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. */
69 tag_type tag; /* Tag associated with this interface. */
71 /* These members are valid only after bridge_reconfigure() causes them to
73 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
74 int ofp_port; /* OpenFlow port number, -1 if unknown. */
75 struct netdev *netdev; /* Network device. */
76 const char *type; /* Usually same as cfg->type. */
77 const struct ovsrec_interface *cfg;
78 bool need_refresh; /* Refresh iface after create. */
82 struct uuid uuid; /* UUID of this "mirror" record in database. */
83 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
84 struct bridge *bridge;
86 const struct ovsrec_mirror *cfg;
90 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
91 struct bridge *bridge;
94 const struct ovsrec_port *cfg;
96 /* An ordinary bridge port has 1 interface.
97 * A bridge port for bonding has at least 2 interfaces. */
98 struct list ifaces; /* List of "struct iface"s. */
102 struct hmap_node node; /* In 'all_bridges'. */
103 char *name; /* User-specified arbitrary name. */
104 char *type; /* Datapath type. */
105 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
106 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
107 const struct ovsrec_bridge *cfg;
109 /* OpenFlow switch processing. */
110 struct ofproto *ofproto; /* OpenFlow switch. */
113 struct hmap ports; /* "struct port"s indexed by name. */
114 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
115 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
117 /* Port mirroring. */
118 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
120 /* Synthetic local port if necessary. */
121 struct ovsrec_port synth_local_port;
122 struct ovsrec_interface synth_local_iface;
123 struct ovsrec_interface *synth_local_ifacep;
126 /* All bridges, indexed by name. */
127 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
129 /* OVSDB IDL used to obtain configuration. */
130 static struct ovsdb_idl *idl;
132 /* Each time this timer expires, the bridge fetches systems and interface
133 * statistics and pushes them into the database. */
134 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
135 static long long int stats_timer = LLONG_MIN;
137 /* Stores the time after which rate limited statistics may be written to the
138 * database. Only updated when changes to the database require rate limiting.
140 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
141 static long long int db_limiter = LLONG_MIN;
143 static void add_del_bridges(const struct ovsrec_open_vswitch *);
144 static void bridge_del_ofprotos(void);
145 static bool bridge_add_ofprotos(struct bridge *);
146 static void bridge_create(const struct ovsrec_bridge *);
147 static void bridge_destroy(struct bridge *);
148 static struct bridge *bridge_lookup(const char *name);
149 static unixctl_cb_func bridge_unixctl_dump_flows;
150 static unixctl_cb_func bridge_unixctl_reconnect;
151 static size_t bridge_get_controllers(const struct bridge *br,
152 struct ovsrec_controller ***controllersp);
153 static void bridge_add_del_ports(struct bridge *,
154 const unsigned long int *splinter_vlans);
155 static void bridge_add_ofproto_ports(struct bridge *);
156 static void bridge_del_ofproto_ports(struct bridge *);
157 static void bridge_refresh_ofp_port(struct bridge *);
158 static void bridge_configure_datapath_id(struct bridge *);
159 static void bridge_configure_flow_eviction_threshold(struct bridge *);
160 static void bridge_configure_netflow(struct bridge *);
161 static void bridge_configure_forward_bpdu(struct bridge *);
162 static void bridge_configure_mac_idle_time(struct bridge *);
163 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
164 static void bridge_configure_stp(struct bridge *);
165 static void bridge_configure_tables(struct bridge *);
166 static void bridge_configure_remotes(struct bridge *,
167 const struct sockaddr_in *managers,
169 static void bridge_pick_local_hw_addr(struct bridge *,
170 uint8_t ea[ETH_ADDR_LEN],
171 struct iface **hw_addr_iface);
172 static uint64_t bridge_pick_datapath_id(struct bridge *,
173 const uint8_t bridge_ea[ETH_ADDR_LEN],
174 struct iface *hw_addr_iface);
175 static uint64_t dpid_from_hash(const void *, size_t nbytes);
176 static bool bridge_has_bond_fake_iface(const struct bridge *,
178 static bool port_is_bond_fake_iface(const struct port *);
180 static unixctl_cb_func qos_unixctl_show;
182 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
183 static void port_add_ifaces(struct port *);
184 static void port_del_ifaces(struct port *);
185 static void port_destroy(struct port *);
186 static struct port *port_lookup(const struct bridge *, const char *name);
187 static void port_configure(struct port *);
188 static struct lacp_settings *port_configure_lacp(struct port *,
189 struct lacp_settings *);
190 static void port_configure_bond(struct port *, struct bond_settings *,
191 uint32_t *bond_stable_ids);
192 static bool port_is_synthetic(const struct port *);
194 static void bridge_configure_mirrors(struct bridge *);
195 static struct mirror *mirror_create(struct bridge *,
196 const struct ovsrec_mirror *);
197 static void mirror_destroy(struct mirror *);
198 static bool mirror_configure(struct mirror *);
199 static void mirror_refresh_stats(struct mirror *);
201 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
202 static struct iface *iface_create(struct port *port,
203 const struct ovsrec_interface *if_cfg);
204 static void iface_destroy(struct iface *);
205 static struct iface *iface_lookup(const struct bridge *, const char *name);
206 static struct iface *iface_find(const char *name);
207 static struct iface *iface_from_ofp_port(const struct bridge *,
209 static void iface_set_mac(struct iface *);
210 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
211 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
212 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
213 static void iface_configure_cfm(struct iface *);
214 static void iface_refresh_cfm_stats(struct iface *);
215 static void iface_refresh_stats(struct iface *);
216 static void iface_refresh_status(struct iface *);
217 static bool iface_is_synthetic(const struct iface *);
218 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
220 static void shash_to_ovs_idl_map(struct shash *,
221 char ***keys, char ***values, size_t *n);
223 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
225 * This is deprecated. It is only for compatibility with broken device drivers
226 * in old versions of Linux that do not properly support VLANs when VLAN
227 * devices are not used. When broken device drivers are no longer in
228 * widespread use, we will delete these interfaces. */
230 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
231 static bool vlan_splinters_enabled_anywhere;
233 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
234 static unsigned long int *collect_splinter_vlans(
235 const struct ovsrec_open_vswitch *);
236 static void configure_splinter_port(struct port *);
237 static void add_vlan_splinter_ports(struct bridge *,
238 const unsigned long int *splinter_vlans,
239 struct shash *ports);
241 /* Public functions. */
243 /* Initializes the bridge module, configuring it to obtain its configuration
244 * from an OVSDB server accessed over 'remote', which should be a string in a
245 * form acceptable to ovsdb_idl_create(). */
247 bridge_init(const char *remote)
249 /* Create connection to database. */
250 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
251 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
253 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
254 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
255 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
257 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
258 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
259 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
261 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
262 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
263 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
265 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
266 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
267 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
268 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
270 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
271 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
272 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
273 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
274 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
275 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
276 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
277 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
278 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
279 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
280 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
281 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
282 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
283 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
284 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
286 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
287 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
288 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
289 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
291 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
293 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
296 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
298 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
300 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
302 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
303 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
304 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
305 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
306 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
308 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
310 /* Register unixctl commands. */
311 unixctl_command_register("qos/show", "interface", 1, 1,
312 qos_unixctl_show, NULL);
313 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
314 bridge_unixctl_dump_flows, NULL);
315 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
316 bridge_unixctl_reconnect, NULL);
326 struct bridge *br, *next_br;
328 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
331 ovsdb_idl_destroy(idl);
334 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
335 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
336 * responsible for freeing '*managersp' (with free()).
338 * You may be asking yourself "why does ovs-vswitchd care?", because
339 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
340 * should not be and in fact is not directly involved in that. But
341 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
342 * it has to tell in-band control where the managers are to enable that.
343 * (Thus, only managers connected in-band are collected.)
346 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
347 struct sockaddr_in **managersp, size_t *n_managersp)
349 struct sockaddr_in *managers = NULL;
350 size_t n_managers = 0;
354 /* Collect all of the potential targets from the "targets" columns of the
355 * rows pointed to by "manager_options", excluding any that are
358 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
359 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
361 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
362 sset_find_and_delete(&targets, m->target);
364 sset_add(&targets, m->target);
368 /* Now extract the targets' IP addresses. */
369 if (!sset_is_empty(&targets)) {
372 managers = xmalloc(sset_count(&targets) * sizeof *managers);
373 SSET_FOR_EACH (target, &targets) {
374 struct sockaddr_in *sin = &managers[n_managers];
376 if (stream_parse_target_with_default_ports(target,
384 sset_destroy(&targets);
386 *managersp = managers;
387 *n_managersp = n_managers;
391 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
393 unsigned long int *splinter_vlans;
394 struct sockaddr_in *managers;
395 struct bridge *br, *next;
396 int sflow_bridge_number;
399 COVERAGE_INC(bridge_reconfigure);
401 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
402 * iface"s according to 'ovs_cfg', with only very minimal configuration
405 * This is mostly an update to bridge data structures. Very little is
406 * pushed down to ofproto or lower layers. */
407 add_del_bridges(ovs_cfg);
408 splinter_vlans = collect_splinter_vlans(ovs_cfg);
409 HMAP_FOR_EACH (br, node, &all_bridges) {
410 bridge_add_del_ports(br, splinter_vlans);
412 free(splinter_vlans);
414 /* Delete all datapaths and datapath ports that are no longer configured.
416 * The kernel will reject any attempt to add a given port to a datapath if
417 * that port already belongs to a different datapath, so we must do all
418 * port deletions before any port additions. A datapath always has a
419 * "local port" so we must delete not-configured datapaths too. */
420 bridge_del_ofprotos();
421 HMAP_FOR_EACH (br, node, &all_bridges) {
423 bridge_del_ofproto_ports(br);
427 /* Create datapaths and datapath ports that are missing.
429 * After this is done, we have our final set of bridges, ports, and
430 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
431 * has at least one iface, every "struct iface" has a valid ofp_port and
433 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
435 if (bridge_add_ofprotos(br)) {
436 bridge_del_ofproto_ports(br);
442 HMAP_FOR_EACH (br, node, &all_bridges) {
443 bridge_refresh_ofp_port(br);
444 bridge_add_ofproto_ports(br);
447 /* Complete the configuration. */
448 sflow_bridge_number = 0;
449 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
450 HMAP_FOR_EACH (br, node, &all_bridges) {
453 /* We need the datapath ID early to allow LACP ports to use it as the
454 * default system ID. */
455 bridge_configure_datapath_id(br);
457 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
460 port_configure(port);
462 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
463 iface_configure_cfm(iface);
464 iface_configure_qos(iface, port->cfg->qos);
465 iface_set_mac(iface);
468 bridge_configure_mirrors(br);
469 bridge_configure_flow_eviction_threshold(br);
470 bridge_configure_forward_bpdu(br);
471 bridge_configure_mac_idle_time(br);
472 bridge_configure_remotes(br, managers, n_managers);
473 bridge_configure_netflow(br);
474 bridge_configure_sflow(br, &sflow_bridge_number);
475 bridge_configure_stp(br);
476 bridge_configure_tables(br);
480 /* ovs-vswitchd has completed initialization, so allow the process that
481 * forked us to exit successfully. */
482 daemonize_complete();
485 /* Iterate over all ofprotos and delete any of them that do not have a
486 * configured bridge or that are the wrong type. */
488 bridge_del_ofprotos(void)
496 ofproto_enumerate_types(&types);
497 SSET_FOR_EACH (type, &types) {
500 ofproto_enumerate_names(type, &names);
501 SSET_FOR_EACH (name, &names) {
502 struct bridge *br = bridge_lookup(name);
503 if (!br || strcmp(type, br->type)) {
504 ofproto_delete(name, type);
508 sset_destroy(&names);
509 sset_destroy(&types);
513 bridge_add_ofprotos(struct bridge *br)
515 int error = ofproto_create(br->name, br->type, &br->ofproto);
517 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
524 port_configure(struct port *port)
526 const struct ovsrec_port *cfg = port->cfg;
527 struct bond_settings bond_settings;
528 struct lacp_settings lacp_settings;
529 struct ofproto_bundle_settings s;
532 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
533 configure_splinter_port(port);
542 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
543 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
544 s.slaves[s.n_slaves++] = iface->ofp_port;
549 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
553 /* Get VLAN trunks. */
556 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
560 if (cfg->vlan_mode) {
561 if (!strcmp(cfg->vlan_mode, "access")) {
562 s.vlan_mode = PORT_VLAN_ACCESS;
563 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
564 s.vlan_mode = PORT_VLAN_TRUNK;
565 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
566 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
567 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
568 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
570 /* This "can't happen" because ovsdb-server should prevent it. */
571 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
572 s.vlan_mode = PORT_VLAN_TRUNK;
576 s.vlan_mode = PORT_VLAN_ACCESS;
578 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
582 s.vlan_mode = PORT_VLAN_TRUNK;
585 s.use_priority_tags = !strcmp("true", ovsrec_port_get_other_config_value(
586 cfg, "priority-tags", ""));
588 /* Get LACP settings. */
589 s.lacp = port_configure_lacp(port, &lacp_settings);
593 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
594 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
595 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
598 s.lacp_slaves = NULL;
601 /* Get bond settings. */
602 if (s.n_slaves > 1) {
603 s.bond = &bond_settings;
604 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
605 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
608 s.bond_stable_ids = NULL;
610 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
611 netdev_set_miimon_interval(iface->netdev, 0);
616 ofproto_bundle_register(port->bridge->ofproto, port, &s);
622 free(s.bond_stable_ids);
625 /* Pick local port hardware address and datapath ID for 'br'. */
627 bridge_configure_datapath_id(struct bridge *br)
629 uint8_t ea[ETH_ADDR_LEN];
631 struct iface *local_iface;
632 struct iface *hw_addr_iface;
635 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
636 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
638 int error = netdev_set_etheraddr(local_iface->netdev, ea);
640 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
641 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
642 "Ethernet address: %s",
643 br->name, strerror(error));
646 memcpy(br->ea, ea, ETH_ADDR_LEN);
648 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
649 ofproto_set_datapath_id(br->ofproto, dpid);
651 dpid_string = xasprintf("%016"PRIx64, dpid);
652 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
656 /* Set NetFlow configuration on 'br'. */
658 bridge_configure_netflow(struct bridge *br)
660 struct ovsrec_netflow *cfg = br->cfg->netflow;
661 struct netflow_options opts;
664 ofproto_set_netflow(br->ofproto, NULL);
668 memset(&opts, 0, sizeof opts);
670 /* Get default NetFlow configuration from datapath.
671 * Apply overrides from 'cfg'. */
672 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
673 if (cfg->engine_type) {
674 opts.engine_type = *cfg->engine_type;
676 if (cfg->engine_id) {
677 opts.engine_id = *cfg->engine_id;
680 /* Configure active timeout interval. */
681 opts.active_timeout = cfg->active_timeout;
682 if (!opts.active_timeout) {
683 opts.active_timeout = -1;
684 } else if (opts.active_timeout < 0) {
685 VLOG_WARN("bridge %s: active timeout interval set to negative "
686 "value, using default instead (%d seconds)", br->name,
687 NF_ACTIVE_TIMEOUT_DEFAULT);
688 opts.active_timeout = -1;
691 /* Add engine ID to interface number to disambiguate bridgs? */
692 opts.add_id_to_iface = cfg->add_id_to_interface;
693 if (opts.add_id_to_iface) {
694 if (opts.engine_id > 0x7f) {
695 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
696 "another vswitch, choose an engine id less than 128",
699 if (hmap_count(&br->ports) > 508) {
700 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
701 "another port when more than 508 ports are used",
707 sset_init(&opts.collectors);
708 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
711 if (ofproto_set_netflow(br->ofproto, &opts)) {
712 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
714 sset_destroy(&opts.collectors);
717 /* Set sFlow configuration on 'br'. */
719 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
721 const struct ovsrec_sflow *cfg = br->cfg->sflow;
722 struct ovsrec_controller **controllers;
723 struct ofproto_sflow_options oso;
724 size_t n_controllers;
728 ofproto_set_sflow(br->ofproto, NULL);
732 memset(&oso, 0, sizeof oso);
734 sset_init(&oso.targets);
735 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
737 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
739 oso.sampling_rate = *cfg->sampling;
742 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
744 oso.polling_interval = *cfg->polling;
747 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
749 oso.header_len = *cfg->header;
752 oso.sub_id = (*sflow_bridge_number)++;
753 oso.agent_device = cfg->agent;
755 oso.control_ip = NULL;
756 n_controllers = bridge_get_controllers(br, &controllers);
757 for (i = 0; i < n_controllers; i++) {
758 if (controllers[i]->local_ip) {
759 oso.control_ip = controllers[i]->local_ip;
763 ofproto_set_sflow(br->ofproto, &oso);
765 sset_destroy(&oso.targets);
769 port_configure_stp(const struct ofproto *ofproto, struct port *port,
770 struct ofproto_port_stp_settings *port_s,
771 int *port_num_counter, unsigned long *port_num_bitmap)
773 const char *config_str;
776 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-enable",
778 if (config_str && !strcmp(config_str, "false")) {
779 port_s->enable = false;
782 port_s->enable = true;
785 /* STP over bonds is not supported. */
786 if (!list_is_singleton(&port->ifaces)) {
787 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
789 port_s->enable = false;
793 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
795 /* Internal ports shouldn't participate in spanning tree, so
797 if (!strcmp(iface->type, "internal")) {
798 VLOG_DBG("port %s: disable STP on internal ports", port->name);
799 port_s->enable = false;
803 /* STP on mirror output ports is not supported. */
804 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
805 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
806 port_s->enable = false;
810 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-port-num",
813 unsigned long int port_num = strtoul(config_str, NULL, 0);
814 int port_idx = port_num - 1;
816 if (port_num < 1 || port_num > STP_MAX_PORTS) {
817 VLOG_ERR("port %s: invalid stp-port-num", port->name);
818 port_s->enable = false;
822 if (bitmap_is_set(port_num_bitmap, port_idx)) {
823 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
824 port->name, port_num);
825 port_s->enable = false;
828 bitmap_set1(port_num_bitmap, port_idx);
829 port_s->port_num = port_idx;
831 if (*port_num_counter > STP_MAX_PORTS) {
832 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
833 port_s->enable = false;
837 port_s->port_num = (*port_num_counter)++;
840 config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-path-cost",
843 port_s->path_cost = strtoul(config_str, NULL, 10);
845 enum netdev_features current;
847 if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) {
848 /* Couldn't get speed, so assume 100Mb/s. */
849 port_s->path_cost = 19;
853 mbps = netdev_features_to_bps(current) / 1000000;
854 port_s->path_cost = stp_convert_speed_to_cost(mbps);
858 config_str = ovsrec_port_get_other_config_value(port->cfg,
862 port_s->priority = strtoul(config_str, NULL, 0);
864 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
868 /* Set spanning tree configuration on 'br'. */
870 bridge_configure_stp(struct bridge *br)
872 if (!br->cfg->stp_enable) {
873 ofproto_set_stp(br->ofproto, NULL);
875 struct ofproto_stp_settings br_s;
876 const char *config_str;
878 int port_num_counter;
879 unsigned long *port_num_bitmap;
881 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
885 uint8_t ea[ETH_ADDR_LEN];
887 if (eth_addr_from_string(config_str, ea)) {
888 br_s.system_id = eth_addr_to_uint64(ea);
890 br_s.system_id = eth_addr_to_uint64(br->ea);
891 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
892 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
895 br_s.system_id = eth_addr_to_uint64(br->ea);
898 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
902 br_s.priority = strtoul(config_str, NULL, 0);
904 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
907 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
911 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
913 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
916 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
920 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
922 br_s.max_age = STP_DEFAULT_MAX_AGE;
925 config_str = ovsrec_bridge_get_other_config_value(br->cfg,
929 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
931 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
934 /* Configure STP on the bridge. */
935 if (ofproto_set_stp(br->ofproto, &br_s)) {
936 VLOG_ERR("bridge %s: could not enable STP", br->name);
940 /* Users must either set the port number with the "stp-port-num"
941 * configuration on all ports or none. If manual configuration
942 * is not done, then we allocate them sequentially. */
943 port_num_counter = 0;
944 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
945 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
946 struct ofproto_port_stp_settings port_s;
949 port_configure_stp(br->ofproto, port, &port_s,
950 &port_num_counter, port_num_bitmap);
952 /* As bonds are not supported, just apply configuration to
954 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
955 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
957 VLOG_ERR("port %s: could not enable STP", port->name);
963 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
964 && port_num_counter) {
965 VLOG_ERR("bridge %s: must manually configure all STP port "
966 "IDs or none, disabling", br->name);
967 ofproto_set_stp(br->ofproto, NULL);
969 bitmap_free(port_num_bitmap);
974 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
976 const struct port *port = port_lookup(br, name);
977 return port && port_is_bond_fake_iface(port);
981 port_is_bond_fake_iface(const struct port *port)
983 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
987 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
989 struct bridge *br, *next;
993 /* Collect new bridges' names and types. */
995 for (i = 0; i < cfg->n_bridges; i++) {
996 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
997 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
999 if (strchr(br_cfg->name, '/')) {
1000 /* Prevent remote ovsdb-server users from accessing arbitrary
1001 * directories, e.g. consider a bridge named "../../../etc/". */
1002 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1004 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1005 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1009 /* Get rid of deleted bridges or those whose types have changed.
1010 * Update 'cfg' of bridges that still exist. */
1011 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1012 br->cfg = shash_find_data(&new_br, br->name);
1013 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1014 br->cfg->datapath_type))) {
1019 /* Add new bridges. */
1020 for (i = 0; i < cfg->n_bridges; i++) {
1021 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1022 struct bridge *br = bridge_lookup(br_cfg->name);
1024 bridge_create(br_cfg);
1028 shash_destroy(&new_br);
1031 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
1034 * The kernel will reject any attempt to add a given port to a datapath if that
1035 * port already belongs to a different datapath, so we must do all port
1036 * deletions before any port additions. */
1038 bridge_del_ofproto_ports(struct bridge *br)
1040 struct ofproto_port_dump dump;
1041 struct ofproto_port ofproto_port;
1043 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1044 const char *name = ofproto_port.name;
1045 struct iface *iface;
1049 /* Ignore the local port. We can't change it anyhow. */
1050 if (!strcmp(name, br->name)) {
1054 /* Get the type that 'ofproto_port' should have (ordinarily the
1055 * type of its corresponding iface) or NULL if it should be
1057 iface = iface_lookup(br, name);
1058 type = (iface ? iface->type
1059 : bridge_has_bond_fake_iface(br, name) ? "internal"
1062 /* If it's the wrong type then delete the ofproto port. */
1064 && !strcmp(ofproto_port.type, type)
1065 && (!iface || !iface->netdev
1066 || !strcmp(netdev_get_type(iface->netdev), type))) {
1069 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1071 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
1072 br->name, name, strerror(error));
1075 netdev_close(iface->netdev);
1076 iface->netdev = NULL;
1082 iface_set_ofp_port(struct iface *iface, int ofp_port)
1084 struct bridge *br = iface->port->bridge;
1086 assert(iface->ofp_port < 0 && ofp_port >= 0);
1087 iface->ofp_port = ofp_port;
1088 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1089 iface_set_ofport(iface->cfg, ofp_port);
1093 bridge_refresh_ofp_port(struct bridge *br)
1095 struct ofproto_port_dump dump;
1096 struct ofproto_port ofproto_port;
1099 /* Clear all the "ofp_port"es. */
1100 hmap_clear(&br->ifaces);
1101 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1102 struct iface *iface;
1104 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1105 iface->ofp_port = -1;
1109 /* Obtain the correct "ofp_port"s from ofproto. */
1110 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1111 struct iface *iface = iface_lookup(br, ofproto_port.name);
1113 if (iface->ofp_port >= 0) {
1114 VLOG_WARN("bridge %s: interface %s reported twice",
1115 br->name, ofproto_port.name);
1116 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
1117 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1118 br->name, ofproto_port.ofp_port);
1120 iface_set_ofp_port(iface, ofproto_port.ofp_port);
1126 /* Add an ofproto port for any "struct iface" that doesn't have one.
1127 * Delete any "struct iface" for which this fails.
1128 * Delete any "struct port" that thereby ends up with no ifaces. */
1130 bridge_add_ofproto_ports(struct bridge *br)
1132 struct port *port, *next_port;
1134 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1135 struct iface *iface, *next_iface;
1136 struct ofproto_port ofproto_port;
1138 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
1141 /* Open the netdev. */
1142 if (!iface->netdev) {
1143 error = netdev_open(iface->name, iface->type, &iface->netdev);
1145 VLOG_WARN("could not open network device %s (%s)",
1146 iface->name, strerror(error));
1150 && port->cfg->vlan_mode
1151 && !strcmp(port->cfg->vlan_mode, "splinter")) {
1152 netdev_turn_flags_on(iface->netdev, NETDEV_UP, true);
1158 /* Configure the netdev. */
1159 if (iface->netdev) {
1163 shash_from_ovs_idl_map(iface->cfg->key_options,
1164 iface->cfg->value_options,
1165 iface->cfg->n_options, &args);
1166 error = netdev_set_config(iface->netdev, &args);
1167 shash_destroy(&args);
1170 VLOG_WARN("could not configure network device %s (%s)",
1171 iface->name, strerror(error));
1172 netdev_close(iface->netdev);
1173 iface->netdev = NULL;
1177 /* Add the port, if necessary. */
1178 if (iface->netdev && iface->ofp_port < 0) {
1182 error = ofproto_port_add(br->ofproto, iface->netdev,
1185 iface_set_ofp_port(iface, ofp_port);
1187 netdev_close(iface->netdev);
1188 iface->netdev = NULL;
1192 /* Populate stats columns in new Interface rows. */
1193 if (iface->netdev && iface->need_refresh) {
1194 iface_refresh_stats(iface);
1195 iface_refresh_status(iface);
1196 iface->need_refresh = false;
1199 /* Delete the iface if we failed. */
1200 if (iface->netdev && iface->ofp_port >= 0) {
1201 VLOG_DBG("bridge %s: interface %s is on port %d",
1202 br->name, iface->name, iface->ofp_port);
1204 if (iface->netdev) {
1205 VLOG_ERR("bridge %s: missing %s interface, dropping",
1206 br->name, iface->name);
1208 /* We already reported a related error, don't bother
1209 * duplicating it. */
1211 if (!ofproto_port_query_by_name(br->ofproto, port->name,
1213 ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
1214 ofproto_port_destroy(&ofproto_port);
1216 iface_clear_db_record(iface->cfg);
1217 iface_destroy(iface);
1220 if (list_is_empty(&port->ifaces)) {
1221 VLOG_WARN("%s port has no interfaces, dropping", port->name);
1226 /* Add bond fake iface if necessary. */
1227 if (port_is_bond_fake_iface(port)) {
1228 if (ofproto_port_query_by_name(br->ofproto, port->name,
1230 struct netdev *netdev;
1233 error = netdev_open(port->name, "internal", &netdev);
1235 ofproto_port_add(br->ofproto, netdev, NULL);
1236 netdev_close(netdev);
1238 VLOG_WARN("could not open network device %s (%s)",
1239 port->name, strerror(error));
1242 /* Already exists, nothing to do. */
1243 ofproto_port_destroy(&ofproto_port);
1249 /* Set Flow eviction threshold */
1251 bridge_configure_flow_eviction_threshold(struct bridge *br)
1253 const char *threshold_str;
1257 ovsrec_bridge_get_other_config_value(br->cfg,
1258 "flow-eviction-threshold",
1260 if (threshold_str) {
1261 threshold = strtoul(threshold_str, NULL, 10);
1263 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
1265 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1268 /* Set forward BPDU option. */
1270 bridge_configure_forward_bpdu(struct bridge *br)
1272 const char *forward_bpdu_str;
1273 bool forward_bpdu = false;
1275 forward_bpdu_str = ovsrec_bridge_get_other_config_value(br->cfg,
1278 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1279 forward_bpdu = true;
1281 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1284 /* Set MAC aging time for 'br'. */
1286 bridge_configure_mac_idle_time(struct bridge *br)
1288 const char *idle_time_str;
1291 idle_time_str = ovsrec_bridge_get_other_config_value(br->cfg,
1294 idle_time = (idle_time_str && atoi(idle_time_str)
1295 ? atoi(idle_time_str)
1296 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1297 ofproto_set_mac_idle_time(br->ofproto, idle_time);
1301 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1302 struct iface **hw_addr_iface)
1304 struct hmapx mirror_output_ports;
1307 bool found_addr = false;
1311 *hw_addr_iface = NULL;
1313 /* Did the user request a particular MAC? */
1314 hwaddr = ovsrec_bridge_get_other_config_value(br->cfg, "hwaddr", NULL);
1315 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1316 if (eth_addr_is_multicast(ea)) {
1317 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1318 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1319 } else if (eth_addr_is_zero(ea)) {
1320 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1326 /* Mirror output ports don't participate in picking the local hardware
1327 * address. ofproto can't help us find out whether a given port is a
1328 * mirror output because we haven't configured mirrors yet, so we need to
1329 * accumulate them ourselves. */
1330 hmapx_init(&mirror_output_ports);
1331 for (i = 0; i < br->cfg->n_mirrors; i++) {
1332 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1333 if (m->output_port) {
1334 hmapx_add(&mirror_output_ports, m->output_port);
1338 /* Otherwise choose the minimum non-local MAC address among all of the
1340 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1341 uint8_t iface_ea[ETH_ADDR_LEN];
1342 struct iface *candidate;
1343 struct iface *iface;
1345 /* Mirror output ports don't participate. */
1346 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1350 /* Choose the MAC address to represent the port. */
1352 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1353 /* Find the interface with this Ethernet address (if any) so that
1354 * we can provide the correct devname to the caller. */
1355 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1356 uint8_t candidate_ea[ETH_ADDR_LEN];
1357 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1358 && eth_addr_equals(iface_ea, candidate_ea)) {
1363 /* Choose the interface whose MAC address will represent the port.
1364 * The Linux kernel bonding code always chooses the MAC address of
1365 * the first slave added to a bond, and the Fedora networking
1366 * scripts always add slaves to a bond in alphabetical order, so
1367 * for compatibility we choose the interface with the name that is
1368 * first in alphabetical order. */
1369 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1370 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1375 /* The local port doesn't count (since we're trying to choose its
1376 * MAC address anyway). */
1377 if (iface->ofp_port == OFPP_LOCAL) {
1382 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1388 /* Compare against our current choice. */
1389 if (!eth_addr_is_multicast(iface_ea) &&
1390 !eth_addr_is_local(iface_ea) &&
1391 !eth_addr_is_reserved(iface_ea) &&
1392 !eth_addr_is_zero(iface_ea) &&
1393 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1395 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1396 *hw_addr_iface = iface;
1401 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1402 br->name, ETH_ADDR_ARGS(ea));
1404 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1405 *hw_addr_iface = NULL;
1406 VLOG_WARN("bridge %s: using default bridge Ethernet "
1407 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1410 hmapx_destroy(&mirror_output_ports);
1413 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1414 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1415 * an interface on 'br', then that interface must be passed in as
1416 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1417 * 'hw_addr_iface' must be passed in as a null pointer. */
1419 bridge_pick_datapath_id(struct bridge *br,
1420 const uint8_t bridge_ea[ETH_ADDR_LEN],
1421 struct iface *hw_addr_iface)
1424 * The procedure for choosing a bridge MAC address will, in the most
1425 * ordinary case, also choose a unique MAC that we can use as a datapath
1426 * ID. In some special cases, though, multiple bridges will end up with
1427 * the same MAC address. This is OK for the bridges, but it will confuse
1428 * the OpenFlow controller, because each datapath needs a unique datapath
1431 * Datapath IDs must be unique. It is also very desirable that they be
1432 * stable from one run to the next, so that policy set on a datapath
1435 const char *datapath_id;
1438 datapath_id = ovsrec_bridge_get_other_config_value(br->cfg, "datapath-id",
1440 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1444 if (!hw_addr_iface) {
1446 * A purely internal bridge, that is, one that has no non-virtual
1447 * network devices on it at all, is difficult because it has no
1448 * natural unique identifier at all.
1450 * When the host is a XenServer, we handle this case by hashing the
1451 * host's UUID with the name of the bridge. Names of bridges are
1452 * persistent across XenServer reboots, although they can be reused if
1453 * an internal network is destroyed and then a new one is later
1454 * created, so this is fairly effective.
1456 * When the host is not a XenServer, we punt by using a random MAC
1457 * address on each run.
1459 const char *host_uuid = xenserver_get_host_uuid();
1461 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1462 dpid = dpid_from_hash(combined, strlen(combined));
1468 return eth_addr_to_uint64(bridge_ea);
1472 dpid_from_hash(const void *data, size_t n)
1474 uint8_t hash[SHA1_DIGEST_SIZE];
1476 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1477 sha1_bytes(data, n, hash);
1478 eth_addr_mark_random(hash);
1479 return eth_addr_to_uint64(hash);
1483 iface_refresh_status(struct iface *iface)
1487 enum netdev_features current;
1488 enum netdev_flags flags;
1494 if (iface_is_synthetic(iface)) {
1500 if (!netdev_get_drv_info(iface->netdev, &sh)) {
1502 char **keys, **values;
1504 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1505 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1510 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1513 shash_destroy_free_data(&sh);
1515 error = netdev_get_flags(iface->netdev, &flags);
1517 ovsrec_interface_set_admin_state(iface->cfg,
1518 flags & NETDEV_UP ? "up" : "down");
1521 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1524 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1526 ovsrec_interface_set_duplex(iface->cfg,
1527 netdev_features_is_full_duplex(current)
1529 /* warning: uint64_t -> int64_t conversion */
1530 bps = netdev_features_to_bps(current);
1531 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1534 ovsrec_interface_set_duplex(iface->cfg, NULL);
1535 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1538 error = netdev_get_mtu(iface->netdev, &mtu);
1541 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1544 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1548 /* Writes 'iface''s CFM statistics to the database. */
1550 iface_refresh_cfm_stats(struct iface *iface)
1552 const struct ovsrec_interface *cfg = iface->cfg;
1554 const uint64_t *rmps;
1558 if (iface_is_synthetic(iface)) {
1562 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1565 const char *reasons[CFM_FAULT_N_REASONS];
1566 bool fault_bool = fault;
1570 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1571 int reason = 1 << i;
1572 if (fault & reason) {
1573 reasons[j++] = cfm_fault_reason_to_str(reason);
1577 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1578 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1580 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1581 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1584 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1585 iface->ofp_port, &rmps, &n_rmps);
1587 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1590 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1593 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1596 int64_t cfm_health = health;
1597 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1599 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1604 iface_refresh_stats(struct iface *iface)
1606 #define IFACE_STATS \
1607 IFACE_STAT(rx_packets, "rx_packets") \
1608 IFACE_STAT(tx_packets, "tx_packets") \
1609 IFACE_STAT(rx_bytes, "rx_bytes") \
1610 IFACE_STAT(tx_bytes, "tx_bytes") \
1611 IFACE_STAT(rx_dropped, "rx_dropped") \
1612 IFACE_STAT(tx_dropped, "tx_dropped") \
1613 IFACE_STAT(rx_errors, "rx_errors") \
1614 IFACE_STAT(tx_errors, "tx_errors") \
1615 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1616 IFACE_STAT(rx_over_errors, "rx_over_err") \
1617 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1618 IFACE_STAT(collisions, "collisions")
1620 #define IFACE_STAT(MEMBER, NAME) NAME,
1621 static char *keys[] = { IFACE_STATS };
1623 int64_t values[ARRAY_SIZE(keys)];
1626 struct netdev_stats stats;
1628 if (iface_is_synthetic(iface)) {
1632 /* Intentionally ignore return value, since errors will set 'stats' to
1633 * all-1s, and we will deal with that correctly below. */
1634 netdev_get_stats(iface->netdev, &stats);
1636 /* Copy statistics into values[] array. */
1638 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1641 assert(i == ARRAY_SIZE(keys));
1643 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1649 br_refresh_stp_status(struct bridge *br)
1651 struct ofproto *ofproto = br->ofproto;
1652 struct ofproto_stp_status status;
1653 char *keys[3], *values[3];
1656 if (ofproto_get_stp_status(ofproto, &status)) {
1660 if (!status.enabled) {
1661 ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0);
1665 keys[0] = "stp_bridge_id",
1666 values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id));
1667 keys[1] = "stp_designated_root",
1668 values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root));
1669 keys[2] = "stp_root_path_cost",
1670 values[2] = xasprintf("%d", status.root_path_cost);
1672 ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values));
1674 for (i = 0; i < ARRAY_SIZE(values); i++) {
1680 port_refresh_stp_status(struct port *port)
1682 struct ofproto *ofproto = port->bridge->ofproto;
1683 struct iface *iface;
1684 struct ofproto_port_stp_status status;
1686 char *str_values[4];
1687 int64_t int_values[3];
1690 if (port_is_synthetic(port)) {
1694 /* STP doesn't currently support bonds. */
1695 if (!list_is_singleton(&port->ifaces)) {
1696 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1700 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1702 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1706 if (!status.enabled) {
1707 ovsrec_port_set_status(port->cfg, NULL, NULL, 0);
1708 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1712 /* Set Status column. */
1713 keys[0] = "stp_port_id";
1714 str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id);
1715 keys[1] = "stp_state";
1716 str_values[1] = xstrdup(stp_state_name(status.state));
1717 keys[2] = "stp_sec_in_state";
1718 str_values[2] = xasprintf("%u", status.sec_in_state);
1719 keys[3] = "stp_role";
1720 str_values[3] = xstrdup(stp_role_name(status.role));
1722 ovsrec_port_set_status(port->cfg, keys, str_values,
1723 ARRAY_SIZE(str_values));
1725 for (i = 0; i < ARRAY_SIZE(str_values); i++) {
1726 free(str_values[i]);
1729 /* Set Statistics column. */
1730 keys[0] = "stp_tx_count";
1731 int_values[0] = status.tx_count;
1732 keys[1] = "stp_rx_count";
1733 int_values[1] = status.rx_count;
1734 keys[2] = "stp_error_count";
1735 int_values[2] = status.error_count;
1737 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1738 ARRAY_SIZE(int_values));
1742 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1746 /* Use other-config:enable-system-stats by preference. */
1747 enable = ovsrec_open_vswitch_get_other_config_value(cfg,
1748 "enable-statistics",
1751 return !strcmp(enable, "true");
1754 /* Disable by default. */
1759 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1761 struct ovsdb_datum datum;
1765 if (enable_system_stats(cfg)) {
1766 get_system_stats(&stats);
1769 ovsdb_datum_from_shash(&datum, &stats);
1770 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1774 static inline const char *
1775 nx_role_to_str(enum nx_role role)
1780 case NX_ROLE_MASTER:
1785 return "*** INVALID ROLE ***";
1790 refresh_controller_status(void)
1794 const struct ovsrec_controller *cfg;
1798 /* Accumulate status for controllers on all bridges. */
1799 HMAP_FOR_EACH (br, node, &all_bridges) {
1800 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1803 /* Update each controller in the database with current status. */
1804 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1805 struct ofproto_controller_info *cinfo =
1806 shash_find_data(&info, cfg->target);
1809 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1810 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1811 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1812 (char **) cinfo->pairs.values,
1815 ovsrec_controller_set_is_connected(cfg, false);
1816 ovsrec_controller_set_role(cfg, NULL);
1817 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1821 ofproto_free_ofproto_controller_info(&info);
1825 refresh_cfm_stats(void)
1827 static struct ovsdb_idl_txn *txn = NULL;
1832 txn = ovsdb_idl_txn_create(idl);
1834 HMAP_FOR_EACH (br, node, &all_bridges) {
1835 struct iface *iface;
1837 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1838 iface_refresh_cfm_stats(iface);
1843 if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
1844 ovsdb_idl_txn_destroy(txn);
1849 /* Performs periodic activity required by bridges that needs to be done with
1850 * the least possible latency.
1852 * It makes sense to call this function a couple of times per poll loop, to
1853 * provide a significant performance boost on some benchmarks with ofprotos
1854 * that use the ofproto-dpif implementation. */
1856 bridge_run_fast(void)
1860 HMAP_FOR_EACH (br, node, &all_bridges) {
1861 ofproto_run_fast(br->ofproto);
1868 const struct ovsrec_open_vswitch *cfg;
1870 bool vlan_splinters_changed;
1871 bool database_changed;
1874 /* (Re)configure if necessary. */
1875 database_changed = ovsdb_idl_run(idl);
1876 if (ovsdb_idl_is_lock_contended(idl)) {
1877 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1878 struct bridge *br, *next_br;
1880 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1881 "disabling this process until it goes away");
1883 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1887 } else if (!ovsdb_idl_has_lock(idl)) {
1890 cfg = ovsrec_open_vswitch_first(idl);
1892 /* Let each bridge do the work that it needs to do. */
1893 HMAP_FOR_EACH (br, node, &all_bridges) {
1894 ofproto_run(br->ofproto);
1897 /* Re-configure SSL. We do this on every trip through the main loop,
1898 * instead of just when the database changes, because the contents of the
1899 * key and certificate files can change without the database changing.
1901 * We do this before bridge_reconfigure() because that function might
1902 * initiate SSL connections and thus requires SSL to be configured. */
1903 if (cfg && cfg->ssl) {
1904 const struct ovsrec_ssl *ssl = cfg->ssl;
1906 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1907 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1910 /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage
1912 vlan_splinters_changed = false;
1913 if (vlan_splinters_enabled_anywhere) {
1914 HMAP_FOR_EACH (br, node, &all_bridges) {
1915 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
1916 vlan_splinters_changed = true;
1922 if (database_changed || vlan_splinters_changed) {
1924 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1926 bridge_reconfigure(cfg);
1928 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1929 ovsdb_idl_txn_commit(txn);
1930 ovsdb_idl_txn_destroy(txn); /* XXX */
1932 /* We still need to reconfigure to avoid dangling pointers to
1933 * now-destroyed ovsrec structures inside bridge data. */
1934 static const struct ovsrec_open_vswitch null_cfg;
1936 bridge_reconfigure(&null_cfg);
1940 /* Refresh system and interface stats if necessary. */
1941 if (time_msec() >= stats_timer) {
1943 struct ovsdb_idl_txn *txn;
1945 txn = ovsdb_idl_txn_create(idl);
1946 HMAP_FOR_EACH (br, node, &all_bridges) {
1950 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1951 struct iface *iface;
1953 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1954 iface_refresh_stats(iface);
1955 iface_refresh_status(iface);
1959 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
1960 mirror_refresh_stats(m);
1964 refresh_system_stats(cfg);
1965 refresh_controller_status();
1966 ovsdb_idl_txn_commit(txn);
1967 ovsdb_idl_txn_destroy(txn); /* XXX */
1970 stats_timer = time_msec() + STATS_INTERVAL;
1973 if (time_msec() >= db_limiter) {
1974 struct ovsdb_idl_txn *txn;
1976 txn = ovsdb_idl_txn_create(idl);
1977 HMAP_FOR_EACH (br, node, &all_bridges) {
1978 struct iface *iface;
1981 br_refresh_stp_status(br);
1983 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1984 port_refresh_stp_status(port);
1987 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
1988 const char *link_state;
1989 int64_t link_resets;
1992 if (iface_is_synthetic(iface)) {
1996 current = ofproto_port_is_lacp_current(br->ofproto,
2000 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2002 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2005 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2006 ovsrec_interface_set_link_state(iface->cfg, link_state);
2008 link_resets = netdev_get_carrier_resets(iface->netdev);
2009 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2013 if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) {
2014 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
2016 ovsdb_idl_txn_destroy(txn);
2019 refresh_cfm_stats();
2025 ovsdb_idl_wait(idl);
2026 if (!hmap_is_empty(&all_bridges)) {
2029 HMAP_FOR_EACH (br, node, &all_bridges) {
2030 ofproto_wait(br->ofproto);
2032 poll_timer_wait_until(stats_timer);
2034 if (db_limiter > time_msec()) {
2035 poll_timer_wait_until(db_limiter);
2040 /* QoS unixctl user interface functions. */
2042 struct qos_unixctl_show_cbdata {
2044 struct iface *iface;
2048 qos_unixctl_show_cb(unsigned int queue_id,
2049 const struct shash *details,
2052 struct qos_unixctl_show_cbdata *data = aux;
2053 struct ds *ds = data->ds;
2054 struct iface *iface = data->iface;
2055 struct netdev_queue_stats stats;
2056 struct shash_node *node;
2059 ds_put_cstr(ds, "\n");
2061 ds_put_format(ds, "Queue %u:\n", queue_id);
2063 ds_put_cstr(ds, "Default:\n");
2066 SHASH_FOR_EACH (node, details) {
2067 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
2070 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2072 if (stats.tx_packets != UINT64_MAX) {
2073 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2076 if (stats.tx_bytes != UINT64_MAX) {
2077 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2080 if (stats.tx_errors != UINT64_MAX) {
2081 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2084 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2085 queue_id, strerror(error));
2090 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2091 const char *argv[], void *aux OVS_UNUSED)
2093 struct ds ds = DS_EMPTY_INITIALIZER;
2094 struct shash sh = SHASH_INITIALIZER(&sh);
2095 struct iface *iface;
2097 struct shash_node *node;
2098 struct qos_unixctl_show_cbdata data;
2101 iface = iface_find(argv[1]);
2103 unixctl_command_reply_error(conn, "no such interface");
2107 netdev_get_qos(iface->netdev, &type, &sh);
2109 if (*type != '\0') {
2110 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2112 SHASH_FOR_EACH (node, &sh) {
2113 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
2118 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2121 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2123 unixctl_command_reply(conn, ds_cstr(&ds));
2125 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2126 unixctl_command_reply_error(conn, ds_cstr(&ds));
2129 shash_destroy_free_data(&sh);
2133 /* Bridge reconfiguration functions. */
2135 bridge_create(const struct ovsrec_bridge *br_cfg)
2139 assert(!bridge_lookup(br_cfg->name));
2140 br = xzalloc(sizeof *br);
2142 br->name = xstrdup(br_cfg->name);
2143 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2146 /* Derive the default Ethernet address from the bridge's UUID. This should
2147 * be unique and it will be stable between ovs-vswitchd runs. */
2148 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2149 eth_addr_mark_random(br->default_ea);
2151 hmap_init(&br->ports);
2152 hmap_init(&br->ifaces);
2153 hmap_init(&br->iface_by_name);
2154 hmap_init(&br->mirrors);
2156 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2160 bridge_destroy(struct bridge *br)
2163 struct mirror *mirror, *next_mirror;
2164 struct port *port, *next_port;
2166 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2169 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2170 mirror_destroy(mirror);
2172 hmap_remove(&all_bridges, &br->node);
2173 ofproto_destroy(br->ofproto);
2174 hmap_destroy(&br->ifaces);
2175 hmap_destroy(&br->ports);
2176 hmap_destroy(&br->iface_by_name);
2177 hmap_destroy(&br->mirrors);
2184 static struct bridge *
2185 bridge_lookup(const char *name)
2189 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2190 if (!strcmp(br->name, name)) {
2197 /* Handle requests for a listing of all flows known by the OpenFlow
2198 * stack, including those normally hidden. */
2200 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2201 const char *argv[], void *aux OVS_UNUSED)
2206 br = bridge_lookup(argv[1]);
2208 unixctl_command_reply_error(conn, "Unknown bridge");
2213 ofproto_get_all_flows(br->ofproto, &results);
2215 unixctl_command_reply(conn, ds_cstr(&results));
2216 ds_destroy(&results);
2219 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2220 * connections and reconnect. If BRIDGE is not specified, then all bridges
2221 * drop their controller connections and reconnect. */
2223 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2224 const char *argv[], void *aux OVS_UNUSED)
2228 br = bridge_lookup(argv[1]);
2230 unixctl_command_reply_error(conn, "Unknown bridge");
2233 ofproto_reconnect_controllers(br->ofproto);
2235 HMAP_FOR_EACH (br, node, &all_bridges) {
2236 ofproto_reconnect_controllers(br->ofproto);
2239 unixctl_command_reply(conn, NULL);
2243 bridge_get_controllers(const struct bridge *br,
2244 struct ovsrec_controller ***controllersp)
2246 struct ovsrec_controller **controllers;
2247 size_t n_controllers;
2249 controllers = br->cfg->controller;
2250 n_controllers = br->cfg->n_controller;
2252 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2258 *controllersp = controllers;
2260 return n_controllers;
2263 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
2264 * those configured in 'br->cfg'. */
2266 bridge_add_del_ports(struct bridge *br,
2267 const unsigned long int *splinter_vlans)
2269 struct port *port, *next;
2270 struct shash_node *node;
2271 struct shash new_ports;
2274 /* Collect new ports. */
2275 shash_init(&new_ports);
2276 for (i = 0; i < br->cfg->n_ports; i++) {
2277 const char *name = br->cfg->ports[i]->name;
2278 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2279 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2283 if (bridge_get_controllers(br, NULL)
2284 && !shash_find(&new_ports, br->name)) {
2285 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2286 br->name, br->name);
2288 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2289 br->synth_local_port.n_interfaces = 1;
2290 br->synth_local_port.name = br->name;
2292 br->synth_local_iface.name = br->name;
2293 br->synth_local_iface.type = "internal";
2295 br->synth_local_ifacep = &br->synth_local_iface;
2297 shash_add(&new_ports, br->name, &br->synth_local_port);
2300 if (splinter_vlans) {
2301 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2304 /* Get rid of deleted ports.
2305 * Get rid of deleted interfaces on ports that still exist. */
2306 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2307 port->cfg = shash_find_data(&new_ports, port->name);
2311 port_del_ifaces(port);
2315 /* Create new ports.
2316 * Add new interfaces to existing ports. */
2317 SHASH_FOR_EACH (node, &new_ports) {
2318 struct port *port = port_lookup(br, node->name);
2320 struct ovsrec_port *cfg = node->data;
2321 port = port_create(br, cfg);
2323 port_add_ifaces(port);
2324 if (list_is_empty(&port->ifaces)) {
2325 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
2326 br->name, port->name);
2330 shash_destroy(&new_ports);
2333 /* Initializes 'oc' appropriately as a management service controller for
2336 * The caller must free oc->target when it is no longer needed. */
2338 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2339 struct ofproto_controller *oc)
2341 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2342 oc->max_backoff = 0;
2343 oc->probe_interval = 60;
2344 oc->band = OFPROTO_OUT_OF_BAND;
2346 oc->burst_limit = 0;
2347 oc->enable_async_msgs = true;
2350 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2352 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2353 struct ofproto_controller *oc)
2355 const char *config_str;
2357 oc->target = c->target;
2358 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2359 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2360 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2361 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2362 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2363 oc->burst_limit = (c->controller_burst_limit
2364 ? *c->controller_burst_limit : 0);
2365 oc->enable_async_msgs = (!c->enable_async_messages
2366 || *c->enable_async_messages);
2367 config_str = ovsrec_controller_get_other_config_value(c, "dscp", NULL);
2369 oc->dscp = atoi(config_str);
2371 oc->dscp = DSCP_DEFAULT;
2375 /* Configures the IP stack for 'br''s local interface properly according to the
2376 * configuration in 'c'. */
2378 bridge_configure_local_iface_netdev(struct bridge *br,
2379 struct ovsrec_controller *c)
2381 struct netdev *netdev;
2382 struct in_addr mask, gateway;
2384 struct iface *local_iface;
2387 /* If there's no local interface or no IP address, give up. */
2388 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2389 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2393 /* Bring up the local interface. */
2394 netdev = local_iface->netdev;
2395 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2397 /* Configure the IP address and netmask. */
2398 if (!c->local_netmask
2399 || !inet_aton(c->local_netmask, &mask)
2401 mask.s_addr = guess_netmask(ip.s_addr);
2403 if (!netdev_set_in4(netdev, ip, mask)) {
2404 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2405 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2408 /* Configure the default gateway. */
2409 if (c->local_gateway
2410 && inet_aton(c->local_gateway, &gateway)
2411 && gateway.s_addr) {
2412 if (!netdev_add_router(netdev, gateway)) {
2413 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2414 br->name, IP_ARGS(&gateway.s_addr));
2419 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2420 * in either string are treated as equal to any number of slashes in the other,
2421 * e.g. "x///y" is equal to "x/y". */
2423 equal_pathnames(const char *a, const char *b)
2427 a += strspn(a, "/");
2428 b += strspn(b, "/");
2429 } else if (*a == '\0') {
2440 bridge_configure_remotes(struct bridge *br,
2441 const struct sockaddr_in *managers, size_t n_managers)
2443 const char *disable_ib_str, *queue_id_str;
2444 bool disable_in_band = false;
2447 struct ovsrec_controller **controllers;
2448 size_t n_controllers;
2450 enum ofproto_fail_mode fail_mode;
2452 struct ofproto_controller *ocs;
2456 /* Check if we should disable in-band control on this bridge. */
2457 disable_ib_str = ovsrec_bridge_get_other_config_value(br->cfg,
2460 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2461 disable_in_band = true;
2464 /* Set OpenFlow queue ID for in-band control. */
2465 queue_id_str = ovsrec_bridge_get_other_config_value(br->cfg,
2468 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2469 ofproto_set_in_band_queue(br->ofproto, queue_id);
2471 if (disable_in_band) {
2472 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2474 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2477 n_controllers = bridge_get_controllers(br, &controllers);
2479 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2482 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2483 for (i = 0; i < n_controllers; i++) {
2484 struct ovsrec_controller *c = controllers[i];
2486 if (!strncmp(c->target, "punix:", 6)
2487 || !strncmp(c->target, "unix:", 5)) {
2488 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2491 whitelist = xasprintf("unix:%s/%s.controller",
2492 ovs_rundir(), br->name);
2493 if (!equal_pathnames(c->target, whitelist)) {
2494 /* Prevent remote ovsdb-server users from accessing arbitrary
2495 * Unix domain sockets and overwriting arbitrary local
2497 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2498 "controller \"%s\" due to possibility for remote "
2499 "exploit. Instead, specify whitelisted \"%s\" or "
2500 "connect to \"unix:%s/%s.mgmt\" (which is always "
2501 "available without special configuration).",
2502 br->name, c->target, whitelist,
2503 ovs_rundir(), br->name);
2511 bridge_configure_local_iface_netdev(br, c);
2512 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2513 if (disable_in_band) {
2514 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2519 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2520 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2523 /* Set the fail-mode. */
2524 fail_mode = !br->cfg->fail_mode
2525 || !strcmp(br->cfg->fail_mode, "standalone")
2526 ? OFPROTO_FAIL_STANDALONE
2527 : OFPROTO_FAIL_SECURE;
2528 ofproto_set_fail_mode(br->ofproto, fail_mode);
2530 /* Configure OpenFlow controller connection snooping. */
2531 if (!ofproto_has_snoops(br->ofproto)) {
2535 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2536 ovs_rundir(), br->name));
2537 ofproto_set_snoops(br->ofproto, &snoops);
2538 sset_destroy(&snoops);
2543 bridge_configure_tables(struct bridge *br)
2545 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2549 n_tables = ofproto_get_n_tables(br->ofproto);
2551 for (i = 0; i < n_tables; i++) {
2552 struct ofproto_table_settings s;
2555 s.max_flows = UINT_MAX;
2559 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2560 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2563 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2564 s.max_flows = *cfg->flow_limit;
2566 if (cfg->overflow_policy
2567 && !strcmp(cfg->overflow_policy, "evict")) {
2570 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2571 for (k = 0; k < cfg->n_groups; k++) {
2572 const char *string = cfg->groups[k];
2575 msg = mf_parse_subfield__(&s.groups[k], &string);
2577 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
2578 "'groups' (%s)", br->name, i, msg);
2580 } else if (*string) {
2581 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
2582 "element '%s' contains trailing garbage",
2583 br->name, i, cfg->groups[k]);
2591 ofproto_configure_table(br->ofproto, i, &s);
2595 for (; j < br->cfg->n_flow_tables; j++) {
2596 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
2597 "%"PRId64" not supported by this datapath", br->name,
2598 br->cfg->key_flow_tables[j]);
2602 /* Port functions. */
2604 static struct port *
2605 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2609 port = xzalloc(sizeof *port);
2611 port->name = xstrdup(cfg->name);
2613 list_init(&port->ifaces);
2615 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2617 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2622 /* Deletes interfaces from 'port' that are no longer configured for it. */
2624 port_del_ifaces(struct port *port)
2626 struct iface *iface, *next;
2627 struct sset new_ifaces;
2630 /* Collect list of new interfaces. */
2631 sset_init(&new_ifaces);
2632 for (i = 0; i < port->cfg->n_interfaces; i++) {
2633 const char *name = port->cfg->interfaces[i]->name;
2634 const char *type = port->cfg->interfaces[i]->name;
2635 if (strcmp(type, "null")) {
2636 sset_add(&new_ifaces, name);
2640 /* Get rid of deleted interfaces. */
2641 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2642 if (!sset_contains(&new_ifaces, iface->name)) {
2643 iface_destroy(iface);
2647 sset_destroy(&new_ifaces);
2650 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2653 port_add_ifaces(struct port *port)
2655 struct shash new_ifaces;
2656 struct shash_node *node;
2659 /* Collect new ifaces. */
2660 shash_init(&new_ifaces);
2661 for (i = 0; i < port->cfg->n_interfaces; i++) {
2662 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2663 if (strcmp(cfg->type, "null")
2664 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2665 VLOG_WARN("port %s: %s specified twice as port interface",
2666 port->name, cfg->name);
2667 iface_clear_db_record(cfg);
2671 /* Create new interfaces.
2672 * Update interface types and 'cfg' members. */
2673 SHASH_FOR_EACH (node, &new_ifaces) {
2674 const struct ovsrec_interface *cfg = node->data;
2675 const char *iface_name = node->name;
2676 struct iface *iface;
2678 iface = iface_lookup(port->bridge, iface_name);
2680 iface = iface_create(port, cfg);
2685 /* Determine interface type. The local port always has type
2686 * "internal". Other ports take their type from the database and
2687 * default to "system" if none is specified. */
2688 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2689 : cfg->type[0] ? cfg->type
2692 shash_destroy(&new_ifaces);
2696 port_destroy(struct port *port)
2699 struct bridge *br = port->bridge;
2700 struct iface *iface, *next;
2703 ofproto_bundle_unregister(br->ofproto, port);
2706 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2707 iface_destroy(iface);
2710 hmap_remove(&br->ports, &port->hmap_node);
2712 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2719 static struct port *
2720 port_lookup(const struct bridge *br, const char *name)
2724 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2726 if (!strcmp(port->name, name)) {
2734 enable_lacp(struct port *port, bool *activep)
2736 if (!port->cfg->lacp) {
2737 /* XXX when LACP implementation has been sufficiently tested, enable by
2738 * default and make active on bonded ports. */
2740 } else if (!strcmp(port->cfg->lacp, "off")) {
2742 } else if (!strcmp(port->cfg->lacp, "active")) {
2745 } else if (!strcmp(port->cfg->lacp, "passive")) {
2749 VLOG_WARN("port %s: unknown LACP mode %s",
2750 port->name, port->cfg->lacp);
2755 static struct lacp_settings *
2756 port_configure_lacp(struct port *port, struct lacp_settings *s)
2758 const char *lacp_time, *system_id;
2759 long long int custom_time;
2762 if (!enable_lacp(port, &s->active)) {
2766 s->name = port->name;
2768 system_id = ovsrec_port_get_other_config_value(port->cfg, "lacp-system-id",
2771 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
2772 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
2773 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
2774 " address.", port->name, system_id);
2778 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2781 if (eth_addr_is_zero(s->id)) {
2782 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
2786 /* Prefer bondable links if unspecified. */
2787 priority = atoi(ovsrec_port_get_other_config_value(port->cfg,
2788 "lacp-system-priority",
2790 s->priority = (priority > 0 && priority <= UINT16_MAX
2792 : UINT16_MAX - !list_is_short(&port->ifaces));
2794 s->heartbeat = !strcmp(ovsrec_port_get_other_config_value(port->cfg,
2799 lacp_time = ovsrec_port_get_other_config_value(port->cfg, "lacp-time",
2801 custom_time = atoi(lacp_time);
2802 if (!strcmp(lacp_time, "fast")) {
2803 s->lacp_time = LACP_TIME_FAST;
2804 } else if (!strcmp(lacp_time, "slow")) {
2805 s->lacp_time = LACP_TIME_SLOW;
2806 } else if (custom_time > 0) {
2807 s->lacp_time = LACP_TIME_CUSTOM;
2808 s->custom_time = custom_time;
2810 s->lacp_time = LACP_TIME_SLOW;
2817 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2819 int priority, portid, key;
2821 portid = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2825 atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2826 "lacp-port-priority",
2828 key = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
2829 "lacp-aggregation-key",
2832 if (portid <= 0 || portid > UINT16_MAX) {
2833 portid = iface->ofp_port;
2836 if (priority <= 0 || priority > UINT16_MAX) {
2837 priority = UINT16_MAX;
2840 if (key < 0 || key > UINT16_MAX) {
2844 s->name = iface->name;
2846 s->priority = priority;
2851 port_configure_bond(struct port *port, struct bond_settings *s,
2852 uint32_t *bond_stable_ids)
2854 const char *detect_s;
2855 struct iface *iface;
2856 int miimon_interval;
2859 s->name = port->name;
2861 if (port->cfg->bond_mode) {
2862 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2863 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2864 port->name, port->cfg->bond_mode,
2865 bond_mode_to_string(s->balance));
2868 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2870 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
2871 * active-backup. At some point we should remove this warning. */
2872 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
2873 " in previous versions, the default bond_mode was"
2874 " balance-slb", port->name,
2875 bond_mode_to_string(s->balance));
2877 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2878 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2879 "please use another bond type or disable flood_vlans",
2884 atoi(ovsrec_port_get_other_config_value(port->cfg,
2885 "bond-miimon-interval", "0"));
2886 if (miimon_interval <= 0) {
2887 miimon_interval = 200;
2890 detect_s = ovsrec_port_get_other_config_value(port->cfg,
2893 if (!strcmp(detect_s, "carrier")) {
2894 miimon_interval = 0;
2895 } else if (strcmp(detect_s, "miimon")) {
2896 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2897 "defaulting to carrier", port->name, detect_s);
2898 miimon_interval = 0;
2901 s->up_delay = MAX(0, port->cfg->bond_updelay);
2902 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2903 s->basis = atoi(ovsrec_port_get_other_config_value(port->cfg,
2906 s->rebalance_interval = atoi(
2907 ovsrec_port_get_other_config_value(port->cfg,
2908 "bond-rebalance-interval",
2910 if (s->rebalance_interval && s->rebalance_interval < 1000) {
2911 s->rebalance_interval = 1000;
2914 s->fake_iface = port->cfg->bond_fake_iface;
2917 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2918 long long stable_id;
2921 atoll(ovsrec_interface_get_other_config_value(iface->cfg,
2924 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2925 stable_id = iface->ofp_port;
2927 bond_stable_ids[i++] = stable_id;
2929 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2933 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
2934 * instead of obtaining it from the database. */
2936 port_is_synthetic(const struct port *port)
2938 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
2941 /* Interface functions. */
2943 static struct iface *
2944 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2946 struct bridge *br = port->bridge;
2947 struct iface *iface;
2948 char *name = if_cfg->name;
2950 iface = xzalloc(sizeof *iface);
2952 iface->name = xstrdup(name);
2953 iface->ofp_port = -1;
2954 iface->tag = tag_create_random();
2955 iface->netdev = NULL;
2956 iface->cfg = if_cfg;
2957 iface->need_refresh = true;
2959 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2961 list_push_back(&port->ifaces, &iface->port_elem);
2963 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2969 iface_destroy(struct iface *iface)
2972 struct port *port = iface->port;
2973 struct bridge *br = port->bridge;
2975 if (br->ofproto && iface->ofp_port >= 0) {
2976 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2979 if (iface->ofp_port >= 0) {
2980 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2983 list_remove(&iface->port_elem);
2984 hmap_remove(&br->iface_by_name, &iface->name_node);
2986 netdev_close(iface->netdev);
2993 static struct iface *
2994 iface_lookup(const struct bridge *br, const char *name)
2996 struct iface *iface;
2998 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2999 &br->iface_by_name) {
3000 if (!strcmp(iface->name, name)) {
3008 static struct iface *
3009 iface_find(const char *name)
3011 const struct bridge *br;
3013 HMAP_FOR_EACH (br, node, &all_bridges) {
3014 struct iface *iface = iface_lookup(br, name);
3023 static struct iface *
3024 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3026 struct iface *iface;
3028 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3029 hash_int(ofp_port, 0), &br->ifaces) {
3030 if (iface->ofp_port == ofp_port) {
3037 /* Set Ethernet address of 'iface', if one is specified in the configuration
3040 iface_set_mac(struct iface *iface)
3042 uint8_t ea[ETH_ADDR_LEN];
3044 if (!strcmp(iface->type, "internal")
3045 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3046 if (iface->ofp_port == OFPP_LOCAL) {
3047 VLOG_ERR("interface %s: ignoring mac in Interface record "
3048 "(use Bridge record to set local port's mac)",
3050 } else if (eth_addr_is_multicast(ea)) {
3051 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3054 int error = netdev_set_etheraddr(iface->netdev, ea);
3056 VLOG_ERR("interface %s: setting MAC failed (%s)",
3057 iface->name, strerror(error));
3063 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3065 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3067 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3068 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3072 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3073 * sets the "ofport" field to -1.
3075 * This is appropriate when 'if_cfg''s interface cannot be created or is
3076 * otherwise invalid. */
3078 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3080 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3081 iface_set_ofport(if_cfg, -1);
3082 ovsrec_interface_set_status(if_cfg, NULL, NULL, 0);
3083 ovsrec_interface_set_admin_state(if_cfg, NULL);
3084 ovsrec_interface_set_duplex(if_cfg, NULL);
3085 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3086 ovsrec_interface_set_link_state(if_cfg, NULL);
3087 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3088 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3089 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3090 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3091 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3092 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3096 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3098 * The value strings in '*shash' are taken directly from values[], not copied,
3099 * so the caller should not modify or free them. */
3101 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3102 struct shash *shash)
3107 for (i = 0; i < n; i++) {
3108 shash_add(shash, keys[i], values[i]);
3112 /* Creates 'keys' and 'values' arrays from 'shash'.
3114 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3115 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3116 * are populated with with strings taken directly from 'shash' and thus have
3117 * the same ownership of the key-value pairs in shash.
3120 shash_to_ovs_idl_map(struct shash *shash,
3121 char ***keys, char ***values, size_t *n)
3125 struct shash_node *sn;
3127 count = shash_count(shash);
3129 k = xmalloc(count * sizeof *k);
3130 v = xmalloc(count * sizeof *v);
3133 SHASH_FOR_EACH(sn, shash) {
3144 struct iface_delete_queues_cbdata {
3145 struct netdev *netdev;
3146 const struct ovsdb_datum *queues;
3150 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3152 union ovsdb_atom atom;
3154 atom.integer = target;
3155 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3159 iface_delete_queues(unsigned int queue_id,
3160 const struct shash *details OVS_UNUSED, void *cbdata_)
3162 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3164 if (!queue_ids_include(cbdata->queues, queue_id)) {
3165 netdev_delete_queue(cbdata->netdev, queue_id);
3170 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3172 struct ofpbuf queues_buf;
3174 ofpbuf_init(&queues_buf, 0);
3176 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3177 netdev_set_qos(iface->netdev, NULL, NULL);
3179 struct iface_delete_queues_cbdata cbdata;
3180 struct shash details;
3184 /* Configure top-level Qos for 'iface'. */
3185 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3186 qos->n_other_config, &details);
3187 netdev_set_qos(iface->netdev, qos->type, &details);
3188 shash_destroy(&details);
3190 /* Deconfigure queues that were deleted. */
3191 cbdata.netdev = iface->netdev;
3192 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3194 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3196 /* Configure queues for 'iface'. */
3198 for (i = 0; i < qos->n_queues; i++) {
3199 const struct ovsrec_queue *queue = qos->value_queues[i];
3200 unsigned int queue_id = qos->key_queues[i];
3202 if (queue_id == 0) {
3206 if (queue->n_dscp == 1) {
3207 struct ofproto_port_queue *port_queue;
3209 port_queue = ofpbuf_put_uninit(&queues_buf,
3210 sizeof *port_queue);
3211 port_queue->queue = queue_id;
3212 port_queue->dscp = queue->dscp[0];
3215 shash_from_ovs_idl_map(queue->key_other_config,
3216 queue->value_other_config,
3217 queue->n_other_config, &details);
3218 netdev_set_queue(iface->netdev, queue_id, &details);
3219 shash_destroy(&details);
3222 shash_init(&details);
3223 netdev_set_queue(iface->netdev, 0, &details);
3224 shash_destroy(&details);
3228 if (iface->ofp_port >= 0) {
3229 const struct ofproto_port_queue *port_queues = queues_buf.data;
3230 size_t n_queues = queues_buf.size / sizeof *port_queues;
3232 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3233 port_queues, n_queues);
3236 netdev_set_policing(iface->netdev,
3237 iface->cfg->ingress_policing_rate,
3238 iface->cfg->ingress_policing_burst);
3240 ofpbuf_uninit(&queues_buf);
3244 iface_configure_cfm(struct iface *iface)
3246 const struct ovsrec_interface *cfg = iface->cfg;
3247 const char *extended_str, *opstate_str;
3248 const char *cfm_ccm_vlan;
3249 struct cfm_settings s;
3251 if (!cfg->n_cfm_mpid) {
3252 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3256 s.mpid = *cfg->cfm_mpid;
3257 s.interval = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3260 cfm_ccm_vlan = ovsrec_interface_get_other_config_value(iface->cfg,
3263 s.ccm_pcp = atoi(ovsrec_interface_get_other_config_value(iface->cfg,
3266 if (s.interval <= 0) {
3270 if (!strcasecmp("random", cfm_ccm_vlan)) {
3271 s.ccm_vlan = CFM_RANDOM_VLAN;
3273 s.ccm_vlan = atoi(cfm_ccm_vlan);
3274 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3279 extended_str = ovsrec_interface_get_other_config_value(iface->cfg,
3282 s.extended = !strcasecmp("true", extended_str);
3284 opstate_str = ovsrec_interface_get_other_config_value(iface->cfg,
3287 s.opup = !strcasecmp("up", opstate_str);
3289 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3292 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3293 * instead of obtaining it from the database. */
3295 iface_is_synthetic(const struct iface *iface)
3297 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3301 /* Port mirroring. */
3303 static struct mirror *
3304 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3308 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3309 if (uuid_equals(uuid, &m->uuid)) {
3317 bridge_configure_mirrors(struct bridge *br)
3319 const struct ovsdb_datum *mc;
3320 unsigned long *flood_vlans;
3321 struct mirror *m, *next;
3324 /* Get rid of deleted mirrors. */
3325 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3326 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3327 union ovsdb_atom atom;
3329 atom.uuid = m->uuid;
3330 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3335 /* Add new mirrors and reconfigure existing ones. */
3336 for (i = 0; i < br->cfg->n_mirrors; i++) {
3337 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3338 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3340 m = mirror_create(br, cfg);
3343 if (!mirror_configure(m)) {
3348 /* Update flooded vlans (for RSPAN). */
3349 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3350 br->cfg->n_flood_vlans);
3351 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3352 bitmap_free(flood_vlans);
3355 static struct mirror *
3356 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3360 m = xzalloc(sizeof *m);
3361 m->uuid = cfg->header_.uuid;
3362 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3364 m->name = xstrdup(cfg->name);
3370 mirror_destroy(struct mirror *m)
3373 struct bridge *br = m->bridge;
3376 ofproto_mirror_unregister(br->ofproto, m);
3379 hmap_remove(&br->mirrors, &m->hmap_node);
3386 mirror_collect_ports(struct mirror *m,
3387 struct ovsrec_port **in_ports, int n_in_ports,
3388 void ***out_portsp, size_t *n_out_portsp)
3390 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3391 size_t n_out_ports = 0;
3394 for (i = 0; i < n_in_ports; i++) {
3395 const char *name = in_ports[i]->name;
3396 struct port *port = port_lookup(m->bridge, name);
3398 out_ports[n_out_ports++] = port;
3400 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3401 "port %s", m->bridge->name, m->name, name);
3404 *out_portsp = out_ports;
3405 *n_out_portsp = n_out_ports;
3409 mirror_configure(struct mirror *m)
3411 const struct ovsrec_mirror *cfg = m->cfg;
3412 struct ofproto_mirror_settings s;
3415 if (strcmp(cfg->name, m->name)) {
3417 m->name = xstrdup(cfg->name);
3421 /* Get output port or VLAN. */
3422 if (cfg->output_port) {
3423 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3424 if (!s.out_bundle) {
3425 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3426 m->bridge->name, m->name);
3429 s.out_vlan = UINT16_MAX;
3431 if (cfg->output_vlan) {
3432 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3433 "output vlan; ignoring output vlan",
3434 m->bridge->name, m->name);
3436 } else if (cfg->output_vlan) {
3437 /* The database should prevent invalid VLAN values. */
3438 s.out_bundle = NULL;
3439 s.out_vlan = *cfg->output_vlan;
3441 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3442 m->bridge->name, m->name);
3446 /* Get port selection. */
3447 if (cfg->select_all) {
3448 size_t n_ports = hmap_count(&m->bridge->ports);
3449 void **ports = xmalloc(n_ports * sizeof *ports);
3454 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3464 /* Get ports, dropping ports that don't exist.
3465 * The IDL ensures that there are no duplicates. */
3466 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3467 &s.srcs, &s.n_srcs);
3468 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3469 &s.dsts, &s.n_dsts);
3472 /* Get VLAN selection. */
3473 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3476 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3479 if (s.srcs != s.dsts) {
3488 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3490 * This is deprecated. It is only for compatibility with broken device drivers
3491 * in old versions of Linux that do not properly support VLANs when VLAN
3492 * devices are not used. When broken device drivers are no longer in
3493 * widespread use, we will delete these interfaces. */
3495 static void **blocks;
3496 static size_t n_blocks, allocated_blocks;
3498 /* Adds 'block' to a list of blocks that have to be freed with free() when the
3499 * VLAN splinters are reconfigured. */
3501 register_block(void *block)
3503 if (n_blocks >= allocated_blocks) {
3504 blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks);
3506 blocks[n_blocks++] = block;
3509 /* Frees all of the blocks registered with register_block(). */
3511 free_registered_blocks(void)
3515 for (i = 0; i < n_blocks; i++) {
3521 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3524 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3528 value = ovsrec_interface_get_other_config_value(iface_cfg,
3529 "enable-vlan-splinters",
3531 return !strcmp(value, "true");
3534 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3537 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3538 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3539 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3542 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3543 * use, returns NULL.
3545 * Updates 'vlan_splinters_enabled_anywhere'. */
3546 static unsigned long int *
3547 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3549 unsigned long int *splinter_vlans;
3550 struct sset splinter_ifaces;
3551 const char *real_dev_name;
3552 struct shash *real_devs;
3553 struct shash_node *node;
3557 /* Free space allocated for synthesized ports and interfaces, since we're
3558 * in the process of reconstructing all of them. */
3559 free_registered_blocks();
3561 splinter_vlans = bitmap_allocate(4096);
3562 sset_init(&splinter_ifaces);
3563 vlan_splinters_enabled_anywhere = false;
3564 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3565 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3568 for (j = 0; j < br_cfg->n_ports; j++) {
3569 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3572 for (k = 0; k < port_cfg->n_interfaces; k++) {
3573 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3575 if (vlan_splinters_is_enabled(iface_cfg)) {
3576 vlan_splinters_enabled_anywhere = true;
3577 sset_add(&splinter_ifaces, iface_cfg->name);
3578 vlan_bitmap_from_array__(port_cfg->trunks,
3584 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3585 bitmap_set1(splinter_vlans, *port_cfg->tag);
3590 if (!vlan_splinters_enabled_anywhere) {
3591 free(splinter_vlans);
3592 sset_destroy(&splinter_ifaces);
3596 HMAP_FOR_EACH (br, node, &all_bridges) {
3598 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3602 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3603 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3604 * device to be created for it. */
3605 bitmap_set0(splinter_vlans, 0);
3606 bitmap_set0(splinter_vlans, 4095);
3608 /* Delete all VLAN devices that we don't need. */
3610 real_devs = vlandev_get_real_devs();
3611 SHASH_FOR_EACH (node, real_devs) {
3612 const struct vlan_real_dev *real_dev = node->data;
3613 const struct vlan_dev *vlan_dev;
3614 bool real_dev_has_splinters;
3616 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3618 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3619 if (!real_dev_has_splinters
3620 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3621 struct netdev *netdev;
3623 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3624 if (!netdev_get_in4(netdev, NULL, NULL) ||
3625 !netdev_get_in6(netdev, NULL)) {
3626 vlandev_del(vlan_dev->name);
3628 /* It has an IP address configured, so we don't own
3629 * it. Don't delete it. */
3631 netdev_close(netdev);
3638 /* Add all VLAN devices that we need. */
3639 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3642 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3643 if (!vlandev_get_name(real_dev_name, vid)) {
3644 vlandev_add(real_dev_name, vid);
3651 sset_destroy(&splinter_ifaces);
3653 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3654 free(splinter_vlans);
3657 return splinter_vlans;
3660 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3663 configure_splinter_port(struct port *port)
3665 struct ofproto *ofproto = port->bridge->ofproto;
3666 uint16_t realdev_ofp_port;
3667 const char *realdev_name;
3668 struct iface *vlandev, *realdev;
3670 ofproto_bundle_unregister(port->bridge->ofproto, port);
3672 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3675 realdev_name = ovsrec_port_get_other_config_value(port->cfg,
3677 realdev = iface_lookup(port->bridge, realdev_name);
3678 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3680 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
3684 static struct ovsrec_port *
3685 synthesize_splinter_port(const char *real_dev_name,
3686 const char *vlan_dev_name, int vid)
3688 struct ovsrec_interface *iface;
3689 struct ovsrec_port *port;
3691 iface = xzalloc(sizeof *iface);
3692 iface->name = xstrdup(vlan_dev_name);
3693 iface->type = "system";
3695 port = xzalloc(sizeof *port);
3696 port->interfaces = xmemdup(&iface, sizeof iface);
3697 port->n_interfaces = 1;
3698 port->name = xstrdup(vlan_dev_name);
3699 port->vlan_mode = "splinter";
3700 port->tag = xmalloc(sizeof *port->tag);
3702 port->key_other_config = xmalloc(sizeof *port->key_other_config);
3703 port->key_other_config[0] = "realdev";
3704 port->value_other_config = xmalloc(sizeof *port->value_other_config);
3705 port->value_other_config[0] = xstrdup(real_dev_name);
3706 port->n_other_config = 1;
3708 register_block(iface);
3709 register_block(iface->name);
3710 register_block(port);
3711 register_block(port->interfaces);
3712 register_block(port->name);
3713 register_block(port->tag);
3714 register_block(port->key_other_config);
3715 register_block(port->value_other_config);
3716 register_block(port->value_other_config[0]);
3721 /* For each interface with 'br' that has VLAN splinters enabled, adds a
3722 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
3723 * 1-bit in the 'splinter_vlans' bitmap. */
3725 add_vlan_splinter_ports(struct bridge *br,
3726 const unsigned long int *splinter_vlans,
3727 struct shash *ports)
3731 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
3732 * we're modifying 'ports'. */
3733 for (i = 0; i < br->cfg->n_ports; i++) {
3734 const char *name = br->cfg->ports[i]->name;
3735 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
3738 for (j = 0; j < port_cfg->n_interfaces; j++) {
3739 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
3741 if (vlan_splinters_is_enabled(iface_cfg)) {
3742 const char *real_dev_name;
3745 real_dev_name = iface_cfg->name;
3746 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3747 const char *vlan_dev_name;
3749 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
3751 && !shash_find(ports, vlan_dev_name)) {
3752 shash_add(ports, vlan_dev_name,
3753 synthesize_splinter_port(
3754 real_dev_name, vlan_dev_name, vid));
3763 mirror_refresh_stats(struct mirror *m)
3765 struct ofproto *ofproto = m->bridge->ofproto;
3766 uint64_t tx_packets, tx_bytes;
3769 size_t stat_cnt = 0;
3771 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
3772 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
3776 if (tx_packets != UINT64_MAX) {
3777 keys[stat_cnt] = "tx_packets";
3778 values[stat_cnt] = tx_packets;
3781 if (tx_bytes != UINT64_MAX) {
3782 keys[stat_cnt] = "tx_bytes";
3783 values[stat_cnt] = tx_bytes;
3787 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);