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_clear_db_record(const struct ovsrec_interface *if_cfg);
199 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
200 static void iface_configure_cfm(struct iface *);
201 static void iface_refresh_cfm_stats(struct iface *);
202 static void iface_refresh_stats(struct iface *);
203 static void iface_refresh_status(struct iface *);
204 static bool iface_get_carrier(const struct iface *);
205 static bool iface_is_synthetic(const struct iface *);
207 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
209 static void shash_to_ovs_idl_map(struct shash *,
210 char ***keys, char ***values, size_t *n);
212 /* Public functions. */
214 /* Initializes the bridge module, configuring it to obtain its configuration
215 * from an OVSDB server accessed over 'remote', which should be a string in a
216 * form acceptable to ovsdb_idl_create(). */
218 bridge_init(const char *remote)
220 /* Create connection to database. */
221 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
222 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
224 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
225 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
226 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
227 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
228 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
229 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
230 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
232 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
233 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
235 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
236 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
238 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
239 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
240 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
241 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
242 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
243 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
244 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
245 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
246 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
247 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
248 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
249 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
251 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
252 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
253 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
254 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
258 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
260 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
262 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
264 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
266 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
267 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
268 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
269 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
270 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
272 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
274 /* Register unixctl commands. */
275 unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL);
276 unixctl_command_register("bridge/dump-flows", "bridge",
277 bridge_unixctl_dump_flows, NULL);
278 unixctl_command_register("bridge/reconnect", "[bridge]",
279 bridge_unixctl_reconnect, NULL);
288 struct bridge *br, *next_br;
290 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
293 ovsdb_idl_destroy(idl);
296 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
297 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
298 * responsible for freeing '*managersp' (with free()).
300 * You may be asking yourself "why does ovs-vswitchd care?", because
301 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
302 * should not be and in fact is not directly involved in that. But
303 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
304 * it has to tell in-band control where the managers are to enable that.
305 * (Thus, only managers connected in-band are collected.)
308 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
309 struct sockaddr_in **managersp, size_t *n_managersp)
311 struct sockaddr_in *managers = NULL;
312 size_t n_managers = 0;
316 /* Collect all of the potential targets from the "targets" columns of the
317 * rows pointed to by "manager_options", excluding any that are
320 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
321 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
323 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
324 sset_find_and_delete(&targets, m->target);
326 sset_add(&targets, m->target);
330 /* Now extract the targets' IP addresses. */
331 if (!sset_is_empty(&targets)) {
334 managers = xmalloc(sset_count(&targets) * sizeof *managers);
335 SSET_FOR_EACH (target, &targets) {
336 struct sockaddr_in *sin = &managers[n_managers];
338 if ((!strncmp(target, "tcp:", 4)
339 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
340 (!strncmp(target, "ssl:", 4)
341 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
346 sset_destroy(&targets);
348 *managersp = managers;
349 *n_managersp = n_managers;
353 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
355 struct sockaddr_in *managers;
356 struct bridge *br, *next;
357 int sflow_bridge_number;
360 COVERAGE_INC(bridge_reconfigure);
362 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
363 * iface"s according to 'ovs_cfg', with only very minimal configuration
366 * This is purely an update to bridge data structures. Nothing is pushed
367 * down to ofproto or lower layers. */
368 add_del_bridges(ovs_cfg);
369 HMAP_FOR_EACH (br, node, &all_bridges) {
370 bridge_add_del_ports(br);
373 /* Delete all datapaths and datapath ports that are no longer configured.
375 * The kernel will reject any attempt to add a given port to a datapath if
376 * that port already belongs to a different datapath, so we must do all
377 * port deletions before any port additions. A datapath always has a
378 * "local port" so we must delete not-configured datapaths too. */
379 bridge_del_ofprotos();
380 HMAP_FOR_EACH (br, node, &all_bridges) {
382 bridge_del_ofproto_ports(br);
386 /* Create datapaths and datapath ports that are missing.
388 * After this is done, we have our final set of bridges, ports, and
389 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
390 * has at least one iface, every "struct iface" has a valid ofp_port and
392 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
393 if (!br->ofproto && !bridge_add_ofprotos(br)) {
397 HMAP_FOR_EACH (br, node, &all_bridges) {
398 bridge_refresh_ofp_port(br);
399 bridge_add_ofproto_ports(br);
402 /* Complete the configuration. */
403 sflow_bridge_number = 0;
404 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
405 HMAP_FOR_EACH (br, node, &all_bridges) {
408 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
411 port_configure(port);
413 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
414 iface_configure_cfm(iface);
415 iface_configure_qos(iface, port->cfg->qos);
416 iface_set_mac(iface);
419 bridge_configure_mirrors(br);
420 bridge_configure_datapath_id(br);
421 bridge_configure_flow_eviction_threshold(br);
422 bridge_configure_forward_bpdu(br);
423 bridge_configure_remotes(br, managers, n_managers);
424 bridge_configure_netflow(br);
425 bridge_configure_sflow(br, &sflow_bridge_number);
429 /* ovs-vswitchd has completed initialization, so allow the process that
430 * forked us to exit successfully. */
431 daemonize_complete();
434 /* Iterate over all ofprotos and delete any of them that do not have a
435 * configured bridge or that are the wrong type. */
437 bridge_del_ofprotos(void)
445 ofproto_enumerate_types(&types);
446 SSET_FOR_EACH (type, &types) {
449 ofproto_enumerate_names(type, &names);
450 SSET_FOR_EACH (name, &names) {
451 struct bridge *br = bridge_lookup(name);
452 if (!br || strcmp(type, br->type)) {
453 ofproto_delete(name, type);
457 sset_destroy(&names);
458 sset_destroy(&types);
462 bridge_add_ofprotos(struct bridge *br)
464 int error = ofproto_create(br->name, br->type, &br->ofproto);
466 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
473 port_configure(struct port *port)
475 const struct ovsrec_port *cfg = port->cfg;
476 struct bond_settings bond_settings;
477 struct lacp_settings lacp_settings;
478 struct ofproto_bundle_settings s;
486 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
487 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
488 s.slaves[s.n_slaves++] = iface->ofp_port;
494 if (list_is_short(&port->ifaces)) {
495 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
499 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
500 * they even work as-is. But they have not been tested. */
501 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
506 /* Get VLAN trunks. */
509 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
513 if (cfg->vlan_mode) {
514 if (!strcmp(cfg->vlan_mode, "access")) {
515 s.vlan_mode = PORT_VLAN_ACCESS;
516 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
517 s.vlan_mode = PORT_VLAN_TRUNK;
518 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
519 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
520 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
521 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
523 /* This "can't happen" because ovsdb-server should prevent it. */
524 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
525 s.vlan_mode = PORT_VLAN_TRUNK;
529 s.vlan_mode = PORT_VLAN_ACCESS;
531 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
535 s.vlan_mode = PORT_VLAN_TRUNK;
539 /* Get LACP settings. */
540 s.lacp = port_configure_lacp(port, &lacp_settings);
544 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
545 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
546 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
549 s.lacp_slaves = NULL;
552 /* Get bond settings. */
553 if (s.n_slaves > 1) {
554 s.bond = &bond_settings;
555 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
556 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
559 s.bond_stable_ids = NULL;
561 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
562 netdev_set_miimon_interval(iface->netdev, 0);
567 ofproto_bundle_register(port->bridge->ofproto, port, &s);
573 free(s.bond_stable_ids);
576 /* Pick local port hardware address and datapath ID for 'br'. */
578 bridge_configure_datapath_id(struct bridge *br)
580 uint8_t ea[ETH_ADDR_LEN];
582 struct iface *local_iface;
583 struct iface *hw_addr_iface;
586 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
587 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
589 int error = netdev_set_etheraddr(local_iface->netdev, ea);
591 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
592 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
593 "Ethernet address: %s",
594 br->name, strerror(error));
597 memcpy(br->ea, ea, ETH_ADDR_LEN);
599 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
600 ofproto_set_datapath_id(br->ofproto, dpid);
602 dpid_string = xasprintf("%016"PRIx64, dpid);
603 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
607 /* Set NetFlow configuration on 'br'. */
609 bridge_configure_netflow(struct bridge *br)
611 struct ovsrec_netflow *cfg = br->cfg->netflow;
612 struct netflow_options opts;
615 ofproto_set_netflow(br->ofproto, NULL);
619 memset(&opts, 0, sizeof opts);
621 /* Get default NetFlow configuration from datapath.
622 * Apply overrides from 'cfg'. */
623 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
624 if (cfg->engine_type) {
625 opts.engine_type = *cfg->engine_type;
627 if (cfg->engine_id) {
628 opts.engine_id = *cfg->engine_id;
631 /* Configure active timeout interval. */
632 opts.active_timeout = cfg->active_timeout;
633 if (!opts.active_timeout) {
634 opts.active_timeout = -1;
635 } else if (opts.active_timeout < 0) {
636 VLOG_WARN("bridge %s: active timeout interval set to negative "
637 "value, using default instead (%d seconds)", br->name,
638 NF_ACTIVE_TIMEOUT_DEFAULT);
639 opts.active_timeout = -1;
642 /* Add engine ID to interface number to disambiguate bridgs? */
643 opts.add_id_to_iface = cfg->add_id_to_interface;
644 if (opts.add_id_to_iface) {
645 if (opts.engine_id > 0x7f) {
646 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
647 "another vswitch, choose an engine id less than 128",
650 if (hmap_count(&br->ports) > 508) {
651 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
652 "another port when more than 508 ports are used",
658 sset_init(&opts.collectors);
659 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
662 if (ofproto_set_netflow(br->ofproto, &opts)) {
663 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
665 sset_destroy(&opts.collectors);
668 /* Set sFlow configuration on 'br'. */
670 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
672 const struct ovsrec_sflow *cfg = br->cfg->sflow;
673 struct ovsrec_controller **controllers;
674 struct ofproto_sflow_options oso;
675 size_t n_controllers;
679 ofproto_set_sflow(br->ofproto, NULL);
683 memset(&oso, 0, sizeof oso);
685 sset_init(&oso.targets);
686 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
688 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
690 oso.sampling_rate = *cfg->sampling;
693 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
695 oso.polling_interval = *cfg->polling;
698 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
700 oso.header_len = *cfg->header;
703 oso.sub_id = (*sflow_bridge_number)++;
704 oso.agent_device = cfg->agent;
706 oso.control_ip = NULL;
707 n_controllers = bridge_get_controllers(br, &controllers);
708 for (i = 0; i < n_controllers; i++) {
709 if (controllers[i]->local_ip) {
710 oso.control_ip = controllers[i]->local_ip;
714 ofproto_set_sflow(br->ofproto, &oso);
716 sset_destroy(&oso.targets);
720 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
722 const struct port *port = port_lookup(br, name);
723 return port && port_is_bond_fake_iface(port);
727 port_is_bond_fake_iface(const struct port *port)
729 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
733 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
735 struct bridge *br, *next;
739 /* Collect new bridges' names and types. */
741 for (i = 0; i < cfg->n_bridges; i++) {
742 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
743 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
744 VLOG_WARN("bridge %s specified twice", br_cfg->name);
748 /* Get rid of deleted bridges or those whose types have changed.
749 * Update 'cfg' of bridges that still exist. */
750 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
751 br->cfg = shash_find_data(&new_br, br->name);
752 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
753 br->cfg->datapath_type))) {
758 /* Add new bridges. */
759 for (i = 0; i < cfg->n_bridges; i++) {
760 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
761 struct bridge *br = bridge_lookup(br_cfg->name);
763 bridge_create(br_cfg);
767 shash_destroy(&new_br);
770 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
773 * The kernel will reject any attempt to add a given port to a datapath if that
774 * port already belongs to a different datapath, so we must do all port
775 * deletions before any port additions. */
777 bridge_del_ofproto_ports(struct bridge *br)
779 struct ofproto_port_dump dump;
780 struct ofproto_port ofproto_port;
782 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
783 const char *name = ofproto_port.name;
788 /* Ignore the local port. We can't change it anyhow. */
789 if (!strcmp(name, br->name)) {
793 /* Get the type that 'ofproto_port' should have (ordinarily the
794 * type of its corresponding iface) or NULL if it should be
796 iface = iface_lookup(br, name);
797 type = (iface ? iface->type
798 : bridge_has_bond_fake_iface(br, name) ? "internal"
801 /* If it's the wrong type then delete the ofproto port. */
803 && !strcmp(ofproto_port.type, type)
804 && (!iface || !iface->netdev
805 || !strcmp(netdev_get_type(iface->netdev), type))) {
808 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
810 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
811 br->name, name, strerror(error));
814 netdev_close(iface->netdev);
815 iface->netdev = NULL;
821 iface_set_ofp_port(struct iface *iface, int ofp_port)
823 struct bridge *br = iface->port->bridge;
825 assert(iface->ofp_port < 0 && ofp_port >= 0);
826 iface->ofp_port = ofp_port;
827 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
828 iface_set_ofport(iface->cfg, ofp_port);
832 bridge_refresh_ofp_port(struct bridge *br)
834 struct ofproto_port_dump dump;
835 struct ofproto_port ofproto_port;
838 /* Clear all the "ofp_port"es. */
839 hmap_clear(&br->ifaces);
840 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
843 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
844 iface->ofp_port = -1;
848 /* Obtain the correct "ofp_port"s from ofproto. */
849 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
850 struct iface *iface = iface_lookup(br, ofproto_port.name);
852 if (iface->ofp_port >= 0) {
853 VLOG_WARN("bridge %s: interface %s reported twice",
854 br->name, ofproto_port.name);
855 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
856 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
857 br->name, ofproto_port.ofp_port);
859 iface_set_ofp_port(iface, ofproto_port.ofp_port);
865 /* Add an ofproto port for any "struct iface" that doesn't have one.
866 * Delete any "struct iface" for which this fails.
867 * Delete any "struct port" that thereby ends up with no ifaces. */
869 bridge_add_ofproto_ports(struct bridge *br)
871 struct port *port, *next_port;
873 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
874 struct iface *iface, *next_iface;
875 struct ofproto_port ofproto_port;
877 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
880 /* Open the netdev. */
881 if (!iface->netdev) {
882 error = netdev_open(iface->name, iface->type, &iface->netdev);
884 VLOG_WARN("could not open network device %s (%s)",
885 iface->name, strerror(error));
891 /* Configure the netdev. */
896 shash_from_ovs_idl_map(iface->cfg->key_options,
897 iface->cfg->value_options,
898 iface->cfg->n_options, &args);
899 error = netdev_set_config(iface->netdev, &args);
900 shash_destroy(&args);
903 VLOG_WARN("could not configure network device %s (%s)",
904 iface->name, strerror(error));
905 netdev_close(iface->netdev);
906 iface->netdev = NULL;
910 /* Add the port, if necessary. */
911 if (iface->netdev && iface->ofp_port < 0) {
915 error = ofproto_port_add(br->ofproto, iface->netdev,
918 iface_set_ofp_port(iface, ofp_port);
920 netdev_close(iface->netdev);
921 iface->netdev = NULL;
925 /* Populate stats columns in new Interface rows. */
926 if (iface->netdev && !iface->cfg->mtu) {
927 iface_refresh_stats(iface);
928 iface_refresh_status(iface);
931 /* Delete the iface if we failed. */
932 if (iface->netdev && iface->ofp_port >= 0) {
933 VLOG_DBG("bridge %s: interface %s is on port %d",
934 br->name, iface->name, iface->ofp_port);
937 VLOG_ERR("bridge %s: missing %s interface, dropping",
938 br->name, iface->name);
940 /* We already reported a related error, don't bother
943 iface_clear_db_record(iface->cfg);
944 iface_destroy(iface);
947 if (list_is_empty(&port->ifaces)) {
948 VLOG_WARN("%s port has no interfaces, dropping", port->name);
953 /* Add bond fake iface if necessary. */
954 if (port_is_bond_fake_iface(port)) {
955 if (ofproto_port_query_by_name(br->ofproto, port->name,
957 struct netdev *netdev;
960 error = netdev_open(port->name, "internal", &netdev);
962 ofproto_port_add(br->ofproto, netdev, NULL);
963 netdev_close(netdev);
965 VLOG_WARN("could not open network device %s (%s)",
966 port->name, strerror(error));
969 /* Already exists, nothing to do. */
970 ofproto_port_destroy(&ofproto_port);
977 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
978 const struct ovsdb_idl_column *column,
981 const struct ovsdb_datum *datum;
982 union ovsdb_atom atom;
985 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
986 atom.string = (char *) key;
987 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
988 return idx == UINT_MAX ? NULL : datum->values[idx].string;
992 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
994 return get_ovsrec_key_value(&br_cfg->header_,
995 &ovsrec_bridge_col_other_config, key);
998 /* Set Flow eviction threshold */
1000 bridge_configure_flow_eviction_threshold(struct bridge *br)
1002 const char *threshold_str;
1005 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1006 if (threshold_str) {
1007 threshold = strtoul(threshold_str, NULL, 10);
1009 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1011 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1014 /* Set forward BPDU option. */
1016 bridge_configure_forward_bpdu(struct bridge *br)
1018 const char *forward_bpdu_str;
1019 bool forward_bpdu = false;
1021 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1022 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1023 forward_bpdu = true;
1025 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1029 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1030 struct iface **hw_addr_iface)
1034 bool found_addr = false;
1037 *hw_addr_iface = NULL;
1039 /* Did the user request a particular MAC? */
1040 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1041 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1042 if (eth_addr_is_multicast(ea)) {
1043 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1044 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1045 } else if (eth_addr_is_zero(ea)) {
1046 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1052 /* Otherwise choose the minimum non-local MAC address among all of the
1054 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1055 uint8_t iface_ea[ETH_ADDR_LEN];
1056 struct iface *candidate;
1057 struct iface *iface;
1059 /* Mirror output ports don't participate. */
1060 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1064 /* Choose the MAC address to represent the port. */
1066 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1067 /* Find the interface with this Ethernet address (if any) so that
1068 * we can provide the correct devname to the caller. */
1069 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1070 uint8_t candidate_ea[ETH_ADDR_LEN];
1071 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1072 && eth_addr_equals(iface_ea, candidate_ea)) {
1077 /* Choose the interface whose MAC address will represent the port.
1078 * The Linux kernel bonding code always chooses the MAC address of
1079 * the first slave added to a bond, and the Fedora networking
1080 * scripts always add slaves to a bond in alphabetical order, so
1081 * for compatibility we choose the interface with the name that is
1082 * first in alphabetical order. */
1083 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1084 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1089 /* The local port doesn't count (since we're trying to choose its
1090 * MAC address anyway). */
1091 if (iface->ofp_port == OFPP_LOCAL) {
1096 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1102 /* Compare against our current choice. */
1103 if (!eth_addr_is_multicast(iface_ea) &&
1104 !eth_addr_is_local(iface_ea) &&
1105 !eth_addr_is_reserved(iface_ea) &&
1106 !eth_addr_is_zero(iface_ea) &&
1107 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1109 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1110 *hw_addr_iface = iface;
1115 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1116 br->name, ETH_ADDR_ARGS(ea));
1118 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1119 *hw_addr_iface = NULL;
1120 VLOG_WARN("bridge %s: using default bridge Ethernet "
1121 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1125 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1126 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1127 * an interface on 'br', then that interface must be passed in as
1128 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1129 * 'hw_addr_iface' must be passed in as a null pointer. */
1131 bridge_pick_datapath_id(struct bridge *br,
1132 const uint8_t bridge_ea[ETH_ADDR_LEN],
1133 struct iface *hw_addr_iface)
1136 * The procedure for choosing a bridge MAC address will, in the most
1137 * ordinary case, also choose a unique MAC that we can use as a datapath
1138 * ID. In some special cases, though, multiple bridges will end up with
1139 * the same MAC address. This is OK for the bridges, but it will confuse
1140 * the OpenFlow controller, because each datapath needs a unique datapath
1143 * Datapath IDs must be unique. It is also very desirable that they be
1144 * stable from one run to the next, so that policy set on a datapath
1147 const char *datapath_id;
1150 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1151 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1155 if (hw_addr_iface) {
1157 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1159 * A bridge whose MAC address is taken from a VLAN network device
1160 * (that is, a network device created with vconfig(8) or similar
1161 * tool) will have the same MAC address as a bridge on the VLAN
1162 * device's physical network device.
1164 * Handle this case by hashing the physical network device MAC
1165 * along with the VLAN identifier.
1167 uint8_t buf[ETH_ADDR_LEN + 2];
1168 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1169 buf[ETH_ADDR_LEN] = vlan >> 8;
1170 buf[ETH_ADDR_LEN + 1] = vlan;
1171 return dpid_from_hash(buf, sizeof buf);
1174 * Assume that this bridge's MAC address is unique, since it
1175 * doesn't fit any of the cases we handle specially.
1180 * A purely internal bridge, that is, one that has no non-virtual
1181 * network devices on it at all, is more difficult because it has no
1182 * natural unique identifier at all.
1184 * When the host is a XenServer, we handle this case by hashing the
1185 * host's UUID with the name of the bridge. Names of bridges are
1186 * persistent across XenServer reboots, although they can be reused if
1187 * an internal network is destroyed and then a new one is later
1188 * created, so this is fairly effective.
1190 * When the host is not a XenServer, we punt by using a random MAC
1191 * address on each run.
1193 const char *host_uuid = xenserver_get_host_uuid();
1195 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1196 dpid = dpid_from_hash(combined, strlen(combined));
1202 return eth_addr_to_uint64(bridge_ea);
1206 dpid_from_hash(const void *data, size_t n)
1208 uint8_t hash[SHA1_DIGEST_SIZE];
1210 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1211 sha1_bytes(data, n, hash);
1212 eth_addr_mark_random(hash);
1213 return eth_addr_to_uint64(hash);
1217 iface_refresh_status(struct iface *iface)
1221 enum netdev_flags flags;
1228 if (iface_is_synthetic(iface)) {
1234 if (!netdev_get_status(iface->netdev, &sh)) {
1236 char **keys, **values;
1238 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1239 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1244 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1247 shash_destroy_free_data(&sh);
1249 error = netdev_get_flags(iface->netdev, &flags);
1251 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1254 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1257 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1259 ovsrec_interface_set_duplex(iface->cfg,
1260 netdev_features_is_full_duplex(current)
1262 /* warning: uint64_t -> int64_t conversion */
1263 bps = netdev_features_to_bps(current);
1264 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1267 ovsrec_interface_set_duplex(iface->cfg, NULL);
1268 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1271 ovsrec_interface_set_link_state(iface->cfg,
1272 iface_get_carrier(iface) ? "up" : "down");
1274 error = netdev_get_mtu(iface->netdev, &mtu);
1277 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1280 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1284 /* Writes 'iface''s CFM statistics to the database. */
1286 iface_refresh_cfm_stats(struct iface *iface)
1288 const struct ovsrec_interface *cfg = iface->cfg;
1290 const uint64_t *rmps;
1293 if (iface_is_synthetic(iface)) {
1297 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1300 bool fault_bool = fault;
1301 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1303 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1306 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1307 iface->ofp_port, &rmps, &n_rmps);
1309 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1312 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1317 iface_refresh_lacp_stats(struct iface *iface)
1319 struct ofproto *ofproto = iface->port->bridge->ofproto;
1320 int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
1321 int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
1325 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
1331 iface_refresh_stats(struct iface *iface)
1333 #define IFACE_STATS \
1334 IFACE_STAT(rx_packets, "rx_packets") \
1335 IFACE_STAT(tx_packets, "tx_packets") \
1336 IFACE_STAT(rx_bytes, "rx_bytes") \
1337 IFACE_STAT(tx_bytes, "tx_bytes") \
1338 IFACE_STAT(rx_dropped, "rx_dropped") \
1339 IFACE_STAT(tx_dropped, "tx_dropped") \
1340 IFACE_STAT(rx_errors, "rx_errors") \
1341 IFACE_STAT(tx_errors, "tx_errors") \
1342 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1343 IFACE_STAT(rx_over_errors, "rx_over_err") \
1344 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1345 IFACE_STAT(collisions, "collisions")
1347 #define IFACE_STAT(MEMBER, NAME) NAME,
1348 static char *keys[] = { IFACE_STATS };
1350 int64_t values[ARRAY_SIZE(keys)];
1353 struct netdev_stats stats;
1355 if (iface_is_synthetic(iface)) {
1359 /* Intentionally ignore return value, since errors will set 'stats' to
1360 * all-1s, and we will deal with that correctly below. */
1361 netdev_get_stats(iface->netdev, &stats);
1363 /* Copy statistics into values[] array. */
1365 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1368 assert(i == ARRAY_SIZE(keys));
1370 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1375 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1379 /* Use other-config:enable-system-stats by preference. */
1380 enable = get_ovsrec_key_value(&cfg->header_,
1381 &ovsrec_open_vswitch_col_other_config,
1382 "enable-statistics");
1384 return !strcmp(enable, "true");
1387 /* Disable by default. */
1392 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1394 struct ovsdb_datum datum;
1398 if (enable_system_stats(cfg)) {
1399 get_system_stats(&stats);
1402 ovsdb_datum_from_shash(&datum, &stats);
1403 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1407 static inline const char *
1408 nx_role_to_str(enum nx_role role)
1413 case NX_ROLE_MASTER:
1418 return "*** INVALID ROLE ***";
1423 refresh_controller_status(void)
1427 const struct ovsrec_controller *cfg;
1431 /* Accumulate status for controllers on all bridges. */
1432 HMAP_FOR_EACH (br, node, &all_bridges) {
1433 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1436 /* Update each controller in the database with current status. */
1437 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1438 struct ofproto_controller_info *cinfo =
1439 shash_find_data(&info, cfg->target);
1442 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1443 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1444 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1445 (char **) cinfo->pairs.values,
1448 ovsrec_controller_set_is_connected(cfg, false);
1449 ovsrec_controller_set_role(cfg, NULL);
1450 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1454 ofproto_free_ofproto_controller_info(&info);
1458 refresh_cfm_stats(void)
1460 static struct ovsdb_idl_txn *txn = NULL;
1465 txn = ovsdb_idl_txn_create(idl);
1467 HMAP_FOR_EACH (br, node, &all_bridges) {
1468 struct iface *iface;
1470 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1471 iface_refresh_cfm_stats(iface);
1476 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1477 ovsdb_idl_txn_destroy(txn);
1485 const struct ovsrec_open_vswitch *cfg;
1487 bool datapath_destroyed;
1488 bool database_changed;
1491 /* (Re)configure if necessary. */
1492 database_changed = ovsdb_idl_run(idl);
1493 if (ovsdb_idl_is_lock_contended(idl)) {
1494 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1495 struct bridge *br, *next_br;
1497 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1498 "disabling this process until it goes away");
1500 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1504 } else if (!ovsdb_idl_has_lock(idl)) {
1507 cfg = ovsrec_open_vswitch_first(idl);
1509 /* Let each bridge do the work that it needs to do. */
1510 datapath_destroyed = false;
1511 HMAP_FOR_EACH (br, node, &all_bridges) {
1512 int error = ofproto_run(br->ofproto);
1514 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1515 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1516 "forcing reconfiguration", br->name);
1517 datapath_destroyed = true;
1521 /* Re-configure SSL. We do this on every trip through the main loop,
1522 * instead of just when the database changes, because the contents of the
1523 * key and certificate files can change without the database changing.
1525 * We do this before bridge_reconfigure() because that function might
1526 * initiate SSL connections and thus requires SSL to be configured. */
1527 if (cfg && cfg->ssl) {
1528 const struct ovsrec_ssl *ssl = cfg->ssl;
1530 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1531 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1534 if (database_changed || datapath_destroyed) {
1536 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1538 bridge_reconfigure(cfg);
1540 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1541 ovsdb_idl_txn_commit(txn);
1542 ovsdb_idl_txn_destroy(txn); /* XXX */
1544 /* We still need to reconfigure to avoid dangling pointers to
1545 * now-destroyed ovsrec structures inside bridge data. */
1546 static const struct ovsrec_open_vswitch null_cfg;
1548 bridge_reconfigure(&null_cfg);
1552 /* Refresh system and interface stats if necessary. */
1553 if (time_msec() >= stats_timer) {
1555 struct ovsdb_idl_txn *txn;
1557 txn = ovsdb_idl_txn_create(idl);
1558 HMAP_FOR_EACH (br, node, &all_bridges) {
1561 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1562 struct iface *iface;
1564 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1565 iface_refresh_stats(iface);
1566 iface_refresh_status(iface);
1570 refresh_system_stats(cfg);
1571 refresh_controller_status();
1572 ovsdb_idl_txn_commit(txn);
1573 ovsdb_idl_txn_destroy(txn); /* XXX */
1576 stats_timer = time_msec() + STATS_INTERVAL;
1579 if (time_msec() >= db_limiter) {
1580 struct ovsdb_idl_txn *txn;
1581 bool changed = false;
1583 txn = ovsdb_idl_txn_create(idl);
1584 HMAP_FOR_EACH (br, node, &all_bridges) {
1587 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1588 struct iface *iface;
1590 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1591 /* XXX: Eventually we need to remove the lacp_current flag
1592 * from the database so that we can completely get rid of
1593 * this rate limiter code. */
1594 changed = iface_refresh_lacp_stats(iface) || changed;
1600 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1603 ovsdb_idl_txn_commit(txn);
1604 ovsdb_idl_txn_destroy(txn);
1607 refresh_cfm_stats();
1613 ovsdb_idl_wait(idl);
1614 if (!hmap_is_empty(&all_bridges)) {
1617 HMAP_FOR_EACH (br, node, &all_bridges) {
1618 ofproto_wait(br->ofproto);
1620 poll_timer_wait_until(stats_timer);
1622 if (db_limiter > time_msec()) {
1623 poll_timer_wait_until(db_limiter);
1628 /* QoS unixctl user interface functions. */
1630 struct qos_unixctl_show_cbdata {
1632 struct iface *iface;
1636 qos_unixctl_show_cb(unsigned int queue_id,
1637 const struct shash *details,
1640 struct qos_unixctl_show_cbdata *data = aux;
1641 struct ds *ds = data->ds;
1642 struct iface *iface = data->iface;
1643 struct netdev_queue_stats stats;
1644 struct shash_node *node;
1647 ds_put_cstr(ds, "\n");
1649 ds_put_format(ds, "Queue %u:\n", queue_id);
1651 ds_put_cstr(ds, "Default:\n");
1654 SHASH_FOR_EACH (node, details) {
1655 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1658 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1660 if (stats.tx_packets != UINT64_MAX) {
1661 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1664 if (stats.tx_bytes != UINT64_MAX) {
1665 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1668 if (stats.tx_errors != UINT64_MAX) {
1669 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1672 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1673 queue_id, strerror(error));
1678 qos_unixctl_show(struct unixctl_conn *conn,
1679 const char *args, void *aux OVS_UNUSED)
1681 struct ds ds = DS_EMPTY_INITIALIZER;
1682 struct shash sh = SHASH_INITIALIZER(&sh);
1683 struct iface *iface;
1685 struct shash_node *node;
1686 struct qos_unixctl_show_cbdata data;
1689 iface = iface_find(args);
1691 unixctl_command_reply(conn, 501, "no such interface");
1695 netdev_get_qos(iface->netdev, &type, &sh);
1697 if (*type != '\0') {
1698 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1700 SHASH_FOR_EACH (node, &sh) {
1701 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1706 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1709 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1711 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1713 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1714 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1717 shash_destroy_free_data(&sh);
1721 /* Bridge reconfiguration functions. */
1723 bridge_create(const struct ovsrec_bridge *br_cfg)
1727 assert(!bridge_lookup(br_cfg->name));
1728 br = xzalloc(sizeof *br);
1730 br->name = xstrdup(br_cfg->name);
1731 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1734 /* Derive the default Ethernet address from the bridge's UUID. This should
1735 * be unique and it will be stable between ovs-vswitchd runs. */
1736 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1737 eth_addr_mark_random(br->default_ea);
1739 hmap_init(&br->ports);
1740 hmap_init(&br->ifaces);
1741 hmap_init(&br->iface_by_name);
1742 hmap_init(&br->mirrors);
1744 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1748 bridge_destroy(struct bridge *br)
1751 struct mirror *mirror, *next_mirror;
1752 struct port *port, *next_port;
1754 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1757 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1758 mirror_destroy(mirror);
1760 hmap_remove(&all_bridges, &br->node);
1761 ofproto_destroy(br->ofproto);
1762 hmap_destroy(&br->ifaces);
1763 hmap_destroy(&br->ports);
1764 hmap_destroy(&br->iface_by_name);
1765 hmap_destroy(&br->mirrors);
1772 static struct bridge *
1773 bridge_lookup(const char *name)
1777 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1778 if (!strcmp(br->name, name)) {
1785 /* Handle requests for a listing of all flows known by the OpenFlow
1786 * stack, including those normally hidden. */
1788 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1789 const char *args, void *aux OVS_UNUSED)
1794 br = bridge_lookup(args);
1796 unixctl_command_reply(conn, 501, "Unknown bridge");
1801 ofproto_get_all_flows(br->ofproto, &results);
1803 unixctl_command_reply(conn, 200, ds_cstr(&results));
1804 ds_destroy(&results);
1807 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1808 * connections and reconnect. If BRIDGE is not specified, then all bridges
1809 * drop their controller connections and reconnect. */
1811 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1812 const char *args, void *aux OVS_UNUSED)
1815 if (args[0] != '\0') {
1816 br = bridge_lookup(args);
1818 unixctl_command_reply(conn, 501, "Unknown bridge");
1821 ofproto_reconnect_controllers(br->ofproto);
1823 HMAP_FOR_EACH (br, node, &all_bridges) {
1824 ofproto_reconnect_controllers(br->ofproto);
1827 unixctl_command_reply(conn, 200, NULL);
1831 bridge_get_controllers(const struct bridge *br,
1832 struct ovsrec_controller ***controllersp)
1834 struct ovsrec_controller **controllers;
1835 size_t n_controllers;
1837 controllers = br->cfg->controller;
1838 n_controllers = br->cfg->n_controller;
1840 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1846 *controllersp = controllers;
1848 return n_controllers;
1851 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1852 * those configured in 'br->cfg'. */
1854 bridge_add_del_ports(struct bridge *br)
1856 struct port *port, *next;
1857 struct shash_node *node;
1858 struct shash new_ports;
1861 /* Collect new ports. */
1862 shash_init(&new_ports);
1863 for (i = 0; i < br->cfg->n_ports; i++) {
1864 const char *name = br->cfg->ports[i]->name;
1865 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1866 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1870 if (bridge_get_controllers(br, NULL)
1871 && !shash_find(&new_ports, br->name)) {
1872 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1873 br->name, br->name);
1875 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1876 br->synth_local_port.n_interfaces = 1;
1877 br->synth_local_port.name = br->name;
1879 br->synth_local_iface.name = br->name;
1880 br->synth_local_iface.type = "internal";
1882 br->synth_local_ifacep = &br->synth_local_iface;
1884 shash_add(&new_ports, br->name, &br->synth_local_port);
1887 /* Get rid of deleted ports.
1888 * Get rid of deleted interfaces on ports that still exist.
1889 * Update 'cfg' of ports that still exist. */
1890 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1891 port->cfg = shash_find_data(&new_ports, port->name);
1895 port_del_ifaces(port);
1899 /* Create new ports.
1900 * Add new interfaces to existing ports. */
1901 SHASH_FOR_EACH (node, &new_ports) {
1902 struct port *port = port_lookup(br, node->name);
1904 struct ovsrec_port *cfg = node->data;
1905 port = port_create(br, cfg);
1907 port_add_ifaces(port);
1908 if (list_is_empty(&port->ifaces)) {
1909 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1910 br->name, port->name);
1914 shash_destroy(&new_ports);
1917 /* Initializes 'oc' appropriately as a management service controller for
1920 * The caller must free oc->target when it is no longer needed. */
1922 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1923 struct ofproto_controller *oc)
1925 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1926 oc->max_backoff = 0;
1927 oc->probe_interval = 60;
1928 oc->band = OFPROTO_OUT_OF_BAND;
1930 oc->burst_limit = 0;
1933 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1935 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1936 struct ofproto_controller *oc)
1938 oc->target = c->target;
1939 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1940 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1941 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1942 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1943 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1944 oc->burst_limit = (c->controller_burst_limit
1945 ? *c->controller_burst_limit : 0);
1948 /* Configures the IP stack for 'br''s local interface properly according to the
1949 * configuration in 'c'. */
1951 bridge_configure_local_iface_netdev(struct bridge *br,
1952 struct ovsrec_controller *c)
1954 struct netdev *netdev;
1955 struct in_addr mask, gateway;
1957 struct iface *local_iface;
1960 /* If there's no local interface or no IP address, give up. */
1961 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1962 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1966 /* Bring up the local interface. */
1967 netdev = local_iface->netdev;
1968 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1970 /* Configure the IP address and netmask. */
1971 if (!c->local_netmask
1972 || !inet_aton(c->local_netmask, &mask)
1974 mask.s_addr = guess_netmask(ip.s_addr);
1976 if (!netdev_set_in4(netdev, ip, mask)) {
1977 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1978 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1981 /* Configure the default gateway. */
1982 if (c->local_gateway
1983 && inet_aton(c->local_gateway, &gateway)
1984 && gateway.s_addr) {
1985 if (!netdev_add_router(netdev, gateway)) {
1986 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1987 br->name, IP_ARGS(&gateway.s_addr));
1993 bridge_configure_remotes(struct bridge *br,
1994 const struct sockaddr_in *managers, size_t n_managers)
1996 const char *disable_ib_str, *queue_id_str;
1997 bool disable_in_band = false;
2000 struct ovsrec_controller **controllers;
2001 size_t n_controllers;
2003 enum ofproto_fail_mode fail_mode;
2005 struct ofproto_controller *ocs;
2009 /* Check if we should disable in-band control on this bridge. */
2010 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2011 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2012 disable_in_band = true;
2015 /* Set OpenFlow queue ID for in-band control. */
2016 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2017 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2018 ofproto_set_in_band_queue(br->ofproto, queue_id);
2020 if (disable_in_band) {
2021 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2023 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2026 n_controllers = bridge_get_controllers(br, &controllers);
2028 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2031 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2032 for (i = 0; i < n_controllers; i++) {
2033 struct ovsrec_controller *c = controllers[i];
2035 if (!strncmp(c->target, "punix:", 6)
2036 || !strncmp(c->target, "unix:", 5)) {
2037 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2039 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2040 * domain sockets and overwriting arbitrary local files. */
2041 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
2042 "controller \"%s\" due to possibility for remote "
2043 "exploit", br->name, c->target);
2047 bridge_configure_local_iface_netdev(br, c);
2048 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2049 if (disable_in_band) {
2050 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2055 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2056 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2059 /* Set the fail-mode. */
2060 fail_mode = !br->cfg->fail_mode
2061 || !strcmp(br->cfg->fail_mode, "standalone")
2062 ? OFPROTO_FAIL_STANDALONE
2063 : OFPROTO_FAIL_SECURE;
2064 ofproto_set_fail_mode(br->ofproto, fail_mode);
2066 /* Configure OpenFlow controller connection snooping. */
2067 if (!ofproto_has_snoops(br->ofproto)) {
2071 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2072 ovs_rundir(), br->name));
2073 ofproto_set_snoops(br->ofproto, &snoops);
2074 sset_destroy(&snoops);
2078 /* Port functions. */
2080 static struct port *
2081 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2085 port = xzalloc(sizeof *port);
2087 port->name = xstrdup(cfg->name);
2089 list_init(&port->ifaces);
2091 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2093 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2099 get_port_other_config(const struct ovsrec_port *port, const char *key,
2100 const char *default_value)
2104 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2106 return value ? value : default_value;
2110 get_interface_other_config(const struct ovsrec_interface *iface,
2111 const char *key, const char *default_value)
2115 value = get_ovsrec_key_value(&iface->header_,
2116 &ovsrec_interface_col_other_config, key);
2117 return value ? value : default_value;
2120 /* Deletes interfaces from 'port' that are no longer configured for it. */
2122 port_del_ifaces(struct port *port)
2124 struct iface *iface, *next;
2125 struct sset new_ifaces;
2128 /* Collect list of new interfaces. */
2129 sset_init(&new_ifaces);
2130 for (i = 0; i < port->cfg->n_interfaces; i++) {
2131 const char *name = port->cfg->interfaces[i]->name;
2132 const char *type = port->cfg->interfaces[i]->name;
2133 if (strcmp(type, "null")) {
2134 sset_add(&new_ifaces, name);
2138 /* Get rid of deleted interfaces. */
2139 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2140 if (!sset_contains(&new_ifaces, iface->name)) {
2141 iface_destroy(iface);
2145 sset_destroy(&new_ifaces);
2148 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2151 port_add_ifaces(struct port *port)
2153 struct shash new_ifaces;
2154 struct shash_node *node;
2157 /* Collect new ifaces. */
2158 shash_init(&new_ifaces);
2159 for (i = 0; i < port->cfg->n_interfaces; i++) {
2160 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2161 if (strcmp(cfg->type, "null")
2162 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2163 VLOG_WARN("port %s: %s specified twice as port interface",
2164 port->name, cfg->name);
2165 iface_clear_db_record(cfg);
2169 /* Create new interfaces.
2170 * Update interface types and 'cfg' members. */
2171 SHASH_FOR_EACH (node, &new_ifaces) {
2172 const struct ovsrec_interface *cfg = node->data;
2173 const char *iface_name = node->name;
2174 struct iface *iface;
2176 iface = iface_lookup(port->bridge, iface_name);
2178 iface = iface_create(port, cfg);
2183 /* Determine interface type. The local port always has type
2184 * "internal". Other ports take their type from the database and
2185 * default to "system" if none is specified. */
2186 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2187 : cfg->type[0] ? cfg->type
2190 shash_destroy(&new_ifaces);
2194 port_destroy(struct port *port)
2197 struct bridge *br = port->bridge;
2198 struct iface *iface, *next;
2201 ofproto_bundle_unregister(br->ofproto, port);
2204 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2205 iface_destroy(iface);
2208 hmap_remove(&br->ports, &port->hmap_node);
2210 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2217 static struct port *
2218 port_lookup(const struct bridge *br, const char *name)
2222 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2224 if (!strcmp(port->name, name)) {
2232 enable_lacp(struct port *port, bool *activep)
2234 if (!port->cfg->lacp) {
2235 /* XXX when LACP implementation has been sufficiently tested, enable by
2236 * default and make active on bonded ports. */
2238 } else if (!strcmp(port->cfg->lacp, "off")) {
2240 } else if (!strcmp(port->cfg->lacp, "active")) {
2243 } else if (!strcmp(port->cfg->lacp, "passive")) {
2247 VLOG_WARN("port %s: unknown LACP mode %s",
2248 port->name, port->cfg->lacp);
2253 static struct lacp_settings *
2254 port_configure_lacp(struct port *port, struct lacp_settings *s)
2256 const char *lacp_time;
2257 long long int custom_time;
2260 if (!enable_lacp(port, &s->active)) {
2264 s->name = port->name;
2265 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2267 /* Prefer bondable links if unspecified. */
2268 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2270 s->priority = (priority > 0 && priority <= UINT16_MAX
2272 : UINT16_MAX - !list_is_short(&port->ifaces));
2274 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2279 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2280 custom_time = atoi(lacp_time);
2281 if (!strcmp(lacp_time, "fast")) {
2282 s->lacp_time = LACP_TIME_FAST;
2283 } else if (!strcmp(lacp_time, "slow")) {
2284 s->lacp_time = LACP_TIME_SLOW;
2285 } else if (custom_time > 0) {
2286 s->lacp_time = LACP_TIME_CUSTOM;
2287 s->custom_time = custom_time;
2289 s->lacp_time = LACP_TIME_SLOW;
2296 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2298 int priority, portid, key;
2300 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2301 priority = atoi(get_interface_other_config(iface->cfg,
2302 "lacp-port-priority", "0"));
2303 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2306 if (portid <= 0 || portid > UINT16_MAX) {
2307 portid = iface->ofp_port;
2310 if (priority <= 0 || priority > UINT16_MAX) {
2311 priority = UINT16_MAX;
2314 if (key < 0 || key > UINT16_MAX) {
2318 s->name = iface->name;
2320 s->priority = priority;
2325 port_configure_bond(struct port *port, struct bond_settings *s,
2326 uint32_t *bond_stable_ids)
2328 const char *detect_s;
2329 struct iface *iface;
2330 int miimon_interval;
2333 s->name = port->name;
2334 s->balance = BM_SLB;
2335 if (port->cfg->bond_mode
2336 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2337 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2338 port->name, port->cfg->bond_mode,
2339 bond_mode_to_string(s->balance));
2341 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2342 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2343 "please use another bond type or disable flood_vlans",
2347 miimon_interval = atoi(get_port_other_config(port->cfg,
2348 "bond-miimon-interval", "0"));
2349 if (miimon_interval <= 0) {
2350 miimon_interval = 200;
2353 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2354 if (!strcmp(detect_s, "carrier")) {
2355 miimon_interval = 0;
2356 } else if (strcmp(detect_s, "miimon")) {
2357 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2358 "defaulting to carrier", port->name, detect_s);
2359 miimon_interval = 0;
2362 s->up_delay = MAX(0, port->cfg->bond_updelay);
2363 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2364 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2365 s->rebalance_interval = atoi(
2366 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2367 if (s->rebalance_interval < 1000) {
2368 s->rebalance_interval = 1000;
2371 s->fake_iface = port->cfg->bond_fake_iface;
2374 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2375 long long stable_id;
2377 stable_id = atoll(get_interface_other_config(iface->cfg,
2378 "bond-stable-id", "0"));
2379 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2380 stable_id = iface->ofp_port;
2382 bond_stable_ids[i++] = stable_id;
2384 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2388 /* Interface functions. */
2390 static struct iface *
2391 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2393 struct bridge *br = port->bridge;
2394 struct iface *iface;
2395 char *name = if_cfg->name;
2397 iface = xzalloc(sizeof *iface);
2399 iface->name = xstrdup(name);
2400 iface->ofp_port = -1;
2401 iface->tag = tag_create_random();
2402 iface->netdev = NULL;
2403 iface->cfg = if_cfg;
2405 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2407 list_push_back(&port->ifaces, &iface->port_elem);
2409 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2415 iface_destroy(struct iface *iface)
2418 struct port *port = iface->port;
2419 struct bridge *br = port->bridge;
2421 if (br->ofproto && iface->ofp_port >= 0) {
2422 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2425 if (iface->ofp_port >= 0) {
2426 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2429 list_remove(&iface->port_elem);
2430 hmap_remove(&br->iface_by_name, &iface->name_node);
2432 netdev_close(iface->netdev);
2439 static struct iface *
2440 iface_lookup(const struct bridge *br, const char *name)
2442 struct iface *iface;
2444 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2445 &br->iface_by_name) {
2446 if (!strcmp(iface->name, name)) {
2454 static struct iface *
2455 iface_find(const char *name)
2457 const struct bridge *br;
2459 HMAP_FOR_EACH (br, node, &all_bridges) {
2460 struct iface *iface = iface_lookup(br, name);
2469 static struct iface *
2470 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2472 struct iface *iface;
2474 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2475 hash_int(ofp_port, 0), &br->ifaces) {
2476 if (iface->ofp_port == ofp_port) {
2483 /* Set Ethernet address of 'iface', if one is specified in the configuration
2486 iface_set_mac(struct iface *iface)
2488 uint8_t ea[ETH_ADDR_LEN];
2490 if (!strcmp(iface->type, "internal")
2491 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2492 if (iface->ofp_port == OFPP_LOCAL) {
2493 VLOG_ERR("interface %s: ignoring mac in Interface record "
2494 "(use Bridge record to set local port's mac)",
2496 } else if (eth_addr_is_multicast(ea)) {
2497 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2500 int error = netdev_set_etheraddr(iface->netdev, ea);
2502 VLOG_ERR("interface %s: setting MAC failed (%s)",
2503 iface->name, strerror(error));
2509 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2511 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2513 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2514 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2518 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
2519 * sets the "ofport" field to -1.
2521 * This is appropriate when 'if_cfg''s interface cannot be created or is
2522 * otherwise invalid. */
2524 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
2526 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2527 iface_set_ofport(if_cfg, -1);
2528 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
2529 ovsrec_interface_set_admin_state(if_cfg, NULL);
2530 ovsrec_interface_set_duplex(if_cfg, NULL);
2531 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
2532 ovsrec_interface_set_link_state(if_cfg, NULL);
2533 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
2534 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
2535 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
2536 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
2537 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
2541 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2543 * The value strings in '*shash' are taken directly from values[], not copied,
2544 * so the caller should not modify or free them. */
2546 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2547 struct shash *shash)
2552 for (i = 0; i < n; i++) {
2553 shash_add(shash, keys[i], values[i]);
2557 /* Creates 'keys' and 'values' arrays from 'shash'.
2559 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2560 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2561 * are populated with with strings taken directly from 'shash' and thus have
2562 * the same ownership of the key-value pairs in shash.
2565 shash_to_ovs_idl_map(struct shash *shash,
2566 char ***keys, char ***values, size_t *n)
2570 struct shash_node *sn;
2572 count = shash_count(shash);
2574 k = xmalloc(count * sizeof *k);
2575 v = xmalloc(count * sizeof *v);
2578 SHASH_FOR_EACH(sn, shash) {
2589 struct iface_delete_queues_cbdata {
2590 struct netdev *netdev;
2591 const struct ovsdb_datum *queues;
2595 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2597 union ovsdb_atom atom;
2599 atom.integer = target;
2600 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2604 iface_delete_queues(unsigned int queue_id,
2605 const struct shash *details OVS_UNUSED, void *cbdata_)
2607 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2609 if (!queue_ids_include(cbdata->queues, queue_id)) {
2610 netdev_delete_queue(cbdata->netdev, queue_id);
2615 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2617 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2618 netdev_set_qos(iface->netdev, NULL, NULL);
2620 struct iface_delete_queues_cbdata cbdata;
2621 struct shash details;
2625 /* Configure top-level Qos for 'iface'. */
2626 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2627 qos->n_other_config, &details);
2628 netdev_set_qos(iface->netdev, qos->type, &details);
2629 shash_destroy(&details);
2631 /* Deconfigure queues that were deleted. */
2632 cbdata.netdev = iface->netdev;
2633 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2635 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2637 /* Configure queues for 'iface'. */
2639 for (i = 0; i < qos->n_queues; i++) {
2640 const struct ovsrec_queue *queue = qos->value_queues[i];
2641 unsigned int queue_id = qos->key_queues[i];
2643 if (queue_id == 0) {
2647 shash_from_ovs_idl_map(queue->key_other_config,
2648 queue->value_other_config,
2649 queue->n_other_config, &details);
2650 netdev_set_queue(iface->netdev, queue_id, &details);
2651 shash_destroy(&details);
2654 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2655 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2656 "queue (queue 0). Packets not directed to a "
2657 "correctly configured queue may be dropped.",
2662 netdev_set_policing(iface->netdev,
2663 iface->cfg->ingress_policing_rate,
2664 iface->cfg->ingress_policing_burst);
2668 iface_configure_cfm(struct iface *iface)
2670 const struct ovsrec_interface *cfg = iface->cfg;
2671 const char *extended_str;
2672 struct cfm_settings s;
2674 if (!cfg->n_cfm_mpid) {
2675 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2679 s.mpid = *cfg->cfm_mpid;
2680 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2682 if (s.interval <= 0) {
2686 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
2688 s.extended = !strcasecmp("true", extended_str);
2690 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2693 /* Read carrier or miimon status directly from 'iface''s netdev, according to
2694 * how 'iface''s port is configured.
2696 * Returns true if 'iface' is up, false otherwise. */
2698 iface_get_carrier(const struct iface *iface)
2701 return netdev_get_carrier(iface->netdev);
2704 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2705 * instead of obtaining it from the database. */
2707 iface_is_synthetic(const struct iface *iface)
2709 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2712 /* Port mirroring. */
2714 static struct mirror *
2715 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2719 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2720 if (uuid_equals(uuid, &m->uuid)) {
2728 bridge_configure_mirrors(struct bridge *br)
2730 const struct ovsdb_datum *mc;
2731 unsigned long *flood_vlans;
2732 struct mirror *m, *next;
2735 /* Get rid of deleted mirrors. */
2736 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2737 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2738 union ovsdb_atom atom;
2740 atom.uuid = m->uuid;
2741 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2746 /* Add new mirrors and reconfigure existing ones. */
2747 for (i = 0; i < br->cfg->n_mirrors; i++) {
2748 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2749 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2751 m = mirror_create(br, cfg);
2753 if (!mirror_configure(m, cfg)) {
2758 /* Update flooded vlans (for RSPAN). */
2759 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2760 br->cfg->n_flood_vlans);
2761 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2762 bitmap_free(flood_vlans);
2765 static struct mirror *
2766 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2770 m = xzalloc(sizeof *m);
2771 m->uuid = cfg->header_.uuid;
2772 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2774 m->name = xstrdup(cfg->name);
2780 mirror_destroy(struct mirror *m)
2783 struct bridge *br = m->bridge;
2786 ofproto_mirror_unregister(br->ofproto, m);
2789 hmap_remove(&br->mirrors, &m->hmap_node);
2796 mirror_collect_ports(struct mirror *m,
2797 struct ovsrec_port **in_ports, int n_in_ports,
2798 void ***out_portsp, size_t *n_out_portsp)
2800 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2801 size_t n_out_ports = 0;
2804 for (i = 0; i < n_in_ports; i++) {
2805 const char *name = in_ports[i]->name;
2806 struct port *port = port_lookup(m->bridge, name);
2808 out_ports[n_out_ports++] = port;
2810 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2811 "port %s", m->bridge->name, m->name, name);
2814 *out_portsp = out_ports;
2815 *n_out_portsp = n_out_ports;
2819 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2821 struct ofproto_mirror_settings s;
2824 if (strcmp(cfg->name, m->name)) {
2826 m->name = xstrdup(cfg->name);
2830 /* Get output port or VLAN. */
2831 if (cfg->output_port) {
2832 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2833 if (!s.out_bundle) {
2834 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2835 m->bridge->name, m->name);
2838 s.out_vlan = UINT16_MAX;
2840 if (cfg->output_vlan) {
2841 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2842 "output vlan; ignoring output vlan",
2843 m->bridge->name, m->name);
2845 } else if (cfg->output_vlan) {
2846 /* The database should prevent invalid VLAN values. */
2847 s.out_bundle = NULL;
2848 s.out_vlan = *cfg->output_vlan;
2850 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2851 m->bridge->name, m->name);
2855 /* Get port selection. */
2856 if (cfg->select_all) {
2857 size_t n_ports = hmap_count(&m->bridge->ports);
2858 void **ports = xmalloc(n_ports * sizeof *ports);
2863 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2873 /* Get ports, dropping ports that don't exist.
2874 * The IDL ensures that there are no duplicates. */
2875 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2876 &s.srcs, &s.n_srcs);
2877 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2878 &s.dsts, &s.n_dsts);
2881 /* Get VLAN selection. */
2882 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2885 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2888 if (s.srcs != s.dsts) {