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 "meta-flow.h"
37 #include "ofp-print.h"
39 #include "ofproto/ofproto.h"
40 #include "poll-loop.h"
43 #include "socket-util.h"
45 #include "stream-ssl.h"
47 #include "system-stats.h"
52 #include "vswitchd/vswitch-idl.h"
53 #include "xenserver.h"
55 #include "sflow_api.h"
56 #include "vlan-bitmap.h"
58 VLOG_DEFINE_THIS_MODULE(bridge);
60 COVERAGE_DEFINE(bridge_reconfigure);
63 /* These members are always valid. */
64 struct list port_elem; /* Element in struct port's "ifaces" list. */
65 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
66 struct port *port; /* Containing port. */
67 char *name; /* Host network device name. */
68 tag_type tag; /* Tag associated with this interface. */
70 /* These members are valid only after bridge_reconfigure() causes them to
72 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
73 int ofp_port; /* OpenFlow port number, -1 if unknown. */
74 struct netdev *netdev; /* Network device. */
75 const char *type; /* Usually same as cfg->type. */
76 const struct ovsrec_interface *cfg;
80 struct uuid uuid; /* UUID of this "mirror" record in database. */
81 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
82 struct bridge *bridge;
84 const struct ovsrec_mirror *cfg;
88 struct bridge *bridge;
89 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
92 const struct ovsrec_port *cfg;
94 /* An ordinary bridge port has 1 interface.
95 * A bridge port for bonding has at least 2 interfaces. */
96 struct list ifaces; /* List of "struct iface"s. */
100 struct hmap_node node; /* In 'all_bridges'. */
101 char *name; /* User-specified arbitrary name. */
102 char *type; /* Datapath type. */
103 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
104 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
105 const struct ovsrec_bridge *cfg;
107 /* OpenFlow switch processing. */
108 struct ofproto *ofproto; /* OpenFlow switch. */
111 struct hmap ports; /* "struct port"s indexed by name. */
112 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
113 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
115 /* Port mirroring. */
116 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
118 /* Synthetic local port if necessary. */
119 struct ovsrec_port synth_local_port;
120 struct ovsrec_interface synth_local_iface;
121 struct ovsrec_interface *synth_local_ifacep;
124 /* All bridges, indexed by name. */
125 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
127 /* OVSDB IDL used to obtain configuration. */
128 static struct ovsdb_idl *idl;
130 /* Each time this timer expires, the bridge fetches systems and interface
131 * statistics and pushes them into the database. */
132 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
133 static long long int stats_timer = LLONG_MIN;
135 /* Stores the time after which rate limited statistics may be written to the
136 * database. Only updated when changes to the database require rate limiting.
138 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
139 static long long int db_limiter = LLONG_MIN;
141 static void add_del_bridges(const struct ovsrec_open_vswitch *);
142 static void bridge_del_ofprotos(void);
143 static bool bridge_add_ofprotos(struct bridge *);
144 static void bridge_create(const struct ovsrec_bridge *);
145 static void bridge_destroy(struct bridge *);
146 static struct bridge *bridge_lookup(const char *name);
147 static unixctl_cb_func bridge_unixctl_dump_flows;
148 static unixctl_cb_func bridge_unixctl_reconnect;
149 static size_t bridge_get_controllers(const struct bridge *br,
150 struct ovsrec_controller ***controllersp);
151 static void bridge_add_del_ports(struct bridge *,
152 const unsigned long int *splinter_vlans);
153 static void bridge_add_ofproto_ports(struct bridge *);
154 static void bridge_del_ofproto_ports(struct bridge *);
155 static void bridge_refresh_ofp_port(struct bridge *);
156 static void bridge_configure_datapath_id(struct bridge *);
157 static void bridge_configure_flow_eviction_threshold(struct bridge *);
158 static void bridge_configure_netflow(struct bridge *);
159 static void bridge_configure_forward_bpdu(struct bridge *);
160 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
161 static void bridge_configure_stp(struct bridge *);
162 static void bridge_configure_tables(struct bridge *);
163 static void bridge_configure_remotes(struct bridge *,
164 const struct sockaddr_in *managers,
166 static void bridge_pick_local_hw_addr(struct bridge *,
167 uint8_t ea[ETH_ADDR_LEN],
168 struct iface **hw_addr_iface);
169 static uint64_t bridge_pick_datapath_id(struct bridge *,
170 const uint8_t bridge_ea[ETH_ADDR_LEN],
171 struct iface *hw_addr_iface);
172 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
174 static const char *get_port_other_config(const struct ovsrec_port *,
176 const char *default_value);
177 static uint64_t dpid_from_hash(const void *, size_t nbytes);
178 static bool bridge_has_bond_fake_iface(const struct bridge *,
180 static bool port_is_bond_fake_iface(const struct port *);
182 static unixctl_cb_func qos_unixctl_show;
184 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
185 static void port_add_ifaces(struct port *);
186 static void port_del_ifaces(struct port *);
187 static void port_destroy(struct port *);
188 static struct port *port_lookup(const struct bridge *, const char *name);
189 static void port_configure(struct port *);
190 static struct lacp_settings *port_configure_lacp(struct port *,
191 struct lacp_settings *);
192 static void port_configure_bond(struct port *, struct bond_settings *,
193 uint32_t *bond_stable_ids);
194 static bool port_is_synthetic(const struct port *);
196 static void bridge_configure_mirrors(struct bridge *);
197 static struct mirror *mirror_create(struct bridge *,
198 const struct ovsrec_mirror *);
199 static void mirror_destroy(struct mirror *);
200 static bool mirror_configure(struct mirror *);
201 static void mirror_refresh_stats(struct mirror *);
203 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
204 static struct iface *iface_create(struct port *port,
205 const struct ovsrec_interface *if_cfg);
206 static void iface_destroy(struct iface *);
207 static struct iface *iface_lookup(const struct bridge *, const char *name);
208 static struct iface *iface_find(const char *name);
209 static struct iface *iface_from_ofp_port(const struct bridge *,
211 static void iface_set_mac(struct iface *);
212 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
213 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
214 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
215 static void iface_configure_cfm(struct iface *);
216 static void iface_refresh_cfm_stats(struct iface *);
217 static void iface_refresh_stats(struct iface *);
218 static void iface_refresh_status(struct iface *);
219 static bool iface_is_synthetic(const struct iface *);
220 static const char *get_interface_other_config(const struct ovsrec_interface *,
222 const char *default_value);
224 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
226 static void shash_to_ovs_idl_map(struct shash *,
227 char ***keys, char ***values, size_t *n);
229 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
231 * This is deprecated. It is only for compatibility with broken device drivers
232 * in old versions of Linux that do not properly support VLANs when VLAN
233 * devices are not used. When broken device drivers are no longer in
234 * widespread use, we will delete these interfaces. */
236 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
237 static bool vlan_splinters_enabled_anywhere;
239 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
240 static unsigned long int *collect_splinter_vlans(
241 const struct ovsrec_open_vswitch *);
242 static void configure_splinter_port(struct port *);
243 static void add_vlan_splinter_ports(struct bridge *,
244 const unsigned long int *splinter_vlans,
245 struct shash *ports);
247 /* Public functions. */
249 /* Initializes the bridge module, configuring it to obtain its configuration
250 * from an OVSDB server accessed over 'remote', which should be a string in a
251 * form acceptable to ovsdb_idl_create(). */
253 bridge_init(const char *remote)
255 /* Create connection to database. */
256 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
257 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
259 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
260 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
261 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
262 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
263 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
264 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
265 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
267 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
268 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
269 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
271 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
272 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
273 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
274 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
276 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
277 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
283 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
284 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
285 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
286 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
287 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
288 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
290 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
291 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
292 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
293 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
297 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
299 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
300 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
302 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
306 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
307 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
308 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
309 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
310 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
312 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
314 /* Register unixctl commands. */
315 unixctl_command_register("qos/show", "interface", 1, 1,
316 qos_unixctl_show, NULL);
317 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
318 bridge_unixctl_dump_flows, NULL);
319 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
320 bridge_unixctl_reconnect, NULL);
330 struct bridge *br, *next_br;
332 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
335 ovsdb_idl_destroy(idl);
338 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
339 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
340 * responsible for freeing '*managersp' (with free()).
342 * You may be asking yourself "why does ovs-vswitchd care?", because
343 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
344 * should not be and in fact is not directly involved in that. But
345 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
346 * it has to tell in-band control where the managers are to enable that.
347 * (Thus, only managers connected in-band are collected.)
350 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
351 struct sockaddr_in **managersp, size_t *n_managersp)
353 struct sockaddr_in *managers = NULL;
354 size_t n_managers = 0;
358 /* Collect all of the potential targets from the "targets" columns of the
359 * rows pointed to by "manager_options", excluding any that are
362 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
363 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
365 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
366 sset_find_and_delete(&targets, m->target);
368 sset_add(&targets, m->target);
372 /* Now extract the targets' IP addresses. */
373 if (!sset_is_empty(&targets)) {
376 managers = xmalloc(sset_count(&targets) * sizeof *managers);
377 SSET_FOR_EACH (target, &targets) {
378 struct sockaddr_in *sin = &managers[n_managers];
380 if (stream_parse_target_with_default_ports(target,
388 sset_destroy(&targets);
390 *managersp = managers;
391 *n_managersp = n_managers;
395 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
397 unsigned long int *splinter_vlans;
398 struct sockaddr_in *managers;
399 struct bridge *br, *next;
400 int sflow_bridge_number;
403 COVERAGE_INC(bridge_reconfigure);
405 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
406 * iface"s according to 'ovs_cfg', with only very minimal configuration
409 * This is mostly an update to bridge data structures. Very little is
410 * pushed down to ofproto or lower layers. */
411 add_del_bridges(ovs_cfg);
412 splinter_vlans = collect_splinter_vlans(ovs_cfg);
413 HMAP_FOR_EACH (br, node, &all_bridges) {
414 bridge_add_del_ports(br, splinter_vlans);
416 free(splinter_vlans);
418 /* Delete all datapaths and datapath ports that are no longer configured.
420 * The kernel will reject any attempt to add a given port to a datapath if
421 * that port already belongs to a different datapath, so we must do all
422 * port deletions before any port additions. A datapath always has a
423 * "local port" so we must delete not-configured datapaths too. */
424 bridge_del_ofprotos();
425 HMAP_FOR_EACH (br, node, &all_bridges) {
427 bridge_del_ofproto_ports(br);
431 /* Create datapaths and datapath ports that are missing.
433 * After this is done, we have our final set of bridges, ports, and
434 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
435 * has at least one iface, every "struct iface" has a valid ofp_port and
437 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
438 if (!br->ofproto && !bridge_add_ofprotos(br)) {
442 HMAP_FOR_EACH (br, node, &all_bridges) {
443 bridge_refresh_ofp_port(br);
444 bridge_add_ofproto_ports(br);
447 /* Complete the configuration. */
448 sflow_bridge_number = 0;
449 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
450 HMAP_FOR_EACH (br, node, &all_bridges) {
453 /* We need the datapath ID early to allow LACP ports to use it as the
454 * default system ID. */
455 bridge_configure_datapath_id(br);
457 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
460 port_configure(port);
462 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
463 iface_configure_cfm(iface);
464 iface_configure_qos(iface, port->cfg->qos);
465 iface_set_mac(iface);
468 bridge_configure_mirrors(br);
469 bridge_configure_flow_eviction_threshold(br);
470 bridge_configure_forward_bpdu(br);
471 bridge_configure_remotes(br, managers, n_managers);
472 bridge_configure_netflow(br);
473 bridge_configure_sflow(br, &sflow_bridge_number);
474 bridge_configure_stp(br);
475 bridge_configure_tables(br);
479 /* ovs-vswitchd has completed initialization, so allow the process that
480 * forked us to exit successfully. */
481 daemonize_complete();
484 /* Iterate over all ofprotos and delete any of them that do not have a
485 * configured bridge or that are the wrong type. */
487 bridge_del_ofprotos(void)
495 ofproto_enumerate_types(&types);
496 SSET_FOR_EACH (type, &types) {
499 ofproto_enumerate_names(type, &names);
500 SSET_FOR_EACH (name, &names) {
501 struct bridge *br = bridge_lookup(name);
502 if (!br || strcmp(type, br->type)) {
503 ofproto_delete(name, type);
507 sset_destroy(&names);
508 sset_destroy(&types);
512 bridge_add_ofprotos(struct bridge *br)
514 int error = ofproto_create(br->name, br->type, &br->ofproto);
516 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
523 port_configure(struct port *port)
525 const struct ovsrec_port *cfg = port->cfg;
526 struct bond_settings bond_settings;
527 struct lacp_settings lacp_settings;
528 struct ofproto_bundle_settings s;
531 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
532 configure_splinter_port(port);
541 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
542 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
543 s.slaves[s.n_slaves++] = iface->ofp_port;
548 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
552 /* Get VLAN trunks. */
555 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
559 if (cfg->vlan_mode) {
560 if (!strcmp(cfg->vlan_mode, "access")) {
561 s.vlan_mode = PORT_VLAN_ACCESS;
562 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
563 s.vlan_mode = PORT_VLAN_TRUNK;
564 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
565 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
566 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
567 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
569 /* This "can't happen" because ovsdb-server should prevent it. */
570 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
571 s.vlan_mode = PORT_VLAN_TRUNK;
575 s.vlan_mode = PORT_VLAN_ACCESS;
577 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
581 s.vlan_mode = PORT_VLAN_TRUNK;
584 s.use_priority_tags = !strcmp("true", get_port_other_config(
585 cfg, "priority-tags", ""));
587 /* Get LACP settings. */
588 s.lacp = port_configure_lacp(port, &lacp_settings);
592 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
593 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
594 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
597 s.lacp_slaves = NULL;
600 /* Get bond settings. */
601 if (s.n_slaves > 1) {
602 s.bond = &bond_settings;
603 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
604 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
607 s.bond_stable_ids = NULL;
609 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
610 netdev_set_miimon_interval(iface->netdev, 0);
615 ofproto_bundle_register(port->bridge->ofproto, port, &s);
621 free(s.bond_stable_ids);
624 /* Pick local port hardware address and datapath ID for 'br'. */
626 bridge_configure_datapath_id(struct bridge *br)
628 uint8_t ea[ETH_ADDR_LEN];
630 struct iface *local_iface;
631 struct iface *hw_addr_iface;
634 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
635 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
637 int error = netdev_set_etheraddr(local_iface->netdev, ea);
639 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
640 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
641 "Ethernet address: %s",
642 br->name, strerror(error));
645 memcpy(br->ea, ea, ETH_ADDR_LEN);
647 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
648 ofproto_set_datapath_id(br->ofproto, dpid);
650 dpid_string = xasprintf("%016"PRIx64, dpid);
651 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
655 /* Set NetFlow configuration on 'br'. */
657 bridge_configure_netflow(struct bridge *br)
659 struct ovsrec_netflow *cfg = br->cfg->netflow;
660 struct netflow_options opts;
663 ofproto_set_netflow(br->ofproto, NULL);
667 memset(&opts, 0, sizeof opts);
669 /* Get default NetFlow configuration from datapath.
670 * Apply overrides from 'cfg'. */
671 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
672 if (cfg->engine_type) {
673 opts.engine_type = *cfg->engine_type;
675 if (cfg->engine_id) {
676 opts.engine_id = *cfg->engine_id;
679 /* Configure active timeout interval. */
680 opts.active_timeout = cfg->active_timeout;
681 if (!opts.active_timeout) {
682 opts.active_timeout = -1;
683 } else if (opts.active_timeout < 0) {
684 VLOG_WARN("bridge %s: active timeout interval set to negative "
685 "value, using default instead (%d seconds)", br->name,
686 NF_ACTIVE_TIMEOUT_DEFAULT);
687 opts.active_timeout = -1;
690 /* Add engine ID to interface number to disambiguate bridgs? */
691 opts.add_id_to_iface = cfg->add_id_to_interface;
692 if (opts.add_id_to_iface) {
693 if (opts.engine_id > 0x7f) {
694 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
695 "another vswitch, choose an engine id less than 128",
698 if (hmap_count(&br->ports) > 508) {
699 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
700 "another port when more than 508 ports are used",
706 sset_init(&opts.collectors);
707 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
710 if (ofproto_set_netflow(br->ofproto, &opts)) {
711 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
713 sset_destroy(&opts.collectors);
716 /* Set sFlow configuration on 'br'. */
718 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
720 const struct ovsrec_sflow *cfg = br->cfg->sflow;
721 struct ovsrec_controller **controllers;
722 struct ofproto_sflow_options oso;
723 size_t n_controllers;
727 ofproto_set_sflow(br->ofproto, NULL);
731 memset(&oso, 0, sizeof oso);
733 sset_init(&oso.targets);
734 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
736 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
738 oso.sampling_rate = *cfg->sampling;
741 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
743 oso.polling_interval = *cfg->polling;
746 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
748 oso.header_len = *cfg->header;
751 oso.sub_id = (*sflow_bridge_number)++;
752 oso.agent_device = cfg->agent;
754 oso.control_ip = NULL;
755 n_controllers = bridge_get_controllers(br, &controllers);
756 for (i = 0; i < n_controllers; i++) {
757 if (controllers[i]->local_ip) {
758 oso.control_ip = controllers[i]->local_ip;
762 ofproto_set_sflow(br->ofproto, &oso);
764 sset_destroy(&oso.targets);
768 port_configure_stp(const struct ofproto *ofproto, struct port *port,
769 struct ofproto_port_stp_settings *port_s,
770 int *port_num_counter, unsigned long *port_num_bitmap)
772 const char *config_str;
775 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
776 if (config_str && !strcmp(config_str, "false")) {
777 port_s->enable = false;
780 port_s->enable = true;
783 /* STP over bonds is not supported. */
784 if (!list_is_singleton(&port->ifaces)) {
785 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
787 port_s->enable = false;
791 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
793 /* Internal ports shouldn't participate in spanning tree, so
795 if (!strcmp(iface->type, "internal")) {
796 VLOG_DBG("port %s: disable STP on internal ports", port->name);
797 port_s->enable = false;
801 /* STP on mirror output ports is not supported. */
802 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
803 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
804 port_s->enable = false;
808 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
810 unsigned long int port_num = strtoul(config_str, NULL, 0);
811 int port_idx = port_num - 1;
813 if (port_num < 1 || port_num > STP_MAX_PORTS) {
814 VLOG_ERR("port %s: invalid stp-port-num", port->name);
815 port_s->enable = false;
819 if (bitmap_is_set(port_num_bitmap, port_idx)) {
820 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
821 port->name, port_num);
822 port_s->enable = false;
825 bitmap_set1(port_num_bitmap, port_idx);
826 port_s->port_num = port_idx;
828 if (*port_num_counter > STP_MAX_PORTS) {
829 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
830 port_s->enable = false;
834 port_s->port_num = (*port_num_counter)++;
837 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
839 port_s->path_cost = strtoul(config_str, NULL, 10);
843 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
844 /* Couldn't get speed, so assume 100Mb/s. */
845 port_s->path_cost = 19;
849 mbps = netdev_features_to_bps(current) / 1000000;
850 port_s->path_cost = stp_convert_speed_to_cost(mbps);
854 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
856 port_s->priority = strtoul(config_str, NULL, 0);
858 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
862 /* Set spanning tree configuration on 'br'. */
864 bridge_configure_stp(struct bridge *br)
866 if (!br->cfg->stp_enable) {
867 ofproto_set_stp(br->ofproto, NULL);
869 struct ofproto_stp_settings br_s;
870 const char *config_str;
872 int port_num_counter;
873 unsigned long *port_num_bitmap;
875 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
877 uint8_t ea[ETH_ADDR_LEN];
879 if (eth_addr_from_string(config_str, ea)) {
880 br_s.system_id = eth_addr_to_uint64(ea);
882 br_s.system_id = eth_addr_to_uint64(br->ea);
883 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
884 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
887 br_s.system_id = eth_addr_to_uint64(br->ea);
890 config_str = bridge_get_other_config(br->cfg, "stp-priority");
892 br_s.priority = strtoul(config_str, NULL, 0);
894 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
897 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
899 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
901 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
904 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
906 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
908 br_s.max_age = STP_DEFAULT_MAX_AGE;
911 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
913 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
915 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
918 /* Configure STP on the bridge. */
919 if (ofproto_set_stp(br->ofproto, &br_s)) {
920 VLOG_ERR("bridge %s: could not enable STP", br->name);
924 /* Users must either set the port number with the "stp-port-num"
925 * configuration on all ports or none. If manual configuration
926 * is not done, then we allocate them sequentially. */
927 port_num_counter = 0;
928 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
929 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
930 struct ofproto_port_stp_settings port_s;
933 port_configure_stp(br->ofproto, port, &port_s,
934 &port_num_counter, port_num_bitmap);
936 /* As bonds are not supported, just apply configuration to
938 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
939 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
941 VLOG_ERR("port %s: could not enable STP", port->name);
947 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
948 && port_num_counter) {
949 VLOG_ERR("bridge %s: must manually configure all STP port "
950 "IDs or none, disabling", br->name);
951 ofproto_set_stp(br->ofproto, NULL);
953 bitmap_free(port_num_bitmap);
958 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
960 const struct port *port = port_lookup(br, name);
961 return port && port_is_bond_fake_iface(port);
965 port_is_bond_fake_iface(const struct port *port)
967 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
971 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
973 struct bridge *br, *next;
977 /* Collect new bridges' names and types. */
979 for (i = 0; i < cfg->n_bridges; i++) {
980 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
981 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
983 if (strchr(br_cfg->name, '/')) {
984 /* Prevent remote ovsdb-server users from accessing arbitrary
985 * directories, e.g. consider a bridge named "../../../etc/". */
986 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
988 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
989 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
993 /* Get rid of deleted bridges or those whose types have changed.
994 * Update 'cfg' of bridges that still exist. */
995 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
996 br->cfg = shash_find_data(&new_br, br->name);
997 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
998 br->cfg->datapath_type))) {
1003 /* Add new bridges. */
1004 for (i = 0; i < cfg->n_bridges; i++) {
1005 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1006 struct bridge *br = bridge_lookup(br_cfg->name);
1008 bridge_create(br_cfg);
1012 shash_destroy(&new_br);
1015 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1018 * The kernel will reject any attempt to add a given port to a datapath if that
1019 * port already belongs to a different datapath, so we must do all port
1020 * deletions before any port additions. */
1022 bridge_del_ofproto_ports(struct bridge *br)
1024 struct ofproto_port_dump dump;
1025 struct ofproto_port ofproto_port;
1027 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1028 const char *name = ofproto_port.name;
1029 struct iface *iface;
1033 /* Ignore the local port. We can't change it anyhow. */
1034 if (!strcmp(name, br->name)) {
1038 /* Get the type that 'ofproto_port' should have (ordinarily the
1039 * type of its corresponding iface) or NULL if it should be
1041 iface = iface_lookup(br, name);
1042 type = (iface ? iface->type
1043 : bridge_has_bond_fake_iface(br, name) ? "internal"
1046 /* If it's the wrong type then delete the ofproto port. */
1048 && !strcmp(ofproto_port.type, type)
1049 && (!iface || !iface->netdev
1050 || !strcmp(netdev_get_type(iface->netdev), type))) {
1053 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1055 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1056 br->name, name, strerror(error));
1059 netdev_close(iface->netdev);
1060 iface->netdev = NULL;
1066 iface_set_ofp_port(struct iface *iface, int ofp_port)
1068 struct bridge *br = iface->port->bridge;
1070 assert(iface->ofp_port < 0 && ofp_port >= 0);
1071 iface->ofp_port = ofp_port;
1072 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1073 iface_set_ofport(iface->cfg, ofp_port);
1077 bridge_refresh_ofp_port(struct bridge *br)
1079 struct ofproto_port_dump dump;
1080 struct ofproto_port ofproto_port;
1083 /* Clear all the "ofp_port"es. */
1084 hmap_clear(&br->ifaces);
1085 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1086 struct iface *iface;
1088 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1089 iface->ofp_port = -1;
1093 /* Obtain the correct "ofp_port"s from ofproto. */
1094 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1095 struct iface *iface = iface_lookup(br, ofproto_port.name);
1097 if (iface->ofp_port >= 0) {
1098 VLOG_WARN("bridge %s: interface %s reported twice",
1099 br->name, ofproto_port.name);
1100 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1101 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1102 br->name, ofproto_port.ofp_port);
1104 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1110 /* Add an ofproto port for any "struct iface" that doesn't have one.
1111 * Delete any "struct iface" for which this fails.
1112 * Delete any "struct port" that thereby ends up with no ifaces. */
1114 bridge_add_ofproto_ports(struct bridge *br)
1116 struct port *port, *next_port;
1118 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1119 struct iface *iface, *next_iface;
1120 struct ofproto_port ofproto_port;
1122 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1125 /* Open the netdev. */
1126 if (!iface->netdev) {
1127 error = netdev_open(iface->name, iface->type, &iface->netdev);
1129 VLOG_WARN("could not open network device %s (%s)",
1130 iface->name, strerror(error));
1134 && port->cfg->vlan_mode
1135 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1136 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1142 /* Configure the netdev. */
1143 if (iface->netdev) {
1147 shash_from_ovs_idl_map(iface->cfg->key_options,
1148 iface->cfg->value_options,
1149 iface->cfg->n_options, &args);
1150 error = netdev_set_config(iface->netdev, &args);
1151 shash_destroy(&args);
1154 VLOG_WARN("could not configure network device %s (%s)",
1155 iface->name, strerror(error));
1156 netdev_close(iface->netdev);
1157 iface->netdev = NULL;
1161 /* Add the port, if necessary. */
1162 if (iface->netdev && iface->ofp_port < 0) {
1166 error = ofproto_port_add(br->ofproto, iface->netdev,
1169 iface_set_ofp_port(iface, ofp_port);
1171 netdev_close(iface->netdev);
1172 iface->netdev = NULL;
1176 /* Populate stats columns in new Interface rows. */
1177 if (iface->netdev && !iface->cfg->mtu) {
1178 iface_refresh_stats(iface);
1179 iface_refresh_status(iface);
1182 /* Delete the iface if we failed. */
1183 if (iface->netdev && iface->ofp_port >= 0) {
1184 VLOG_DBG("bridge %s: interface %s is on port %d",
1185 br->name, iface->name, iface->ofp_port);
1187 if (iface->netdev) {
1188 VLOG_ERR("bridge %s: missing %s interface, dropping",
1189 br->name, iface->name);
1191 /* We already reported a related error, don't bother
1192 * duplicating it. */
1194 iface_clear_db_record(iface->cfg);
1195 iface_destroy(iface);
1198 if (list_is_empty(&port->ifaces)) {
1199 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1204 /* Add bond fake iface if necessary. */
1205 if (port_is_bond_fake_iface(port)) {
1206 if (ofproto_port_query_by_name(br->ofproto, port->name,
1208 struct netdev *netdev;
1211 error = netdev_open(port->name, "internal", &netdev);
1213 ofproto_port_add(br->ofproto, netdev, NULL);
1214 netdev_close(netdev);
1216 VLOG_WARN("could not open network device %s (%s)",
1217 port->name, strerror(error));
1220 /* Already exists, nothing to do. */
1221 ofproto_port_destroy(&ofproto_port);
1228 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1232 for (i = 0; i < n; i++) {
1233 if (!strcmp(keys[i], key)) {
1241 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1243 return get_ovsrec_key_value(br_cfg->key_other_config,
1244 br_cfg->value_other_config,
1245 br_cfg->n_other_config, key);
1248 /* Set Flow eviction threshold */
1250 bridge_configure_flow_eviction_threshold(struct bridge *br)
1252 const char *threshold_str;
1255 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1256 if (threshold_str) {
1257 threshold = strtoul(threshold_str, NULL, 10);
1259 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1261 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1264 /* Set forward BPDU option. */
1266 bridge_configure_forward_bpdu(struct bridge *br)
1268 const char *forward_bpdu_str;
1269 bool forward_bpdu = false;
1271 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1272 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1273 forward_bpdu = true;
1275 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1279 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1280 struct iface **hw_addr_iface)
1282 struct hmapx mirror_output_ports;
1285 bool found_addr = false;
1289 *hw_addr_iface = NULL;
1291 /* Did the user request a particular MAC? */
1292 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1293 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1294 if (eth_addr_is_multicast(ea)) {
1295 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1296 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1297 } else if (eth_addr_is_zero(ea)) {
1298 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1304 /* Mirror output ports don't participate in picking the local hardware
1305 * address. ofproto can't help us find out whether a given port is a
1306 * mirror output because we haven't configured mirrors yet, so we need to
1307 * accumulate them ourselves. */
1308 hmapx_init(&mirror_output_ports);
1309 for (i = 0; i < br->cfg->n_mirrors; i++) {
1310 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1311 if (m->output_port) {
1312 hmapx_add(&mirror_output_ports, m->output_port);
1316 /* Otherwise choose the minimum non-local MAC address among all of the
1318 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1319 uint8_t iface_ea[ETH_ADDR_LEN];
1320 struct iface *candidate;
1321 struct iface *iface;
1323 /* Mirror output ports don't participate. */
1324 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1328 /* Choose the MAC address to represent the port. */
1330 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1331 /* Find the interface with this Ethernet address (if any) so that
1332 * we can provide the correct devname to the caller. */
1333 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1334 uint8_t candidate_ea[ETH_ADDR_LEN];
1335 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1336 && eth_addr_equals(iface_ea, candidate_ea)) {
1341 /* Choose the interface whose MAC address will represent the port.
1342 * The Linux kernel bonding code always chooses the MAC address of
1343 * the first slave added to a bond, and the Fedora networking
1344 * scripts always add slaves to a bond in alphabetical order, so
1345 * for compatibility we choose the interface with the name that is
1346 * first in alphabetical order. */
1347 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1348 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1353 /* The local port doesn't count (since we're trying to choose its
1354 * MAC address anyway). */
1355 if (iface->ofp_port == OFPP_LOCAL) {
1360 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1366 /* Compare against our current choice. */
1367 if (!eth_addr_is_multicast(iface_ea) &&
1368 !eth_addr_is_local(iface_ea) &&
1369 !eth_addr_is_reserved(iface_ea) &&
1370 !eth_addr_is_zero(iface_ea) &&
1371 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1373 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1374 *hw_addr_iface = iface;
1379 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1380 br->name, ETH_ADDR_ARGS(ea));
1382 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1383 *hw_addr_iface = NULL;
1384 VLOG_WARN("bridge %s: using default bridge Ethernet "
1385 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1388 hmapx_destroy(&mirror_output_ports);
1391 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1392 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1393 * an interface on 'br', then that interface must be passed in as
1394 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1395 * 'hw_addr_iface' must be passed in as a null pointer. */
1397 bridge_pick_datapath_id(struct bridge *br,
1398 const uint8_t bridge_ea[ETH_ADDR_LEN],
1399 struct iface *hw_addr_iface)
1402 * The procedure for choosing a bridge MAC address will, in the most
1403 * ordinary case, also choose a unique MAC that we can use as a datapath
1404 * ID. In some special cases, though, multiple bridges will end up with
1405 * the same MAC address. This is OK for the bridges, but it will confuse
1406 * the OpenFlow controller, because each datapath needs a unique datapath
1409 * Datapath IDs must be unique. It is also very desirable that they be
1410 * stable from one run to the next, so that policy set on a datapath
1413 const char *datapath_id;
1416 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1417 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1421 if (!hw_addr_iface) {
1423 * A purely internal bridge, that is, one that has no non-virtual
1424 * network devices on it at all, is difficult because it has no
1425 * natural unique identifier at all.
1427 * When the host is a XenServer, we handle this case by hashing the
1428 * host's UUID with the name of the bridge. Names of bridges are
1429 * persistent across XenServer reboots, although they can be reused if
1430 * an internal network is destroyed and then a new one is later
1431 * created, so this is fairly effective.
1433 * When the host is not a XenServer, we punt by using a random MAC
1434 * address on each run.
1436 const char *host_uuid = xenserver_get_host_uuid();
1438 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1439 dpid = dpid_from_hash(combined, strlen(combined));
1445 return eth_addr_to_uint64(bridge_ea);
1449 dpid_from_hash(const void *data, size_t n)
1451 uint8_t hash[SHA1_DIGEST_SIZE];
1453 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1454 sha1_bytes(data, n, hash);
1455 eth_addr_mark_random(hash);
1456 return eth_addr_to_uint64(hash);
1460 iface_refresh_status(struct iface *iface)
1464 enum netdev_flags flags;
1471 if (iface_is_synthetic(iface)) {
1477 if (!netdev_get_status(iface->netdev, &sh)) {
1479 char **keys, **values;
1481 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1482 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1487 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1490 shash_destroy_free_data(&sh);
1492 error = netdev_get_flags(iface->netdev, &flags);
1494 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1497 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1500 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1502 ovsrec_interface_set_duplex(iface->cfg,
1503 netdev_features_is_full_duplex(current)
1505 /* warning: uint64_t -> int64_t conversion */
1506 bps = netdev_features_to_bps(current);
1507 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1510 ovsrec_interface_set_duplex(iface->cfg, NULL);
1511 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1514 error = netdev_get_mtu(iface->netdev, &mtu);
1517 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1520 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1524 /* Writes 'iface''s CFM statistics to the database. */
1526 iface_refresh_cfm_stats(struct iface *iface)
1528 const struct ovsrec_interface *cfg = iface->cfg;
1530 const uint64_t *rmps;
1533 if (iface_is_synthetic(iface)) {
1537 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1540 bool fault_bool = fault;
1541 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1543 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1546 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1547 iface->ofp_port, &rmps, &n_rmps);
1549 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1552 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1557 iface_refresh_stats(struct iface *iface)
1559 #define IFACE_STATS \
1560 IFACE_STAT(rx_packets, "rx_packets") \
1561 IFACE_STAT(tx_packets, "tx_packets") \
1562 IFACE_STAT(rx_bytes, "rx_bytes") \
1563 IFACE_STAT(tx_bytes, "tx_bytes") \
1564 IFACE_STAT(rx_dropped, "rx_dropped") \
1565 IFACE_STAT(tx_dropped, "tx_dropped") \
1566 IFACE_STAT(rx_errors, "rx_errors") \
1567 IFACE_STAT(tx_errors, "tx_errors") \
1568 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1569 IFACE_STAT(rx_over_errors, "rx_over_err") \
1570 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1571 IFACE_STAT(collisions, "collisions")
1573 #define IFACE_STAT(MEMBER, NAME) NAME,
1574 static char *keys[] = { IFACE_STATS };
1576 int64_t values[ARRAY_SIZE(keys)];
1579 struct netdev_stats stats;
1581 if (iface_is_synthetic(iface)) {
1585 /* Intentionally ignore return value, since errors will set 'stats' to
1586 * all-1s, and we will deal with that correctly below. */
1587 netdev_get_stats(iface->netdev, &stats);
1589 /* Copy statistics into values[] array. */
1591 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1594 assert(i == ARRAY_SIZE(keys));
1596 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1601 br_refresh_stp_status(struct bridge *br)
1603 struct ofproto *ofproto = br->ofproto;
1604 struct ofproto_stp_status status;
1605 char *keys[3], *values[3];
1608 if (ofproto_get_stp_status(ofproto, &status)) {
1612 if (!status.enabled) {
1613 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1617 keys[0] = "stp_bridge_id",
1618 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1619 keys[1] = "stp_designated_root",
1620 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1621 keys[2] = "stp_root_path_cost",
1622 values[2] = xasprintf("%d", status.root_path_cost);
1624 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1626 for (i = 0; i < ARRAY_SIZE(values); i++) {
1632 port_refresh_stp_status(struct port *port)
1634 struct ofproto *ofproto = port->bridge->ofproto;
1635 struct iface *iface;
1636 struct ofproto_port_stp_status status;
1638 char *str_values[4];
1639 int64_t int_values[3];
1642 if (port_is_synthetic(port)) {
1646 /* STP doesn't currently support bonds. */
1647 if (!list_is_singleton(&port->ifaces)) {
1648 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1652 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1654 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1658 if (!status.enabled) {
1659 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1660 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1664 /* Set Status column. */
1665 keys[0] = "stp_port_id";
1666 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1667 keys[1] = "stp_state";
1668 str_values[1] = xstrdup(stp_state_name(status.state));
1669 keys[2] = "stp_sec_in_state";
1670 str_values[2] = xasprintf("%u", status.sec_in_state);
1671 keys[3] = "stp_role";
1672 str_values[3] = xstrdup(stp_role_name(status.role));
1674 ovsrec_port_set_status(port->cfg, keys, str_values,
1675 ARRAY_SIZE(str_values));
1677 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1678 free(str_values[i]);
1681 /* Set Statistics column. */
1682 keys[0] = "stp_tx_count";
1683 int_values[0] = status.tx_count;
1684 keys[1] = "stp_rx_count";
1685 int_values[1] = status.rx_count;
1686 keys[2] = "stp_error_count";
1687 int_values[2] = status.error_count;
1689 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1690 ARRAY_SIZE(int_values));
1694 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1698 /* Use other-config:enable-system-stats by preference. */
1699 enable = get_ovsrec_key_value(cfg->key_other_config,
1700 cfg->value_other_config,
1701 cfg->n_other_config,
1702 "enable-statistics");
1704 return !strcmp(enable, "true");
1707 /* Disable by default. */
1712 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1714 struct ovsdb_datum datum;
1718 if (enable_system_stats(cfg)) {
1719 get_system_stats(&stats);
1722 ovsdb_datum_from_shash(&datum, &stats);
1723 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1727 static inline const char *
1728 nx_role_to_str(enum nx_role role)
1733 case NX_ROLE_MASTER:
1738 return "*** INVALID ROLE ***";
1743 refresh_controller_status(void)
1747 const struct ovsrec_controller *cfg;
1751 /* Accumulate status for controllers on all bridges. */
1752 HMAP_FOR_EACH (br, node, &all_bridges) {
1753 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1756 /* Update each controller in the database with current status. */
1757 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1758 struct ofproto_controller_info *cinfo =
1759 shash_find_data(&info, cfg->target);
1762 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1763 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1764 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1765 (char **) cinfo->pairs.values,
1768 ovsrec_controller_set_is_connected(cfg, false);
1769 ovsrec_controller_set_role(cfg, NULL);
1770 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1774 ofproto_free_ofproto_controller_info(&info);
1778 refresh_cfm_stats(void)
1780 static struct ovsdb_idl_txn *txn = NULL;
1785 txn = ovsdb_idl_txn_create(idl);
1787 HMAP_FOR_EACH (br, node, &all_bridges) {
1788 struct iface *iface;
1790 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1791 iface_refresh_cfm_stats(iface);
1796 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1797 ovsdb_idl_txn_destroy(txn);
1802 /* Performs periodic activity required by bridges that needs to be done with
1803 * the least possible latency.
1805 * It makes sense to call this function a couple of times per poll loop, to
1806 * provide a significant performance boost on some benchmarks with ofprotos
1807 * that use the ofproto-dpif implementation. */
1809 bridge_run_fast(void)
1813 HMAP_FOR_EACH (br, node, &all_bridges) {
1814 ofproto_run_fast(br->ofproto);
1821 const struct ovsrec_open_vswitch *cfg;
1823 bool vlan_splinters_changed;
1824 bool database_changed;
1827 /* (Re)configure if necessary. */
1828 database_changed = ovsdb_idl_run(idl);
1829 if (ovsdb_idl_is_lock_contended(idl)) {
1830 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1831 struct bridge *br, *next_br;
1833 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1834 "disabling this process until it goes away");
1836 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1840 } else if (!ovsdb_idl_has_lock(idl)) {
1843 cfg = ovsrec_open_vswitch_first(idl);
1845 /* Let each bridge do the work that it needs to do. */
1846 HMAP_FOR_EACH (br, node, &all_bridges) {
1847 ofproto_run(br->ofproto);
1850 /* Re-configure SSL. We do this on every trip through the main loop,
1851 * instead of just when the database changes, because the contents of the
1852 * key and certificate files can change without the database changing.
1854 * We do this before bridge_reconfigure() because that function might
1855 * initiate SSL connections and thus requires SSL to be configured. */
1856 if (cfg && cfg->ssl) {
1857 const struct ovsrec_ssl *ssl = cfg->ssl;
1859 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1860 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1863 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1865 vlan_splinters_changed = false;
1866 if (vlan_splinters_enabled_anywhere) {
1867 HMAP_FOR_EACH (br, node, &all_bridges) {
1868 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1869 vlan_splinters_changed = true;
1875 if (database_changed || vlan_splinters_changed) {
1877 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1879 bridge_reconfigure(cfg);
1881 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1882 ovsdb_idl_txn_commit(txn);
1883 ovsdb_idl_txn_destroy(txn); /* XXX */
1885 /* We still need to reconfigure to avoid dangling pointers to
1886 * now-destroyed ovsrec structures inside bridge data. */
1887 static const struct ovsrec_open_vswitch null_cfg;
1889 bridge_reconfigure(&null_cfg);
1893 /* Refresh system and interface stats if necessary. */
1894 if (time_msec() >= stats_timer) {
1896 struct ovsdb_idl_txn *txn;
1898 txn = ovsdb_idl_txn_create(idl);
1899 HMAP_FOR_EACH (br, node, &all_bridges) {
1903 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1904 struct iface *iface;
1906 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1907 iface_refresh_stats(iface);
1908 iface_refresh_status(iface);
1912 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1913 mirror_refresh_stats(m);
1917 refresh_system_stats(cfg);
1918 refresh_controller_status();
1919 ovsdb_idl_txn_commit(txn);
1920 ovsdb_idl_txn_destroy(txn); /* XXX */
1923 stats_timer = time_msec() + STATS_INTERVAL;
1926 if (time_msec() >= db_limiter) {
1927 struct ovsdb_idl_txn *txn;
1929 txn = ovsdb_idl_txn_create(idl);
1930 HMAP_FOR_EACH (br, node, &all_bridges) {
1931 struct iface *iface;
1934 br_refresh_stp_status(br);
1936 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1937 port_refresh_stp_status(port);
1940 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1941 const char *link_state;
1942 int64_t link_resets;
1945 if (iface_is_synthetic(iface)) {
1949 current = ofproto_port_is_lacp_current(br->ofproto,
1953 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1955 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1958 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1959 ovsrec_interface_set_link_state(iface->cfg, link_state);
1961 link_resets = netdev_get_carrier_resets(iface->netdev);
1962 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1966 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
1967 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1969 ovsdb_idl_txn_destroy(txn);
1972 refresh_cfm_stats();
1978 ovsdb_idl_wait(idl);
1979 if (!hmap_is_empty(&all_bridges)) {
1982 HMAP_FOR_EACH (br, node, &all_bridges) {
1983 ofproto_wait(br->ofproto);
1985 poll_timer_wait_until(stats_timer);
1987 if (db_limiter > time_msec()) {
1988 poll_timer_wait_until(db_limiter);
1993 /* QoS unixctl user interface functions. */
1995 struct qos_unixctl_show_cbdata {
1997 struct iface *iface;
2001 qos_unixctl_show_cb(unsigned int queue_id,
2002 const struct shash *details,
2005 struct qos_unixctl_show_cbdata *data = aux;
2006 struct ds *ds = data->ds;
2007 struct iface *iface = data->iface;
2008 struct netdev_queue_stats stats;
2009 struct shash_node *node;
2012 ds_put_cstr(ds, "\n");
2014 ds_put_format(ds, "Queue %u:\n", queue_id);
2016 ds_put_cstr(ds, "Default:\n");
2019 SHASH_FOR_EACH (node, details) {
2020 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2023 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2025 if (stats.tx_packets != UINT64_MAX) {
2026 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2029 if (stats.tx_bytes != UINT64_MAX) {
2030 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2033 if (stats.tx_errors != UINT64_MAX) {
2034 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2037 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2038 queue_id, strerror(error));
2043 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2044 const char *argv[], void *aux OVS_UNUSED)
2046 struct ds ds = DS_EMPTY_INITIALIZER;
2047 struct shash sh = SHASH_INITIALIZER(&sh);
2048 struct iface *iface;
2050 struct shash_node *node;
2051 struct qos_unixctl_show_cbdata data;
2054 iface = iface_find(argv[1]);
2056 unixctl_command_reply(conn, 501, "no such interface");
2060 netdev_get_qos(iface->netdev, &type, &sh);
2062 if (*type != '\0') {
2063 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2065 SHASH_FOR_EACH (node, &sh) {
2066 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2071 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2074 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2076 unixctl_command_reply(conn, 200, ds_cstr(&ds));
2078 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2079 unixctl_command_reply(conn, 501, ds_cstr(&ds));
2082 shash_destroy_free_data(&sh);
2086 /* Bridge reconfiguration functions. */
2088 bridge_create(const struct ovsrec_bridge *br_cfg)
2092 assert(!bridge_lookup(br_cfg->name));
2093 br = xzalloc(sizeof *br);
2095 br->name = xstrdup(br_cfg->name);
2096 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2099 /* Derive the default Ethernet address from the bridge's UUID. This should
2100 * be unique and it will be stable between ovs-vswitchd runs. */
2101 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2102 eth_addr_mark_random(br->default_ea);
2104 hmap_init(&br->ports);
2105 hmap_init(&br->ifaces);
2106 hmap_init(&br->iface_by_name);
2107 hmap_init(&br->mirrors);
2109 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2113 bridge_destroy(struct bridge *br)
2116 struct mirror *mirror, *next_mirror;
2117 struct port *port, *next_port;
2119 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2122 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2123 mirror_destroy(mirror);
2125 hmap_remove(&all_bridges, &br->node);
2126 ofproto_destroy(br->ofproto);
2127 hmap_destroy(&br->ifaces);
2128 hmap_destroy(&br->ports);
2129 hmap_destroy(&br->iface_by_name);
2130 hmap_destroy(&br->mirrors);
2137 static struct bridge *
2138 bridge_lookup(const char *name)
2142 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2143 if (!strcmp(br->name, name)) {
2150 /* Handle requests for a listing of all flows known by the OpenFlow
2151 * stack, including those normally hidden. */
2153 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2154 const char *argv[], void *aux OVS_UNUSED)
2159 br = bridge_lookup(argv[1]);
2161 unixctl_command_reply(conn, 501, "Unknown bridge");
2166 ofproto_get_all_flows(br->ofproto, &results);
2168 unixctl_command_reply(conn, 200, ds_cstr(&results));
2169 ds_destroy(&results);
2172 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2173 * connections and reconnect. If BRIDGE is not specified, then all bridges
2174 * drop their controller connections and reconnect. */
2176 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2177 const char *argv[], void *aux OVS_UNUSED)
2181 br = bridge_lookup(argv[1]);
2183 unixctl_command_reply(conn, 501, "Unknown bridge");
2186 ofproto_reconnect_controllers(br->ofproto);
2188 HMAP_FOR_EACH (br, node, &all_bridges) {
2189 ofproto_reconnect_controllers(br->ofproto);
2192 unixctl_command_reply(conn, 200, NULL);
2196 bridge_get_controllers(const struct bridge *br,
2197 struct ovsrec_controller ***controllersp)
2199 struct ovsrec_controller **controllers;
2200 size_t n_controllers;
2202 controllers = br->cfg->controller;
2203 n_controllers = br->cfg->n_controller;
2205 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2211 *controllersp = controllers;
2213 return n_controllers;
2216 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2217 * those configured in 'br->cfg'. */
2219 bridge_add_del_ports(struct bridge *br,
2220 const unsigned long int *splinter_vlans)
2222 struct port *port, *next;
2223 struct shash_node *node;
2224 struct shash new_ports;
2227 /* Collect new ports. */
2228 shash_init(&new_ports);
2229 for (i = 0; i < br->cfg->n_ports; i++) {
2230 const char *name = br->cfg->ports[i]->name;
2231 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2232 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2236 if (bridge_get_controllers(br, NULL)
2237 && !shash_find(&new_ports, br->name)) {
2238 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2239 br->name, br->name);
2241 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2242 br->synth_local_port.n_interfaces = 1;
2243 br->synth_local_port.name = br->name;
2245 br->synth_local_iface.name = br->name;
2246 br->synth_local_iface.type = "internal";
2248 br->synth_local_ifacep = &br->synth_local_iface;
2250 shash_add(&new_ports, br->name, &br->synth_local_port);
2253 if (splinter_vlans) {
2254 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2257 /* Get rid of deleted ports.
2258 * Get rid of deleted interfaces on ports that still exist. */
2259 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2260 port->cfg = shash_find_data(&new_ports, port->name);
2264 port_del_ifaces(port);
2268 /* Create new ports.
2269 * Add new interfaces to existing ports. */
2270 SHASH_FOR_EACH (node, &new_ports) {
2271 struct port *port = port_lookup(br, node->name);
2273 struct ovsrec_port *cfg = node->data;
2274 port = port_create(br, cfg);
2276 port_add_ifaces(port);
2277 if (list_is_empty(&port->ifaces)) {
2278 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2279 br->name, port->name);
2283 shash_destroy(&new_ports);
2286 /* Initializes 'oc' appropriately as a management service controller for
2289 * The caller must free oc->target when it is no longer needed. */
2291 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2292 struct ofproto_controller *oc)
2294 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2295 oc->max_backoff = 0;
2296 oc->probe_interval = 60;
2297 oc->band = OFPROTO_OUT_OF_BAND;
2299 oc->burst_limit = 0;
2302 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2304 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2305 struct ofproto_controller *oc)
2307 oc->target = c->target;
2308 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2309 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2310 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2311 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2312 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2313 oc->burst_limit = (c->controller_burst_limit
2314 ? *c->controller_burst_limit : 0);
2317 /* Configures the IP stack for 'br''s local interface properly according to the
2318 * configuration in 'c'. */
2320 bridge_configure_local_iface_netdev(struct bridge *br,
2321 struct ovsrec_controller *c)
2323 struct netdev *netdev;
2324 struct in_addr mask, gateway;
2326 struct iface *local_iface;
2329 /* If there's no local interface or no IP address, give up. */
2330 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2331 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2335 /* Bring up the local interface. */
2336 netdev = local_iface->netdev;
2337 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2339 /* Configure the IP address and netmask. */
2340 if (!c->local_netmask
2341 || !inet_aton(c->local_netmask, &mask)
2343 mask.s_addr = guess_netmask(ip.s_addr);
2345 if (!netdev_set_in4(netdev, ip, mask)) {
2346 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2347 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2350 /* Configure the default gateway. */
2351 if (c->local_gateway
2352 && inet_aton(c->local_gateway, &gateway)
2353 && gateway.s_addr) {
2354 if (!netdev_add_router(netdev, gateway)) {
2355 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2356 br->name, IP_ARGS(&gateway.s_addr));
2361 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2362 * in either string are treated as equal to any number of slashes in the other,
2363 * e.g. "x///y" is equal to "x/y". */
2365 equal_pathnames(const char *a, const char *b)
2369 a += strspn(a, "/");
2370 b += strspn(b, "/");
2371 } else if (*a == '\0') {
2382 bridge_configure_remotes(struct bridge *br,
2383 const struct sockaddr_in *managers, size_t n_managers)
2385 const char *disable_ib_str, *queue_id_str;
2386 bool disable_in_band = false;
2389 struct ovsrec_controller **controllers;
2390 size_t n_controllers;
2392 enum ofproto_fail_mode fail_mode;
2394 struct ofproto_controller *ocs;
2398 /* Check if we should disable in-band control on this bridge. */
2399 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2400 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2401 disable_in_band = true;
2404 /* Set OpenFlow queue ID for in-band control. */
2405 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2406 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2407 ofproto_set_in_band_queue(br->ofproto, queue_id);
2409 if (disable_in_band) {
2410 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2412 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2415 n_controllers = bridge_get_controllers(br, &controllers);
2417 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2420 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2421 for (i = 0; i < n_controllers; i++) {
2422 struct ovsrec_controller *c = controllers[i];
2424 if (!strncmp(c->target, "punix:", 6)
2425 || !strncmp(c->target, "unix:", 5)) {
2426 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2429 whitelist = xasprintf("unix:%s/%s.controller",
2430 ovs_rundir(), br->name);
2431 if (!equal_pathnames(c->target, whitelist)) {
2432 /* Prevent remote ovsdb-server users from accessing arbitrary
2433 * Unix domain sockets and overwriting arbitrary local
2435 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2436 "controller \"%s\" due to possibility for remote "
2437 "exploit. Instead, specify whitelisted \"%s\" or "
2438 "connect to \"unix:%s/%s.mgmt\" (which is always "
2439 "available without special configuration).",
2440 br->name, c->target, whitelist,
2441 ovs_rundir(), br->name);
2449 bridge_configure_local_iface_netdev(br, c);
2450 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2451 if (disable_in_band) {
2452 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2457 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2458 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2461 /* Set the fail-mode. */
2462 fail_mode = !br->cfg->fail_mode
2463 || !strcmp(br->cfg->fail_mode, "standalone")
2464 ? OFPROTO_FAIL_STANDALONE
2465 : OFPROTO_FAIL_SECURE;
2466 ofproto_set_fail_mode(br->ofproto, fail_mode);
2468 /* Configure OpenFlow controller connection snooping. */
2469 if (!ofproto_has_snoops(br->ofproto)) {
2473 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2474 ovs_rundir(), br->name));
2475 ofproto_set_snoops(br->ofproto, &snoops);
2476 sset_destroy(&snoops);
2481 bridge_configure_tables(struct bridge *br)
2483 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2487 n_tables = ofproto_get_n_tables(br->ofproto);
2489 for (i = 0; i < n_tables; i++) {
2490 struct ofproto_table_settings s;
2493 s.max_flows = UINT_MAX;
2497 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2498 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2501 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2502 s.max_flows = *cfg->flow_limit;
2504 if (cfg->overflow_policy
2505 && !strcmp(cfg->overflow_policy, "evict")) {
2508 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2509 for (k = 0; k < cfg->n_groups; k++) {
2510 const char *string = cfg->groups[k];
2513 msg = mf_parse_subfield__(&s.groups[k], &string);
2515 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2516 "'groups' (%s)", br->name, i, msg);
2518 } else if (*string) {
2519 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2520 "element '%s' contains trailing garbage",
2521 br->name, i, cfg->groups[k]);
2529 ofproto_configure_table(br->ofproto, i, &s);
2533 for (; j < br->cfg->n_flow_tables; j++) {
2534 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2535 "%"PRId64" not supported by this datapath", br->name,
2536 br->cfg->key_flow_tables[j]);
2540 /* Port functions. */
2542 static struct port *
2543 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2547 port = xzalloc(sizeof *port);
2549 port->name = xstrdup(cfg->name);
2551 list_init(&port->ifaces);
2553 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2555 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2561 get_port_other_config(const struct ovsrec_port *port, const char *key,
2562 const char *default_value)
2566 value = get_ovsrec_key_value(port->key_other_config,
2567 port->value_other_config,
2568 port->n_other_config, key);
2569 return value ? value : default_value;
2573 get_interface_other_config(const struct ovsrec_interface *iface,
2574 const char *key, const char *default_value)
2578 value = get_ovsrec_key_value(iface->key_other_config,
2579 iface->value_other_config,
2580 iface->n_other_config, key);
2581 return value ? value : default_value;
2584 /* Deletes interfaces from 'port' that are no longer configured for it. */
2586 port_del_ifaces(struct port *port)
2588 struct iface *iface, *next;
2589 struct sset new_ifaces;
2592 /* Collect list of new interfaces. */
2593 sset_init(&new_ifaces);
2594 for (i = 0; i < port->cfg->n_interfaces; i++) {
2595 const char *name = port->cfg->interfaces[i]->name;
2596 const char *type = port->cfg->interfaces[i]->name;
2597 if (strcmp(type, "null")) {
2598 sset_add(&new_ifaces, name);
2602 /* Get rid of deleted interfaces. */
2603 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2604 if (!sset_contains(&new_ifaces, iface->name)) {
2605 iface_destroy(iface);
2609 sset_destroy(&new_ifaces);
2612 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2615 port_add_ifaces(struct port *port)
2617 struct shash new_ifaces;
2618 struct shash_node *node;
2621 /* Collect new ifaces. */
2622 shash_init(&new_ifaces);
2623 for (i = 0; i < port->cfg->n_interfaces; i++) {
2624 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2625 if (strcmp(cfg->type, "null")
2626 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2627 VLOG_WARN("port %s: %s specified twice as port interface",
2628 port->name, cfg->name);
2629 iface_clear_db_record(cfg);
2633 /* Create new interfaces.
2634 * Update interface types and 'cfg' members. */
2635 SHASH_FOR_EACH (node, &new_ifaces) {
2636 const struct ovsrec_interface *cfg = node->data;
2637 const char *iface_name = node->name;
2638 struct iface *iface;
2640 iface = iface_lookup(port->bridge, iface_name);
2642 iface = iface_create(port, cfg);
2647 /* Determine interface type. The local port always has type
2648 * "internal". Other ports take their type from the database and
2649 * default to "system" if none is specified. */
2650 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2651 : cfg->type[0] ? cfg->type
2654 shash_destroy(&new_ifaces);
2658 port_destroy(struct port *port)
2661 struct bridge *br = port->bridge;
2662 struct iface *iface, *next;
2665 ofproto_bundle_unregister(br->ofproto, port);
2668 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2669 iface_destroy(iface);
2672 hmap_remove(&br->ports, &port->hmap_node);
2674 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2681 static struct port *
2682 port_lookup(const struct bridge *br, const char *name)
2686 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2688 if (!strcmp(port->name, name)) {
2696 enable_lacp(struct port *port, bool *activep)
2698 if (!port->cfg->lacp) {
2699 /* XXX when LACP implementation has been sufficiently tested, enable by
2700 * default and make active on bonded ports. */
2702 } else if (!strcmp(port->cfg->lacp, "off")) {
2704 } else if (!strcmp(port->cfg->lacp, "active")) {
2707 } else if (!strcmp(port->cfg->lacp, "passive")) {
2711 VLOG_WARN("port %s: unknown LACP mode %s",
2712 port->name, port->cfg->lacp);
2717 static struct lacp_settings *
2718 port_configure_lacp(struct port *port, struct lacp_settings *s)
2720 const char *lacp_time, *system_id;
2721 long long int custom_time;
2724 if (!enable_lacp(port, &s->active)) {
2728 s->name = port->name;
2730 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2732 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2733 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2734 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2735 " address.", port->name, system_id);
2739 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2742 if (eth_addr_is_zero(s->id)) {
2743 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2747 /* Prefer bondable links if unspecified. */
2748 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2750 s->priority = (priority > 0 && priority <= UINT16_MAX
2752 : UINT16_MAX - !list_is_short(&port->ifaces));
2754 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2758 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2759 custom_time = atoi(lacp_time);
2760 if (!strcmp(lacp_time, "fast")) {
2761 s->lacp_time = LACP_TIME_FAST;
2762 } else if (!strcmp(lacp_time, "slow")) {
2763 s->lacp_time = LACP_TIME_SLOW;
2764 } else if (custom_time > 0) {
2765 s->lacp_time = LACP_TIME_CUSTOM;
2766 s->custom_time = custom_time;
2768 s->lacp_time = LACP_TIME_SLOW;
2775 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2777 int priority, portid, key;
2779 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2780 priority = atoi(get_interface_other_config(iface->cfg,
2781 "lacp-port-priority", "0"));
2782 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2785 if (portid <= 0 || portid > UINT16_MAX) {
2786 portid = iface->ofp_port;
2789 if (priority <= 0 || priority > UINT16_MAX) {
2790 priority = UINT16_MAX;
2793 if (key < 0 || key > UINT16_MAX) {
2797 s->name = iface->name;
2799 s->priority = priority;
2804 port_configure_bond(struct port *port, struct bond_settings *s,
2805 uint32_t *bond_stable_ids)
2807 const char *detect_s;
2808 struct iface *iface;
2809 int miimon_interval;
2812 s->name = port->name;
2814 if (port->cfg->bond_mode) {
2815 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2816 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2817 port->name, port->cfg->bond_mode,
2818 bond_mode_to_string(s->balance));
2821 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2823 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2824 * active-backup. At some point we should remove this warning. */
2825 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2826 " in previous versions, the default bond_mode was"
2827 " balance-slb", port->name,
2828 bond_mode_to_string(s->balance));
2830 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2831 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2832 "please use another bond type or disable flood_vlans",
2836 miimon_interval = atoi(get_port_other_config(port->cfg,
2837 "bond-miimon-interval", "0"));
2838 if (miimon_interval <= 0) {
2839 miimon_interval = 200;
2842 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2843 if (!strcmp(detect_s, "carrier")) {
2844 miimon_interval = 0;
2845 } else if (strcmp(detect_s, "miimon")) {
2846 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2847 "defaulting to carrier", port->name, detect_s);
2848 miimon_interval = 0;
2851 s->up_delay = MAX(0, port->cfg->bond_updelay);
2852 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2853 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2854 s->rebalance_interval = atoi(
2855 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2856 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2857 s->rebalance_interval = 1000;
2860 s->fake_iface = port->cfg->bond_fake_iface;
2863 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2864 long long stable_id;
2866 stable_id = atoll(get_interface_other_config(iface->cfg,
2867 "bond-stable-id", "0"));
2868 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2869 stable_id = iface->ofp_port;
2871 bond_stable_ids[i++] = stable_id;
2873 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2877 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2878 * instead of obtaining it from the database. */
2880 port_is_synthetic(const struct port *port)
2882 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2885 /* Interface functions. */
2887 static struct iface *
2888 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2890 struct bridge *br = port->bridge;
2891 struct iface *iface;
2892 char *name = if_cfg->name;
2894 iface = xzalloc(sizeof *iface);
2896 iface->name = xstrdup(name);
2897 iface->ofp_port = -1;
2898 iface->tag = tag_create_random();
2899 iface->netdev = NULL;
2900 iface->cfg = if_cfg;
2902 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2904 list_push_back(&port->ifaces, &iface->port_elem);
2906 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2912 iface_destroy(struct iface *iface)
2915 struct port *port = iface->port;
2916 struct bridge *br = port->bridge;
2918 if (br->ofproto && iface->ofp_port >= 0) {
2919 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2922 if (iface->ofp_port >= 0) {
2923 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2926 list_remove(&iface->port_elem);
2927 hmap_remove(&br->iface_by_name, &iface->name_node);
2929 netdev_close(iface->netdev);
2936 static struct iface *
2937 iface_lookup(const struct bridge *br, const char *name)
2939 struct iface *iface;
2941 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2942 &br->iface_by_name) {
2943 if (!strcmp(iface->name, name)) {
2951 static struct iface *
2952 iface_find(const char *name)
2954 const struct bridge *br;
2956 HMAP_FOR_EACH (br, node, &all_bridges) {
2957 struct iface *iface = iface_lookup(br, name);
2966 static struct iface *
2967 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2969 struct iface *iface;
2971 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2972 hash_int(ofp_port, 0), &br->ifaces) {
2973 if (iface->ofp_port == ofp_port) {
2980 /* Set Ethernet address of 'iface', if one is specified in the configuration
2983 iface_set_mac(struct iface *iface)
2985 uint8_t ea[ETH_ADDR_LEN];
2987 if (!strcmp(iface->type, "internal")
2988 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2989 if (iface->ofp_port == OFPP_LOCAL) {
2990 VLOG_ERR("interface %s: ignoring mac in Interface record "
2991 "(use Bridge record to set local port's mac)",
2993 } else if (eth_addr_is_multicast(ea)) {
2994 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2997 int error = netdev_set_etheraddr(iface->netdev, ea);
2999 VLOG_ERR("interface %s: setting MAC failed (%s)",
3000 iface->name, strerror(error));
3006 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3008 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3010 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3011 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3015 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3016 * sets the "ofport" field to -1.
3018 * This is appropriate when 'if_cfg''s interface cannot be created or is
3019 * otherwise invalid. */
3021 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3023 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3024 iface_set_ofport(if_cfg, -1);
3025 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3026 ovsrec_interface_set_admin_state(if_cfg, NULL);
3027 ovsrec_interface_set_duplex(if_cfg, NULL);
3028 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3029 ovsrec_interface_set_link_state(if_cfg, NULL);
3030 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3031 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3032 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3033 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3034 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3038 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3040 * The value strings in '*shash' are taken directly from values[], not copied,
3041 * so the caller should not modify or free them. */
3043 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3044 struct shash *shash)
3049 for (i = 0; i < n; i++) {
3050 shash_add(shash, keys[i], values[i]);
3054 /* Creates 'keys' and 'values' arrays from 'shash'.
3056 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3057 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3058 * are populated with with strings taken directly from 'shash' and thus have
3059 * the same ownership of the key-value pairs in shash.
3062 shash_to_ovs_idl_map(struct shash *shash,
3063 char ***keys, char ***values, size_t *n)
3067 struct shash_node *sn;
3069 count = shash_count(shash);
3071 k = xmalloc(count * sizeof *k);
3072 v = xmalloc(count * sizeof *v);
3075 SHASH_FOR_EACH(sn, shash) {
3086 struct iface_delete_queues_cbdata {
3087 struct netdev *netdev;
3088 const struct ovsdb_datum *queues;
3092 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3094 union ovsdb_atom atom;
3096 atom.integer = target;
3097 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3101 iface_delete_queues(unsigned int queue_id,
3102 const struct shash *details OVS_UNUSED, void *cbdata_)
3104 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3106 if (!queue_ids_include(cbdata->queues, queue_id)) {
3107 netdev_delete_queue(cbdata->netdev, queue_id);
3112 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3114 struct ofpbuf queues_buf;
3116 ofpbuf_init(&queues_buf, 0);
3118 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3119 netdev_set_qos(iface->netdev, NULL, NULL);
3121 struct iface_delete_queues_cbdata cbdata;
3122 struct shash details;
3126 /* Configure top-level Qos for 'iface'. */
3127 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3128 qos->n_other_config, &details);
3129 netdev_set_qos(iface->netdev, qos->type, &details);
3130 shash_destroy(&details);
3132 /* Deconfigure queues that were deleted. */
3133 cbdata.netdev = iface->netdev;
3134 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3136 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3138 /* Configure queues for 'iface'. */
3140 for (i = 0; i < qos->n_queues; i++) {
3141 const struct ovsrec_queue *queue = qos->value_queues[i];
3142 unsigned int queue_id = qos->key_queues[i];
3144 if (queue_id == 0) {
3148 if (queue->n_dscp == 1) {
3149 struct ofproto_port_queue *port_queue;
3151 port_queue = ofpbuf_put_uninit(&queues_buf,
3152 sizeof *port_queue);
3153 port_queue->queue = queue_id;
3154 port_queue->dscp = queue->dscp[0];
3157 shash_from_ovs_idl_map(queue->key_other_config,
3158 queue->value_other_config,
3159 queue->n_other_config, &details);
3160 netdev_set_queue(iface->netdev, queue_id, &details);
3161 shash_destroy(&details);
3164 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3165 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
3166 "queue (queue 0). Packets not directed to a "
3167 "correctly configured queue may be dropped.",
3172 if (iface->ofp_port >= 0) {
3173 const struct ofproto_port_queue *port_queues = queues_buf.data;
3174 size_t n_queues = queues_buf.size / sizeof *port_queues;
3176 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3177 port_queues, n_queues);
3180 netdev_set_policing(iface->netdev,
3181 iface->cfg->ingress_policing_rate,
3182 iface->cfg->ingress_policing_burst);
3184 ofpbuf_uninit(&queues_buf);
3188 iface_configure_cfm(struct iface *iface)
3190 const struct ovsrec_interface *cfg = iface->cfg;
3191 const char *extended_str, *opstate_str;
3192 struct cfm_settings s;
3194 if (!cfg->n_cfm_mpid) {
3195 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3199 s.mpid = *cfg->cfm_mpid;
3200 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3202 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3204 if (s.interval <= 0) {
3208 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3210 s.extended = !strcasecmp("true", extended_str);
3212 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3213 s.opup = !strcasecmp("up", opstate_str);
3215 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3218 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3219 * instead of obtaining it from the database. */
3221 iface_is_synthetic(const struct iface *iface)
3223 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3226 /* Port mirroring. */
3228 static struct mirror *
3229 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3233 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3234 if (uuid_equals(uuid, &m->uuid)) {
3242 bridge_configure_mirrors(struct bridge *br)
3244 const struct ovsdb_datum *mc;
3245 unsigned long *flood_vlans;
3246 struct mirror *m, *next;
3249 /* Get rid of deleted mirrors. */
3250 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3251 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3252 union ovsdb_atom atom;
3254 atom.uuid = m->uuid;
3255 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3260 /* Add new mirrors and reconfigure existing ones. */
3261 for (i = 0; i < br->cfg->n_mirrors; i++) {
3262 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3263 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3265 m = mirror_create(br, cfg);
3268 if (!mirror_configure(m)) {
3273 /* Update flooded vlans (for RSPAN). */
3274 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3275 br->cfg->n_flood_vlans);
3276 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3277 bitmap_free(flood_vlans);
3280 static struct mirror *
3281 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3285 m = xzalloc(sizeof *m);
3286 m->uuid = cfg->header_.uuid;
3287 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3289 m->name = xstrdup(cfg->name);
3295 mirror_destroy(struct mirror *m)
3298 struct bridge *br = m->bridge;
3301 ofproto_mirror_unregister(br->ofproto, m);
3304 hmap_remove(&br->mirrors, &m->hmap_node);
3311 mirror_collect_ports(struct mirror *m,
3312 struct ovsrec_port **in_ports, int n_in_ports,
3313 void ***out_portsp, size_t *n_out_portsp)
3315 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3316 size_t n_out_ports = 0;
3319 for (i = 0; i < n_in_ports; i++) {
3320 const char *name = in_ports[i]->name;
3321 struct port *port = port_lookup(m->bridge, name);
3323 out_ports[n_out_ports++] = port;
3325 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3326 "port %s", m->bridge->name, m->name, name);
3329 *out_portsp = out_ports;
3330 *n_out_portsp = n_out_ports;
3334 mirror_configure(struct mirror *m)
3336 const struct ovsrec_mirror *cfg = m->cfg;
3337 struct ofproto_mirror_settings s;
3340 if (strcmp(cfg->name, m->name)) {
3342 m->name = xstrdup(cfg->name);
3346 /* Get output port or VLAN. */
3347 if (cfg->output_port) {
3348 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3349 if (!s.out_bundle) {
3350 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3351 m->bridge->name, m->name);
3354 s.out_vlan = UINT16_MAX;
3356 if (cfg->output_vlan) {
3357 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3358 "output vlan; ignoring output vlan",
3359 m->bridge->name, m->name);
3361 } else if (cfg->output_vlan) {
3362 /* The database should prevent invalid VLAN values. */
3363 s.out_bundle = NULL;
3364 s.out_vlan = *cfg->output_vlan;
3366 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3367 m->bridge->name, m->name);
3371 /* Get port selection. */
3372 if (cfg->select_all) {
3373 size_t n_ports = hmap_count(&m->bridge->ports);
3374 void **ports = xmalloc(n_ports * sizeof *ports);
3379 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3389 /* Get ports, dropping ports that don't exist.
3390 * The IDL ensures that there are no duplicates. */
3391 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3392 &s.srcs, &s.n_srcs);
3393 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3394 &s.dsts, &s.n_dsts);
3397 /* Get VLAN selection. */
3398 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3401 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3404 if (s.srcs != s.dsts) {
3413 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3415 * This is deprecated. It is only for compatibility with broken device drivers
3416 * in old versions of Linux that do not properly support VLANs when VLAN
3417 * devices are not used. When broken device drivers are no longer in
3418 * widespread use, we will delete these interfaces. */
3420 static void **blocks;
3421 static size_t n_blocks, allocated_blocks;
3423 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3424 * VLAN splinters are reconfigured. */
3426 register_block(void *block)
3428 if (n_blocks >= allocated_blocks) {
3429 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3431 blocks[n_blocks++] = block;
3434 /* Frees all of the blocks registered with register_block(). */
3436 free_registered_blocks(void)
3440 for (i = 0; i < n_blocks; i++) {
3446 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3449 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3453 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3454 return !strcmp(value, "true");
3457 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3460 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3461 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3462 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3465 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3466 * use, returns NULL.
3468 * Updates 'vlan_splinters_enabled_anywhere'. */
3469 static unsigned long int *
3470 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3472 unsigned long int *splinter_vlans;
3473 struct sset splinter_ifaces;
3474 const char *real_dev_name;
3475 struct shash *real_devs;
3476 struct shash_node *node;
3480 /* Free space allocated for synthesized ports and interfaces, since we're
3481 * in the process of reconstructing all of them. */
3482 free_registered_blocks();
3484 splinter_vlans = bitmap_allocate(4096);
3485 sset_init(&splinter_ifaces);
3486 vlan_splinters_enabled_anywhere = false;
3487 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3488 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3491 for (j = 0; j < br_cfg->n_ports; j++) {
3492 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3495 for (k = 0; k < port_cfg->n_interfaces; k++) {
3496 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3498 if (vlan_splinters_is_enabled(iface_cfg)) {
3499 vlan_splinters_enabled_anywhere = true;
3500 sset_add(&splinter_ifaces, iface_cfg->name);
3501 vlan_bitmap_from_array__(port_cfg->trunks,
3507 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3508 bitmap_set1(splinter_vlans, *port_cfg->tag);
3513 if (!vlan_splinters_enabled_anywhere) {
3514 free(splinter_vlans);
3515 sset_destroy(&splinter_ifaces);
3519 HMAP_FOR_EACH (br, node, &all_bridges) {
3521 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3525 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3526 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3527 * device to be created for it. */
3528 bitmap_set0(splinter_vlans, 0);
3529 bitmap_set0(splinter_vlans, 4095);
3531 /* Delete all VLAN devices that we don't need. */
3533 real_devs = vlandev_get_real_devs();
3534 SHASH_FOR_EACH (node, real_devs) {
3535 const struct vlan_real_dev *real_dev = node->data;
3536 const struct vlan_dev *vlan_dev;
3537 bool real_dev_has_splinters;
3539 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3541 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3542 if (!real_dev_has_splinters
3543 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3544 struct netdev *netdev;
3546 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3547 if (!netdev_get_in4(netdev, NULL, NULL) ||
3548 !netdev_get_in6(netdev, NULL)) {
3549 vlandev_del(vlan_dev->name);
3551 /* It has an IP address configured, so we don't own
3552 * it. Don't delete it. */
3554 netdev_close(netdev);
3561 /* Add all VLAN devices that we need. */
3562 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3565 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3566 if (!vlandev_get_name(real_dev_name, vid)) {
3567 vlandev_add(real_dev_name, vid);
3574 sset_destroy(&splinter_ifaces);
3576 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3577 free(splinter_vlans);
3580 return splinter_vlans;
3583 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3586 configure_splinter_port(struct port *port)
3588 struct ofproto *ofproto = port->bridge->ofproto;
3589 uint16_t realdev_ofp_port;
3590 const char *realdev_name;
3591 struct iface *vlandev, *realdev;
3593 ofproto_bundle_unregister(port->bridge->ofproto, port);
3595 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3598 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3599 realdev = iface_lookup(port->bridge, realdev_name);
3600 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3602 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3606 static struct ovsrec_port *
3607 synthesize_splinter_port(const char *real_dev_name,
3608 const char *vlan_dev_name, int vid)
3610 struct ovsrec_interface *iface;
3611 struct ovsrec_port *port;
3613 iface = xzalloc(sizeof *iface);
3614 iface->name = xstrdup(vlan_dev_name);
3615 iface->type = "system";
3617 port = xzalloc(sizeof *port);
3618 port->interfaces = xmemdup(&iface, sizeof iface);
3619 port->n_interfaces = 1;
3620 port->name = xstrdup(vlan_dev_name);
3621 port->vlan_mode = "splinter";
3622 port->tag = xmalloc(sizeof *port->tag);
3624 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3625 port->key_other_config[0] = "realdev";
3626 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3627 port->value_other_config[0] = xstrdup(real_dev_name);
3628 port->n_other_config = 1;
3630 register_block(iface);
3631 register_block(iface->name);
3632 register_block(port);
3633 register_block(port->interfaces);
3634 register_block(port->name);
3635 register_block(port->tag);
3636 register_block(port->key_other_config);
3637 register_block(port->value_other_config);
3638 register_block(port->value_other_config[0]);
3643 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3644 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3645 * 1-bit in the 'splinter_vlans' bitmap. */
3647 add_vlan_splinter_ports(struct bridge *br,
3648 const unsigned long int *splinter_vlans,
3649 struct shash *ports)
3653 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3654 * we're modifying 'ports'. */
3655 for (i = 0; i < br->cfg->n_ports; i++) {
3656 const char *name = br->cfg->ports[i]->name;
3657 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3660 for (j = 0; j < port_cfg->n_interfaces; j++) {
3661 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3663 if (vlan_splinters_is_enabled(iface_cfg)) {
3664 const char *real_dev_name;
3667 real_dev_name = iface_cfg->name;
3668 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3669 const char *vlan_dev_name;
3671 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3673 && !shash_find(ports, vlan_dev_name)) {
3674 shash_add(ports, vlan_dev_name,
3675 synthesize_splinter_port(
3676 real_dev_name, vlan_dev_name, vid));
3685 mirror_refresh_stats(struct mirror *m)
3687 struct ofproto *ofproto = m->bridge->ofproto;
3688 uint64_t tx_packets, tx_bytes;
3691 size_t stat_cnt = 0;
3693 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3694 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3698 if (tx_packets != UINT64_MAX) {
3699 keys[stat_cnt] = "tx_packets";
3700 values[stat_cnt] = tx_packets;
3703 if (tx_bytes != UINT64_MAX) {
3704 keys[stat_cnt] = "tx_bytes";
3705 values[stat_cnt] = tx_bytes;
3709 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);