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 "lib/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. */
70 /* These members are valid only after bridge_reconfigure() causes them to
72 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
73 int ofp_port; /* OpenFlow port number, -1 if unknown. */
74 struct netdev *netdev; /* Network device. */
75 const char *type; /* Usually same as cfg->type. */
76 const struct ovsrec_interface *cfg;
77 bool need_refresh; /* Refresh iface after create. */
81 struct uuid uuid; /* UUID of this "mirror" record in database. */
82 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
83 struct bridge *bridge;
85 const struct ovsrec_mirror *cfg;
89 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
90 struct bridge *bridge;
93 const struct ovsrec_port *cfg;
95 /* An ordinary bridge port has 1 interface.
96 * A bridge port for bonding has at least 2 interfaces. */
97 struct list ifaces; /* List of "struct iface"s. */
101 struct hmap_node node; /* In 'all_bridges'. */
102 char *name; /* User-specified arbitrary name. */
103 char *type; /* Datapath type. */
104 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
105 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
106 const struct ovsrec_bridge *cfg;
108 /* OpenFlow switch processing. */
109 struct ofproto *ofproto; /* OpenFlow switch. */
112 struct hmap ports; /* "struct port"s indexed by name. */
113 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
114 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
116 /* Port mirroring. */
117 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
119 /* Synthetic local port if necessary. */
120 struct ovsrec_port synth_local_port;
121 struct ovsrec_interface synth_local_iface;
122 struct ovsrec_interface *synth_local_ifacep;
125 /* All bridges, indexed by name. */
126 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
128 /* OVSDB IDL used to obtain configuration. */
129 static struct ovsdb_idl *idl;
131 /* Most recently processed IDL sequence number. */
132 static unsigned int idl_seqno;
134 /* Each time this timer expires, the bridge fetches systems and interface
135 * statistics and pushes them into the database. */
136 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
137 static long long int stats_timer = LLONG_MIN;
139 /* Stores the time after which rate limited statistics may be written to the
140 * database. Only updated when changes to the database require rate limiting.
142 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
143 static long long int db_limiter = LLONG_MIN;
145 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
146 * expensive. This can cause bridge_reconfigure() to take a long time during
147 * which no other work can be done. To deal with this problem, we limit port
148 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
149 * call to bridge_reconfigure(). If there is more work to do after the limit
150 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
151 * This allows the rest of the code to catch up on important things like
152 * forwarding packets. */
153 #define OFP_PORT_ACTION_WINDOW 10
154 static bool need_reconfigure = false;
156 static void add_del_bridges(const struct ovsrec_open_vswitch *);
157 static void bridge_del_ofprotos(void);
158 static bool bridge_add_ofprotos(struct bridge *);
159 static void bridge_create(const struct ovsrec_bridge *);
160 static void bridge_destroy(struct bridge *);
161 static struct bridge *bridge_lookup(const char *name);
162 static unixctl_cb_func bridge_unixctl_dump_flows;
163 static unixctl_cb_func bridge_unixctl_reconnect;
164 static size_t bridge_get_controllers(const struct bridge *br,
165 struct ovsrec_controller ***controllersp);
166 static void bridge_add_del_ports(struct bridge *,
167 const unsigned long int *splinter_vlans);
168 static void bridge_add_ofproto_ports(struct bridge *,
169 long long int *port_action_timer);
170 static void bridge_del_ofproto_ports(struct bridge *,
171 long long int *port_action_timer);
172 static void bridge_refresh_ofp_port(struct bridge *);
173 static void bridge_configure_datapath_id(struct bridge *);
174 static void bridge_configure_flow_eviction_threshold(struct bridge *);
175 static void bridge_configure_netflow(struct bridge *);
176 static void bridge_configure_forward_bpdu(struct bridge *);
177 static void bridge_configure_mac_idle_time(struct bridge *);
178 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
179 static void bridge_configure_stp(struct bridge *);
180 static void bridge_configure_tables(struct bridge *);
181 static void bridge_configure_remotes(struct bridge *,
182 const struct sockaddr_in *managers,
184 static void bridge_pick_local_hw_addr(struct bridge *,
185 uint8_t ea[ETH_ADDR_LEN],
186 struct iface **hw_addr_iface);
187 static uint64_t bridge_pick_datapath_id(struct bridge *,
188 const uint8_t bridge_ea[ETH_ADDR_LEN],
189 struct iface *hw_addr_iface);
190 static uint64_t dpid_from_hash(const void *, size_t nbytes);
191 static bool bridge_has_bond_fake_iface(const struct bridge *,
193 static bool port_is_bond_fake_iface(const struct port *);
195 static unixctl_cb_func qos_unixctl_show;
197 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
198 static void port_add_ifaces(struct port *);
199 static void port_del_ifaces(struct port *);
200 static void port_destroy(struct port *);
201 static struct port *port_lookup(const struct bridge *, const char *name);
202 static void port_configure(struct port *);
203 static struct lacp_settings *port_configure_lacp(struct port *,
204 struct lacp_settings *);
205 static void port_configure_bond(struct port *, struct bond_settings *,
206 uint32_t *bond_stable_ids);
207 static bool port_is_synthetic(const struct port *);
209 static void bridge_configure_mirrors(struct bridge *);
210 static struct mirror *mirror_create(struct bridge *,
211 const struct ovsrec_mirror *);
212 static void mirror_destroy(struct mirror *);
213 static bool mirror_configure(struct mirror *);
214 static void mirror_refresh_stats(struct mirror *);
216 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
217 static struct iface *iface_create(struct port *port,
218 const struct ovsrec_interface *if_cfg);
219 static void iface_destroy(struct iface *);
220 static struct iface *iface_lookup(const struct bridge *, const char *name);
221 static struct iface *iface_find(const char *name);
222 static struct iface *iface_from_ofp_port(const struct bridge *,
224 static void iface_set_mac(struct iface *);
225 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
226 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
227 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
228 static void iface_configure_cfm(struct iface *);
229 static void iface_refresh_cfm_stats(struct iface *);
230 static void iface_refresh_stats(struct iface *);
231 static void iface_refresh_status(struct iface *);
232 static bool iface_is_synthetic(const struct iface *);
233 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
235 static void shash_to_ovs_idl_map(struct shash *,
236 char ***keys, char ***values, size_t *n);
238 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
240 * This is deprecated. It is only for compatibility with broken device drivers
241 * in old versions of Linux that do not properly support VLANs when VLAN
242 * devices are not used. When broken device drivers are no longer in
243 * widespread use, we will delete these interfaces. */
245 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
246 static bool vlan_splinters_enabled_anywhere;
248 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
249 static unsigned long int *collect_splinter_vlans(
250 const struct ovsrec_open_vswitch *);
251 static void configure_splinter_port(struct port *);
252 static void add_vlan_splinter_ports(struct bridge *,
253 const unsigned long int *splinter_vlans,
254 struct shash *ports);
256 /* Public functions. */
258 /* Initializes the bridge module, configuring it to obtain its configuration
259 * from an OVSDB server accessed over 'remote', which should be a string in a
260 * form acceptable to ovsdb_idl_create(). */
262 bridge_init(const char *remote)
264 /* Create connection to database. */
265 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
266 idl_seqno = ovsdb_idl_get_seqno(idl);
267 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
269 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
270 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
271 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
272 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
273 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
274 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
275 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
277 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
278 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
279 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
281 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
282 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
283 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
284 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
286 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
287 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
288 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
289 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
290 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
291 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
292 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
293 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
294 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
295 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
296 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
297 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
300 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
302 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
303 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
304 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
305 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
307 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
309 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
311 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
312 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
314 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
316 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
318 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
319 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
320 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
321 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
322 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
324 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
326 /* Register unixctl commands. */
327 unixctl_command_register("qos/show", "interface", 1, 1,
328 qos_unixctl_show, NULL);
329 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
330 bridge_unixctl_dump_flows, NULL);
331 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
332 bridge_unixctl_reconnect, NULL);
342 struct bridge *br, *next_br;
344 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
347 ovsdb_idl_destroy(idl);
350 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
351 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
352 * responsible for freeing '*managersp' (with free()).
354 * You may be asking yourself "why does ovs-vswitchd care?", because
355 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
356 * should not be and in fact is not directly involved in that. But
357 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
358 * it has to tell in-band control where the managers are to enable that.
359 * (Thus, only managers connected in-band are collected.)
362 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
363 struct sockaddr_in **managersp, size_t *n_managersp)
365 struct sockaddr_in *managers = NULL;
366 size_t n_managers = 0;
370 /* Collect all of the potential targets from the "targets" columns of the
371 * rows pointed to by "manager_options", excluding any that are
374 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
375 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
377 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
378 sset_find_and_delete(&targets, m->target);
380 sset_add(&targets, m->target);
384 /* Now extract the targets' IP addresses. */
385 if (!sset_is_empty(&targets)) {
388 managers = xmalloc(sset_count(&targets) * sizeof *managers);
389 SSET_FOR_EACH (target, &targets) {
390 struct sockaddr_in *sin = &managers[n_managers];
392 if (stream_parse_target_with_default_ports(target,
400 sset_destroy(&targets);
402 *managersp = managers;
403 *n_managersp = n_managers;
407 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
409 unsigned long int *splinter_vlans;
410 long long int port_action_timer;
411 struct sockaddr_in *managers;
412 struct bridge *br, *next;
413 int sflow_bridge_number;
416 COVERAGE_INC(bridge_reconfigure);
418 port_action_timer = LLONG_MAX;
419 need_reconfigure = false;
421 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
422 * iface"s according to 'ovs_cfg', with only very minimal configuration
425 * This is mostly an update to bridge data structures. Very little is
426 * pushed down to ofproto or lower layers. */
427 add_del_bridges(ovs_cfg);
428 splinter_vlans = collect_splinter_vlans(ovs_cfg);
429 HMAP_FOR_EACH (br, node, &all_bridges) {
430 bridge_add_del_ports(br, splinter_vlans);
432 free(splinter_vlans);
434 /* Delete all datapaths and datapath ports that are no longer configured.
436 * The kernel will reject any attempt to add a given port to a datapath if
437 * that port already belongs to a different datapath, so we must do all
438 * port deletions before any port additions. A datapath always has a
439 * "local port" so we must delete not-configured datapaths too. */
440 bridge_del_ofprotos();
441 HMAP_FOR_EACH (br, node, &all_bridges) {
443 bridge_del_ofproto_ports(br, &port_action_timer);
447 /* Create datapaths and datapath ports that are missing.
449 * After this is done, we have our final set of bridges, ports, and
450 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
451 * has at least one iface, every "struct iface" has a valid ofp_port and
453 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
455 if (bridge_add_ofprotos(br)) {
456 bridge_del_ofproto_ports(br, &port_action_timer);
462 HMAP_FOR_EACH (br, node, &all_bridges) {
463 bridge_refresh_ofp_port(br);
464 bridge_add_ofproto_ports(br, &port_action_timer);
467 /* Complete the configuration. */
468 sflow_bridge_number = 0;
469 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
470 HMAP_FOR_EACH (br, node, &all_bridges) {
473 /* We need the datapath ID early to allow LACP ports to use it as the
474 * default system ID. */
475 bridge_configure_datapath_id(br);
477 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
480 port_configure(port);
482 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
483 iface_configure_cfm(iface);
484 iface_configure_qos(iface, port->cfg->qos);
485 iface_set_mac(iface);
488 bridge_configure_mirrors(br);
489 bridge_configure_flow_eviction_threshold(br);
490 bridge_configure_forward_bpdu(br);
491 bridge_configure_mac_idle_time(br);
492 bridge_configure_remotes(br, managers, n_managers);
493 bridge_configure_netflow(br);
494 bridge_configure_sflow(br, &sflow_bridge_number);
495 bridge_configure_stp(br);
496 bridge_configure_tables(br);
500 if (!need_reconfigure) {
501 /* ovs-vswitchd has completed initialization, so allow the process that
502 * forked us to exit successfully. */
503 daemonize_complete();
507 /* Iterate over all ofprotos and delete any of them that do not have a
508 * configured bridge or that are the wrong type. */
510 bridge_del_ofprotos(void)
518 ofproto_enumerate_types(&types);
519 SSET_FOR_EACH (type, &types) {
522 ofproto_enumerate_names(type, &names);
523 SSET_FOR_EACH (name, &names) {
524 struct bridge *br = bridge_lookup(name);
525 if (!br || strcmp(type, br->type)) {
526 ofproto_delete(name, type);
530 sset_destroy(&names);
531 sset_destroy(&types);
535 bridge_add_ofprotos(struct bridge *br)
537 int error = ofproto_create(br->name, br->type, &br->ofproto);
539 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
546 port_configure(struct port *port)
548 const struct ovsrec_port *cfg = port->cfg;
549 struct bond_settings bond_settings;
550 struct lacp_settings lacp_settings;
551 struct ofproto_bundle_settings s;
554 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
555 configure_splinter_port(port);
564 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
565 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
566 s.slaves[s.n_slaves++] = iface->ofp_port;
571 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
575 /* Get VLAN trunks. */
578 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
582 if (cfg->vlan_mode) {
583 if (!strcmp(cfg->vlan_mode, "access")) {
584 s.vlan_mode = PORT_VLAN_ACCESS;
585 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
586 s.vlan_mode = PORT_VLAN_TRUNK;
587 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
588 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
589 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
590 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
592 /* This "can't happen" because ovsdb-server should prevent it. */
593 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
594 s.vlan_mode = PORT_VLAN_TRUNK;
598 s.vlan_mode = PORT_VLAN_ACCESS;
600 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
604 s.vlan_mode = PORT_VLAN_TRUNK;
607 s.use_priority_tags = !strcmp("true", ovsrec_port_get_other_config_value(
608 cfg, "priority-tags", ""));
610 /* Get LACP settings. */
611 s.lacp = port_configure_lacp(port, &lacp_settings);
615 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
616 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
617 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
620 s.lacp_slaves = NULL;
623 /* Get bond settings. */
624 if (s.n_slaves > 1) {
625 s.bond = &bond_settings;
626 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
627 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
630 s.bond_stable_ids = NULL;
632 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
633 netdev_set_miimon_interval(iface->netdev, 0);
638 ofproto_bundle_register(port->bridge->ofproto, port, &s);
644 free(s.bond_stable_ids);
647 /* Pick local port hardware address and datapath ID for 'br'. */
649 bridge_configure_datapath_id(struct bridge *br)
651 uint8_t ea[ETH_ADDR_LEN];
653 struct iface *local_iface;
654 struct iface *hw_addr_iface;
657 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
658 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
660 int error = netdev_set_etheraddr(local_iface->netdev, ea);
662 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
663 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
664 "Ethernet address: %s",
665 br->name, strerror(error));
668 memcpy(br->ea, ea, ETH_ADDR_LEN);
670 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
671 ofproto_set_datapath_id(br->ofproto, dpid);
673 dpid_string = xasprintf("%016"PRIx64, dpid);
674 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
678 /* Set NetFlow configuration on 'br'. */
680 bridge_configure_netflow(struct bridge *br)
682 struct ovsrec_netflow *cfg = br->cfg->netflow;
683 struct netflow_options opts;
686 ofproto_set_netflow(br->ofproto, NULL);
690 memset(&opts, 0, sizeof opts);
692 /* Get default NetFlow configuration from datapath.
693 * Apply overrides from 'cfg'. */
694 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
695 if (cfg->engine_type) {
696 opts.engine_type = *cfg->engine_type;
698 if (cfg->engine_id) {
699 opts.engine_id = *cfg->engine_id;
702 /* Configure active timeout interval. */
703 opts.active_timeout = cfg->active_timeout;
704 if (!opts.active_timeout) {
705 opts.active_timeout = -1;
706 } else if (opts.active_timeout < 0) {
707 VLOG_WARN("bridge %s: active timeout interval set to negative "
708 "value, using default instead (%d seconds)", br->name,
709 NF_ACTIVE_TIMEOUT_DEFAULT);
710 opts.active_timeout = -1;
713 /* Add engine ID to interface number to disambiguate bridgs? */
714 opts.add_id_to_iface = cfg->add_id_to_interface;
715 if (opts.add_id_to_iface) {
716 if (opts.engine_id > 0x7f) {
717 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
718 "another vswitch, choose an engine id less than 128",
721 if (hmap_count(&br->ports) > 508) {
722 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
723 "another port when more than 508 ports are used",
729 sset_init(&opts.collectors);
730 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
733 if (ofproto_set_netflow(br->ofproto, &opts)) {
734 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
736 sset_destroy(&opts.collectors);
739 /* Set sFlow configuration on 'br'. */
741 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
743 const struct ovsrec_sflow *cfg = br->cfg->sflow;
744 struct ovsrec_controller **controllers;
745 struct ofproto_sflow_options oso;
746 size_t n_controllers;
750 ofproto_set_sflow(br->ofproto, NULL);
754 memset(&oso, 0, sizeof oso);
756 sset_init(&oso.targets);
757 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
759 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
761 oso.sampling_rate = *cfg->sampling;
764 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
766 oso.polling_interval = *cfg->polling;
769 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
771 oso.header_len = *cfg->header;
774 oso.sub_id = (*sflow_bridge_number)++;
775 oso.agent_device = cfg->agent;
777 oso.control_ip = NULL;
778 n_controllers = bridge_get_controllers(br, &controllers);
779 for (i = 0; i < n_controllers; i++) {
780 if (controllers[i]->local_ip) {
781 oso.control_ip = controllers[i]->local_ip;
785 ofproto_set_sflow(br->ofproto, &oso);
787 sset_destroy(&oso.targets);
791 port_configure_stp(const struct ofproto *ofproto, struct port *port,
792 struct ofproto_port_stp_settings *port_s,
793 int *port_num_counter, unsigned long *port_num_bitmap)
795 const char *config_str;
798 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-enable",
800 if (config_str && !strcmp(config_str, "false")) {
801 port_s->enable = false;
804 port_s->enable = true;
807 /* STP over bonds is not supported. */
808 if (!list_is_singleton(&port->ifaces)) {
809 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
811 port_s->enable = false;
815 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
817 /* Internal ports shouldn't participate in spanning tree, so
819 if (!strcmp(iface->type, "internal")) {
820 VLOG_DBG("port %s: disable STP on internal ports", port->name);
821 port_s->enable = false;
825 /* STP on mirror output ports is not supported. */
826 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
827 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
828 port_s->enable = false;
832 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-port-num",
835 unsigned long int port_num = strtoul(config_str, NULL, 0);
836 int port_idx = port_num - 1;
838 if (port_num < 1 || port_num > STP_MAX_PORTS) {
839 VLOG_ERR("port %s: invalid stp-port-num", port->name);
840 port_s->enable = false;
844 if (bitmap_is_set(port_num_bitmap, port_idx)) {
845 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
846 port->name, port_num);
847 port_s->enable = false;
850 bitmap_set1(port_num_bitmap, port_idx);
851 port_s->port_num = port_idx;
853 if (*port_num_counter > STP_MAX_PORTS) {
854 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
855 port_s->enable = false;
859 port_s->port_num = (*port_num_counter)++;
862 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-path-cost",
865 port_s->path_cost = strtoul(config_str, NULL, 10);
867 enum netdev_features current;
869 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
870 /* Couldn't get speed, so assume 100Mb/s. */
871 port_s->path_cost = 19;
875 mbps = netdev_features_to_bps(current) / 1000000;
876 port_s->path_cost = stp_convert_speed_to_cost(mbps);
880 config_str = ovsrec_port_get_other_config_value(port->cfg,
884 port_s->priority = strtoul(config_str, NULL, 0);
886 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
890 /* Set spanning tree configuration on 'br'. */
892 bridge_configure_stp(struct bridge *br)
894 if (!br->cfg->stp_enable) {
895 ofproto_set_stp(br->ofproto, NULL);
897 struct ofproto_stp_settings br_s;
898 const char *config_str;
900 int port_num_counter;
901 unsigned long *port_num_bitmap;
903 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
907 uint8_t ea[ETH_ADDR_LEN];
909 if (eth_addr_from_string(config_str, ea)) {
910 br_s.system_id = eth_addr_to_uint64(ea);
912 br_s.system_id = eth_addr_to_uint64(br->ea);
913 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
914 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
917 br_s.system_id = eth_addr_to_uint64(br->ea);
920 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
924 br_s.priority = strtoul(config_str, NULL, 0);
926 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
929 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
933 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
935 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
938 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
942 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
944 br_s.max_age = STP_DEFAULT_MAX_AGE;
947 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
951 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
953 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
956 /* Configure STP on the bridge. */
957 if (ofproto_set_stp(br->ofproto, &br_s)) {
958 VLOG_ERR("bridge %s: could not enable STP", br->name);
962 /* Users must either set the port number with the "stp-port-num"
963 * configuration on all ports or none. If manual configuration
964 * is not done, then we allocate them sequentially. */
965 port_num_counter = 0;
966 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
967 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
968 struct ofproto_port_stp_settings port_s;
971 port_configure_stp(br->ofproto, port, &port_s,
972 &port_num_counter, port_num_bitmap);
974 /* As bonds are not supported, just apply configuration to
976 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
977 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
979 VLOG_ERR("port %s: could not enable STP", port->name);
985 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
986 && port_num_counter) {
987 VLOG_ERR("bridge %s: must manually configure all STP port "
988 "IDs or none, disabling", br->name);
989 ofproto_set_stp(br->ofproto, NULL);
991 bitmap_free(port_num_bitmap);
996 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
998 const struct port *port = port_lookup(br, name);
999 return port && port_is_bond_fake_iface(port);
1003 port_is_bond_fake_iface(const struct port *port)
1005 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1009 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1011 struct bridge *br, *next;
1012 struct shash new_br;
1015 /* Collect new bridges' names and types. */
1016 shash_init(&new_br);
1017 for (i = 0; i < cfg->n_bridges; i++) {
1018 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1019 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1021 if (strchr(br_cfg->name, '/')) {
1022 /* Prevent remote ovsdb-server users from accessing arbitrary
1023 * directories, e.g. consider a bridge named "../../../etc/". */
1024 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1026 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1027 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1031 /* Get rid of deleted bridges or those whose types have changed.
1032 * Update 'cfg' of bridges that still exist. */
1033 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1034 br->cfg = shash_find_data(&new_br, br->name);
1035 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1036 br->cfg->datapath_type))) {
1041 /* Add new bridges. */
1042 for (i = 0; i < cfg->n_bridges; i++) {
1043 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1044 struct bridge *br = bridge_lookup(br_cfg->name);
1046 bridge_create(br_cfg);
1050 shash_destroy(&new_br);
1054 may_port_action(long long int *port_action_timer)
1056 if (need_reconfigure) {
1061 if (*port_action_timer == LLONG_MAX) {
1062 *port_action_timer = time_msec() + OFP_PORT_ACTION_WINDOW;
1063 } else if (time_msec() > *port_action_timer) {
1064 need_reconfigure = true;
1070 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1073 * The kernel will reject any attempt to add a given port to a datapath if that
1074 * port already belongs to a different datapath, so we must do all port
1075 * deletions before any port additions. */
1077 bridge_del_ofproto_ports(struct bridge *br,
1078 long long int *port_action_timer)
1080 struct ofproto_port_dump dump;
1081 struct ofproto_port ofproto_port;
1083 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1084 const char *name = ofproto_port.name;
1085 struct iface *iface;
1089 /* Ignore the local port. We can't change it anyhow. */
1090 if (!strcmp(name, br->name)) {
1094 /* Get the type that 'ofproto_port' should have (ordinarily the
1095 * type of its corresponding iface) or NULL if it should be
1097 iface = iface_lookup(br, name);
1098 type = (iface ? iface->type
1099 : bridge_has_bond_fake_iface(br, name) ? "internal"
1102 /* If it's the wrong type then delete the ofproto port. */
1104 && !strcmp(ofproto_port.type, type)
1105 && (!iface || !iface->netdev
1106 || !strcmp(netdev_get_type(iface->netdev), type))) {
1110 if (may_port_action(port_action_timer)) {
1111 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1113 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1114 br->name, name, strerror(error));
1116 VLOG_INFO("bridge %s: removed interface %s (%d)", br->name,
1117 name, ofproto_port.ofp_port);
1121 netdev_close(iface->netdev);
1122 iface->netdev = NULL;
1128 iface_set_ofp_port(struct iface *iface, int ofp_port)
1130 struct bridge *br = iface->port->bridge;
1132 assert(iface->ofp_port < 0 && ofp_port >= 0);
1133 iface->ofp_port = ofp_port;
1134 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1135 iface_set_ofport(iface->cfg, ofp_port);
1139 bridge_refresh_ofp_port(struct bridge *br)
1141 struct ofproto_port_dump dump;
1142 struct ofproto_port ofproto_port;
1145 /* Clear all the "ofp_port"es. */
1146 hmap_clear(&br->ifaces);
1147 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1148 struct iface *iface;
1150 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1151 iface->ofp_port = -1;
1155 /* Obtain the correct "ofp_port"s from ofproto. */
1156 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1157 struct iface *iface = iface_lookup(br, ofproto_port.name);
1159 if (iface->ofp_port >= 0) {
1160 VLOG_WARN("bridge %s: interface %s reported twice",
1161 br->name, ofproto_port.name);
1162 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1163 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1164 br->name, ofproto_port.ofp_port);
1166 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1172 /* Add an ofproto port for any "struct iface" that doesn't have one.
1173 * Delete any "struct iface" for which this fails.
1174 * Delete any "struct port" that thereby ends up with no ifaces. */
1176 bridge_add_ofproto_ports(struct bridge *br,
1177 long long int *port_action_timer)
1179 struct port *port, *next_port;
1181 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1182 struct iface *iface, *next_iface;
1183 struct ofproto_port ofproto_port;
1185 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1188 if (iface->ofp_port < 0 && !may_port_action(port_action_timer)) {
1189 iface_clear_db_record(iface->cfg);
1190 iface_destroy(iface);
1194 /* Open the netdev. */
1195 if (!iface->netdev) {
1196 error = netdev_open(iface->name, iface->type, &iface->netdev);
1198 VLOG_WARN("could not open network device %s (%s)",
1199 iface->name, strerror(error));
1203 && port->cfg->vlan_mode
1204 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1205 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1211 /* Configure the netdev. */
1212 if (iface->netdev) {
1216 shash_from_ovs_idl_map(iface->cfg->key_options,
1217 iface->cfg->value_options,
1218 iface->cfg->n_options, &args);
1219 error = netdev_set_config(iface->netdev, &args);
1220 shash_destroy(&args);
1223 VLOG_WARN("could not configure network device %s (%s)",
1224 iface->name, strerror(error));
1225 netdev_close(iface->netdev);
1226 iface->netdev = NULL;
1230 /* Add the port, if necessary. */
1231 if (iface->netdev && iface->ofp_port < 0) {
1235 error = ofproto_port_add(br->ofproto, iface->netdev,
1238 VLOG_INFO("bridge %s: added interface %s (%d)", br->name,
1239 iface->name, ofp_port);
1240 iface_set_ofp_port(iface, ofp_port);
1242 netdev_close(iface->netdev);
1243 iface->netdev = NULL;
1247 /* Populate stats columns in new Interface rows. */
1248 if (iface->netdev && iface->need_refresh) {
1249 iface_refresh_stats(iface);
1250 iface_refresh_status(iface);
1251 iface->need_refresh = false;
1254 /* Delete the iface if we failed. */
1255 if (iface->netdev && iface->ofp_port >= 0) {
1256 VLOG_DBG("bridge %s: interface %s is on port %d",
1257 br->name, iface->name, iface->ofp_port);
1259 if (iface->netdev) {
1260 VLOG_ERR("bridge %s: missing %s interface, dropping",
1261 br->name, iface->name);
1263 /* We already reported a related error, don't bother
1264 * duplicating it. */
1266 if (!ofproto_port_query_by_name(br->ofproto, port->name,
1268 VLOG_INFO("bridge %s: removed interface %s (%d)",
1269 br->name, port->name, ofproto_port.ofp_port);
1270 ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1271 ofproto_port_destroy(&ofproto_port);
1273 iface_clear_db_record(iface->cfg);
1274 iface_destroy(iface);
1278 if (list_is_empty(&port->ifaces)) {
1279 if (!need_reconfigure) {
1280 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1286 /* Add bond fake iface if necessary. */
1287 if (port_is_bond_fake_iface(port)) {
1288 if (ofproto_port_query_by_name(br->ofproto, port->name,
1290 struct netdev *netdev;
1293 error = netdev_open(port->name, "internal", &netdev);
1295 /* There are unlikely to be a great number of fake
1296 * interfaces so we don't bother rate limiting their
1298 ofproto_port_add(br->ofproto, netdev, NULL);
1299 netdev_close(netdev);
1301 VLOG_WARN("could not open network device %s (%s)",
1302 port->name, strerror(error));
1305 /* Already exists, nothing to do. */
1306 ofproto_port_destroy(&ofproto_port);
1312 /* Set Flow eviction threshold */
1314 bridge_configure_flow_eviction_threshold(struct bridge *br)
1316 const char *threshold_str;
1320 ovsrec_bridge_get_other_config_value(br->cfg,
1321 "flow-eviction-threshold",
1323 if (threshold_str) {
1324 threshold = strtoul(threshold_str, NULL, 10);
1326 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1328 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1331 /* Set forward BPDU option. */
1333 bridge_configure_forward_bpdu(struct bridge *br)
1335 const char *forward_bpdu_str;
1336 bool forward_bpdu = false;
1338 forward_bpdu_str = ovsrec_bridge_get_other_config_value(br->cfg,
1341 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1342 forward_bpdu = true;
1344 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1347 /* Set MAC aging time for 'br'. */
1349 bridge_configure_mac_idle_time(struct bridge *br)
1351 const char *idle_time_str;
1354 idle_time_str = ovsrec_bridge_get_other_config_value(br->cfg,
1357 idle_time = (idle_time_str && atoi(idle_time_str)
1358 ? atoi(idle_time_str)
1359 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1360 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1364 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1365 struct iface **hw_addr_iface)
1367 struct hmapx mirror_output_ports;
1370 bool found_addr = false;
1374 *hw_addr_iface = NULL;
1376 /* Did the user request a particular MAC? */
1377 hwaddr = ovsrec_bridge_get_other_config_value(br->cfg, "hwaddr", NULL);
1378 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1379 if (eth_addr_is_multicast(ea)) {
1380 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1381 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1382 } else if (eth_addr_is_zero(ea)) {
1383 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1389 /* Mirror output ports don't participate in picking the local hardware
1390 * address. ofproto can't help us find out whether a given port is a
1391 * mirror output because we haven't configured mirrors yet, so we need to
1392 * accumulate them ourselves. */
1393 hmapx_init(&mirror_output_ports);
1394 for (i = 0; i < br->cfg->n_mirrors; i++) {
1395 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1396 if (m->output_port) {
1397 hmapx_add(&mirror_output_ports, m->output_port);
1401 /* Otherwise choose the minimum non-local MAC address among all of the
1403 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1404 uint8_t iface_ea[ETH_ADDR_LEN];
1405 struct iface *candidate;
1406 struct iface *iface;
1408 /* Mirror output ports don't participate. */
1409 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1413 /* Choose the MAC address to represent the port. */
1415 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1416 /* Find the interface with this Ethernet address (if any) so that
1417 * we can provide the correct devname to the caller. */
1418 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1419 uint8_t candidate_ea[ETH_ADDR_LEN];
1420 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1421 && eth_addr_equals(iface_ea, candidate_ea)) {
1426 /* Choose the interface whose MAC address will represent the port.
1427 * The Linux kernel bonding code always chooses the MAC address of
1428 * the first slave added to a bond, and the Fedora networking
1429 * scripts always add slaves to a bond in alphabetical order, so
1430 * for compatibility we choose the interface with the name that is
1431 * first in alphabetical order. */
1432 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1433 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1438 /* The local port doesn't count (since we're trying to choose its
1439 * MAC address anyway). */
1440 if (iface->ofp_port == OFPP_LOCAL) {
1445 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1451 /* Compare against our current choice. */
1452 if (!eth_addr_is_multicast(iface_ea) &&
1453 !eth_addr_is_local(iface_ea) &&
1454 !eth_addr_is_reserved(iface_ea) &&
1455 !eth_addr_is_zero(iface_ea) &&
1456 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1458 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1459 *hw_addr_iface = iface;
1464 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1465 br->name, ETH_ADDR_ARGS(ea));
1467 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
1468 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1469 *hw_addr_iface = NULL;
1470 VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
1471 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1474 hmapx_destroy(&mirror_output_ports);
1477 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1478 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1479 * an interface on 'br', then that interface must be passed in as
1480 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1481 * 'hw_addr_iface' must be passed in as a null pointer. */
1483 bridge_pick_datapath_id(struct bridge *br,
1484 const uint8_t bridge_ea[ETH_ADDR_LEN],
1485 struct iface *hw_addr_iface)
1488 * The procedure for choosing a bridge MAC address will, in the most
1489 * ordinary case, also choose a unique MAC that we can use as a datapath
1490 * ID. In some special cases, though, multiple bridges will end up with
1491 * the same MAC address. This is OK for the bridges, but it will confuse
1492 * the OpenFlow controller, because each datapath needs a unique datapath
1495 * Datapath IDs must be unique. It is also very desirable that they be
1496 * stable from one run to the next, so that policy set on a datapath
1499 const char *datapath_id;
1502 datapath_id = ovsrec_bridge_get_other_config_value(br->cfg, "datapath-id",
1504 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1508 if (!hw_addr_iface) {
1510 * A purely internal bridge, that is, one that has no non-virtual
1511 * network devices on it at all, is difficult because it has no
1512 * natural unique identifier at all.
1514 * When the host is a XenServer, we handle this case by hashing the
1515 * host's UUID with the name of the bridge. Names of bridges are
1516 * persistent across XenServer reboots, although they can be reused if
1517 * an internal network is destroyed and then a new one is later
1518 * created, so this is fairly effective.
1520 * When the host is not a XenServer, we punt by using a random MAC
1521 * address on each run.
1523 const char *host_uuid = xenserver_get_host_uuid();
1525 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1526 dpid = dpid_from_hash(combined, strlen(combined));
1532 return eth_addr_to_uint64(bridge_ea);
1536 dpid_from_hash(const void *data, size_t n)
1538 uint8_t hash[SHA1_DIGEST_SIZE];
1540 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1541 sha1_bytes(data, n, hash);
1542 eth_addr_mark_random(hash);
1543 return eth_addr_to_uint64(hash);
1547 iface_refresh_status(struct iface *iface)
1551 enum netdev_features current;
1552 enum netdev_flags flags;
1558 if (iface_is_synthetic(iface)) {
1564 if (!netdev_get_drv_info(iface->netdev, &sh)) {
1566 char **keys, **values;
1568 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1569 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1574 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1577 shash_destroy_free_data(&sh);
1579 error = netdev_get_flags(iface->netdev, &flags);
1581 ovsrec_interface_set_admin_state(iface->cfg,
1582 flags & NETDEV_UP ? "up" : "down");
1585 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1588 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1590 ovsrec_interface_set_duplex(iface->cfg,
1591 netdev_features_is_full_duplex(current)
1593 /* warning: uint64_t -> int64_t conversion */
1594 bps = netdev_features_to_bps(current);
1595 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1598 ovsrec_interface_set_duplex(iface->cfg, NULL);
1599 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1602 error = netdev_get_mtu(iface->netdev, &mtu);
1605 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1608 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1612 /* Writes 'iface''s CFM statistics to the database. */
1614 iface_refresh_cfm_stats(struct iface *iface)
1616 const struct ovsrec_interface *cfg = iface->cfg;
1618 const uint64_t *rmps;
1622 if (iface_is_synthetic(iface)) {
1626 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1629 const char *reasons[CFM_FAULT_N_REASONS];
1630 bool fault_bool = fault;
1634 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1635 int reason = 1 << i;
1636 if (fault & reason) {
1637 reasons[j++] = cfm_fault_reason_to_str(reason);
1641 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1642 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1644 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1645 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1648 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1649 iface->ofp_port, &rmps, &n_rmps);
1651 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1654 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1657 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1660 int64_t cfm_health = health;
1661 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1663 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1668 iface_refresh_stats(struct iface *iface)
1670 #define IFACE_STATS \
1671 IFACE_STAT(rx_packets, "rx_packets") \
1672 IFACE_STAT(tx_packets, "tx_packets") \
1673 IFACE_STAT(rx_bytes, "rx_bytes") \
1674 IFACE_STAT(tx_bytes, "tx_bytes") \
1675 IFACE_STAT(rx_dropped, "rx_dropped") \
1676 IFACE_STAT(tx_dropped, "tx_dropped") \
1677 IFACE_STAT(rx_errors, "rx_errors") \
1678 IFACE_STAT(tx_errors, "tx_errors") \
1679 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1680 IFACE_STAT(rx_over_errors, "rx_over_err") \
1681 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1682 IFACE_STAT(collisions, "collisions")
1684 #define IFACE_STAT(MEMBER, NAME) NAME,
1685 static char *keys[] = { IFACE_STATS };
1687 int64_t values[ARRAY_SIZE(keys)];
1690 struct netdev_stats stats;
1692 if (iface_is_synthetic(iface)) {
1696 /* Intentionally ignore return value, since errors will set 'stats' to
1697 * all-1s, and we will deal with that correctly below. */
1698 netdev_get_stats(iface->netdev, &stats);
1700 /* Copy statistics into values[] array. */
1702 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1705 assert(i == ARRAY_SIZE(keys));
1707 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1713 br_refresh_stp_status(struct bridge *br)
1715 struct ofproto *ofproto = br->ofproto;
1716 struct ofproto_stp_status status;
1717 char *keys[3], *values[3];
1720 if (ofproto_get_stp_status(ofproto, &status)) {
1724 if (!status.enabled) {
1725 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1729 keys[0] = "stp_bridge_id",
1730 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1731 keys[1] = "stp_designated_root",
1732 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1733 keys[2] = "stp_root_path_cost",
1734 values[2] = xasprintf("%d", status.root_path_cost);
1736 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1738 for (i = 0; i < ARRAY_SIZE(values); i++) {
1744 port_refresh_stp_status(struct port *port)
1746 struct ofproto *ofproto = port->bridge->ofproto;
1747 struct iface *iface;
1748 struct ofproto_port_stp_status status;
1750 char *str_values[4];
1751 int64_t int_values[3];
1754 if (port_is_synthetic(port)) {
1758 /* STP doesn't currently support bonds. */
1759 if (!list_is_singleton(&port->ifaces)) {
1760 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1764 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1766 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1770 if (!status.enabled) {
1771 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1772 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1776 /* Set Status column. */
1777 keys[0] = "stp_port_id";
1778 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1779 keys[1] = "stp_state";
1780 str_values[1] = xstrdup(stp_state_name(status.state));
1781 keys[2] = "stp_sec_in_state";
1782 str_values[2] = xasprintf("%u", status.sec_in_state);
1783 keys[3] = "stp_role";
1784 str_values[3] = xstrdup(stp_role_name(status.role));
1786 ovsrec_port_set_status(port->cfg, keys, str_values,
1787 ARRAY_SIZE(str_values));
1789 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1790 free(str_values[i]);
1793 /* Set Statistics column. */
1794 keys[0] = "stp_tx_count";
1795 int_values[0] = status.tx_count;
1796 keys[1] = "stp_rx_count";
1797 int_values[1] = status.rx_count;
1798 keys[2] = "stp_error_count";
1799 int_values[2] = status.error_count;
1801 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1802 ARRAY_SIZE(int_values));
1806 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1810 /* Use other-config:enable-system-stats by preference. */
1811 enable = ovsrec_open_vswitch_get_other_config_value(cfg,
1812 "enable-statistics",
1815 return !strcmp(enable, "true");
1818 /* Disable by default. */
1823 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1825 struct ovsdb_datum datum;
1829 if (enable_system_stats(cfg)) {
1830 get_system_stats(&stats);
1833 ovsdb_datum_from_shash(&datum, &stats);
1834 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1838 static inline const char *
1839 nx_role_to_str(enum nx_role role)
1844 case NX_ROLE_MASTER:
1849 return "*** INVALID ROLE ***";
1854 refresh_controller_status(void)
1858 const struct ovsrec_controller *cfg;
1862 /* Accumulate status for controllers on all bridges. */
1863 HMAP_FOR_EACH (br, node, &all_bridges) {
1864 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1867 /* Update each controller in the database with current status. */
1868 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1869 struct ofproto_controller_info *cinfo =
1870 shash_find_data(&info, cfg->target);
1873 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1874 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1875 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1876 (char **) cinfo->pairs.values,
1879 ovsrec_controller_set_is_connected(cfg, false);
1880 ovsrec_controller_set_role(cfg, NULL);
1881 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1885 ofproto_free_ofproto_controller_info(&info);
1889 refresh_cfm_stats(void)
1891 static struct ovsdb_idl_txn *txn = NULL;
1896 txn = ovsdb_idl_txn_create(idl);
1898 HMAP_FOR_EACH (br, node, &all_bridges) {
1899 struct iface *iface;
1901 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1902 iface_refresh_cfm_stats(iface);
1907 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1908 ovsdb_idl_txn_destroy(txn);
1913 /* Performs periodic activity required by bridges that needs to be done with
1914 * the least possible latency.
1916 * It makes sense to call this function a couple of times per poll loop, to
1917 * provide a significant performance boost on some benchmarks with ofprotos
1918 * that use the ofproto-dpif implementation. */
1920 bridge_run_fast(void)
1924 HMAP_FOR_EACH (br, node, &all_bridges) {
1925 ofproto_run_fast(br->ofproto);
1932 const struct ovsrec_open_vswitch *cfg;
1934 bool vlan_splinters_changed;
1937 /* (Re)configure if necessary. */
1939 if (ovsdb_idl_is_lock_contended(idl)) {
1940 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1941 struct bridge *br, *next_br;
1943 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1944 "disabling this process until it goes away");
1946 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1950 } else if (!ovsdb_idl_has_lock(idl)) {
1953 cfg = ovsrec_open_vswitch_first(idl);
1955 /* Let each bridge do the work that it needs to do. */
1956 HMAP_FOR_EACH (br, node, &all_bridges) {
1957 ofproto_run(br->ofproto);
1960 /* Re-configure SSL. We do this on every trip through the main loop,
1961 * instead of just when the database changes, because the contents of the
1962 * key and certificate files can change without the database changing.
1964 * We do this before bridge_reconfigure() because that function might
1965 * initiate SSL connections and thus requires SSL to be configured. */
1966 if (cfg && cfg->ssl) {
1967 const struct ovsrec_ssl *ssl = cfg->ssl;
1969 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1970 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1973 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1975 vlan_splinters_changed = false;
1976 if (vlan_splinters_enabled_anywhere) {
1977 HMAP_FOR_EACH (br, node, &all_bridges) {
1978 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1979 vlan_splinters_changed = true;
1985 if (need_reconfigure || ovsdb_idl_get_seqno(idl) != idl_seqno
1986 || vlan_splinters_changed) {
1987 idl_seqno = ovsdb_idl_get_seqno(idl);
1989 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1991 bridge_reconfigure(cfg);
1993 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1994 ovsdb_idl_txn_commit(txn);
1995 ovsdb_idl_txn_destroy(txn); /* XXX */
1997 /* We still need to reconfigure to avoid dangling pointers to
1998 * now-destroyed ovsrec structures inside bridge data. */
1999 static const struct ovsrec_open_vswitch null_cfg;
2001 bridge_reconfigure(&null_cfg);
2005 /* Refresh system and interface stats if necessary. */
2006 if (time_msec() >= stats_timer) {
2008 struct ovsdb_idl_txn *txn;
2010 txn = ovsdb_idl_txn_create(idl);
2011 HMAP_FOR_EACH (br, node, &all_bridges) {
2015 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2016 struct iface *iface;
2018 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2019 iface_refresh_stats(iface);
2020 iface_refresh_status(iface);
2024 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2025 mirror_refresh_stats(m);
2029 refresh_system_stats(cfg);
2030 refresh_controller_status();
2031 ovsdb_idl_txn_commit(txn);
2032 ovsdb_idl_txn_destroy(txn); /* XXX */
2035 stats_timer = time_msec() + STATS_INTERVAL;
2038 if (time_msec() >= db_limiter) {
2039 struct ovsdb_idl_txn *txn;
2041 txn = ovsdb_idl_txn_create(idl);
2042 HMAP_FOR_EACH (br, node, &all_bridges) {
2043 struct iface *iface;
2046 br_refresh_stp_status(br);
2048 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2049 port_refresh_stp_status(port);
2052 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2053 const char *link_state;
2054 int64_t link_resets;
2057 if (iface_is_synthetic(iface)) {
2061 current = ofproto_port_is_lacp_current(br->ofproto,
2065 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2067 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2070 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2071 ovsrec_interface_set_link_state(iface->cfg, link_state);
2073 link_resets = netdev_get_carrier_resets(iface->netdev);
2074 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2078 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2079 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2081 ovsdb_idl_txn_destroy(txn);
2084 refresh_cfm_stats();
2090 ovsdb_idl_wait(idl);
2092 if (need_reconfigure) {
2093 poll_immediate_wake();
2096 if (!hmap_is_empty(&all_bridges)) {
2099 HMAP_FOR_EACH (br, node, &all_bridges) {
2100 ofproto_wait(br->ofproto);
2102 poll_timer_wait_until(stats_timer);
2104 if (db_limiter > time_msec()) {
2105 poll_timer_wait_until(db_limiter);
2110 /* QoS unixctl user interface functions. */
2112 struct qos_unixctl_show_cbdata {
2114 struct iface *iface;
2118 qos_unixctl_show_cb(unsigned int queue_id,
2119 const struct shash *details,
2122 struct qos_unixctl_show_cbdata *data = aux;
2123 struct ds *ds = data->ds;
2124 struct iface *iface = data->iface;
2125 struct netdev_queue_stats stats;
2126 struct shash_node *node;
2129 ds_put_cstr(ds, "\n");
2131 ds_put_format(ds, "Queue %u:\n", queue_id);
2133 ds_put_cstr(ds, "Default:\n");
2136 SHASH_FOR_EACH (node, details) {
2137 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2140 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2142 if (stats.tx_packets != UINT64_MAX) {
2143 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2146 if (stats.tx_bytes != UINT64_MAX) {
2147 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2150 if (stats.tx_errors != UINT64_MAX) {
2151 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2154 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2155 queue_id, strerror(error));
2160 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2161 const char *argv[], void *aux OVS_UNUSED)
2163 struct ds ds = DS_EMPTY_INITIALIZER;
2164 struct shash sh = SHASH_INITIALIZER(&sh);
2165 struct iface *iface;
2167 struct shash_node *node;
2168 struct qos_unixctl_show_cbdata data;
2171 iface = iface_find(argv[1]);
2173 unixctl_command_reply_error(conn, "no such interface");
2177 netdev_get_qos(iface->netdev, &type, &sh);
2179 if (*type != '\0') {
2180 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2182 SHASH_FOR_EACH (node, &sh) {
2183 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2188 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2191 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2193 unixctl_command_reply(conn, ds_cstr(&ds));
2195 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2196 unixctl_command_reply_error(conn, ds_cstr(&ds));
2199 shash_destroy_free_data(&sh);
2203 /* Bridge reconfiguration functions. */
2205 bridge_create(const struct ovsrec_bridge *br_cfg)
2209 assert(!bridge_lookup(br_cfg->name));
2210 br = xzalloc(sizeof *br);
2212 br->name = xstrdup(br_cfg->name);
2213 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2216 /* Derive the default Ethernet address from the bridge's UUID. This should
2217 * be unique and it will be stable between ovs-vswitchd runs. */
2218 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2219 eth_addr_mark_random(br->default_ea);
2221 hmap_init(&br->ports);
2222 hmap_init(&br->ifaces);
2223 hmap_init(&br->iface_by_name);
2224 hmap_init(&br->mirrors);
2226 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2230 bridge_destroy(struct bridge *br)
2233 struct mirror *mirror, *next_mirror;
2234 struct port *port, *next_port;
2236 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2239 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2240 mirror_destroy(mirror);
2242 hmap_remove(&all_bridges, &br->node);
2243 ofproto_destroy(br->ofproto);
2244 hmap_destroy(&br->ifaces);
2245 hmap_destroy(&br->ports);
2246 hmap_destroy(&br->iface_by_name);
2247 hmap_destroy(&br->mirrors);
2254 static struct bridge *
2255 bridge_lookup(const char *name)
2259 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2260 if (!strcmp(br->name, name)) {
2267 /* Handle requests for a listing of all flows known by the OpenFlow
2268 * stack, including those normally hidden. */
2270 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2271 const char *argv[], void *aux OVS_UNUSED)
2276 br = bridge_lookup(argv[1]);
2278 unixctl_command_reply_error(conn, "Unknown bridge");
2283 ofproto_get_all_flows(br->ofproto, &results);
2285 unixctl_command_reply(conn, ds_cstr(&results));
2286 ds_destroy(&results);
2289 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2290 * connections and reconnect. If BRIDGE is not specified, then all bridges
2291 * drop their controller connections and reconnect. */
2293 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2294 const char *argv[], void *aux OVS_UNUSED)
2298 br = bridge_lookup(argv[1]);
2300 unixctl_command_reply_error(conn, "Unknown bridge");
2303 ofproto_reconnect_controllers(br->ofproto);
2305 HMAP_FOR_EACH (br, node, &all_bridges) {
2306 ofproto_reconnect_controllers(br->ofproto);
2309 unixctl_command_reply(conn, NULL);
2313 bridge_get_controllers(const struct bridge *br,
2314 struct ovsrec_controller ***controllersp)
2316 struct ovsrec_controller **controllers;
2317 size_t n_controllers;
2319 controllers = br->cfg->controller;
2320 n_controllers = br->cfg->n_controller;
2322 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2328 *controllersp = controllers;
2330 return n_controllers;
2333 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2334 * those configured in 'br->cfg'. */
2336 bridge_add_del_ports(struct bridge *br,
2337 const unsigned long int *splinter_vlans)
2339 struct port *port, *next;
2340 struct shash_node *node;
2341 struct shash new_ports;
2344 /* Collect new ports. */
2345 shash_init(&new_ports);
2346 for (i = 0; i < br->cfg->n_ports; i++) {
2347 const char *name = br->cfg->ports[i]->name;
2348 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2349 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2353 if (bridge_get_controllers(br, NULL)
2354 && !shash_find(&new_ports, br->name)) {
2355 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2356 br->name, br->name);
2358 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2359 br->synth_local_port.n_interfaces = 1;
2360 br->synth_local_port.name = br->name;
2362 br->synth_local_iface.name = br->name;
2363 br->synth_local_iface.type = "internal";
2365 br->synth_local_ifacep = &br->synth_local_iface;
2367 shash_add(&new_ports, br->name, &br->synth_local_port);
2370 if (splinter_vlans) {
2371 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2374 /* Get rid of deleted ports.
2375 * Get rid of deleted interfaces on ports that still exist. */
2376 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2377 port->cfg = shash_find_data(&new_ports, port->name);
2381 port_del_ifaces(port);
2385 /* Create new ports.
2386 * Add new interfaces to existing ports. */
2387 SHASH_FOR_EACH (node, &new_ports) {
2388 struct port *port = port_lookup(br, node->name);
2390 struct ovsrec_port *cfg = node->data;
2391 port = port_create(br, cfg);
2393 port_add_ifaces(port);
2394 if (list_is_empty(&port->ifaces)) {
2395 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2396 br->name, port->name);
2400 shash_destroy(&new_ports);
2403 /* Initializes 'oc' appropriately as a management service controller for
2406 * The caller must free oc->target when it is no longer needed. */
2408 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2409 struct ofproto_controller *oc)
2411 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2412 oc->max_backoff = 0;
2413 oc->probe_interval = 60;
2414 oc->band = OFPROTO_OUT_OF_BAND;
2416 oc->burst_limit = 0;
2417 oc->enable_async_msgs = true;
2420 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2422 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2423 struct ofproto_controller *oc)
2425 const char *config_str;
2427 oc->target = c->target;
2428 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2429 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2430 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2431 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2432 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2433 oc->burst_limit = (c->controller_burst_limit
2434 ? *c->controller_burst_limit : 0);
2435 oc->enable_async_msgs = (!c->enable_async_messages
2436 || *c->enable_async_messages);
2437 config_str = ovsrec_controller_get_other_config_value(c, "dscp", NULL);
2439 oc->dscp = DSCP_DEFAULT;
2441 int dscp = atoi(config_str);
2443 if (dscp >= 0 && dscp <= 63) {
2449 /* Configures the IP stack for 'br''s local interface properly according to the
2450 * configuration in 'c'. */
2452 bridge_configure_local_iface_netdev(struct bridge *br,
2453 struct ovsrec_controller *c)
2455 struct netdev *netdev;
2456 struct in_addr mask, gateway;
2458 struct iface *local_iface;
2461 /* If there's no local interface or no IP address, give up. */
2462 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2463 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2467 /* Bring up the local interface. */
2468 netdev = local_iface->netdev;
2469 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2471 /* Configure the IP address and netmask. */
2472 if (!c->local_netmask
2473 || !inet_aton(c->local_netmask, &mask)
2475 mask.s_addr = guess_netmask(ip.s_addr);
2477 if (!netdev_set_in4(netdev, ip, mask)) {
2478 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2479 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2482 /* Configure the default gateway. */
2483 if (c->local_gateway
2484 && inet_aton(c->local_gateway, &gateway)
2485 && gateway.s_addr) {
2486 if (!netdev_add_router(netdev, gateway)) {
2487 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2488 br->name, IP_ARGS(&gateway.s_addr));
2493 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2494 * in either string are treated as equal to any number of slashes in the other,
2495 * e.g. "x///y" is equal to "x/y". */
2497 equal_pathnames(const char *a, const char *b)
2501 a += strspn(a, "/");
2502 b += strspn(b, "/");
2503 } else if (*a == '\0') {
2514 bridge_configure_remotes(struct bridge *br,
2515 const struct sockaddr_in *managers, size_t n_managers)
2517 const char *disable_ib_str, *queue_id_str;
2518 bool disable_in_band = false;
2521 struct ovsrec_controller **controllers;
2522 size_t n_controllers;
2524 enum ofproto_fail_mode fail_mode;
2526 struct ofproto_controller *ocs;
2530 /* Check if we should disable in-band control on this bridge. */
2531 disable_ib_str = ovsrec_bridge_get_other_config_value(br->cfg,
2534 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2535 disable_in_band = true;
2538 /* Set OpenFlow queue ID for in-band control. */
2539 queue_id_str = ovsrec_bridge_get_other_config_value(br->cfg,
2542 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2543 ofproto_set_in_band_queue(br->ofproto, queue_id);
2545 if (disable_in_band) {
2546 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2548 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2551 n_controllers = bridge_get_controllers(br, &controllers);
2553 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2556 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2557 for (i = 0; i < n_controllers; i++) {
2558 struct ovsrec_controller *c = controllers[i];
2560 if (!strncmp(c->target, "punix:", 6)
2561 || !strncmp(c->target, "unix:", 5)) {
2562 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2565 whitelist = xasprintf("unix:%s/%s.controller",
2566 ovs_rundir(), br->name);
2567 if (!equal_pathnames(c->target, whitelist)) {
2568 /* Prevent remote ovsdb-server users from accessing arbitrary
2569 * Unix domain sockets and overwriting arbitrary local
2571 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2572 "controller \"%s\" due to possibility for remote "
2573 "exploit. Instead, specify whitelisted \"%s\" or "
2574 "connect to \"unix:%s/%s.mgmt\" (which is always "
2575 "available without special configuration).",
2576 br->name, c->target, whitelist,
2577 ovs_rundir(), br->name);
2585 bridge_configure_local_iface_netdev(br, c);
2586 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2587 if (disable_in_band) {
2588 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2593 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2594 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2597 /* Set the fail-mode. */
2598 fail_mode = !br->cfg->fail_mode
2599 || !strcmp(br->cfg->fail_mode, "standalone")
2600 ? OFPROTO_FAIL_STANDALONE
2601 : OFPROTO_FAIL_SECURE;
2602 ofproto_set_fail_mode(br->ofproto, fail_mode);
2604 /* Configure OpenFlow controller connection snooping. */
2605 if (!ofproto_has_snoops(br->ofproto)) {
2609 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2610 ovs_rundir(), br->name));
2611 ofproto_set_snoops(br->ofproto, &snoops);
2612 sset_destroy(&snoops);
2617 bridge_configure_tables(struct bridge *br)
2619 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2623 n_tables = ofproto_get_n_tables(br->ofproto);
2625 for (i = 0; i < n_tables; i++) {
2626 struct ofproto_table_settings s;
2629 s.max_flows = UINT_MAX;
2633 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2634 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2637 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2638 s.max_flows = *cfg->flow_limit;
2640 if (cfg->overflow_policy
2641 && !strcmp(cfg->overflow_policy, "evict")) {
2644 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2645 for (k = 0; k < cfg->n_groups; k++) {
2646 const char *string = cfg->groups[k];
2649 msg = mf_parse_subfield__(&s.groups[k], &string);
2651 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2652 "'groups' (%s)", br->name, i, msg);
2654 } else if (*string) {
2655 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2656 "element '%s' contains trailing garbage",
2657 br->name, i, cfg->groups[k]);
2665 ofproto_configure_table(br->ofproto, i, &s);
2669 for (; j < br->cfg->n_flow_tables; j++) {
2670 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2671 "%"PRId64" not supported by this datapath", br->name,
2672 br->cfg->key_flow_tables[j]);
2676 /* Port functions. */
2678 static struct port *
2679 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2683 port = xzalloc(sizeof *port);
2685 port->name = xstrdup(cfg->name);
2687 list_init(&port->ifaces);
2689 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2693 /* Deletes interfaces from 'port' that are no longer configured for it. */
2695 port_del_ifaces(struct port *port)
2697 struct iface *iface, *next;
2698 struct sset new_ifaces;
2701 /* Collect list of new interfaces. */
2702 sset_init(&new_ifaces);
2703 for (i = 0; i < port->cfg->n_interfaces; i++) {
2704 const char *name = port->cfg->interfaces[i]->name;
2705 const char *type = port->cfg->interfaces[i]->name;
2706 if (strcmp(type, "null")) {
2707 sset_add(&new_ifaces, name);
2711 /* Get rid of deleted interfaces. */
2712 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2713 if (!sset_contains(&new_ifaces, iface->name)) {
2714 iface_destroy(iface);
2718 sset_destroy(&new_ifaces);
2721 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2724 port_add_ifaces(struct port *port)
2726 struct shash new_ifaces;
2727 struct shash_node *node;
2730 /* Collect new ifaces. */
2731 shash_init(&new_ifaces);
2732 for (i = 0; i < port->cfg->n_interfaces; i++) {
2733 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2734 if (strcmp(cfg->type, "null")
2735 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2736 VLOG_WARN("port %s: %s specified twice as port interface",
2737 port->name, cfg->name);
2738 iface_clear_db_record(cfg);
2742 /* Create new interfaces.
2743 * Update interface types and 'cfg' members. */
2744 SHASH_FOR_EACH (node, &new_ifaces) {
2745 const struct ovsrec_interface *cfg = node->data;
2746 const char *iface_name = node->name;
2747 struct iface *iface;
2749 iface = iface_lookup(port->bridge, iface_name);
2751 iface = iface_create(port, cfg);
2756 /* Determine interface type. The local port always has type
2757 * "internal". Other ports take their type from the database and
2758 * default to "system" if none is specified. */
2759 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2760 : cfg->type[0] ? cfg->type
2763 shash_destroy(&new_ifaces);
2767 port_destroy(struct port *port)
2770 struct bridge *br = port->bridge;
2771 struct iface *iface, *next;
2774 ofproto_bundle_unregister(br->ofproto, port);
2777 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2778 iface_destroy(iface);
2781 hmap_remove(&br->ports, &port->hmap_node);
2787 static struct port *
2788 port_lookup(const struct bridge *br, const char *name)
2792 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2794 if (!strcmp(port->name, name)) {
2802 enable_lacp(struct port *port, bool *activep)
2804 if (!port->cfg->lacp) {
2805 /* XXX when LACP implementation has been sufficiently tested, enable by
2806 * default and make active on bonded ports. */
2808 } else if (!strcmp(port->cfg->lacp, "off")) {
2810 } else if (!strcmp(port->cfg->lacp, "active")) {
2813 } else if (!strcmp(port->cfg->lacp, "passive")) {
2817 VLOG_WARN("port %s: unknown LACP mode %s",
2818 port->name, port->cfg->lacp);
2823 static struct lacp_settings *
2824 port_configure_lacp(struct port *port, struct lacp_settings *s)
2826 const char *lacp_time, *system_id;
2829 if (!enable_lacp(port, &s->active)) {
2833 s->name = port->name;
2835 system_id = ovsrec_port_get_other_config_value(port->cfg, "lacp-system-id",
2838 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2839 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2840 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2841 " address.", port->name, system_id);
2845 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2848 if (eth_addr_is_zero(s->id)) {
2849 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2853 /* Prefer bondable links if unspecified. */
2854 priority = atoi(ovsrec_port_get_other_config_value(port->cfg,
2855 "lacp-system-priority",
2857 s->priority = (priority > 0 && priority <= UINT16_MAX
2859 : UINT16_MAX - !list_is_short(&port->ifaces));
2861 lacp_time = ovsrec_port_get_other_config_value(port->cfg, "lacp-time",
2863 s->fast = !strcasecmp(lacp_time, "fast");
2868 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2870 int priority, portid, key;
2872 portid = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2876 atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2877 "lacp-port-priority",
2879 key = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2880 "lacp-aggregation-key",
2883 if (portid <= 0 || portid > UINT16_MAX) {
2884 portid = iface->ofp_port;
2887 if (priority <= 0 || priority > UINT16_MAX) {
2888 priority = UINT16_MAX;
2891 if (key < 0 || key > UINT16_MAX) {
2895 s->name = iface->name;
2897 s->priority = priority;
2902 port_configure_bond(struct port *port, struct bond_settings *s,
2903 uint32_t *bond_stable_ids)
2905 const char *detect_s;
2906 struct iface *iface;
2907 int miimon_interval;
2910 s->name = port->name;
2912 if (port->cfg->bond_mode) {
2913 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2914 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2915 port->name, port->cfg->bond_mode,
2916 bond_mode_to_string(s->balance));
2919 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2921 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2922 * active-backup. At some point we should remove this warning. */
2923 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2924 " in previous versions, the default bond_mode was"
2925 " balance-slb", port->name,
2926 bond_mode_to_string(s->balance));
2928 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2929 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2930 "please use another bond type or disable flood_vlans",
2935 atoi(ovsrec_port_get_other_config_value(port->cfg,
2936 "bond-miimon-interval", "0"));
2937 if (miimon_interval <= 0) {
2938 miimon_interval = 200;
2941 detect_s = ovsrec_port_get_other_config_value(port->cfg,
2944 if (!strcmp(detect_s, "carrier")) {
2945 miimon_interval = 0;
2946 } else if (strcmp(detect_s, "miimon")) {
2947 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2948 "defaulting to carrier", port->name, detect_s);
2949 miimon_interval = 0;
2952 s->up_delay = MAX(0, port->cfg->bond_updelay);
2953 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2954 s->basis = atoi(ovsrec_port_get_other_config_value(port->cfg,
2957 s->rebalance_interval = atoi(
2958 ovsrec_port_get_other_config_value(port->cfg,
2959 "bond-rebalance-interval",
2961 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2962 s->rebalance_interval = 1000;
2965 s->fake_iface = port->cfg->bond_fake_iface;
2968 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2969 long long stable_id;
2972 atoll(ovsrec_interface_get_other_config_value(iface->cfg,
2975 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2976 stable_id = iface->ofp_port;
2978 bond_stable_ids[i++] = stable_id;
2980 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2984 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2985 * instead of obtaining it from the database. */
2987 port_is_synthetic(const struct port *port)
2989 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2992 /* Interface functions. */
2994 static struct iface *
2995 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2997 struct bridge *br = port->bridge;
2998 struct iface *iface;
2999 char *name = if_cfg->name;
3001 iface = xzalloc(sizeof *iface);
3003 iface->name = xstrdup(name);
3004 iface->ofp_port = -1;
3005 iface->netdev = NULL;
3006 iface->cfg = if_cfg;
3007 iface->need_refresh = true;
3009 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
3011 list_push_back(&port->ifaces, &iface->port_elem);
3013 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3019 iface_destroy(struct iface *iface)
3022 struct port *port = iface->port;
3023 struct bridge *br = port->bridge;
3025 if (br->ofproto && iface->ofp_port >= 0) {
3026 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3029 if (iface->ofp_port >= 0) {
3030 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3033 list_remove(&iface->port_elem);
3034 hmap_remove(&br->iface_by_name, &iface->name_node);
3036 netdev_close(iface->netdev);
3043 static struct iface *
3044 iface_lookup(const struct bridge *br, const char *name)
3046 struct iface *iface;
3048 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3049 &br->iface_by_name) {
3050 if (!strcmp(iface->name, name)) {
3058 static struct iface *
3059 iface_find(const char *name)
3061 const struct bridge *br;
3063 HMAP_FOR_EACH (br, node, &all_bridges) {
3064 struct iface *iface = iface_lookup(br, name);
3073 static struct iface *
3074 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3076 struct iface *iface;
3078 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3079 hash_int(ofp_port, 0), &br->ifaces) {
3080 if (iface->ofp_port == ofp_port) {
3087 /* Set Ethernet address of 'iface', if one is specified in the configuration
3090 iface_set_mac(struct iface *iface)
3092 uint8_t ea[ETH_ADDR_LEN];
3094 if (!strcmp(iface->type, "internal")
3095 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3096 if (iface->ofp_port == OFPP_LOCAL) {
3097 VLOG_ERR("interface %s: ignoring mac in Interface record "
3098 "(use Bridge record to set local port's mac)",
3100 } else if (eth_addr_is_multicast(ea)) {
3101 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3104 int error = netdev_set_etheraddr(iface->netdev, ea);
3106 VLOG_ERR("interface %s: setting MAC failed (%s)",
3107 iface->name, strerror(error));
3113 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3115 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3117 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3118 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3122 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3123 * sets the "ofport" field to -1.
3125 * This is appropriate when 'if_cfg''s interface cannot be created or is
3126 * otherwise invalid. */
3128 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3130 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3131 iface_set_ofport(if_cfg, -1);
3132 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3133 ovsrec_interface_set_admin_state(if_cfg, NULL);
3134 ovsrec_interface_set_duplex(if_cfg, NULL);
3135 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3136 ovsrec_interface_set_link_state(if_cfg, NULL);
3137 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3138 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3139 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3140 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3141 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3142 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3146 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3148 * The value strings in '*shash' are taken directly from values[], not copied,
3149 * so the caller should not modify or free them. */
3151 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3152 struct shash *shash)
3157 for (i = 0; i < n; i++) {
3158 shash_add(shash, keys[i], values[i]);
3162 /* Creates 'keys' and 'values' arrays from 'shash'.
3164 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3165 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3166 * are populated with with strings taken directly from 'shash' and thus have
3167 * the same ownership of the key-value pairs in shash.
3170 shash_to_ovs_idl_map(struct shash *shash,
3171 char ***keys, char ***values, size_t *n)
3175 struct shash_node *sn;
3177 count = shash_count(shash);
3179 k = xmalloc(count * sizeof *k);
3180 v = xmalloc(count * sizeof *v);
3183 SHASH_FOR_EACH(sn, shash) {
3194 struct iface_delete_queues_cbdata {
3195 struct netdev *netdev;
3196 const struct ovsdb_datum *queues;
3200 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3202 union ovsdb_atom atom;
3204 atom.integer = target;
3205 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3209 iface_delete_queues(unsigned int queue_id,
3210 const struct shash *details OVS_UNUSED, void *cbdata_)
3212 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3214 if (!queue_ids_include(cbdata->queues, queue_id)) {
3215 netdev_delete_queue(cbdata->netdev, queue_id);
3220 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3222 struct ofpbuf queues_buf;
3224 ofpbuf_init(&queues_buf, 0);
3226 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3227 netdev_set_qos(iface->netdev, NULL, NULL);
3229 struct iface_delete_queues_cbdata cbdata;
3230 struct shash details;
3234 /* Configure top-level Qos for 'iface'. */
3235 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3236 qos->n_other_config, &details);
3237 netdev_set_qos(iface->netdev, qos->type, &details);
3238 shash_destroy(&details);
3240 /* Deconfigure queues that were deleted. */
3241 cbdata.netdev = iface->netdev;
3242 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3244 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3246 /* Configure queues for 'iface'. */
3248 for (i = 0; i < qos->n_queues; i++) {
3249 const struct ovsrec_queue *queue = qos->value_queues[i];
3250 unsigned int queue_id = qos->key_queues[i];
3252 if (queue_id == 0) {
3256 if (queue->n_dscp == 1) {
3257 struct ofproto_port_queue *port_queue;
3259 port_queue = ofpbuf_put_uninit(&queues_buf,
3260 sizeof *port_queue);
3261 port_queue->queue = queue_id;
3262 port_queue->dscp = queue->dscp[0];
3265 shash_from_ovs_idl_map(queue->key_other_config,
3266 queue->value_other_config,
3267 queue->n_other_config, &details);
3268 netdev_set_queue(iface->netdev, queue_id, &details);
3269 shash_destroy(&details);
3272 shash_init(&details);
3273 netdev_set_queue(iface->netdev, 0, &details);
3274 shash_destroy(&details);
3278 if (iface->ofp_port >= 0) {
3279 const struct ofproto_port_queue *port_queues = queues_buf.data;
3280 size_t n_queues = queues_buf.size / sizeof *port_queues;
3282 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3283 port_queues, n_queues);
3286 netdev_set_policing(iface->netdev,
3287 iface->cfg->ingress_policing_rate,
3288 iface->cfg->ingress_policing_burst);
3290 ofpbuf_uninit(&queues_buf);
3294 iface_configure_cfm(struct iface *iface)
3296 const struct ovsrec_interface *cfg = iface->cfg;
3297 const char *extended_str, *opstate_str;
3298 const char *cfm_ccm_vlan;
3299 struct cfm_settings s;
3301 if (!cfg->n_cfm_mpid) {
3302 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3306 s.mpid = *cfg->cfm_mpid;
3307 s.interval = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3310 cfm_ccm_vlan = ovsrec_interface_get_other_config_value(iface->cfg,
3313 s.ccm_pcp = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3316 if (s.interval <= 0) {
3320 if (!strcasecmp("random", cfm_ccm_vlan)) {
3321 s.ccm_vlan = CFM_RANDOM_VLAN;
3323 s.ccm_vlan = atoi(cfm_ccm_vlan);
3324 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3329 extended_str = ovsrec_interface_get_other_config_value(iface->cfg,
3332 s.extended = !strcasecmp("true", extended_str);
3334 opstate_str = ovsrec_interface_get_other_config_value(iface->cfg,
3337 s.opup = !strcasecmp("up", opstate_str);
3339 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3342 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3343 * instead of obtaining it from the database. */
3345 iface_is_synthetic(const struct iface *iface)
3347 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3351 /* Port mirroring. */
3353 static struct mirror *
3354 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3358 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3359 if (uuid_equals(uuid, &m->uuid)) {
3367 bridge_configure_mirrors(struct bridge *br)
3369 const struct ovsdb_datum *mc;
3370 unsigned long *flood_vlans;
3371 struct mirror *m, *next;
3374 /* Get rid of deleted mirrors. */
3375 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3376 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3377 union ovsdb_atom atom;
3379 atom.uuid = m->uuid;
3380 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3385 /* Add new mirrors and reconfigure existing ones. */
3386 for (i = 0; i < br->cfg->n_mirrors; i++) {
3387 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3388 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3390 m = mirror_create(br, cfg);
3393 if (!mirror_configure(m)) {
3398 /* Update flooded vlans (for RSPAN). */
3399 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3400 br->cfg->n_flood_vlans);
3401 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3402 bitmap_free(flood_vlans);
3405 static struct mirror *
3406 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3410 m = xzalloc(sizeof *m);
3411 m->uuid = cfg->header_.uuid;
3412 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3414 m->name = xstrdup(cfg->name);
3420 mirror_destroy(struct mirror *m)
3423 struct bridge *br = m->bridge;
3426 ofproto_mirror_unregister(br->ofproto, m);
3429 hmap_remove(&br->mirrors, &m->hmap_node);
3436 mirror_collect_ports(struct mirror *m,
3437 struct ovsrec_port **in_ports, int n_in_ports,
3438 void ***out_portsp, size_t *n_out_portsp)
3440 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3441 size_t n_out_ports = 0;
3444 for (i = 0; i < n_in_ports; i++) {
3445 const char *name = in_ports[i]->name;
3446 struct port *port = port_lookup(m->bridge, name);
3448 out_ports[n_out_ports++] = port;
3450 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3451 "port %s", m->bridge->name, m->name, name);
3454 *out_portsp = out_ports;
3455 *n_out_portsp = n_out_ports;
3459 mirror_configure(struct mirror *m)
3461 const struct ovsrec_mirror *cfg = m->cfg;
3462 struct ofproto_mirror_settings s;
3465 if (strcmp(cfg->name, m->name)) {
3467 m->name = xstrdup(cfg->name);
3471 /* Get output port or VLAN. */
3472 if (cfg->output_port) {
3473 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3474 if (!s.out_bundle) {
3475 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3476 m->bridge->name, m->name);
3479 s.out_vlan = UINT16_MAX;
3481 if (cfg->output_vlan) {
3482 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3483 "output vlan; ignoring output vlan",
3484 m->bridge->name, m->name);
3486 } else if (cfg->output_vlan) {
3487 /* The database should prevent invalid VLAN values. */
3488 s.out_bundle = NULL;
3489 s.out_vlan = *cfg->output_vlan;
3491 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3492 m->bridge->name, m->name);
3496 /* Get port selection. */
3497 if (cfg->select_all) {
3498 size_t n_ports = hmap_count(&m->bridge->ports);
3499 void **ports = xmalloc(n_ports * sizeof *ports);
3504 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3514 /* Get ports, dropping ports that don't exist.
3515 * The IDL ensures that there are no duplicates. */
3516 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3517 &s.srcs, &s.n_srcs);
3518 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3519 &s.dsts, &s.n_dsts);
3522 /* Get VLAN selection. */
3523 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3526 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3529 if (s.srcs != s.dsts) {
3538 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3540 * This is deprecated. It is only for compatibility with broken device drivers
3541 * in old versions of Linux that do not properly support VLANs when VLAN
3542 * devices are not used. When broken device drivers are no longer in
3543 * widespread use, we will delete these interfaces. */
3545 static void **blocks;
3546 static size_t n_blocks, allocated_blocks;
3548 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3549 * VLAN splinters are reconfigured. */
3551 register_block(void *block)
3553 if (n_blocks >= allocated_blocks) {
3554 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3556 blocks[n_blocks++] = block;
3559 /* Frees all of the blocks registered with register_block(). */
3561 free_registered_blocks(void)
3565 for (i = 0; i < n_blocks; i++) {
3571 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3574 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3578 value = ovsrec_interface_get_other_config_value(iface_cfg,
3579 "enable-vlan-splinters",
3581 return !strcmp(value, "true");
3584 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3587 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3588 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3589 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3592 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3593 * use, returns NULL.
3595 * Updates 'vlan_splinters_enabled_anywhere'. */
3596 static unsigned long int *
3597 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3599 unsigned long int *splinter_vlans;
3600 struct sset splinter_ifaces;
3601 const char *real_dev_name;
3602 struct shash *real_devs;
3603 struct shash_node *node;
3607 /* Free space allocated for synthesized ports and interfaces, since we're
3608 * in the process of reconstructing all of them. */
3609 free_registered_blocks();
3611 splinter_vlans = bitmap_allocate(4096);
3612 sset_init(&splinter_ifaces);
3613 vlan_splinters_enabled_anywhere = false;
3614 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3615 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3618 for (j = 0; j < br_cfg->n_ports; j++) {
3619 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3622 for (k = 0; k < port_cfg->n_interfaces; k++) {
3623 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3625 if (vlan_splinters_is_enabled(iface_cfg)) {
3626 vlan_splinters_enabled_anywhere = true;
3627 sset_add(&splinter_ifaces, iface_cfg->name);
3628 vlan_bitmap_from_array__(port_cfg->trunks,
3634 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3635 bitmap_set1(splinter_vlans, *port_cfg->tag);
3640 if (!vlan_splinters_enabled_anywhere) {
3641 free(splinter_vlans);
3642 sset_destroy(&splinter_ifaces);
3646 HMAP_FOR_EACH (br, node, &all_bridges) {
3648 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3652 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3653 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3654 * device to be created for it. */
3655 bitmap_set0(splinter_vlans, 0);
3656 bitmap_set0(splinter_vlans, 4095);
3658 /* Delete all VLAN devices that we don't need. */
3660 real_devs = vlandev_get_real_devs();
3661 SHASH_FOR_EACH (node, real_devs) {
3662 const struct vlan_real_dev *real_dev = node->data;
3663 const struct vlan_dev *vlan_dev;
3664 bool real_dev_has_splinters;
3666 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3668 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3669 if (!real_dev_has_splinters
3670 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3671 struct netdev *netdev;
3673 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3674 if (!netdev_get_in4(netdev, NULL, NULL) ||
3675 !netdev_get_in6(netdev, NULL)) {
3676 vlandev_del(vlan_dev->name);
3678 /* It has an IP address configured, so we don't own
3679 * it. Don't delete it. */
3681 netdev_close(netdev);
3688 /* Add all VLAN devices that we need. */
3689 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3692 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3693 if (!vlandev_get_name(real_dev_name, vid)) {
3694 vlandev_add(real_dev_name, vid);
3701 sset_destroy(&splinter_ifaces);
3703 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3704 free(splinter_vlans);
3707 return splinter_vlans;
3710 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3713 configure_splinter_port(struct port *port)
3715 struct ofproto *ofproto = port->bridge->ofproto;
3716 uint16_t realdev_ofp_port;
3717 const char *realdev_name;
3718 struct iface *vlandev, *realdev;
3720 ofproto_bundle_unregister(port->bridge->ofproto, port);
3722 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3725 realdev_name = ovsrec_port_get_other_config_value(port->cfg,
3727 realdev = iface_lookup(port->bridge, realdev_name);
3728 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3730 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3734 static struct ovsrec_port *
3735 synthesize_splinter_port(const char *real_dev_name,
3736 const char *vlan_dev_name, int vid)
3738 struct ovsrec_interface *iface;
3739 struct ovsrec_port *port;
3741 iface = xzalloc(sizeof *iface);
3742 iface->name = xstrdup(vlan_dev_name);
3743 iface->type = "system";
3745 port = xzalloc(sizeof *port);
3746 port->interfaces = xmemdup(&iface, sizeof iface);
3747 port->n_interfaces = 1;
3748 port->name = xstrdup(vlan_dev_name);
3749 port->vlan_mode = "splinter";
3750 port->tag = xmalloc(sizeof *port->tag);
3752 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3753 port->key_other_config[0] = "realdev";
3754 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3755 port->value_other_config[0] = xstrdup(real_dev_name);
3756 port->n_other_config = 1;
3758 register_block(iface);
3759 register_block(iface->name);
3760 register_block(port);
3761 register_block(port->interfaces);
3762 register_block(port->name);
3763 register_block(port->tag);
3764 register_block(port->key_other_config);
3765 register_block(port->value_other_config);
3766 register_block(port->value_other_config[0]);
3771 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3772 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3773 * 1-bit in the 'splinter_vlans' bitmap. */
3775 add_vlan_splinter_ports(struct bridge *br,
3776 const unsigned long int *splinter_vlans,
3777 struct shash *ports)
3781 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3782 * we're modifying 'ports'. */
3783 for (i = 0; i < br->cfg->n_ports; i++) {
3784 const char *name = br->cfg->ports[i]->name;
3785 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3788 for (j = 0; j < port_cfg->n_interfaces; j++) {
3789 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3791 if (vlan_splinters_is_enabled(iface_cfg)) {
3792 const char *real_dev_name;
3795 real_dev_name = iface_cfg->name;
3796 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3797 const char *vlan_dev_name;
3799 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3801 && !shash_find(ports, vlan_dev_name)) {
3802 shash_add(ports, vlan_dev_name,
3803 synthesize_splinter_port(
3804 real_dev_name, vlan_dev_name, vid));
3813 mirror_refresh_stats(struct mirror *m)
3815 struct ofproto *ofproto = m->bridge->ofproto;
3816 uint64_t tx_packets, tx_bytes;
3819 size_t stat_cnt = 0;
3821 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3822 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3826 if (tx_packets != UINT64_MAX) {
3827 keys[stat_cnt] = "tx_packets";
3828 values[stat_cnt] = tx_packets;
3831 if (tx_bytes != UINT64_MAX) {
3832 keys[stat_cnt] = "tx_bytes";
3833 values[stat_cnt] = tx_bytes;
3837 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);