1 /* Copyright (c) 2008, 2009, 2010, 2011 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"
35 #include "ofp-print.h"
37 #include "ofproto/ofproto.h"
38 #include "poll-loop.h"
41 #include "socket-util.h"
42 #include "stream-ssl.h"
44 #include "system-stats.h"
49 #include "vswitchd/vswitch-idl.h"
50 #include "xenserver.h"
52 #include "sflow_api.h"
53 #include "vlan-bitmap.h"
55 VLOG_DEFINE_THIS_MODULE(bridge);
57 COVERAGE_DEFINE(bridge_reconfigure);
60 /* These members are always valid. */
61 struct list port_elem; /* Element in struct port's "ifaces" list. */
62 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
63 struct port *port; /* Containing port. */
64 char *name; /* Host network device name. */
65 tag_type tag; /* Tag associated with this interface. */
67 /* These members are valid only after bridge_reconfigure() causes them to
69 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
70 int ofp_port; /* OpenFlow port number, -1 if unknown. */
71 struct netdev *netdev; /* Network device. */
72 const char *type; /* Usually same as cfg->type. */
73 const struct ovsrec_interface *cfg;
77 struct uuid uuid; /* UUID of this "mirror" record in database. */
78 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
79 struct bridge *bridge;
84 struct bridge *bridge;
85 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
88 const struct ovsrec_port *cfg;
90 /* An ordinary bridge port has 1 interface.
91 * A bridge port for bonding has at least 2 interfaces. */
92 struct list ifaces; /* List of "struct iface"s. */
96 struct hmap_node node; /* In 'all_bridges'. */
97 char *name; /* User-specified arbitrary name. */
98 char *type; /* Datapath type. */
99 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
100 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
101 const struct ovsrec_bridge *cfg;
103 /* OpenFlow switch processing. */
104 struct ofproto *ofproto; /* OpenFlow switch. */
107 struct hmap ports; /* "struct port"s indexed by name. */
108 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
109 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
111 /* Port mirroring. */
112 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
114 /* Synthetic local port if necessary. */
115 struct ovsrec_port synth_local_port;
116 struct ovsrec_interface synth_local_iface;
117 struct ovsrec_interface *synth_local_ifacep;
120 /* All bridges, indexed by name. */
121 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
123 /* OVSDB IDL used to obtain configuration. */
124 static struct ovsdb_idl *idl;
126 /* Each time this timer expires, the bridge fetches systems and interface
127 * statistics and pushes them into the database. */
128 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
129 static long long int stats_timer = LLONG_MIN;
131 /* Stores the time after which rate limited statistics may be written to the
132 * database. Only updated when changes to the database require rate limiting.
134 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
135 static long long int db_limiter = LLONG_MIN;
137 static void add_del_bridges(const struct ovsrec_open_vswitch *);
138 static void bridge_del_ofprotos(void);
139 static bool bridge_add_ofprotos(struct bridge *);
140 static void bridge_create(const struct ovsrec_bridge *);
141 static void bridge_destroy(struct bridge *);
142 static struct bridge *bridge_lookup(const char *name);
143 static unixctl_cb_func bridge_unixctl_dump_flows;
144 static unixctl_cb_func bridge_unixctl_reconnect;
145 static size_t bridge_get_controllers(const struct bridge *br,
146 struct ovsrec_controller ***controllersp);
147 static void bridge_add_del_ports(struct bridge *,
148 const unsigned long int *splinter_vlans);
149 static void bridge_add_ofproto_ports(struct bridge *);
150 static void bridge_del_ofproto_ports(struct bridge *);
151 static void bridge_refresh_ofp_port(struct bridge *);
152 static void bridge_configure_datapath_id(struct bridge *);
153 static void bridge_configure_flow_eviction_threshold(struct bridge *);
154 static void bridge_configure_netflow(struct bridge *);
155 static void bridge_configure_forward_bpdu(struct bridge *);
156 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
157 static void bridge_configure_stp(struct bridge *);
158 static void bridge_configure_remotes(struct bridge *,
159 const struct sockaddr_in *managers,
161 static void bridge_pick_local_hw_addr(struct bridge *,
162 uint8_t ea[ETH_ADDR_LEN],
163 struct iface **hw_addr_iface);
164 static uint64_t bridge_pick_datapath_id(struct bridge *,
165 const uint8_t bridge_ea[ETH_ADDR_LEN],
166 struct iface *hw_addr_iface);
167 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
169 static const char *get_port_other_config(const struct ovsrec_port *,
171 const char *default_value);
172 static uint64_t dpid_from_hash(const void *, size_t nbytes);
173 static bool bridge_has_bond_fake_iface(const struct bridge *,
175 static bool port_is_bond_fake_iface(const struct port *);
177 static unixctl_cb_func qos_unixctl_show;
179 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
180 static void port_add_ifaces(struct port *);
181 static void port_del_ifaces(struct port *);
182 static void port_destroy(struct port *);
183 static struct port *port_lookup(const struct bridge *, const char *name);
184 static void port_configure(struct port *);
185 static struct lacp_settings *port_configure_lacp(struct port *,
186 struct lacp_settings *);
187 static void port_configure_bond(struct port *, struct bond_settings *,
188 uint32_t *bond_stable_ids);
189 static bool port_is_synthetic(const struct port *);
191 static void bridge_configure_mirrors(struct bridge *);
192 static struct mirror *mirror_create(struct bridge *,
193 const struct ovsrec_mirror *);
194 static void mirror_destroy(struct mirror *);
195 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
197 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
198 static struct iface *iface_create(struct port *port,
199 const struct ovsrec_interface *if_cfg);
200 static void iface_destroy(struct iface *);
201 static struct iface *iface_lookup(const struct bridge *, const char *name);
202 static struct iface *iface_find(const char *name);
203 static struct iface *iface_from_ofp_port(const struct bridge *,
205 static void iface_set_mac(struct iface *);
206 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
207 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
208 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
209 static void iface_configure_cfm(struct iface *);
210 static void iface_refresh_cfm_stats(struct iface *);
211 static void iface_refresh_stats(struct iface *);
212 static void iface_refresh_status(struct iface *);
213 static bool iface_is_synthetic(const struct iface *);
214 static const char *get_interface_other_config(const struct ovsrec_interface *,
216 const char *default_value);
218 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
220 static void shash_to_ovs_idl_map(struct shash *,
221 char ***keys, char ***values, size_t *n);
223 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
225 * This is deprecated. It is only for compatibility with broken device drivers
226 * in old versions of Linux that do not properly support VLANs when VLAN
227 * devices are not used. When broken device drivers are no longer in
228 * widespread use, we will delete these interfaces. */
230 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
231 static bool vlan_splinters_enabled_anywhere;
233 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
234 static unsigned long int *collect_splinter_vlans(
235 const struct ovsrec_open_vswitch *);
236 static void configure_splinter_port(struct port *);
237 static void add_vlan_splinter_ports(struct bridge *,
238 const unsigned long int *splinter_vlans,
239 struct shash *ports);
241 /* Public functions. */
243 /* Initializes the bridge module, configuring it to obtain its configuration
244 * from an OVSDB server accessed over 'remote', which should be a string in a
245 * form acceptable to ovsdb_idl_create(). */
247 bridge_init(const char *remote)
249 /* Create connection to database. */
250 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
251 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
253 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
254 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
255 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
257 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
258 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
259 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
261 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
262 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
263 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
265 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
266 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
267 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
268 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
270 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
271 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
272 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
273 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
274 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
275 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
276 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
277 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
282 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
284 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
285 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
286 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
287 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
289 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
291 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
293 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
297 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
299 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
300 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
301 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
302 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
303 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
305 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
307 /* Register unixctl commands. */
308 unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL);
309 unixctl_command_register("bridge/dump-flows", "bridge",
310 bridge_unixctl_dump_flows, NULL);
311 unixctl_command_register("bridge/reconnect", "[bridge]",
312 bridge_unixctl_reconnect, NULL);
321 struct bridge *br, *next_br;
323 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
326 ovsdb_idl_destroy(idl);
329 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
330 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
331 * responsible for freeing '*managersp' (with free()).
333 * You may be asking yourself "why does ovs-vswitchd care?", because
334 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
335 * should not be and in fact is not directly involved in that. But
336 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
337 * it has to tell in-band control where the managers are to enable that.
338 * (Thus, only managers connected in-band are collected.)
341 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
342 struct sockaddr_in **managersp, size_t *n_managersp)
344 struct sockaddr_in *managers = NULL;
345 size_t n_managers = 0;
349 /* Collect all of the potential targets from the "targets" columns of the
350 * rows pointed to by "manager_options", excluding any that are
353 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
354 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
356 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
357 sset_find_and_delete(&targets, m->target);
359 sset_add(&targets, m->target);
363 /* Now extract the targets' IP addresses. */
364 if (!sset_is_empty(&targets)) {
367 managers = xmalloc(sset_count(&targets) * sizeof *managers);
368 SSET_FOR_EACH (target, &targets) {
369 struct sockaddr_in *sin = &managers[n_managers];
371 if ((!strncmp(target, "tcp:", 4)
372 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
373 (!strncmp(target, "ssl:", 4)
374 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
379 sset_destroy(&targets);
381 *managersp = managers;
382 *n_managersp = n_managers;
386 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
388 unsigned long int *splinter_vlans;
389 struct sockaddr_in *managers;
390 struct bridge *br, *next;
391 int sflow_bridge_number;
394 COVERAGE_INC(bridge_reconfigure);
396 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
397 * iface"s according to 'ovs_cfg', with only very minimal configuration
400 * This is mostly an update to bridge data structures. Very little is
401 * pushed down to ofproto or lower layers. */
402 add_del_bridges(ovs_cfg);
403 splinter_vlans = collect_splinter_vlans(ovs_cfg);
404 HMAP_FOR_EACH (br, node, &all_bridges) {
405 bridge_add_del_ports(br, splinter_vlans);
407 free(splinter_vlans);
409 /* Delete all datapaths and datapath ports that are no longer configured.
411 * The kernel will reject any attempt to add a given port to a datapath if
412 * that port already belongs to a different datapath, so we must do all
413 * port deletions before any port additions. A datapath always has a
414 * "local port" so we must delete not-configured datapaths too. */
415 bridge_del_ofprotos();
416 HMAP_FOR_EACH (br, node, &all_bridges) {
418 bridge_del_ofproto_ports(br);
422 /* Create datapaths and datapath ports that are missing.
424 * After this is done, we have our final set of bridges, ports, and
425 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
426 * has at least one iface, every "struct iface" has a valid ofp_port and
428 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
429 if (!br->ofproto && !bridge_add_ofprotos(br)) {
433 HMAP_FOR_EACH (br, node, &all_bridges) {
434 bridge_refresh_ofp_port(br);
435 bridge_add_ofproto_ports(br);
438 /* Complete the configuration. */
439 sflow_bridge_number = 0;
440 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
441 HMAP_FOR_EACH (br, node, &all_bridges) {
444 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
447 port_configure(port);
449 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
450 iface_configure_cfm(iface);
451 iface_configure_qos(iface, port->cfg->qos);
452 iface_set_mac(iface);
455 bridge_configure_mirrors(br);
456 bridge_configure_datapath_id(br);
457 bridge_configure_flow_eviction_threshold(br);
458 bridge_configure_forward_bpdu(br);
459 bridge_configure_remotes(br, managers, n_managers);
460 bridge_configure_netflow(br);
461 bridge_configure_sflow(br, &sflow_bridge_number);
462 bridge_configure_stp(br);
466 /* ovs-vswitchd has completed initialization, so allow the process that
467 * forked us to exit successfully. */
468 daemonize_complete();
471 /* Iterate over all ofprotos and delete any of them that do not have a
472 * configured bridge or that are the wrong type. */
474 bridge_del_ofprotos(void)
482 ofproto_enumerate_types(&types);
483 SSET_FOR_EACH (type, &types) {
486 ofproto_enumerate_names(type, &names);
487 SSET_FOR_EACH (name, &names) {
488 struct bridge *br = bridge_lookup(name);
489 if (!br || strcmp(type, br->type)) {
490 ofproto_delete(name, type);
494 sset_destroy(&names);
495 sset_destroy(&types);
499 bridge_add_ofprotos(struct bridge *br)
501 int error = ofproto_create(br->name, br->type, &br->ofproto);
503 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
510 port_configure(struct port *port)
512 const struct ovsrec_port *cfg = port->cfg;
513 struct bond_settings bond_settings;
514 struct lacp_settings lacp_settings;
515 struct ofproto_bundle_settings s;
518 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
519 configure_splinter_port(port);
528 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
529 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
530 s.slaves[s.n_slaves++] = iface->ofp_port;
536 if (list_is_short(&port->ifaces)) {
537 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
541 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
542 * they even work as-is. But they have not been tested. */
543 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
548 /* Get VLAN trunks. */
551 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
555 if (cfg->vlan_mode) {
556 if (!strcmp(cfg->vlan_mode, "access")) {
557 s.vlan_mode = PORT_VLAN_ACCESS;
558 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
559 s.vlan_mode = PORT_VLAN_TRUNK;
560 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
561 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
562 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
563 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
565 /* This "can't happen" because ovsdb-server should prevent it. */
566 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
567 s.vlan_mode = PORT_VLAN_TRUNK;
571 s.vlan_mode = PORT_VLAN_ACCESS;
573 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
577 s.vlan_mode = PORT_VLAN_TRUNK;
580 s.use_priority_tags = !strcmp("true", get_port_other_config(
581 cfg, "priority-tags", ""));
583 /* Get LACP settings. */
584 s.lacp = port_configure_lacp(port, &lacp_settings);
588 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
589 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
590 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
593 s.lacp_slaves = NULL;
596 /* Get bond settings. */
597 if (s.n_slaves > 1) {
598 s.bond = &bond_settings;
599 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
600 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
603 s.bond_stable_ids = NULL;
605 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
606 netdev_set_miimon_interval(iface->netdev, 0);
611 ofproto_bundle_register(port->bridge->ofproto, port, &s);
617 free(s.bond_stable_ids);
620 /* Pick local port hardware address and datapath ID for 'br'. */
622 bridge_configure_datapath_id(struct bridge *br)
624 uint8_t ea[ETH_ADDR_LEN];
626 struct iface *local_iface;
627 struct iface *hw_addr_iface;
630 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
631 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
633 int error = netdev_set_etheraddr(local_iface->netdev, ea);
635 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
636 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
637 "Ethernet address: %s",
638 br->name, strerror(error));
641 memcpy(br->ea, ea, ETH_ADDR_LEN);
643 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
644 ofproto_set_datapath_id(br->ofproto, dpid);
646 dpid_string = xasprintf("%016"PRIx64, dpid);
647 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
651 /* Set NetFlow configuration on 'br'. */
653 bridge_configure_netflow(struct bridge *br)
655 struct ovsrec_netflow *cfg = br->cfg->netflow;
656 struct netflow_options opts;
659 ofproto_set_netflow(br->ofproto, NULL);
663 memset(&opts, 0, sizeof opts);
665 /* Get default NetFlow configuration from datapath.
666 * Apply overrides from 'cfg'. */
667 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
668 if (cfg->engine_type) {
669 opts.engine_type = *cfg->engine_type;
671 if (cfg->engine_id) {
672 opts.engine_id = *cfg->engine_id;
675 /* Configure active timeout interval. */
676 opts.active_timeout = cfg->active_timeout;
677 if (!opts.active_timeout) {
678 opts.active_timeout = -1;
679 } else if (opts.active_timeout < 0) {
680 VLOG_WARN("bridge %s: active timeout interval set to negative "
681 "value, using default instead (%d seconds)", br->name,
682 NF_ACTIVE_TIMEOUT_DEFAULT);
683 opts.active_timeout = -1;
686 /* Add engine ID to interface number to disambiguate bridgs? */
687 opts.add_id_to_iface = cfg->add_id_to_interface;
688 if (opts.add_id_to_iface) {
689 if (opts.engine_id > 0x7f) {
690 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
691 "another vswitch, choose an engine id less than 128",
694 if (hmap_count(&br->ports) > 508) {
695 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
696 "another port when more than 508 ports are used",
702 sset_init(&opts.collectors);
703 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
706 if (ofproto_set_netflow(br->ofproto, &opts)) {
707 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
709 sset_destroy(&opts.collectors);
712 /* Set sFlow configuration on 'br'. */
714 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
716 const struct ovsrec_sflow *cfg = br->cfg->sflow;
717 struct ovsrec_controller **controllers;
718 struct ofproto_sflow_options oso;
719 size_t n_controllers;
723 ofproto_set_sflow(br->ofproto, NULL);
727 memset(&oso, 0, sizeof oso);
729 sset_init(&oso.targets);
730 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
732 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
734 oso.sampling_rate = *cfg->sampling;
737 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
739 oso.polling_interval = *cfg->polling;
742 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
744 oso.header_len = *cfg->header;
747 oso.sub_id = (*sflow_bridge_number)++;
748 oso.agent_device = cfg->agent;
750 oso.control_ip = NULL;
751 n_controllers = bridge_get_controllers(br, &controllers);
752 for (i = 0; i < n_controllers; i++) {
753 if (controllers[i]->local_ip) {
754 oso.control_ip = controllers[i]->local_ip;
758 ofproto_set_sflow(br->ofproto, &oso);
760 sset_destroy(&oso.targets);
764 port_configure_stp(const struct ofproto *ofproto, struct port *port,
765 struct ofproto_port_stp_settings *port_s,
766 int *port_num_counter, unsigned long *port_num_bitmap)
768 const char *config_str;
771 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
772 if (config_str && !strcmp(config_str, "false")) {
773 port_s->enable = false;
776 port_s->enable = true;
779 /* STP over bonds is not supported. */
780 if (!list_is_singleton(&port->ifaces)) {
781 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
783 port_s->enable = false;
787 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
789 /* Internal ports shouldn't participate in spanning tree, so
791 if (!strcmp(iface->type, "internal")) {
792 VLOG_DBG("port %s: disable STP on internal ports", port->name);
793 port_s->enable = false;
797 /* STP on mirror output ports is not supported. */
798 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
799 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
800 port_s->enable = false;
804 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
806 unsigned long int port_num = strtoul(config_str, NULL, 0);
807 int port_idx = port_num - 1;
809 if (port_num < 1 || port_num > STP_MAX_PORTS) {
810 VLOG_ERR("port %s: invalid stp-port-num", port->name);
811 port_s->enable = false;
815 if (bitmap_is_set(port_num_bitmap, port_idx)) {
816 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
817 port->name, port_num);
818 port_s->enable = false;
821 bitmap_set1(port_num_bitmap, port_idx);
822 port_s->port_num = port_idx;
824 if (*port_num_counter > STP_MAX_PORTS) {
825 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
826 port_s->enable = false;
830 port_s->port_num = (*port_num_counter)++;
833 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
835 port_s->path_cost = strtoul(config_str, NULL, 10);
839 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
840 /* Couldn't get speed, so assume 100Mb/s. */
841 port_s->path_cost = 19;
845 mbps = netdev_features_to_bps(current) / 1000000;
846 port_s->path_cost = stp_convert_speed_to_cost(mbps);
850 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
852 port_s->priority = strtoul(config_str, NULL, 0);
854 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
858 /* Set spanning tree configuration on 'br'. */
860 bridge_configure_stp(struct bridge *br)
862 if (!br->cfg->stp_enable) {
863 ofproto_set_stp(br->ofproto, NULL);
865 struct ofproto_stp_settings br_s;
866 const char *config_str;
868 int port_num_counter;
869 unsigned long *port_num_bitmap;
871 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
873 uint8_t ea[ETH_ADDR_LEN];
875 if (eth_addr_from_string(config_str, ea)) {
876 br_s.system_id = eth_addr_to_uint64(ea);
878 br_s.system_id = eth_addr_to_uint64(br->ea);
879 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
880 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
883 br_s.system_id = eth_addr_to_uint64(br->ea);
886 config_str = bridge_get_other_config(br->cfg, "stp-priority");
888 br_s.priority = strtoul(config_str, NULL, 0);
890 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
893 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
895 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
897 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
900 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
902 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
904 br_s.max_age = STP_DEFAULT_MAX_AGE;
907 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
909 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
911 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
914 /* Configure STP on the bridge. */
915 if (ofproto_set_stp(br->ofproto, &br_s)) {
916 VLOG_ERR("bridge %s: could not enable STP", br->name);
920 /* Users must either set the port number with the "stp-port-num"
921 * configuration on all ports or none. If manual configuration
922 * is not done, then we allocate them sequentially. */
923 port_num_counter = 0;
924 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
925 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
926 struct ofproto_port_stp_settings port_s;
929 port_configure_stp(br->ofproto, port, &port_s,
930 &port_num_counter, port_num_bitmap);
932 /* As bonds are not supported, just apply configuration to
934 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
935 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
937 VLOG_ERR("port %s: could not enable STP", port->name);
943 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
944 && port_num_counter) {
945 VLOG_ERR("bridge %s: must manually configure all STP port "
946 "IDs or none, disabling", br->name);
947 ofproto_set_stp(br->ofproto, NULL);
949 bitmap_free(port_num_bitmap);
954 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
956 const struct port *port = port_lookup(br, name);
957 return port && port_is_bond_fake_iface(port);
961 port_is_bond_fake_iface(const struct port *port)
963 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
967 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
969 struct bridge *br, *next;
973 /* Collect new bridges' names and types. */
975 for (i = 0; i < cfg->n_bridges; i++) {
976 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
977 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
979 if (strchr(br_cfg->name, '/')) {
980 /* Prevent remote ovsdb-server users from accessing arbitrary
981 * directories, e.g. consider a bridge named "../../../etc/". */
982 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
984 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
985 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
989 /* Get rid of deleted bridges or those whose types have changed.
990 * Update 'cfg' of bridges that still exist. */
991 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
992 br->cfg = shash_find_data(&new_br, br->name);
993 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
994 br->cfg->datapath_type))) {
999 /* Add new bridges. */
1000 for (i = 0; i < cfg->n_bridges; i++) {
1001 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1002 struct bridge *br = bridge_lookup(br_cfg->name);
1004 bridge_create(br_cfg);
1008 shash_destroy(&new_br);
1011 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1014 * The kernel will reject any attempt to add a given port to a datapath if that
1015 * port already belongs to a different datapath, so we must do all port
1016 * deletions before any port additions. */
1018 bridge_del_ofproto_ports(struct bridge *br)
1020 struct ofproto_port_dump dump;
1021 struct ofproto_port ofproto_port;
1023 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1024 const char *name = ofproto_port.name;
1025 struct iface *iface;
1029 /* Ignore the local port. We can't change it anyhow. */
1030 if (!strcmp(name, br->name)) {
1034 /* Get the type that 'ofproto_port' should have (ordinarily the
1035 * type of its corresponding iface) or NULL if it should be
1037 iface = iface_lookup(br, name);
1038 type = (iface ? iface->type
1039 : bridge_has_bond_fake_iface(br, name) ? "internal"
1042 /* If it's the wrong type then delete the ofproto port. */
1044 && !strcmp(ofproto_port.type, type)
1045 && (!iface || !iface->netdev
1046 || !strcmp(netdev_get_type(iface->netdev), type))) {
1049 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1051 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1052 br->name, name, strerror(error));
1055 netdev_close(iface->netdev);
1056 iface->netdev = NULL;
1062 iface_set_ofp_port(struct iface *iface, int ofp_port)
1064 struct bridge *br = iface->port->bridge;
1066 assert(iface->ofp_port < 0 && ofp_port >= 0);
1067 iface->ofp_port = ofp_port;
1068 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1069 iface_set_ofport(iface->cfg, ofp_port);
1073 bridge_refresh_ofp_port(struct bridge *br)
1075 struct ofproto_port_dump dump;
1076 struct ofproto_port ofproto_port;
1079 /* Clear all the "ofp_port"es. */
1080 hmap_clear(&br->ifaces);
1081 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1082 struct iface *iface;
1084 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1085 iface->ofp_port = -1;
1089 /* Obtain the correct "ofp_port"s from ofproto. */
1090 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1091 struct iface *iface = iface_lookup(br, ofproto_port.name);
1093 if (iface->ofp_port >= 0) {
1094 VLOG_WARN("bridge %s: interface %s reported twice",
1095 br->name, ofproto_port.name);
1096 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1097 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1098 br->name, ofproto_port.ofp_port);
1100 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1106 /* Add an ofproto port for any "struct iface" that doesn't have one.
1107 * Delete any "struct iface" for which this fails.
1108 * Delete any "struct port" that thereby ends up with no ifaces. */
1110 bridge_add_ofproto_ports(struct bridge *br)
1112 struct port *port, *next_port;
1114 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1115 struct iface *iface, *next_iface;
1116 struct ofproto_port ofproto_port;
1118 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1121 /* Open the netdev. */
1122 if (!iface->netdev) {
1123 error = netdev_open(iface->name, iface->type, &iface->netdev);
1125 VLOG_WARN("could not open network device %s (%s)",
1126 iface->name, strerror(error));
1130 && port->cfg->vlan_mode
1131 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1132 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1138 /* Configure the netdev. */
1139 if (iface->netdev) {
1143 shash_from_ovs_idl_map(iface->cfg->key_options,
1144 iface->cfg->value_options,
1145 iface->cfg->n_options, &args);
1146 error = netdev_set_config(iface->netdev, &args);
1147 shash_destroy(&args);
1150 VLOG_WARN("could not configure network device %s (%s)",
1151 iface->name, strerror(error));
1152 netdev_close(iface->netdev);
1153 iface->netdev = NULL;
1157 /* Add the port, if necessary. */
1158 if (iface->netdev && iface->ofp_port < 0) {
1162 error = ofproto_port_add(br->ofproto, iface->netdev,
1165 iface_set_ofp_port(iface, ofp_port);
1167 netdev_close(iface->netdev);
1168 iface->netdev = NULL;
1172 /* Populate stats columns in new Interface rows. */
1173 if (iface->netdev && !iface->cfg->mtu) {
1174 iface_refresh_stats(iface);
1175 iface_refresh_status(iface);
1178 /* Delete the iface if we failed. */
1179 if (iface->netdev && iface->ofp_port >= 0) {
1180 VLOG_DBG("bridge %s: interface %s is on port %d",
1181 br->name, iface->name, iface->ofp_port);
1183 if (iface->netdev) {
1184 VLOG_ERR("bridge %s: missing %s interface, dropping",
1185 br->name, iface->name);
1187 /* We already reported a related error, don't bother
1188 * duplicating it. */
1190 iface_clear_db_record(iface->cfg);
1191 iface_destroy(iface);
1194 if (list_is_empty(&port->ifaces)) {
1195 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1200 /* Add bond fake iface if necessary. */
1201 if (port_is_bond_fake_iface(port)) {
1202 if (ofproto_port_query_by_name(br->ofproto, port->name,
1204 struct netdev *netdev;
1207 error = netdev_open(port->name, "internal", &netdev);
1209 ofproto_port_add(br->ofproto, netdev, NULL);
1210 netdev_close(netdev);
1212 VLOG_WARN("could not open network device %s (%s)",
1213 port->name, strerror(error));
1216 /* Already exists, nothing to do. */
1217 ofproto_port_destroy(&ofproto_port);
1224 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1228 for (i = 0; i < n; i++) {
1229 if (!strcmp(keys[i], key)) {
1237 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1239 return get_ovsrec_key_value(br_cfg->key_other_config,
1240 br_cfg->value_other_config,
1241 br_cfg->n_other_config, key);
1244 /* Set Flow eviction threshold */
1246 bridge_configure_flow_eviction_threshold(struct bridge *br)
1248 const char *threshold_str;
1251 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1252 if (threshold_str) {
1253 threshold = strtoul(threshold_str, NULL, 10);
1255 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1257 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1260 /* Set forward BPDU option. */
1262 bridge_configure_forward_bpdu(struct bridge *br)
1264 const char *forward_bpdu_str;
1265 bool forward_bpdu = false;
1267 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1268 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1269 forward_bpdu = true;
1271 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1275 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1276 struct iface **hw_addr_iface)
1280 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 /* Otherwise choose the minimum non-local MAC address among all of the
1300 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1301 uint8_t iface_ea[ETH_ADDR_LEN];
1302 struct iface *candidate;
1303 struct iface *iface;
1305 /* Mirror output ports don't participate. */
1306 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1310 /* Choose the MAC address to represent the port. */
1312 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1313 /* Find the interface with this Ethernet address (if any) so that
1314 * we can provide the correct devname to the caller. */
1315 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1316 uint8_t candidate_ea[ETH_ADDR_LEN];
1317 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1318 && eth_addr_equals(iface_ea, candidate_ea)) {
1323 /* Choose the interface whose MAC address will represent the port.
1324 * The Linux kernel bonding code always chooses the MAC address of
1325 * the first slave added to a bond, and the Fedora networking
1326 * scripts always add slaves to a bond in alphabetical order, so
1327 * for compatibility we choose the interface with the name that is
1328 * first in alphabetical order. */
1329 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1330 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1335 /* The local port doesn't count (since we're trying to choose its
1336 * MAC address anyway). */
1337 if (iface->ofp_port == OFPP_LOCAL) {
1342 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1348 /* Compare against our current choice. */
1349 if (!eth_addr_is_multicast(iface_ea) &&
1350 !eth_addr_is_local(iface_ea) &&
1351 !eth_addr_is_reserved(iface_ea) &&
1352 !eth_addr_is_zero(iface_ea) &&
1353 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1355 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1356 *hw_addr_iface = iface;
1361 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1362 br->name, ETH_ADDR_ARGS(ea));
1364 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1365 *hw_addr_iface = NULL;
1366 VLOG_WARN("bridge %s: using default bridge Ethernet "
1367 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1371 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1372 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1373 * an interface on 'br', then that interface must be passed in as
1374 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1375 * 'hw_addr_iface' must be passed in as a null pointer. */
1377 bridge_pick_datapath_id(struct bridge *br,
1378 const uint8_t bridge_ea[ETH_ADDR_LEN],
1379 struct iface *hw_addr_iface)
1382 * The procedure for choosing a bridge MAC address will, in the most
1383 * ordinary case, also choose a unique MAC that we can use as a datapath
1384 * ID. In some special cases, though, multiple bridges will end up with
1385 * the same MAC address. This is OK for the bridges, but it will confuse
1386 * the OpenFlow controller, because each datapath needs a unique datapath
1389 * Datapath IDs must be unique. It is also very desirable that they be
1390 * stable from one run to the next, so that policy set on a datapath
1393 const char *datapath_id;
1396 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1397 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1401 if (!hw_addr_iface) {
1403 * A purely internal bridge, that is, one that has no non-virtual
1404 * network devices on it at all, is difficult because it has no
1405 * natural unique identifier at all.
1407 * When the host is a XenServer, we handle this case by hashing the
1408 * host's UUID with the name of the bridge. Names of bridges are
1409 * persistent across XenServer reboots, although they can be reused if
1410 * an internal network is destroyed and then a new one is later
1411 * created, so this is fairly effective.
1413 * When the host is not a XenServer, we punt by using a random MAC
1414 * address on each run.
1416 const char *host_uuid = xenserver_get_host_uuid();
1418 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1419 dpid = dpid_from_hash(combined, strlen(combined));
1425 return eth_addr_to_uint64(bridge_ea);
1429 dpid_from_hash(const void *data, size_t n)
1431 uint8_t hash[SHA1_DIGEST_SIZE];
1433 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1434 sha1_bytes(data, n, hash);
1435 eth_addr_mark_random(hash);
1436 return eth_addr_to_uint64(hash);
1440 iface_refresh_status(struct iface *iface)
1444 enum netdev_flags flags;
1451 if (iface_is_synthetic(iface)) {
1457 if (!netdev_get_status(iface->netdev, &sh)) {
1459 char **keys, **values;
1461 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1462 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1467 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1470 shash_destroy_free_data(&sh);
1472 error = netdev_get_flags(iface->netdev, &flags);
1474 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1477 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1480 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1482 ovsrec_interface_set_duplex(iface->cfg,
1483 netdev_features_is_full_duplex(current)
1485 /* warning: uint64_t -> int64_t conversion */
1486 bps = netdev_features_to_bps(current);
1487 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1490 ovsrec_interface_set_duplex(iface->cfg, NULL);
1491 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1494 error = netdev_get_mtu(iface->netdev, &mtu);
1497 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1500 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1504 /* Writes 'iface''s CFM statistics to the database. */
1506 iface_refresh_cfm_stats(struct iface *iface)
1508 const struct ovsrec_interface *cfg = iface->cfg;
1510 const uint64_t *rmps;
1513 if (iface_is_synthetic(iface)) {
1517 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1520 bool fault_bool = fault;
1521 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1523 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1526 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1527 iface->ofp_port, &rmps, &n_rmps);
1529 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1532 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1537 iface_refresh_stats(struct iface *iface)
1539 #define IFACE_STATS \
1540 IFACE_STAT(rx_packets, "rx_packets") \
1541 IFACE_STAT(tx_packets, "tx_packets") \
1542 IFACE_STAT(rx_bytes, "rx_bytes") \
1543 IFACE_STAT(tx_bytes, "tx_bytes") \
1544 IFACE_STAT(rx_dropped, "rx_dropped") \
1545 IFACE_STAT(tx_dropped, "tx_dropped") \
1546 IFACE_STAT(rx_errors, "rx_errors") \
1547 IFACE_STAT(tx_errors, "tx_errors") \
1548 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1549 IFACE_STAT(rx_over_errors, "rx_over_err") \
1550 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1551 IFACE_STAT(collisions, "collisions")
1553 #define IFACE_STAT(MEMBER, NAME) NAME,
1554 static char *keys[] = { IFACE_STATS };
1556 int64_t values[ARRAY_SIZE(keys)];
1559 struct netdev_stats stats;
1561 if (iface_is_synthetic(iface)) {
1565 /* Intentionally ignore return value, since errors will set 'stats' to
1566 * all-1s, and we will deal with that correctly below. */
1567 netdev_get_stats(iface->netdev, &stats);
1569 /* Copy statistics into values[] array. */
1571 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1574 assert(i == ARRAY_SIZE(keys));
1576 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1581 br_refresh_stp_status(struct bridge *br)
1583 struct ofproto *ofproto = br->ofproto;
1584 struct ofproto_stp_status status;
1585 char *keys[3], *values[3];
1588 if (ofproto_get_stp_status(ofproto, &status)) {
1592 if (!status.enabled) {
1593 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1597 keys[0] = "stp_bridge_id",
1598 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1599 keys[1] = "stp_designated_root",
1600 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1601 keys[2] = "stp_root_path_cost",
1602 values[2] = xasprintf("%d", status.root_path_cost);
1604 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1606 for (i = 0; i < ARRAY_SIZE(values); i++) {
1612 port_refresh_stp_status(struct port *port)
1614 struct ofproto *ofproto = port->bridge->ofproto;
1615 struct iface *iface;
1616 struct ofproto_port_stp_status status;
1618 char *str_values[4];
1619 int64_t int_values[3];
1622 if (port_is_synthetic(port)) {
1626 /* STP doesn't currently support bonds. */
1627 if (!list_is_singleton(&port->ifaces)) {
1628 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1632 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1634 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1638 if (!status.enabled) {
1639 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1640 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1644 /* Set Status column. */
1645 keys[0] = "stp_port_id";
1646 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1647 keys[1] = "stp_state";
1648 str_values[1] = xstrdup(stp_state_name(status.state));
1649 keys[2] = "stp_sec_in_state";
1650 str_values[2] = xasprintf("%u", status.sec_in_state);
1651 keys[3] = "stp_role";
1652 str_values[3] = xstrdup(stp_role_name(status.role));
1654 ovsrec_port_set_status(port->cfg, keys, str_values,
1655 ARRAY_SIZE(str_values));
1657 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1658 free(str_values[i]);
1661 /* Set Statistics column. */
1662 keys[0] = "stp_tx_count";
1663 int_values[0] = status.tx_count;
1664 keys[1] = "stp_rx_count";
1665 int_values[1] = status.rx_count;
1666 keys[2] = "stp_error_count";
1667 int_values[2] = status.error_count;
1669 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1670 ARRAY_SIZE(int_values));
1674 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1678 /* Use other-config:enable-system-stats by preference. */
1679 enable = get_ovsrec_key_value(cfg->key_other_config,
1680 cfg->value_other_config,
1681 cfg->n_other_config,
1682 "enable-statistics");
1684 return !strcmp(enable, "true");
1687 /* Disable by default. */
1692 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1694 struct ovsdb_datum datum;
1698 if (enable_system_stats(cfg)) {
1699 get_system_stats(&stats);
1702 ovsdb_datum_from_shash(&datum, &stats);
1703 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1707 static inline const char *
1708 nx_role_to_str(enum nx_role role)
1713 case NX_ROLE_MASTER:
1718 return "*** INVALID ROLE ***";
1723 refresh_controller_status(void)
1727 const struct ovsrec_controller *cfg;
1731 /* Accumulate status for controllers on all bridges. */
1732 HMAP_FOR_EACH (br, node, &all_bridges) {
1733 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1736 /* Update each controller in the database with current status. */
1737 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1738 struct ofproto_controller_info *cinfo =
1739 shash_find_data(&info, cfg->target);
1742 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1743 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1744 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1745 (char **) cinfo->pairs.values,
1748 ovsrec_controller_set_is_connected(cfg, false);
1749 ovsrec_controller_set_role(cfg, NULL);
1750 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1754 ofproto_free_ofproto_controller_info(&info);
1758 refresh_cfm_stats(void)
1760 static struct ovsdb_idl_txn *txn = NULL;
1765 txn = ovsdb_idl_txn_create(idl);
1767 HMAP_FOR_EACH (br, node, &all_bridges) {
1768 struct iface *iface;
1770 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1771 iface_refresh_cfm_stats(iface);
1776 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1777 ovsdb_idl_txn_destroy(txn);
1785 const struct ovsrec_open_vswitch *cfg;
1787 bool vlan_splinters_changed;
1788 bool datapath_destroyed;
1789 bool database_changed;
1792 /* (Re)configure if necessary. */
1793 database_changed = ovsdb_idl_run(idl);
1794 if (ovsdb_idl_is_lock_contended(idl)) {
1795 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1796 struct bridge *br, *next_br;
1798 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1799 "disabling this process until it goes away");
1801 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1805 } else if (!ovsdb_idl_has_lock(idl)) {
1808 cfg = ovsrec_open_vswitch_first(idl);
1810 /* Let each bridge do the work that it needs to do. */
1811 datapath_destroyed = false;
1812 HMAP_FOR_EACH (br, node, &all_bridges) {
1813 int error = ofproto_run(br->ofproto);
1815 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1816 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1817 "forcing reconfiguration", br->name);
1818 datapath_destroyed = true;
1822 /* Re-configure SSL. We do this on every trip through the main loop,
1823 * instead of just when the database changes, because the contents of the
1824 * key and certificate files can change without the database changing.
1826 * We do this before bridge_reconfigure() because that function might
1827 * initiate SSL connections and thus requires SSL to be configured. */
1828 if (cfg && cfg->ssl) {
1829 const struct ovsrec_ssl *ssl = cfg->ssl;
1831 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1832 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1835 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1837 vlan_splinters_changed = false;
1838 if (vlan_splinters_enabled_anywhere) {
1839 HMAP_FOR_EACH (br, node, &all_bridges) {
1840 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1841 vlan_splinters_changed = true;
1847 if (database_changed || datapath_destroyed || vlan_splinters_changed) {
1849 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1851 bridge_reconfigure(cfg);
1853 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1854 ovsdb_idl_txn_commit(txn);
1855 ovsdb_idl_txn_destroy(txn); /* XXX */
1857 /* We still need to reconfigure to avoid dangling pointers to
1858 * now-destroyed ovsrec structures inside bridge data. */
1859 static const struct ovsrec_open_vswitch null_cfg;
1861 bridge_reconfigure(&null_cfg);
1865 /* Refresh system and interface stats if necessary. */
1866 if (time_msec() >= stats_timer) {
1868 struct ovsdb_idl_txn *txn;
1870 txn = ovsdb_idl_txn_create(idl);
1871 HMAP_FOR_EACH (br, node, &all_bridges) {
1874 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1875 struct iface *iface;
1877 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1878 iface_refresh_stats(iface);
1879 iface_refresh_status(iface);
1883 refresh_system_stats(cfg);
1884 refresh_controller_status();
1885 ovsdb_idl_txn_commit(txn);
1886 ovsdb_idl_txn_destroy(txn); /* XXX */
1889 stats_timer = time_msec() + STATS_INTERVAL;
1892 if (time_msec() >= db_limiter) {
1893 struct ovsdb_idl_txn *txn;
1895 txn = ovsdb_idl_txn_create(idl);
1896 HMAP_FOR_EACH (br, node, &all_bridges) {
1897 struct iface *iface;
1900 br_refresh_stp_status(br);
1902 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1903 port_refresh_stp_status(port);
1906 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1907 const char *link_state;
1908 int64_t link_resets;
1911 if (iface_is_synthetic(iface)) {
1915 current = ofproto_port_is_lacp_current(br->ofproto,
1919 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1921 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1924 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1925 ovsrec_interface_set_link_state(iface->cfg, link_state);
1927 link_resets = netdev_get_carrier_resets(iface->netdev);
1928 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1932 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
1933 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1935 ovsdb_idl_txn_destroy(txn);
1938 refresh_cfm_stats();
1944 ovsdb_idl_wait(idl);
1945 if (!hmap_is_empty(&all_bridges)) {
1948 HMAP_FOR_EACH (br, node, &all_bridges) {
1949 ofproto_wait(br->ofproto);
1951 poll_timer_wait_until(stats_timer);
1953 if (db_limiter > time_msec()) {
1954 poll_timer_wait_until(db_limiter);
1959 /* QoS unixctl user interface functions. */
1961 struct qos_unixctl_show_cbdata {
1963 struct iface *iface;
1967 qos_unixctl_show_cb(unsigned int queue_id,
1968 const struct shash *details,
1971 struct qos_unixctl_show_cbdata *data = aux;
1972 struct ds *ds = data->ds;
1973 struct iface *iface = data->iface;
1974 struct netdev_queue_stats stats;
1975 struct shash_node *node;
1978 ds_put_cstr(ds, "\n");
1980 ds_put_format(ds, "Queue %u:\n", queue_id);
1982 ds_put_cstr(ds, "Default:\n");
1985 SHASH_FOR_EACH (node, details) {
1986 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1989 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1991 if (stats.tx_packets != UINT64_MAX) {
1992 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1995 if (stats.tx_bytes != UINT64_MAX) {
1996 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1999 if (stats.tx_errors != UINT64_MAX) {
2000 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2003 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2004 queue_id, strerror(error));
2009 qos_unixctl_show(struct unixctl_conn *conn,
2010 const char *args, void *aux OVS_UNUSED)
2012 struct ds ds = DS_EMPTY_INITIALIZER;
2013 struct shash sh = SHASH_INITIALIZER(&sh);
2014 struct iface *iface;
2016 struct shash_node *node;
2017 struct qos_unixctl_show_cbdata data;
2020 iface = iface_find(args);
2022 unixctl_command_reply(conn, 501, "no such interface");
2026 netdev_get_qos(iface->netdev, &type, &sh);
2028 if (*type != '\0') {
2029 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2031 SHASH_FOR_EACH (node, &sh) {
2032 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2037 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2040 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2042 unixctl_command_reply(conn, 200, ds_cstr(&ds));
2044 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2045 unixctl_command_reply(conn, 501, ds_cstr(&ds));
2048 shash_destroy_free_data(&sh);
2052 /* Bridge reconfiguration functions. */
2054 bridge_create(const struct ovsrec_bridge *br_cfg)
2058 assert(!bridge_lookup(br_cfg->name));
2059 br = xzalloc(sizeof *br);
2061 br->name = xstrdup(br_cfg->name);
2062 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2065 /* Derive the default Ethernet address from the bridge's UUID. This should
2066 * be unique and it will be stable between ovs-vswitchd runs. */
2067 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2068 eth_addr_mark_random(br->default_ea);
2070 hmap_init(&br->ports);
2071 hmap_init(&br->ifaces);
2072 hmap_init(&br->iface_by_name);
2073 hmap_init(&br->mirrors);
2075 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2079 bridge_destroy(struct bridge *br)
2082 struct mirror *mirror, *next_mirror;
2083 struct port *port, *next_port;
2085 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2088 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2089 mirror_destroy(mirror);
2091 hmap_remove(&all_bridges, &br->node);
2092 ofproto_destroy(br->ofproto);
2093 hmap_destroy(&br->ifaces);
2094 hmap_destroy(&br->ports);
2095 hmap_destroy(&br->iface_by_name);
2096 hmap_destroy(&br->mirrors);
2103 static struct bridge *
2104 bridge_lookup(const char *name)
2108 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2109 if (!strcmp(br->name, name)) {
2116 /* Handle requests for a listing of all flows known by the OpenFlow
2117 * stack, including those normally hidden. */
2119 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
2120 const char *args, void *aux OVS_UNUSED)
2125 br = bridge_lookup(args);
2127 unixctl_command_reply(conn, 501, "Unknown bridge");
2132 ofproto_get_all_flows(br->ofproto, &results);
2134 unixctl_command_reply(conn, 200, ds_cstr(&results));
2135 ds_destroy(&results);
2138 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2139 * connections and reconnect. If BRIDGE is not specified, then all bridges
2140 * drop their controller connections and reconnect. */
2142 bridge_unixctl_reconnect(struct unixctl_conn *conn,
2143 const char *args, void *aux OVS_UNUSED)
2146 if (args[0] != '\0') {
2147 br = bridge_lookup(args);
2149 unixctl_command_reply(conn, 501, "Unknown bridge");
2152 ofproto_reconnect_controllers(br->ofproto);
2154 HMAP_FOR_EACH (br, node, &all_bridges) {
2155 ofproto_reconnect_controllers(br->ofproto);
2158 unixctl_command_reply(conn, 200, NULL);
2162 bridge_get_controllers(const struct bridge *br,
2163 struct ovsrec_controller ***controllersp)
2165 struct ovsrec_controller **controllers;
2166 size_t n_controllers;
2168 controllers = br->cfg->controller;
2169 n_controllers = br->cfg->n_controller;
2171 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2177 *controllersp = controllers;
2179 return n_controllers;
2182 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2183 * those configured in 'br->cfg'. */
2185 bridge_add_del_ports(struct bridge *br,
2186 const unsigned long int *splinter_vlans)
2188 struct port *port, *next;
2189 struct shash_node *node;
2190 struct shash new_ports;
2193 /* Collect new ports. */
2194 shash_init(&new_ports);
2195 for (i = 0; i < br->cfg->n_ports; i++) {
2196 const char *name = br->cfg->ports[i]->name;
2197 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2198 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2202 if (bridge_get_controllers(br, NULL)
2203 && !shash_find(&new_ports, br->name)) {
2204 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2205 br->name, br->name);
2207 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2208 br->synth_local_port.n_interfaces = 1;
2209 br->synth_local_port.name = br->name;
2211 br->synth_local_iface.name = br->name;
2212 br->synth_local_iface.type = "internal";
2214 br->synth_local_ifacep = &br->synth_local_iface;
2216 shash_add(&new_ports, br->name, &br->synth_local_port);
2219 if (splinter_vlans) {
2220 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2223 /* Get rid of deleted ports.
2224 * Get rid of deleted interfaces on ports that still exist. */
2225 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2226 port->cfg = shash_find_data(&new_ports, port->name);
2230 port_del_ifaces(port);
2234 /* Create new ports.
2235 * Add new interfaces to existing ports. */
2236 SHASH_FOR_EACH (node, &new_ports) {
2237 struct port *port = port_lookup(br, node->name);
2239 struct ovsrec_port *cfg = node->data;
2240 port = port_create(br, cfg);
2242 port_add_ifaces(port);
2243 if (list_is_empty(&port->ifaces)) {
2244 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2245 br->name, port->name);
2249 shash_destroy(&new_ports);
2252 /* Initializes 'oc' appropriately as a management service controller for
2255 * The caller must free oc->target when it is no longer needed. */
2257 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2258 struct ofproto_controller *oc)
2260 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2261 oc->max_backoff = 0;
2262 oc->probe_interval = 60;
2263 oc->band = OFPROTO_OUT_OF_BAND;
2265 oc->burst_limit = 0;
2268 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2270 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2271 struct ofproto_controller *oc)
2273 oc->target = c->target;
2274 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2275 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2276 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2277 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2278 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2279 oc->burst_limit = (c->controller_burst_limit
2280 ? *c->controller_burst_limit : 0);
2283 /* Configures the IP stack for 'br''s local interface properly according to the
2284 * configuration in 'c'. */
2286 bridge_configure_local_iface_netdev(struct bridge *br,
2287 struct ovsrec_controller *c)
2289 struct netdev *netdev;
2290 struct in_addr mask, gateway;
2292 struct iface *local_iface;
2295 /* If there's no local interface or no IP address, give up. */
2296 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2297 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2301 /* Bring up the local interface. */
2302 netdev = local_iface->netdev;
2303 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2305 /* Configure the IP address and netmask. */
2306 if (!c->local_netmask
2307 || !inet_aton(c->local_netmask, &mask)
2309 mask.s_addr = guess_netmask(ip.s_addr);
2311 if (!netdev_set_in4(netdev, ip, mask)) {
2312 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2313 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2316 /* Configure the default gateway. */
2317 if (c->local_gateway
2318 && inet_aton(c->local_gateway, &gateway)
2319 && gateway.s_addr) {
2320 if (!netdev_add_router(netdev, gateway)) {
2321 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2322 br->name, IP_ARGS(&gateway.s_addr));
2327 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2328 * in either string are treated as equal to any number of slashes in the other,
2329 * e.g. "x///y" is equal to "x/y". */
2331 equal_pathnames(const char *a, const char *b)
2335 a += strspn(a, "/");
2336 b += strspn(b, "/");
2337 } else if (*a == '\0') {
2348 bridge_configure_remotes(struct bridge *br,
2349 const struct sockaddr_in *managers, size_t n_managers)
2351 const char *disable_ib_str, *queue_id_str;
2352 bool disable_in_band = false;
2355 struct ovsrec_controller **controllers;
2356 size_t n_controllers;
2358 enum ofproto_fail_mode fail_mode;
2360 struct ofproto_controller *ocs;
2364 /* Check if we should disable in-band control on this bridge. */
2365 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2366 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2367 disable_in_band = true;
2370 /* Set OpenFlow queue ID for in-band control. */
2371 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2372 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2373 ofproto_set_in_band_queue(br->ofproto, queue_id);
2375 if (disable_in_band) {
2376 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2378 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2381 n_controllers = bridge_get_controllers(br, &controllers);
2383 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2386 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2387 for (i = 0; i < n_controllers; i++) {
2388 struct ovsrec_controller *c = controllers[i];
2390 if (!strncmp(c->target, "punix:", 6)
2391 || !strncmp(c->target, "unix:", 5)) {
2392 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2395 whitelist = xasprintf("unix:%s/%s.controller",
2396 ovs_rundir(), br->name);
2397 if (!equal_pathnames(c->target, whitelist)) {
2398 /* Prevent remote ovsdb-server users from accessing arbitrary
2399 * Unix domain sockets and overwriting arbitrary local
2401 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2402 "controller \"%s\" due to possibility for remote "
2403 "exploit. Instead, specify whitelisted \"%s\" or "
2404 "connect to \"unix:%s/%s.mgmt\" (which is always "
2405 "available without special configuration).",
2406 br->name, c->target, whitelist,
2407 ovs_rundir(), br->name);
2415 bridge_configure_local_iface_netdev(br, c);
2416 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2417 if (disable_in_band) {
2418 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2423 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2424 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2427 /* Set the fail-mode. */
2428 fail_mode = !br->cfg->fail_mode
2429 || !strcmp(br->cfg->fail_mode, "standalone")
2430 ? OFPROTO_FAIL_STANDALONE
2431 : OFPROTO_FAIL_SECURE;
2432 ofproto_set_fail_mode(br->ofproto, fail_mode);
2434 /* Configure OpenFlow controller connection snooping. */
2435 if (!ofproto_has_snoops(br->ofproto)) {
2439 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2440 ovs_rundir(), br->name));
2441 ofproto_set_snoops(br->ofproto, &snoops);
2442 sset_destroy(&snoops);
2446 /* Port functions. */
2448 static struct port *
2449 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2453 port = xzalloc(sizeof *port);
2455 port->name = xstrdup(cfg->name);
2457 list_init(&port->ifaces);
2459 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2461 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2467 get_port_other_config(const struct ovsrec_port *port, const char *key,
2468 const char *default_value)
2472 value = get_ovsrec_key_value(port->key_other_config,
2473 port->value_other_config,
2474 port->n_other_config, key);
2475 return value ? value : default_value;
2479 get_interface_other_config(const struct ovsrec_interface *iface,
2480 const char *key, const char *default_value)
2484 value = get_ovsrec_key_value(iface->key_other_config,
2485 iface->value_other_config,
2486 iface->n_other_config, key);
2487 return value ? value : default_value;
2490 /* Deletes interfaces from 'port' that are no longer configured for it. */
2492 port_del_ifaces(struct port *port)
2494 struct iface *iface, *next;
2495 struct sset new_ifaces;
2498 /* Collect list of new interfaces. */
2499 sset_init(&new_ifaces);
2500 for (i = 0; i < port->cfg->n_interfaces; i++) {
2501 const char *name = port->cfg->interfaces[i]->name;
2502 const char *type = port->cfg->interfaces[i]->name;
2503 if (strcmp(type, "null")) {
2504 sset_add(&new_ifaces, name);
2508 /* Get rid of deleted interfaces. */
2509 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2510 if (!sset_contains(&new_ifaces, iface->name)) {
2511 iface_destroy(iface);
2515 sset_destroy(&new_ifaces);
2518 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2521 port_add_ifaces(struct port *port)
2523 struct shash new_ifaces;
2524 struct shash_node *node;
2527 /* Collect new ifaces. */
2528 shash_init(&new_ifaces);
2529 for (i = 0; i < port->cfg->n_interfaces; i++) {
2530 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2531 if (strcmp(cfg->type, "null")
2532 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2533 VLOG_WARN("port %s: %s specified twice as port interface",
2534 port->name, cfg->name);
2535 iface_clear_db_record(cfg);
2539 /* Create new interfaces.
2540 * Update interface types and 'cfg' members. */
2541 SHASH_FOR_EACH (node, &new_ifaces) {
2542 const struct ovsrec_interface *cfg = node->data;
2543 const char *iface_name = node->name;
2544 struct iface *iface;
2546 iface = iface_lookup(port->bridge, iface_name);
2548 iface = iface_create(port, cfg);
2553 /* Determine interface type. The local port always has type
2554 * "internal". Other ports take their type from the database and
2555 * default to "system" if none is specified. */
2556 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2557 : cfg->type[0] ? cfg->type
2560 shash_destroy(&new_ifaces);
2564 port_destroy(struct port *port)
2567 struct bridge *br = port->bridge;
2568 struct iface *iface, *next;
2571 ofproto_bundle_unregister(br->ofproto, port);
2574 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2575 iface_destroy(iface);
2578 hmap_remove(&br->ports, &port->hmap_node);
2580 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2587 static struct port *
2588 port_lookup(const struct bridge *br, const char *name)
2592 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2594 if (!strcmp(port->name, name)) {
2602 enable_lacp(struct port *port, bool *activep)
2604 if (!port->cfg->lacp) {
2605 /* XXX when LACP implementation has been sufficiently tested, enable by
2606 * default and make active on bonded ports. */
2608 } else if (!strcmp(port->cfg->lacp, "off")) {
2610 } else if (!strcmp(port->cfg->lacp, "active")) {
2613 } else if (!strcmp(port->cfg->lacp, "passive")) {
2617 VLOG_WARN("port %s: unknown LACP mode %s",
2618 port->name, port->cfg->lacp);
2623 static struct lacp_settings *
2624 port_configure_lacp(struct port *port, struct lacp_settings *s)
2626 const char *lacp_time;
2627 long long int custom_time;
2630 if (!enable_lacp(port, &s->active)) {
2634 s->name = port->name;
2635 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2637 /* Prefer bondable links if unspecified. */
2638 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2640 s->priority = (priority > 0 && priority <= UINT16_MAX
2642 : UINT16_MAX - !list_is_short(&port->ifaces));
2644 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2649 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2650 custom_time = atoi(lacp_time);
2651 if (!strcmp(lacp_time, "fast")) {
2652 s->lacp_time = LACP_TIME_FAST;
2653 } else if (!strcmp(lacp_time, "slow")) {
2654 s->lacp_time = LACP_TIME_SLOW;
2655 } else if (custom_time > 0) {
2656 s->lacp_time = LACP_TIME_CUSTOM;
2657 s->custom_time = custom_time;
2659 s->lacp_time = LACP_TIME_SLOW;
2666 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2668 int priority, portid, key;
2670 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2671 priority = atoi(get_interface_other_config(iface->cfg,
2672 "lacp-port-priority", "0"));
2673 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2676 if (portid <= 0 || portid > UINT16_MAX) {
2677 portid = iface->ofp_port;
2680 if (priority <= 0 || priority > UINT16_MAX) {
2681 priority = UINT16_MAX;
2684 if (key < 0 || key > UINT16_MAX) {
2688 s->name = iface->name;
2690 s->priority = priority;
2695 port_configure_bond(struct port *port, struct bond_settings *s,
2696 uint32_t *bond_stable_ids)
2698 const char *detect_s;
2699 struct iface *iface;
2700 int miimon_interval;
2703 s->name = port->name;
2704 s->balance = BM_SLB;
2705 if (port->cfg->bond_mode
2706 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2707 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2708 port->name, port->cfg->bond_mode,
2709 bond_mode_to_string(s->balance));
2711 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2712 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2713 "please use another bond type or disable flood_vlans",
2717 miimon_interval = atoi(get_port_other_config(port->cfg,
2718 "bond-miimon-interval", "0"));
2719 if (miimon_interval <= 0) {
2720 miimon_interval = 200;
2723 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2724 if (!strcmp(detect_s, "carrier")) {
2725 miimon_interval = 0;
2726 } else if (strcmp(detect_s, "miimon")) {
2727 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2728 "defaulting to carrier", port->name, detect_s);
2729 miimon_interval = 0;
2732 s->up_delay = MAX(0, port->cfg->bond_updelay);
2733 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2734 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2735 s->rebalance_interval = atoi(
2736 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2737 if (s->rebalance_interval < 1000) {
2738 s->rebalance_interval = 1000;
2741 s->fake_iface = port->cfg->bond_fake_iface;
2744 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2745 long long stable_id;
2747 stable_id = atoll(get_interface_other_config(iface->cfg,
2748 "bond-stable-id", "0"));
2749 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2750 stable_id = iface->ofp_port;
2752 bond_stable_ids[i++] = stable_id;
2754 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2758 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2759 * instead of obtaining it from the database. */
2761 port_is_synthetic(const struct port *port)
2763 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2766 /* Interface functions. */
2768 static struct iface *
2769 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2771 struct bridge *br = port->bridge;
2772 struct iface *iface;
2773 char *name = if_cfg->name;
2775 iface = xzalloc(sizeof *iface);
2777 iface->name = xstrdup(name);
2778 iface->ofp_port = -1;
2779 iface->tag = tag_create_random();
2780 iface->netdev = NULL;
2781 iface->cfg = if_cfg;
2783 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2785 list_push_back(&port->ifaces, &iface->port_elem);
2787 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2793 iface_destroy(struct iface *iface)
2796 struct port *port = iface->port;
2797 struct bridge *br = port->bridge;
2799 if (br->ofproto && iface->ofp_port >= 0) {
2800 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2803 if (iface->ofp_port >= 0) {
2804 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2807 list_remove(&iface->port_elem);
2808 hmap_remove(&br->iface_by_name, &iface->name_node);
2810 netdev_close(iface->netdev);
2817 static struct iface *
2818 iface_lookup(const struct bridge *br, const char *name)
2820 struct iface *iface;
2822 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2823 &br->iface_by_name) {
2824 if (!strcmp(iface->name, name)) {
2832 static struct iface *
2833 iface_find(const char *name)
2835 const struct bridge *br;
2837 HMAP_FOR_EACH (br, node, &all_bridges) {
2838 struct iface *iface = iface_lookup(br, name);
2847 static struct iface *
2848 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2850 struct iface *iface;
2852 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2853 hash_int(ofp_port, 0), &br->ifaces) {
2854 if (iface->ofp_port == ofp_port) {
2861 /* Set Ethernet address of 'iface', if one is specified in the configuration
2864 iface_set_mac(struct iface *iface)
2866 uint8_t ea[ETH_ADDR_LEN];
2868 if (!strcmp(iface->type, "internal")
2869 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2870 if (iface->ofp_port == OFPP_LOCAL) {
2871 VLOG_ERR("interface %s: ignoring mac in Interface record "
2872 "(use Bridge record to set local port's mac)",
2874 } else if (eth_addr_is_multicast(ea)) {
2875 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2878 int error = netdev_set_etheraddr(iface->netdev, ea);
2880 VLOG_ERR("interface %s: setting MAC failed (%s)",
2881 iface->name, strerror(error));
2887 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2889 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2891 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2892 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2896 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
2897 * sets the "ofport" field to -1.
2899 * This is appropriate when 'if_cfg''s interface cannot be created or is
2900 * otherwise invalid. */
2902 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
2904 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2905 iface_set_ofport(if_cfg, -1);
2906 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
2907 ovsrec_interface_set_admin_state(if_cfg, NULL);
2908 ovsrec_interface_set_duplex(if_cfg, NULL);
2909 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
2910 ovsrec_interface_set_link_state(if_cfg, NULL);
2911 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
2912 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
2913 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
2914 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
2915 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
2919 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2921 * The value strings in '*shash' are taken directly from values[], not copied,
2922 * so the caller should not modify or free them. */
2924 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2925 struct shash *shash)
2930 for (i = 0; i < n; i++) {
2931 shash_add(shash, keys[i], values[i]);
2935 /* Creates 'keys' and 'values' arrays from 'shash'.
2937 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2938 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2939 * are populated with with strings taken directly from 'shash' and thus have
2940 * the same ownership of the key-value pairs in shash.
2943 shash_to_ovs_idl_map(struct shash *shash,
2944 char ***keys, char ***values, size_t *n)
2948 struct shash_node *sn;
2950 count = shash_count(shash);
2952 k = xmalloc(count * sizeof *k);
2953 v = xmalloc(count * sizeof *v);
2956 SHASH_FOR_EACH(sn, shash) {
2967 struct iface_delete_queues_cbdata {
2968 struct netdev *netdev;
2969 const struct ovsdb_datum *queues;
2973 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2975 union ovsdb_atom atom;
2977 atom.integer = target;
2978 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2982 iface_delete_queues(unsigned int queue_id,
2983 const struct shash *details OVS_UNUSED, void *cbdata_)
2985 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2987 if (!queue_ids_include(cbdata->queues, queue_id)) {
2988 netdev_delete_queue(cbdata->netdev, queue_id);
2993 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2995 struct ofpbuf queues_buf;
2997 ofpbuf_init(&queues_buf, 0);
2999 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3000 netdev_set_qos(iface->netdev, NULL, NULL);
3002 struct iface_delete_queues_cbdata cbdata;
3003 struct shash details;
3007 /* Configure top-level Qos for 'iface'. */
3008 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3009 qos->n_other_config, &details);
3010 netdev_set_qos(iface->netdev, qos->type, &details);
3011 shash_destroy(&details);
3013 /* Deconfigure queues that were deleted. */
3014 cbdata.netdev = iface->netdev;
3015 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3017 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3019 /* Configure queues for 'iface'. */
3021 for (i = 0; i < qos->n_queues; i++) {
3022 const struct ovsrec_queue *queue = qos->value_queues[i];
3023 unsigned int queue_id = qos->key_queues[i];
3025 if (queue_id == 0) {
3029 if (queue->n_dscp == 1) {
3030 struct ofproto_port_queue *port_queue;
3032 port_queue = ofpbuf_put_uninit(&queues_buf,
3033 sizeof *port_queue);
3034 port_queue->queue = queue_id;
3035 port_queue->dscp = queue->dscp[0];
3038 shash_from_ovs_idl_map(queue->key_other_config,
3039 queue->value_other_config,
3040 queue->n_other_config, &details);
3041 netdev_set_queue(iface->netdev, queue_id, &details);
3042 shash_destroy(&details);
3045 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3046 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
3047 "queue (queue 0). Packets not directed to a "
3048 "correctly configured queue may be dropped.",
3053 if (iface->ofp_port >= 0) {
3054 const struct ofproto_port_queue *port_queues = queues_buf.data;
3055 size_t n_queues = queues_buf.size / sizeof *port_queues;
3057 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3058 port_queues, n_queues);
3061 netdev_set_policing(iface->netdev,
3062 iface->cfg->ingress_policing_rate,
3063 iface->cfg->ingress_policing_burst);
3065 ofpbuf_uninit(&queues_buf);
3069 iface_configure_cfm(struct iface *iface)
3071 const struct ovsrec_interface *cfg = iface->cfg;
3072 const char *extended_str, *opstate_str;
3073 struct cfm_settings s;
3075 if (!cfg->n_cfm_mpid) {
3076 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3080 s.mpid = *cfg->cfm_mpid;
3081 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3083 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3085 if (s.interval <= 0) {
3089 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3091 s.extended = !strcasecmp("true", extended_str);
3093 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3094 s.opup = !strcasecmp("up", opstate_str);
3096 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3099 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3100 * instead of obtaining it from the database. */
3102 iface_is_synthetic(const struct iface *iface)
3104 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3107 /* Port mirroring. */
3109 static struct mirror *
3110 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3114 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3115 if (uuid_equals(uuid, &m->uuid)) {
3123 bridge_configure_mirrors(struct bridge *br)
3125 const struct ovsdb_datum *mc;
3126 unsigned long *flood_vlans;
3127 struct mirror *m, *next;
3130 /* Get rid of deleted mirrors. */
3131 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3132 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3133 union ovsdb_atom atom;
3135 atom.uuid = m->uuid;
3136 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3141 /* Add new mirrors and reconfigure existing ones. */
3142 for (i = 0; i < br->cfg->n_mirrors; i++) {
3143 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3144 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3146 m = mirror_create(br, cfg);
3148 if (!mirror_configure(m, cfg)) {
3153 /* Update flooded vlans (for RSPAN). */
3154 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3155 br->cfg->n_flood_vlans);
3156 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3157 bitmap_free(flood_vlans);
3160 static struct mirror *
3161 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3165 m = xzalloc(sizeof *m);
3166 m->uuid = cfg->header_.uuid;
3167 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3169 m->name = xstrdup(cfg->name);
3175 mirror_destroy(struct mirror *m)
3178 struct bridge *br = m->bridge;
3181 ofproto_mirror_unregister(br->ofproto, m);
3184 hmap_remove(&br->mirrors, &m->hmap_node);
3191 mirror_collect_ports(struct mirror *m,
3192 struct ovsrec_port **in_ports, int n_in_ports,
3193 void ***out_portsp, size_t *n_out_portsp)
3195 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3196 size_t n_out_ports = 0;
3199 for (i = 0; i < n_in_ports; i++) {
3200 const char *name = in_ports[i]->name;
3201 struct port *port = port_lookup(m->bridge, name);
3203 out_ports[n_out_ports++] = port;
3205 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3206 "port %s", m->bridge->name, m->name, name);
3209 *out_portsp = out_ports;
3210 *n_out_portsp = n_out_ports;
3214 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
3216 struct ofproto_mirror_settings s;
3219 if (strcmp(cfg->name, m->name)) {
3221 m->name = xstrdup(cfg->name);
3225 /* Get output port or VLAN. */
3226 if (cfg->output_port) {
3227 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3228 if (!s.out_bundle) {
3229 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3230 m->bridge->name, m->name);
3233 s.out_vlan = UINT16_MAX;
3235 if (cfg->output_vlan) {
3236 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3237 "output vlan; ignoring output vlan",
3238 m->bridge->name, m->name);
3240 } else if (cfg->output_vlan) {
3241 /* The database should prevent invalid VLAN values. */
3242 s.out_bundle = NULL;
3243 s.out_vlan = *cfg->output_vlan;
3245 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3246 m->bridge->name, m->name);
3250 /* Get port selection. */
3251 if (cfg->select_all) {
3252 size_t n_ports = hmap_count(&m->bridge->ports);
3253 void **ports = xmalloc(n_ports * sizeof *ports);
3258 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3268 /* Get ports, dropping ports that don't exist.
3269 * The IDL ensures that there are no duplicates. */
3270 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3271 &s.srcs, &s.n_srcs);
3272 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3273 &s.dsts, &s.n_dsts);
3276 /* Get VLAN selection. */
3277 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3280 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3283 if (s.srcs != s.dsts) {
3292 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3294 * This is deprecated. It is only for compatibility with broken device drivers
3295 * in old versions of Linux that do not properly support VLANs when VLAN
3296 * devices are not used. When broken device drivers are no longer in
3297 * widespread use, we will delete these interfaces. */
3299 static void **blocks;
3300 static size_t n_blocks, allocated_blocks;
3302 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3303 * VLAN splinters are reconfigured. */
3305 register_block(void *block)
3307 if (n_blocks >= allocated_blocks) {
3308 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3310 blocks[n_blocks++] = block;
3313 /* Frees all of the blocks registered with register_block(). */
3315 free_registered_blocks(void)
3319 for (i = 0; i < n_blocks; i++) {
3325 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3328 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3332 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3333 return !strcmp(value, "true");
3336 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3339 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3340 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3341 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3344 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3345 * use, returns NULL.
3347 * Updates 'vlan_splinters_enabled_anywhere'. */
3348 static unsigned long int *
3349 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3351 unsigned long int *splinter_vlans;
3352 struct sset splinter_ifaces;
3353 const char *real_dev_name;
3354 struct shash *real_devs;
3355 struct shash_node *node;
3359 splinter_vlans = NULL;
3360 sset_init(&splinter_ifaces);
3361 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3362 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3365 for (j = 0; j < br_cfg->n_ports; j++) {
3366 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3369 for (k = 0; k < port_cfg->n_interfaces; k++) {
3370 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3372 if (vlan_splinters_is_enabled(iface_cfg)) {
3373 sset_add(&splinter_ifaces, iface_cfg->name);
3375 if (!splinter_vlans) {
3376 splinter_vlans = bitmap_allocate(4096);
3378 vlan_bitmap_from_array__(port_cfg->trunks,
3386 vlan_splinters_enabled_anywhere = splinter_vlans != NULL;
3387 if (!splinter_vlans) {
3388 sset_destroy(&splinter_ifaces);
3392 HMAP_FOR_EACH (br, node, &all_bridges) {
3394 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3398 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3399 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3400 * device to be created for it. */
3401 bitmap_set0(splinter_vlans, 0);
3402 bitmap_set0(splinter_vlans, 4095);
3404 /* Delete all VLAN devices that we don't need. */
3406 real_devs = vlandev_get_real_devs();
3407 SHASH_FOR_EACH (node, real_devs) {
3408 const struct vlan_real_dev *real_dev = node->data;
3409 const struct vlan_dev *vlan_dev;
3410 bool real_dev_has_splinters;
3412 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3414 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3415 if (!real_dev_has_splinters
3416 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3417 struct netdev *netdev;
3419 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3420 if (!netdev_get_in4(netdev, NULL, NULL) ||
3421 !netdev_get_in6(netdev, NULL)) {
3422 vlandev_del(vlan_dev->name);
3424 /* It has an IP address configured, so we don't own
3425 * it. Don't delete it. */
3427 netdev_close(netdev);
3434 /* Add all VLAN devices that we need. */
3435 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3438 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3439 if (!vlandev_get_name(real_dev_name, vid)) {
3440 vlandev_add(real_dev_name, vid);
3447 sset_destroy(&splinter_ifaces);
3449 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3450 free(splinter_vlans);
3453 return splinter_vlans;
3456 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3459 configure_splinter_port(struct port *port)
3461 struct ofproto *ofproto = port->bridge->ofproto;
3462 uint16_t realdev_ofp_port;
3463 const char *realdev_name;
3464 struct iface *vlandev, *realdev;
3466 ofproto_bundle_unregister(port->bridge->ofproto, port);
3468 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3471 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3472 realdev = iface_lookup(port->bridge, realdev_name);
3473 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3475 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3479 static struct ovsrec_port *
3480 synthesize_splinter_port(const char *real_dev_name,
3481 const char *vlan_dev_name, int vid)
3483 struct ovsrec_interface *iface;
3484 struct ovsrec_port *port;
3486 iface = xzalloc(sizeof *iface);
3487 iface->name = xstrdup(vlan_dev_name);
3488 iface->type = "system";
3490 port = xzalloc(sizeof *port);
3491 port->interfaces = xmemdup(&iface, sizeof iface);
3492 port->n_interfaces = 1;
3493 port->name = xstrdup(vlan_dev_name);
3494 port->vlan_mode = "splinter";
3495 port->tag = xmalloc(sizeof *port->tag);
3497 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3498 port->key_other_config[0] = "realdev";
3499 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3500 port->value_other_config[0] = xstrdup(real_dev_name);
3501 port->n_other_config = 1;
3503 register_block(iface);
3504 register_block(iface->name);
3505 register_block(port);
3506 register_block(port->interfaces);
3507 register_block(port->name);
3508 register_block(port->tag);
3509 register_block(port->key_other_config);
3510 register_block(port->value_other_config);
3511 register_block(port->value_other_config[0]);
3516 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3517 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3518 * 1-bit in the 'splinter_vlans' bitmap. */
3520 add_vlan_splinter_ports(struct bridge *br,
3521 const unsigned long int *splinter_vlans,
3522 struct shash *ports)
3526 free_registered_blocks();
3528 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3529 * we're modifying 'ports'. */
3530 for (i = 0; i < br->cfg->n_ports; i++) {
3531 const char *name = br->cfg->ports[i]->name;
3532 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3535 for (j = 0; j < port_cfg->n_interfaces; j++) {
3536 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3538 if (vlan_splinters_is_enabled(iface_cfg)) {
3539 const char *real_dev_name;
3542 real_dev_name = iface_cfg->name;
3543 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3544 const char *vlan_dev_name;
3546 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3548 && !shash_find(ports, vlan_dev_name)) {
3549 shash_add(ports, vlan_dev_name,
3550 synthesize_splinter_port(
3551 real_dev_name, vlan_dev_name, vid));