]> git.proxmox.com Git - mirror_ovs.git/blob - vswitchd/bridge.c
bridge: Fix controller status update to passive connections
[mirror_ovs.git] / vswitchd / bridge.c
1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 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
22 #include "async-append.h"
23 #include "bfd.h"
24 #include "bitmap.h"
25 #include "cfm.h"
26 #include "connectivity.h"
27 #include "coverage.h"
28 #include "daemon.h"
29 #include "dirs.h"
30 #include "dpif.h"
31 #include "dpdk.h"
32 #include "hash.h"
33 #include "openvswitch/hmap.h"
34 #include "hmapx.h"
35 #include "if-notifier.h"
36 #include "jsonrpc.h"
37 #include "lacp.h"
38 #include "mac-learning.h"
39 #include "mcast-snooping.h"
40 #include "netdev.h"
41 #include "nx-match.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "openvswitch/dynamic-string.h"
45 #include "openvswitch/list.h"
46 #include "openvswitch/meta-flow.h"
47 #include "openvswitch/ofp-print.h"
48 #include "openvswitch/ofp-util.h"
49 #include "openvswitch/ofpbuf.h"
50 #include "openvswitch/vlog.h"
51 #include "ovs-lldp.h"
52 #include "ovs-numa.h"
53 #include "packets.h"
54 #include "poll-loop.h"
55 #include "seq.h"
56 #include "sflow_api.h"
57 #include "sha1.h"
58 #include "openvswitch/shash.h"
59 #include "smap.h"
60 #include "socket-util.h"
61 #include "stream.h"
62 #include "stream-ssl.h"
63 #include "sset.h"
64 #include "system-stats.h"
65 #include "timeval.h"
66 #include "util.h"
67 #include "unixctl.h"
68 #include "lib/vswitch-idl.h"
69 #include "xenserver.h"
70 #include "vlan-bitmap.h"
71
72 VLOG_DEFINE_THIS_MODULE(bridge);
73
74 COVERAGE_DEFINE(bridge_reconfigure);
75
76 struct iface {
77 /* These members are always valid.
78 *
79 * They are immutable: they never change between iface_create() and
80 * iface_destroy(). */
81 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
86 struct netdev *netdev; /* Network device. */
87 ofp_port_t ofp_port; /* OpenFlow port number. */
88 uint64_t change_seq;
89
90 /* These members are valid only within bridge_reconfigure(). */
91 const char *type; /* Usually same as cfg->type. */
92 const char *netdev_type; /* type that should be used for netdev_open. */
93 const struct ovsrec_interface *cfg;
94 };
95
96 struct mirror {
97 struct uuid uuid; /* UUID of this "mirror" record in database. */
98 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
99 struct bridge *bridge;
100 char *name;
101 const struct ovsrec_mirror *cfg;
102 };
103
104 struct port {
105 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
106 struct bridge *bridge;
107 char *name;
108
109 const struct ovsrec_port *cfg;
110
111 /* An ordinary bridge port has 1 interface.
112 * A bridge port for bonding has at least 2 interfaces. */
113 struct ovs_list ifaces; /* List of "struct iface"s. */
114 };
115
116 struct bridge {
117 struct hmap_node node; /* In 'all_bridges'. */
118 char *name; /* User-specified arbitrary name. */
119 char *type; /* Datapath type. */
120 struct eth_addr ea; /* Bridge Ethernet Address. */
121 struct eth_addr default_ea; /* Default MAC. */
122 const struct ovsrec_bridge *cfg;
123
124 /* OpenFlow switch processing. */
125 struct ofproto *ofproto; /* OpenFlow switch. */
126
127 /* Bridge ports. */
128 struct hmap ports; /* "struct port"s indexed by name. */
129 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
130 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
131
132 /* Port mirroring. */
133 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
134
135 /* Auto Attach */
136 struct hmap mappings; /* "struct" indexed by UUID */
137
138 /* Used during reconfiguration. */
139 struct shash wanted_ports;
140
141 /* Synthetic local port if necessary. */
142 struct ovsrec_port synth_local_port;
143 struct ovsrec_interface synth_local_iface;
144 struct ovsrec_interface *synth_local_ifacep;
145 };
146
147 struct aa_mapping {
148 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
149 struct bridge *bridge;
150 uint32_t isid;
151 uint16_t vlan;
152 char *br_name;
153 };
154
155 /* All bridges, indexed by name. */
156 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
157
158 /* OVSDB IDL used to obtain configuration. */
159 static struct ovsdb_idl *idl;
160
161 /* We want to complete daemonization, fully detaching from our parent process,
162 * only after we have completed our initial configuration, committed our state
163 * to the database, and received confirmation back from the database server
164 * that it applied the commit. This allows our parent process to know that,
165 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
166 * to have already been filled in. (It is only "very likely" rather than
167 * certain because there is always a slim possibility that the transaction will
168 * fail or that some other client has added new bridges, ports, etc. while
169 * ovs-vswitchd was configuring using an old configuration.)
170 *
171 * We only need to do this once for our initial configuration at startup, so
172 * 'initial_config_done' tracks whether we've already done it. While we are
173 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
174 * itself and is otherwise NULL. */
175 static bool initial_config_done;
176 static struct ovsdb_idl_txn *daemonize_txn;
177
178 /* Most recently processed IDL sequence number. */
179 static unsigned int idl_seqno;
180
181 /* Track changes to port connectivity. */
182 static uint64_t connectivity_seqno = LLONG_MIN;
183
184 /* Status update to database.
185 *
186 * Some information in the database must be kept as up-to-date as possible to
187 * allow controllers to respond rapidly to network outages. Those status are
188 * updated via the 'status_txn'.
189 *
190 * We use the global connectivity sequence number to detect the status change.
191 * Also, to prevent the status update from sending too much to the database,
192 * we check the return status of each update transaction and do not start new
193 * update if the previous transaction status is 'TXN_INCOMPLETE'.
194 *
195 * 'statux_txn' is NULL if there is no ongoing status update.
196 *
197 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
198 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
199 * which will cause the main thread wake up soon and retry the status update.
200 */
201 static struct ovsdb_idl_txn *status_txn;
202 static bool status_txn_try_again;
203
204 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
205 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
206 #define STATUS_CHECK_AGAIN_MSEC 100
207
208 /* Statistics update to database. */
209 static struct ovsdb_idl_txn *stats_txn;
210
211 /* Each time this timer expires, the bridge fetches interface and mirror
212 * statistics and pushes them into the database. */
213 static int stats_timer_interval;
214 static long long int stats_timer = LLONG_MIN;
215
216 /* Each time this timer expires, the bridge fetches the list of port/VLAN
217 * membership that has been modified by the AA.
218 */
219 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
220 static long long int aa_refresh_timer = LLONG_MIN;
221
222 /* Whenever system interfaces are added, removed or change state, the bridge
223 * will be reconfigured.
224 */
225 static struct if_notifier *ifnotifier;
226 static struct seq *ifaces_changed;
227 static uint64_t last_ifaces_changed;
228
229 static void add_del_bridges(const struct ovsrec_open_vswitch *);
230 static void bridge_run__(void);
231 static void bridge_create(const struct ovsrec_bridge *);
232 static void bridge_destroy(struct bridge *, bool del);
233 static struct bridge *bridge_lookup(const char *name);
234 static unixctl_cb_func bridge_unixctl_dump_flows;
235 static unixctl_cb_func bridge_unixctl_reconnect;
236 static size_t bridge_get_controllers(const struct bridge *br,
237 struct ovsrec_controller ***controllersp);
238 static void bridge_collect_wanted_ports(struct bridge *,
239 struct shash *wanted_ports);
240 static void bridge_delete_ofprotos(void);
241 static void bridge_delete_or_reconfigure_ports(struct bridge *);
242 static void bridge_del_ports(struct bridge *,
243 const struct shash *wanted_ports);
244 static void bridge_add_ports(struct bridge *,
245 const struct shash *wanted_ports);
246
247 static void bridge_configure_datapath_id(struct bridge *);
248 static void bridge_configure_netflow(struct bridge *);
249 static void bridge_configure_forward_bpdu(struct bridge *);
250 static void bridge_configure_mac_table(struct bridge *);
251 static void bridge_configure_mcast_snooping(struct bridge *);
252 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
253 static void bridge_configure_ipfix(struct bridge *);
254 static void bridge_configure_spanning_tree(struct bridge *);
255 static void bridge_configure_tables(struct bridge *);
256 static void bridge_configure_dp_desc(struct bridge *);
257 static void bridge_configure_aa(struct bridge *);
258 static void bridge_aa_refresh_queued(struct bridge *);
259 static bool bridge_aa_need_refresh(struct bridge *);
260 static void bridge_configure_remotes(struct bridge *,
261 const struct sockaddr_in *managers,
262 size_t n_managers);
263 static void bridge_pick_local_hw_addr(struct bridge *, struct eth_addr *ea,
264 struct iface **hw_addr_iface);
265 static uint64_t bridge_pick_datapath_id(struct bridge *,
266 const struct eth_addr bridge_ea,
267 struct iface *hw_addr_iface);
268 static uint64_t dpid_from_hash(const void *, size_t nbytes);
269 static bool bridge_has_bond_fake_iface(const struct bridge *,
270 const char *name);
271 static bool port_is_bond_fake_iface(const struct port *);
272
273 static unixctl_cb_func qos_unixctl_show_types;
274 static unixctl_cb_func qos_unixctl_show;
275
276 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
277 static void port_del_ifaces(struct port *);
278 static void port_destroy(struct port *);
279 static struct port *port_lookup(const struct bridge *, const char *name);
280 static void port_configure(struct port *);
281 static struct lacp_settings *port_configure_lacp(struct port *,
282 struct lacp_settings *);
283 static void port_configure_bond(struct port *, struct bond_settings *);
284 static bool port_is_synthetic(const struct port *);
285
286 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
287 static void run_system_stats(void);
288
289 static void bridge_configure_mirrors(struct bridge *);
290 static struct mirror *mirror_create(struct bridge *,
291 const struct ovsrec_mirror *);
292 static void mirror_destroy(struct mirror *);
293 static bool mirror_configure(struct mirror *);
294 static void mirror_refresh_stats(struct mirror *);
295
296 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
297 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
298 const struct ovsrec_port *);
299 static bool iface_is_internal(const struct ovsrec_interface *iface,
300 const struct ovsrec_bridge *br);
301 static const char *iface_get_type(const struct ovsrec_interface *,
302 const struct ovsrec_bridge *);
303 static void iface_destroy(struct iface *);
304 static void iface_destroy__(struct iface *);
305 static struct iface *iface_lookup(const struct bridge *, const char *name);
306 static struct iface *iface_find(const char *name);
307 static struct iface *iface_from_ofp_port(const struct bridge *,
308 ofp_port_t ofp_port);
309 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
310 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
311 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
312 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
313 static void iface_configure_cfm(struct iface *);
314 static void iface_refresh_cfm_stats(struct iface *);
315 static void iface_refresh_stats(struct iface *);
316 static void iface_refresh_netdev_status(struct iface *);
317 static void iface_refresh_ofproto_status(struct iface *);
318 static bool iface_is_synthetic(const struct iface *);
319 static ofp_port_t iface_get_requested_ofp_port(
320 const struct ovsrec_interface *);
321 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
322
323
324 static void discover_types(const struct ovsrec_open_vswitch *cfg);
325
326 static void
327 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
328 {
329 struct shash iface_hints;
330 static bool initialized = false;
331 int i;
332
333 if (initialized) {
334 return;
335 }
336
337 shash_init(&iface_hints);
338
339 if (cfg) {
340 for (i = 0; i < cfg->n_bridges; i++) {
341 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
342 int j;
343
344 for (j = 0; j < br_cfg->n_ports; j++) {
345 struct ovsrec_port *port_cfg = br_cfg->ports[j];
346 int k;
347
348 for (k = 0; k < port_cfg->n_interfaces; k++) {
349 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
350 struct iface_hint *iface_hint;
351
352 iface_hint = xmalloc(sizeof *iface_hint);
353 iface_hint->br_name = br_cfg->name;
354 iface_hint->br_type = br_cfg->datapath_type;
355 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
356
357 shash_add(&iface_hints, if_cfg->name, iface_hint);
358 }
359 }
360 }
361 }
362
363 ofproto_init(&iface_hints);
364
365 shash_destroy_free_data(&iface_hints);
366 initialized = true;
367 }
368
369 static void
370 if_change_cb(void *aux OVS_UNUSED)
371 {
372 seq_change(ifaces_changed);
373 }
374
375 static bool
376 if_notifier_changed(struct if_notifier *notifier OVS_UNUSED)
377 {
378 uint64_t new_seq;
379 bool changed = false;
380 new_seq = seq_read(ifaces_changed);
381 if (new_seq != last_ifaces_changed) {
382 changed = true;
383 last_ifaces_changed = new_seq;
384 }
385 seq_wait(ifaces_changed, last_ifaces_changed);
386 return changed;
387 }
388 \f
389 /* Public functions. */
390
391 /* Initializes the bridge module, configuring it to obtain its configuration
392 * from an OVSDB server accessed over 'remote', which should be a string in a
393 * form acceptable to ovsdb_idl_create(). */
394 void
395 bridge_init(const char *remote)
396 {
397 /* Create connection to database. */
398 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
399 idl_seqno = ovsdb_idl_get_seqno(idl);
400 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
401 ovsdb_idl_verify_write_only(idl);
402
403 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
404 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
405 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
406 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
407 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
408 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
409 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
410 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
411 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
412
413 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
414 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
415 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
416 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
417 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
418 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
419 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
420
421 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
422 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
423 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
424 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
425 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
426 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
427 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
428 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
429 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
430 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
440 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
441 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
442 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
443 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
444 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
445 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
446 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
447 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
448 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
449 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
450
451 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
452 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
453 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
454 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
455
456 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
457
458 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
459
460 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
461 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
462
463 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
464 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
465 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
466 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
467
468 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
469 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
470 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
471 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
472 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
473
474 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
475
476 /* Register unixctl commands. */
477 unixctl_command_register("qos/show-types", "interface", 1, 1,
478 qos_unixctl_show_types, NULL);
479 unixctl_command_register("qos/show", "interface", 1, 1,
480 qos_unixctl_show, NULL);
481 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
482 bridge_unixctl_dump_flows, NULL);
483 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
484 bridge_unixctl_reconnect, NULL);
485 lacp_init();
486 bond_init();
487 cfm_init();
488 bfd_init();
489 ovs_numa_init();
490 stp_init();
491 lldp_init();
492 rstp_init();
493 ifaces_changed = seq_create();
494 last_ifaces_changed = seq_read(ifaces_changed);
495 ifnotifier = if_notifier_create(if_change_cb, NULL);
496 }
497
498 void
499 bridge_exit(bool delete_datapath)
500 {
501 struct bridge *br, *next_br;
502
503 if_notifier_destroy(ifnotifier);
504 seq_destroy(ifaces_changed);
505 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
506 bridge_destroy(br, delete_datapath);
507 }
508 ovsdb_idl_destroy(idl);
509 }
510
511 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
512 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
513 * responsible for freeing '*managersp' (with free()).
514 *
515 * You may be asking yourself "why does ovs-vswitchd care?", because
516 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
517 * should not be and in fact is not directly involved in that. But
518 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
519 * it has to tell in-band control where the managers are to enable that.
520 * (Thus, only managers connected in-band and with non-loopback addresses
521 * are collected.)
522 */
523 static void
524 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
525 struct sockaddr_in **managersp, size_t *n_managersp)
526 {
527 struct sockaddr_in *managers = NULL;
528 size_t n_managers = 0;
529 struct sset targets;
530 size_t i;
531
532 /* Collect all of the potential targets from the "targets" columns of the
533 * rows pointed to by "manager_options", excluding any that are
534 * out-of-band. */
535 sset_init(&targets);
536 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
537 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
538
539 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
540 sset_find_and_delete(&targets, m->target);
541 } else {
542 sset_add(&targets, m->target);
543 }
544 }
545
546 /* Now extract the targets' IP addresses. */
547 if (!sset_is_empty(&targets)) {
548 const char *target;
549
550 managers = xmalloc(sset_count(&targets) * sizeof *managers);
551 SSET_FOR_EACH (target, &targets) {
552 union {
553 struct sockaddr_storage ss;
554 struct sockaddr_in in;
555 } sa;
556
557 /* Ignore loopback. */
558 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
559 &sa.ss)
560 && sa.ss.ss_family == AF_INET
561 && sa.in.sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
562 managers[n_managers++] = sa.in;
563 }
564 }
565 }
566 sset_destroy(&targets);
567
568 *managersp = managers;
569 *n_managersp = n_managers;
570 }
571
572 static void
573 config_ofproto_types(const struct smap *other_config)
574 {
575 struct sset types;
576 const char *type;
577
578 /* Pass custom configuration to datapath types. */
579 sset_init(&types);
580 ofproto_enumerate_types(&types);
581 SSET_FOR_EACH (type, &types) {
582 ofproto_type_set_config(type, other_config);
583 }
584 sset_destroy(&types);
585 }
586
587 static void
588 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
589 {
590 struct sockaddr_in *managers;
591 struct bridge *br, *next;
592 int sflow_bridge_number;
593 size_t n_managers;
594
595 COVERAGE_INC(bridge_reconfigure);
596
597 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
598 OFPROTO_FLOW_LIMIT_DEFAULT));
599 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
600 OFPROTO_MAX_IDLE_DEFAULT));
601 ofproto_set_vlan_limit(smap_get_int(&ovs_cfg->other_config, "vlan-limit",
602 LEGACY_MAX_VLAN_HEADERS));
603
604 ofproto_set_threads(
605 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
606 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
607
608 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
609 * to 'ovs_cfg', with only very minimal configuration otherwise.
610 *
611 * This is mostly an update to bridge data structures. Nothing is pushed
612 * down to ofproto or lower layers. */
613 add_del_bridges(ovs_cfg);
614 HMAP_FOR_EACH (br, node, &all_bridges) {
615 bridge_collect_wanted_ports(br, &br->wanted_ports);
616 bridge_del_ports(br, &br->wanted_ports);
617 }
618
619 /* Start pushing configuration changes down to the ofproto layer:
620 *
621 * - Delete ofprotos that are no longer configured.
622 *
623 * - Delete ports that are no longer configured.
624 *
625 * - Reconfigure existing ports to their desired configurations, or
626 * delete them if not possible.
627 *
628 * We have to do all the deletions before we can do any additions, because
629 * the ports to be added might require resources that will be freed up by
630 * deletions (they might especially overlap in name). */
631 bridge_delete_ofprotos();
632 HMAP_FOR_EACH (br, node, &all_bridges) {
633 if (br->ofproto) {
634 bridge_delete_or_reconfigure_ports(br);
635 }
636 }
637
638 /* Finish pushing configuration changes to the ofproto layer:
639 *
640 * - Create ofprotos that are missing.
641 *
642 * - Add ports that are missing. */
643 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
644 if (!br->ofproto) {
645 int error;
646
647 error = ofproto_create(br->name, br->type, &br->ofproto);
648 if (error) {
649 VLOG_ERR("failed to create bridge %s: %s", br->name,
650 ovs_strerror(error));
651 shash_destroy(&br->wanted_ports);
652 bridge_destroy(br, true);
653 } else {
654 /* Trigger storing datapath version. */
655 seq_change(connectivity_seq_get());
656 }
657 }
658 }
659
660 config_ofproto_types(&ovs_cfg->other_config);
661
662 HMAP_FOR_EACH (br, node, &all_bridges) {
663 bridge_add_ports(br, &br->wanted_ports);
664 shash_destroy(&br->wanted_ports);
665 }
666
667 reconfigure_system_stats(ovs_cfg);
668
669 /* Complete the configuration. */
670 sflow_bridge_number = 0;
671 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
672 HMAP_FOR_EACH (br, node, &all_bridges) {
673 struct port *port;
674
675 /* We need the datapath ID early to allow LACP ports to use it as the
676 * default system ID. */
677 bridge_configure_datapath_id(br);
678
679 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
680 struct iface *iface;
681
682 port_configure(port);
683
684 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
685 iface_set_ofport(iface->cfg, iface->ofp_port);
686 /* Clear eventual previous errors */
687 ovsrec_interface_set_error(iface->cfg, NULL);
688 iface_configure_cfm(iface);
689 iface_configure_qos(iface, port->cfg->qos);
690 iface_set_mac(br, port, iface);
691 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
692 &iface->cfg->bfd);
693 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
694 &iface->cfg->lldp);
695 ofproto_port_set_config(br->ofproto, iface->ofp_port,
696 &iface->cfg->other_config);
697 }
698 }
699 bridge_configure_mirrors(br);
700 bridge_configure_forward_bpdu(br);
701 bridge_configure_mac_table(br);
702 bridge_configure_mcast_snooping(br);
703 bridge_configure_remotes(br, managers, n_managers);
704 bridge_configure_netflow(br);
705 bridge_configure_sflow(br, &sflow_bridge_number);
706 bridge_configure_ipfix(br);
707 bridge_configure_spanning_tree(br);
708 bridge_configure_tables(br);
709 bridge_configure_dp_desc(br);
710 bridge_configure_aa(br);
711 }
712 free(managers);
713
714 /* The ofproto-dpif provider does some final reconfiguration in its
715 * ->type_run() function. We have to call it before notifying the database
716 * client that reconfiguration is complete, otherwise there is a very
717 * narrow race window in which e.g. ofproto/trace will not recognize the
718 * new configuration (sometimes this causes unit test failures). */
719 bridge_run__();
720 }
721
722 /* Delete ofprotos which aren't configured or have the wrong type. Create
723 * ofprotos which don't exist but need to. */
724 static void
725 bridge_delete_ofprotos(void)
726 {
727 struct bridge *br;
728 struct sset names;
729 struct sset types;
730 const char *type;
731
732 /* Delete ofprotos with no bridge or with the wrong type. */
733 sset_init(&names);
734 sset_init(&types);
735 ofproto_enumerate_types(&types);
736 SSET_FOR_EACH (type, &types) {
737 const char *name;
738
739 ofproto_enumerate_names(type, &names);
740 SSET_FOR_EACH (name, &names) {
741 br = bridge_lookup(name);
742 if (!br || strcmp(type, br->type)) {
743 ofproto_delete(name, type);
744 }
745 }
746 }
747 sset_destroy(&names);
748 sset_destroy(&types);
749 }
750
751 static ofp_port_t *
752 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
753 {
754 if (*n >= *allocated) {
755 ports = x2nrealloc(ports, allocated, sizeof *ports);
756 }
757 ports[(*n)++] = port;
758 return ports;
759 }
760
761 /* Configures the MTU of 'netdev' based on the "mtu_request" column
762 * in 'iface_cfg'. */
763 static int
764 iface_set_netdev_mtu(const struct ovsrec_interface *iface_cfg,
765 struct netdev *netdev)
766 {
767 if (iface_cfg->n_mtu_request == 1) {
768 /* The user explicitly asked for this MTU. */
769 netdev_mtu_user_config(netdev, true);
770 /* Try to set the MTU to the requested value. */
771 return netdev_set_mtu(netdev, *iface_cfg->mtu_request);
772 }
773
774 /* The user didn't explicitly asked for any MTU. */
775 netdev_mtu_user_config(netdev, false);
776 return 0;
777 }
778
779 static void
780 bridge_delete_or_reconfigure_ports(struct bridge *br)
781 {
782 struct ofproto_port ofproto_port;
783 struct ofproto_port_dump dump;
784
785 struct sset ofproto_ports;
786 struct port *port, *port_next;
787
788 /* List of "ofp_port"s to delete. We make a list instead of deleting them
789 * right away because ofproto implementations aren't necessarily able to
790 * iterate through a changing list of ports in an entirely robust way. */
791 ofp_port_t *del;
792 size_t n, allocated;
793 size_t i;
794
795 del = NULL;
796 n = allocated = 0;
797 sset_init(&ofproto_ports);
798
799 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
800 * that are not configured in the database. (This commonly happens when
801 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
802 *
803 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
804 * that have the wrong OpenFlow port number (and arrange to add them back
805 * with the correct OpenFlow port number). */
806 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
807 ofp_port_t requested_ofp_port;
808 struct iface *iface;
809
810 sset_add(&ofproto_ports, ofproto_port.name);
811
812 iface = iface_lookup(br, ofproto_port.name);
813 if (!iface) {
814 /* No such iface is configured, so we should delete this
815 * ofproto_port.
816 *
817 * As a corner case exception, keep the port if it's a bond fake
818 * interface. */
819 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
820 && !strcmp(ofproto_port.type, "internal")) {
821 continue;
822 }
823 goto delete;
824 }
825
826 if (strcmp(ofproto_port.type, iface->netdev_type)
827 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
828 /* The interface is the wrong type or can't be configured.
829 * Delete it. */
830 goto delete;
831 }
832
833 iface_set_netdev_mtu(iface->cfg, iface->netdev);
834
835 /* If the requested OpenFlow port for 'iface' changed, and it's not
836 * already the correct port, then we might want to temporarily delete
837 * this interface, so we can add it back again with the new OpenFlow
838 * port number. */
839 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
840 if (iface->ofp_port != OFPP_LOCAL &&
841 requested_ofp_port != OFPP_NONE &&
842 requested_ofp_port != iface->ofp_port) {
843 ofp_port_t victim_request;
844 struct iface *victim;
845
846 /* Check for an existing OpenFlow port currently occupying
847 * 'iface''s requested port number. If there isn't one, then
848 * delete this port. Otherwise we need to consider further. */
849 victim = iface_from_ofp_port(br, requested_ofp_port);
850 if (!victim) {
851 goto delete;
852 }
853
854 /* 'victim' is a port currently using 'iface''s requested port
855 * number. Unless 'victim' specifically requested that port
856 * number, too, then we can delete both 'iface' and 'victim'
857 * temporarily. (We'll add both of them back again later with new
858 * OpenFlow port numbers.)
859 *
860 * If 'victim' did request port number 'requested_ofp_port', just
861 * like 'iface', then that's a configuration inconsistency that we
862 * can't resolve. We might as well let it keep its current port
863 * number. */
864 victim_request = iface_get_requested_ofp_port(victim->cfg);
865 if (victim_request != requested_ofp_port) {
866 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
867 iface_destroy(victim);
868 goto delete;
869 }
870 }
871
872 /* Keep it. */
873 continue;
874
875 delete:
876 iface_destroy(iface);
877 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
878 }
879 for (i = 0; i < n; i++) {
880 ofproto_port_del(br->ofproto, del[i]);
881 }
882 free(del);
883
884 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
885 * that we didn't see when we iterated through the datapath, i.e. ports
886 * that disappeared underneath use. This is an unusual situation, but it
887 * can happen in some cases:
888 *
889 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
890 * idea but could happen).
891 *
892 * - The port represented a device that disappeared, e.g. a tuntap
893 * device destroyed via "tunctl -d", a physical Ethernet device
894 * whose module was just unloaded via "rmmod", or a virtual NIC for a
895 * VM whose VM was just terminated. */
896 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
897 struct iface *iface, *iface_next;
898
899 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
900 if (!sset_contains(&ofproto_ports, iface->name)) {
901 iface_destroy__(iface);
902 }
903 }
904
905 if (ovs_list_is_empty(&port->ifaces)) {
906 port_destroy(port);
907 }
908 }
909 sset_destroy(&ofproto_ports);
910 }
911
912 static void
913 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
914 bool with_requested_port)
915 {
916 struct shash_node *port_node;
917
918 SHASH_FOR_EACH (port_node, wanted_ports) {
919 const struct ovsrec_port *port_cfg = port_node->data;
920 size_t i;
921
922 for (i = 0; i < port_cfg->n_interfaces; i++) {
923 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
924 ofp_port_t requested_ofp_port;
925
926 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
927 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
928 struct iface *iface = iface_lookup(br, iface_cfg->name);
929
930 if (!iface) {
931 iface_create(br, iface_cfg, port_cfg);
932 }
933 }
934 }
935 }
936 }
937
938 static void
939 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
940 {
941 /* First add interfaces that request a particular port number. */
942 bridge_add_ports__(br, wanted_ports, true);
943
944 /* Then add interfaces that want automatic port number assignment.
945 * We add these afterward to avoid accidentally taking a specifically
946 * requested port number. */
947 bridge_add_ports__(br, wanted_ports, false);
948 }
949
950 static void
951 port_configure(struct port *port)
952 {
953 const struct ovsrec_port *cfg = port->cfg;
954 struct bond_settings bond_settings;
955 struct lacp_settings lacp_settings;
956 struct ofproto_bundle_settings s;
957 struct iface *iface;
958
959 /* Get name. */
960 s.name = port->name;
961
962 /* Get slaves. */
963 s.n_slaves = 0;
964 s.slaves = xmalloc(ovs_list_size(&port->ifaces) * sizeof *s.slaves);
965 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
966 s.slaves[s.n_slaves++] = iface->ofp_port;
967 }
968
969 /* Get VLAN tag. */
970 s.vlan = -1;
971 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
972 s.vlan = *cfg->tag;
973 }
974
975 /* Get VLAN trunks. */
976 s.trunks = NULL;
977 if (cfg->n_trunks) {
978 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
979 }
980
981 s.cvlans = NULL;
982 if (cfg->n_cvlans) {
983 s.cvlans = vlan_bitmap_from_array(cfg->cvlans, cfg->n_cvlans);
984 }
985
986 /* Get VLAN mode. */
987 if (cfg->vlan_mode) {
988 if (!strcmp(cfg->vlan_mode, "access")) {
989 s.vlan_mode = PORT_VLAN_ACCESS;
990 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
991 s.vlan_mode = PORT_VLAN_TRUNK;
992 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
993 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
994 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
995 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
996 } else if (!strcmp(cfg->vlan_mode, "dot1q-tunnel")) {
997 s.vlan_mode = PORT_VLAN_DOT1Q_TUNNEL;
998 } else {
999 /* This "can't happen" because ovsdb-server should prevent it. */
1000 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
1001 "back to trunk mode", port->name, cfg->vlan_mode);
1002 s.vlan_mode = PORT_VLAN_TRUNK;
1003 }
1004 } else {
1005 if (s.vlan >= 0) {
1006 s.vlan_mode = PORT_VLAN_ACCESS;
1007 if (cfg->n_trunks || cfg->n_cvlans) {
1008 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
1009 port->name);
1010 }
1011 } else {
1012 s.vlan_mode = PORT_VLAN_TRUNK;
1013 }
1014 }
1015
1016 const char *qe = smap_get_def(&cfg->other_config, "qinq-ethtype", "");
1017 s.qinq_ethtype = (!strcmp(qe, "802.1q")
1018 ? ETH_TYPE_VLAN_8021Q
1019 : ETH_TYPE_VLAN_8021AD);
1020
1021 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
1022 false);
1023
1024 /* Get LACP settings. */
1025 s.lacp = port_configure_lacp(port, &lacp_settings);
1026 if (s.lacp) {
1027 size_t i = 0;
1028
1029 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
1030 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1031 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
1032 }
1033 } else {
1034 s.lacp_slaves = NULL;
1035 }
1036
1037 /* Get bond settings. */
1038 if (s.n_slaves > 1) {
1039 s.bond = &bond_settings;
1040 port_configure_bond(port, &bond_settings);
1041 } else {
1042 s.bond = NULL;
1043 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1044 netdev_set_miimon_interval(iface->netdev, 0);
1045 }
1046 }
1047
1048 /* Protected port mode */
1049 s.protected = cfg->protected;
1050
1051 /* Register. */
1052 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1053
1054 /* Clean up. */
1055 free(s.cvlans);
1056 free(s.slaves);
1057 free(s.trunks);
1058 free(s.lacp_slaves);
1059 }
1060
1061 /* Pick local port hardware address and datapath ID for 'br'. */
1062 static void
1063 bridge_configure_datapath_id(struct bridge *br)
1064 {
1065 struct eth_addr ea;
1066 uint64_t dpid;
1067 struct iface *local_iface;
1068 struct iface *hw_addr_iface;
1069 char *dpid_string;
1070
1071 bridge_pick_local_hw_addr(br, &ea, &hw_addr_iface);
1072 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1073 if (local_iface) {
1074 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1075 if (error) {
1076 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1077 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1078 "Ethernet address: %s",
1079 br->name, ovs_strerror(error));
1080 }
1081 }
1082 br->ea = ea;
1083
1084 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1085 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1086 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1087 ofproto_set_datapath_id(br->ofproto, dpid);
1088 }
1089
1090 dpid_string = xasprintf("%016"PRIx64, dpid);
1091 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1092 free(dpid_string);
1093 }
1094
1095 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1096 * 'br'. */
1097 static uint32_t
1098 bridge_get_allowed_versions(struct bridge *br)
1099 {
1100 if (!br->cfg->n_protocols) {
1101 return 0;
1102 }
1103
1104 return ofputil_versions_from_strings(br->cfg->protocols,
1105 br->cfg->n_protocols);
1106 }
1107
1108 /* Set NetFlow configuration on 'br'. */
1109 static void
1110 bridge_configure_netflow(struct bridge *br)
1111 {
1112 struct ovsrec_netflow *cfg = br->cfg->netflow;
1113 struct netflow_options opts;
1114
1115 if (!cfg) {
1116 ofproto_set_netflow(br->ofproto, NULL);
1117 return;
1118 }
1119
1120 memset(&opts, 0, sizeof opts);
1121
1122 /* Get default NetFlow configuration from datapath.
1123 * Apply overrides from 'cfg'. */
1124 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1125 if (cfg->engine_type) {
1126 opts.engine_type = *cfg->engine_type;
1127 }
1128 if (cfg->engine_id) {
1129 opts.engine_id = *cfg->engine_id;
1130 }
1131
1132 /* Configure active timeout interval. */
1133 opts.active_timeout = cfg->active_timeout;
1134 if (!opts.active_timeout) {
1135 opts.active_timeout = -1;
1136 } else if (opts.active_timeout < 0) {
1137 VLOG_WARN("bridge %s: active timeout interval set to negative "
1138 "value, using default instead (%d seconds)", br->name,
1139 NF_ACTIVE_TIMEOUT_DEFAULT);
1140 opts.active_timeout = -1;
1141 }
1142
1143 /* Add engine ID to interface number to disambiguate bridgs? */
1144 opts.add_id_to_iface = cfg->add_id_to_interface;
1145 if (opts.add_id_to_iface) {
1146 if (opts.engine_id > 0x7f) {
1147 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1148 "another vswitch, choose an engine id less than 128",
1149 br->name);
1150 }
1151 if (hmap_count(&br->ports) > 508) {
1152 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1153 "another port when more than 508 ports are used",
1154 br->name);
1155 }
1156 }
1157
1158 /* Collectors. */
1159 sset_init(&opts.collectors);
1160 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1161
1162 /* Configure. */
1163 if (ofproto_set_netflow(br->ofproto, &opts)) {
1164 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1165 }
1166 sset_destroy(&opts.collectors);
1167 }
1168
1169 /* Set sFlow configuration on 'br'. */
1170 static void
1171 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1172 {
1173 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1174 struct ovsrec_controller **controllers;
1175 struct ofproto_sflow_options oso;
1176 size_t n_controllers;
1177 size_t i;
1178
1179 if (!cfg) {
1180 ofproto_set_sflow(br->ofproto, NULL);
1181 return;
1182 }
1183
1184 memset(&oso, 0, sizeof oso);
1185
1186 sset_init(&oso.targets);
1187 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1188
1189 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1190 if (cfg->sampling) {
1191 oso.sampling_rate = *cfg->sampling;
1192 }
1193
1194 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1195 if (cfg->polling) {
1196 oso.polling_interval = *cfg->polling;
1197 }
1198
1199 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1200 if (cfg->header) {
1201 oso.header_len = *cfg->header;
1202 }
1203
1204 oso.sub_id = (*sflow_bridge_number)++;
1205 oso.agent_device = cfg->agent;
1206
1207 oso.control_ip = NULL;
1208 n_controllers = bridge_get_controllers(br, &controllers);
1209 for (i = 0; i < n_controllers; i++) {
1210 if (controllers[i]->local_ip) {
1211 oso.control_ip = controllers[i]->local_ip;
1212 break;
1213 }
1214 }
1215 ofproto_set_sflow(br->ofproto, &oso);
1216
1217 sset_destroy(&oso.targets);
1218 }
1219
1220 /* Returns whether a IPFIX row is valid. */
1221 static bool
1222 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1223 {
1224 return ipfix && ipfix->n_targets > 0;
1225 }
1226
1227 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1228 static bool
1229 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1230 const struct bridge *br)
1231 {
1232 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1233 }
1234
1235 /* Set IPFIX configuration on 'br'. */
1236 static void
1237 bridge_configure_ipfix(struct bridge *br)
1238 {
1239 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1240 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1241 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1242 struct ofproto_ipfix_bridge_exporter_options be_opts;
1243 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1244 size_t n_fe_opts = 0;
1245 const char *virtual_obs_id;
1246
1247 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1248 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1249 n_fe_opts++;
1250 }
1251 }
1252
1253 if (!valid_be_cfg && n_fe_opts == 0) {
1254 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1255 return;
1256 }
1257
1258 if (valid_be_cfg) {
1259 memset(&be_opts, 0, sizeof be_opts);
1260
1261 sset_init(&be_opts.targets);
1262 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1263
1264 if (be_cfg->sampling) {
1265 be_opts.sampling_rate = *be_cfg->sampling;
1266 } else {
1267 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1268 }
1269 if (be_cfg->obs_domain_id) {
1270 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1271 }
1272 if (be_cfg->obs_point_id) {
1273 be_opts.obs_point_id = *be_cfg->obs_point_id;
1274 }
1275 if (be_cfg->cache_active_timeout) {
1276 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1277 }
1278 if (be_cfg->cache_max_flows) {
1279 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1280 }
1281
1282 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1283 "enable-tunnel-sampling", true);
1284
1285 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1286 "enable-input-sampling", false);
1287
1288 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1289 "enable-output-sampling", false);
1290
1291 virtual_obs_id = smap_get(&be_cfg->other_config, "virtual_obs_id");
1292 be_opts.virtual_obs_id = nullable_xstrdup(virtual_obs_id);
1293 }
1294
1295 if (n_fe_opts > 0) {
1296 struct ofproto_ipfix_flow_exporter_options *opts;
1297 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1298 opts = fe_opts;
1299 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1300 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1301 opts->collector_set_id = fe_cfg->id;
1302 sset_init(&opts->targets);
1303 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1304 fe_cfg->ipfix->n_targets);
1305 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1306 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1307 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1308 ? *fe_cfg->ipfix->cache_max_flows : 0;
1309 opts->enable_tunnel_sampling = smap_get_bool(
1310 &fe_cfg->ipfix->other_config,
1311 "enable-tunnel-sampling", true);
1312 virtual_obs_id = smap_get(&fe_cfg->ipfix->other_config,
1313 "virtual_obs_id");
1314 opts->virtual_obs_id = nullable_xstrdup(virtual_obs_id);
1315 opts++;
1316 }
1317 }
1318 }
1319
1320 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1321 n_fe_opts);
1322
1323 if (valid_be_cfg) {
1324 sset_destroy(&be_opts.targets);
1325 free(be_opts.virtual_obs_id);
1326 }
1327
1328 if (n_fe_opts > 0) {
1329 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1330 size_t i;
1331 for (i = 0; i < n_fe_opts; i++) {
1332 sset_destroy(&opts->targets);
1333 free(opts->virtual_obs_id);
1334 opts++;
1335 }
1336 free(fe_opts);
1337 }
1338 }
1339
1340 static void
1341 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1342 struct ofproto_port_stp_settings *port_s,
1343 int *port_num_counter, unsigned long *port_num_bitmap)
1344 {
1345 const char *config_str;
1346 struct iface *iface;
1347
1348 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1349 port_s->enable = false;
1350 return;
1351 } else {
1352 port_s->enable = true;
1353 }
1354
1355 /* STP over bonds is not supported. */
1356 if (!ovs_list_is_singleton(&port->ifaces)) {
1357 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1358 port->name);
1359 port_s->enable = false;
1360 return;
1361 }
1362
1363 iface = CONTAINER_OF(ovs_list_front(&port->ifaces), struct iface, port_elem);
1364
1365 /* Internal ports shouldn't participate in spanning tree, so
1366 * skip them. */
1367 if (!strcmp(iface->type, "internal")) {
1368 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1369 port_s->enable = false;
1370 return;
1371 }
1372
1373 /* STP on mirror output ports is not supported. */
1374 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1375 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1376 port_s->enable = false;
1377 return;
1378 }
1379
1380 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1381 if (config_str) {
1382 unsigned long int port_num = strtoul(config_str, NULL, 0);
1383 int port_idx = port_num - 1;
1384
1385 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1386 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1387 port_s->enable = false;
1388 return;
1389 }
1390
1391 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1392 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1393 port->name, port_num);
1394 port_s->enable = false;
1395 return;
1396 }
1397 bitmap_set1(port_num_bitmap, port_idx);
1398 port_s->port_num = port_idx;
1399 } else {
1400 if (*port_num_counter >= STP_MAX_PORTS) {
1401 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1402 port_s->enable = false;
1403 return;
1404 }
1405
1406 port_s->port_num = (*port_num_counter)++;
1407 }
1408
1409 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1410 if (config_str) {
1411 port_s->path_cost = strtoul(config_str, NULL, 10);
1412 } else {
1413 enum netdev_features current;
1414 unsigned int mbps;
1415
1416 netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1417 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1418 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1419 }
1420
1421 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1422 if (config_str) {
1423 port_s->priority = strtoul(config_str, NULL, 0);
1424 } else {
1425 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1426 }
1427 }
1428
1429 static void
1430 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1431 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1432 {
1433 const char *config_str;
1434 struct iface *iface;
1435
1436 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1437 port_s->enable = false;
1438 return;
1439 } else {
1440 port_s->enable = true;
1441 }
1442
1443 /* RSTP over bonds is not supported. */
1444 if (!ovs_list_is_singleton(&port->ifaces)) {
1445 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1446 port->name);
1447 port_s->enable = false;
1448 return;
1449 }
1450
1451 iface = CONTAINER_OF(ovs_list_front(&port->ifaces), struct iface, port_elem);
1452
1453 /* Internal ports shouldn't participate in spanning tree, so
1454 * skip them. */
1455 if (!strcmp(iface->type, "internal")) {
1456 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1457 port_s->enable = false;
1458 return;
1459 }
1460
1461 /* RSTP on mirror output ports is not supported. */
1462 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1463 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1464 port_s->enable = false;
1465 return;
1466 }
1467
1468 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1469 if (config_str) {
1470 unsigned long int port_num = strtoul(config_str, NULL, 0);
1471 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1472 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1473 port_s->enable = false;
1474 return;
1475 }
1476 port_s->port_num = port_num;
1477 } else {
1478 if (*port_num_counter >= RSTP_MAX_PORTS) {
1479 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1480 port_s->enable = false;
1481 return;
1482 }
1483 /* If rstp-port-num is not specified, use 0.
1484 * rstp_port_set_port_number() will look for the first free one. */
1485 port_s->port_num = 0;
1486 }
1487
1488 /* Increment the port num counter, because we only support
1489 * RSTP_MAX_PORTS rstp ports. */
1490 (*port_num_counter)++;
1491
1492 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1493 if (config_str) {
1494 port_s->path_cost = strtoul(config_str, NULL, 10);
1495 } else {
1496 enum netdev_features current;
1497 unsigned int mbps;
1498
1499 netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1500 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1501 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1502 }
1503
1504 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1505 if (config_str) {
1506 port_s->priority = strtoul(config_str, NULL, 0);
1507 } else {
1508 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1509 }
1510
1511 port_s->admin_p2p_mac_state = smap_get_ullong(
1512 &port->cfg->other_config, "rstp-admin-p2p-mac",
1513 RSTP_ADMIN_P2P_MAC_FORCE_TRUE);
1514
1515 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1516 "rstp-admin-port-state", true);
1517
1518 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1519 "rstp-port-admin-edge", false);
1520 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1521 "rstp-port-auto-edge", true);
1522 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1523 "rstp-port-mcheck", false);
1524 }
1525
1526 /* Set spanning tree configuration on 'br'. */
1527 static void
1528 bridge_configure_stp(struct bridge *br, bool enable_stp)
1529 {
1530 if (!enable_stp) {
1531 ofproto_set_stp(br->ofproto, NULL);
1532 } else {
1533 struct ofproto_stp_settings br_s;
1534 const char *config_str;
1535 struct port *port;
1536 int port_num_counter;
1537 unsigned long *port_num_bitmap;
1538
1539 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1540 if (config_str) {
1541 struct eth_addr ea;
1542
1543 if (eth_addr_from_string(config_str, &ea)) {
1544 br_s.system_id = eth_addr_to_uint64(ea);
1545 } else {
1546 br_s.system_id = eth_addr_to_uint64(br->ea);
1547 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1548 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1549 }
1550 } else {
1551 br_s.system_id = eth_addr_to_uint64(br->ea);
1552 }
1553
1554 br_s.priority = smap_get_ullong(&br->cfg->other_config, "stp-priority",
1555 STP_DEFAULT_BRIDGE_PRIORITY);
1556 br_s.hello_time = smap_get_ullong(&br->cfg->other_config,
1557 "stp-hello-time",
1558 STP_DEFAULT_HELLO_TIME);
1559
1560 br_s.max_age = smap_get_ullong(&br->cfg->other_config, "stp-max-age",
1561 STP_DEFAULT_MAX_AGE / 1000) * 1000;
1562 br_s.fwd_delay = smap_get_ullong(&br->cfg->other_config,
1563 "stp-forward-delay",
1564 STP_DEFAULT_FWD_DELAY / 1000) * 1000;
1565
1566 /* Configure STP on the bridge. */
1567 if (ofproto_set_stp(br->ofproto, &br_s)) {
1568 VLOG_ERR("bridge %s: could not enable STP", br->name);
1569 return;
1570 }
1571
1572 /* Users must either set the port number with the "stp-port-num"
1573 * configuration on all ports or none. If manual configuration
1574 * is not done, then we allocate them sequentially. */
1575 port_num_counter = 0;
1576 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1577 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1578 struct ofproto_port_stp_settings port_s;
1579 struct iface *iface;
1580
1581 port_configure_stp(br->ofproto, port, &port_s,
1582 &port_num_counter, port_num_bitmap);
1583
1584 /* As bonds are not supported, just apply configuration to
1585 * all interfaces. */
1586 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1587 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1588 &port_s)) {
1589 VLOG_ERR("port %s: could not enable STP", port->name);
1590 continue;
1591 }
1592 }
1593 }
1594
1595 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1596 && port_num_counter) {
1597 VLOG_ERR("bridge %s: must manually configure all STP port "
1598 "IDs or none, disabling", br->name);
1599 ofproto_set_stp(br->ofproto, NULL);
1600 }
1601 bitmap_free(port_num_bitmap);
1602 }
1603 }
1604
1605 static void
1606 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1607 {
1608 if (!enable_rstp) {
1609 ofproto_set_rstp(br->ofproto, NULL);
1610 } else {
1611 struct ofproto_rstp_settings br_s;
1612 const char *config_str;
1613 struct port *port;
1614 int port_num_counter;
1615
1616 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1617 if (config_str) {
1618 struct eth_addr ea;
1619
1620 if (eth_addr_from_string(config_str, &ea)) {
1621 br_s.address = eth_addr_to_uint64(ea);
1622 }
1623 else {
1624 br_s.address = eth_addr_to_uint64(br->ea);
1625 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1626 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1627 }
1628 }
1629 else {
1630 br_s.address = eth_addr_to_uint64(br->ea);
1631 }
1632
1633 const struct smap *oc = &br->cfg->other_config;
1634 br_s.priority = smap_get_ullong(oc, "rstp-priority",
1635 RSTP_DEFAULT_PRIORITY);
1636 br_s.ageing_time = smap_get_ullong(oc, "rstp-ageing-time",
1637 RSTP_DEFAULT_AGEING_TIME);
1638 br_s.force_protocol_version = smap_get_ullong(
1639 oc, "rstp-force-protocol-version", FPV_DEFAULT);
1640 br_s.bridge_max_age = smap_get_ullong(oc, "rstp-max-age",
1641 RSTP_DEFAULT_BRIDGE_MAX_AGE);
1642 br_s.bridge_forward_delay = smap_get_ullong(
1643 oc, "rstp-forward-delay", RSTP_DEFAULT_BRIDGE_FORWARD_DELAY);
1644 br_s.transmit_hold_count = smap_get_ullong(
1645 oc, "rstp-transmit-hold-count", RSTP_DEFAULT_TRANSMIT_HOLD_COUNT);
1646
1647 /* Configure RSTP on the bridge. */
1648 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1649 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1650 return;
1651 }
1652
1653 port_num_counter = 0;
1654 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1655 struct ofproto_port_rstp_settings port_s;
1656 struct iface *iface;
1657
1658 port_configure_rstp(br->ofproto, port, &port_s,
1659 &port_num_counter);
1660
1661 /* As bonds are not supported, just apply configuration to
1662 * all interfaces. */
1663 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1664 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1665 &port_s)) {
1666 VLOG_ERR("port %s: could not enable RSTP", port->name);
1667 continue;
1668 }
1669 }
1670 }
1671 }
1672 }
1673
1674 static void
1675 bridge_configure_spanning_tree(struct bridge *br)
1676 {
1677 bool enable_rstp = br->cfg->rstp_enable;
1678 bool enable_stp = br->cfg->stp_enable;
1679
1680 if (enable_rstp && enable_stp) {
1681 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1682 "configured; enabling RSTP", br->name);
1683 enable_stp = false;
1684 }
1685
1686 bridge_configure_stp(br, enable_stp);
1687 bridge_configure_rstp(br, enable_rstp);
1688 }
1689
1690 static bool
1691 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1692 {
1693 const struct port *port = port_lookup(br, name);
1694 return port && port_is_bond_fake_iface(port);
1695 }
1696
1697 static bool
1698 port_is_bond_fake_iface(const struct port *port)
1699 {
1700 return port->cfg->bond_fake_iface && !ovs_list_is_short(&port->ifaces);
1701 }
1702
1703 static void
1704 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1705 {
1706 struct bridge *br, *next;
1707 struct shash_node *node;
1708 struct shash new_br;
1709 size_t i;
1710
1711 /* Collect new bridges' names and types. */
1712 shash_init(&new_br);
1713 for (i = 0; i < cfg->n_bridges; i++) {
1714 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1715 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1716
1717 if (strchr(br_cfg->name, '/') || strchr(br_cfg->name, '\\')) {
1718 /* Prevent remote ovsdb-server users from accessing arbitrary
1719 * directories, e.g. consider a bridge named "../../../etc/".
1720 *
1721 * Prohibiting "\" is only necessary on Windows but it's no great
1722 * loss elsewhere. */
1723 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1724 br_cfg->name);
1725 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1726 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1727 }
1728 }
1729
1730 /* Get rid of deleted bridges or those whose types have changed.
1731 * Update 'cfg' of bridges that still exist. */
1732 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1733 br->cfg = shash_find_data(&new_br, br->name);
1734 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1735 br->cfg->datapath_type))) {
1736 bridge_destroy(br, true);
1737 }
1738 }
1739
1740 /* Add new bridges. */
1741 SHASH_FOR_EACH(node, &new_br) {
1742 const struct ovsrec_bridge *br_cfg = node->data;
1743 if (!bridge_lookup(br_cfg->name)) {
1744 bridge_create(br_cfg);
1745 }
1746 }
1747
1748 shash_destroy(&new_br);
1749 }
1750
1751 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1752 * Returns 0 if successful, otherwise a positive errno value. */
1753 static int
1754 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1755 struct netdev *netdev, char **errp)
1756 {
1757 return netdev_set_config(netdev, &iface_cfg->options, errp);
1758 }
1759
1760 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1761 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1762 *
1763 * If successful, returns 0 and stores the network device in '*netdevp'. On
1764 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1765 static int
1766 iface_do_create(const struct bridge *br,
1767 const struct ovsrec_interface *iface_cfg,
1768 ofp_port_t *ofp_portp, struct netdev **netdevp,
1769 char **errp)
1770 {
1771 struct netdev *netdev = NULL;
1772 int error;
1773 const char *type;
1774
1775 if (netdev_is_reserved_name(iface_cfg->name)) {
1776 VLOG_WARN("could not create interface %s, name is reserved",
1777 iface_cfg->name);
1778 error = EINVAL;
1779 goto error;
1780 }
1781
1782 type = ofproto_port_open_type(br->cfg->datapath_type,
1783 iface_get_type(iface_cfg, br->cfg));
1784 error = netdev_open(iface_cfg->name, type, &netdev);
1785 if (error) {
1786 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1787 iface_cfg->name, ovs_strerror(error));
1788 goto error;
1789 }
1790
1791 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1792 if (error) {
1793 goto error;
1794 }
1795
1796 iface_set_netdev_mtu(iface_cfg, netdev);
1797
1798 *ofp_portp = iface_pick_ofport(iface_cfg);
1799 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1800 if (error) {
1801 VLOG_WARN_BUF(errp, "could not add network device %s to ofproto (%s)",
1802 iface_cfg->name, ovs_strerror(error));
1803 goto error;
1804 }
1805
1806 VLOG_INFO("bridge %s: added interface %s on port %d",
1807 br->name, iface_cfg->name, *ofp_portp);
1808
1809 *netdevp = netdev;
1810 return 0;
1811
1812 error:
1813 *netdevp = NULL;
1814 netdev_close(netdev);
1815 return error;
1816 }
1817
1818 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1819 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1820 * automatically allocated for the iface. Takes ownership of and
1821 * deallocates 'if_cfg'.
1822 *
1823 * Return true if an iface is successfully created, false otherwise. */
1824 static bool
1825 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1826 const struct ovsrec_port *port_cfg)
1827 {
1828 struct netdev *netdev;
1829 struct iface *iface;
1830 ofp_port_t ofp_port;
1831 struct port *port;
1832 char *errp = NULL;
1833 int error;
1834
1835 /* Do the bits that can fail up front. */
1836 ovs_assert(!iface_lookup(br, iface_cfg->name));
1837 error = iface_do_create(br, iface_cfg, &ofp_port, &netdev, &errp);
1838 if (error) {
1839 iface_clear_db_record(iface_cfg, errp);
1840 free(errp);
1841 return false;
1842 }
1843
1844 /* Get or create the port structure. */
1845 port = port_lookup(br, port_cfg->name);
1846 if (!port) {
1847 port = port_create(br, port_cfg);
1848 }
1849
1850 /* Create the iface structure. */
1851 iface = xzalloc(sizeof *iface);
1852 ovs_list_push_back(&port->ifaces, &iface->port_elem);
1853 hmap_insert(&br->iface_by_name, &iface->name_node,
1854 hash_string(iface_cfg->name, 0));
1855 iface->port = port;
1856 iface->name = xstrdup(iface_cfg->name);
1857 iface->ofp_port = ofp_port;
1858 iface->netdev = netdev;
1859 iface->type = iface_get_type(iface_cfg, br->cfg);
1860 iface->netdev_type = ofproto_port_open_type(br->cfg->datapath_type,
1861 iface->type);
1862 iface->cfg = iface_cfg;
1863 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1864 hash_ofp_port(ofp_port));
1865
1866 /* Populate initial status in database. */
1867 iface_refresh_stats(iface);
1868 iface_refresh_netdev_status(iface);
1869
1870 /* Add bond fake iface if necessary. */
1871 if (port_is_bond_fake_iface(port)) {
1872 struct ofproto_port ofproto_port;
1873
1874 if (ofproto_port_query_by_name(br->ofproto, port->name,
1875 &ofproto_port)) {
1876 error = netdev_open(port->name, "internal", &netdev);
1877 if (!error) {
1878 ofp_port_t fake_ofp_port = OFPP_NONE;
1879 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1880 netdev_close(netdev);
1881 } else {
1882 VLOG_WARN("could not open network device %s (%s)",
1883 port->name, ovs_strerror(error));
1884 }
1885 } else {
1886 /* Already exists, nothing to do. */
1887 ofproto_port_destroy(&ofproto_port);
1888 }
1889 }
1890
1891 return true;
1892 }
1893
1894 /* Set forward BPDU option. */
1895 static void
1896 bridge_configure_forward_bpdu(struct bridge *br)
1897 {
1898 ofproto_set_forward_bpdu(br->ofproto,
1899 smap_get_bool(&br->cfg->other_config,
1900 "forward-bpdu",
1901 false));
1902 }
1903
1904 /* Set MAC learning table configuration for 'br'. */
1905 static void
1906 bridge_configure_mac_table(struct bridge *br)
1907 {
1908 const struct smap *oc = &br->cfg->other_config;
1909 int idle_time = smap_get_int(oc, "mac-aging-time", 0);
1910 if (!idle_time) {
1911 idle_time = MAC_ENTRY_DEFAULT_IDLE_TIME;
1912 }
1913
1914 int mac_table_size = smap_get_int(oc, "mac-table-size", 0);
1915 if (!mac_table_size) {
1916 mac_table_size = MAC_DEFAULT_MAX;
1917 }
1918
1919 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1920 }
1921
1922 /* Set multicast snooping table configuration for 'br'. */
1923 static void
1924 bridge_configure_mcast_snooping(struct bridge *br)
1925 {
1926 if (!br->cfg->mcast_snooping_enable) {
1927 ofproto_set_mcast_snooping(br->ofproto, NULL);
1928 } else {
1929 struct port *port;
1930 struct ofproto_mcast_snooping_settings br_s;
1931
1932 const struct smap *oc = &br->cfg->other_config;
1933 int idle_time = smap_get_int(oc, "mcast-snooping-aging-time", 0);
1934 br_s.idle_time = idle_time ? idle_time : MCAST_ENTRY_DEFAULT_IDLE_TIME;
1935 int max_entries = smap_get_int(oc, "mcast-snooping-table-size", 0);
1936 br_s.max_entries = (max_entries
1937 ? max_entries
1938 : MCAST_DEFAULT_MAX_ENTRIES);
1939
1940 br_s.flood_unreg = !smap_get_bool(
1941 oc, "mcast-snooping-disable-flood-unregistered", false);
1942
1943 /* Configure multicast snooping on the bridge */
1944 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1945 VLOG_ERR("bridge %s: could not enable multicast snooping",
1946 br->name);
1947 return;
1948 }
1949
1950 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1951 struct ofproto_mcast_snooping_port_settings port_s;
1952 port_s.flood = smap_get_bool(&port->cfg->other_config,
1953 "mcast-snooping-flood", false);
1954 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1955 "mcast-snooping-flood-reports", false);
1956 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1957 VLOG_ERR("port %s: could not configure mcast snooping",
1958 port->name);
1959 }
1960 }
1961 }
1962 }
1963
1964 static void
1965 find_local_hw_addr(const struct bridge *br, struct eth_addr *ea,
1966 const struct port *fake_br, struct iface **hw_addr_iface)
1967 {
1968 struct hmapx mirror_output_ports;
1969 struct port *port;
1970 bool found_addr = false;
1971 int error;
1972 int i;
1973
1974 /* Mirror output ports don't participate in picking the local hardware
1975 * address. ofproto can't help us find out whether a given port is a
1976 * mirror output because we haven't configured mirrors yet, so we need to
1977 * accumulate them ourselves. */
1978 hmapx_init(&mirror_output_ports);
1979 for (i = 0; i < br->cfg->n_mirrors; i++) {
1980 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1981 if (m->output_port) {
1982 hmapx_add(&mirror_output_ports, m->output_port);
1983 }
1984 }
1985
1986 /* Otherwise choose the minimum non-local MAC address among all of the
1987 * interfaces. */
1988 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1989 struct eth_addr iface_ea;
1990 struct iface *candidate;
1991 struct iface *iface;
1992
1993 /* Mirror output ports don't participate. */
1994 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1995 continue;
1996 }
1997
1998 /* Choose the MAC address to represent the port. */
1999 iface = NULL;
2000 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac,
2001 &iface_ea)) {
2002 /* Find the interface with this Ethernet address (if any) so that
2003 * we can provide the correct devname to the caller. */
2004 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2005 struct eth_addr candidate_ea;
2006 if (!netdev_get_etheraddr(candidate->netdev, &candidate_ea)
2007 && eth_addr_equals(iface_ea, candidate_ea)) {
2008 iface = candidate;
2009 }
2010 }
2011 } else {
2012 /* Choose the interface whose MAC address will represent the port.
2013 * The Linux kernel bonding code always chooses the MAC address of
2014 * the first slave added to a bond, and the Fedora networking
2015 * scripts always add slaves to a bond in alphabetical order, so
2016 * for compatibility we choose the interface with the name that is
2017 * first in alphabetical order. */
2018 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2019 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2020 iface = candidate;
2021 }
2022 }
2023
2024 /* A port always has at least one interface. */
2025 ovs_assert(iface != NULL);
2026
2027 /* The local port doesn't count (since we're trying to choose its
2028 * MAC address anyway). */
2029 if (iface->ofp_port == OFPP_LOCAL) {
2030 continue;
2031 }
2032
2033 /* For fake bridges we only choose from ports with the same tag */
2034 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2035 if (!port->cfg->tag) {
2036 continue;
2037 }
2038 if (*port->cfg->tag != *fake_br->cfg->tag) {
2039 continue;
2040 }
2041 }
2042
2043 /* Grab MAC. */
2044 error = netdev_get_etheraddr(iface->netdev, &iface_ea);
2045 if (error) {
2046 continue;
2047 }
2048 }
2049
2050 /* Compare against our current choice. */
2051 if (!eth_addr_is_multicast(iface_ea) &&
2052 !eth_addr_is_local(iface_ea) &&
2053 !eth_addr_is_reserved(iface_ea) &&
2054 !eth_addr_is_zero(iface_ea) &&
2055 (!found_addr || eth_addr_compare_3way(iface_ea, *ea) < 0))
2056 {
2057 *ea = iface_ea;
2058 *hw_addr_iface = iface;
2059 found_addr = true;
2060 }
2061 }
2062
2063 if (!found_addr) {
2064 *ea = br->default_ea;
2065 *hw_addr_iface = NULL;
2066 }
2067
2068 hmapx_destroy(&mirror_output_ports);
2069 }
2070
2071 static void
2072 bridge_pick_local_hw_addr(struct bridge *br, struct eth_addr *ea,
2073 struct iface **hw_addr_iface)
2074 {
2075 *hw_addr_iface = NULL;
2076
2077 /* Did the user request a particular MAC? */
2078 const char *hwaddr = smap_get_def(&br->cfg->other_config, "hwaddr", "");
2079 if (eth_addr_from_string(hwaddr, ea)) {
2080 if (eth_addr_is_multicast(*ea)) {
2081 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2082 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(*ea));
2083 } else if (eth_addr_is_zero(*ea)) {
2084 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2085 } else {
2086 return;
2087 }
2088 }
2089
2090 /* Find a local hw address */
2091 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2092 }
2093
2094 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2095 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2096 * an interface on 'br', then that interface must be passed in as
2097 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2098 * 'hw_addr_iface' must be passed in as a null pointer. */
2099 static uint64_t
2100 bridge_pick_datapath_id(struct bridge *br,
2101 const struct eth_addr bridge_ea,
2102 struct iface *hw_addr_iface)
2103 {
2104 /*
2105 * The procedure for choosing a bridge MAC address will, in the most
2106 * ordinary case, also choose a unique MAC that we can use as a datapath
2107 * ID. In some special cases, though, multiple bridges will end up with
2108 * the same MAC address. This is OK for the bridges, but it will confuse
2109 * the OpenFlow controller, because each datapath needs a unique datapath
2110 * ID.
2111 *
2112 * Datapath IDs must be unique. It is also very desirable that they be
2113 * stable from one run to the next, so that policy set on a datapath
2114 * "sticks".
2115 */
2116 const char *datapath_id;
2117 uint64_t dpid;
2118
2119 datapath_id = smap_get_def(&br->cfg->other_config, "datapath-id", "");
2120 if (dpid_from_string(datapath_id, &dpid)) {
2121 return dpid;
2122 }
2123
2124 if (!hw_addr_iface) {
2125 /*
2126 * A purely internal bridge, that is, one that has no non-virtual
2127 * network devices on it at all, is difficult because it has no
2128 * natural unique identifier at all.
2129 *
2130 * When the host is a XenServer, we handle this case by hashing the
2131 * host's UUID with the name of the bridge. Names of bridges are
2132 * persistent across XenServer reboots, although they can be reused if
2133 * an internal network is destroyed and then a new one is later
2134 * created, so this is fairly effective.
2135 *
2136 * When the host is not a XenServer, we punt by using a random MAC
2137 * address on each run.
2138 */
2139 const char *host_uuid = xenserver_get_host_uuid();
2140 if (host_uuid) {
2141 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2142 dpid = dpid_from_hash(combined, strlen(combined));
2143 free(combined);
2144 return dpid;
2145 }
2146 }
2147
2148 return eth_addr_to_uint64(bridge_ea);
2149 }
2150
2151 static uint64_t
2152 dpid_from_hash(const void *data, size_t n)
2153 {
2154 union {
2155 uint8_t bytes[SHA1_DIGEST_SIZE];
2156 struct eth_addr ea;
2157 } hash;
2158
2159 sha1_bytes(data, n, hash.bytes);
2160 eth_addr_mark_random(&hash.ea);
2161 return eth_addr_to_uint64(hash.ea);
2162 }
2163
2164 static void
2165 iface_refresh_netdev_status(struct iface *iface)
2166 {
2167 struct smap smap;
2168
2169 enum netdev_features current;
2170 enum netdev_flags flags;
2171 const char *link_state;
2172 struct eth_addr mac;
2173 int64_t bps, mtu_64, ifindex64, link_resets;
2174 int mtu, error;
2175
2176 if (iface_is_synthetic(iface)) {
2177 return;
2178 }
2179
2180 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2181 && !status_txn_try_again) {
2182 return;
2183 }
2184
2185 iface->change_seq = netdev_get_change_seq(iface->netdev);
2186
2187 smap_init(&smap);
2188
2189 if (!netdev_get_status(iface->netdev, &smap)) {
2190 ovsrec_interface_set_status(iface->cfg, &smap);
2191 } else {
2192 ovsrec_interface_set_status(iface->cfg, NULL);
2193 }
2194
2195 smap_destroy(&smap);
2196
2197 error = netdev_get_flags(iface->netdev, &flags);
2198 if (!error) {
2199 const char *state = flags & NETDEV_UP ? "up" : "down";
2200
2201 ovsrec_interface_set_admin_state(iface->cfg, state);
2202 } else {
2203 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2204 }
2205
2206 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2207 ovsrec_interface_set_link_state(iface->cfg, link_state);
2208
2209 link_resets = netdev_get_carrier_resets(iface->netdev);
2210 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2211
2212 error = netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
2213 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2214 if (bps) {
2215 ovsrec_interface_set_duplex(iface->cfg,
2216 netdev_features_is_full_duplex(current)
2217 ? "full" : "half");
2218 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2219 } else {
2220 ovsrec_interface_set_duplex(iface->cfg, NULL);
2221 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2222 }
2223
2224 error = netdev_get_mtu(iface->netdev, &mtu);
2225 if (!error) {
2226 mtu_64 = mtu;
2227 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2228 } else {
2229 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2230 }
2231
2232 error = netdev_get_etheraddr(iface->netdev, &mac);
2233 if (!error) {
2234 char mac_string[ETH_ADDR_STRLEN + 1];
2235
2236 snprintf(mac_string, sizeof mac_string,
2237 ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2238 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2239 } else {
2240 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2241 }
2242
2243 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2244 * if there is no valid ifindex number. */
2245 ifindex64 = netdev_get_ifindex(iface->netdev);
2246 if (ifindex64 < 0) {
2247 ifindex64 = 0;
2248 }
2249 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2250 }
2251
2252 static void
2253 iface_refresh_ofproto_status(struct iface *iface)
2254 {
2255 int current;
2256
2257 if (iface_is_synthetic(iface)) {
2258 return;
2259 }
2260
2261 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2262 iface->ofp_port);
2263 if (current >= 0) {
2264 bool bl = current;
2265 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2266 } else {
2267 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2268 }
2269
2270 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2271 iface->ofp_port)
2272 || status_txn_try_again) {
2273 iface_refresh_cfm_stats(iface);
2274 }
2275
2276 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2277 iface->ofp_port)
2278 || status_txn_try_again) {
2279 struct smap smap;
2280
2281 smap_init(&smap);
2282 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2283 iface->ofp_port, &smap);
2284 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2285 smap_destroy(&smap);
2286 }
2287 }
2288
2289 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2290 * synthetic. */
2291 static void
2292 iface_refresh_cfm_stats(struct iface *iface)
2293 {
2294 const struct ovsrec_interface *cfg = iface->cfg;
2295 struct cfm_status status;
2296 int error;
2297
2298 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2299 iface->ofp_port, &status);
2300 if (error > 0) {
2301 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2302 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2303 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2304 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2305 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2306 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2307 } else {
2308 const char *reasons[CFM_FAULT_N_REASONS];
2309 int64_t cfm_health = status.health;
2310 int64_t cfm_flap_count = status.flap_count;
2311 bool faulted = status.faults != 0;
2312 size_t i, j;
2313
2314 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2315
2316 j = 0;
2317 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2318 int reason = 1 << i;
2319 if (status.faults & reason) {
2320 reasons[j++] = cfm_fault_reason_to_str(reason);
2321 }
2322 }
2323 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2324
2325 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2326
2327 if (status.remote_opstate >= 0) {
2328 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2329 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2330 } else {
2331 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2332 }
2333
2334 ovsrec_interface_set_cfm_remote_mpids(cfg,
2335 (const int64_t *)status.rmps,
2336 status.n_rmps);
2337 if (cfm_health >= 0) {
2338 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2339 } else {
2340 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2341 }
2342
2343 free(status.rmps);
2344 }
2345 }
2346
2347 static void
2348 iface_refresh_stats(struct iface *iface)
2349 {
2350 #define IFACE_STATS \
2351 IFACE_STAT(rx_packets, "rx_packets") \
2352 IFACE_STAT(tx_packets, "tx_packets") \
2353 IFACE_STAT(rx_bytes, "rx_bytes") \
2354 IFACE_STAT(tx_bytes, "tx_bytes") \
2355 IFACE_STAT(rx_dropped, "rx_dropped") \
2356 IFACE_STAT(tx_dropped, "tx_dropped") \
2357 IFACE_STAT(rx_errors, "rx_errors") \
2358 IFACE_STAT(tx_errors, "tx_errors") \
2359 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2360 IFACE_STAT(rx_over_errors, "rx_over_err") \
2361 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2362 IFACE_STAT(collisions, "collisions") \
2363 IFACE_STAT(rx_1_to_64_packets, "rx_1_to_64_packets") \
2364 IFACE_STAT(rx_65_to_127_packets, "rx_65_to_127_packets") \
2365 IFACE_STAT(rx_128_to_255_packets, "rx_128_to_255_packets") \
2366 IFACE_STAT(rx_256_to_511_packets, "rx_256_to_511_packets") \
2367 IFACE_STAT(rx_512_to_1023_packets, "rx_512_to_1023_packets") \
2368 IFACE_STAT(rx_1024_to_1522_packets, "rx_1024_to_1522_packets") \
2369 IFACE_STAT(rx_1523_to_max_packets, "rx_1523_to_max_packets") \
2370 IFACE_STAT(tx_1_to_64_packets, "tx_1_to_64_packets") \
2371 IFACE_STAT(tx_65_to_127_packets, "tx_65_to_127_packets") \
2372 IFACE_STAT(tx_128_to_255_packets, "tx_128_to_255_packets") \
2373 IFACE_STAT(tx_256_to_511_packets, "tx_256_to_511_packets") \
2374 IFACE_STAT(tx_512_to_1023_packets, "tx_512_to_1023_packets") \
2375 IFACE_STAT(tx_1024_to_1522_packets, "tx_1024_to_1522_packets") \
2376 IFACE_STAT(tx_1523_to_max_packets, "tx_1523_to_max_packets") \
2377 IFACE_STAT(tx_multicast_packets, "tx_multicast_packets") \
2378 IFACE_STAT(rx_broadcast_packets, "rx_broadcast_packets") \
2379 IFACE_STAT(tx_broadcast_packets, "tx_broadcast_packets") \
2380 IFACE_STAT(rx_undersized_errors, "rx_undersized_errors") \
2381 IFACE_STAT(rx_oversize_errors, "rx_oversize_errors") \
2382 IFACE_STAT(rx_fragmented_errors, "rx_fragmented_errors") \
2383 IFACE_STAT(rx_jabber_errors, "rx_jabber_errors")
2384
2385 #define IFACE_STAT(MEMBER, NAME) + 1
2386 enum { N_IFACE_STATS = IFACE_STATS };
2387 #undef IFACE_STAT
2388 int64_t values[N_IFACE_STATS];
2389 const char *keys[N_IFACE_STATS];
2390 int n;
2391
2392 struct netdev_stats stats;
2393
2394 if (iface_is_synthetic(iface)) {
2395 return;
2396 }
2397
2398 /* Intentionally ignore return value, since errors will set 'stats' to
2399 * all-1s, and we will deal with that correctly below. */
2400 netdev_get_stats(iface->netdev, &stats);
2401
2402 /* Copy statistics into keys[] and values[]. */
2403 n = 0;
2404 #define IFACE_STAT(MEMBER, NAME) \
2405 if (stats.MEMBER != UINT64_MAX) { \
2406 keys[n] = NAME; \
2407 values[n] = stats.MEMBER; \
2408 n++; \
2409 }
2410 IFACE_STATS;
2411 #undef IFACE_STAT
2412 ovs_assert(n <= N_IFACE_STATS);
2413
2414 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2415 #undef IFACE_STATS
2416 }
2417
2418 static void
2419 br_refresh_datapath_info(struct bridge *br)
2420 {
2421 const char *version;
2422
2423 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2424 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2425 : NULL);
2426
2427 ovsrec_bridge_set_datapath_version(br->cfg,
2428 version ? version : "<unknown>");
2429 }
2430
2431 static void
2432 br_refresh_stp_status(struct bridge *br)
2433 {
2434 struct smap smap = SMAP_INITIALIZER(&smap);
2435 struct ofproto *ofproto = br->ofproto;
2436 struct ofproto_stp_status status;
2437
2438 if (ofproto_get_stp_status(ofproto, &status)) {
2439 return;
2440 }
2441
2442 if (!status.enabled) {
2443 ovsrec_bridge_set_status(br->cfg, NULL);
2444 return;
2445 }
2446
2447 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2448 STP_ID_ARGS(status.bridge_id));
2449 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2450 STP_ID_ARGS(status.designated_root));
2451 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2452
2453 ovsrec_bridge_set_status(br->cfg, &smap);
2454 smap_destroy(&smap);
2455 }
2456
2457 static void
2458 port_refresh_stp_status(struct port *port)
2459 {
2460 struct ofproto *ofproto = port->bridge->ofproto;
2461 struct iface *iface;
2462 struct ofproto_port_stp_status status;
2463 struct smap smap;
2464
2465 if (port_is_synthetic(port)) {
2466 return;
2467 }
2468
2469 /* STP doesn't currently support bonds. */
2470 if (!ovs_list_is_singleton(&port->ifaces)) {
2471 ovsrec_port_set_status(port->cfg, NULL);
2472 return;
2473 }
2474
2475 iface = CONTAINER_OF(ovs_list_front(&port->ifaces), struct iface, port_elem);
2476 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2477 return;
2478 }
2479
2480 if (!status.enabled) {
2481 ovsrec_port_set_status(port->cfg, NULL);
2482 return;
2483 }
2484
2485 /* Set Status column. */
2486 smap_init(&smap);
2487 smap_add_format(&smap, "stp_port_id", "%d", status.port_id);
2488 smap_add(&smap, "stp_state", stp_state_name(status.state));
2489 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2490 smap_add(&smap, "stp_role", stp_role_name(status.role));
2491 ovsrec_port_set_status(port->cfg, &smap);
2492 smap_destroy(&smap);
2493 }
2494
2495 static void
2496 port_refresh_stp_stats(struct port *port)
2497 {
2498 struct ofproto *ofproto = port->bridge->ofproto;
2499 struct iface *iface;
2500 struct ofproto_port_stp_stats stats;
2501 const char *keys[3];
2502 int64_t int_values[3];
2503
2504 if (port_is_synthetic(port)) {
2505 return;
2506 }
2507
2508 /* STP doesn't currently support bonds. */
2509 if (!ovs_list_is_singleton(&port->ifaces)) {
2510 return;
2511 }
2512
2513 iface = CONTAINER_OF(ovs_list_front(&port->ifaces), struct iface, port_elem);
2514 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2515 return;
2516 }
2517
2518 if (!stats.enabled) {
2519 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2520 return;
2521 }
2522
2523 /* Set Statistics column. */
2524 keys[0] = "stp_tx_count";
2525 int_values[0] = stats.tx_count;
2526 keys[1] = "stp_rx_count";
2527 int_values[1] = stats.rx_count;
2528 keys[2] = "stp_error_count";
2529 int_values[2] = stats.error_count;
2530
2531 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2532 ARRAY_SIZE(int_values));
2533 }
2534
2535 static void
2536 br_refresh_rstp_status(struct bridge *br)
2537 {
2538 struct smap smap = SMAP_INITIALIZER(&smap);
2539 struct ofproto *ofproto = br->ofproto;
2540 struct ofproto_rstp_status status;
2541
2542 if (ofproto_get_rstp_status(ofproto, &status)) {
2543 return;
2544 }
2545 if (!status.enabled) {
2546 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2547 return;
2548 }
2549 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2550 RSTP_ID_ARGS(status.bridge_id));
2551 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2552 status.root_path_cost);
2553 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2554 RSTP_ID_ARGS(status.root_id));
2555 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2556 RSTP_ID_ARGS(status.designated_id));
2557 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2558 status.designated_port_id);
2559 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2560 status.bridge_port_id);
2561 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2562 smap_destroy(&smap);
2563 }
2564
2565 static void
2566 port_refresh_rstp_status(struct port *port)
2567 {
2568 struct ofproto *ofproto = port->bridge->ofproto;
2569 struct iface *iface;
2570 struct ofproto_port_rstp_status status;
2571 const char *keys[4];
2572 int64_t int_values[4];
2573 struct smap smap;
2574
2575 if (port_is_synthetic(port)) {
2576 return;
2577 }
2578
2579 /* RSTP doesn't currently support bonds. */
2580 if (!ovs_list_is_singleton(&port->ifaces)) {
2581 ovsrec_port_set_rstp_status(port->cfg, NULL);
2582 return;
2583 }
2584
2585 iface = CONTAINER_OF(ovs_list_front(&port->ifaces), struct iface, port_elem);
2586 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2587 return;
2588 }
2589
2590 if (!status.enabled) {
2591 ovsrec_port_set_rstp_status(port->cfg, NULL);
2592 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2593 return;
2594 }
2595 /* Set Status column. */
2596 smap_init(&smap);
2597
2598 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2599 status.port_id);
2600 smap_add_format(&smap, "rstp_port_role", "%s",
2601 rstp_port_role_name(status.role));
2602 smap_add_format(&smap, "rstp_port_state", "%s",
2603 rstp_state_name(status.state));
2604 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2605 RSTP_ID_ARGS(status.designated_bridge_id));
2606 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2607 status.designated_port_id);
2608 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2609 status.designated_path_cost);
2610
2611 ovsrec_port_set_rstp_status(port->cfg, &smap);
2612 smap_destroy(&smap);
2613
2614 /* Set Statistics column. */
2615 keys[0] = "rstp_tx_count";
2616 int_values[0] = status.tx_count;
2617 keys[1] = "rstp_rx_count";
2618 int_values[1] = status.rx_count;
2619 keys[2] = "rstp_uptime";
2620 int_values[2] = status.uptime;
2621 keys[3] = "rstp_error_count";
2622 int_values[3] = status.error_count;
2623 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2624 ARRAY_SIZE(int_values));
2625 }
2626
2627 static void
2628 port_refresh_bond_status(struct port *port, bool force_update)
2629 {
2630 struct eth_addr mac;
2631
2632 /* Return if port is not a bond */
2633 if (ovs_list_is_singleton(&port->ifaces)) {
2634 return;
2635 }
2636
2637 if (bond_get_changed_active_slave(port->name, &mac, force_update)) {
2638 struct ds mac_s;
2639
2640 ds_init(&mac_s);
2641 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2642 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2643 ds_destroy(&mac_s);
2644 }
2645 }
2646
2647 static bool
2648 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2649 {
2650 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2651 }
2652
2653 static void
2654 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2655 {
2656 bool enable = enable_system_stats(cfg);
2657
2658 system_stats_enable(enable);
2659 if (!enable) {
2660 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2661 }
2662 }
2663
2664 static void
2665 run_system_stats(void)
2666 {
2667 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2668 struct smap *stats;
2669
2670 stats = system_stats_run();
2671 if (stats && cfg) {
2672 struct ovsdb_idl_txn *txn;
2673 struct ovsdb_datum datum;
2674
2675 txn = ovsdb_idl_txn_create(idl);
2676 ovsdb_datum_from_smap(&datum, stats);
2677 smap_destroy(stats);
2678 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2679 &datum);
2680 ovsdb_idl_txn_commit(txn);
2681 ovsdb_idl_txn_destroy(txn);
2682
2683 free(stats);
2684 }
2685 }
2686
2687 static const char *
2688 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2689 {
2690 switch (role) {
2691 case OFPCR12_ROLE_EQUAL:
2692 return "other";
2693 case OFPCR12_ROLE_MASTER:
2694 return "master";
2695 case OFPCR12_ROLE_SLAVE:
2696 return "slave";
2697 case OFPCR12_ROLE_NOCHANGE:
2698 default:
2699 return "*** INVALID ROLE ***";
2700 }
2701 }
2702
2703 static void
2704 refresh_controller_status(void)
2705 {
2706 struct bridge *br;
2707
2708 /* Accumulate status for controllers on all bridges. */
2709 HMAP_FOR_EACH (br, node, &all_bridges) {
2710 struct shash info = SHASH_INITIALIZER(&info);
2711 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2712
2713 /* Update each controller of the bridge in the database with
2714 * current status. */
2715 struct ovsrec_controller **controllers;
2716 size_t n_controllers = bridge_get_controllers(br, &controllers);
2717 size_t i;
2718 for (i = 0; i < n_controllers; i++) {
2719 struct ovsrec_controller *cfg = controllers[i];
2720 struct ofproto_controller_info *cinfo =
2721 shash_find_data(&info, cfg->target);
2722
2723 /* cinfo is NULL when 'cfg->target' is a passive connection. */
2724 if (cinfo) {
2725 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2726 const char *role = ofp12_controller_role_to_str(cinfo->role);
2727 ovsrec_controller_set_role(cfg, role);
2728 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2729 } else {
2730 ovsrec_controller_set_is_connected(cfg, false);
2731 ovsrec_controller_set_role(cfg, NULL);
2732 ovsrec_controller_set_status(cfg, NULL);
2733 }
2734 }
2735
2736 ofproto_free_ofproto_controller_info(&info);
2737 }
2738 }
2739 \f
2740 /* Update interface and mirror statistics if necessary. */
2741 static void
2742 run_stats_update(void)
2743 {
2744 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2745 int stats_interval;
2746
2747 if (!cfg) {
2748 return;
2749 }
2750
2751 /* Statistics update interval should always be greater than or equal to
2752 * 5000 ms. */
2753 stats_interval = MAX(smap_get_int(&cfg->other_config,
2754 "stats-update-interval",
2755 5000), 5000);
2756 if (stats_timer_interval != stats_interval) {
2757 stats_timer_interval = stats_interval;
2758 stats_timer = LLONG_MIN;
2759 }
2760
2761 if (time_msec() >= stats_timer) {
2762 enum ovsdb_idl_txn_status status;
2763
2764 /* Rate limit the update. Do not start a new update if the
2765 * previous one is not done. */
2766 if (!stats_txn) {
2767 struct bridge *br;
2768
2769 stats_txn = ovsdb_idl_txn_create(idl);
2770 HMAP_FOR_EACH (br, node, &all_bridges) {
2771 struct port *port;
2772 struct mirror *m;
2773
2774 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2775 struct iface *iface;
2776
2777 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2778 iface_refresh_stats(iface);
2779 }
2780 port_refresh_stp_stats(port);
2781 }
2782 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2783 mirror_refresh_stats(m);
2784 }
2785 }
2786 refresh_controller_status();
2787 }
2788
2789 status = ovsdb_idl_txn_commit(stats_txn);
2790 if (status != TXN_INCOMPLETE) {
2791 stats_timer = time_msec() + stats_timer_interval;
2792 ovsdb_idl_txn_destroy(stats_txn);
2793 stats_txn = NULL;
2794 }
2795 }
2796 }
2797
2798 static void
2799 stats_update_wait(void)
2800 {
2801 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2802 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2803 if (stats_txn) {
2804 ovsdb_idl_txn_wait(stats_txn);
2805 } else {
2806 poll_timer_wait_until(stats_timer);
2807 }
2808 }
2809
2810 /* Update bridge/port/interface status if necessary. */
2811 static void
2812 run_status_update(void)
2813 {
2814 if (!status_txn) {
2815 uint64_t seq;
2816
2817 /* Rate limit the update. Do not start a new update if the
2818 * previous one is not done. */
2819 seq = seq_read(connectivity_seq_get());
2820 if (seq != connectivity_seqno || status_txn_try_again) {
2821 struct bridge *br;
2822
2823 connectivity_seqno = seq;
2824 status_txn = ovsdb_idl_txn_create(idl);
2825 HMAP_FOR_EACH (br, node, &all_bridges) {
2826 struct port *port;
2827
2828 br_refresh_stp_status(br);
2829 br_refresh_rstp_status(br);
2830 br_refresh_datapath_info(br);
2831 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2832 struct iface *iface;
2833
2834 port_refresh_stp_status(port);
2835 port_refresh_rstp_status(port);
2836 port_refresh_bond_status(port, status_txn_try_again);
2837 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2838 iface_refresh_netdev_status(iface);
2839 iface_refresh_ofproto_status(iface);
2840 }
2841 }
2842 }
2843 }
2844 }
2845
2846 /* Commit the transaction and get the status. If the transaction finishes,
2847 * then destroy the transaction. Otherwise, keep it so that we can check
2848 * progress the next time that this function is called. */
2849 if (status_txn) {
2850 enum ovsdb_idl_txn_status status;
2851
2852 status = ovsdb_idl_txn_commit(status_txn);
2853 if (status != TXN_INCOMPLETE) {
2854 ovsdb_idl_txn_destroy(status_txn);
2855 status_txn = NULL;
2856
2857 /* Sets the 'status_txn_try_again' if the transaction fails. */
2858 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2859 status_txn_try_again = false;
2860 } else {
2861 status_txn_try_again = true;
2862 }
2863 }
2864 }
2865
2866 /* Refresh AA port status if necessary. */
2867 if (time_msec() >= aa_refresh_timer) {
2868 struct bridge *br;
2869
2870 HMAP_FOR_EACH (br, node, &all_bridges) {
2871 if (bridge_aa_need_refresh(br)) {
2872 struct ovsdb_idl_txn *txn;
2873
2874 txn = ovsdb_idl_txn_create(idl);
2875 bridge_aa_refresh_queued(br);
2876 ovsdb_idl_txn_commit(txn);
2877 ovsdb_idl_txn_destroy(txn);
2878 }
2879 }
2880
2881 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2882 }
2883 }
2884
2885 static void
2886 status_update_wait(void)
2887 {
2888 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2889 * transaction to complete. If the status update to database needs to be
2890 * run again (transaction fails), registers a timeout in
2891 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2892 * sequence number. */
2893 if (status_txn) {
2894 ovsdb_idl_txn_wait(status_txn);
2895 } else if (status_txn_try_again) {
2896 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2897 } else {
2898 seq_wait(connectivity_seq_get(), connectivity_seqno);
2899 }
2900 }
2901
2902 static void
2903 bridge_run__(void)
2904 {
2905 struct bridge *br;
2906 struct sset types;
2907 const char *type;
2908
2909 /* Let each datapath type do the work that it needs to do. */
2910 sset_init(&types);
2911 ofproto_enumerate_types(&types);
2912 SSET_FOR_EACH (type, &types) {
2913 ofproto_type_run(type);
2914 }
2915 sset_destroy(&types);
2916
2917 /* Let each bridge do the work that it needs to do. */
2918 HMAP_FOR_EACH (br, node, &all_bridges) {
2919 ofproto_run(br->ofproto);
2920 }
2921 }
2922
2923 void
2924 bridge_run(void)
2925 {
2926 static struct ovsrec_open_vswitch null_cfg;
2927 const struct ovsrec_open_vswitch *cfg;
2928
2929 ovsrec_open_vswitch_init(&null_cfg);
2930
2931 ovsdb_idl_run(idl);
2932
2933 if_notifier_run();
2934
2935 if (ovsdb_idl_is_lock_contended(idl)) {
2936 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2937 struct bridge *br, *next_br;
2938
2939 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2940 "disabling this process (pid %ld) until it goes away",
2941 (long int) getpid());
2942
2943 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2944 bridge_destroy(br, false);
2945 }
2946 /* Since we will not be running system_stats_run() in this process
2947 * with the current situation of multiple ovs-vswitchd daemons,
2948 * disable system stats collection. */
2949 system_stats_enable(false);
2950 return;
2951 } else if (!ovsdb_idl_has_lock(idl)
2952 || !ovsdb_idl_has_ever_connected(idl)) {
2953 /* Returns if not holding the lock or not done retrieving db
2954 * contents. */
2955 return;
2956 }
2957 cfg = ovsrec_open_vswitch_first(idl);
2958
2959 if (cfg) {
2960 netdev_set_flow_api_enabled(&cfg->other_config);
2961 dpdk_init(&cfg->other_config);
2962 }
2963
2964 /* Initialize the ofproto library. This only needs to run once, but
2965 * it must be done after the configuration is set. If the
2966 * initialization has already occurred, bridge_init_ofproto()
2967 * returns immediately. */
2968 bridge_init_ofproto(cfg);
2969
2970 /* Once the value of flow-restore-wait is false, we no longer should
2971 * check its value from the database. */
2972 if (cfg && ofproto_get_flow_restore_wait()) {
2973 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2974 "flow-restore-wait", false));
2975 }
2976
2977 bridge_run__();
2978
2979 /* Re-configure SSL. We do this on every trip through the main loop,
2980 * instead of just when the database changes, because the contents of the
2981 * key and certificate files can change without the database changing.
2982 *
2983 * We do this before bridge_reconfigure() because that function might
2984 * initiate SSL connections and thus requires SSL to be configured. */
2985 if (cfg && cfg->ssl) {
2986 const struct ovsrec_ssl *ssl = cfg->ssl;
2987
2988 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2989 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2990 }
2991
2992 if (ovsdb_idl_get_seqno(idl) != idl_seqno ||
2993 if_notifier_changed(ifnotifier)) {
2994 struct ovsdb_idl_txn *txn;
2995
2996 idl_seqno = ovsdb_idl_get_seqno(idl);
2997 txn = ovsdb_idl_txn_create(idl);
2998 bridge_reconfigure(cfg ? cfg : &null_cfg);
2999
3000 if (cfg) {
3001 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
3002 discover_types(cfg);
3003 }
3004
3005 /* If we are completing our initial configuration for this run
3006 * of ovs-vswitchd, then keep the transaction around to monitor
3007 * it for completion. */
3008 if (initial_config_done) {
3009 /* Always sets the 'status_txn_try_again' to check again,
3010 * in case that this transaction fails. */
3011 status_txn_try_again = true;
3012 ovsdb_idl_txn_commit(txn);
3013 ovsdb_idl_txn_destroy(txn);
3014 } else {
3015 initial_config_done = true;
3016 daemonize_txn = txn;
3017 }
3018 }
3019
3020 if (daemonize_txn) {
3021 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
3022 if (status != TXN_INCOMPLETE) {
3023 ovsdb_idl_txn_destroy(daemonize_txn);
3024 daemonize_txn = NULL;
3025
3026 /* ovs-vswitchd has completed initialization, so allow the
3027 * process that forked us to exit successfully. */
3028 daemonize_complete();
3029
3030 vlog_enable_async();
3031
3032 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
3033 }
3034 }
3035
3036 run_stats_update();
3037 run_status_update();
3038 run_system_stats();
3039 }
3040
3041 void
3042 bridge_wait(void)
3043 {
3044 struct sset types;
3045 const char *type;
3046
3047 ovsdb_idl_wait(idl);
3048 if (daemonize_txn) {
3049 ovsdb_idl_txn_wait(daemonize_txn);
3050 }
3051
3052 if_notifier_wait();
3053
3054 sset_init(&types);
3055 ofproto_enumerate_types(&types);
3056 SSET_FOR_EACH (type, &types) {
3057 ofproto_type_wait(type);
3058 }
3059 sset_destroy(&types);
3060
3061 if (!hmap_is_empty(&all_bridges)) {
3062 struct bridge *br;
3063
3064 HMAP_FOR_EACH (br, node, &all_bridges) {
3065 ofproto_wait(br->ofproto);
3066 }
3067 stats_update_wait();
3068 status_update_wait();
3069 }
3070
3071 system_stats_wait();
3072 }
3073
3074 /* Adds some memory usage statistics for bridges into 'usage', for use with
3075 * memory_report(). */
3076 void
3077 bridge_get_memory_usage(struct simap *usage)
3078 {
3079 struct bridge *br;
3080 struct sset types;
3081 const char *type;
3082
3083 sset_init(&types);
3084 ofproto_enumerate_types(&types);
3085 SSET_FOR_EACH (type, &types) {
3086 ofproto_type_get_memory_usage(type, usage);
3087 }
3088 sset_destroy(&types);
3089
3090 HMAP_FOR_EACH (br, node, &all_bridges) {
3091 ofproto_get_memory_usage(br->ofproto, usage);
3092 }
3093 }
3094 \f
3095 /* QoS unixctl user interface functions. */
3096
3097 struct qos_unixctl_show_cbdata {
3098 struct ds *ds;
3099 struct iface *iface;
3100 };
3101
3102 static void
3103 qos_unixctl_show_queue(unsigned int queue_id,
3104 const struct smap *details,
3105 struct iface *iface,
3106 struct ds *ds)
3107 {
3108 struct netdev_queue_stats stats;
3109 struct smap_node *node;
3110 int error;
3111
3112 ds_put_cstr(ds, "\n");
3113 if (queue_id) {
3114 ds_put_format(ds, "Queue %u:\n", queue_id);
3115 } else {
3116 ds_put_cstr(ds, "Default:\n");
3117 }
3118
3119 SMAP_FOR_EACH (node, details) {
3120 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3121 }
3122
3123 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3124 if (!error) {
3125 if (stats.tx_packets != UINT64_MAX) {
3126 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3127 }
3128
3129 if (stats.tx_bytes != UINT64_MAX) {
3130 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3131 }
3132
3133 if (stats.tx_errors != UINT64_MAX) {
3134 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3135 }
3136 } else {
3137 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3138 queue_id, ovs_strerror(error));
3139 }
3140 }
3141
3142 static void
3143 qos_unixctl_show_types(struct unixctl_conn *conn, int argc OVS_UNUSED,
3144 const char *argv[], void *aux OVS_UNUSED)
3145 {
3146 struct ds ds = DS_EMPTY_INITIALIZER;
3147 struct sset types = SSET_INITIALIZER(&types);
3148 struct iface *iface;
3149 const char * types_name;
3150 int error;
3151
3152 iface = iface_find(argv[1]);
3153 if (!iface) {
3154 unixctl_command_reply_error(conn, "no such interface");
3155 return;
3156 }
3157
3158 error = netdev_get_qos_types(iface->netdev, &types);
3159 if (!error) {
3160 if (!sset_is_empty(&types)) {
3161 SSET_FOR_EACH (types_name, &types) {
3162 ds_put_format(&ds, "QoS type: %s\n", types_name);
3163 }
3164 unixctl_command_reply(conn, ds_cstr(&ds));
3165 } else {
3166 ds_put_format(&ds, "No QoS types supported for interface: %s\n",
3167 iface->name);
3168 unixctl_command_reply(conn, ds_cstr(&ds));
3169 }
3170 } else {
3171 ds_put_format(&ds, "%s: failed to retrieve supported QoS types (%s)",
3172 iface->name, ovs_strerror(error));
3173 unixctl_command_reply_error(conn, ds_cstr(&ds));
3174 }
3175
3176 sset_destroy(&types);
3177 ds_destroy(&ds);
3178 }
3179
3180 static void
3181 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3182 const char *argv[], void *aux OVS_UNUSED)
3183 {
3184 struct ds ds = DS_EMPTY_INITIALIZER;
3185 struct smap smap = SMAP_INITIALIZER(&smap);
3186 struct iface *iface;
3187 const char *type;
3188 struct smap_node *node;
3189 int error;
3190
3191 iface = iface_find(argv[1]);
3192 if (!iface) {
3193 unixctl_command_reply_error(conn, "no such interface");
3194 return;
3195 }
3196
3197 error = netdev_get_qos(iface->netdev, &type, &smap);
3198 if (!error) {
3199 if (*type != '\0') {
3200 struct netdev_queue_dump dump;
3201 struct smap details;
3202 unsigned int queue_id;
3203
3204 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3205
3206 SMAP_FOR_EACH (node, &smap) {
3207 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3208 }
3209
3210 smap_init(&details);
3211 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3212 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3213 }
3214 smap_destroy(&details);
3215
3216 unixctl_command_reply(conn, ds_cstr(&ds));
3217 } else {
3218 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3219 unixctl_command_reply(conn, ds_cstr(&ds));
3220 }
3221 } else {
3222 ds_put_format(&ds, "%s: failed to retrieve QOS configuration (%s)\n",
3223 iface->name, ovs_strerror(error));
3224 unixctl_command_reply_error(conn, ds_cstr(&ds));
3225 }
3226
3227 smap_destroy(&smap);
3228 ds_destroy(&ds);
3229 }
3230 \f
3231 /* Bridge reconfiguration functions. */
3232 static void
3233 bridge_create(const struct ovsrec_bridge *br_cfg)
3234 {
3235 struct bridge *br;
3236
3237 ovs_assert(!bridge_lookup(br_cfg->name));
3238 br = xzalloc(sizeof *br);
3239
3240 br->name = xstrdup(br_cfg->name);
3241 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3242 br->cfg = br_cfg;
3243
3244 /* Derive the default Ethernet address from the bridge's UUID. This should
3245 * be unique and it will be stable between ovs-vswitchd runs. */
3246 memcpy(&br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3247 eth_addr_mark_random(&br->default_ea);
3248
3249 hmap_init(&br->ports);
3250 hmap_init(&br->ifaces);
3251 hmap_init(&br->iface_by_name);
3252 hmap_init(&br->mirrors);
3253
3254 hmap_init(&br->mappings);
3255 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3256 }
3257
3258 static void
3259 bridge_destroy(struct bridge *br, bool del)
3260 {
3261 if (br) {
3262 struct mirror *mirror, *next_mirror;
3263 struct port *port, *next_port;
3264
3265 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3266 port_destroy(port);
3267 }
3268 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3269 mirror_destroy(mirror);
3270 }
3271
3272 hmap_remove(&all_bridges, &br->node);
3273 ofproto_destroy(br->ofproto, del);
3274 hmap_destroy(&br->ifaces);
3275 hmap_destroy(&br->ports);
3276 hmap_destroy(&br->iface_by_name);
3277 hmap_destroy(&br->mirrors);
3278 hmap_destroy(&br->mappings);
3279 free(br->name);
3280 free(br->type);
3281 free(br);
3282 }
3283 }
3284
3285 static struct bridge *
3286 bridge_lookup(const char *name)
3287 {
3288 struct bridge *br;
3289
3290 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3291 if (!strcmp(br->name, name)) {
3292 return br;
3293 }
3294 }
3295 return NULL;
3296 }
3297
3298 /* Handle requests for a listing of all flows known by the OpenFlow
3299 * stack, including those normally hidden. */
3300 static void
3301 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3302 const char *argv[], void *aux OVS_UNUSED)
3303 {
3304 struct bridge *br;
3305 struct ds results;
3306
3307 br = bridge_lookup(argv[1]);
3308 if (!br) {
3309 unixctl_command_reply_error(conn, "Unknown bridge");
3310 return;
3311 }
3312
3313 ds_init(&results);
3314 ofproto_get_all_flows(br->ofproto, &results);
3315
3316 unixctl_command_reply(conn, ds_cstr(&results));
3317 ds_destroy(&results);
3318 }
3319
3320 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3321 * connections and reconnect. If BRIDGE is not specified, then all bridges
3322 * drop their controller connections and reconnect. */
3323 static void
3324 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3325 const char *argv[], void *aux OVS_UNUSED)
3326 {
3327 struct bridge *br;
3328 if (argc > 1) {
3329 br = bridge_lookup(argv[1]);
3330 if (!br) {
3331 unixctl_command_reply_error(conn, "Unknown bridge");
3332 return;
3333 }
3334 ofproto_reconnect_controllers(br->ofproto);
3335 } else {
3336 HMAP_FOR_EACH (br, node, &all_bridges) {
3337 ofproto_reconnect_controllers(br->ofproto);
3338 }
3339 }
3340 unixctl_command_reply(conn, NULL);
3341 }
3342
3343 static size_t
3344 bridge_get_controllers(const struct bridge *br,
3345 struct ovsrec_controller ***controllersp)
3346 {
3347 struct ovsrec_controller **controllers;
3348 size_t n_controllers;
3349
3350 controllers = br->cfg->controller;
3351 n_controllers = br->cfg->n_controller;
3352
3353 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3354 controllers = NULL;
3355 n_controllers = 0;
3356 }
3357
3358 if (controllersp) {
3359 *controllersp = controllers;
3360 }
3361 return n_controllers;
3362 }
3363
3364 static void
3365 bridge_collect_wanted_ports(struct bridge *br,
3366 struct shash *wanted_ports)
3367 {
3368 size_t i;
3369
3370 shash_init(wanted_ports);
3371
3372 for (i = 0; i < br->cfg->n_ports; i++) {
3373 const char *name = br->cfg->ports[i]->name;
3374 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3375 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3376 br->name, name);
3377 }
3378 }
3379 if (bridge_get_controllers(br, NULL)
3380 && !shash_find(wanted_ports, br->name)) {
3381 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3382 br->name, br->name);
3383
3384 ovsrec_interface_init(&br->synth_local_iface);
3385 ovsrec_port_init(&br->synth_local_port);
3386
3387 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3388 br->synth_local_port.n_interfaces = 1;
3389 br->synth_local_port.name = br->name;
3390
3391 br->synth_local_iface.name = br->name;
3392 br->synth_local_iface.type = "internal";
3393
3394 br->synth_local_ifacep = &br->synth_local_iface;
3395
3396 shash_add(wanted_ports, br->name, &br->synth_local_port);
3397 }
3398 }
3399
3400 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3401 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3402 * 'br' needs to complete its configuration. */
3403 static void
3404 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3405 {
3406 struct shash_node *port_node;
3407 struct port *port, *next;
3408
3409 /* Get rid of deleted ports.
3410 * Get rid of deleted interfaces on ports that still exist. */
3411 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3412 port->cfg = shash_find_data(wanted_ports, port->name);
3413 if (!port->cfg) {
3414 port_destroy(port);
3415 } else {
3416 port_del_ifaces(port);
3417 }
3418 }
3419
3420 /* Update iface->cfg and iface->type in interfaces that still exist. */
3421 SHASH_FOR_EACH (port_node, wanted_ports) {
3422 const struct ovsrec_port *port_rec = port_node->data;
3423 size_t i;
3424
3425 for (i = 0; i < port_rec->n_interfaces; i++) {
3426 const struct ovsrec_interface *cfg = port_rec->interfaces[i];
3427 struct iface *iface = iface_lookup(br, cfg->name);
3428 const char *type = iface_get_type(cfg, br->cfg);
3429 const char *dp_type = br->cfg->datapath_type;
3430 const char *netdev_type = ofproto_port_open_type(dp_type, type);
3431
3432 if (iface) {
3433 iface->cfg = cfg;
3434 iface->type = type;
3435 iface->netdev_type = netdev_type;
3436 } else if (!strcmp(type, "null")) {
3437 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3438 " may be removed in February 2013. Please email"
3439 " dev@openvswitch.org with concerns.",
3440 cfg->name);
3441 } else {
3442 /* We will add new interfaces later. */
3443 }
3444 }
3445 }
3446 }
3447
3448 /* Initializes 'oc' appropriately as a management service controller for
3449 * 'br'.
3450 *
3451 * The caller must free oc->target when it is no longer needed. */
3452 static void
3453 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3454 struct ofproto_controller *oc)
3455 {
3456 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3457 oc->max_backoff = 0;
3458 oc->probe_interval = 60;
3459 oc->band = OFPROTO_OUT_OF_BAND;
3460 oc->rate_limit = 0;
3461 oc->burst_limit = 0;
3462 oc->enable_async_msgs = true;
3463 oc->dscp = 0;
3464 }
3465
3466 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3467 static void
3468 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3469 struct ofproto_controller *oc)
3470 {
3471 int dscp;
3472
3473 oc->target = c->target;
3474 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3475 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3476 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3477 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3478 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3479 oc->burst_limit = (c->controller_burst_limit
3480 ? *c->controller_burst_limit : 0);
3481 oc->enable_async_msgs = (!c->enable_async_messages
3482 || *c->enable_async_messages);
3483 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3484 if (dscp < 0 || dscp > 63) {
3485 dscp = DSCP_DEFAULT;
3486 }
3487 oc->dscp = dscp;
3488 }
3489
3490 /* Configures the IP stack for 'br''s local interface properly according to the
3491 * configuration in 'c'. */
3492 static void
3493 bridge_configure_local_iface_netdev(struct bridge *br,
3494 struct ovsrec_controller *c)
3495 {
3496 struct netdev *netdev;
3497 struct in_addr mask, gateway;
3498
3499 struct iface *local_iface;
3500 struct in_addr ip;
3501
3502 /* If there's no local interface or no IP address, give up. */
3503 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3504 if (!local_iface || !c->local_ip || !ip_parse(c->local_ip, &ip.s_addr)) {
3505 return;
3506 }
3507
3508 /* Bring up the local interface. */
3509 netdev = local_iface->netdev;
3510 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3511
3512 /* Configure the IP address and netmask. */
3513 if (!c->local_netmask
3514 || !ip_parse(c->local_netmask, &mask.s_addr)
3515 || !mask.s_addr) {
3516 mask.s_addr = guess_netmask(ip.s_addr);
3517 }
3518 if (!netdev_set_in4(netdev, ip, mask)) {
3519 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3520 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3521 }
3522
3523 /* Configure the default gateway. */
3524 if (c->local_gateway
3525 && ip_parse(c->local_gateway, &gateway.s_addr)
3526 && gateway.s_addr) {
3527 if (!netdev_add_router(netdev, gateway)) {
3528 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3529 br->name, IP_ARGS(gateway.s_addr));
3530 }
3531 }
3532 }
3533
3534 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3535 * in either string are treated as equal to any number of slashes in the other,
3536 * e.g. "x///y" is equal to "x/y".
3537 *
3538 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3539 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3540 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3541 * 'b' against a prefix of 'a'.
3542 */
3543 static bool
3544 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3545 {
3546 const char *b_start = b;
3547 for (;;) {
3548 if (b - b_start >= b_stoplen) {
3549 return true;
3550 } else if (*a != *b) {
3551 return false;
3552 } else if (*a == '/') {
3553 a += strspn(a, "/");
3554 b += strspn(b, "/");
3555 } else if (*a == '\0') {
3556 return true;
3557 } else {
3558 a++;
3559 b++;
3560 }
3561 }
3562 }
3563
3564 static void
3565 bridge_configure_remotes(struct bridge *br,
3566 const struct sockaddr_in *managers, size_t n_managers)
3567 {
3568 bool disable_in_band;
3569
3570 struct ovsrec_controller **controllers;
3571 size_t n_controllers;
3572
3573 enum ofproto_fail_mode fail_mode;
3574
3575 struct ofproto_controller *ocs;
3576 size_t n_ocs;
3577 size_t i;
3578
3579 /* Check if we should disable in-band control on this bridge. */
3580 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3581 false);
3582
3583 /* Set OpenFlow queue ID for in-band control. */
3584 ofproto_set_in_band_queue(br->ofproto,
3585 smap_get_int(&br->cfg->other_config,
3586 "in-band-queue", -1));
3587
3588 if (disable_in_band) {
3589 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3590 } else {
3591 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3592 }
3593
3594 n_controllers = bridge_get_controllers(br, &controllers);
3595
3596 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3597 n_ocs = 0;
3598
3599 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3600 for (i = 0; i < n_controllers; i++) {
3601 struct ovsrec_controller *c = controllers[i];
3602
3603 if (daemon_should_self_confine()
3604 && (!strncmp(c->target, "punix:", 6)
3605 || !strncmp(c->target, "unix:", 5))) {
3606 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3607 char *whitelist;
3608
3609 if (!strncmp(c->target, "unix:", 5)) {
3610 /* Connect to a listening socket */
3611 whitelist = xasprintf("unix:%s/", ovs_rundir());
3612 if (strchr(c->target, '/') &&
3613 !equal_pathnames(c->target, whitelist,
3614 strlen(whitelist))) {
3615 /* Absolute path specified, but not in ovs_rundir */
3616 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3617 "controller \"%s\" due to possibility for "
3618 "remote exploit. Instead, specify socket "
3619 "in whitelisted \"%s\" or connect to "
3620 "\"unix:%s/%s.mgmt\" (which is always "
3621 "available without special configuration).",
3622 br->name, c->target, whitelist,
3623 ovs_rundir(), br->name);
3624 free(whitelist);
3625 continue;
3626 }
3627 } else {
3628 whitelist = xasprintf("punix:%s/%s.",
3629 ovs_rundir(), br->name);
3630 if (!equal_pathnames(c->target, whitelist, strlen(whitelist))
3631 || strchr(c->target + strlen(whitelist), '/')) {
3632 /* Prevent remote ovsdb-server users from accessing
3633 * arbitrary Unix domain sockets and overwriting arbitrary
3634 * local files. */
3635 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3636 "controller \"%s\" due to possibility of "
3637 "overwriting local files. Instead, specify "
3638 "path in whitelisted format \"%s*\" or "
3639 "connect to \"unix:%s/%s.mgmt\" (which is "
3640 "always available without special "
3641 "configuration).",
3642 br->name, c->target, whitelist,
3643 ovs_rundir(), br->name);
3644 free(whitelist);
3645 continue;
3646 }
3647 }
3648
3649 free(whitelist);
3650 }
3651
3652 bridge_configure_local_iface_netdev(br, c);
3653 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3654 if (disable_in_band) {
3655 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3656 }
3657 n_ocs++;
3658 }
3659
3660 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3661 bridge_get_allowed_versions(br));
3662 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3663 free(ocs);
3664
3665 /* Set the fail-mode. */
3666 fail_mode = !br->cfg->fail_mode
3667 || !strcmp(br->cfg->fail_mode, "standalone")
3668 ? OFPROTO_FAIL_STANDALONE
3669 : OFPROTO_FAIL_SECURE;
3670 ofproto_set_fail_mode(br->ofproto, fail_mode);
3671
3672 /* Configure OpenFlow controller connection snooping. */
3673 if (!ofproto_has_snoops(br->ofproto)) {
3674 struct sset snoops;
3675
3676 sset_init(&snoops);
3677 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3678 ovs_rundir(), br->name));
3679 ofproto_set_snoops(br->ofproto, &snoops);
3680 sset_destroy(&snoops);
3681 }
3682 }
3683
3684 static void
3685 bridge_configure_tables(struct bridge *br)
3686 {
3687 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3688 int n_tables;
3689 int i, j;
3690
3691 n_tables = ofproto_get_n_tables(br->ofproto);
3692 j = 0;
3693 for (i = 0; i < n_tables; i++) {
3694 struct ofproto_table_settings s;
3695 bool use_default_prefixes = true;
3696
3697 s.name = NULL;
3698 s.max_flows = UINT_MAX;
3699 s.groups = NULL;
3700 s.enable_eviction = false;
3701 s.n_groups = 0;
3702 s.n_prefix_fields = 0;
3703 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3704
3705 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3706 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3707
3708 s.name = cfg->name;
3709 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3710 s.max_flows = *cfg->flow_limit;
3711 }
3712
3713 s.enable_eviction = (cfg->overflow_policy
3714 && !strcmp(cfg->overflow_policy, "evict"));
3715 if (cfg->n_groups) {
3716 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3717 for (int k = 0; k < cfg->n_groups; k++) {
3718 const char *string = cfg->groups[k];
3719 char *msg;
3720
3721 msg = mf_parse_subfield__(&s.groups[k], &string);
3722 if (msg) {
3723 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3724 "'groups' (%s)", br->name, i, msg);
3725 free(msg);
3726 } else if (*string) {
3727 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3728 "element '%s' contains trailing garbage",
3729 br->name, i, cfg->groups[k]);
3730 } else {
3731 s.n_groups++;
3732 }
3733 }
3734 }
3735
3736 /* Prefix lookup fields. */
3737 s.n_prefix_fields = 0;
3738 for (int k = 0; k < cfg->n_prefixes; k++) {
3739 const char *name = cfg->prefixes[k];
3740 const struct mf_field *mf;
3741
3742 if (strcmp(name, "none") == 0) {
3743 use_default_prefixes = false;
3744 s.n_prefix_fields = 0;
3745 break;
3746 }
3747 mf = mf_from_name(name);
3748 if (!mf) {
3749 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3750 br->name, name);
3751 continue;
3752 }
3753 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3754 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3755 "%s", br->name, name);
3756 continue;
3757 }
3758 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3759 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3760 "field not used: %s", br->name, name);
3761 continue;
3762 }
3763 use_default_prefixes = false;
3764 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3765 }
3766 }
3767 if (use_default_prefixes) {
3768 /* Use default values. */
3769 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3770 memcpy(s.prefix_fields, default_prefix_fields,
3771 sizeof default_prefix_fields);
3772 } else {
3773 struct ds ds = DS_EMPTY_INITIALIZER;
3774 for (int k = 0; k < s.n_prefix_fields; k++) {
3775 if (k) {
3776 ds_put_char(&ds, ',');
3777 }
3778 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3779 }
3780 if (s.n_prefix_fields == 0) {
3781 ds_put_cstr(&ds, "none");
3782 }
3783 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3784 br->name, i, ds_cstr(&ds));
3785 ds_destroy(&ds);
3786 }
3787
3788 ofproto_configure_table(br->ofproto, i, &s);
3789
3790 free(s.groups);
3791 }
3792 for (; j < br->cfg->n_flow_tables; j++) {
3793 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3794 "%"PRId64" not supported by this datapath", br->name,
3795 br->cfg->key_flow_tables[j]);
3796 }
3797 }
3798
3799 static void
3800 bridge_configure_dp_desc(struct bridge *br)
3801 {
3802 ofproto_set_dp_desc(br->ofproto,
3803 smap_get(&br->cfg->other_config, "dp-desc"));
3804 }
3805
3806 static struct aa_mapping *
3807 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3808 {
3809 struct aa_mapping *m;
3810
3811 HMAP_FOR_EACH_IN_BUCKET (m,
3812 hmap_node,
3813 hash_bytes(&isid, sizeof isid, 0),
3814 &br->mappings) {
3815 if (isid == m->isid) {
3816 return m;
3817 }
3818 }
3819 return NULL;
3820 }
3821
3822 static struct aa_mapping *
3823 bridge_aa_mapping_create(struct bridge *br,
3824 const int64_t isid,
3825 const int64_t vlan)
3826 {
3827 struct aa_mapping *m;
3828
3829 m = xzalloc(sizeof *m);
3830 m->bridge = br;
3831 m->isid = isid;
3832 m->vlan = vlan;
3833 m->br_name = xstrdup(br->name);
3834 hmap_insert(&br->mappings,
3835 &m->hmap_node,
3836 hash_bytes(&isid, sizeof isid, 0));
3837
3838 return m;
3839 }
3840
3841 static void
3842 bridge_aa_mapping_destroy(struct aa_mapping *m)
3843 {
3844 if (m) {
3845 struct bridge *br = m->bridge;
3846
3847 if (br->ofproto) {
3848 ofproto_aa_mapping_unregister(br->ofproto, m);
3849 }
3850
3851 hmap_remove(&br->mappings, &m->hmap_node);
3852 if (m->br_name) {
3853 free(m->br_name);
3854 }
3855 free(m);
3856 }
3857 }
3858
3859 static bool
3860 bridge_aa_mapping_configure(struct aa_mapping *m)
3861 {
3862 struct aa_mapping_settings s;
3863
3864 s.isid = m->isid;
3865 s.vlan = m->vlan;
3866
3867 /* Configure. */
3868 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3869
3870 return true;
3871 }
3872
3873 static void
3874 bridge_configure_aa(struct bridge *br)
3875 {
3876 const struct ovsdb_datum *mc;
3877 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3878 struct aa_settings aa_s;
3879 struct aa_mapping *m, *next;
3880 size_t i;
3881
3882 if (!auto_attach) {
3883 ofproto_set_aa(br->ofproto, NULL, NULL);
3884 return;
3885 }
3886
3887 memset(&aa_s, 0, sizeof aa_s);
3888 aa_s.system_description = auto_attach->system_description;
3889 aa_s.system_name = auto_attach->system_name;
3890 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3891
3892 mc = ovsrec_autoattach_get_mappings(auto_attach,
3893 OVSDB_TYPE_INTEGER,
3894 OVSDB_TYPE_INTEGER);
3895 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3896 union ovsdb_atom atom;
3897
3898 atom.integer = m->isid;
3899 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_INTEGER) == UINT_MAX) {
3900 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3901 m->isid, m->vlan);
3902 bridge_aa_mapping_destroy(m);
3903 }
3904 }
3905
3906 /* Add new mappings and reconfigure existing ones. */
3907 for (i = 0; i < auto_attach->n_mappings; ++i) {
3908 m = bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3909
3910 if (!m) {
3911 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3912 auto_attach->key_mappings[i],
3913 auto_attach->value_mappings[i]);
3914 m = bridge_aa_mapping_create(br,
3915 auto_attach->key_mappings[i],
3916 auto_attach->value_mappings[i]);
3917
3918 if (!bridge_aa_mapping_configure(m)) {
3919 bridge_aa_mapping_destroy(m);
3920 }
3921 }
3922 }
3923 }
3924
3925 static bool
3926 bridge_aa_need_refresh(struct bridge *br)
3927 {
3928 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3929 }
3930
3931 static void
3932 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3933 {
3934 int64_t *trunks = NULL;
3935 unsigned int i = 0;
3936 bool found = false, reconfigure = false;
3937
3938 for (i = 0; i < port->cfg->n_trunks; i++) {
3939 if (port->cfg->trunks[i] == m->vlan) {
3940 found = true;
3941 break;
3942 }
3943 }
3944
3945 switch (m->oper) {
3946 case BRIDGE_AA_VLAN_OPER_ADD:
3947 if (!found) {
3948 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3949
3950 for (i = 0; i < port->cfg->n_trunks; i++) {
3951 trunks[i] = port->cfg->trunks[i];
3952 }
3953 trunks[i++] = m->vlan;
3954 reconfigure = true;
3955 }
3956
3957 break;
3958
3959 case BRIDGE_AA_VLAN_OPER_REMOVE:
3960 if (found) {
3961 unsigned int j = 0;
3962
3963 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3964
3965 for (i = 0; i < port->cfg->n_trunks; i++) {
3966 if (port->cfg->trunks[i] != m->vlan) {
3967 trunks[j++] = port->cfg->trunks[i];
3968 }
3969 }
3970 i = j;
3971 reconfigure = true;
3972 }
3973
3974 break;
3975
3976 case BRIDGE_AA_VLAN_OPER_UNDEF:
3977 default:
3978 VLOG_WARN("unrecognized operation %u", m->oper);
3979 break;
3980 }
3981
3982 if (reconfigure) {
3983 /* VLAN switching under trunk mode cause the trunk port to switch all
3984 * VLANs, see ovs-vswitchd.conf.db
3985 */
3986 if (i == 0) {
3987 static char *vlan_mode_access = "access";
3988 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3989 }
3990
3991 if (i == 1) {
3992 static char *vlan_mode_trunk = "trunk";
3993 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3994 }
3995
3996 ovsrec_port_set_trunks(port->cfg, trunks, i);
3997
3998 /* Force reconfigure of the port. */
3999 port_configure(port);
4000 }
4001
4002 free(trunks);
4003 }
4004
4005 static void
4006 bridge_aa_refresh_queued(struct bridge *br)
4007 {
4008 struct ovs_list *list = xmalloc(sizeof *list);
4009 struct bridge_aa_vlan *node, *next;
4010
4011 ovs_list_init(list);
4012 ofproto_aa_vlan_get_queued(br->ofproto, list);
4013
4014 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
4015 struct port *port;
4016
4017 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
4018 node->oper);
4019
4020 port = port_lookup(br, node->port_name);
4021 if (port) {
4022 bridge_aa_update_trunks(port, node);
4023 }
4024
4025 ovs_list_remove(&node->list_node);
4026 free(node->port_name);
4027 free(node);
4028 }
4029
4030 free(list);
4031 }
4032
4033 \f
4034 /* Port functions. */
4035
4036 static struct port *
4037 port_create(struct bridge *br, const struct ovsrec_port *cfg)
4038 {
4039 struct port *port;
4040
4041 port = xzalloc(sizeof *port);
4042 port->bridge = br;
4043 port->name = xstrdup(cfg->name);
4044 port->cfg = cfg;
4045 ovs_list_init(&port->ifaces);
4046
4047 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
4048 return port;
4049 }
4050
4051 /* Deletes interfaces from 'port' that are no longer configured for it. */
4052 static void
4053 port_del_ifaces(struct port *port)
4054 {
4055 struct iface *iface, *next;
4056 struct sset new_ifaces;
4057 size_t i;
4058
4059 /* Collect list of new interfaces. */
4060 sset_init(&new_ifaces);
4061 for (i = 0; i < port->cfg->n_interfaces; i++) {
4062 const char *name = port->cfg->interfaces[i]->name;
4063 const char *type = port->cfg->interfaces[i]->type;
4064 if (strcmp(type, "null")) {
4065 sset_add(&new_ifaces, name);
4066 }
4067 }
4068
4069 /* Get rid of deleted interfaces. */
4070 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4071 if (!sset_contains(&new_ifaces, iface->name)) {
4072 iface_destroy(iface);
4073 }
4074 }
4075
4076 sset_destroy(&new_ifaces);
4077 }
4078
4079 static void
4080 port_destroy(struct port *port)
4081 {
4082 if (port) {
4083 struct bridge *br = port->bridge;
4084 struct iface *iface, *next;
4085
4086 if (br->ofproto) {
4087 ofproto_bundle_unregister(br->ofproto, port);
4088 }
4089
4090 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4091 iface_destroy__(iface);
4092 }
4093
4094 hmap_remove(&br->ports, &port->hmap_node);
4095 free(port->name);
4096 free(port);
4097 }
4098 }
4099
4100 static struct port *
4101 port_lookup(const struct bridge *br, const char *name)
4102 {
4103 struct port *port;
4104
4105 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4106 &br->ports) {
4107 if (!strcmp(port->name, name)) {
4108 return port;
4109 }
4110 }
4111 return NULL;
4112 }
4113
4114 static bool
4115 enable_lacp(struct port *port, bool *activep)
4116 {
4117 if (!port->cfg->lacp) {
4118 /* XXX when LACP implementation has been sufficiently tested, enable by
4119 * default and make active on bonded ports. */
4120 return false;
4121 } else if (!strcmp(port->cfg->lacp, "off")) {
4122 return false;
4123 } else if (!strcmp(port->cfg->lacp, "active")) {
4124 *activep = true;
4125 return true;
4126 } else if (!strcmp(port->cfg->lacp, "passive")) {
4127 *activep = false;
4128 return true;
4129 } else {
4130 VLOG_WARN("port %s: unknown LACP mode %s",
4131 port->name, port->cfg->lacp);
4132 return false;
4133 }
4134 }
4135
4136 static struct lacp_settings *
4137 port_configure_lacp(struct port *port, struct lacp_settings *s)
4138 {
4139 const char *lacp_time, *system_id;
4140 int priority;
4141
4142 if (!enable_lacp(port, &s->active)) {
4143 return NULL;
4144 }
4145
4146 s->name = port->name;
4147
4148 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4149 if (system_id) {
4150 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4151 ETH_ADDR_SCAN_ARGS(s->id))) {
4152 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4153 " address.", port->name, system_id);
4154 return NULL;
4155 }
4156 } else {
4157 s->id = port->bridge->ea;
4158 }
4159
4160 if (eth_addr_is_zero(s->id)) {
4161 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4162 return NULL;
4163 }
4164
4165 /* Prefer bondable links if unspecified. */
4166 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4167 0);
4168 s->priority = (priority > 0 && priority <= UINT16_MAX
4169 ? priority
4170 : UINT16_MAX - !ovs_list_is_short(&port->ifaces));
4171
4172 lacp_time = smap_get_def(&port->cfg->other_config, "lacp-time", "");
4173 s->fast = !strcasecmp(lacp_time, "fast");
4174
4175 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4176 "lacp-fallback-ab", false);
4177
4178 return s;
4179 }
4180
4181 static void
4182 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4183 {
4184 int priority, portid, key;
4185
4186 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4187 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4188 0);
4189 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4190
4191 if (portid <= 0 || portid > UINT16_MAX) {
4192 portid = ofp_to_u16(iface->ofp_port);
4193 }
4194
4195 if (priority <= 0 || priority > UINT16_MAX) {
4196 priority = UINT16_MAX;
4197 }
4198
4199 if (key < 0 || key > UINT16_MAX) {
4200 key = 0;
4201 }
4202
4203 s->name = iface->name;
4204 s->id = portid;
4205 s->priority = priority;
4206 s->key = key;
4207 }
4208
4209 static void
4210 port_configure_bond(struct port *port, struct bond_settings *s)
4211 {
4212 const char *detect_s;
4213 struct iface *iface;
4214 const char *mac_s;
4215 int miimon_interval;
4216
4217 s->name = port->name;
4218 s->balance = BM_AB;
4219 if (port->cfg->bond_mode) {
4220 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4221 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4222 port->name, port->cfg->bond_mode,
4223 bond_mode_to_string(s->balance));
4224 }
4225 } else {
4226 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4227
4228 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4229 * active-backup. At some point we should remove this warning. */
4230 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4231 " in previous versions, the default bond_mode was"
4232 " balance-slb", port->name,
4233 bond_mode_to_string(s->balance));
4234 }
4235 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4236 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4237 "please use another bond type or disable flood_vlans",
4238 port->name);
4239 }
4240
4241 miimon_interval = smap_get_int(&port->cfg->other_config,
4242 "bond-miimon-interval", 0);
4243 if (miimon_interval <= 0) {
4244 miimon_interval = 200;
4245 }
4246
4247 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4248 if (!detect_s || !strcmp(detect_s, "carrier")) {
4249 miimon_interval = 0;
4250 } else if (strcmp(detect_s, "miimon")) {
4251 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4252 "defaulting to carrier", port->name, detect_s);
4253 miimon_interval = 0;
4254 }
4255
4256 s->up_delay = MAX(0, port->cfg->bond_updelay);
4257 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4258 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4259 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4260 "bond-rebalance-interval", 10000);
4261 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4262 s->rebalance_interval = 1000;
4263 }
4264
4265 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4266 "lacp-fallback-ab", false);
4267
4268 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4269 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4270 }
4271
4272 mac_s = port->cfg->bond_active_slave;
4273 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4274 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4275 /* OVSDB did not store the last active interface */
4276 s->active_slave_mac = eth_addr_zero;
4277 }
4278 }
4279
4280 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4281 * instead of obtaining it from the database. */
4282 static bool
4283 port_is_synthetic(const struct port *port)
4284 {
4285 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4286 }
4287 \f
4288 /* Interface functions. */
4289
4290 static bool
4291 iface_is_internal(const struct ovsrec_interface *iface,
4292 const struct ovsrec_bridge *br)
4293 {
4294 /* The local port and "internal" ports are always "internal". */
4295 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4296 }
4297
4298 /* Returns the correct network device type for interface 'iface' in bridge
4299 * 'br'. */
4300 static const char *
4301 iface_get_type(const struct ovsrec_interface *iface,
4302 const struct ovsrec_bridge *br)
4303 {
4304 const char *type;
4305
4306 /* The local port always has type "internal". Other ports take
4307 * their type from the database and default to "system" if none is
4308 * specified. */
4309 if (iface_is_internal(iface, br)) {
4310 type = "internal";
4311 } else {
4312 type = iface->type[0] ? iface->type : "system";
4313 }
4314
4315 return type;
4316 }
4317
4318 static void
4319 iface_destroy__(struct iface *iface)
4320 {
4321 if (iface) {
4322 struct port *port = iface->port;
4323 struct bridge *br = port->bridge;
4324
4325 VLOG_INFO("bridge %s: deleted interface %s on port %d",
4326 br->name, iface->name, iface->ofp_port);
4327
4328 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4329 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4330 }
4331
4332 if (iface->ofp_port != OFPP_NONE) {
4333 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4334 }
4335
4336 ovs_list_remove(&iface->port_elem);
4337 hmap_remove(&br->iface_by_name, &iface->name_node);
4338
4339 /* The user is changing configuration here, so netdev_remove needs to be
4340 * used as opposed to netdev_close */
4341 netdev_remove(iface->netdev);
4342
4343 free(iface->name);
4344 free(iface);
4345 }
4346 }
4347
4348 static void
4349 iface_destroy(struct iface *iface)
4350 {
4351 if (iface) {
4352 struct port *port = iface->port;
4353
4354 iface_destroy__(iface);
4355 if (ovs_list_is_empty(&port->ifaces)) {
4356 port_destroy(port);
4357 }
4358 }
4359 }
4360
4361 static struct iface *
4362 iface_lookup(const struct bridge *br, const char *name)
4363 {
4364 struct iface *iface;
4365
4366 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4367 &br->iface_by_name) {
4368 if (!strcmp(iface->name, name)) {
4369 return iface;
4370 }
4371 }
4372
4373 return NULL;
4374 }
4375
4376 static struct iface *
4377 iface_find(const char *name)
4378 {
4379 const struct bridge *br;
4380
4381 HMAP_FOR_EACH (br, node, &all_bridges) {
4382 struct iface *iface = iface_lookup(br, name);
4383
4384 if (iface) {
4385 return iface;
4386 }
4387 }
4388 return NULL;
4389 }
4390
4391 static struct iface *
4392 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4393 {
4394 struct iface *iface;
4395
4396 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4397 &br->ifaces) {
4398 if (iface->ofp_port == ofp_port) {
4399 return iface;
4400 }
4401 }
4402 return NULL;
4403 }
4404
4405 /* Set Ethernet address of 'iface', if one is specified in the configuration
4406 * file. */
4407 static void
4408 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4409 {
4410 struct eth_addr ea, *mac = NULL;
4411 struct iface *hw_addr_iface;
4412
4413 if (strcmp(iface->type, "internal")) {
4414 return;
4415 }
4416
4417 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, &ea)) {
4418 mac = &ea;
4419 } else if (port->cfg->fake_bridge) {
4420 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4421 find_local_hw_addr(br, &ea, port, &hw_addr_iface);
4422 mac = &ea;
4423 }
4424
4425 if (mac) {
4426 if (iface->ofp_port == OFPP_LOCAL) {
4427 VLOG_ERR("interface %s: ignoring mac in Interface record "
4428 "(use Bridge record to set local port's mac)",
4429 iface->name);
4430 } else if (eth_addr_is_multicast(*mac)) {
4431 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4432 iface->name);
4433 } else if (eth_addr_is_zero(*mac)) {
4434 VLOG_ERR("interface %s: cannot set MAC to all zero address",
4435 iface->name);
4436 } else {
4437 int error = netdev_set_etheraddr(iface->netdev, *mac);
4438 if (error) {
4439 VLOG_ERR("interface %s: setting MAC failed (%s)",
4440 iface->name, ovs_strerror(error));
4441 }
4442 }
4443 }
4444 }
4445
4446 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4447 static void
4448 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4449 {
4450 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4451 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4452 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4453 }
4454 }
4455
4456 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4457 * sets the "ofport" field to -1.
4458 *
4459 * This is appropriate when 'if_cfg''s interface cannot be created or is
4460 * otherwise invalid. */
4461 static void
4462 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4463 {
4464 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4465 iface_set_ofport(if_cfg, OFPP_NONE);
4466 ovsrec_interface_set_error(if_cfg, errp);
4467 ovsrec_interface_set_status(if_cfg, NULL);
4468 ovsrec_interface_set_admin_state(if_cfg, NULL);
4469 ovsrec_interface_set_duplex(if_cfg, NULL);
4470 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4471 ovsrec_interface_set_link_state(if_cfg, NULL);
4472 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4473 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4474 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4475 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4476 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4477 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4478 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4479 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4480 }
4481 }
4482
4483 static bool
4484 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4485 {
4486 union ovsdb_atom atom;
4487
4488 atom.integer = target;
4489 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4490 }
4491
4492 static void
4493 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4494 {
4495 struct ofpbuf queues_buf;
4496
4497 ofpbuf_init(&queues_buf, 0);
4498
4499 if (!qos || qos->type[0] == '\0') {
4500 netdev_set_qos(iface->netdev, NULL, NULL);
4501 } else {
4502 const struct ovsdb_datum *queues;
4503 struct netdev_queue_dump dump;
4504 unsigned int queue_id;
4505 struct smap details;
4506 bool queue_zero;
4507 size_t i;
4508
4509 /* Configure top-level Qos for 'iface'. */
4510 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4511
4512 /* Deconfigure queues that were deleted. */
4513 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4514 OVSDB_TYPE_UUID);
4515 smap_init(&details);
4516 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4517 if (!queue_ids_include(queues, queue_id)) {
4518 netdev_delete_queue(iface->netdev, queue_id);
4519 }
4520 }
4521 smap_destroy(&details);
4522
4523 /* Configure queues for 'iface'. */
4524 queue_zero = false;
4525 for (i = 0; i < qos->n_queues; i++) {
4526 const struct ovsrec_queue *queue = qos->value_queues[i];
4527 queue_id = qos->key_queues[i];
4528
4529 if (queue_id == 0) {
4530 queue_zero = true;
4531 }
4532
4533 if (queue->n_dscp == 1) {
4534 struct ofproto_port_queue *port_queue;
4535
4536 port_queue = ofpbuf_put_uninit(&queues_buf,
4537 sizeof *port_queue);
4538 port_queue->queue = queue_id;
4539 port_queue->dscp = queue->dscp[0];
4540 }
4541
4542 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4543 }
4544 if (!queue_zero) {
4545 smap_init(&details);
4546 netdev_set_queue(iface->netdev, 0, &details);
4547 smap_destroy(&details);
4548 }
4549 }
4550
4551 if (iface->ofp_port != OFPP_NONE) {
4552 const struct ofproto_port_queue *port_queues = queues_buf.data;
4553 size_t n_queues = queues_buf.size / sizeof *port_queues;
4554
4555 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4556 port_queues, n_queues);
4557 }
4558
4559 netdev_set_policing(iface->netdev,
4560 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4561 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4562
4563 ofpbuf_uninit(&queues_buf);
4564 }
4565
4566 static void
4567 iface_configure_cfm(struct iface *iface)
4568 {
4569 const struct ovsrec_interface *cfg = iface->cfg;
4570 const char *opstate_str;
4571 const char *cfm_ccm_vlan;
4572 struct cfm_settings s;
4573 struct smap netdev_args;
4574
4575 if (!cfg->n_cfm_mpid) {
4576 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4577 return;
4578 }
4579
4580 s.check_tnl_key = false;
4581 smap_init(&netdev_args);
4582 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4583 const char *key = smap_get(&netdev_args, "key");
4584 const char *in_key = smap_get(&netdev_args, "in_key");
4585
4586 s.check_tnl_key = (key && !strcmp(key, "flow"))
4587 || (in_key && !strcmp(in_key, "flow"));
4588 }
4589 smap_destroy(&netdev_args);
4590
4591 s.mpid = *cfg->cfm_mpid;
4592 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4593 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4594 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4595
4596 if (s.interval <= 0) {
4597 s.interval = 1000;
4598 }
4599
4600 if (!cfm_ccm_vlan) {
4601 s.ccm_vlan = 0;
4602 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4603 s.ccm_vlan = CFM_RANDOM_VLAN;
4604 } else {
4605 s.ccm_vlan = atoi(cfm_ccm_vlan);
4606 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4607 s.ccm_vlan = 0;
4608 }
4609 }
4610
4611 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4612 false);
4613 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4614
4615 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4616 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4617
4618 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4619 }
4620
4621 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4622 * instead of obtaining it from the database. */
4623 static bool
4624 iface_is_synthetic(const struct iface *iface)
4625 {
4626 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4627 }
4628
4629 static ofp_port_t
4630 iface_validate_ofport__(size_t n, int64_t *ofport)
4631 {
4632 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4633 ? u16_to_ofp(*ofport)
4634 : OFPP_NONE);
4635 }
4636
4637 static ofp_port_t
4638 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4639 {
4640 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4641 }
4642
4643 static ofp_port_t
4644 iface_pick_ofport(const struct ovsrec_interface *cfg)
4645 {
4646 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4647 return (requested_ofport != OFPP_NONE
4648 ? requested_ofport
4649 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4650 }
4651 \f
4652 /* Port mirroring. */
4653
4654 static struct mirror *
4655 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4656 {
4657 struct mirror *m;
4658
4659 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4660 if (uuid_equals(uuid, &m->uuid)) {
4661 return m;
4662 }
4663 }
4664 return NULL;
4665 }
4666
4667 static void
4668 bridge_configure_mirrors(struct bridge *br)
4669 {
4670 const struct ovsdb_datum *mc;
4671 unsigned long *flood_vlans;
4672 struct mirror *m, *next;
4673 size_t i;
4674
4675 /* Get rid of deleted mirrors. */
4676 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4677 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4678 union ovsdb_atom atom;
4679
4680 atom.uuid = m->uuid;
4681 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4682 mirror_destroy(m);
4683 }
4684 }
4685
4686 /* Add new mirrors and reconfigure existing ones. */
4687 for (i = 0; i < br->cfg->n_mirrors; i++) {
4688 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4689 m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4690 if (!m) {
4691 m = mirror_create(br, cfg);
4692 }
4693 m->cfg = cfg;
4694 if (!mirror_configure(m)) {
4695 mirror_destroy(m);
4696 }
4697 }
4698
4699 /* Update flooded vlans (for RSPAN). */
4700 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4701 br->cfg->n_flood_vlans);
4702 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4703 bitmap_free(flood_vlans);
4704 }
4705
4706 static struct mirror *
4707 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4708 {
4709 struct mirror *m;
4710
4711 m = xzalloc(sizeof *m);
4712 m->uuid = cfg->header_.uuid;
4713 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4714 m->bridge = br;
4715 m->name = xstrdup(cfg->name);
4716
4717 return m;
4718 }
4719
4720 static void
4721 mirror_destroy(struct mirror *m)
4722 {
4723 if (m) {
4724 struct bridge *br = m->bridge;
4725
4726 if (br->ofproto) {
4727 ofproto_mirror_unregister(br->ofproto, m);
4728 }
4729
4730 hmap_remove(&br->mirrors, &m->hmap_node);
4731 free(m->name);
4732 free(m);
4733 }
4734 }
4735
4736 static void
4737 mirror_collect_ports(struct mirror *m,
4738 struct ovsrec_port **in_ports, int n_in_ports,
4739 void ***out_portsp, size_t *n_out_portsp)
4740 {
4741 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4742 size_t n_out_ports = 0;
4743 size_t i;
4744
4745 for (i = 0; i < n_in_ports; i++) {
4746 const char *name = in_ports[i]->name;
4747 struct port *port = port_lookup(m->bridge, name);
4748 if (port) {
4749 out_ports[n_out_ports++] = port;
4750 } else {
4751 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4752 "port %s", m->bridge->name, m->name, name);
4753 }
4754 }
4755 *out_portsp = out_ports;
4756 *n_out_portsp = n_out_ports;
4757 }
4758
4759 static bool
4760 mirror_configure(struct mirror *m)
4761 {
4762 const struct ovsrec_mirror *cfg = m->cfg;
4763 struct ofproto_mirror_settings s;
4764
4765 /* Set name. */
4766 if (strcmp(cfg->name, m->name)) {
4767 free(m->name);
4768 m->name = xstrdup(cfg->name);
4769 }
4770 s.name = m->name;
4771
4772 /* Get output port or VLAN. */
4773 if (cfg->output_port) {
4774 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4775 if (!s.out_bundle) {
4776 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4777 m->bridge->name, m->name);
4778 return false;
4779 }
4780 s.out_vlan = UINT16_MAX;
4781
4782 if (cfg->output_vlan) {
4783 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4784 "output vlan; ignoring output vlan",
4785 m->bridge->name, m->name);
4786 }
4787 } else if (cfg->output_vlan) {
4788 /* The database should prevent invalid VLAN values. */
4789 s.out_bundle = NULL;
4790 s.out_vlan = *cfg->output_vlan;
4791 } else {
4792 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4793 m->bridge->name, m->name);
4794 return false;
4795 }
4796
4797 if (cfg->snaplen) {
4798 s.snaplen = *cfg->snaplen;
4799 } else {
4800 s.snaplen = 0;
4801 }
4802
4803 /* Get port selection. */
4804 if (cfg->select_all) {
4805 size_t n_ports = hmap_count(&m->bridge->ports);
4806 void **ports = xmalloc(n_ports * sizeof *ports);
4807 struct port *port;
4808 size_t i;
4809
4810 i = 0;
4811 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4812 ports[i++] = port;
4813 }
4814
4815 s.srcs = ports;
4816 s.n_srcs = n_ports;
4817
4818 s.dsts = ports;
4819 s.n_dsts = n_ports;
4820 } else {
4821 /* Get ports, dropping ports that don't exist.
4822 * The IDL ensures that there are no duplicates. */
4823 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4824 &s.srcs, &s.n_srcs);
4825 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4826 &s.dsts, &s.n_dsts);
4827 }
4828
4829 /* Get VLAN selection. */
4830 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4831
4832 /* Configure. */
4833 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4834
4835 /* Clean up. */
4836 if (s.srcs != s.dsts) {
4837 free(s.dsts);
4838 }
4839 free(s.srcs);
4840 free(s.src_vlans);
4841
4842 return true;
4843 }
4844
4845 \f
4846 static void
4847 mirror_refresh_stats(struct mirror *m)
4848 {
4849 struct ofproto *ofproto = m->bridge->ofproto;
4850 uint64_t tx_packets, tx_bytes;
4851 const char *keys[2];
4852 int64_t values[2];
4853 size_t stat_cnt = 0;
4854
4855 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4856 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4857 return;
4858 }
4859
4860 if (tx_packets != UINT64_MAX) {
4861 keys[stat_cnt] = "tx_packets";
4862 values[stat_cnt] = tx_packets;
4863 stat_cnt++;
4864 }
4865 if (tx_bytes != UINT64_MAX) {
4866 keys[stat_cnt] = "tx_bytes";
4867 values[stat_cnt] = tx_bytes;
4868 stat_cnt++;
4869 }
4870
4871 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
4872 }
4873
4874 /*
4875 * Add registered netdev and dpif types to ovsdb to allow external
4876 * applications to query the capabilities of the Open vSwitch instance
4877 * running on the node.
4878 */
4879 static void
4880 discover_types(const struct ovsrec_open_vswitch *cfg)
4881 {
4882 struct sset types;
4883
4884 /* Datapath types. */
4885 sset_init(&types);
4886 dp_enumerate_types(&types);
4887 const char **datapath_types = sset_array(&types);
4888 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
4889 sset_count(&types));
4890 free(datapath_types);
4891 sset_destroy(&types);
4892
4893 /* Port types. */
4894 sset_init(&types);
4895 netdev_enumerate_types(&types);
4896 const char **iface_types = sset_array(&types);
4897 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
4898 free(iface_types);
4899 sset_destroy(&types);
4900 }