1 /* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "ofp-print.h"
37 #include "ofproto/ofproto.h"
38 #include "poll-loop.h"
41 #include "socket-util.h"
42 #include "stream-ssl.h"
44 #include "system-stats.h"
48 #include "vswitchd/vswitch-idl.h"
49 #include "xenserver.h"
51 #include "sflow_api.h"
52 #include "vlan-bitmap.h"
54 VLOG_DEFINE_THIS_MODULE(bridge);
56 COVERAGE_DEFINE(bridge_reconfigure);
59 /* These members are always valid. */
60 struct list port_elem; /* Element in struct port's "ifaces" list. */
61 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
62 struct port *port; /* Containing port. */
63 char *name; /* Host network device name. */
64 tag_type tag; /* Tag associated with this interface. */
66 /* These members are valid only after bridge_reconfigure() causes them to
68 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
69 int ofp_port; /* OpenFlow port number, -1 if unknown. */
70 struct netdev *netdev; /* Network device. */
71 const char *type; /* Usually same as cfg->type. */
72 const struct ovsrec_interface *cfg;
76 struct uuid uuid; /* UUID of this "mirror" record in database. */
77 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
78 struct bridge *bridge;
83 struct bridge *bridge;
84 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
87 const struct ovsrec_port *cfg;
89 /* An ordinary bridge port has 1 interface.
90 * A bridge port for bonding has at least 2 interfaces. */
91 struct list ifaces; /* List of "struct iface"s. */
95 struct hmap_node node; /* In 'all_bridges'. */
96 char *name; /* User-specified arbitrary name. */
97 char *type; /* Datapath type. */
98 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
99 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
100 const struct ovsrec_bridge *cfg;
102 /* OpenFlow switch processing. */
103 struct ofproto *ofproto; /* OpenFlow switch. */
106 struct hmap ports; /* "struct port"s indexed by name. */
107 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
108 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
110 /* Port mirroring. */
111 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
113 /* Synthetic local port if necessary. */
114 struct ovsrec_port synth_local_port;
115 struct ovsrec_interface synth_local_iface;
116 struct ovsrec_interface *synth_local_ifacep;
119 /* All bridges, indexed by name. */
120 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
122 /* OVSDB IDL used to obtain configuration. */
123 static struct ovsdb_idl *idl;
125 /* Each time this timer expires, the bridge fetches systems and interface
126 * statistics and pushes them into the database. */
127 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
128 static long long int stats_timer = LLONG_MIN;
130 /* Stores the time after which rate limited statistics may be written to the
131 * database. Only updated when changes to the database require rate limiting.
133 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
134 static long long int db_limiter = LLONG_MIN;
136 static void add_del_bridges(const struct ovsrec_open_vswitch *);
137 static void bridge_del_ofprotos(void);
138 static bool bridge_add_ofprotos(struct bridge *);
139 static void bridge_create(const struct ovsrec_bridge *);
140 static void bridge_destroy(struct bridge *);
141 static struct bridge *bridge_lookup(const char *name);
142 static unixctl_cb_func bridge_unixctl_dump_flows;
143 static unixctl_cb_func bridge_unixctl_reconnect;
144 static size_t bridge_get_controllers(const struct bridge *br,
145 struct ovsrec_controller ***controllersp);
146 static void bridge_add_del_ports(struct bridge *);
147 static void bridge_add_ofproto_ports(struct bridge *);
148 static void bridge_del_ofproto_ports(struct bridge *);
149 static void bridge_refresh_ofp_port(struct bridge *);
150 static void bridge_configure_datapath_id(struct bridge *);
151 static void bridge_configure_flow_eviction_threshold(struct bridge *);
152 static void bridge_configure_netflow(struct bridge *);
153 static void bridge_configure_forward_bpdu(struct bridge *);
154 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
155 static void bridge_configure_remotes(struct bridge *,
156 const struct sockaddr_in *managers,
158 static void bridge_pick_local_hw_addr(struct bridge *,
159 uint8_t ea[ETH_ADDR_LEN],
160 struct iface **hw_addr_iface);
161 static uint64_t bridge_pick_datapath_id(struct bridge *,
162 const uint8_t bridge_ea[ETH_ADDR_LEN],
163 struct iface *hw_addr_iface);
164 static uint64_t dpid_from_hash(const void *, size_t nbytes);
165 static bool bridge_has_bond_fake_iface(const struct bridge *,
167 static bool port_is_bond_fake_iface(const struct port *);
169 static unixctl_cb_func qos_unixctl_show;
171 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
172 static void port_add_ifaces(struct port *);
173 static void port_del_ifaces(struct port *);
174 static void port_destroy(struct port *);
175 static struct port *port_lookup(const struct bridge *, const char *name);
176 static void port_configure(struct port *);
177 static struct lacp_settings *port_configure_lacp(struct port *,
178 struct lacp_settings *);
179 static void port_configure_bond(struct port *, struct bond_settings *,
180 uint32_t *bond_stable_ids);
182 static void bridge_configure_mirrors(struct bridge *);
183 static struct mirror *mirror_create(struct bridge *,
184 const struct ovsrec_mirror *);
185 static void mirror_destroy(struct mirror *);
186 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
188 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
189 static struct iface *iface_create(struct port *port,
190 const struct ovsrec_interface *if_cfg);
191 static void iface_destroy(struct iface *);
192 static struct iface *iface_lookup(const struct bridge *, const char *name);
193 static struct iface *iface_find(const char *name);
194 static struct iface *iface_from_ofp_port(const struct bridge *,
196 static void iface_set_mac(struct iface *);
197 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
198 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
199 static void iface_configure_cfm(struct iface *);
200 static void iface_refresh_cfm_stats(struct iface *);
201 static void iface_refresh_stats(struct iface *);
202 static void iface_refresh_status(struct iface *);
203 static bool iface_get_carrier(const struct iface *);
204 static bool iface_is_synthetic(const struct iface *);
206 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
208 static void shash_to_ovs_idl_map(struct shash *,
209 char ***keys, char ***values, size_t *n);
211 /* Public functions. */
213 /* Initializes the bridge module, configuring it to obtain its configuration
214 * from an OVSDB server accessed over 'remote', which should be a string in a
215 * form acceptable to ovsdb_idl_create(). */
217 bridge_init(const char *remote)
219 /* Create connection to database. */
220 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
221 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
223 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
224 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
225 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
226 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
227 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
228 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
229 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
231 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
232 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
234 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
235 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
237 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
238 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
239 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
240 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
241 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
242 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
243 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
244 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
245 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
246 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
247 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
249 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
250 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
251 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
252 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
254 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
258 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
260 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
262 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
264 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
265 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
266 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
267 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
268 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
270 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
272 /* Register unixctl commands. */
273 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
274 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
276 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
286 struct bridge *br, *next_br;
288 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
291 ovsdb_idl_destroy(idl);
294 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
295 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
296 * responsible for freeing '*managersp' (with free()).
298 * You may be asking yourself "why does ovs-vswitchd care?", because
299 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
300 * should not be and in fact is not directly involved in that. But
301 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
302 * it has to tell in-band control where the managers are to enable that.
303 * (Thus, only managers connected in-band are collected.)
306 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
307 struct sockaddr_in **managersp, size_t *n_managersp)
309 struct sockaddr_in *managers = NULL;
310 size_t n_managers = 0;
314 /* Collect all of the potential targets from the "targets" columns of the
315 * rows pointed to by "manager_options", excluding any that are
318 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
319 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
321 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
322 sset_find_and_delete(&targets, m->target);
324 sset_add(&targets, m->target);
328 /* Now extract the targets' IP addresses. */
329 if (!sset_is_empty(&targets)) {
332 managers = xmalloc(sset_count(&targets) * sizeof *managers);
333 SSET_FOR_EACH (target, &targets) {
334 struct sockaddr_in *sin = &managers[n_managers];
336 if ((!strncmp(target, "tcp:", 4)
337 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
338 (!strncmp(target, "ssl:", 4)
339 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
344 sset_destroy(&targets);
346 *managersp = managers;
347 *n_managersp = n_managers;
351 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
353 struct sockaddr_in *managers;
354 struct bridge *br, *next;
355 int sflow_bridge_number;
358 COVERAGE_INC(bridge_reconfigure);
360 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
361 * iface"s according to 'ovs_cfg', with only very minimal configuration
364 * This is purely an update to bridge data structures. Nothing is pushed
365 * down to ofproto or lower layers. */
366 add_del_bridges(ovs_cfg);
367 HMAP_FOR_EACH (br, node, &all_bridges) {
368 bridge_add_del_ports(br);
371 /* Delete all datapaths and datapath ports that are no longer configured.
373 * The kernel will reject any attempt to add a given port to a datapath if
374 * that port already belongs to a different datapath, so we must do all
375 * port deletions before any port additions. A datapath always has a
376 * "local port" so we must delete not-configured datapaths too. */
377 bridge_del_ofprotos();
378 HMAP_FOR_EACH (br, node, &all_bridges) {
380 bridge_del_ofproto_ports(br);
384 /* Create datapaths and datapath ports that are missing.
386 * After this is done, we have our final set of bridges, ports, and
387 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
388 * has at least one iface, every "struct iface" has a valid ofp_port and
390 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
391 if (!br->ofproto && !bridge_add_ofprotos(br)) {
395 HMAP_FOR_EACH (br, node, &all_bridges) {
396 bridge_refresh_ofp_port(br);
397 bridge_add_ofproto_ports(br);
400 /* Complete the configuration. */
401 sflow_bridge_number = 0;
402 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
403 HMAP_FOR_EACH (br, node, &all_bridges) {
406 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
409 port_configure(port);
411 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
412 iface_configure_cfm(iface);
413 iface_configure_qos(iface, port->cfg->qos);
414 iface_set_mac(iface);
417 bridge_configure_mirrors(br);
418 bridge_configure_datapath_id(br);
419 bridge_configure_flow_eviction_threshold(br);
420 bridge_configure_forward_bpdu(br);
421 bridge_configure_remotes(br, managers, n_managers);
422 bridge_configure_netflow(br);
423 bridge_configure_sflow(br, &sflow_bridge_number);
427 /* ovs-vswitchd has completed initialization, so allow the process that
428 * forked us to exit successfully. */
429 daemonize_complete();
432 /* Iterate over all ofprotos and delete any of them that do not have a
433 * configured bridge or that are the wrong type. */
435 bridge_del_ofprotos(void)
443 ofproto_enumerate_types(&types);
444 SSET_FOR_EACH (type, &types) {
447 ofproto_enumerate_names(type, &names);
448 SSET_FOR_EACH (name, &names) {
449 struct bridge *br = bridge_lookup(name);
450 if (!br || strcmp(type, br->type)) {
451 ofproto_delete(name, type);
455 sset_destroy(&names);
456 sset_destroy(&types);
460 bridge_add_ofprotos(struct bridge *br)
462 int error = ofproto_create(br->name, br->type, &br->ofproto);
464 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
471 port_configure(struct port *port)
473 const struct ovsrec_port *cfg = port->cfg;
474 struct bond_settings bond_settings;
475 struct lacp_settings lacp_settings;
476 struct ofproto_bundle_settings s;
484 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
485 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
486 s.slaves[s.n_slaves++] = iface->ofp_port;
492 if (list_is_short(&port->ifaces)) {
493 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
495 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan);
498 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
499 * they even work as-is. But they have not been tested. */
500 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
505 /* Get VLAN trunks. */
507 if (s.vlan < 0 && cfg->n_trunks) {
508 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
509 } else if (s.vlan >= 0 && cfg->n_trunks) {
510 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
514 /* Get LACP settings. */
515 s.lacp = port_configure_lacp(port, &lacp_settings);
519 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
520 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
521 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
524 s.lacp_slaves = NULL;
527 /* Get bond settings. */
528 if (s.n_slaves > 1) {
529 s.bond = &bond_settings;
530 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
531 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
534 s.bond_stable_ids = NULL;
536 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
537 netdev_set_miimon_interval(iface->netdev, 0);
542 ofproto_bundle_register(port->bridge->ofproto, port, &s);
548 free(s.bond_stable_ids);
551 /* Pick local port hardware address and datapath ID for 'br'. */
553 bridge_configure_datapath_id(struct bridge *br)
555 uint8_t ea[ETH_ADDR_LEN];
557 struct iface *local_iface;
558 struct iface *hw_addr_iface;
561 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
562 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
564 int error = netdev_set_etheraddr(local_iface->netdev, ea);
566 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
567 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
568 "Ethernet address: %s",
569 br->name, strerror(error));
572 memcpy(br->ea, ea, ETH_ADDR_LEN);
574 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
575 ofproto_set_datapath_id(br->ofproto, dpid);
577 dpid_string = xasprintf("%016"PRIx64, dpid);
578 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
582 /* Set NetFlow configuration on 'br'. */
584 bridge_configure_netflow(struct bridge *br)
586 struct ovsrec_netflow *cfg = br->cfg->netflow;
587 struct netflow_options opts;
590 ofproto_set_netflow(br->ofproto, NULL);
594 memset(&opts, 0, sizeof opts);
596 /* Get default NetFlow configuration from datapath.
597 * Apply overrides from 'cfg'. */
598 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
599 if (cfg->engine_type) {
600 opts.engine_type = *cfg->engine_type;
602 if (cfg->engine_id) {
603 opts.engine_id = *cfg->engine_id;
606 /* Configure active timeout interval. */
607 opts.active_timeout = cfg->active_timeout;
608 if (!opts.active_timeout) {
609 opts.active_timeout = -1;
610 } else if (opts.active_timeout < 0) {
611 VLOG_WARN("bridge %s: active timeout interval set to negative "
612 "value, using default instead (%d seconds)", br->name,
613 NF_ACTIVE_TIMEOUT_DEFAULT);
614 opts.active_timeout = -1;
617 /* Add engine ID to interface number to disambiguate bridgs? */
618 opts.add_id_to_iface = cfg->add_id_to_interface;
619 if (opts.add_id_to_iface) {
620 if (opts.engine_id > 0x7f) {
621 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
622 "another vswitch, choose an engine id less than 128",
625 if (hmap_count(&br->ports) > 508) {
626 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
627 "another port when more than 508 ports are used",
633 sset_init(&opts.collectors);
634 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
637 if (ofproto_set_netflow(br->ofproto, &opts)) {
638 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
640 sset_destroy(&opts.collectors);
643 /* Set sFlow configuration on 'br'. */
645 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
647 const struct ovsrec_sflow *cfg = br->cfg->sflow;
648 struct ovsrec_controller **controllers;
649 struct ofproto_sflow_options oso;
650 size_t n_controllers;
654 ofproto_set_sflow(br->ofproto, NULL);
658 memset(&oso, 0, sizeof oso);
660 sset_init(&oso.targets);
661 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
663 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
665 oso.sampling_rate = *cfg->sampling;
668 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
670 oso.polling_interval = *cfg->polling;
673 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
675 oso.header_len = *cfg->header;
678 oso.sub_id = (*sflow_bridge_number)++;
679 oso.agent_device = cfg->agent;
681 oso.control_ip = NULL;
682 n_controllers = bridge_get_controllers(br, &controllers);
683 for (i = 0; i < n_controllers; i++) {
684 if (controllers[i]->local_ip) {
685 oso.control_ip = controllers[i]->local_ip;
689 ofproto_set_sflow(br->ofproto, &oso);
691 sset_destroy(&oso.targets);
695 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
697 const struct port *port = port_lookup(br, name);
698 return port && port_is_bond_fake_iface(port);
702 port_is_bond_fake_iface(const struct port *port)
704 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
708 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
710 struct bridge *br, *next;
714 /* Collect new bridges' names and types. */
716 for (i = 0; i < cfg->n_bridges; i++) {
717 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
718 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
719 VLOG_WARN("bridge %s specified twice", br_cfg->name);
723 /* Get rid of deleted bridges or those whose types have changed.
724 * Update 'cfg' of bridges that still exist. */
725 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
726 br->cfg = shash_find_data(&new_br, br->name);
727 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
728 br->cfg->datapath_type))) {
733 /* Add new bridges. */
734 for (i = 0; i < cfg->n_bridges; i++) {
735 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
736 struct bridge *br = bridge_lookup(br_cfg->name);
738 bridge_create(br_cfg);
742 shash_destroy(&new_br);
745 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
748 * The kernel will reject any attempt to add a given port to a datapath if that
749 * port already belongs to a different datapath, so we must do all port
750 * deletions before any port additions. */
752 bridge_del_ofproto_ports(struct bridge *br)
754 struct ofproto_port_dump dump;
755 struct ofproto_port ofproto_port;
757 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
758 const char *name = ofproto_port.name;
763 /* Ignore the local port. We can't change it anyhow. */
764 if (!strcmp(name, br->name)) {
768 /* Get the type that 'ofproto_port' should have (ordinarily the
769 * type of its corresponding iface) or NULL if it should be
771 iface = iface_lookup(br, name);
772 type = (iface ? iface->type
773 : bridge_has_bond_fake_iface(br, name) ? "internal"
776 /* If it's the wrong type then delete the ofproto port. */
778 && !strcmp(ofproto_port.type, type)
779 && (!iface || !iface->netdev
780 || !strcmp(netdev_get_type(iface->netdev), type))) {
783 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
785 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
786 br->name, name, strerror(error));
789 netdev_close(iface->netdev);
790 iface->netdev = NULL;
796 iface_set_ofp_port(struct iface *iface, int ofp_port)
798 struct bridge *br = iface->port->bridge;
800 assert(iface->ofp_port < 0 && ofp_port >= 0);
801 iface->ofp_port = ofp_port;
802 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
803 iface_set_ofport(iface->cfg, ofp_port);
807 bridge_refresh_ofp_port(struct bridge *br)
809 struct ofproto_port_dump dump;
810 struct ofproto_port ofproto_port;
813 /* Clear all the "ofp_port"es. */
814 hmap_clear(&br->ifaces);
815 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
818 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
819 iface->ofp_port = -1;
823 /* Obtain the correct "ofp_port"s from ofproto. */
824 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
825 struct iface *iface = iface_lookup(br, ofproto_port.name);
827 if (iface->ofp_port >= 0) {
828 VLOG_WARN("bridge %s: interface %s reported twice",
829 br->name, ofproto_port.name);
830 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
831 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
832 br->name, ofproto_port.ofp_port);
834 iface_set_ofp_port(iface, ofproto_port.ofp_port);
840 /* Add an ofproto port for any "struct iface" that doesn't have one.
841 * Delete any "struct iface" for which this fails.
842 * Delete any "struct port" that thereby ends up with no ifaces. */
844 bridge_add_ofproto_ports(struct bridge *br)
846 struct port *port, *next_port;
848 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
849 struct iface *iface, *next_iface;
850 struct ofproto_port ofproto_port;
852 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
855 /* Open the netdev. */
856 if (!iface->netdev) {
857 error = netdev_open(iface->name, iface->type, &iface->netdev);
859 VLOG_WARN("could not open network device %s (%s)",
860 iface->name, strerror(error));
866 /* Configure the netdev. */
871 shash_from_ovs_idl_map(iface->cfg->key_options,
872 iface->cfg->value_options,
873 iface->cfg->n_options, &args);
874 error = netdev_set_config(iface->netdev, &args);
875 shash_destroy(&args);
878 VLOG_WARN("could not configure network device %s (%s)",
879 iface->name, strerror(error));
880 netdev_close(iface->netdev);
881 iface->netdev = NULL;
885 /* Add the port, if necessary. */
886 if (iface->netdev && iface->ofp_port < 0) {
890 error = ofproto_port_add(br->ofproto, iface->netdev,
893 iface_set_ofp_port(iface, ofp_port);
895 netdev_close(iface->netdev);
896 iface->netdev = NULL;
900 /* Populate stats columns in new Interface rows. */
901 if (iface->netdev && !iface->cfg->mtu) {
902 iface_refresh_stats(iface);
903 iface_refresh_status(iface);
906 /* Delete the iface if we failed. */
907 if (iface->netdev && iface->ofp_port >= 0) {
908 VLOG_DBG("bridge %s: interface %s is on port %d",
909 br->name, iface->name, iface->ofp_port);
912 VLOG_ERR("bridge %s: missing %s interface, dropping",
913 br->name, iface->name);
915 /* We already reported a related error, don't bother
918 iface_set_ofport(iface->cfg, -1);
919 iface_destroy(iface);
922 if (list_is_empty(&port->ifaces)) {
923 VLOG_WARN("%s port has no interfaces, dropping", port->name);
928 /* Add bond fake iface if necessary. */
929 if (port_is_bond_fake_iface(port)) {
930 if (ofproto_port_query_by_name(br->ofproto, port->name,
932 struct netdev *netdev;
935 error = netdev_open(port->name, "internal", &netdev);
937 ofproto_port_add(br->ofproto, netdev, NULL);
938 netdev_close(netdev);
940 VLOG_WARN("could not open network device %s (%s)",
941 port->name, strerror(error));
944 /* Already exists, nothing to do. */
945 ofproto_port_destroy(&ofproto_port);
952 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
953 const struct ovsdb_idl_column *column,
956 const struct ovsdb_datum *datum;
957 union ovsdb_atom atom;
960 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
961 atom.string = (char *) key;
962 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
963 return idx == UINT_MAX ? NULL : datum->values[idx].string;
967 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
969 return get_ovsrec_key_value(&br_cfg->header_,
970 &ovsrec_bridge_col_other_config, key);
973 /* Set Flow eviction threshold */
975 bridge_configure_flow_eviction_threshold(struct bridge *br)
977 const char *threshold_str;
980 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
982 threshold = strtoul(threshold_str, NULL, 10);
984 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
986 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
989 /* Set forward BPDU option. */
991 bridge_configure_forward_bpdu(struct bridge *br)
993 const char *forward_bpdu_str;
994 bool forward_bpdu = false;
996 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
997 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1000 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1004 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1005 struct iface **hw_addr_iface)
1009 bool found_addr = false;
1012 *hw_addr_iface = NULL;
1014 /* Did the user request a particular MAC? */
1015 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1016 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1017 if (eth_addr_is_multicast(ea)) {
1018 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1019 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1020 } else if (eth_addr_is_zero(ea)) {
1021 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1027 /* Otherwise choose the minimum non-local MAC address among all of the
1029 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1030 uint8_t iface_ea[ETH_ADDR_LEN];
1031 struct iface *candidate;
1032 struct iface *iface;
1034 /* Mirror output ports don't participate. */
1035 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1039 /* Choose the MAC address to represent the port. */
1041 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1042 /* Find the interface with this Ethernet address (if any) so that
1043 * we can provide the correct devname to the caller. */
1044 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1045 uint8_t candidate_ea[ETH_ADDR_LEN];
1046 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1047 && eth_addr_equals(iface_ea, candidate_ea)) {
1052 /* Choose the interface whose MAC address will represent the port.
1053 * The Linux kernel bonding code always chooses the MAC address of
1054 * the first slave added to a bond, and the Fedora networking
1055 * scripts always add slaves to a bond in alphabetical order, so
1056 * for compatibility we choose the interface with the name that is
1057 * first in alphabetical order. */
1058 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1059 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1064 /* The local port doesn't count (since we're trying to choose its
1065 * MAC address anyway). */
1066 if (iface->ofp_port == OFPP_LOCAL) {
1071 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1077 /* Compare against our current choice. */
1078 if (!eth_addr_is_multicast(iface_ea) &&
1079 !eth_addr_is_local(iface_ea) &&
1080 !eth_addr_is_reserved(iface_ea) &&
1081 !eth_addr_is_zero(iface_ea) &&
1082 eth_addr_compare_3way(iface_ea, ea) < 0)
1084 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1085 *hw_addr_iface = iface;
1090 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1091 br->name, ETH_ADDR_ARGS(ea));
1093 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1094 *hw_addr_iface = NULL;
1095 VLOG_WARN("bridge %s: using default bridge Ethernet "
1096 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1100 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1101 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1102 * an interface on 'br', then that interface must be passed in as
1103 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1104 * 'hw_addr_iface' must be passed in as a null pointer. */
1106 bridge_pick_datapath_id(struct bridge *br,
1107 const uint8_t bridge_ea[ETH_ADDR_LEN],
1108 struct iface *hw_addr_iface)
1111 * The procedure for choosing a bridge MAC address will, in the most
1112 * ordinary case, also choose a unique MAC that we can use as a datapath
1113 * ID. In some special cases, though, multiple bridges will end up with
1114 * the same MAC address. This is OK for the bridges, but it will confuse
1115 * the OpenFlow controller, because each datapath needs a unique datapath
1118 * Datapath IDs must be unique. It is also very desirable that they be
1119 * stable from one run to the next, so that policy set on a datapath
1122 const char *datapath_id;
1125 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1126 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1130 if (hw_addr_iface) {
1132 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1134 * A bridge whose MAC address is taken from a VLAN network device
1135 * (that is, a network device created with vconfig(8) or similar
1136 * tool) will have the same MAC address as a bridge on the VLAN
1137 * device's physical network device.
1139 * Handle this case by hashing the physical network device MAC
1140 * along with the VLAN identifier.
1142 uint8_t buf[ETH_ADDR_LEN + 2];
1143 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1144 buf[ETH_ADDR_LEN] = vlan >> 8;
1145 buf[ETH_ADDR_LEN + 1] = vlan;
1146 return dpid_from_hash(buf, sizeof buf);
1149 * Assume that this bridge's MAC address is unique, since it
1150 * doesn't fit any of the cases we handle specially.
1155 * A purely internal bridge, that is, one that has no non-virtual
1156 * network devices on it at all, is more difficult because it has no
1157 * natural unique identifier at all.
1159 * When the host is a XenServer, we handle this case by hashing the
1160 * host's UUID with the name of the bridge. Names of bridges are
1161 * persistent across XenServer reboots, although they can be reused if
1162 * an internal network is destroyed and then a new one is later
1163 * created, so this is fairly effective.
1165 * When the host is not a XenServer, we punt by using a random MAC
1166 * address on each run.
1168 const char *host_uuid = xenserver_get_host_uuid();
1170 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1171 dpid = dpid_from_hash(combined, strlen(combined));
1177 return eth_addr_to_uint64(bridge_ea);
1181 dpid_from_hash(const void *data, size_t n)
1183 uint8_t hash[SHA1_DIGEST_SIZE];
1185 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1186 sha1_bytes(data, n, hash);
1187 eth_addr_mark_random(hash);
1188 return eth_addr_to_uint64(hash);
1192 iface_refresh_status(struct iface *iface)
1196 enum netdev_flags flags;
1203 if (iface_is_synthetic(iface)) {
1209 if (!netdev_get_status(iface->netdev, &sh)) {
1211 char **keys, **values;
1213 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1214 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1219 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1222 shash_destroy_free_data(&sh);
1224 error = netdev_get_flags(iface->netdev, &flags);
1226 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1229 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1232 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1234 ovsrec_interface_set_duplex(iface->cfg,
1235 netdev_features_is_full_duplex(current)
1237 /* warning: uint64_t -> int64_t conversion */
1238 bps = netdev_features_to_bps(current);
1239 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1242 ovsrec_interface_set_duplex(iface->cfg, NULL);
1243 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1246 ovsrec_interface_set_link_state(iface->cfg,
1247 iface_get_carrier(iface) ? "up" : "down");
1249 error = netdev_get_mtu(iface->netdev, &mtu);
1250 if (!error && mtu != INT_MAX) {
1252 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1255 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1259 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1260 * changed, false otherwise. */
1262 iface_refresh_cfm_stats(struct iface *iface)
1264 const struct ovsrec_interface *cfg = iface->cfg;
1267 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1270 bool fault_bool = fault;
1271 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1273 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1278 iface_refresh_lacp_stats(struct iface *iface)
1280 struct ofproto *ofproto = iface->port->bridge->ofproto;
1281 int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
1282 int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
1286 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
1292 iface_refresh_stats(struct iface *iface)
1294 #define IFACE_STATS \
1295 IFACE_STAT(rx_packets, "rx_packets") \
1296 IFACE_STAT(tx_packets, "tx_packets") \
1297 IFACE_STAT(rx_bytes, "rx_bytes") \
1298 IFACE_STAT(tx_bytes, "tx_bytes") \
1299 IFACE_STAT(rx_dropped, "rx_dropped") \
1300 IFACE_STAT(tx_dropped, "tx_dropped") \
1301 IFACE_STAT(rx_errors, "rx_errors") \
1302 IFACE_STAT(tx_errors, "tx_errors") \
1303 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1304 IFACE_STAT(rx_over_errors, "rx_over_err") \
1305 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1306 IFACE_STAT(collisions, "collisions")
1308 #define IFACE_STAT(MEMBER, NAME) NAME,
1309 static char *keys[] = { IFACE_STATS };
1311 int64_t values[ARRAY_SIZE(keys)];
1314 struct netdev_stats stats;
1316 if (iface_is_synthetic(iface)) {
1320 /* Intentionally ignore return value, since errors will set 'stats' to
1321 * all-1s, and we will deal with that correctly below. */
1322 netdev_get_stats(iface->netdev, &stats);
1324 /* Copy statistics into values[] array. */
1326 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1329 assert(i == ARRAY_SIZE(keys));
1331 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1336 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1340 /* Use other-config:enable-system-stats by preference. */
1341 enable = get_ovsrec_key_value(&cfg->header_,
1342 &ovsrec_open_vswitch_col_other_config,
1343 "enable-statistics");
1345 return !strcmp(enable, "true");
1348 /* Disable by default. */
1353 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1355 struct ovsdb_datum datum;
1359 if (enable_system_stats(cfg)) {
1360 get_system_stats(&stats);
1363 ovsdb_datum_from_shash(&datum, &stats);
1364 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1368 static inline const char *
1369 nx_role_to_str(enum nx_role role)
1374 case NX_ROLE_MASTER:
1379 return "*** INVALID ROLE ***";
1384 refresh_controller_status(void)
1388 const struct ovsrec_controller *cfg;
1392 /* Accumulate status for controllers on all bridges. */
1393 HMAP_FOR_EACH (br, node, &all_bridges) {
1394 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1397 /* Update each controller in the database with current status. */
1398 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1399 struct ofproto_controller_info *cinfo =
1400 shash_find_data(&info, cfg->target);
1403 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1404 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1405 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1406 (char **) cinfo->pairs.values,
1409 ovsrec_controller_set_is_connected(cfg, false);
1410 ovsrec_controller_set_role(cfg, NULL);
1411 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1415 ofproto_free_ofproto_controller_info(&info);
1419 refresh_cfm_stats(void)
1421 static struct ovsdb_idl_txn *txn = NULL;
1426 txn = ovsdb_idl_txn_create(idl);
1428 HMAP_FOR_EACH (br, node, &all_bridges) {
1429 struct iface *iface;
1431 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1432 iface_refresh_cfm_stats(iface);
1437 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1438 ovsdb_idl_txn_destroy(txn);
1446 const struct ovsrec_open_vswitch *cfg;
1448 bool datapath_destroyed;
1449 bool database_changed;
1452 /* (Re)configure if necessary. */
1453 database_changed = ovsdb_idl_run(idl);
1454 if (ovsdb_idl_is_lock_contended(idl)) {
1455 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1456 struct bridge *br, *next_br;
1458 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1459 "disabling this process until it goes away");
1461 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1465 } else if (!ovsdb_idl_has_lock(idl)) {
1468 cfg = ovsrec_open_vswitch_first(idl);
1470 /* Let each bridge do the work that it needs to do. */
1471 datapath_destroyed = false;
1472 HMAP_FOR_EACH (br, node, &all_bridges) {
1473 int error = ofproto_run(br->ofproto);
1475 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1476 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1477 "forcing reconfiguration", br->name);
1478 datapath_destroyed = true;
1482 /* Re-configure SSL. We do this on every trip through the main loop,
1483 * instead of just when the database changes, because the contents of the
1484 * key and certificate files can change without the database changing.
1486 * We do this before bridge_reconfigure() because that function might
1487 * initiate SSL connections and thus requires SSL to be configured. */
1488 if (cfg && cfg->ssl) {
1489 const struct ovsrec_ssl *ssl = cfg->ssl;
1491 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1492 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1495 if (database_changed || datapath_destroyed) {
1497 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1499 bridge_reconfigure(cfg);
1501 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1502 ovsdb_idl_txn_commit(txn);
1503 ovsdb_idl_txn_destroy(txn); /* XXX */
1505 /* We still need to reconfigure to avoid dangling pointers to
1506 * now-destroyed ovsrec structures inside bridge data. */
1507 static const struct ovsrec_open_vswitch null_cfg;
1509 bridge_reconfigure(&null_cfg);
1513 /* Refresh system and interface stats if necessary. */
1514 if (time_msec() >= stats_timer) {
1516 struct ovsdb_idl_txn *txn;
1518 txn = ovsdb_idl_txn_create(idl);
1519 HMAP_FOR_EACH (br, node, &all_bridges) {
1522 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1523 struct iface *iface;
1525 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1526 iface_refresh_stats(iface);
1527 iface_refresh_status(iface);
1531 refresh_system_stats(cfg);
1532 refresh_controller_status();
1533 ovsdb_idl_txn_commit(txn);
1534 ovsdb_idl_txn_destroy(txn); /* XXX */
1537 stats_timer = time_msec() + STATS_INTERVAL;
1540 if (time_msec() >= db_limiter) {
1541 struct ovsdb_idl_txn *txn;
1542 bool changed = false;
1544 txn = ovsdb_idl_txn_create(idl);
1545 HMAP_FOR_EACH (br, node, &all_bridges) {
1548 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1549 struct iface *iface;
1551 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1552 /* XXX: Eventually we need to remove the lacp_current flag
1553 * from the database so that we can completely get rid of
1554 * this rate limiter code. */
1555 changed = iface_refresh_lacp_stats(iface) || changed;
1561 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1564 ovsdb_idl_txn_commit(txn);
1565 ovsdb_idl_txn_destroy(txn);
1568 refresh_cfm_stats();
1574 ovsdb_idl_wait(idl);
1575 if (!hmap_is_empty(&all_bridges)) {
1578 HMAP_FOR_EACH (br, node, &all_bridges) {
1579 ofproto_wait(br->ofproto);
1581 poll_timer_wait_until(stats_timer);
1583 if (db_limiter > time_msec()) {
1584 poll_timer_wait_until(db_limiter);
1589 /* QoS unixctl user interface functions. */
1591 struct qos_unixctl_show_cbdata {
1593 struct iface *iface;
1597 qos_unixctl_show_cb(unsigned int queue_id,
1598 const struct shash *details,
1601 struct qos_unixctl_show_cbdata *data = aux;
1602 struct ds *ds = data->ds;
1603 struct iface *iface = data->iface;
1604 struct netdev_queue_stats stats;
1605 struct shash_node *node;
1608 ds_put_cstr(ds, "\n");
1610 ds_put_format(ds, "Queue %u:\n", queue_id);
1612 ds_put_cstr(ds, "Default:\n");
1615 SHASH_FOR_EACH (node, details) {
1616 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1619 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1621 if (stats.tx_packets != UINT64_MAX) {
1622 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1625 if (stats.tx_bytes != UINT64_MAX) {
1626 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1629 if (stats.tx_errors != UINT64_MAX) {
1630 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1633 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1634 queue_id, strerror(error));
1639 qos_unixctl_show(struct unixctl_conn *conn,
1640 const char *args, void *aux OVS_UNUSED)
1642 struct ds ds = DS_EMPTY_INITIALIZER;
1643 struct shash sh = SHASH_INITIALIZER(&sh);
1644 struct iface *iface;
1646 struct shash_node *node;
1647 struct qos_unixctl_show_cbdata data;
1650 iface = iface_find(args);
1652 unixctl_command_reply(conn, 501, "no such interface");
1656 netdev_get_qos(iface->netdev, &type, &sh);
1658 if (*type != '\0') {
1659 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1661 SHASH_FOR_EACH (node, &sh) {
1662 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1667 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1670 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1672 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1674 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1675 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1678 shash_destroy_free_data(&sh);
1682 /* Bridge reconfiguration functions. */
1684 bridge_create(const struct ovsrec_bridge *br_cfg)
1688 assert(!bridge_lookup(br_cfg->name));
1689 br = xzalloc(sizeof *br);
1691 br->name = xstrdup(br_cfg->name);
1692 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1695 /* Derive the default Ethernet address from the bridge's UUID. This should
1696 * be unique and it will be stable between ovs-vswitchd runs. */
1697 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1698 eth_addr_mark_random(br->default_ea);
1700 hmap_init(&br->ports);
1701 hmap_init(&br->ifaces);
1702 hmap_init(&br->iface_by_name);
1703 hmap_init(&br->mirrors);
1705 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1709 bridge_destroy(struct bridge *br)
1712 struct mirror *mirror, *next_mirror;
1713 struct port *port, *next_port;
1715 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1718 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1719 mirror_destroy(mirror);
1721 hmap_remove(&all_bridges, &br->node);
1722 ofproto_destroy(br->ofproto);
1723 hmap_destroy(&br->ifaces);
1724 hmap_destroy(&br->ports);
1725 hmap_destroy(&br->iface_by_name);
1726 hmap_destroy(&br->mirrors);
1733 static struct bridge *
1734 bridge_lookup(const char *name)
1738 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1739 if (!strcmp(br->name, name)) {
1746 /* Handle requests for a listing of all flows known by the OpenFlow
1747 * stack, including those normally hidden. */
1749 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1750 const char *args, void *aux OVS_UNUSED)
1755 br = bridge_lookup(args);
1757 unixctl_command_reply(conn, 501, "Unknown bridge");
1762 ofproto_get_all_flows(br->ofproto, &results);
1764 unixctl_command_reply(conn, 200, ds_cstr(&results));
1765 ds_destroy(&results);
1768 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1769 * connections and reconnect. If BRIDGE is not specified, then all bridges
1770 * drop their controller connections and reconnect. */
1772 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1773 const char *args, void *aux OVS_UNUSED)
1776 if (args[0] != '\0') {
1777 br = bridge_lookup(args);
1779 unixctl_command_reply(conn, 501, "Unknown bridge");
1782 ofproto_reconnect_controllers(br->ofproto);
1784 HMAP_FOR_EACH (br, node, &all_bridges) {
1785 ofproto_reconnect_controllers(br->ofproto);
1788 unixctl_command_reply(conn, 200, NULL);
1792 bridge_get_controllers(const struct bridge *br,
1793 struct ovsrec_controller ***controllersp)
1795 struct ovsrec_controller **controllers;
1796 size_t n_controllers;
1798 controllers = br->cfg->controller;
1799 n_controllers = br->cfg->n_controller;
1801 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1807 *controllersp = controllers;
1809 return n_controllers;
1812 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1813 * those configured in 'br->cfg'. */
1815 bridge_add_del_ports(struct bridge *br)
1817 struct port *port, *next;
1818 struct shash_node *node;
1819 struct shash new_ports;
1822 /* Collect new ports. */
1823 shash_init(&new_ports);
1824 for (i = 0; i < br->cfg->n_ports; i++) {
1825 const char *name = br->cfg->ports[i]->name;
1826 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1827 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1831 if (bridge_get_controllers(br, NULL)
1832 && !shash_find(&new_ports, br->name)) {
1833 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1834 br->name, br->name);
1836 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1837 br->synth_local_port.n_interfaces = 1;
1838 br->synth_local_port.name = br->name;
1840 br->synth_local_iface.name = br->name;
1841 br->synth_local_iface.type = "internal";
1843 br->synth_local_ifacep = &br->synth_local_iface;
1845 shash_add(&new_ports, br->name, &br->synth_local_port);
1848 /* Get rid of deleted ports.
1849 * Get rid of deleted interfaces on ports that still exist.
1850 * Update 'cfg' of ports that still exist. */
1851 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1852 port->cfg = shash_find_data(&new_ports, port->name);
1856 port_del_ifaces(port);
1860 /* Create new ports.
1861 * Add new interfaces to existing ports. */
1862 SHASH_FOR_EACH (node, &new_ports) {
1863 struct port *port = port_lookup(br, node->name);
1865 struct ovsrec_port *cfg = node->data;
1866 port = port_create(br, cfg);
1868 port_add_ifaces(port);
1869 if (list_is_empty(&port->ifaces)) {
1870 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1871 br->name, port->name);
1875 shash_destroy(&new_ports);
1878 /* Initializes 'oc' appropriately as a management service controller for
1881 * The caller must free oc->target when it is no longer needed. */
1883 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1884 struct ofproto_controller *oc)
1886 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1887 oc->max_backoff = 0;
1888 oc->probe_interval = 60;
1889 oc->band = OFPROTO_OUT_OF_BAND;
1891 oc->burst_limit = 0;
1894 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1896 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1897 struct ofproto_controller *oc)
1899 oc->target = c->target;
1900 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1901 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1902 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1903 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1904 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1905 oc->burst_limit = (c->controller_burst_limit
1906 ? *c->controller_burst_limit : 0);
1909 /* Configures the IP stack for 'br''s local interface properly according to the
1910 * configuration in 'c'. */
1912 bridge_configure_local_iface_netdev(struct bridge *br,
1913 struct ovsrec_controller *c)
1915 struct netdev *netdev;
1916 struct in_addr mask, gateway;
1918 struct iface *local_iface;
1921 /* If there's no local interface or no IP address, give up. */
1922 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1923 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1927 /* Bring up the local interface. */
1928 netdev = local_iface->netdev;
1929 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1931 /* Configure the IP address and netmask. */
1932 if (!c->local_netmask
1933 || !inet_aton(c->local_netmask, &mask)
1935 mask.s_addr = guess_netmask(ip.s_addr);
1937 if (!netdev_set_in4(netdev, ip, mask)) {
1938 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1939 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1942 /* Configure the default gateway. */
1943 if (c->local_gateway
1944 && inet_aton(c->local_gateway, &gateway)
1945 && gateway.s_addr) {
1946 if (!netdev_add_router(netdev, gateway)) {
1947 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1948 br->name, IP_ARGS(&gateway.s_addr));
1954 bridge_configure_remotes(struct bridge *br,
1955 const struct sockaddr_in *managers, size_t n_managers)
1957 const char *disable_ib_str, *queue_id_str;
1958 bool disable_in_band = false;
1961 struct ovsrec_controller **controllers;
1962 size_t n_controllers;
1964 enum ofproto_fail_mode fail_mode;
1966 struct ofproto_controller *ocs;
1970 /* Check if we should disable in-band control on this bridge. */
1971 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1972 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1973 disable_in_band = true;
1976 /* Set OpenFlow queue ID for in-band control. */
1977 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1978 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1979 ofproto_set_in_band_queue(br->ofproto, queue_id);
1981 if (disable_in_band) {
1982 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1984 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1987 n_controllers = bridge_get_controllers(br, &controllers);
1989 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1992 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1993 for (i = 0; i < n_controllers; i++) {
1994 struct ovsrec_controller *c = controllers[i];
1996 if (!strncmp(c->target, "punix:", 6)
1997 || !strncmp(c->target, "unix:", 5)) {
1998 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2000 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2001 * domain sockets and overwriting arbitrary local files. */
2002 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
2003 "controller \"%s\" due to possibility for remote "
2004 "exploit", br->name, c->target);
2008 bridge_configure_local_iface_netdev(br, c);
2009 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2010 if (disable_in_band) {
2011 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2016 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2017 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2020 /* Set the fail-mode. */
2021 fail_mode = !br->cfg->fail_mode
2022 || !strcmp(br->cfg->fail_mode, "standalone")
2023 ? OFPROTO_FAIL_STANDALONE
2024 : OFPROTO_FAIL_SECURE;
2025 ofproto_set_fail_mode(br->ofproto, fail_mode);
2027 /* Configure OpenFlow controller connection snooping. */
2028 if (!ofproto_has_snoops(br->ofproto)) {
2032 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2033 ovs_rundir(), br->name));
2034 ofproto_set_snoops(br->ofproto, &snoops);
2035 sset_destroy(&snoops);
2039 /* Port functions. */
2041 static struct port *
2042 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2046 port = xzalloc(sizeof *port);
2048 port->name = xstrdup(cfg->name);
2050 list_init(&port->ifaces);
2052 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2054 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2060 get_port_other_config(const struct ovsrec_port *port, const char *key,
2061 const char *default_value)
2065 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2067 return value ? value : default_value;
2071 get_interface_other_config(const struct ovsrec_interface *iface,
2072 const char *key, const char *default_value)
2076 value = get_ovsrec_key_value(&iface->header_,
2077 &ovsrec_interface_col_other_config, key);
2078 return value ? value : default_value;
2081 /* Deletes interfaces from 'port' that are no longer configured for it. */
2083 port_del_ifaces(struct port *port)
2085 struct iface *iface, *next;
2086 struct sset new_ifaces;
2089 /* Collect list of new interfaces. */
2090 sset_init(&new_ifaces);
2091 for (i = 0; i < port->cfg->n_interfaces; i++) {
2092 const char *name = port->cfg->interfaces[i]->name;
2093 const char *type = port->cfg->interfaces[i]->name;
2094 if (strcmp(type, "null")) {
2095 sset_add(&new_ifaces, name);
2099 /* Get rid of deleted interfaces. */
2100 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2101 if (!sset_contains(&new_ifaces, iface->name)) {
2102 iface_destroy(iface);
2106 sset_destroy(&new_ifaces);
2109 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2112 port_add_ifaces(struct port *port)
2114 struct shash new_ifaces;
2115 struct shash_node *node;
2118 /* Collect new ifaces. */
2119 shash_init(&new_ifaces);
2120 for (i = 0; i < port->cfg->n_interfaces; i++) {
2121 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2122 if (strcmp(cfg->type, "null")
2123 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2124 VLOG_WARN("port %s: %s specified twice as port interface",
2125 port->name, cfg->name);
2126 iface_set_ofport(cfg, -1);
2130 /* Create new interfaces.
2131 * Update interface types and 'cfg' members. */
2132 SHASH_FOR_EACH (node, &new_ifaces) {
2133 const struct ovsrec_interface *cfg = node->data;
2134 const char *iface_name = node->name;
2135 struct iface *iface;
2137 iface = iface_lookup(port->bridge, iface_name);
2139 iface = iface_create(port, cfg);
2144 /* Determine interface type. The local port always has type
2145 * "internal". Other ports take their type from the database and
2146 * default to "system" if none is specified. */
2147 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2148 : cfg->type[0] ? cfg->type
2151 shash_destroy(&new_ifaces);
2155 port_destroy(struct port *port)
2158 struct bridge *br = port->bridge;
2159 struct iface *iface, *next;
2162 ofproto_bundle_unregister(br->ofproto, port);
2165 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2166 iface_destroy(iface);
2169 hmap_remove(&br->ports, &port->hmap_node);
2171 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2178 static struct port *
2179 port_lookup(const struct bridge *br, const char *name)
2183 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2185 if (!strcmp(port->name, name)) {
2193 enable_lacp(struct port *port, bool *activep)
2195 if (!port->cfg->lacp) {
2196 /* XXX when LACP implementation has been sufficiently tested, enable by
2197 * default and make active on bonded ports. */
2199 } else if (!strcmp(port->cfg->lacp, "off")) {
2201 } else if (!strcmp(port->cfg->lacp, "active")) {
2204 } else if (!strcmp(port->cfg->lacp, "passive")) {
2208 VLOG_WARN("port %s: unknown LACP mode %s",
2209 port->name, port->cfg->lacp);
2214 static struct lacp_settings *
2215 port_configure_lacp(struct port *port, struct lacp_settings *s)
2217 const char *lacp_time;
2218 long long int custom_time;
2221 if (!enable_lacp(port, &s->active)) {
2225 s->name = port->name;
2226 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2228 /* Prefer bondable links if unspecified. */
2229 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2231 s->priority = (priority > 0 && priority <= UINT16_MAX
2233 : UINT16_MAX - !list_is_short(&port->ifaces));
2235 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2240 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2241 custom_time = atoi(lacp_time);
2242 if (!strcmp(lacp_time, "fast")) {
2243 s->lacp_time = LACP_TIME_FAST;
2244 } else if (!strcmp(lacp_time, "slow")) {
2245 s->lacp_time = LACP_TIME_SLOW;
2246 } else if (custom_time > 0) {
2247 s->lacp_time = LACP_TIME_CUSTOM;
2248 s->custom_time = custom_time;
2250 s->lacp_time = LACP_TIME_SLOW;
2257 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2259 int priority, portid, key;
2261 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2262 priority = atoi(get_interface_other_config(iface->cfg,
2263 "lacp-port-priority", "0"));
2264 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2267 if (portid <= 0 || portid > UINT16_MAX) {
2268 portid = iface->ofp_port;
2271 if (priority <= 0 || priority > UINT16_MAX) {
2272 priority = UINT16_MAX;
2275 if (key < 0 || key > UINT16_MAX) {
2279 s->name = iface->name;
2281 s->priority = priority;
2286 port_configure_bond(struct port *port, struct bond_settings *s,
2287 uint32_t *bond_stable_ids)
2289 const char *detect_s;
2290 struct iface *iface;
2291 int miimon_interval;
2294 s->name = port->name;
2295 s->balance = BM_SLB;
2296 if (port->cfg->bond_mode
2297 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2298 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2299 port->name, port->cfg->bond_mode,
2300 bond_mode_to_string(s->balance));
2302 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2303 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2304 "please use another bond type or disable flood_vlans",
2308 miimon_interval = atoi(get_port_other_config(port->cfg,
2309 "bond-miimon-interval", "0"));
2310 if (miimon_interval <= 0) {
2311 miimon_interval = 200;
2314 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2315 if (!strcmp(detect_s, "carrier")) {
2316 miimon_interval = 0;
2317 } else if (strcmp(detect_s, "miimon")) {
2318 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2319 "defaulting to carrier", port->name, detect_s);
2320 miimon_interval = 0;
2323 s->up_delay = MAX(0, port->cfg->bond_updelay);
2324 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2325 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2326 s->rebalance_interval = atoi(
2327 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2328 if (s->rebalance_interval < 1000) {
2329 s->rebalance_interval = 1000;
2332 s->fake_iface = port->cfg->bond_fake_iface;
2335 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2336 long long stable_id;
2338 stable_id = atoll(get_interface_other_config(iface->cfg,
2339 "bond-stable-id", "0"));
2340 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2341 stable_id = iface->ofp_port;
2343 bond_stable_ids[i++] = stable_id;
2345 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2349 /* Interface functions. */
2351 static struct iface *
2352 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2354 struct bridge *br = port->bridge;
2355 struct iface *iface;
2356 char *name = if_cfg->name;
2358 iface = xzalloc(sizeof *iface);
2360 iface->name = xstrdup(name);
2361 iface->ofp_port = -1;
2362 iface->tag = tag_create_random();
2363 iface->netdev = NULL;
2364 iface->cfg = if_cfg;
2366 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2368 list_push_back(&port->ifaces, &iface->port_elem);
2370 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2376 iface_destroy(struct iface *iface)
2379 struct port *port = iface->port;
2380 struct bridge *br = port->bridge;
2382 if (br->ofproto && iface->ofp_port >= 0) {
2383 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2386 if (iface->ofp_port >= 0) {
2387 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2390 list_remove(&iface->port_elem);
2391 hmap_remove(&br->iface_by_name, &iface->name_node);
2393 netdev_close(iface->netdev);
2400 static struct iface *
2401 iface_lookup(const struct bridge *br, const char *name)
2403 struct iface *iface;
2405 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2406 &br->iface_by_name) {
2407 if (!strcmp(iface->name, name)) {
2415 static struct iface *
2416 iface_find(const char *name)
2418 const struct bridge *br;
2420 HMAP_FOR_EACH (br, node, &all_bridges) {
2421 struct iface *iface = iface_lookup(br, name);
2430 static struct iface *
2431 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2433 struct iface *iface;
2435 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2436 hash_int(ofp_port, 0), &br->ifaces) {
2437 if (iface->ofp_port == ofp_port) {
2444 /* Set Ethernet address of 'iface', if one is specified in the configuration
2447 iface_set_mac(struct iface *iface)
2449 uint8_t ea[ETH_ADDR_LEN];
2451 if (!strcmp(iface->type, "internal")
2452 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2453 if (iface->ofp_port == OFPP_LOCAL) {
2454 VLOG_ERR("interface %s: ignoring mac in Interface record "
2455 "(use Bridge record to set local port's mac)",
2457 } else if (eth_addr_is_multicast(ea)) {
2458 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2461 int error = netdev_set_etheraddr(iface->netdev, ea);
2463 VLOG_ERR("interface %s: setting MAC failed (%s)",
2464 iface->name, strerror(error));
2470 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2472 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2474 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2475 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2479 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2481 * The value strings in '*shash' are taken directly from values[], not copied,
2482 * so the caller should not modify or free them. */
2484 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2485 struct shash *shash)
2490 for (i = 0; i < n; i++) {
2491 shash_add(shash, keys[i], values[i]);
2495 /* Creates 'keys' and 'values' arrays from 'shash'.
2497 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2498 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2499 * are populated with with strings taken directly from 'shash' and thus have
2500 * the same ownership of the key-value pairs in shash.
2503 shash_to_ovs_idl_map(struct shash *shash,
2504 char ***keys, char ***values, size_t *n)
2508 struct shash_node *sn;
2510 count = shash_count(shash);
2512 k = xmalloc(count * sizeof *k);
2513 v = xmalloc(count * sizeof *v);
2516 SHASH_FOR_EACH(sn, shash) {
2527 struct iface_delete_queues_cbdata {
2528 struct netdev *netdev;
2529 const struct ovsdb_datum *queues;
2533 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2535 union ovsdb_atom atom;
2537 atom.integer = target;
2538 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2542 iface_delete_queues(unsigned int queue_id,
2543 const struct shash *details OVS_UNUSED, void *cbdata_)
2545 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2547 if (!queue_ids_include(cbdata->queues, queue_id)) {
2548 netdev_delete_queue(cbdata->netdev, queue_id);
2553 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2555 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2556 netdev_set_qos(iface->netdev, NULL, NULL);
2558 struct iface_delete_queues_cbdata cbdata;
2559 struct shash details;
2563 /* Configure top-level Qos for 'iface'. */
2564 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2565 qos->n_other_config, &details);
2566 netdev_set_qos(iface->netdev, qos->type, &details);
2567 shash_destroy(&details);
2569 /* Deconfigure queues that were deleted. */
2570 cbdata.netdev = iface->netdev;
2571 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2573 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2575 /* Configure queues for 'iface'. */
2577 for (i = 0; i < qos->n_queues; i++) {
2578 const struct ovsrec_queue *queue = qos->value_queues[i];
2579 unsigned int queue_id = qos->key_queues[i];
2581 if (queue_id == 0) {
2585 shash_from_ovs_idl_map(queue->key_other_config,
2586 queue->value_other_config,
2587 queue->n_other_config, &details);
2588 netdev_set_queue(iface->netdev, queue_id, &details);
2589 shash_destroy(&details);
2592 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2593 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2594 "queue (queue 0). Packets not directed to a "
2595 "correctly configured queue may be dropped.",
2600 netdev_set_policing(iface->netdev,
2601 iface->cfg->ingress_policing_rate,
2602 iface->cfg->ingress_policing_burst);
2606 iface_configure_cfm(struct iface *iface)
2608 const struct ovsrec_interface *cfg = iface->cfg;
2609 struct cfm_settings s;
2611 if (!cfg->n_cfm_mpid) {
2612 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2616 s.mpid = *cfg->cfm_mpid;
2617 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2619 if (s.interval <= 0) {
2623 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2626 /* Read carrier or miimon status directly from 'iface''s netdev, according to
2627 * how 'iface''s port is configured.
2629 * Returns true if 'iface' is up, false otherwise. */
2631 iface_get_carrier(const struct iface *iface)
2634 return netdev_get_carrier(iface->netdev);
2637 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2638 * instead of obtaining it from the database. */
2640 iface_is_synthetic(const struct iface *iface)
2642 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2645 /* Port mirroring. */
2647 static struct mirror *
2648 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2652 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2653 if (uuid_equals(uuid, &m->uuid)) {
2661 bridge_configure_mirrors(struct bridge *br)
2663 const struct ovsdb_datum *mc;
2664 unsigned long *flood_vlans;
2665 struct mirror *m, *next;
2668 /* Get rid of deleted mirrors. */
2669 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2670 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2671 union ovsdb_atom atom;
2673 atom.uuid = m->uuid;
2674 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2679 /* Add new mirrors and reconfigure existing ones. */
2680 for (i = 0; i < br->cfg->n_mirrors; i++) {
2681 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2682 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2684 m = mirror_create(br, cfg);
2686 if (!mirror_configure(m, cfg)) {
2691 /* Update flooded vlans (for RSPAN). */
2692 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2693 br->cfg->n_flood_vlans);
2694 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2695 bitmap_free(flood_vlans);
2698 static struct mirror *
2699 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2703 m = xzalloc(sizeof *m);
2704 m->uuid = cfg->header_.uuid;
2705 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2707 m->name = xstrdup(cfg->name);
2713 mirror_destroy(struct mirror *m)
2716 struct bridge *br = m->bridge;
2719 ofproto_mirror_unregister(br->ofproto, m);
2722 hmap_remove(&br->mirrors, &m->hmap_node);
2729 mirror_collect_ports(struct mirror *m,
2730 struct ovsrec_port **in_ports, int n_in_ports,
2731 void ***out_portsp, size_t *n_out_portsp)
2733 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2734 size_t n_out_ports = 0;
2737 for (i = 0; i < n_in_ports; i++) {
2738 const char *name = in_ports[i]->name;
2739 struct port *port = port_lookup(m->bridge, name);
2741 out_ports[n_out_ports++] = port;
2743 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2744 "port %s", m->bridge->name, m->name, name);
2747 *out_portsp = out_ports;
2748 *n_out_portsp = n_out_ports;
2752 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2754 struct ofproto_mirror_settings s;
2757 if (strcmp(cfg->name, m->name)) {
2759 m->name = xstrdup(cfg->name);
2763 /* Get output port or VLAN. */
2764 if (cfg->output_port) {
2765 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2766 if (!s.out_bundle) {
2767 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2768 m->bridge->name, m->name);
2771 s.out_vlan = UINT16_MAX;
2773 if (cfg->output_vlan) {
2774 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2775 "output vlan; ignoring output vlan",
2776 m->bridge->name, m->name);
2778 } else if (cfg->output_vlan) {
2779 /* The database should prevent invalid VLAN values. */
2780 s.out_bundle = NULL;
2781 s.out_vlan = *cfg->output_vlan;
2783 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2784 m->bridge->name, m->name);
2788 /* Get port selection. */
2789 if (cfg->select_all) {
2790 size_t n_ports = hmap_count(&m->bridge->ports);
2791 void **ports = xmalloc(n_ports * sizeof *ports);
2796 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2806 /* Get ports, dropping ports that don't exist.
2807 * The IDL ensures that there are no duplicates. */
2808 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2809 &s.srcs, &s.n_srcs);
2810 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2811 &s.dsts, &s.n_dsts);
2814 /* Get VLAN selection. */
2815 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2818 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2821 if (s.srcs != s.dsts) {