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"
48 #include "vswitchd/vswitch-idl.h"
49 #include "xenserver.h"
51 #include "sflow_api.h"
52 #include "vlan-bitmap.h"
54 VLOG_DEFINE_THIS_MODULE(bridge);
56 COVERAGE_DEFINE(bridge_reconfigure);
59 /* These members are always valid. */
60 struct list port_elem; /* Element in struct port's "ifaces" list. */
61 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
62 struct port *port; /* Containing port. */
63 char *name; /* Host network device name. */
64 tag_type tag; /* Tag associated with this interface. */
66 /* These members are valid only after bridge_reconfigure() causes them to
68 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
69 int ofp_port; /* OpenFlow port number, -1 if unknown. */
70 struct netdev *netdev; /* Network device. */
71 const char *type; /* Usually same as cfg->type. */
72 const struct ovsrec_interface *cfg;
76 struct uuid uuid; /* UUID of this "mirror" record in database. */
77 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
78 struct bridge *bridge;
83 struct bridge *bridge;
84 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
87 const struct ovsrec_port *cfg;
89 /* An ordinary bridge port has 1 interface.
90 * A bridge port for bonding has at least 2 interfaces. */
91 struct list ifaces; /* List of "struct iface"s. */
95 struct hmap_node node; /* In 'all_bridges'. */
96 char *name; /* User-specified arbitrary name. */
97 char *type; /* Datapath type. */
98 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
99 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
100 const struct ovsrec_bridge *cfg;
102 /* OpenFlow switch processing. */
103 struct ofproto *ofproto; /* OpenFlow switch. */
106 struct hmap ports; /* "struct port"s indexed by name. */
107 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
108 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
110 /* Port mirroring. */
111 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
113 /* Synthetic local port if necessary. */
114 struct ovsrec_port synth_local_port;
115 struct ovsrec_interface synth_local_iface;
116 struct ovsrec_interface *synth_local_ifacep;
119 /* All bridges, indexed by name. */
120 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
122 /* OVSDB IDL used to obtain configuration. */
123 static struct ovsdb_idl *idl;
125 /* Each time this timer expires, the bridge fetches systems and interface
126 * statistics and pushes them into the database. */
127 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
128 static long long int stats_timer = LLONG_MIN;
130 /* Stores the time after which rate limited statistics may be written to the
131 * database. Only updated when changes to the database require rate limiting.
133 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
134 static long long int db_limiter = LLONG_MIN;
136 static void add_del_bridges(const struct ovsrec_open_vswitch *);
137 static void bridge_del_ofprotos(void);
138 static bool bridge_add_ofprotos(struct bridge *);
139 static void bridge_create(const struct ovsrec_bridge *);
140 static void bridge_destroy(struct bridge *);
141 static struct bridge *bridge_lookup(const char *name);
142 static unixctl_cb_func bridge_unixctl_dump_flows;
143 static unixctl_cb_func bridge_unixctl_reconnect;
144 static size_t bridge_get_controllers(const struct bridge *br,
145 struct ovsrec_controller ***controllersp);
146 static void bridge_add_del_ports(struct bridge *);
147 static void bridge_add_ofproto_ports(struct bridge *);
148 static void bridge_del_ofproto_ports(struct bridge *);
149 static void bridge_refresh_ofp_port(struct bridge *);
150 static void bridge_configure_datapath_id(struct bridge *);
151 static void bridge_configure_flow_eviction_threshold(struct bridge *);
152 static void bridge_configure_netflow(struct bridge *);
153 static void bridge_configure_forward_bpdu(struct bridge *);
154 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
155 static void bridge_configure_remotes(struct bridge *,
156 const struct sockaddr_in *managers,
158 static void bridge_pick_local_hw_addr(struct bridge *,
159 uint8_t ea[ETH_ADDR_LEN],
160 struct iface **hw_addr_iface);
161 static uint64_t bridge_pick_datapath_id(struct bridge *,
162 const uint8_t bridge_ea[ETH_ADDR_LEN],
163 struct iface *hw_addr_iface);
164 static uint64_t dpid_from_hash(const void *, size_t nbytes);
165 static bool bridge_has_bond_fake_iface(const struct bridge *,
167 static bool port_is_bond_fake_iface(const struct port *);
169 static unixctl_cb_func qos_unixctl_show;
171 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
172 static void port_add_ifaces(struct port *);
173 static void port_del_ifaces(struct port *);
174 static void port_destroy(struct port *);
175 static struct port *port_lookup(const struct bridge *, const char *name);
176 static void port_configure(struct port *);
177 static struct lacp_settings *port_configure_lacp(struct port *,
178 struct lacp_settings *);
179 static void port_configure_bond(struct port *, struct bond_settings *,
180 uint32_t *bond_stable_ids);
182 static void bridge_configure_mirrors(struct bridge *);
183 static struct mirror *mirror_create(struct bridge *,
184 const struct ovsrec_mirror *);
185 static void mirror_destroy(struct mirror *);
186 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
188 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
189 static struct iface *iface_create(struct port *port,
190 const struct ovsrec_interface *if_cfg);
191 static void iface_destroy(struct iface *);
192 static struct iface *iface_lookup(const struct bridge *, const char *name);
193 static struct iface *iface_find(const char *name);
194 static struct iface *iface_from_ofp_port(const struct bridge *,
196 static void iface_set_mac(struct iface *);
197 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
198 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
199 static void iface_configure_cfm(struct iface *);
200 static bool iface_refresh_cfm_stats(struct iface *);
201 static void iface_refresh_stats(struct iface *);
202 static void iface_refresh_status(struct iface *);
203 static bool iface_get_carrier(const struct iface *);
204 static bool iface_is_synthetic(const struct iface *);
206 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
208 static void shash_to_ovs_idl_map(struct shash *,
209 char ***keys, char ***values, size_t *n);
211 /* Public functions. */
213 /* Initializes the bridge module, configuring it to obtain its configuration
214 * from an OVSDB server accessed over 'remote', which should be a string in a
215 * form acceptable to ovsdb_idl_create(). */
217 bridge_init(const char *remote)
219 /* Create connection to database. */
220 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
221 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
223 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
224 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
225 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
226 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
227 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
228 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
229 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
231 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
232 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
234 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
235 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
237 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
238 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
239 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
240 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
241 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
242 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
243 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
244 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
245 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
247 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
248 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
249 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
250 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
252 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
254 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
258 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
260 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
262 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
263 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
264 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
265 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
266 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
268 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
270 /* Register unixctl commands. */
271 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
272 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
274 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
284 struct bridge *br, *next_br;
286 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
289 ovsdb_idl_destroy(idl);
292 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
293 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
294 * responsible for freeing '*managersp' (with free()).
296 * You may be asking yourself "why does ovs-vswitchd care?", because
297 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
298 * should not be and in fact is not directly involved in that. But
299 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
300 * it has to tell in-band control where the managers are to enable that.
301 * (Thus, only managers connected in-band are collected.)
304 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
305 struct sockaddr_in **managersp, size_t *n_managersp)
307 struct sockaddr_in *managers = NULL;
308 size_t n_managers = 0;
312 /* Collect all of the potential targets from the "targets" columns of the
313 * rows pointed to by "manager_options", excluding any that are
316 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
317 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
319 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
320 sset_find_and_delete(&targets, m->target);
322 sset_add(&targets, m->target);
326 /* Now extract the targets' IP addresses. */
327 if (!sset_is_empty(&targets)) {
330 managers = xmalloc(sset_count(&targets) * sizeof *managers);
331 SSET_FOR_EACH (target, &targets) {
332 struct sockaddr_in *sin = &managers[n_managers];
334 if ((!strncmp(target, "tcp:", 4)
335 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
336 (!strncmp(target, "ssl:", 4)
337 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
342 sset_destroy(&targets);
344 *managersp = managers;
345 *n_managersp = n_managers;
349 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
351 struct sockaddr_in *managers;
352 struct bridge *br, *next;
353 int sflow_bridge_number;
356 COVERAGE_INC(bridge_reconfigure);
358 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
359 * iface"s according to 'ovs_cfg', with only very minimal configuration
362 * This is purely an update to bridge data structures. Nothing is pushed
363 * down to ofproto or lower layers. */
364 add_del_bridges(ovs_cfg);
365 HMAP_FOR_EACH (br, node, &all_bridges) {
366 bridge_add_del_ports(br);
369 /* Delete all datapaths and datapath ports that are no longer configured.
371 * The kernel will reject any attempt to add a given port to a datapath if
372 * that port already belongs to a different datapath, so we must do all
373 * port deletions before any port additions. A datapath always has a
374 * "local port" so we must delete not-configured datapaths too. */
375 bridge_del_ofprotos();
376 HMAP_FOR_EACH (br, node, &all_bridges) {
378 bridge_del_ofproto_ports(br);
382 /* Create datapaths and datapath ports that are missing.
384 * After this is done, we have our final set of bridges, ports, and
385 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
386 * has at least one iface, every "struct iface" has a valid ofp_port and
388 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
389 if (!br->ofproto && !bridge_add_ofprotos(br)) {
393 HMAP_FOR_EACH (br, node, &all_bridges) {
394 bridge_refresh_ofp_port(br);
395 bridge_add_ofproto_ports(br);
398 /* Complete the configuration. */
399 sflow_bridge_number = 0;
400 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
401 HMAP_FOR_EACH (br, node, &all_bridges) {
404 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
407 port_configure(port);
409 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
410 iface_configure_cfm(iface);
411 iface_configure_qos(iface, port->cfg->qos);
412 iface_set_mac(iface);
415 bridge_configure_mirrors(br);
416 bridge_configure_datapath_id(br);
417 bridge_configure_flow_eviction_threshold(br);
418 bridge_configure_forward_bpdu(br);
419 bridge_configure_remotes(br, managers, n_managers);
420 bridge_configure_netflow(br);
421 bridge_configure_sflow(br, &sflow_bridge_number);
425 /* ovs-vswitchd has completed initialization, so allow the process that
426 * forked us to exit successfully. */
427 daemonize_complete();
430 /* Iterate over all ofprotos and delete any of them that do not have a
431 * configured bridge or that are the wrong type. */
433 bridge_del_ofprotos(void)
441 ofproto_enumerate_types(&types);
442 SSET_FOR_EACH (type, &types) {
445 ofproto_enumerate_names(type, &names);
446 SSET_FOR_EACH (name, &names) {
447 struct bridge *br = bridge_lookup(name);
448 if (!br || strcmp(type, br->type)) {
449 ofproto_delete(name, type);
453 sset_destroy(&names);
454 sset_destroy(&types);
458 bridge_add_ofprotos(struct bridge *br)
460 int error = ofproto_create(br->name, br->type, &br->ofproto);
462 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
469 port_configure(struct port *port)
471 const struct ovsrec_port *cfg = port->cfg;
472 struct bond_settings bond_settings;
473 struct lacp_settings lacp_settings;
474 struct ofproto_bundle_settings s;
482 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
483 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
484 s.slaves[s.n_slaves++] = iface->ofp_port;
490 if (list_is_short(&port->ifaces)) {
491 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
493 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan);
496 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
497 * they even work as-is. But they have not been tested. */
498 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
503 /* Get VLAN trunks. */
505 if (s.vlan < 0 && cfg->n_trunks) {
506 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
507 } else if (s.vlan >= 0 && cfg->n_trunks) {
508 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
512 /* Get LACP settings. */
513 s.lacp = port_configure_lacp(port, &lacp_settings);
517 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
518 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
519 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
522 s.lacp_slaves = NULL;
525 /* Get bond settings. */
526 if (s.n_slaves > 1) {
527 s.bond = &bond_settings;
528 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
529 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
532 s.bond_stable_ids = NULL;
534 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
535 netdev_set_miimon_interval(iface->netdev, 0);
540 ofproto_bundle_register(port->bridge->ofproto, port, &s);
546 free(s.bond_stable_ids);
549 /* Pick local port hardware address and datapath ID for 'br'. */
551 bridge_configure_datapath_id(struct bridge *br)
553 uint8_t ea[ETH_ADDR_LEN];
555 struct iface *local_iface;
556 struct iface *hw_addr_iface;
559 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
560 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
562 int error = netdev_set_etheraddr(local_iface->netdev, ea);
564 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
565 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
566 "Ethernet address: %s",
567 br->name, strerror(error));
570 memcpy(br->ea, ea, ETH_ADDR_LEN);
572 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
573 ofproto_set_datapath_id(br->ofproto, dpid);
575 dpid_string = xasprintf("%016"PRIx64, dpid);
576 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
580 /* Set NetFlow configuration on 'br'. */
582 bridge_configure_netflow(struct bridge *br)
584 struct ovsrec_netflow *cfg = br->cfg->netflow;
585 struct netflow_options opts;
588 ofproto_set_netflow(br->ofproto, NULL);
592 memset(&opts, 0, sizeof opts);
594 /* Get default NetFlow configuration from datapath.
595 * Apply overrides from 'cfg'. */
596 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
597 if (cfg->engine_type) {
598 opts.engine_type = *cfg->engine_type;
600 if (cfg->engine_id) {
601 opts.engine_id = *cfg->engine_id;
604 /* Configure active timeout interval. */
605 opts.active_timeout = cfg->active_timeout;
606 if (!opts.active_timeout) {
607 opts.active_timeout = -1;
608 } else if (opts.active_timeout < 0) {
609 VLOG_WARN("bridge %s: active timeout interval set to negative "
610 "value, using default instead (%d seconds)", br->name,
611 NF_ACTIVE_TIMEOUT_DEFAULT);
612 opts.active_timeout = -1;
615 /* Add engine ID to interface number to disambiguate bridgs? */
616 opts.add_id_to_iface = cfg->add_id_to_interface;
617 if (opts.add_id_to_iface) {
618 if (opts.engine_id > 0x7f) {
619 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
620 "another vswitch, choose an engine id less than 128",
623 if (hmap_count(&br->ports) > 508) {
624 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
625 "another port when more than 508 ports are used",
631 sset_init(&opts.collectors);
632 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
635 if (ofproto_set_netflow(br->ofproto, &opts)) {
636 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
638 sset_destroy(&opts.collectors);
641 /* Set sFlow configuration on 'br'. */
643 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
645 const struct ovsrec_sflow *cfg = br->cfg->sflow;
646 struct ovsrec_controller **controllers;
647 struct ofproto_sflow_options oso;
648 size_t n_controllers;
652 ofproto_set_sflow(br->ofproto, NULL);
656 memset(&oso, 0, sizeof oso);
658 sset_init(&oso.targets);
659 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
661 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
663 oso.sampling_rate = *cfg->sampling;
666 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
668 oso.polling_interval = *cfg->polling;
671 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
673 oso.header_len = *cfg->header;
676 oso.sub_id = (*sflow_bridge_number)++;
677 oso.agent_device = cfg->agent;
679 oso.control_ip = NULL;
680 n_controllers = bridge_get_controllers(br, &controllers);
681 for (i = 0; i < n_controllers; i++) {
682 if (controllers[i]->local_ip) {
683 oso.control_ip = controllers[i]->local_ip;
687 ofproto_set_sflow(br->ofproto, &oso);
689 sset_destroy(&oso.targets);
693 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
695 const struct port *port = port_lookup(br, name);
696 return port && port_is_bond_fake_iface(port);
700 port_is_bond_fake_iface(const struct port *port)
702 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
706 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
708 struct bridge *br, *next;
712 /* Collect new bridges' names and types. */
714 for (i = 0; i < cfg->n_bridges; i++) {
715 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
716 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
717 VLOG_WARN("bridge %s specified twice", br_cfg->name);
721 /* Get rid of deleted bridges or those whose types have changed.
722 * Update 'cfg' of bridges that still exist. */
723 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
724 br->cfg = shash_find_data(&new_br, br->name);
725 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
726 br->cfg->datapath_type))) {
731 /* Add new bridges. */
732 for (i = 0; i < cfg->n_bridges; i++) {
733 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
734 struct bridge *br = bridge_lookup(br_cfg->name);
736 bridge_create(br_cfg);
740 shash_destroy(&new_br);
743 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
746 * The kernel will reject any attempt to add a given port to a datapath if that
747 * port already belongs to a different datapath, so we must do all port
748 * deletions before any port additions. */
750 bridge_del_ofproto_ports(struct bridge *br)
752 struct ofproto_port_dump dump;
753 struct ofproto_port ofproto_port;
755 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
756 const char *name = ofproto_port.name;
761 /* Ignore the local port. We can't change it anyhow. */
762 if (!strcmp(name, br->name)) {
766 /* Get the type that 'ofproto_port' should have (ordinarily the
767 * type of its corresponding iface) or NULL if it should be
769 iface = iface_lookup(br, name);
770 type = (iface ? iface->type
771 : bridge_has_bond_fake_iface(br, name) ? "internal"
774 /* If it's the wrong type then delete the ofproto port. */
776 && !strcmp(ofproto_port.type, type)
777 && (!iface || !iface->netdev
778 || !strcmp(netdev_get_type(iface->netdev), type))) {
781 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
783 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
784 br->name, name, strerror(error));
787 netdev_close(iface->netdev);
788 iface->netdev = NULL;
794 iface_set_ofp_port(struct iface *iface, int ofp_port)
796 struct bridge *br = iface->port->bridge;
798 assert(iface->ofp_port < 0 && ofp_port >= 0);
799 iface->ofp_port = ofp_port;
800 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
801 iface_set_ofport(iface->cfg, ofp_port);
805 bridge_refresh_ofp_port(struct bridge *br)
807 struct ofproto_port_dump dump;
808 struct ofproto_port ofproto_port;
811 /* Clear all the "ofp_port"es. */
812 hmap_clear(&br->ifaces);
813 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
816 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
817 iface->ofp_port = -1;
821 /* Obtain the correct "ofp_port"s from ofproto. */
822 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
823 struct iface *iface = iface_lookup(br, ofproto_port.name);
825 if (iface->ofp_port >= 0) {
826 VLOG_WARN("bridge %s: interface %s reported twice",
827 br->name, ofproto_port.name);
828 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
829 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
830 br->name, ofproto_port.ofp_port);
832 iface_set_ofp_port(iface, ofproto_port.ofp_port);
838 /* Add an ofproto port for any "struct iface" that doesn't have one.
839 * Delete any "struct iface" for which this fails.
840 * Delete any "struct port" that thereby ends up with no ifaces. */
842 bridge_add_ofproto_ports(struct bridge *br)
844 struct port *port, *next_port;
846 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
847 struct iface *iface, *next_iface;
848 struct ofproto_port ofproto_port;
850 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
853 /* Open the netdev. */
854 if (!iface->netdev) {
855 error = netdev_open(iface->name, iface->type, &iface->netdev);
857 VLOG_WARN("could not open network device %s (%s)",
858 iface->name, strerror(error));
864 /* Configure the netdev. */
869 shash_from_ovs_idl_map(iface->cfg->key_options,
870 iface->cfg->value_options,
871 iface->cfg->n_options, &args);
872 error = netdev_set_config(iface->netdev, &args);
873 shash_destroy(&args);
876 VLOG_WARN("could not configure network device %s (%s)",
877 iface->name, strerror(error));
878 netdev_close(iface->netdev);
879 iface->netdev = NULL;
883 /* Add the port, if necessary. */
884 if (iface->netdev && iface->ofp_port < 0) {
888 error = ofproto_port_add(br->ofproto, iface->netdev,
891 iface_set_ofp_port(iface, ofp_port);
893 netdev_close(iface->netdev);
894 iface->netdev = NULL;
898 /* Populate stats columns in new Interface rows. */
899 if (iface->netdev && !iface->cfg->mtu) {
900 iface_refresh_stats(iface);
901 iface_refresh_status(iface);
904 /* Delete the iface if we failed. */
905 if (iface->netdev && iface->ofp_port >= 0) {
906 VLOG_DBG("bridge %s: interface %s is on port %d",
907 br->name, iface->name, iface->ofp_port);
910 VLOG_ERR("bridge %s: missing %s interface, dropping",
911 br->name, iface->name);
913 /* We already reported a related error, don't bother
916 iface_set_ofport(iface->cfg, -1);
917 iface_destroy(iface);
920 if (list_is_empty(&port->ifaces)) {
921 VLOG_WARN("%s port has no interfaces, dropping", port->name);
926 /* Add bond fake iface if necessary. */
927 if (port_is_bond_fake_iface(port)) {
928 if (ofproto_port_query_by_name(br->ofproto, port->name,
930 struct netdev *netdev;
933 error = netdev_open(port->name, "internal", &netdev);
935 ofproto_port_add(br->ofproto, netdev, NULL);
936 netdev_close(netdev);
938 VLOG_WARN("could not open network device %s (%s)",
939 port->name, strerror(error));
942 /* Already exists, nothing to do. */
943 ofproto_port_destroy(&ofproto_port);
950 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
951 const struct ovsdb_idl_column *column,
954 const struct ovsdb_datum *datum;
955 union ovsdb_atom atom;
958 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
959 atom.string = (char *) key;
960 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
961 return idx == UINT_MAX ? NULL : datum->values[idx].string;
965 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
967 return get_ovsrec_key_value(&br_cfg->header_,
968 &ovsrec_bridge_col_other_config, key);
971 /* Set Flow eviction threshold */
973 bridge_configure_flow_eviction_threshold(struct bridge *br)
975 const char *threshold_str;
978 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
980 threshold = strtoul(threshold_str, NULL, 10);
982 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
984 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
987 /* Set forward BPDU option. */
989 bridge_configure_forward_bpdu(struct bridge *br)
991 const char *forward_bpdu_str;
992 bool forward_bpdu = false;
994 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
995 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
998 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1002 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1003 struct iface **hw_addr_iface)
1009 *hw_addr_iface = NULL;
1011 /* Did the user request a particular MAC? */
1012 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1013 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1014 if (eth_addr_is_multicast(ea)) {
1015 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1016 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1017 } else if (eth_addr_is_zero(ea)) {
1018 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1024 /* Otherwise choose the minimum non-local MAC address among all of the
1026 memset(ea, 0xff, ETH_ADDR_LEN);
1027 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1028 uint8_t iface_ea[ETH_ADDR_LEN];
1029 struct iface *candidate;
1030 struct iface *iface;
1032 /* Mirror output ports don't participate. */
1033 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1037 /* Choose the MAC address to represent the port. */
1039 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1040 /* Find the interface with this Ethernet address (if any) so that
1041 * we can provide the correct devname to the caller. */
1042 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1043 uint8_t candidate_ea[ETH_ADDR_LEN];
1044 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1045 && eth_addr_equals(iface_ea, candidate_ea)) {
1050 /* Choose the interface whose MAC address will represent the port.
1051 * The Linux kernel bonding code always chooses the MAC address of
1052 * the first slave added to a bond, and the Fedora networking
1053 * scripts always add slaves to a bond in alphabetical order, so
1054 * for compatibility we choose the interface with the name that is
1055 * first in alphabetical order. */
1056 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1057 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1062 /* The local port doesn't count (since we're trying to choose its
1063 * MAC address anyway). */
1064 if (iface->ofp_port == OFPP_LOCAL) {
1069 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1075 /* Compare against our current choice. */
1076 if (!eth_addr_is_multicast(iface_ea) &&
1077 !eth_addr_is_local(iface_ea) &&
1078 !eth_addr_is_reserved(iface_ea) &&
1079 !eth_addr_is_zero(iface_ea) &&
1080 eth_addr_compare_3way(iface_ea, ea) < 0)
1082 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1083 *hw_addr_iface = iface;
1086 if (eth_addr_is_multicast(ea)) {
1087 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1088 *hw_addr_iface = NULL;
1089 VLOG_WARN("bridge %s: using default bridge Ethernet "
1090 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1092 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1093 br->name, ETH_ADDR_ARGS(ea));
1097 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1098 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1099 * an interface on 'br', then that interface must be passed in as
1100 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1101 * 'hw_addr_iface' must be passed in as a null pointer. */
1103 bridge_pick_datapath_id(struct bridge *br,
1104 const uint8_t bridge_ea[ETH_ADDR_LEN],
1105 struct iface *hw_addr_iface)
1108 * The procedure for choosing a bridge MAC address will, in the most
1109 * ordinary case, also choose a unique MAC that we can use as a datapath
1110 * ID. In some special cases, though, multiple bridges will end up with
1111 * the same MAC address. This is OK for the bridges, but it will confuse
1112 * the OpenFlow controller, because each datapath needs a unique datapath
1115 * Datapath IDs must be unique. It is also very desirable that they be
1116 * stable from one run to the next, so that policy set on a datapath
1119 const char *datapath_id;
1122 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1123 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1127 if (hw_addr_iface) {
1129 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1131 * A bridge whose MAC address is taken from a VLAN network device
1132 * (that is, a network device created with vconfig(8) or similar
1133 * tool) will have the same MAC address as a bridge on the VLAN
1134 * device's physical network device.
1136 * Handle this case by hashing the physical network device MAC
1137 * along with the VLAN identifier.
1139 uint8_t buf[ETH_ADDR_LEN + 2];
1140 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1141 buf[ETH_ADDR_LEN] = vlan >> 8;
1142 buf[ETH_ADDR_LEN + 1] = vlan;
1143 return dpid_from_hash(buf, sizeof buf);
1146 * Assume that this bridge's MAC address is unique, since it
1147 * doesn't fit any of the cases we handle specially.
1152 * A purely internal bridge, that is, one that has no non-virtual
1153 * network devices on it at all, is more difficult because it has no
1154 * natural unique identifier at all.
1156 * When the host is a XenServer, we handle this case by hashing the
1157 * host's UUID with the name of the bridge. Names of bridges are
1158 * persistent across XenServer reboots, although they can be reused if
1159 * an internal network is destroyed and then a new one is later
1160 * created, so this is fairly effective.
1162 * When the host is not a XenServer, we punt by using a random MAC
1163 * address on each run.
1165 const char *host_uuid = xenserver_get_host_uuid();
1167 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1168 dpid = dpid_from_hash(combined, strlen(combined));
1174 return eth_addr_to_uint64(bridge_ea);
1178 dpid_from_hash(const void *data, size_t n)
1180 uint8_t hash[SHA1_DIGEST_SIZE];
1182 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1183 sha1_bytes(data, n, hash);
1184 eth_addr_mark_random(hash);
1185 return eth_addr_to_uint64(hash);
1189 iface_refresh_status(struct iface *iface)
1193 enum netdev_flags flags;
1200 if (iface_is_synthetic(iface)) {
1206 if (!netdev_get_status(iface->netdev, &sh)) {
1208 char **keys, **values;
1210 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1211 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1216 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1219 shash_destroy_free_data(&sh);
1221 error = netdev_get_flags(iface->netdev, &flags);
1223 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1226 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1229 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1231 ovsrec_interface_set_duplex(iface->cfg,
1232 netdev_features_is_full_duplex(current)
1234 /* warning: uint64_t -> int64_t conversion */
1235 bps = netdev_features_to_bps(current);
1236 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1239 ovsrec_interface_set_duplex(iface->cfg, NULL);
1240 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1243 ovsrec_interface_set_link_state(iface->cfg,
1244 iface_get_carrier(iface) ? "up" : "down");
1246 error = netdev_get_mtu(iface->netdev, &mtu);
1247 if (!error && mtu != INT_MAX) {
1249 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1252 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1256 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1257 * changed, false otherwise. */
1259 iface_refresh_cfm_stats(struct iface *iface)
1261 const struct ovsrec_interface *cfg = iface->cfg;
1262 bool changed = false;
1265 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1272 if (cfg->n_cfm_fault != 1 || cfg->cfm_fault[0] != fault) {
1273 bool fault_bool = fault;
1274 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1282 iface_refresh_lacp_stats(struct iface *iface)
1284 struct ofproto *ofproto = iface->port->bridge->ofproto;
1285 int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
1286 int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
1290 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
1296 iface_refresh_stats(struct iface *iface)
1298 #define IFACE_STATS \
1299 IFACE_STAT(rx_packets, "rx_packets") \
1300 IFACE_STAT(tx_packets, "tx_packets") \
1301 IFACE_STAT(rx_bytes, "rx_bytes") \
1302 IFACE_STAT(tx_bytes, "tx_bytes") \
1303 IFACE_STAT(rx_dropped, "rx_dropped") \
1304 IFACE_STAT(tx_dropped, "tx_dropped") \
1305 IFACE_STAT(rx_errors, "rx_errors") \
1306 IFACE_STAT(tx_errors, "tx_errors") \
1307 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1308 IFACE_STAT(rx_over_errors, "rx_over_err") \
1309 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1310 IFACE_STAT(collisions, "collisions")
1312 #define IFACE_STAT(MEMBER, NAME) NAME,
1313 static char *keys[] = { IFACE_STATS };
1315 int64_t values[ARRAY_SIZE(keys)];
1318 struct netdev_stats stats;
1320 if (iface_is_synthetic(iface)) {
1324 /* Intentionally ignore return value, since errors will set 'stats' to
1325 * all-1s, and we will deal with that correctly below. */
1326 netdev_get_stats(iface->netdev, &stats);
1328 /* Copy statistics into values[] array. */
1330 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1333 assert(i == ARRAY_SIZE(keys));
1335 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1340 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1344 /* Use other-config:enable-system-stats by preference. */
1345 enable = get_ovsrec_key_value(&cfg->header_,
1346 &ovsrec_open_vswitch_col_other_config,
1347 "enable-statistics");
1349 return !strcmp(enable, "true");
1352 /* Disable by default. */
1357 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1359 struct ovsdb_datum datum;
1363 if (enable_system_stats(cfg)) {
1364 get_system_stats(&stats);
1367 ovsdb_datum_from_shash(&datum, &stats);
1368 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1372 static inline const char *
1373 nx_role_to_str(enum nx_role role)
1378 case NX_ROLE_MASTER:
1383 return "*** INVALID ROLE ***";
1388 refresh_controller_status(void)
1392 const struct ovsrec_controller *cfg;
1396 /* Accumulate status for controllers on all bridges. */
1397 HMAP_FOR_EACH (br, node, &all_bridges) {
1398 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1401 /* Update each controller in the database with current status. */
1402 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1403 struct ofproto_controller_info *cinfo =
1404 shash_find_data(&info, cfg->target);
1407 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1408 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1409 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1410 (char **) cinfo->pairs.values,
1413 ovsrec_controller_set_is_connected(cfg, false);
1414 ovsrec_controller_set_role(cfg, NULL);
1415 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1419 ofproto_free_ofproto_controller_info(&info);
1425 const struct ovsrec_open_vswitch *cfg;
1427 bool datapath_destroyed;
1428 bool database_changed;
1431 /* (Re)configure if necessary. */
1432 database_changed = ovsdb_idl_run(idl);
1433 if (ovsdb_idl_is_lock_contended(idl)) {
1434 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1435 struct bridge *br, *next_br;
1437 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1438 "disabling this process until it goes away");
1440 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1444 } else if (!ovsdb_idl_has_lock(idl)) {
1447 cfg = ovsrec_open_vswitch_first(idl);
1449 /* Let each bridge do the work that it needs to do. */
1450 datapath_destroyed = false;
1451 HMAP_FOR_EACH (br, node, &all_bridges) {
1452 int error = ofproto_run(br->ofproto);
1454 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1455 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1456 "forcing reconfiguration", br->name);
1457 datapath_destroyed = true;
1461 /* Re-configure SSL. We do this on every trip through the main loop,
1462 * instead of just when the database changes, because the contents of the
1463 * key and certificate files can change without the database changing.
1465 * We do this before bridge_reconfigure() because that function might
1466 * initiate SSL connections and thus requires SSL to be configured. */
1467 if (cfg && cfg->ssl) {
1468 const struct ovsrec_ssl *ssl = cfg->ssl;
1470 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1471 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1474 if (database_changed || datapath_destroyed) {
1476 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1478 bridge_reconfigure(cfg);
1480 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1481 ovsdb_idl_txn_commit(txn);
1482 ovsdb_idl_txn_destroy(txn); /* XXX */
1484 /* We still need to reconfigure to avoid dangling pointers to
1485 * now-destroyed ovsrec structures inside bridge data. */
1486 static const struct ovsrec_open_vswitch null_cfg;
1488 bridge_reconfigure(&null_cfg);
1492 /* Refresh system and interface stats if necessary. */
1493 if (time_msec() >= stats_timer) {
1495 struct ovsdb_idl_txn *txn;
1497 txn = ovsdb_idl_txn_create(idl);
1498 HMAP_FOR_EACH (br, node, &all_bridges) {
1501 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1502 struct iface *iface;
1504 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1505 iface_refresh_stats(iface);
1506 iface_refresh_status(iface);
1510 refresh_system_stats(cfg);
1511 refresh_controller_status();
1512 ovsdb_idl_txn_commit(txn);
1513 ovsdb_idl_txn_destroy(txn); /* XXX */
1516 stats_timer = time_msec() + STATS_INTERVAL;
1519 if (time_msec() >= db_limiter) {
1520 struct ovsdb_idl_txn *txn;
1521 bool changed = false;
1523 txn = ovsdb_idl_txn_create(idl);
1524 HMAP_FOR_EACH (br, node, &all_bridges) {
1527 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1528 struct iface *iface;
1530 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1531 changed = iface_refresh_cfm_stats(iface) || changed;
1532 changed = iface_refresh_lacp_stats(iface) || changed;
1538 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1541 ovsdb_idl_txn_commit(txn);
1542 ovsdb_idl_txn_destroy(txn);
1549 ovsdb_idl_wait(idl);
1550 if (!hmap_is_empty(&all_bridges)) {
1553 HMAP_FOR_EACH (br, node, &all_bridges) {
1554 ofproto_wait(br->ofproto);
1556 poll_timer_wait_until(stats_timer);
1558 if (db_limiter > time_msec()) {
1559 poll_timer_wait_until(db_limiter);
1564 /* QoS unixctl user interface functions. */
1566 struct qos_unixctl_show_cbdata {
1568 struct iface *iface;
1572 qos_unixctl_show_cb(unsigned int queue_id,
1573 const struct shash *details,
1576 struct qos_unixctl_show_cbdata *data = aux;
1577 struct ds *ds = data->ds;
1578 struct iface *iface = data->iface;
1579 struct netdev_queue_stats stats;
1580 struct shash_node *node;
1583 ds_put_cstr(ds, "\n");
1585 ds_put_format(ds, "Queue %u:\n", queue_id);
1587 ds_put_cstr(ds, "Default:\n");
1590 SHASH_FOR_EACH (node, details) {
1591 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1594 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1596 if (stats.tx_packets != UINT64_MAX) {
1597 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1600 if (stats.tx_bytes != UINT64_MAX) {
1601 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1604 if (stats.tx_errors != UINT64_MAX) {
1605 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1608 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1609 queue_id, strerror(error));
1614 qos_unixctl_show(struct unixctl_conn *conn,
1615 const char *args, void *aux OVS_UNUSED)
1617 struct ds ds = DS_EMPTY_INITIALIZER;
1618 struct shash sh = SHASH_INITIALIZER(&sh);
1619 struct iface *iface;
1621 struct shash_node *node;
1622 struct qos_unixctl_show_cbdata data;
1625 iface = iface_find(args);
1627 unixctl_command_reply(conn, 501, "no such interface");
1631 netdev_get_qos(iface->netdev, &type, &sh);
1633 if (*type != '\0') {
1634 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1636 SHASH_FOR_EACH (node, &sh) {
1637 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1642 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1645 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1647 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1649 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1650 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1653 shash_destroy_free_data(&sh);
1657 /* Bridge reconfiguration functions. */
1659 bridge_create(const struct ovsrec_bridge *br_cfg)
1663 assert(!bridge_lookup(br_cfg->name));
1664 br = xzalloc(sizeof *br);
1666 br->name = xstrdup(br_cfg->name);
1667 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1670 /* Derive the default Ethernet address from the bridge's UUID. This should
1671 * be unique and it will be stable between ovs-vswitchd runs. */
1672 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1673 eth_addr_mark_random(br->default_ea);
1675 hmap_init(&br->ports);
1676 hmap_init(&br->ifaces);
1677 hmap_init(&br->iface_by_name);
1678 hmap_init(&br->mirrors);
1680 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1684 bridge_destroy(struct bridge *br)
1687 struct mirror *mirror, *next_mirror;
1688 struct port *port, *next_port;
1690 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1693 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1694 mirror_destroy(mirror);
1696 hmap_remove(&all_bridges, &br->node);
1697 ofproto_destroy(br->ofproto);
1698 hmap_destroy(&br->ifaces);
1699 hmap_destroy(&br->ports);
1700 hmap_destroy(&br->iface_by_name);
1701 hmap_destroy(&br->mirrors);
1708 static struct bridge *
1709 bridge_lookup(const char *name)
1713 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1714 if (!strcmp(br->name, name)) {
1721 /* Handle requests for a listing of all flows known by the OpenFlow
1722 * stack, including those normally hidden. */
1724 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1725 const char *args, void *aux OVS_UNUSED)
1730 br = bridge_lookup(args);
1732 unixctl_command_reply(conn, 501, "Unknown bridge");
1737 ofproto_get_all_flows(br->ofproto, &results);
1739 unixctl_command_reply(conn, 200, ds_cstr(&results));
1740 ds_destroy(&results);
1743 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1744 * connections and reconnect. If BRIDGE is not specified, then all bridges
1745 * drop their controller connections and reconnect. */
1747 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1748 const char *args, void *aux OVS_UNUSED)
1751 if (args[0] != '\0') {
1752 br = bridge_lookup(args);
1754 unixctl_command_reply(conn, 501, "Unknown bridge");
1757 ofproto_reconnect_controllers(br->ofproto);
1759 HMAP_FOR_EACH (br, node, &all_bridges) {
1760 ofproto_reconnect_controllers(br->ofproto);
1763 unixctl_command_reply(conn, 200, NULL);
1767 bridge_get_controllers(const struct bridge *br,
1768 struct ovsrec_controller ***controllersp)
1770 struct ovsrec_controller **controllers;
1771 size_t n_controllers;
1773 controllers = br->cfg->controller;
1774 n_controllers = br->cfg->n_controller;
1776 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1782 *controllersp = controllers;
1784 return n_controllers;
1787 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1788 * those configured in 'br->cfg'. */
1790 bridge_add_del_ports(struct bridge *br)
1792 struct port *port, *next;
1793 struct shash_node *node;
1794 struct shash new_ports;
1797 /* Collect new ports. */
1798 shash_init(&new_ports);
1799 for (i = 0; i < br->cfg->n_ports; i++) {
1800 const char *name = br->cfg->ports[i]->name;
1801 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1802 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1806 if (bridge_get_controllers(br, NULL)
1807 && !shash_find(&new_ports, br->name)) {
1808 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1809 br->name, br->name);
1811 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1812 br->synth_local_port.n_interfaces = 1;
1813 br->synth_local_port.name = br->name;
1815 br->synth_local_iface.name = br->name;
1816 br->synth_local_iface.type = "internal";
1818 br->synth_local_ifacep = &br->synth_local_iface;
1820 shash_add(&new_ports, br->name, &br->synth_local_port);
1823 /* Get rid of deleted ports.
1824 * Get rid of deleted interfaces on ports that still exist.
1825 * Update 'cfg' of ports that still exist. */
1826 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1827 port->cfg = shash_find_data(&new_ports, port->name);
1831 port_del_ifaces(port);
1835 /* Create new ports.
1836 * Add new interfaces to existing ports. */
1837 SHASH_FOR_EACH (node, &new_ports) {
1838 struct port *port = port_lookup(br, node->name);
1840 struct ovsrec_port *cfg = node->data;
1841 port = port_create(br, cfg);
1843 port_add_ifaces(port);
1844 if (list_is_empty(&port->ifaces)) {
1845 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1846 br->name, port->name);
1850 shash_destroy(&new_ports);
1853 /* Initializes 'oc' appropriately as a management service controller for
1856 * The caller must free oc->target when it is no longer needed. */
1858 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1859 struct ofproto_controller *oc)
1861 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1862 oc->max_backoff = 0;
1863 oc->probe_interval = 60;
1864 oc->band = OFPROTO_OUT_OF_BAND;
1866 oc->burst_limit = 0;
1869 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1871 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1872 struct ofproto_controller *oc)
1874 oc->target = c->target;
1875 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1876 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1877 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1878 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1879 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1880 oc->burst_limit = (c->controller_burst_limit
1881 ? *c->controller_burst_limit : 0);
1884 /* Configures the IP stack for 'br''s local interface properly according to the
1885 * configuration in 'c'. */
1887 bridge_configure_local_iface_netdev(struct bridge *br,
1888 struct ovsrec_controller *c)
1890 struct netdev *netdev;
1891 struct in_addr mask, gateway;
1893 struct iface *local_iface;
1896 /* If there's no local interface or no IP address, give up. */
1897 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1898 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1902 /* Bring up the local interface. */
1903 netdev = local_iface->netdev;
1904 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1906 /* Configure the IP address and netmask. */
1907 if (!c->local_netmask
1908 || !inet_aton(c->local_netmask, &mask)
1910 mask.s_addr = guess_netmask(ip.s_addr);
1912 if (!netdev_set_in4(netdev, ip, mask)) {
1913 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1914 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1917 /* Configure the default gateway. */
1918 if (c->local_gateway
1919 && inet_aton(c->local_gateway, &gateway)
1920 && gateway.s_addr) {
1921 if (!netdev_add_router(netdev, gateway)) {
1922 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1923 br->name, IP_ARGS(&gateway.s_addr));
1929 bridge_configure_remotes(struct bridge *br,
1930 const struct sockaddr_in *managers, size_t n_managers)
1932 const char *disable_ib_str, *queue_id_str;
1933 bool disable_in_band = false;
1936 struct ovsrec_controller **controllers;
1937 size_t n_controllers;
1939 enum ofproto_fail_mode fail_mode;
1941 struct ofproto_controller *ocs;
1945 /* Check if we should disable in-band control on this bridge. */
1946 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1947 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1948 disable_in_band = true;
1951 /* Set OpenFlow queue ID for in-band control. */
1952 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1953 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1954 ofproto_set_in_band_queue(br->ofproto, queue_id);
1956 if (disable_in_band) {
1957 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1959 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1962 n_controllers = bridge_get_controllers(br, &controllers);
1964 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1967 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1968 for (i = 0; i < n_controllers; i++) {
1969 struct ovsrec_controller *c = controllers[i];
1971 if (!strncmp(c->target, "punix:", 6)
1972 || !strncmp(c->target, "unix:", 5)) {
1973 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1975 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
1976 * domain sockets and overwriting arbitrary local files. */
1977 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
1978 "controller \"%s\" due to possibility for remote "
1979 "exploit", br->name, c->target);
1983 bridge_configure_local_iface_netdev(br, c);
1984 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
1985 if (disable_in_band) {
1986 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
1991 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
1992 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
1995 /* Set the fail-mode. */
1996 fail_mode = !br->cfg->fail_mode
1997 || !strcmp(br->cfg->fail_mode, "standalone")
1998 ? OFPROTO_FAIL_STANDALONE
1999 : OFPROTO_FAIL_SECURE;
2000 ofproto_set_fail_mode(br->ofproto, fail_mode);
2002 /* Configure OpenFlow controller connection snooping. */
2003 if (!ofproto_has_snoops(br->ofproto)) {
2007 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2008 ovs_rundir(), br->name));
2009 ofproto_set_snoops(br->ofproto, &snoops);
2010 sset_destroy(&snoops);
2014 /* Port functions. */
2016 static struct port *
2017 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2021 port = xzalloc(sizeof *port);
2023 port->name = xstrdup(cfg->name);
2025 list_init(&port->ifaces);
2027 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2029 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2035 get_port_other_config(const struct ovsrec_port *port, const char *key,
2036 const char *default_value)
2040 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2042 return value ? value : default_value;
2046 get_interface_other_config(const struct ovsrec_interface *iface,
2047 const char *key, const char *default_value)
2051 value = get_ovsrec_key_value(&iface->header_,
2052 &ovsrec_interface_col_other_config, key);
2053 return value ? value : default_value;
2056 /* Deletes interfaces from 'port' that are no longer configured for it. */
2058 port_del_ifaces(struct port *port)
2060 struct iface *iface, *next;
2061 struct sset new_ifaces;
2064 /* Collect list of new interfaces. */
2065 sset_init(&new_ifaces);
2066 for (i = 0; i < port->cfg->n_interfaces; i++) {
2067 const char *name = port->cfg->interfaces[i]->name;
2068 const char *type = port->cfg->interfaces[i]->name;
2069 if (strcmp(type, "null")) {
2070 sset_add(&new_ifaces, name);
2074 /* Get rid of deleted interfaces. */
2075 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2076 if (!sset_contains(&new_ifaces, iface->name)) {
2077 iface_destroy(iface);
2081 sset_destroy(&new_ifaces);
2084 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2087 port_add_ifaces(struct port *port)
2089 struct shash new_ifaces;
2090 struct shash_node *node;
2093 /* Collect new ifaces. */
2094 shash_init(&new_ifaces);
2095 for (i = 0; i < port->cfg->n_interfaces; i++) {
2096 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2097 if (strcmp(cfg->type, "null")
2098 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2099 VLOG_WARN("port %s: %s specified twice as port interface",
2100 port->name, cfg->name);
2101 iface_set_ofport(cfg, -1);
2105 /* Create new interfaces.
2106 * Update interface types and 'cfg' members. */
2107 SHASH_FOR_EACH (node, &new_ifaces) {
2108 const struct ovsrec_interface *cfg = node->data;
2109 const char *iface_name = node->name;
2110 struct iface *iface;
2112 iface = iface_lookup(port->bridge, iface_name);
2114 iface = iface_create(port, cfg);
2119 /* Determine interface type. The local port always has type
2120 * "internal". Other ports take their type from the database and
2121 * default to "system" if none is specified. */
2122 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2123 : cfg->type[0] ? cfg->type
2126 shash_destroy(&new_ifaces);
2130 port_destroy(struct port *port)
2133 struct bridge *br = port->bridge;
2134 struct iface *iface, *next;
2137 ofproto_bundle_unregister(br->ofproto, port);
2140 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2141 iface_destroy(iface);
2144 hmap_remove(&br->ports, &port->hmap_node);
2146 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2153 static struct port *
2154 port_lookup(const struct bridge *br, const char *name)
2158 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2160 if (!strcmp(port->name, name)) {
2168 enable_lacp(struct port *port, bool *activep)
2170 if (!port->cfg->lacp) {
2171 /* XXX when LACP implementation has been sufficiently tested, enable by
2172 * default and make active on bonded ports. */
2174 } else if (!strcmp(port->cfg->lacp, "off")) {
2176 } else if (!strcmp(port->cfg->lacp, "active")) {
2179 } else if (!strcmp(port->cfg->lacp, "passive")) {
2183 VLOG_WARN("port %s: unknown LACP mode %s",
2184 port->name, port->cfg->lacp);
2189 static struct lacp_settings *
2190 port_configure_lacp(struct port *port, struct lacp_settings *s)
2192 const char *lacp_time;
2193 long long int custom_time;
2196 if (!enable_lacp(port, &s->active)) {
2200 s->name = port->name;
2201 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2203 /* Prefer bondable links if unspecified. */
2204 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2206 s->priority = (priority > 0 && priority <= UINT16_MAX
2208 : UINT16_MAX - !list_is_short(&port->ifaces));
2210 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2215 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2216 custom_time = atoi(lacp_time);
2217 if (!strcmp(lacp_time, "fast")) {
2218 s->lacp_time = LACP_TIME_FAST;
2219 } else if (!strcmp(lacp_time, "slow")) {
2220 s->lacp_time = LACP_TIME_SLOW;
2221 } else if (custom_time > 0) {
2222 s->lacp_time = LACP_TIME_CUSTOM;
2223 s->custom_time = custom_time;
2225 s->lacp_time = LACP_TIME_SLOW;
2232 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2234 int priority, portid, key;
2236 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2237 priority = atoi(get_interface_other_config(iface->cfg,
2238 "lacp-port-priority", "0"));
2239 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2242 if (portid <= 0 || portid > UINT16_MAX) {
2243 portid = iface->ofp_port;
2246 if (priority <= 0 || priority > UINT16_MAX) {
2247 priority = UINT16_MAX;
2250 if (key < 0 || key > UINT16_MAX) {
2254 s->name = iface->name;
2256 s->priority = priority;
2261 port_configure_bond(struct port *port, struct bond_settings *s,
2262 uint32_t *bond_stable_ids)
2264 const char *detect_s;
2265 struct iface *iface;
2266 int miimon_interval;
2269 s->name = port->name;
2270 s->balance = BM_SLB;
2271 if (port->cfg->bond_mode
2272 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2273 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2274 port->name, port->cfg->bond_mode,
2275 bond_mode_to_string(s->balance));
2277 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2278 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2279 "please use another bond type or disable flood_vlans",
2283 miimon_interval = atoi(get_port_other_config(port->cfg,
2284 "bond-miimon-interval", "0"));
2285 if (miimon_interval <= 0) {
2286 miimon_interval = 200;
2289 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2290 if (!strcmp(detect_s, "carrier")) {
2291 miimon_interval = 0;
2292 } else if (strcmp(detect_s, "miimon")) {
2293 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2294 "defaulting to carrier", port->name, detect_s);
2295 miimon_interval = 0;
2298 s->up_delay = MAX(0, port->cfg->bond_updelay);
2299 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2300 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2301 s->rebalance_interval = atoi(
2302 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2303 if (s->rebalance_interval < 1000) {
2304 s->rebalance_interval = 1000;
2307 s->fake_iface = port->cfg->bond_fake_iface;
2310 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2311 long long stable_id;
2313 stable_id = atoll(get_interface_other_config(iface->cfg,
2314 "bond-stable-id", "0"));
2315 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2316 stable_id = iface->ofp_port;
2318 bond_stable_ids[i++] = stable_id;
2320 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2324 /* Interface functions. */
2326 static struct iface *
2327 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2329 struct bridge *br = port->bridge;
2330 struct iface *iface;
2331 char *name = if_cfg->name;
2333 iface = xzalloc(sizeof *iface);
2335 iface->name = xstrdup(name);
2336 iface->ofp_port = -1;
2337 iface->tag = tag_create_random();
2338 iface->netdev = NULL;
2339 iface->cfg = if_cfg;
2341 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2343 list_push_back(&port->ifaces, &iface->port_elem);
2345 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2351 iface_destroy(struct iface *iface)
2354 struct port *port = iface->port;
2355 struct bridge *br = port->bridge;
2357 if (br->ofproto && iface->ofp_port >= 0) {
2358 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2361 if (iface->ofp_port >= 0) {
2362 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2365 list_remove(&iface->port_elem);
2366 hmap_remove(&br->iface_by_name, &iface->name_node);
2368 netdev_close(iface->netdev);
2375 static struct iface *
2376 iface_lookup(const struct bridge *br, const char *name)
2378 struct iface *iface;
2380 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2381 &br->iface_by_name) {
2382 if (!strcmp(iface->name, name)) {
2390 static struct iface *
2391 iface_find(const char *name)
2393 const struct bridge *br;
2395 HMAP_FOR_EACH (br, node, &all_bridges) {
2396 struct iface *iface = iface_lookup(br, name);
2405 static struct iface *
2406 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2408 struct iface *iface;
2410 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2411 hash_int(ofp_port, 0), &br->ifaces) {
2412 if (iface->ofp_port == ofp_port) {
2419 /* Set Ethernet address of 'iface', if one is specified in the configuration
2422 iface_set_mac(struct iface *iface)
2424 uint8_t ea[ETH_ADDR_LEN];
2426 if (!strcmp(iface->type, "internal")
2427 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2428 if (iface->ofp_port == OFPP_LOCAL) {
2429 VLOG_ERR("interface %s: ignoring mac in Interface record "
2430 "(use Bridge record to set local port's mac)",
2432 } else if (eth_addr_is_multicast(ea)) {
2433 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2436 int error = netdev_set_etheraddr(iface->netdev, ea);
2438 VLOG_ERR("interface %s: setting MAC failed (%s)",
2439 iface->name, strerror(error));
2445 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2447 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2449 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2450 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2454 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2456 * The value strings in '*shash' are taken directly from values[], not copied,
2457 * so the caller should not modify or free them. */
2459 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2460 struct shash *shash)
2465 for (i = 0; i < n; i++) {
2466 shash_add(shash, keys[i], values[i]);
2470 /* Creates 'keys' and 'values' arrays from 'shash'.
2472 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2473 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2474 * are populated with with strings taken directly from 'shash' and thus have
2475 * the same ownership of the key-value pairs in shash.
2478 shash_to_ovs_idl_map(struct shash *shash,
2479 char ***keys, char ***values, size_t *n)
2483 struct shash_node *sn;
2485 count = shash_count(shash);
2487 k = xmalloc(count * sizeof *k);
2488 v = xmalloc(count * sizeof *v);
2491 SHASH_FOR_EACH(sn, shash) {
2502 struct iface_delete_queues_cbdata {
2503 struct netdev *netdev;
2504 const struct ovsdb_datum *queues;
2508 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2510 union ovsdb_atom atom;
2512 atom.integer = target;
2513 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2517 iface_delete_queues(unsigned int queue_id,
2518 const struct shash *details OVS_UNUSED, void *cbdata_)
2520 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2522 if (!queue_ids_include(cbdata->queues, queue_id)) {
2523 netdev_delete_queue(cbdata->netdev, queue_id);
2528 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2530 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2531 netdev_set_qos(iface->netdev, NULL, NULL);
2533 struct iface_delete_queues_cbdata cbdata;
2534 struct shash details;
2538 /* Configure top-level Qos for 'iface'. */
2539 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2540 qos->n_other_config, &details);
2541 netdev_set_qos(iface->netdev, qos->type, &details);
2542 shash_destroy(&details);
2544 /* Deconfigure queues that were deleted. */
2545 cbdata.netdev = iface->netdev;
2546 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2548 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2550 /* Configure queues for 'iface'. */
2552 for (i = 0; i < qos->n_queues; i++) {
2553 const struct ovsrec_queue *queue = qos->value_queues[i];
2554 unsigned int queue_id = qos->key_queues[i];
2556 if (queue_id == 0) {
2560 shash_from_ovs_idl_map(queue->key_other_config,
2561 queue->value_other_config,
2562 queue->n_other_config, &details);
2563 netdev_set_queue(iface->netdev, queue_id, &details);
2564 shash_destroy(&details);
2567 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2568 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2569 "queue (queue 0). Packets not directed to a "
2570 "correctly configured queue may be dropped.",
2575 netdev_set_policing(iface->netdev,
2576 iface->cfg->ingress_policing_rate,
2577 iface->cfg->ingress_policing_burst);
2581 iface_configure_cfm(struct iface *iface)
2583 const struct ovsrec_interface *cfg = iface->cfg;
2584 struct cfm_settings s;
2585 uint16_t remote_mpid;
2587 if (!cfg->n_cfm_mpid || !cfg->n_cfm_remote_mpid) {
2588 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2592 s.mpid = *cfg->cfm_mpid;
2593 remote_mpid = *cfg->cfm_remote_mpid;
2594 s.remote_mpids = &remote_mpid;
2595 s.n_remote_mpids = 1;
2597 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2599 if (s.interval <= 0) {
2603 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2606 /* Read carrier or miimon status directly from 'iface''s netdev, according to
2607 * how 'iface''s port is configured.
2609 * Returns true if 'iface' is up, false otherwise. */
2611 iface_get_carrier(const struct iface *iface)
2614 return netdev_get_carrier(iface->netdev);
2617 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2618 * instead of obtaining it from the database. */
2620 iface_is_synthetic(const struct iface *iface)
2622 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2625 /* Port mirroring. */
2627 static struct mirror *
2628 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2632 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2633 if (uuid_equals(uuid, &m->uuid)) {
2641 bridge_configure_mirrors(struct bridge *br)
2643 const struct ovsdb_datum *mc;
2644 unsigned long *flood_vlans;
2645 struct mirror *m, *next;
2648 /* Get rid of deleted mirrors. */
2649 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2650 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2651 union ovsdb_atom atom;
2653 atom.uuid = m->uuid;
2654 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2659 /* Add new mirrors and reconfigure existing ones. */
2660 for (i = 0; i < br->cfg->n_mirrors; i++) {
2661 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2662 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2664 m = mirror_create(br, cfg);
2666 if (!mirror_configure(m, cfg)) {
2671 /* Update flooded vlans (for RSPAN). */
2672 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2673 br->cfg->n_flood_vlans);
2674 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2675 bitmap_free(flood_vlans);
2678 static struct mirror *
2679 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2683 m = xzalloc(sizeof *m);
2684 m->uuid = cfg->header_.uuid;
2685 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2687 m->name = xstrdup(cfg->name);
2693 mirror_destroy(struct mirror *m)
2696 struct bridge *br = m->bridge;
2699 ofproto_mirror_unregister(br->ofproto, m);
2702 hmap_remove(&br->mirrors, &m->hmap_node);
2709 mirror_collect_ports(struct mirror *m,
2710 struct ovsrec_port **in_ports, int n_in_ports,
2711 void ***out_portsp, size_t *n_out_portsp)
2713 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2714 size_t n_out_ports = 0;
2717 for (i = 0; i < n_in_ports; i++) {
2718 const char *name = in_ports[i]->name;
2719 struct port *port = port_lookup(m->bridge, name);
2721 out_ports[n_out_ports++] = port;
2723 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2724 "port %s", m->bridge->name, m->name, name);
2727 *out_portsp = out_ports;
2728 *n_out_portsp = n_out_ports;
2732 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2734 struct ofproto_mirror_settings s;
2737 if (strcmp(cfg->name, m->name)) {
2739 m->name = xstrdup(cfg->name);
2743 /* Get output port or VLAN. */
2744 if (cfg->output_port) {
2745 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2746 if (!s.out_bundle) {
2747 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2748 m->bridge->name, m->name);
2751 s.out_vlan = UINT16_MAX;
2753 if (cfg->output_vlan) {
2754 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2755 "output vlan; ignoring output vlan",
2756 m->bridge->name, m->name);
2758 } else if (cfg->output_vlan) {
2759 /* The database should prevent invalid VLAN values. */
2760 s.out_bundle = NULL;
2761 s.out_vlan = *cfg->output_vlan;
2763 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2764 m->bridge->name, m->name);
2768 /* Get port selection. */
2769 if (cfg->select_all) {
2770 size_t n_ports = hmap_count(&m->bridge->ports);
2771 void **ports = xmalloc(n_ports * sizeof *ports);
2776 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2786 /* Get ports, dropping ports that don't exist.
2787 * The IDL ensures that there are no duplicates. */
2788 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2789 &s.srcs, &s.n_srcs);
2790 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2791 &s.dsts, &s.n_dsts);
2794 /* Get VLAN selection. */
2795 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2798 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2801 if (s.srcs != s.dsts) {