1 /* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "ofp-print.h"
37 #include "ofproto/ofproto.h"
38 #include "poll-loop.h"
41 #include "socket-util.h"
42 #include "stream-ssl.h"
44 #include "system-stats.h"
48 #include "vswitchd/vswitch-idl.h"
49 #include "xenserver.h"
51 #include "sflow_api.h"
52 #include "vlan-bitmap.h"
54 VLOG_DEFINE_THIS_MODULE(bridge);
56 COVERAGE_DEFINE(bridge_reconfigure);
59 /* These members are always valid. */
60 struct list port_elem; /* Element in struct port's "ifaces" list. */
61 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
62 struct port *port; /* Containing port. */
63 char *name; /* Host network device name. */
64 tag_type tag; /* Tag associated with this interface. */
66 /* These members are valid only after bridge_reconfigure() causes them to
68 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
69 int ofp_port; /* OpenFlow port number, -1 if unknown. */
70 struct netdev *netdev; /* Network device. */
71 const char *type; /* Usually same as cfg->type. */
72 const struct ovsrec_interface *cfg;
76 struct uuid uuid; /* UUID of this "mirror" record in database. */
77 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
78 struct bridge *bridge;
83 struct bridge *bridge;
84 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
87 const struct ovsrec_port *cfg;
89 /* An ordinary bridge port has 1 interface.
90 * A bridge port for bonding has at least 2 interfaces. */
91 struct list ifaces; /* List of "struct iface"s. */
95 struct hmap_node node; /* In 'all_bridges'. */
96 char *name; /* User-specified arbitrary name. */
97 char *type; /* Datapath type. */
98 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
99 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
100 const struct ovsrec_bridge *cfg;
102 /* OpenFlow switch processing. */
103 struct ofproto *ofproto; /* OpenFlow switch. */
106 struct hmap ports; /* "struct port"s indexed by name. */
107 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
108 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
110 /* Port mirroring. */
111 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
113 /* Synthetic local port if necessary. */
114 struct ovsrec_port synth_local_port;
115 struct ovsrec_interface synth_local_iface;
116 struct ovsrec_interface *synth_local_ifacep;
119 /* All bridges, indexed by name. */
120 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
122 /* OVSDB IDL used to obtain configuration. */
123 static struct ovsdb_idl *idl;
125 /* Each time this timer expires, the bridge fetches systems and interface
126 * statistics and pushes them into the database. */
127 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
128 static long long int stats_timer = LLONG_MIN;
130 /* Stores the time after which rate limited statistics may be written to the
131 * database. Only updated when changes to the database require rate limiting.
133 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
134 static long long int db_limiter = LLONG_MIN;
136 static void add_del_bridges(const struct ovsrec_open_vswitch *);
137 static void bridge_del_ofprotos(void);
138 static bool bridge_add_ofprotos(struct bridge *);
139 static void bridge_create(const struct ovsrec_bridge *);
140 static void bridge_destroy(struct bridge *);
141 static struct bridge *bridge_lookup(const char *name);
142 static unixctl_cb_func bridge_unixctl_dump_flows;
143 static unixctl_cb_func bridge_unixctl_reconnect;
144 static size_t bridge_get_controllers(const struct bridge *br,
145 struct ovsrec_controller ***controllersp);
146 static void bridge_add_del_ports(struct bridge *);
147 static void bridge_add_ofproto_ports(struct bridge *);
148 static void bridge_del_ofproto_ports(struct bridge *);
149 static void bridge_refresh_ofp_port(struct bridge *);
150 static void bridge_configure_datapath_id(struct bridge *);
151 static void bridge_configure_flow_eviction_threshold(struct bridge *);
152 static void bridge_configure_netflow(struct bridge *);
153 static void bridge_configure_forward_bpdu(struct bridge *);
154 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
155 static void bridge_configure_remotes(struct bridge *,
156 const struct sockaddr_in *managers,
158 static void bridge_pick_local_hw_addr(struct bridge *,
159 uint8_t ea[ETH_ADDR_LEN],
160 struct iface **hw_addr_iface);
161 static uint64_t bridge_pick_datapath_id(struct bridge *,
162 const uint8_t bridge_ea[ETH_ADDR_LEN],
163 struct iface *hw_addr_iface);
164 static uint64_t dpid_from_hash(const void *, size_t nbytes);
165 static bool bridge_has_bond_fake_iface(const struct bridge *,
167 static bool port_is_bond_fake_iface(const struct port *);
169 static unixctl_cb_func qos_unixctl_show;
171 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
172 static void port_add_ifaces(struct port *);
173 static void port_del_ifaces(struct port *);
174 static void port_destroy(struct port *);
175 static struct port *port_lookup(const struct bridge *, const char *name);
176 static void port_configure(struct port *);
177 static struct lacp_settings *port_configure_lacp(struct port *,
178 struct lacp_settings *);
179 static void port_configure_bond(struct port *, struct bond_settings *,
180 uint32_t *bond_stable_ids);
182 static void bridge_configure_mirrors(struct bridge *);
183 static struct mirror *mirror_create(struct bridge *,
184 const struct ovsrec_mirror *);
185 static void mirror_destroy(struct mirror *);
186 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
188 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
189 static struct iface *iface_create(struct port *port,
190 const struct ovsrec_interface *if_cfg);
191 static void iface_destroy(struct iface *);
192 static struct iface *iface_lookup(const struct bridge *, const char *name);
193 static struct iface *iface_find(const char *name);
194 static struct iface *iface_from_ofp_port(const struct bridge *,
196 static void iface_set_mac(struct iface *);
197 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
198 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
199 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
200 static void iface_configure_cfm(struct iface *);
201 static void iface_refresh_cfm_stats(struct iface *);
202 static void iface_refresh_stats(struct iface *);
203 static void iface_refresh_status(struct iface *);
204 static bool iface_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_cfm_remote_mpids);
247 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
248 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
250 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
251 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
252 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
253 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
255 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
257 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
259 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
261 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
263 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
265 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
266 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
267 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
268 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
269 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
271 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
273 /* Register unixctl commands. */
274 unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL);
275 unixctl_command_register("bridge/dump-flows", "bridge",
276 bridge_unixctl_dump_flows, NULL);
277 unixctl_command_register("bridge/reconnect", "[bridge]",
278 bridge_unixctl_reconnect, NULL);
287 struct bridge *br, *next_br;
289 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
292 ovsdb_idl_destroy(idl);
295 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
296 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
297 * responsible for freeing '*managersp' (with free()).
299 * You may be asking yourself "why does ovs-vswitchd care?", because
300 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
301 * should not be and in fact is not directly involved in that. But
302 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
303 * it has to tell in-band control where the managers are to enable that.
304 * (Thus, only managers connected in-band are collected.)
307 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
308 struct sockaddr_in **managersp, size_t *n_managersp)
310 struct sockaddr_in *managers = NULL;
311 size_t n_managers = 0;
315 /* Collect all of the potential targets from the "targets" columns of the
316 * rows pointed to by "manager_options", excluding any that are
319 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
320 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
322 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
323 sset_find_and_delete(&targets, m->target);
325 sset_add(&targets, m->target);
329 /* Now extract the targets' IP addresses. */
330 if (!sset_is_empty(&targets)) {
333 managers = xmalloc(sset_count(&targets) * sizeof *managers);
334 SSET_FOR_EACH (target, &targets) {
335 struct sockaddr_in *sin = &managers[n_managers];
337 if ((!strncmp(target, "tcp:", 4)
338 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
339 (!strncmp(target, "ssl:", 4)
340 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
345 sset_destroy(&targets);
347 *managersp = managers;
348 *n_managersp = n_managers;
352 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
354 struct sockaddr_in *managers;
355 struct bridge *br, *next;
356 int sflow_bridge_number;
359 COVERAGE_INC(bridge_reconfigure);
361 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
362 * iface"s according to 'ovs_cfg', with only very minimal configuration
365 * This is purely an update to bridge data structures. Nothing is pushed
366 * down to ofproto or lower layers. */
367 add_del_bridges(ovs_cfg);
368 HMAP_FOR_EACH (br, node, &all_bridges) {
369 bridge_add_del_ports(br);
372 /* Delete all datapaths and datapath ports that are no longer configured.
374 * The kernel will reject any attempt to add a given port to a datapath if
375 * that port already belongs to a different datapath, so we must do all
376 * port deletions before any port additions. A datapath always has a
377 * "local port" so we must delete not-configured datapaths too. */
378 bridge_del_ofprotos();
379 HMAP_FOR_EACH (br, node, &all_bridges) {
381 bridge_del_ofproto_ports(br);
385 /* Create datapaths and datapath ports that are missing.
387 * After this is done, we have our final set of bridges, ports, and
388 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
389 * has at least one iface, every "struct iface" has a valid ofp_port and
391 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
392 if (!br->ofproto && !bridge_add_ofprotos(br)) {
396 HMAP_FOR_EACH (br, node, &all_bridges) {
397 bridge_refresh_ofp_port(br);
398 bridge_add_ofproto_ports(br);
401 /* Complete the configuration. */
402 sflow_bridge_number = 0;
403 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
404 HMAP_FOR_EACH (br, node, &all_bridges) {
407 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
410 port_configure(port);
412 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
413 iface_configure_cfm(iface);
414 iface_configure_qos(iface, port->cfg->qos);
415 iface_set_mac(iface);
418 bridge_configure_mirrors(br);
419 bridge_configure_datapath_id(br);
420 bridge_configure_flow_eviction_threshold(br);
421 bridge_configure_forward_bpdu(br);
422 bridge_configure_remotes(br, managers, n_managers);
423 bridge_configure_netflow(br);
424 bridge_configure_sflow(br, &sflow_bridge_number);
428 /* ovs-vswitchd has completed initialization, so allow the process that
429 * forked us to exit successfully. */
430 daemonize_complete();
433 /* Iterate over all ofprotos and delete any of them that do not have a
434 * configured bridge or that are the wrong type. */
436 bridge_del_ofprotos(void)
444 ofproto_enumerate_types(&types);
445 SSET_FOR_EACH (type, &types) {
448 ofproto_enumerate_names(type, &names);
449 SSET_FOR_EACH (name, &names) {
450 struct bridge *br = bridge_lookup(name);
451 if (!br || strcmp(type, br->type)) {
452 ofproto_delete(name, type);
456 sset_destroy(&names);
457 sset_destroy(&types);
461 bridge_add_ofprotos(struct bridge *br)
463 int error = ofproto_create(br->name, br->type, &br->ofproto);
465 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
472 port_configure(struct port *port)
474 const struct ovsrec_port *cfg = port->cfg;
475 struct bond_settings bond_settings;
476 struct lacp_settings lacp_settings;
477 struct ofproto_bundle_settings s;
485 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
486 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
487 s.slaves[s.n_slaves++] = iface->ofp_port;
493 if (list_is_short(&port->ifaces)) {
494 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
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. */
508 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
512 if (cfg->vlan_mode) {
513 if (!strcmp(cfg->vlan_mode, "access")) {
514 s.vlan_mode = PORT_VLAN_ACCESS;
515 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
516 s.vlan_mode = PORT_VLAN_TRUNK;
517 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
518 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
519 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
520 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
522 /* This "can't happen" because ovsdb-server should prevent it. */
523 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
524 s.vlan_mode = PORT_VLAN_TRUNK;
528 s.vlan_mode = PORT_VLAN_ACCESS;
530 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
534 s.vlan_mode = PORT_VLAN_TRUNK;
538 /* Get LACP settings. */
539 s.lacp = port_configure_lacp(port, &lacp_settings);
543 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
544 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
545 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
548 s.lacp_slaves = NULL;
551 /* Get bond settings. */
552 if (s.n_slaves > 1) {
553 s.bond = &bond_settings;
554 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
555 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
558 s.bond_stable_ids = NULL;
560 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
561 netdev_set_miimon_interval(iface->netdev, 0);
566 ofproto_bundle_register(port->bridge->ofproto, port, &s);
572 free(s.bond_stable_ids);
575 /* Pick local port hardware address and datapath ID for 'br'. */
577 bridge_configure_datapath_id(struct bridge *br)
579 uint8_t ea[ETH_ADDR_LEN];
581 struct iface *local_iface;
582 struct iface *hw_addr_iface;
585 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
586 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
588 int error = netdev_set_etheraddr(local_iface->netdev, ea);
590 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
591 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
592 "Ethernet address: %s",
593 br->name, strerror(error));
596 memcpy(br->ea, ea, ETH_ADDR_LEN);
598 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
599 ofproto_set_datapath_id(br->ofproto, dpid);
601 dpid_string = xasprintf("%016"PRIx64, dpid);
602 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
606 /* Set NetFlow configuration on 'br'. */
608 bridge_configure_netflow(struct bridge *br)
610 struct ovsrec_netflow *cfg = br->cfg->netflow;
611 struct netflow_options opts;
614 ofproto_set_netflow(br->ofproto, NULL);
618 memset(&opts, 0, sizeof opts);
620 /* Get default NetFlow configuration from datapath.
621 * Apply overrides from 'cfg'. */
622 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
623 if (cfg->engine_type) {
624 opts.engine_type = *cfg->engine_type;
626 if (cfg->engine_id) {
627 opts.engine_id = *cfg->engine_id;
630 /* Configure active timeout interval. */
631 opts.active_timeout = cfg->active_timeout;
632 if (!opts.active_timeout) {
633 opts.active_timeout = -1;
634 } else if (opts.active_timeout < 0) {
635 VLOG_WARN("bridge %s: active timeout interval set to negative "
636 "value, using default instead (%d seconds)", br->name,
637 NF_ACTIVE_TIMEOUT_DEFAULT);
638 opts.active_timeout = -1;
641 /* Add engine ID to interface number to disambiguate bridgs? */
642 opts.add_id_to_iface = cfg->add_id_to_interface;
643 if (opts.add_id_to_iface) {
644 if (opts.engine_id > 0x7f) {
645 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
646 "another vswitch, choose an engine id less than 128",
649 if (hmap_count(&br->ports) > 508) {
650 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
651 "another port when more than 508 ports are used",
657 sset_init(&opts.collectors);
658 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
661 if (ofproto_set_netflow(br->ofproto, &opts)) {
662 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
664 sset_destroy(&opts.collectors);
667 /* Set sFlow configuration on 'br'. */
669 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
671 const struct ovsrec_sflow *cfg = br->cfg->sflow;
672 struct ovsrec_controller **controllers;
673 struct ofproto_sflow_options oso;
674 size_t n_controllers;
678 ofproto_set_sflow(br->ofproto, NULL);
682 memset(&oso, 0, sizeof oso);
684 sset_init(&oso.targets);
685 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
687 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
689 oso.sampling_rate = *cfg->sampling;
692 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
694 oso.polling_interval = *cfg->polling;
697 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
699 oso.header_len = *cfg->header;
702 oso.sub_id = (*sflow_bridge_number)++;
703 oso.agent_device = cfg->agent;
705 oso.control_ip = NULL;
706 n_controllers = bridge_get_controllers(br, &controllers);
707 for (i = 0; i < n_controllers; i++) {
708 if (controllers[i]->local_ip) {
709 oso.control_ip = controllers[i]->local_ip;
713 ofproto_set_sflow(br->ofproto, &oso);
715 sset_destroy(&oso.targets);
719 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
721 const struct port *port = port_lookup(br, name);
722 return port && port_is_bond_fake_iface(port);
726 port_is_bond_fake_iface(const struct port *port)
728 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
732 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
734 struct bridge *br, *next;
738 /* Collect new bridges' names and types. */
740 for (i = 0; i < cfg->n_bridges; i++) {
741 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
742 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
744 if (strchr(br_cfg->name, '/')) {
745 /* Prevent remote ovsdb-server users from accessing arbitrary
746 * directories, e.g. consider a bridge named "../../../etc/". */
747 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
749 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
750 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
754 /* Get rid of deleted bridges or those whose types have changed.
755 * Update 'cfg' of bridges that still exist. */
756 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
757 br->cfg = shash_find_data(&new_br, br->name);
758 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
759 br->cfg->datapath_type))) {
764 /* Add new bridges. */
765 for (i = 0; i < cfg->n_bridges; i++) {
766 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
767 struct bridge *br = bridge_lookup(br_cfg->name);
769 bridge_create(br_cfg);
773 shash_destroy(&new_br);
776 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
779 * The kernel will reject any attempt to add a given port to a datapath if that
780 * port already belongs to a different datapath, so we must do all port
781 * deletions before any port additions. */
783 bridge_del_ofproto_ports(struct bridge *br)
785 struct ofproto_port_dump dump;
786 struct ofproto_port ofproto_port;
788 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
789 const char *name = ofproto_port.name;
794 /* Ignore the local port. We can't change it anyhow. */
795 if (!strcmp(name, br->name)) {
799 /* Get the type that 'ofproto_port' should have (ordinarily the
800 * type of its corresponding iface) or NULL if it should be
802 iface = iface_lookup(br, name);
803 type = (iface ? iface->type
804 : bridge_has_bond_fake_iface(br, name) ? "internal"
807 /* If it's the wrong type then delete the ofproto port. */
809 && !strcmp(ofproto_port.type, type)
810 && (!iface || !iface->netdev
811 || !strcmp(netdev_get_type(iface->netdev), type))) {
814 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
816 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
817 br->name, name, strerror(error));
820 netdev_close(iface->netdev);
821 iface->netdev = NULL;
827 iface_set_ofp_port(struct iface *iface, int ofp_port)
829 struct bridge *br = iface->port->bridge;
831 assert(iface->ofp_port < 0 && ofp_port >= 0);
832 iface->ofp_port = ofp_port;
833 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
834 iface_set_ofport(iface->cfg, ofp_port);
838 bridge_refresh_ofp_port(struct bridge *br)
840 struct ofproto_port_dump dump;
841 struct ofproto_port ofproto_port;
844 /* Clear all the "ofp_port"es. */
845 hmap_clear(&br->ifaces);
846 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
849 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
850 iface->ofp_port = -1;
854 /* Obtain the correct "ofp_port"s from ofproto. */
855 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
856 struct iface *iface = iface_lookup(br, ofproto_port.name);
858 if (iface->ofp_port >= 0) {
859 VLOG_WARN("bridge %s: interface %s reported twice",
860 br->name, ofproto_port.name);
861 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
862 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
863 br->name, ofproto_port.ofp_port);
865 iface_set_ofp_port(iface, ofproto_port.ofp_port);
871 /* Add an ofproto port for any "struct iface" that doesn't have one.
872 * Delete any "struct iface" for which this fails.
873 * Delete any "struct port" that thereby ends up with no ifaces. */
875 bridge_add_ofproto_ports(struct bridge *br)
877 struct port *port, *next_port;
879 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
880 struct iface *iface, *next_iface;
881 struct ofproto_port ofproto_port;
883 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
886 /* Open the netdev. */
887 if (!iface->netdev) {
888 error = netdev_open(iface->name, iface->type, &iface->netdev);
890 VLOG_WARN("could not open network device %s (%s)",
891 iface->name, strerror(error));
897 /* Configure the netdev. */
902 shash_from_ovs_idl_map(iface->cfg->key_options,
903 iface->cfg->value_options,
904 iface->cfg->n_options, &args);
905 error = netdev_set_config(iface->netdev, &args);
906 shash_destroy(&args);
909 VLOG_WARN("could not configure network device %s (%s)",
910 iface->name, strerror(error));
911 netdev_close(iface->netdev);
912 iface->netdev = NULL;
916 /* Add the port, if necessary. */
917 if (iface->netdev && iface->ofp_port < 0) {
921 error = ofproto_port_add(br->ofproto, iface->netdev,
924 iface_set_ofp_port(iface, ofp_port);
926 netdev_close(iface->netdev);
927 iface->netdev = NULL;
931 /* Populate stats columns in new Interface rows. */
932 if (iface->netdev && !iface->cfg->mtu) {
933 iface_refresh_stats(iface);
934 iface_refresh_status(iface);
937 /* Delete the iface if we failed. */
938 if (iface->netdev && iface->ofp_port >= 0) {
939 VLOG_DBG("bridge %s: interface %s is on port %d",
940 br->name, iface->name, iface->ofp_port);
943 VLOG_ERR("bridge %s: missing %s interface, dropping",
944 br->name, iface->name);
946 /* We already reported a related error, don't bother
949 iface_clear_db_record(iface->cfg);
950 iface_destroy(iface);
953 if (list_is_empty(&port->ifaces)) {
954 VLOG_WARN("%s port has no interfaces, dropping", port->name);
959 /* Add bond fake iface if necessary. */
960 if (port_is_bond_fake_iface(port)) {
961 if (ofproto_port_query_by_name(br->ofproto, port->name,
963 struct netdev *netdev;
966 error = netdev_open(port->name, "internal", &netdev);
968 ofproto_port_add(br->ofproto, netdev, NULL);
969 netdev_close(netdev);
971 VLOG_WARN("could not open network device %s (%s)",
972 port->name, strerror(error));
975 /* Already exists, nothing to do. */
976 ofproto_port_destroy(&ofproto_port);
983 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
984 const struct ovsdb_idl_column *column,
987 const struct ovsdb_datum *datum;
988 union ovsdb_atom atom;
991 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
992 atom.string = (char *) key;
993 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
994 return idx == UINT_MAX ? NULL : datum->values[idx].string;
998 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1000 return get_ovsrec_key_value(&br_cfg->header_,
1001 &ovsrec_bridge_col_other_config, key);
1004 /* Set Flow eviction threshold */
1006 bridge_configure_flow_eviction_threshold(struct bridge *br)
1008 const char *threshold_str;
1011 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1012 if (threshold_str) {
1013 threshold = strtoul(threshold_str, NULL, 10);
1015 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1017 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1020 /* Set forward BPDU option. */
1022 bridge_configure_forward_bpdu(struct bridge *br)
1024 const char *forward_bpdu_str;
1025 bool forward_bpdu = false;
1027 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1028 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1029 forward_bpdu = true;
1031 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1035 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1036 struct iface **hw_addr_iface)
1040 bool found_addr = false;
1043 *hw_addr_iface = NULL;
1045 /* Did the user request a particular MAC? */
1046 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1047 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1048 if (eth_addr_is_multicast(ea)) {
1049 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1050 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1051 } else if (eth_addr_is_zero(ea)) {
1052 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1058 /* Otherwise choose the minimum non-local MAC address among all of the
1060 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1061 uint8_t iface_ea[ETH_ADDR_LEN];
1062 struct iface *candidate;
1063 struct iface *iface;
1065 /* Mirror output ports don't participate. */
1066 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1070 /* Choose the MAC address to represent the port. */
1072 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1073 /* Find the interface with this Ethernet address (if any) so that
1074 * we can provide the correct devname to the caller. */
1075 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1076 uint8_t candidate_ea[ETH_ADDR_LEN];
1077 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1078 && eth_addr_equals(iface_ea, candidate_ea)) {
1083 /* Choose the interface whose MAC address will represent the port.
1084 * The Linux kernel bonding code always chooses the MAC address of
1085 * the first slave added to a bond, and the Fedora networking
1086 * scripts always add slaves to a bond in alphabetical order, so
1087 * for compatibility we choose the interface with the name that is
1088 * first in alphabetical order. */
1089 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1090 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1095 /* The local port doesn't count (since we're trying to choose its
1096 * MAC address anyway). */
1097 if (iface->ofp_port == OFPP_LOCAL) {
1102 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1108 /* Compare against our current choice. */
1109 if (!eth_addr_is_multicast(iface_ea) &&
1110 !eth_addr_is_local(iface_ea) &&
1111 !eth_addr_is_reserved(iface_ea) &&
1112 !eth_addr_is_zero(iface_ea) &&
1113 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1115 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1116 *hw_addr_iface = iface;
1121 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1122 br->name, ETH_ADDR_ARGS(ea));
1124 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1125 *hw_addr_iface = NULL;
1126 VLOG_WARN("bridge %s: using default bridge Ethernet "
1127 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1131 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1132 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1133 * an interface on 'br', then that interface must be passed in as
1134 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1135 * 'hw_addr_iface' must be passed in as a null pointer. */
1137 bridge_pick_datapath_id(struct bridge *br,
1138 const uint8_t bridge_ea[ETH_ADDR_LEN],
1139 struct iface *hw_addr_iface)
1142 * The procedure for choosing a bridge MAC address will, in the most
1143 * ordinary case, also choose a unique MAC that we can use as a datapath
1144 * ID. In some special cases, though, multiple bridges will end up with
1145 * the same MAC address. This is OK for the bridges, but it will confuse
1146 * the OpenFlow controller, because each datapath needs a unique datapath
1149 * Datapath IDs must be unique. It is also very desirable that they be
1150 * stable from one run to the next, so that policy set on a datapath
1153 const char *datapath_id;
1156 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1157 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1161 if (hw_addr_iface) {
1163 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1165 * A bridge whose MAC address is taken from a VLAN network device
1166 * (that is, a network device created with vconfig(8) or similar
1167 * tool) will have the same MAC address as a bridge on the VLAN
1168 * device's physical network device.
1170 * Handle this case by hashing the physical network device MAC
1171 * along with the VLAN identifier.
1173 uint8_t buf[ETH_ADDR_LEN + 2];
1174 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1175 buf[ETH_ADDR_LEN] = vlan >> 8;
1176 buf[ETH_ADDR_LEN + 1] = vlan;
1177 return dpid_from_hash(buf, sizeof buf);
1180 * Assume that this bridge's MAC address is unique, since it
1181 * doesn't fit any of the cases we handle specially.
1186 * A purely internal bridge, that is, one that has no non-virtual
1187 * network devices on it at all, is more difficult because it has no
1188 * natural unique identifier at all.
1190 * When the host is a XenServer, we handle this case by hashing the
1191 * host's UUID with the name of the bridge. Names of bridges are
1192 * persistent across XenServer reboots, although they can be reused if
1193 * an internal network is destroyed and then a new one is later
1194 * created, so this is fairly effective.
1196 * When the host is not a XenServer, we punt by using a random MAC
1197 * address on each run.
1199 const char *host_uuid = xenserver_get_host_uuid();
1201 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1202 dpid = dpid_from_hash(combined, strlen(combined));
1208 return eth_addr_to_uint64(bridge_ea);
1212 dpid_from_hash(const void *data, size_t n)
1214 uint8_t hash[SHA1_DIGEST_SIZE];
1216 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1217 sha1_bytes(data, n, hash);
1218 eth_addr_mark_random(hash);
1219 return eth_addr_to_uint64(hash);
1223 iface_refresh_status(struct iface *iface)
1227 enum netdev_flags flags;
1234 if (iface_is_synthetic(iface)) {
1240 if (!netdev_get_status(iface->netdev, &sh)) {
1242 char **keys, **values;
1244 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1245 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1250 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1253 shash_destroy_free_data(&sh);
1255 error = netdev_get_flags(iface->netdev, &flags);
1257 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1260 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1263 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1265 ovsrec_interface_set_duplex(iface->cfg,
1266 netdev_features_is_full_duplex(current)
1268 /* warning: uint64_t -> int64_t conversion */
1269 bps = netdev_features_to_bps(current);
1270 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1273 ovsrec_interface_set_duplex(iface->cfg, NULL);
1274 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1277 error = netdev_get_mtu(iface->netdev, &mtu);
1280 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1283 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1287 /* Writes 'iface''s CFM statistics to the database. */
1289 iface_refresh_cfm_stats(struct iface *iface)
1291 const struct ovsrec_interface *cfg = iface->cfg;
1293 const uint64_t *rmps;
1296 if (iface_is_synthetic(iface)) {
1300 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1303 bool fault_bool = fault;
1304 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1306 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1309 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1310 iface->ofp_port, &rmps, &n_rmps);
1312 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1315 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1320 iface_refresh_stats(struct iface *iface)
1322 #define IFACE_STATS \
1323 IFACE_STAT(rx_packets, "rx_packets") \
1324 IFACE_STAT(tx_packets, "tx_packets") \
1325 IFACE_STAT(rx_bytes, "rx_bytes") \
1326 IFACE_STAT(tx_bytes, "tx_bytes") \
1327 IFACE_STAT(rx_dropped, "rx_dropped") \
1328 IFACE_STAT(tx_dropped, "tx_dropped") \
1329 IFACE_STAT(rx_errors, "rx_errors") \
1330 IFACE_STAT(tx_errors, "tx_errors") \
1331 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1332 IFACE_STAT(rx_over_errors, "rx_over_err") \
1333 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1334 IFACE_STAT(collisions, "collisions")
1336 #define IFACE_STAT(MEMBER, NAME) NAME,
1337 static char *keys[] = { IFACE_STATS };
1339 int64_t values[ARRAY_SIZE(keys)];
1342 struct netdev_stats stats;
1344 if (iface_is_synthetic(iface)) {
1348 /* Intentionally ignore return value, since errors will set 'stats' to
1349 * all-1s, and we will deal with that correctly below. */
1350 netdev_get_stats(iface->netdev, &stats);
1352 /* Copy statistics into values[] array. */
1354 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1357 assert(i == ARRAY_SIZE(keys));
1359 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1364 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1368 /* Use other-config:enable-system-stats by preference. */
1369 enable = get_ovsrec_key_value(&cfg->header_,
1370 &ovsrec_open_vswitch_col_other_config,
1371 "enable-statistics");
1373 return !strcmp(enable, "true");
1376 /* Disable by default. */
1381 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1383 struct ovsdb_datum datum;
1387 if (enable_system_stats(cfg)) {
1388 get_system_stats(&stats);
1391 ovsdb_datum_from_shash(&datum, &stats);
1392 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1396 static inline const char *
1397 nx_role_to_str(enum nx_role role)
1402 case NX_ROLE_MASTER:
1407 return "*** INVALID ROLE ***";
1412 refresh_controller_status(void)
1416 const struct ovsrec_controller *cfg;
1420 /* Accumulate status for controllers on all bridges. */
1421 HMAP_FOR_EACH (br, node, &all_bridges) {
1422 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1425 /* Update each controller in the database with current status. */
1426 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1427 struct ofproto_controller_info *cinfo =
1428 shash_find_data(&info, cfg->target);
1431 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1432 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1433 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1434 (char **) cinfo->pairs.values,
1437 ovsrec_controller_set_is_connected(cfg, false);
1438 ovsrec_controller_set_role(cfg, NULL);
1439 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1443 ofproto_free_ofproto_controller_info(&info);
1447 refresh_cfm_stats(void)
1449 static struct ovsdb_idl_txn *txn = NULL;
1454 txn = ovsdb_idl_txn_create(idl);
1456 HMAP_FOR_EACH (br, node, &all_bridges) {
1457 struct iface *iface;
1459 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1460 iface_refresh_cfm_stats(iface);
1465 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1466 ovsdb_idl_txn_destroy(txn);
1474 const struct ovsrec_open_vswitch *cfg;
1476 bool datapath_destroyed;
1477 bool database_changed;
1480 /* (Re)configure if necessary. */
1481 database_changed = ovsdb_idl_run(idl);
1482 if (ovsdb_idl_is_lock_contended(idl)) {
1483 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1484 struct bridge *br, *next_br;
1486 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1487 "disabling this process until it goes away");
1489 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1493 } else if (!ovsdb_idl_has_lock(idl)) {
1496 cfg = ovsrec_open_vswitch_first(idl);
1498 /* Let each bridge do the work that it needs to do. */
1499 datapath_destroyed = false;
1500 HMAP_FOR_EACH (br, node, &all_bridges) {
1501 int error = ofproto_run(br->ofproto);
1503 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1504 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1505 "forcing reconfiguration", br->name);
1506 datapath_destroyed = true;
1510 /* Re-configure SSL. We do this on every trip through the main loop,
1511 * instead of just when the database changes, because the contents of the
1512 * key and certificate files can change without the database changing.
1514 * We do this before bridge_reconfigure() because that function might
1515 * initiate SSL connections and thus requires SSL to be configured. */
1516 if (cfg && cfg->ssl) {
1517 const struct ovsrec_ssl *ssl = cfg->ssl;
1519 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1520 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1523 if (database_changed || datapath_destroyed) {
1525 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1527 bridge_reconfigure(cfg);
1529 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1530 ovsdb_idl_txn_commit(txn);
1531 ovsdb_idl_txn_destroy(txn); /* XXX */
1533 /* We still need to reconfigure to avoid dangling pointers to
1534 * now-destroyed ovsrec structures inside bridge data. */
1535 static const struct ovsrec_open_vswitch null_cfg;
1537 bridge_reconfigure(&null_cfg);
1541 /* Refresh system and interface stats if necessary. */
1542 if (time_msec() >= stats_timer) {
1544 struct ovsdb_idl_txn *txn;
1546 txn = ovsdb_idl_txn_create(idl);
1547 HMAP_FOR_EACH (br, node, &all_bridges) {
1550 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1551 struct iface *iface;
1553 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1554 iface_refresh_stats(iface);
1555 iface_refresh_status(iface);
1559 refresh_system_stats(cfg);
1560 refresh_controller_status();
1561 ovsdb_idl_txn_commit(txn);
1562 ovsdb_idl_txn_destroy(txn); /* XXX */
1565 stats_timer = time_msec() + STATS_INTERVAL;
1568 if (time_msec() >= db_limiter) {
1569 struct ovsdb_idl_txn *txn;
1571 txn = ovsdb_idl_txn_create(idl);
1572 HMAP_FOR_EACH (br, node, &all_bridges) {
1573 struct iface *iface;
1575 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1576 const char *link_state;
1579 if (iface_is_synthetic(iface)) {
1583 current = ofproto_port_is_lacp_current(br->ofproto,
1587 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1589 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1592 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1593 ovsrec_interface_set_link_state(iface->cfg, link_state);
1597 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
1598 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1600 ovsdb_idl_txn_destroy(txn);
1603 refresh_cfm_stats();
1609 ovsdb_idl_wait(idl);
1610 if (!hmap_is_empty(&all_bridges)) {
1613 HMAP_FOR_EACH (br, node, &all_bridges) {
1614 ofproto_wait(br->ofproto);
1616 poll_timer_wait_until(stats_timer);
1618 if (db_limiter > time_msec()) {
1619 poll_timer_wait_until(db_limiter);
1624 /* QoS unixctl user interface functions. */
1626 struct qos_unixctl_show_cbdata {
1628 struct iface *iface;
1632 qos_unixctl_show_cb(unsigned int queue_id,
1633 const struct shash *details,
1636 struct qos_unixctl_show_cbdata *data = aux;
1637 struct ds *ds = data->ds;
1638 struct iface *iface = data->iface;
1639 struct netdev_queue_stats stats;
1640 struct shash_node *node;
1643 ds_put_cstr(ds, "\n");
1645 ds_put_format(ds, "Queue %u:\n", queue_id);
1647 ds_put_cstr(ds, "Default:\n");
1650 SHASH_FOR_EACH (node, details) {
1651 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1654 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1656 if (stats.tx_packets != UINT64_MAX) {
1657 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1660 if (stats.tx_bytes != UINT64_MAX) {
1661 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1664 if (stats.tx_errors != UINT64_MAX) {
1665 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1668 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1669 queue_id, strerror(error));
1674 qos_unixctl_show(struct unixctl_conn *conn,
1675 const char *args, void *aux OVS_UNUSED)
1677 struct ds ds = DS_EMPTY_INITIALIZER;
1678 struct shash sh = SHASH_INITIALIZER(&sh);
1679 struct iface *iface;
1681 struct shash_node *node;
1682 struct qos_unixctl_show_cbdata data;
1685 iface = iface_find(args);
1687 unixctl_command_reply(conn, 501, "no such interface");
1691 netdev_get_qos(iface->netdev, &type, &sh);
1693 if (*type != '\0') {
1694 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1696 SHASH_FOR_EACH (node, &sh) {
1697 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1702 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1705 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1707 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1709 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1710 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1713 shash_destroy_free_data(&sh);
1717 /* Bridge reconfiguration functions. */
1719 bridge_create(const struct ovsrec_bridge *br_cfg)
1723 assert(!bridge_lookup(br_cfg->name));
1724 br = xzalloc(sizeof *br);
1726 br->name = xstrdup(br_cfg->name);
1727 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1730 /* Derive the default Ethernet address from the bridge's UUID. This should
1731 * be unique and it will be stable between ovs-vswitchd runs. */
1732 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1733 eth_addr_mark_random(br->default_ea);
1735 hmap_init(&br->ports);
1736 hmap_init(&br->ifaces);
1737 hmap_init(&br->iface_by_name);
1738 hmap_init(&br->mirrors);
1740 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1744 bridge_destroy(struct bridge *br)
1747 struct mirror *mirror, *next_mirror;
1748 struct port *port, *next_port;
1750 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1753 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1754 mirror_destroy(mirror);
1756 hmap_remove(&all_bridges, &br->node);
1757 ofproto_destroy(br->ofproto);
1758 hmap_destroy(&br->ifaces);
1759 hmap_destroy(&br->ports);
1760 hmap_destroy(&br->iface_by_name);
1761 hmap_destroy(&br->mirrors);
1768 static struct bridge *
1769 bridge_lookup(const char *name)
1773 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1774 if (!strcmp(br->name, name)) {
1781 /* Handle requests for a listing of all flows known by the OpenFlow
1782 * stack, including those normally hidden. */
1784 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1785 const char *args, void *aux OVS_UNUSED)
1790 br = bridge_lookup(args);
1792 unixctl_command_reply(conn, 501, "Unknown bridge");
1797 ofproto_get_all_flows(br->ofproto, &results);
1799 unixctl_command_reply(conn, 200, ds_cstr(&results));
1800 ds_destroy(&results);
1803 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1804 * connections and reconnect. If BRIDGE is not specified, then all bridges
1805 * drop their controller connections and reconnect. */
1807 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1808 const char *args, void *aux OVS_UNUSED)
1811 if (args[0] != '\0') {
1812 br = bridge_lookup(args);
1814 unixctl_command_reply(conn, 501, "Unknown bridge");
1817 ofproto_reconnect_controllers(br->ofproto);
1819 HMAP_FOR_EACH (br, node, &all_bridges) {
1820 ofproto_reconnect_controllers(br->ofproto);
1823 unixctl_command_reply(conn, 200, NULL);
1827 bridge_get_controllers(const struct bridge *br,
1828 struct ovsrec_controller ***controllersp)
1830 struct ovsrec_controller **controllers;
1831 size_t n_controllers;
1833 controllers = br->cfg->controller;
1834 n_controllers = br->cfg->n_controller;
1836 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1842 *controllersp = controllers;
1844 return n_controllers;
1847 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1848 * those configured in 'br->cfg'. */
1850 bridge_add_del_ports(struct bridge *br)
1852 struct port *port, *next;
1853 struct shash_node *node;
1854 struct shash new_ports;
1857 /* Collect new ports. */
1858 shash_init(&new_ports);
1859 for (i = 0; i < br->cfg->n_ports; i++) {
1860 const char *name = br->cfg->ports[i]->name;
1861 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1862 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1866 if (bridge_get_controllers(br, NULL)
1867 && !shash_find(&new_ports, br->name)) {
1868 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1869 br->name, br->name);
1871 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1872 br->synth_local_port.n_interfaces = 1;
1873 br->synth_local_port.name = br->name;
1875 br->synth_local_iface.name = br->name;
1876 br->synth_local_iface.type = "internal";
1878 br->synth_local_ifacep = &br->synth_local_iface;
1880 shash_add(&new_ports, br->name, &br->synth_local_port);
1883 /* Get rid of deleted ports.
1884 * Get rid of deleted interfaces on ports that still exist.
1885 * Update 'cfg' of ports that still exist. */
1886 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1887 port->cfg = shash_find_data(&new_ports, port->name);
1891 port_del_ifaces(port);
1895 /* Create new ports.
1896 * Add new interfaces to existing ports. */
1897 SHASH_FOR_EACH (node, &new_ports) {
1898 struct port *port = port_lookup(br, node->name);
1900 struct ovsrec_port *cfg = node->data;
1901 port = port_create(br, cfg);
1903 port_add_ifaces(port);
1904 if (list_is_empty(&port->ifaces)) {
1905 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1906 br->name, port->name);
1910 shash_destroy(&new_ports);
1913 /* Initializes 'oc' appropriately as a management service controller for
1916 * The caller must free oc->target when it is no longer needed. */
1918 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1919 struct ofproto_controller *oc)
1921 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1922 oc->max_backoff = 0;
1923 oc->probe_interval = 60;
1924 oc->band = OFPROTO_OUT_OF_BAND;
1926 oc->burst_limit = 0;
1929 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1931 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1932 struct ofproto_controller *oc)
1934 oc->target = c->target;
1935 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1936 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1937 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1938 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1939 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1940 oc->burst_limit = (c->controller_burst_limit
1941 ? *c->controller_burst_limit : 0);
1944 /* Configures the IP stack for 'br''s local interface properly according to the
1945 * configuration in 'c'. */
1947 bridge_configure_local_iface_netdev(struct bridge *br,
1948 struct ovsrec_controller *c)
1950 struct netdev *netdev;
1951 struct in_addr mask, gateway;
1953 struct iface *local_iface;
1956 /* If there's no local interface or no IP address, give up. */
1957 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1958 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1962 /* Bring up the local interface. */
1963 netdev = local_iface->netdev;
1964 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1966 /* Configure the IP address and netmask. */
1967 if (!c->local_netmask
1968 || !inet_aton(c->local_netmask, &mask)
1970 mask.s_addr = guess_netmask(ip.s_addr);
1972 if (!netdev_set_in4(netdev, ip, mask)) {
1973 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1974 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1977 /* Configure the default gateway. */
1978 if (c->local_gateway
1979 && inet_aton(c->local_gateway, &gateway)
1980 && gateway.s_addr) {
1981 if (!netdev_add_router(netdev, gateway)) {
1982 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1983 br->name, IP_ARGS(&gateway.s_addr));
1989 bridge_configure_remotes(struct bridge *br,
1990 const struct sockaddr_in *managers, size_t n_managers)
1992 const char *disable_ib_str, *queue_id_str;
1993 bool disable_in_band = false;
1996 struct ovsrec_controller **controllers;
1997 size_t n_controllers;
1999 enum ofproto_fail_mode fail_mode;
2001 struct ofproto_controller *ocs;
2005 /* Check if we should disable in-band control on this bridge. */
2006 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2007 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2008 disable_in_band = true;
2011 /* Set OpenFlow queue ID for in-band control. */
2012 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2013 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2014 ofproto_set_in_band_queue(br->ofproto, queue_id);
2016 if (disable_in_band) {
2017 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2019 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2022 n_controllers = bridge_get_controllers(br, &controllers);
2024 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2027 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2028 for (i = 0; i < n_controllers; i++) {
2029 struct ovsrec_controller *c = controllers[i];
2031 if (!strncmp(c->target, "punix:", 6)
2032 || !strncmp(c->target, "unix:", 5)) {
2033 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2035 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2036 * domain sockets and overwriting arbitrary local files. */
2037 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
2038 "controller \"%s\" due to possibility for remote "
2039 "exploit", br->name, c->target);
2043 bridge_configure_local_iface_netdev(br, c);
2044 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2045 if (disable_in_band) {
2046 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2051 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2052 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2055 /* Set the fail-mode. */
2056 fail_mode = !br->cfg->fail_mode
2057 || !strcmp(br->cfg->fail_mode, "standalone")
2058 ? OFPROTO_FAIL_STANDALONE
2059 : OFPROTO_FAIL_SECURE;
2060 ofproto_set_fail_mode(br->ofproto, fail_mode);
2062 /* Configure OpenFlow controller connection snooping. */
2063 if (!ofproto_has_snoops(br->ofproto)) {
2067 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2068 ovs_rundir(), br->name));
2069 ofproto_set_snoops(br->ofproto, &snoops);
2070 sset_destroy(&snoops);
2074 /* Port functions. */
2076 static struct port *
2077 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2081 port = xzalloc(sizeof *port);
2083 port->name = xstrdup(cfg->name);
2085 list_init(&port->ifaces);
2087 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2089 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2095 get_port_other_config(const struct ovsrec_port *port, const char *key,
2096 const char *default_value)
2100 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2102 return value ? value : default_value;
2106 get_interface_other_config(const struct ovsrec_interface *iface,
2107 const char *key, const char *default_value)
2111 value = get_ovsrec_key_value(&iface->header_,
2112 &ovsrec_interface_col_other_config, key);
2113 return value ? value : default_value;
2116 /* Deletes interfaces from 'port' that are no longer configured for it. */
2118 port_del_ifaces(struct port *port)
2120 struct iface *iface, *next;
2121 struct sset new_ifaces;
2124 /* Collect list of new interfaces. */
2125 sset_init(&new_ifaces);
2126 for (i = 0; i < port->cfg->n_interfaces; i++) {
2127 const char *name = port->cfg->interfaces[i]->name;
2128 const char *type = port->cfg->interfaces[i]->name;
2129 if (strcmp(type, "null")) {
2130 sset_add(&new_ifaces, name);
2134 /* Get rid of deleted interfaces. */
2135 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2136 if (!sset_contains(&new_ifaces, iface->name)) {
2137 iface_destroy(iface);
2141 sset_destroy(&new_ifaces);
2144 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2147 port_add_ifaces(struct port *port)
2149 struct shash new_ifaces;
2150 struct shash_node *node;
2153 /* Collect new ifaces. */
2154 shash_init(&new_ifaces);
2155 for (i = 0; i < port->cfg->n_interfaces; i++) {
2156 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2157 if (strcmp(cfg->type, "null")
2158 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2159 VLOG_WARN("port %s: %s specified twice as port interface",
2160 port->name, cfg->name);
2161 iface_clear_db_record(cfg);
2165 /* Create new interfaces.
2166 * Update interface types and 'cfg' members. */
2167 SHASH_FOR_EACH (node, &new_ifaces) {
2168 const struct ovsrec_interface *cfg = node->data;
2169 const char *iface_name = node->name;
2170 struct iface *iface;
2172 iface = iface_lookup(port->bridge, iface_name);
2174 iface = iface_create(port, cfg);
2179 /* Determine interface type. The local port always has type
2180 * "internal". Other ports take their type from the database and
2181 * default to "system" if none is specified. */
2182 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2183 : cfg->type[0] ? cfg->type
2186 shash_destroy(&new_ifaces);
2190 port_destroy(struct port *port)
2193 struct bridge *br = port->bridge;
2194 struct iface *iface, *next;
2197 ofproto_bundle_unregister(br->ofproto, port);
2200 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2201 iface_destroy(iface);
2204 hmap_remove(&br->ports, &port->hmap_node);
2206 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2213 static struct port *
2214 port_lookup(const struct bridge *br, const char *name)
2218 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2220 if (!strcmp(port->name, name)) {
2228 enable_lacp(struct port *port, bool *activep)
2230 if (!port->cfg->lacp) {
2231 /* XXX when LACP implementation has been sufficiently tested, enable by
2232 * default and make active on bonded ports. */
2234 } else if (!strcmp(port->cfg->lacp, "off")) {
2236 } else if (!strcmp(port->cfg->lacp, "active")) {
2239 } else if (!strcmp(port->cfg->lacp, "passive")) {
2243 VLOG_WARN("port %s: unknown LACP mode %s",
2244 port->name, port->cfg->lacp);
2249 static struct lacp_settings *
2250 port_configure_lacp(struct port *port, struct lacp_settings *s)
2252 const char *lacp_time;
2253 long long int custom_time;
2256 if (!enable_lacp(port, &s->active)) {
2260 s->name = port->name;
2261 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2263 /* Prefer bondable links if unspecified. */
2264 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2266 s->priority = (priority > 0 && priority <= UINT16_MAX
2268 : UINT16_MAX - !list_is_short(&port->ifaces));
2270 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2275 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2276 custom_time = atoi(lacp_time);
2277 if (!strcmp(lacp_time, "fast")) {
2278 s->lacp_time = LACP_TIME_FAST;
2279 } else if (!strcmp(lacp_time, "slow")) {
2280 s->lacp_time = LACP_TIME_SLOW;
2281 } else if (custom_time > 0) {
2282 s->lacp_time = LACP_TIME_CUSTOM;
2283 s->custom_time = custom_time;
2285 s->lacp_time = LACP_TIME_SLOW;
2292 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2294 int priority, portid, key;
2296 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2297 priority = atoi(get_interface_other_config(iface->cfg,
2298 "lacp-port-priority", "0"));
2299 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2302 if (portid <= 0 || portid > UINT16_MAX) {
2303 portid = iface->ofp_port;
2306 if (priority <= 0 || priority > UINT16_MAX) {
2307 priority = UINT16_MAX;
2310 if (key < 0 || key > UINT16_MAX) {
2314 s->name = iface->name;
2316 s->priority = priority;
2321 port_configure_bond(struct port *port, struct bond_settings *s,
2322 uint32_t *bond_stable_ids)
2324 const char *detect_s;
2325 struct iface *iface;
2326 int miimon_interval;
2329 s->name = port->name;
2330 s->balance = BM_SLB;
2331 if (port->cfg->bond_mode
2332 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2333 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2334 port->name, port->cfg->bond_mode,
2335 bond_mode_to_string(s->balance));
2337 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2338 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2339 "please use another bond type or disable flood_vlans",
2343 miimon_interval = atoi(get_port_other_config(port->cfg,
2344 "bond-miimon-interval", "0"));
2345 if (miimon_interval <= 0) {
2346 miimon_interval = 200;
2349 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2350 if (!strcmp(detect_s, "carrier")) {
2351 miimon_interval = 0;
2352 } else if (strcmp(detect_s, "miimon")) {
2353 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2354 "defaulting to carrier", port->name, detect_s);
2355 miimon_interval = 0;
2358 s->up_delay = MAX(0, port->cfg->bond_updelay);
2359 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2360 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2361 s->rebalance_interval = atoi(
2362 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2363 if (s->rebalance_interval < 1000) {
2364 s->rebalance_interval = 1000;
2367 s->fake_iface = port->cfg->bond_fake_iface;
2370 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2371 long long stable_id;
2373 stable_id = atoll(get_interface_other_config(iface->cfg,
2374 "bond-stable-id", "0"));
2375 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2376 stable_id = iface->ofp_port;
2378 bond_stable_ids[i++] = stable_id;
2380 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2384 /* Interface functions. */
2386 static struct iface *
2387 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2389 struct bridge *br = port->bridge;
2390 struct iface *iface;
2391 char *name = if_cfg->name;
2393 iface = xzalloc(sizeof *iface);
2395 iface->name = xstrdup(name);
2396 iface->ofp_port = -1;
2397 iface->tag = tag_create_random();
2398 iface->netdev = NULL;
2399 iface->cfg = if_cfg;
2401 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2403 list_push_back(&port->ifaces, &iface->port_elem);
2405 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2411 iface_destroy(struct iface *iface)
2414 struct port *port = iface->port;
2415 struct bridge *br = port->bridge;
2417 if (br->ofproto && iface->ofp_port >= 0) {
2418 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2421 if (iface->ofp_port >= 0) {
2422 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2425 list_remove(&iface->port_elem);
2426 hmap_remove(&br->iface_by_name, &iface->name_node);
2428 netdev_close(iface->netdev);
2435 static struct iface *
2436 iface_lookup(const struct bridge *br, const char *name)
2438 struct iface *iface;
2440 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2441 &br->iface_by_name) {
2442 if (!strcmp(iface->name, name)) {
2450 static struct iface *
2451 iface_find(const char *name)
2453 const struct bridge *br;
2455 HMAP_FOR_EACH (br, node, &all_bridges) {
2456 struct iface *iface = iface_lookup(br, name);
2465 static struct iface *
2466 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2468 struct iface *iface;
2470 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2471 hash_int(ofp_port, 0), &br->ifaces) {
2472 if (iface->ofp_port == ofp_port) {
2479 /* Set Ethernet address of 'iface', if one is specified in the configuration
2482 iface_set_mac(struct iface *iface)
2484 uint8_t ea[ETH_ADDR_LEN];
2486 if (!strcmp(iface->type, "internal")
2487 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2488 if (iface->ofp_port == OFPP_LOCAL) {
2489 VLOG_ERR("interface %s: ignoring mac in Interface record "
2490 "(use Bridge record to set local port's mac)",
2492 } else if (eth_addr_is_multicast(ea)) {
2493 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2496 int error = netdev_set_etheraddr(iface->netdev, ea);
2498 VLOG_ERR("interface %s: setting MAC failed (%s)",
2499 iface->name, strerror(error));
2505 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2507 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2509 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2510 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2514 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
2515 * sets the "ofport" field to -1.
2517 * This is appropriate when 'if_cfg''s interface cannot be created or is
2518 * otherwise invalid. */
2520 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
2522 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2523 iface_set_ofport(if_cfg, -1);
2524 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
2525 ovsrec_interface_set_admin_state(if_cfg, NULL);
2526 ovsrec_interface_set_duplex(if_cfg, NULL);
2527 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
2528 ovsrec_interface_set_link_state(if_cfg, NULL);
2529 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
2530 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
2531 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
2532 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
2533 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
2537 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2539 * The value strings in '*shash' are taken directly from values[], not copied,
2540 * so the caller should not modify or free them. */
2542 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2543 struct shash *shash)
2548 for (i = 0; i < n; i++) {
2549 shash_add(shash, keys[i], values[i]);
2553 /* Creates 'keys' and 'values' arrays from 'shash'.
2555 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2556 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2557 * are populated with with strings taken directly from 'shash' and thus have
2558 * the same ownership of the key-value pairs in shash.
2561 shash_to_ovs_idl_map(struct shash *shash,
2562 char ***keys, char ***values, size_t *n)
2566 struct shash_node *sn;
2568 count = shash_count(shash);
2570 k = xmalloc(count * sizeof *k);
2571 v = xmalloc(count * sizeof *v);
2574 SHASH_FOR_EACH(sn, shash) {
2585 struct iface_delete_queues_cbdata {
2586 struct netdev *netdev;
2587 const struct ovsdb_datum *queues;
2591 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2593 union ovsdb_atom atom;
2595 atom.integer = target;
2596 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2600 iface_delete_queues(unsigned int queue_id,
2601 const struct shash *details OVS_UNUSED, void *cbdata_)
2603 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2605 if (!queue_ids_include(cbdata->queues, queue_id)) {
2606 netdev_delete_queue(cbdata->netdev, queue_id);
2611 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2613 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2614 netdev_set_qos(iface->netdev, NULL, NULL);
2616 struct iface_delete_queues_cbdata cbdata;
2617 struct shash details;
2621 /* Configure top-level Qos for 'iface'. */
2622 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2623 qos->n_other_config, &details);
2624 netdev_set_qos(iface->netdev, qos->type, &details);
2625 shash_destroy(&details);
2627 /* Deconfigure queues that were deleted. */
2628 cbdata.netdev = iface->netdev;
2629 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2631 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2633 /* Configure queues for 'iface'. */
2635 for (i = 0; i < qos->n_queues; i++) {
2636 const struct ovsrec_queue *queue = qos->value_queues[i];
2637 unsigned int queue_id = qos->key_queues[i];
2639 if (queue_id == 0) {
2643 shash_from_ovs_idl_map(queue->key_other_config,
2644 queue->value_other_config,
2645 queue->n_other_config, &details);
2646 netdev_set_queue(iface->netdev, queue_id, &details);
2647 shash_destroy(&details);
2650 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2651 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2652 "queue (queue 0). Packets not directed to a "
2653 "correctly configured queue may be dropped.",
2658 netdev_set_policing(iface->netdev,
2659 iface->cfg->ingress_policing_rate,
2660 iface->cfg->ingress_policing_burst);
2664 iface_configure_cfm(struct iface *iface)
2666 const struct ovsrec_interface *cfg = iface->cfg;
2667 const char *extended_str, *opstate_str;
2668 struct cfm_settings s;
2670 if (!cfg->n_cfm_mpid) {
2671 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2675 s.mpid = *cfg->cfm_mpid;
2676 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2678 if (s.interval <= 0) {
2682 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
2684 s.extended = !strcasecmp("true", extended_str);
2686 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
2687 s.opup = !strcasecmp("up", opstate_str);
2689 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2692 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2693 * instead of obtaining it from the database. */
2695 iface_is_synthetic(const struct iface *iface)
2697 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2700 /* Port mirroring. */
2702 static struct mirror *
2703 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2707 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2708 if (uuid_equals(uuid, &m->uuid)) {
2716 bridge_configure_mirrors(struct bridge *br)
2718 const struct ovsdb_datum *mc;
2719 unsigned long *flood_vlans;
2720 struct mirror *m, *next;
2723 /* Get rid of deleted mirrors. */
2724 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2725 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2726 union ovsdb_atom atom;
2728 atom.uuid = m->uuid;
2729 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2734 /* Add new mirrors and reconfigure existing ones. */
2735 for (i = 0; i < br->cfg->n_mirrors; i++) {
2736 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2737 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2739 m = mirror_create(br, cfg);
2741 if (!mirror_configure(m, cfg)) {
2746 /* Update flooded vlans (for RSPAN). */
2747 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2748 br->cfg->n_flood_vlans);
2749 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2750 bitmap_free(flood_vlans);
2753 static struct mirror *
2754 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2758 m = xzalloc(sizeof *m);
2759 m->uuid = cfg->header_.uuid;
2760 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2762 m->name = xstrdup(cfg->name);
2768 mirror_destroy(struct mirror *m)
2771 struct bridge *br = m->bridge;
2774 ofproto_mirror_unregister(br->ofproto, m);
2777 hmap_remove(&br->mirrors, &m->hmap_node);
2784 mirror_collect_ports(struct mirror *m,
2785 struct ovsrec_port **in_ports, int n_in_ports,
2786 void ***out_portsp, size_t *n_out_portsp)
2788 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2789 size_t n_out_ports = 0;
2792 for (i = 0; i < n_in_ports; i++) {
2793 const char *name = in_ports[i]->name;
2794 struct port *port = port_lookup(m->bridge, name);
2796 out_ports[n_out_ports++] = port;
2798 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2799 "port %s", m->bridge->name, m->name, name);
2802 *out_portsp = out_ports;
2803 *n_out_portsp = n_out_ports;
2807 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2809 struct ofproto_mirror_settings s;
2812 if (strcmp(cfg->name, m->name)) {
2814 m->name = xstrdup(cfg->name);
2818 /* Get output port or VLAN. */
2819 if (cfg->output_port) {
2820 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2821 if (!s.out_bundle) {
2822 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2823 m->bridge->name, m->name);
2826 s.out_vlan = UINT16_MAX;
2828 if (cfg->output_vlan) {
2829 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2830 "output vlan; ignoring output vlan",
2831 m->bridge->name, m->name);
2833 } else if (cfg->output_vlan) {
2834 /* The database should prevent invalid VLAN values. */
2835 s.out_bundle = NULL;
2836 s.out_vlan = *cfg->output_vlan;
2838 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2839 m->bridge->name, m->name);
2843 /* Get port selection. */
2844 if (cfg->select_all) {
2845 size_t n_ports = hmap_count(&m->bridge->ports);
2846 void **ports = xmalloc(n_ports * sizeof *ports);
2851 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2861 /* Get ports, dropping ports that don't exist.
2862 * The IDL ensures that there are no duplicates. */
2863 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2864 &s.srcs, &s.n_srcs);
2865 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2866 &s.dsts, &s.n_dsts);
2869 /* Get VLAN selection. */
2870 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2873 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2876 if (s.srcs != s.dsts) {