1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 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 "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
44 #include "socket-util.h"
46 #include "stream-ssl.h"
48 #include "system-stats.h"
53 #include "vswitchd/vswitch-idl.h"
54 #include "xenserver.h"
56 #include "sflow_api.h"
57 #include "vlan-bitmap.h"
59 VLOG_DEFINE_THIS_MODULE(bridge);
61 COVERAGE_DEFINE(bridge_reconfigure);
64 /* These members are always valid. */
65 struct list port_elem; /* Element in struct port's "ifaces" list. */
66 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
67 struct port *port; /* Containing port. */
68 char *name; /* Host network device name. */
69 tag_type tag; /* Tag associated with this interface. */
71 /* These members are valid only after bridge_reconfigure() causes them to
73 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
74 int ofp_port; /* OpenFlow port number, -1 if unknown. */
75 struct netdev *netdev; /* Network device. */
76 const char *type; /* Usually same as cfg->type. */
77 const struct ovsrec_interface *cfg;
81 struct uuid uuid; /* UUID of this "mirror" record in database. */
82 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
83 struct bridge *bridge;
85 const struct ovsrec_mirror *cfg;
89 struct bridge *bridge;
90 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
93 const struct ovsrec_port *cfg;
95 /* An ordinary bridge port has 1 interface.
96 * A bridge port for bonding has at least 2 interfaces. */
97 struct list ifaces; /* List of "struct iface"s. */
101 struct hmap_node node; /* In 'all_bridges'. */
102 char *name; /* User-specified arbitrary name. */
103 char *type; /* Datapath type. */
104 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
105 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
106 const struct ovsrec_bridge *cfg;
108 /* OpenFlow switch processing. */
109 struct ofproto *ofproto; /* OpenFlow switch. */
112 struct hmap ports; /* "struct port"s indexed by name. */
113 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
114 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
116 /* Port mirroring. */
117 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
119 /* Synthetic local port if necessary. */
120 struct ovsrec_port synth_local_port;
121 struct ovsrec_interface synth_local_iface;
122 struct ovsrec_interface *synth_local_ifacep;
125 /* All bridges, indexed by name. */
126 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
128 /* OVSDB IDL used to obtain configuration. */
129 static struct ovsdb_idl *idl;
131 /* Each time this timer expires, the bridge fetches systems and interface
132 * statistics and pushes them into the database. */
133 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
134 static long long int stats_timer = LLONG_MIN;
136 /* Stores the time after which rate limited statistics may be written to the
137 * database. Only updated when changes to the database require rate limiting.
139 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
140 static long long int db_limiter = LLONG_MIN;
142 static void add_del_bridges(const struct ovsrec_open_vswitch *);
143 static void bridge_del_ofprotos(void);
144 static bool bridge_add_ofprotos(struct bridge *);
145 static void bridge_create(const struct ovsrec_bridge *);
146 static void bridge_destroy(struct bridge *);
147 static struct bridge *bridge_lookup(const char *name);
148 static unixctl_cb_func bridge_unixctl_dump_flows;
149 static unixctl_cb_func bridge_unixctl_reconnect;
150 static size_t bridge_get_controllers(const struct bridge *br,
151 struct ovsrec_controller ***controllersp);
152 static void bridge_add_del_ports(struct bridge *,
153 const unsigned long int *splinter_vlans);
154 static void bridge_add_ofproto_ports(struct bridge *);
155 static void bridge_del_ofproto_ports(struct bridge *);
156 static void bridge_refresh_ofp_port(struct bridge *);
157 static void bridge_configure_datapath_id(struct bridge *);
158 static void bridge_configure_flow_eviction_threshold(struct bridge *);
159 static void bridge_configure_netflow(struct bridge *);
160 static void bridge_configure_forward_bpdu(struct bridge *);
161 static void bridge_configure_mac_idle_time(struct bridge *);
162 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
163 static void bridge_configure_stp(struct bridge *);
164 static void bridge_configure_tables(struct bridge *);
165 static void bridge_configure_remotes(struct bridge *,
166 const struct sockaddr_in *managers,
168 static void bridge_pick_local_hw_addr(struct bridge *,
169 uint8_t ea[ETH_ADDR_LEN],
170 struct iface **hw_addr_iface);
171 static uint64_t bridge_pick_datapath_id(struct bridge *,
172 const uint8_t bridge_ea[ETH_ADDR_LEN],
173 struct iface *hw_addr_iface);
174 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
176 static const char *get_port_other_config(const struct ovsrec_port *,
178 const char *default_value);
179 static uint64_t dpid_from_hash(const void *, size_t nbytes);
180 static bool bridge_has_bond_fake_iface(const struct bridge *,
182 static bool port_is_bond_fake_iface(const struct port *);
184 static unixctl_cb_func qos_unixctl_show;
186 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
187 static void port_add_ifaces(struct port *);
188 static void port_del_ifaces(struct port *);
189 static void port_destroy(struct port *);
190 static struct port *port_lookup(const struct bridge *, const char *name);
191 static void port_configure(struct port *);
192 static struct lacp_settings *port_configure_lacp(struct port *,
193 struct lacp_settings *);
194 static void port_configure_bond(struct port *, struct bond_settings *,
195 uint32_t *bond_stable_ids);
196 static bool port_is_synthetic(const struct port *);
198 static void bridge_configure_mirrors(struct bridge *);
199 static struct mirror *mirror_create(struct bridge *,
200 const struct ovsrec_mirror *);
201 static void mirror_destroy(struct mirror *);
202 static bool mirror_configure(struct mirror *);
203 static void mirror_refresh_stats(struct mirror *);
205 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
206 static struct iface *iface_create(struct port *port,
207 const struct ovsrec_interface *if_cfg);
208 static void iface_destroy(struct iface *);
209 static struct iface *iface_lookup(const struct bridge *, const char *name);
210 static struct iface *iface_find(const char *name);
211 static struct iface *iface_from_ofp_port(const struct bridge *,
213 static void iface_set_mac(struct iface *);
214 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
215 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
216 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
217 static void iface_configure_cfm(struct iface *);
218 static void iface_refresh_cfm_stats(struct iface *);
219 static void iface_refresh_stats(struct iface *);
220 static void iface_refresh_status(struct iface *);
221 static bool iface_is_synthetic(const struct iface *);
222 static const char *get_interface_other_config(const struct ovsrec_interface *,
224 const char *default_value);
226 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
228 static void shash_to_ovs_idl_map(struct shash *,
229 char ***keys, char ***values, size_t *n);
231 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
233 * This is deprecated. It is only for compatibility with broken device drivers
234 * in old versions of Linux that do not properly support VLANs when VLAN
235 * devices are not used. When broken device drivers are no longer in
236 * widespread use, we will delete these interfaces. */
238 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
239 static bool vlan_splinters_enabled_anywhere;
241 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
242 static unsigned long int *collect_splinter_vlans(
243 const struct ovsrec_open_vswitch *);
244 static void configure_splinter_port(struct port *);
245 static void add_vlan_splinter_ports(struct bridge *,
246 const unsigned long int *splinter_vlans,
247 struct shash *ports);
249 /* Public functions. */
251 /* Initializes the bridge module, configuring it to obtain its configuration
252 * from an OVSDB server accessed over 'remote', which should be a string in a
253 * form acceptable to ovsdb_idl_create(). */
255 bridge_init(const char *remote)
257 /* Create connection to database. */
258 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
259 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
261 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
262 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
263 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
264 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
265 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
266 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
267 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
269 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
270 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
271 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
273 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
274 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
275 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
276 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
283 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
284 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
285 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
286 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
287 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
288 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
289 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
290 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
291 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
293 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
294 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
295 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
296 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
298 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
300 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
302 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
303 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
305 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
307 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
309 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
310 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
311 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
312 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
313 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
315 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
317 /* Register unixctl commands. */
318 unixctl_command_register("qos/show", "interface", 1, 1,
319 qos_unixctl_show, NULL);
320 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
321 bridge_unixctl_dump_flows, NULL);
322 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
323 bridge_unixctl_reconnect, NULL);
333 struct bridge *br, *next_br;
335 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
338 ovsdb_idl_destroy(idl);
341 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
342 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
343 * responsible for freeing '*managersp' (with free()).
345 * You may be asking yourself "why does ovs-vswitchd care?", because
346 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
347 * should not be and in fact is not directly involved in that. But
348 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
349 * it has to tell in-band control where the managers are to enable that.
350 * (Thus, only managers connected in-band are collected.)
353 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
354 struct sockaddr_in **managersp, size_t *n_managersp)
356 struct sockaddr_in *managers = NULL;
357 size_t n_managers = 0;
361 /* Collect all of the potential targets from the "targets" columns of the
362 * rows pointed to by "manager_options", excluding any that are
365 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
366 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
368 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
369 sset_find_and_delete(&targets, m->target);
371 sset_add(&targets, m->target);
375 /* Now extract the targets' IP addresses. */
376 if (!sset_is_empty(&targets)) {
379 managers = xmalloc(sset_count(&targets) * sizeof *managers);
380 SSET_FOR_EACH (target, &targets) {
381 struct sockaddr_in *sin = &managers[n_managers];
383 if (stream_parse_target_with_default_ports(target,
391 sset_destroy(&targets);
393 *managersp = managers;
394 *n_managersp = n_managers;
398 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
400 unsigned long int *splinter_vlans;
401 struct sockaddr_in *managers;
402 struct bridge *br, *next;
403 int sflow_bridge_number;
406 COVERAGE_INC(bridge_reconfigure);
408 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
409 * iface"s according to 'ovs_cfg', with only very minimal configuration
412 * This is mostly an update to bridge data structures. Very little is
413 * pushed down to ofproto or lower layers. */
414 add_del_bridges(ovs_cfg);
415 splinter_vlans = collect_splinter_vlans(ovs_cfg);
416 HMAP_FOR_EACH (br, node, &all_bridges) {
417 bridge_add_del_ports(br, splinter_vlans);
419 free(splinter_vlans);
421 /* Delete all datapaths and datapath ports that are no longer configured.
423 * The kernel will reject any attempt to add a given port to a datapath if
424 * that port already belongs to a different datapath, so we must do all
425 * port deletions before any port additions. A datapath always has a
426 * "local port" so we must delete not-configured datapaths too. */
427 bridge_del_ofprotos();
428 HMAP_FOR_EACH (br, node, &all_bridges) {
430 bridge_del_ofproto_ports(br);
434 /* Create datapaths and datapath ports that are missing.
436 * After this is done, we have our final set of bridges, ports, and
437 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
438 * has at least one iface, every "struct iface" has a valid ofp_port and
440 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
442 if (bridge_add_ofprotos(br)) {
443 bridge_del_ofproto_ports(br);
449 HMAP_FOR_EACH (br, node, &all_bridges) {
450 bridge_refresh_ofp_port(br);
451 bridge_add_ofproto_ports(br);
454 /* Complete the configuration. */
455 sflow_bridge_number = 0;
456 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
457 HMAP_FOR_EACH (br, node, &all_bridges) {
460 /* We need the datapath ID early to allow LACP ports to use it as the
461 * default system ID. */
462 bridge_configure_datapath_id(br);
464 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
467 port_configure(port);
469 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
470 iface_configure_cfm(iface);
471 iface_configure_qos(iface, port->cfg->qos);
472 iface_set_mac(iface);
475 bridge_configure_mirrors(br);
476 bridge_configure_flow_eviction_threshold(br);
477 bridge_configure_forward_bpdu(br);
478 bridge_configure_mac_idle_time(br);
479 bridge_configure_remotes(br, managers, n_managers);
480 bridge_configure_netflow(br);
481 bridge_configure_sflow(br, &sflow_bridge_number);
482 bridge_configure_stp(br);
483 bridge_configure_tables(br);
487 /* ovs-vswitchd has completed initialization, so allow the process that
488 * forked us to exit successfully. */
489 daemonize_complete();
492 /* Iterate over all ofprotos and delete any of them that do not have a
493 * configured bridge or that are the wrong type. */
495 bridge_del_ofprotos(void)
503 ofproto_enumerate_types(&types);
504 SSET_FOR_EACH (type, &types) {
507 ofproto_enumerate_names(type, &names);
508 SSET_FOR_EACH (name, &names) {
509 struct bridge *br = bridge_lookup(name);
510 if (!br || strcmp(type, br->type)) {
511 ofproto_delete(name, type);
515 sset_destroy(&names);
516 sset_destroy(&types);
520 bridge_add_ofprotos(struct bridge *br)
522 int error = ofproto_create(br->name, br->type, &br->ofproto);
524 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
531 port_configure(struct port *port)
533 const struct ovsrec_port *cfg = port->cfg;
534 struct bond_settings bond_settings;
535 struct lacp_settings lacp_settings;
536 struct ofproto_bundle_settings s;
539 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
540 configure_splinter_port(port);
549 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
550 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
551 s.slaves[s.n_slaves++] = iface->ofp_port;
556 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
560 /* Get VLAN trunks. */
563 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
567 if (cfg->vlan_mode) {
568 if (!strcmp(cfg->vlan_mode, "access")) {
569 s.vlan_mode = PORT_VLAN_ACCESS;
570 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
571 s.vlan_mode = PORT_VLAN_TRUNK;
572 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
573 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
574 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
575 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
577 /* This "can't happen" because ovsdb-server should prevent it. */
578 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
579 s.vlan_mode = PORT_VLAN_TRUNK;
583 s.vlan_mode = PORT_VLAN_ACCESS;
585 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
589 s.vlan_mode = PORT_VLAN_TRUNK;
592 s.use_priority_tags = !strcmp("true", get_port_other_config(
593 cfg, "priority-tags", ""));
595 /* Get LACP settings. */
596 s.lacp = port_configure_lacp(port, &lacp_settings);
600 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
601 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
602 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
605 s.lacp_slaves = NULL;
608 /* Get bond settings. */
609 if (s.n_slaves > 1) {
610 s.bond = &bond_settings;
611 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
612 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
615 s.bond_stable_ids = NULL;
617 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
618 netdev_set_miimon_interval(iface->netdev, 0);
623 ofproto_bundle_register(port->bridge->ofproto, port, &s);
629 free(s.bond_stable_ids);
632 /* Pick local port hardware address and datapath ID for 'br'. */
634 bridge_configure_datapath_id(struct bridge *br)
636 uint8_t ea[ETH_ADDR_LEN];
638 struct iface *local_iface;
639 struct iface *hw_addr_iface;
642 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
643 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
645 int error = netdev_set_etheraddr(local_iface->netdev, ea);
647 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
648 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
649 "Ethernet address: %s",
650 br->name, strerror(error));
653 memcpy(br->ea, ea, ETH_ADDR_LEN);
655 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
656 ofproto_set_datapath_id(br->ofproto, dpid);
658 dpid_string = xasprintf("%016"PRIx64, dpid);
659 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
663 /* Set NetFlow configuration on 'br'. */
665 bridge_configure_netflow(struct bridge *br)
667 struct ovsrec_netflow *cfg = br->cfg->netflow;
668 struct netflow_options opts;
671 ofproto_set_netflow(br->ofproto, NULL);
675 memset(&opts, 0, sizeof opts);
677 /* Get default NetFlow configuration from datapath.
678 * Apply overrides from 'cfg'. */
679 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
680 if (cfg->engine_type) {
681 opts.engine_type = *cfg->engine_type;
683 if (cfg->engine_id) {
684 opts.engine_id = *cfg->engine_id;
687 /* Configure active timeout interval. */
688 opts.active_timeout = cfg->active_timeout;
689 if (!opts.active_timeout) {
690 opts.active_timeout = -1;
691 } else if (opts.active_timeout < 0) {
692 VLOG_WARN("bridge %s: active timeout interval set to negative "
693 "value, using default instead (%d seconds)", br->name,
694 NF_ACTIVE_TIMEOUT_DEFAULT);
695 opts.active_timeout = -1;
698 /* Add engine ID to interface number to disambiguate bridgs? */
699 opts.add_id_to_iface = cfg->add_id_to_interface;
700 if (opts.add_id_to_iface) {
701 if (opts.engine_id > 0x7f) {
702 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
703 "another vswitch, choose an engine id less than 128",
706 if (hmap_count(&br->ports) > 508) {
707 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
708 "another port when more than 508 ports are used",
714 sset_init(&opts.collectors);
715 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
718 if (ofproto_set_netflow(br->ofproto, &opts)) {
719 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
721 sset_destroy(&opts.collectors);
724 /* Set sFlow configuration on 'br'. */
726 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
728 const struct ovsrec_sflow *cfg = br->cfg->sflow;
729 struct ovsrec_controller **controllers;
730 struct ofproto_sflow_options oso;
731 size_t n_controllers;
735 ofproto_set_sflow(br->ofproto, NULL);
739 memset(&oso, 0, sizeof oso);
741 sset_init(&oso.targets);
742 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
744 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
746 oso.sampling_rate = *cfg->sampling;
749 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
751 oso.polling_interval = *cfg->polling;
754 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
756 oso.header_len = *cfg->header;
759 oso.sub_id = (*sflow_bridge_number)++;
760 oso.agent_device = cfg->agent;
762 oso.control_ip = NULL;
763 n_controllers = bridge_get_controllers(br, &controllers);
764 for (i = 0; i < n_controllers; i++) {
765 if (controllers[i]->local_ip) {
766 oso.control_ip = controllers[i]->local_ip;
770 ofproto_set_sflow(br->ofproto, &oso);
772 sset_destroy(&oso.targets);
776 port_configure_stp(const struct ofproto *ofproto, struct port *port,
777 struct ofproto_port_stp_settings *port_s,
778 int *port_num_counter, unsigned long *port_num_bitmap)
780 const char *config_str;
783 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
784 if (config_str && !strcmp(config_str, "false")) {
785 port_s->enable = false;
788 port_s->enable = true;
791 /* STP over bonds is not supported. */
792 if (!list_is_singleton(&port->ifaces)) {
793 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
795 port_s->enable = false;
799 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
801 /* Internal ports shouldn't participate in spanning tree, so
803 if (!strcmp(iface->type, "internal")) {
804 VLOG_DBG("port %s: disable STP on internal ports", port->name);
805 port_s->enable = false;
809 /* STP on mirror output ports is not supported. */
810 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
811 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
812 port_s->enable = false;
816 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
818 unsigned long int port_num = strtoul(config_str, NULL, 0);
819 int port_idx = port_num - 1;
821 if (port_num < 1 || port_num > STP_MAX_PORTS) {
822 VLOG_ERR("port %s: invalid stp-port-num", port->name);
823 port_s->enable = false;
827 if (bitmap_is_set(port_num_bitmap, port_idx)) {
828 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
829 port->name, port_num);
830 port_s->enable = false;
833 bitmap_set1(port_num_bitmap, port_idx);
834 port_s->port_num = port_idx;
836 if (*port_num_counter > STP_MAX_PORTS) {
837 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
838 port_s->enable = false;
842 port_s->port_num = (*port_num_counter)++;
845 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
847 port_s->path_cost = strtoul(config_str, NULL, 10);
851 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
852 /* Couldn't get speed, so assume 100Mb/s. */
853 port_s->path_cost = 19;
857 mbps = netdev_features_to_bps(current) / 1000000;
858 port_s->path_cost = stp_convert_speed_to_cost(mbps);
862 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
864 port_s->priority = strtoul(config_str, NULL, 0);
866 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
870 /* Set spanning tree configuration on 'br'. */
872 bridge_configure_stp(struct bridge *br)
874 if (!br->cfg->stp_enable) {
875 ofproto_set_stp(br->ofproto, NULL);
877 struct ofproto_stp_settings br_s;
878 const char *config_str;
880 int port_num_counter;
881 unsigned long *port_num_bitmap;
883 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
885 uint8_t ea[ETH_ADDR_LEN];
887 if (eth_addr_from_string(config_str, ea)) {
888 br_s.system_id = eth_addr_to_uint64(ea);
890 br_s.system_id = eth_addr_to_uint64(br->ea);
891 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
892 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
895 br_s.system_id = eth_addr_to_uint64(br->ea);
898 config_str = bridge_get_other_config(br->cfg, "stp-priority");
900 br_s.priority = strtoul(config_str, NULL, 0);
902 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
905 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
907 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
909 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
912 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
914 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
916 br_s.max_age = STP_DEFAULT_MAX_AGE;
919 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
921 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
923 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
926 /* Configure STP on the bridge. */
927 if (ofproto_set_stp(br->ofproto, &br_s)) {
928 VLOG_ERR("bridge %s: could not enable STP", br->name);
932 /* Users must either set the port number with the "stp-port-num"
933 * configuration on all ports or none. If manual configuration
934 * is not done, then we allocate them sequentially. */
935 port_num_counter = 0;
936 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
937 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
938 struct ofproto_port_stp_settings port_s;
941 port_configure_stp(br->ofproto, port, &port_s,
942 &port_num_counter, port_num_bitmap);
944 /* As bonds are not supported, just apply configuration to
946 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
947 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
949 VLOG_ERR("port %s: could not enable STP", port->name);
955 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
956 && port_num_counter) {
957 VLOG_ERR("bridge %s: must manually configure all STP port "
958 "IDs or none, disabling", br->name);
959 ofproto_set_stp(br->ofproto, NULL);
961 bitmap_free(port_num_bitmap);
966 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
968 const struct port *port = port_lookup(br, name);
969 return port && port_is_bond_fake_iface(port);
973 port_is_bond_fake_iface(const struct port *port)
975 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
979 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
981 struct bridge *br, *next;
985 /* Collect new bridges' names and types. */
987 for (i = 0; i < cfg->n_bridges; i++) {
988 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
989 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
991 if (strchr(br_cfg->name, '/')) {
992 /* Prevent remote ovsdb-server users from accessing arbitrary
993 * directories, e.g. consider a bridge named "../../../etc/". */
994 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
996 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
997 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1001 /* Get rid of deleted bridges or those whose types have changed.
1002 * Update 'cfg' of bridges that still exist. */
1003 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1004 br->cfg = shash_find_data(&new_br, br->name);
1005 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1006 br->cfg->datapath_type))) {
1011 /* Add new bridges. */
1012 for (i = 0; i < cfg->n_bridges; i++) {
1013 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1014 struct bridge *br = bridge_lookup(br_cfg->name);
1016 bridge_create(br_cfg);
1020 shash_destroy(&new_br);
1023 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1026 * The kernel will reject any attempt to add a given port to a datapath if that
1027 * port already belongs to a different datapath, so we must do all port
1028 * deletions before any port additions. */
1030 bridge_del_ofproto_ports(struct bridge *br)
1032 struct ofproto_port_dump dump;
1033 struct ofproto_port ofproto_port;
1035 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1036 const char *name = ofproto_port.name;
1037 struct iface *iface;
1041 /* Ignore the local port. We can't change it anyhow. */
1042 if (!strcmp(name, br->name)) {
1046 /* Get the type that 'ofproto_port' should have (ordinarily the
1047 * type of its corresponding iface) or NULL if it should be
1049 iface = iface_lookup(br, name);
1050 type = (iface ? iface->type
1051 : bridge_has_bond_fake_iface(br, name) ? "internal"
1054 /* If it's the wrong type then delete the ofproto port. */
1056 && !strcmp(ofproto_port.type, type)
1057 && (!iface || !iface->netdev
1058 || !strcmp(netdev_get_type(iface->netdev), type))) {
1061 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1063 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1064 br->name, name, strerror(error));
1067 netdev_close(iface->netdev);
1068 iface->netdev = NULL;
1074 iface_set_ofp_port(struct iface *iface, int ofp_port)
1076 struct bridge *br = iface->port->bridge;
1078 assert(iface->ofp_port < 0 && ofp_port >= 0);
1079 iface->ofp_port = ofp_port;
1080 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1081 iface_set_ofport(iface->cfg, ofp_port);
1085 bridge_refresh_ofp_port(struct bridge *br)
1087 struct ofproto_port_dump dump;
1088 struct ofproto_port ofproto_port;
1091 /* Clear all the "ofp_port"es. */
1092 hmap_clear(&br->ifaces);
1093 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1094 struct iface *iface;
1096 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1097 iface->ofp_port = -1;
1101 /* Obtain the correct "ofp_port"s from ofproto. */
1102 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1103 struct iface *iface = iface_lookup(br, ofproto_port.name);
1105 if (iface->ofp_port >= 0) {
1106 VLOG_WARN("bridge %s: interface %s reported twice",
1107 br->name, ofproto_port.name);
1108 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1109 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1110 br->name, ofproto_port.ofp_port);
1112 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1118 /* Add an ofproto port for any "struct iface" that doesn't have one.
1119 * Delete any "struct iface" for which this fails.
1120 * Delete any "struct port" that thereby ends up with no ifaces. */
1122 bridge_add_ofproto_ports(struct bridge *br)
1124 struct port *port, *next_port;
1126 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1127 struct iface *iface, *next_iface;
1128 struct ofproto_port ofproto_port;
1130 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1133 /* Open the netdev. */
1134 if (!iface->netdev) {
1135 error = netdev_open(iface->name, iface->type, &iface->netdev);
1137 VLOG_WARN("could not open network device %s (%s)",
1138 iface->name, strerror(error));
1142 && port->cfg->vlan_mode
1143 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1144 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1150 /* Configure the netdev. */
1151 if (iface->netdev) {
1155 shash_from_ovs_idl_map(iface->cfg->key_options,
1156 iface->cfg->value_options,
1157 iface->cfg->n_options, &args);
1158 error = netdev_set_config(iface->netdev, &args);
1159 shash_destroy(&args);
1162 VLOG_WARN("could not configure network device %s (%s)",
1163 iface->name, strerror(error));
1164 netdev_close(iface->netdev);
1165 iface->netdev = NULL;
1169 /* Add the port, if necessary. */
1170 if (iface->netdev && iface->ofp_port < 0) {
1174 error = ofproto_port_add(br->ofproto, iface->netdev,
1177 iface_set_ofp_port(iface, ofp_port);
1179 netdev_close(iface->netdev);
1180 iface->netdev = NULL;
1184 /* Populate stats columns in new Interface rows. */
1185 if (iface->netdev && !iface->cfg->mtu) {
1186 iface_refresh_stats(iface);
1187 iface_refresh_status(iface);
1190 /* Delete the iface if we failed. */
1191 if (iface->netdev && iface->ofp_port >= 0) {
1192 VLOG_DBG("bridge %s: interface %s is on port %d",
1193 br->name, iface->name, iface->ofp_port);
1195 if (iface->netdev) {
1196 VLOG_ERR("bridge %s: missing %s interface, dropping",
1197 br->name, iface->name);
1199 /* We already reported a related error, don't bother
1200 * duplicating it. */
1202 iface_clear_db_record(iface->cfg);
1203 iface_destroy(iface);
1206 if (list_is_empty(&port->ifaces)) {
1207 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1212 /* Add bond fake iface if necessary. */
1213 if (port_is_bond_fake_iface(port)) {
1214 if (ofproto_port_query_by_name(br->ofproto, port->name,
1216 struct netdev *netdev;
1219 error = netdev_open(port->name, "internal", &netdev);
1221 ofproto_port_add(br->ofproto, netdev, NULL);
1222 netdev_close(netdev);
1224 VLOG_WARN("could not open network device %s (%s)",
1225 port->name, strerror(error));
1228 /* Already exists, nothing to do. */
1229 ofproto_port_destroy(&ofproto_port);
1236 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1240 for (i = 0; i < n; i++) {
1241 if (!strcmp(keys[i], key)) {
1249 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1251 return get_ovsrec_key_value(br_cfg->key_other_config,
1252 br_cfg->value_other_config,
1253 br_cfg->n_other_config, key);
1256 /* Set Flow eviction threshold */
1258 bridge_configure_flow_eviction_threshold(struct bridge *br)
1260 const char *threshold_str;
1263 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1264 if (threshold_str) {
1265 threshold = strtoul(threshold_str, NULL, 10);
1267 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1269 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1272 /* Set forward BPDU option. */
1274 bridge_configure_forward_bpdu(struct bridge *br)
1276 const char *forward_bpdu_str;
1277 bool forward_bpdu = false;
1279 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1280 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1281 forward_bpdu = true;
1283 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1286 /* Set MAC aging time for 'br'. */
1288 bridge_configure_mac_idle_time(struct bridge *br)
1290 const char *idle_time_str;
1293 idle_time_str = bridge_get_other_config(br->cfg, "mac-aging-time");
1294 idle_time = (idle_time_str && atoi(idle_time_str)
1295 ? atoi(idle_time_str)
1296 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1297 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1301 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1302 struct iface **hw_addr_iface)
1304 struct hmapx mirror_output_ports;
1307 bool found_addr = false;
1311 *hw_addr_iface = NULL;
1313 /* Did the user request a particular MAC? */
1314 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1315 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1316 if (eth_addr_is_multicast(ea)) {
1317 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1318 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1319 } else if (eth_addr_is_zero(ea)) {
1320 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1326 /* Mirror output ports don't participate in picking the local hardware
1327 * address. ofproto can't help us find out whether a given port is a
1328 * mirror output because we haven't configured mirrors yet, so we need to
1329 * accumulate them ourselves. */
1330 hmapx_init(&mirror_output_ports);
1331 for (i = 0; i < br->cfg->n_mirrors; i++) {
1332 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1333 if (m->output_port) {
1334 hmapx_add(&mirror_output_ports, m->output_port);
1338 /* Otherwise choose the minimum non-local MAC address among all of the
1340 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1341 uint8_t iface_ea[ETH_ADDR_LEN];
1342 struct iface *candidate;
1343 struct iface *iface;
1345 /* Mirror output ports don't participate. */
1346 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1350 /* Choose the MAC address to represent the port. */
1352 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1353 /* Find the interface with this Ethernet address (if any) so that
1354 * we can provide the correct devname to the caller. */
1355 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1356 uint8_t candidate_ea[ETH_ADDR_LEN];
1357 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1358 && eth_addr_equals(iface_ea, candidate_ea)) {
1363 /* Choose the interface whose MAC address will represent the port.
1364 * The Linux kernel bonding code always chooses the MAC address of
1365 * the first slave added to a bond, and the Fedora networking
1366 * scripts always add slaves to a bond in alphabetical order, so
1367 * for compatibility we choose the interface with the name that is
1368 * first in alphabetical order. */
1369 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1370 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1375 /* The local port doesn't count (since we're trying to choose its
1376 * MAC address anyway). */
1377 if (iface->ofp_port == OFPP_LOCAL) {
1382 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1388 /* Compare against our current choice. */
1389 if (!eth_addr_is_multicast(iface_ea) &&
1390 !eth_addr_is_local(iface_ea) &&
1391 !eth_addr_is_reserved(iface_ea) &&
1392 !eth_addr_is_zero(iface_ea) &&
1393 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1395 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1396 *hw_addr_iface = iface;
1401 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1402 br->name, ETH_ADDR_ARGS(ea));
1404 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1405 *hw_addr_iface = NULL;
1406 VLOG_WARN("bridge %s: using default bridge Ethernet "
1407 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1410 hmapx_destroy(&mirror_output_ports);
1413 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1414 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1415 * an interface on 'br', then that interface must be passed in as
1416 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1417 * 'hw_addr_iface' must be passed in as a null pointer. */
1419 bridge_pick_datapath_id(struct bridge *br,
1420 const uint8_t bridge_ea[ETH_ADDR_LEN],
1421 struct iface *hw_addr_iface)
1424 * The procedure for choosing a bridge MAC address will, in the most
1425 * ordinary case, also choose a unique MAC that we can use as a datapath
1426 * ID. In some special cases, though, multiple bridges will end up with
1427 * the same MAC address. This is OK for the bridges, but it will confuse
1428 * the OpenFlow controller, because each datapath needs a unique datapath
1431 * Datapath IDs must be unique. It is also very desirable that they be
1432 * stable from one run to the next, so that policy set on a datapath
1435 const char *datapath_id;
1438 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1439 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1443 if (!hw_addr_iface) {
1445 * A purely internal bridge, that is, one that has no non-virtual
1446 * network devices on it at all, is difficult because it has no
1447 * natural unique identifier at all.
1449 * When the host is a XenServer, we handle this case by hashing the
1450 * host's UUID with the name of the bridge. Names of bridges are
1451 * persistent across XenServer reboots, although they can be reused if
1452 * an internal network is destroyed and then a new one is later
1453 * created, so this is fairly effective.
1455 * When the host is not a XenServer, we punt by using a random MAC
1456 * address on each run.
1458 const char *host_uuid = xenserver_get_host_uuid();
1460 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1461 dpid = dpid_from_hash(combined, strlen(combined));
1467 return eth_addr_to_uint64(bridge_ea);
1471 dpid_from_hash(const void *data, size_t n)
1473 uint8_t hash[SHA1_DIGEST_SIZE];
1475 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1476 sha1_bytes(data, n, hash);
1477 eth_addr_mark_random(hash);
1478 return eth_addr_to_uint64(hash);
1482 iface_refresh_status(struct iface *iface)
1486 enum netdev_flags flags;
1493 if (iface_is_synthetic(iface)) {
1499 if (!netdev_get_status(iface->netdev, &sh)) {
1501 char **keys, **values;
1503 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1504 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1509 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1512 shash_destroy_free_data(&sh);
1514 error = netdev_get_flags(iface->netdev, &flags);
1516 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1519 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1522 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1524 ovsrec_interface_set_duplex(iface->cfg,
1525 netdev_features_is_full_duplex(current)
1527 /* warning: uint64_t -> int64_t conversion */
1528 bps = netdev_features_to_bps(current);
1529 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1532 ovsrec_interface_set_duplex(iface->cfg, NULL);
1533 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1536 error = netdev_get_mtu(iface->netdev, &mtu);
1539 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1542 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1546 /* Writes 'iface''s CFM statistics to the database. */
1548 iface_refresh_cfm_stats(struct iface *iface)
1550 const struct ovsrec_interface *cfg = iface->cfg;
1552 const uint64_t *rmps;
1555 if (iface_is_synthetic(iface)) {
1559 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1562 const char *reasons[CFM_FAULT_N_REASONS];
1563 bool fault_bool = fault;
1567 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1568 int reason = 1 << i;
1569 if (fault & reason) {
1570 reasons[j++] = cfm_fault_reason_to_str(reason);
1574 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1575 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1577 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1578 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1581 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1582 iface->ofp_port, &rmps, &n_rmps);
1584 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1587 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1592 iface_refresh_stats(struct iface *iface)
1594 #define IFACE_STATS \
1595 IFACE_STAT(rx_packets, "rx_packets") \
1596 IFACE_STAT(tx_packets, "tx_packets") \
1597 IFACE_STAT(rx_bytes, "rx_bytes") \
1598 IFACE_STAT(tx_bytes, "tx_bytes") \
1599 IFACE_STAT(rx_dropped, "rx_dropped") \
1600 IFACE_STAT(tx_dropped, "tx_dropped") \
1601 IFACE_STAT(rx_errors, "rx_errors") \
1602 IFACE_STAT(tx_errors, "tx_errors") \
1603 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1604 IFACE_STAT(rx_over_errors, "rx_over_err") \
1605 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1606 IFACE_STAT(collisions, "collisions")
1608 #define IFACE_STAT(MEMBER, NAME) NAME,
1609 static char *keys[] = { IFACE_STATS };
1611 int64_t values[ARRAY_SIZE(keys)];
1614 struct netdev_stats stats;
1616 if (iface_is_synthetic(iface)) {
1620 /* Intentionally ignore return value, since errors will set 'stats' to
1621 * all-1s, and we will deal with that correctly below. */
1622 netdev_get_stats(iface->netdev, &stats);
1624 /* Copy statistics into values[] array. */
1626 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1629 assert(i == ARRAY_SIZE(keys));
1631 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1636 br_refresh_stp_status(struct bridge *br)
1638 struct ofproto *ofproto = br->ofproto;
1639 struct ofproto_stp_status status;
1640 char *keys[3], *values[3];
1643 if (ofproto_get_stp_status(ofproto, &status)) {
1647 if (!status.enabled) {
1648 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1652 keys[0] = "stp_bridge_id",
1653 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1654 keys[1] = "stp_designated_root",
1655 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1656 keys[2] = "stp_root_path_cost",
1657 values[2] = xasprintf("%d", status.root_path_cost);
1659 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1661 for (i = 0; i < ARRAY_SIZE(values); i++) {
1667 port_refresh_stp_status(struct port *port)
1669 struct ofproto *ofproto = port->bridge->ofproto;
1670 struct iface *iface;
1671 struct ofproto_port_stp_status status;
1673 char *str_values[4];
1674 int64_t int_values[3];
1677 if (port_is_synthetic(port)) {
1681 /* STP doesn't currently support bonds. */
1682 if (!list_is_singleton(&port->ifaces)) {
1683 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1687 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1689 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1693 if (!status.enabled) {
1694 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1695 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1699 /* Set Status column. */
1700 keys[0] = "stp_port_id";
1701 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1702 keys[1] = "stp_state";
1703 str_values[1] = xstrdup(stp_state_name(status.state));
1704 keys[2] = "stp_sec_in_state";
1705 str_values[2] = xasprintf("%u", status.sec_in_state);
1706 keys[3] = "stp_role";
1707 str_values[3] = xstrdup(stp_role_name(status.role));
1709 ovsrec_port_set_status(port->cfg, keys, str_values,
1710 ARRAY_SIZE(str_values));
1712 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1713 free(str_values[i]);
1716 /* Set Statistics column. */
1717 keys[0] = "stp_tx_count";
1718 int_values[0] = status.tx_count;
1719 keys[1] = "stp_rx_count";
1720 int_values[1] = status.rx_count;
1721 keys[2] = "stp_error_count";
1722 int_values[2] = status.error_count;
1724 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1725 ARRAY_SIZE(int_values));
1729 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1733 /* Use other-config:enable-system-stats by preference. */
1734 enable = get_ovsrec_key_value(cfg->key_other_config,
1735 cfg->value_other_config,
1736 cfg->n_other_config,
1737 "enable-statistics");
1739 return !strcmp(enable, "true");
1742 /* Disable by default. */
1747 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1749 struct ovsdb_datum datum;
1753 if (enable_system_stats(cfg)) {
1754 get_system_stats(&stats);
1757 ovsdb_datum_from_shash(&datum, &stats);
1758 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1762 static inline const char *
1763 nx_role_to_str(enum nx_role role)
1768 case NX_ROLE_MASTER:
1773 return "*** INVALID ROLE ***";
1778 refresh_controller_status(void)
1782 const struct ovsrec_controller *cfg;
1786 /* Accumulate status for controllers on all bridges. */
1787 HMAP_FOR_EACH (br, node, &all_bridges) {
1788 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1791 /* Update each controller in the database with current status. */
1792 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1793 struct ofproto_controller_info *cinfo =
1794 shash_find_data(&info, cfg->target);
1797 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1798 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1799 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1800 (char **) cinfo->pairs.values,
1803 ovsrec_controller_set_is_connected(cfg, false);
1804 ovsrec_controller_set_role(cfg, NULL);
1805 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1809 ofproto_free_ofproto_controller_info(&info);
1813 refresh_cfm_stats(void)
1815 static struct ovsdb_idl_txn *txn = NULL;
1820 txn = ovsdb_idl_txn_create(idl);
1822 HMAP_FOR_EACH (br, node, &all_bridges) {
1823 struct iface *iface;
1825 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1826 iface_refresh_cfm_stats(iface);
1831 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1832 ovsdb_idl_txn_destroy(txn);
1837 /* Performs periodic activity required by bridges that needs to be done with
1838 * the least possible latency.
1840 * It makes sense to call this function a couple of times per poll loop, to
1841 * provide a significant performance boost on some benchmarks with ofprotos
1842 * that use the ofproto-dpif implementation. */
1844 bridge_run_fast(void)
1848 HMAP_FOR_EACH (br, node, &all_bridges) {
1849 ofproto_run_fast(br->ofproto);
1856 const struct ovsrec_open_vswitch *cfg;
1858 bool vlan_splinters_changed;
1859 bool database_changed;
1862 /* (Re)configure if necessary. */
1863 database_changed = ovsdb_idl_run(idl);
1864 if (ovsdb_idl_is_lock_contended(idl)) {
1865 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1866 struct bridge *br, *next_br;
1868 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1869 "disabling this process until it goes away");
1871 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1875 } else if (!ovsdb_idl_has_lock(idl)) {
1878 cfg = ovsrec_open_vswitch_first(idl);
1880 /* Let each bridge do the work that it needs to do. */
1881 HMAP_FOR_EACH (br, node, &all_bridges) {
1882 ofproto_run(br->ofproto);
1885 /* Re-configure SSL. We do this on every trip through the main loop,
1886 * instead of just when the database changes, because the contents of the
1887 * key and certificate files can change without the database changing.
1889 * We do this before bridge_reconfigure() because that function might
1890 * initiate SSL connections and thus requires SSL to be configured. */
1891 if (cfg && cfg->ssl) {
1892 const struct ovsrec_ssl *ssl = cfg->ssl;
1894 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1895 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1898 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1900 vlan_splinters_changed = false;
1901 if (vlan_splinters_enabled_anywhere) {
1902 HMAP_FOR_EACH (br, node, &all_bridges) {
1903 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1904 vlan_splinters_changed = true;
1910 if (database_changed || vlan_splinters_changed) {
1912 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1914 bridge_reconfigure(cfg);
1916 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1917 ovsdb_idl_txn_commit(txn);
1918 ovsdb_idl_txn_destroy(txn); /* XXX */
1920 /* We still need to reconfigure to avoid dangling pointers to
1921 * now-destroyed ovsrec structures inside bridge data. */
1922 static const struct ovsrec_open_vswitch null_cfg;
1924 bridge_reconfigure(&null_cfg);
1928 /* Refresh system and interface stats if necessary. */
1929 if (time_msec() >= stats_timer) {
1931 struct ovsdb_idl_txn *txn;
1933 txn = ovsdb_idl_txn_create(idl);
1934 HMAP_FOR_EACH (br, node, &all_bridges) {
1938 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1939 struct iface *iface;
1941 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1942 iface_refresh_stats(iface);
1943 iface_refresh_status(iface);
1947 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1948 mirror_refresh_stats(m);
1952 refresh_system_stats(cfg);
1953 refresh_controller_status();
1954 ovsdb_idl_txn_commit(txn);
1955 ovsdb_idl_txn_destroy(txn); /* XXX */
1958 stats_timer = time_msec() + STATS_INTERVAL;
1961 if (time_msec() >= db_limiter) {
1962 struct ovsdb_idl_txn *txn;
1964 txn = ovsdb_idl_txn_create(idl);
1965 HMAP_FOR_EACH (br, node, &all_bridges) {
1966 struct iface *iface;
1969 br_refresh_stp_status(br);
1971 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1972 port_refresh_stp_status(port);
1975 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1976 const char *link_state;
1977 int64_t link_resets;
1980 if (iface_is_synthetic(iface)) {
1984 current = ofproto_port_is_lacp_current(br->ofproto,
1988 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1990 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1993 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1994 ovsrec_interface_set_link_state(iface->cfg, link_state);
1996 link_resets = netdev_get_carrier_resets(iface->netdev);
1997 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2001 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2002 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2004 ovsdb_idl_txn_destroy(txn);
2007 refresh_cfm_stats();
2013 ovsdb_idl_wait(idl);
2014 if (!hmap_is_empty(&all_bridges)) {
2017 HMAP_FOR_EACH (br, node, &all_bridges) {
2018 ofproto_wait(br->ofproto);
2020 poll_timer_wait_until(stats_timer);
2022 if (db_limiter > time_msec()) {
2023 poll_timer_wait_until(db_limiter);
2028 /* QoS unixctl user interface functions. */
2030 struct qos_unixctl_show_cbdata {
2032 struct iface *iface;
2036 qos_unixctl_show_cb(unsigned int queue_id,
2037 const struct shash *details,
2040 struct qos_unixctl_show_cbdata *data = aux;
2041 struct ds *ds = data->ds;
2042 struct iface *iface = data->iface;
2043 struct netdev_queue_stats stats;
2044 struct shash_node *node;
2047 ds_put_cstr(ds, "\n");
2049 ds_put_format(ds, "Queue %u:\n", queue_id);
2051 ds_put_cstr(ds, "Default:\n");
2054 SHASH_FOR_EACH (node, details) {
2055 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2058 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2060 if (stats.tx_packets != UINT64_MAX) {
2061 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2064 if (stats.tx_bytes != UINT64_MAX) {
2065 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2068 if (stats.tx_errors != UINT64_MAX) {
2069 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2072 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2073 queue_id, strerror(error));
2078 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2079 const char *argv[], void *aux OVS_UNUSED)
2081 struct ds ds = DS_EMPTY_INITIALIZER;
2082 struct shash sh = SHASH_INITIALIZER(&sh);
2083 struct iface *iface;
2085 struct shash_node *node;
2086 struct qos_unixctl_show_cbdata data;
2089 iface = iface_find(argv[1]);
2091 unixctl_command_reply_error(conn, "no such interface");
2095 netdev_get_qos(iface->netdev, &type, &sh);
2097 if (*type != '\0') {
2098 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2100 SHASH_FOR_EACH (node, &sh) {
2101 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2106 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2109 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2111 unixctl_command_reply(conn, ds_cstr(&ds));
2113 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2114 unixctl_command_reply_error(conn, ds_cstr(&ds));
2117 shash_destroy_free_data(&sh);
2121 /* Bridge reconfiguration functions. */
2123 bridge_create(const struct ovsrec_bridge *br_cfg)
2127 assert(!bridge_lookup(br_cfg->name));
2128 br = xzalloc(sizeof *br);
2130 br->name = xstrdup(br_cfg->name);
2131 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2134 /* Derive the default Ethernet address from the bridge's UUID. This should
2135 * be unique and it will be stable between ovs-vswitchd runs. */
2136 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2137 eth_addr_mark_random(br->default_ea);
2139 hmap_init(&br->ports);
2140 hmap_init(&br->ifaces);
2141 hmap_init(&br->iface_by_name);
2142 hmap_init(&br->mirrors);
2144 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2148 bridge_destroy(struct bridge *br)
2151 struct mirror *mirror, *next_mirror;
2152 struct port *port, *next_port;
2154 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2157 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2158 mirror_destroy(mirror);
2160 hmap_remove(&all_bridges, &br->node);
2161 ofproto_destroy(br->ofproto);
2162 hmap_destroy(&br->ifaces);
2163 hmap_destroy(&br->ports);
2164 hmap_destroy(&br->iface_by_name);
2165 hmap_destroy(&br->mirrors);
2172 static struct bridge *
2173 bridge_lookup(const char *name)
2177 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2178 if (!strcmp(br->name, name)) {
2185 /* Handle requests for a listing of all flows known by the OpenFlow
2186 * stack, including those normally hidden. */
2188 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2189 const char *argv[], void *aux OVS_UNUSED)
2194 br = bridge_lookup(argv[1]);
2196 unixctl_command_reply_error(conn, "Unknown bridge");
2201 ofproto_get_all_flows(br->ofproto, &results);
2203 unixctl_command_reply(conn, ds_cstr(&results));
2204 ds_destroy(&results);
2207 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2208 * connections and reconnect. If BRIDGE is not specified, then all bridges
2209 * drop their controller connections and reconnect. */
2211 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2212 const char *argv[], void *aux OVS_UNUSED)
2216 br = bridge_lookup(argv[1]);
2218 unixctl_command_reply_error(conn, "Unknown bridge");
2221 ofproto_reconnect_controllers(br->ofproto);
2223 HMAP_FOR_EACH (br, node, &all_bridges) {
2224 ofproto_reconnect_controllers(br->ofproto);
2227 unixctl_command_reply(conn, NULL);
2231 bridge_get_controllers(const struct bridge *br,
2232 struct ovsrec_controller ***controllersp)
2234 struct ovsrec_controller **controllers;
2235 size_t n_controllers;
2237 controllers = br->cfg->controller;
2238 n_controllers = br->cfg->n_controller;
2240 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2246 *controllersp = controllers;
2248 return n_controllers;
2251 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2252 * those configured in 'br->cfg'. */
2254 bridge_add_del_ports(struct bridge *br,
2255 const unsigned long int *splinter_vlans)
2257 struct port *port, *next;
2258 struct shash_node *node;
2259 struct shash new_ports;
2262 /* Collect new ports. */
2263 shash_init(&new_ports);
2264 for (i = 0; i < br->cfg->n_ports; i++) {
2265 const char *name = br->cfg->ports[i]->name;
2266 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2267 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2271 if (bridge_get_controllers(br, NULL)
2272 && !shash_find(&new_ports, br->name)) {
2273 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2274 br->name, br->name);
2276 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2277 br->synth_local_port.n_interfaces = 1;
2278 br->synth_local_port.name = br->name;
2280 br->synth_local_iface.name = br->name;
2281 br->synth_local_iface.type = "internal";
2283 br->synth_local_ifacep = &br->synth_local_iface;
2285 shash_add(&new_ports, br->name, &br->synth_local_port);
2288 if (splinter_vlans) {
2289 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2292 /* Get rid of deleted ports.
2293 * Get rid of deleted interfaces on ports that still exist. */
2294 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2295 port->cfg = shash_find_data(&new_ports, port->name);
2299 port_del_ifaces(port);
2303 /* Create new ports.
2304 * Add new interfaces to existing ports. */
2305 SHASH_FOR_EACH (node, &new_ports) {
2306 struct port *port = port_lookup(br, node->name);
2308 struct ovsrec_port *cfg = node->data;
2309 port = port_create(br, cfg);
2311 port_add_ifaces(port);
2312 if (list_is_empty(&port->ifaces)) {
2313 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2314 br->name, port->name);
2318 shash_destroy(&new_ports);
2321 /* Initializes 'oc' appropriately as a management service controller for
2324 * The caller must free oc->target when it is no longer needed. */
2326 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2327 struct ofproto_controller *oc)
2329 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2330 oc->max_backoff = 0;
2331 oc->probe_interval = 60;
2332 oc->band = OFPROTO_OUT_OF_BAND;
2334 oc->burst_limit = 0;
2335 oc->enable_async_msgs = true;
2338 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2340 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2341 struct ofproto_controller *oc)
2343 oc->target = c->target;
2344 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2345 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2346 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2347 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2348 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2349 oc->burst_limit = (c->controller_burst_limit
2350 ? *c->controller_burst_limit : 0);
2351 oc->enable_async_msgs = (!c->enable_async_messages
2352 || *c->enable_async_messages);
2355 /* Configures the IP stack for 'br''s local interface properly according to the
2356 * configuration in 'c'. */
2358 bridge_configure_local_iface_netdev(struct bridge *br,
2359 struct ovsrec_controller *c)
2361 struct netdev *netdev;
2362 struct in_addr mask, gateway;
2364 struct iface *local_iface;
2367 /* If there's no local interface or no IP address, give up. */
2368 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2369 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2373 /* Bring up the local interface. */
2374 netdev = local_iface->netdev;
2375 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2377 /* Configure the IP address and netmask. */
2378 if (!c->local_netmask
2379 || !inet_aton(c->local_netmask, &mask)
2381 mask.s_addr = guess_netmask(ip.s_addr);
2383 if (!netdev_set_in4(netdev, ip, mask)) {
2384 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2385 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2388 /* Configure the default gateway. */
2389 if (c->local_gateway
2390 && inet_aton(c->local_gateway, &gateway)
2391 && gateway.s_addr) {
2392 if (!netdev_add_router(netdev, gateway)) {
2393 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2394 br->name, IP_ARGS(&gateway.s_addr));
2399 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2400 * in either string are treated as equal to any number of slashes in the other,
2401 * e.g. "x///y" is equal to "x/y". */
2403 equal_pathnames(const char *a, const char *b)
2407 a += strspn(a, "/");
2408 b += strspn(b, "/");
2409 } else if (*a == '\0') {
2420 bridge_configure_remotes(struct bridge *br,
2421 const struct sockaddr_in *managers, size_t n_managers)
2423 const char *disable_ib_str, *queue_id_str;
2424 bool disable_in_band = false;
2427 struct ovsrec_controller **controllers;
2428 size_t n_controllers;
2430 enum ofproto_fail_mode fail_mode;
2432 struct ofproto_controller *ocs;
2436 /* Check if we should disable in-band control on this bridge. */
2437 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2438 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2439 disable_in_band = true;
2442 /* Set OpenFlow queue ID for in-band control. */
2443 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2444 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2445 ofproto_set_in_band_queue(br->ofproto, queue_id);
2447 if (disable_in_band) {
2448 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2450 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2453 n_controllers = bridge_get_controllers(br, &controllers);
2455 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2458 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2459 for (i = 0; i < n_controllers; i++) {
2460 struct ovsrec_controller *c = controllers[i];
2462 if (!strncmp(c->target, "punix:", 6)
2463 || !strncmp(c->target, "unix:", 5)) {
2464 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2467 whitelist = xasprintf("unix:%s/%s.controller",
2468 ovs_rundir(), br->name);
2469 if (!equal_pathnames(c->target, whitelist)) {
2470 /* Prevent remote ovsdb-server users from accessing arbitrary
2471 * Unix domain sockets and overwriting arbitrary local
2473 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2474 "controller \"%s\" due to possibility for remote "
2475 "exploit. Instead, specify whitelisted \"%s\" or "
2476 "connect to \"unix:%s/%s.mgmt\" (which is always "
2477 "available without special configuration).",
2478 br->name, c->target, whitelist,
2479 ovs_rundir(), br->name);
2487 bridge_configure_local_iface_netdev(br, c);
2488 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2489 if (disable_in_band) {
2490 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2495 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2496 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2499 /* Set the fail-mode. */
2500 fail_mode = !br->cfg->fail_mode
2501 || !strcmp(br->cfg->fail_mode, "standalone")
2502 ? OFPROTO_FAIL_STANDALONE
2503 : OFPROTO_FAIL_SECURE;
2504 ofproto_set_fail_mode(br->ofproto, fail_mode);
2506 /* Configure OpenFlow controller connection snooping. */
2507 if (!ofproto_has_snoops(br->ofproto)) {
2511 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2512 ovs_rundir(), br->name));
2513 ofproto_set_snoops(br->ofproto, &snoops);
2514 sset_destroy(&snoops);
2519 bridge_configure_tables(struct bridge *br)
2521 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2525 n_tables = ofproto_get_n_tables(br->ofproto);
2527 for (i = 0; i < n_tables; i++) {
2528 struct ofproto_table_settings s;
2531 s.max_flows = UINT_MAX;
2535 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2536 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2539 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2540 s.max_flows = *cfg->flow_limit;
2542 if (cfg->overflow_policy
2543 && !strcmp(cfg->overflow_policy, "evict")) {
2546 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2547 for (k = 0; k < cfg->n_groups; k++) {
2548 const char *string = cfg->groups[k];
2551 msg = mf_parse_subfield__(&s.groups[k], &string);
2553 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2554 "'groups' (%s)", br->name, i, msg);
2556 } else if (*string) {
2557 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2558 "element '%s' contains trailing garbage",
2559 br->name, i, cfg->groups[k]);
2567 ofproto_configure_table(br->ofproto, i, &s);
2571 for (; j < br->cfg->n_flow_tables; j++) {
2572 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2573 "%"PRId64" not supported by this datapath", br->name,
2574 br->cfg->key_flow_tables[j]);
2578 /* Port functions. */
2580 static struct port *
2581 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2585 port = xzalloc(sizeof *port);
2587 port->name = xstrdup(cfg->name);
2589 list_init(&port->ifaces);
2591 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2593 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2599 get_port_other_config(const struct ovsrec_port *port, const char *key,
2600 const char *default_value)
2604 value = get_ovsrec_key_value(port->key_other_config,
2605 port->value_other_config,
2606 port->n_other_config, key);
2607 return value ? value : default_value;
2611 get_interface_other_config(const struct ovsrec_interface *iface,
2612 const char *key, const char *default_value)
2616 value = get_ovsrec_key_value(iface->key_other_config,
2617 iface->value_other_config,
2618 iface->n_other_config, key);
2619 return value ? value : default_value;
2622 /* Deletes interfaces from 'port' that are no longer configured for it. */
2624 port_del_ifaces(struct port *port)
2626 struct iface *iface, *next;
2627 struct sset new_ifaces;
2630 /* Collect list of new interfaces. */
2631 sset_init(&new_ifaces);
2632 for (i = 0; i < port->cfg->n_interfaces; i++) {
2633 const char *name = port->cfg->interfaces[i]->name;
2634 const char *type = port->cfg->interfaces[i]->name;
2635 if (strcmp(type, "null")) {
2636 sset_add(&new_ifaces, name);
2640 /* Get rid of deleted interfaces. */
2641 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2642 if (!sset_contains(&new_ifaces, iface->name)) {
2643 iface_destroy(iface);
2647 sset_destroy(&new_ifaces);
2650 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2653 port_add_ifaces(struct port *port)
2655 struct shash new_ifaces;
2656 struct shash_node *node;
2659 /* Collect new ifaces. */
2660 shash_init(&new_ifaces);
2661 for (i = 0; i < port->cfg->n_interfaces; i++) {
2662 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2663 if (strcmp(cfg->type, "null")
2664 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2665 VLOG_WARN("port %s: %s specified twice as port interface",
2666 port->name, cfg->name);
2667 iface_clear_db_record(cfg);
2671 /* Create new interfaces.
2672 * Update interface types and 'cfg' members. */
2673 SHASH_FOR_EACH (node, &new_ifaces) {
2674 const struct ovsrec_interface *cfg = node->data;
2675 const char *iface_name = node->name;
2676 struct iface *iface;
2678 iface = iface_lookup(port->bridge, iface_name);
2680 iface = iface_create(port, cfg);
2685 /* Determine interface type. The local port always has type
2686 * "internal". Other ports take their type from the database and
2687 * default to "system" if none is specified. */
2688 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2689 : cfg->type[0] ? cfg->type
2692 shash_destroy(&new_ifaces);
2696 port_destroy(struct port *port)
2699 struct bridge *br = port->bridge;
2700 struct iface *iface, *next;
2703 ofproto_bundle_unregister(br->ofproto, port);
2706 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2707 iface_destroy(iface);
2710 hmap_remove(&br->ports, &port->hmap_node);
2712 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2719 static struct port *
2720 port_lookup(const struct bridge *br, const char *name)
2724 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2726 if (!strcmp(port->name, name)) {
2734 enable_lacp(struct port *port, bool *activep)
2736 if (!port->cfg->lacp) {
2737 /* XXX when LACP implementation has been sufficiently tested, enable by
2738 * default and make active on bonded ports. */
2740 } else if (!strcmp(port->cfg->lacp, "off")) {
2742 } else if (!strcmp(port->cfg->lacp, "active")) {
2745 } else if (!strcmp(port->cfg->lacp, "passive")) {
2749 VLOG_WARN("port %s: unknown LACP mode %s",
2750 port->name, port->cfg->lacp);
2755 static struct lacp_settings *
2756 port_configure_lacp(struct port *port, struct lacp_settings *s)
2758 const char *lacp_time, *system_id;
2759 long long int custom_time;
2762 if (!enable_lacp(port, &s->active)) {
2766 s->name = port->name;
2768 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2770 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2771 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2772 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2773 " address.", port->name, system_id);
2777 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2780 if (eth_addr_is_zero(s->id)) {
2781 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2785 /* Prefer bondable links if unspecified. */
2786 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2788 s->priority = (priority > 0 && priority <= UINT16_MAX
2790 : UINT16_MAX - !list_is_short(&port->ifaces));
2792 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2796 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2797 custom_time = atoi(lacp_time);
2798 if (!strcmp(lacp_time, "fast")) {
2799 s->lacp_time = LACP_TIME_FAST;
2800 } else if (!strcmp(lacp_time, "slow")) {
2801 s->lacp_time = LACP_TIME_SLOW;
2802 } else if (custom_time > 0) {
2803 s->lacp_time = LACP_TIME_CUSTOM;
2804 s->custom_time = custom_time;
2806 s->lacp_time = LACP_TIME_SLOW;
2813 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2815 int priority, portid, key;
2817 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2818 priority = atoi(get_interface_other_config(iface->cfg,
2819 "lacp-port-priority", "0"));
2820 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2823 if (portid <= 0 || portid > UINT16_MAX) {
2824 portid = iface->ofp_port;
2827 if (priority <= 0 || priority > UINT16_MAX) {
2828 priority = UINT16_MAX;
2831 if (key < 0 || key > UINT16_MAX) {
2835 s->name = iface->name;
2837 s->priority = priority;
2842 port_configure_bond(struct port *port, struct bond_settings *s,
2843 uint32_t *bond_stable_ids)
2845 const char *detect_s;
2846 struct iface *iface;
2847 int miimon_interval;
2850 s->name = port->name;
2852 if (port->cfg->bond_mode) {
2853 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2854 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2855 port->name, port->cfg->bond_mode,
2856 bond_mode_to_string(s->balance));
2859 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2861 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2862 * active-backup. At some point we should remove this warning. */
2863 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2864 " in previous versions, the default bond_mode was"
2865 " balance-slb", port->name,
2866 bond_mode_to_string(s->balance));
2868 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2869 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2870 "please use another bond type or disable flood_vlans",
2874 miimon_interval = atoi(get_port_other_config(port->cfg,
2875 "bond-miimon-interval", "0"));
2876 if (miimon_interval <= 0) {
2877 miimon_interval = 200;
2880 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2881 if (!strcmp(detect_s, "carrier")) {
2882 miimon_interval = 0;
2883 } else if (strcmp(detect_s, "miimon")) {
2884 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2885 "defaulting to carrier", port->name, detect_s);
2886 miimon_interval = 0;
2889 s->up_delay = MAX(0, port->cfg->bond_updelay);
2890 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2891 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2892 s->rebalance_interval = atoi(
2893 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2894 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2895 s->rebalance_interval = 1000;
2898 s->fake_iface = port->cfg->bond_fake_iface;
2901 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2902 long long stable_id;
2904 stable_id = atoll(get_interface_other_config(iface->cfg,
2905 "bond-stable-id", "0"));
2906 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2907 stable_id = iface->ofp_port;
2909 bond_stable_ids[i++] = stable_id;
2911 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2915 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2916 * instead of obtaining it from the database. */
2918 port_is_synthetic(const struct port *port)
2920 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2923 /* Interface functions. */
2925 static struct iface *
2926 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2928 struct bridge *br = port->bridge;
2929 struct iface *iface;
2930 char *name = if_cfg->name;
2932 iface = xzalloc(sizeof *iface);
2934 iface->name = xstrdup(name);
2935 iface->ofp_port = -1;
2936 iface->tag = tag_create_random();
2937 iface->netdev = NULL;
2938 iface->cfg = if_cfg;
2940 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2942 list_push_back(&port->ifaces, &iface->port_elem);
2944 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2950 iface_destroy(struct iface *iface)
2953 struct port *port = iface->port;
2954 struct bridge *br = port->bridge;
2956 if (br->ofproto && iface->ofp_port >= 0) {
2957 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2960 if (iface->ofp_port >= 0) {
2961 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2964 list_remove(&iface->port_elem);
2965 hmap_remove(&br->iface_by_name, &iface->name_node);
2967 netdev_close(iface->netdev);
2974 static struct iface *
2975 iface_lookup(const struct bridge *br, const char *name)
2977 struct iface *iface;
2979 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2980 &br->iface_by_name) {
2981 if (!strcmp(iface->name, name)) {
2989 static struct iface *
2990 iface_find(const char *name)
2992 const struct bridge *br;
2994 HMAP_FOR_EACH (br, node, &all_bridges) {
2995 struct iface *iface = iface_lookup(br, name);
3004 static struct iface *
3005 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3007 struct iface *iface;
3009 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3010 hash_int(ofp_port, 0), &br->ifaces) {
3011 if (iface->ofp_port == ofp_port) {
3018 /* Set Ethernet address of 'iface', if one is specified in the configuration
3021 iface_set_mac(struct iface *iface)
3023 uint8_t ea[ETH_ADDR_LEN];
3025 if (!strcmp(iface->type, "internal")
3026 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3027 if (iface->ofp_port == OFPP_LOCAL) {
3028 VLOG_ERR("interface %s: ignoring mac in Interface record "
3029 "(use Bridge record to set local port's mac)",
3031 } else if (eth_addr_is_multicast(ea)) {
3032 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3035 int error = netdev_set_etheraddr(iface->netdev, ea);
3037 VLOG_ERR("interface %s: setting MAC failed (%s)",
3038 iface->name, strerror(error));
3044 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3046 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3048 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3049 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3053 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3054 * sets the "ofport" field to -1.
3056 * This is appropriate when 'if_cfg''s interface cannot be created or is
3057 * otherwise invalid. */
3059 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3061 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3062 iface_set_ofport(if_cfg, -1);
3063 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3064 ovsrec_interface_set_admin_state(if_cfg, NULL);
3065 ovsrec_interface_set_duplex(if_cfg, NULL);
3066 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3067 ovsrec_interface_set_link_state(if_cfg, NULL);
3068 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3069 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3070 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3071 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3072 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3073 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3077 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3079 * The value strings in '*shash' are taken directly from values[], not copied,
3080 * so the caller should not modify or free them. */
3082 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3083 struct shash *shash)
3088 for (i = 0; i < n; i++) {
3089 shash_add(shash, keys[i], values[i]);
3093 /* Creates 'keys' and 'values' arrays from 'shash'.
3095 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3096 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3097 * are populated with with strings taken directly from 'shash' and thus have
3098 * the same ownership of the key-value pairs in shash.
3101 shash_to_ovs_idl_map(struct shash *shash,
3102 char ***keys, char ***values, size_t *n)
3106 struct shash_node *sn;
3108 count = shash_count(shash);
3110 k = xmalloc(count * sizeof *k);
3111 v = xmalloc(count * sizeof *v);
3114 SHASH_FOR_EACH(sn, shash) {
3125 struct iface_delete_queues_cbdata {
3126 struct netdev *netdev;
3127 const struct ovsdb_datum *queues;
3131 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3133 union ovsdb_atom atom;
3135 atom.integer = target;
3136 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3140 iface_delete_queues(unsigned int queue_id,
3141 const struct shash *details OVS_UNUSED, void *cbdata_)
3143 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3145 if (!queue_ids_include(cbdata->queues, queue_id)) {
3146 netdev_delete_queue(cbdata->netdev, queue_id);
3151 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3153 struct ofpbuf queues_buf;
3155 ofpbuf_init(&queues_buf, 0);
3157 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3158 netdev_set_qos(iface->netdev, NULL, NULL);
3160 struct iface_delete_queues_cbdata cbdata;
3161 struct shash details;
3165 /* Configure top-level Qos for 'iface'. */
3166 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3167 qos->n_other_config, &details);
3168 netdev_set_qos(iface->netdev, qos->type, &details);
3169 shash_destroy(&details);
3171 /* Deconfigure queues that were deleted. */
3172 cbdata.netdev = iface->netdev;
3173 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3175 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3177 /* Configure queues for 'iface'. */
3179 for (i = 0; i < qos->n_queues; i++) {
3180 const struct ovsrec_queue *queue = qos->value_queues[i];
3181 unsigned int queue_id = qos->key_queues[i];
3183 if (queue_id == 0) {
3187 if (queue->n_dscp == 1) {
3188 struct ofproto_port_queue *port_queue;
3190 port_queue = ofpbuf_put_uninit(&queues_buf,
3191 sizeof *port_queue);
3192 port_queue->queue = queue_id;
3193 port_queue->dscp = queue->dscp[0];
3196 shash_from_ovs_idl_map(queue->key_other_config,
3197 queue->value_other_config,
3198 queue->n_other_config, &details);
3199 netdev_set_queue(iface->netdev, queue_id, &details);
3200 shash_destroy(&details);
3203 shash_init(&details);
3204 netdev_set_queue(iface->netdev, 0, &details);
3205 shash_destroy(&details);
3209 if (iface->ofp_port >= 0) {
3210 const struct ofproto_port_queue *port_queues = queues_buf.data;
3211 size_t n_queues = queues_buf.size / sizeof *port_queues;
3213 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3214 port_queues, n_queues);
3217 netdev_set_policing(iface->netdev,
3218 iface->cfg->ingress_policing_rate,
3219 iface->cfg->ingress_policing_burst);
3221 ofpbuf_uninit(&queues_buf);
3225 iface_configure_cfm(struct iface *iface)
3227 const struct ovsrec_interface *cfg = iface->cfg;
3228 const char *extended_str, *opstate_str;
3229 struct cfm_settings s;
3231 if (!cfg->n_cfm_mpid) {
3232 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3236 s.mpid = *cfg->cfm_mpid;
3237 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3239 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3241 s.ccm_pcp = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_pcp",
3243 if (s.interval <= 0) {
3247 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3249 s.extended = !strcasecmp("true", extended_str);
3251 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3252 s.opup = !strcasecmp("up", opstate_str);
3254 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3257 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3258 * instead of obtaining it from the database. */
3260 iface_is_synthetic(const struct iface *iface)
3262 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3265 /* Port mirroring. */
3267 static struct mirror *
3268 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3272 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3273 if (uuid_equals(uuid, &m->uuid)) {
3281 bridge_configure_mirrors(struct bridge *br)
3283 const struct ovsdb_datum *mc;
3284 unsigned long *flood_vlans;
3285 struct mirror *m, *next;
3288 /* Get rid of deleted mirrors. */
3289 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3290 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3291 union ovsdb_atom atom;
3293 atom.uuid = m->uuid;
3294 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3299 /* Add new mirrors and reconfigure existing ones. */
3300 for (i = 0; i < br->cfg->n_mirrors; i++) {
3301 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3302 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3304 m = mirror_create(br, cfg);
3307 if (!mirror_configure(m)) {
3312 /* Update flooded vlans (for RSPAN). */
3313 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3314 br->cfg->n_flood_vlans);
3315 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3316 bitmap_free(flood_vlans);
3319 static struct mirror *
3320 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3324 m = xzalloc(sizeof *m);
3325 m->uuid = cfg->header_.uuid;
3326 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3328 m->name = xstrdup(cfg->name);
3334 mirror_destroy(struct mirror *m)
3337 struct bridge *br = m->bridge;
3340 ofproto_mirror_unregister(br->ofproto, m);
3343 hmap_remove(&br->mirrors, &m->hmap_node);
3350 mirror_collect_ports(struct mirror *m,
3351 struct ovsrec_port **in_ports, int n_in_ports,
3352 void ***out_portsp, size_t *n_out_portsp)
3354 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3355 size_t n_out_ports = 0;
3358 for (i = 0; i < n_in_ports; i++) {
3359 const char *name = in_ports[i]->name;
3360 struct port *port = port_lookup(m->bridge, name);
3362 out_ports[n_out_ports++] = port;
3364 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3365 "port %s", m->bridge->name, m->name, name);
3368 *out_portsp = out_ports;
3369 *n_out_portsp = n_out_ports;
3373 mirror_configure(struct mirror *m)
3375 const struct ovsrec_mirror *cfg = m->cfg;
3376 struct ofproto_mirror_settings s;
3379 if (strcmp(cfg->name, m->name)) {
3381 m->name = xstrdup(cfg->name);
3385 /* Get output port or VLAN. */
3386 if (cfg->output_port) {
3387 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3388 if (!s.out_bundle) {
3389 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3390 m->bridge->name, m->name);
3393 s.out_vlan = UINT16_MAX;
3395 if (cfg->output_vlan) {
3396 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3397 "output vlan; ignoring output vlan",
3398 m->bridge->name, m->name);
3400 } else if (cfg->output_vlan) {
3401 /* The database should prevent invalid VLAN values. */
3402 s.out_bundle = NULL;
3403 s.out_vlan = *cfg->output_vlan;
3405 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3406 m->bridge->name, m->name);
3410 /* Get port selection. */
3411 if (cfg->select_all) {
3412 size_t n_ports = hmap_count(&m->bridge->ports);
3413 void **ports = xmalloc(n_ports * sizeof *ports);
3418 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3428 /* Get ports, dropping ports that don't exist.
3429 * The IDL ensures that there are no duplicates. */
3430 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3431 &s.srcs, &s.n_srcs);
3432 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3433 &s.dsts, &s.n_dsts);
3436 /* Get VLAN selection. */
3437 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3440 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3443 if (s.srcs != s.dsts) {
3452 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3454 * This is deprecated. It is only for compatibility with broken device drivers
3455 * in old versions of Linux that do not properly support VLANs when VLAN
3456 * devices are not used. When broken device drivers are no longer in
3457 * widespread use, we will delete these interfaces. */
3459 static void **blocks;
3460 static size_t n_blocks, allocated_blocks;
3462 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3463 * VLAN splinters are reconfigured. */
3465 register_block(void *block)
3467 if (n_blocks >= allocated_blocks) {
3468 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3470 blocks[n_blocks++] = block;
3473 /* Frees all of the blocks registered with register_block(). */
3475 free_registered_blocks(void)
3479 for (i = 0; i < n_blocks; i++) {
3485 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3488 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3492 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3493 return !strcmp(value, "true");
3496 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3499 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3500 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3501 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3504 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3505 * use, returns NULL.
3507 * Updates 'vlan_splinters_enabled_anywhere'. */
3508 static unsigned long int *
3509 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3511 unsigned long int *splinter_vlans;
3512 struct sset splinter_ifaces;
3513 const char *real_dev_name;
3514 struct shash *real_devs;
3515 struct shash_node *node;
3519 /* Free space allocated for synthesized ports and interfaces, since we're
3520 * in the process of reconstructing all of them. */
3521 free_registered_blocks();
3523 splinter_vlans = bitmap_allocate(4096);
3524 sset_init(&splinter_ifaces);
3525 vlan_splinters_enabled_anywhere = false;
3526 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3527 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3530 for (j = 0; j < br_cfg->n_ports; j++) {
3531 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3534 for (k = 0; k < port_cfg->n_interfaces; k++) {
3535 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3537 if (vlan_splinters_is_enabled(iface_cfg)) {
3538 vlan_splinters_enabled_anywhere = true;
3539 sset_add(&splinter_ifaces, iface_cfg->name);
3540 vlan_bitmap_from_array__(port_cfg->trunks,
3546 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3547 bitmap_set1(splinter_vlans, *port_cfg->tag);
3552 if (!vlan_splinters_enabled_anywhere) {
3553 free(splinter_vlans);
3554 sset_destroy(&splinter_ifaces);
3558 HMAP_FOR_EACH (br, node, &all_bridges) {
3560 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3564 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3565 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3566 * device to be created for it. */
3567 bitmap_set0(splinter_vlans, 0);
3568 bitmap_set0(splinter_vlans, 4095);
3570 /* Delete all VLAN devices that we don't need. */
3572 real_devs = vlandev_get_real_devs();
3573 SHASH_FOR_EACH (node, real_devs) {
3574 const struct vlan_real_dev *real_dev = node->data;
3575 const struct vlan_dev *vlan_dev;
3576 bool real_dev_has_splinters;
3578 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3580 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3581 if (!real_dev_has_splinters
3582 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3583 struct netdev *netdev;
3585 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3586 if (!netdev_get_in4(netdev, NULL, NULL) ||
3587 !netdev_get_in6(netdev, NULL)) {
3588 vlandev_del(vlan_dev->name);
3590 /* It has an IP address configured, so we don't own
3591 * it. Don't delete it. */
3593 netdev_close(netdev);
3600 /* Add all VLAN devices that we need. */
3601 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3604 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3605 if (!vlandev_get_name(real_dev_name, vid)) {
3606 vlandev_add(real_dev_name, vid);
3613 sset_destroy(&splinter_ifaces);
3615 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3616 free(splinter_vlans);
3619 return splinter_vlans;
3622 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3625 configure_splinter_port(struct port *port)
3627 struct ofproto *ofproto = port->bridge->ofproto;
3628 uint16_t realdev_ofp_port;
3629 const char *realdev_name;
3630 struct iface *vlandev, *realdev;
3632 ofproto_bundle_unregister(port->bridge->ofproto, port);
3634 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3637 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3638 realdev = iface_lookup(port->bridge, realdev_name);
3639 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3641 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3645 static struct ovsrec_port *
3646 synthesize_splinter_port(const char *real_dev_name,
3647 const char *vlan_dev_name, int vid)
3649 struct ovsrec_interface *iface;
3650 struct ovsrec_port *port;
3652 iface = xzalloc(sizeof *iface);
3653 iface->name = xstrdup(vlan_dev_name);
3654 iface->type = "system";
3656 port = xzalloc(sizeof *port);
3657 port->interfaces = xmemdup(&iface, sizeof iface);
3658 port->n_interfaces = 1;
3659 port->name = xstrdup(vlan_dev_name);
3660 port->vlan_mode = "splinter";
3661 port->tag = xmalloc(sizeof *port->tag);
3663 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3664 port->key_other_config[0] = "realdev";
3665 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3666 port->value_other_config[0] = xstrdup(real_dev_name);
3667 port->n_other_config = 1;
3669 register_block(iface);
3670 register_block(iface->name);
3671 register_block(port);
3672 register_block(port->interfaces);
3673 register_block(port->name);
3674 register_block(port->tag);
3675 register_block(port->key_other_config);
3676 register_block(port->value_other_config);
3677 register_block(port->value_other_config[0]);
3682 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3683 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3684 * 1-bit in the 'splinter_vlans' bitmap. */
3686 add_vlan_splinter_ports(struct bridge *br,
3687 const unsigned long int *splinter_vlans,
3688 struct shash *ports)
3692 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3693 * we're modifying 'ports'. */
3694 for (i = 0; i < br->cfg->n_ports; i++) {
3695 const char *name = br->cfg->ports[i]->name;
3696 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3699 for (j = 0; j < port_cfg->n_interfaces; j++) {
3700 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3702 if (vlan_splinters_is_enabled(iface_cfg)) {
3703 const char *real_dev_name;
3706 real_dev_name = iface_cfg->name;
3707 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3708 const char *vlan_dev_name;
3710 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3712 && !shash_find(ports, vlan_dev_name)) {
3713 shash_add(ports, vlan_dev_name,
3714 synthesize_splinter_port(
3715 real_dev_name, vlan_dev_name, vid));
3724 mirror_refresh_stats(struct mirror *m)
3726 struct ofproto *ofproto = m->bridge->ofproto;
3727 uint64_t tx_packets, tx_bytes;
3730 size_t stat_cnt = 0;
3732 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3733 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3737 if (tx_packets != UINT64_MAX) {
3738 keys[stat_cnt] = "tx_packets";
3739 values[stat_cnt] = tx_packets;
3742 if (tx_bytes != UINT64_MAX) {
3743 keys[stat_cnt] = "tx_bytes";
3744 values[stat_cnt] = tx_bytes;
3748 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);