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_sflow(struct bridge *, int *sflow_bridge_number);
154 static void bridge_configure_remotes(struct bridge *,
155 const struct sockaddr_in *managers,
157 static void bridge_pick_local_hw_addr(struct bridge *,
158 uint8_t ea[ETH_ADDR_LEN],
159 struct iface **hw_addr_iface);
160 static uint64_t bridge_pick_datapath_id(struct bridge *,
161 const uint8_t bridge_ea[ETH_ADDR_LEN],
162 struct iface *hw_addr_iface);
163 static uint64_t dpid_from_hash(const void *, size_t nbytes);
164 static bool bridge_has_bond_fake_iface(const struct bridge *,
166 static bool port_is_bond_fake_iface(const struct port *);
168 static unixctl_cb_func qos_unixctl_show;
170 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
171 static void port_add_ifaces(struct port *);
172 static void port_del_ifaces(struct port *);
173 static void port_destroy(struct port *);
174 static struct port *port_lookup(const struct bridge *, const char *name);
175 static void port_configure(struct port *);
176 static struct lacp_settings *port_configure_lacp(struct port *,
177 struct lacp_settings *);
178 static void port_configure_bond(struct port *, struct bond_settings *,
179 uint32_t *bond_stable_ids);
181 static void bridge_configure_mirrors(struct bridge *);
182 static struct mirror *mirror_create(struct bridge *,
183 const struct ovsrec_mirror *);
184 static void mirror_destroy(struct mirror *);
185 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
187 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
188 static struct iface *iface_create(struct port *port,
189 const struct ovsrec_interface *if_cfg);
190 static void iface_destroy(struct iface *);
191 static struct iface *iface_lookup(const struct bridge *, const char *name);
192 static struct iface *iface_find(const char *name);
193 static struct iface *iface_from_ofp_port(const struct bridge *,
195 static void iface_set_mac(struct iface *);
196 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
197 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
198 static void iface_configure_cfm(struct iface *);
199 static bool iface_refresh_cfm_stats(struct iface *);
200 static void iface_refresh_stats(struct iface *);
201 static void iface_refresh_status(struct iface *);
202 static bool iface_get_carrier(const struct iface *);
203 static bool iface_is_synthetic(const struct iface *);
205 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
207 static void shash_to_ovs_idl_map(struct shash *,
208 char ***keys, char ***values, size_t *n);
210 /* Public functions. */
212 /* Initializes the bridge module, configuring it to obtain its configuration
213 * from an OVSDB server accessed over 'remote', which should be a string in a
214 * form acceptable to ovsdb_idl_create(). */
216 bridge_init(const char *remote)
218 /* Create connection to database. */
219 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
220 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
222 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
223 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
224 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
225 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
226 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
227 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
228 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
230 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
231 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
233 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
234 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
236 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
237 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
238 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
239 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
240 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
241 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
242 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
243 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
244 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
246 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
247 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
248 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
249 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
251 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
253 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
255 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
257 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
259 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
261 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
262 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
263 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
264 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
265 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
267 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
269 /* Register unixctl commands. */
270 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
271 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
273 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
283 struct bridge *br, *next_br;
285 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
288 ovsdb_idl_destroy(idl);
291 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
292 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
293 * responsible for freeing '*managersp' (with free()).
295 * You may be asking yourself "why does ovs-vswitchd care?", because
296 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
297 * should not be and in fact is not directly involved in that. But
298 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
299 * it has to tell in-band control where the managers are to enable that.
300 * (Thus, only managers connected in-band are collected.)
303 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
304 struct sockaddr_in **managersp, size_t *n_managersp)
306 struct sockaddr_in *managers = NULL;
307 size_t n_managers = 0;
311 /* Collect all of the potential targets from the "targets" columns of the
312 * rows pointed to by "manager_options", excluding any that are
315 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
316 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
318 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
319 sset_find_and_delete(&targets, m->target);
321 sset_add(&targets, m->target);
325 /* Now extract the targets' IP addresses. */
326 if (!sset_is_empty(&targets)) {
329 managers = xmalloc(sset_count(&targets) * sizeof *managers);
330 SSET_FOR_EACH (target, &targets) {
331 struct sockaddr_in *sin = &managers[n_managers];
333 if ((!strncmp(target, "tcp:", 4)
334 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
335 (!strncmp(target, "ssl:", 4)
336 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
341 sset_destroy(&targets);
343 *managersp = managers;
344 *n_managersp = n_managers;
348 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
350 struct sockaddr_in *managers;
351 struct bridge *br, *next;
352 int sflow_bridge_number;
355 COVERAGE_INC(bridge_reconfigure);
357 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
358 * iface"s according to 'ovs_cfg', with only very minimal configuration
361 * This is purely an update to bridge data structures. Nothing is pushed
362 * down to ofproto or lower layers. */
363 add_del_bridges(ovs_cfg);
364 HMAP_FOR_EACH (br, node, &all_bridges) {
365 bridge_add_del_ports(br);
368 /* Delete all datapaths and datapath ports that are no longer configured.
370 * The kernel will reject any attempt to add a given port to a datapath if
371 * that port already belongs to a different datapath, so we must do all
372 * port deletions before any port additions. A datapath always has a
373 * "local port" so we must delete not-configured datapaths too. */
374 bridge_del_ofprotos();
375 HMAP_FOR_EACH (br, node, &all_bridges) {
377 bridge_del_ofproto_ports(br);
381 /* Create datapaths and datapath ports that are missing.
383 * After this is done, we have our final set of bridges, ports, and
384 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
385 * has at least one iface, every "struct iface" has a valid ofp_port and
387 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
388 if (!br->ofproto && !bridge_add_ofprotos(br)) {
392 HMAP_FOR_EACH (br, node, &all_bridges) {
393 bridge_refresh_ofp_port(br);
394 bridge_add_ofproto_ports(br);
397 /* Complete the configuration. */
398 sflow_bridge_number = 0;
399 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
400 HMAP_FOR_EACH (br, node, &all_bridges) {
403 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
406 port_configure(port);
408 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
409 iface_configure_cfm(iface);
410 iface_configure_qos(iface, port->cfg->qos);
411 iface_set_mac(iface);
414 bridge_configure_mirrors(br);
415 bridge_configure_datapath_id(br);
416 bridge_configure_flow_eviction_threshold(br);
417 bridge_configure_remotes(br, managers, n_managers);
418 bridge_configure_netflow(br);
419 bridge_configure_sflow(br, &sflow_bridge_number);
423 /* ovs-vswitchd has completed initialization, so allow the process that
424 * forked us to exit successfully. */
425 daemonize_complete();
428 /* Iterate over all ofprotos and delete any of them that do not have a
429 * configured bridge or that are the wrong type. */
431 bridge_del_ofprotos(void)
439 ofproto_enumerate_types(&types);
440 SSET_FOR_EACH (type, &types) {
443 ofproto_enumerate_names(type, &names);
444 SSET_FOR_EACH (name, &names) {
445 struct bridge *br = bridge_lookup(name);
446 if (!br || strcmp(type, br->type)) {
447 ofproto_delete(name, type);
451 sset_destroy(&names);
452 sset_destroy(&types);
456 bridge_add_ofprotos(struct bridge *br)
458 int error = ofproto_create(br->name, br->type, &br->ofproto);
460 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
467 port_configure(struct port *port)
469 const struct ovsrec_port *cfg = port->cfg;
470 struct bond_settings bond_settings;
471 struct lacp_settings lacp_settings;
472 struct ofproto_bundle_settings s;
480 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
481 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
482 s.slaves[s.n_slaves++] = iface->ofp_port;
488 if (list_is_short(&port->ifaces)) {
489 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
491 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan);
494 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
495 * they even work as-is. But they have not been tested. */
496 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
501 /* Get VLAN trunks. */
503 if (s.vlan < 0 && cfg->n_trunks) {
504 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
505 } else if (s.vlan >= 0 && cfg->n_trunks) {
506 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
510 /* Get LACP settings. */
511 s.lacp = port_configure_lacp(port, &lacp_settings);
515 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
516 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
517 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
520 s.lacp_slaves = NULL;
523 /* Get bond settings. */
524 if (s.n_slaves > 1) {
525 s.bond = &bond_settings;
526 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
527 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
530 s.bond_stable_ids = NULL;
532 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
533 netdev_set_miimon_interval(iface->netdev, 0);
538 ofproto_bundle_register(port->bridge->ofproto, port, &s);
544 free(s.bond_stable_ids);
547 /* Pick local port hardware address and datapath ID for 'br'. */
549 bridge_configure_datapath_id(struct bridge *br)
551 uint8_t ea[ETH_ADDR_LEN];
553 struct iface *local_iface;
554 struct iface *hw_addr_iface;
557 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
558 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
560 int error = netdev_set_etheraddr(local_iface->netdev, ea);
562 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
563 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
564 "Ethernet address: %s",
565 br->name, strerror(error));
568 memcpy(br->ea, ea, ETH_ADDR_LEN);
570 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
571 ofproto_set_datapath_id(br->ofproto, dpid);
573 dpid_string = xasprintf("%016"PRIx64, dpid);
574 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
578 /* Set NetFlow configuration on 'br'. */
580 bridge_configure_netflow(struct bridge *br)
582 struct ovsrec_netflow *cfg = br->cfg->netflow;
583 struct netflow_options opts;
586 ofproto_set_netflow(br->ofproto, NULL);
590 memset(&opts, 0, sizeof opts);
592 /* Get default NetFlow configuration from datapath.
593 * Apply overrides from 'cfg'. */
594 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
595 if (cfg->engine_type) {
596 opts.engine_type = *cfg->engine_type;
598 if (cfg->engine_id) {
599 opts.engine_id = *cfg->engine_id;
602 /* Configure active timeout interval. */
603 opts.active_timeout = cfg->active_timeout;
604 if (!opts.active_timeout) {
605 opts.active_timeout = -1;
606 } else if (opts.active_timeout < 0) {
607 VLOG_WARN("bridge %s: active timeout interval set to negative "
608 "value, using default instead (%d seconds)", br->name,
609 NF_ACTIVE_TIMEOUT_DEFAULT);
610 opts.active_timeout = -1;
613 /* Add engine ID to interface number to disambiguate bridgs? */
614 opts.add_id_to_iface = cfg->add_id_to_interface;
615 if (opts.add_id_to_iface) {
616 if (opts.engine_id > 0x7f) {
617 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
618 "another vswitch, choose an engine id less than 128",
621 if (hmap_count(&br->ports) > 508) {
622 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
623 "another port when more than 508 ports are used",
629 sset_init(&opts.collectors);
630 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
633 if (ofproto_set_netflow(br->ofproto, &opts)) {
634 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
636 sset_destroy(&opts.collectors);
639 /* Set sFlow configuration on 'br'. */
641 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
643 const struct ovsrec_sflow *cfg = br->cfg->sflow;
644 struct ovsrec_controller **controllers;
645 struct ofproto_sflow_options oso;
646 size_t n_controllers;
650 ofproto_set_sflow(br->ofproto, NULL);
654 memset(&oso, 0, sizeof oso);
656 sset_init(&oso.targets);
657 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
659 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
661 oso.sampling_rate = *cfg->sampling;
664 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
666 oso.polling_interval = *cfg->polling;
669 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
671 oso.header_len = *cfg->header;
674 oso.sub_id = (*sflow_bridge_number)++;
675 oso.agent_device = cfg->agent;
677 oso.control_ip = NULL;
678 n_controllers = bridge_get_controllers(br, &controllers);
679 for (i = 0; i < n_controllers; i++) {
680 if (controllers[i]->local_ip) {
681 oso.control_ip = controllers[i]->local_ip;
685 ofproto_set_sflow(br->ofproto, &oso);
687 sset_destroy(&oso.targets);
691 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
693 const struct port *port = port_lookup(br, name);
694 return port && port_is_bond_fake_iface(port);
698 port_is_bond_fake_iface(const struct port *port)
700 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
704 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
706 struct bridge *br, *next;
710 /* Collect new bridges' names and types. */
712 for (i = 0; i < cfg->n_bridges; i++) {
713 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
714 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
715 VLOG_WARN("bridge %s specified twice", br_cfg->name);
719 /* Get rid of deleted bridges or those whose types have changed.
720 * Update 'cfg' of bridges that still exist. */
721 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
722 br->cfg = shash_find_data(&new_br, br->name);
723 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
724 br->cfg->datapath_type))) {
729 /* Add new bridges. */
730 for (i = 0; i < cfg->n_bridges; i++) {
731 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
732 struct bridge *br = bridge_lookup(br_cfg->name);
734 bridge_create(br_cfg);
738 shash_destroy(&new_br);
741 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
744 * The kernel will reject any attempt to add a given port to a datapath if that
745 * port already belongs to a different datapath, so we must do all port
746 * deletions before any port additions. */
748 bridge_del_ofproto_ports(struct bridge *br)
750 struct ofproto_port_dump dump;
751 struct ofproto_port ofproto_port;
753 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
754 const char *name = ofproto_port.name;
759 /* Ignore the local port. We can't change it anyhow. */
760 if (!strcmp(name, br->name)) {
764 /* Get the type that 'ofproto_port' should have (ordinarily the
765 * type of its corresponding iface) or NULL if it should be
767 iface = iface_lookup(br, name);
768 type = (iface ? iface->type
769 : bridge_has_bond_fake_iface(br, name) ? "internal"
772 /* If it's the wrong type then delete the ofproto port. */
774 && !strcmp(ofproto_port.type, type)
775 && (!iface || !iface->netdev
776 || !strcmp(netdev_get_type(iface->netdev), type))) {
779 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
781 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
782 br->name, name, strerror(error));
785 netdev_close(iface->netdev);
786 iface->netdev = NULL;
792 iface_set_ofp_port(struct iface *iface, int ofp_port)
794 struct bridge *br = iface->port->bridge;
796 assert(iface->ofp_port < 0 && ofp_port >= 0);
797 iface->ofp_port = ofp_port;
798 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
799 iface_set_ofport(iface->cfg, ofp_port);
803 bridge_refresh_ofp_port(struct bridge *br)
805 struct ofproto_port_dump dump;
806 struct ofproto_port ofproto_port;
809 /* Clear all the "ofp_port"es. */
810 hmap_clear(&br->ifaces);
811 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
814 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
815 iface->ofp_port = -1;
819 /* Obtain the correct "ofp_port"s from ofproto. */
820 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
821 struct iface *iface = iface_lookup(br, ofproto_port.name);
823 if (iface->ofp_port >= 0) {
824 VLOG_WARN("bridge %s: interface %s reported twice",
825 br->name, ofproto_port.name);
826 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
827 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
828 br->name, ofproto_port.ofp_port);
830 iface_set_ofp_port(iface, ofproto_port.ofp_port);
836 /* Add an ofproto port for any "struct iface" that doesn't have one.
837 * Delete any "struct iface" for which this fails.
838 * Delete any "struct port" that thereby ends up with no ifaces. */
840 bridge_add_ofproto_ports(struct bridge *br)
842 struct port *port, *next_port;
844 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
845 struct iface *iface, *next_iface;
846 struct ofproto_port ofproto_port;
848 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
852 /* Open the netdev or reconfigure it. */
854 shash_from_ovs_idl_map(iface->cfg->key_options,
855 iface->cfg->value_options,
856 iface->cfg->n_options, &args);
857 if (!iface->netdev) {
858 struct netdev_options options;
859 options.name = iface->name;
860 options.type = iface->type;
861 options.args = &args;
862 error = netdev_open(&options, &iface->netdev);
864 error = netdev_set_config(iface->netdev, &args);
866 shash_destroy(&args);
868 VLOG_WARN("could not %s network device %s (%s)",
869 iface->netdev ? "reconfigure" : "open",
870 iface->name, strerror(error));
873 /* Add the port, if necessary. */
874 if (iface->netdev && iface->ofp_port < 0) {
878 error = ofproto_port_add(br->ofproto, iface->netdev,
881 iface_set_ofp_port(iface, ofp_port);
883 netdev_close(iface->netdev);
884 iface->netdev = NULL;
888 /* Populate stats columns in new Interface rows. */
889 if (iface->netdev && !iface->cfg->mtu) {
890 iface_refresh_stats(iface);
891 iface_refresh_status(iface);
894 /* Delete the iface if */
895 if (iface->netdev && iface->ofp_port >= 0) {
896 VLOG_DBG("bridge %s: interface %s is on port %d",
897 br->name, iface->name, iface->ofp_port);
900 VLOG_ERR("bridge %s: missing %s interface, dropping",
901 br->name, iface->name);
903 /* We already reported a related error, don't bother
906 iface_set_ofport(iface->cfg, -1);
907 iface_destroy(iface);
910 if (list_is_empty(&port->ifaces)) {
911 VLOG_WARN("%s port has no interfaces, dropping", port->name);
916 /* Add bond fake iface if necessary. */
917 if (port_is_bond_fake_iface(port)) {
918 if (ofproto_port_query_by_name(br->ofproto, port->name,
920 struct netdev_options options;
921 struct netdev *netdev;
924 options.name = port->name;
925 options.type = "internal";
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);
965 /* Set Flow eviction threshold */
967 bridge_configure_flow_eviction_threshold(struct bridge *br)
969 const char *threshold_str;
972 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
974 threshold = strtoul(threshold_str, NULL, 10);
976 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
978 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
982 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
983 struct iface **hw_addr_iface)
989 *hw_addr_iface = NULL;
991 /* Did the user request a particular MAC? */
992 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
993 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
994 if (eth_addr_is_multicast(ea)) {
995 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
996 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
997 } else if (eth_addr_is_zero(ea)) {
998 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1004 /* Otherwise choose the minimum non-local MAC address among all of the
1006 memset(ea, 0xff, ETH_ADDR_LEN);
1007 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1008 uint8_t iface_ea[ETH_ADDR_LEN];
1009 struct iface *candidate;
1010 struct iface *iface;
1012 /* Mirror output ports don't participate. */
1013 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1017 /* Choose the MAC address to represent the port. */
1019 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1020 /* Find the interface with this Ethernet address (if any) so that
1021 * we can provide the correct devname to the caller. */
1022 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1023 uint8_t candidate_ea[ETH_ADDR_LEN];
1024 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1025 && eth_addr_equals(iface_ea, candidate_ea)) {
1030 /* Choose the interface whose MAC address will represent the port.
1031 * The Linux kernel bonding code always chooses the MAC address of
1032 * the first slave added to a bond, and the Fedora networking
1033 * scripts always add slaves to a bond in alphabetical order, so
1034 * for compatibility we choose the interface with the name that is
1035 * first in alphabetical order. */
1036 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1037 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1042 /* The local port doesn't count (since we're trying to choose its
1043 * MAC address anyway). */
1044 if (iface->ofp_port == OFPP_LOCAL) {
1049 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1055 /* Compare against our current choice. */
1056 if (!eth_addr_is_multicast(iface_ea) &&
1057 !eth_addr_is_local(iface_ea) &&
1058 !eth_addr_is_reserved(iface_ea) &&
1059 !eth_addr_is_zero(iface_ea) &&
1060 eth_addr_compare_3way(iface_ea, ea) < 0)
1062 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1063 *hw_addr_iface = iface;
1066 if (eth_addr_is_multicast(ea)) {
1067 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1068 *hw_addr_iface = NULL;
1069 VLOG_WARN("bridge %s: using default bridge Ethernet "
1070 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1072 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1073 br->name, ETH_ADDR_ARGS(ea));
1077 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1078 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1079 * an interface on 'br', then that interface must be passed in as
1080 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1081 * 'hw_addr_iface' must be passed in as a null pointer. */
1083 bridge_pick_datapath_id(struct bridge *br,
1084 const uint8_t bridge_ea[ETH_ADDR_LEN],
1085 struct iface *hw_addr_iface)
1088 * The procedure for choosing a bridge MAC address will, in the most
1089 * ordinary case, also choose a unique MAC that we can use as a datapath
1090 * ID. In some special cases, though, multiple bridges will end up with
1091 * the same MAC address. This is OK for the bridges, but it will confuse
1092 * the OpenFlow controller, because each datapath needs a unique datapath
1095 * Datapath IDs must be unique. It is also very desirable that they be
1096 * stable from one run to the next, so that policy set on a datapath
1099 const char *datapath_id;
1102 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1103 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1107 if (hw_addr_iface) {
1109 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1111 * A bridge whose MAC address is taken from a VLAN network device
1112 * (that is, a network device created with vconfig(8) or similar
1113 * tool) will have the same MAC address as a bridge on the VLAN
1114 * device's physical network device.
1116 * Handle this case by hashing the physical network device MAC
1117 * along with the VLAN identifier.
1119 uint8_t buf[ETH_ADDR_LEN + 2];
1120 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1121 buf[ETH_ADDR_LEN] = vlan >> 8;
1122 buf[ETH_ADDR_LEN + 1] = vlan;
1123 return dpid_from_hash(buf, sizeof buf);
1126 * Assume that this bridge's MAC address is unique, since it
1127 * doesn't fit any of the cases we handle specially.
1132 * A purely internal bridge, that is, one that has no non-virtual
1133 * network devices on it at all, is more difficult because it has no
1134 * natural unique identifier at all.
1136 * When the host is a XenServer, we handle this case by hashing the
1137 * host's UUID with the name of the bridge. Names of bridges are
1138 * persistent across XenServer reboots, although they can be reused if
1139 * an internal network is destroyed and then a new one is later
1140 * created, so this is fairly effective.
1142 * When the host is not a XenServer, we punt by using a random MAC
1143 * address on each run.
1145 const char *host_uuid = xenserver_get_host_uuid();
1147 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1148 dpid = dpid_from_hash(combined, strlen(combined));
1154 return eth_addr_to_uint64(bridge_ea);
1158 dpid_from_hash(const void *data, size_t n)
1160 uint8_t hash[SHA1_DIGEST_SIZE];
1162 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1163 sha1_bytes(data, n, hash);
1164 eth_addr_mark_random(hash);
1165 return eth_addr_to_uint64(hash);
1169 iface_refresh_status(struct iface *iface)
1173 enum netdev_flags flags;
1180 if (iface_is_synthetic(iface)) {
1186 if (!netdev_get_status(iface->netdev, &sh)) {
1188 char **keys, **values;
1190 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1191 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1196 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1199 shash_destroy_free_data(&sh);
1201 error = netdev_get_flags(iface->netdev, &flags);
1203 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1206 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1209 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1211 ovsrec_interface_set_duplex(iface->cfg,
1212 netdev_features_is_full_duplex(current)
1214 /* warning: uint64_t -> int64_t conversion */
1215 bps = netdev_features_to_bps(current);
1216 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1219 ovsrec_interface_set_duplex(iface->cfg, NULL);
1220 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1223 ovsrec_interface_set_link_state(iface->cfg,
1224 iface_get_carrier(iface) ? "up" : "down");
1226 error = netdev_get_mtu(iface->netdev, &mtu);
1227 if (!error && mtu != INT_MAX) {
1229 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1232 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1236 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1237 * changed, false otherwise. */
1239 iface_refresh_cfm_stats(struct iface *iface)
1241 const struct ovsrec_interface *cfg = iface->cfg;
1242 bool changed = false;
1245 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1252 if (cfg->n_cfm_fault != 1 || cfg->cfm_fault[0] != fault) {
1253 bool fault_bool = fault;
1254 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1262 iface_refresh_lacp_stats(struct iface *iface)
1264 struct ofproto *ofproto = iface->port->bridge->ofproto;
1265 int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
1266 int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
1270 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
1276 iface_refresh_stats(struct iface *iface)
1278 #define IFACE_STATS \
1279 IFACE_STAT(rx_packets, "rx_packets") \
1280 IFACE_STAT(tx_packets, "tx_packets") \
1281 IFACE_STAT(rx_bytes, "rx_bytes") \
1282 IFACE_STAT(tx_bytes, "tx_bytes") \
1283 IFACE_STAT(rx_dropped, "rx_dropped") \
1284 IFACE_STAT(tx_dropped, "tx_dropped") \
1285 IFACE_STAT(rx_errors, "rx_errors") \
1286 IFACE_STAT(tx_errors, "tx_errors") \
1287 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1288 IFACE_STAT(rx_over_errors, "rx_over_err") \
1289 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1290 IFACE_STAT(collisions, "collisions")
1292 #define IFACE_STAT(MEMBER, NAME) NAME,
1293 static char *keys[] = { IFACE_STATS };
1295 int64_t values[ARRAY_SIZE(keys)];
1298 struct netdev_stats stats;
1300 if (iface_is_synthetic(iface)) {
1304 /* Intentionally ignore return value, since errors will set 'stats' to
1305 * all-1s, and we will deal with that correctly below. */
1306 netdev_get_stats(iface->netdev, &stats);
1308 /* Copy statistics into values[] array. */
1310 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1313 assert(i == ARRAY_SIZE(keys));
1315 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1320 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1324 /* Use other-config:enable-system-stats by preference. */
1325 enable = get_ovsrec_key_value(&cfg->header_,
1326 &ovsrec_open_vswitch_col_other_config,
1327 "enable-statistics");
1329 return !strcmp(enable, "true");
1332 /* Disable by default. */
1337 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1339 struct ovsdb_datum datum;
1343 if (enable_system_stats(cfg)) {
1344 get_system_stats(&stats);
1347 ovsdb_datum_from_shash(&datum, &stats);
1348 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1352 static inline const char *
1353 nx_role_to_str(enum nx_role role)
1358 case NX_ROLE_MASTER:
1363 return "*** INVALID ROLE ***";
1368 refresh_controller_status(void)
1372 const struct ovsrec_controller *cfg;
1376 /* Accumulate status for controllers on all bridges. */
1377 HMAP_FOR_EACH (br, node, &all_bridges) {
1378 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1381 /* Update each controller in the database with current status. */
1382 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1383 struct ofproto_controller_info *cinfo =
1384 shash_find_data(&info, cfg->target);
1387 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1388 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1389 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1390 (char **) cinfo->pairs.values,
1393 ovsrec_controller_set_is_connected(cfg, false);
1394 ovsrec_controller_set_role(cfg, NULL);
1395 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1399 ofproto_free_ofproto_controller_info(&info);
1405 const struct ovsrec_open_vswitch *cfg;
1407 bool datapath_destroyed;
1408 bool database_changed;
1411 /* (Re)configure if necessary. */
1412 database_changed = ovsdb_idl_run(idl);
1413 if (ovsdb_idl_is_lock_contended(idl)) {
1414 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1415 struct bridge *br, *next_br;
1417 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1418 "disabling this process until it goes away");
1420 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1424 } else if (!ovsdb_idl_has_lock(idl)) {
1427 cfg = ovsrec_open_vswitch_first(idl);
1429 /* Let each bridge do the work that it needs to do. */
1430 datapath_destroyed = false;
1431 HMAP_FOR_EACH (br, node, &all_bridges) {
1432 int error = ofproto_run(br->ofproto);
1434 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1435 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1436 "forcing reconfiguration", br->name);
1437 datapath_destroyed = true;
1441 /* Re-configure SSL. We do this on every trip through the main loop,
1442 * instead of just when the database changes, because the contents of the
1443 * key and certificate files can change without the database changing.
1445 * We do this before bridge_reconfigure() because that function might
1446 * initiate SSL connections and thus requires SSL to be configured. */
1447 if (cfg && cfg->ssl) {
1448 const struct ovsrec_ssl *ssl = cfg->ssl;
1450 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1451 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1454 if (database_changed || datapath_destroyed) {
1456 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1458 bridge_reconfigure(cfg);
1460 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1461 ovsdb_idl_txn_commit(txn);
1462 ovsdb_idl_txn_destroy(txn); /* XXX */
1464 /* We still need to reconfigure to avoid dangling pointers to
1465 * now-destroyed ovsrec structures inside bridge data. */
1466 static const struct ovsrec_open_vswitch null_cfg;
1468 bridge_reconfigure(&null_cfg);
1472 /* Refresh system and interface stats if necessary. */
1473 if (time_msec() >= stats_timer) {
1475 struct ovsdb_idl_txn *txn;
1477 txn = ovsdb_idl_txn_create(idl);
1478 HMAP_FOR_EACH (br, node, &all_bridges) {
1481 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1482 struct iface *iface;
1484 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1485 iface_refresh_stats(iface);
1486 iface_refresh_status(iface);
1490 refresh_system_stats(cfg);
1491 refresh_controller_status();
1492 ovsdb_idl_txn_commit(txn);
1493 ovsdb_idl_txn_destroy(txn); /* XXX */
1496 stats_timer = time_msec() + STATS_INTERVAL;
1499 if (time_msec() >= db_limiter) {
1500 struct ovsdb_idl_txn *txn;
1501 bool changed = false;
1503 txn = ovsdb_idl_txn_create(idl);
1504 HMAP_FOR_EACH (br, node, &all_bridges) {
1507 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1508 struct iface *iface;
1510 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1511 changed = iface_refresh_cfm_stats(iface) || changed;
1512 changed = iface_refresh_lacp_stats(iface) || changed;
1518 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1521 ovsdb_idl_txn_commit(txn);
1522 ovsdb_idl_txn_destroy(txn);
1529 ovsdb_idl_wait(idl);
1530 if (!hmap_is_empty(&all_bridges)) {
1533 HMAP_FOR_EACH (br, node, &all_bridges) {
1534 ofproto_wait(br->ofproto);
1536 poll_timer_wait_until(stats_timer);
1538 if (db_limiter > time_msec()) {
1539 poll_timer_wait_until(db_limiter);
1544 /* QoS unixctl user interface functions. */
1546 struct qos_unixctl_show_cbdata {
1548 struct iface *iface;
1552 qos_unixctl_show_cb(unsigned int queue_id,
1553 const struct shash *details,
1556 struct qos_unixctl_show_cbdata *data = aux;
1557 struct ds *ds = data->ds;
1558 struct iface *iface = data->iface;
1559 struct netdev_queue_stats stats;
1560 struct shash_node *node;
1563 ds_put_cstr(ds, "\n");
1565 ds_put_format(ds, "Queue %u:\n", queue_id);
1567 ds_put_cstr(ds, "Default:\n");
1570 SHASH_FOR_EACH (node, details) {
1571 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1574 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1576 if (stats.tx_packets != UINT64_MAX) {
1577 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1580 if (stats.tx_bytes != UINT64_MAX) {
1581 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1584 if (stats.tx_errors != UINT64_MAX) {
1585 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1588 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1589 queue_id, strerror(error));
1594 qos_unixctl_show(struct unixctl_conn *conn,
1595 const char *args, void *aux OVS_UNUSED)
1597 struct ds ds = DS_EMPTY_INITIALIZER;
1598 struct shash sh = SHASH_INITIALIZER(&sh);
1599 struct iface *iface;
1601 struct shash_node *node;
1602 struct qos_unixctl_show_cbdata data;
1605 iface = iface_find(args);
1607 unixctl_command_reply(conn, 501, "no such interface");
1611 netdev_get_qos(iface->netdev, &type, &sh);
1613 if (*type != '\0') {
1614 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1616 SHASH_FOR_EACH (node, &sh) {
1617 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1622 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1625 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1627 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1629 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1630 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1633 shash_destroy_free_data(&sh);
1637 /* Bridge reconfiguration functions. */
1639 bridge_create(const struct ovsrec_bridge *br_cfg)
1643 assert(!bridge_lookup(br_cfg->name));
1644 br = xzalloc(sizeof *br);
1646 br->name = xstrdup(br_cfg->name);
1647 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1650 /* Derive the default Ethernet address from the bridge's UUID. This should
1651 * be unique and it will be stable between ovs-vswitchd runs. */
1652 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1653 eth_addr_mark_random(br->default_ea);
1655 hmap_init(&br->ports);
1656 hmap_init(&br->ifaces);
1657 hmap_init(&br->iface_by_name);
1658 hmap_init(&br->mirrors);
1660 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1664 bridge_destroy(struct bridge *br)
1667 struct mirror *mirror, *next_mirror;
1668 struct port *port, *next_port;
1670 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1673 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1674 mirror_destroy(mirror);
1676 hmap_remove(&all_bridges, &br->node);
1677 ofproto_destroy(br->ofproto);
1678 hmap_destroy(&br->ifaces);
1679 hmap_destroy(&br->ports);
1680 hmap_destroy(&br->iface_by_name);
1681 hmap_destroy(&br->mirrors);
1688 static struct bridge *
1689 bridge_lookup(const char *name)
1693 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1694 if (!strcmp(br->name, name)) {
1701 /* Handle requests for a listing of all flows known by the OpenFlow
1702 * stack, including those normally hidden. */
1704 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1705 const char *args, void *aux OVS_UNUSED)
1710 br = bridge_lookup(args);
1712 unixctl_command_reply(conn, 501, "Unknown bridge");
1717 ofproto_get_all_flows(br->ofproto, &results);
1719 unixctl_command_reply(conn, 200, ds_cstr(&results));
1720 ds_destroy(&results);
1723 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1724 * connections and reconnect. If BRIDGE is not specified, then all bridges
1725 * drop their controller connections and reconnect. */
1727 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1728 const char *args, void *aux OVS_UNUSED)
1731 if (args[0] != '\0') {
1732 br = bridge_lookup(args);
1734 unixctl_command_reply(conn, 501, "Unknown bridge");
1737 ofproto_reconnect_controllers(br->ofproto);
1739 HMAP_FOR_EACH (br, node, &all_bridges) {
1740 ofproto_reconnect_controllers(br->ofproto);
1743 unixctl_command_reply(conn, 200, NULL);
1747 bridge_get_controllers(const struct bridge *br,
1748 struct ovsrec_controller ***controllersp)
1750 struct ovsrec_controller **controllers;
1751 size_t n_controllers;
1753 controllers = br->cfg->controller;
1754 n_controllers = br->cfg->n_controller;
1756 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1762 *controllersp = controllers;
1764 return n_controllers;
1767 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1768 * those configured in 'br->cfg'. */
1770 bridge_add_del_ports(struct bridge *br)
1772 struct port *port, *next;
1773 struct shash_node *node;
1774 struct shash new_ports;
1777 /* Collect new ports. */
1778 shash_init(&new_ports);
1779 for (i = 0; i < br->cfg->n_ports; i++) {
1780 const char *name = br->cfg->ports[i]->name;
1781 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1782 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1786 if (bridge_get_controllers(br, NULL)
1787 && !shash_find(&new_ports, br->name)) {
1788 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1789 br->name, br->name);
1791 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1792 br->synth_local_port.n_interfaces = 1;
1793 br->synth_local_port.name = br->name;
1795 br->synth_local_iface.name = br->name;
1796 br->synth_local_iface.type = "internal";
1798 br->synth_local_ifacep = &br->synth_local_iface;
1800 shash_add(&new_ports, br->name, &br->synth_local_port);
1803 /* Get rid of deleted ports.
1804 * Get rid of deleted interfaces on ports that still exist.
1805 * Update 'cfg' of ports that still exist. */
1806 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1807 port->cfg = shash_find_data(&new_ports, port->name);
1811 port_del_ifaces(port);
1815 /* Create new ports.
1816 * Add new interfaces to existing ports. */
1817 SHASH_FOR_EACH (node, &new_ports) {
1818 struct port *port = port_lookup(br, node->name);
1820 struct ovsrec_port *cfg = node->data;
1821 port = port_create(br, cfg);
1823 port_add_ifaces(port);
1824 if (list_is_empty(&port->ifaces)) {
1825 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1826 br->name, port->name);
1830 shash_destroy(&new_ports);
1833 /* Initializes 'oc' appropriately as a management service controller for
1836 * The caller must free oc->target when it is no longer needed. */
1838 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1839 struct ofproto_controller *oc)
1841 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1842 oc->max_backoff = 0;
1843 oc->probe_interval = 60;
1844 oc->band = OFPROTO_OUT_OF_BAND;
1846 oc->burst_limit = 0;
1849 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1851 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1852 struct ofproto_controller *oc)
1854 oc->target = c->target;
1855 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1856 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1857 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1858 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1859 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1860 oc->burst_limit = (c->controller_burst_limit
1861 ? *c->controller_burst_limit : 0);
1864 /* Configures the IP stack for 'br''s local interface properly according to the
1865 * configuration in 'c'. */
1867 bridge_configure_local_iface_netdev(struct bridge *br,
1868 struct ovsrec_controller *c)
1870 struct netdev *netdev;
1871 struct in_addr mask, gateway;
1873 struct iface *local_iface;
1876 /* If there's no local interface or no IP address, give up. */
1877 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1878 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1882 /* Bring up the local interface. */
1883 netdev = local_iface->netdev;
1884 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1886 /* Configure the IP address and netmask. */
1887 if (!c->local_netmask
1888 || !inet_aton(c->local_netmask, &mask)
1890 mask.s_addr = guess_netmask(ip.s_addr);
1892 if (!netdev_set_in4(netdev, ip, mask)) {
1893 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1894 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1897 /* Configure the default gateway. */
1898 if (c->local_gateway
1899 && inet_aton(c->local_gateway, &gateway)
1900 && gateway.s_addr) {
1901 if (!netdev_add_router(netdev, gateway)) {
1902 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1903 br->name, IP_ARGS(&gateway.s_addr));
1909 bridge_configure_remotes(struct bridge *br,
1910 const struct sockaddr_in *managers, size_t n_managers)
1912 const char *disable_ib_str, *queue_id_str;
1913 bool disable_in_band = false;
1916 struct ovsrec_controller **controllers;
1917 size_t n_controllers;
1919 enum ofproto_fail_mode fail_mode;
1921 struct ofproto_controller *ocs;
1925 /* Check if we should disable in-band control on this bridge. */
1926 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1927 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1928 disable_in_band = true;
1931 /* Set OpenFlow queue ID for in-band control. */
1932 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1933 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1934 ofproto_set_in_band_queue(br->ofproto, queue_id);
1936 if (disable_in_band) {
1937 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1939 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1942 n_controllers = bridge_get_controllers(br, &controllers);
1944 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1947 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1948 for (i = 0; i < n_controllers; i++) {
1949 struct ovsrec_controller *c = controllers[i];
1951 if (!strncmp(c->target, "punix:", 6)
1952 || !strncmp(c->target, "unix:", 5)) {
1953 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1955 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
1956 * domain sockets and overwriting arbitrary local files. */
1957 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
1958 "controller \"%s\" due to possibility for remote "
1959 "exploit", br->name, c->target);
1963 bridge_configure_local_iface_netdev(br, c);
1964 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
1965 if (disable_in_band) {
1966 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
1971 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
1972 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
1975 /* Set the fail-mode. */
1976 fail_mode = !br->cfg->fail_mode
1977 || !strcmp(br->cfg->fail_mode, "standalone")
1978 ? OFPROTO_FAIL_STANDALONE
1979 : OFPROTO_FAIL_SECURE;
1980 ofproto_set_fail_mode(br->ofproto, fail_mode);
1982 /* Configure OpenFlow controller connection snooping. */
1983 if (!ofproto_has_snoops(br->ofproto)) {
1987 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1988 ovs_rundir(), br->name));
1989 ofproto_set_snoops(br->ofproto, &snoops);
1990 sset_destroy(&snoops);
1994 /* Port functions. */
1996 static struct port *
1997 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2001 port = xzalloc(sizeof *port);
2003 port->name = xstrdup(cfg->name);
2005 list_init(&port->ifaces);
2007 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2009 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2015 get_port_other_config(const struct ovsrec_port *port, const char *key,
2016 const char *default_value)
2020 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2022 return value ? value : default_value;
2026 get_interface_other_config(const struct ovsrec_interface *iface,
2027 const char *key, const char *default_value)
2031 value = get_ovsrec_key_value(&iface->header_,
2032 &ovsrec_interface_col_other_config, key);
2033 return value ? value : default_value;
2036 /* Deletes interfaces from 'port' that are no longer configured for it. */
2038 port_del_ifaces(struct port *port)
2040 struct iface *iface, *next;
2041 struct sset new_ifaces;
2044 /* Collect list of new interfaces. */
2045 sset_init(&new_ifaces);
2046 for (i = 0; i < port->cfg->n_interfaces; i++) {
2047 const char *name = port->cfg->interfaces[i]->name;
2048 const char *type = port->cfg->interfaces[i]->name;
2049 if (strcmp(type, "null")) {
2050 sset_add(&new_ifaces, name);
2054 /* Get rid of deleted interfaces. */
2055 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2056 if (!sset_contains(&new_ifaces, iface->name)) {
2057 iface_destroy(iface);
2061 sset_destroy(&new_ifaces);
2064 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2067 port_add_ifaces(struct port *port)
2069 struct shash new_ifaces;
2070 struct shash_node *node;
2073 /* Collect new ifaces. */
2074 shash_init(&new_ifaces);
2075 for (i = 0; i < port->cfg->n_interfaces; i++) {
2076 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2077 if (strcmp(cfg->type, "null")
2078 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2079 VLOG_WARN("port %s: %s specified twice as port interface",
2080 port->name, cfg->name);
2081 iface_set_ofport(cfg, -1);
2085 /* Create new interfaces.
2086 * Update interface types and 'cfg' members. */
2087 SHASH_FOR_EACH (node, &new_ifaces) {
2088 const struct ovsrec_interface *cfg = node->data;
2089 const char *iface_name = node->name;
2090 struct iface *iface;
2092 iface = iface_lookup(port->bridge, iface_name);
2094 iface = iface_create(port, cfg);
2099 /* Determine interface type. The local port always has type
2100 * "internal". Other ports take their type from the database and
2101 * default to "system" if none is specified. */
2102 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2103 : cfg->type[0] ? cfg->type
2106 shash_destroy(&new_ifaces);
2110 port_destroy(struct port *port)
2113 struct bridge *br = port->bridge;
2114 struct iface *iface, *next;
2117 ofproto_bundle_unregister(br->ofproto, port);
2120 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2121 iface_destroy(iface);
2124 hmap_remove(&br->ports, &port->hmap_node);
2126 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2133 static struct port *
2134 port_lookup(const struct bridge *br, const char *name)
2138 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2140 if (!strcmp(port->name, name)) {
2148 enable_lacp(struct port *port, bool *activep)
2150 if (!port->cfg->lacp) {
2151 /* XXX when LACP implementation has been sufficiently tested, enable by
2152 * default and make active on bonded ports. */
2154 } else if (!strcmp(port->cfg->lacp, "off")) {
2156 } else if (!strcmp(port->cfg->lacp, "active")) {
2159 } else if (!strcmp(port->cfg->lacp, "passive")) {
2163 VLOG_WARN("port %s: unknown LACP mode %s",
2164 port->name, port->cfg->lacp);
2169 static struct lacp_settings *
2170 port_configure_lacp(struct port *port, struct lacp_settings *s)
2172 const char *lacp_time;
2173 long long int custom_time;
2176 if (!enable_lacp(port, &s->active)) {
2180 s->name = port->name;
2181 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2183 /* Prefer bondable links if unspecified. */
2184 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2186 s->priority = (priority > 0 && priority <= UINT16_MAX
2188 : UINT16_MAX - !list_is_short(&port->ifaces));
2190 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2195 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2196 custom_time = atoi(lacp_time);
2197 if (!strcmp(lacp_time, "fast")) {
2198 s->lacp_time = LACP_TIME_FAST;
2199 } else if (!strcmp(lacp_time, "slow")) {
2200 s->lacp_time = LACP_TIME_SLOW;
2201 } else if (custom_time > 0) {
2202 s->lacp_time = LACP_TIME_CUSTOM;
2203 s->custom_time = custom_time;
2205 s->lacp_time = LACP_TIME_SLOW;
2212 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2214 int priority, portid, key;
2216 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2217 priority = atoi(get_interface_other_config(iface->cfg,
2218 "lacp-port-priority", "0"));
2219 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2222 if (portid <= 0 || portid > UINT16_MAX) {
2223 portid = iface->ofp_port;
2226 if (priority <= 0 || priority > UINT16_MAX) {
2227 priority = UINT16_MAX;
2230 if (key < 0 || key > UINT16_MAX) {
2234 s->name = iface->name;
2236 s->priority = priority;
2241 port_configure_bond(struct port *port, struct bond_settings *s,
2242 uint32_t *bond_stable_ids)
2244 const char *detect_s;
2245 struct iface *iface;
2246 int miimon_interval;
2249 s->name = port->name;
2250 s->balance = BM_SLB;
2251 if (port->cfg->bond_mode
2252 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2253 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2254 port->name, port->cfg->bond_mode,
2255 bond_mode_to_string(s->balance));
2257 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2258 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2259 "please use another bond type or disable flood_vlans",
2263 miimon_interval = atoi(get_port_other_config(port->cfg,
2264 "bond-miimon-interval", "0"));
2265 if (miimon_interval <= 0) {
2266 miimon_interval = 200;
2269 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2270 if (!strcmp(detect_s, "carrier")) {
2271 miimon_interval = 0;
2272 } else if (strcmp(detect_s, "miimon")) {
2273 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2274 "defaulting to carrier", port->name, detect_s);
2275 miimon_interval = 0;
2278 s->up_delay = MAX(0, port->cfg->bond_updelay);
2279 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2280 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2281 s->rebalance_interval = atoi(
2282 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2283 if (s->rebalance_interval < 1000) {
2284 s->rebalance_interval = 1000;
2287 s->fake_iface = port->cfg->bond_fake_iface;
2290 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2291 long long stable_id;
2293 stable_id = atoll(get_interface_other_config(iface->cfg,
2294 "bond-stable-id", "0"));
2295 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2296 stable_id = iface->ofp_port;
2298 bond_stable_ids[i++] = stable_id;
2300 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2304 /* Interface functions. */
2306 static struct iface *
2307 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2309 struct bridge *br = port->bridge;
2310 struct iface *iface;
2311 char *name = if_cfg->name;
2313 iface = xzalloc(sizeof *iface);
2315 iface->name = xstrdup(name);
2316 iface->ofp_port = -1;
2317 iface->tag = tag_create_random();
2318 iface->netdev = NULL;
2319 iface->cfg = if_cfg;
2321 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2323 list_push_back(&port->ifaces, &iface->port_elem);
2325 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2331 iface_destroy(struct iface *iface)
2334 struct port *port = iface->port;
2335 struct bridge *br = port->bridge;
2337 if (br->ofproto && iface->ofp_port >= 0) {
2338 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2341 if (iface->ofp_port >= 0) {
2342 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2345 list_remove(&iface->port_elem);
2346 hmap_remove(&br->iface_by_name, &iface->name_node);
2348 netdev_close(iface->netdev);
2355 static struct iface *
2356 iface_lookup(const struct bridge *br, const char *name)
2358 struct iface *iface;
2360 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2361 &br->iface_by_name) {
2362 if (!strcmp(iface->name, name)) {
2370 static struct iface *
2371 iface_find(const char *name)
2373 const struct bridge *br;
2375 HMAP_FOR_EACH (br, node, &all_bridges) {
2376 struct iface *iface = iface_lookup(br, name);
2385 static struct iface *
2386 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2388 struct iface *iface;
2390 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2391 hash_int(ofp_port, 0), &br->ifaces) {
2392 if (iface->ofp_port == ofp_port) {
2399 /* Set Ethernet address of 'iface', if one is specified in the configuration
2402 iface_set_mac(struct iface *iface)
2404 uint8_t ea[ETH_ADDR_LEN];
2406 if (!strcmp(iface->type, "internal")
2407 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2408 if (iface->ofp_port == OFPP_LOCAL) {
2409 VLOG_ERR("interface %s: ignoring mac in Interface record "
2410 "(use Bridge record to set local port's mac)",
2412 } else if (eth_addr_is_multicast(ea)) {
2413 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2416 int error = netdev_set_etheraddr(iface->netdev, ea);
2418 VLOG_ERR("interface %s: setting MAC failed (%s)",
2419 iface->name, strerror(error));
2425 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2427 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2429 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2430 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2434 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2436 * The value strings in '*shash' are taken directly from values[], not copied,
2437 * so the caller should not modify or free them. */
2439 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2440 struct shash *shash)
2445 for (i = 0; i < n; i++) {
2446 shash_add(shash, keys[i], values[i]);
2450 /* Creates 'keys' and 'values' arrays from 'shash'.
2452 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2453 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2454 * are populated with with strings taken directly from 'shash' and thus have
2455 * the same ownership of the key-value pairs in shash.
2458 shash_to_ovs_idl_map(struct shash *shash,
2459 char ***keys, char ***values, size_t *n)
2463 struct shash_node *sn;
2465 count = shash_count(shash);
2467 k = xmalloc(count * sizeof *k);
2468 v = xmalloc(count * sizeof *v);
2471 SHASH_FOR_EACH(sn, shash) {
2482 struct iface_delete_queues_cbdata {
2483 struct netdev *netdev;
2484 const struct ovsdb_datum *queues;
2488 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2490 union ovsdb_atom atom;
2492 atom.integer = target;
2493 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2497 iface_delete_queues(unsigned int queue_id,
2498 const struct shash *details OVS_UNUSED, void *cbdata_)
2500 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2502 if (!queue_ids_include(cbdata->queues, queue_id)) {
2503 netdev_delete_queue(cbdata->netdev, queue_id);
2508 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2510 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2511 netdev_set_qos(iface->netdev, NULL, NULL);
2513 struct iface_delete_queues_cbdata cbdata;
2514 struct shash details;
2518 /* Configure top-level Qos for 'iface'. */
2519 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2520 qos->n_other_config, &details);
2521 netdev_set_qos(iface->netdev, qos->type, &details);
2522 shash_destroy(&details);
2524 /* Deconfigure queues that were deleted. */
2525 cbdata.netdev = iface->netdev;
2526 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2528 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2530 /* Configure queues for 'iface'. */
2532 for (i = 0; i < qos->n_queues; i++) {
2533 const struct ovsrec_queue *queue = qos->value_queues[i];
2534 unsigned int queue_id = qos->key_queues[i];
2536 if (queue_id == 0) {
2540 shash_from_ovs_idl_map(queue->key_other_config,
2541 queue->value_other_config,
2542 queue->n_other_config, &details);
2543 netdev_set_queue(iface->netdev, queue_id, &details);
2544 shash_destroy(&details);
2547 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2548 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2549 "queue (queue 0). Packets not directed to a "
2550 "correctly configured queue may be dropped.",
2555 netdev_set_policing(iface->netdev,
2556 iface->cfg->ingress_policing_rate,
2557 iface->cfg->ingress_policing_burst);
2561 iface_configure_cfm(struct iface *iface)
2563 const struct ovsrec_interface *cfg = iface->cfg;
2564 struct cfm_settings s;
2565 uint16_t remote_mpid;
2567 if (!cfg->n_cfm_mpid || !cfg->n_cfm_remote_mpid) {
2568 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2572 s.mpid = *cfg->cfm_mpid;
2573 remote_mpid = *cfg->cfm_remote_mpid;
2574 s.remote_mpids = &remote_mpid;
2575 s.n_remote_mpids = 1;
2577 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2579 if (s.interval <= 0) {
2583 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2586 /* Read carrier or miimon status directly from 'iface''s netdev, according to
2587 * how 'iface''s port is configured.
2589 * Returns true if 'iface' is up, false otherwise. */
2591 iface_get_carrier(const struct iface *iface)
2594 return netdev_get_carrier(iface->netdev);
2597 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2598 * instead of obtaining it from the database. */
2600 iface_is_synthetic(const struct iface *iface)
2602 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2605 /* Port mirroring. */
2607 static struct mirror *
2608 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2612 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2613 if (uuid_equals(uuid, &m->uuid)) {
2621 bridge_configure_mirrors(struct bridge *br)
2623 const struct ovsdb_datum *mc;
2624 unsigned long *flood_vlans;
2625 struct mirror *m, *next;
2628 /* Get rid of deleted mirrors. */
2629 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2630 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2631 union ovsdb_atom atom;
2633 atom.uuid = m->uuid;
2634 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2639 /* Add new mirrors and reconfigure existing ones. */
2640 for (i = 0; i < br->cfg->n_mirrors; i++) {
2641 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2642 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2644 m = mirror_create(br, cfg);
2646 if (!mirror_configure(m, cfg)) {
2651 /* Update flooded vlans (for RSPAN). */
2652 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2653 br->cfg->n_flood_vlans);
2654 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2655 bitmap_free(flood_vlans);
2658 static struct mirror *
2659 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2663 m = xzalloc(sizeof *m);
2664 m->uuid = cfg->header_.uuid;
2665 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2667 m->name = xstrdup(cfg->name);
2673 mirror_destroy(struct mirror *m)
2676 struct bridge *br = m->bridge;
2679 ofproto_mirror_unregister(br->ofproto, m);
2682 hmap_remove(&br->mirrors, &m->hmap_node);
2689 mirror_collect_ports(struct mirror *m,
2690 struct ovsrec_port **in_ports, int n_in_ports,
2691 void ***out_portsp, size_t *n_out_portsp)
2693 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2694 size_t n_out_ports = 0;
2697 for (i = 0; i < n_in_ports; i++) {
2698 const char *name = in_ports[i]->name;
2699 struct port *port = port_lookup(m->bridge, name);
2701 out_ports[n_out_ports++] = port;
2703 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2704 "port %s", m->bridge->name, m->name, name);
2707 *out_portsp = out_ports;
2708 *n_out_portsp = n_out_ports;
2712 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2714 struct ofproto_mirror_settings s;
2717 if (strcmp(cfg->name, m->name)) {
2719 m->name = xstrdup(cfg->name);
2723 /* Get output port or VLAN. */
2724 if (cfg->output_port) {
2725 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2726 if (!s.out_bundle) {
2727 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2728 m->bridge->name, m->name);
2731 s.out_vlan = UINT16_MAX;
2733 if (cfg->output_vlan) {
2734 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2735 "output vlan; ignoring output vlan",
2736 m->bridge->name, m->name);
2738 } else if (cfg->output_vlan) {
2739 /* The database should prevent invalid VLAN values. */
2740 s.out_bundle = NULL;
2741 s.out_vlan = *cfg->output_vlan;
2743 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2744 m->bridge->name, m->name);
2748 /* Get port selection. */
2749 if (cfg->select_all) {
2750 size_t n_ports = hmap_count(&m->bridge->ports);
2751 void **ports = xmalloc(n_ports * sizeof *ports);
2756 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2766 /* Get ports, dropping ports that don't exist.
2767 * The IDL ensures that there are no duplicates. */
2768 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2769 &s.srcs, &s.n_srcs);
2770 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2771 &s.dsts, &s.n_dsts);
2774 /* Get VLAN selection. */
2775 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2778 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2781 if (s.srcs != s.dsts) {