1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
44 #include "socket-util.h"
46 #include "stream-ssl.h"
48 #include "system-stats.h"
53 #include "vswitchd/vswitch-idl.h"
54 #include "xenserver.h"
56 #include "sflow_api.h"
57 #include "vlan-bitmap.h"
59 VLOG_DEFINE_THIS_MODULE(bridge);
61 COVERAGE_DEFINE(bridge_reconfigure);
64 /* These members are always valid. */
65 struct list port_elem; /* Element in struct port's "ifaces" list. */
66 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
67 struct port *port; /* Containing port. */
68 char *name; /* Host network device name. */
69 tag_type tag; /* Tag associated with this interface. */
71 /* These members are valid only after bridge_reconfigure() causes them to
73 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
74 int ofp_port; /* OpenFlow port number, -1 if unknown. */
75 struct netdev *netdev; /* Network device. */
76 const char *type; /* Usually same as cfg->type. */
77 const struct ovsrec_interface *cfg;
78 bool need_refresh; /* Refresh iface after create. */
82 struct uuid uuid; /* UUID of this "mirror" record in database. */
83 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
84 struct bridge *bridge;
86 const struct ovsrec_mirror *cfg;
90 struct bridge *bridge;
91 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
94 const struct ovsrec_port *cfg;
96 /* An ordinary bridge port has 1 interface.
97 * A bridge port for bonding has at least 2 interfaces. */
98 struct list ifaces; /* List of "struct iface"s. */
102 struct hmap_node node; /* In 'all_bridges'. */
103 char *name; /* User-specified arbitrary name. */
104 char *type; /* Datapath type. */
105 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
106 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
107 const struct ovsrec_bridge *cfg;
109 /* OpenFlow switch processing. */
110 struct ofproto *ofproto; /* OpenFlow switch. */
113 struct hmap ports; /* "struct port"s indexed by name. */
114 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
115 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
117 /* Port mirroring. */
118 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
120 /* Synthetic local port if necessary. */
121 struct ovsrec_port synth_local_port;
122 struct ovsrec_interface synth_local_iface;
123 struct ovsrec_interface *synth_local_ifacep;
126 /* All bridges, indexed by name. */
127 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
129 /* OVSDB IDL used to obtain configuration. */
130 static struct ovsdb_idl *idl;
132 /* Each time this timer expires, the bridge fetches systems and interface
133 * statistics and pushes them into the database. */
134 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
135 static long long int stats_timer = LLONG_MIN;
137 /* Stores the time after which rate limited statistics may be written to the
138 * database. Only updated when changes to the database require rate limiting.
140 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
141 static long long int db_limiter = LLONG_MIN;
143 static void add_del_bridges(const struct ovsrec_open_vswitch *);
144 static void bridge_del_ofprotos(void);
145 static bool bridge_add_ofprotos(struct bridge *);
146 static void bridge_create(const struct ovsrec_bridge *);
147 static void bridge_destroy(struct bridge *);
148 static struct bridge *bridge_lookup(const char *name);
149 static unixctl_cb_func bridge_unixctl_dump_flows;
150 static unixctl_cb_func bridge_unixctl_reconnect;
151 static size_t bridge_get_controllers(const struct bridge *br,
152 struct ovsrec_controller ***controllersp);
153 static void bridge_add_del_ports(struct bridge *,
154 const unsigned long int *splinter_vlans);
155 static void bridge_add_ofproto_ports(struct bridge *);
156 static void bridge_del_ofproto_ports(struct bridge *);
157 static void bridge_refresh_ofp_port(struct bridge *);
158 static void bridge_configure_datapath_id(struct bridge *);
159 static void bridge_configure_flow_eviction_threshold(struct bridge *);
160 static void bridge_configure_netflow(struct bridge *);
161 static void bridge_configure_forward_bpdu(struct bridge *);
162 static void bridge_configure_mac_idle_time(struct bridge *);
163 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
164 static void bridge_configure_stp(struct bridge *);
165 static void bridge_configure_tables(struct bridge *);
166 static void bridge_configure_remotes(struct bridge *,
167 const struct sockaddr_in *managers,
169 static void bridge_pick_local_hw_addr(struct bridge *,
170 uint8_t ea[ETH_ADDR_LEN],
171 struct iface **hw_addr_iface);
172 static uint64_t bridge_pick_datapath_id(struct bridge *,
173 const uint8_t bridge_ea[ETH_ADDR_LEN],
174 struct iface *hw_addr_iface);
175 static const char *bridge_get_other_config(const struct ovsrec_bridge *,
177 static const char *get_port_other_config(const struct ovsrec_port *,
179 const char *default_value);
180 static uint64_t dpid_from_hash(const void *, size_t nbytes);
181 static bool bridge_has_bond_fake_iface(const struct bridge *,
183 static bool port_is_bond_fake_iface(const struct port *);
185 static unixctl_cb_func qos_unixctl_show;
187 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
188 static void port_add_ifaces(struct port *);
189 static void port_del_ifaces(struct port *);
190 static void port_destroy(struct port *);
191 static struct port *port_lookup(const struct bridge *, const char *name);
192 static void port_configure(struct port *);
193 static struct lacp_settings *port_configure_lacp(struct port *,
194 struct lacp_settings *);
195 static void port_configure_bond(struct port *, struct bond_settings *,
196 uint32_t *bond_stable_ids);
197 static bool port_is_synthetic(const struct port *);
199 static void bridge_configure_mirrors(struct bridge *);
200 static struct mirror *mirror_create(struct bridge *,
201 const struct ovsrec_mirror *);
202 static void mirror_destroy(struct mirror *);
203 static bool mirror_configure(struct mirror *);
204 static void mirror_refresh_stats(struct mirror *);
206 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
207 static struct iface *iface_create(struct port *port,
208 const struct ovsrec_interface *if_cfg);
209 static void iface_destroy(struct iface *);
210 static struct iface *iface_lookup(const struct bridge *, const char *name);
211 static struct iface *iface_find(const char *name);
212 static struct iface *iface_from_ofp_port(const struct bridge *,
214 static void iface_set_mac(struct iface *);
215 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
216 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
217 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
218 static void iface_configure_cfm(struct iface *);
219 static void iface_refresh_cfm_stats(struct iface *);
220 static void iface_refresh_stats(struct iface *);
221 static void iface_refresh_status(struct iface *);
222 static bool iface_is_synthetic(const struct iface *);
223 static const char *get_interface_other_config(const struct ovsrec_interface *,
225 const char *default_value);
227 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
229 static void shash_to_ovs_idl_map(struct shash *,
230 char ***keys, char ***values, size_t *n);
232 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
234 * This is deprecated. It is only for compatibility with broken device drivers
235 * in old versions of Linux that do not properly support VLANs when VLAN
236 * devices are not used. When broken device drivers are no longer in
237 * widespread use, we will delete these interfaces. */
239 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
240 static bool vlan_splinters_enabled_anywhere;
242 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
243 static unsigned long int *collect_splinter_vlans(
244 const struct ovsrec_open_vswitch *);
245 static void configure_splinter_port(struct port *);
246 static void add_vlan_splinter_ports(struct bridge *,
247 const unsigned long int *splinter_vlans,
248 struct shash *ports);
250 /* Public functions. */
252 /* Initializes the bridge module, configuring it to obtain its configuration
253 * from an OVSDB server accessed over 'remote', which should be a string in a
254 * form acceptable to ovsdb_idl_create(). */
256 bridge_init(const char *remote)
258 /* Create connection to database. */
259 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
260 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
262 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
263 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
264 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
265 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
266 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
267 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
268 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
270 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
271 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
272 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
274 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
275 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
276 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
277 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
283 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
284 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
285 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
286 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
287 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
288 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
289 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
290 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
291 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
292 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
294 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
295 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
296 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
297 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
299 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
301 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
303 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
304 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
306 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
308 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
310 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
311 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
312 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
313 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
314 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
316 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
318 /* Register unixctl commands. */
319 unixctl_command_register("qos/show", "interface", 1, 1,
320 qos_unixctl_show, NULL);
321 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
322 bridge_unixctl_dump_flows, NULL);
323 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
324 bridge_unixctl_reconnect, NULL);
334 struct bridge *br, *next_br;
336 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
339 ovsdb_idl_destroy(idl);
342 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
343 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
344 * responsible for freeing '*managersp' (with free()).
346 * You may be asking yourself "why does ovs-vswitchd care?", because
347 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
348 * should not be and in fact is not directly involved in that. But
349 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
350 * it has to tell in-band control where the managers are to enable that.
351 * (Thus, only managers connected in-band are collected.)
354 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
355 struct sockaddr_in **managersp, size_t *n_managersp)
357 struct sockaddr_in *managers = NULL;
358 size_t n_managers = 0;
362 /* Collect all of the potential targets from the "targets" columns of the
363 * rows pointed to by "manager_options", excluding any that are
366 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
367 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
369 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
370 sset_find_and_delete(&targets, m->target);
372 sset_add(&targets, m->target);
376 /* Now extract the targets' IP addresses. */
377 if (!sset_is_empty(&targets)) {
380 managers = xmalloc(sset_count(&targets) * sizeof *managers);
381 SSET_FOR_EACH (target, &targets) {
382 struct sockaddr_in *sin = &managers[n_managers];
384 if (stream_parse_target_with_default_ports(target,
392 sset_destroy(&targets);
394 *managersp = managers;
395 *n_managersp = n_managers;
399 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
401 unsigned long int *splinter_vlans;
402 struct sockaddr_in *managers;
403 struct bridge *br, *next;
404 int sflow_bridge_number;
407 COVERAGE_INC(bridge_reconfigure);
409 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
410 * iface"s according to 'ovs_cfg', with only very minimal configuration
413 * This is mostly an update to bridge data structures. Very little is
414 * pushed down to ofproto or lower layers. */
415 add_del_bridges(ovs_cfg);
416 splinter_vlans = collect_splinter_vlans(ovs_cfg);
417 HMAP_FOR_EACH (br, node, &all_bridges) {
418 bridge_add_del_ports(br, splinter_vlans);
420 free(splinter_vlans);
422 /* Delete all datapaths and datapath ports that are no longer configured.
424 * The kernel will reject any attempt to add a given port to a datapath if
425 * that port already belongs to a different datapath, so we must do all
426 * port deletions before any port additions. A datapath always has a
427 * "local port" so we must delete not-configured datapaths too. */
428 bridge_del_ofprotos();
429 HMAP_FOR_EACH (br, node, &all_bridges) {
431 bridge_del_ofproto_ports(br);
435 /* Create datapaths and datapath ports that are missing.
437 * After this is done, we have our final set of bridges, ports, and
438 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
439 * has at least one iface, every "struct iface" has a valid ofp_port and
441 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
443 if (bridge_add_ofprotos(br)) {
444 bridge_del_ofproto_ports(br);
450 HMAP_FOR_EACH (br, node, &all_bridges) {
451 bridge_refresh_ofp_port(br);
452 bridge_add_ofproto_ports(br);
455 /* Complete the configuration. */
456 sflow_bridge_number = 0;
457 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
458 HMAP_FOR_EACH (br, node, &all_bridges) {
461 /* We need the datapath ID early to allow LACP ports to use it as the
462 * default system ID. */
463 bridge_configure_datapath_id(br);
465 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
468 port_configure(port);
470 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
471 iface_configure_cfm(iface);
472 iface_configure_qos(iface, port->cfg->qos);
473 iface_set_mac(iface);
476 bridge_configure_mirrors(br);
477 bridge_configure_flow_eviction_threshold(br);
478 bridge_configure_forward_bpdu(br);
479 bridge_configure_mac_idle_time(br);
480 bridge_configure_remotes(br, managers, n_managers);
481 bridge_configure_netflow(br);
482 bridge_configure_sflow(br, &sflow_bridge_number);
483 bridge_configure_stp(br);
484 bridge_configure_tables(br);
488 /* ovs-vswitchd has completed initialization, so allow the process that
489 * forked us to exit successfully. */
490 daemonize_complete();
493 /* Iterate over all ofprotos and delete any of them that do not have a
494 * configured bridge or that are the wrong type. */
496 bridge_del_ofprotos(void)
504 ofproto_enumerate_types(&types);
505 SSET_FOR_EACH (type, &types) {
508 ofproto_enumerate_names(type, &names);
509 SSET_FOR_EACH (name, &names) {
510 struct bridge *br = bridge_lookup(name);
511 if (!br || strcmp(type, br->type)) {
512 ofproto_delete(name, type);
516 sset_destroy(&names);
517 sset_destroy(&types);
521 bridge_add_ofprotos(struct bridge *br)
523 int error = ofproto_create(br->name, br->type, &br->ofproto);
525 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
532 port_configure(struct port *port)
534 const struct ovsrec_port *cfg = port->cfg;
535 struct bond_settings bond_settings;
536 struct lacp_settings lacp_settings;
537 struct ofproto_bundle_settings s;
540 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
541 configure_splinter_port(port);
550 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
551 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
552 s.slaves[s.n_slaves++] = iface->ofp_port;
557 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
561 /* Get VLAN trunks. */
564 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
568 if (cfg->vlan_mode) {
569 if (!strcmp(cfg->vlan_mode, "access")) {
570 s.vlan_mode = PORT_VLAN_ACCESS;
571 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
572 s.vlan_mode = PORT_VLAN_TRUNK;
573 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
574 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
575 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
576 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
578 /* This "can't happen" because ovsdb-server should prevent it. */
579 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
580 s.vlan_mode = PORT_VLAN_TRUNK;
584 s.vlan_mode = PORT_VLAN_ACCESS;
586 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
590 s.vlan_mode = PORT_VLAN_TRUNK;
593 s.use_priority_tags = !strcmp("true", get_port_other_config(
594 cfg, "priority-tags", ""));
596 /* Get LACP settings. */
597 s.lacp = port_configure_lacp(port, &lacp_settings);
601 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
602 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
603 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
606 s.lacp_slaves = NULL;
609 /* Get bond settings. */
610 if (s.n_slaves > 1) {
611 s.bond = &bond_settings;
612 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
613 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
616 s.bond_stable_ids = NULL;
618 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
619 netdev_set_miimon_interval(iface->netdev, 0);
624 ofproto_bundle_register(port->bridge->ofproto, port, &s);
630 free(s.bond_stable_ids);
633 /* Pick local port hardware address and datapath ID for 'br'. */
635 bridge_configure_datapath_id(struct bridge *br)
637 uint8_t ea[ETH_ADDR_LEN];
639 struct iface *local_iface;
640 struct iface *hw_addr_iface;
643 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
644 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
646 int error = netdev_set_etheraddr(local_iface->netdev, ea);
648 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
649 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
650 "Ethernet address: %s",
651 br->name, strerror(error));
654 memcpy(br->ea, ea, ETH_ADDR_LEN);
656 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
657 ofproto_set_datapath_id(br->ofproto, dpid);
659 dpid_string = xasprintf("%016"PRIx64, dpid);
660 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
664 /* Set NetFlow configuration on 'br'. */
666 bridge_configure_netflow(struct bridge *br)
668 struct ovsrec_netflow *cfg = br->cfg->netflow;
669 struct netflow_options opts;
672 ofproto_set_netflow(br->ofproto, NULL);
676 memset(&opts, 0, sizeof opts);
678 /* Get default NetFlow configuration from datapath.
679 * Apply overrides from 'cfg'. */
680 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
681 if (cfg->engine_type) {
682 opts.engine_type = *cfg->engine_type;
684 if (cfg->engine_id) {
685 opts.engine_id = *cfg->engine_id;
688 /* Configure active timeout interval. */
689 opts.active_timeout = cfg->active_timeout;
690 if (!opts.active_timeout) {
691 opts.active_timeout = -1;
692 } else if (opts.active_timeout < 0) {
693 VLOG_WARN("bridge %s: active timeout interval set to negative "
694 "value, using default instead (%d seconds)", br->name,
695 NF_ACTIVE_TIMEOUT_DEFAULT);
696 opts.active_timeout = -1;
699 /* Add engine ID to interface number to disambiguate bridgs? */
700 opts.add_id_to_iface = cfg->add_id_to_interface;
701 if (opts.add_id_to_iface) {
702 if (opts.engine_id > 0x7f) {
703 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
704 "another vswitch, choose an engine id less than 128",
707 if (hmap_count(&br->ports) > 508) {
708 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
709 "another port when more than 508 ports are used",
715 sset_init(&opts.collectors);
716 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
719 if (ofproto_set_netflow(br->ofproto, &opts)) {
720 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
722 sset_destroy(&opts.collectors);
725 /* Set sFlow configuration on 'br'. */
727 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
729 const struct ovsrec_sflow *cfg = br->cfg->sflow;
730 struct ovsrec_controller **controllers;
731 struct ofproto_sflow_options oso;
732 size_t n_controllers;
736 ofproto_set_sflow(br->ofproto, NULL);
740 memset(&oso, 0, sizeof oso);
742 sset_init(&oso.targets);
743 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
745 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
747 oso.sampling_rate = *cfg->sampling;
750 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
752 oso.polling_interval = *cfg->polling;
755 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
757 oso.header_len = *cfg->header;
760 oso.sub_id = (*sflow_bridge_number)++;
761 oso.agent_device = cfg->agent;
763 oso.control_ip = NULL;
764 n_controllers = bridge_get_controllers(br, &controllers);
765 for (i = 0; i < n_controllers; i++) {
766 if (controllers[i]->local_ip) {
767 oso.control_ip = controllers[i]->local_ip;
771 ofproto_set_sflow(br->ofproto, &oso);
773 sset_destroy(&oso.targets);
777 port_configure_stp(const struct ofproto *ofproto, struct port *port,
778 struct ofproto_port_stp_settings *port_s,
779 int *port_num_counter, unsigned long *port_num_bitmap)
781 const char *config_str;
784 config_str = get_port_other_config(port->cfg, "stp-enable", NULL);
785 if (config_str && !strcmp(config_str, "false")) {
786 port_s->enable = false;
789 port_s->enable = true;
792 /* STP over bonds is not supported. */
793 if (!list_is_singleton(&port->ifaces)) {
794 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
796 port_s->enable = false;
800 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
802 /* Internal ports shouldn't participate in spanning tree, so
804 if (!strcmp(iface->type, "internal")) {
805 VLOG_DBG("port %s: disable STP on internal ports", port->name);
806 port_s->enable = false;
810 /* STP on mirror output ports is not supported. */
811 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
812 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
813 port_s->enable = false;
817 config_str = get_port_other_config(port->cfg, "stp-port-num", NULL);
819 unsigned long int port_num = strtoul(config_str, NULL, 0);
820 int port_idx = port_num - 1;
822 if (port_num < 1 || port_num > STP_MAX_PORTS) {
823 VLOG_ERR("port %s: invalid stp-port-num", port->name);
824 port_s->enable = false;
828 if (bitmap_is_set(port_num_bitmap, port_idx)) {
829 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
830 port->name, port_num);
831 port_s->enable = false;
834 bitmap_set1(port_num_bitmap, port_idx);
835 port_s->port_num = port_idx;
837 if (*port_num_counter > STP_MAX_PORTS) {
838 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
839 port_s->enable = false;
843 port_s->port_num = (*port_num_counter)++;
846 config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL);
848 port_s->path_cost = strtoul(config_str, NULL, 10);
852 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
853 /* Couldn't get speed, so assume 100Mb/s. */
854 port_s->path_cost = 19;
858 mbps = netdev_features_to_bps(current) / 1000000;
859 port_s->path_cost = stp_convert_speed_to_cost(mbps);
863 config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL);
865 port_s->priority = strtoul(config_str, NULL, 0);
867 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
871 /* Set spanning tree configuration on 'br'. */
873 bridge_configure_stp(struct bridge *br)
875 if (!br->cfg->stp_enable) {
876 ofproto_set_stp(br->ofproto, NULL);
878 struct ofproto_stp_settings br_s;
879 const char *config_str;
881 int port_num_counter;
882 unsigned long *port_num_bitmap;
884 config_str = bridge_get_other_config(br->cfg, "stp-system-id");
886 uint8_t ea[ETH_ADDR_LEN];
888 if (eth_addr_from_string(config_str, ea)) {
889 br_s.system_id = eth_addr_to_uint64(ea);
891 br_s.system_id = eth_addr_to_uint64(br->ea);
892 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
893 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
896 br_s.system_id = eth_addr_to_uint64(br->ea);
899 config_str = bridge_get_other_config(br->cfg, "stp-priority");
901 br_s.priority = strtoul(config_str, NULL, 0);
903 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
906 config_str = bridge_get_other_config(br->cfg, "stp-hello-time");
908 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
910 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
913 config_str = bridge_get_other_config(br->cfg, "stp-max-age");
915 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
917 br_s.max_age = STP_DEFAULT_MAX_AGE;
920 config_str = bridge_get_other_config(br->cfg, "stp-forward-delay");
922 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
924 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
927 /* Configure STP on the bridge. */
928 if (ofproto_set_stp(br->ofproto, &br_s)) {
929 VLOG_ERR("bridge %s: could not enable STP", br->name);
933 /* Users must either set the port number with the "stp-port-num"
934 * configuration on all ports or none. If manual configuration
935 * is not done, then we allocate them sequentially. */
936 port_num_counter = 0;
937 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
938 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
939 struct ofproto_port_stp_settings port_s;
942 port_configure_stp(br->ofproto, port, &port_s,
943 &port_num_counter, port_num_bitmap);
945 /* As bonds are not supported, just apply configuration to
947 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
948 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
950 VLOG_ERR("port %s: could not enable STP", port->name);
956 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
957 && port_num_counter) {
958 VLOG_ERR("bridge %s: must manually configure all STP port "
959 "IDs or none, disabling", br->name);
960 ofproto_set_stp(br->ofproto, NULL);
962 bitmap_free(port_num_bitmap);
967 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
969 const struct port *port = port_lookup(br, name);
970 return port && port_is_bond_fake_iface(port);
974 port_is_bond_fake_iface(const struct port *port)
976 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
980 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
982 struct bridge *br, *next;
986 /* Collect new bridges' names and types. */
988 for (i = 0; i < cfg->n_bridges; i++) {
989 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
990 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
992 if (strchr(br_cfg->name, '/')) {
993 /* Prevent remote ovsdb-server users from accessing arbitrary
994 * directories, e.g. consider a bridge named "../../../etc/". */
995 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
997 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
998 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1002 /* Get rid of deleted bridges or those whose types have changed.
1003 * Update 'cfg' of bridges that still exist. */
1004 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1005 br->cfg = shash_find_data(&new_br, br->name);
1006 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1007 br->cfg->datapath_type))) {
1012 /* Add new bridges. */
1013 for (i = 0; i < cfg->n_bridges; i++) {
1014 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1015 struct bridge *br = bridge_lookup(br_cfg->name);
1017 bridge_create(br_cfg);
1021 shash_destroy(&new_br);
1024 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1027 * The kernel will reject any attempt to add a given port to a datapath if that
1028 * port already belongs to a different datapath, so we must do all port
1029 * deletions before any port additions. */
1031 bridge_del_ofproto_ports(struct bridge *br)
1033 struct ofproto_port_dump dump;
1034 struct ofproto_port ofproto_port;
1036 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1037 const char *name = ofproto_port.name;
1038 struct iface *iface;
1042 /* Ignore the local port. We can't change it anyhow. */
1043 if (!strcmp(name, br->name)) {
1047 /* Get the type that 'ofproto_port' should have (ordinarily the
1048 * type of its corresponding iface) or NULL if it should be
1050 iface = iface_lookup(br, name);
1051 type = (iface ? iface->type
1052 : bridge_has_bond_fake_iface(br, name) ? "internal"
1055 /* If it's the wrong type then delete the ofproto port. */
1057 && !strcmp(ofproto_port.type, type)
1058 && (!iface || !iface->netdev
1059 || !strcmp(netdev_get_type(iface->netdev), type))) {
1062 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1064 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1065 br->name, name, strerror(error));
1068 netdev_close(iface->netdev);
1069 iface->netdev = NULL;
1075 iface_set_ofp_port(struct iface *iface, int ofp_port)
1077 struct bridge *br = iface->port->bridge;
1079 assert(iface->ofp_port < 0 && ofp_port >= 0);
1080 iface->ofp_port = ofp_port;
1081 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1082 iface_set_ofport(iface->cfg, ofp_port);
1086 bridge_refresh_ofp_port(struct bridge *br)
1088 struct ofproto_port_dump dump;
1089 struct ofproto_port ofproto_port;
1092 /* Clear all the "ofp_port"es. */
1093 hmap_clear(&br->ifaces);
1094 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1095 struct iface *iface;
1097 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1098 iface->ofp_port = -1;
1102 /* Obtain the correct "ofp_port"s from ofproto. */
1103 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1104 struct iface *iface = iface_lookup(br, ofproto_port.name);
1106 if (iface->ofp_port >= 0) {
1107 VLOG_WARN("bridge %s: interface %s reported twice",
1108 br->name, ofproto_port.name);
1109 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1110 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1111 br->name, ofproto_port.ofp_port);
1113 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1119 /* Add an ofproto port for any "struct iface" that doesn't have one.
1120 * Delete any "struct iface" for which this fails.
1121 * Delete any "struct port" that thereby ends up with no ifaces. */
1123 bridge_add_ofproto_ports(struct bridge *br)
1125 struct port *port, *next_port;
1127 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1128 struct iface *iface, *next_iface;
1129 struct ofproto_port ofproto_port;
1131 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1134 /* Open the netdev. */
1135 if (!iface->netdev) {
1136 error = netdev_open(iface->name, iface->type, &iface->netdev);
1138 VLOG_WARN("could not open network device %s (%s)",
1139 iface->name, strerror(error));
1143 && port->cfg->vlan_mode
1144 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1145 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1151 /* Configure the netdev. */
1152 if (iface->netdev) {
1156 shash_from_ovs_idl_map(iface->cfg->key_options,
1157 iface->cfg->value_options,
1158 iface->cfg->n_options, &args);
1159 error = netdev_set_config(iface->netdev, &args);
1160 shash_destroy(&args);
1163 VLOG_WARN("could not configure network device %s (%s)",
1164 iface->name, strerror(error));
1165 netdev_close(iface->netdev);
1166 iface->netdev = NULL;
1170 /* Add the port, if necessary. */
1171 if (iface->netdev && iface->ofp_port < 0) {
1175 error = ofproto_port_add(br->ofproto, iface->netdev,
1178 iface_set_ofp_port(iface, ofp_port);
1180 netdev_close(iface->netdev);
1181 iface->netdev = NULL;
1185 /* Populate stats columns in new Interface rows. */
1186 if (iface->netdev && iface->need_refresh) {
1187 iface_refresh_stats(iface);
1188 iface_refresh_status(iface);
1189 iface->need_refresh = false;
1192 /* Delete the iface if we failed. */
1193 if (iface->netdev && iface->ofp_port >= 0) {
1194 VLOG_DBG("bridge %s: interface %s is on port %d",
1195 br->name, iface->name, iface->ofp_port);
1197 if (iface->netdev) {
1198 VLOG_ERR("bridge %s: missing %s interface, dropping",
1199 br->name, iface->name);
1201 /* We already reported a related error, don't bother
1202 * duplicating it. */
1204 iface_clear_db_record(iface->cfg);
1205 iface_destroy(iface);
1208 if (list_is_empty(&port->ifaces)) {
1209 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1214 /* Add bond fake iface if necessary. */
1215 if (port_is_bond_fake_iface(port)) {
1216 if (ofproto_port_query_by_name(br->ofproto, port->name,
1218 struct netdev *netdev;
1221 error = netdev_open(port->name, "internal", &netdev);
1223 ofproto_port_add(br->ofproto, netdev, NULL);
1224 netdev_close(netdev);
1226 VLOG_WARN("could not open network device %s (%s)",
1227 port->name, strerror(error));
1230 /* Already exists, nothing to do. */
1231 ofproto_port_destroy(&ofproto_port);
1238 get_ovsrec_key_value(char **keys, char **values, size_t n, const char *key)
1242 for (i = 0; i < n; i++) {
1243 if (!strcmp(keys[i], key)) {
1251 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
1253 return get_ovsrec_key_value(br_cfg->key_other_config,
1254 br_cfg->value_other_config,
1255 br_cfg->n_other_config, key);
1258 /* Set Flow eviction threshold */
1260 bridge_configure_flow_eviction_threshold(struct bridge *br)
1262 const char *threshold_str;
1265 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
1266 if (threshold_str) {
1267 threshold = strtoul(threshold_str, NULL, 10);
1269 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1271 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1274 /* Set forward BPDU option. */
1276 bridge_configure_forward_bpdu(struct bridge *br)
1278 const char *forward_bpdu_str;
1279 bool forward_bpdu = false;
1281 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1282 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1283 forward_bpdu = true;
1285 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1288 /* Set MAC aging time for 'br'. */
1290 bridge_configure_mac_idle_time(struct bridge *br)
1292 const char *idle_time_str;
1295 idle_time_str = bridge_get_other_config(br->cfg, "mac-aging-time");
1296 idle_time = (idle_time_str && atoi(idle_time_str)
1297 ? atoi(idle_time_str)
1298 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1299 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1303 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1304 struct iface **hw_addr_iface)
1306 struct hmapx mirror_output_ports;
1309 bool found_addr = false;
1313 *hw_addr_iface = NULL;
1315 /* Did the user request a particular MAC? */
1316 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1317 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1318 if (eth_addr_is_multicast(ea)) {
1319 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1320 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1321 } else if (eth_addr_is_zero(ea)) {
1322 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1328 /* Mirror output ports don't participate in picking the local hardware
1329 * address. ofproto can't help us find out whether a given port is a
1330 * mirror output because we haven't configured mirrors yet, so we need to
1331 * accumulate them ourselves. */
1332 hmapx_init(&mirror_output_ports);
1333 for (i = 0; i < br->cfg->n_mirrors; i++) {
1334 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1335 if (m->output_port) {
1336 hmapx_add(&mirror_output_ports, m->output_port);
1340 /* Otherwise choose the minimum non-local MAC address among all of the
1342 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1343 uint8_t iface_ea[ETH_ADDR_LEN];
1344 struct iface *candidate;
1345 struct iface *iface;
1347 /* Mirror output ports don't participate. */
1348 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1352 /* Choose the MAC address to represent the port. */
1354 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1355 /* Find the interface with this Ethernet address (if any) so that
1356 * we can provide the correct devname to the caller. */
1357 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1358 uint8_t candidate_ea[ETH_ADDR_LEN];
1359 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1360 && eth_addr_equals(iface_ea, candidate_ea)) {
1365 /* Choose the interface whose MAC address will represent the port.
1366 * The Linux kernel bonding code always chooses the MAC address of
1367 * the first slave added to a bond, and the Fedora networking
1368 * scripts always add slaves to a bond in alphabetical order, so
1369 * for compatibility we choose the interface with the name that is
1370 * first in alphabetical order. */
1371 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1372 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1377 /* The local port doesn't count (since we're trying to choose its
1378 * MAC address anyway). */
1379 if (iface->ofp_port == OFPP_LOCAL) {
1384 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1390 /* Compare against our current choice. */
1391 if (!eth_addr_is_multicast(iface_ea) &&
1392 !eth_addr_is_local(iface_ea) &&
1393 !eth_addr_is_reserved(iface_ea) &&
1394 !eth_addr_is_zero(iface_ea) &&
1395 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1397 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1398 *hw_addr_iface = iface;
1403 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1404 br->name, ETH_ADDR_ARGS(ea));
1406 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1407 *hw_addr_iface = NULL;
1408 VLOG_WARN("bridge %s: using default bridge Ethernet "
1409 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1412 hmapx_destroy(&mirror_output_ports);
1415 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1416 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1417 * an interface on 'br', then that interface must be passed in as
1418 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1419 * 'hw_addr_iface' must be passed in as a null pointer. */
1421 bridge_pick_datapath_id(struct bridge *br,
1422 const uint8_t bridge_ea[ETH_ADDR_LEN],
1423 struct iface *hw_addr_iface)
1426 * The procedure for choosing a bridge MAC address will, in the most
1427 * ordinary case, also choose a unique MAC that we can use as a datapath
1428 * ID. In some special cases, though, multiple bridges will end up with
1429 * the same MAC address. This is OK for the bridges, but it will confuse
1430 * the OpenFlow controller, because each datapath needs a unique datapath
1433 * Datapath IDs must be unique. It is also very desirable that they be
1434 * stable from one run to the next, so that policy set on a datapath
1437 const char *datapath_id;
1440 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1441 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1445 if (!hw_addr_iface) {
1447 * A purely internal bridge, that is, one that has no non-virtual
1448 * network devices on it at all, is difficult because it has no
1449 * natural unique identifier at all.
1451 * When the host is a XenServer, we handle this case by hashing the
1452 * host's UUID with the name of the bridge. Names of bridges are
1453 * persistent across XenServer reboots, although they can be reused if
1454 * an internal network is destroyed and then a new one is later
1455 * created, so this is fairly effective.
1457 * When the host is not a XenServer, we punt by using a random MAC
1458 * address on each run.
1460 const char *host_uuid = xenserver_get_host_uuid();
1462 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1463 dpid = dpid_from_hash(combined, strlen(combined));
1469 return eth_addr_to_uint64(bridge_ea);
1473 dpid_from_hash(const void *data, size_t n)
1475 uint8_t hash[SHA1_DIGEST_SIZE];
1477 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1478 sha1_bytes(data, n, hash);
1479 eth_addr_mark_random(hash);
1480 return eth_addr_to_uint64(hash);
1484 iface_refresh_status(struct iface *iface)
1488 enum netdev_flags flags;
1495 if (iface_is_synthetic(iface)) {
1501 if (!netdev_get_status(iface->netdev, &sh)) {
1503 char **keys, **values;
1505 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1506 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1511 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1514 shash_destroy_free_data(&sh);
1516 error = netdev_get_flags(iface->netdev, &flags);
1518 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1521 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1524 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1526 ovsrec_interface_set_duplex(iface->cfg,
1527 netdev_features_is_full_duplex(current)
1529 /* warning: uint64_t -> int64_t conversion */
1530 bps = netdev_features_to_bps(current);
1531 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1534 ovsrec_interface_set_duplex(iface->cfg, NULL);
1535 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1538 error = netdev_get_mtu(iface->netdev, &mtu);
1541 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1544 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1548 /* Writes 'iface''s CFM statistics to the database. */
1550 iface_refresh_cfm_stats(struct iface *iface)
1552 const struct ovsrec_interface *cfg = iface->cfg;
1554 const uint64_t *rmps;
1557 if (iface_is_synthetic(iface)) {
1561 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1564 const char *reasons[CFM_FAULT_N_REASONS];
1565 bool fault_bool = fault;
1569 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1570 int reason = 1 << i;
1571 if (fault & reason) {
1572 reasons[j++] = cfm_fault_reason_to_str(reason);
1576 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1577 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1579 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1580 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1583 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1584 iface->ofp_port, &rmps, &n_rmps);
1586 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1589 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1594 iface_refresh_stats(struct iface *iface)
1596 #define IFACE_STATS \
1597 IFACE_STAT(rx_packets, "rx_packets") \
1598 IFACE_STAT(tx_packets, "tx_packets") \
1599 IFACE_STAT(rx_bytes, "rx_bytes") \
1600 IFACE_STAT(tx_bytes, "tx_bytes") \
1601 IFACE_STAT(rx_dropped, "rx_dropped") \
1602 IFACE_STAT(tx_dropped, "tx_dropped") \
1603 IFACE_STAT(rx_errors, "rx_errors") \
1604 IFACE_STAT(tx_errors, "tx_errors") \
1605 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1606 IFACE_STAT(rx_over_errors, "rx_over_err") \
1607 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1608 IFACE_STAT(collisions, "collisions")
1610 #define IFACE_STAT(MEMBER, NAME) NAME,
1611 static char *keys[] = { IFACE_STATS };
1613 int64_t values[ARRAY_SIZE(keys)];
1616 struct netdev_stats stats;
1618 if (iface_is_synthetic(iface)) {
1622 /* Intentionally ignore return value, since errors will set 'stats' to
1623 * all-1s, and we will deal with that correctly below. */
1624 netdev_get_stats(iface->netdev, &stats);
1626 /* Copy statistics into values[] array. */
1628 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1631 assert(i == ARRAY_SIZE(keys));
1633 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1638 br_refresh_stp_status(struct bridge *br)
1640 struct ofproto *ofproto = br->ofproto;
1641 struct ofproto_stp_status status;
1642 char *keys[3], *values[3];
1645 if (ofproto_get_stp_status(ofproto, &status)) {
1649 if (!status.enabled) {
1650 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1654 keys[0] = "stp_bridge_id",
1655 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1656 keys[1] = "stp_designated_root",
1657 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1658 keys[2] = "stp_root_path_cost",
1659 values[2] = xasprintf("%d", status.root_path_cost);
1661 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1663 for (i = 0; i < ARRAY_SIZE(values); i++) {
1669 port_refresh_stp_status(struct port *port)
1671 struct ofproto *ofproto = port->bridge->ofproto;
1672 struct iface *iface;
1673 struct ofproto_port_stp_status status;
1675 char *str_values[4];
1676 int64_t int_values[3];
1679 if (port_is_synthetic(port)) {
1683 /* STP doesn't currently support bonds. */
1684 if (!list_is_singleton(&port->ifaces)) {
1685 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1689 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1691 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1695 if (!status.enabled) {
1696 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1697 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1701 /* Set Status column. */
1702 keys[0] = "stp_port_id";
1703 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1704 keys[1] = "stp_state";
1705 str_values[1] = xstrdup(stp_state_name(status.state));
1706 keys[2] = "stp_sec_in_state";
1707 str_values[2] = xasprintf("%u", status.sec_in_state);
1708 keys[3] = "stp_role";
1709 str_values[3] = xstrdup(stp_role_name(status.role));
1711 ovsrec_port_set_status(port->cfg, keys, str_values,
1712 ARRAY_SIZE(str_values));
1714 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1715 free(str_values[i]);
1718 /* Set Statistics column. */
1719 keys[0] = "stp_tx_count";
1720 int_values[0] = status.tx_count;
1721 keys[1] = "stp_rx_count";
1722 int_values[1] = status.rx_count;
1723 keys[2] = "stp_error_count";
1724 int_values[2] = status.error_count;
1726 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1727 ARRAY_SIZE(int_values));
1731 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1735 /* Use other-config:enable-system-stats by preference. */
1736 enable = get_ovsrec_key_value(cfg->key_other_config,
1737 cfg->value_other_config,
1738 cfg->n_other_config,
1739 "enable-statistics");
1741 return !strcmp(enable, "true");
1744 /* Disable by default. */
1749 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1751 struct ovsdb_datum datum;
1755 if (enable_system_stats(cfg)) {
1756 get_system_stats(&stats);
1759 ovsdb_datum_from_shash(&datum, &stats);
1760 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1764 static inline const char *
1765 nx_role_to_str(enum nx_role role)
1770 case NX_ROLE_MASTER:
1775 return "*** INVALID ROLE ***";
1780 refresh_controller_status(void)
1784 const struct ovsrec_controller *cfg;
1788 /* Accumulate status for controllers on all bridges. */
1789 HMAP_FOR_EACH (br, node, &all_bridges) {
1790 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1793 /* Update each controller in the database with current status. */
1794 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1795 struct ofproto_controller_info *cinfo =
1796 shash_find_data(&info, cfg->target);
1799 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1800 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1801 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1802 (char **) cinfo->pairs.values,
1805 ovsrec_controller_set_is_connected(cfg, false);
1806 ovsrec_controller_set_role(cfg, NULL);
1807 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1811 ofproto_free_ofproto_controller_info(&info);
1815 refresh_cfm_stats(void)
1817 static struct ovsdb_idl_txn *txn = NULL;
1822 txn = ovsdb_idl_txn_create(idl);
1824 HMAP_FOR_EACH (br, node, &all_bridges) {
1825 struct iface *iface;
1827 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1828 iface_refresh_cfm_stats(iface);
1833 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1834 ovsdb_idl_txn_destroy(txn);
1839 /* Performs periodic activity required by bridges that needs to be done with
1840 * the least possible latency.
1842 * It makes sense to call this function a couple of times per poll loop, to
1843 * provide a significant performance boost on some benchmarks with ofprotos
1844 * that use the ofproto-dpif implementation. */
1846 bridge_run_fast(void)
1850 HMAP_FOR_EACH (br, node, &all_bridges) {
1851 ofproto_run_fast(br->ofproto);
1858 const struct ovsrec_open_vswitch *cfg;
1860 bool vlan_splinters_changed;
1861 bool database_changed;
1864 /* (Re)configure if necessary. */
1865 database_changed = ovsdb_idl_run(idl);
1866 if (ovsdb_idl_is_lock_contended(idl)) {
1867 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1868 struct bridge *br, *next_br;
1870 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1871 "disabling this process until it goes away");
1873 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1877 } else if (!ovsdb_idl_has_lock(idl)) {
1880 cfg = ovsrec_open_vswitch_first(idl);
1882 /* Let each bridge do the work that it needs to do. */
1883 HMAP_FOR_EACH (br, node, &all_bridges) {
1884 ofproto_run(br->ofproto);
1887 /* Re-configure SSL. We do this on every trip through the main loop,
1888 * instead of just when the database changes, because the contents of the
1889 * key and certificate files can change without the database changing.
1891 * We do this before bridge_reconfigure() because that function might
1892 * initiate SSL connections and thus requires SSL to be configured. */
1893 if (cfg && cfg->ssl) {
1894 const struct ovsrec_ssl *ssl = cfg->ssl;
1896 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1897 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1900 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1902 vlan_splinters_changed = false;
1903 if (vlan_splinters_enabled_anywhere) {
1904 HMAP_FOR_EACH (br, node, &all_bridges) {
1905 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1906 vlan_splinters_changed = true;
1912 if (database_changed || vlan_splinters_changed) {
1914 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1916 bridge_reconfigure(cfg);
1918 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1919 ovsdb_idl_txn_commit(txn);
1920 ovsdb_idl_txn_destroy(txn); /* XXX */
1922 /* We still need to reconfigure to avoid dangling pointers to
1923 * now-destroyed ovsrec structures inside bridge data. */
1924 static const struct ovsrec_open_vswitch null_cfg;
1926 bridge_reconfigure(&null_cfg);
1930 /* Refresh system and interface stats if necessary. */
1931 if (time_msec() >= stats_timer) {
1933 struct ovsdb_idl_txn *txn;
1935 txn = ovsdb_idl_txn_create(idl);
1936 HMAP_FOR_EACH (br, node, &all_bridges) {
1940 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1941 struct iface *iface;
1943 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1944 iface_refresh_stats(iface);
1945 iface_refresh_status(iface);
1949 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1950 mirror_refresh_stats(m);
1954 refresh_system_stats(cfg);
1955 refresh_controller_status();
1956 ovsdb_idl_txn_commit(txn);
1957 ovsdb_idl_txn_destroy(txn); /* XXX */
1960 stats_timer = time_msec() + STATS_INTERVAL;
1963 if (time_msec() >= db_limiter) {
1964 struct ovsdb_idl_txn *txn;
1966 txn = ovsdb_idl_txn_create(idl);
1967 HMAP_FOR_EACH (br, node, &all_bridges) {
1968 struct iface *iface;
1971 br_refresh_stp_status(br);
1973 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1974 port_refresh_stp_status(port);
1977 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1978 const char *link_state;
1979 int64_t link_resets;
1982 if (iface_is_synthetic(iface)) {
1986 current = ofproto_port_is_lacp_current(br->ofproto,
1990 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1992 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1995 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1996 ovsrec_interface_set_link_state(iface->cfg, link_state);
1998 link_resets = netdev_get_carrier_resets(iface->netdev);
1999 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2003 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2004 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2006 ovsdb_idl_txn_destroy(txn);
2009 refresh_cfm_stats();
2015 ovsdb_idl_wait(idl);
2016 if (!hmap_is_empty(&all_bridges)) {
2019 HMAP_FOR_EACH (br, node, &all_bridges) {
2020 ofproto_wait(br->ofproto);
2022 poll_timer_wait_until(stats_timer);
2024 if (db_limiter > time_msec()) {
2025 poll_timer_wait_until(db_limiter);
2030 /* QoS unixctl user interface functions. */
2032 struct qos_unixctl_show_cbdata {
2034 struct iface *iface;
2038 qos_unixctl_show_cb(unsigned int queue_id,
2039 const struct shash *details,
2042 struct qos_unixctl_show_cbdata *data = aux;
2043 struct ds *ds = data->ds;
2044 struct iface *iface = data->iface;
2045 struct netdev_queue_stats stats;
2046 struct shash_node *node;
2049 ds_put_cstr(ds, "\n");
2051 ds_put_format(ds, "Queue %u:\n", queue_id);
2053 ds_put_cstr(ds, "Default:\n");
2056 SHASH_FOR_EACH (node, details) {
2057 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2060 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2062 if (stats.tx_packets != UINT64_MAX) {
2063 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2066 if (stats.tx_bytes != UINT64_MAX) {
2067 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2070 if (stats.tx_errors != UINT64_MAX) {
2071 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2074 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2075 queue_id, strerror(error));
2080 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2081 const char *argv[], void *aux OVS_UNUSED)
2083 struct ds ds = DS_EMPTY_INITIALIZER;
2084 struct shash sh = SHASH_INITIALIZER(&sh);
2085 struct iface *iface;
2087 struct shash_node *node;
2088 struct qos_unixctl_show_cbdata data;
2091 iface = iface_find(argv[1]);
2093 unixctl_command_reply_error(conn, "no such interface");
2097 netdev_get_qos(iface->netdev, &type, &sh);
2099 if (*type != '\0') {
2100 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2102 SHASH_FOR_EACH (node, &sh) {
2103 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2108 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2111 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2113 unixctl_command_reply(conn, ds_cstr(&ds));
2115 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2116 unixctl_command_reply_error(conn, ds_cstr(&ds));
2119 shash_destroy_free_data(&sh);
2123 /* Bridge reconfiguration functions. */
2125 bridge_create(const struct ovsrec_bridge *br_cfg)
2129 assert(!bridge_lookup(br_cfg->name));
2130 br = xzalloc(sizeof *br);
2132 br->name = xstrdup(br_cfg->name);
2133 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2136 /* Derive the default Ethernet address from the bridge's UUID. This should
2137 * be unique and it will be stable between ovs-vswitchd runs. */
2138 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2139 eth_addr_mark_random(br->default_ea);
2141 hmap_init(&br->ports);
2142 hmap_init(&br->ifaces);
2143 hmap_init(&br->iface_by_name);
2144 hmap_init(&br->mirrors);
2146 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2150 bridge_destroy(struct bridge *br)
2153 struct mirror *mirror, *next_mirror;
2154 struct port *port, *next_port;
2156 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2159 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2160 mirror_destroy(mirror);
2162 hmap_remove(&all_bridges, &br->node);
2163 ofproto_destroy(br->ofproto);
2164 hmap_destroy(&br->ifaces);
2165 hmap_destroy(&br->ports);
2166 hmap_destroy(&br->iface_by_name);
2167 hmap_destroy(&br->mirrors);
2174 static struct bridge *
2175 bridge_lookup(const char *name)
2179 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2180 if (!strcmp(br->name, name)) {
2187 /* Handle requests for a listing of all flows known by the OpenFlow
2188 * stack, including those normally hidden. */
2190 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2191 const char *argv[], void *aux OVS_UNUSED)
2196 br = bridge_lookup(argv[1]);
2198 unixctl_command_reply_error(conn, "Unknown bridge");
2203 ofproto_get_all_flows(br->ofproto, &results);
2205 unixctl_command_reply(conn, ds_cstr(&results));
2206 ds_destroy(&results);
2209 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2210 * connections and reconnect. If BRIDGE is not specified, then all bridges
2211 * drop their controller connections and reconnect. */
2213 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2214 const char *argv[], void *aux OVS_UNUSED)
2218 br = bridge_lookup(argv[1]);
2220 unixctl_command_reply_error(conn, "Unknown bridge");
2223 ofproto_reconnect_controllers(br->ofproto);
2225 HMAP_FOR_EACH (br, node, &all_bridges) {
2226 ofproto_reconnect_controllers(br->ofproto);
2229 unixctl_command_reply(conn, NULL);
2233 bridge_get_controllers(const struct bridge *br,
2234 struct ovsrec_controller ***controllersp)
2236 struct ovsrec_controller **controllers;
2237 size_t n_controllers;
2239 controllers = br->cfg->controller;
2240 n_controllers = br->cfg->n_controller;
2242 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2248 *controllersp = controllers;
2250 return n_controllers;
2253 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2254 * those configured in 'br->cfg'. */
2256 bridge_add_del_ports(struct bridge *br,
2257 const unsigned long int *splinter_vlans)
2259 struct port *port, *next;
2260 struct shash_node *node;
2261 struct shash new_ports;
2264 /* Collect new ports. */
2265 shash_init(&new_ports);
2266 for (i = 0; i < br->cfg->n_ports; i++) {
2267 const char *name = br->cfg->ports[i]->name;
2268 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2269 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2273 if (bridge_get_controllers(br, NULL)
2274 && !shash_find(&new_ports, br->name)) {
2275 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2276 br->name, br->name);
2278 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2279 br->synth_local_port.n_interfaces = 1;
2280 br->synth_local_port.name = br->name;
2282 br->synth_local_iface.name = br->name;
2283 br->synth_local_iface.type = "internal";
2285 br->synth_local_ifacep = &br->synth_local_iface;
2287 shash_add(&new_ports, br->name, &br->synth_local_port);
2290 if (splinter_vlans) {
2291 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2294 /* Get rid of deleted ports.
2295 * Get rid of deleted interfaces on ports that still exist. */
2296 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2297 port->cfg = shash_find_data(&new_ports, port->name);
2301 port_del_ifaces(port);
2305 /* Create new ports.
2306 * Add new interfaces to existing ports. */
2307 SHASH_FOR_EACH (node, &new_ports) {
2308 struct port *port = port_lookup(br, node->name);
2310 struct ovsrec_port *cfg = node->data;
2311 port = port_create(br, cfg);
2313 port_add_ifaces(port);
2314 if (list_is_empty(&port->ifaces)) {
2315 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2316 br->name, port->name);
2320 shash_destroy(&new_ports);
2323 /* Initializes 'oc' appropriately as a management service controller for
2326 * The caller must free oc->target when it is no longer needed. */
2328 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2329 struct ofproto_controller *oc)
2331 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2332 oc->max_backoff = 0;
2333 oc->probe_interval = 60;
2334 oc->band = OFPROTO_OUT_OF_BAND;
2336 oc->burst_limit = 0;
2337 oc->enable_async_msgs = true;
2340 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2342 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2343 struct ofproto_controller *oc)
2345 oc->target = c->target;
2346 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2347 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2348 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2349 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2350 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2351 oc->burst_limit = (c->controller_burst_limit
2352 ? *c->controller_burst_limit : 0);
2353 oc->enable_async_msgs = (!c->enable_async_messages
2354 || *c->enable_async_messages);
2357 /* Configures the IP stack for 'br''s local interface properly according to the
2358 * configuration in 'c'. */
2360 bridge_configure_local_iface_netdev(struct bridge *br,
2361 struct ovsrec_controller *c)
2363 struct netdev *netdev;
2364 struct in_addr mask, gateway;
2366 struct iface *local_iface;
2369 /* If there's no local interface or no IP address, give up. */
2370 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2371 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2375 /* Bring up the local interface. */
2376 netdev = local_iface->netdev;
2377 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2379 /* Configure the IP address and netmask. */
2380 if (!c->local_netmask
2381 || !inet_aton(c->local_netmask, &mask)
2383 mask.s_addr = guess_netmask(ip.s_addr);
2385 if (!netdev_set_in4(netdev, ip, mask)) {
2386 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2387 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2390 /* Configure the default gateway. */
2391 if (c->local_gateway
2392 && inet_aton(c->local_gateway, &gateway)
2393 && gateway.s_addr) {
2394 if (!netdev_add_router(netdev, gateway)) {
2395 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2396 br->name, IP_ARGS(&gateway.s_addr));
2401 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2402 * in either string are treated as equal to any number of slashes in the other,
2403 * e.g. "x///y" is equal to "x/y". */
2405 equal_pathnames(const char *a, const char *b)
2409 a += strspn(a, "/");
2410 b += strspn(b, "/");
2411 } else if (*a == '\0') {
2422 bridge_configure_remotes(struct bridge *br,
2423 const struct sockaddr_in *managers, size_t n_managers)
2425 const char *disable_ib_str, *queue_id_str;
2426 bool disable_in_band = false;
2429 struct ovsrec_controller **controllers;
2430 size_t n_controllers;
2432 enum ofproto_fail_mode fail_mode;
2434 struct ofproto_controller *ocs;
2438 /* Check if we should disable in-band control on this bridge. */
2439 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2440 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2441 disable_in_band = true;
2444 /* Set OpenFlow queue ID for in-band control. */
2445 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2446 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2447 ofproto_set_in_band_queue(br->ofproto, queue_id);
2449 if (disable_in_band) {
2450 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2452 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2455 n_controllers = bridge_get_controllers(br, &controllers);
2457 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2460 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2461 for (i = 0; i < n_controllers; i++) {
2462 struct ovsrec_controller *c = controllers[i];
2464 if (!strncmp(c->target, "punix:", 6)
2465 || !strncmp(c->target, "unix:", 5)) {
2466 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2469 whitelist = xasprintf("unix:%s/%s.controller",
2470 ovs_rundir(), br->name);
2471 if (!equal_pathnames(c->target, whitelist)) {
2472 /* Prevent remote ovsdb-server users from accessing arbitrary
2473 * Unix domain sockets and overwriting arbitrary local
2475 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2476 "controller \"%s\" due to possibility for remote "
2477 "exploit. Instead, specify whitelisted \"%s\" or "
2478 "connect to \"unix:%s/%s.mgmt\" (which is always "
2479 "available without special configuration).",
2480 br->name, c->target, whitelist,
2481 ovs_rundir(), br->name);
2489 bridge_configure_local_iface_netdev(br, c);
2490 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2491 if (disable_in_band) {
2492 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2497 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2498 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2501 /* Set the fail-mode. */
2502 fail_mode = !br->cfg->fail_mode
2503 || !strcmp(br->cfg->fail_mode, "standalone")
2504 ? OFPROTO_FAIL_STANDALONE
2505 : OFPROTO_FAIL_SECURE;
2506 ofproto_set_fail_mode(br->ofproto, fail_mode);
2508 /* Configure OpenFlow controller connection snooping. */
2509 if (!ofproto_has_snoops(br->ofproto)) {
2513 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2514 ovs_rundir(), br->name));
2515 ofproto_set_snoops(br->ofproto, &snoops);
2516 sset_destroy(&snoops);
2521 bridge_configure_tables(struct bridge *br)
2523 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2527 n_tables = ofproto_get_n_tables(br->ofproto);
2529 for (i = 0; i < n_tables; i++) {
2530 struct ofproto_table_settings s;
2533 s.max_flows = UINT_MAX;
2537 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2538 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2541 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2542 s.max_flows = *cfg->flow_limit;
2544 if (cfg->overflow_policy
2545 && !strcmp(cfg->overflow_policy, "evict")) {
2548 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2549 for (k = 0; k < cfg->n_groups; k++) {
2550 const char *string = cfg->groups[k];
2553 msg = mf_parse_subfield__(&s.groups[k], &string);
2555 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2556 "'groups' (%s)", br->name, i, msg);
2558 } else if (*string) {
2559 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2560 "element '%s' contains trailing garbage",
2561 br->name, i, cfg->groups[k]);
2569 ofproto_configure_table(br->ofproto, i, &s);
2573 for (; j < br->cfg->n_flow_tables; j++) {
2574 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2575 "%"PRId64" not supported by this datapath", br->name,
2576 br->cfg->key_flow_tables[j]);
2580 /* Port functions. */
2582 static struct port *
2583 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2587 port = xzalloc(sizeof *port);
2589 port->name = xstrdup(cfg->name);
2591 list_init(&port->ifaces);
2593 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2595 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2601 get_port_other_config(const struct ovsrec_port *port, const char *key,
2602 const char *default_value)
2606 value = get_ovsrec_key_value(port->key_other_config,
2607 port->value_other_config,
2608 port->n_other_config, key);
2609 return value ? value : default_value;
2613 get_interface_other_config(const struct ovsrec_interface *iface,
2614 const char *key, const char *default_value)
2618 value = get_ovsrec_key_value(iface->key_other_config,
2619 iface->value_other_config,
2620 iface->n_other_config, key);
2621 return value ? value : default_value;
2624 /* Deletes interfaces from 'port' that are no longer configured for it. */
2626 port_del_ifaces(struct port *port)
2628 struct iface *iface, *next;
2629 struct sset new_ifaces;
2632 /* Collect list of new interfaces. */
2633 sset_init(&new_ifaces);
2634 for (i = 0; i < port->cfg->n_interfaces; i++) {
2635 const char *name = port->cfg->interfaces[i]->name;
2636 const char *type = port->cfg->interfaces[i]->name;
2637 if (strcmp(type, "null")) {
2638 sset_add(&new_ifaces, name);
2642 /* Get rid of deleted interfaces. */
2643 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2644 if (!sset_contains(&new_ifaces, iface->name)) {
2645 iface_destroy(iface);
2649 sset_destroy(&new_ifaces);
2652 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2655 port_add_ifaces(struct port *port)
2657 struct shash new_ifaces;
2658 struct shash_node *node;
2661 /* Collect new ifaces. */
2662 shash_init(&new_ifaces);
2663 for (i = 0; i < port->cfg->n_interfaces; i++) {
2664 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2665 if (strcmp(cfg->type, "null")
2666 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2667 VLOG_WARN("port %s: %s specified twice as port interface",
2668 port->name, cfg->name);
2669 iface_clear_db_record(cfg);
2673 /* Create new interfaces.
2674 * Update interface types and 'cfg' members. */
2675 SHASH_FOR_EACH (node, &new_ifaces) {
2676 const struct ovsrec_interface *cfg = node->data;
2677 const char *iface_name = node->name;
2678 struct iface *iface;
2680 iface = iface_lookup(port->bridge, iface_name);
2682 iface = iface_create(port, cfg);
2687 /* Determine interface type. The local port always has type
2688 * "internal". Other ports take their type from the database and
2689 * default to "system" if none is specified. */
2690 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2691 : cfg->type[0] ? cfg->type
2694 shash_destroy(&new_ifaces);
2698 port_destroy(struct port *port)
2701 struct bridge *br = port->bridge;
2702 struct iface *iface, *next;
2705 ofproto_bundle_unregister(br->ofproto, port);
2708 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2709 iface_destroy(iface);
2712 hmap_remove(&br->ports, &port->hmap_node);
2714 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2721 static struct port *
2722 port_lookup(const struct bridge *br, const char *name)
2726 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2728 if (!strcmp(port->name, name)) {
2736 enable_lacp(struct port *port, bool *activep)
2738 if (!port->cfg->lacp) {
2739 /* XXX when LACP implementation has been sufficiently tested, enable by
2740 * default and make active on bonded ports. */
2742 } else if (!strcmp(port->cfg->lacp, "off")) {
2744 } else if (!strcmp(port->cfg->lacp, "active")) {
2747 } else if (!strcmp(port->cfg->lacp, "passive")) {
2751 VLOG_WARN("port %s: unknown LACP mode %s",
2752 port->name, port->cfg->lacp);
2757 static struct lacp_settings *
2758 port_configure_lacp(struct port *port, struct lacp_settings *s)
2760 const char *lacp_time, *system_id;
2761 long long int custom_time;
2764 if (!enable_lacp(port, &s->active)) {
2768 s->name = port->name;
2770 system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL);
2772 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2773 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2774 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2775 " address.", port->name, system_id);
2779 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2782 if (eth_addr_is_zero(s->id)) {
2783 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2787 /* Prefer bondable links if unspecified. */
2788 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2790 s->priority = (priority > 0 && priority <= UINT16_MAX
2792 : UINT16_MAX - !list_is_short(&port->ifaces));
2794 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2798 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2799 custom_time = atoi(lacp_time);
2800 if (!strcmp(lacp_time, "fast")) {
2801 s->lacp_time = LACP_TIME_FAST;
2802 } else if (!strcmp(lacp_time, "slow")) {
2803 s->lacp_time = LACP_TIME_SLOW;
2804 } else if (custom_time > 0) {
2805 s->lacp_time = LACP_TIME_CUSTOM;
2806 s->custom_time = custom_time;
2808 s->lacp_time = LACP_TIME_SLOW;
2815 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2817 int priority, portid, key;
2819 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2820 priority = atoi(get_interface_other_config(iface->cfg,
2821 "lacp-port-priority", "0"));
2822 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2825 if (portid <= 0 || portid > UINT16_MAX) {
2826 portid = iface->ofp_port;
2829 if (priority <= 0 || priority > UINT16_MAX) {
2830 priority = UINT16_MAX;
2833 if (key < 0 || key > UINT16_MAX) {
2837 s->name = iface->name;
2839 s->priority = priority;
2844 port_configure_bond(struct port *port, struct bond_settings *s,
2845 uint32_t *bond_stable_ids)
2847 const char *detect_s;
2848 struct iface *iface;
2849 int miimon_interval;
2852 s->name = port->name;
2854 if (port->cfg->bond_mode) {
2855 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2856 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2857 port->name, port->cfg->bond_mode,
2858 bond_mode_to_string(s->balance));
2861 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2863 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2864 * active-backup. At some point we should remove this warning. */
2865 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2866 " in previous versions, the default bond_mode was"
2867 " balance-slb", port->name,
2868 bond_mode_to_string(s->balance));
2870 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2871 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2872 "please use another bond type or disable flood_vlans",
2876 miimon_interval = atoi(get_port_other_config(port->cfg,
2877 "bond-miimon-interval", "0"));
2878 if (miimon_interval <= 0) {
2879 miimon_interval = 200;
2882 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2883 if (!strcmp(detect_s, "carrier")) {
2884 miimon_interval = 0;
2885 } else if (strcmp(detect_s, "miimon")) {
2886 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2887 "defaulting to carrier", port->name, detect_s);
2888 miimon_interval = 0;
2891 s->up_delay = MAX(0, port->cfg->bond_updelay);
2892 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2893 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2894 s->rebalance_interval = atoi(
2895 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2896 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2897 s->rebalance_interval = 1000;
2900 s->fake_iface = port->cfg->bond_fake_iface;
2903 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2904 long long stable_id;
2906 stable_id = atoll(get_interface_other_config(iface->cfg,
2907 "bond-stable-id", "0"));
2908 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2909 stable_id = iface->ofp_port;
2911 bond_stable_ids[i++] = stable_id;
2913 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2917 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2918 * instead of obtaining it from the database. */
2920 port_is_synthetic(const struct port *port)
2922 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2925 /* Interface functions. */
2927 static struct iface *
2928 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2930 struct bridge *br = port->bridge;
2931 struct iface *iface;
2932 char *name = if_cfg->name;
2934 iface = xzalloc(sizeof *iface);
2936 iface->name = xstrdup(name);
2937 iface->ofp_port = -1;
2938 iface->tag = tag_create_random();
2939 iface->netdev = NULL;
2940 iface->cfg = if_cfg;
2941 iface->need_refresh = true;
2943 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2945 list_push_back(&port->ifaces, &iface->port_elem);
2947 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2953 iface_destroy(struct iface *iface)
2956 struct port *port = iface->port;
2957 struct bridge *br = port->bridge;
2959 if (br->ofproto && iface->ofp_port >= 0) {
2960 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2963 if (iface->ofp_port >= 0) {
2964 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2967 list_remove(&iface->port_elem);
2968 hmap_remove(&br->iface_by_name, &iface->name_node);
2970 netdev_close(iface->netdev);
2977 static struct iface *
2978 iface_lookup(const struct bridge *br, const char *name)
2980 struct iface *iface;
2982 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2983 &br->iface_by_name) {
2984 if (!strcmp(iface->name, name)) {
2992 static struct iface *
2993 iface_find(const char *name)
2995 const struct bridge *br;
2997 HMAP_FOR_EACH (br, node, &all_bridges) {
2998 struct iface *iface = iface_lookup(br, name);
3007 static struct iface *
3008 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3010 struct iface *iface;
3012 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3013 hash_int(ofp_port, 0), &br->ifaces) {
3014 if (iface->ofp_port == ofp_port) {
3021 /* Set Ethernet address of 'iface', if one is specified in the configuration
3024 iface_set_mac(struct iface *iface)
3026 uint8_t ea[ETH_ADDR_LEN];
3028 if (!strcmp(iface->type, "internal")
3029 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3030 if (iface->ofp_port == OFPP_LOCAL) {
3031 VLOG_ERR("interface %s: ignoring mac in Interface record "
3032 "(use Bridge record to set local port's mac)",
3034 } else if (eth_addr_is_multicast(ea)) {
3035 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3038 int error = netdev_set_etheraddr(iface->netdev, ea);
3040 VLOG_ERR("interface %s: setting MAC failed (%s)",
3041 iface->name, strerror(error));
3047 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3049 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3051 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3052 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3056 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3057 * sets the "ofport" field to -1.
3059 * This is appropriate when 'if_cfg''s interface cannot be created or is
3060 * otherwise invalid. */
3062 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3064 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3065 iface_set_ofport(if_cfg, -1);
3066 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3067 ovsrec_interface_set_admin_state(if_cfg, NULL);
3068 ovsrec_interface_set_duplex(if_cfg, NULL);
3069 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3070 ovsrec_interface_set_link_state(if_cfg, NULL);
3071 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3072 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3073 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3074 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3075 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3076 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3080 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3082 * The value strings in '*shash' are taken directly from values[], not copied,
3083 * so the caller should not modify or free them. */
3085 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3086 struct shash *shash)
3091 for (i = 0; i < n; i++) {
3092 shash_add(shash, keys[i], values[i]);
3096 /* Creates 'keys' and 'values' arrays from 'shash'.
3098 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3099 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3100 * are populated with with strings taken directly from 'shash' and thus have
3101 * the same ownership of the key-value pairs in shash.
3104 shash_to_ovs_idl_map(struct shash *shash,
3105 char ***keys, char ***values, size_t *n)
3109 struct shash_node *sn;
3111 count = shash_count(shash);
3113 k = xmalloc(count * sizeof *k);
3114 v = xmalloc(count * sizeof *v);
3117 SHASH_FOR_EACH(sn, shash) {
3128 struct iface_delete_queues_cbdata {
3129 struct netdev *netdev;
3130 const struct ovsdb_datum *queues;
3134 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3136 union ovsdb_atom atom;
3138 atom.integer = target;
3139 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3143 iface_delete_queues(unsigned int queue_id,
3144 const struct shash *details OVS_UNUSED, void *cbdata_)
3146 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3148 if (!queue_ids_include(cbdata->queues, queue_id)) {
3149 netdev_delete_queue(cbdata->netdev, queue_id);
3154 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3156 struct ofpbuf queues_buf;
3158 ofpbuf_init(&queues_buf, 0);
3160 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3161 netdev_set_qos(iface->netdev, NULL, NULL);
3163 struct iface_delete_queues_cbdata cbdata;
3164 struct shash details;
3168 /* Configure top-level Qos for 'iface'. */
3169 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3170 qos->n_other_config, &details);
3171 netdev_set_qos(iface->netdev, qos->type, &details);
3172 shash_destroy(&details);
3174 /* Deconfigure queues that were deleted. */
3175 cbdata.netdev = iface->netdev;
3176 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3178 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3180 /* Configure queues for 'iface'. */
3182 for (i = 0; i < qos->n_queues; i++) {
3183 const struct ovsrec_queue *queue = qos->value_queues[i];
3184 unsigned int queue_id = qos->key_queues[i];
3186 if (queue_id == 0) {
3190 if (queue->n_dscp == 1) {
3191 struct ofproto_port_queue *port_queue;
3193 port_queue = ofpbuf_put_uninit(&queues_buf,
3194 sizeof *port_queue);
3195 port_queue->queue = queue_id;
3196 port_queue->dscp = queue->dscp[0];
3199 shash_from_ovs_idl_map(queue->key_other_config,
3200 queue->value_other_config,
3201 queue->n_other_config, &details);
3202 netdev_set_queue(iface->netdev, queue_id, &details);
3203 shash_destroy(&details);
3206 shash_init(&details);
3207 netdev_set_queue(iface->netdev, 0, &details);
3208 shash_destroy(&details);
3212 if (iface->ofp_port >= 0) {
3213 const struct ofproto_port_queue *port_queues = queues_buf.data;
3214 size_t n_queues = queues_buf.size / sizeof *port_queues;
3216 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3217 port_queues, n_queues);
3220 netdev_set_policing(iface->netdev,
3221 iface->cfg->ingress_policing_rate,
3222 iface->cfg->ingress_policing_burst);
3224 ofpbuf_uninit(&queues_buf);
3228 iface_configure_cfm(struct iface *iface)
3230 const struct ovsrec_interface *cfg = iface->cfg;
3231 const char *extended_str, *opstate_str;
3232 struct cfm_settings s;
3234 if (!cfg->n_cfm_mpid) {
3235 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3239 s.mpid = *cfg->cfm_mpid;
3240 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
3242 s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan",
3244 s.ccm_pcp = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_pcp",
3246 if (s.interval <= 0) {
3250 extended_str = get_interface_other_config(iface->cfg, "cfm_extended",
3252 s.extended = !strcasecmp("true", extended_str);
3254 opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up");
3255 s.opup = !strcasecmp("up", opstate_str);
3257 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3260 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3261 * instead of obtaining it from the database. */
3263 iface_is_synthetic(const struct iface *iface)
3265 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3268 /* Port mirroring. */
3270 static struct mirror *
3271 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3275 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3276 if (uuid_equals(uuid, &m->uuid)) {
3284 bridge_configure_mirrors(struct bridge *br)
3286 const struct ovsdb_datum *mc;
3287 unsigned long *flood_vlans;
3288 struct mirror *m, *next;
3291 /* Get rid of deleted mirrors. */
3292 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3293 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3294 union ovsdb_atom atom;
3296 atom.uuid = m->uuid;
3297 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3302 /* Add new mirrors and reconfigure existing ones. */
3303 for (i = 0; i < br->cfg->n_mirrors; i++) {
3304 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3305 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3307 m = mirror_create(br, cfg);
3310 if (!mirror_configure(m)) {
3315 /* Update flooded vlans (for RSPAN). */
3316 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3317 br->cfg->n_flood_vlans);
3318 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3319 bitmap_free(flood_vlans);
3322 static struct mirror *
3323 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3327 m = xzalloc(sizeof *m);
3328 m->uuid = cfg->header_.uuid;
3329 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3331 m->name = xstrdup(cfg->name);
3337 mirror_destroy(struct mirror *m)
3340 struct bridge *br = m->bridge;
3343 ofproto_mirror_unregister(br->ofproto, m);
3346 hmap_remove(&br->mirrors, &m->hmap_node);
3353 mirror_collect_ports(struct mirror *m,
3354 struct ovsrec_port **in_ports, int n_in_ports,
3355 void ***out_portsp, size_t *n_out_portsp)
3357 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3358 size_t n_out_ports = 0;
3361 for (i = 0; i < n_in_ports; i++) {
3362 const char *name = in_ports[i]->name;
3363 struct port *port = port_lookup(m->bridge, name);
3365 out_ports[n_out_ports++] = port;
3367 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3368 "port %s", m->bridge->name, m->name, name);
3371 *out_portsp = out_ports;
3372 *n_out_portsp = n_out_ports;
3376 mirror_configure(struct mirror *m)
3378 const struct ovsrec_mirror *cfg = m->cfg;
3379 struct ofproto_mirror_settings s;
3382 if (strcmp(cfg->name, m->name)) {
3384 m->name = xstrdup(cfg->name);
3388 /* Get output port or VLAN. */
3389 if (cfg->output_port) {
3390 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3391 if (!s.out_bundle) {
3392 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3393 m->bridge->name, m->name);
3396 s.out_vlan = UINT16_MAX;
3398 if (cfg->output_vlan) {
3399 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3400 "output vlan; ignoring output vlan",
3401 m->bridge->name, m->name);
3403 } else if (cfg->output_vlan) {
3404 /* The database should prevent invalid VLAN values. */
3405 s.out_bundle = NULL;
3406 s.out_vlan = *cfg->output_vlan;
3408 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3409 m->bridge->name, m->name);
3413 /* Get port selection. */
3414 if (cfg->select_all) {
3415 size_t n_ports = hmap_count(&m->bridge->ports);
3416 void **ports = xmalloc(n_ports * sizeof *ports);
3421 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3431 /* Get ports, dropping ports that don't exist.
3432 * The IDL ensures that there are no duplicates. */
3433 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3434 &s.srcs, &s.n_srcs);
3435 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3436 &s.dsts, &s.n_dsts);
3439 /* Get VLAN selection. */
3440 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3443 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3446 if (s.srcs != s.dsts) {
3455 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3457 * This is deprecated. It is only for compatibility with broken device drivers
3458 * in old versions of Linux that do not properly support VLANs when VLAN
3459 * devices are not used. When broken device drivers are no longer in
3460 * widespread use, we will delete these interfaces. */
3462 static void **blocks;
3463 static size_t n_blocks, allocated_blocks;
3465 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3466 * VLAN splinters are reconfigured. */
3468 register_block(void *block)
3470 if (n_blocks >= allocated_blocks) {
3471 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3473 blocks[n_blocks++] = block;
3476 /* Frees all of the blocks registered with register_block(). */
3478 free_registered_blocks(void)
3482 for (i = 0; i < n_blocks; i++) {
3488 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3491 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3495 value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", "");
3496 return !strcmp(value, "true");
3499 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3502 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3503 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3504 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3507 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3508 * use, returns NULL.
3510 * Updates 'vlan_splinters_enabled_anywhere'. */
3511 static unsigned long int *
3512 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3514 unsigned long int *splinter_vlans;
3515 struct sset splinter_ifaces;
3516 const char *real_dev_name;
3517 struct shash *real_devs;
3518 struct shash_node *node;
3522 /* Free space allocated for synthesized ports and interfaces, since we're
3523 * in the process of reconstructing all of them. */
3524 free_registered_blocks();
3526 splinter_vlans = bitmap_allocate(4096);
3527 sset_init(&splinter_ifaces);
3528 vlan_splinters_enabled_anywhere = false;
3529 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3530 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3533 for (j = 0; j < br_cfg->n_ports; j++) {
3534 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3537 for (k = 0; k < port_cfg->n_interfaces; k++) {
3538 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3540 if (vlan_splinters_is_enabled(iface_cfg)) {
3541 vlan_splinters_enabled_anywhere = true;
3542 sset_add(&splinter_ifaces, iface_cfg->name);
3543 vlan_bitmap_from_array__(port_cfg->trunks,
3549 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3550 bitmap_set1(splinter_vlans, *port_cfg->tag);
3555 if (!vlan_splinters_enabled_anywhere) {
3556 free(splinter_vlans);
3557 sset_destroy(&splinter_ifaces);
3561 HMAP_FOR_EACH (br, node, &all_bridges) {
3563 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3567 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3568 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3569 * device to be created for it. */
3570 bitmap_set0(splinter_vlans, 0);
3571 bitmap_set0(splinter_vlans, 4095);
3573 /* Delete all VLAN devices that we don't need. */
3575 real_devs = vlandev_get_real_devs();
3576 SHASH_FOR_EACH (node, real_devs) {
3577 const struct vlan_real_dev *real_dev = node->data;
3578 const struct vlan_dev *vlan_dev;
3579 bool real_dev_has_splinters;
3581 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3583 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3584 if (!real_dev_has_splinters
3585 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3586 struct netdev *netdev;
3588 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3589 if (!netdev_get_in4(netdev, NULL, NULL) ||
3590 !netdev_get_in6(netdev, NULL)) {
3591 vlandev_del(vlan_dev->name);
3593 /* It has an IP address configured, so we don't own
3594 * it. Don't delete it. */
3596 netdev_close(netdev);
3603 /* Add all VLAN devices that we need. */
3604 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3607 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3608 if (!vlandev_get_name(real_dev_name, vid)) {
3609 vlandev_add(real_dev_name, vid);
3616 sset_destroy(&splinter_ifaces);
3618 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3619 free(splinter_vlans);
3622 return splinter_vlans;
3625 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3628 configure_splinter_port(struct port *port)
3630 struct ofproto *ofproto = port->bridge->ofproto;
3631 uint16_t realdev_ofp_port;
3632 const char *realdev_name;
3633 struct iface *vlandev, *realdev;
3635 ofproto_bundle_unregister(port->bridge->ofproto, port);
3637 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3640 realdev_name = get_port_other_config(port->cfg, "realdev", NULL);
3641 realdev = iface_lookup(port->bridge, realdev_name);
3642 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3644 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3648 static struct ovsrec_port *
3649 synthesize_splinter_port(const char *real_dev_name,
3650 const char *vlan_dev_name, int vid)
3652 struct ovsrec_interface *iface;
3653 struct ovsrec_port *port;
3655 iface = xzalloc(sizeof *iface);
3656 iface->name = xstrdup(vlan_dev_name);
3657 iface->type = "system";
3659 port = xzalloc(sizeof *port);
3660 port->interfaces = xmemdup(&iface, sizeof iface);
3661 port->n_interfaces = 1;
3662 port->name = xstrdup(vlan_dev_name);
3663 port->vlan_mode = "splinter";
3664 port->tag = xmalloc(sizeof *port->tag);
3666 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3667 port->key_other_config[0] = "realdev";
3668 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3669 port->value_other_config[0] = xstrdup(real_dev_name);
3670 port->n_other_config = 1;
3672 register_block(iface);
3673 register_block(iface->name);
3674 register_block(port);
3675 register_block(port->interfaces);
3676 register_block(port->name);
3677 register_block(port->tag);
3678 register_block(port->key_other_config);
3679 register_block(port->value_other_config);
3680 register_block(port->value_other_config[0]);
3685 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3686 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3687 * 1-bit in the 'splinter_vlans' bitmap. */
3689 add_vlan_splinter_ports(struct bridge *br,
3690 const unsigned long int *splinter_vlans,
3691 struct shash *ports)
3695 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3696 * we're modifying 'ports'. */
3697 for (i = 0; i < br->cfg->n_ports; i++) {
3698 const char *name = br->cfg->ports[i]->name;
3699 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3702 for (j = 0; j < port_cfg->n_interfaces; j++) {
3703 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3705 if (vlan_splinters_is_enabled(iface_cfg)) {
3706 const char *real_dev_name;
3709 real_dev_name = iface_cfg->name;
3710 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3711 const char *vlan_dev_name;
3713 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3715 && !shash_find(ports, vlan_dev_name)) {
3716 shash_add(ports, vlan_dev_name,
3717 synthesize_splinter_port(
3718 real_dev_name, vlan_dev_name, vid));
3727 mirror_refresh_stats(struct mirror *m)
3729 struct ofproto *ofproto = m->bridge->ofproto;
3730 uint64_t tx_packets, tx_bytes;
3733 size_t stat_cnt = 0;
3735 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3736 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3740 if (tx_packets != UINT64_MAX) {
3741 keys[stat_cnt] = "tx_packets";
3742 values[stat_cnt] = tx_packets;
3745 if (tx_bytes != UINT64_MAX) {
3746 keys[stat_cnt] = "tx_bytes";
3747 values[stat_cnt] = tx_bytes;
3751 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);