]> git.proxmox.com Git - mirror_ovs.git/blob - vswitchd/bridge.c
ovs-lldp: Use better types for ISID and VLANs.
[mirror_ovs.git] / vswitchd / bridge.c
1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
2 *
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 *
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
14 */
15
16 #include <config.h>
17 #include "bridge.h"
18 #include <errno.h>
19 #include <inttypes.h>
20 #include <stdlib.h>
21 #include "async-append.h"
22 #include "bfd.h"
23 #include "bitmap.h"
24 #include "cfm.h"
25 #include "connectivity.h"
26 #include "coverage.h"
27 #include "daemon.h"
28 #include "dirs.h"
29 #include "dynamic-string.h"
30 #include "hash.h"
31 #include "hmap.h"
32 #include "hmapx.h"
33 #include "jsonrpc.h"
34 #include "lacp.h"
35 #include "list.h"
36 #include "ovs-lldp.h"
37 #include "mac-learning.h"
38 #include "mcast-snooping.h"
39 #include "meta-flow.h"
40 #include "netdev.h"
41 #include "nx-match.h"
42 #include "ofp-print.h"
43 #include "ofp-util.h"
44 #include "ofpbuf.h"
45 #include "ofproto/bond.h"
46 #include "ofproto/ofproto.h"
47 #include "ovs-numa.h"
48 #include "poll-loop.h"
49 #include "seq.h"
50 #include "sha1.h"
51 #include "shash.h"
52 #include "smap.h"
53 #include "socket-util.h"
54 #include "stream.h"
55 #include "stream-ssl.h"
56 #include "sset.h"
57 #include "system-stats.h"
58 #include "timeval.h"
59 #include "util.h"
60 #include "unixctl.h"
61 #include "vlandev.h"
62 #include "lib/vswitch-idl.h"
63 #include "xenserver.h"
64 #include "openvswitch/vlog.h"
65 #include "sflow_api.h"
66 #include "vlan-bitmap.h"
67 #include "packets.h"
68
69 VLOG_DEFINE_THIS_MODULE(bridge);
70
71 COVERAGE_DEFINE(bridge_reconfigure);
72
73 struct iface {
74 /* These members are always valid.
75 *
76 * They are immutable: they never change between iface_create() and
77 * iface_destroy(). */
78 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
79 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
80 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
81 struct port *port; /* Containing port. */
82 char *name; /* Host network device name. */
83 struct netdev *netdev; /* Network device. */
84 ofp_port_t ofp_port; /* OpenFlow port number. */
85 uint64_t change_seq;
86
87 /* These members are valid only within bridge_reconfigure(). */
88 const char *type; /* Usually same as cfg->type. */
89 const struct ovsrec_interface *cfg;
90 };
91
92 struct mirror {
93 struct uuid uuid; /* UUID of this "mirror" record in database. */
94 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
95 struct bridge *bridge;
96 char *name;
97 const struct ovsrec_mirror *cfg;
98 };
99
100 struct port {
101 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
102 struct bridge *bridge;
103 char *name;
104
105 const struct ovsrec_port *cfg;
106
107 /* An ordinary bridge port has 1 interface.
108 * A bridge port for bonding has at least 2 interfaces. */
109 struct ovs_list ifaces; /* List of "struct iface"s. */
110 };
111
112 struct bridge {
113 struct hmap_node node; /* In 'all_bridges'. */
114 char *name; /* User-specified arbitrary name. */
115 char *type; /* Datapath type. */
116 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
117 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
118 const struct ovsrec_bridge *cfg;
119
120 /* OpenFlow switch processing. */
121 struct ofproto *ofproto; /* OpenFlow switch. */
122
123 /* Bridge ports. */
124 struct hmap ports; /* "struct port"s indexed by name. */
125 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
126 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
127
128 /* Port mirroring. */
129 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
130
131 /* Auto Attach */
132 struct hmap mappings; /* "struct" indexed by UUID */
133
134 /* Used during reconfiguration. */
135 struct shash wanted_ports;
136
137 /* Synthetic local port if necessary. */
138 struct ovsrec_port synth_local_port;
139 struct ovsrec_interface synth_local_iface;
140 struct ovsrec_interface *synth_local_ifacep;
141 };
142
143 struct aa_mapping {
144 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
145 struct bridge *bridge;
146 uint32_t isid;
147 uint16_t vlan;
148 char *br_name;
149 };
150
151 /* All bridges, indexed by name. */
152 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
153
154 /* OVSDB IDL used to obtain configuration. */
155 static struct ovsdb_idl *idl;
156
157 /* We want to complete daemonization, fully detaching from our parent process,
158 * only after we have completed our initial configuration, committed our state
159 * to the database, and received confirmation back from the database server
160 * that it applied the commit. This allows our parent process to know that,
161 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
162 * to have already been filled in. (It is only "very likely" rather than
163 * certain because there is always a slim possibility that the transaction will
164 * fail or that some other client has added new bridges, ports, etc. while
165 * ovs-vswitchd was configuring using an old configuration.)
166 *
167 * We only need to do this once for our initial configuration at startup, so
168 * 'initial_config_done' tracks whether we've already done it. While we are
169 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
170 * itself and is otherwise NULL. */
171 static bool initial_config_done;
172 static struct ovsdb_idl_txn *daemonize_txn;
173
174 /* Most recently processed IDL sequence number. */
175 static unsigned int idl_seqno;
176
177 /* Track changes to port connectivity. */
178 static uint64_t connectivity_seqno = LLONG_MIN;
179
180 /* Status update to database.
181 *
182 * Some information in the database must be kept as up-to-date as possible to
183 * allow controllers to respond rapidly to network outages. Those status are
184 * updated via the 'status_txn'.
185 *
186 * We use the global connectivity sequence number to detect the status change.
187 * Also, to prevent the status update from sending too much to the database,
188 * we check the return status of each update transaction and do not start new
189 * update if the previous transaction status is 'TXN_INCOMPLETE'.
190 *
191 * 'statux_txn' is NULL if there is no ongoing status update.
192 *
193 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
194 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
195 * which will cause the main thread wake up soon and retry the status update.
196 */
197 static struct ovsdb_idl_txn *status_txn;
198 static bool status_txn_try_again;
199
200 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
201 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
202 #define STATUS_CHECK_AGAIN_MSEC 100
203
204 /* Each time this timer expires, the bridge fetches interface and mirror
205 * statistics and pushes them into the database. */
206 static int stats_timer_interval;
207 static long long int stats_timer = LLONG_MIN;
208
209 /* Each time this timer expires, the bridge fetches the list of port/VLAN
210 * membership that has been modified by the AA.
211 */
212 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
213 static long long int aa_refresh_timer = LLONG_MIN;
214
215 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
216 * expensive. This can cause bridge_reconfigure() to take a long time during
217 * which no other work can be done. To deal with this problem, we limit port
218 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
219 * call to bridge_reconfigure(). If there is more work to do after the limit
220 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
221 * This allows the rest of the code to catch up on important things like
222 * forwarding packets. */
223 #define OFP_PORT_ACTION_WINDOW 10
224
225 static void add_del_bridges(const struct ovsrec_open_vswitch *);
226 static void bridge_run__(void);
227 static void bridge_create(const struct ovsrec_bridge *);
228 static void bridge_destroy(struct bridge *);
229 static struct bridge *bridge_lookup(const char *name);
230 static unixctl_cb_func bridge_unixctl_dump_flows;
231 static unixctl_cb_func bridge_unixctl_reconnect;
232 static size_t bridge_get_controllers(const struct bridge *br,
233 struct ovsrec_controller ***controllersp);
234 static void bridge_collect_wanted_ports(struct bridge *,
235 const unsigned long *splinter_vlans,
236 struct shash *wanted_ports);
237 static void bridge_delete_ofprotos(void);
238 static void bridge_delete_or_reconfigure_ports(struct bridge *);
239 static void bridge_del_ports(struct bridge *,
240 const struct shash *wanted_ports);
241 static void bridge_add_ports(struct bridge *,
242 const struct shash *wanted_ports);
243
244 static void bridge_configure_datapath_id(struct bridge *);
245 static void bridge_configure_netflow(struct bridge *);
246 static void bridge_configure_forward_bpdu(struct bridge *);
247 static void bridge_configure_mac_table(struct bridge *);
248 static void bridge_configure_mcast_snooping(struct bridge *);
249 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
250 static void bridge_configure_ipfix(struct bridge *);
251 static void bridge_configure_stp(struct bridge *);
252 static void bridge_configure_rstp(struct bridge *);
253 static void bridge_configure_tables(struct bridge *);
254 static void bridge_configure_dp_desc(struct bridge *);
255 static void bridge_configure_aa(struct bridge *);
256 static void bridge_aa_refresh_queued(struct bridge *);
257 static bool bridge_aa_need_refresh(struct bridge *);
258 static void bridge_configure_remotes(struct bridge *,
259 const struct sockaddr_in *managers,
260 size_t n_managers);
261 static void bridge_pick_local_hw_addr(struct bridge *,
262 uint8_t ea[ETH_ADDR_LEN],
263 struct iface **hw_addr_iface);
264 static uint64_t bridge_pick_datapath_id(struct bridge *,
265 const uint8_t bridge_ea[ETH_ADDR_LEN],
266 struct iface *hw_addr_iface);
267 static uint64_t dpid_from_hash(const void *, size_t nbytes);
268 static bool bridge_has_bond_fake_iface(const struct bridge *,
269 const char *name);
270 static bool port_is_bond_fake_iface(const struct port *);
271
272 static unixctl_cb_func qos_unixctl_show;
273
274 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
275 static void port_del_ifaces(struct port *);
276 static void port_destroy(struct port *);
277 static struct port *port_lookup(const struct bridge *, const char *name);
278 static void port_configure(struct port *);
279 static struct lacp_settings *port_configure_lacp(struct port *,
280 struct lacp_settings *);
281 static void port_configure_bond(struct port *, struct bond_settings *);
282 static bool port_is_synthetic(const struct port *);
283
284 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
285 static void run_system_stats(void);
286
287 static void bridge_configure_mirrors(struct bridge *);
288 static struct mirror *mirror_create(struct bridge *,
289 const struct ovsrec_mirror *);
290 static void mirror_destroy(struct mirror *);
291 static bool mirror_configure(struct mirror *);
292 static void mirror_refresh_stats(struct mirror *);
293
294 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
295 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
296 const struct ovsrec_port *);
297 static bool iface_is_internal(const struct ovsrec_interface *iface,
298 const struct ovsrec_bridge *br);
299 static const char *iface_get_type(const struct ovsrec_interface *,
300 const struct ovsrec_bridge *);
301 static void iface_destroy(struct iface *);
302 static void iface_destroy__(struct iface *);
303 static struct iface *iface_lookup(const struct bridge *, const char *name);
304 static struct iface *iface_find(const char *name);
305 static struct iface *iface_from_ofp_port(const struct bridge *,
306 ofp_port_t ofp_port);
307 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
308 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
309 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
310 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
311 static void iface_configure_cfm(struct iface *);
312 static void iface_refresh_cfm_stats(struct iface *);
313 static void iface_refresh_stats(struct iface *);
314 static void iface_refresh_netdev_status(struct iface *);
315 static void iface_refresh_ofproto_status(struct iface *);
316 static bool iface_is_synthetic(const struct iface *);
317 static ofp_port_t iface_get_requested_ofp_port(
318 const struct ovsrec_interface *);
319 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
320
321 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
322 *
323 * This is deprecated. It is only for compatibility with broken device drivers
324 * in old versions of Linux that do not properly support VLANs when VLAN
325 * devices are not used. When broken device drivers are no longer in
326 * widespread use, we will delete these interfaces. */
327
328 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
329 static bool vlan_splinters_enabled_anywhere;
330
331 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
332 static unsigned long int *collect_splinter_vlans(
333 const struct ovsrec_open_vswitch *);
334 static void configure_splinter_port(struct port *);
335 static void add_vlan_splinter_ports(struct bridge *,
336 const unsigned long int *splinter_vlans,
337 struct shash *ports);
338
339 static void
340 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
341 {
342 struct shash iface_hints;
343 static bool initialized = false;
344 int i;
345
346 if (initialized) {
347 return;
348 }
349
350 shash_init(&iface_hints);
351
352 if (cfg) {
353 for (i = 0; i < cfg->n_bridges; i++) {
354 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
355 int j;
356
357 for (j = 0; j < br_cfg->n_ports; j++) {
358 struct ovsrec_port *port_cfg = br_cfg->ports[j];
359 int k;
360
361 for (k = 0; k < port_cfg->n_interfaces; k++) {
362 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
363 struct iface_hint *iface_hint;
364
365 iface_hint = xmalloc(sizeof *iface_hint);
366 iface_hint->br_name = br_cfg->name;
367 iface_hint->br_type = br_cfg->datapath_type;
368 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
369
370 shash_add(&iface_hints, if_cfg->name, iface_hint);
371 }
372 }
373 }
374 }
375
376 ofproto_init(&iface_hints);
377
378 shash_destroy_free_data(&iface_hints);
379 initialized = true;
380 }
381 \f
382 /* Public functions. */
383
384 /* Initializes the bridge module, configuring it to obtain its configuration
385 * from an OVSDB server accessed over 'remote', which should be a string in a
386 * form acceptable to ovsdb_idl_create(). */
387 void
388 bridge_init(const char *remote)
389 {
390 /* Create connection to database. */
391 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
392 idl_seqno = ovsdb_idl_get_seqno(idl);
393 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
394 ovsdb_idl_verify_write_only(idl);
395
396 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
397 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
398 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
400 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
401 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
402 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
403
404 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
405 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
406 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
407 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
408 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
409 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
410 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
411
412 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
413 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
414 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
415 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
416 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
417 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
418 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
419 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
420 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
421 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
422 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
423 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
424 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
425 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
426 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
427 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
428 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
429 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
430 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
440 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
441
442 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
443 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
444 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
445 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
446
447 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
448
449 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
450
451 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
452 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
453
454 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
455 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
456 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
457 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
458
459 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
460 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
461 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
462 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
463 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
464
465 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
466
467 /* Register unixctl commands. */
468 unixctl_command_register("qos/show", "interface", 1, 1,
469 qos_unixctl_show, NULL);
470 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
471 bridge_unixctl_dump_flows, NULL);
472 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
473 bridge_unixctl_reconnect, NULL);
474 lacp_init();
475 bond_init();
476 cfm_init();
477 ovs_numa_init();
478 stp_init();
479 lldp_init();
480 rstp_init();
481 }
482
483 void
484 bridge_exit(void)
485 {
486 struct bridge *br, *next_br;
487
488 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
489 bridge_destroy(br);
490 }
491 ovsdb_idl_destroy(idl);
492 }
493
494 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
495 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
496 * responsible for freeing '*managersp' (with free()).
497 *
498 * You may be asking yourself "why does ovs-vswitchd care?", because
499 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
500 * should not be and in fact is not directly involved in that. But
501 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
502 * it has to tell in-band control where the managers are to enable that.
503 * (Thus, only managers connected in-band are collected.)
504 */
505 static void
506 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
507 struct sockaddr_in **managersp, size_t *n_managersp)
508 {
509 struct sockaddr_in *managers = NULL;
510 size_t n_managers = 0;
511 struct sset targets;
512 size_t i;
513
514 /* Collect all of the potential targets from the "targets" columns of the
515 * rows pointed to by "manager_options", excluding any that are
516 * out-of-band. */
517 sset_init(&targets);
518 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
519 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
520
521 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
522 sset_find_and_delete(&targets, m->target);
523 } else {
524 sset_add(&targets, m->target);
525 }
526 }
527
528 /* Now extract the targets' IP addresses. */
529 if (!sset_is_empty(&targets)) {
530 const char *target;
531
532 managers = xmalloc(sset_count(&targets) * sizeof *managers);
533 SSET_FOR_EACH (target, &targets) {
534 union {
535 struct sockaddr_storage ss;
536 struct sockaddr_in in;
537 } sa;
538
539 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
540 &sa.ss)
541 && sa.ss.ss_family == AF_INET) {
542 managers[n_managers++] = sa.in;
543 }
544 }
545 }
546 sset_destroy(&targets);
547
548 *managersp = managers;
549 *n_managersp = n_managers;
550 }
551
552 static void
553 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
554 {
555 unsigned long int *splinter_vlans;
556 struct sockaddr_in *managers;
557 struct bridge *br, *next;
558 int sflow_bridge_number;
559 size_t n_managers;
560
561 COVERAGE_INC(bridge_reconfigure);
562
563 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
564 OFPROTO_FLOW_LIMIT_DEFAULT));
565 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
566 OFPROTO_MAX_IDLE_DEFAULT));
567 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
568 "n-dpdk-rxqs", 0));
569 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
570
571 ofproto_set_threads(
572 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
573 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
574
575 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
576 * to 'ovs_cfg', with only very minimal configuration otherwise.
577 *
578 * This is mostly an update to bridge data structures. Nothing is pushed
579 * down to ofproto or lower layers. */
580 add_del_bridges(ovs_cfg);
581 splinter_vlans = collect_splinter_vlans(ovs_cfg);
582 HMAP_FOR_EACH (br, node, &all_bridges) {
583 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
584 bridge_del_ports(br, &br->wanted_ports);
585 }
586 free(splinter_vlans);
587
588 /* Start pushing configuration changes down to the ofproto layer:
589 *
590 * - Delete ofprotos that are no longer configured.
591 *
592 * - Delete ports that are no longer configured.
593 *
594 * - Reconfigure existing ports to their desired configurations, or
595 * delete them if not possible.
596 *
597 * We have to do all the deletions before we can do any additions, because
598 * the ports to be added might require resources that will be freed up by
599 * deletions (they might especially overlap in name). */
600 bridge_delete_ofprotos();
601 HMAP_FOR_EACH (br, node, &all_bridges) {
602 if (br->ofproto) {
603 bridge_delete_or_reconfigure_ports(br);
604 }
605 }
606
607 /* Finish pushing configuration changes to the ofproto layer:
608 *
609 * - Create ofprotos that are missing.
610 *
611 * - Add ports that are missing. */
612 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
613 if (!br->ofproto) {
614 int error;
615
616 error = ofproto_create(br->name, br->type, &br->ofproto);
617 if (error) {
618 VLOG_ERR("failed to create bridge %s: %s", br->name,
619 ovs_strerror(error));
620 shash_destroy(&br->wanted_ports);
621 bridge_destroy(br);
622 } else {
623 /* Trigger storing datapath version. */
624 seq_change(connectivity_seq_get());
625 }
626 }
627 }
628 HMAP_FOR_EACH (br, node, &all_bridges) {
629 bridge_add_ports(br, &br->wanted_ports);
630 shash_destroy(&br->wanted_ports);
631 }
632
633 reconfigure_system_stats(ovs_cfg);
634
635 /* Complete the configuration. */
636 sflow_bridge_number = 0;
637 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
638 HMAP_FOR_EACH (br, node, &all_bridges) {
639 struct port *port;
640
641 /* We need the datapath ID early to allow LACP ports to use it as the
642 * default system ID. */
643 bridge_configure_datapath_id(br);
644
645 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
646 struct iface *iface;
647
648 port_configure(port);
649
650 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
651 iface_set_ofport(iface->cfg, iface->ofp_port);
652 /* Clear eventual previous errors */
653 ovsrec_interface_set_error(iface->cfg, NULL);
654 iface_configure_cfm(iface);
655 iface_configure_qos(iface, port->cfg->qos);
656 iface_set_mac(br, port, iface);
657 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
658 &iface->cfg->bfd);
659 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
660 &iface->cfg->lldp);
661 }
662 }
663 bridge_configure_mirrors(br);
664 bridge_configure_forward_bpdu(br);
665 bridge_configure_mac_table(br);
666 bridge_configure_mcast_snooping(br);
667 bridge_configure_remotes(br, managers, n_managers);
668 bridge_configure_netflow(br);
669 bridge_configure_sflow(br, &sflow_bridge_number);
670 bridge_configure_ipfix(br);
671 bridge_configure_stp(br);
672 bridge_configure_rstp(br);
673 bridge_configure_tables(br);
674 bridge_configure_dp_desc(br);
675 bridge_configure_aa(br);
676 }
677 free(managers);
678
679 /* The ofproto-dpif provider does some final reconfiguration in its
680 * ->type_run() function. We have to call it before notifying the database
681 * client that reconfiguration is complete, otherwise there is a very
682 * narrow race window in which e.g. ofproto/trace will not recognize the
683 * new configuration (sometimes this causes unit test failures). */
684 bridge_run__();
685 }
686
687 /* Delete ofprotos which aren't configured or have the wrong type. Create
688 * ofprotos which don't exist but need to. */
689 static void
690 bridge_delete_ofprotos(void)
691 {
692 struct bridge *br;
693 struct sset names;
694 struct sset types;
695 const char *type;
696
697 /* Delete ofprotos with no bridge or with the wrong type. */
698 sset_init(&names);
699 sset_init(&types);
700 ofproto_enumerate_types(&types);
701 SSET_FOR_EACH (type, &types) {
702 const char *name;
703
704 ofproto_enumerate_names(type, &names);
705 SSET_FOR_EACH (name, &names) {
706 br = bridge_lookup(name);
707 if (!br || strcmp(type, br->type)) {
708 ofproto_delete(name, type);
709 }
710 }
711 }
712 sset_destroy(&names);
713 sset_destroy(&types);
714 }
715
716 static ofp_port_t *
717 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
718 {
719 if (*n >= *allocated) {
720 ports = x2nrealloc(ports, allocated, sizeof *ports);
721 }
722 ports[(*n)++] = port;
723 return ports;
724 }
725
726 static void
727 bridge_delete_or_reconfigure_ports(struct bridge *br)
728 {
729 struct ofproto_port ofproto_port;
730 struct ofproto_port_dump dump;
731
732 struct sset ofproto_ports;
733 struct port *port, *port_next;
734
735 /* List of "ofp_port"s to delete. We make a list instead of deleting them
736 * right away because ofproto implementations aren't necessarily able to
737 * iterate through a changing list of ports in an entirely robust way. */
738 ofp_port_t *del;
739 size_t n, allocated;
740 size_t i;
741
742 del = NULL;
743 n = allocated = 0;
744 sset_init(&ofproto_ports);
745
746 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
747 * that are not configured in the database. (This commonly happens when
748 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
749 *
750 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
751 * that have the wrong OpenFlow port number (and arrange to add them back
752 * with the correct OpenFlow port number). */
753 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
754 ofp_port_t requested_ofp_port;
755 struct iface *iface;
756
757 sset_add(&ofproto_ports, ofproto_port.name);
758
759 iface = iface_lookup(br, ofproto_port.name);
760 if (!iface) {
761 /* No such iface is configured, so we should delete this
762 * ofproto_port.
763 *
764 * As a corner case exception, keep the port if it's a bond fake
765 * interface. */
766 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
767 && !strcmp(ofproto_port.type, "internal")) {
768 continue;
769 }
770 goto delete;
771 }
772
773 if (strcmp(ofproto_port.type, iface->type)
774 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
775 /* The interface is the wrong type or can't be configured.
776 * Delete it. */
777 goto delete;
778 }
779
780 /* If the requested OpenFlow port for 'iface' changed, and it's not
781 * already the correct port, then we might want to temporarily delete
782 * this interface, so we can add it back again with the new OpenFlow
783 * port number. */
784 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
785 if (iface->ofp_port != OFPP_LOCAL &&
786 requested_ofp_port != OFPP_NONE &&
787 requested_ofp_port != iface->ofp_port) {
788 ofp_port_t victim_request;
789 struct iface *victim;
790
791 /* Check for an existing OpenFlow port currently occupying
792 * 'iface''s requested port number. If there isn't one, then
793 * delete this port. Otherwise we need to consider further. */
794 victim = iface_from_ofp_port(br, requested_ofp_port);
795 if (!victim) {
796 goto delete;
797 }
798
799 /* 'victim' is a port currently using 'iface''s requested port
800 * number. Unless 'victim' specifically requested that port
801 * number, too, then we can delete both 'iface' and 'victim'
802 * temporarily. (We'll add both of them back again later with new
803 * OpenFlow port numbers.)
804 *
805 * If 'victim' did request port number 'requested_ofp_port', just
806 * like 'iface', then that's a configuration inconsistency that we
807 * can't resolve. We might as well let it keep its current port
808 * number. */
809 victim_request = iface_get_requested_ofp_port(victim->cfg);
810 if (victim_request != requested_ofp_port) {
811 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
812 iface_destroy(victim);
813 goto delete;
814 }
815 }
816
817 /* Keep it. */
818 continue;
819
820 delete:
821 iface_destroy(iface);
822 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
823 }
824 for (i = 0; i < n; i++) {
825 ofproto_port_del(br->ofproto, del[i]);
826 }
827 free(del);
828
829 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
830 * that we didn't see when we iterated through the datapath, i.e. ports
831 * that disappeared underneath use. This is an unusual situation, but it
832 * can happen in some cases:
833 *
834 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
835 * idea but could happen).
836 *
837 * - The port represented a device that disappeared, e.g. a tuntap
838 * device destroyed via "tunctl -d", a physical Ethernet device
839 * whose module was just unloaded via "rmmod", or a virtual NIC for a
840 * VM whose VM was just terminated. */
841 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
842 struct iface *iface, *iface_next;
843
844 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
845 if (!sset_contains(&ofproto_ports, iface->name)) {
846 iface_destroy__(iface);
847 }
848 }
849
850 if (list_is_empty(&port->ifaces)) {
851 port_destroy(port);
852 }
853 }
854 sset_destroy(&ofproto_ports);
855 }
856
857 static void
858 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
859 bool with_requested_port)
860 {
861 struct shash_node *port_node;
862
863 SHASH_FOR_EACH (port_node, wanted_ports) {
864 const struct ovsrec_port *port_cfg = port_node->data;
865 size_t i;
866
867 for (i = 0; i < port_cfg->n_interfaces; i++) {
868 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
869 ofp_port_t requested_ofp_port;
870
871 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
872 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
873 struct iface *iface = iface_lookup(br, iface_cfg->name);
874
875 if (!iface) {
876 iface_create(br, iface_cfg, port_cfg);
877 }
878 }
879 }
880 }
881 }
882
883 static void
884 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
885 {
886 /* First add interfaces that request a particular port number. */
887 bridge_add_ports__(br, wanted_ports, true);
888
889 /* Then add interfaces that want automatic port number assignment.
890 * We add these afterward to avoid accidentally taking a specifically
891 * requested port number. */
892 bridge_add_ports__(br, wanted_ports, false);
893 }
894
895 static void
896 port_configure(struct port *port)
897 {
898 const struct ovsrec_port *cfg = port->cfg;
899 struct bond_settings bond_settings;
900 struct lacp_settings lacp_settings;
901 struct ofproto_bundle_settings s;
902 struct iface *iface;
903
904 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
905 configure_splinter_port(port);
906 return;
907 }
908
909 /* Get name. */
910 s.name = port->name;
911
912 /* Get slaves. */
913 s.n_slaves = 0;
914 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
915 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
916 s.slaves[s.n_slaves++] = iface->ofp_port;
917 }
918
919 /* Get VLAN tag. */
920 s.vlan = -1;
921 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
922 s.vlan = *cfg->tag;
923 }
924
925 /* Get VLAN trunks. */
926 s.trunks = NULL;
927 if (cfg->n_trunks) {
928 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
929 }
930
931 /* Get VLAN mode. */
932 if (cfg->vlan_mode) {
933 if (!strcmp(cfg->vlan_mode, "access")) {
934 s.vlan_mode = PORT_VLAN_ACCESS;
935 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
936 s.vlan_mode = PORT_VLAN_TRUNK;
937 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
938 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
939 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
940 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
941 } else {
942 /* This "can't happen" because ovsdb-server should prevent it. */
943 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
944 "back to trunk mode", port->name, cfg->vlan_mode);
945 s.vlan_mode = PORT_VLAN_TRUNK;
946 }
947 } else {
948 if (s.vlan >= 0) {
949 s.vlan_mode = PORT_VLAN_ACCESS;
950 if (cfg->n_trunks) {
951 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
952 port->name);
953 }
954 } else {
955 s.vlan_mode = PORT_VLAN_TRUNK;
956 }
957 }
958 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
959 false);
960
961 /* Get LACP settings. */
962 s.lacp = port_configure_lacp(port, &lacp_settings);
963 if (s.lacp) {
964 size_t i = 0;
965
966 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
967 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
968 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
969 }
970 } else {
971 s.lacp_slaves = NULL;
972 }
973
974 /* Get bond settings. */
975 if (s.n_slaves > 1) {
976 s.bond = &bond_settings;
977 port_configure_bond(port, &bond_settings);
978 } else {
979 s.bond = NULL;
980 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
981 netdev_set_miimon_interval(iface->netdev, 0);
982 }
983 }
984
985 /* Register. */
986 ofproto_bundle_register(port->bridge->ofproto, port, &s);
987
988 /* Clean up. */
989 free(s.slaves);
990 free(s.trunks);
991 free(s.lacp_slaves);
992 }
993
994 /* Pick local port hardware address and datapath ID for 'br'. */
995 static void
996 bridge_configure_datapath_id(struct bridge *br)
997 {
998 uint8_t ea[ETH_ADDR_LEN];
999 uint64_t dpid;
1000 struct iface *local_iface;
1001 struct iface *hw_addr_iface;
1002 char *dpid_string;
1003
1004 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1005 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1006 if (local_iface) {
1007 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1008 if (error) {
1009 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1010 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1011 "Ethernet address: %s",
1012 br->name, ovs_strerror(error));
1013 }
1014 }
1015 memcpy(br->ea, ea, ETH_ADDR_LEN);
1016
1017 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1018 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1019 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1020 ofproto_set_datapath_id(br->ofproto, dpid);
1021 }
1022
1023 dpid_string = xasprintf("%016"PRIx64, dpid);
1024 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1025 free(dpid_string);
1026 }
1027
1028 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1029 * 'br'. */
1030 static uint32_t
1031 bridge_get_allowed_versions(struct bridge *br)
1032 {
1033 if (!br->cfg->n_protocols)
1034 return 0;
1035
1036 return ofputil_versions_from_strings(br->cfg->protocols,
1037 br->cfg->n_protocols);
1038 }
1039
1040 /* Set NetFlow configuration on 'br'. */
1041 static void
1042 bridge_configure_netflow(struct bridge *br)
1043 {
1044 struct ovsrec_netflow *cfg = br->cfg->netflow;
1045 struct netflow_options opts;
1046
1047 if (!cfg) {
1048 ofproto_set_netflow(br->ofproto, NULL);
1049 return;
1050 }
1051
1052 memset(&opts, 0, sizeof opts);
1053
1054 /* Get default NetFlow configuration from datapath.
1055 * Apply overrides from 'cfg'. */
1056 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1057 if (cfg->engine_type) {
1058 opts.engine_type = *cfg->engine_type;
1059 }
1060 if (cfg->engine_id) {
1061 opts.engine_id = *cfg->engine_id;
1062 }
1063
1064 /* Configure active timeout interval. */
1065 opts.active_timeout = cfg->active_timeout;
1066 if (!opts.active_timeout) {
1067 opts.active_timeout = -1;
1068 } else if (opts.active_timeout < 0) {
1069 VLOG_WARN("bridge %s: active timeout interval set to negative "
1070 "value, using default instead (%d seconds)", br->name,
1071 NF_ACTIVE_TIMEOUT_DEFAULT);
1072 opts.active_timeout = -1;
1073 }
1074
1075 /* Add engine ID to interface number to disambiguate bridgs? */
1076 opts.add_id_to_iface = cfg->add_id_to_interface;
1077 if (opts.add_id_to_iface) {
1078 if (opts.engine_id > 0x7f) {
1079 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1080 "another vswitch, choose an engine id less than 128",
1081 br->name);
1082 }
1083 if (hmap_count(&br->ports) > 508) {
1084 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1085 "another port when more than 508 ports are used",
1086 br->name);
1087 }
1088 }
1089
1090 /* Collectors. */
1091 sset_init(&opts.collectors);
1092 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1093
1094 /* Configure. */
1095 if (ofproto_set_netflow(br->ofproto, &opts)) {
1096 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1097 }
1098 sset_destroy(&opts.collectors);
1099 }
1100
1101 /* Set sFlow configuration on 'br'. */
1102 static void
1103 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1104 {
1105 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1106 struct ovsrec_controller **controllers;
1107 struct ofproto_sflow_options oso;
1108 size_t n_controllers;
1109 size_t i;
1110
1111 if (!cfg) {
1112 ofproto_set_sflow(br->ofproto, NULL);
1113 return;
1114 }
1115
1116 memset(&oso, 0, sizeof oso);
1117
1118 sset_init(&oso.targets);
1119 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1120
1121 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1122 if (cfg->sampling) {
1123 oso.sampling_rate = *cfg->sampling;
1124 }
1125
1126 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1127 if (cfg->polling) {
1128 oso.polling_interval = *cfg->polling;
1129 }
1130
1131 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1132 if (cfg->header) {
1133 oso.header_len = *cfg->header;
1134 }
1135
1136 oso.sub_id = (*sflow_bridge_number)++;
1137 oso.agent_device = cfg->agent;
1138
1139 oso.control_ip = NULL;
1140 n_controllers = bridge_get_controllers(br, &controllers);
1141 for (i = 0; i < n_controllers; i++) {
1142 if (controllers[i]->local_ip) {
1143 oso.control_ip = controllers[i]->local_ip;
1144 break;
1145 }
1146 }
1147 ofproto_set_sflow(br->ofproto, &oso);
1148
1149 sset_destroy(&oso.targets);
1150 }
1151
1152 /* Returns whether a IPFIX row is valid. */
1153 static bool
1154 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1155 {
1156 return ipfix && ipfix->n_targets > 0;
1157 }
1158
1159 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1160 static bool
1161 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1162 const struct bridge *br)
1163 {
1164 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1165 }
1166
1167 /* Set IPFIX configuration on 'br'. */
1168 static void
1169 bridge_configure_ipfix(struct bridge *br)
1170 {
1171 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1172 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1173 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1174 struct ofproto_ipfix_bridge_exporter_options be_opts;
1175 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1176 size_t n_fe_opts = 0;
1177
1178 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1179 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1180 n_fe_opts++;
1181 }
1182 }
1183
1184 if (!valid_be_cfg && n_fe_opts == 0) {
1185 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1186 return;
1187 }
1188
1189 if (valid_be_cfg) {
1190 memset(&be_opts, 0, sizeof be_opts);
1191
1192 sset_init(&be_opts.targets);
1193 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1194
1195 if (be_cfg->sampling) {
1196 be_opts.sampling_rate = *be_cfg->sampling;
1197 } else {
1198 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1199 }
1200 if (be_cfg->obs_domain_id) {
1201 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1202 }
1203 if (be_cfg->obs_point_id) {
1204 be_opts.obs_point_id = *be_cfg->obs_point_id;
1205 }
1206 if (be_cfg->cache_active_timeout) {
1207 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1208 }
1209 if (be_cfg->cache_max_flows) {
1210 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1211 }
1212
1213 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1214 "enable-tunnel-sampling", true);
1215
1216 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1217 "enable-input-sampling", false);
1218
1219 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1220 "enable-output-sampling", false);
1221 }
1222
1223 if (n_fe_opts > 0) {
1224 struct ofproto_ipfix_flow_exporter_options *opts;
1225 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1226 opts = fe_opts;
1227 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1228 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1229 opts->collector_set_id = fe_cfg->id;
1230 sset_init(&opts->targets);
1231 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1232 fe_cfg->ipfix->n_targets);
1233 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1234 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1235 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1236 ? *fe_cfg->ipfix->cache_max_flows : 0;
1237 opts++;
1238 }
1239 }
1240 }
1241
1242 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1243 n_fe_opts);
1244
1245 if (valid_be_cfg) {
1246 sset_destroy(&be_opts.targets);
1247 }
1248
1249 if (n_fe_opts > 0) {
1250 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1251 size_t i;
1252 for (i = 0; i < n_fe_opts; i++) {
1253 sset_destroy(&opts->targets);
1254 opts++;
1255 }
1256 free(fe_opts);
1257 }
1258 }
1259
1260 static void
1261 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1262 struct ofproto_port_stp_settings *port_s,
1263 int *port_num_counter, unsigned long *port_num_bitmap)
1264 {
1265 const char *config_str;
1266 struct iface *iface;
1267
1268 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1269 port_s->enable = false;
1270 return;
1271 } else {
1272 port_s->enable = true;
1273 }
1274
1275 /* STP over bonds is not supported. */
1276 if (!list_is_singleton(&port->ifaces)) {
1277 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1278 port->name);
1279 port_s->enable = false;
1280 return;
1281 }
1282
1283 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1284
1285 /* Internal ports shouldn't participate in spanning tree, so
1286 * skip them. */
1287 if (!strcmp(iface->type, "internal")) {
1288 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1289 port_s->enable = false;
1290 return;
1291 }
1292
1293 /* STP on mirror output ports is not supported. */
1294 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1295 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1296 port_s->enable = false;
1297 return;
1298 }
1299
1300 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1301 if (config_str) {
1302 unsigned long int port_num = strtoul(config_str, NULL, 0);
1303 int port_idx = port_num - 1;
1304
1305 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1306 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1307 port_s->enable = false;
1308 return;
1309 }
1310
1311 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1312 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1313 port->name, port_num);
1314 port_s->enable = false;
1315 return;
1316 }
1317 bitmap_set1(port_num_bitmap, port_idx);
1318 port_s->port_num = port_idx;
1319 } else {
1320 if (*port_num_counter >= STP_MAX_PORTS) {
1321 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1322 port_s->enable = false;
1323 return;
1324 }
1325
1326 port_s->port_num = (*port_num_counter)++;
1327 }
1328
1329 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1330 if (config_str) {
1331 port_s->path_cost = strtoul(config_str, NULL, 10);
1332 } else {
1333 enum netdev_features current;
1334 unsigned int mbps;
1335
1336 netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1337 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1338 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1339 }
1340
1341 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1342 if (config_str) {
1343 port_s->priority = strtoul(config_str, NULL, 0);
1344 } else {
1345 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1346 }
1347 }
1348
1349 static void
1350 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1351 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1352 {
1353 const char *config_str;
1354 struct iface *iface;
1355
1356 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1357 port_s->enable = false;
1358 return;
1359 } else {
1360 port_s->enable = true;
1361 }
1362
1363 /* RSTP over bonds is not supported. */
1364 if (!list_is_singleton(&port->ifaces)) {
1365 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1366 port->name);
1367 port_s->enable = false;
1368 return;
1369 }
1370
1371 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1372
1373 /* Internal ports shouldn't participate in spanning tree, so
1374 * skip them. */
1375 if (!strcmp(iface->type, "internal")) {
1376 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1377 port_s->enable = false;
1378 return;
1379 }
1380
1381 /* RSTP on mirror output ports is not supported. */
1382 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1383 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1384 port_s->enable = false;
1385 return;
1386 }
1387
1388 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1389 if (config_str) {
1390 unsigned long int port_num = strtoul(config_str, NULL, 0);
1391 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1392 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1393 port_s->enable = false;
1394 return;
1395 }
1396 port_s->port_num = port_num;
1397 } else {
1398 if (*port_num_counter >= RSTP_MAX_PORTS) {
1399 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1400 port_s->enable = false;
1401 return;
1402 }
1403 /* If rstp-port-num is not specified, use 0.
1404 * rstp_port_set_port_number() will look for the first free one. */
1405 port_s->port_num = 0;
1406 }
1407
1408 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1409 if (config_str) {
1410 port_s->path_cost = strtoul(config_str, NULL, 10);
1411 } else {
1412 enum netdev_features current;
1413 unsigned int mbps;
1414
1415 netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
1416 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1417 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1418 }
1419
1420 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1421 if (config_str) {
1422 port_s->priority = strtoul(config_str, NULL, 0);
1423 } else {
1424 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1425 }
1426
1427 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1428 if (config_str) {
1429 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1430 } else {
1431 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1432 }
1433
1434 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1435 "rstp-admin-port-state", true);
1436
1437 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1438 "rstp-port-admin-edge", false);
1439 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1440 "rstp-port-auto-edge", true);
1441 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1442 "rstp-port-mcheck", false);
1443 }
1444
1445 /* Set spanning tree configuration on 'br'. */
1446 static void
1447 bridge_configure_stp(struct bridge *br)
1448 {
1449 struct ofproto_rstp_status rstp_status;
1450
1451 ofproto_get_rstp_status(br->ofproto, &rstp_status);
1452 if (!br->cfg->stp_enable) {
1453 ofproto_set_stp(br->ofproto, NULL);
1454 } else if (rstp_status.enabled) {
1455 /* Do not activate STP if RSTP is enabled. */
1456 VLOG_ERR("STP cannot be enabled if RSTP is running.");
1457 ofproto_set_stp(br->ofproto, NULL);
1458 ovsrec_bridge_set_stp_enable(br->cfg, false);
1459 } else {
1460 struct ofproto_stp_settings br_s;
1461 const char *config_str;
1462 struct port *port;
1463 int port_num_counter;
1464 unsigned long *port_num_bitmap;
1465
1466 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1467 if (config_str) {
1468 uint8_t ea[ETH_ADDR_LEN];
1469
1470 if (eth_addr_from_string(config_str, ea)) {
1471 br_s.system_id = eth_addr_to_uint64(ea);
1472 } else {
1473 br_s.system_id = eth_addr_to_uint64(br->ea);
1474 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1475 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1476 }
1477 } else {
1478 br_s.system_id = eth_addr_to_uint64(br->ea);
1479 }
1480
1481 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1482 if (config_str) {
1483 br_s.priority = strtoul(config_str, NULL, 0);
1484 } else {
1485 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1486 }
1487
1488 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1489 if (config_str) {
1490 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1491 } else {
1492 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1493 }
1494
1495 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1496 if (config_str) {
1497 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1498 } else {
1499 br_s.max_age = STP_DEFAULT_MAX_AGE;
1500 }
1501
1502 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1503 if (config_str) {
1504 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1505 } else {
1506 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1507 }
1508
1509 /* Configure STP on the bridge. */
1510 if (ofproto_set_stp(br->ofproto, &br_s)) {
1511 VLOG_ERR("bridge %s: could not enable STP", br->name);
1512 return;
1513 }
1514
1515 /* Users must either set the port number with the "stp-port-num"
1516 * configuration on all ports or none. If manual configuration
1517 * is not done, then we allocate them sequentially. */
1518 port_num_counter = 0;
1519 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1520 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1521 struct ofproto_port_stp_settings port_s;
1522 struct iface *iface;
1523
1524 port_configure_stp(br->ofproto, port, &port_s,
1525 &port_num_counter, port_num_bitmap);
1526
1527 /* As bonds are not supported, just apply configuration to
1528 * all interfaces. */
1529 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1530 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1531 &port_s)) {
1532 VLOG_ERR("port %s: could not enable STP", port->name);
1533 continue;
1534 }
1535 }
1536 }
1537
1538 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1539 && port_num_counter) {
1540 VLOG_ERR("bridge %s: must manually configure all STP port "
1541 "IDs or none, disabling", br->name);
1542 ofproto_set_stp(br->ofproto, NULL);
1543 }
1544 bitmap_free(port_num_bitmap);
1545 }
1546 }
1547
1548 static void
1549 bridge_configure_rstp(struct bridge *br)
1550 {
1551 struct ofproto_stp_status stp_status;
1552
1553 ofproto_get_stp_status(br->ofproto, &stp_status);
1554 if (!br->cfg->rstp_enable) {
1555 ofproto_set_rstp(br->ofproto, NULL);
1556 } else if (stp_status.enabled) {
1557 /* Do not activate RSTP if STP is enabled. */
1558 VLOG_ERR("RSTP cannot be enabled if STP is running.");
1559 ofproto_set_rstp(br->ofproto, NULL);
1560 ovsrec_bridge_set_rstp_enable(br->cfg, false);
1561 } else {
1562 struct ofproto_rstp_settings br_s;
1563 const char *config_str;
1564 struct port *port;
1565 int port_num_counter;
1566
1567 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1568 if (config_str) {
1569 uint8_t ea[ETH_ADDR_LEN];
1570
1571 if (eth_addr_from_string(config_str, ea)) {
1572 br_s.address = eth_addr_to_uint64(ea);
1573 }
1574 else {
1575 br_s.address = eth_addr_to_uint64(br->ea);
1576 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1577 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1578 }
1579 }
1580 else {
1581 br_s.address = eth_addr_to_uint64(br->ea);
1582 }
1583
1584 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1585 if (config_str) {
1586 br_s.priority = strtoul(config_str, NULL, 0);
1587 } else {
1588 br_s.priority = RSTP_DEFAULT_PRIORITY;
1589 }
1590
1591 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1592 if (config_str) {
1593 br_s.ageing_time = strtoul(config_str, NULL, 0);
1594 } else {
1595 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1596 }
1597
1598 config_str = smap_get(&br->cfg->other_config,
1599 "rstp-force-protocol-version");
1600 if (config_str) {
1601 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1602 } else {
1603 br_s.force_protocol_version = FPV_DEFAULT;
1604 }
1605
1606 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1607 if (config_str) {
1608 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1609 } else {
1610 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1611 }
1612
1613 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1614 if (config_str) {
1615 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1616 } else {
1617 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1618 }
1619
1620 config_str = smap_get(&br->cfg->other_config,
1621 "rstp-transmit-hold-count");
1622 if (config_str) {
1623 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1624 } else {
1625 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1626 }
1627
1628 /* Configure RSTP on the bridge. */
1629 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1630 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1631 return;
1632 }
1633
1634 port_num_counter = 0;
1635 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1636 struct ofproto_port_rstp_settings port_s;
1637 struct iface *iface;
1638
1639 port_configure_rstp(br->ofproto, port, &port_s,
1640 &port_num_counter);
1641
1642 /* As bonds are not supported, just apply configuration to
1643 * all interfaces. */
1644 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1645 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1646 &port_s)) {
1647 VLOG_ERR("port %s: could not enable RSTP", port->name);
1648 continue;
1649 }
1650 }
1651 }
1652 }
1653 }
1654
1655 static bool
1656 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1657 {
1658 const struct port *port = port_lookup(br, name);
1659 return port && port_is_bond_fake_iface(port);
1660 }
1661
1662 static bool
1663 port_is_bond_fake_iface(const struct port *port)
1664 {
1665 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1666 }
1667
1668 static void
1669 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1670 {
1671 struct bridge *br, *next;
1672 struct shash new_br;
1673 size_t i;
1674
1675 /* Collect new bridges' names and types. */
1676 shash_init(&new_br);
1677 for (i = 0; i < cfg->n_bridges; i++) {
1678 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1679 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1680
1681 if (strchr(br_cfg->name, '/')) {
1682 /* Prevent remote ovsdb-server users from accessing arbitrary
1683 * directories, e.g. consider a bridge named "../../../etc/". */
1684 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1685 br_cfg->name);
1686 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1687 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1688 }
1689 }
1690
1691 /* Get rid of deleted bridges or those whose types have changed.
1692 * Update 'cfg' of bridges that still exist. */
1693 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1694 br->cfg = shash_find_data(&new_br, br->name);
1695 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1696 br->cfg->datapath_type))) {
1697 bridge_destroy(br);
1698 }
1699 }
1700
1701 /* Add new bridges. */
1702 for (i = 0; i < cfg->n_bridges; i++) {
1703 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1704 struct bridge *br = bridge_lookup(br_cfg->name);
1705 if (!br) {
1706 bridge_create(br_cfg);
1707 }
1708 }
1709
1710 shash_destroy(&new_br);
1711 }
1712
1713 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1714 * Returns 0 if successful, otherwise a positive errno value. */
1715 static int
1716 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1717 struct netdev *netdev, char **errp)
1718 {
1719 return netdev_set_config(netdev, &iface_cfg->options, errp);
1720 }
1721
1722 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1723 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1724 *
1725 * If successful, returns 0 and stores the network device in '*netdevp'. On
1726 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1727 static int
1728 iface_do_create(const struct bridge *br,
1729 const struct ovsrec_interface *iface_cfg,
1730 const struct ovsrec_port *port_cfg,
1731 ofp_port_t *ofp_portp, struct netdev **netdevp,
1732 char **errp)
1733 {
1734 struct netdev *netdev = NULL;
1735 int error;
1736
1737 if (netdev_is_reserved_name(iface_cfg->name)) {
1738 VLOG_WARN("could not create interface %s, name is reserved",
1739 iface_cfg->name);
1740 error = EINVAL;
1741 goto error;
1742 }
1743
1744 error = netdev_open(iface_cfg->name,
1745 iface_get_type(iface_cfg, br->cfg), &netdev);
1746 if (error) {
1747 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1748 iface_cfg->name, ovs_strerror(error));
1749 goto error;
1750 }
1751
1752 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1753 if (error) {
1754 goto error;
1755 }
1756
1757 *ofp_portp = iface_pick_ofport(iface_cfg);
1758 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1759 if (error) {
1760 goto error;
1761 }
1762
1763 VLOG_INFO("bridge %s: added interface %s on port %d",
1764 br->name, iface_cfg->name, *ofp_portp);
1765
1766 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1767 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1768 }
1769
1770 *netdevp = netdev;
1771 return 0;
1772
1773 error:
1774 *netdevp = NULL;
1775 netdev_close(netdev);
1776 return error;
1777 }
1778
1779 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1780 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1781 * automatically allocated for the iface. Takes ownership of and
1782 * deallocates 'if_cfg'.
1783 *
1784 * Return true if an iface is successfully created, false otherwise. */
1785 static bool
1786 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1787 const struct ovsrec_port *port_cfg)
1788 {
1789 struct netdev *netdev;
1790 struct iface *iface;
1791 ofp_port_t ofp_port;
1792 struct port *port;
1793 char *errp = NULL;
1794 int error;
1795
1796 /* Do the bits that can fail up front. */
1797 ovs_assert(!iface_lookup(br, iface_cfg->name));
1798 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1799 if (error) {
1800 iface_clear_db_record(iface_cfg, errp);
1801 free(errp);
1802 return false;
1803 }
1804
1805 /* Get or create the port structure. */
1806 port = port_lookup(br, port_cfg->name);
1807 if (!port) {
1808 port = port_create(br, port_cfg);
1809 }
1810
1811 /* Create the iface structure. */
1812 iface = xzalloc(sizeof *iface);
1813 list_push_back(&port->ifaces, &iface->port_elem);
1814 hmap_insert(&br->iface_by_name, &iface->name_node,
1815 hash_string(iface_cfg->name, 0));
1816 iface->port = port;
1817 iface->name = xstrdup(iface_cfg->name);
1818 iface->ofp_port = ofp_port;
1819 iface->netdev = netdev;
1820 iface->type = iface_get_type(iface_cfg, br->cfg);
1821 iface->cfg = iface_cfg;
1822 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1823 hash_ofp_port(ofp_port));
1824
1825 /* Populate initial status in database. */
1826 iface_refresh_stats(iface);
1827 iface_refresh_netdev_status(iface);
1828
1829 /* Add bond fake iface if necessary. */
1830 if (port_is_bond_fake_iface(port)) {
1831 struct ofproto_port ofproto_port;
1832
1833 if (ofproto_port_query_by_name(br->ofproto, port->name,
1834 &ofproto_port)) {
1835 struct netdev *netdev;
1836 int error;
1837
1838 error = netdev_open(port->name, "internal", &netdev);
1839 if (!error) {
1840 ofp_port_t fake_ofp_port = OFPP_NONE;
1841 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1842 netdev_close(netdev);
1843 } else {
1844 VLOG_WARN("could not open network device %s (%s)",
1845 port->name, ovs_strerror(error));
1846 }
1847 } else {
1848 /* Already exists, nothing to do. */
1849 ofproto_port_destroy(&ofproto_port);
1850 }
1851 }
1852
1853 return true;
1854 }
1855
1856 /* Set forward BPDU option. */
1857 static void
1858 bridge_configure_forward_bpdu(struct bridge *br)
1859 {
1860 ofproto_set_forward_bpdu(br->ofproto,
1861 smap_get_bool(&br->cfg->other_config,
1862 "forward-bpdu",
1863 false));
1864 }
1865
1866 /* Set MAC learning table configuration for 'br'. */
1867 static void
1868 bridge_configure_mac_table(struct bridge *br)
1869 {
1870 const char *idle_time_str;
1871 int idle_time;
1872
1873 const char *mac_table_size_str;
1874 int mac_table_size;
1875
1876 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1877 idle_time = (idle_time_str && atoi(idle_time_str)
1878 ? atoi(idle_time_str)
1879 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1880
1881 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1882 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1883 ? atoi(mac_table_size_str)
1884 : MAC_DEFAULT_MAX);
1885
1886 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1887 }
1888
1889 /* Set multicast snooping table configuration for 'br'. */
1890 static void
1891 bridge_configure_mcast_snooping(struct bridge *br)
1892 {
1893 if (!br->cfg->mcast_snooping_enable) {
1894 ofproto_set_mcast_snooping(br->ofproto, NULL);
1895 } else {
1896 struct port *port;
1897 struct ofproto_mcast_snooping_settings br_s;
1898 const char *idle_time_str;
1899 const char *max_entries_str;
1900
1901 idle_time_str = smap_get(&br->cfg->other_config,
1902 "mcast-snooping-aging-time");
1903 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1904 ? atoi(idle_time_str)
1905 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1906
1907 max_entries_str = smap_get(&br->cfg->other_config,
1908 "mcast-snooping-table-size");
1909 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1910 ? atoi(max_entries_str)
1911 : MCAST_DEFAULT_MAX_ENTRIES);
1912
1913 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1914 "mcast-snooping-disable-flood-unregistered",
1915 false);
1916
1917 /* Configure multicast snooping on the bridge */
1918 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1919 VLOG_ERR("bridge %s: could not enable multicast snooping",
1920 br->name);
1921 return;
1922 }
1923
1924 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1925 struct ofproto_mcast_snooping_port_settings port_s;
1926 port_s.flood = smap_get_bool(&port->cfg->other_config,
1927 "mcast-snooping-flood", false);
1928 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1929 "mcast-snooping-flood-reports", false);
1930 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1931 VLOG_ERR("port %s: could not configure mcast snooping",
1932 port->name);
1933 }
1934 }
1935 }
1936 }
1937
1938 static void
1939 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1940 const struct port *fake_br, struct iface **hw_addr_iface)
1941 {
1942 struct hmapx mirror_output_ports;
1943 struct port *port;
1944 bool found_addr = false;
1945 int error;
1946 int i;
1947
1948 /* Mirror output ports don't participate in picking the local hardware
1949 * address. ofproto can't help us find out whether a given port is a
1950 * mirror output because we haven't configured mirrors yet, so we need to
1951 * accumulate them ourselves. */
1952 hmapx_init(&mirror_output_ports);
1953 for (i = 0; i < br->cfg->n_mirrors; i++) {
1954 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1955 if (m->output_port) {
1956 hmapx_add(&mirror_output_ports, m->output_port);
1957 }
1958 }
1959
1960 /* Otherwise choose the minimum non-local MAC address among all of the
1961 * interfaces. */
1962 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1963 uint8_t iface_ea[ETH_ADDR_LEN];
1964 struct iface *candidate;
1965 struct iface *iface;
1966
1967 /* Mirror output ports don't participate. */
1968 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1969 continue;
1970 }
1971
1972 /* Choose the MAC address to represent the port. */
1973 iface = NULL;
1974 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1975 /* Find the interface with this Ethernet address (if any) so that
1976 * we can provide the correct devname to the caller. */
1977 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1978 uint8_t candidate_ea[ETH_ADDR_LEN];
1979 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1980 && eth_addr_equals(iface_ea, candidate_ea)) {
1981 iface = candidate;
1982 }
1983 }
1984 } else {
1985 /* Choose the interface whose MAC address will represent the port.
1986 * The Linux kernel bonding code always chooses the MAC address of
1987 * the first slave added to a bond, and the Fedora networking
1988 * scripts always add slaves to a bond in alphabetical order, so
1989 * for compatibility we choose the interface with the name that is
1990 * first in alphabetical order. */
1991 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1992 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1993 iface = candidate;
1994 }
1995 }
1996
1997 /* The local port doesn't count (since we're trying to choose its
1998 * MAC address anyway). */
1999 if (iface->ofp_port == OFPP_LOCAL) {
2000 continue;
2001 }
2002
2003 /* For fake bridges we only choose from ports with the same tag */
2004 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2005 if (!port->cfg->tag) {
2006 continue;
2007 }
2008 if (*port->cfg->tag != *fake_br->cfg->tag) {
2009 continue;
2010 }
2011 }
2012
2013 /* Grab MAC. */
2014 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2015 if (error) {
2016 continue;
2017 }
2018 }
2019
2020 /* Compare against our current choice. */
2021 if (!eth_addr_is_multicast(iface_ea) &&
2022 !eth_addr_is_local(iface_ea) &&
2023 !eth_addr_is_reserved(iface_ea) &&
2024 !eth_addr_is_zero(iface_ea) &&
2025 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2026 {
2027 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2028 *hw_addr_iface = iface;
2029 found_addr = true;
2030 }
2031 }
2032
2033 if (!found_addr) {
2034 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2035 *hw_addr_iface = NULL;
2036 }
2037
2038 hmapx_destroy(&mirror_output_ports);
2039 }
2040
2041 static void
2042 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2043 struct iface **hw_addr_iface)
2044 {
2045 const char *hwaddr;
2046 *hw_addr_iface = NULL;
2047
2048 /* Did the user request a particular MAC? */
2049 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2050 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2051 if (eth_addr_is_multicast(ea)) {
2052 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2053 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2054 } else if (eth_addr_is_zero(ea)) {
2055 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2056 } else {
2057 return;
2058 }
2059 }
2060
2061 /* Find a local hw address */
2062 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2063 }
2064
2065 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2066 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2067 * an interface on 'br', then that interface must be passed in as
2068 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2069 * 'hw_addr_iface' must be passed in as a null pointer. */
2070 static uint64_t
2071 bridge_pick_datapath_id(struct bridge *br,
2072 const uint8_t bridge_ea[ETH_ADDR_LEN],
2073 struct iface *hw_addr_iface)
2074 {
2075 /*
2076 * The procedure for choosing a bridge MAC address will, in the most
2077 * ordinary case, also choose a unique MAC that we can use as a datapath
2078 * ID. In some special cases, though, multiple bridges will end up with
2079 * the same MAC address. This is OK for the bridges, but it will confuse
2080 * the OpenFlow controller, because each datapath needs a unique datapath
2081 * ID.
2082 *
2083 * Datapath IDs must be unique. It is also very desirable that they be
2084 * stable from one run to the next, so that policy set on a datapath
2085 * "sticks".
2086 */
2087 const char *datapath_id;
2088 uint64_t dpid;
2089
2090 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2091 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2092 return dpid;
2093 }
2094
2095 if (!hw_addr_iface) {
2096 /*
2097 * A purely internal bridge, that is, one that has no non-virtual
2098 * network devices on it at all, is difficult because it has no
2099 * natural unique identifier at all.
2100 *
2101 * When the host is a XenServer, we handle this case by hashing the
2102 * host's UUID with the name of the bridge. Names of bridges are
2103 * persistent across XenServer reboots, although they can be reused if
2104 * an internal network is destroyed and then a new one is later
2105 * created, so this is fairly effective.
2106 *
2107 * When the host is not a XenServer, we punt by using a random MAC
2108 * address on each run.
2109 */
2110 const char *host_uuid = xenserver_get_host_uuid();
2111 if (host_uuid) {
2112 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2113 dpid = dpid_from_hash(combined, strlen(combined));
2114 free(combined);
2115 return dpid;
2116 }
2117 }
2118
2119 return eth_addr_to_uint64(bridge_ea);
2120 }
2121
2122 static uint64_t
2123 dpid_from_hash(const void *data, size_t n)
2124 {
2125 uint8_t hash[SHA1_DIGEST_SIZE];
2126
2127 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2128 sha1_bytes(data, n, hash);
2129 eth_addr_mark_random(hash);
2130 return eth_addr_to_uint64(hash);
2131 }
2132
2133 static void
2134 iface_refresh_netdev_status(struct iface *iface)
2135 {
2136 struct smap smap;
2137
2138 enum netdev_features current;
2139 enum netdev_flags flags;
2140 const char *link_state;
2141 uint8_t mac[ETH_ADDR_LEN];
2142 int64_t bps, mtu_64, ifindex64, link_resets;
2143 int mtu, error;
2144
2145 if (iface_is_synthetic(iface)) {
2146 return;
2147 }
2148
2149 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2150 && !status_txn_try_again) {
2151 return;
2152 }
2153
2154 iface->change_seq = netdev_get_change_seq(iface->netdev);
2155
2156 smap_init(&smap);
2157
2158 if (!netdev_get_status(iface->netdev, &smap)) {
2159 ovsrec_interface_set_status(iface->cfg, &smap);
2160 } else {
2161 ovsrec_interface_set_status(iface->cfg, NULL);
2162 }
2163
2164 smap_destroy(&smap);
2165
2166 error = netdev_get_flags(iface->netdev, &flags);
2167 if (!error) {
2168 const char *state = flags & NETDEV_UP ? "up" : "down";
2169
2170 ovsrec_interface_set_admin_state(iface->cfg, state);
2171 } else {
2172 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2173 }
2174
2175 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2176 ovsrec_interface_set_link_state(iface->cfg, link_state);
2177
2178 link_resets = netdev_get_carrier_resets(iface->netdev);
2179 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2180
2181 error = netdev_get_features(iface->netdev, &current, NULL, NULL, NULL);
2182 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2183 if (bps) {
2184 ovsrec_interface_set_duplex(iface->cfg,
2185 netdev_features_is_full_duplex(current)
2186 ? "full" : "half");
2187 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2188 } else {
2189 ovsrec_interface_set_duplex(iface->cfg, NULL);
2190 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2191 }
2192
2193 error = netdev_get_mtu(iface->netdev, &mtu);
2194 if (!error) {
2195 mtu_64 = mtu;
2196 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2197 } else {
2198 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2199 }
2200
2201 error = netdev_get_etheraddr(iface->netdev, mac);
2202 if (!error) {
2203 char mac_string[32];
2204
2205 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2206 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2207 } else {
2208 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2209 }
2210
2211 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2212 * if there is no valid ifindex number. */
2213 ifindex64 = netdev_get_ifindex(iface->netdev);
2214 if (ifindex64 < 0) {
2215 ifindex64 = 0;
2216 }
2217 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2218 }
2219
2220 static void
2221 iface_refresh_ofproto_status(struct iface *iface)
2222 {
2223 int current;
2224
2225 if (iface_is_synthetic(iface)) {
2226 return;
2227 }
2228
2229 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2230 iface->ofp_port);
2231 if (current >= 0) {
2232 bool bl = current;
2233 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2234 } else {
2235 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2236 }
2237
2238 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2239 iface->ofp_port)
2240 || status_txn_try_again) {
2241 iface_refresh_cfm_stats(iface);
2242 }
2243
2244 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2245 iface->ofp_port)
2246 || status_txn_try_again) {
2247 struct smap smap;
2248
2249 smap_init(&smap);
2250 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2251 iface->ofp_port, &smap);
2252 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2253 smap_destroy(&smap);
2254 }
2255 }
2256
2257 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2258 * synthetic. */
2259 static void
2260 iface_refresh_cfm_stats(struct iface *iface)
2261 {
2262 const struct ovsrec_interface *cfg = iface->cfg;
2263 struct cfm_status status;
2264 int error;
2265
2266 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2267 iface->ofp_port, &status);
2268 if (error > 0) {
2269 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2270 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2271 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2272 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2273 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2274 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2275 } else {
2276 const char *reasons[CFM_FAULT_N_REASONS];
2277 int64_t cfm_health = status.health;
2278 int64_t cfm_flap_count = status.flap_count;
2279 bool faulted = status.faults != 0;
2280 size_t i, j;
2281
2282 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2283
2284 j = 0;
2285 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2286 int reason = 1 << i;
2287 if (status.faults & reason) {
2288 reasons[j++] = cfm_fault_reason_to_str(reason);
2289 }
2290 }
2291 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2292
2293 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2294
2295 if (status.remote_opstate >= 0) {
2296 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2297 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2298 } else {
2299 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2300 }
2301
2302 ovsrec_interface_set_cfm_remote_mpids(cfg,
2303 (const int64_t *)status.rmps,
2304 status.n_rmps);
2305 if (cfm_health >= 0) {
2306 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2307 } else {
2308 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2309 }
2310
2311 free(status.rmps);
2312 }
2313 }
2314
2315 static void
2316 iface_refresh_stats(struct iface *iface)
2317 {
2318 #define IFACE_STATS \
2319 IFACE_STAT(rx_packets, "rx_packets") \
2320 IFACE_STAT(tx_packets, "tx_packets") \
2321 IFACE_STAT(rx_bytes, "rx_bytes") \
2322 IFACE_STAT(tx_bytes, "tx_bytes") \
2323 IFACE_STAT(rx_dropped, "rx_dropped") \
2324 IFACE_STAT(tx_dropped, "tx_dropped") \
2325 IFACE_STAT(rx_errors, "rx_errors") \
2326 IFACE_STAT(tx_errors, "tx_errors") \
2327 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2328 IFACE_STAT(rx_over_errors, "rx_over_err") \
2329 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2330 IFACE_STAT(collisions, "collisions")
2331
2332 #define IFACE_STAT(MEMBER, NAME) + 1
2333 enum { N_IFACE_STATS = IFACE_STATS };
2334 #undef IFACE_STAT
2335 int64_t values[N_IFACE_STATS];
2336 const char *keys[N_IFACE_STATS];
2337 int n;
2338
2339 struct netdev_stats stats;
2340
2341 if (iface_is_synthetic(iface)) {
2342 return;
2343 }
2344
2345 /* Intentionally ignore return value, since errors will set 'stats' to
2346 * all-1s, and we will deal with that correctly below. */
2347 netdev_get_stats(iface->netdev, &stats);
2348
2349 /* Copy statistics into keys[] and values[]. */
2350 n = 0;
2351 #define IFACE_STAT(MEMBER, NAME) \
2352 if (stats.MEMBER != UINT64_MAX) { \
2353 keys[n] = NAME; \
2354 values[n] = stats.MEMBER; \
2355 n++; \
2356 }
2357 IFACE_STATS;
2358 #undef IFACE_STAT
2359 ovs_assert(n <= N_IFACE_STATS);
2360
2361 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2362 #undef IFACE_STATS
2363 }
2364
2365 static void
2366 br_refresh_datapath_info(struct bridge *br)
2367 {
2368 const char *version;
2369
2370 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2371 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2372 : NULL);
2373
2374 ovsrec_bridge_set_datapath_version(br->cfg,
2375 version ? version : "<unknown>");
2376 }
2377
2378 static void
2379 br_refresh_stp_status(struct bridge *br)
2380 {
2381 struct smap smap = SMAP_INITIALIZER(&smap);
2382 struct ofproto *ofproto = br->ofproto;
2383 struct ofproto_stp_status status;
2384
2385 if (ofproto_get_stp_status(ofproto, &status)) {
2386 return;
2387 }
2388
2389 if (!status.enabled) {
2390 ovsrec_bridge_set_status(br->cfg, NULL);
2391 return;
2392 }
2393
2394 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2395 STP_ID_ARGS(status.bridge_id));
2396 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2397 STP_ID_ARGS(status.designated_root));
2398 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2399
2400 ovsrec_bridge_set_status(br->cfg, &smap);
2401 smap_destroy(&smap);
2402 }
2403
2404 static void
2405 port_refresh_stp_status(struct port *port)
2406 {
2407 struct ofproto *ofproto = port->bridge->ofproto;
2408 struct iface *iface;
2409 struct ofproto_port_stp_status status;
2410 struct smap smap;
2411
2412 if (port_is_synthetic(port)) {
2413 return;
2414 }
2415
2416 /* STP doesn't currently support bonds. */
2417 if (!list_is_singleton(&port->ifaces)) {
2418 ovsrec_port_set_status(port->cfg, NULL);
2419 return;
2420 }
2421
2422 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2423 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2424 return;
2425 }
2426
2427 if (!status.enabled) {
2428 ovsrec_port_set_status(port->cfg, NULL);
2429 return;
2430 }
2431
2432 /* Set Status column. */
2433 smap_init(&smap);
2434 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2435 smap_add(&smap, "stp_state", stp_state_name(status.state));
2436 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2437 smap_add(&smap, "stp_role", stp_role_name(status.role));
2438 ovsrec_port_set_status(port->cfg, &smap);
2439 smap_destroy(&smap);
2440 }
2441
2442 static void
2443 port_refresh_stp_stats(struct port *port)
2444 {
2445 struct ofproto *ofproto = port->bridge->ofproto;
2446 struct iface *iface;
2447 struct ofproto_port_stp_stats stats;
2448 const char *keys[3];
2449 int64_t int_values[3];
2450
2451 if (port_is_synthetic(port)) {
2452 return;
2453 }
2454
2455 /* STP doesn't currently support bonds. */
2456 if (!list_is_singleton(&port->ifaces)) {
2457 return;
2458 }
2459
2460 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2461 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2462 return;
2463 }
2464
2465 if (!stats.enabled) {
2466 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2467 return;
2468 }
2469
2470 /* Set Statistics column. */
2471 keys[0] = "stp_tx_count";
2472 int_values[0] = stats.tx_count;
2473 keys[1] = "stp_rx_count";
2474 int_values[1] = stats.rx_count;
2475 keys[2] = "stp_error_count";
2476 int_values[2] = stats.error_count;
2477
2478 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2479 ARRAY_SIZE(int_values));
2480 }
2481
2482 static void
2483 br_refresh_rstp_status(struct bridge *br)
2484 {
2485 struct smap smap = SMAP_INITIALIZER(&smap);
2486 struct ofproto *ofproto = br->ofproto;
2487 struct ofproto_rstp_status status;
2488
2489 if (ofproto_get_rstp_status(ofproto, &status)) {
2490 return;
2491 }
2492 if (!status.enabled) {
2493 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2494 return;
2495 }
2496 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2497 RSTP_ID_ARGS(status.bridge_id));
2498 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2499 status.root_path_cost);
2500 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2501 RSTP_ID_ARGS(status.root_id));
2502 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2503 RSTP_ID_ARGS(status.designated_id));
2504 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2505 status.designated_port_id);
2506 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2507 status.bridge_port_id);
2508 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2509 smap_destroy(&smap);
2510 }
2511
2512 static void
2513 port_refresh_rstp_status(struct port *port)
2514 {
2515 struct ofproto *ofproto = port->bridge->ofproto;
2516 struct iface *iface;
2517 struct ofproto_port_rstp_status status;
2518 const char *keys[4];
2519 int64_t int_values[4];
2520 struct smap smap;
2521
2522 if (port_is_synthetic(port)) {
2523 return;
2524 }
2525
2526 /* RSTP doesn't currently support bonds. */
2527 if (!list_is_singleton(&port->ifaces)) {
2528 ovsrec_port_set_rstp_status(port->cfg, NULL);
2529 return;
2530 }
2531
2532 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2533 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2534 return;
2535 }
2536
2537 if (!status.enabled) {
2538 ovsrec_port_set_rstp_status(port->cfg, NULL);
2539 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2540 return;
2541 }
2542 /* Set Status column. */
2543 smap_init(&smap);
2544
2545 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2546 status.port_id);
2547 smap_add_format(&smap, "rstp_port_role", "%s",
2548 rstp_port_role_name(status.role));
2549 smap_add_format(&smap, "rstp_port_state", "%s",
2550 rstp_state_name(status.state));
2551 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2552 RSTP_ID_ARGS(status.designated_bridge_id));
2553 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2554 status.designated_port_id);
2555 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2556 status.designated_path_cost);
2557
2558 ovsrec_port_set_rstp_status(port->cfg, &smap);
2559 smap_destroy(&smap);
2560
2561 /* Set Statistics column. */
2562 keys[0] = "rstp_tx_count";
2563 int_values[0] = status.tx_count;
2564 keys[1] = "rstp_rx_count";
2565 int_values[1] = status.rx_count;
2566 keys[2] = "rstp_uptime";
2567 int_values[2] = status.uptime;
2568 keys[3] = "rstp_error_count";
2569 int_values[3] = status.error_count;
2570 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2571 ARRAY_SIZE(int_values));
2572 }
2573
2574 static void
2575 port_refresh_bond_status(struct port *port, bool force_update)
2576 {
2577 uint8_t mac[ETH_ADDR_LEN];
2578
2579 /* Return if port is not a bond */
2580 if (list_is_singleton(&port->ifaces)) {
2581 return;
2582 }
2583
2584 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2585 struct ds mac_s;
2586
2587 ds_init(&mac_s);
2588 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2589 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2590 ds_destroy(&mac_s);
2591 }
2592 }
2593
2594 static bool
2595 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2596 {
2597 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2598 }
2599
2600 static void
2601 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2602 {
2603 bool enable = enable_system_stats(cfg);
2604
2605 system_stats_enable(enable);
2606 if (!enable) {
2607 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2608 }
2609 }
2610
2611 static void
2612 run_system_stats(void)
2613 {
2614 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2615 struct smap *stats;
2616
2617 stats = system_stats_run();
2618 if (stats && cfg) {
2619 struct ovsdb_idl_txn *txn;
2620 struct ovsdb_datum datum;
2621
2622 txn = ovsdb_idl_txn_create(idl);
2623 ovsdb_datum_from_smap(&datum, stats);
2624 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2625 &datum);
2626 ovsdb_idl_txn_commit(txn);
2627 ovsdb_idl_txn_destroy(txn);
2628
2629 free(stats);
2630 }
2631 }
2632
2633 static const char *
2634 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2635 {
2636 switch (role) {
2637 case OFPCR12_ROLE_EQUAL:
2638 return "other";
2639 case OFPCR12_ROLE_MASTER:
2640 return "master";
2641 case OFPCR12_ROLE_SLAVE:
2642 return "slave";
2643 case OFPCR12_ROLE_NOCHANGE:
2644 default:
2645 return "*** INVALID ROLE ***";
2646 }
2647 }
2648
2649 static void
2650 refresh_controller_status(void)
2651 {
2652 struct bridge *br;
2653 struct shash info;
2654 const struct ovsrec_controller *cfg;
2655
2656 shash_init(&info);
2657
2658 /* Accumulate status for controllers on all bridges. */
2659 HMAP_FOR_EACH (br, node, &all_bridges) {
2660 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2661 }
2662
2663 /* Update each controller in the database with current status. */
2664 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2665 struct ofproto_controller_info *cinfo =
2666 shash_find_data(&info, cfg->target);
2667
2668 if (cinfo) {
2669 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2670 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2671 cinfo->role));
2672 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2673 } else {
2674 ovsrec_controller_set_is_connected(cfg, false);
2675 ovsrec_controller_set_role(cfg, NULL);
2676 ovsrec_controller_set_status(cfg, NULL);
2677 }
2678 }
2679
2680 ofproto_free_ofproto_controller_info(&info);
2681 }
2682 \f
2683 /* Update interface and mirror statistics if necessary. */
2684 static void
2685 run_stats_update(void)
2686 {
2687 static struct ovsdb_idl_txn *stats_txn;
2688 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2689 int stats_interval;
2690
2691 if (!cfg) {
2692 return;
2693 }
2694
2695 /* Statistics update interval should always be greater than or equal to
2696 * 5000 ms. */
2697 stats_interval = MAX(smap_get_int(&cfg->other_config,
2698 "stats-update-interval",
2699 5000), 5000);
2700 if (stats_timer_interval != stats_interval) {
2701 stats_timer_interval = stats_interval;
2702 stats_timer = LLONG_MIN;
2703 }
2704
2705 if (time_msec() >= stats_timer) {
2706 enum ovsdb_idl_txn_status status;
2707
2708 /* Rate limit the update. Do not start a new update if the
2709 * previous one is not done. */
2710 if (!stats_txn) {
2711 struct bridge *br;
2712
2713 stats_txn = ovsdb_idl_txn_create(idl);
2714 HMAP_FOR_EACH (br, node, &all_bridges) {
2715 struct port *port;
2716 struct mirror *m;
2717
2718 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2719 struct iface *iface;
2720
2721 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2722 iface_refresh_stats(iface);
2723 }
2724 port_refresh_stp_stats(port);
2725 }
2726 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2727 mirror_refresh_stats(m);
2728 }
2729 }
2730 refresh_controller_status();
2731 }
2732
2733 status = ovsdb_idl_txn_commit(stats_txn);
2734 if (status != TXN_INCOMPLETE) {
2735 stats_timer = time_msec() + stats_timer_interval;
2736 ovsdb_idl_txn_destroy(stats_txn);
2737 stats_txn = NULL;
2738 }
2739 }
2740 }
2741
2742 /* Update bridge/port/interface status if necessary. */
2743 static void
2744 run_status_update(void)
2745 {
2746 if (!status_txn) {
2747 uint64_t seq;
2748
2749 /* Rate limit the update. Do not start a new update if the
2750 * previous one is not done. */
2751 seq = seq_read(connectivity_seq_get());
2752 if (seq != connectivity_seqno || status_txn_try_again) {
2753 struct bridge *br;
2754
2755 connectivity_seqno = seq;
2756 status_txn = ovsdb_idl_txn_create(idl);
2757 HMAP_FOR_EACH (br, node, &all_bridges) {
2758 struct port *port;
2759
2760 br_refresh_stp_status(br);
2761 br_refresh_rstp_status(br);
2762 br_refresh_datapath_info(br);
2763 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2764 struct iface *iface;
2765
2766 port_refresh_stp_status(port);
2767 port_refresh_rstp_status(port);
2768 port_refresh_bond_status(port, status_txn_try_again);
2769 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2770 iface_refresh_netdev_status(iface);
2771 iface_refresh_ofproto_status(iface);
2772 }
2773 }
2774 }
2775 }
2776 }
2777
2778 /* Commit the transaction and get the status. If the transaction finishes,
2779 * then destroy the transaction. Otherwise, keep it so that we can check
2780 * progress the next time that this function is called. */
2781 if (status_txn) {
2782 enum ovsdb_idl_txn_status status;
2783
2784 status = ovsdb_idl_txn_commit(status_txn);
2785 if (status != TXN_INCOMPLETE) {
2786 ovsdb_idl_txn_destroy(status_txn);
2787 status_txn = NULL;
2788
2789 /* Sets the 'status_txn_try_again' if the transaction fails. */
2790 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2791 status_txn_try_again = false;
2792 } else {
2793 status_txn_try_again = true;
2794 }
2795 }
2796 }
2797
2798 /* Refresh AA port status if necessary. */
2799 if (time_msec() >= aa_refresh_timer) {
2800 struct bridge *br;
2801
2802 HMAP_FOR_EACH (br, node, &all_bridges) {
2803 if (bridge_aa_need_refresh(br)) {
2804 struct ovsdb_idl_txn *txn;
2805
2806 txn = ovsdb_idl_txn_create(idl);
2807 bridge_aa_refresh_queued(br);
2808 ovsdb_idl_txn_commit(txn);
2809 ovsdb_idl_txn_destroy(txn);
2810 }
2811 }
2812
2813 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2814 }
2815 }
2816
2817 static void
2818 status_update_wait(void)
2819 {
2820 /* This prevents the process from constantly waking up on
2821 * connectivity seq, when there is no connection to ovsdb. */
2822 if (!ovsdb_idl_has_lock(idl)) {
2823 return;
2824 }
2825
2826 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2827 * transaction to complete. If the status update to database needs to be
2828 * run again (transaction fails), registers a timeout in
2829 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2830 * sequence number. */
2831 if (status_txn) {
2832 ovsdb_idl_txn_wait(status_txn);
2833 } else if (status_txn_try_again) {
2834 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2835 } else {
2836 seq_wait(connectivity_seq_get(), connectivity_seqno);
2837 }
2838 }
2839
2840 static void
2841 bridge_run__(void)
2842 {
2843 struct bridge *br;
2844 struct sset types;
2845 const char *type;
2846
2847 /* Let each datapath type do the work that it needs to do. */
2848 sset_init(&types);
2849 ofproto_enumerate_types(&types);
2850 SSET_FOR_EACH (type, &types) {
2851 ofproto_type_run(type);
2852 }
2853 sset_destroy(&types);
2854
2855 /* Let each bridge do the work that it needs to do. */
2856 HMAP_FOR_EACH (br, node, &all_bridges) {
2857 ofproto_run(br->ofproto);
2858 }
2859 }
2860
2861 void
2862 bridge_run(void)
2863 {
2864 static struct ovsrec_open_vswitch null_cfg;
2865 const struct ovsrec_open_vswitch *cfg;
2866
2867 bool vlan_splinters_changed;
2868
2869 ovsrec_open_vswitch_init(&null_cfg);
2870
2871 ovsdb_idl_run(idl);
2872
2873 if (ovsdb_idl_is_lock_contended(idl)) {
2874 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2875 struct bridge *br, *next_br;
2876
2877 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2878 "disabling this process (pid %ld) until it goes away",
2879 (long int) getpid());
2880
2881 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2882 bridge_destroy(br);
2883 }
2884 /* Since we will not be running system_stats_run() in this process
2885 * with the current situation of multiple ovs-vswitchd daemons,
2886 * disable system stats collection. */
2887 system_stats_enable(false);
2888 return;
2889 } else if (!ovsdb_idl_has_lock(idl)) {
2890 return;
2891 }
2892 cfg = ovsrec_open_vswitch_first(idl);
2893
2894 /* Initialize the ofproto library. This only needs to run once, but
2895 * it must be done after the configuration is set. If the
2896 * initialization has already occurred, bridge_init_ofproto()
2897 * returns immediately. */
2898 bridge_init_ofproto(cfg);
2899
2900 /* Once the value of flow-restore-wait is false, we no longer should
2901 * check its value from the database. */
2902 if (cfg && ofproto_get_flow_restore_wait()) {
2903 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2904 "flow-restore-wait", false));
2905 }
2906
2907 bridge_run__();
2908
2909 /* Re-configure SSL. We do this on every trip through the main loop,
2910 * instead of just when the database changes, because the contents of the
2911 * key and certificate files can change without the database changing.
2912 *
2913 * We do this before bridge_reconfigure() because that function might
2914 * initiate SSL connections and thus requires SSL to be configured. */
2915 if (cfg && cfg->ssl) {
2916 const struct ovsrec_ssl *ssl = cfg->ssl;
2917
2918 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2919 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2920 }
2921
2922 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2923 * usage has changed. */
2924 vlan_splinters_changed = false;
2925 if (vlan_splinters_enabled_anywhere) {
2926 struct bridge *br;
2927
2928 HMAP_FOR_EACH (br, node, &all_bridges) {
2929 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2930 vlan_splinters_changed = true;
2931 break;
2932 }
2933 }
2934 }
2935
2936 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2937 struct ovsdb_idl_txn *txn;
2938
2939 idl_seqno = ovsdb_idl_get_seqno(idl);
2940 txn = ovsdb_idl_txn_create(idl);
2941 bridge_reconfigure(cfg ? cfg : &null_cfg);
2942
2943 if (cfg) {
2944 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2945 }
2946
2947 /* If we are completing our initial configuration for this run
2948 * of ovs-vswitchd, then keep the transaction around to monitor
2949 * it for completion. */
2950 if (initial_config_done) {
2951 /* Always sets the 'status_txn_try_again' to check again,
2952 * in case that this transaction fails. */
2953 status_txn_try_again = true;
2954 ovsdb_idl_txn_commit(txn);
2955 ovsdb_idl_txn_destroy(txn);
2956 } else {
2957 initial_config_done = true;
2958 daemonize_txn = txn;
2959 }
2960 }
2961
2962 if (daemonize_txn) {
2963 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2964 if (status != TXN_INCOMPLETE) {
2965 ovsdb_idl_txn_destroy(daemonize_txn);
2966 daemonize_txn = NULL;
2967
2968 /* ovs-vswitchd has completed initialization, so allow the
2969 * process that forked us to exit successfully. */
2970 daemonize_complete();
2971
2972 vlog_enable_async();
2973
2974 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2975 }
2976 }
2977
2978 run_stats_update();
2979 run_status_update();
2980 run_system_stats();
2981 }
2982
2983 void
2984 bridge_wait(void)
2985 {
2986 struct sset types;
2987 const char *type;
2988
2989 ovsdb_idl_wait(idl);
2990 if (daemonize_txn) {
2991 ovsdb_idl_txn_wait(daemonize_txn);
2992 }
2993
2994 sset_init(&types);
2995 ofproto_enumerate_types(&types);
2996 SSET_FOR_EACH (type, &types) {
2997 ofproto_type_wait(type);
2998 }
2999 sset_destroy(&types);
3000
3001 if (!hmap_is_empty(&all_bridges)) {
3002 struct bridge *br;
3003
3004 HMAP_FOR_EACH (br, node, &all_bridges) {
3005 ofproto_wait(br->ofproto);
3006 }
3007
3008 poll_timer_wait_until(stats_timer);
3009 }
3010
3011 status_update_wait();
3012 system_stats_wait();
3013 }
3014
3015 /* Adds some memory usage statistics for bridges into 'usage', for use with
3016 * memory_report(). */
3017 void
3018 bridge_get_memory_usage(struct simap *usage)
3019 {
3020 struct bridge *br;
3021 struct sset types;
3022 const char *type;
3023
3024 sset_init(&types);
3025 ofproto_enumerate_types(&types);
3026 SSET_FOR_EACH (type, &types) {
3027 ofproto_type_get_memory_usage(type, usage);
3028 }
3029 sset_destroy(&types);
3030
3031 HMAP_FOR_EACH (br, node, &all_bridges) {
3032 ofproto_get_memory_usage(br->ofproto, usage);
3033 }
3034 }
3035 \f
3036 /* QoS unixctl user interface functions. */
3037
3038 struct qos_unixctl_show_cbdata {
3039 struct ds *ds;
3040 struct iface *iface;
3041 };
3042
3043 static void
3044 qos_unixctl_show_queue(unsigned int queue_id,
3045 const struct smap *details,
3046 struct iface *iface,
3047 struct ds *ds)
3048 {
3049 struct netdev_queue_stats stats;
3050 struct smap_node *node;
3051 int error;
3052
3053 ds_put_cstr(ds, "\n");
3054 if (queue_id) {
3055 ds_put_format(ds, "Queue %u:\n", queue_id);
3056 } else {
3057 ds_put_cstr(ds, "Default:\n");
3058 }
3059
3060 SMAP_FOR_EACH (node, details) {
3061 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3062 }
3063
3064 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3065 if (!error) {
3066 if (stats.tx_packets != UINT64_MAX) {
3067 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3068 }
3069
3070 if (stats.tx_bytes != UINT64_MAX) {
3071 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3072 }
3073
3074 if (stats.tx_errors != UINT64_MAX) {
3075 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3076 }
3077 } else {
3078 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3079 queue_id, ovs_strerror(error));
3080 }
3081 }
3082
3083 static void
3084 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3085 const char *argv[], void *aux OVS_UNUSED)
3086 {
3087 struct ds ds = DS_EMPTY_INITIALIZER;
3088 struct smap smap = SMAP_INITIALIZER(&smap);
3089 struct iface *iface;
3090 const char *type;
3091 struct smap_node *node;
3092
3093 iface = iface_find(argv[1]);
3094 if (!iface) {
3095 unixctl_command_reply_error(conn, "no such interface");
3096 return;
3097 }
3098
3099 netdev_get_qos(iface->netdev, &type, &smap);
3100
3101 if (*type != '\0') {
3102 struct netdev_queue_dump dump;
3103 struct smap details;
3104 unsigned int queue_id;
3105
3106 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3107
3108 SMAP_FOR_EACH (node, &smap) {
3109 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3110 }
3111
3112 smap_init(&details);
3113 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3114 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3115 }
3116 smap_destroy(&details);
3117
3118 unixctl_command_reply(conn, ds_cstr(&ds));
3119 } else {
3120 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3121 unixctl_command_reply_error(conn, ds_cstr(&ds));
3122 }
3123
3124 smap_destroy(&smap);
3125 ds_destroy(&ds);
3126 }
3127 \f
3128 /* Bridge reconfiguration functions. */
3129 static void
3130 bridge_create(const struct ovsrec_bridge *br_cfg)
3131 {
3132 struct bridge *br;
3133
3134 ovs_assert(!bridge_lookup(br_cfg->name));
3135 br = xzalloc(sizeof *br);
3136
3137 br->name = xstrdup(br_cfg->name);
3138 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3139 br->cfg = br_cfg;
3140
3141 /* Derive the default Ethernet address from the bridge's UUID. This should
3142 * be unique and it will be stable between ovs-vswitchd runs. */
3143 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3144 eth_addr_mark_random(br->default_ea);
3145
3146 hmap_init(&br->ports);
3147 hmap_init(&br->ifaces);
3148 hmap_init(&br->iface_by_name);
3149 hmap_init(&br->mirrors);
3150
3151 hmap_init(&br->mappings);
3152 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3153 }
3154
3155 static void
3156 bridge_destroy(struct bridge *br)
3157 {
3158 if (br) {
3159 struct mirror *mirror, *next_mirror;
3160 struct port *port, *next_port;
3161
3162 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3163 port_destroy(port);
3164 }
3165 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3166 mirror_destroy(mirror);
3167 }
3168
3169 hmap_remove(&all_bridges, &br->node);
3170 ofproto_destroy(br->ofproto);
3171 hmap_destroy(&br->ifaces);
3172 hmap_destroy(&br->ports);
3173 hmap_destroy(&br->iface_by_name);
3174 hmap_destroy(&br->mirrors);
3175 hmap_destroy(&br->mappings);
3176 free(br->name);
3177 free(br->type);
3178 free(br);
3179 }
3180 }
3181
3182 static struct bridge *
3183 bridge_lookup(const char *name)
3184 {
3185 struct bridge *br;
3186
3187 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3188 if (!strcmp(br->name, name)) {
3189 return br;
3190 }
3191 }
3192 return NULL;
3193 }
3194
3195 /* Handle requests for a listing of all flows known by the OpenFlow
3196 * stack, including those normally hidden. */
3197 static void
3198 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3199 const char *argv[], void *aux OVS_UNUSED)
3200 {
3201 struct bridge *br;
3202 struct ds results;
3203
3204 br = bridge_lookup(argv[1]);
3205 if (!br) {
3206 unixctl_command_reply_error(conn, "Unknown bridge");
3207 return;
3208 }
3209
3210 ds_init(&results);
3211 ofproto_get_all_flows(br->ofproto, &results);
3212
3213 unixctl_command_reply(conn, ds_cstr(&results));
3214 ds_destroy(&results);
3215 }
3216
3217 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3218 * connections and reconnect. If BRIDGE is not specified, then all bridges
3219 * drop their controller connections and reconnect. */
3220 static void
3221 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3222 const char *argv[], void *aux OVS_UNUSED)
3223 {
3224 struct bridge *br;
3225 if (argc > 1) {
3226 br = bridge_lookup(argv[1]);
3227 if (!br) {
3228 unixctl_command_reply_error(conn, "Unknown bridge");
3229 return;
3230 }
3231 ofproto_reconnect_controllers(br->ofproto);
3232 } else {
3233 HMAP_FOR_EACH (br, node, &all_bridges) {
3234 ofproto_reconnect_controllers(br->ofproto);
3235 }
3236 }
3237 unixctl_command_reply(conn, NULL);
3238 }
3239
3240 static size_t
3241 bridge_get_controllers(const struct bridge *br,
3242 struct ovsrec_controller ***controllersp)
3243 {
3244 struct ovsrec_controller **controllers;
3245 size_t n_controllers;
3246
3247 controllers = br->cfg->controller;
3248 n_controllers = br->cfg->n_controller;
3249
3250 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3251 controllers = NULL;
3252 n_controllers = 0;
3253 }
3254
3255 if (controllersp) {
3256 *controllersp = controllers;
3257 }
3258 return n_controllers;
3259 }
3260
3261 static void
3262 bridge_collect_wanted_ports(struct bridge *br,
3263 const unsigned long int *splinter_vlans,
3264 struct shash *wanted_ports)
3265 {
3266 size_t i;
3267
3268 shash_init(wanted_ports);
3269
3270 for (i = 0; i < br->cfg->n_ports; i++) {
3271 const char *name = br->cfg->ports[i]->name;
3272 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3273 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3274 br->name, name);
3275 }
3276 }
3277 if (bridge_get_controllers(br, NULL)
3278 && !shash_find(wanted_ports, br->name)) {
3279 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3280 br->name, br->name);
3281
3282 ovsrec_interface_init(&br->synth_local_iface);
3283 ovsrec_port_init(&br->synth_local_port);
3284
3285 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3286 br->synth_local_port.n_interfaces = 1;
3287 br->synth_local_port.name = br->name;
3288
3289 br->synth_local_iface.name = br->name;
3290 br->synth_local_iface.type = "internal";
3291
3292 br->synth_local_ifacep = &br->synth_local_iface;
3293
3294 shash_add(wanted_ports, br->name, &br->synth_local_port);
3295 }
3296
3297 if (splinter_vlans) {
3298 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3299 }
3300 }
3301
3302 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3303 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3304 * 'br' needs to complete its configuration. */
3305 static void
3306 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3307 {
3308 struct shash_node *port_node;
3309 struct port *port, *next;
3310
3311 /* Get rid of deleted ports.
3312 * Get rid of deleted interfaces on ports that still exist. */
3313 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3314 port->cfg = shash_find_data(wanted_ports, port->name);
3315 if (!port->cfg) {
3316 port_destroy(port);
3317 } else {
3318 port_del_ifaces(port);
3319 }
3320 }
3321
3322 /* Update iface->cfg and iface->type in interfaces that still exist. */
3323 SHASH_FOR_EACH (port_node, wanted_ports) {
3324 const struct ovsrec_port *port = port_node->data;
3325 size_t i;
3326
3327 for (i = 0; i < port->n_interfaces; i++) {
3328 const struct ovsrec_interface *cfg = port->interfaces[i];
3329 struct iface *iface = iface_lookup(br, cfg->name);
3330 const char *type = iface_get_type(cfg, br->cfg);
3331
3332 if (iface) {
3333 iface->cfg = cfg;
3334 iface->type = type;
3335 } else if (!strcmp(type, "null")) {
3336 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3337 " may be removed in February 2013. Please email"
3338 " dev@openvswitch.org with concerns.",
3339 cfg->name);
3340 } else {
3341 /* We will add new interfaces later. */
3342 }
3343 }
3344 }
3345 }
3346
3347 /* Initializes 'oc' appropriately as a management service controller for
3348 * 'br'.
3349 *
3350 * The caller must free oc->target when it is no longer needed. */
3351 static void
3352 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3353 struct ofproto_controller *oc)
3354 {
3355 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3356 oc->max_backoff = 0;
3357 oc->probe_interval = 60;
3358 oc->band = OFPROTO_OUT_OF_BAND;
3359 oc->rate_limit = 0;
3360 oc->burst_limit = 0;
3361 oc->enable_async_msgs = true;
3362 oc->dscp = 0;
3363 }
3364
3365 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3366 static void
3367 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3368 struct ofproto_controller *oc)
3369 {
3370 int dscp;
3371
3372 oc->target = c->target;
3373 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3374 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3375 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3376 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3377 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3378 oc->burst_limit = (c->controller_burst_limit
3379 ? *c->controller_burst_limit : 0);
3380 oc->enable_async_msgs = (!c->enable_async_messages
3381 || *c->enable_async_messages);
3382 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3383 if (dscp < 0 || dscp > 63) {
3384 dscp = DSCP_DEFAULT;
3385 }
3386 oc->dscp = dscp;
3387 }
3388
3389 /* Configures the IP stack for 'br''s local interface properly according to the
3390 * configuration in 'c'. */
3391 static void
3392 bridge_configure_local_iface_netdev(struct bridge *br,
3393 struct ovsrec_controller *c)
3394 {
3395 struct netdev *netdev;
3396 struct in_addr mask, gateway;
3397
3398 struct iface *local_iface;
3399 struct in_addr ip;
3400
3401 /* If there's no local interface or no IP address, give up. */
3402 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3403 if (!local_iface || !c->local_ip
3404 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3405 return;
3406 }
3407
3408 /* Bring up the local interface. */
3409 netdev = local_iface->netdev;
3410 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3411
3412 /* Configure the IP address and netmask. */
3413 if (!c->local_netmask
3414 || !inet_pton(AF_INET, c->local_netmask, &mask)
3415 || !mask.s_addr) {
3416 mask.s_addr = guess_netmask(ip.s_addr);
3417 }
3418 if (!netdev_set_in4(netdev, ip, mask)) {
3419 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3420 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3421 }
3422
3423 /* Configure the default gateway. */
3424 if (c->local_gateway
3425 && inet_pton(AF_INET, c->local_gateway, &gateway)
3426 && gateway.s_addr) {
3427 if (!netdev_add_router(netdev, gateway)) {
3428 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3429 br->name, IP_ARGS(gateway.s_addr));
3430 }
3431 }
3432 }
3433
3434 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3435 * in either string are treated as equal to any number of slashes in the other,
3436 * e.g. "x///y" is equal to "x/y".
3437 *
3438 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3439 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3440 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3441 * 'b' against a prefix of 'a'.
3442 */
3443 static bool
3444 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3445 {
3446 const char *b_start = b;
3447 for (;;) {
3448 if (b - b_start >= b_stoplen) {
3449 return true;
3450 } else if (*a != *b) {
3451 return false;
3452 } else if (*a == '/') {
3453 a += strspn(a, "/");
3454 b += strspn(b, "/");
3455 } else if (*a == '\0') {
3456 return true;
3457 } else {
3458 a++;
3459 b++;
3460 }
3461 }
3462 }
3463
3464 static void
3465 bridge_configure_remotes(struct bridge *br,
3466 const struct sockaddr_in *managers, size_t n_managers)
3467 {
3468 bool disable_in_band;
3469
3470 struct ovsrec_controller **controllers;
3471 size_t n_controllers;
3472
3473 enum ofproto_fail_mode fail_mode;
3474
3475 struct ofproto_controller *ocs;
3476 size_t n_ocs;
3477 size_t i;
3478
3479 /* Check if we should disable in-band control on this bridge. */
3480 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3481 false);
3482
3483 /* Set OpenFlow queue ID for in-band control. */
3484 ofproto_set_in_band_queue(br->ofproto,
3485 smap_get_int(&br->cfg->other_config,
3486 "in-band-queue", -1));
3487
3488 if (disable_in_band) {
3489 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3490 } else {
3491 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3492 }
3493
3494 n_controllers = bridge_get_controllers(br, &controllers);
3495
3496 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3497 n_ocs = 0;
3498
3499 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3500 for (i = 0; i < n_controllers; i++) {
3501 struct ovsrec_controller *c = controllers[i];
3502
3503 if (!strncmp(c->target, "punix:", 6)
3504 || !strncmp(c->target, "unix:", 5)) {
3505 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3506 char *whitelist;
3507
3508 if (!strncmp(c->target, "unix:", 5)) {
3509 /* Connect to a listening socket */
3510 whitelist = xasprintf("unix:%s/", ovs_rundir());
3511 if (strchr(c->target, '/') &&
3512 !equal_pathnames(c->target, whitelist,
3513 strlen(whitelist))) {
3514 /* Absolute path specified, but not in ovs_rundir */
3515 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3516 "controller \"%s\" due to possibility for "
3517 "remote exploit. Instead, specify socket "
3518 "in whitelisted \"%s\" or connect to "
3519 "\"unix:%s/%s.mgmt\" (which is always "
3520 "available without special configuration).",
3521 br->name, c->target, whitelist,
3522 ovs_rundir(), br->name);
3523 free(whitelist);
3524 continue;
3525 }
3526 } else {
3527 whitelist = xasprintf("punix:%s/%s.controller",
3528 ovs_rundir(), br->name);
3529 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3530 /* Prevent remote ovsdb-server users from accessing
3531 * arbitrary Unix domain sockets and overwriting arbitrary
3532 * local files. */
3533 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3534 "controller \"%s\" due to possibility of "
3535 "overwriting local files. Instead, specify "
3536 "whitelisted \"%s\" or connect to "
3537 "\"unix:%s/%s.mgmt\" (which is always "
3538 "available without special configuration).",
3539 br->name, c->target, whitelist,
3540 ovs_rundir(), br->name);
3541 free(whitelist);
3542 continue;
3543 }
3544 }
3545
3546 free(whitelist);
3547 }
3548
3549 bridge_configure_local_iface_netdev(br, c);
3550 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3551 if (disable_in_band) {
3552 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3553 }
3554 n_ocs++;
3555 }
3556
3557 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3558 bridge_get_allowed_versions(br));
3559 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3560 free(ocs);
3561
3562 /* Set the fail-mode. */
3563 fail_mode = !br->cfg->fail_mode
3564 || !strcmp(br->cfg->fail_mode, "standalone")
3565 ? OFPROTO_FAIL_STANDALONE
3566 : OFPROTO_FAIL_SECURE;
3567 ofproto_set_fail_mode(br->ofproto, fail_mode);
3568
3569 /* Configure OpenFlow controller connection snooping. */
3570 if (!ofproto_has_snoops(br->ofproto)) {
3571 struct sset snoops;
3572
3573 sset_init(&snoops);
3574 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3575 ovs_rundir(), br->name));
3576 ofproto_set_snoops(br->ofproto, &snoops);
3577 sset_destroy(&snoops);
3578 }
3579 }
3580
3581 static void
3582 bridge_configure_tables(struct bridge *br)
3583 {
3584 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3585 int n_tables;
3586 int i, j, k;
3587
3588 n_tables = ofproto_get_n_tables(br->ofproto);
3589 j = 0;
3590 for (i = 0; i < n_tables; i++) {
3591 struct ofproto_table_settings s;
3592 bool use_default_prefixes = true;
3593
3594 s.name = NULL;
3595 s.max_flows = UINT_MAX;
3596 s.groups = NULL;
3597 s.n_groups = 0;
3598 s.n_prefix_fields = 0;
3599 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3600
3601 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3602 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3603
3604 s.name = cfg->name;
3605 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3606 s.max_flows = *cfg->flow_limit;
3607 }
3608 if (cfg->overflow_policy
3609 && !strcmp(cfg->overflow_policy, "evict")) {
3610
3611 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3612 for (k = 0; k < cfg->n_groups; k++) {
3613 const char *string = cfg->groups[k];
3614 char *msg;
3615
3616 msg = mf_parse_subfield__(&s.groups[k], &string);
3617 if (msg) {
3618 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3619 "'groups' (%s)", br->name, i, msg);
3620 free(msg);
3621 } else if (*string) {
3622 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3623 "element '%s' contains trailing garbage",
3624 br->name, i, cfg->groups[k]);
3625 } else {
3626 s.n_groups++;
3627 }
3628 }
3629 }
3630 /* Prefix lookup fields. */
3631 s.n_prefix_fields = 0;
3632 for (k = 0; k < cfg->n_prefixes; k++) {
3633 const char *name = cfg->prefixes[k];
3634 const struct mf_field *mf;
3635
3636 if (strcmp(name, "none") == 0) {
3637 use_default_prefixes = false;
3638 s.n_prefix_fields = 0;
3639 break;
3640 }
3641 mf = mf_from_name(name);
3642 if (!mf) {
3643 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3644 br->name, name);
3645 continue;
3646 }
3647 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3648 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3649 "%s", br->name, name);
3650 continue;
3651 }
3652 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3653 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3654 "field not used: %s", br->name, name);
3655 continue;
3656 }
3657 use_default_prefixes = false;
3658 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3659 }
3660 }
3661 if (use_default_prefixes) {
3662 /* Use default values. */
3663 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3664 memcpy(s.prefix_fields, default_prefix_fields,
3665 sizeof default_prefix_fields);
3666 } else {
3667 int k;
3668 struct ds ds = DS_EMPTY_INITIALIZER;
3669 for (k = 0; k < s.n_prefix_fields; k++) {
3670 if (k) {
3671 ds_put_char(&ds, ',');
3672 }
3673 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3674 }
3675 if (s.n_prefix_fields == 0) {
3676 ds_put_cstr(&ds, "none");
3677 }
3678 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3679 br->name, i, ds_cstr(&ds));
3680 ds_destroy(&ds);
3681 }
3682
3683 ofproto_configure_table(br->ofproto, i, &s);
3684
3685 free(s.groups);
3686 }
3687 for (; j < br->cfg->n_flow_tables; j++) {
3688 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3689 "%"PRId64" not supported by this datapath", br->name,
3690 br->cfg->key_flow_tables[j]);
3691 }
3692 }
3693
3694 static void
3695 bridge_configure_dp_desc(struct bridge *br)
3696 {
3697 ofproto_set_dp_desc(br->ofproto,
3698 smap_get(&br->cfg->other_config, "dp-desc"));
3699 }
3700
3701 static struct aa_mapping *
3702 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3703 {
3704 struct aa_mapping *m;
3705
3706 HMAP_FOR_EACH_IN_BUCKET (m,
3707 hmap_node,
3708 hash_bytes(&isid, sizeof isid, 0),
3709 &br->mappings) {
3710 if (isid == m->isid) {
3711 return m;
3712 }
3713 }
3714 return NULL;
3715 }
3716
3717 static struct aa_mapping *
3718 bridge_aa_mapping_create(struct bridge *br,
3719 const int64_t isid,
3720 const int64_t vlan)
3721 {
3722 struct aa_mapping *m;
3723
3724 m = xzalloc(sizeof *m);
3725 m->bridge = br;
3726 m->isid = isid;
3727 m->vlan = vlan;
3728 m->br_name = xstrdup(br->name);
3729 hmap_insert(&br->mappings,
3730 &m->hmap_node,
3731 hash_bytes(&isid, sizeof isid, 0));
3732
3733 return m;
3734 }
3735
3736 static void
3737 bridge_aa_mapping_destroy(struct aa_mapping *m)
3738 {
3739 if (m) {
3740 struct bridge *br = m->bridge;
3741
3742 if (br->ofproto) {
3743 ofproto_aa_mapping_unregister(br->ofproto, m);
3744 }
3745
3746 hmap_remove(&br->mappings, &m->hmap_node);
3747 if (m->br_name) {
3748 free(m->br_name);
3749 }
3750 free(m);
3751 }
3752 }
3753
3754 static bool
3755 bridge_aa_mapping_configure(struct aa_mapping *m)
3756 {
3757 struct aa_mapping_settings s;
3758
3759 s.isid = m->isid;
3760 s.vlan = m->vlan;
3761
3762 /* Configure. */
3763 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3764
3765 return true;
3766 }
3767
3768 static void
3769 bridge_configure_aa(struct bridge *br)
3770 {
3771 const struct ovsdb_datum *mc;
3772 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3773 struct aa_settings aa_s;
3774 struct aa_mapping *m, *next;
3775 size_t i;
3776
3777 if (!auto_attach) {
3778 ofproto_set_aa(br->ofproto, NULL, NULL);
3779 return;
3780 }
3781
3782 memset(&aa_s, 0, sizeof aa_s);
3783 aa_s.system_description = auto_attach->system_description;
3784 aa_s.system_name = auto_attach->system_name;
3785 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3786
3787 mc = ovsrec_autoattach_get_mappings(auto_attach,
3788 OVSDB_TYPE_INTEGER,
3789 OVSDB_TYPE_INTEGER);
3790 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3791 union ovsdb_atom atom;
3792
3793 atom.integer = m->isid;
3794 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3795 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3796 m->isid, m->vlan);
3797 bridge_aa_mapping_destroy(m);
3798 }
3799 }
3800
3801 /* Add new mappings and reconfigure existing ones. */
3802 for (i = 0; i < auto_attach->n_mappings; ++i) {
3803 struct aa_mapping *m =
3804 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3805
3806 if (!m) {
3807 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3808 auto_attach->key_mappings[i],
3809 auto_attach->value_mappings[i]);
3810 m = bridge_aa_mapping_create(br,
3811 auto_attach->key_mappings[i],
3812 auto_attach->value_mappings[i]);
3813
3814 if (!bridge_aa_mapping_configure(m)) {
3815 bridge_aa_mapping_destroy(m);
3816 }
3817 }
3818 }
3819 }
3820
3821 static bool
3822 bridge_aa_need_refresh(struct bridge *br)
3823 {
3824 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3825 }
3826
3827 static void
3828 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3829 {
3830 int64_t *trunks = NULL;
3831 unsigned int i = 0;
3832 bool found = false, reconfigure = false;
3833
3834 for (i = 0; i < port->cfg->n_trunks; i++) {
3835 if (port->cfg->trunks[i] == m->vlan) {
3836 found = true;
3837 break;
3838 }
3839 }
3840
3841 switch (m->oper) {
3842 case BRIDGE_AA_VLAN_OPER_ADD:
3843 if (!found) {
3844 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3845
3846 for (i = 0; i < port->cfg->n_trunks; i++) {
3847 trunks[i] = port->cfg->trunks[i];
3848 }
3849 trunks[i++] = m->vlan;
3850 reconfigure = true;
3851 }
3852
3853 break;
3854
3855 case BRIDGE_AA_VLAN_OPER_REMOVE:
3856 if (found) {
3857 unsigned int j = 0;
3858
3859 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3860
3861 for (i = 0; i < port->cfg->n_trunks; i++) {
3862 if (port->cfg->trunks[i] != m->vlan) {
3863 trunks[j++] = port->cfg->trunks[i];
3864 }
3865 }
3866 i = j;
3867 reconfigure = true;
3868 }
3869
3870 break;
3871
3872 case BRIDGE_AA_VLAN_OPER_UNDEF:
3873 default:
3874 VLOG_WARN("unrecognized operation %u", m->oper);
3875 break;
3876 }
3877
3878 if (reconfigure) {
3879 /* VLAN switching under trunk mode cause the trunk port to switch all
3880 * VLANs, see ovs-vswitchd.conf.db
3881 */
3882 if (i == 0) {
3883 static char *vlan_mode_access = "access";
3884 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3885 }
3886
3887 if (i == 1) {
3888 static char *vlan_mode_trunk = "trunk";
3889 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3890 }
3891
3892 ovsrec_port_set_trunks(port->cfg, trunks, i);
3893
3894 /* Force reconfigure of the port. */
3895 port_configure(port);
3896 }
3897 }
3898
3899 static void
3900 bridge_aa_refresh_queued(struct bridge *br)
3901 {
3902 struct ovs_list *list = xmalloc(sizeof *list);
3903 struct bridge_aa_vlan *node;
3904
3905 list_init(list);
3906 ofproto_aa_vlan_get_queued(br->ofproto, list);
3907
3908 LIST_FOR_EACH(node, list_node, list) {
3909 struct port *port;
3910
3911 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3912 node->oper);
3913
3914 port = port_lookup(br, node->port_name);
3915 if (port) {
3916 bridge_aa_update_trunks(port, node);
3917 }
3918
3919 list_remove(&node->list_node);
3920 free(node->port_name);
3921 free(node);
3922 }
3923
3924 free(list);
3925 }
3926
3927 \f
3928 /* Port functions. */
3929
3930 static struct port *
3931 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3932 {
3933 struct port *port;
3934
3935 port = xzalloc(sizeof *port);
3936 port->bridge = br;
3937 port->name = xstrdup(cfg->name);
3938 port->cfg = cfg;
3939 list_init(&port->ifaces);
3940
3941 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3942 return port;
3943 }
3944
3945 /* Deletes interfaces from 'port' that are no longer configured for it. */
3946 static void
3947 port_del_ifaces(struct port *port)
3948 {
3949 struct iface *iface, *next;
3950 struct sset new_ifaces;
3951 size_t i;
3952
3953 /* Collect list of new interfaces. */
3954 sset_init(&new_ifaces);
3955 for (i = 0; i < port->cfg->n_interfaces; i++) {
3956 const char *name = port->cfg->interfaces[i]->name;
3957 const char *type = port->cfg->interfaces[i]->type;
3958 if (strcmp(type, "null")) {
3959 sset_add(&new_ifaces, name);
3960 }
3961 }
3962
3963 /* Get rid of deleted interfaces. */
3964 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3965 if (!sset_contains(&new_ifaces, iface->name)) {
3966 iface_destroy(iface);
3967 }
3968 }
3969
3970 sset_destroy(&new_ifaces);
3971 }
3972
3973 static void
3974 port_destroy(struct port *port)
3975 {
3976 if (port) {
3977 struct bridge *br = port->bridge;
3978 struct iface *iface, *next;
3979
3980 if (br->ofproto) {
3981 ofproto_bundle_unregister(br->ofproto, port);
3982 }
3983
3984 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3985 iface_destroy__(iface);
3986 }
3987
3988 hmap_remove(&br->ports, &port->hmap_node);
3989 free(port->name);
3990 free(port);
3991 }
3992 }
3993
3994 static struct port *
3995 port_lookup(const struct bridge *br, const char *name)
3996 {
3997 struct port *port;
3998
3999 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4000 &br->ports) {
4001 if (!strcmp(port->name, name)) {
4002 return port;
4003 }
4004 }
4005 return NULL;
4006 }
4007
4008 static bool
4009 enable_lacp(struct port *port, bool *activep)
4010 {
4011 if (!port->cfg->lacp) {
4012 /* XXX when LACP implementation has been sufficiently tested, enable by
4013 * default and make active on bonded ports. */
4014 return false;
4015 } else if (!strcmp(port->cfg->lacp, "off")) {
4016 return false;
4017 } else if (!strcmp(port->cfg->lacp, "active")) {
4018 *activep = true;
4019 return true;
4020 } else if (!strcmp(port->cfg->lacp, "passive")) {
4021 *activep = false;
4022 return true;
4023 } else {
4024 VLOG_WARN("port %s: unknown LACP mode %s",
4025 port->name, port->cfg->lacp);
4026 return false;
4027 }
4028 }
4029
4030 static struct lacp_settings *
4031 port_configure_lacp(struct port *port, struct lacp_settings *s)
4032 {
4033 const char *lacp_time, *system_id;
4034 int priority;
4035
4036 if (!enable_lacp(port, &s->active)) {
4037 return NULL;
4038 }
4039
4040 s->name = port->name;
4041
4042 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4043 if (system_id) {
4044 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4045 ETH_ADDR_SCAN_ARGS(s->id))) {
4046 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4047 " address.", port->name, system_id);
4048 return NULL;
4049 }
4050 } else {
4051 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4052 }
4053
4054 if (eth_addr_is_zero(s->id)) {
4055 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4056 return NULL;
4057 }
4058
4059 /* Prefer bondable links if unspecified. */
4060 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4061 0);
4062 s->priority = (priority > 0 && priority <= UINT16_MAX
4063 ? priority
4064 : UINT16_MAX - !list_is_short(&port->ifaces));
4065
4066 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4067 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4068
4069 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4070 "lacp-fallback-ab", false);
4071
4072 return s;
4073 }
4074
4075 static void
4076 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4077 {
4078 int priority, portid, key;
4079
4080 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4081 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4082 0);
4083 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4084
4085 if (portid <= 0 || portid > UINT16_MAX) {
4086 portid = ofp_to_u16(iface->ofp_port);
4087 }
4088
4089 if (priority <= 0 || priority > UINT16_MAX) {
4090 priority = UINT16_MAX;
4091 }
4092
4093 if (key < 0 || key > UINT16_MAX) {
4094 key = 0;
4095 }
4096
4097 s->name = iface->name;
4098 s->id = portid;
4099 s->priority = priority;
4100 s->key = key;
4101 }
4102
4103 static void
4104 port_configure_bond(struct port *port, struct bond_settings *s)
4105 {
4106 const char *detect_s;
4107 struct iface *iface;
4108 const char *mac_s;
4109 int miimon_interval;
4110
4111 s->name = port->name;
4112 s->balance = BM_AB;
4113 if (port->cfg->bond_mode) {
4114 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4115 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4116 port->name, port->cfg->bond_mode,
4117 bond_mode_to_string(s->balance));
4118 }
4119 } else {
4120 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4121
4122 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4123 * active-backup. At some point we should remove this warning. */
4124 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4125 " in previous versions, the default bond_mode was"
4126 " balance-slb", port->name,
4127 bond_mode_to_string(s->balance));
4128 }
4129 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4130 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4131 "please use another bond type or disable flood_vlans",
4132 port->name);
4133 }
4134
4135 miimon_interval = smap_get_int(&port->cfg->other_config,
4136 "bond-miimon-interval", 0);
4137 if (miimon_interval <= 0) {
4138 miimon_interval = 200;
4139 }
4140
4141 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4142 if (!detect_s || !strcmp(detect_s, "carrier")) {
4143 miimon_interval = 0;
4144 } else if (strcmp(detect_s, "miimon")) {
4145 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4146 "defaulting to carrier", port->name, detect_s);
4147 miimon_interval = 0;
4148 }
4149
4150 s->up_delay = MAX(0, port->cfg->bond_updelay);
4151 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4152 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4153 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4154 "bond-rebalance-interval", 10000);
4155 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4156 s->rebalance_interval = 1000;
4157 }
4158
4159 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4160 "lacp-fallback-ab", false);
4161
4162 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4163 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4164 }
4165
4166 mac_s = port->cfg->bond_active_slave;
4167 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4168 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4169 /* OVSDB did not store the last active interface */
4170 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4171 }
4172 }
4173
4174 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4175 * instead of obtaining it from the database. */
4176 static bool
4177 port_is_synthetic(const struct port *port)
4178 {
4179 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4180 }
4181 \f
4182 /* Interface functions. */
4183
4184 static bool
4185 iface_is_internal(const struct ovsrec_interface *iface,
4186 const struct ovsrec_bridge *br)
4187 {
4188 /* The local port and "internal" ports are always "internal". */
4189 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4190 }
4191
4192 /* Returns the correct network device type for interface 'iface' in bridge
4193 * 'br'. */
4194 static const char *
4195 iface_get_type(const struct ovsrec_interface *iface,
4196 const struct ovsrec_bridge *br)
4197 {
4198 const char *type;
4199
4200 /* The local port always has type "internal". Other ports take
4201 * their type from the database and default to "system" if none is
4202 * specified. */
4203 if (iface_is_internal(iface, br)) {
4204 type = "internal";
4205 } else {
4206 type = iface->type[0] ? iface->type : "system";
4207 }
4208
4209 return ofproto_port_open_type(br->datapath_type, type);
4210 }
4211
4212 static void
4213 iface_destroy__(struct iface *iface)
4214 {
4215 if (iface) {
4216 struct port *port = iface->port;
4217 struct bridge *br = port->bridge;
4218
4219 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4220 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4221 }
4222
4223 if (iface->ofp_port != OFPP_NONE) {
4224 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4225 }
4226
4227 list_remove(&iface->port_elem);
4228 hmap_remove(&br->iface_by_name, &iface->name_node);
4229
4230 /* The user is changing configuration here, so netdev_remove needs to be
4231 * used as opposed to netdev_close */
4232 netdev_remove(iface->netdev);
4233
4234 free(iface->name);
4235 free(iface);
4236 }
4237 }
4238
4239 static void
4240 iface_destroy(struct iface *iface)
4241 {
4242 if (iface) {
4243 struct port *port = iface->port;
4244
4245 iface_destroy__(iface);
4246 if (list_is_empty(&port->ifaces)) {
4247 port_destroy(port);
4248 }
4249 }
4250 }
4251
4252 static struct iface *
4253 iface_lookup(const struct bridge *br, const char *name)
4254 {
4255 struct iface *iface;
4256
4257 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4258 &br->iface_by_name) {
4259 if (!strcmp(iface->name, name)) {
4260 return iface;
4261 }
4262 }
4263
4264 return NULL;
4265 }
4266
4267 static struct iface *
4268 iface_find(const char *name)
4269 {
4270 const struct bridge *br;
4271
4272 HMAP_FOR_EACH (br, node, &all_bridges) {
4273 struct iface *iface = iface_lookup(br, name);
4274
4275 if (iface) {
4276 return iface;
4277 }
4278 }
4279 return NULL;
4280 }
4281
4282 static struct iface *
4283 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4284 {
4285 struct iface *iface;
4286
4287 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4288 &br->ifaces) {
4289 if (iface->ofp_port == ofp_port) {
4290 return iface;
4291 }
4292 }
4293 return NULL;
4294 }
4295
4296 /* Set Ethernet address of 'iface', if one is specified in the configuration
4297 * file. */
4298 static void
4299 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4300 {
4301 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4302 struct iface *hw_addr_iface;
4303
4304 if (strcmp(iface->type, "internal")) {
4305 return;
4306 }
4307
4308 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4309 mac = ea;
4310 } else if (port->cfg->fake_bridge) {
4311 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4312 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4313 mac = ea;
4314 }
4315
4316 if (mac) {
4317 if (iface->ofp_port == OFPP_LOCAL) {
4318 VLOG_ERR("interface %s: ignoring mac in Interface record "
4319 "(use Bridge record to set local port's mac)",
4320 iface->name);
4321 } else if (eth_addr_is_multicast(mac)) {
4322 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4323 iface->name);
4324 } else {
4325 int error = netdev_set_etheraddr(iface->netdev, mac);
4326 if (error) {
4327 VLOG_ERR("interface %s: setting MAC failed (%s)",
4328 iface->name, ovs_strerror(error));
4329 }
4330 }
4331 }
4332 }
4333
4334 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4335 static void
4336 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4337 {
4338 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4339 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4340 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4341 }
4342 }
4343
4344 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4345 * sets the "ofport" field to -1.
4346 *
4347 * This is appropriate when 'if_cfg''s interface cannot be created or is
4348 * otherwise invalid. */
4349 static void
4350 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4351 {
4352 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4353 iface_set_ofport(if_cfg, OFPP_NONE);
4354 ovsrec_interface_set_error(if_cfg, errp);
4355 ovsrec_interface_set_status(if_cfg, NULL);
4356 ovsrec_interface_set_admin_state(if_cfg, NULL);
4357 ovsrec_interface_set_duplex(if_cfg, NULL);
4358 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4359 ovsrec_interface_set_link_state(if_cfg, NULL);
4360 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4361 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4362 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4363 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4364 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4365 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4366 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4367 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4368 }
4369 }
4370
4371 static bool
4372 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4373 {
4374 union ovsdb_atom atom;
4375
4376 atom.integer = target;
4377 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4378 }
4379
4380 static void
4381 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4382 {
4383 struct ofpbuf queues_buf;
4384
4385 ofpbuf_init(&queues_buf, 0);
4386
4387 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
4388 netdev_set_qos(iface->netdev, NULL, NULL);
4389 } else {
4390 const struct ovsdb_datum *queues;
4391 struct netdev_queue_dump dump;
4392 unsigned int queue_id;
4393 struct smap details;
4394 bool queue_zero;
4395 size_t i;
4396
4397 /* Configure top-level Qos for 'iface'. */
4398 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4399
4400 /* Deconfigure queues that were deleted. */
4401 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4402 OVSDB_TYPE_UUID);
4403 smap_init(&details);
4404 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4405 if (!queue_ids_include(queues, queue_id)) {
4406 netdev_delete_queue(iface->netdev, queue_id);
4407 }
4408 }
4409 smap_destroy(&details);
4410
4411 /* Configure queues for 'iface'. */
4412 queue_zero = false;
4413 for (i = 0; i < qos->n_queues; i++) {
4414 const struct ovsrec_queue *queue = qos->value_queues[i];
4415 unsigned int queue_id = qos->key_queues[i];
4416
4417 if (queue_id == 0) {
4418 queue_zero = true;
4419 }
4420
4421 if (queue->n_dscp == 1) {
4422 struct ofproto_port_queue *port_queue;
4423
4424 port_queue = ofpbuf_put_uninit(&queues_buf,
4425 sizeof *port_queue);
4426 port_queue->queue = queue_id;
4427 port_queue->dscp = queue->dscp[0];
4428 }
4429
4430 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4431 }
4432 if (!queue_zero) {
4433 struct smap details;
4434
4435 smap_init(&details);
4436 netdev_set_queue(iface->netdev, 0, &details);
4437 smap_destroy(&details);
4438 }
4439 }
4440
4441 if (iface->ofp_port != OFPP_NONE) {
4442 const struct ofproto_port_queue *port_queues = queues_buf.data;
4443 size_t n_queues = queues_buf.size / sizeof *port_queues;
4444
4445 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4446 port_queues, n_queues);
4447 }
4448
4449 netdev_set_policing(iface->netdev,
4450 iface->cfg->ingress_policing_rate,
4451 iface->cfg->ingress_policing_burst);
4452
4453 ofpbuf_uninit(&queues_buf);
4454 }
4455
4456 static void
4457 iface_configure_cfm(struct iface *iface)
4458 {
4459 const struct ovsrec_interface *cfg = iface->cfg;
4460 const char *opstate_str;
4461 const char *cfm_ccm_vlan;
4462 struct cfm_settings s;
4463 struct smap netdev_args;
4464
4465 if (!cfg->n_cfm_mpid) {
4466 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4467 return;
4468 }
4469
4470 s.check_tnl_key = false;
4471 smap_init(&netdev_args);
4472 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4473 const char *key = smap_get(&netdev_args, "key");
4474 const char *in_key = smap_get(&netdev_args, "in_key");
4475
4476 s.check_tnl_key = (key && !strcmp(key, "flow"))
4477 || (in_key && !strcmp(in_key, "flow"));
4478 }
4479 smap_destroy(&netdev_args);
4480
4481 s.mpid = *cfg->cfm_mpid;
4482 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4483 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4484 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4485
4486 if (s.interval <= 0) {
4487 s.interval = 1000;
4488 }
4489
4490 if (!cfm_ccm_vlan) {
4491 s.ccm_vlan = 0;
4492 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4493 s.ccm_vlan = CFM_RANDOM_VLAN;
4494 } else {
4495 s.ccm_vlan = atoi(cfm_ccm_vlan);
4496 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4497 s.ccm_vlan = 0;
4498 }
4499 }
4500
4501 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4502 false);
4503 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4504
4505 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4506 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4507
4508 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4509 }
4510
4511 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4512 * instead of obtaining it from the database. */
4513 static bool
4514 iface_is_synthetic(const struct iface *iface)
4515 {
4516 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4517 }
4518
4519 static ofp_port_t
4520 iface_validate_ofport__(size_t n, int64_t *ofport)
4521 {
4522 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4523 ? u16_to_ofp(*ofport)
4524 : OFPP_NONE);
4525 }
4526
4527 static ofp_port_t
4528 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4529 {
4530 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4531 }
4532
4533 static ofp_port_t
4534 iface_pick_ofport(const struct ovsrec_interface *cfg)
4535 {
4536 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4537 return (requested_ofport != OFPP_NONE
4538 ? requested_ofport
4539 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4540 }
4541 \f
4542 /* Port mirroring. */
4543
4544 static struct mirror *
4545 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4546 {
4547 struct mirror *m;
4548
4549 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4550 if (uuid_equals(uuid, &m->uuid)) {
4551 return m;
4552 }
4553 }
4554 return NULL;
4555 }
4556
4557 static void
4558 bridge_configure_mirrors(struct bridge *br)
4559 {
4560 const struct ovsdb_datum *mc;
4561 unsigned long *flood_vlans;
4562 struct mirror *m, *next;
4563 size_t i;
4564
4565 /* Get rid of deleted mirrors. */
4566 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4567 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4568 union ovsdb_atom atom;
4569
4570 atom.uuid = m->uuid;
4571 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4572 mirror_destroy(m);
4573 }
4574 }
4575
4576 /* Add new mirrors and reconfigure existing ones. */
4577 for (i = 0; i < br->cfg->n_mirrors; i++) {
4578 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4579 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4580 if (!m) {
4581 m = mirror_create(br, cfg);
4582 }
4583 m->cfg = cfg;
4584 if (!mirror_configure(m)) {
4585 mirror_destroy(m);
4586 }
4587 }
4588
4589 /* Update flooded vlans (for RSPAN). */
4590 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4591 br->cfg->n_flood_vlans);
4592 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4593 bitmap_free(flood_vlans);
4594 }
4595
4596 static struct mirror *
4597 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4598 {
4599 struct mirror *m;
4600
4601 m = xzalloc(sizeof *m);
4602 m->uuid = cfg->header_.uuid;
4603 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4604 m->bridge = br;
4605 m->name = xstrdup(cfg->name);
4606
4607 return m;
4608 }
4609
4610 static void
4611 mirror_destroy(struct mirror *m)
4612 {
4613 if (m) {
4614 struct bridge *br = m->bridge;
4615
4616 if (br->ofproto) {
4617 ofproto_mirror_unregister(br->ofproto, m);
4618 }
4619
4620 hmap_remove(&br->mirrors, &m->hmap_node);
4621 free(m->name);
4622 free(m);
4623 }
4624 }
4625
4626 static void
4627 mirror_collect_ports(struct mirror *m,
4628 struct ovsrec_port **in_ports, int n_in_ports,
4629 void ***out_portsp, size_t *n_out_portsp)
4630 {
4631 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4632 size_t n_out_ports = 0;
4633 size_t i;
4634
4635 for (i = 0; i < n_in_ports; i++) {
4636 const char *name = in_ports[i]->name;
4637 struct port *port = port_lookup(m->bridge, name);
4638 if (port) {
4639 out_ports[n_out_ports++] = port;
4640 } else {
4641 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4642 "port %s", m->bridge->name, m->name, name);
4643 }
4644 }
4645 *out_portsp = out_ports;
4646 *n_out_portsp = n_out_ports;
4647 }
4648
4649 static bool
4650 mirror_configure(struct mirror *m)
4651 {
4652 const struct ovsrec_mirror *cfg = m->cfg;
4653 struct ofproto_mirror_settings s;
4654
4655 /* Set name. */
4656 if (strcmp(cfg->name, m->name)) {
4657 free(m->name);
4658 m->name = xstrdup(cfg->name);
4659 }
4660 s.name = m->name;
4661
4662 /* Get output port or VLAN. */
4663 if (cfg->output_port) {
4664 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4665 if (!s.out_bundle) {
4666 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4667 m->bridge->name, m->name);
4668 return false;
4669 }
4670 s.out_vlan = UINT16_MAX;
4671
4672 if (cfg->output_vlan) {
4673 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4674 "output vlan; ignoring output vlan",
4675 m->bridge->name, m->name);
4676 }
4677 } else if (cfg->output_vlan) {
4678 /* The database should prevent invalid VLAN values. */
4679 s.out_bundle = NULL;
4680 s.out_vlan = *cfg->output_vlan;
4681 } else {
4682 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4683 m->bridge->name, m->name);
4684 return false;
4685 }
4686
4687 /* Get port selection. */
4688 if (cfg->select_all) {
4689 size_t n_ports = hmap_count(&m->bridge->ports);
4690 void **ports = xmalloc(n_ports * sizeof *ports);
4691 struct port *port;
4692 size_t i;
4693
4694 i = 0;
4695 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4696 ports[i++] = port;
4697 }
4698
4699 s.srcs = ports;
4700 s.n_srcs = n_ports;
4701
4702 s.dsts = ports;
4703 s.n_dsts = n_ports;
4704 } else {
4705 /* Get ports, dropping ports that don't exist.
4706 * The IDL ensures that there are no duplicates. */
4707 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4708 &s.srcs, &s.n_srcs);
4709 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4710 &s.dsts, &s.n_dsts);
4711 }
4712
4713 /* Get VLAN selection. */
4714 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4715
4716 /* Configure. */
4717 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4718
4719 /* Clean up. */
4720 if (s.srcs != s.dsts) {
4721 free(s.dsts);
4722 }
4723 free(s.srcs);
4724 free(s.src_vlans);
4725
4726 return true;
4727 }
4728 \f
4729 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4730 *
4731 * This is deprecated. It is only for compatibility with broken device drivers
4732 * in old versions of Linux that do not properly support VLANs when VLAN
4733 * devices are not used. When broken device drivers are no longer in
4734 * widespread use, we will delete these interfaces. */
4735
4736 static struct ovsrec_port **recs;
4737 static size_t n_recs, allocated_recs;
4738
4739 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4740 * splinters are reconfigured. */
4741 static void
4742 register_rec(struct ovsrec_port *rec)
4743 {
4744 if (n_recs >= allocated_recs) {
4745 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4746 }
4747 recs[n_recs++] = rec;
4748 }
4749
4750 /* Frees all of the ports registered with register_reg(). */
4751 static void
4752 free_registered_recs(void)
4753 {
4754 size_t i;
4755
4756 for (i = 0; i < n_recs; i++) {
4757 struct ovsrec_port *port = recs[i];
4758 size_t j;
4759
4760 for (j = 0; j < port->n_interfaces; j++) {
4761 struct ovsrec_interface *iface = port->interfaces[j];
4762 free(iface->name);
4763 free(iface);
4764 }
4765
4766 smap_destroy(&port->other_config);
4767 free(port->interfaces);
4768 free(port->name);
4769 free(port->tag);
4770 free(port);
4771 }
4772 n_recs = 0;
4773 }
4774
4775 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4776 * otherwise. */
4777 static bool
4778 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4779 {
4780 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4781 false);
4782 }
4783
4784 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4785 * splinters.
4786 *
4787 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4788 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4789 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4790 * with free().
4791 *
4792 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4793 * use, returns NULL.
4794 *
4795 * Updates 'vlan_splinters_enabled_anywhere'. */
4796 static unsigned long int *
4797 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4798 {
4799 unsigned long int *splinter_vlans;
4800 struct sset splinter_ifaces;
4801 const char *real_dev_name;
4802 struct shash *real_devs;
4803 struct shash_node *node;
4804 struct bridge *br;
4805 size_t i;
4806
4807 /* Free space allocated for synthesized ports and interfaces, since we're
4808 * in the process of reconstructing all of them. */
4809 free_registered_recs();
4810
4811 splinter_vlans = bitmap_allocate(4096);
4812 sset_init(&splinter_ifaces);
4813 vlan_splinters_enabled_anywhere = false;
4814 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4815 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4816 size_t j;
4817
4818 for (j = 0; j < br_cfg->n_ports; j++) {
4819 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4820 int k;
4821
4822 for (k = 0; k < port_cfg->n_interfaces; k++) {
4823 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4824
4825 if (vlan_splinters_is_enabled(iface_cfg)) {
4826 vlan_splinters_enabled_anywhere = true;
4827 sset_add(&splinter_ifaces, iface_cfg->name);
4828 vlan_bitmap_from_array__(port_cfg->trunks,
4829 port_cfg->n_trunks,
4830 splinter_vlans);
4831 }
4832 }
4833
4834 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4835 bitmap_set1(splinter_vlans, *port_cfg->tag);
4836 }
4837 }
4838 }
4839
4840 if (!vlan_splinters_enabled_anywhere) {
4841 free(splinter_vlans);
4842 sset_destroy(&splinter_ifaces);
4843 return NULL;
4844 }
4845
4846 HMAP_FOR_EACH (br, node, &all_bridges) {
4847 if (br->ofproto) {
4848 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4849 }
4850 }
4851
4852 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4853 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4854 * device to be created for it. */
4855 bitmap_set0(splinter_vlans, 0);
4856 bitmap_set0(splinter_vlans, 4095);
4857
4858 /* Delete all VLAN devices that we don't need. */
4859 vlandev_refresh();
4860 real_devs = vlandev_get_real_devs();
4861 SHASH_FOR_EACH (node, real_devs) {
4862 const struct vlan_real_dev *real_dev = node->data;
4863 const struct vlan_dev *vlan_dev;
4864 bool real_dev_has_splinters;
4865
4866 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4867 real_dev->name);
4868 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4869 if (!real_dev_has_splinters
4870 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4871 struct netdev *netdev;
4872
4873 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4874 if (!netdev_get_in4(netdev, NULL, NULL) ||
4875 !netdev_get_in6(netdev, NULL)) {
4876 /* It has an IP address configured, so we don't own
4877 * it. Don't delete it. */
4878 } else {
4879 vlandev_del(vlan_dev->name);
4880 }
4881 netdev_close(netdev);
4882 }
4883 }
4884
4885 }
4886 }
4887
4888 /* Add all VLAN devices that we need. */
4889 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4890 int vid;
4891
4892 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4893 if (!vlandev_get_name(real_dev_name, vid)) {
4894 vlandev_add(real_dev_name, vid);
4895 }
4896 }
4897 }
4898
4899 vlandev_refresh();
4900
4901 sset_destroy(&splinter_ifaces);
4902
4903 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4904 free(splinter_vlans);
4905 return NULL;
4906 }
4907 return splinter_vlans;
4908 }
4909
4910 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4911 * ofproto. */
4912 static void
4913 configure_splinter_port(struct port *port)
4914 {
4915 struct ofproto *ofproto = port->bridge->ofproto;
4916 ofp_port_t realdev_ofp_port;
4917 const char *realdev_name;
4918 struct iface *vlandev, *realdev;
4919
4920 ofproto_bundle_unregister(port->bridge->ofproto, port);
4921
4922 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4923 port_elem);
4924
4925 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4926 realdev = iface_lookup(port->bridge, realdev_name);
4927 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4928
4929 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4930 *port->cfg->tag);
4931 }
4932
4933 static struct ovsrec_port *
4934 synthesize_splinter_port(const char *real_dev_name,
4935 const char *vlan_dev_name, int vid)
4936 {
4937 struct ovsrec_interface *iface;
4938 struct ovsrec_port *port;
4939
4940 iface = xmalloc(sizeof *iface);
4941 ovsrec_interface_init(iface);
4942 iface->name = xstrdup(vlan_dev_name);
4943 iface->type = "system";
4944
4945 port = xmalloc(sizeof *port);
4946 ovsrec_port_init(port);
4947 port->interfaces = xmemdup(&iface, sizeof iface);
4948 port->n_interfaces = 1;
4949 port->name = xstrdup(vlan_dev_name);
4950 port->vlan_mode = "splinter";
4951 port->tag = xmalloc(sizeof *port->tag);
4952 *port->tag = vid;
4953
4954 smap_add(&port->other_config, "realdev", real_dev_name);
4955
4956 register_rec(port);
4957 return port;
4958 }
4959
4960 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4961 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4962 * 1-bit in the 'splinter_vlans' bitmap. */
4963 static void
4964 add_vlan_splinter_ports(struct bridge *br,
4965 const unsigned long int *splinter_vlans,
4966 struct shash *ports)
4967 {
4968 size_t i;
4969
4970 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4971 * we're modifying 'ports'. */
4972 for (i = 0; i < br->cfg->n_ports; i++) {
4973 const char *name = br->cfg->ports[i]->name;
4974 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4975 size_t j;
4976
4977 for (j = 0; j < port_cfg->n_interfaces; j++) {
4978 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4979
4980 if (vlan_splinters_is_enabled(iface_cfg)) {
4981 const char *real_dev_name;
4982 uint16_t vid;
4983
4984 real_dev_name = iface_cfg->name;
4985 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4986 const char *vlan_dev_name;
4987
4988 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4989 if (vlan_dev_name
4990 && !shash_find(ports, vlan_dev_name)) {
4991 shash_add(ports, vlan_dev_name,
4992 synthesize_splinter_port(
4993 real_dev_name, vlan_dev_name, vid));
4994 }
4995 }
4996 }
4997 }
4998 }
4999 }
5000
5001 static void
5002 mirror_refresh_stats(struct mirror *m)
5003 {
5004 struct ofproto *ofproto = m->bridge->ofproto;
5005 uint64_t tx_packets, tx_bytes;
5006 const char *keys[2];
5007 int64_t values[2];
5008 size_t stat_cnt = 0;
5009
5010 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5011 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5012 return;
5013 }
5014
5015 if (tx_packets != UINT64_MAX) {
5016 keys[stat_cnt] = "tx_packets";
5017 values[stat_cnt] = tx_packets;
5018 stat_cnt++;
5019 }
5020 if (tx_bytes != UINT64_MAX) {
5021 keys[stat_cnt] = "tx_bytes";
5022 values[stat_cnt] = tx_bytes;
5023 stat_cnt++;
5024 }
5025
5026 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5027 }