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_netflow(struct bridge *);
152 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
153 static void bridge_configure_remotes(struct bridge *,
154 const struct sockaddr_in *managers,
156 static void bridge_pick_local_hw_addr(struct bridge *,
157 uint8_t ea[ETH_ADDR_LEN],
158 struct iface **hw_addr_iface);
159 static uint64_t bridge_pick_datapath_id(struct bridge *,
160 const uint8_t bridge_ea[ETH_ADDR_LEN],
161 struct iface *hw_addr_iface);
162 static uint64_t dpid_from_hash(const void *, size_t nbytes);
163 static bool bridge_has_bond_fake_iface(const struct bridge *,
165 static bool port_is_bond_fake_iface(const struct port *);
167 static unixctl_cb_func qos_unixctl_show;
169 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
170 static void port_add_ifaces(struct port *);
171 static void port_del_ifaces(struct port *);
172 static void port_destroy(struct port *);
173 static struct port *port_lookup(const struct bridge *, const char *name);
174 static void port_configure(struct port *);
175 static struct lacp_settings *port_configure_lacp(struct port *,
176 struct lacp_settings *);
177 static void port_configure_bond(struct port *, struct bond_settings *,
178 uint32_t *bond_stable_ids);
180 static void bridge_configure_mirrors(struct bridge *);
181 static struct mirror *mirror_create(struct bridge *,
182 const struct ovsrec_mirror *);
183 static void mirror_destroy(struct mirror *);
184 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
186 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
187 static struct iface *iface_create(struct port *port,
188 const struct ovsrec_interface *if_cfg);
189 static void iface_destroy(struct iface *);
190 static struct iface *iface_lookup(const struct bridge *, const char *name);
191 static struct iface *iface_find(const char *name);
192 static struct iface *iface_from_ofp_port(const struct bridge *,
194 static void iface_set_mac(struct iface *);
195 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
196 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
197 static void iface_configure_cfm(struct iface *);
198 static bool iface_refresh_cfm_stats(struct iface *);
199 static void iface_refresh_stats(struct iface *);
200 static void iface_refresh_status(struct iface *);
201 static bool iface_get_carrier(const struct iface *);
202 static bool iface_is_synthetic(const struct iface *);
204 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
206 static void shash_to_ovs_idl_map(struct shash *,
207 char ***keys, char ***values, size_t *n);
209 /* Public functions. */
211 /* Initializes the bridge module, configuring it to obtain its configuration
212 * from an OVSDB server accessed over 'remote', which should be a string in a
213 * form acceptable to ovsdb_idl_create(). */
215 bridge_init(const char *remote)
217 /* Create connection to database. */
218 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
219 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
221 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
222 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
223 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
224 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
225 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
226 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
227 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
229 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
230 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
232 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
233 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
235 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
236 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
237 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
238 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
239 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
240 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
241 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
242 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
243 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
245 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
246 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
247 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
248 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
250 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
252 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
254 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
258 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
260 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
261 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
262 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
263 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
264 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
266 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
268 /* Register unixctl commands. */
269 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
270 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
272 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
282 struct bridge *br, *next_br;
284 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
287 ovsdb_idl_destroy(idl);
290 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
291 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
292 * responsible for freeing '*managersp' (with free()).
294 * You may be asking yourself "why does ovs-vswitchd care?", because
295 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
296 * should not be and in fact is not directly involved in that. But
297 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
298 * it has to tell in-band control where the managers are to enable that.
299 * (Thus, only managers connected in-band are collected.)
302 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
303 struct sockaddr_in **managersp, size_t *n_managersp)
305 struct sockaddr_in *managers = NULL;
306 size_t n_managers = 0;
310 /* Collect all of the potential targets from the "targets" columns of the
311 * rows pointed to by "manager_options", excluding any that are
314 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
315 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
317 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
318 sset_find_and_delete(&targets, m->target);
320 sset_add(&targets, m->target);
324 /* Now extract the targets' IP addresses. */
325 if (!sset_is_empty(&targets)) {
328 managers = xmalloc(sset_count(&targets) * sizeof *managers);
329 SSET_FOR_EACH (target, &targets) {
330 struct sockaddr_in *sin = &managers[n_managers];
332 if ((!strncmp(target, "tcp:", 4)
333 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
334 (!strncmp(target, "ssl:", 4)
335 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
340 sset_destroy(&targets);
342 *managersp = managers;
343 *n_managersp = n_managers;
347 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
349 struct sockaddr_in *managers;
350 struct bridge *br, *next;
351 int sflow_bridge_number;
354 COVERAGE_INC(bridge_reconfigure);
356 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
357 * iface"s according to 'ovs_cfg', with only very minimal configuration
360 * This is purely an update to bridge data structures. Nothing is pushed
361 * down to ofproto or lower layers. */
362 add_del_bridges(ovs_cfg);
363 HMAP_FOR_EACH (br, node, &all_bridges) {
364 bridge_add_del_ports(br);
367 /* Delete all datapaths and datapath ports that are no longer configured.
369 * The kernel will reject any attempt to add a given port to a datapath if
370 * that port already belongs to a different datapath, so we must do all
371 * port deletions before any port additions. A datapath always has a
372 * "local port" so we must delete not-configured datapaths too. */
373 bridge_del_ofprotos();
374 HMAP_FOR_EACH (br, node, &all_bridges) {
376 bridge_del_ofproto_ports(br);
380 /* Create datapaths and datapath ports that are missing.
382 * After this is done, we have our final set of bridges, ports, and
383 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
384 * has at least one iface, every "struct iface" has a valid ofp_port and
386 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
387 if (!br->ofproto && !bridge_add_ofprotos(br)) {
391 HMAP_FOR_EACH (br, node, &all_bridges) {
392 bridge_refresh_ofp_port(br);
393 bridge_add_ofproto_ports(br);
396 /* Complete the configuration. */
397 sflow_bridge_number = 0;
398 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
399 HMAP_FOR_EACH (br, node, &all_bridges) {
402 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
405 port_configure(port);
407 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
408 iface_configure_cfm(iface);
409 iface_configure_qos(iface, port->cfg->qos);
410 iface_set_mac(iface);
413 bridge_configure_mirrors(br);
414 bridge_configure_datapath_id(br);
415 bridge_configure_remotes(br, managers, n_managers);
416 bridge_configure_netflow(br);
417 bridge_configure_sflow(br, &sflow_bridge_number);
421 /* ovs-vswitchd has completed initialization, so allow the process that
422 * forked us to exit successfully. */
423 daemonize_complete();
426 /* Iterate over all ofprotos and delete any of them that do not have a
427 * configured bridge or that are the wrong type. */
429 bridge_del_ofprotos(void)
437 ofproto_enumerate_types(&types);
438 SSET_FOR_EACH (type, &types) {
441 ofproto_enumerate_names(type, &names);
442 SSET_FOR_EACH (name, &names) {
443 struct bridge *br = bridge_lookup(name);
444 if (!br || strcmp(type, br->type)) {
445 ofproto_delete(name, type);
449 sset_destroy(&names);
450 sset_destroy(&types);
454 bridge_add_ofprotos(struct bridge *br)
456 int error = ofproto_create(br->name, br->type, &br->ofproto);
458 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
465 port_configure(struct port *port)
467 const struct ovsrec_port *cfg = port->cfg;
468 struct bond_settings bond_settings;
469 struct lacp_settings lacp_settings;
470 struct ofproto_bundle_settings s;
478 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
479 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
480 s.slaves[s.n_slaves++] = iface->ofp_port;
486 if (list_is_short(&port->ifaces)) {
487 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
489 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan);
492 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
493 * they even work as-is. But they have not been tested. */
494 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
499 /* Get VLAN trunks. */
501 if (s.vlan < 0 && cfg->n_trunks) {
502 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
503 } else if (s.vlan >= 0 && cfg->n_trunks) {
504 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
508 /* Get LACP settings. */
509 s.lacp = port_configure_lacp(port, &lacp_settings);
513 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
514 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
515 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
518 s.lacp_slaves = NULL;
521 /* Get bond settings. */
522 if (s.n_slaves > 1) {
523 s.bond = &bond_settings;
524 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
525 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
528 s.bond_stable_ids = NULL;
530 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
531 netdev_set_miimon_interval(iface->netdev, 0);
536 ofproto_bundle_register(port->bridge->ofproto, port, &s);
542 free(s.bond_stable_ids);
545 /* Pick local port hardware address and datapath ID for 'br'. */
547 bridge_configure_datapath_id(struct bridge *br)
549 uint8_t ea[ETH_ADDR_LEN];
551 struct iface *local_iface;
552 struct iface *hw_addr_iface;
555 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
556 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
558 int error = netdev_set_etheraddr(local_iface->netdev, ea);
560 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
561 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
562 "Ethernet address: %s",
563 br->name, strerror(error));
566 memcpy(br->ea, ea, ETH_ADDR_LEN);
568 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
569 ofproto_set_datapath_id(br->ofproto, dpid);
571 dpid_string = xasprintf("%016"PRIx64, dpid);
572 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
576 /* Set NetFlow configuration on 'br'. */
578 bridge_configure_netflow(struct bridge *br)
580 struct ovsrec_netflow *cfg = br->cfg->netflow;
581 struct netflow_options opts;
584 ofproto_set_netflow(br->ofproto, NULL);
588 memset(&opts, 0, sizeof opts);
590 /* Get default NetFlow configuration from datapath.
591 * Apply overrides from 'cfg'. */
592 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
593 if (cfg->engine_type) {
594 opts.engine_type = *cfg->engine_type;
596 if (cfg->engine_id) {
597 opts.engine_id = *cfg->engine_id;
600 /* Configure active timeout interval. */
601 opts.active_timeout = cfg->active_timeout;
602 if (!opts.active_timeout) {
603 opts.active_timeout = -1;
604 } else if (opts.active_timeout < 0) {
605 VLOG_WARN("bridge %s: active timeout interval set to negative "
606 "value, using default instead (%d seconds)", br->name,
607 NF_ACTIVE_TIMEOUT_DEFAULT);
608 opts.active_timeout = -1;
611 /* Add engine ID to interface number to disambiguate bridgs? */
612 opts.add_id_to_iface = cfg->add_id_to_interface;
613 if (opts.add_id_to_iface) {
614 if (opts.engine_id > 0x7f) {
615 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
616 "another vswitch, choose an engine id less than 128",
619 if (hmap_count(&br->ports) > 508) {
620 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
621 "another port when more than 508 ports are used",
627 sset_init(&opts.collectors);
628 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
631 if (ofproto_set_netflow(br->ofproto, &opts)) {
632 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
634 sset_destroy(&opts.collectors);
637 /* Set sFlow configuration on 'br'. */
639 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
641 const struct ovsrec_sflow *cfg = br->cfg->sflow;
642 struct ovsrec_controller **controllers;
643 struct ofproto_sflow_options oso;
644 size_t n_controllers;
648 ofproto_set_sflow(br->ofproto, NULL);
652 memset(&oso, 0, sizeof oso);
654 sset_init(&oso.targets);
655 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
657 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
659 oso.sampling_rate = *cfg->sampling;
662 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
664 oso.polling_interval = *cfg->polling;
667 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
669 oso.header_len = *cfg->header;
672 oso.sub_id = (*sflow_bridge_number)++;
673 oso.agent_device = cfg->agent;
675 oso.control_ip = NULL;
676 n_controllers = bridge_get_controllers(br, &controllers);
677 for (i = 0; i < n_controllers; i++) {
678 if (controllers[i]->local_ip) {
679 oso.control_ip = controllers[i]->local_ip;
683 ofproto_set_sflow(br->ofproto, &oso);
685 sset_destroy(&oso.targets);
689 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
691 const struct port *port = port_lookup(br, name);
692 return port && port_is_bond_fake_iface(port);
696 port_is_bond_fake_iface(const struct port *port)
698 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
702 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
704 struct bridge *br, *next;
708 /* Collect new bridges' names and types. */
710 for (i = 0; i < cfg->n_bridges; i++) {
711 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
712 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
713 VLOG_WARN("bridge %s specified twice", br_cfg->name);
717 /* Get rid of deleted bridges or those whose types have changed.
718 * Update 'cfg' of bridges that still exist. */
719 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
720 br->cfg = shash_find_data(&new_br, br->name);
721 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
722 br->cfg->datapath_type))) {
727 /* Add new bridges. */
728 for (i = 0; i < cfg->n_bridges; i++) {
729 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
730 struct bridge *br = bridge_lookup(br_cfg->name);
732 bridge_create(br_cfg);
736 shash_destroy(&new_br);
739 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
742 * The kernel will reject any attempt to add a given port to a datapath if that
743 * port already belongs to a different datapath, so we must do all port
744 * deletions before any port additions. */
746 bridge_del_ofproto_ports(struct bridge *br)
748 struct ofproto_port_dump dump;
749 struct ofproto_port ofproto_port;
751 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
752 const char *name = ofproto_port.name;
757 /* Ignore the local port. We can't change it anyhow. */
758 if (!strcmp(name, br->name)) {
762 /* Get the type that 'ofproto_port' should have (ordinarily the
763 * type of its corresponding iface) or NULL if it should be
765 iface = iface_lookup(br, name);
766 type = (iface ? iface->type
767 : bridge_has_bond_fake_iface(br, name) ? "internal"
770 /* If it's the wrong type then delete the ofproto port. */
772 && !strcmp(ofproto_port.type, type)
773 && (!iface || !iface->netdev
774 || !strcmp(netdev_get_type(iface->netdev), type))) {
777 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
779 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
780 br->name, name, strerror(error));
783 netdev_close(iface->netdev);
784 iface->netdev = NULL;
790 iface_set_ofp_port(struct iface *iface, int ofp_port)
792 struct bridge *br = iface->port->bridge;
794 assert(iface->ofp_port < 0 && ofp_port >= 0);
795 iface->ofp_port = ofp_port;
796 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
797 iface_set_ofport(iface->cfg, ofp_port);
801 bridge_refresh_ofp_port(struct bridge *br)
803 struct ofproto_port_dump dump;
804 struct ofproto_port ofproto_port;
807 /* Clear all the "ofp_port"es. */
808 hmap_clear(&br->ifaces);
809 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
812 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
813 iface->ofp_port = -1;
817 /* Obtain the correct "ofp_port"s from ofproto. */
818 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
819 struct iface *iface = iface_lookup(br, ofproto_port.name);
821 if (iface->ofp_port >= 0) {
822 VLOG_WARN("bridge %s: interface %s reported twice",
823 br->name, ofproto_port.name);
824 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
825 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
826 br->name, ofproto_port.ofp_port);
828 iface_set_ofp_port(iface, ofproto_port.ofp_port);
834 /* Add an ofproto port for any "struct iface" that doesn't have one.
835 * Delete any "struct iface" for which this fails.
836 * Delete any "struct port" that thereby ends up with no ifaces. */
838 bridge_add_ofproto_ports(struct bridge *br)
840 struct port *port, *next_port;
842 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
843 struct iface *iface, *next_iface;
844 struct ofproto_port ofproto_port;
846 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
850 /* Open the netdev or reconfigure it. */
852 shash_from_ovs_idl_map(iface->cfg->key_options,
853 iface->cfg->value_options,
854 iface->cfg->n_options, &args);
855 if (!iface->netdev) {
856 struct netdev_options options;
857 options.name = iface->name;
858 options.type = iface->type;
859 options.args = &args;
860 options.ethertype = NETDEV_ETH_TYPE_NONE;
861 error = netdev_open(&options, &iface->netdev);
863 error = netdev_set_config(iface->netdev, &args);
865 shash_destroy(&args);
867 VLOG_WARN("could not %s network device %s (%s)",
868 iface->netdev ? "reconfigure" : "open",
869 iface->name, strerror(error));
872 /* Populate stats columns in new Interface rows. */
873 if (iface->netdev && !iface->cfg->mtu) {
874 iface_refresh_stats(iface);
875 iface_refresh_status(iface);
878 /* Add the port, if necessary. */
879 if (iface->netdev && iface->ofp_port < 0) {
883 error = ofproto_port_add(br->ofproto, iface->netdev,
886 iface_set_ofp_port(iface, ofp_port);
888 netdev_close(iface->netdev);
889 iface->netdev = NULL;
893 /* Delete the iface if */
894 if (iface->netdev && iface->ofp_port >= 0) {
895 VLOG_DBG("bridge %s: interface %s is on port %d",
896 br->name, iface->name, iface->ofp_port);
899 VLOG_ERR("bridge %s: missing %s interface, dropping",
900 br->name, iface->name);
902 /* We already reported a related error, don't bother
905 iface_set_ofport(iface->cfg, -1);
906 iface_destroy(iface);
909 if (list_is_empty(&port->ifaces)) {
910 VLOG_WARN("%s port has no interfaces, dropping", port->name);
915 /* Add bond fake iface if necessary. */
916 if (port_is_bond_fake_iface(port)) {
917 if (ofproto_port_query_by_name(br->ofproto, port->name,
919 struct netdev_options options;
920 struct netdev *netdev;
923 options.name = port->name;
924 options.type = "internal";
926 options.ethertype = NETDEV_ETH_TYPE_NONE;
927 error = netdev_open(&options, &netdev);
929 ofproto_port_add(br->ofproto, netdev, NULL);
930 netdev_close(netdev);
932 VLOG_WARN("could not open network device %s (%s)",
933 port->name, strerror(error));
936 /* Already exists, nothing to do. */
937 ofproto_port_destroy(&ofproto_port);
944 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
945 const struct ovsdb_idl_column *column,
948 const struct ovsdb_datum *datum;
949 union ovsdb_atom atom;
952 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
953 atom.string = (char *) key;
954 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
955 return idx == UINT_MAX ? NULL : datum->values[idx].string;
959 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
961 return get_ovsrec_key_value(&br_cfg->header_,
962 &ovsrec_bridge_col_other_config, key);
966 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
967 struct iface **hw_addr_iface)
973 *hw_addr_iface = NULL;
975 /* Did the user request a particular MAC? */
976 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
977 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
978 if (eth_addr_is_multicast(ea)) {
979 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
980 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
981 } else if (eth_addr_is_zero(ea)) {
982 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
988 /* Otherwise choose the minimum non-local MAC address among all of the
990 memset(ea, 0xff, ETH_ADDR_LEN);
991 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
992 uint8_t iface_ea[ETH_ADDR_LEN];
993 struct iface *candidate;
996 /* Mirror output ports don't participate. */
997 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1001 /* Choose the MAC address to represent the port. */
1003 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1004 /* Find the interface with this Ethernet address (if any) so that
1005 * we can provide the correct devname to the caller. */
1006 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1007 uint8_t candidate_ea[ETH_ADDR_LEN];
1008 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1009 && eth_addr_equals(iface_ea, candidate_ea)) {
1014 /* Choose the interface whose MAC address will represent the port.
1015 * The Linux kernel bonding code always chooses the MAC address of
1016 * the first slave added to a bond, and the Fedora networking
1017 * scripts always add slaves to a bond in alphabetical order, so
1018 * for compatibility we choose the interface with the name that is
1019 * first in alphabetical order. */
1020 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1021 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1026 /* The local port doesn't count (since we're trying to choose its
1027 * MAC address anyway). */
1028 if (iface->ofp_port == OFPP_LOCAL) {
1033 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1039 /* Compare against our current choice. */
1040 if (!eth_addr_is_multicast(iface_ea) &&
1041 !eth_addr_is_local(iface_ea) &&
1042 !eth_addr_is_reserved(iface_ea) &&
1043 !eth_addr_is_zero(iface_ea) &&
1044 eth_addr_compare_3way(iface_ea, ea) < 0)
1046 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1047 *hw_addr_iface = iface;
1050 if (eth_addr_is_multicast(ea)) {
1051 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1052 *hw_addr_iface = NULL;
1053 VLOG_WARN("bridge %s: using default bridge Ethernet "
1054 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1056 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1057 br->name, ETH_ADDR_ARGS(ea));
1061 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1062 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1063 * an interface on 'br', then that interface must be passed in as
1064 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1065 * 'hw_addr_iface' must be passed in as a null pointer. */
1067 bridge_pick_datapath_id(struct bridge *br,
1068 const uint8_t bridge_ea[ETH_ADDR_LEN],
1069 struct iface *hw_addr_iface)
1072 * The procedure for choosing a bridge MAC address will, in the most
1073 * ordinary case, also choose a unique MAC that we can use as a datapath
1074 * ID. In some special cases, though, multiple bridges will end up with
1075 * the same MAC address. This is OK for the bridges, but it will confuse
1076 * the OpenFlow controller, because each datapath needs a unique datapath
1079 * Datapath IDs must be unique. It is also very desirable that they be
1080 * stable from one run to the next, so that policy set on a datapath
1083 const char *datapath_id;
1086 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1087 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1091 if (hw_addr_iface) {
1093 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1095 * A bridge whose MAC address is taken from a VLAN network device
1096 * (that is, a network device created with vconfig(8) or similar
1097 * tool) will have the same MAC address as a bridge on the VLAN
1098 * device's physical network device.
1100 * Handle this case by hashing the physical network device MAC
1101 * along with the VLAN identifier.
1103 uint8_t buf[ETH_ADDR_LEN + 2];
1104 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1105 buf[ETH_ADDR_LEN] = vlan >> 8;
1106 buf[ETH_ADDR_LEN + 1] = vlan;
1107 return dpid_from_hash(buf, sizeof buf);
1110 * Assume that this bridge's MAC address is unique, since it
1111 * doesn't fit any of the cases we handle specially.
1116 * A purely internal bridge, that is, one that has no non-virtual
1117 * network devices on it at all, is more difficult because it has no
1118 * natural unique identifier at all.
1120 * When the host is a XenServer, we handle this case by hashing the
1121 * host's UUID with the name of the bridge. Names of bridges are
1122 * persistent across XenServer reboots, although they can be reused if
1123 * an internal network is destroyed and then a new one is later
1124 * created, so this is fairly effective.
1126 * When the host is not a XenServer, we punt by using a random MAC
1127 * address on each run.
1129 const char *host_uuid = xenserver_get_host_uuid();
1131 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1132 dpid = dpid_from_hash(combined, strlen(combined));
1138 return eth_addr_to_uint64(bridge_ea);
1142 dpid_from_hash(const void *data, size_t n)
1144 uint8_t hash[SHA1_DIGEST_SIZE];
1146 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1147 sha1_bytes(data, n, hash);
1148 eth_addr_mark_random(hash);
1149 return eth_addr_to_uint64(hash);
1153 iface_refresh_status(struct iface *iface)
1157 enum netdev_flags flags;
1164 if (iface_is_synthetic(iface)) {
1170 if (!netdev_get_status(iface->netdev, &sh)) {
1172 char **keys, **values;
1174 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1175 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1180 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1183 shash_destroy_free_data(&sh);
1185 error = netdev_get_flags(iface->netdev, &flags);
1187 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1190 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1193 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1195 ovsrec_interface_set_duplex(iface->cfg,
1196 netdev_features_is_full_duplex(current)
1198 /* warning: uint64_t -> int64_t conversion */
1199 bps = netdev_features_to_bps(current);
1200 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1203 ovsrec_interface_set_duplex(iface->cfg, NULL);
1204 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1207 ovsrec_interface_set_link_state(iface->cfg,
1208 iface_get_carrier(iface) ? "up" : "down");
1210 error = netdev_get_mtu(iface->netdev, &mtu);
1211 if (!error && mtu != INT_MAX) {
1213 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1216 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1220 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1221 * changed, false otherwise. */
1223 iface_refresh_cfm_stats(struct iface *iface)
1225 const struct ovsrec_interface *cfg = iface->cfg;
1226 bool changed = false;
1229 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1236 if (cfg->n_cfm_fault != 1 || cfg->cfm_fault[0] != fault) {
1237 bool fault_bool = fault;
1238 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1246 iface_refresh_lacp_stats(struct iface *iface)
1248 struct ofproto *ofproto = iface->port->bridge->ofproto;
1249 int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
1250 int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
1254 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
1260 iface_refresh_stats(struct iface *iface)
1262 #define IFACE_STATS \
1263 IFACE_STAT(rx_packets, "rx_packets") \
1264 IFACE_STAT(tx_packets, "tx_packets") \
1265 IFACE_STAT(rx_bytes, "rx_bytes") \
1266 IFACE_STAT(tx_bytes, "tx_bytes") \
1267 IFACE_STAT(rx_dropped, "rx_dropped") \
1268 IFACE_STAT(tx_dropped, "tx_dropped") \
1269 IFACE_STAT(rx_errors, "rx_errors") \
1270 IFACE_STAT(tx_errors, "tx_errors") \
1271 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1272 IFACE_STAT(rx_over_errors, "rx_over_err") \
1273 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1274 IFACE_STAT(collisions, "collisions")
1276 #define IFACE_STAT(MEMBER, NAME) NAME,
1277 static char *keys[] = { IFACE_STATS };
1279 int64_t values[ARRAY_SIZE(keys)];
1282 struct netdev_stats stats;
1284 if (iface_is_synthetic(iface)) {
1288 /* Intentionally ignore return value, since errors will set 'stats' to
1289 * all-1s, and we will deal with that correctly below. */
1290 netdev_get_stats(iface->netdev, &stats);
1292 /* Copy statistics into values[] array. */
1294 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1297 assert(i == ARRAY_SIZE(keys));
1299 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1304 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1308 /* Use other-config:enable-system-stats by preference. */
1309 enable = get_ovsrec_key_value(&cfg->header_,
1310 &ovsrec_open_vswitch_col_other_config,
1311 "enable-statistics");
1313 return !strcmp(enable, "true");
1316 /* Disable by default. */
1321 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1323 struct ovsdb_datum datum;
1327 if (enable_system_stats(cfg)) {
1328 get_system_stats(&stats);
1331 ovsdb_datum_from_shash(&datum, &stats);
1332 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1336 static inline const char *
1337 nx_role_to_str(enum nx_role role)
1342 case NX_ROLE_MASTER:
1347 return "*** INVALID ROLE ***";
1352 refresh_controller_status(void)
1356 const struct ovsrec_controller *cfg;
1360 /* Accumulate status for controllers on all bridges. */
1361 HMAP_FOR_EACH (br, node, &all_bridges) {
1362 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1365 /* Update each controller in the database with current status. */
1366 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1367 struct ofproto_controller_info *cinfo =
1368 shash_find_data(&info, cfg->target);
1371 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1372 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1373 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1374 (char **) cinfo->pairs.values,
1377 ovsrec_controller_set_is_connected(cfg, false);
1378 ovsrec_controller_set_role(cfg, NULL);
1379 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1383 ofproto_free_ofproto_controller_info(&info);
1389 const struct ovsrec_open_vswitch *cfg;
1391 bool datapath_destroyed;
1392 bool database_changed;
1395 /* (Re)configure if necessary. */
1396 database_changed = ovsdb_idl_run(idl);
1397 if (ovsdb_idl_is_lock_contended(idl)) {
1398 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1399 struct bridge *br, *next_br;
1401 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1402 "disabling this process until it goes away");
1404 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1408 } else if (!ovsdb_idl_has_lock(idl)) {
1411 cfg = ovsrec_open_vswitch_first(idl);
1413 /* Let each bridge do the work that it needs to do. */
1414 datapath_destroyed = false;
1415 HMAP_FOR_EACH (br, node, &all_bridges) {
1416 int error = ofproto_run(br->ofproto);
1418 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1419 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1420 "forcing reconfiguration", br->name);
1421 datapath_destroyed = true;
1425 /* Re-configure SSL. We do this on every trip through the main loop,
1426 * instead of just when the database changes, because the contents of the
1427 * key and certificate files can change without the database changing.
1429 * We do this before bridge_reconfigure() because that function might
1430 * initiate SSL connections and thus requires SSL to be configured. */
1431 if (cfg && cfg->ssl) {
1432 const struct ovsrec_ssl *ssl = cfg->ssl;
1434 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1435 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1438 if (database_changed || datapath_destroyed) {
1440 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1442 bridge_reconfigure(cfg);
1444 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1445 ovsdb_idl_txn_commit(txn);
1446 ovsdb_idl_txn_destroy(txn); /* XXX */
1448 /* We still need to reconfigure to avoid dangling pointers to
1449 * now-destroyed ovsrec structures inside bridge data. */
1450 static const struct ovsrec_open_vswitch null_cfg;
1452 bridge_reconfigure(&null_cfg);
1456 /* Refresh system and interface stats if necessary. */
1457 if (time_msec() >= stats_timer) {
1459 struct ovsdb_idl_txn *txn;
1461 txn = ovsdb_idl_txn_create(idl);
1462 HMAP_FOR_EACH (br, node, &all_bridges) {
1465 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1466 struct iface *iface;
1468 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1469 iface_refresh_stats(iface);
1470 iface_refresh_status(iface);
1474 refresh_system_stats(cfg);
1475 refresh_controller_status();
1476 ovsdb_idl_txn_commit(txn);
1477 ovsdb_idl_txn_destroy(txn); /* XXX */
1480 stats_timer = time_msec() + STATS_INTERVAL;
1483 if (time_msec() >= db_limiter) {
1484 struct ovsdb_idl_txn *txn;
1485 bool changed = false;
1487 txn = ovsdb_idl_txn_create(idl);
1488 HMAP_FOR_EACH (br, node, &all_bridges) {
1491 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1492 struct iface *iface;
1494 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1495 changed = iface_refresh_cfm_stats(iface) || changed;
1496 changed = iface_refresh_lacp_stats(iface) || changed;
1502 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1505 ovsdb_idl_txn_commit(txn);
1506 ovsdb_idl_txn_destroy(txn);
1513 ovsdb_idl_wait(idl);
1514 if (!hmap_is_empty(&all_bridges)) {
1517 HMAP_FOR_EACH (br, node, &all_bridges) {
1518 ofproto_wait(br->ofproto);
1520 poll_timer_wait_until(stats_timer);
1522 if (db_limiter > time_msec()) {
1523 poll_timer_wait_until(db_limiter);
1528 /* QoS unixctl user interface functions. */
1530 struct qos_unixctl_show_cbdata {
1532 struct iface *iface;
1536 qos_unixctl_show_cb(unsigned int queue_id,
1537 const struct shash *details,
1540 struct qos_unixctl_show_cbdata *data = aux;
1541 struct ds *ds = data->ds;
1542 struct iface *iface = data->iface;
1543 struct netdev_queue_stats stats;
1544 struct shash_node *node;
1547 ds_put_cstr(ds, "\n");
1549 ds_put_format(ds, "Queue %u:\n", queue_id);
1551 ds_put_cstr(ds, "Default:\n");
1554 SHASH_FOR_EACH (node, details) {
1555 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1558 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1560 if (stats.tx_packets != UINT64_MAX) {
1561 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1564 if (stats.tx_bytes != UINT64_MAX) {
1565 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1568 if (stats.tx_errors != UINT64_MAX) {
1569 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1572 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1573 queue_id, strerror(error));
1578 qos_unixctl_show(struct unixctl_conn *conn,
1579 const char *args, void *aux OVS_UNUSED)
1581 struct ds ds = DS_EMPTY_INITIALIZER;
1582 struct shash sh = SHASH_INITIALIZER(&sh);
1583 struct iface *iface;
1585 struct shash_node *node;
1586 struct qos_unixctl_show_cbdata data;
1589 iface = iface_find(args);
1591 unixctl_command_reply(conn, 501, "no such interface");
1595 netdev_get_qos(iface->netdev, &type, &sh);
1597 if (*type != '\0') {
1598 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1600 SHASH_FOR_EACH (node, &sh) {
1601 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1606 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1609 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1611 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1613 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1614 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1617 shash_destroy_free_data(&sh);
1621 /* Bridge reconfiguration functions. */
1623 bridge_create(const struct ovsrec_bridge *br_cfg)
1627 assert(!bridge_lookup(br_cfg->name));
1628 br = xzalloc(sizeof *br);
1630 br->name = xstrdup(br_cfg->name);
1631 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1634 /* Derive the default Ethernet address from the bridge's UUID. This should
1635 * be unique and it will be stable between ovs-vswitchd runs. */
1636 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1637 eth_addr_mark_random(br->default_ea);
1639 hmap_init(&br->ports);
1640 hmap_init(&br->ifaces);
1641 hmap_init(&br->iface_by_name);
1642 hmap_init(&br->mirrors);
1644 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1648 bridge_destroy(struct bridge *br)
1651 struct mirror *mirror, *next_mirror;
1652 struct port *port, *next_port;
1654 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1657 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1658 mirror_destroy(mirror);
1660 hmap_remove(&all_bridges, &br->node);
1661 ofproto_destroy(br->ofproto);
1662 hmap_destroy(&br->ifaces);
1663 hmap_destroy(&br->ports);
1664 hmap_destroy(&br->iface_by_name);
1665 hmap_destroy(&br->mirrors);
1672 static struct bridge *
1673 bridge_lookup(const char *name)
1677 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1678 if (!strcmp(br->name, name)) {
1685 /* Handle requests for a listing of all flows known by the OpenFlow
1686 * stack, including those normally hidden. */
1688 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1689 const char *args, void *aux OVS_UNUSED)
1694 br = bridge_lookup(args);
1696 unixctl_command_reply(conn, 501, "Unknown bridge");
1701 ofproto_get_all_flows(br->ofproto, &results);
1703 unixctl_command_reply(conn, 200, ds_cstr(&results));
1704 ds_destroy(&results);
1707 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1708 * connections and reconnect. If BRIDGE is not specified, then all bridges
1709 * drop their controller connections and reconnect. */
1711 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1712 const char *args, void *aux OVS_UNUSED)
1715 if (args[0] != '\0') {
1716 br = bridge_lookup(args);
1718 unixctl_command_reply(conn, 501, "Unknown bridge");
1721 ofproto_reconnect_controllers(br->ofproto);
1723 HMAP_FOR_EACH (br, node, &all_bridges) {
1724 ofproto_reconnect_controllers(br->ofproto);
1727 unixctl_command_reply(conn, 200, NULL);
1731 bridge_get_controllers(const struct bridge *br,
1732 struct ovsrec_controller ***controllersp)
1734 struct ovsrec_controller **controllers;
1735 size_t n_controllers;
1737 controllers = br->cfg->controller;
1738 n_controllers = br->cfg->n_controller;
1740 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1746 *controllersp = controllers;
1748 return n_controllers;
1751 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1752 * those configured in 'br->cfg'. */
1754 bridge_add_del_ports(struct bridge *br)
1756 struct port *port, *next;
1757 struct shash_node *node;
1758 struct shash new_ports;
1761 /* Collect new ports. */
1762 shash_init(&new_ports);
1763 for (i = 0; i < br->cfg->n_ports; i++) {
1764 const char *name = br->cfg->ports[i]->name;
1765 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1766 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1770 if (bridge_get_controllers(br, NULL)
1771 && !shash_find(&new_ports, br->name)) {
1772 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1773 br->name, br->name);
1775 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1776 br->synth_local_port.n_interfaces = 1;
1777 br->synth_local_port.name = br->name;
1779 br->synth_local_iface.name = br->name;
1780 br->synth_local_iface.type = "internal";
1782 br->synth_local_ifacep = &br->synth_local_iface;
1784 shash_add(&new_ports, br->name, &br->synth_local_port);
1787 /* Get rid of deleted ports.
1788 * Get rid of deleted interfaces on ports that still exist.
1789 * Update 'cfg' of ports that still exist. */
1790 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1791 port->cfg = shash_find_data(&new_ports, port->name);
1795 port_del_ifaces(port);
1799 /* Create new ports.
1800 * Add new interfaces to existing ports. */
1801 SHASH_FOR_EACH (node, &new_ports) {
1802 struct port *port = port_lookup(br, node->name);
1804 struct ovsrec_port *cfg = node->data;
1805 port = port_create(br, cfg);
1807 port_add_ifaces(port);
1808 if (list_is_empty(&port->ifaces)) {
1809 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1810 br->name, port->name);
1814 shash_destroy(&new_ports);
1817 /* Initializes 'oc' appropriately as a management service controller for
1820 * The caller must free oc->target when it is no longer needed. */
1822 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1823 struct ofproto_controller *oc)
1825 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1826 oc->max_backoff = 0;
1827 oc->probe_interval = 60;
1828 oc->band = OFPROTO_OUT_OF_BAND;
1830 oc->burst_limit = 0;
1833 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1835 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1836 struct ofproto_controller *oc)
1838 oc->target = c->target;
1839 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1840 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1841 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1842 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1843 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1844 oc->burst_limit = (c->controller_burst_limit
1845 ? *c->controller_burst_limit : 0);
1848 /* Configures the IP stack for 'br''s local interface properly according to the
1849 * configuration in 'c'. */
1851 bridge_configure_local_iface_netdev(struct bridge *br,
1852 struct ovsrec_controller *c)
1854 struct netdev *netdev;
1855 struct in_addr mask, gateway;
1857 struct iface *local_iface;
1860 /* If there's no local interface or no IP address, give up. */
1861 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1862 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1866 /* Bring up the local interface. */
1867 netdev = local_iface->netdev;
1868 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1870 /* Configure the IP address and netmask. */
1871 if (!c->local_netmask
1872 || !inet_aton(c->local_netmask, &mask)
1874 mask.s_addr = guess_netmask(ip.s_addr);
1876 if (!netdev_set_in4(netdev, ip, mask)) {
1877 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1878 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1881 /* Configure the default gateway. */
1882 if (c->local_gateway
1883 && inet_aton(c->local_gateway, &gateway)
1884 && gateway.s_addr) {
1885 if (!netdev_add_router(netdev, gateway)) {
1886 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1887 br->name, IP_ARGS(&gateway.s_addr));
1893 bridge_configure_remotes(struct bridge *br,
1894 const struct sockaddr_in *managers, size_t n_managers)
1896 const char *disable_ib_str, *queue_id_str;
1897 bool disable_in_band = false;
1900 struct ovsrec_controller **controllers;
1901 size_t n_controllers;
1903 enum ofproto_fail_mode fail_mode;
1905 struct ofproto_controller *ocs;
1909 /* Check if we should disable in-band control on this bridge. */
1910 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1911 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1912 disable_in_band = true;
1915 /* Set OpenFlow queue ID for in-band control. */
1916 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1917 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1918 ofproto_set_in_band_queue(br->ofproto, queue_id);
1920 if (disable_in_band) {
1921 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1923 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1926 n_controllers = bridge_get_controllers(br, &controllers);
1928 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1931 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1932 for (i = 0; i < n_controllers; i++) {
1933 struct ovsrec_controller *c = controllers[i];
1935 if (!strncmp(c->target, "punix:", 6)
1936 || !strncmp(c->target, "unix:", 5)) {
1937 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1939 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
1940 * domain sockets and overwriting arbitrary local files. */
1941 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
1942 "controller \"%s\" due to possibility for remote "
1943 "exploit", br->name, c->target);
1947 bridge_configure_local_iface_netdev(br, c);
1948 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
1949 if (disable_in_band) {
1950 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
1955 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
1956 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
1959 /* Set the fail-mode. */
1960 fail_mode = !br->cfg->fail_mode
1961 || !strcmp(br->cfg->fail_mode, "standalone")
1962 ? OFPROTO_FAIL_STANDALONE
1963 : OFPROTO_FAIL_SECURE;
1964 ofproto_set_fail_mode(br->ofproto, fail_mode);
1966 /* Configure OpenFlow controller connection snooping. */
1967 if (!ofproto_has_snoops(br->ofproto)) {
1971 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1972 ovs_rundir(), br->name));
1973 ofproto_set_snoops(br->ofproto, &snoops);
1974 sset_destroy(&snoops);
1978 /* Port functions. */
1980 static struct port *
1981 port_create(struct bridge *br, const struct ovsrec_port *cfg)
1985 port = xzalloc(sizeof *port);
1987 port->name = xstrdup(cfg->name);
1989 list_init(&port->ifaces);
1991 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
1993 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
1999 get_port_other_config(const struct ovsrec_port *port, const char *key,
2000 const char *default_value)
2004 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2006 return value ? value : default_value;
2010 get_interface_other_config(const struct ovsrec_interface *iface,
2011 const char *key, const char *default_value)
2015 value = get_ovsrec_key_value(&iface->header_,
2016 &ovsrec_interface_col_other_config, key);
2017 return value ? value : default_value;
2020 /* Deletes interfaces from 'port' that are no longer configured for it. */
2022 port_del_ifaces(struct port *port)
2024 struct iface *iface, *next;
2025 struct sset new_ifaces;
2028 /* Collect list of new interfaces. */
2029 sset_init(&new_ifaces);
2030 for (i = 0; i < port->cfg->n_interfaces; i++) {
2031 const char *name = port->cfg->interfaces[i]->name;
2032 const char *type = port->cfg->interfaces[i]->name;
2033 if (strcmp(type, "null")) {
2034 sset_add(&new_ifaces, name);
2038 /* Get rid of deleted interfaces. */
2039 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2040 if (!sset_contains(&new_ifaces, iface->name)) {
2041 iface_destroy(iface);
2045 sset_destroy(&new_ifaces);
2048 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2051 port_add_ifaces(struct port *port)
2053 struct shash new_ifaces;
2054 struct shash_node *node;
2057 /* Collect new ifaces. */
2058 shash_init(&new_ifaces);
2059 for (i = 0; i < port->cfg->n_interfaces; i++) {
2060 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2061 if (strcmp(cfg->type, "null")
2062 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2063 VLOG_WARN("port %s: %s specified twice as port interface",
2064 port->name, cfg->name);
2065 iface_set_ofport(cfg, -1);
2069 /* Create new interfaces.
2070 * Update interface types and 'cfg' members. */
2071 SHASH_FOR_EACH (node, &new_ifaces) {
2072 const struct ovsrec_interface *cfg = node->data;
2073 const char *iface_name = node->name;
2074 struct iface *iface;
2076 iface = iface_lookup(port->bridge, iface_name);
2078 iface = iface_create(port, cfg);
2083 /* Determine interface type. The local port always has type
2084 * "internal". Other ports take their type from the database and
2085 * default to "system" if none is specified. */
2086 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2087 : cfg->type[0] ? cfg->type
2090 shash_destroy(&new_ifaces);
2094 port_destroy(struct port *port)
2097 struct bridge *br = port->bridge;
2098 struct iface *iface, *next;
2101 ofproto_bundle_unregister(br->ofproto, port);
2104 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2105 iface_destroy(iface);
2108 hmap_remove(&br->ports, &port->hmap_node);
2110 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2117 static struct port *
2118 port_lookup(const struct bridge *br, const char *name)
2122 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2124 if (!strcmp(port->name, name)) {
2132 enable_lacp(struct port *port, bool *activep)
2134 if (!port->cfg->lacp) {
2135 /* XXX when LACP implementation has been sufficiently tested, enable by
2136 * default and make active on bonded ports. */
2138 } else if (!strcmp(port->cfg->lacp, "off")) {
2140 } else if (!strcmp(port->cfg->lacp, "active")) {
2143 } else if (!strcmp(port->cfg->lacp, "passive")) {
2147 VLOG_WARN("port %s: unknown LACP mode %s",
2148 port->name, port->cfg->lacp);
2153 static struct lacp_settings *
2154 port_configure_lacp(struct port *port, struct lacp_settings *s)
2156 const char *lacp_time;
2157 long long int custom_time;
2160 if (!enable_lacp(port, &s->active)) {
2164 s->name = port->name;
2165 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2167 /* Prefer bondable links if unspecified. */
2168 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2170 s->priority = (priority > 0 && priority <= UINT16_MAX
2172 : UINT16_MAX - !list_is_short(&port->ifaces));
2174 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2179 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2180 custom_time = atoi(lacp_time);
2181 if (!strcmp(lacp_time, "fast")) {
2182 s->lacp_time = LACP_TIME_FAST;
2183 } else if (!strcmp(lacp_time, "slow")) {
2184 s->lacp_time = LACP_TIME_SLOW;
2185 } else if (custom_time > 0) {
2186 s->lacp_time = LACP_TIME_CUSTOM;
2187 s->custom_time = custom_time;
2189 s->lacp_time = LACP_TIME_SLOW;
2196 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2198 int priority, portid, key;
2200 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2201 priority = atoi(get_interface_other_config(iface->cfg,
2202 "lacp-port-priority", "0"));
2203 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2206 if (portid <= 0 || portid > UINT16_MAX) {
2207 portid = iface->ofp_port;
2210 if (priority <= 0 || priority > UINT16_MAX) {
2211 priority = UINT16_MAX;
2214 if (key < 0 || key > UINT16_MAX) {
2218 s->name = iface->name;
2220 s->priority = priority;
2225 port_configure_bond(struct port *port, struct bond_settings *s,
2226 uint32_t *bond_stable_ids)
2228 const char *detect_s;
2229 struct iface *iface;
2230 int miimon_interval;
2233 s->name = port->name;
2234 s->balance = BM_SLB;
2235 if (port->cfg->bond_mode
2236 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2237 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2238 port->name, port->cfg->bond_mode,
2239 bond_mode_to_string(s->balance));
2242 miimon_interval = atoi(get_port_other_config(port->cfg,
2243 "bond-miimon-interval", "0"));
2244 if (miimon_interval <= 0) {
2245 miimon_interval = 200;
2248 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2249 if (!strcmp(detect_s, "carrier")) {
2250 miimon_interval = 0;
2251 } else if (strcmp(detect_s, "miimon")) {
2252 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2253 "defaulting to carrier", port->name, detect_s);
2254 miimon_interval = 0;
2257 s->up_delay = MAX(0, port->cfg->bond_updelay);
2258 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2259 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2260 s->rebalance_interval = atoi(
2261 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2262 if (s->rebalance_interval < 1000) {
2263 s->rebalance_interval = 1000;
2266 s->fake_iface = port->cfg->bond_fake_iface;
2269 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2270 long long stable_id;
2272 stable_id = atoll(get_interface_other_config(iface->cfg,
2273 "bond-stable-id", "0"));
2274 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2275 stable_id = iface->ofp_port;
2277 bond_stable_ids[i++] = stable_id;
2279 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2283 /* Interface functions. */
2285 static struct iface *
2286 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2288 struct bridge *br = port->bridge;
2289 struct iface *iface;
2290 char *name = if_cfg->name;
2292 iface = xzalloc(sizeof *iface);
2294 iface->name = xstrdup(name);
2295 iface->ofp_port = -1;
2296 iface->tag = tag_create_random();
2297 iface->netdev = NULL;
2298 iface->cfg = if_cfg;
2300 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2302 list_push_back(&port->ifaces, &iface->port_elem);
2304 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2310 iface_destroy(struct iface *iface)
2313 struct port *port = iface->port;
2314 struct bridge *br = port->bridge;
2316 if (br->ofproto && iface->ofp_port >= 0) {
2317 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2320 if (iface->ofp_port >= 0) {
2321 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2324 list_remove(&iface->port_elem);
2325 hmap_remove(&br->iface_by_name, &iface->name_node);
2327 netdev_close(iface->netdev);
2334 static struct iface *
2335 iface_lookup(const struct bridge *br, const char *name)
2337 struct iface *iface;
2339 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2340 &br->iface_by_name) {
2341 if (!strcmp(iface->name, name)) {
2349 static struct iface *
2350 iface_find(const char *name)
2352 const struct bridge *br;
2354 HMAP_FOR_EACH (br, node, &all_bridges) {
2355 struct iface *iface = iface_lookup(br, name);
2364 static struct iface *
2365 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2367 struct iface *iface;
2369 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2370 hash_int(ofp_port, 0), &br->ifaces) {
2371 if (iface->ofp_port == ofp_port) {
2378 /* Set Ethernet address of 'iface', if one is specified in the configuration
2381 iface_set_mac(struct iface *iface)
2383 uint8_t ea[ETH_ADDR_LEN];
2385 if (!strcmp(iface->type, "internal")
2386 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2387 if (iface->ofp_port == OFPP_LOCAL) {
2388 VLOG_ERR("interface %s: ignoring mac in Interface record "
2389 "(use Bridge record to set local port's mac)",
2391 } else if (eth_addr_is_multicast(ea)) {
2392 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2395 int error = netdev_set_etheraddr(iface->netdev, ea);
2397 VLOG_ERR("interface %s: setting MAC failed (%s)",
2398 iface->name, strerror(error));
2404 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2406 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2408 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2409 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2413 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2415 * The value strings in '*shash' are taken directly from values[], not copied,
2416 * so the caller should not modify or free them. */
2418 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2419 struct shash *shash)
2424 for (i = 0; i < n; i++) {
2425 shash_add(shash, keys[i], values[i]);
2429 /* Creates 'keys' and 'values' arrays from 'shash'.
2431 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2432 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2433 * are populated with with strings taken directly from 'shash' and thus have
2434 * the same ownership of the key-value pairs in shash.
2437 shash_to_ovs_idl_map(struct shash *shash,
2438 char ***keys, char ***values, size_t *n)
2442 struct shash_node *sn;
2444 count = shash_count(shash);
2446 k = xmalloc(count * sizeof *k);
2447 v = xmalloc(count * sizeof *v);
2450 SHASH_FOR_EACH(sn, shash) {
2461 struct iface_delete_queues_cbdata {
2462 struct netdev *netdev;
2463 const struct ovsdb_datum *queues;
2467 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2469 union ovsdb_atom atom;
2471 atom.integer = target;
2472 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2476 iface_delete_queues(unsigned int queue_id,
2477 const struct shash *details OVS_UNUSED, void *cbdata_)
2479 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2481 if (!queue_ids_include(cbdata->queues, queue_id)) {
2482 netdev_delete_queue(cbdata->netdev, queue_id);
2487 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2489 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2490 netdev_set_qos(iface->netdev, NULL, NULL);
2492 struct iface_delete_queues_cbdata cbdata;
2493 struct shash details;
2497 /* Configure top-level Qos for 'iface'. */
2498 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2499 qos->n_other_config, &details);
2500 netdev_set_qos(iface->netdev, qos->type, &details);
2501 shash_destroy(&details);
2503 /* Deconfigure queues that were deleted. */
2504 cbdata.netdev = iface->netdev;
2505 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2507 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2509 /* Configure queues for 'iface'. */
2511 for (i = 0; i < qos->n_queues; i++) {
2512 const struct ovsrec_queue *queue = qos->value_queues[i];
2513 unsigned int queue_id = qos->key_queues[i];
2515 if (queue_id == 0) {
2519 shash_from_ovs_idl_map(queue->key_other_config,
2520 queue->value_other_config,
2521 queue->n_other_config, &details);
2522 netdev_set_queue(iface->netdev, queue_id, &details);
2523 shash_destroy(&details);
2526 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2527 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2528 "queue (queue 0). Packets not directed to a "
2529 "correctly configured queue may be dropped.",
2534 netdev_set_policing(iface->netdev,
2535 iface->cfg->ingress_policing_rate,
2536 iface->cfg->ingress_policing_burst);
2540 iface_configure_cfm(struct iface *iface)
2542 const struct ovsrec_interface *cfg = iface->cfg;
2543 struct cfm_settings s;
2544 uint16_t remote_mpid;
2546 if (!cfg->n_cfm_mpid || !cfg->n_cfm_remote_mpid) {
2547 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2551 s.mpid = *cfg->cfm_mpid;
2552 remote_mpid = *cfg->cfm_remote_mpid;
2553 s.remote_mpids = &remote_mpid;
2554 s.n_remote_mpids = 1;
2556 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2558 if (s.interval <= 0) {
2562 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2565 /* Read carrier or miimon status directly from 'iface''s netdev, according to
2566 * how 'iface''s port is configured.
2568 * Returns true if 'iface' is up, false otherwise. */
2570 iface_get_carrier(const struct iface *iface)
2573 return netdev_get_carrier(iface->netdev);
2576 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2577 * instead of obtaining it from the database. */
2579 iface_is_synthetic(const struct iface *iface)
2581 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2584 /* Port mirroring. */
2586 static struct mirror *
2587 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2591 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2592 if (uuid_equals(uuid, &m->uuid)) {
2600 bridge_configure_mirrors(struct bridge *br)
2602 const struct ovsdb_datum *mc;
2603 unsigned long *flood_vlans;
2604 struct mirror *m, *next;
2607 /* Get rid of deleted mirrors. */
2608 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2609 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2610 union ovsdb_atom atom;
2612 atom.uuid = m->uuid;
2613 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2618 /* Add new mirrors and reconfigure existing ones. */
2619 for (i = 0; i < br->cfg->n_mirrors; i++) {
2620 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2621 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2623 m = mirror_create(br, cfg);
2625 if (!mirror_configure(m, cfg)) {
2630 /* Update flooded vlans (for RSPAN). */
2631 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2632 br->cfg->n_flood_vlans);
2633 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2634 bitmap_free(flood_vlans);
2637 static struct mirror *
2638 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2642 m = xzalloc(sizeof *m);
2643 m->uuid = cfg->header_.uuid;
2644 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2646 m->name = xstrdup(cfg->name);
2652 mirror_destroy(struct mirror *m)
2655 struct bridge *br = m->bridge;
2658 ofproto_mirror_unregister(br->ofproto, m);
2661 hmap_remove(&br->mirrors, &m->hmap_node);
2668 mirror_collect_ports(struct mirror *m,
2669 struct ovsrec_port **in_ports, int n_in_ports,
2670 void ***out_portsp, size_t *n_out_portsp)
2672 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2673 size_t n_out_ports = 0;
2676 for (i = 0; i < n_in_ports; i++) {
2677 const char *name = in_ports[i]->name;
2678 struct port *port = port_lookup(m->bridge, name);
2680 out_ports[n_out_ports++] = port;
2682 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2683 "port %s", m->bridge->name, m->name, name);
2686 *out_portsp = out_ports;
2687 *n_out_portsp = n_out_ports;
2691 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2693 struct ofproto_mirror_settings s;
2696 if (strcmp(cfg->name, m->name)) {
2698 m->name = xstrdup(cfg->name);
2702 /* Get output port or VLAN. */
2703 if (cfg->output_port) {
2704 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2705 if (!s.out_bundle) {
2706 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2707 m->bridge->name, m->name);
2710 s.out_vlan = UINT16_MAX;
2712 if (cfg->output_vlan) {
2713 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2714 "output vlan; ignoring output vlan",
2715 m->bridge->name, m->name);
2717 } else if (cfg->output_vlan) {
2718 /* The database should prevent invalid VLAN values. */
2719 s.out_bundle = NULL;
2720 s.out_vlan = *cfg->output_vlan;
2722 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2723 m->bridge->name, m->name);
2727 /* Get port selection. */
2728 if (cfg->select_all) {
2729 size_t n_ports = hmap_count(&m->bridge->ports);
2730 void **ports = xmalloc(n_ports * sizeof *ports);
2735 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2745 /* Get ports, dropping ports that don't exist.
2746 * The IDL ensures that there are no duplicates. */
2747 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2748 &s.srcs, &s.n_srcs);
2749 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2750 &s.dsts, &s.n_dsts);
2754 /* Get VLAN selection. */
2755 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2758 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2761 if (s.srcs != s.dsts) {