, const char *name);
252 static struct iface *iface_find(const char *name);
253 static struct iface *iface_from_ofp_port(const struct bridge *,
254 ofp_port_t ofp_port);
255 static void iface_set_mac(struct iface *);
256 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
257 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
258 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
259 static void iface_configure_cfm(struct iface *);
260 static void iface_refresh_cfm_stats(struct iface *);
261 static void iface_refresh_stats(struct iface *);
262 static void iface_refresh_status(struct iface *);
263 static bool iface_is_synthetic(const struct iface *);
264 static ofp_port_t iface_get_requested_ofp_port(
265 const struct ovsrec_interface *);
266 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
267
268 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
269 *
270 * This is deprecated. It is only for compatibility with broken device drivers
271 * in old versions of Linux that do not properly support VLANs when VLAN
272 * devices are not used. When broken device drivers are no longer in
273 * widespread use, we will delete these interfaces. */
274
275 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
276 static bool vlan_splinters_enabled_anywhere;
277
278 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
279 static unsigned long int *collect_splinter_vlans(
280 const struct ovsrec_open_vswitch *);
281 static void configure_splinter_port(struct port *);
282 static void add_vlan_splinter_ports(struct bridge *,
283 const unsigned long int *splinter_vlans,
284 struct shash *ports);
285
286 static void
287 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
288 {
289 struct shash iface_hints;
290 static bool initialized = false;
291 int i;
292
293 if (initialized) {
294 return;
295 }
296
297 shash_init(&iface_hints);
298
299 if (cfg) {
300 for (i = 0; i < cfg->n_bridges; i++) {
301 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
302 int j;
303
304 for (j = 0; j < br_cfg->n_ports; j++) {
305 struct ovsrec_port *port_cfg = br_cfg->ports[j];
306 int k;
307
308 for (k = 0; k < port_cfg->n_interfaces; k++) {
309 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
310 struct iface_hint *iface_hint;
311
312 iface_hint = xmalloc(sizeof *iface_hint);
313 iface_hint->br_name = br_cfg->name;
314 iface_hint->br_type = br_cfg->datapath_type;
315 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
316
317 shash_add(&iface_hints, if_cfg->name, iface_hint);
318 }
319 }
320 }
321 }
322
323 ofproto_init(&iface_hints);
324
325 shash_destroy_free_data(&iface_hints);
326 initialized = true;
327 }
328 \f
329 /* Public functions. */
330
331 /* Initializes the bridge module, configuring it to obtain its configuration
332 * from an OVSDB server accessed over 'remote', which should be a string in a
333 * form acceptable to ovsdb_idl_create(). */
334 void
335 bridge_init(const char *remote)
336 {
337 /* Create connection to database. */
338 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
339 idl_seqno = ovsdb_idl_get_seqno(idl);
340 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
341 ovsdb_idl_verify_write_only(idl);
342
343 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
344 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
345 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
350
351 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
352 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
353 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
354
355 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
356 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
357 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
358 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
359
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
379 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
380
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
383 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
384 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
385
386 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
387
388 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
389
390 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
391 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
392
393 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
397
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
403
404 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
405
406 /* Register unixctl commands. */
407 unixctl_command_register("qos/show", "interface", 1, 1,
408 qos_unixctl_show, NULL);
409 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
410 bridge_unixctl_dump_flows, NULL);
411 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
412 bridge_unixctl_reconnect, NULL);
413 lacp_init();
414 bond_init();
415 cfm_init();
416 stp_init();
417 }
418
419 void
420 bridge_exit(void)
421 {
422 struct bridge *br, *next_br;
423
424 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
425 bridge_destroy(br);
426 }
427 ovsdb_idl_destroy(idl);
428 }
429
430 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
431 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
432 * responsible for freeing '*managersp' (with free()).
433 *
434 * You may be asking yourself "why does ovs-vswitchd care?", because
435 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
436 * should not be and in fact is not directly involved in that. But
437 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
438 * it has to tell in-band control where the managers are to enable that.
439 * (Thus, only managers connected in-band are collected.)
440 */
441 static void
442 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
443 struct sockaddr_in **managersp, size_t *n_managersp)
444 {
445 struct sockaddr_in *managers = NULL;
446 size_t n_managers = 0;
447 struct sset targets;
448 size_t i;
449
450 /* Collect all of the potential targets from the "targets" columns of the
451 * rows pointed to by "manager_options", excluding any that are
452 * out-of-band. */
453 sset_init(&targets);
454 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
455 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
456
457 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
458 sset_find_and_delete(&targets, m->target);
459 } else {
460 sset_add(&targets, m->target);
461 }
462 }
463
464 /* Now extract the targets' IP addresses. */
465 if (!sset_is_empty(&targets)) {
466 const char *target;
467
468 managers = xmalloc(sset_count(&targets) * sizeof *managers);
469 SSET_FOR_EACH (target, &targets) {
470 struct sockaddr_in *sin = &managers[n_managers];
471
472 if (stream_parse_target_with_default_port(target,
473 OVSDB_OLD_PORT,
474 sin)) {
475 n_managers++;
476 }
477 }
478 }
479 sset_destroy(&targets);
480
481 *managersp = managers;
482 *n_managersp = n_managers;
483 }
484
485 static void
486 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
487 {
488 unsigned long int *splinter_vlans;
489 struct sockaddr_in *managers;
490 struct bridge *br, *next;
491 int sflow_bridge_number;
492 size_t n_managers;
493
494 COVERAGE_INC(bridge_reconfigure);
495
496 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
497 OFPROTO_FLOW_LIMIT_DEFAULT));
498
499 ofproto_set_threads(
500 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
501 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
502
503 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
504 "force-miss-model"));
505
506 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
507 * to 'ovs_cfg', with only very minimal configuration otherwise.
508 *
509 * This is mostly an update to bridge data structures. Nothing is pushed
510 * down to ofproto or lower layers. */
511 add_del_bridges(ovs_cfg);
512 splinter_vlans = collect_splinter_vlans(ovs_cfg);
513 HMAP_FOR_EACH (br, node, &all_bridges) {
514 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
515 bridge_del_ports(br, &br->wanted_ports);
516 }
517 free(splinter_vlans);
518
519 /* Start pushing configuration changes down to the ofproto layer:
520 *
521 * - Delete ofprotos that are no longer configured.
522 *
523 * - Delete ports that are no longer configured.
524 *
525 * - Reconfigure existing ports to their desired configurations, or
526 * delete them if not possible.
527 *
528 * We have to do all the deletions before we can do any additions, because
529 * the ports to be added might require resources that will be freed up by
530 * deletions (they might especially overlap in name). */
531 bridge_delete_ofprotos();
532 HMAP_FOR_EACH (br, node, &all_bridges) {
533 if (br->ofproto) {
534 bridge_delete_or_reconfigure_ports(br);
535 }
536 }
537
538 /* Finish pushing configuration changes to the ofproto layer:
539 *
540 * - Create ofprotos that are missing.
541 *
542 * - Add ports that are missing. */
543 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
544 if (!br->ofproto) {
545 int error;
546
547 error = ofproto_create(br->name, br->type, &br->ofproto);
548 if (error) {
549 VLOG_ERR("failed to create bridge %s: %s", br->name,
550 ovs_strerror(error));
551 shash_destroy(&br->wanted_ports);
552 bridge_destroy(br);
553 }
554 }
555 }
556 HMAP_FOR_EACH (br, node, &all_bridges) {
557 bridge_add_ports(br, &br->wanted_ports);
558 shash_destroy(&br->wanted_ports);
559 }
560
561 reconfigure_system_stats(ovs_cfg);
562
563 /* Complete the configuration. */
564 sflow_bridge_number = 0;
565 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
566 HMAP_FOR_EACH (br, node, &all_bridges) {
567 struct port *port;
568
569 /* We need the datapath ID early to allow LACP ports to use it as the
570 * default system ID. */
571 bridge_configure_datapath_id(br);
572
573 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
574 struct iface *iface;
575
576 port_configure(port);
577
578 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
579 iface_configure_cfm(iface);
580 iface_configure_qos(iface, port->cfg->qos);
581 iface_set_mac(iface);
582 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
583 &iface->cfg->bfd);
584 }
585 }
586 bridge_configure_mirrors(br);
587 bridge_configure_forward_bpdu(br);
588 bridge_configure_mac_table(br);
589 bridge_configure_remotes(br, managers, n_managers);
590 bridge_configure_netflow(br);
591 bridge_configure_sflow(br, &sflow_bridge_number);
592 bridge_configure_ipfix(br);
593 bridge_configure_stp(br);
594 bridge_configure_tables(br);
595 bridge_configure_dp_desc(br);
596
597 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
598 /* XXX: Remove this warning message eventually. */
599 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
600 " moved to the Open_vSwitch table. Ignoring its"
601 " setting in the bridge table.");
602 }
603 }
604 free(managers);
605
606 /* The ofproto-dpif provider does some final reconfiguration in its
607 * ->type_run() function. We have to call it before notifying the database
608 * client that reconfiguration is complete, otherwise there is a very
609 * narrow race window in which e.g. ofproto/trace will not recognize the
610 * new configuration (sometimes this causes unit test failures). */
611 bridge_run__();
612 }
613
614 /* Delete ofprotos which aren't configured or have the wrong type. Create
615 * ofprotos which don't exist but need to. */
616 static void
617 bridge_delete_ofprotos(void)
618 {
619 struct bridge *br;
620 struct sset names;
621 struct sset types;
622 const char *type;
623
624 /* Delete ofprotos with no bridge or with the wrong type. */
625 sset_init(&names);
626 sset_init(&types);
627 ofproto_enumerate_types(&types);
628 SSET_FOR_EACH (type, &types) {
629 const char *name;
630
631 ofproto_enumerate_names(type, &names);
632 SSET_FOR_EACH (name, &names) {
633 br = bridge_lookup(name);
634 if (!br || strcmp(type, br->type)) {
635 ofproto_delete(name, type);
636 }
637 }
638 }
639 sset_destroy(&names);
640 sset_destroy(&types);
641 }
642
643 static ofp_port_t *
644 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
645 {
646 if (*n >= *allocated) {
647 ports = x2nrealloc(ports, allocated, sizeof *ports);
648 }
649 ports[(*n)++] = port;
650 return ports;
651 }
652
653 static void
654 bridge_delete_or_reconfigure_ports(struct bridge *br)
655 {
656 struct ofproto_port ofproto_port;
657 struct ofproto_port_dump dump;
658
659 /* List of "ofp_port"s to delete. We make a list instead of deleting them
660 * right away because ofproto implementations aren't necessarily able to
661 * iterate through a changing list of ports in an entirely robust way. */
662 ofp_port_t *del;
663 size_t n, allocated;
664 size_t i;
665
666 del = NULL;
667 n = allocated = 0;
668
669 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
670 ofp_port_t requested_ofp_port;
671 struct iface *iface;
672
673 iface = iface_lookup(br, ofproto_port.name);
674 if (!iface) {
675 /* No such iface is configured, so we should delete this
676 * ofproto_port.
677 *
678 * As a corner case exception, keep the port if it's a bond fake
679 * interface. */
680 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
681 && !strcmp(ofproto_port.type, "internal")) {
682 continue;
683 }
684 goto delete;
685 }
686
687 if (strcmp(ofproto_port.type, iface->type)
688 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
689 /* The interface is the wrong type or can't be configured.
690 * Delete it. */
691 goto delete;
692 }
693
694 /* If the requested OpenFlow port for 'iface' changed, and it's not
695 * already the correct port, then we might want to temporarily delete
696 * this interface, so we can add it back again with the new OpenFlow
697 * port number. */
698 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
699 if (iface->ofp_port != OFPP_LOCAL &&
700 requested_ofp_port != OFPP_NONE &&
701 requested_ofp_port != iface->ofp_port) {
702 ofp_port_t victim_request;
703 struct iface *victim;
704
705 /* Check for an existing OpenFlow port currently occupying
706 * 'iface''s requested port number. If there isn't one, then
707 * delete this port. Otherwise we need to consider further. */
708 victim = iface_from_ofp_port(br, requested_ofp_port);
709 if (!victim) {
710 goto delete;
711 }
712
713 /* 'victim' is a port currently using 'iface''s requested port
714 * number. Unless 'victim' specifically requested that port
715 * number, too, then we can delete both 'iface' and 'victim'
716 * temporarily. (We'll add both of them back again later with new
717 * OpenFlow port numbers.)
718 *
719 * If 'victim' did request port number 'requested_ofp_port', just
720 * like 'iface', then that's a configuration inconsistency that we
721 * can't resolve. We might as well let it keep its current port
722 * number. */
723 victim_request = iface_get_requested_ofp_port(victim->cfg);
724 if (victim_request != requested_ofp_port) {
725 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
726 iface_destroy(victim);
727 goto delete;
728 }
729 }
730
731 /* Keep it. */
732 continue;
733
734 delete:
735 iface_destroy(iface);
736 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
737 }
738
739 for (i = 0; i < n; i++) {
740 ofproto_port_del(br->ofproto, del[i]);
741 }
742 free(del);
743 }
744
745 static void
746 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
747 bool with_requested_port)
748 {
749 struct shash_node *port_node;
750
751 SHASH_FOR_EACH (port_node, wanted_ports) {
752 const struct ovsrec_port *port_cfg = port_node->data;
753 size_t i;
754
755 for (i = 0; i < port_cfg->n_interfaces; i++) {
756 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
757 ofp_port_t requested_ofp_port;
758
759 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
760 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
761 struct iface *iface = iface_lookup(br, iface_cfg->name);
762
763 if (!iface) {
764 iface_create(br, iface_cfg, port_cfg);
765 }
766 }
767 }
768 }
769 }
770
771 static void
772 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
773 {
774 /* First add interfaces that request a particular port number. */
775 bridge_add_ports__(br, wanted_ports, true);
776
777 /* Then add interfaces that want automatic port number assignment.
778 * We add these afterward to avoid accidentally taking a specifically
779 * requested port number. */
780 bridge_add_ports__(br, wanted_ports, false);
781 }
782
783 static void
784 port_configure(struct port *port)
785 {
786 const struct ovsrec_port *cfg = port->cfg;
787 struct bond_settings bond_settings;
788 struct lacp_settings lacp_settings;
789 struct ofproto_bundle_settings s;
790 struct iface *iface;
791
792 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
793 configure_splinter_port(port);
794 return;
795 }
796
797 /* Get name. */
798 s.name = port->name;
799
800 /* Get slaves. */
801 s.n_slaves = 0;
802 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
803 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
804 s.slaves[s.n_slaves++] = iface->ofp_port;
805 }
806
807 /* Get VLAN tag. */
808 s.vlan = -1;
809 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
810 s.vlan = *cfg->tag;
811 }
812
813 /* Get VLAN trunks. */
814 s.trunks = NULL;
815 if (cfg->n_trunks) {
816 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
817 }
818
819 /* Get VLAN mode. */
820 if (cfg->vlan_mode) {
821 if (!strcmp(cfg->vlan_mode, "access")) {
822 s.vlan_mode = PORT_VLAN_ACCESS;
823 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
824 s.vlan_mode = PORT_VLAN_TRUNK;
825 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
826 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
827 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
828 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
829 } else {
830 /* This "can't happen" because ovsdb-server should prevent it. */
831 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
832 s.vlan_mode = PORT_VLAN_TRUNK;
833 }
834 } else {
835 if (s.vlan >= 0) {
836 s.vlan_mode = PORT_VLAN_ACCESS;
837 if (cfg->n_trunks) {
838 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
839 port->name);
840 }
841 } else {
842 s.vlan_mode = PORT_VLAN_TRUNK;
843 }
844 }
845 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
846 false);
847
848 /* Get LACP settings. */
849 s.lacp = port_configure_lacp(port, &lacp_settings);
850 if (s.lacp) {
851 size_t i = 0;
852
853 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
854 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
855 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
856 }
857 } else {
858 s.lacp_slaves = NULL;
859 }
860
861 /* Get bond settings. */
862 if (s.n_slaves > 1) {
863 s.bond = &bond_settings;
864 port_configure_bond(port, &bond_settings);
865 } else {
866 s.bond = NULL;
867 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
868 netdev_set_miimon_interval(iface->netdev, 0);
869 }
870 }
871
872 /* Register. */
873 ofproto_bundle_register(port->bridge->ofproto, port, &s);
874
875 /* Clean up. */
876 free(s.slaves);
877 free(s.trunks);
878 free(s.lacp_slaves);
879 }
880
881 static void
882 bridge_configure_flow_miss_model(const char *opt)
883 {
884 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
885
886 if (opt) {
887 if (!strcmp(opt, "with-facets")) {
888 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
889 } else if (!strcmp(opt, "without-facets")) {
890 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
891 }
892 }
893
894 ofproto_set_flow_miss_model(model);
895 }
896
897 /* Pick local port hardware address and datapath ID for 'br'. */
898 static void
899 bridge_configure_datapath_id(struct bridge *br)
900 {
901 uint8_t ea[ETH_ADDR_LEN];
902 uint64_t dpid;
903 struct iface *local_iface;
904 struct iface *hw_addr_iface;
905 char *dpid_string;
906
907 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
908 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
909 if (local_iface) {
910 int error = netdev_set_etheraddr(local_iface->netdev, ea);
911 if (error) {
912 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
913 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
914 "Ethernet address: %s",
915 br->name, ovs_strerror(error));
916 }
917 }
918 memcpy(br->ea, ea, ETH_ADDR_LEN);
919
920 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
921 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
922 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
923 ofproto_set_datapath_id(br->ofproto, dpid);
924 }
925
926 dpid_string = xasprintf("%016"PRIx64, dpid);
927 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
928 free(dpid_string);
929 }
930
931 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
932 * 'br'. */
933 static uint32_t
934 bridge_get_allowed_versions(struct bridge *br)
935 {
936 if (!br->cfg->n_protocols)
937 return 0;
938
939 return ofputil_versions_from_strings(br->cfg->protocols,
940 br->cfg->n_protocols);
941 }
942
943 /* Set NetFlow configuration on 'br'. */
944 static void
945 bridge_configure_netflow(struct bridge *br)
946 {
947 struct ovsrec_netflow *cfg = br->cfg->netflow;
948 struct netflow_options opts;
949
950 if (!cfg) {
951 ofproto_set_netflow(br->ofproto, NULL);
952 return;
953 }
954
955 memset(&opts, 0, sizeof opts);
956
957 /* Get default NetFlow configuration from datapath.
958 * Apply overrides from 'cfg'. */
959 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
960 if (cfg->engine_type) {
961 opts.engine_type = *cfg->engine_type;
962 }
963 if (cfg->engine_id) {
964 opts.engine_id = *cfg->engine_id;
965 }
966
967 /* Configure active timeout interval. */
968 opts.active_timeout = cfg->active_timeout;
969 if (!opts.active_timeout) {
970 opts.active_timeout = -1;
971 } else if (opts.active_timeout < 0) {
972 VLOG_WARN("bridge %s: active timeout interval set to negative "
973 "value, using default instead (%d seconds)", br->name,
974 NF_ACTIVE_TIMEOUT_DEFAULT);
975 opts.active_timeout = -1;
976 }
977
978 /* Add engine ID to interface number to disambiguate bridgs? */
979 opts.add_id_to_iface = cfg->add_id_to_interface;
980 if (opts.add_id_to_iface) {
981 if (opts.engine_id > 0x7f) {
982 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
983 "another vswitch, choose an engine id less than 128",
984 br->name);
985 }
986 if (hmap_count(&br->ports) > 508) {
987 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
988 "another port when more than 508 ports are used",
989 br->name);
990 }
991 }
992
993 /* Collectors. */
994 sset_init(&opts.collectors);
995 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
996
997 /* Configure. */
998 if (ofproto_set_netflow(br->ofproto, &opts)) {
999 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1000 }
1001 sset_destroy(&opts.collectors);
1002 }
1003
1004 /* Set sFlow configuration on 'br'. */
1005 static void
1006 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1007 {
1008 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1009 struct ovsrec_controller **controllers;
1010 struct ofproto_sflow_options oso;
1011 size_t n_controllers;
1012 size_t i;
1013
1014 if (!cfg) {
1015 ofproto_set_sflow(br->ofproto, NULL);
1016 return;
1017 }
1018
1019 memset(&oso, 0, sizeof oso);
1020
1021 sset_init(&oso.targets);
1022 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1023
1024 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1025 if (cfg->sampling) {
1026 oso.sampling_rate = *cfg->sampling;
1027 }
1028
1029 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1030 if (cfg->polling) {
1031 oso.polling_interval = *cfg->polling;
1032 }
1033
1034 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1035 if (cfg->header) {
1036 oso.header_len = *cfg->header;
1037 }
1038
1039 oso.sub_id = (*sflow_bridge_number)++;
1040 oso.agent_device = cfg->agent;
1041
1042 oso.control_ip = NULL;
1043 n_controllers = bridge_get_controllers(br, &controllers);
1044 for (i = 0; i < n_controllers; i++) {
1045 if (controllers[i]->local_ip) {
1046 oso.control_ip = controllers[i]->local_ip;
1047 break;
1048 }
1049 }
1050 ofproto_set_sflow(br->ofproto, &oso);
1051
1052 sset_destroy(&oso.targets);
1053 }
1054
1055 /* Returns whether a IPFIX row is valid. */
1056 static bool
1057 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1058 {
1059 return ipfix && ipfix->n_targets > 0;
1060 }
1061
1062 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1063 static bool
1064 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1065 const struct bridge *br)
1066 {
1067 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1068 }
1069
1070 /* Set IPFIX configuration on 'br'. */
1071 static void
1072 bridge_configure_ipfix(struct bridge *br)
1073 {
1074 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1075 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1076 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1077 struct ofproto_ipfix_bridge_exporter_options be_opts;
1078 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1079 size_t n_fe_opts = 0;
1080
1081 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1082 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1083 n_fe_opts++;
1084 }
1085 }
1086
1087 if (!valid_be_cfg && n_fe_opts == 0) {
1088 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1089 return;
1090 }
1091
1092 if (valid_be_cfg) {
1093 memset(&be_opts, 0, sizeof be_opts);
1094
1095 sset_init(&be_opts.targets);
1096 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1097
1098 if (be_cfg->sampling) {
1099 be_opts.sampling_rate = *be_cfg->sampling;
1100 } else {
1101 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1102 }
1103 if (be_cfg->obs_domain_id) {
1104 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1105 }
1106 if (be_cfg->obs_point_id) {
1107 be_opts.obs_point_id = *be_cfg->obs_point_id;
1108 }
1109 if (be_cfg->cache_active_timeout) {
1110 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1111 }
1112 if (be_cfg->cache_max_flows) {
1113 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1114 }
1115 }
1116
1117 if (n_fe_opts > 0) {
1118 struct ofproto_ipfix_flow_exporter_options *opts;
1119 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1120 opts = fe_opts;
1121 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1122 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1123 opts->collector_set_id = fe_cfg->id;
1124 sset_init(&opts->targets);
1125 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1126 fe_cfg->ipfix->n_targets);
1127 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1128 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1129 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1130 ? *fe_cfg->ipfix->cache_max_flows : 0;
1131 opts++;
1132 }
1133 }
1134 }
1135
1136 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1137 n_fe_opts);
1138
1139 if (valid_be_cfg) {
1140 sset_destroy(&be_opts.targets);
1141 }
1142
1143 if (n_fe_opts > 0) {
1144 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1145 size_t i;
1146 for (i = 0; i < n_fe_opts; i++) {
1147 sset_destroy(&opts->targets);
1148 opts++;
1149 }
1150 free(fe_opts);
1151 }
1152 }
1153
1154 static void
1155 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1156 struct ofproto_port_stp_settings *port_s,
1157 int *port_num_counter, unsigned long *port_num_bitmap)
1158 {
1159 const char *config_str;
1160 struct iface *iface;
1161
1162 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1163 port_s->enable = false;
1164 return;
1165 } else {
1166 port_s->enable = true;
1167 }
1168
1169 /* STP over bonds is not supported. */
1170 if (!list_is_singleton(&port->ifaces)) {
1171 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1172 port->name);
1173 port_s->enable = false;
1174 return;
1175 }
1176
1177 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1178
1179 /* Internal ports shouldn't participate in spanning tree, so
1180 * skip them. */
1181 if (!strcmp(iface->type, "internal")) {
1182 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1183 port_s->enable = false;
1184 return;
1185 }
1186
1187 /* STP on mirror output ports is not supported. */
1188 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1189 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1190 port_s->enable = false;
1191 return;
1192 }
1193
1194 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1195 if (config_str) {
1196 unsigned long int port_num = strtoul(config_str, NULL, 0);
1197 int port_idx = port_num - 1;
1198
1199 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1200 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1201 port_s->enable = false;
1202 return;
1203 }
1204
1205 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1206 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1207 port->name, port_num);
1208 port_s->enable = false;
1209 return;
1210 }
1211 bitmap_set1(port_num_bitmap, port_idx);
1212 port_s->port_num = port_idx;
1213 } else {
1214 if (*port_num_counter >= STP_MAX_PORTS) {
1215 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1216 port_s->enable = false;
1217 return;
1218 }
1219
1220 port_s->port_num = (*port_num_counter)++;
1221 }
1222
1223 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1224 if (config_str) {
1225 port_s->path_cost = strtoul(config_str, NULL, 10);
1226 } else {
1227 enum netdev_features current;
1228 unsigned int mbps;
1229
1230 netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1231 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1232 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1233 }
1234
1235 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1236 if (config_str) {
1237 port_s->priority = strtoul(config_str, NULL, 0);
1238 } else {
1239 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1240 }
1241 }
1242
1243 /* Set spanning tree configuration on 'br'. */
1244 static void
1245 bridge_configure_stp(struct bridge *br)
1246 {
1247 if (!br->cfg->stp_enable) {
1248 ofproto_set_stp(br->ofproto, NULL);
1249 } else {
1250 struct ofproto_stp_settings br_s;
1251 const char *config_str;
1252 struct port *port;
1253 int port_num_counter;
1254 unsigned long *port_num_bitmap;
1255
1256 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1257 if (config_str) {
1258 uint8_t ea[ETH_ADDR_LEN];
1259
1260 if (eth_addr_from_string(config_str, ea)) {
1261 br_s.system_id = eth_addr_to_uint64(ea);
1262 } else {
1263 br_s.system_id = eth_addr_to_uint64(br->ea);
1264 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1265 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1266 }
1267 } else {
1268 br_s.system_id = eth_addr_to_uint64(br->ea);
1269 }
1270
1271 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1272 if (config_str) {
1273 br_s.priority = strtoul(config_str, NULL, 0);
1274 } else {
1275 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1276 }
1277
1278 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1279 if (config_str) {
1280 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1281 } else {
1282 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1283 }
1284
1285 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1286 if (config_str) {
1287 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1288 } else {
1289 br_s.max_age = STP_DEFAULT_MAX_AGE;
1290 }
1291
1292 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1293 if (config_str) {
1294 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1295 } else {
1296 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1297 }
1298
1299 /* Configure STP on the bridge. */
1300 if (ofproto_set_stp(br->ofproto, &br_s)) {
1301 VLOG_ERR("bridge %s: could not enable STP", br->name);
1302 return;
1303 }
1304
1305 /* Users must either set the port number with the "stp-port-num"
1306 * configuration on all ports or none. If manual configuration
1307 * is not done, then we allocate them sequentially. */
1308 port_num_counter = 0;
1309 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1310 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1311 struct ofproto_port_stp_settings port_s;
1312 struct iface *iface;
1313
1314 port_configure_stp(br->ofproto, port, &port_s,
1315 &port_num_counter, port_num_bitmap);
1316
1317 /* As bonds are not supported, just apply configuration to
1318 * all interfaces. */
1319 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1320 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1321 &port_s)) {
1322 VLOG_ERR("port %s: could not enable STP", port->name);
1323 continue;
1324 }
1325 }
1326 }
1327
1328 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1329 && port_num_counter) {
1330 VLOG_ERR("bridge %s: must manually configure all STP port "
1331 "IDs or none, disabling", br->name);
1332 ofproto_set_stp(br->ofproto, NULL);
1333 }
1334 bitmap_free(port_num_bitmap);
1335 }
1336 }
1337
1338 static bool
1339 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1340 {
1341 const struct port *port = port_lookup(br, name);
1342 return port && port_is_bond_fake_iface(port);
1343 }
1344
1345 static bool
1346 port_is_bond_fake_iface(const struct port *port)
1347 {
1348 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1349 }
1350
1351 static void
1352 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1353 {
1354 struct bridge *br, *next;
1355 struct shash new_br;
1356 size_t i;
1357
1358 /* Collect new bridges' names and types. */
1359 shash_init(&new_br);
1360 for (i = 0; i < cfg->n_bridges; i++) {
1361 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1362 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1363
1364 if (strchr(br_cfg->name, '/')) {
1365 /* Prevent remote ovsdb-server users from accessing arbitrary
1366 * directories, e.g. consider a bridge named "../../../etc/". */
1367 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1368 br_cfg->name);
1369 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1370 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1371 }
1372 }
1373
1374 /* Get rid of deleted bridges or those whose types have changed.
1375 * Update 'cfg' of bridges that still exist. */
1376 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1377 br->cfg = shash_find_data(&new_br, br->name);
1378 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1379 br->cfg->datapath_type))) {
1380 bridge_destroy(br);
1381 }
1382 }
1383
1384 /* Add new bridges. */
1385 for (i = 0; i < cfg->n_bridges; i++) {
1386 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1387 struct bridge *br = bridge_lookup(br_cfg->name);
1388 if (!br) {
1389 bridge_create(br_cfg);
1390 }
1391 }
1392
1393 shash_destroy(&new_br);
1394 }
1395
1396 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1397 * Returns 0 if successful, otherwise a positive errno value. */
1398 static int
1399 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1400 struct netdev *netdev)
1401 {
1402 return netdev_set_config(netdev, &iface_cfg->options);
1403 }
1404
1405 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1406 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1407 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1408 * untouched.
1409 *
1410 * If successful, returns 0 and stores the network device in '*netdevp'. On
1411 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1412 static int
1413 iface_do_create(const struct bridge *br,
1414 const struct ovsrec_interface *iface_cfg,
1415 const struct ovsrec_port *port_cfg,
1416 ofp_port_t *ofp_portp, struct netdev **netdevp)
1417 {
1418 struct netdev *netdev = NULL;
1419 int error;
1420
1421 if (netdev_is_reserved_name(iface_cfg->name)) {
1422 VLOG_WARN("could not create interface %s, name is reserved",
1423 iface_cfg->name);
1424 error = EINVAL;
1425 goto error;
1426 }
1427
1428 error = netdev_open(iface_cfg->name,
1429 iface_get_type(iface_cfg, br->cfg), &netdev);
1430 if (error) {
1431 VLOG_WARN("could not open network device %s (%s)",
1432 iface_cfg->name, ovs_strerror(error));
1433 goto error;
1434 }
1435
1436 error = iface_set_netdev_config(iface_cfg, netdev);
1437 if (error) {
1438 goto error;
1439 }
1440
1441 *ofp_portp = iface_pick_ofport(iface_cfg);
1442 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1443 if (error) {
1444 goto error;
1445 }
1446
1447 VLOG_INFO("bridge %s: added interface %s on port %d",
1448 br->name, iface_cfg->name, *ofp_portp);
1449
1450 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1451 || iface_is_internal(iface_cfg, br->cfg)) {
1452 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1453 }
1454
1455 *netdevp = netdev;
1456 return 0;
1457
1458 error:
1459 *netdevp = NULL;
1460 netdev_close(netdev);
1461 return error;
1462 }
1463
1464 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1465 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1466 * automatically allocated for the iface. Takes ownership of and
1467 * deallocates 'if_cfg'.
1468 *
1469 * Return true if an iface is successfully created, false otherwise. */
1470 static bool
1471 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1472 const struct ovsrec_port *port_cfg)
1473 {
1474 struct netdev *netdev;
1475 struct iface *iface;
1476 ofp_port_t ofp_port;
1477 struct port *port;
1478 int error;
1479
1480 /* Do the bits that can fail up front. */
1481 ovs_assert(!iface_lookup(br, iface_cfg->name));
1482 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1483 if (error) {
1484 iface_set_ofport(iface_cfg, OFPP_NONE);
1485 iface_clear_db_record(iface_cfg);
1486 return false;
1487 }
1488
1489 /* Get or create the port structure. */
1490 port = port_lookup(br, port_cfg->name);
1491 if (!port) {
1492 port = port_create(br, port_cfg);
1493 }
1494
1495 /* Create the iface structure. */
1496 iface = xzalloc(sizeof *iface);
1497 list_push_back(&port->ifaces, &iface->port_elem);
1498 hmap_insert(&br->iface_by_name, &iface->name_node,
1499 hash_string(iface_cfg->name, 0));
1500 iface->port = port;
1501 iface->name = xstrdup(iface_cfg->name);
1502 iface->ofp_port = ofp_port;
1503 iface->netdev = netdev;
1504 iface->type = iface_get_type(iface_cfg, br->cfg);
1505 iface->cfg = iface_cfg;
1506 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1507 hash_ofp_port(ofp_port));
1508
1509 iface_set_ofport(iface->cfg, ofp_port);
1510
1511 /* Populate initial status in database. */
1512 iface_refresh_stats(iface);
1513 iface_refresh_status(iface);
1514
1515 /* Add bond fake iface if necessary. */
1516 if (port_is_bond_fake_iface(port)) {
1517 struct ofproto_port ofproto_port;
1518
1519 if (ofproto_port_query_by_name(br->ofproto, port->name,
1520 &ofproto_port)) {
1521 struct netdev *netdev;
1522 int error;
1523
1524 error = netdev_open(port->name, "internal", &netdev);
1525 if (!error) {
1526 ofp_port_t fake_ofp_port = OFPP_NONE;
1527 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1528 netdev_close(netdev);
1529 } else {
1530 VLOG_WARN("could not open network device %s (%s)",
1531 port->name, ovs_strerror(error));
1532 }
1533 } else {
1534 /* Already exists, nothing to do. */
1535 ofproto_port_destroy(&ofproto_port);
1536 }
1537 }
1538
1539 return true;
1540 }
1541
1542 /* Set forward BPDU option. */
1543 static void
1544 bridge_configure_forward_bpdu(struct bridge *br)
1545 {
1546 ofproto_set_forward_bpdu(br->ofproto,
1547 smap_get_bool(&br->cfg->other_config,
1548 "forward-bpdu",
1549 false));
1550 }
1551
1552 /* Set MAC learning table configuration for 'br'. */
1553 static void
1554 bridge_configure_mac_table(struct bridge *br)
1555 {
1556 const char *idle_time_str;
1557 int idle_time;
1558
1559 const char *mac_table_size_str;
1560 int mac_table_size;
1561
1562 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1563 idle_time = (idle_time_str && atoi(idle_time_str)
1564 ? atoi(idle_time_str)
1565 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1566
1567 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1568 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1569 ? atoi(mac_table_size_str)
1570 : MAC_DEFAULT_MAX);
1571
1572 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1573 }
1574
1575 static void
1576 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1577 struct iface **hw_addr_iface)
1578 {
1579 struct hmapx mirror_output_ports;
1580 const char *hwaddr;
1581 struct port *port;
1582 bool found_addr = false;
1583 int error;
1584 int i;
1585
1586 *hw_addr_iface = NULL;
1587
1588 /* Did the user request a particular MAC? */
1589 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1590 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1591 if (eth_addr_is_multicast(ea)) {
1592 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1593 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1594 } else if (eth_addr_is_zero(ea)) {
1595 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1596 } else {
1597 return;
1598 }
1599 }
1600
1601 /* Mirror output ports don't participate in picking the local hardware
1602 * address. ofproto can't help us find out whether a given port is a
1603 * mirror output because we haven't configured mirrors yet, so we need to
1604 * accumulate them ourselves. */
1605 hmapx_init(&mirror_output_ports);
1606 for (i = 0; i < br->cfg->n_mirrors; i++) {
1607 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1608 if (m->output_port) {
1609 hmapx_add(&mirror_output_ports, m->output_port);
1610 }
1611 }
1612
1613 /* Otherwise choose the minimum non-local MAC address among all of the
1614 * interfaces. */
1615 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1616 uint8_t iface_ea[ETH_ADDR_LEN];
1617 struct iface *candidate;
1618 struct iface *iface;
1619
1620 /* Mirror output ports don't participate. */
1621 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1622 continue;
1623 }
1624
1625 /* Choose the MAC address to represent the port. */
1626 iface = NULL;
1627 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1628 /* Find the interface with this Ethernet address (if any) so that
1629 * we can provide the correct devname to the caller. */
1630 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1631 uint8_t candidate_ea[ETH_ADDR_LEN];
1632 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1633 && eth_addr_equals(iface_ea, candidate_ea)) {
1634 iface = candidate;
1635 }
1636 }
1637 } else {
1638 /* Choose the interface whose MAC address will represent the port.
1639 * The Linux kernel bonding code always chooses the MAC address of
1640 * the first slave added to a bond, and the Fedora networking
1641 * scripts always add slaves to a bond in alphabetical order, so
1642 * for compatibility we choose the interface with the name that is
1643 * first in alphabetical order. */
1644 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1645 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1646 iface = candidate;
1647 }
1648 }
1649
1650 /* The local port doesn't count (since we're trying to choose its
1651 * MAC address anyway). */
1652 if (iface->ofp_port == OFPP_LOCAL) {
1653 continue;
1654 }
1655
1656 /* Grab MAC. */
1657 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1658 if (error) {
1659 continue;
1660 }
1661 }
1662
1663 /* Compare against our current choice. */
1664 if (!eth_addr_is_multicast(iface_ea) &&
1665 !eth_addr_is_local(iface_ea) &&
1666 !eth_addr_is_reserved(iface_ea) &&
1667 !eth_addr_is_zero(iface_ea) &&
1668 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1669 {
1670 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1671 *hw_addr_iface = iface;
1672 found_addr = true;
1673 }
1674 }
1675
1676 if (!found_addr) {
1677 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1678 *hw_addr_iface = NULL;
1679 }
1680
1681 hmapx_destroy(&mirror_output_ports);
1682 }
1683
1684 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1685 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1686 * an interface on 'br', then that interface must be passed in as
1687 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1688 * 'hw_addr_iface' must be passed in as a null pointer. */
1689 static uint64_t
1690 bridge_pick_datapath_id(struct bridge *br,
1691 const uint8_t bridge_ea[ETH_ADDR_LEN],
1692 struct iface *hw_addr_iface)
1693 {
1694 /*
1695 * The procedure for choosing a bridge MAC address will, in the most
1696 * ordinary case, also choose a unique MAC that we can use as a datapath
1697 * ID. In some special cases, though, multiple bridges will end up with
1698 * the same MAC address. This is OK for the bridges, but it will confuse
1699 * the OpenFlow controller, because each datapath needs a unique datapath
1700 * ID.
1701 *
1702 * Datapath IDs must be unique. It is also very desirable that they be
1703 * stable from one run to the next, so that policy set on a datapath
1704 * "sticks".
1705 */
1706 const char *datapath_id;
1707 uint64_t dpid;
1708
1709 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1710 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1711 return dpid;
1712 }
1713
1714 if (!hw_addr_iface) {
1715 /*
1716 * A purely internal bridge, that is, one that has no non-virtual
1717 * network devices on it at all, is difficult because it has no
1718 * natural unique identifier at all.
1719 *
1720 * When the host is a XenServer, we handle this case by hashing the
1721 * host's UUID with the name of the bridge. Names of bridges are
1722 * persistent across XenServer reboots, although they can be reused if
1723 * an internal network is destroyed and then a new one is later
1724 * created, so this is fairly effective.
1725 *
1726 * When the host is not a XenServer, we punt by using a random MAC
1727 * address on each run.
1728 */
1729 const char *host_uuid = xenserver_get_host_uuid();
1730 if (host_uuid) {
1731 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1732 dpid = dpid_from_hash(combined, strlen(combined));
1733 free(combined);
1734 return dpid;
1735 }
1736 }
1737
1738 return eth_addr_to_uint64(bridge_ea);
1739 }
1740
1741 static uint64_t
1742 dpid_from_hash(const void *data, size_t n)
1743 {
1744 uint8_t hash[SHA1_DIGEST_SIZE];
1745
1746 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1747 sha1_bytes(data, n, hash);
1748 eth_addr_mark_random(hash);
1749 return eth_addr_to_uint64(hash);
1750 }
1751
1752 static void
1753 iface_refresh_status(struct iface *iface)
1754 {
1755 struct smap smap;
1756
1757 enum netdev_features current;
1758 int64_t bps;
1759 int mtu;
1760 int64_t mtu_64;
1761 uint8_t mac[ETH_ADDR_LEN];
1762 int64_t ifindex64;
1763 int error;
1764
1765 if (iface_is_synthetic(iface)) {
1766 return;
1767 }
1768
1769 smap_init(&smap);
1770
1771 if (!netdev_get_status(iface->netdev, &smap)) {
1772 ovsrec_interface_set_status(iface->cfg, &smap);
1773 } else {
1774 ovsrec_interface_set_status(iface->cfg, NULL);
1775 }
1776
1777 smap_destroy(&smap);
1778
1779 error = netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1780 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1781 if (bps) {
1782 ovsrec_interface_set_duplex(iface->cfg,
1783 netdev_features_is_full_duplex(current)
1784 ? "full" : "half");
1785 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1786 } else {
1787 ovsrec_interface_set_duplex(iface->cfg, NULL);
1788 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1789 }
1790
1791 error = netdev_get_mtu(iface->netdev, &mtu);
1792 if (!error) {
1793 mtu_64 = mtu;
1794 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1795 } else {
1796 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1797 }
1798
1799 error = netdev_get_etheraddr(iface->netdev, mac);
1800 if (!error) {
1801 char mac_string[32];
1802
1803 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1804 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1805 } else {
1806 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1807 }
1808
1809 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1810 * if there is no valid ifindex number. */
1811 ifindex64 = netdev_get_ifindex(iface->netdev);
1812 if (ifindex64 < 0) {
1813 ifindex64 = 0;
1814 }
1815 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1816 }
1817
1818 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1819 * synthetic. */
1820 static void
1821 iface_refresh_cfm_stats(struct iface *iface)
1822 {
1823 const struct ovsrec_interface *cfg = iface->cfg;
1824 struct ofproto_cfm_status status;
1825
1826 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1827 iface->ofp_port, &status)) {
1828 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1829 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1830 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1831 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1832 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1833 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1834 } else {
1835 const char *reasons[CFM_FAULT_N_REASONS];
1836 int64_t cfm_health = status.health;
1837 int64_t cfm_flap_count = status.flap_count;
1838 bool faulted = status.faults != 0;
1839 size_t i, j;
1840
1841 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1842
1843 j = 0;
1844 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1845 int reason = 1 << i;
1846 if (status.faults & reason) {
1847 reasons[j++] = cfm_fault_reason_to_str(reason);
1848 }
1849 }
1850 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1851
1852 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1853
1854 if (status.remote_opstate >= 0) {
1855 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1856 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1857 } else {
1858 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1859 }
1860
1861 ovsrec_interface_set_cfm_remote_mpids(cfg,
1862 (const int64_t *)status.rmps,
1863 status.n_rmps);
1864 if (cfm_health >= 0) {
1865 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1866 } else {
1867 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1868 }
1869
1870 free(status.rmps);
1871 }
1872 }
1873
1874 static void
1875 iface_refresh_stats(struct iface *iface)
1876 {
1877 #define IFACE_STATS \
1878 IFACE_STAT(rx_packets, "rx_packets") \
1879 IFACE_STAT(tx_packets, "tx_packets") \
1880 IFACE_STAT(rx_bytes, "rx_bytes") \
1881 IFACE_STAT(tx_bytes, "tx_bytes") \
1882 IFACE_STAT(rx_dropped, "rx_dropped") \
1883 IFACE_STAT(tx_dropped, "tx_dropped") \
1884 IFACE_STAT(rx_errors, "rx_errors") \
1885 IFACE_STAT(tx_errors, "tx_errors") \
1886 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1887 IFACE_STAT(rx_over_errors, "rx_over_err") \
1888 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1889 IFACE_STAT(collisions, "collisions")
1890
1891 #define IFACE_STAT(MEMBER, NAME) + 1
1892 enum { N_IFACE_STATS = IFACE_STATS };
1893 #undef IFACE_STAT
1894 int64_t values[N_IFACE_STATS];
1895 char *keys[N_IFACE_STATS];
1896 int n;
1897
1898 struct netdev_stats stats;
1899
1900 if (iface_is_synthetic(iface)) {
1901 return;
1902 }
1903
1904 /* Intentionally ignore return value, since errors will set 'stats' to
1905 * all-1s, and we will deal with that correctly below. */
1906 netdev_get_stats(iface->netdev, &stats);
1907
1908 /* Copy statistics into keys[] and values[]. */
1909 n = 0;
1910 #define IFACE_STAT(MEMBER, NAME) \
1911 if (stats.MEMBER != UINT64_MAX) { \
1912 keys[n] = NAME; \
1913 values[n] = stats.MEMBER; \
1914 n++; \
1915 }
1916 IFACE_STATS;
1917 #undef IFACE_STAT
1918 ovs_assert(n <= N_IFACE_STATS);
1919
1920 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1921 #undef IFACE_STATS
1922 }
1923
1924 static void
1925 br_refresh_stp_status(struct bridge *br)
1926 {
1927 struct smap smap = SMAP_INITIALIZER(&smap);
1928 struct ofproto *ofproto = br->ofproto;
1929 struct ofproto_stp_status status;
1930
1931 if (ofproto_get_stp_status(ofproto, &status)) {
1932 return;
1933 }
1934
1935 if (!status.enabled) {
1936 ovsrec_bridge_set_status(br->cfg, NULL);
1937 return;
1938 }
1939
1940 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1941 STP_ID_ARGS(status.bridge_id));
1942 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1943 STP_ID_ARGS(status.designated_root));
1944 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1945
1946 ovsrec_bridge_set_status(br->cfg, &smap);
1947 smap_destroy(&smap);
1948 }
1949
1950 static void
1951 port_refresh_stp_status(struct port *port)
1952 {
1953 struct ofproto *ofproto = port->bridge->ofproto;
1954 struct iface *iface;
1955 struct ofproto_port_stp_status status;
1956 struct smap smap;
1957
1958 if (port_is_synthetic(port)) {
1959 return;
1960 }
1961
1962 /* STP doesn't currently support bonds. */
1963 if (!list_is_singleton(&port->ifaces)) {
1964 ovsrec_port_set_status(port->cfg, NULL);
1965 return;
1966 }
1967
1968 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1969 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1970 return;
1971 }
1972
1973 if (!status.enabled) {
1974 ovsrec_port_set_status(port->cfg, NULL);
1975 return;
1976 }
1977
1978 /* Set Status column. */
1979 smap_init(&smap);
1980 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1981 smap_add(&smap, "stp_state", stp_state_name(status.state));
1982 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1983 smap_add(&smap, "stp_role", stp_role_name(status.role));
1984 ovsrec_port_set_status(port->cfg, &smap);
1985 smap_destroy(&smap);
1986 }
1987
1988 static void
1989 port_refresh_stp_stats(struct port *port)
1990 {
1991 struct ofproto *ofproto = port->bridge->ofproto;
1992 struct iface *iface;
1993 struct ofproto_port_stp_stats stats;
1994 char *keys[3];
1995 int64_t int_values[3];
1996
1997 if (port_is_synthetic(port)) {
1998 return;
1999 }
2000
2001 /* STP doesn't currently support bonds. */
2002 if (!list_is_singleton(&port->ifaces)) {
2003 return;
2004 }
2005
2006 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2007 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2008 return;
2009 }
2010
2011 if (!stats.enabled) {
2012 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2013 return;
2014 }
2015
2016 /* Set Statistics column. */
2017 keys[0] = "stp_tx_count";
2018 int_values[0] = stats.tx_count;
2019 keys[1] = "stp_rx_count";
2020 int_values[1] = stats.rx_count;
2021 keys[2] = "stp_error_count";
2022 int_values[2] = stats.error_count;
2023
2024 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2025 ARRAY_SIZE(int_values));
2026 }
2027
2028 static bool
2029 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2030 {
2031 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2032 }
2033
2034 static void
2035 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2036 {
2037 bool enable = enable_system_stats(cfg);
2038
2039 system_stats_enable(enable);
2040 if (!enable) {
2041 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2042 }
2043 }
2044
2045 static void
2046 run_system_stats(void)
2047 {
2048 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2049 struct smap *stats;
2050
2051 stats = system_stats_run();
2052 if (stats && cfg) {
2053 struct ovsdb_idl_txn *txn;
2054 struct ovsdb_datum datum;
2055
2056 txn = ovsdb_idl_txn_create(idl);
2057 ovsdb_datum_from_smap(&datum, stats);
2058 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2059 &datum);
2060 ovsdb_idl_txn_commit(txn);
2061 ovsdb_idl_txn_destroy(txn);
2062
2063 free(stats);
2064 }
2065 }
2066
2067 static const char *
2068 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2069 {
2070 switch (role) {
2071 case OFPCR12_ROLE_EQUAL:
2072 return "other";
2073 case OFPCR12_ROLE_MASTER:
2074 return "master";
2075 case OFPCR12_ROLE_SLAVE:
2076 return "slave";
2077 case OFPCR12_ROLE_NOCHANGE:
2078 default:
2079 return "*** INVALID ROLE ***";
2080 }
2081 }
2082
2083 static void
2084 refresh_controller_status(void)
2085 {
2086 struct bridge *br;
2087 struct shash info;
2088 const struct ovsrec_controller *cfg;
2089
2090 shash_init(&info);
2091
2092 /* Accumulate status for controllers on all bridges. */
2093 HMAP_FOR_EACH (br, node, &all_bridges) {
2094 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2095 }
2096
2097 /* Update each controller in the database with current status. */
2098 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2099 struct ofproto_controller_info *cinfo =
2100 shash_find_data(&info, cfg->target);
2101
2102 if (cinfo) {
2103 struct smap smap = SMAP_INITIALIZER(&smap);
2104 const char **values = cinfo->pairs.values;
2105 const char **keys = cinfo->pairs.keys;
2106 size_t i;
2107
2108 for (i = 0; i < cinfo->pairs.n; i++) {
2109 smap_add(&smap, keys[i], values[i]);
2110 }
2111
2112 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2113 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2114 cinfo->role));
2115 ovsrec_controller_set_status(cfg, &smap);
2116 smap_destroy(&smap);
2117 } else {
2118 ovsrec_controller_set_is_connected(cfg, false);
2119 ovsrec_controller_set_role(cfg, NULL);
2120 ovsrec_controller_set_status(cfg, NULL);
2121 }
2122 }
2123
2124 ofproto_free_ofproto_controller_info(&info);
2125 }
2126 \f
2127 /* "Instant" stats.
2128 *
2129 * Some information in the database must be kept as up-to-date as possible to
2130 * allow controllers to respond rapidly to network outages. We call these
2131 * statistics "instant" stats.
2132 *
2133 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2134 * assuming that they've changed. The only means we have to determine whether
2135 * they have changed are:
2136 *
2137 * - Try to commit changes to the database. If nothing changed, then
2138 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2139 * value.
2140 *
2141 * - instant_stats_run() is called late in the run loop, after anything that
2142 * might change any of the instant stats.
2143 *
2144 * We use these two facts together to avoid waking the process up every
2145 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2146 */
2147
2148 /* Minimum interval between writing updates to the instant stats to the
2149 * database. */
2150 #define INSTANT_INTERVAL_MSEC 100
2151
2152 /* Current instant stats database transaction, NULL if there is no ongoing
2153 * transaction. */
2154 static struct ovsdb_idl_txn *instant_txn;
2155
2156 /* Next time (in msec on monotonic clock) at which we will update the instant
2157 * stats. */
2158 static long long int instant_next_txn = LLONG_MIN;
2159
2160 /* True if the run loop has run since we last saw that the instant stats were
2161 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2162 * to refresh the instant stats. */
2163 static bool instant_stats_could_have_changed;
2164
2165 static void
2166 instant_stats_run(void)
2167 {
2168 enum ovsdb_idl_txn_status status;
2169
2170 instant_stats_could_have_changed = true;
2171
2172 if (!instant_txn) {
2173 struct bridge *br;
2174 uint64_t seq;
2175
2176 if (time_msec() < instant_next_txn) {
2177 return;
2178 }
2179 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2180
2181 seq = seq_read(connectivity_seq_get());
2182 if (seq == connectivity_seqno) {
2183 return;
2184 }
2185 connectivity_seqno = seq;
2186
2187 instant_txn = ovsdb_idl_txn_create(idl);
2188 HMAP_FOR_EACH (br, node, &all_bridges) {
2189 struct iface *iface;
2190 struct port *port;
2191
2192 br_refresh_stp_status(br);
2193
2194 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2195 port_refresh_stp_status(port);
2196 }
2197
2198 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2199 enum netdev_flags flags;
2200 struct smap smap;
2201 const char *link_state;
2202 int64_t link_resets;
2203 int current, error;
2204
2205 if (iface_is_synthetic(iface)) {
2206 continue;
2207 }
2208
2209 current = ofproto_port_is_lacp_current(br->ofproto,
2210 iface->ofp_port);
2211 if (current >= 0) {
2212 bool bl = current;
2213 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2214 } else {
2215 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2216 }
2217
2218 error = netdev_get_flags(iface->netdev, &flags);
2219 if (!error) {
2220 const char *state = flags & NETDEV_UP ? "up" : "down";
2221 ovsrec_interface_set_admin_state(iface->cfg, state);
2222 } else {
2223 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2224 }
2225
2226 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2227 ovsrec_interface_set_link_state(iface->cfg, link_state);
2228
2229 link_resets = netdev_get_carrier_resets(iface->netdev);
2230 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2231
2232 iface_refresh_cfm_stats(iface);
2233
2234 smap_init(&smap);
2235 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2236 &smap);
2237 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2238 smap_destroy(&smap);
2239 }
2240 }
2241 }
2242
2243 status = ovsdb_idl_txn_commit(instant_txn);
2244 if (status != TXN_INCOMPLETE) {
2245 ovsdb_idl_txn_destroy(instant_txn);
2246 instant_txn = NULL;
2247 }
2248 if (status == TXN_UNCHANGED) {
2249 instant_stats_could_have_changed = false;
2250 }
2251 }
2252
2253 static void
2254 instant_stats_wait(void)
2255 {
2256 if (instant_txn) {
2257 ovsdb_idl_txn_wait(instant_txn);
2258 } else if (instant_stats_could_have_changed) {
2259 poll_timer_wait_until(instant_next_txn);
2260 }
2261 }
2262 \f
2263 static void
2264 bridge_run__(void)
2265 {
2266 struct bridge *br;
2267 struct sset types;
2268 const char *type;
2269
2270 /* Let each datapath type do the work that it needs to do. */
2271 sset_init(&types);
2272 ofproto_enumerate_types(&types);
2273 SSET_FOR_EACH (type, &types) {
2274 ofproto_type_run(type);
2275 }
2276 sset_destroy(&types);
2277
2278 /* Let each bridge do the work that it needs to do. */
2279 HMAP_FOR_EACH (br, node, &all_bridges) {
2280 ofproto_run(br->ofproto);
2281 }
2282 }
2283
2284 void
2285 bridge_run(void)
2286 {
2287 static struct ovsrec_open_vswitch null_cfg;
2288 const struct ovsrec_open_vswitch *cfg;
2289
2290 bool vlan_splinters_changed;
2291 struct bridge *br;
2292
2293 ovsrec_open_vswitch_init(&null_cfg);
2294
2295 ovsdb_idl_run(idl);
2296
2297 if (ovsdb_idl_is_lock_contended(idl)) {
2298 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2299 struct bridge *br, *next_br;
2300
2301 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2302 "disabling this process (pid %ld) until it goes away",
2303 (long int) getpid());
2304
2305 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2306 bridge_destroy(br);
2307 }
2308 /* Since we will not be running system_stats_run() in this process
2309 * with the current situation of multiple ovs-vswitchd daemons,
2310 * disable system stats collection. */
2311 system_stats_enable(false);
2312 return;
2313 } else if (!ovsdb_idl_has_lock(idl)) {
2314 return;
2315 }
2316 cfg = ovsrec_open_vswitch_first(idl);
2317
2318 /* Initialize the ofproto library. This only needs to run once, but
2319 * it must be done after the configuration is set. If the
2320 * initialization has already occurred, bridge_init_ofproto()
2321 * returns immediately. */
2322 bridge_init_ofproto(cfg);
2323
2324 /* Once the value of flow-restore-wait is false, we no longer should
2325 * check its value from the database. */
2326 if (cfg && ofproto_get_flow_restore_wait()) {
2327 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2328 "flow-restore-wait", false));
2329 }
2330
2331 bridge_run__();
2332
2333 /* Re-configure SSL. We do this on every trip through the main loop,
2334 * instead of just when the database changes, because the contents of the
2335 * key and certificate files can change without the database changing.
2336 *
2337 * We do this before bridge_reconfigure() because that function might
2338 * initiate SSL connections and thus requires SSL to be configured. */
2339 if (cfg && cfg->ssl) {
2340 const struct ovsrec_ssl *ssl = cfg->ssl;
2341
2342 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2343 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2344 }
2345
2346 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2347 * usage has changed. */
2348 vlan_splinters_changed = false;
2349 if (vlan_splinters_enabled_anywhere) {
2350 HMAP_FOR_EACH (br, node, &all_bridges) {
2351 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2352 vlan_splinters_changed = true;
2353 break;
2354 }
2355 }
2356 }
2357
2358 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2359 struct ovsdb_idl_txn *txn;
2360
2361 idl_seqno = ovsdb_idl_get_seqno(idl);
2362 txn = ovsdb_idl_txn_create(idl);
2363 bridge_reconfigure(cfg ? cfg : &null_cfg);
2364
2365 if (cfg) {
2366 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2367 }
2368
2369 /* If we are completing our initial configuration for this run
2370 * of ovs-vswitchd, then keep the transaction around to monitor
2371 * it for completion. */
2372 if (initial_config_done) {
2373 ovsdb_idl_txn_commit(txn);
2374 ovsdb_idl_txn_destroy(txn);
2375 } else {
2376 initial_config_done = true;
2377 daemonize_txn = txn;
2378 }
2379 }
2380
2381 if (daemonize_txn) {
2382 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2383 if (status != TXN_INCOMPLETE) {
2384 ovsdb_idl_txn_destroy(daemonize_txn);
2385 daemonize_txn = NULL;
2386
2387 /* ovs-vswitchd has completed initialization, so allow the
2388 * process that forked us to exit successfully. */
2389 daemonize_complete();
2390
2391 vlog_enable_async();
2392
2393 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2394 }
2395 }
2396
2397 /* Refresh interface and mirror stats if necessary. */
2398 if (time_msec() >= iface_stats_timer) {
2399 if (cfg) {
2400 struct ovsdb_idl_txn *txn;
2401
2402 txn = ovsdb_idl_txn_create(idl);
2403 HMAP_FOR_EACH (br, node, &all_bridges) {
2404 struct port *port;
2405 struct mirror *m;
2406
2407 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2408 struct iface *iface;
2409
2410 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2411 iface_refresh_stats(iface);
2412 iface_refresh_status(iface);
2413 }
2414
2415 port_refresh_stp_stats(port);
2416 }
2417
2418 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2419 mirror_refresh_stats(m);
2420 }
2421
2422 }
2423 refresh_controller_status();
2424 ovsdb_idl_txn_commit(txn);
2425 ovsdb_idl_txn_destroy(txn); /* XXX */
2426 }
2427
2428 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2429 }
2430
2431 run_system_stats();
2432 instant_stats_run();
2433 }
2434
2435 void
2436 bridge_wait(void)
2437 {
2438 struct sset types;
2439 const char *type;
2440
2441 ovsdb_idl_wait(idl);
2442 if (daemonize_txn) {
2443 ovsdb_idl_txn_wait(daemonize_txn);
2444 }
2445
2446 sset_init(&types);
2447 ofproto_enumerate_types(&types);
2448 SSET_FOR_EACH (type, &types) {
2449 ofproto_type_wait(type);
2450 }
2451 sset_destroy(&types);
2452
2453 if (!hmap_is_empty(&all_bridges)) {
2454 struct bridge *br;
2455
2456 HMAP_FOR_EACH (br, node, &all_bridges) {
2457 ofproto_wait(br->ofproto);
2458 }
2459 poll_timer_wait_until(iface_stats_timer);
2460 }
2461
2462 system_stats_wait();
2463 instant_stats_wait();
2464 }
2465
2466 /* Adds some memory usage statistics for bridges into 'usage', for use with
2467 * memory_report(). */
2468 void
2469 bridge_get_memory_usage(struct simap *usage)
2470 {
2471 struct bridge *br;
2472 struct sset types;
2473 const char *type;
2474
2475 sset_init(&types);
2476 ofproto_enumerate_types(&types);
2477 SSET_FOR_EACH (type, &types) {
2478 ofproto_type_get_memory_usage(type, usage);
2479 }
2480 sset_destroy(&types);
2481
2482 HMAP_FOR_EACH (br, node, &all_bridges) {
2483 ofproto_get_memory_usage(br->ofproto, usage);
2484 }
2485 }
2486 \f
2487 /* QoS unixctl user interface functions. */
2488
2489 struct qos_unixctl_show_cbdata {
2490 struct ds *ds;
2491 struct iface *iface;
2492 };
2493
2494 static void
2495 qos_unixctl_show_queue(unsigned int queue_id,
2496 const struct smap *details,
2497 struct iface *iface,
2498 struct ds *ds)
2499 {
2500 struct netdev_queue_stats stats;
2501 struct smap_node *node;
2502 int error;
2503
2504 ds_put_cstr(ds, "\n");
2505 if (queue_id) {
2506 ds_put_format(ds, "Queue %u:\n", queue_id);
2507 } else {
2508 ds_put_cstr(ds, "Default:\n");
2509 }
2510
2511 SMAP_FOR_EACH (node, details) {
2512 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2513 }
2514
2515 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2516 if (!error) {
2517 if (stats.tx_packets != UINT64_MAX) {
2518 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2519 }
2520
2521 if (stats.tx_bytes != UINT64_MAX) {
2522 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2523 }
2524
2525 if (stats.tx_errors != UINT64_MAX) {
2526 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2527 }
2528 } else {
2529 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2530 queue_id, ovs_strerror(error));
2531 }
2532 }
2533
2534 static void
2535 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2536 const char *argv[], void *aux OVS_UNUSED)
2537 {
2538 struct ds ds = DS_EMPTY_INITIALIZER;
2539 struct smap smap = SMAP_INITIALIZER(&smap);
2540 struct iface *iface;
2541 const char *type;
2542 struct smap_node *node;
2543
2544 iface = iface_find(argv[1]);
2545 if (!iface) {
2546 unixctl_command_reply_error(conn, "no such interface");
2547 return;
2548 }
2549
2550 netdev_get_qos(iface->netdev, &type, &smap);
2551
2552 if (*type != '\0') {
2553 struct netdev_queue_dump dump;
2554 struct smap details;
2555 unsigned int queue_id;
2556
2557 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2558
2559 SMAP_FOR_EACH (node, &smap) {
2560 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2561 }
2562
2563 smap_init(&details);
2564 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2565 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2566 }
2567 smap_destroy(&details);
2568
2569 unixctl_command_reply(conn, ds_cstr(&ds));
2570 } else {
2571 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2572 unixctl_command_reply_error(conn, ds_cstr(&ds));
2573 }
2574
2575 smap_destroy(&smap);
2576 ds_destroy(&ds);
2577 }
2578 \f
2579 /* Bridge reconfiguration functions. */
2580 static void
2581 bridge_create(const struct ovsrec_bridge *br_cfg)
2582 {
2583 struct bridge *br;
2584
2585 ovs_assert(!bridge_lookup(br_cfg->name));
2586 br = xzalloc(sizeof *br);
2587
2588 br->name = xstrdup(br_cfg->name);
2589 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2590 br->cfg = br_cfg;
2591
2592 /* Derive the default Ethernet address from the bridge's UUID. This should
2593 * be unique and it will be stable between ovs-vswitchd runs. */
2594 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2595 eth_addr_mark_random(br->default_ea);
2596
2597 hmap_init(&br->ports);
2598 hmap_init(&br->ifaces);
2599 hmap_init(&br->iface_by_name);
2600 hmap_init(&br->mirrors);
2601
2602 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2603 }
2604
2605 static void
2606 bridge_destroy(struct bridge *br)
2607 {
2608 if (br) {
2609 struct mirror *mirror, *next_mirror;
2610 struct port *port, *next_port;
2611
2612 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2613 port_destroy(port);
2614 }
2615 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2616 mirror_destroy(mirror);
2617 }
2618
2619 hmap_remove(&all_bridges, &br->node);
2620 ofproto_destroy(br->ofproto);
2621 hmap_destroy(&br->ifaces);
2622 hmap_destroy(&br->ports);
2623 hmap_destroy(&br->iface_by_name);
2624 hmap_destroy(&br->mirrors);
2625 free(br->name);
2626 free(br->type);
2627 free(br);
2628 }
2629 }
2630
2631 static struct bridge *
2632 bridge_lookup(const char *name)
2633 {
2634 struct bridge *br;
2635
2636 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2637 if (!strcmp(br->name, name)) {
2638 return br;
2639 }
2640 }
2641 return NULL;
2642 }
2643
2644 /* Handle requests for a listing of all flows known by the OpenFlow
2645 * stack, including those normally hidden. */
2646 static void
2647 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2648 const char *argv[], void *aux OVS_UNUSED)
2649 {
2650 struct bridge *br;
2651 struct ds results;
2652
2653 br = bridge_lookup(argv[1]);
2654 if (!br) {
2655 unixctl_command_reply_error(conn, "Unknown bridge");
2656 return;
2657 }
2658
2659 ds_init(&results);
2660 ofproto_get_all_flows(br->ofproto, &results);
2661
2662 unixctl_command_reply(conn, ds_cstr(&results));
2663 ds_destroy(&results);
2664 }
2665
2666 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2667 * connections and reconnect. If BRIDGE is not specified, then all bridges
2668 * drop their controller connections and reconnect. */
2669 static void
2670 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2671 const char *argv[], void *aux OVS_UNUSED)
2672 {
2673 struct bridge *br;
2674 if (argc > 1) {
2675 br = bridge_lookup(argv[1]);
2676 if (!br) {
2677 unixctl_command_reply_error(conn, "Unknown bridge");
2678 return;
2679 }
2680 ofproto_reconnect_controllers(br->ofproto);
2681 } else {
2682 HMAP_FOR_EACH (br, node, &all_bridges) {
2683 ofproto_reconnect_controllers(br->ofproto);
2684 }
2685 }
2686 unixctl_command_reply(conn, NULL);
2687 }
2688
2689 static size_t
2690 bridge_get_controllers(const struct bridge *br,
2691 struct ovsrec_controller ***controllersp)
2692 {
2693 struct ovsrec_controller **controllers;
2694 size_t n_controllers;
2695
2696 controllers = br->cfg->controller;
2697 n_controllers = br->cfg->n_controller;
2698
2699 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2700 controllers = NULL;
2701 n_controllers = 0;
2702 }
2703
2704 if (controllersp) {
2705 *controllersp = controllers;
2706 }
2707 return n_controllers;
2708 }
2709
2710 static void
2711 bridge_collect_wanted_ports(struct bridge *br,
2712 const unsigned long int *splinter_vlans,
2713 struct shash *wanted_ports)
2714 {
2715 size_t i;
2716
2717 shash_init(wanted_ports);
2718
2719 for (i = 0; i < br->cfg->n_ports; i++) {
2720 const char *name = br->cfg->ports[i]->name;
2721 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2722 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2723 br->name, name);
2724 }
2725 }
2726 if (bridge_get_controllers(br, NULL)
2727 && !shash_find(wanted_ports, br->name)) {
2728 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2729 br->name, br->name);
2730
2731 ovsrec_interface_init(&br->synth_local_iface);
2732 ovsrec_port_init(&br->synth_local_port);
2733
2734 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2735 br->synth_local_port.n_interfaces = 1;
2736 br->synth_local_port.name = br->name;
2737
2738 br->synth_local_iface.name = br->name;
2739 br->synth_local_iface.type = "internal";
2740
2741 br->synth_local_ifacep = &br->synth_local_iface;
2742
2743 shash_add(wanted_ports, br->name, &br->synth_local_port);
2744 }
2745
2746 if (splinter_vlans) {
2747 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2748 }
2749 }
2750
2751 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2752 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2753 * 'br' needs to complete its configuration. */
2754 static void
2755 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2756 {
2757 struct shash_node *port_node;
2758 struct port *port, *next;
2759
2760 /* Get rid of deleted ports.
2761 * Get rid of deleted interfaces on ports that still exist. */
2762 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2763 port->cfg = shash_find_data(wanted_ports, port->name);
2764 if (!port->cfg) {
2765 port_destroy(port);
2766 } else {
2767 port_del_ifaces(port);
2768 }
2769 }
2770
2771 /* Update iface->cfg and iface->type in interfaces that still exist. */
2772 SHASH_FOR_EACH (port_node, wanted_ports) {
2773 const struct ovsrec_port *port = port_node->data;
2774 size_t i;
2775
2776 for (i = 0; i < port->n_interfaces; i++) {
2777 const struct ovsrec_interface *cfg = port->interfaces[i];
2778 struct iface *iface = iface_lookup(br, cfg->name);
2779 const char *type = iface_get_type(cfg, br->cfg);
2780
2781 if (iface) {
2782 iface->cfg = cfg;
2783 iface->type = type;
2784 } else if (!strcmp(type, "null")) {
2785 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2786 " may be removed in February 2013. Please email"
2787 " dev@openvswitch.org with concerns.",
2788 cfg->name);
2789 } else {
2790 /* We will add new interfaces later. */
2791 }
2792 }
2793 }
2794 }
2795
2796 /* Initializes 'oc' appropriately as a management service controller for
2797 * 'br'.
2798 *
2799 * The caller must free oc->target when it is no longer needed. */
2800 static void
2801 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2802 struct ofproto_controller *oc)
2803 {
2804 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2805 oc->max_backoff = 0;
2806 oc->probe_interval = 60;
2807 oc->band = OFPROTO_OUT_OF_BAND;
2808 oc->rate_limit = 0;
2809 oc->burst_limit = 0;
2810 oc->enable_async_msgs = true;
2811 }
2812
2813 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2814 static void
2815 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2816 struct ofproto_controller *oc)
2817 {
2818 int dscp;
2819
2820 oc->target = c->target;
2821 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2822 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2823 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2824 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2825 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2826 oc->burst_limit = (c->controller_burst_limit
2827 ? *c->controller_burst_limit : 0);
2828 oc->enable_async_msgs = (!c->enable_async_messages
2829 || *c->enable_async_messages);
2830 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2831 if (dscp < 0 || dscp > 63) {
2832 dscp = DSCP_DEFAULT;
2833 }
2834 oc->dscp = dscp;
2835 }
2836
2837 /* Configures the IP stack for 'br''s local interface properly according to the
2838 * configuration in 'c'. */
2839 static void
2840 bridge_configure_local_iface_netdev(struct bridge *br,
2841 struct ovsrec_controller *c)
2842 {
2843 struct netdev *netdev;
2844 struct in_addr mask, gateway;
2845
2846 struct iface *local_iface;
2847 struct in_addr ip;
2848
2849 /* If there's no local interface or no IP address, give up. */
2850 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2851 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2852 return;
2853 }
2854
2855 /* Bring up the local interface. */
2856 netdev = local_iface->netdev;
2857 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2858
2859 /* Configure the IP address and netmask. */
2860 if (!c->local_netmask
2861 || !inet_aton(c->local_netmask, &mask)
2862 || !mask.s_addr) {
2863 mask.s_addr = guess_netmask(ip.s_addr);
2864 }
2865 if (!netdev_set_in4(netdev, ip, mask)) {
2866 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2867 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2868 }
2869
2870 /* Configure the default gateway. */
2871 if (c->local_gateway
2872 && inet_aton(c->local_gateway, &gateway)
2873 && gateway.s_addr) {
2874 if (!netdev_add_router(netdev, gateway)) {
2875 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2876 br->name, IP_ARGS(gateway.s_addr));
2877 }
2878 }
2879 }
2880
2881 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2882 * in either string are treated as equal to any number of slashes in the other,
2883 * e.g. "x///y" is equal to "x/y".
2884 *
2885 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2886 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2887 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2888 * 'b' against a prefix of 'a'.
2889 */
2890 static bool
2891 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2892 {
2893 const char *b_start = b;
2894 for (;;) {
2895 if (b - b_start >= b_stoplen) {
2896 return true;
2897 } else if (*a != *b) {
2898 return false;
2899 } else if (*a == '/') {
2900 a += strspn(a, "/");
2901 b += strspn(b, "/");
2902 } else if (*a == '\0') {
2903 return true;
2904 } else {
2905 a++;
2906 b++;
2907 }
2908 }
2909 }
2910
2911 static void
2912 bridge_configure_remotes(struct bridge *br,
2913 const struct sockaddr_in *managers, size_t n_managers)
2914 {
2915 bool disable_in_band;
2916
2917 struct ovsrec_controller **controllers;
2918 size_t n_controllers;
2919
2920 enum ofproto_fail_mode fail_mode;
2921
2922 struct ofproto_controller *ocs;
2923 size_t n_ocs;
2924 size_t i;
2925
2926 /* Check if we should disable in-band control on this bridge. */
2927 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2928 false);
2929
2930 /* Set OpenFlow queue ID for in-band control. */
2931 ofproto_set_in_band_queue(br->ofproto,
2932 smap_get_int(&br->cfg->other_config,
2933 "in-band-queue", -1));
2934
2935 if (disable_in_band) {
2936 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2937 } else {
2938 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2939 }
2940
2941 n_controllers = bridge_get_controllers(br, &controllers);
2942
2943 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2944 n_ocs = 0;
2945
2946 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2947 for (i = 0; i < n_controllers; i++) {
2948 struct ovsrec_controller *c = controllers[i];
2949
2950 if (!strncmp(c->target, "punix:", 6)
2951 || !strncmp(c->target, "unix:", 5)) {
2952 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2953 char *whitelist;
2954
2955 if (!strncmp(c->target, "unix:", 5)) {
2956 /* Connect to a listening socket */
2957 whitelist = xasprintf("unix:%s/", ovs_rundir());
2958 if (strchr(c->target, '/') &&
2959 !equal_pathnames(c->target, whitelist,
2960 strlen(whitelist))) {
2961 /* Absolute path specified, but not in ovs_rundir */
2962 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2963 "controller \"%s\" due to possibility for "
2964 "remote exploit. Instead, specify socket "
2965 "in whitelisted \"%s\" or connect to "
2966 "\"unix:%s/%s.mgmt\" (which is always "
2967 "available without special configuration).",
2968 br->name, c->target, whitelist,
2969 ovs_rundir(), br->name);
2970 free(whitelist);
2971 continue;
2972 }
2973 } else {
2974 whitelist = xasprintf("punix:%s/%s.controller",
2975 ovs_rundir(), br->name);
2976 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2977 /* Prevent remote ovsdb-server users from accessing
2978 * arbitrary Unix domain sockets and overwriting arbitrary
2979 * local files. */
2980 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2981 "controller \"%s\" due to possibility of "
2982 "overwriting local files. Instead, specify "
2983 "whitelisted \"%s\" or connect to "
2984 "\"unix:%s/%s.mgmt\" (which is always "
2985 "available without special configuration).",
2986 br->name, c->target, whitelist,
2987 ovs_rundir(), br->name);
2988 free(whitelist);
2989 continue;
2990 }
2991 }
2992
2993 free(whitelist);
2994 }
2995
2996 bridge_configure_local_iface_netdev(br, c);
2997 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2998 if (disable_in_band) {
2999 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3000 }
3001 n_ocs++;
3002 }
3003
3004 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3005 bridge_get_allowed_versions(br));
3006 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3007 free(ocs);
3008
3009 /* Set the fail-mode. */
3010 fail_mode = !br->cfg->fail_mode
3011 || !strcmp(br->cfg->fail_mode, "standalone")
3012 ? OFPROTO_FAIL_STANDALONE
3013 : OFPROTO_FAIL_SECURE;
3014 ofproto_set_fail_mode(br->ofproto, fail_mode);
3015
3016 /* Configure OpenFlow controller connection snooping. */
3017 if (!ofproto_has_snoops(br->ofproto)) {
3018 struct sset snoops;
3019
3020 sset_init(&snoops);
3021 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3022 ovs_rundir(), br->name));
3023 ofproto_set_snoops(br->ofproto, &snoops);
3024 sset_destroy(&snoops);
3025 }
3026 }
3027
3028 static void
3029 bridge_configure_tables(struct bridge *br)
3030 {
3031 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3032 int n_tables;
3033 int i, j, k;
3034
3035 n_tables = ofproto_get_n_tables(br->ofproto);
3036 j = 0;
3037 for (i = 0; i < n_tables; i++) {
3038 struct ofproto_table_settings s;
3039
3040 s.name = NULL;
3041 s.max_flows = UINT_MAX;
3042 s.groups = NULL;
3043 s.n_groups = 0;
3044 s.n_prefix_fields = 0;
3045 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3046
3047 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3048 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3049
3050 s.name = cfg->name;
3051 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3052 s.max_flows = *cfg->flow_limit;
3053 }
3054 if (cfg->overflow_policy
3055 && !strcmp(cfg->overflow_policy, "evict")) {
3056
3057 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3058 for (k = 0; k < cfg->n_groups; k++) {
3059 const char *string = cfg->groups[k];
3060 char *msg;
3061
3062 msg = mf_parse_subfield__(&s.groups[k], &string);
3063 if (msg) {
3064 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3065 "'groups' (%s)", br->name, i, msg);
3066 free(msg);
3067 } else if (*string) {
3068 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3069 "element '%s' contains trailing garbage",
3070 br->name, i, cfg->groups[k]);
3071 } else {
3072 s.n_groups++;
3073 }
3074 }
3075 }
3076 /* Prefix lookup fields. */
3077 s.n_prefix_fields = 0;
3078 for (k = 0; k < cfg->n_prefixes; k++) {
3079 const char *name = cfg->prefixes[k];
3080 const struct mf_field *mf = mf_from_name(name);
3081 if (!mf) {
3082 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3083 br->name, name);
3084 continue;
3085 }
3086 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3087 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3088 "%s", br->name, name);
3089 continue;
3090 }
3091 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3092 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3093 "field not used: %s", br->name, name);
3094 continue;
3095 }
3096 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3097 }
3098 if (s.n_prefix_fields > 0) {
3099 int k;
3100 struct ds ds = DS_EMPTY_INITIALIZER;
3101 for (k = 0; k < s.n_prefix_fields; k++) {
3102 if (k) {
3103 ds_put_char(&ds, ',');
3104 }
3105 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3106 }
3107 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3108 br->name, i, ds_cstr(&ds));
3109 ds_destroy(&ds);
3110 }
3111 }
3112
3113 ofproto_configure_table(br->ofproto, i, &s);
3114
3115 free(s.groups);
3116 }
3117 for (; j < br->cfg->n_flow_tables; j++) {
3118 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3119 "%"PRId64" not supported by this datapath", br->name,
3120 br->cfg->key_flow_tables[j]);
3121 }
3122 }
3123
3124 static void
3125 bridge_configure_dp_desc(struct bridge *br)
3126 {
3127 ofproto_set_dp_desc(br->ofproto,
3128 smap_get(&br->cfg->other_config, "dp-desc"));
3129 }
3130 \f
3131 /* Port functions. */
3132
3133 static void iface_destroy__(struct iface *);
3134
3135 static struct port *
3136 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3137 {
3138 struct port *port;
3139
3140 port = xzalloc(sizeof *port);
3141 port->bridge = br;
3142 port->name = xstrdup(cfg->name);
3143 port->cfg = cfg;
3144 list_init(&port->ifaces);
3145
3146 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3147 return port;
3148 }
3149
3150 /* Deletes interfaces from 'port' that are no longer configured for it. */
3151 static void
3152 port_del_ifaces(struct port *port)
3153 {
3154 struct iface *iface, *next;
3155 struct sset new_ifaces;
3156 size_t i;
3157
3158 /* Collect list of new interfaces. */
3159 sset_init(&new_ifaces);
3160 for (i = 0; i < port->cfg->n_interfaces; i++) {
3161 const char *name = port->cfg->interfaces[i]->name;
3162 const char *type = port->cfg->interfaces[i]->type;
3163 if (strcmp(type, "null")) {
3164 sset_add(&new_ifaces, name);
3165 }
3166 }
3167
3168 /* Get rid of deleted interfaces. */
3169 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3170 if (!sset_contains(&new_ifaces, iface->name)) {
3171 iface_destroy(iface);
3172 }
3173 }
3174
3175 sset_destroy(&new_ifaces);
3176 }
3177
3178 static void
3179 port_destroy(struct port *port)
3180 {
3181 if (port) {
3182 struct bridge *br = port->bridge;
3183 struct iface *iface, *next;
3184
3185 if (br->ofproto) {
3186 ofproto_bundle_unregister(br->ofproto, port);
3187 }
3188
3189 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3190 iface_destroy__(iface);
3191 }
3192
3193 hmap_remove(&br->ports, &port->hmap_node);
3194 free(port->name);
3195 free(port);
3196 }
3197 }
3198
3199 static struct port *
3200 port_lookup(const struct bridge *br, const char *name)
3201 {
3202 struct port *port;
3203
3204 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3205 &br->ports) {
3206 if (!strcmp(port->name, name)) {
3207 return port;
3208 }
3209 }
3210 return NULL;
3211 }
3212
3213 static bool
3214 enable_lacp(struct port *port, bool *activep)
3215 {
3216 if (!port->cfg->lacp) {
3217 /* XXX when LACP implementation has been sufficiently tested, enable by
3218 * default and make active on bonded ports. */
3219 return false;
3220 } else if (!strcmp(port->cfg->lacp, "off")) {
3221 return false;
3222 } else if (!strcmp(port->cfg->lacp, "active")) {
3223 *activep = true;
3224 return true;
3225 } else if (!strcmp(port->cfg->lacp, "passive")) {
3226 *activep = false;
3227 return true;
3228 } else {
3229 VLOG_WARN("port %s: unknown LACP mode %s",
3230 port->name, port->cfg->lacp);
3231 return false;
3232 }
3233 }
3234
3235 static struct lacp_settings *
3236 port_configure_lacp(struct port *port, struct lacp_settings *s)
3237 {
3238 const char *lacp_time, *system_id;
3239 int priority;
3240
3241 if (!enable_lacp(port, &s->active)) {
3242 return NULL;
3243 }
3244
3245 s->name = port->name;
3246
3247 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3248 if (system_id) {
3249 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3250 ETH_ADDR_SCAN_ARGS(s->id))) {
3251 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3252 " address.", port->name, system_id);
3253 return NULL;
3254 }
3255 } else {
3256 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3257 }
3258
3259 if (eth_addr_is_zero(s->id)) {
3260 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3261 return NULL;
3262 }
3263
3264 /* Prefer bondable links if unspecified. */
3265 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3266 0);
3267 s->priority = (priority > 0 && priority <= UINT16_MAX
3268 ? priority
3269 : UINT16_MAX - !list_is_short(&port->ifaces));
3270
3271 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3272 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3273
3274 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3275 "lacp-fallback-ab", false);
3276
3277 return s;
3278 }
3279
3280 static void
3281 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3282 {
3283 int priority, portid, key;
3284
3285 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3286 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3287 0);
3288 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3289
3290 if (portid <= 0 || portid > UINT16_MAX) {
3291 portid = ofp_to_u16(iface->ofp_port);
3292 }
3293
3294 if (priority <= 0 || priority > UINT16_MAX) {
3295 priority = UINT16_MAX;
3296 }
3297
3298 if (key < 0 || key > UINT16_MAX) {
3299 key = 0;
3300 }
3301
3302 s->name = iface->name;
3303 s->id = portid;
3304 s->priority = priority;
3305 s->key = key;
3306 }
3307
3308 static void
3309 port_configure_bond(struct port *port, struct bond_settings *s)
3310 {
3311 const char *detect_s;
3312 struct iface *iface;
3313 int miimon_interval;
3314
3315 s->name = port->name;
3316 s->balance = BM_AB;
3317 if (port->cfg->bond_mode) {
3318 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3319 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3320 port->name, port->cfg->bond_mode,
3321 bond_mode_to_string(s->balance));
3322 }
3323 } else {
3324 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3325
3326 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3327 * active-backup. At some point we should remove this warning. */
3328 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3329 " in previous versions, the default bond_mode was"
3330 " balance-slb", port->name,
3331 bond_mode_to_string(s->balance));
3332 }
3333 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3334 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3335 "please use another bond type or disable flood_vlans",
3336 port->name);
3337 }
3338
3339 miimon_interval = smap_get_int(&port->cfg->other_config,
3340 "bond-miimon-interval", 0);
3341 if (miimon_interval <= 0) {
3342 miimon_interval = 200;
3343 }
3344
3345 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3346 if (!detect_s || !strcmp(detect_s, "carrier")) {
3347 miimon_interval = 0;
3348 } else if (strcmp(detect_s, "miimon")) {
3349 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3350 "defaulting to carrier", port->name, detect_s);
3351 miimon_interval = 0;
3352 }
3353
3354 s->up_delay = MAX(0, port->cfg->bond_updelay);
3355 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3356 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3357 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3358 "bond-rebalance-interval", 10000);
3359 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3360 s->rebalance_interval = 1000;
3361 }
3362
3363 s->fake_iface = port->cfg->bond_fake_iface;
3364
3365 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3366 "lacp-fallback-ab", false);
3367
3368 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3369 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3370 }
3371 }
3372
3373 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3374 * instead of obtaining it from the database. */
3375 static bool
3376 port_is_synthetic(const struct port *port)
3377 {
3378 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3379 }
3380 \f
3381 /* Interface functions. */
3382
3383 static bool
3384 iface_is_internal(const struct ovsrec_interface *iface,
3385 const struct ovsrec_bridge *br)
3386 {
3387 /* The local port and "internal" ports are always "internal". */
3388 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3389 }
3390
3391 /* Returns the correct network device type for interface 'iface' in bridge
3392 * 'br'. */
3393 static const char *
3394 iface_get_type(const struct ovsrec_interface *iface,
3395 const struct ovsrec_bridge *br)
3396 {
3397 const char *type;
3398
3399 /* The local port always has type "internal". Other ports take
3400 * their type from the database and default to "system" if none is
3401 * specified. */
3402 if (iface_is_internal(iface, br)) {
3403 type = "internal";
3404 } else {
3405 type = iface->type[0] ? iface->type : "system";
3406 }
3407
3408 return ofproto_port_open_type(br->datapath_type, type);
3409 }
3410
3411 static void
3412 iface_destroy__(struct iface *iface)
3413 {
3414 if (iface) {
3415 struct port *port = iface->port;
3416 struct bridge *br = port->bridge;
3417
3418 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3419 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3420 }
3421
3422 if (iface->ofp_port != OFPP_NONE) {
3423 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3424 }
3425
3426 list_remove(&iface->port_elem);
3427 hmap_remove(&br->iface_by_name, &iface->name_node);
3428
3429 netdev_close(iface->netdev);
3430
3431 free(iface->name);
3432 free(iface);
3433 }
3434 }
3435
3436 static void
3437 iface_destroy(struct iface *iface)
3438 {
3439 if (iface) {
3440 struct port *port = iface->port;
3441
3442 iface_destroy__(iface);
3443 if (list_is_empty(&port->ifaces)) {
3444 port_destroy(port);
3445 }
3446 }
3447 }
3448
3449 static struct iface *
3450 iface_lookup(const struct bridge *br, const char *name)
3451 {
3452 struct iface *iface;
3453
3454 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3455 &br->iface_by_name) {
3456 if (!strcmp(iface->name, name)) {
3457 return iface;
3458 }
3459 }
3460
3461 return NULL;
3462 }
3463
3464 static struct iface *
3465 iface_find(const char *name)
3466 {
3467 const struct bridge *br;
3468
3469 HMAP_FOR_EACH (br, node, &all_bridges) {
3470 struct iface *iface = iface_lookup(br, name);
3471
3472 if (iface) {
3473 return iface;
3474 }
3475 }
3476 return NULL;
3477 }
3478
3479 static struct iface *
3480 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3481 {
3482 struct iface *iface;
3483
3484 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3485 &br->ifaces) {
3486 if (iface->ofp_port == ofp_port) {
3487 return iface;
3488 }
3489 }
3490 return NULL;
3491 }
3492
3493 /* Set Ethernet address of 'iface', if one is specified in the configuration
3494 * file. */
3495 static void
3496 iface_set_mac(struct iface *iface)
3497 {
3498 uint8_t ea[ETH_ADDR_LEN];
3499
3500 if (!strcmp(iface->type, "internal")
3501 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3502 if (iface->ofp_port == OFPP_LOCAL) {
3503 VLOG_ERR("interface %s: ignoring mac in Interface record "
3504 "(use Bridge record to set local port's mac)",
3505 iface->name);
3506 } else if (eth_addr_is_multicast(ea)) {
3507 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3508 iface->name);
3509 } else {
3510 int error = netdev_set_etheraddr(iface->netdev, ea);
3511 if (error) {
3512 VLOG_ERR("interface %s: setting MAC failed (%s)",
3513 iface->name, ovs_strerror(error));
3514 }
3515 }
3516 }
3517 }
3518
3519 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3520 static void
3521 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3522 {
3523 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3524 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3525 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3526 }
3527 }
3528
3529 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3530 * sets the "ofport" field to -1.
3531 *
3532 * This is appropriate when 'if_cfg''s interface cannot be created or is
3533 * otherwise invalid. */
3534 static void
3535 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3536 {
3537 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3538 ovsrec_interface_set_status(if_cfg, NULL);
3539 ovsrec_interface_set_admin_state(if_cfg, NULL);
3540 ovsrec_interface_set_duplex(if_cfg, NULL);
3541 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3542 ovsrec_interface_set_link_state(if_cfg, NULL);
3543 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3544 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3545 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3546 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3547 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3548 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3549 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3550 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3551 }
3552 }
3553
3554 static bool
3555 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3556 {
3557 union ovsdb_atom atom;
3558
3559 atom.integer = target;
3560 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3561 }
3562
3563 static void
3564 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3565 {
3566 struct ofpbuf queues_buf;
3567
3568 ofpbuf_init(&queues_buf, 0);
3569
3570 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3571 netdev_set_qos(iface->netdev, NULL, NULL);
3572 } else {
3573 const struct ovsdb_datum *queues;
3574 struct netdev_queue_dump dump;
3575 unsigned int queue_id;
3576 struct smap details;
3577 bool queue_zero;
3578 size_t i;
3579
3580 /* Configure top-level Qos for 'iface'. */
3581 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3582
3583 /* Deconfigure queues that were deleted. */
3584 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3585 OVSDB_TYPE_UUID);
3586 smap_init(&details);
3587 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3588 if (!queue_ids_include(queues, queue_id)) {
3589 netdev_delete_queue(iface->netdev, queue_id);
3590 }
3591 }
3592 smap_destroy(&details);
3593
3594 /* Configure queues for 'iface'. */
3595 queue_zero = false;
3596 for (i = 0; i < qos->n_queues; i++) {
3597 const struct ovsrec_queue *queue = qos->value_queues[i];
3598 unsigned int queue_id = qos->key_queues[i];
3599
3600 if (queue_id == 0) {
3601 queue_zero = true;
3602 }
3603
3604 if (queue->n_dscp == 1) {
3605 struct ofproto_port_queue *port_queue;
3606
3607 port_queue = ofpbuf_put_uninit(&queues_buf,
3608 sizeof *port_queue);
3609 port_queue->queue = queue_id;
3610 port_queue->dscp = queue->dscp[0];
3611 }
3612
3613 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3614 }
3615 if (!queue_zero) {
3616 struct smap details;
3617
3618 smap_init(&details);
3619 netdev_set_queue(iface->netdev, 0, &details);
3620 smap_destroy(&details);
3621 }
3622 }
3623
3624 if (iface->ofp_port != OFPP_NONE) {
3625 const struct ofproto_port_queue *port_queues = queues_buf.data;
3626 size_t n_queues = queues_buf.size / sizeof *port_queues;
3627
3628 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3629 port_queues, n_queues);
3630 }
3631
3632 netdev_set_policing(iface->netdev,
3633 iface->cfg->ingress_policing_rate,
3634 iface->cfg->ingress_policing_burst);
3635
3636 ofpbuf_uninit(&queues_buf);
3637 }
3638
3639 static void
3640 iface_configure_cfm(struct iface *iface)
3641 {
3642 const struct ovsrec_interface *cfg = iface->cfg;
3643 const char *opstate_str;
3644 const char *cfm_ccm_vlan;
3645 struct cfm_settings s;
3646 struct smap netdev_args;
3647
3648 if (!cfg->n_cfm_mpid) {
3649 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3650 return;
3651 }
3652
3653 s.check_tnl_key = false;
3654 smap_init(&netdev_args);
3655 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3656 const char *key = smap_get(&netdev_args, "key");
3657 const char *in_key = smap_get(&netdev_args, "in_key");
3658
3659 s.check_tnl_key = (key && !strcmp(key, "flow"))
3660 || (in_key && !strcmp(in_key, "flow"));
3661 }
3662 smap_destroy(&netdev_args);
3663
3664 s.mpid = *cfg->cfm_mpid;
3665 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3666 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3667 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3668
3669 if (s.interval <= 0) {
3670 s.interval = 1000;
3671 }
3672
3673 if (!cfm_ccm_vlan) {
3674 s.ccm_vlan = 0;
3675 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3676 s.ccm_vlan = CFM_RANDOM_VLAN;
3677 } else {
3678 s.ccm_vlan = atoi(cfm_ccm_vlan);
3679 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3680 s.ccm_vlan = 0;
3681 }
3682 }
3683
3684 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3685 false);
3686 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3687
3688 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3689 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3690
3691 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3692 }
3693
3694 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3695 * instead of obtaining it from the database. */
3696 static bool
3697 iface_is_synthetic(const struct iface *iface)
3698 {
3699 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3700 }
3701
3702 static ofp_port_t
3703 iface_validate_ofport__(size_t n, int64_t *ofport)
3704 {
3705 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3706 ? u16_to_ofp(*ofport)
3707 : OFPP_NONE);
3708 }
3709
3710 static ofp_port_t
3711 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3712 {
3713 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3714 }
3715
3716 static ofp_port_t
3717 iface_pick_ofport(const struct ovsrec_interface *cfg)
3718 {
3719 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3720 return (requested_ofport != OFPP_NONE
3721 ? requested_ofport
3722 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3723 }
3724 \f
3725 /* Port mirroring. */
3726
3727 static struct mirror *
3728 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3729 {
3730 struct mirror *m;
3731
3732 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3733 if (uuid_equals(uuid, &m->uuid)) {
3734 return m;
3735 }
3736 }
3737 return NULL;
3738 }
3739
3740 static void
3741 bridge_configure_mirrors(struct bridge *br)
3742 {
3743 const struct ovsdb_datum *mc;
3744 unsigned long *flood_vlans;
3745 struct mirror *m, *next;
3746 size_t i;
3747
3748 /* Get rid of deleted mirrors. */
3749 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3750 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3751 union ovsdb_atom atom;
3752
3753 atom.uuid = m->uuid;
3754 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3755 mirror_destroy(m);
3756 }
3757 }
3758
3759 /* Add new mirrors and reconfigure existing ones. */
3760 for (i = 0; i < br->cfg->n_mirrors; i++) {
3761 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3762 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3763 if (!m) {
3764 m = mirror_create(br, cfg);
3765 }
3766 m->cfg = cfg;
3767 if (!mirror_configure(m)) {
3768 mirror_destroy(m);
3769 }
3770 }
3771
3772 /* Update flooded vlans (for RSPAN). */
3773 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3774 br->cfg->n_flood_vlans);
3775 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3776 bitmap_free(flood_vlans);
3777 }
3778
3779 static struct mirror *
3780 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3781 {
3782 struct mirror *m;
3783
3784 m = xzalloc(sizeof *m);
3785 m->uuid = cfg->header_.uuid;
3786 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3787 m->bridge = br;
3788 m->name = xstrdup(cfg->name);
3789
3790 return m;
3791 }
3792
3793 static void
3794 mirror_destroy(struct mirror *m)
3795 {
3796 if (m) {
3797 struct bridge *br = m->bridge;
3798
3799 if (br->ofproto) {
3800 ofproto_mirror_unregister(br->ofproto, m);
3801 }
3802
3803 hmap_remove(&br->mirrors, &m->hmap_node);
3804 free(m->name);
3805 free(m);
3806 }
3807 }
3808
3809 static void
3810 mirror_collect_ports(struct mirror *m,
3811 struct ovsrec_port **in_ports, int n_in_ports,
3812 void ***out_portsp, size_t *n_out_portsp)
3813 {
3814 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3815 size_t n_out_ports = 0;
3816 size_t i;
3817
3818 for (i = 0; i < n_in_ports; i++) {
3819 const char *name = in_ports[i]->name;
3820 struct port *port = port_lookup(m->bridge, name);
3821 if (port) {
3822 out_ports[n_out_ports++] = port;
3823 } else {
3824 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3825 "port %s", m->bridge->name, m->name, name);
3826 }
3827 }
3828 *out_portsp = out_ports;
3829 *n_out_portsp = n_out_ports;
3830 }
3831
3832 static bool
3833 mirror_configure(struct mirror *m)
3834 {
3835 const struct ovsrec_mirror *cfg = m->cfg;
3836 struct ofproto_mirror_settings s;
3837
3838 /* Set name. */
3839 if (strcmp(cfg->name, m->name)) {
3840 free(m->name);
3841 m->name = xstrdup(cfg->name);
3842 }
3843 s.name = m->name;
3844
3845 /* Get output port or VLAN. */
3846 if (cfg->output_port) {
3847 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3848 if (!s.out_bundle) {
3849 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3850 m->bridge->name, m->name);
3851 return false;
3852 }
3853 s.out_vlan = UINT16_MAX;
3854
3855 if (cfg->output_vlan) {
3856 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3857 "output vlan; ignoring output vlan",
3858 m->bridge->name, m->name);
3859 }
3860 } else if (cfg->output_vlan) {
3861 /* The database should prevent invalid VLAN values. */
3862 s.out_bundle = NULL;
3863 s.out_vlan = *cfg->output_vlan;
3864 } else {
3865 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3866 m->bridge->name, m->name);
3867 return false;
3868 }
3869
3870 /* Get port selection. */
3871 if (cfg->select_all) {
3872 size_t n_ports = hmap_count(&m->bridge->ports);
3873 void **ports = xmalloc(n_ports * sizeof *ports);
3874 struct port *port;
3875 size_t i;
3876
3877 i = 0;
3878 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3879 ports[i++] = port;
3880 }
3881
3882 s.srcs = ports;
3883 s.n_srcs = n_ports;
3884
3885 s.dsts = ports;
3886 s.n_dsts = n_ports;
3887 } else {
3888 /* Get ports, dropping ports that don't exist.
3889 * The IDL ensures that there are no duplicates. */
3890 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3891 &s.srcs, &s.n_srcs);
3892 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3893 &s.dsts, &s.n_dsts);
3894 }
3895
3896 /* Get VLAN selection. */
3897 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3898
3899 /* Configure. */
3900 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3901
3902 /* Clean up. */
3903 if (s.srcs != s.dsts) {
3904 free(s.dsts);
3905 }
3906 free(s.srcs);
3907 free(s.src_vlans);
3908
3909 return true;
3910 }
3911 \f
3912 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3913 *
3914 * This is deprecated. It is only for compatibility with broken device drivers
3915 * in old versions of Linux that do not properly support VLANs when VLAN
3916 * devices are not used. When broken device drivers are no longer in
3917 * widespread use, we will delete these interfaces. */
3918
3919 static struct ovsrec_port **recs;
3920 static size_t n_recs, allocated_recs;
3921
3922 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3923 * splinters are reconfigured. */
3924 static void
3925 register_rec(struct ovsrec_port *rec)
3926 {
3927 if (n_recs >= allocated_recs) {
3928 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3929 }
3930 recs[n_recs++] = rec;
3931 }
3932
3933 /* Frees all of the ports registered with register_reg(). */
3934 static void
3935 free_registered_recs(void)
3936 {
3937 size_t i;
3938
3939 for (i = 0; i < n_recs; i++) {
3940 struct ovsrec_port *port = recs[i];
3941 size_t j;
3942
3943 for (j = 0; j < port->n_interfaces; j++) {
3944 struct ovsrec_interface *iface = port->interfaces[j];
3945 free(iface->name);
3946 free(iface);
3947 }
3948
3949 smap_destroy(&port->other_config);
3950 free(port->interfaces);
3951 free(port->name);
3952 free(port->tag);
3953 free(port);
3954 }
3955 n_recs = 0;
3956 }
3957
3958 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3959 * otherwise. */
3960 static bool
3961 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3962 {
3963 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3964 false);
3965 }
3966
3967 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3968 * splinters.
3969 *
3970 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3971 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3972 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3973 * with free().
3974 *
3975 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3976 * use, returns NULL.
3977 *
3978 * Updates 'vlan_splinters_enabled_anywhere'. */
3979 static unsigned long int *
3980 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3981 {
3982 unsigned long int *splinter_vlans;
3983 struct sset splinter_ifaces;
3984 const char *real_dev_name;
3985 struct shash *real_devs;
3986 struct shash_node *node;
3987 struct bridge *br;
3988 size_t i;
3989
3990 /* Free space allocated for synthesized ports and interfaces, since we're
3991 * in the process of reconstructing all of them. */
3992 free_registered_recs();
3993
3994 splinter_vlans = bitmap_allocate(4096);
3995 sset_init(&splinter_ifaces);
3996 vlan_splinters_enabled_anywhere = false;
3997 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3998 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3999 size_t j;
4000
4001 for (j = 0; j < br_cfg->n_ports; j++) {
4002 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4003 int k;
4004
4005 for (k = 0; k < port_cfg->n_interfaces; k++) {
4006 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4007
4008 if (vlan_splinters_is_enabled(iface_cfg)) {
4009 vlan_splinters_enabled_anywhere = true;
4010 sset_add(&splinter_ifaces, iface_cfg->name);
4011 vlan_bitmap_from_array__(port_cfg->trunks,
4012 port_cfg->n_trunks,
4013 splinter_vlans);
4014 }
4015 }
4016
4017 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4018 bitmap_set1(splinter_vlans, *port_cfg->tag);
4019 }
4020 }
4021 }
4022
4023 if (!vlan_splinters_enabled_anywhere) {
4024 free(splinter_vlans);
4025 sset_destroy(&splinter_ifaces);
4026 return NULL;
4027 }
4028
4029 HMAP_FOR_EACH (br, node, &all_bridges) {
4030 if (br->ofproto) {
4031 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4032 }
4033 }
4034
4035 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4036 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4037 * device to be created for it. */
4038 bitmap_set0(splinter_vlans, 0);
4039 bitmap_set0(splinter_vlans, 4095);
4040
4041 /* Delete all VLAN devices that we don't need. */
4042 vlandev_refresh();
4043 real_devs = vlandev_get_real_devs();
4044 SHASH_FOR_EACH (node, real_devs) {
4045 const struct vlan_real_dev *real_dev = node->data;
4046 const struct vlan_dev *vlan_dev;
4047 bool real_dev_has_splinters;
4048
4049 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4050 real_dev->name);
4051 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4052 if (!real_dev_has_splinters
4053 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4054 struct netdev *netdev;
4055
4056 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4057 if (!netdev_get_in4(netdev, NULL, NULL) ||
4058 !netdev_get_in6(netdev, NULL)) {
4059 /* It has an IP address configured, so we don't own
4060 * it. Don't delete it. */
4061 } else {
4062 vlandev_del(vlan_dev->name);
4063 }
4064 netdev_close(netdev);
4065 }
4066 }
4067
4068 }
4069 }
4070
4071 /* Add all VLAN devices that we need. */
4072 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4073 int vid;
4074
4075 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4076 if (!vlandev_get_name(real_dev_name, vid)) {
4077 vlandev_add(real_dev_name, vid);
4078 }
4079 }
4080 }
4081
4082 vlandev_refresh();
4083
4084 sset_destroy(&splinter_ifaces);
4085
4086 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4087 free(splinter_vlans);
4088 return NULL;
4089 }
4090 return splinter_vlans;
4091 }
4092
4093 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4094 * ofproto. */
4095 static void
4096 configure_splinter_port(struct port *port)
4097 {
4098 struct ofproto *ofproto = port->bridge->ofproto;
4099 ofp_port_t realdev_ofp_port;
4100 const char *realdev_name;
4101 struct iface *vlandev, *realdev;
4102
4103 ofproto_bundle_unregister(port->bridge->ofproto, port);
4104
4105 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4106 port_elem);
4107
4108 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4109 realdev = iface_lookup(port->bridge, realdev_name);
4110 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4111
4112 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4113 *port->cfg->tag);
4114 }
4115
4116 static struct ovsrec_port *
4117 synthesize_splinter_port(const char *real_dev_name,
4118 const char *vlan_dev_name, int vid)
4119 {
4120 struct ovsrec_interface *iface;
4121 struct ovsrec_port *port;
4122
4123 iface = xmalloc(sizeof *iface);
4124 ovsrec_interface_init(iface);
4125 iface->name = xstrdup(vlan_dev_name);
4126 iface->type = "system";
4127
4128 port = xmalloc(sizeof *port);
4129 ovsrec_port_init(port);
4130 port->interfaces = xmemdup(&iface, sizeof iface);
4131 port->n_interfaces = 1;
4132 port->name = xstrdup(vlan_dev_name);
4133 port->vlan_mode = "splinter";
4134 port->tag = xmalloc(sizeof *port->tag);
4135 *port->tag = vid;
4136
4137 smap_add(&port->other_config, "realdev", real_dev_name);
4138
4139 register_rec(port);
4140 return port;
4141 }
4142
4143 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4144 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4145 * 1-bit in the 'splinter_vlans' bitmap. */
4146 static void
4147 add_vlan_splinter_ports(struct bridge *br,
4148 const unsigned long int *splinter_vlans,
4149 struct shash *ports)
4150 {
4151 size_t i;
4152
4153 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4154 * we're modifying 'ports'. */
4155 for (i = 0; i < br->cfg->n_ports; i++) {
4156 const char *name = br->cfg->ports[i]->name;
4157 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4158 size_t j;
4159
4160 for (j = 0; j < port_cfg->n_interfaces; j++) {
4161 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4162
4163 if (vlan_splinters_is_enabled(iface_cfg)) {
4164 const char *real_dev_name;
4165 uint16_t vid;
4166
4167 real_dev_name = iface_cfg->name;
4168 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4169 const char *vlan_dev_name;
4170
4171 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4172 if (vlan_dev_name
4173 && !shash_find(ports, vlan_dev_name)) {
4174 shash_add(ports, vlan_dev_name,
4175 synthesize_splinter_port(
4176 real_dev_name, vlan_dev_name, vid));
4177 }
4178 }
4179 }
4180 }
4181 }
4182 }
4183
4184 static void
4185 mirror_refresh_stats(struct mirror *m)
4186 {
4187 struct ofproto *ofproto = m->bridge->ofproto;
4188 uint64_t tx_packets, tx_bytes;
4189 char *keys[2];
4190 int64_t values[2];
4191 size_t stat_cnt = 0;
4192
4193 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4194 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4195 return;
4196 }
4197
4198 if (tx_packets != UINT64_MAX) {
4199 keys[stat_cnt] = "tx_packets";
4200 values[stat_cnt] = tx_packets;
4201 stat_cnt++;
4202 }
4203 if (tx_bytes != UINT64_MAX) {
4204 keys[stat_cnt] = "tx_bytes";
4205 values[stat_cnt] = tx_bytes;
4206 stat_cnt++;
4207 }
4208
4209 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
4210 }