1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks
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:
7 * http://www.apache.org/licenses/LICENSE-2.0
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.
28 #include "dynamic-string.h"
36 #include "ofp-print.h"
38 #include "ofproto/ofproto.h"
39 #include "poll-loop.h"
42 #include "socket-util.h"
43 #include "stream-ssl.h"
45 #include "system-stats.h"
50 #include "vswitchd/vswitch-idl.h"
51 #include "xenserver.h"
53 #include "sflow_api.h"
54 #include "vlan-bitmap.h"
56 VLOG_DEFINE_THIS_MODULE(bridge);
58 COVERAGE_DEFINE(bridge_reconfigure);
61 /* These members are always valid. */
62 struct list port_elem; /* Element in struct port's "ifaces" list. */
63 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
64 struct port *port; /* Containing port. */
65 char *name; /* Host network device name. */
66 tag_type tag; /* Tag associated with this interface. */
68 /* These members are valid only after bridge_reconfigure() causes them to
70 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
71 int ofp_port; /* OpenFlow port number, -1 if unknown. */
72 struct netdev *netdev; /* Network device. */
73 const char *type; /* Usually same as cfg->type. */
74 const struct ovsrec_interface *cfg;
78 struct uuid uuid; /* UUID of this "mirror" record in database. */
79 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
80 struct bridge *bridge;
82 const struct ovsrec_mirror *cfg;
86 struct bridge *bridge;
87 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
90 const struct ovsrec_port *cfg;
92 /* An ordinary bridge port has 1 interface.
93 * A bridge port for bonding has at least 2 interfaces. */
94 struct list ifaces; /* List of "struct iface"s. */
98 struct hmap_node node; /* In 'all_bridges'. */
99 char *name; /* User-specified arbitrary name. */
100 char *type; /* Datapath type. */
101 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
102 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
103 const struct ovsrec_bridge *cfg;
105 /* OpenFlow switch processing. */
106 struct ofproto *ofproto; /* OpenFlow switch. */
109 struct hmap ports; /* "struct port"s indexed by name. */
110 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
111 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
113 /* Port mirroring. */
114 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
116 /* Synthetic local port if necessary. */
117 struct ovsrec_port synth_local_port;
118 struct ovsrec_interface synth_local_iface;
119 struct ovsrec_interface *synth_local_ifacep;
122 /* All bridges, indexed by name. */
123 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
125 /* OVSDB IDL used to obtain configuration. */
126 static struct ovsdb_idl *idl;
128 /* Each time this timer expires, the bridge fetches systems and interface
129 * statistics and pushes them into the database. */
130 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
131 static long long int stats_timer = LLONG_MIN;
133 /* Stores the time after which rate limited statistics may be written to the
134 * database. Only updated when changes to the database require rate limiting.
136 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
137 static long long int db_limiter = LLONG_MIN;
139 static void add_del_bridges(const struct ovsrec_open_vswitch *);
140 static void bridge_del_ofprotos(void);
141 static bool bridge_add_ofprotos(struct bridge *);
142 static void bridge_create(const struct ovsrec_bridge *);
143 static void bridge_destroy(struct bridge *);
144 static struct bridge *bridge_lookup(const char *name);
145 static unixctl_cb_func bridge_unixctl_dump_flows;
146 static unixctl_cb_func bridge_unixctl_reconnect;
147 static size_t bridge_get_controllers(const struct bridge *br,
148 struct ovsrec_controller ***controllersp);
149 static void bridge_add_del_ports(struct bridge *,
150 const unsigned long int *splinter_vlans);
151 static void bridge_add_ofproto_ports(struct bridge *);
152 static void bridge_del_ofproto_ports(struct bridge *);
153 static void bridge_refresh_ofp_port(struct bridge *);
154 static void bridge_configure_datapath_id(struct bridge *);
155 static void bridge_configure_flow_eviction_threshold(struct bridge *);
156 static void bridge_configure_netflow(struct bridge *);
157 static void bridge_configure_forward_bpdu(struct bridge *);
158 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
159 static void bridge_configure_stp(struct bridge *);
160 static void bridge_configure_remotes(struct bridge *,
161 const struct sockaddr_in *managers,
163 static void bridge_pick_local_hw_addr(struct bridge *,
164 uint8_t ea[ETH_ADDR_LEN],
165 struct iface **hw_addr_iface);
166 static uint64_t bridge_pick_datapath_id(struct bridge *,
167 const uint8_t bridge_ea[ETH_ADDR_LEN],
168 struct iface *hw_addr_iface);
169 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
171 static const char *get_port_other_config(const struct ovsrec_port *,
173 const char *default_value);
174 static uint64_t dpid_from_hash(const void *, size_t nbytes);
175 static bool bridge_has_bond_fake_iface(const struct bridge *,
177 static bool port_is_bond_fake_iface(const struct port *);
179 static unixctl_cb_func qos_unixctl_show;
181 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
182 static void port_add_ifaces(struct port *);
183 static void port_del_ifaces(struct port *);
184 static void port_destroy(struct port *);
185 static struct port *port_lookup(const struct bridge *, const char *name);
186 static void port_configure(struct port *);
187 static struct lacp_settings *port_configure_lacp(struct port *,
188 struct lacp_settings *);
189 static void port_configure_bond(struct port *, struct bond_settings *,
190 uint32_t *bond_stable_ids);
191 static bool port_is_synthetic(const struct port *);
193 static void bridge_configure_mirrors(struct bridge *);
194 static struct mirror *mirror_create(struct bridge *,
195 const struct ovsrec_mirror *);
196 static void mirror_destroy(struct mirror *);
197 static bool mirror_configure(struct mirror *);
198 static void mirror_refresh_stats(struct mirror *);
200 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
201 static struct iface *iface_create(struct port *port,
202 const struct ovsrec_interface *if_cfg);
203 static void iface_destroy(struct iface *);
204 static struct iface *iface_lookup(const struct bridge *, const char *name);
205 static struct iface *iface_find(const char *name);
206 static struct iface *iface_from_ofp_port(const struct bridge *,
208 static void iface_set_mac(struct iface *);
209 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
210 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
211 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
212 static void iface_configure_cfm(struct iface *);
213 static void iface_refresh_cfm_stats(struct iface *);
214 static void iface_refresh_stats(struct iface *);
215 static void iface_refresh_status(struct iface *);
216 static bool iface_is_synthetic(const struct iface *);
217 static const char *get_interface_other_config(const struct ovsrec_interface *,
219 const char *default_value);
221 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
223 static void shash_to_ovs_idl_map(struct shash *,
224 char ***keys, char ***values, size_t *n);
226 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
228 * This is deprecated. It is only for compatibility with broken device drivers
229 * in old versions of Linux that do not properly support VLANs when VLAN
230 * devices are not used. When broken device drivers are no longer in
231 * widespread use, we will delete these interfaces. */
233 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
234 static bool vlan_splinters_enabled_anywhere;
236 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
237 static unsigned long int *collect_splinter_vlans(
238 const struct ovsrec_open_vswitch *);
239 static void configure_splinter_port(struct port *);
240 static void add_vlan_splinter_ports(struct bridge *,
241 const unsigned long int *splinter_vlans,
242 struct shash *ports);
244 /* Public functions. */
246 /* Initializes the bridge module, configuring it to obtain its configuration
247 * from an OVSDB server accessed over 'remote', which should be a string in a
248 * form acceptable to ovsdb_idl_create(). */
250 bridge_init(const char *remote)
252 /* Create connection to database. */
253 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
254 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
256 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
257 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
258 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
259 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
260 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
261 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
262 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
264 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
265 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
266 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
268 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
269 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
270 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
271 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
273 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
274 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
275 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
276 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
277 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
283 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
284 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
285 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
287 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
288 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
289 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
290 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
292 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
294 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
297 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
299 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
301 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
303 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
305 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
306 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
307 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
309 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
311 /* Register unixctl commands. */
312 unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL);
313 unixctl_command_register("bridge/dump-flows", "bridge",
314 bridge_unixctl_dump_flows, NULL);
315 unixctl_command_register("bridge/reconnect", "[bridge]",
316 bridge_unixctl_reconnect, NULL);
325 struct bridge *br, *next_br;
327 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
330 ovsdb_idl_destroy(idl);
333 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
334 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
335 * responsible for freeing '*managersp' (with free()).
337 * You may be asking yourself "why does ovs-vswitchd care?", because
338 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
339 * should not be and in fact is not directly involved in that. But
340 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
341 * it has to tell in-band control where the managers are to enable that.
342 * (Thus, only managers connected in-band are collected.)
345 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
346 struct sockaddr_in **managersp, size_t *n_managersp)
348 struct sockaddr_in *managers = NULL;
349 size_t n_managers = 0;
353 /* Collect all of the potential targets from the "targets" columns of the
354 * rows pointed to by "manager_options", excluding any that are
357 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
358 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
360 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
361 sset_find_and_delete(&targets, m->target);
363 sset_add(&targets, m->target);
367 /* Now extract the targets' IP addresses. */
368 if (!sset_is_empty(&targets)) {
371 managers = xmalloc(sset_count(&targets) * sizeof *managers);
372 SSET_FOR_EACH (target, &targets) {
373 struct sockaddr_in *sin = &managers[n_managers];
375 if ((!strncmp(target, "tcp:", 4)
376 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
377 (!strncmp(target, "ssl:", 4)
378 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
383 sset_destroy(&targets);
385 *managersp = managers;
386 *n_managersp = n_managers;
390 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
392 unsigned long int *splinter_vlans;
393 struct sockaddr_in *managers;
394 struct bridge *br, *next;
395 int sflow_bridge_number;
398 COVERAGE_INC(bridge_reconfigure);
400 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
401 * iface"s according to 'ovs_cfg', with only very minimal configuration
404 * This is mostly an update to bridge data structures. Very little is
405 * pushed down to ofproto or lower layers. */
406 add_del_bridges(ovs_cfg);
407 splinter_vlans = collect_splinter_vlans(ovs_cfg);
408 HMAP_FOR_EACH (br, node, &all_bridges) {
409 bridge_add_del_ports(br, splinter_vlans);
411 free(splinter_vlans);
413 /* Delete all datapaths and datapath ports that are no longer configured.
415 * The kernel will reject any attempt to add a given port to a datapath if
416 * that port already belongs to a different datapath, so we must do all
417 * port deletions before any port additions. A datapath always has a
418 * "local port" so we must delete not-configured datapaths too. */
419 bridge_del_ofprotos();
420 HMAP_FOR_EACH (br, node, &all_bridges) {
422 bridge_del_ofproto_ports(br);
426 /* Create datapaths and datapath ports that are missing.
428 * After this is done, we have our final set of bridges, ports, and
429 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
430 * has at least one iface, every "struct iface" has a valid ofp_port and
432 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
433 if (!br->ofproto && !bridge_add_ofprotos(br)) {
437 HMAP_FOR_EACH (br, node, &all_bridges) {
438 bridge_refresh_ofp_port(br);
439 bridge_add_ofproto_ports(br);
442 /* Complete the configuration. */
443 sflow_bridge_number = 0;
444 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
445 HMAP_FOR_EACH (br, node, &all_bridges) {
448 /* We need the datapath ID early to allow LACP ports to use it as the
449 * default system ID. */
450 bridge_configure_datapath_id(br);
452 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
455 port_configure(port);
457 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
458 iface_configure_cfm(iface);
459 iface_configure_qos(iface, port->cfg->qos);
460 iface_set_mac(iface);
463 bridge_configure_mirrors(br);
464 bridge_configure_flow_eviction_threshold(br);
465 bridge_configure_forward_bpdu(br);
466 bridge_configure_remotes(br, managers, n_managers);
467 bridge_configure_netflow(br);
468 bridge_configure_sflow(br, &sflow_bridge_number);
469 bridge_configure_stp(br);
473 /* ovs-vswitchd has completed initialization, so allow the process that
474 * forked us to exit successfully. */
475 daemonize_complete();
478 /* Iterate over all ofprotos and delete any of them that do not have a
479 * configured bridge or that are the wrong type. */
481 bridge_del_ofprotos(void)
489 ofproto_enumerate_types(&types);
490 SSET_FOR_EACH (type, &types) {
493 ofproto_enumerate_names(type, &names);
494 SSET_FOR_EACH (name, &names) {
495 struct bridge *br = bridge_lookup(name);
496 if (!br || strcmp(type, br->type)) {
497 ofproto_delete(name, type);
501 sset_destroy(&names);
502 sset_destroy(&types);
506 bridge_add_ofprotos(struct bridge *br)
508 int error = ofproto_create(br->name, br->type, &br->ofproto);
510 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
517 port_configure(struct port *port)
519 const struct ovsrec_port *cfg = port->cfg;
520 struct bond_settings bond_settings;
521 struct lacp_settings lacp_settings;
522 struct ofproto_bundle_settings s;
525 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
526 configure_splinter_port(port);
535 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
536 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
537 s.slaves[s.n_slaves++] = iface->ofp_port;
542 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
546 /* Get VLAN trunks. */
549 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
553 if (cfg->vlan_mode) {
554 if (!strcmp(cfg->vlan_mode, "access")) {
555 s.vlan_mode = PORT_VLAN_ACCESS;
556 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
557 s.vlan_mode = PORT_VLAN_TRUNK;
558 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
559 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
560 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
561 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
563 /* This "can't happen" because ovsdb-server should prevent it. */
564 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
565 s.vlan_mode = PORT_VLAN_TRUNK;
569 s.vlan_mode = PORT_VLAN_ACCESS;
571 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
575 s.vlan_mode = PORT_VLAN_TRUNK;
578 s.use_priority_tags = !strcmp("true", get_port_other_config(
579 cfg, "priority-tags", ""));
581 /* Get LACP settings. */
582 s.lacp = port_configure_lacp(port, &lacp_settings);
586 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
587 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
588 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
591 s.lacp_slaves = NULL;
594 /* Get bond settings. */
595 if (s.n_slaves > 1) {
596 s.bond = &bond_settings;
597 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
598 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
601 s.bond_stable_ids = NULL;
603 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
604 netdev_set_miimon_interval(iface->netdev, 0);
609 ofproto_bundle_register(port->bridge->ofproto, port, &s);
615 free(s.bond_stable_ids);
618 /* Pick local port hardware address and datapath ID for 'br'. */
620 bridge_configure_datapath_id(struct bridge *br)
622 uint8_t ea[ETH_ADDR_LEN];
624 struct iface *local_iface;
625 struct iface *hw_addr_iface;
628 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
629 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
631 int error = netdev_set_etheraddr(local_iface->netdev, ea);
633 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
634 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
635 "Ethernet address: %s",
636 br->name, strerror(error));
639 memcpy(br->ea, ea, ETH_ADDR_LEN);
641 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
642 ofproto_set_datapath_id(br->ofproto, dpid);
644 dpid_string = xasprintf("%016"PRIx64, dpid);
645 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
649 /* Set NetFlow configuration on 'br'. */
651 bridge_configure_netflow(struct bridge *br)
653 struct ovsrec_netflow *cfg = br->cfg->netflow;
654 struct netflow_options opts;
657 ofproto_set_netflow(br->ofproto, NULL);
661 memset(&opts, 0, sizeof opts);
663 /* Get default NetFlow configuration from datapath.
664 * Apply overrides from 'cfg'. */
665 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
666 if (cfg->engine_type) {
667 opts.engine_type = *cfg->engine_type;
669 if (cfg->engine_id) {
670 opts.engine_id = *cfg->engine_id;
673 /* Configure active timeout interval. */
674 opts.active_timeout = cfg->active_timeout;
675 if (!opts.active_timeout) {
676 opts.active_timeout = -1;
677 } else if (opts.active_timeout < 0) {
678 VLOG_WARN("bridge %s: active timeout interval set to negative "
679 "value, using default instead (%d seconds)", br->name,
680 NF_ACTIVE_TIMEOUT_DEFAULT);
681 opts.active_timeout = -1;
684 /* Add engine ID to interface number to disambiguate bridgs? */
685 opts.add_id_to_iface = cfg->add_id_to_interface;
686 if (opts.add_id_to_iface) {
687 if (opts.engine_id > 0x7f) {
688 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
689 "another vswitch, choose an engine id less than 128",
692 if (hmap_count(&br->ports) > 508) {
693 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
694 "another port when more than 508 ports are used",
700 sset_init(&opts.collectors);
701 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
704 if (ofproto_set_netflow(br->ofproto, &opts)) {
705 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
707 sset_destroy(&opts.collectors);
710 /* Set sFlow configuration on 'br'. */
712 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
714 const struct ovsrec_sflow *cfg = br->cfg->sflow;
715 struct ovsrec_controller **controllers;
716 struct ofproto_sflow_options oso;
717 size_t n_controllers;
721 ofproto_set_sflow(br->ofproto, NULL);
725 memset(&oso, 0, sizeof oso);
727 sset_init(&oso.targets);
728 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
730 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
732 oso.sampling_rate = *cfg->sampling;
735 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
737 oso.polling_interval = *cfg->polling;
740 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
742 oso.header_len = *cfg->header;
745 oso.sub_id = (*sflow_bridge_number)++;
746 oso.agent_device = cfg->agent;
748 oso.control_ip = NULL;
749 n_controllers = bridge_get_controllers(br, &controllers);
750 for (i = 0; i < n_controllers; i++) {
751 if (controllers[i]->local_ip) {
752 oso.control_ip = controllers[i]->local_ip;
756 ofproto_set_sflow(br->ofproto, &oso);
758 sset_destroy(&oso.targets);
762 port_configure_stp(const struct ofproto *ofproto, struct port *port,
763 struct ofproto_port_stp_settings *port_s,
764 int *port_num_counter, unsigned long *port_num_bitmap)
766 const char *config_str;
769 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
770 if (config_str && !strcmp(config_str, "false")) {
771 port_s->enable = false;
774 port_s->enable = true;
777 /* STP over bonds is not supported. */
778 if (!list_is_singleton(&port->ifaces)) {
779 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
781 port_s->enable = false;
785 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
787 /* Internal ports shouldn't participate in spanning tree, so
789 if (!strcmp(iface->type, "internal")) {
790 VLOG_DBG("port %s: disable STP on internal ports", port->name);
791 port_s->enable = false;
795 /* STP on mirror output ports is not supported. */
796 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
797 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
798 port_s->enable = false;
802 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
804 unsigned long int port_num = strtoul(config_str, NULL, 0);
805 int port_idx = port_num - 1;
807 if (port_num < 1 || port_num > STP_MAX_PORTS) {
808 VLOG_ERR("port %s: invalid stp-port-num", port->name);
809 port_s->enable = false;
813 if (bitmap_is_set(port_num_bitmap, port_idx)) {
814 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
815 port->name, port_num);
816 port_s->enable = false;
819 bitmap_set1(port_num_bitmap, port_idx);
820 port_s->port_num = port_idx;
822 if (*port_num_counter > STP_MAX_PORTS) {
823 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
824 port_s->enable = false;
828 port_s->port_num = (*port_num_counter)++;
831 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
833 port_s->path_cost = strtoul(config_str, NULL, 10);
837 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
838 /* Couldn't get speed, so assume 100Mb/s. */
839 port_s->path_cost = 19;
843 mbps = netdev_features_to_bps(current) / 1000000;
844 port_s->path_cost = stp_convert_speed_to_cost(mbps);
848 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
850 port_s->priority = strtoul(config_str, NULL, 0);
852 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
856 /* Set spanning tree configuration on 'br'. */
858 bridge_configure_stp(struct bridge *br)
860 if (!br->cfg->stp_enable) {
861 ofproto_set_stp(br->ofproto, NULL);
863 struct ofproto_stp_settings br_s;
864 const char *config_str;
866 int port_num_counter;
867 unsigned long *port_num_bitmap;
869 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
871 uint8_t ea[ETH_ADDR_LEN];
873 if (eth_addr_from_string(config_str, ea)) {
874 br_s.system_id = eth_addr_to_uint64(ea);
876 br_s.system_id = eth_addr_to_uint64(br->ea);
877 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
878 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
881 br_s.system_id = eth_addr_to_uint64(br->ea);
884 config_str = bridge_get_other_config(br->cfg, "stp-priority");
886 br_s.priority = strtoul(config_str, NULL, 0);
888 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
891 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
893 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
895 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
898 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
900 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
902 br_s.max_age = STP_DEFAULT_MAX_AGE;
905 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
907 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
909 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
912 /* Configure STP on the bridge. */
913 if (ofproto_set_stp(br->ofproto, &br_s)) {
914 VLOG_ERR("bridge %s: could not enable STP", br->name);
918 /* Users must either set the port number with the "stp-port-num"
919 * configuration on all ports or none. If manual configuration
920 * is not done, then we allocate them sequentially. */
921 port_num_counter = 0;
922 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
923 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
924 struct ofproto_port_stp_settings port_s;
927 port_configure_stp(br->ofproto, port, &port_s,
928 &port_num_counter, port_num_bitmap);
930 /* As bonds are not supported, just apply configuration to
932 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
933 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
935 VLOG_ERR("port %s: could not enable STP", port->name);
941 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
942 && port_num_counter) {
943 VLOG_ERR("bridge %s: must manually configure all STP port "
944 "IDs or none, disabling", br->name);
945 ofproto_set_stp(br->ofproto, NULL);
947 bitmap_free(port_num_bitmap);
952 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
954 const struct port *port = port_lookup(br, name);
955 return port && port_is_bond_fake_iface(port);
959 port_is_bond_fake_iface(const struct port *port)
961 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
965 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
967 struct bridge *br, *next;
971 /* Collect new bridges' names and types. */
973 for (i = 0; i < cfg->n_bridges; i++) {
974 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
975 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
977 if (strchr(br_cfg->name, '/')) {
978 /* Prevent remote ovsdb-server users from accessing arbitrary
979 * directories, e.g. consider a bridge named "../../../etc/". */
980 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
982 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
983 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
987 /* Get rid of deleted bridges or those whose types have changed.
988 * Update 'cfg' of bridges that still exist. */
989 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
990 br->cfg = shash_find_data(&new_br, br->name);
991 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
992 br->cfg->datapath_type))) {
997 /* Add new bridges. */
998 for (i = 0; i < cfg->n_bridges; i++) {
999 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1000 struct bridge *br = bridge_lookup(br_cfg->name);
1002 bridge_create(br_cfg);
1006 shash_destroy(&new_br);
1009 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1012 * The kernel will reject any attempt to add a given port to a datapath if that
1013 * port already belongs to a different datapath, so we must do all port
1014 * deletions before any port additions. */
1016 bridge_del_ofproto_ports(struct bridge *br)
1018 struct ofproto_port_dump dump;
1019 struct ofproto_port ofproto_port;
1021 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1022 const char *name = ofproto_port.name;
1023 struct iface *iface;
1027 /* Ignore the local port. We can't change it anyhow. */
1028 if (!strcmp(name, br->name)) {
1032 /* Get the type that 'ofproto_port' should have (ordinarily the
1033 * type of its corresponding iface) or NULL if it should be
1035 iface = iface_lookup(br, name);
1036 type = (iface ? iface->type
1037 : bridge_has_bond_fake_iface(br, name) ? "internal"
1040 /* If it's the wrong type then delete the ofproto port. */
1042 && !strcmp(ofproto_port.type, type)
1043 && (!iface || !iface->netdev
1044 || !strcmp(netdev_get_type(iface->netdev), type))) {
1047 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1049 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1050 br->name, name, strerror(error));
1053 netdev_close(iface->netdev);
1054 iface->netdev = NULL;
1060 iface_set_ofp_port(struct iface *iface, int ofp_port)
1062 struct bridge *br = iface->port->bridge;
1064 assert(iface->ofp_port < 0 && ofp_port >= 0);
1065 iface->ofp_port = ofp_port;
1066 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1067 iface_set_ofport(iface->cfg, ofp_port);
1071 bridge_refresh_ofp_port(struct bridge *br)
1073 struct ofproto_port_dump dump;
1074 struct ofproto_port ofproto_port;
1077 /* Clear all the "ofp_port"es. */
1078 hmap_clear(&br->ifaces);
1079 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1080 struct iface *iface;
1082 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1083 iface->ofp_port = -1;
1087 /* Obtain the correct "ofp_port"s from ofproto. */
1088 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1089 struct iface *iface = iface_lookup(br, ofproto_port.name);
1091 if (iface->ofp_port >= 0) {
1092 VLOG_WARN("bridge %s: interface %s reported twice",
1093 br->name, ofproto_port.name);
1094 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1095 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1096 br->name, ofproto_port.ofp_port);
1098 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1104 /* Add an ofproto port for any "struct iface" that doesn't have one.
1105 * Delete any "struct iface" for which this fails.
1106 * Delete any "struct port" that thereby ends up with no ifaces. */
1108 bridge_add_ofproto_ports(struct bridge *br)
1110 struct port *port, *next_port;
1112 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1113 struct iface *iface, *next_iface;
1114 struct ofproto_port ofproto_port;
1116 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1119 /* Open the netdev. */
1120 if (!iface->netdev) {
1121 error = netdev_open(iface->name, iface->type, &iface->netdev);
1123 VLOG_WARN("could not open network device %s (%s)",
1124 iface->name, strerror(error));
1128 && port->cfg->vlan_mode
1129 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1130 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1136 /* Configure the netdev. */
1137 if (iface->netdev) {
1141 shash_from_ovs_idl_map(iface->cfg->key_options,
1142 iface->cfg->value_options,
1143 iface->cfg->n_options, &args);
1144 error = netdev_set_config(iface->netdev, &args);
1145 shash_destroy(&args);
1148 VLOG_WARN("could not configure network device %s (%s)",
1149 iface->name, strerror(error));
1150 netdev_close(iface->netdev);
1151 iface->netdev = NULL;
1155 /* Add the port, if necessary. */
1156 if (iface->netdev && iface->ofp_port < 0) {
1160 error = ofproto_port_add(br->ofproto, iface->netdev,
1163 iface_set_ofp_port(iface, ofp_port);
1165 netdev_close(iface->netdev);
1166 iface->netdev = NULL;
1170 /* Populate stats columns in new Interface rows. */
1171 if (iface->netdev && !iface->cfg->mtu) {
1172 iface_refresh_stats(iface);
1173 iface_refresh_status(iface);
1176 /* Delete the iface if we failed. */
1177 if (iface->netdev && iface->ofp_port >= 0) {
1178 VLOG_DBG("bridge %s: interface %s is on port %d",
1179 br->name, iface->name, iface->ofp_port);
1181 if (iface->netdev) {
1182 VLOG_ERR("bridge %s: missing %s interface, dropping",
1183 br->name, iface->name);
1185 /* We already reported a related error, don't bother
1186 * duplicating it. */
1188 iface_clear_db_record(iface->cfg);
1189 iface_destroy(iface);
1192 if (list_is_empty(&port->ifaces)) {
1193 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1198 /* Add bond fake iface if necessary. */
1199 if (port_is_bond_fake_iface(port)) {
1200 if (ofproto_port_query_by_name(br->ofproto, port->name,
1202 struct netdev *netdev;
1205 error = netdev_open(port->name, "internal", &netdev);
1207 ofproto_port_add(br->ofproto, netdev, NULL);
1208 netdev_close(netdev);
1210 VLOG_WARN("could not open network device %s (%s)",
1211 port->name, strerror(error));
1214 /* Already exists, nothing to do. */
1215 ofproto_port_destroy(&ofproto_port);
1222 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1226 for (i = 0; i < n; i++) {
1227 if (!strcmp(keys[i], key)) {
1235 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1237 return get_ovsrec_key_value(br_cfg->key_other_config,
1238 br_cfg->value_other_config,
1239 br_cfg->n_other_config, key);
1242 /* Set Flow eviction threshold */
1244 bridge_configure_flow_eviction_threshold(struct bridge *br)
1246 const char *threshold_str;
1249 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1250 if (threshold_str) {
1251 threshold = strtoul(threshold_str, NULL, 10);
1253 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1255 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1258 /* Set forward BPDU option. */
1260 bridge_configure_forward_bpdu(struct bridge *br)
1262 const char *forward_bpdu_str;
1263 bool forward_bpdu = false;
1265 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1266 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1267 forward_bpdu = true;
1269 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1273 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1274 struct iface **hw_addr_iface)
1276 struct hmapx mirror_output_ports;
1279 bool found_addr = false;
1283 *hw_addr_iface = NULL;
1285 /* Did the user request a particular MAC? */
1286 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1287 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1288 if (eth_addr_is_multicast(ea)) {
1289 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1290 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1291 } else if (eth_addr_is_zero(ea)) {
1292 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1298 /* Mirror output ports don't participate in picking the local hardware
1299 * address. ofproto can't help us find out whether a given port is a
1300 * mirror output because we haven't configured mirrors yet, so we need to
1301 * accumulate them ourselves. */
1302 hmapx_init(&mirror_output_ports);
1303 for (i = 0; i < br->cfg->n_mirrors; i++) {
1304 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1305 if (m->output_port) {
1306 hmapx_add(&mirror_output_ports, m->output_port);
1310 /* Otherwise choose the minimum non-local MAC address among all of the
1312 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1313 uint8_t iface_ea[ETH_ADDR_LEN];
1314 struct iface *candidate;
1315 struct iface *iface;
1317 /* Mirror output ports don't participate. */
1318 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1322 /* Choose the MAC address to represent the port. */
1324 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1325 /* Find the interface with this Ethernet address (if any) so that
1326 * we can provide the correct devname to the caller. */
1327 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1328 uint8_t candidate_ea[ETH_ADDR_LEN];
1329 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1330 && eth_addr_equals(iface_ea, candidate_ea)) {
1335 /* Choose the interface whose MAC address will represent the port.
1336 * The Linux kernel bonding code always chooses the MAC address of
1337 * the first slave added to a bond, and the Fedora networking
1338 * scripts always add slaves to a bond in alphabetical order, so
1339 * for compatibility we choose the interface with the name that is
1340 * first in alphabetical order. */
1341 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1342 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1347 /* The local port doesn't count (since we're trying to choose its
1348 * MAC address anyway). */
1349 if (iface->ofp_port == OFPP_LOCAL) {
1354 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1360 /* Compare against our current choice. */
1361 if (!eth_addr_is_multicast(iface_ea) &&
1362 !eth_addr_is_local(iface_ea) &&
1363 !eth_addr_is_reserved(iface_ea) &&
1364 !eth_addr_is_zero(iface_ea) &&
1365 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1367 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1368 *hw_addr_iface = iface;
1373 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1374 br->name, ETH_ADDR_ARGS(ea));
1376 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1377 *hw_addr_iface = NULL;
1378 VLOG_WARN("bridge %s: using default bridge Ethernet "
1379 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1382 hmapx_destroy(&mirror_output_ports);
1385 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1386 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1387 * an interface on 'br', then that interface must be passed in as
1388 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1389 * 'hw_addr_iface' must be passed in as a null pointer. */
1391 bridge_pick_datapath_id(struct bridge *br,
1392 const uint8_t bridge_ea[ETH_ADDR_LEN],
1393 struct iface *hw_addr_iface)
1396 * The procedure for choosing a bridge MAC address will, in the most
1397 * ordinary case, also choose a unique MAC that we can use as a datapath
1398 * ID. In some special cases, though, multiple bridges will end up with
1399 * the same MAC address. This is OK for the bridges, but it will confuse
1400 * the OpenFlow controller, because each datapath needs a unique datapath
1403 * Datapath IDs must be unique. It is also very desirable that they be
1404 * stable from one run to the next, so that policy set on a datapath
1407 const char *datapath_id;
1410 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1411 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1415 if (!hw_addr_iface) {
1417 * A purely internal bridge, that is, one that has no non-virtual
1418 * network devices on it at all, is difficult because it has no
1419 * natural unique identifier at all.
1421 * When the host is a XenServer, we handle this case by hashing the
1422 * host's UUID with the name of the bridge. Names of bridges are
1423 * persistent across XenServer reboots, although they can be reused if
1424 * an internal network is destroyed and then a new one is later
1425 * created, so this is fairly effective.
1427 * When the host is not a XenServer, we punt by using a random MAC
1428 * address on each run.
1430 const char *host_uuid = xenserver_get_host_uuid();
1432 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1433 dpid = dpid_from_hash(combined, strlen(combined));
1439 return eth_addr_to_uint64(bridge_ea);
1443 dpid_from_hash(const void *data, size_t n)
1445 uint8_t hash[SHA1_DIGEST_SIZE];
1447 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1448 sha1_bytes(data, n, hash);
1449 eth_addr_mark_random(hash);
1450 return eth_addr_to_uint64(hash);
1454 iface_refresh_status(struct iface *iface)
1458 enum netdev_flags flags;
1465 if (iface_is_synthetic(iface)) {
1471 if (!netdev_get_status(iface->netdev, &sh)) {
1473 char **keys, **values;
1475 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1476 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1481 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1484 shash_destroy_free_data(&sh);
1486 error = netdev_get_flags(iface->netdev, &flags);
1488 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1491 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1494 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1496 ovsrec_interface_set_duplex(iface->cfg,
1497 netdev_features_is_full_duplex(current)
1499 /* warning: uint64_t -> int64_t conversion */
1500 bps = netdev_features_to_bps(current);
1501 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1504 ovsrec_interface_set_duplex(iface->cfg, NULL);
1505 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1508 error = netdev_get_mtu(iface->netdev, &mtu);
1511 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1514 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1518 /* Writes 'iface''s CFM statistics to the database. */
1520 iface_refresh_cfm_stats(struct iface *iface)
1522 const struct ovsrec_interface *cfg = iface->cfg;
1524 const uint64_t *rmps;
1527 if (iface_is_synthetic(iface)) {
1531 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1534 bool fault_bool = fault;
1535 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1537 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1540 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1541 iface->ofp_port, &rmps, &n_rmps);
1543 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1546 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1551 iface_refresh_stats(struct iface *iface)
1553 #define IFACE_STATS \
1554 IFACE_STAT(rx_packets, "rx_packets") \
1555 IFACE_STAT(tx_packets, "tx_packets") \
1556 IFACE_STAT(rx_bytes, "rx_bytes") \
1557 IFACE_STAT(tx_bytes, "tx_bytes") \
1558 IFACE_STAT(rx_dropped, "rx_dropped") \
1559 IFACE_STAT(tx_dropped, "tx_dropped") \
1560 IFACE_STAT(rx_errors, "rx_errors") \
1561 IFACE_STAT(tx_errors, "tx_errors") \
1562 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1563 IFACE_STAT(rx_over_errors, "rx_over_err") \
1564 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1565 IFACE_STAT(collisions, "collisions")
1567 #define IFACE_STAT(MEMBER, NAME) NAME,
1568 static char *keys[] = { IFACE_STATS };
1570 int64_t values[ARRAY_SIZE(keys)];
1573 struct netdev_stats stats;
1575 if (iface_is_synthetic(iface)) {
1579 /* Intentionally ignore return value, since errors will set 'stats' to
1580 * all-1s, and we will deal with that correctly below. */
1581 netdev_get_stats(iface->netdev, &stats);
1583 /* Copy statistics into values[] array. */
1585 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1588 assert(i == ARRAY_SIZE(keys));
1590 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1595 br_refresh_stp_status(struct bridge *br)
1597 struct ofproto *ofproto = br->ofproto;
1598 struct ofproto_stp_status status;
1599 char *keys[3], *values[3];
1602 if (ofproto_get_stp_status(ofproto, &status)) {
1606 if (!status.enabled) {
1607 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1611 keys[0] = "stp_bridge_id",
1612 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1613 keys[1] = "stp_designated_root",
1614 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1615 keys[2] = "stp_root_path_cost",
1616 values[2] = xasprintf("%d", status.root_path_cost);
1618 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1620 for (i = 0; i < ARRAY_SIZE(values); i++) {
1626 port_refresh_stp_status(struct port *port)
1628 struct ofproto *ofproto = port->bridge->ofproto;
1629 struct iface *iface;
1630 struct ofproto_port_stp_status status;
1632 char *str_values[4];
1633 int64_t int_values[3];
1636 if (port_is_synthetic(port)) {
1640 /* STP doesn't currently support bonds. */
1641 if (!list_is_singleton(&port->ifaces)) {
1642 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1646 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1648 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1652 if (!status.enabled) {
1653 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1654 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1658 /* Set Status column. */
1659 keys[0] = "stp_port_id";
1660 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1661 keys[1] = "stp_state";
1662 str_values[1] = xstrdup(stp_state_name(status.state));
1663 keys[2] = "stp_sec_in_state";
1664 str_values[2] = xasprintf("%u", status.sec_in_state);
1665 keys[3] = "stp_role";
1666 str_values[3] = xstrdup(stp_role_name(status.role));
1668 ovsrec_port_set_status(port->cfg, keys, str_values,
1669 ARRAY_SIZE(str_values));
1671 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1672 free(str_values[i]);
1675 /* Set Statistics column. */
1676 keys[0] = "stp_tx_count";
1677 int_values[0] = status.tx_count;
1678 keys[1] = "stp_rx_count";
1679 int_values[1] = status.rx_count;
1680 keys[2] = "stp_error_count";
1681 int_values[2] = status.error_count;
1683 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1684 ARRAY_SIZE(int_values));
1688 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1692 /* Use other-config:enable-system-stats by preference. */
1693 enable = get_ovsrec_key_value(cfg->key_other_config,
1694 cfg->value_other_config,
1695 cfg->n_other_config,
1696 "enable-statistics");
1698 return !strcmp(enable, "true");
1701 /* Disable by default. */
1706 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1708 struct ovsdb_datum datum;
1712 if (enable_system_stats(cfg)) {
1713 get_system_stats(&stats);
1716 ovsdb_datum_from_shash(&datum, &stats);
1717 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1721 static inline const char *
1722 nx_role_to_str(enum nx_role role)
1727 case NX_ROLE_MASTER:
1732 return "*** INVALID ROLE ***";
1737 refresh_controller_status(void)
1741 const struct ovsrec_controller *cfg;
1745 /* Accumulate status for controllers on all bridges. */
1746 HMAP_FOR_EACH (br, node, &all_bridges) {
1747 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1750 /* Update each controller in the database with current status. */
1751 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1752 struct ofproto_controller_info *cinfo =
1753 shash_find_data(&info, cfg->target);
1756 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1757 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1758 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1759 (char **) cinfo->pairs.values,
1762 ovsrec_controller_set_is_connected(cfg, false);
1763 ovsrec_controller_set_role(cfg, NULL);
1764 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1768 ofproto_free_ofproto_controller_info(&info);
1772 refresh_cfm_stats(void)
1774 static struct ovsdb_idl_txn *txn = NULL;
1779 txn = ovsdb_idl_txn_create(idl);
1781 HMAP_FOR_EACH (br, node, &all_bridges) {
1782 struct iface *iface;
1784 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1785 iface_refresh_cfm_stats(iface);
1790 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1791 ovsdb_idl_txn_destroy(txn);
1796 /* Performs periodic activity required by bridges that needs to be done with
1797 * the least possible latency.
1799 * It makes sense to call this function a couple of times per poll loop, to
1800 * provide a significant performance boost on some benchmarks with ofprotos
1801 * that use the ofproto-dpif implementation. */
1803 bridge_run_fast(void)
1807 HMAP_FOR_EACH (br, node, &all_bridges) {
1808 ofproto_run_fast(br->ofproto);
1815 const struct ovsrec_open_vswitch *cfg;
1817 bool vlan_splinters_changed;
1818 bool database_changed;
1821 /* (Re)configure if necessary. */
1822 database_changed = ovsdb_idl_run(idl);
1823 if (ovsdb_idl_is_lock_contended(idl)) {
1824 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1825 struct bridge *br, *next_br;
1827 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1828 "disabling this process until it goes away");
1830 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1834 } else if (!ovsdb_idl_has_lock(idl)) {
1837 cfg = ovsrec_open_vswitch_first(idl);
1839 /* Let each bridge do the work that it needs to do. */
1840 HMAP_FOR_EACH (br, node, &all_bridges) {
1841 ofproto_run(br->ofproto);
1844 /* Re-configure SSL. We do this on every trip through the main loop,
1845 * instead of just when the database changes, because the contents of the
1846 * key and certificate files can change without the database changing.
1848 * We do this before bridge_reconfigure() because that function might
1849 * initiate SSL connections and thus requires SSL to be configured. */
1850 if (cfg && cfg->ssl) {
1851 const struct ovsrec_ssl *ssl = cfg->ssl;
1853 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1854 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1857 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1859 vlan_splinters_changed = false;
1860 if (vlan_splinters_enabled_anywhere) {
1861 HMAP_FOR_EACH (br, node, &all_bridges) {
1862 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1863 vlan_splinters_changed = true;
1869 if (database_changed || vlan_splinters_changed) {
1871 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1873 bridge_reconfigure(cfg);
1875 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1876 ovsdb_idl_txn_commit(txn);
1877 ovsdb_idl_txn_destroy(txn); /* XXX */
1879 /* We still need to reconfigure to avoid dangling pointers to
1880 * now-destroyed ovsrec structures inside bridge data. */
1881 static const struct ovsrec_open_vswitch null_cfg;
1883 bridge_reconfigure(&null_cfg);
1887 /* Refresh system and interface stats if necessary. */
1888 if (time_msec() >= stats_timer) {
1890 struct ovsdb_idl_txn *txn;
1892 txn = ovsdb_idl_txn_create(idl);
1893 HMAP_FOR_EACH (br, node, &all_bridges) {
1897 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1898 struct iface *iface;
1900 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1901 iface_refresh_stats(iface);
1902 iface_refresh_status(iface);
1906 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1907 mirror_refresh_stats(m);
1911 refresh_system_stats(cfg);
1912 refresh_controller_status();
1913 ovsdb_idl_txn_commit(txn);
1914 ovsdb_idl_txn_destroy(txn); /* XXX */
1917 stats_timer = time_msec() + STATS_INTERVAL;
1920 if (time_msec() >= db_limiter) {
1921 struct ovsdb_idl_txn *txn;
1923 txn = ovsdb_idl_txn_create(idl);
1924 HMAP_FOR_EACH (br, node, &all_bridges) {
1925 struct iface *iface;
1928 br_refresh_stp_status(br);
1930 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1931 port_refresh_stp_status(port);
1934 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1935 const char *link_state;
1936 int64_t link_resets;
1939 if (iface_is_synthetic(iface)) {
1943 current = ofproto_port_is_lacp_current(br->ofproto,
1947 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1949 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1952 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1953 ovsrec_interface_set_link_state(iface->cfg, link_state);
1955 link_resets = netdev_get_carrier_resets(iface->netdev);
1956 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1960 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
1961 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1963 ovsdb_idl_txn_destroy(txn);
1966 refresh_cfm_stats();
1972 ovsdb_idl_wait(idl);
1973 if (!hmap_is_empty(&all_bridges)) {
1976 HMAP_FOR_EACH (br, node, &all_bridges) {
1977 ofproto_wait(br->ofproto);
1979 poll_timer_wait_until(stats_timer);
1981 if (db_limiter > time_msec()) {
1982 poll_timer_wait_until(db_limiter);
1987 /* QoS unixctl user interface functions. */
1989 struct qos_unixctl_show_cbdata {
1991 struct iface *iface;
1995 qos_unixctl_show_cb(unsigned int queue_id,
1996 const struct shash *details,
1999 struct qos_unixctl_show_cbdata *data = aux;
2000 struct ds *ds = data->ds;
2001 struct iface *iface = data->iface;
2002 struct netdev_queue_stats stats;
2003 struct shash_node *node;
2006 ds_put_cstr(ds, "\n");
2008 ds_put_format(ds, "Queue %u:\n", queue_id);
2010 ds_put_cstr(ds, "Default:\n");
2013 SHASH_FOR_EACH (node, details) {
2014 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2017 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2019 if (stats.tx_packets != UINT64_MAX) {
2020 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2023 if (stats.tx_bytes != UINT64_MAX) {
2024 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2027 if (stats.tx_errors != UINT64_MAX) {
2028 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2031 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2032 queue_id, strerror(error));
2037 qos_unixctl_show(struct unixctl_conn *conn,
2038 const char *args, void *aux OVS_UNUSED)
2040 struct ds ds = DS_EMPTY_INITIALIZER;
2041 struct shash sh = SHASH_INITIALIZER(&sh);
2042 struct iface *iface;
2044 struct shash_node *node;
2045 struct qos_unixctl_show_cbdata data;
2048 iface = iface_find(args);
2050 unixctl_command_reply(conn, 501, "no such interface");
2054 netdev_get_qos(iface->netdev, &type, &sh);
2056 if (*type != '\0') {
2057 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2059 SHASH_FOR_EACH (node, &sh) {
2060 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2065 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2068 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2070 unixctl_command_reply(conn, 200, ds_cstr(&ds));
2072 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2073 unixctl_command_reply(conn, 501, ds_cstr(&ds));
2076 shash_destroy_free_data(&sh);
2080 /* Bridge reconfiguration functions. */
2082 bridge_create(const struct ovsrec_bridge *br_cfg)
2086 assert(!bridge_lookup(br_cfg->name));
2087 br = xzalloc(sizeof *br);
2089 br->name = xstrdup(br_cfg->name);
2090 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2093 /* Derive the default Ethernet address from the bridge's UUID. This should
2094 * be unique and it will be stable between ovs-vswitchd runs. */
2095 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2096 eth_addr_mark_random(br->default_ea);
2098 hmap_init(&br->ports);
2099 hmap_init(&br->ifaces);
2100 hmap_init(&br->iface_by_name);
2101 hmap_init(&br->mirrors);
2103 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2107 bridge_destroy(struct bridge *br)
2110 struct mirror *mirror, *next_mirror;
2111 struct port *port, *next_port;
2113 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2116 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2117 mirror_destroy(mirror);
2119 hmap_remove(&all_bridges, &br->node);
2120 ofproto_destroy(br->ofproto);
2121 hmap_destroy(&br->ifaces);
2122 hmap_destroy(&br->ports);
2123 hmap_destroy(&br->iface_by_name);
2124 hmap_destroy(&br->mirrors);
2131 static struct bridge *
2132 bridge_lookup(const char *name)
2136 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2137 if (!strcmp(br->name, name)) {
2144 /* Handle requests for a listing of all flows known by the OpenFlow
2145 * stack, including those normally hidden. */
2147 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
2148 const char *args, void *aux OVS_UNUSED)
2153 br = bridge_lookup(args);
2155 unixctl_command_reply(conn, 501, "Unknown bridge");
2160 ofproto_get_all_flows(br->ofproto, &results);
2162 unixctl_command_reply(conn, 200, ds_cstr(&results));
2163 ds_destroy(&results);
2166 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2167 * connections and reconnect. If BRIDGE is not specified, then all bridges
2168 * drop their controller connections and reconnect. */
2170 bridge_unixctl_reconnect(struct unixctl_conn *conn,
2171 const char *args, void *aux OVS_UNUSED)
2174 if (args[0] != '\0') {
2175 br = bridge_lookup(args);
2177 unixctl_command_reply(conn, 501, "Unknown bridge");
2180 ofproto_reconnect_controllers(br->ofproto);
2182 HMAP_FOR_EACH (br, node, &all_bridges) {
2183 ofproto_reconnect_controllers(br->ofproto);
2186 unixctl_command_reply(conn, 200, NULL);
2190 bridge_get_controllers(const struct bridge *br,
2191 struct ovsrec_controller ***controllersp)
2193 struct ovsrec_controller **controllers;
2194 size_t n_controllers;
2196 controllers = br->cfg->controller;
2197 n_controllers = br->cfg->n_controller;
2199 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2205 *controllersp = controllers;
2207 return n_controllers;
2210 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2211 * those configured in 'br->cfg'. */
2213 bridge_add_del_ports(struct bridge *br,
2214 const unsigned long int *splinter_vlans)
2216 struct port *port, *next;
2217 struct shash_node *node;
2218 struct shash new_ports;
2221 /* Collect new ports. */
2222 shash_init(&new_ports);
2223 for (i = 0; i < br->cfg->n_ports; i++) {
2224 const char *name = br->cfg->ports[i]->name;
2225 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2226 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2230 if (bridge_get_controllers(br, NULL)
2231 && !shash_find(&new_ports, br->name)) {
2232 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2233 br->name, br->name);
2235 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2236 br->synth_local_port.n_interfaces = 1;
2237 br->synth_local_port.name = br->name;
2239 br->synth_local_iface.name = br->name;
2240 br->synth_local_iface.type = "internal";
2242 br->synth_local_ifacep = &br->synth_local_iface;
2244 shash_add(&new_ports, br->name, &br->synth_local_port);
2247 if (splinter_vlans) {
2248 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2251 /* Get rid of deleted ports.
2252 * Get rid of deleted interfaces on ports that still exist. */
2253 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2254 port->cfg = shash_find_data(&new_ports, port->name);
2258 port_del_ifaces(port);
2262 /* Create new ports.
2263 * Add new interfaces to existing ports. */
2264 SHASH_FOR_EACH (node, &new_ports) {
2265 struct port *port = port_lookup(br, node->name);
2267 struct ovsrec_port *cfg = node->data;
2268 port = port_create(br, cfg);
2270 port_add_ifaces(port);
2271 if (list_is_empty(&port->ifaces)) {
2272 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2273 br->name, port->name);
2277 shash_destroy(&new_ports);
2280 /* Initializes 'oc' appropriately as a management service controller for
2283 * The caller must free oc->target when it is no longer needed. */
2285 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2286 struct ofproto_controller *oc)
2288 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2289 oc->max_backoff = 0;
2290 oc->probe_interval = 60;
2291 oc->band = OFPROTO_OUT_OF_BAND;
2293 oc->burst_limit = 0;
2296 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2298 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2299 struct ofproto_controller *oc)
2301 oc->target = c->target;
2302 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2303 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2304 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2305 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2306 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2307 oc->burst_limit = (c->controller_burst_limit
2308 ? *c->controller_burst_limit : 0);
2311 /* Configures the IP stack for 'br''s local interface properly according to the
2312 * configuration in 'c'. */
2314 bridge_configure_local_iface_netdev(struct bridge *br,
2315 struct ovsrec_controller *c)
2317 struct netdev *netdev;
2318 struct in_addr mask, gateway;
2320 struct iface *local_iface;
2323 /* If there's no local interface or no IP address, give up. */
2324 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2325 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2329 /* Bring up the local interface. */
2330 netdev = local_iface->netdev;
2331 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2333 /* Configure the IP address and netmask. */
2334 if (!c->local_netmask
2335 || !inet_aton(c->local_netmask, &mask)
2337 mask.s_addr = guess_netmask(ip.s_addr);
2339 if (!netdev_set_in4(netdev, ip, mask)) {
2340 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2341 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2344 /* Configure the default gateway. */
2345 if (c->local_gateway
2346 && inet_aton(c->local_gateway, &gateway)
2347 && gateway.s_addr) {
2348 if (!netdev_add_router(netdev, gateway)) {
2349 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2350 br->name, IP_ARGS(&gateway.s_addr));
2355 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2356 * in either string are treated as equal to any number of slashes in the other,
2357 * e.g. "x///y" is equal to "x/y". */
2359 equal_pathnames(const char *a, const char *b)
2363 a += strspn(a, "/");
2364 b += strspn(b, "/");
2365 } else if (*a == '\0') {
2376 bridge_configure_remotes(struct bridge *br,
2377 const struct sockaddr_in *managers, size_t n_managers)
2379 const char *disable_ib_str, *queue_id_str;
2380 bool disable_in_band = false;
2383 struct ovsrec_controller **controllers;
2384 size_t n_controllers;
2386 enum ofproto_fail_mode fail_mode;
2388 struct ofproto_controller *ocs;
2392 /* Check if we should disable in-band control on this bridge. */
2393 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2394 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2395 disable_in_band = true;
2398 /* Set OpenFlow queue ID for in-band control. */
2399 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2400 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2401 ofproto_set_in_band_queue(br->ofproto, queue_id);
2403 if (disable_in_band) {
2404 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2406 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2409 n_controllers = bridge_get_controllers(br, &controllers);
2411 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2414 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2415 for (i = 0; i < n_controllers; i++) {
2416 struct ovsrec_controller *c = controllers[i];
2418 if (!strncmp(c->target, "punix:", 6)
2419 || !strncmp(c->target, "unix:", 5)) {
2420 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2423 whitelist = xasprintf("unix:%s/%s.controller",
2424 ovs_rundir(), br->name);
2425 if (!equal_pathnames(c->target, whitelist)) {
2426 /* Prevent remote ovsdb-server users from accessing arbitrary
2427 * Unix domain sockets and overwriting arbitrary local
2429 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2430 "controller \"%s\" due to possibility for remote "
2431 "exploit. Instead, specify whitelisted \"%s\" or "
2432 "connect to \"unix:%s/%s.mgmt\" (which is always "
2433 "available without special configuration).",
2434 br->name, c->target, whitelist,
2435 ovs_rundir(), br->name);
2443 bridge_configure_local_iface_netdev(br, c);
2444 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2445 if (disable_in_band) {
2446 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2451 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2452 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2455 /* Set the fail-mode. */
2456 fail_mode = !br->cfg->fail_mode
2457 || !strcmp(br->cfg->fail_mode, "standalone")
2458 ? OFPROTO_FAIL_STANDALONE
2459 : OFPROTO_FAIL_SECURE;
2460 ofproto_set_fail_mode(br->ofproto, fail_mode);
2462 /* Configure OpenFlow controller connection snooping. */
2463 if (!ofproto_has_snoops(br->ofproto)) {
2467 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2468 ovs_rundir(), br->name));
2469 ofproto_set_snoops(br->ofproto, &snoops);
2470 sset_destroy(&snoops);
2474 /* Port functions. */
2476 static struct port *
2477 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2481 port = xzalloc(sizeof *port);
2483 port->name = xstrdup(cfg->name);
2485 list_init(&port->ifaces);
2487 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2489 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2495 get_port_other_config(const struct ovsrec_port *port, const char *key,
2496 const char *default_value)
2500 value = get_ovsrec_key_value(port->key_other_config,
2501 port->value_other_config,
2502 port->n_other_config, key);
2503 return value ? value : default_value;
2507 get_interface_other_config(const struct ovsrec_interface *iface,
2508 const char *key, const char *default_value)
2512 value = get_ovsrec_key_value(iface->key_other_config,
2513 iface->value_other_config,
2514 iface->n_other_config, key);
2515 return value ? value : default_value;
2518 /* Deletes interfaces from 'port' that are no longer configured for it. */
2520 port_del_ifaces(struct port *port)
2522 struct iface *iface, *next;
2523 struct sset new_ifaces;
2526 /* Collect list of new interfaces. */
2527 sset_init(&new_ifaces);
2528 for (i = 0; i < port->cfg->n_interfaces; i++) {
2529 const char *name = port->cfg->interfaces[i]->name;
2530 const char *type = port->cfg->interfaces[i]->name;
2531 if (strcmp(type, "null")) {
2532 sset_add(&new_ifaces, name);
2536 /* Get rid of deleted interfaces. */
2537 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2538 if (!sset_contains(&new_ifaces, iface->name)) {
2539 iface_destroy(iface);
2543 sset_destroy(&new_ifaces);
2546 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2549 port_add_ifaces(struct port *port)
2551 struct shash new_ifaces;
2552 struct shash_node *node;
2555 /* Collect new ifaces. */
2556 shash_init(&new_ifaces);
2557 for (i = 0; i < port->cfg->n_interfaces; i++) {
2558 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2559 if (strcmp(cfg->type, "null")
2560 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2561 VLOG_WARN("port %s: %s specified twice as port interface",
2562 port->name, cfg->name);
2563 iface_clear_db_record(cfg);
2567 /* Create new interfaces.
2568 * Update interface types and 'cfg' members. */
2569 SHASH_FOR_EACH (node, &new_ifaces) {
2570 const struct ovsrec_interface *cfg = node->data;
2571 const char *iface_name = node->name;
2572 struct iface *iface;
2574 iface = iface_lookup(port->bridge, iface_name);
2576 iface = iface_create(port, cfg);
2581 /* Determine interface type. The local port always has type
2582 * "internal". Other ports take their type from the database and
2583 * default to "system" if none is specified. */
2584 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2585 : cfg->type[0] ? cfg->type
2588 shash_destroy(&new_ifaces);
2592 port_destroy(struct port *port)
2595 struct bridge *br = port->bridge;
2596 struct iface *iface, *next;
2599 ofproto_bundle_unregister(br->ofproto, port);
2602 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2603 iface_destroy(iface);
2606 hmap_remove(&br->ports, &port->hmap_node);
2608 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2615 static struct port *
2616 port_lookup(const struct bridge *br, const char *name)
2620 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2622 if (!strcmp(port->name, name)) {
2630 enable_lacp(struct port *port, bool *activep)
2632 if (!port->cfg->lacp) {
2633 /* XXX when LACP implementation has been sufficiently tested, enable by
2634 * default and make active on bonded ports. */
2636 } else if (!strcmp(port->cfg->lacp, "off")) {
2638 } else if (!strcmp(port->cfg->lacp, "active")) {
2641 } else if (!strcmp(port->cfg->lacp, "passive")) {
2645 VLOG_WARN("port %s: unknown LACP mode %s",
2646 port->name, port->cfg->lacp);
2651 static struct lacp_settings *
2652 port_configure_lacp(struct port *port, struct lacp_settings *s)
2654 const char *lacp_time, *system_id;
2655 long long int custom_time;
2658 if (!enable_lacp(port, &s->active)) {
2662 s->name = port->name;
2664 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2666 || sscanf(system_id, ETH_ADDR_SCAN_FMT,
2667 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2668 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2671 if (eth_addr_is_zero(s->id)) {
2672 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2676 /* Prefer bondable links if unspecified. */
2677 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2679 s->priority = (priority > 0 && priority <= UINT16_MAX
2681 : UINT16_MAX - !list_is_short(&port->ifaces));
2683 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2687 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2688 custom_time = atoi(lacp_time);
2689 if (!strcmp(lacp_time, "fast")) {
2690 s->lacp_time = LACP_TIME_FAST;
2691 } else if (!strcmp(lacp_time, "slow")) {
2692 s->lacp_time = LACP_TIME_SLOW;
2693 } else if (custom_time > 0) {
2694 s->lacp_time = LACP_TIME_CUSTOM;
2695 s->custom_time = custom_time;
2697 s->lacp_time = LACP_TIME_SLOW;
2704 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2706 int priority, portid, key;
2708 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2709 priority = atoi(get_interface_other_config(iface->cfg,
2710 "lacp-port-priority", "0"));
2711 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2714 if (portid <= 0 || portid > UINT16_MAX) {
2715 portid = iface->ofp_port;
2718 if (priority <= 0 || priority > UINT16_MAX) {
2719 priority = UINT16_MAX;
2722 if (key < 0 || key > UINT16_MAX) {
2726 s->name = iface->name;
2728 s->priority = priority;
2733 port_configure_bond(struct port *port, struct bond_settings *s,
2734 uint32_t *bond_stable_ids)
2736 const char *detect_s;
2737 struct iface *iface;
2738 int miimon_interval;
2741 s->name = port->name;
2742 s->balance = BM_SLB;
2743 if (port->cfg->bond_mode) {
2744 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2745 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2746 port->name, port->cfg->bond_mode,
2747 bond_mode_to_string(s->balance));
2750 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2752 /* XXX: Post version 1.4.*, change the default bond_mode to
2753 * active-backup. Until then, warn that the change is imminent. */
2754 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2755 " in future versions, the default bond_mode is expected"
2756 " to change to active-backup", port->name,
2757 bond_mode_to_string(s->balance));
2759 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2760 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2761 "please use another bond type or disable flood_vlans",
2765 miimon_interval = atoi(get_port_other_config(port->cfg,
2766 "bond-miimon-interval", "0"));
2767 if (miimon_interval <= 0) {
2768 miimon_interval = 200;
2771 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2772 if (!strcmp(detect_s, "carrier")) {
2773 miimon_interval = 0;
2774 } else if (strcmp(detect_s, "miimon")) {
2775 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2776 "defaulting to carrier", port->name, detect_s);
2777 miimon_interval = 0;
2780 s->up_delay = MAX(0, port->cfg->bond_updelay);
2781 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2782 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2783 s->rebalance_interval = atoi(
2784 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2785 if (s->rebalance_interval < 1000) {
2786 s->rebalance_interval = 1000;
2789 s->fake_iface = port->cfg->bond_fake_iface;
2792 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2793 long long stable_id;
2795 stable_id = atoll(get_interface_other_config(iface->cfg,
2796 "bond-stable-id", "0"));
2797 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2798 stable_id = iface->ofp_port;
2800 bond_stable_ids[i++] = stable_id;
2802 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2806 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2807 * instead of obtaining it from the database. */
2809 port_is_synthetic(const struct port *port)
2811 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2814 /* Interface functions. */
2816 static struct iface *
2817 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2819 struct bridge *br = port->bridge;
2820 struct iface *iface;
2821 char *name = if_cfg->name;
2823 iface = xzalloc(sizeof *iface);
2825 iface->name = xstrdup(name);
2826 iface->ofp_port = -1;
2827 iface->tag = tag_create_random();
2828 iface->netdev = NULL;
2829 iface->cfg = if_cfg;
2831 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2833 list_push_back(&port->ifaces, &iface->port_elem);
2835 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2841 iface_destroy(struct iface *iface)
2844 struct port *port = iface->port;
2845 struct bridge *br = port->bridge;
2847 if (br->ofproto && iface->ofp_port >= 0) {
2848 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2851 if (iface->ofp_port >= 0) {
2852 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2855 list_remove(&iface->port_elem);
2856 hmap_remove(&br->iface_by_name, &iface->name_node);
2858 netdev_close(iface->netdev);
2865 static struct iface *
2866 iface_lookup(const struct bridge *br, const char *name)
2868 struct iface *iface;
2870 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2871 &br->iface_by_name) {
2872 if (!strcmp(iface->name, name)) {
2880 static struct iface *
2881 iface_find(const char *name)
2883 const struct bridge *br;
2885 HMAP_FOR_EACH (br, node, &all_bridges) {
2886 struct iface *iface = iface_lookup(br, name);
2895 static struct iface *
2896 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2898 struct iface *iface;
2900 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2901 hash_int(ofp_port, 0), &br->ifaces) {
2902 if (iface->ofp_port == ofp_port) {
2909 /* Set Ethernet address of 'iface', if one is specified in the configuration
2912 iface_set_mac(struct iface *iface)
2914 uint8_t ea[ETH_ADDR_LEN];
2916 if (!strcmp(iface->type, "internal")
2917 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2918 if (iface->ofp_port == OFPP_LOCAL) {
2919 VLOG_ERR("interface %s: ignoring mac in Interface record "
2920 "(use Bridge record to set local port's mac)",
2922 } else if (eth_addr_is_multicast(ea)) {
2923 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2926 int error = netdev_set_etheraddr(iface->netdev, ea);
2928 VLOG_ERR("interface %s: setting MAC failed (%s)",
2929 iface->name, strerror(error));
2935 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2937 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2939 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2940 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2944 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
2945 * sets the "ofport" field to -1.
2947 * This is appropriate when 'if_cfg''s interface cannot be created or is
2948 * otherwise invalid. */
2950 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
2952 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2953 iface_set_ofport(if_cfg, -1);
2954 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
2955 ovsrec_interface_set_admin_state(if_cfg, NULL);
2956 ovsrec_interface_set_duplex(if_cfg, NULL);
2957 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
2958 ovsrec_interface_set_link_state(if_cfg, NULL);
2959 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
2960 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
2961 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
2962 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
2963 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
2967 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2969 * The value strings in '*shash' are taken directly from values[], not copied,
2970 * so the caller should not modify or free them. */
2972 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2973 struct shash *shash)
2978 for (i = 0; i < n; i++) {
2979 shash_add(shash, keys[i], values[i]);
2983 /* Creates 'keys' and 'values' arrays from 'shash'.
2985 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2986 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2987 * are populated with with strings taken directly from 'shash' and thus have
2988 * the same ownership of the key-value pairs in shash.
2991 shash_to_ovs_idl_map(struct shash *shash,
2992 char ***keys, char ***values, size_t *n)
2996 struct shash_node *sn;
2998 count = shash_count(shash);
3000 k = xmalloc(count * sizeof *k);
3001 v = xmalloc(count * sizeof *v);
3004 SHASH_FOR_EACH(sn, shash) {
3015 struct iface_delete_queues_cbdata {
3016 struct netdev *netdev;
3017 const struct ovsdb_datum *queues;
3021 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3023 union ovsdb_atom atom;
3025 atom.integer = target;
3026 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3030 iface_delete_queues(unsigned int queue_id,
3031 const struct shash *details OVS_UNUSED, void *cbdata_)
3033 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3035 if (!queue_ids_include(cbdata->queues, queue_id)) {
3036 netdev_delete_queue(cbdata->netdev, queue_id);
3041 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3043 struct ofpbuf queues_buf;
3045 ofpbuf_init(&queues_buf, 0);
3047 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3048 netdev_set_qos(iface->netdev, NULL, NULL);
3050 struct iface_delete_queues_cbdata cbdata;
3051 struct shash details;
3055 /* Configure top-level Qos for 'iface'. */
3056 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3057 qos->n_other_config, &details);
3058 netdev_set_qos(iface->netdev, qos->type, &details);
3059 shash_destroy(&details);
3061 /* Deconfigure queues that were deleted. */
3062 cbdata.netdev = iface->netdev;
3063 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3065 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3067 /* Configure queues for 'iface'. */
3069 for (i = 0; i < qos->n_queues; i++) {
3070 const struct ovsrec_queue *queue = qos->value_queues[i];
3071 unsigned int queue_id = qos->key_queues[i];
3073 if (queue_id == 0) {
3077 if (queue->n_dscp == 1) {
3078 struct ofproto_port_queue *port_queue;
3080 port_queue = ofpbuf_put_uninit(&queues_buf,
3081 sizeof *port_queue);
3082 port_queue->queue = queue_id;
3083 port_queue->dscp = queue->dscp[0];
3086 shash_from_ovs_idl_map(queue->key_other_config,
3087 queue->value_other_config,
3088 queue->n_other_config, &details);
3089 netdev_set_queue(iface->netdev, queue_id, &details);
3090 shash_destroy(&details);
3093 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3094 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
3095 "queue (queue 0). Packets not directed to a "
3096 "correctly configured queue may be dropped.",
3101 if (iface->ofp_port >= 0) {
3102 const struct ofproto_port_queue *port_queues = queues_buf.data;
3103 size_t n_queues = queues_buf.size / sizeof *port_queues;
3105 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3106 port_queues, n_queues);
3109 netdev_set_policing(iface->netdev,
3110 iface->cfg->ingress_policing_rate,
3111 iface->cfg->ingress_policing_burst);
3113 ofpbuf_uninit(&queues_buf);
3117 iface_configure_cfm(struct iface *iface)
3119 const struct ovsrec_interface *cfg = iface->cfg;
3120 const char *extended_str, *opstate_str;
3121 struct cfm_settings s;
3123 if (!cfg->n_cfm_mpid) {
3124 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3128 s.mpid = *cfg->cfm_mpid;
3129 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3131 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3133 if (s.interval <= 0) {
3137 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3139 s.extended = !strcasecmp("true", extended_str);
3141 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3142 s.opup = !strcasecmp("up", opstate_str);
3144 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3147 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3148 * instead of obtaining it from the database. */
3150 iface_is_synthetic(const struct iface *iface)
3152 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3155 /* Port mirroring. */
3157 static struct mirror *
3158 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3162 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3163 if (uuid_equals(uuid, &m->uuid)) {
3171 bridge_configure_mirrors(struct bridge *br)
3173 const struct ovsdb_datum *mc;
3174 unsigned long *flood_vlans;
3175 struct mirror *m, *next;
3178 /* Get rid of deleted mirrors. */
3179 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3180 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3181 union ovsdb_atom atom;
3183 atom.uuid = m->uuid;
3184 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3189 /* Add new mirrors and reconfigure existing ones. */
3190 for (i = 0; i < br->cfg->n_mirrors; i++) {
3191 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3192 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3194 m = mirror_create(br, cfg);
3197 if (!mirror_configure(m)) {
3202 /* Update flooded vlans (for RSPAN). */
3203 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3204 br->cfg->n_flood_vlans);
3205 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3206 bitmap_free(flood_vlans);
3209 static struct mirror *
3210 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3214 m = xzalloc(sizeof *m);
3215 m->uuid = cfg->header_.uuid;
3216 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3218 m->name = xstrdup(cfg->name);
3224 mirror_destroy(struct mirror *m)
3227 struct bridge *br = m->bridge;
3230 ofproto_mirror_unregister(br->ofproto, m);
3233 hmap_remove(&br->mirrors, &m->hmap_node);
3240 mirror_collect_ports(struct mirror *m,
3241 struct ovsrec_port **in_ports, int n_in_ports,
3242 void ***out_portsp, size_t *n_out_portsp)
3244 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3245 size_t n_out_ports = 0;
3248 for (i = 0; i < n_in_ports; i++) {
3249 const char *name = in_ports[i]->name;
3250 struct port *port = port_lookup(m->bridge, name);
3252 out_ports[n_out_ports++] = port;
3254 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3255 "port %s", m->bridge->name, m->name, name);
3258 *out_portsp = out_ports;
3259 *n_out_portsp = n_out_ports;
3263 mirror_configure(struct mirror *m)
3265 const struct ovsrec_mirror *cfg = m->cfg;
3266 struct ofproto_mirror_settings s;
3269 if (strcmp(cfg->name, m->name)) {
3271 m->name = xstrdup(cfg->name);
3275 /* Get output port or VLAN. */
3276 if (cfg->output_port) {
3277 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3278 if (!s.out_bundle) {
3279 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3280 m->bridge->name, m->name);
3283 s.out_vlan = UINT16_MAX;
3285 if (cfg->output_vlan) {
3286 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3287 "output vlan; ignoring output vlan",
3288 m->bridge->name, m->name);
3290 } else if (cfg->output_vlan) {
3291 /* The database should prevent invalid VLAN values. */
3292 s.out_bundle = NULL;
3293 s.out_vlan = *cfg->output_vlan;
3295 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3296 m->bridge->name, m->name);
3300 /* Get port selection. */
3301 if (cfg->select_all) {
3302 size_t n_ports = hmap_count(&m->bridge->ports);
3303 void **ports = xmalloc(n_ports * sizeof *ports);
3308 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3318 /* Get ports, dropping ports that don't exist.
3319 * The IDL ensures that there are no duplicates. */
3320 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3321 &s.srcs, &s.n_srcs);
3322 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3323 &s.dsts, &s.n_dsts);
3326 /* Get VLAN selection. */
3327 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3330 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3333 if (s.srcs != s.dsts) {
3342 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3344 * This is deprecated. It is only for compatibility with broken device drivers
3345 * in old versions of Linux that do not properly support VLANs when VLAN
3346 * devices are not used. When broken device drivers are no longer in
3347 * widespread use, we will delete these interfaces. */
3349 static void **blocks;
3350 static size_t n_blocks, allocated_blocks;
3352 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3353 * VLAN splinters are reconfigured. */
3355 register_block(void *block)
3357 if (n_blocks >= allocated_blocks) {
3358 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3360 blocks[n_blocks++] = block;
3363 /* Frees all of the blocks registered with register_block(). */
3365 free_registered_blocks(void)
3369 for (i = 0; i < n_blocks; i++) {
3375 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3378 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3382 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3383 return !strcmp(value, "true");
3386 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3389 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3390 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3391 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3394 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3395 * use, returns NULL.
3397 * Updates 'vlan_splinters_enabled_anywhere'. */
3398 static unsigned long int *
3399 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3401 unsigned long int *splinter_vlans;
3402 struct sset splinter_ifaces;
3403 const char *real_dev_name;
3404 struct shash *real_devs;
3405 struct shash_node *node;
3409 /* Free space allocated for synthesized ports and interfaces, since we're
3410 * in the process of reconstructing all of them. */
3411 free_registered_blocks();
3413 splinter_vlans = bitmap_allocate(4096);
3414 sset_init(&splinter_ifaces);
3415 vlan_splinters_enabled_anywhere = false;
3416 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3417 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3420 for (j = 0; j < br_cfg->n_ports; j++) {
3421 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3424 for (k = 0; k < port_cfg->n_interfaces; k++) {
3425 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3427 if (vlan_splinters_is_enabled(iface_cfg)) {
3428 vlan_splinters_enabled_anywhere = true;
3429 sset_add(&splinter_ifaces, iface_cfg->name);
3430 vlan_bitmap_from_array__(port_cfg->trunks,
3436 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3437 bitmap_set1(splinter_vlans, *port_cfg->tag);
3442 if (!vlan_splinters_enabled_anywhere) {
3443 free(splinter_vlans);
3444 sset_destroy(&splinter_ifaces);
3448 HMAP_FOR_EACH (br, node, &all_bridges) {
3450 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3454 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3455 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3456 * device to be created for it. */
3457 bitmap_set0(splinter_vlans, 0);
3458 bitmap_set0(splinter_vlans, 4095);
3460 /* Delete all VLAN devices that we don't need. */
3462 real_devs = vlandev_get_real_devs();
3463 SHASH_FOR_EACH (node, real_devs) {
3464 const struct vlan_real_dev *real_dev = node->data;
3465 const struct vlan_dev *vlan_dev;
3466 bool real_dev_has_splinters;
3468 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3470 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3471 if (!real_dev_has_splinters
3472 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3473 struct netdev *netdev;
3475 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3476 if (!netdev_get_in4(netdev, NULL, NULL) ||
3477 !netdev_get_in6(netdev, NULL)) {
3478 vlandev_del(vlan_dev->name);
3480 /* It has an IP address configured, so we don't own
3481 * it. Don't delete it. */
3483 netdev_close(netdev);
3490 /* Add all VLAN devices that we need. */
3491 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3494 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3495 if (!vlandev_get_name(real_dev_name, vid)) {
3496 vlandev_add(real_dev_name, vid);
3503 sset_destroy(&splinter_ifaces);
3505 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3506 free(splinter_vlans);
3509 return splinter_vlans;
3512 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3515 configure_splinter_port(struct port *port)
3517 struct ofproto *ofproto = port->bridge->ofproto;
3518 uint16_t realdev_ofp_port;
3519 const char *realdev_name;
3520 struct iface *vlandev, *realdev;
3522 ofproto_bundle_unregister(port->bridge->ofproto, port);
3524 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3527 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3528 realdev = iface_lookup(port->bridge, realdev_name);
3529 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3531 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3535 static struct ovsrec_port *
3536 synthesize_splinter_port(const char *real_dev_name,
3537 const char *vlan_dev_name, int vid)
3539 struct ovsrec_interface *iface;
3540 struct ovsrec_port *port;
3542 iface = xzalloc(sizeof *iface);
3543 iface->name = xstrdup(vlan_dev_name);
3544 iface->type = "system";
3546 port = xzalloc(sizeof *port);
3547 port->interfaces = xmemdup(&iface, sizeof iface);
3548 port->n_interfaces = 1;
3549 port->name = xstrdup(vlan_dev_name);
3550 port->vlan_mode = "splinter";
3551 port->tag = xmalloc(sizeof *port->tag);
3553 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3554 port->key_other_config[0] = "realdev";
3555 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3556 port->value_other_config[0] = xstrdup(real_dev_name);
3557 port->n_other_config = 1;
3559 register_block(iface);
3560 register_block(iface->name);
3561 register_block(port);
3562 register_block(port->interfaces);
3563 register_block(port->name);
3564 register_block(port->tag);
3565 register_block(port->key_other_config);
3566 register_block(port->value_other_config);
3567 register_block(port->value_other_config[0]);
3572 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3573 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3574 * 1-bit in the 'splinter_vlans' bitmap. */
3576 add_vlan_splinter_ports(struct bridge *br,
3577 const unsigned long int *splinter_vlans,
3578 struct shash *ports)
3582 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3583 * we're modifying 'ports'. */
3584 for (i = 0; i < br->cfg->n_ports; i++) {
3585 const char *name = br->cfg->ports[i]->name;
3586 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3589 for (j = 0; j < port_cfg->n_interfaces; j++) {
3590 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3592 if (vlan_splinters_is_enabled(iface_cfg)) {
3593 const char *real_dev_name;
3596 real_dev_name = iface_cfg->name;
3597 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3598 const char *vlan_dev_name;
3600 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3602 && !shash_find(ports, vlan_dev_name)) {
3603 shash_add(ports, vlan_dev_name,
3604 synthesize_splinter_port(
3605 real_dev_name, vlan_dev_name, vid));
3614 mirror_refresh_stats(struct mirror *m)
3616 struct ofproto *ofproto = m->bridge->ofproto;
3617 uint64_t tx_packets, tx_bytes;
3620 size_t stat_cnt = 0;
3622 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3623 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3627 if (tx_packets != UINT64_MAX) {
3628 keys[stat_cnt] = "tx_packets";
3629 values[stat_cnt] = tx_packets;
3632 if (tx_bytes != UINT64_MAX) {
3633 keys[stat_cnt] = "tx_bytes";
3634 values[stat_cnt] = tx_bytes;
3638 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);