1 /* Copyright (c) 2008, 2009, 2010, 2011 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.
18 #include "byte-order.h"
21 #include <arpa/inet.h>
24 #include <sys/socket.h>
26 #include <openflow/openflow.h>
31 #include <sys/socket.h>
32 #include <sys/types.h>
37 #include "classifier.h"
42 #include "dynamic-string.h"
49 #include "mac-learning.h"
53 #include "ofp-print.h"
55 #include "ofproto/netflow.h"
56 #include "ofproto/ofproto.h"
57 #include "ovsdb-data.h"
59 #include "poll-loop.h"
63 #include "socket-util.h"
64 #include "stream-ssl.h"
67 #include "system-stats.h"
72 #include "vswitchd/vswitch-idl.h"
73 #include "xenserver.h"
75 #include "sflow_api.h"
77 VLOG_DEFINE_THIS_MODULE(bridge);
79 COVERAGE_DEFINE(bridge_flush);
80 COVERAGE_DEFINE(bridge_process_flow);
81 COVERAGE_DEFINE(bridge_reconfigure);
89 struct dst builtin[32];
94 static void dst_set_init(struct dst_set *);
95 static void dst_set_add(struct dst_set *, const struct dst *);
96 static void dst_set_free(struct dst_set *);
99 /* These members are always valid. */
100 struct list port_elem; /* Element in struct port's "ifaces" list. */
101 struct port *port; /* Containing port. */
102 char *name; /* Host network device name. */
103 tag_type tag; /* Tag associated with this interface. */
105 /* These members are valid only after bridge_reconfigure() causes them to
107 struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
108 int dp_ifidx; /* Index within kernel datapath. */
109 struct netdev *netdev; /* Network device. */
110 const char *type; /* Usually same as cfg->type. */
111 const struct ovsrec_interface *cfg;
114 #define MAX_MIRRORS 32
115 typedef uint32_t mirror_mask_t;
116 #define MIRROR_MASK_C(X) UINT32_C(X)
117 BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS);
119 struct bridge *bridge;
122 struct uuid uuid; /* UUID of this "mirror" record in database. */
124 /* Selection criteria. */
125 struct sset src_ports; /* Source port names. */
126 struct sset dst_ports; /* Destination port names. */
131 struct port *out_port;
135 #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
137 struct bridge *bridge;
138 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
141 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
142 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
143 * NULL if all VLANs are trunked. */
144 const struct ovsrec_port *cfg;
146 /* An ordinary bridge port has 1 interface.
147 * A bridge port for bonding has at least 2 interfaces. */
148 struct list ifaces; /* List of "struct iface"s. */
150 struct lacp *lacp; /* NULL if LACP is not enabled. */
155 /* Port mirroring info. */
156 mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
157 mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
158 bool is_mirror_output_port; /* Does port mirroring send frames here? */
162 struct list node; /* Node in global list of bridges. */
163 char *name; /* User-specified arbitrary name. */
164 struct mac_learning *ml; /* MAC learning table. */
165 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
166 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
167 const struct ovsrec_bridge *cfg;
169 /* OpenFlow switch processing. */
170 struct ofproto *ofproto; /* OpenFlow switch. */
172 /* Kernel datapath information. */
173 struct dpif *dpif; /* Datapath. */
174 struct hmap ifaces; /* "struct iface"s indexed by dp_ifidx. */
177 struct hmap ports; /* "struct port"s indexed by name. */
178 struct shash iface_by_name; /* "struct iface"s indexed by name. */
181 bool has_bonded_ports;
186 /* Port mirroring. */
187 struct mirror *mirrors[MAX_MIRRORS];
189 /* Synthetic local port if necessary. */
190 struct ovsrec_port synth_local_port;
191 struct ovsrec_interface synth_local_iface;
192 struct ovsrec_interface *synth_local_ifacep;
195 /* List of all bridges. */
196 static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
198 /* OVSDB IDL used to obtain configuration. */
199 static struct ovsdb_idl *idl;
201 /* Each time this timer expires, the bridge fetches systems and interface
202 * statistics and pushes them into the database. */
203 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
204 static long long int stats_timer = LLONG_MIN;
206 /* Stores the time after which rate limited statistics may be written to the
207 * database. Only updated when changes to the database require rate limiting.
209 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
210 static long long int db_limiter = LLONG_MIN;
212 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
213 static void bridge_destroy(struct bridge *);
214 static struct bridge *bridge_lookup(const char *name);
215 static unixctl_cb_func bridge_unixctl_dump_flows;
216 static unixctl_cb_func bridge_unixctl_reconnect;
217 static int bridge_run_one(struct bridge *);
218 static size_t bridge_get_controllers(const struct bridge *br,
219 struct ovsrec_controller ***controllersp);
220 static void bridge_reconfigure_one(struct bridge *);
221 static void bridge_reconfigure_remotes(struct bridge *,
222 const struct sockaddr_in *managers,
224 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
225 static void bridge_fetch_dp_ifaces(struct bridge *);
226 static void bridge_flush(struct bridge *);
227 static void bridge_pick_local_hw_addr(struct bridge *,
228 uint8_t ea[ETH_ADDR_LEN],
229 struct iface **hw_addr_iface);
230 static uint64_t bridge_pick_datapath_id(struct bridge *,
231 const uint8_t bridge_ea[ETH_ADDR_LEN],
232 struct iface *hw_addr_iface);
233 static uint64_t dpid_from_hash(const void *, size_t nbytes);
235 static unixctl_cb_func bridge_unixctl_fdb_show;
236 static unixctl_cb_func cfm_unixctl_show;
237 static unixctl_cb_func qos_unixctl_show;
239 static void port_run(struct port *);
240 static void port_wait(struct port *);
241 static struct port *port_create(struct bridge *, const char *name);
242 static void port_reconfigure(struct port *, const struct ovsrec_port *);
243 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
244 static void port_destroy(struct port *);
245 static struct port *port_lookup(const struct bridge *, const char *name);
246 static struct iface *port_get_an_iface(const struct port *);
247 static struct port *port_from_dp_ifidx(const struct bridge *,
249 static void port_reconfigure_lacp(struct port *);
250 static void port_reconfigure_bond(struct port *);
251 static void port_send_learning_packets(struct port *);
253 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
254 static void mirror_destroy(struct mirror *);
255 static void mirror_reconfigure(struct bridge *);
256 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
257 static bool vlan_is_mirrored(const struct mirror *, int vlan);
259 static struct iface *iface_create(struct port *port,
260 const struct ovsrec_interface *if_cfg);
261 static void iface_destroy(struct iface *);
262 static struct iface *iface_lookup(const struct bridge *, const char *name);
263 static struct iface *iface_find(const char *name);
264 static struct iface *iface_from_dp_ifidx(const struct bridge *,
266 static void iface_set_mac(struct iface *);
267 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
268 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
269 static void iface_update_cfm(struct iface *);
270 static bool iface_refresh_cfm_stats(struct iface *iface);
271 static bool iface_get_carrier(const struct iface *);
272 static bool iface_is_synthetic(const struct iface *);
274 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
276 static void shash_to_ovs_idl_map(struct shash *,
277 char ***keys, char ***values, size_t *n);
279 /* Hooks into ofproto processing. */
280 static struct ofhooks bridge_ofhooks;
282 /* Public functions. */
284 /* Initializes the bridge module, configuring it to obtain its configuration
285 * from an OVSDB server accessed over 'remote', which should be a string in a
286 * form acceptable to ovsdb_idl_create(). */
288 bridge_init(const char *remote)
290 /* Create connection to database. */
291 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
293 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
294 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
295 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
297 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
298 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
299 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
301 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
302 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
305 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
315 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
317 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
319 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
320 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
322 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
324 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
326 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
332 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
334 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
336 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
338 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
339 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
340 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
342 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
344 /* Register unixctl commands. */
345 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
346 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
347 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
348 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
350 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
359 struct bridge *br, *next_br;
361 LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
364 ovsdb_idl_destroy(idl);
367 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
368 * but for which the ovs-vswitchd configuration 'cfg' is required. */
370 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
372 static bool already_configured_once;
373 struct sset bridge_names;
374 struct sset dpif_names, dpif_types;
378 /* Only do this once per ovs-vswitchd run. */
379 if (already_configured_once) {
382 already_configured_once = true;
384 stats_timer = time_msec() + STATS_INTERVAL;
386 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
387 sset_init(&bridge_names);
388 for (i = 0; i < cfg->n_bridges; i++) {
389 sset_add(&bridge_names, cfg->bridges[i]->name);
392 /* Iterate over all system dpifs and delete any of them that do not appear
394 sset_init(&dpif_names);
395 sset_init(&dpif_types);
396 dp_enumerate_types(&dpif_types);
397 SSET_FOR_EACH (type, &dpif_types) {
400 dp_enumerate_names(type, &dpif_names);
402 /* Delete each dpif whose name is not in 'bridge_names'. */
403 SSET_FOR_EACH (name, &dpif_names) {
404 if (!sset_contains(&bridge_names, name)) {
408 retval = dpif_open(name, type, &dpif);
416 sset_destroy(&bridge_names);
417 sset_destroy(&dpif_names);
418 sset_destroy(&dpif_types);
421 /* Callback for iterate_and_prune_ifaces(). */
423 check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
425 if (!iface->netdev) {
426 /* We already reported a related error, don't bother duplicating it. */
430 if (iface->dp_ifidx < 0) {
431 VLOG_ERR("%s interface not in %s, dropping",
432 iface->name, dpif_name(br->dpif));
436 VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
437 iface->name, iface->dp_ifidx);
441 /* Callback for iterate_and_prune_ifaces(). */
443 set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
444 void *aux OVS_UNUSED)
446 /* Set policing attributes. */
447 netdev_set_policing(iface->netdev,
448 iface->cfg->ingress_policing_rate,
449 iface->cfg->ingress_policing_burst);
451 /* Set MAC address of internal interfaces other than the local
453 iface_set_mac(iface);
458 /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
459 * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
460 * deletes from 'br' any ports that no longer have any interfaces. */
462 iterate_and_prune_ifaces(struct bridge *br,
463 bool (*cb)(struct bridge *, struct iface *,
467 struct port *port, *next_port;
469 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
470 struct iface *iface, *next_iface;
472 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
473 if (!cb(br, iface, aux)) {
474 iface_set_ofport(iface->cfg, -1);
475 iface_destroy(iface);
479 if (list_is_empty(&port->ifaces)) {
480 VLOG_WARN("%s port has no interfaces, dropping", port->name);
486 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
487 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
488 * responsible for freeing '*managersp' (with free()).
490 * You may be asking yourself "why does ovs-vswitchd care?", because
491 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
492 * should not be and in fact is not directly involved in that. But
493 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
494 * it has to tell in-band control where the managers are to enable that.
495 * (Thus, only managers connected in-band are collected.)
498 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
499 struct sockaddr_in **managersp, size_t *n_managersp)
501 struct sockaddr_in *managers = NULL;
502 size_t n_managers = 0;
506 /* Collect all of the potential targets from the "targets" columns of the
507 * rows pointed to by "manager_options", excluding any that are
510 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
511 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
513 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
514 sset_find_and_delete(&targets, m->target);
516 sset_add(&targets, m->target);
520 /* Now extract the targets' IP addresses. */
521 if (!sset_is_empty(&targets)) {
524 managers = xmalloc(sset_count(&targets) * sizeof *managers);
525 SSET_FOR_EACH (target, &targets) {
526 struct sockaddr_in *sin = &managers[n_managers];
528 if ((!strncmp(target, "tcp:", 4)
529 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
530 (!strncmp(target, "ssl:", 4)
531 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
536 sset_destroy(&targets);
538 *managersp = managers;
539 *n_managersp = n_managers;
543 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
545 struct shash old_br, new_br;
546 struct shash_node *node;
547 struct bridge *br, *next;
548 struct sockaddr_in *managers;
551 int sflow_bridge_number;
553 COVERAGE_INC(bridge_reconfigure);
555 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
557 /* Collect old and new bridges. */
560 LIST_FOR_EACH (br, node, &all_bridges) {
561 shash_add(&old_br, br->name, br);
563 for (i = 0; i < ovs_cfg->n_bridges; i++) {
564 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
565 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
566 VLOG_WARN("more than one bridge named %s", br_cfg->name);
570 /* Get rid of deleted bridges and add new bridges. */
571 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
572 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
579 SHASH_FOR_EACH (node, &new_br) {
580 const char *br_name = node->name;
581 const struct ovsrec_bridge *br_cfg = node->data;
582 br = shash_find_data(&old_br, br_name);
584 /* If the bridge datapath type has changed, we need to tear it
585 * down and recreate. */
586 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
588 bridge_create(br_cfg);
591 bridge_create(br_cfg);
594 shash_destroy(&old_br);
595 shash_destroy(&new_br);
597 /* Reconfigure all bridges. */
598 LIST_FOR_EACH (br, node, &all_bridges) {
599 bridge_reconfigure_one(br);
602 /* Add and delete ports on all datapaths.
604 * The kernel will reject any attempt to add a given port to a datapath if
605 * that port already belongs to a different datapath, so we must do all
606 * port deletions before any port additions. */
607 LIST_FOR_EACH (br, node, &all_bridges) {
608 struct dpif_port_dump dump;
609 struct shash want_ifaces;
610 struct dpif_port dpif_port;
612 bridge_get_all_ifaces(br, &want_ifaces);
613 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
614 if (!shash_find(&want_ifaces, dpif_port.name)
615 && strcmp(dpif_port.name, br->name)) {
616 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
618 VLOG_WARN("failed to remove %s interface from %s: %s",
619 dpif_port.name, dpif_name(br->dpif),
624 shash_destroy(&want_ifaces);
626 LIST_FOR_EACH (br, node, &all_bridges) {
627 struct shash cur_ifaces, want_ifaces;
628 struct dpif_port_dump dump;
629 struct dpif_port dpif_port;
631 /* Get the set of interfaces currently in this datapath. */
632 shash_init(&cur_ifaces);
633 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
634 struct dpif_port *port_info = xmalloc(sizeof *port_info);
635 dpif_port_clone(port_info, &dpif_port);
636 shash_add(&cur_ifaces, dpif_port.name, port_info);
639 /* Get the set of interfaces we want on this datapath. */
640 bridge_get_all_ifaces(br, &want_ifaces);
642 hmap_clear(&br->ifaces);
643 SHASH_FOR_EACH (node, &want_ifaces) {
644 const char *if_name = node->name;
645 struct iface *iface = node->data;
646 struct dpif_port *dpif_port;
650 type = iface ? iface->type : "internal";
651 dpif_port = shash_find_data(&cur_ifaces, if_name);
653 /* If we have a port or a netdev already, and it's not the type we
654 * want, then delete the port (if any) and close the netdev (if
656 if ((dpif_port && strcmp(dpif_port->type, type))
657 || (iface && iface->netdev
658 && strcmp(type, netdev_get_type(iface->netdev)))) {
660 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
667 netdev_close(iface->netdev);
668 iface->netdev = NULL;
672 /* If the port doesn't exist or we don't have the netdev open,
673 * we need to do more work. */
674 if (!dpif_port || (iface && !iface->netdev)) {
675 struct netdev_options options;
676 struct netdev *netdev;
679 /* First open the network device. */
680 options.name = if_name;
682 options.args = &args;
683 options.ethertype = NETDEV_ETH_TYPE_NONE;
687 shash_from_ovs_idl_map(iface->cfg->key_options,
688 iface->cfg->value_options,
689 iface->cfg->n_options, &args);
691 error = netdev_open(&options, &netdev);
692 shash_destroy(&args);
695 VLOG_WARN("could not open network device %s (%s)",
696 if_name, strerror(error));
700 /* Then add the port if we haven't already. */
702 error = dpif_port_add(br->dpif, netdev, NULL);
704 netdev_close(netdev);
705 if (error == EFBIG) {
706 VLOG_ERR("ran out of valid port numbers on %s",
707 dpif_name(br->dpif));
710 VLOG_WARN("failed to add %s interface to %s: %s",
711 if_name, dpif_name(br->dpif),
718 /* Update 'iface'. */
720 iface->netdev = netdev;
722 } else if (iface && iface->netdev) {
726 shash_from_ovs_idl_map(iface->cfg->key_options,
727 iface->cfg->value_options,
728 iface->cfg->n_options, &args);
729 netdev_set_config(iface->netdev, &args);
730 shash_destroy(&args);
733 shash_destroy(&want_ifaces);
735 SHASH_FOR_EACH (node, &cur_ifaces) {
736 struct dpif_port *port_info = node->data;
737 dpif_port_destroy(port_info);
740 shash_destroy(&cur_ifaces);
742 sflow_bridge_number = 0;
743 LIST_FOR_EACH (br, node, &all_bridges) {
744 uint8_t ea[ETH_ADDR_LEN];
746 struct iface *local_iface;
747 struct iface *hw_addr_iface;
750 bridge_fetch_dp_ifaces(br);
752 /* Delete interfaces that cannot be opened.
754 * From this point forward we are guaranteed that every "struct iface"
755 * has nonnull 'netdev' and correct 'dp_ifidx'. */
756 iterate_and_prune_ifaces(br, check_iface, NULL);
758 /* Pick local port hardware address, datapath ID. */
759 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
760 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
762 int error = netdev_set_etheraddr(local_iface->netdev, ea);
764 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
765 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
766 "Ethernet address: %s",
767 br->name, strerror(error));
770 memcpy(br->ea, ea, ETH_ADDR_LEN);
772 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
773 ofproto_set_datapath_id(br->ofproto, dpid);
775 dpid_string = xasprintf("%016"PRIx64, dpid);
776 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
779 /* Set NetFlow configuration on this bridge. */
780 if (br->cfg->netflow) {
781 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
782 struct netflow_options opts;
784 memset(&opts, 0, sizeof opts);
786 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
787 if (nf_cfg->engine_type) {
788 opts.engine_type = *nf_cfg->engine_type;
790 if (nf_cfg->engine_id) {
791 opts.engine_id = *nf_cfg->engine_id;
794 opts.active_timeout = nf_cfg->active_timeout;
795 if (!opts.active_timeout) {
796 opts.active_timeout = -1;
797 } else if (opts.active_timeout < 0) {
798 VLOG_WARN("bridge %s: active timeout interval set to negative "
799 "value, using default instead (%d seconds)", br->name,
800 NF_ACTIVE_TIMEOUT_DEFAULT);
801 opts.active_timeout = -1;
804 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
805 if (opts.add_id_to_iface) {
806 if (opts.engine_id > 0x7f) {
807 VLOG_WARN("bridge %s: netflow port mangling may conflict "
808 "with another vswitch, choose an engine id less "
809 "than 128", br->name);
811 if (hmap_count(&br->ports) > 508) {
812 VLOG_WARN("bridge %s: netflow port mangling will conflict "
813 "with another port when more than 508 ports are "
818 sset_init(&opts.collectors);
819 sset_add_array(&opts.collectors,
820 nf_cfg->targets, nf_cfg->n_targets);
821 if (ofproto_set_netflow(br->ofproto, &opts)) {
822 VLOG_ERR("bridge %s: problem setting netflow collectors",
825 sset_destroy(&opts.collectors);
827 ofproto_set_netflow(br->ofproto, NULL);
830 /* Set sFlow configuration on this bridge. */
831 if (br->cfg->sflow) {
832 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
833 struct ovsrec_controller **controllers;
834 struct ofproto_sflow_options oso;
835 size_t n_controllers;
837 memset(&oso, 0, sizeof oso);
839 sset_init(&oso.targets);
840 sset_add_array(&oso.targets,
841 sflow_cfg->targets, sflow_cfg->n_targets);
843 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
844 if (sflow_cfg->sampling) {
845 oso.sampling_rate = *sflow_cfg->sampling;
848 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
849 if (sflow_cfg->polling) {
850 oso.polling_interval = *sflow_cfg->polling;
853 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
854 if (sflow_cfg->header) {
855 oso.header_len = *sflow_cfg->header;
858 oso.sub_id = sflow_bridge_number++;
859 oso.agent_device = sflow_cfg->agent;
861 oso.control_ip = NULL;
862 n_controllers = bridge_get_controllers(br, &controllers);
863 for (i = 0; i < n_controllers; i++) {
864 if (controllers[i]->local_ip) {
865 oso.control_ip = controllers[i]->local_ip;
869 ofproto_set_sflow(br->ofproto, &oso);
871 sset_destroy(&oso.targets);
873 ofproto_set_sflow(br->ofproto, NULL);
876 /* Update the controller and related settings. It would be more
877 * straightforward to call this from bridge_reconfigure_one(), but we
878 * can't do it there for two reasons. First, and most importantly, at
879 * that point we don't know the dp_ifidx of any interfaces that have
880 * been added to the bridge (because we haven't actually added them to
881 * the datapath). Second, at that point we haven't set the datapath ID
882 * yet; when a controller is configured, resetting the datapath ID will
883 * immediately disconnect from the controller, so it's better to set
884 * the datapath ID before the controller. */
885 bridge_reconfigure_remotes(br, managers, n_managers);
887 LIST_FOR_EACH (br, node, &all_bridges) {
890 br->has_bonded_ports = false;
891 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
894 port_reconfigure_lacp(port);
895 port_reconfigure_bond(port);
897 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
898 iface_update_qos(iface, port->cfg->qos);
902 LIST_FOR_EACH (br, node, &all_bridges) {
903 iterate_and_prune_ifaces(br, set_iface_properties, NULL);
906 /* Some reconfiguration operations require the bridge to have been run at
908 LIST_FOR_EACH (br, node, &all_bridges) {
913 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
914 iface_update_cfm(iface);
920 /* ovs-vswitchd has completed initialization, so allow the process that
921 * forked us to exit successfully. */
922 daemonize_complete();
926 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
927 const struct ovsdb_idl_column *column,
930 const struct ovsdb_datum *datum;
931 union ovsdb_atom atom;
934 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
935 atom.string = (char *) key;
936 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
937 return idx == UINT_MAX ? NULL : datum->values[idx].string;
941 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
943 return get_ovsrec_key_value(&br_cfg->header_,
944 &ovsrec_bridge_col_other_config, key);
948 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
949 struct iface **hw_addr_iface)
955 *hw_addr_iface = NULL;
957 /* Did the user request a particular MAC? */
958 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
959 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
960 if (eth_addr_is_multicast(ea)) {
961 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
962 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
963 } else if (eth_addr_is_zero(ea)) {
964 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
970 /* Otherwise choose the minimum non-local MAC address among all of the
972 memset(ea, 0xff, ETH_ADDR_LEN);
973 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
974 uint8_t iface_ea[ETH_ADDR_LEN];
975 struct iface *candidate;
978 /* Mirror output ports don't participate. */
979 if (port->is_mirror_output_port) {
983 /* Choose the MAC address to represent the port. */
985 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
986 /* Find the interface with this Ethernet address (if any) so that
987 * we can provide the correct devname to the caller. */
988 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
989 uint8_t candidate_ea[ETH_ADDR_LEN];
990 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
991 && eth_addr_equals(iface_ea, candidate_ea)) {
996 /* Choose the interface whose MAC address will represent the port.
997 * The Linux kernel bonding code always chooses the MAC address of
998 * the first slave added to a bond, and the Fedora networking
999 * scripts always add slaves to a bond in alphabetical order, so
1000 * for compatibility we choose the interface with the name that is
1001 * first in alphabetical order. */
1002 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1003 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1008 /* The local port doesn't count (since we're trying to choose its
1009 * MAC address anyway). */
1010 if (iface->dp_ifidx == ODPP_LOCAL) {
1015 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1017 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1018 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
1019 iface->name, strerror(error));
1024 /* Compare against our current choice. */
1025 if (!eth_addr_is_multicast(iface_ea) &&
1026 !eth_addr_is_local(iface_ea) &&
1027 !eth_addr_is_reserved(iface_ea) &&
1028 !eth_addr_is_zero(iface_ea) &&
1029 eth_addr_compare_3way(iface_ea, ea) < 0)
1031 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1032 *hw_addr_iface = iface;
1035 if (eth_addr_is_multicast(ea)) {
1036 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1037 *hw_addr_iface = NULL;
1038 VLOG_WARN("bridge %s: using default bridge Ethernet "
1039 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1041 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1042 br->name, ETH_ADDR_ARGS(ea));
1046 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1047 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1048 * an interface on 'br', then that interface must be passed in as
1049 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1050 * 'hw_addr_iface' must be passed in as a null pointer. */
1052 bridge_pick_datapath_id(struct bridge *br,
1053 const uint8_t bridge_ea[ETH_ADDR_LEN],
1054 struct iface *hw_addr_iface)
1057 * The procedure for choosing a bridge MAC address will, in the most
1058 * ordinary case, also choose a unique MAC that we can use as a datapath
1059 * ID. In some special cases, though, multiple bridges will end up with
1060 * the same MAC address. This is OK for the bridges, but it will confuse
1061 * the OpenFlow controller, because each datapath needs a unique datapath
1064 * Datapath IDs must be unique. It is also very desirable that they be
1065 * stable from one run to the next, so that policy set on a datapath
1068 const char *datapath_id;
1071 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1072 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1076 if (hw_addr_iface) {
1078 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1080 * A bridge whose MAC address is taken from a VLAN network device
1081 * (that is, a network device created with vconfig(8) or similar
1082 * tool) will have the same MAC address as a bridge on the VLAN
1083 * device's physical network device.
1085 * Handle this case by hashing the physical network device MAC
1086 * along with the VLAN identifier.
1088 uint8_t buf[ETH_ADDR_LEN + 2];
1089 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1090 buf[ETH_ADDR_LEN] = vlan >> 8;
1091 buf[ETH_ADDR_LEN + 1] = vlan;
1092 return dpid_from_hash(buf, sizeof buf);
1095 * Assume that this bridge's MAC address is unique, since it
1096 * doesn't fit any of the cases we handle specially.
1101 * A purely internal bridge, that is, one that has no non-virtual
1102 * network devices on it at all, is more difficult because it has no
1103 * natural unique identifier at all.
1105 * When the host is a XenServer, we handle this case by hashing the
1106 * host's UUID with the name of the bridge. Names of bridges are
1107 * persistent across XenServer reboots, although they can be reused if
1108 * an internal network is destroyed and then a new one is later
1109 * created, so this is fairly effective.
1111 * When the host is not a XenServer, we punt by using a random MAC
1112 * address on each run.
1114 const char *host_uuid = xenserver_get_host_uuid();
1116 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1117 dpid = dpid_from_hash(combined, strlen(combined));
1123 return eth_addr_to_uint64(bridge_ea);
1127 dpid_from_hash(const void *data, size_t n)
1129 uint8_t hash[SHA1_DIGEST_SIZE];
1131 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1132 sha1_bytes(data, n, hash);
1133 eth_addr_mark_random(hash);
1134 return eth_addr_to_uint64(hash);
1138 iface_refresh_status(struct iface *iface)
1142 enum netdev_flags flags;
1149 if (iface_is_synthetic(iface)) {
1155 if (!netdev_get_status(iface->netdev, &sh)) {
1157 char **keys, **values;
1159 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1160 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1165 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1168 shash_destroy_free_data(&sh);
1170 error = netdev_get_flags(iface->netdev, &flags);
1172 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1175 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1178 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1180 ovsrec_interface_set_duplex(iface->cfg,
1181 netdev_features_is_full_duplex(current)
1183 /* warning: uint64_t -> int64_t conversion */
1184 bps = netdev_features_to_bps(current);
1185 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1188 ovsrec_interface_set_duplex(iface->cfg, NULL);
1189 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1193 ovsrec_interface_set_link_state(iface->cfg,
1194 iface_get_carrier(iface) ? "up" : "down");
1196 error = netdev_get_mtu(iface->netdev, &mtu);
1197 if (!error && mtu != INT_MAX) {
1199 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1202 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1206 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1207 * changed, false otherwise. */
1209 iface_refresh_cfm_stats(struct iface *iface)
1211 const struct ovsrec_monitor *mon;
1212 const struct cfm *cfm;
1213 bool changed = false;
1216 mon = iface->cfg->monitor;
1217 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1223 for (i = 0; i < mon->n_remote_mps; i++) {
1224 const struct ovsrec_maintenance_point *mp;
1225 const struct remote_mp *rmp;
1227 mp = mon->remote_mps[i];
1228 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1230 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1231 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1236 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1237 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1245 iface_refresh_lacp_stats(struct iface *iface)
1247 bool *db_current = iface->cfg->lacp_current;
1248 bool changed = false;
1250 if (iface->port->lacp) {
1251 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1253 if (!db_current || *db_current != current) {
1255 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1257 } else if (db_current) {
1259 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1266 iface_refresh_stats(struct iface *iface)
1272 static const struct iface_stat iface_stats[] = {
1273 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1274 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1275 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1276 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1277 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1278 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1279 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1280 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1281 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1282 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1283 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1284 { "collisions", offsetof(struct netdev_stats, collisions) },
1286 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1287 const struct iface_stat *s;
1289 char *keys[N_STATS];
1290 int64_t values[N_STATS];
1293 struct netdev_stats stats;
1295 if (iface_is_synthetic(iface)) {
1299 /* Intentionally ignore return value, since errors will set 'stats' to
1300 * all-1s, and we will deal with that correctly below. */
1301 netdev_get_stats(iface->netdev, &stats);
1304 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1305 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1306 if (value != UINT64_MAX) {
1313 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1317 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1319 struct ovsdb_datum datum;
1323 get_system_stats(&stats);
1325 ovsdb_datum_from_shash(&datum, &stats);
1326 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1330 static inline const char *
1331 nx_role_to_str(enum nx_role role)
1336 case NX_ROLE_MASTER:
1341 return "*** INVALID ROLE ***";
1346 bridge_refresh_controller_status(const struct bridge *br)
1349 const struct ovsrec_controller *cfg;
1351 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1353 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1354 struct ofproto_controller_info *cinfo =
1355 shash_find_data(&info, cfg->target);
1358 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1359 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1360 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1361 (char **) cinfo->pairs.values,
1364 ovsrec_controller_set_is_connected(cfg, false);
1365 ovsrec_controller_set_role(cfg, NULL);
1366 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1370 ofproto_free_ofproto_controller_info(&info);
1376 const struct ovsrec_open_vswitch *cfg;
1378 bool datapath_destroyed;
1379 bool database_changed;
1382 /* Let each bridge do the work that it needs to do. */
1383 datapath_destroyed = false;
1384 LIST_FOR_EACH (br, node, &all_bridges) {
1385 int error = bridge_run_one(br);
1387 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1388 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1389 "forcing reconfiguration", br->name);
1390 datapath_destroyed = true;
1394 /* (Re)configure if necessary. */
1395 database_changed = ovsdb_idl_run(idl);
1396 cfg = ovsrec_open_vswitch_first(idl);
1398 /* Re-configure SSL. We do this on every trip through the main loop,
1399 * instead of just when the database changes, because the contents of the
1400 * key and certificate files can change without the database changing.
1402 * We do this before bridge_reconfigure() because that function might
1403 * initiate SSL connections and thus requires SSL to be configured. */
1404 if (cfg && cfg->ssl) {
1405 const struct ovsrec_ssl *ssl = cfg->ssl;
1407 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1408 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1411 if (database_changed || datapath_destroyed) {
1413 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1415 bridge_configure_once(cfg);
1416 bridge_reconfigure(cfg);
1418 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1419 ovsdb_idl_txn_commit(txn);
1420 ovsdb_idl_txn_destroy(txn); /* XXX */
1422 /* We still need to reconfigure to avoid dangling pointers to
1423 * now-destroyed ovsrec structures inside bridge data. */
1424 static const struct ovsrec_open_vswitch null_cfg;
1426 bridge_reconfigure(&null_cfg);
1430 /* Refresh system and interface stats if necessary. */
1431 if (time_msec() >= stats_timer) {
1433 struct ovsdb_idl_txn *txn;
1435 txn = ovsdb_idl_txn_create(idl);
1436 LIST_FOR_EACH (br, node, &all_bridges) {
1439 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1440 struct iface *iface;
1442 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1443 iface_refresh_stats(iface);
1444 iface_refresh_status(iface);
1447 bridge_refresh_controller_status(br);
1449 refresh_system_stats(cfg);
1450 ovsdb_idl_txn_commit(txn);
1451 ovsdb_idl_txn_destroy(txn); /* XXX */
1454 stats_timer = time_msec() + STATS_INTERVAL;
1457 if (time_msec() >= db_limiter) {
1458 struct ovsdb_idl_txn *txn;
1459 bool changed = false;
1461 txn = ovsdb_idl_txn_create(idl);
1462 LIST_FOR_EACH (br, node, &all_bridges) {
1465 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1466 struct iface *iface;
1468 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1469 changed = iface_refresh_cfm_stats(iface) || changed;
1470 changed = iface_refresh_lacp_stats(iface) || changed;
1476 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1479 ovsdb_idl_txn_commit(txn);
1480 ovsdb_idl_txn_destroy(txn);
1489 LIST_FOR_EACH (br, node, &all_bridges) {
1492 ofproto_wait(br->ofproto);
1493 mac_learning_wait(br->ml);
1494 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1498 ovsdb_idl_wait(idl);
1499 poll_timer_wait_until(stats_timer);
1501 if (db_limiter > time_msec()) {
1502 poll_timer_wait_until(db_limiter);
1506 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1507 * configuration changes. */
1509 bridge_flush(struct bridge *br)
1511 COVERAGE_INC(bridge_flush);
1515 /* Bridge unixctl user interface functions. */
1517 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1518 const char *args, void *aux OVS_UNUSED)
1520 struct ds ds = DS_EMPTY_INITIALIZER;
1521 const struct bridge *br;
1522 const struct mac_entry *e;
1524 br = bridge_lookup(args);
1526 unixctl_command_reply(conn, 501, "no such bridge");
1530 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1531 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1532 struct port *port = e->port.p;
1533 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1534 port_get_an_iface(port)->dp_ifidx,
1535 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1537 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1541 /* CFM unixctl user interface functions. */
1543 cfm_unixctl_show(struct unixctl_conn *conn,
1544 const char *args, void *aux OVS_UNUSED)
1546 struct ds ds = DS_EMPTY_INITIALIZER;
1547 struct iface *iface;
1548 const struct cfm *cfm;
1550 iface = iface_find(args);
1552 unixctl_command_reply(conn, 501, "no such interface");
1556 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1559 unixctl_command_reply(conn, 501, "CFM not enabled");
1563 cfm_dump_ds(cfm, &ds);
1564 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1568 /* QoS unixctl user interface functions. */
1570 struct qos_unixctl_show_cbdata {
1572 struct iface *iface;
1576 qos_unixctl_show_cb(unsigned int queue_id,
1577 const struct shash *details,
1580 struct qos_unixctl_show_cbdata *data = aux;
1581 struct ds *ds = data->ds;
1582 struct iface *iface = data->iface;
1583 struct netdev_queue_stats stats;
1584 struct shash_node *node;
1587 ds_put_cstr(ds, "\n");
1589 ds_put_format(ds, "Queue %u:\n", queue_id);
1591 ds_put_cstr(ds, "Default:\n");
1594 SHASH_FOR_EACH (node, details) {
1595 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1598 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1600 if (stats.tx_packets != UINT64_MAX) {
1601 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1604 if (stats.tx_bytes != UINT64_MAX) {
1605 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1608 if (stats.tx_errors != UINT64_MAX) {
1609 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1612 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1613 queue_id, strerror(error));
1618 qos_unixctl_show(struct unixctl_conn *conn,
1619 const char *args, void *aux OVS_UNUSED)
1621 struct ds ds = DS_EMPTY_INITIALIZER;
1622 struct shash sh = SHASH_INITIALIZER(&sh);
1623 struct iface *iface;
1625 struct shash_node *node;
1626 struct qos_unixctl_show_cbdata data;
1629 iface = iface_find(args);
1631 unixctl_command_reply(conn, 501, "no such interface");
1635 netdev_get_qos(iface->netdev, &type, &sh);
1637 if (*type != '\0') {
1638 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1640 SHASH_FOR_EACH (node, &sh) {
1641 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1646 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1649 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1651 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1653 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1654 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1657 shash_destroy_free_data(&sh);
1661 /* Bridge reconfiguration functions. */
1662 static struct bridge *
1663 bridge_create(const struct ovsrec_bridge *br_cfg)
1668 assert(!bridge_lookup(br_cfg->name));
1669 br = xzalloc(sizeof *br);
1671 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1678 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1681 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1683 dpif_delete(br->dpif);
1684 dpif_close(br->dpif);
1689 br->name = xstrdup(br_cfg->name);
1691 br->ml = mac_learning_create();
1692 eth_addr_nicira_random(br->default_ea);
1694 hmap_init(&br->ports);
1695 hmap_init(&br->ifaces);
1696 shash_init(&br->iface_by_name);
1700 list_push_back(&all_bridges, &br->node);
1702 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1708 bridge_destroy(struct bridge *br)
1711 struct port *port, *next;
1715 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1718 for (i = 0; i < MAX_MIRRORS; i++) {
1719 mirror_destroy(br->mirrors[i]);
1721 list_remove(&br->node);
1722 ofproto_destroy(br->ofproto);
1723 error = dpif_delete(br->dpif);
1724 if (error && error != ENOENT) {
1725 VLOG_ERR("failed to delete %s: %s",
1726 dpif_name(br->dpif), strerror(error));
1728 dpif_close(br->dpif);
1729 mac_learning_destroy(br->ml);
1730 hmap_destroy(&br->ifaces);
1731 hmap_destroy(&br->ports);
1732 shash_destroy(&br->iface_by_name);
1733 free(br->synth_local_iface.type);
1739 static struct bridge *
1740 bridge_lookup(const char *name)
1744 LIST_FOR_EACH (br, node, &all_bridges) {
1745 if (!strcmp(br->name, name)) {
1752 /* Handle requests for a listing of all flows known by the OpenFlow
1753 * stack, including those normally hidden. */
1755 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1756 const char *args, void *aux OVS_UNUSED)
1761 br = bridge_lookup(args);
1763 unixctl_command_reply(conn, 501, "Unknown bridge");
1768 ofproto_get_all_flows(br->ofproto, &results);
1770 unixctl_command_reply(conn, 200, ds_cstr(&results));
1771 ds_destroy(&results);
1774 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1775 * connections and reconnect. If BRIDGE is not specified, then all bridges
1776 * drop their controller connections and reconnect. */
1778 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1779 const char *args, void *aux OVS_UNUSED)
1782 if (args[0] != '\0') {
1783 br = bridge_lookup(args);
1785 unixctl_command_reply(conn, 501, "Unknown bridge");
1788 ofproto_reconnect_controllers(br->ofproto);
1790 LIST_FOR_EACH (br, node, &all_bridges) {
1791 ofproto_reconnect_controllers(br->ofproto);
1794 unixctl_command_reply(conn, 200, NULL);
1798 bridge_run_one(struct bridge *br)
1803 error = ofproto_run1(br->ofproto);
1808 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1810 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1814 error = ofproto_run2(br->ofproto, br->flush);
1821 bridge_get_controllers(const struct bridge *br,
1822 struct ovsrec_controller ***controllersp)
1824 struct ovsrec_controller **controllers;
1825 size_t n_controllers;
1827 controllers = br->cfg->controller;
1828 n_controllers = br->cfg->n_controller;
1830 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1836 *controllersp = controllers;
1838 return n_controllers;
1842 bridge_reconfigure_one(struct bridge *br)
1844 enum ofproto_fail_mode fail_mode;
1845 struct port *port, *next;
1846 struct shash_node *node;
1847 struct shash new_ports;
1850 /* Collect new ports. */
1851 shash_init(&new_ports);
1852 for (i = 0; i < br->cfg->n_ports; i++) {
1853 const char *name = br->cfg->ports[i]->name;
1854 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1855 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1859 if (!shash_find(&new_ports, br->name)) {
1860 struct dpif_port dpif_port;
1863 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1864 br->name, br->name);
1866 dpif_port_query_by_number(br->dpif, ODPP_LOCAL, &dpif_port);
1867 type = xstrdup(dpif_port.type ? dpif_port.type : "internal");
1868 dpif_port_destroy(&dpif_port);
1870 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1871 br->synth_local_port.n_interfaces = 1;
1872 br->synth_local_port.name = br->name;
1874 br->synth_local_iface.name = br->name;
1875 free(br->synth_local_iface.type);
1876 br->synth_local_iface.type = type;
1878 br->synth_local_ifacep = &br->synth_local_iface;
1880 shash_add(&new_ports, br->name, &br->synth_local_port);
1883 /* Get rid of deleted ports.
1884 * Get rid of deleted interfaces on ports that still exist. */
1885 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1886 const struct ovsrec_port *port_cfg;
1888 port_cfg = shash_find_data(&new_ports, port->name);
1892 port_del_ifaces(port, port_cfg);
1896 /* Create new ports.
1897 * Add new interfaces to existing ports.
1898 * Reconfigure existing ports. */
1899 SHASH_FOR_EACH (node, &new_ports) {
1900 struct port *port = port_lookup(br, node->name);
1902 port = port_create(br, node->name);
1905 port_reconfigure(port, node->data);
1906 if (list_is_empty(&port->ifaces)) {
1907 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1908 br->name, port->name);
1912 shash_destroy(&new_ports);
1914 /* Set the fail-mode */
1915 fail_mode = !br->cfg->fail_mode
1916 || !strcmp(br->cfg->fail_mode, "standalone")
1917 ? OFPROTO_FAIL_STANDALONE
1918 : OFPROTO_FAIL_SECURE;
1919 if (ofproto_get_fail_mode(br->ofproto) != fail_mode
1920 && !ofproto_has_primary_controller(br->ofproto)) {
1921 ofproto_flush_flows(br->ofproto);
1923 ofproto_set_fail_mode(br->ofproto, fail_mode);
1925 /* Delete all flows if we're switching from connected to standalone or vice
1926 * versa. (XXX Should we delete all flows if we are switching from one
1927 * controller to another?) */
1929 /* Configure OpenFlow controller connection snooping. */
1930 if (!ofproto_has_snoops(br->ofproto)) {
1934 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1935 ovs_rundir(), br->name));
1936 ofproto_set_snoops(br->ofproto, &snoops);
1937 sset_destroy(&snoops);
1940 mirror_reconfigure(br);
1943 /* Initializes 'oc' appropriately as a management service controller for
1946 * The caller must free oc->target when it is no longer needed. */
1948 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1949 struct ofproto_controller *oc)
1951 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1952 oc->max_backoff = 0;
1953 oc->probe_interval = 60;
1954 oc->band = OFPROTO_OUT_OF_BAND;
1956 oc->burst_limit = 0;
1959 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1961 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1962 struct ofproto_controller *oc)
1964 oc->target = c->target;
1965 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1966 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1967 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1968 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1969 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1970 oc->burst_limit = (c->controller_burst_limit
1971 ? *c->controller_burst_limit : 0);
1974 /* Configures the IP stack for 'br''s local interface properly according to the
1975 * configuration in 'c'. */
1977 bridge_configure_local_iface_netdev(struct bridge *br,
1978 struct ovsrec_controller *c)
1980 struct netdev *netdev;
1981 struct in_addr mask, gateway;
1983 struct iface *local_iface;
1986 /* If there's no local interface or no IP address, give up. */
1987 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1988 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1992 /* Bring up the local interface. */
1993 netdev = local_iface->netdev;
1994 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1996 /* Configure the IP address and netmask. */
1997 if (!c->local_netmask
1998 || !inet_aton(c->local_netmask, &mask)
2000 mask.s_addr = guess_netmask(ip.s_addr);
2002 if (!netdev_set_in4(netdev, ip, mask)) {
2003 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2004 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2007 /* Configure the default gateway. */
2008 if (c->local_gateway
2009 && inet_aton(c->local_gateway, &gateway)
2010 && gateway.s_addr) {
2011 if (!netdev_add_router(netdev, gateway)) {
2012 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2013 br->name, IP_ARGS(&gateway.s_addr));
2019 bridge_reconfigure_remotes(struct bridge *br,
2020 const struct sockaddr_in *managers,
2023 const char *disable_ib_str, *queue_id_str;
2024 bool disable_in_band = false;
2027 struct ovsrec_controller **controllers;
2028 size_t n_controllers;
2031 struct ofproto_controller *ocs;
2035 /* Check if we should disable in-band control on this bridge. */
2036 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2037 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2038 disable_in_band = true;
2041 /* Set OpenFlow queue ID for in-band control. */
2042 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2043 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2044 ofproto_set_in_band_queue(br->ofproto, queue_id);
2046 if (disable_in_band) {
2047 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2049 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2051 had_primary = ofproto_has_primary_controller(br->ofproto);
2053 n_controllers = bridge_get_controllers(br, &controllers);
2055 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2058 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2059 for (i = 0; i < n_controllers; i++) {
2060 struct ovsrec_controller *c = controllers[i];
2062 if (!strncmp(c->target, "punix:", 6)
2063 || !strncmp(c->target, "unix:", 5)) {
2064 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2066 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2067 * domain sockets and overwriting arbitrary local files. */
2068 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
2069 "\"%s\" due to possibility for remote exploit",
2070 dpif_name(br->dpif), c->target);
2074 bridge_configure_local_iface_netdev(br, c);
2075 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2076 if (disable_in_band) {
2077 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2082 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2083 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2086 if (had_primary != ofproto_has_primary_controller(br->ofproto)) {
2087 ofproto_flush_flows(br->ofproto);
2090 /* If there are no controllers and the bridge is in standalone
2091 * mode, set up a flow that matches every packet and directs
2092 * them to OFPP_NORMAL (which goes to us). Otherwise, the
2093 * switch is in secure mode and we won't pass any traffic until
2094 * a controller has been defined and it tells us to do so. */
2096 && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) {
2097 union ofp_action action;
2098 struct cls_rule rule;
2100 memset(&action, 0, sizeof action);
2101 action.type = htons(OFPAT_OUTPUT);
2102 action.output.len = htons(sizeof action);
2103 action.output.port = htons(OFPP_NORMAL);
2104 cls_rule_init_catchall(&rule, 0);
2105 ofproto_add_flow(br->ofproto, &rule, &action, 1);
2110 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2115 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2116 struct iface *iface;
2118 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2119 shash_add_once(ifaces, iface->name, iface);
2121 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2122 shash_add_once(ifaces, port->name, NULL);
2127 /* For robustness, in case the administrator moves around datapath ports behind
2128 * our back, we re-check all the datapath port numbers here.
2130 * This function will set the 'dp_ifidx' members of interfaces that have
2131 * disappeared to -1, so only call this function from a context where those
2132 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2133 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2134 * datapath, which doesn't support UINT16_MAX+1 ports. */
2136 bridge_fetch_dp_ifaces(struct bridge *br)
2138 struct dpif_port_dump dump;
2139 struct dpif_port dpif_port;
2142 /* Reset all interface numbers. */
2143 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2144 struct iface *iface;
2146 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2147 iface->dp_ifidx = -1;
2150 hmap_clear(&br->ifaces);
2152 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2153 struct iface *iface = iface_lookup(br, dpif_port.name);
2155 if (iface->dp_ifidx >= 0) {
2156 VLOG_WARN("%s reported interface %s twice",
2157 dpif_name(br->dpif), dpif_port.name);
2158 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2159 VLOG_WARN("%s reported interface %"PRIu16" twice",
2160 dpif_name(br->dpif), dpif_port.port_no);
2162 iface->dp_ifidx = dpif_port.port_no;
2163 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2164 hash_int(iface->dp_ifidx, 0));
2167 iface_set_ofport(iface->cfg,
2168 (iface->dp_ifidx >= 0
2169 ? odp_port_to_ofp_port(iface->dp_ifidx)
2175 /* Bridge packet processing functions. */
2178 set_dst(struct dst *dst, const struct flow *flow,
2179 const struct port *in_port, const struct port *out_port,
2182 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2183 : in_port->vlan >= 0 ? in_port->vlan
2184 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2185 : vlan_tci_to_vid(flow->vlan_tci));
2187 dst->iface = (!out_port->bond
2188 ? port_get_an_iface(out_port)
2189 : bond_choose_output_slave(out_port->bond, flow,
2192 return dst->iface != NULL;
2196 mirror_mask_ffs(mirror_mask_t mask)
2198 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2203 dst_set_init(struct dst_set *set)
2205 set->dsts = set->builtin;
2207 set->allocated = ARRAY_SIZE(set->builtin);
2211 dst_set_add(struct dst_set *set, const struct dst *dst)
2213 if (set->n >= set->allocated) {
2214 size_t new_allocated;
2215 struct dst *new_dsts;
2217 new_allocated = set->allocated * 2;
2218 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2219 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2223 set->dsts = new_dsts;
2224 set->allocated = new_allocated;
2226 set->dsts[set->n++] = *dst;
2230 dst_set_free(struct dst_set *set)
2232 if (set->dsts != set->builtin) {
2238 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2241 for (i = 0; i < set->n; i++) {
2242 if (set->dsts[i].vlan == test->vlan
2243 && set->dsts[i].iface == test->iface) {
2251 port_trunks_vlan(const struct port *port, uint16_t vlan)
2253 return (port->vlan < 0
2254 && (!port->trunks || bitmap_is_set(port->trunks, vlan)));
2258 port_includes_vlan(const struct port *port, uint16_t vlan)
2260 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2264 port_is_floodable(const struct port *port)
2266 struct iface *iface;
2268 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2269 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2277 /* Returns an arbitrary interface within 'port'. */
2278 static struct iface *
2279 port_get_an_iface(const struct port *port)
2281 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2285 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2286 const struct port *in_port, const struct port *out_port,
2287 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2291 if (out_port == FLOOD_PORT) {
2294 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2296 && port_is_floodable(port)
2297 && port_includes_vlan(port, vlan)
2298 && !port->is_mirror_output_port
2299 && set_dst(&dst, flow, in_port, port, tags)) {
2300 dst_set_add(set, &dst);
2303 *nf_output_iface = NF_OUT_FLOOD;
2304 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2305 dst_set_add(set, &dst);
2306 *nf_output_iface = dst.iface->dp_ifidx;
2311 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2312 uint16_t vlan, const struct port *in_port,
2313 struct dst_set *set, tag_type *tags)
2315 mirror_mask_t mirrors;
2319 mirrors = in_port->src_mirrors;
2320 for (i = 0; i < set->n; i++) {
2321 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2328 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2329 if (flow_vlan == 0) {
2330 flow_vlan = OFP_VLAN_NONE;
2334 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2335 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2339 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2340 && !dst_is_duplicate(set, &dst)) {
2341 dst_set_add(set, &dst);
2346 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2347 if (port_includes_vlan(port, m->out_vlan)
2348 && set_dst(&dst, flow, in_port, port, tags))
2350 if (port->vlan < 0) {
2351 dst.vlan = m->out_vlan;
2353 if (dst_is_duplicate(set, &dst)) {
2357 /* Use the vlan tag on the original flow instead of
2358 * the one passed in the vlan parameter. This ensures
2359 * that we compare the vlan from before any implicit
2360 * tagging tags place. This is necessary because
2361 * dst->vlan is the final vlan, after removing implicit
2363 if (port == in_port && dst.vlan == flow_vlan) {
2364 /* Don't send out input port on same VLAN. */
2367 dst_set_add(set, &dst);
2372 mirrors &= mirrors - 1;
2377 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2378 const struct port *in_port, const struct port *out_port,
2379 tag_type *tags, struct ofpbuf *actions,
2380 uint16_t *nf_output_iface)
2382 uint16_t initial_vlan, cur_vlan;
2383 const struct dst *dst;
2387 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2389 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2391 /* Output all the packets we can without having to change the VLAN. */
2392 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2393 if (initial_vlan == 0) {
2394 initial_vlan = OFP_VLAN_NONE;
2396 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2397 if (dst->vlan != initial_vlan) {
2400 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2403 /* Then output the rest. */
2404 cur_vlan = initial_vlan;
2405 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2406 if (dst->vlan == initial_vlan) {
2409 if (dst->vlan != cur_vlan) {
2410 if (dst->vlan == OFP_VLAN_NONE) {
2411 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2414 tci = htons(dst->vlan & VLAN_VID_MASK);
2415 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2416 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2418 cur_vlan = dst->vlan;
2420 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2426 /* Returns the effective vlan of a packet, taking into account both the
2427 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2428 * the packet is untagged and -1 indicates it has an invalid header and
2429 * should be dropped. */
2430 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2431 struct port *in_port, bool have_packet)
2433 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2434 if (in_port->vlan >= 0) {
2437 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2438 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2439 "packet received on port %s configured with "
2440 "implicit VLAN %"PRIu16,
2441 br->name, vlan, in_port->name, in_port->vlan);
2445 vlan = in_port->vlan;
2447 if (!port_includes_vlan(in_port, vlan)) {
2449 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2450 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2451 "packet received on port %s not configured for "
2453 br->name, vlan, in_port->name, vlan);
2462 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2463 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2464 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2466 is_gratuitous_arp(const struct flow *flow)
2468 return (flow->dl_type == htons(ETH_TYPE_ARP)
2469 && eth_addr_is_broadcast(flow->dl_dst)
2470 && (flow->nw_proto == ARP_OP_REPLY
2471 || (flow->nw_proto == ARP_OP_REQUEST
2472 && flow->nw_src == flow->nw_dst)));
2476 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2477 struct port *in_port)
2479 struct mac_entry *mac;
2481 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2485 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2486 if (is_gratuitous_arp(flow)) {
2487 /* We don't want to learn from gratuitous ARP packets that are
2488 * reflected back over bond slaves so we lock the learning table. */
2489 if (!in_port->bond) {
2490 mac_entry_set_grat_arp_lock(mac);
2491 } else if (mac_entry_is_grat_arp_locked(mac)) {
2496 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2497 /* The log messages here could actually be useful in debugging,
2498 * so keep the rate limit relatively high. */
2499 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2500 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2501 "on port %s in VLAN %d",
2502 br->name, ETH_ADDR_ARGS(flow->dl_src),
2503 in_port->name, vlan);
2505 mac->port.p = in_port;
2506 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2510 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2511 * dropped. Returns true if they may be forwarded, false if they should be
2514 * If 'have_packet' is true, it indicates that the caller is processing a
2515 * received packet. If 'have_packet' is false, then the caller is just
2516 * revalidating an existing flow because configuration has changed. Either
2517 * way, 'have_packet' only affects logging (there is no point in logging errors
2518 * during revalidation).
2520 * Sets '*in_portp' to the input port. This will be a null pointer if
2521 * flow->in_port does not designate a known input port (in which case
2522 * is_admissible() returns false).
2524 * When returning true, sets '*vlanp' to the effective VLAN of the input
2525 * packet, as returned by flow_get_vlan().
2527 * May also add tags to '*tags', although the current implementation only does
2528 * so in one special case.
2531 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2532 tag_type *tags, int *vlanp, struct port **in_portp)
2534 struct iface *in_iface;
2535 struct port *in_port;
2538 /* Find the interface and port structure for the received packet. */
2539 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2541 /* No interface? Something fishy... */
2543 /* Odd. A few possible reasons here:
2545 * - We deleted an interface but there are still a few packets
2546 * queued up from it.
2548 * - Someone externally added an interface (e.g. with "ovs-dpctl
2549 * add-if") that we don't know about.
2551 * - Packet arrived on the local port but the local port is not
2552 * one of our bridge ports.
2554 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2556 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2557 "interface %"PRIu16, br->name, flow->in_port);
2563 *in_portp = in_port = in_iface->port;
2564 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2569 /* Drop frames for reserved multicast addresses. */
2570 if (eth_addr_is_reserved(flow->dl_dst)) {
2574 /* Drop frames on ports reserved for mirroring. */
2575 if (in_port->is_mirror_output_port) {
2577 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2578 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2579 "%s, which is reserved exclusively for mirroring",
2580 br->name, in_port->name);
2585 if (in_port->bond) {
2586 struct mac_entry *mac;
2588 switch (bond_check_admissibility(in_port->bond, in_iface,
2589 flow->dl_dst, tags)) {
2596 case BV_DROP_IF_MOVED:
2597 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2598 if (mac && mac->port.p != in_port &&
2599 (!is_gratuitous_arp(flow)
2600 || mac_entry_is_grat_arp_locked(mac))) {
2610 /* If the composed actions may be applied to any packet in the given 'flow',
2611 * returns true. Otherwise, the actions should only be applied to 'packet', or
2612 * not at all, if 'packet' was NULL. */
2614 process_flow(struct bridge *br, const struct flow *flow,
2615 const struct ofpbuf *packet, struct ofpbuf *actions,
2616 tag_type *tags, uint16_t *nf_output_iface)
2618 struct port *in_port;
2619 struct port *out_port;
2620 struct mac_entry *mac;
2623 /* Check whether we should drop packets in this flow. */
2624 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2629 /* Learn source MAC (but don't try to learn from revalidation). */
2631 update_learning_table(br, flow, vlan, in_port);
2634 /* Determine output port. */
2635 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2637 out_port = mac->port.p;
2638 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2639 /* If we are revalidating but don't have a learning entry then
2640 * eject the flow. Installing a flow that floods packets opens
2641 * up a window of time where we could learn from a packet reflected
2642 * on a bond and blackhole packets before the learning table is
2643 * updated to reflect the correct port. */
2646 out_port = FLOOD_PORT;
2649 /* Don't send packets out their input ports. */
2650 if (in_port == out_port) {
2656 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2664 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2665 struct ofpbuf *actions, tag_type *tags,
2666 uint16_t *nf_output_iface, void *br_)
2668 struct bridge *br = br_;
2670 COVERAGE_INC(bridge_process_flow);
2671 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2675 bridge_special_ofhook_cb(const struct flow *flow,
2676 const struct ofpbuf *packet, void *br_)
2678 struct iface *iface;
2679 struct bridge *br = br_;
2681 iface = iface_from_dp_ifidx(br, flow->in_port);
2683 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2684 if (iface && iface->port->lacp && packet) {
2685 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2687 lacp_process_pdu(iface->port->lacp, iface, pdu);
2697 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2698 const struct nlattr *actions,
2700 uint64_t n_bytes, void *br_)
2702 struct bridge *br = br_;
2703 const struct nlattr *a;
2704 struct port *in_port;
2709 /* Feed information from the active flows back into the learning table to
2710 * ensure that table is always in sync with what is actually flowing
2711 * through the datapath.
2713 * We test that 'tags' is nonzero to ensure that only flows that include an
2714 * OFPP_NORMAL action are used for learning. This works because
2715 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2716 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2717 update_learning_table(br, flow, vlan, in_port);
2720 /* Account for bond slave utilization. */
2721 if (!br->has_bonded_ports) {
2724 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2725 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2726 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2727 if (out_port && out_port->bond) {
2728 uint16_t vlan = (flow->vlan_tci
2729 ? vlan_tci_to_vid(flow->vlan_tci)
2731 bond_account(out_port->bond, flow, vlan, n_bytes);
2738 bridge_account_checkpoint_ofhook_cb(void *br_)
2740 struct bridge *br = br_;
2743 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2745 bond_rebalance(port->bond,
2746 ofproto_get_revalidate_set(br->ofproto));
2752 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2753 tag_type *tags, void *br_)
2755 struct bridge *br = br_;
2756 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2757 struct port *port = port_from_dp_ifidx(br, odp_port);
2762 } else if (list_is_short(&port->ifaces)) {
2765 struct iface *iface;
2767 /* Autopath does not support VLAN hashing. */
2768 iface = bond_choose_output_slave(port->bond, flow,
2769 OFP_VLAN_NONE, tags);
2770 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2773 return odp_port_to_ofp_port(ret);
2776 static struct ofhooks bridge_ofhooks = {
2777 bridge_normal_ofhook_cb,
2778 bridge_special_ofhook_cb,
2779 bridge_account_flow_ofhook_cb,
2780 bridge_account_checkpoint_ofhook_cb,
2781 bridge_autopath_ofhook_cb,
2784 /* Port functions. */
2787 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2789 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2790 struct iface *iface = iface_;
2791 uint8_t ea[ETH_ADDR_LEN];
2794 error = netdev_get_etheraddr(iface->netdev, ea);
2796 struct lacp_pdu *packet_pdu;
2797 struct ofpbuf packet;
2799 ofpbuf_init(&packet, 0);
2800 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2801 sizeof *packet_pdu);
2803 error = netdev_send(iface->netdev, &packet);
2805 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2806 "(%s)", iface->port->name, iface->name,
2809 ofpbuf_uninit(&packet);
2811 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2812 "%s (%s)", iface->port->name, iface->name,
2818 port_run(struct port *port)
2821 lacp_run(port->lacp, lacp_send_pdu_cb);
2825 struct iface *iface;
2827 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2828 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2829 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2832 bond_run(port->bond,
2833 ofproto_get_revalidate_set(port->bridge->ofproto),
2834 lacp_negotiated(port->lacp));
2835 if (bond_should_send_learning_packets(port->bond)) {
2836 port_send_learning_packets(port);
2842 port_wait(struct port *port)
2845 lacp_wait(port->lacp);
2849 bond_wait(port->bond);
2853 static struct port *
2854 port_create(struct bridge *br, const char *name)
2858 port = xzalloc(sizeof *port);
2861 port->trunks = NULL;
2862 port->name = xstrdup(name);
2863 list_init(&port->ifaces);
2865 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2867 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2874 get_port_other_config(const struct ovsrec_port *port, const char *key,
2875 const char *default_value)
2879 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2881 return value ? value : default_value;
2885 get_interface_other_config(const struct ovsrec_interface *iface,
2886 const char *key, const char *default_value)
2890 value = get_ovsrec_key_value(&iface->header_,
2891 &ovsrec_interface_col_other_config, key);
2892 return value ? value : default_value;
2896 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2898 struct iface *iface, *next;
2899 struct sset new_ifaces;
2902 /* Collect list of new interfaces. */
2903 sset_init(&new_ifaces);
2904 for (i = 0; i < cfg->n_interfaces; i++) {
2905 const char *name = cfg->interfaces[i]->name;
2906 sset_add(&new_ifaces, name);
2909 /* Get rid of deleted interfaces. */
2910 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2911 if (!sset_contains(&new_ifaces, iface->name)) {
2912 iface_destroy(iface);
2916 sset_destroy(&new_ifaces);
2919 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2920 * to revalidate every flow. */
2922 port_flush_macs(struct port *port)
2924 struct bridge *br = port->bridge;
2925 struct mac_learning *ml = br->ml;
2926 struct mac_entry *mac, *next_mac;
2929 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2930 if (mac->port.p == port) {
2931 mac_learning_expire(ml, mac);
2937 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2939 struct sset new_ifaces;
2940 bool need_flush = false;
2941 unsigned long *trunks;
2948 /* Add new interfaces and update 'cfg' member of existing ones. */
2949 sset_init(&new_ifaces);
2950 for (i = 0; i < cfg->n_interfaces; i++) {
2951 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2952 struct iface *iface;
2954 if (!sset_add(&new_ifaces, if_cfg->name)) {
2955 VLOG_WARN("port %s: %s specified twice as port interface",
2956 port->name, if_cfg->name);
2957 iface_set_ofport(if_cfg, -1);
2961 iface = iface_lookup(port->bridge, if_cfg->name);
2963 if (iface->port != port) {
2964 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2966 port->bridge->name, if_cfg->name, iface->port->name);
2969 iface->cfg = if_cfg;
2971 iface = iface_create(port, if_cfg);
2974 /* Determine interface type. The local port always has type
2975 * "internal". Other ports take their type from the database and
2976 * default to "system" if none is specified. */
2977 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2978 : if_cfg->type[0] ? if_cfg->type
2981 sset_destroy(&new_ifaces);
2986 if (list_is_short(&port->ifaces)) {
2988 if (vlan >= 0 && vlan <= 4095) {
2989 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2994 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2995 * they even work as-is. But they have not been tested. */
2996 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
3000 if (port->vlan != vlan) {
3005 /* Get trunked VLANs. */
3007 if (vlan < 0 && cfg->n_trunks) {
3010 trunks = bitmap_allocate(4096);
3012 for (i = 0; i < cfg->n_trunks; i++) {
3013 int trunk = cfg->trunks[i];
3015 bitmap_set1(trunks, trunk);
3021 VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
3022 port->name, cfg->n_trunks);
3024 if (n_errors == cfg->n_trunks) {
3025 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
3027 bitmap_free(trunks);
3030 } else if (vlan >= 0 && cfg->n_trunks) {
3031 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
3035 ? port->trunks != NULL
3036 : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
3039 bitmap_free(port->trunks);
3040 port->trunks = trunks;
3043 port_flush_macs(port);
3048 port_destroy(struct port *port)
3051 struct bridge *br = port->bridge;
3052 struct iface *iface, *next;
3055 for (i = 0; i < MAX_MIRRORS; i++) {
3056 struct mirror *m = br->mirrors[i];
3057 if (m && m->out_port == port) {
3062 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3063 iface_destroy(iface);
3066 hmap_remove(&br->ports, &port->hmap_node);
3068 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
3070 bond_destroy(port->bond);
3071 lacp_destroy(port->lacp);
3072 port_flush_macs(port);
3074 bitmap_free(port->trunks);
3080 static struct port *
3081 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3083 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
3084 return iface ? iface->port : NULL;
3087 static struct port *
3088 port_lookup(const struct bridge *br, const char *name)
3092 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3094 if (!strcmp(port->name, name)) {
3102 enable_lacp(struct port *port, bool *activep)
3104 if (!port->cfg->lacp) {
3105 /* XXX when LACP implementation has been sufficiently tested, enable by
3106 * default and make active on bonded ports. */
3108 } else if (!strcmp(port->cfg->lacp, "off")) {
3110 } else if (!strcmp(port->cfg->lacp, "active")) {
3113 } else if (!strcmp(port->cfg->lacp, "passive")) {
3117 VLOG_WARN("port %s: unknown LACP mode %s",
3118 port->name, port->cfg->lacp);
3124 iface_reconfigure_lacp(struct iface *iface)
3126 struct lacp_slave_settings s;
3127 int priority, portid;
3129 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
3130 priority = atoi(get_interface_other_config(iface->cfg,
3131 "lacp-port-priority", "0"));
3133 if (portid <= 0 || portid > UINT16_MAX) {
3134 portid = iface->dp_ifidx;
3137 if (priority <= 0 || priority > UINT16_MAX) {
3138 priority = UINT16_MAX;
3141 s.name = iface->name;
3143 s.priority = priority;
3144 lacp_slave_register(iface->port->lacp, iface, &s);
3148 port_reconfigure_lacp(struct port *port)
3150 static struct lacp_settings s;
3151 struct iface *iface;
3152 uint8_t sysid[ETH_ADDR_LEN];
3153 const char *sysid_str;
3154 const char *lacp_time;
3155 long long int custom_time;
3158 if (!enable_lacp(port, &s.active)) {
3159 lacp_destroy(port->lacp);
3164 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3165 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3166 memcpy(s.id, sysid, ETH_ADDR_LEN);
3168 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3171 s.name = port->name;
3173 /* Prefer bondable links if unspecified. */
3174 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3176 s.priority = (priority > 0 && priority <= UINT16_MAX
3178 : UINT16_MAX - !list_is_short(&port->ifaces));
3180 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3184 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3185 custom_time = atoi(lacp_time);
3186 if (!strcmp(lacp_time, "fast")) {
3187 s.lacp_time = LACP_TIME_FAST;
3188 } else if (!strcmp(lacp_time, "slow")) {
3189 s.lacp_time = LACP_TIME_SLOW;
3190 } else if (custom_time > 0) {
3191 s.lacp_time = LACP_TIME_CUSTOM;
3192 s.custom_time = custom_time;
3194 s.lacp_time = LACP_TIME_SLOW;
3198 port->lacp = lacp_create();
3201 lacp_configure(port->lacp, &s);
3203 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3204 iface_reconfigure_lacp(iface);
3209 port_reconfigure_bond(struct port *port)
3211 struct bond_settings s;
3212 const char *detect_s;
3213 struct iface *iface;
3215 if (list_is_short(&port->ifaces)) {
3216 /* Not a bonded port. */
3217 bond_destroy(port->bond);
3222 port->bridge->has_bonded_ports = true;
3224 s.name = port->name;
3226 if (port->cfg->bond_mode
3227 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3228 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3229 port->name, port->cfg->bond_mode,
3230 bond_mode_to_string(s.balance));
3233 s.detect = BLSM_CARRIER;
3234 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3235 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3236 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3238 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3241 s.miimon_interval = atoi(
3242 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3243 if (s.miimon_interval < 100) {
3244 s.miimon_interval = 100;
3247 s.up_delay = MAX(0, port->cfg->bond_updelay);
3248 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3249 s.rebalance_interval = atoi(
3250 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3251 if (s.rebalance_interval < 1000) {
3252 s.rebalance_interval = 1000;
3255 s.fake_iface = port->cfg->bond_fake_iface;
3258 port->bond = bond_create(&s);
3260 if (bond_reconfigure(port->bond, &s)) {
3261 bridge_flush(port->bridge);
3265 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3266 uint16_t stable_id = (port->lacp
3267 ? lacp_slave_get_port_id(port->lacp, iface)
3269 bond_slave_register(iface->port->bond, iface, stable_id,
3275 port_send_learning_packets(struct port *port)
3277 struct bridge *br = port->bridge;
3278 int error, n_packets, n_errors;
3279 struct mac_entry *e;
3281 error = n_packets = n_errors = 0;
3282 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3283 if (e->port.p != port) {
3284 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3294 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3295 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3296 "packets, last error was: %s",
3297 port->name, n_errors, n_packets, strerror(error));
3299 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3300 port->name, n_packets);
3304 /* Interface functions. */
3306 static struct iface *
3307 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3309 struct bridge *br = port->bridge;
3310 struct iface *iface;
3311 char *name = if_cfg->name;
3313 iface = xzalloc(sizeof *iface);
3315 iface->name = xstrdup(name);
3316 iface->dp_ifidx = -1;
3317 iface->tag = tag_create_random();
3318 iface->netdev = NULL;
3319 iface->cfg = if_cfg;
3321 shash_add_assert(&br->iface_by_name, iface->name, iface);
3323 list_push_back(&port->ifaces, &iface->port_elem);
3325 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3333 iface_destroy(struct iface *iface)
3336 struct port *port = iface->port;
3337 struct bridge *br = port->bridge;
3340 bond_slave_unregister(port->bond, iface);
3344 lacp_slave_unregister(port->lacp, iface);
3347 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3349 if (iface->dp_ifidx >= 0) {
3350 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3353 list_remove(&iface->port_elem);
3355 netdev_close(iface->netdev);
3360 bridge_flush(port->bridge);
3364 static struct iface *
3365 iface_lookup(const struct bridge *br, const char *name)
3367 return shash_find_data(&br->iface_by_name, name);
3370 static struct iface *
3371 iface_find(const char *name)
3373 const struct bridge *br;
3375 LIST_FOR_EACH (br, node, &all_bridges) {
3376 struct iface *iface = iface_lookup(br, name);
3385 static struct iface *
3386 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3388 struct iface *iface;
3390 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3391 hash_int(dp_ifidx, 0), &br->ifaces) {
3392 if (iface->dp_ifidx == dp_ifidx) {
3399 /* Set Ethernet address of 'iface', if one is specified in the configuration
3402 iface_set_mac(struct iface *iface)
3404 uint8_t ea[ETH_ADDR_LEN];
3406 if (!strcmp(iface->type, "internal")
3407 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3408 if (iface->dp_ifidx == ODPP_LOCAL) {
3409 VLOG_ERR("interface %s: ignoring mac in Interface record "
3410 "(use Bridge record to set local port's mac)",
3412 } else if (eth_addr_is_multicast(ea)) {
3413 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3416 int error = netdev_set_etheraddr(iface->netdev, ea);
3418 VLOG_ERR("interface %s: setting MAC failed (%s)",
3419 iface->name, strerror(error));
3425 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3427 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3429 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3430 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3434 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3436 * The value strings in '*shash' are taken directly from values[], not copied,
3437 * so the caller should not modify or free them. */
3439 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3440 struct shash *shash)
3445 for (i = 0; i < n; i++) {
3446 shash_add(shash, keys[i], values[i]);
3450 /* Creates 'keys' and 'values' arrays from 'shash'.
3452 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3453 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3454 * are populated with with strings taken directly from 'shash' and thus have
3455 * the same ownership of the key-value pairs in shash.
3458 shash_to_ovs_idl_map(struct shash *shash,
3459 char ***keys, char ***values, size_t *n)
3463 struct shash_node *sn;
3465 count = shash_count(shash);
3467 k = xmalloc(count * sizeof *k);
3468 v = xmalloc(count * sizeof *v);
3471 SHASH_FOR_EACH(sn, shash) {
3482 struct iface_delete_queues_cbdata {
3483 struct netdev *netdev;
3484 const struct ovsdb_datum *queues;
3488 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3490 union ovsdb_atom atom;
3492 atom.integer = target;
3493 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3497 iface_delete_queues(unsigned int queue_id,
3498 const struct shash *details OVS_UNUSED, void *cbdata_)
3500 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3502 if (!queue_ids_include(cbdata->queues, queue_id)) {
3503 netdev_delete_queue(cbdata->netdev, queue_id);
3508 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3510 if (!qos || qos->type[0] == '\0') {
3511 netdev_set_qos(iface->netdev, NULL, NULL);
3513 struct iface_delete_queues_cbdata cbdata;
3514 struct shash details;
3517 /* Configure top-level Qos for 'iface'. */
3518 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3519 qos->n_other_config, &details);
3520 netdev_set_qos(iface->netdev, qos->type, &details);
3521 shash_destroy(&details);
3523 /* Deconfigure queues that were deleted. */
3524 cbdata.netdev = iface->netdev;
3525 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3527 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3529 /* Configure queues for 'iface'. */
3530 for (i = 0; i < qos->n_queues; i++) {
3531 const struct ovsrec_queue *queue = qos->value_queues[i];
3532 unsigned int queue_id = qos->key_queues[i];
3534 shash_from_ovs_idl_map(queue->key_other_config,
3535 queue->value_other_config,
3536 queue->n_other_config, &details);
3537 netdev_set_queue(iface->netdev, queue_id, &details);
3538 shash_destroy(&details);
3544 iface_update_cfm(struct iface *iface)
3548 uint16_t *remote_mps;
3549 struct ovsrec_monitor *mon;
3550 uint8_t maid[CCM_MAID_LEN];
3552 mon = iface->cfg->monitor;
3555 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3559 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3560 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3564 cfm.mpid = mon->mpid;
3565 cfm.interval = mon->interval ? *mon->interval : 1000;
3567 memcpy(cfm.maid, maid, sizeof cfm.maid);
3569 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3570 for(i = 0; i < mon->n_remote_mps; i++) {
3571 remote_mps[i] = mon->remote_mps[i]->mpid;
3574 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3575 &cfm, remote_mps, mon->n_remote_mps);
3579 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3580 * how 'iface''s port is configured.
3582 * Returns true if 'iface' is up, false otherwise. */
3584 iface_get_carrier(const struct iface *iface)
3587 return netdev_get_carrier(iface->netdev);
3590 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3591 * instead of obtaining it from the database. */
3593 iface_is_synthetic(const struct iface *iface)
3595 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3598 /* Port mirroring. */
3600 static struct mirror *
3601 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3605 for (i = 0; i < MAX_MIRRORS; i++) {
3606 struct mirror *m = br->mirrors[i];
3607 if (m && uuid_equals(uuid, &m->uuid)) {
3615 mirror_reconfigure(struct bridge *br)
3617 unsigned long *rspan_vlans;
3621 /* Get rid of deleted mirrors. */
3622 for (i = 0; i < MAX_MIRRORS; i++) {
3623 struct mirror *m = br->mirrors[i];
3625 const struct ovsdb_datum *mc;
3626 union ovsdb_atom atom;
3628 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3629 atom.uuid = br->mirrors[i]->uuid;
3630 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3636 /* Add new mirrors and reconfigure existing ones. */
3637 for (i = 0; i < br->cfg->n_mirrors; i++) {
3638 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3639 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3641 mirror_reconfigure_one(m, cfg);
3643 mirror_create(br, cfg);
3647 /* Update port reserved status. */
3648 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3649 port->is_mirror_output_port = false;
3651 for (i = 0; i < MAX_MIRRORS; i++) {
3652 struct mirror *m = br->mirrors[i];
3653 if (m && m->out_port) {
3654 m->out_port->is_mirror_output_port = true;
3658 /* Update flooded vlans (for RSPAN). */
3660 if (br->cfg->n_flood_vlans) {
3661 rspan_vlans = bitmap_allocate(4096);
3663 for (i = 0; i < br->cfg->n_flood_vlans; i++) {
3664 int64_t vlan = br->cfg->flood_vlans[i];
3665 if (vlan >= 0 && vlan < 4096) {
3666 bitmap_set1(rspan_vlans, vlan);
3667 VLOG_INFO("bridge %s: disabling learning on vlan %"PRId64,
3670 VLOG_ERR("bridge %s: invalid value %"PRId64 "for flood VLAN",
3675 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3677 mac_learning_flush(br->ml);
3682 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3687 for (i = 0; ; i++) {
3688 if (i >= MAX_MIRRORS) {
3689 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3690 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3693 if (!br->mirrors[i]) {
3698 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3700 mac_learning_flush(br->ml);
3702 br->mirrors[i] = m = xzalloc(sizeof *m);
3703 m->uuid = cfg->header_.uuid;
3706 m->name = xstrdup(cfg->name);
3707 sset_init(&m->src_ports);
3708 sset_init(&m->dst_ports);
3714 mirror_reconfigure_one(m, cfg);
3718 mirror_destroy(struct mirror *m)
3721 struct bridge *br = m->bridge;
3724 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3725 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3726 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3729 sset_destroy(&m->src_ports);
3730 sset_destroy(&m->dst_ports);
3733 m->bridge->mirrors[m->idx] = NULL;
3738 mac_learning_flush(br->ml);
3743 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3748 for (i = 0; i < n_ports; i++) {
3749 const char *name = ports[i]->name;
3750 if (port_lookup(m->bridge, name)) {
3751 sset_add(names, name);
3753 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3754 "port %s", m->bridge->name, m->name, name);
3760 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3766 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3768 for (i = 0; i < cfg->n_select_vlan; i++) {
3769 int64_t vlan = cfg->select_vlan[i];
3770 if (vlan < 0 || vlan > 4095) {
3771 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3772 m->bridge->name, m->name, vlan);
3774 (*vlans)[n_vlans++] = vlan;
3781 vlan_is_mirrored(const struct mirror *m, int vlan)
3785 for (i = 0; i < m->n_vlans; i++) {
3786 if (m->vlans[i] == vlan) {
3794 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3796 struct sset src_ports, dst_ports;
3797 mirror_mask_t mirror_bit;
3798 struct port *out_port;
3805 if (strcmp(cfg->name, m->name)) {
3807 m->name = xstrdup(cfg->name);
3810 /* Get output port. */
3811 if (cfg->output_port) {
3812 out_port = port_lookup(m->bridge, cfg->output_port->name);
3814 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3815 m->bridge->name, m->name);
3821 if (cfg->output_vlan) {
3822 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3823 "output vlan; ignoring output vlan",
3824 m->bridge->name, m->name);
3826 } else if (cfg->output_vlan) {
3828 out_vlan = *cfg->output_vlan;
3830 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3831 m->bridge->name, m->name);
3836 sset_init(&src_ports);
3837 sset_init(&dst_ports);
3838 if (cfg->select_all) {
3839 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3840 sset_add(&src_ports, port->name);
3841 sset_add(&dst_ports, port->name);
3846 /* Get ports, and drop duplicates and ports that don't exist. */
3847 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3849 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3852 /* Get all the vlans, and drop duplicate and invalid vlans. */
3853 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3856 /* Update mirror data. */
3857 if (!sset_equals(&m->src_ports, &src_ports)
3858 || !sset_equals(&m->dst_ports, &dst_ports)
3859 || m->n_vlans != n_vlans
3860 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3861 || m->out_port != out_port
3862 || m->out_vlan != out_vlan) {
3863 bridge_flush(m->bridge);
3864 mac_learning_flush(m->bridge->ml);
3866 sset_swap(&m->src_ports, &src_ports);
3867 sset_swap(&m->dst_ports, &dst_ports);
3870 m->n_vlans = n_vlans;
3871 m->out_port = out_port;
3872 m->out_vlan = out_vlan;
3875 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3876 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3877 if (sset_contains(&m->src_ports, port->name)) {
3878 port->src_mirrors |= mirror_bit;
3880 port->src_mirrors &= ~mirror_bit;
3883 if (sset_contains(&m->dst_ports, port->name)) {
3884 port->dst_mirrors |= mirror_bit;
3886 port->dst_mirrors &= ~mirror_bit;
3891 sset_destroy(&src_ports);
3892 sset_destroy(&dst_ports);