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];
190 /* List of all bridges. */
191 static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
193 /* OVSDB IDL used to obtain configuration. */
194 static struct ovsdb_idl *idl;
196 /* Each time this timer expires, the bridge fetches systems and interface
197 * statistics and pushes them into the database. */
198 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
199 static long long int stats_timer = LLONG_MIN;
201 /* Stores the time after which rate limited statistics may be written to the
202 * database. Only updated when changes to the database require rate limiting.
204 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
205 static long long int db_limiter = LLONG_MIN;
207 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
208 static void bridge_destroy(struct bridge *);
209 static struct bridge *bridge_lookup(const char *name);
210 static unixctl_cb_func bridge_unixctl_dump_flows;
211 static unixctl_cb_func bridge_unixctl_reconnect;
212 static int bridge_run_one(struct bridge *);
213 static size_t bridge_get_controllers(const struct bridge *br,
214 struct ovsrec_controller ***controllersp);
215 static void bridge_reconfigure_one(struct bridge *);
216 static void bridge_reconfigure_remotes(struct bridge *,
217 const struct sockaddr_in *managers,
219 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
220 static void bridge_fetch_dp_ifaces(struct bridge *);
221 static void bridge_flush(struct bridge *);
222 static void bridge_pick_local_hw_addr(struct bridge *,
223 uint8_t ea[ETH_ADDR_LEN],
224 struct iface **hw_addr_iface);
225 static uint64_t bridge_pick_datapath_id(struct bridge *,
226 const uint8_t bridge_ea[ETH_ADDR_LEN],
227 struct iface *hw_addr_iface);
228 static uint64_t dpid_from_hash(const void *, size_t nbytes);
230 static unixctl_cb_func bridge_unixctl_fdb_show;
231 static unixctl_cb_func cfm_unixctl_show;
232 static unixctl_cb_func qos_unixctl_show;
234 static void port_run(struct port *);
235 static void port_wait(struct port *);
236 static struct port *port_create(struct bridge *, const char *name);
237 static void port_reconfigure(struct port *, const struct ovsrec_port *);
238 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
239 static void port_destroy(struct port *);
240 static struct port *port_lookup(const struct bridge *, const char *name);
241 static struct iface *port_get_an_iface(const struct port *);
242 static struct port *port_from_dp_ifidx(const struct bridge *,
244 static void port_reconfigure_lacp(struct port *);
245 static void port_reconfigure_bond(struct port *);
246 static void port_send_learning_packets(struct port *);
248 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
249 static void mirror_destroy(struct mirror *);
250 static void mirror_reconfigure(struct bridge *);
251 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
252 static bool vlan_is_mirrored(const struct mirror *, int vlan);
254 static struct iface *iface_create(struct port *port,
255 const struct ovsrec_interface *if_cfg);
256 static void iface_destroy(struct iface *);
257 static struct iface *iface_lookup(const struct bridge *, const char *name);
258 static struct iface *iface_find(const char *name);
259 static struct iface *iface_from_dp_ifidx(const struct bridge *,
261 static void iface_set_mac(struct iface *);
262 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
263 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
264 static void iface_update_cfm(struct iface *);
265 static bool iface_refresh_cfm_stats(struct iface *iface);
266 static bool iface_get_carrier(const struct iface *);
268 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
270 static void shash_to_ovs_idl_map(struct shash *,
271 char ***keys, char ***values, size_t *n);
273 /* Hooks into ofproto processing. */
274 static struct ofhooks bridge_ofhooks;
276 /* Public functions. */
278 /* Initializes the bridge module, configuring it to obtain its configuration
279 * from an OVSDB server accessed over 'remote', which should be a string in a
280 * form acceptable to ovsdb_idl_create(). */
282 bridge_init(const char *remote)
284 /* Create connection to database. */
285 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
287 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
288 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
289 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
290 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
291 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
292 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
293 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
295 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
296 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
298 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
299 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
305 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
309 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
311 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
312 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
313 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
314 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
316 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
318 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
320 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
322 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
324 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
326 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
332 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
333 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
334 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
336 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
338 /* Register unixctl commands. */
339 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
340 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
341 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
342 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
344 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
353 struct bridge *br, *next_br;
355 LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
358 ovsdb_idl_destroy(idl);
361 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
362 * but for which the ovs-vswitchd configuration 'cfg' is required. */
364 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
366 static bool already_configured_once;
367 struct sset bridge_names;
368 struct sset dpif_names, dpif_types;
372 /* Only do this once per ovs-vswitchd run. */
373 if (already_configured_once) {
376 already_configured_once = true;
378 stats_timer = time_msec() + STATS_INTERVAL;
380 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
381 sset_init(&bridge_names);
382 for (i = 0; i < cfg->n_bridges; i++) {
383 sset_add(&bridge_names, cfg->bridges[i]->name);
386 /* Iterate over all system dpifs and delete any of them that do not appear
388 sset_init(&dpif_names);
389 sset_init(&dpif_types);
390 dp_enumerate_types(&dpif_types);
391 SSET_FOR_EACH (type, &dpif_types) {
394 dp_enumerate_names(type, &dpif_names);
396 /* Delete each dpif whose name is not in 'bridge_names'. */
397 SSET_FOR_EACH (name, &dpif_names) {
398 if (!sset_contains(&bridge_names, name)) {
402 retval = dpif_open(name, type, &dpif);
410 sset_destroy(&bridge_names);
411 sset_destroy(&dpif_names);
412 sset_destroy(&dpif_types);
415 /* Callback for iterate_and_prune_ifaces(). */
417 check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
419 if (!iface->netdev) {
420 /* We already reported a related error, don't bother duplicating it. */
424 if (iface->dp_ifidx < 0) {
425 VLOG_ERR("%s interface not in %s, dropping",
426 iface->name, dpif_name(br->dpif));
430 VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
431 iface->name, iface->dp_ifidx);
435 /* Callback for iterate_and_prune_ifaces(). */
437 set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
438 void *aux OVS_UNUSED)
440 /* Set policing attributes. */
441 netdev_set_policing(iface->netdev,
442 iface->cfg->ingress_policing_rate,
443 iface->cfg->ingress_policing_burst);
445 /* Set MAC address of internal interfaces other than the local
447 if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) {
448 iface_set_mac(iface);
454 /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
455 * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
456 * deletes from 'br' any ports that no longer have any interfaces. */
458 iterate_and_prune_ifaces(struct bridge *br,
459 bool (*cb)(struct bridge *, struct iface *,
463 struct port *port, *next_port;
465 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
466 struct iface *iface, *next_iface;
468 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
469 if (!cb(br, iface, aux)) {
470 iface_set_ofport(iface->cfg, -1);
471 iface_destroy(iface);
475 if (list_is_empty(&port->ifaces)) {
476 VLOG_WARN("%s port has no interfaces, dropping", port->name);
482 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
483 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
484 * responsible for freeing '*managersp' (with free()).
486 * You may be asking yourself "why does ovs-vswitchd care?", because
487 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
488 * should not be and in fact is not directly involved in that. But
489 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
490 * it has to tell in-band control where the managers are to enable that.
491 * (Thus, only managers connected in-band are collected.)
494 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
495 struct sockaddr_in **managersp, size_t *n_managersp)
497 struct sockaddr_in *managers = NULL;
498 size_t n_managers = 0;
502 /* Collect all of the potential targets from the "targets" columns of the
503 * rows pointed to by "manager_options", excluding any that are
506 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
507 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
509 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
510 sset_find_and_delete(&targets, m->target);
512 sset_add(&targets, m->target);
516 /* Now extract the targets' IP addresses. */
517 if (!sset_is_empty(&targets)) {
520 managers = xmalloc(sset_count(&targets) * sizeof *managers);
521 SSET_FOR_EACH (target, &targets) {
522 struct sockaddr_in *sin = &managers[n_managers];
524 if ((!strncmp(target, "tcp:", 4)
525 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
526 (!strncmp(target, "ssl:", 4)
527 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
532 sset_destroy(&targets);
534 *managersp = managers;
535 *n_managersp = n_managers;
539 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
541 struct shash old_br, new_br;
542 struct shash_node *node;
543 struct bridge *br, *next;
544 struct sockaddr_in *managers;
547 int sflow_bridge_number;
549 COVERAGE_INC(bridge_reconfigure);
551 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
553 /* Collect old and new bridges. */
556 LIST_FOR_EACH (br, node, &all_bridges) {
557 shash_add(&old_br, br->name, br);
559 for (i = 0; i < ovs_cfg->n_bridges; i++) {
560 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
561 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
562 VLOG_WARN("more than one bridge named %s", br_cfg->name);
566 /* Get rid of deleted bridges and add new bridges. */
567 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
568 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
575 SHASH_FOR_EACH (node, &new_br) {
576 const char *br_name = node->name;
577 const struct ovsrec_bridge *br_cfg = node->data;
578 br = shash_find_data(&old_br, br_name);
580 /* If the bridge datapath type has changed, we need to tear it
581 * down and recreate. */
582 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
584 bridge_create(br_cfg);
587 bridge_create(br_cfg);
590 shash_destroy(&old_br);
591 shash_destroy(&new_br);
593 /* Reconfigure all bridges. */
594 LIST_FOR_EACH (br, node, &all_bridges) {
595 bridge_reconfigure_one(br);
598 /* Add and delete ports on all datapaths.
600 * The kernel will reject any attempt to add a given port to a datapath if
601 * that port already belongs to a different datapath, so we must do all
602 * port deletions before any port additions. */
603 LIST_FOR_EACH (br, node, &all_bridges) {
604 struct dpif_port_dump dump;
605 struct shash want_ifaces;
606 struct dpif_port dpif_port;
608 bridge_get_all_ifaces(br, &want_ifaces);
609 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
610 if (!shash_find(&want_ifaces, dpif_port.name)
611 && strcmp(dpif_port.name, br->name)) {
612 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
614 VLOG_WARN("failed to remove %s interface from %s: %s",
615 dpif_port.name, dpif_name(br->dpif),
620 shash_destroy(&want_ifaces);
622 LIST_FOR_EACH (br, node, &all_bridges) {
623 struct shash cur_ifaces, want_ifaces;
624 struct dpif_port_dump dump;
625 struct dpif_port dpif_port;
627 /* Get the set of interfaces currently in this datapath. */
628 shash_init(&cur_ifaces);
629 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
630 struct dpif_port *port_info = xmalloc(sizeof *port_info);
631 dpif_port_clone(port_info, &dpif_port);
632 shash_add(&cur_ifaces, dpif_port.name, port_info);
635 /* Get the set of interfaces we want on this datapath. */
636 bridge_get_all_ifaces(br, &want_ifaces);
638 hmap_clear(&br->ifaces);
639 SHASH_FOR_EACH (node, &want_ifaces) {
640 const char *if_name = node->name;
641 struct iface *iface = node->data;
642 struct dpif_port *dpif_port;
646 type = iface ? iface->type : "internal";
647 dpif_port = shash_find_data(&cur_ifaces, if_name);
649 /* If we have a port or a netdev already, and it's not the type we
650 * want, then delete the port (if any) and close the netdev (if
652 if ((dpif_port && strcmp(dpif_port->type, type))
653 || (iface && iface->netdev
654 && strcmp(type, netdev_get_type(iface->netdev)))) {
656 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
663 netdev_close(iface->netdev);
664 iface->netdev = NULL;
668 /* If the port doesn't exist or we don't have the netdev open,
669 * we need to do more work. */
670 if (!dpif_port || (iface && !iface->netdev)) {
671 struct netdev_options options;
672 struct netdev *netdev;
675 /* First open the network device. */
676 options.name = if_name;
678 options.args = &args;
679 options.ethertype = NETDEV_ETH_TYPE_NONE;
683 shash_from_ovs_idl_map(iface->cfg->key_options,
684 iface->cfg->value_options,
685 iface->cfg->n_options, &args);
687 error = netdev_open(&options, &netdev);
688 shash_destroy(&args);
691 VLOG_WARN("could not open network device %s (%s)",
692 if_name, strerror(error));
696 /* Then add the port if we haven't already. */
698 error = dpif_port_add(br->dpif, netdev, NULL);
700 netdev_close(netdev);
701 if (error == EFBIG) {
702 VLOG_ERR("ran out of valid port numbers on %s",
703 dpif_name(br->dpif));
706 VLOG_WARN("failed to add %s interface to %s: %s",
707 if_name, dpif_name(br->dpif),
714 /* Update 'iface'. */
716 iface->netdev = netdev;
718 } else if (iface && iface->netdev) {
722 shash_from_ovs_idl_map(iface->cfg->key_options,
723 iface->cfg->value_options,
724 iface->cfg->n_options, &args);
725 netdev_set_config(iface->netdev, &args);
726 shash_destroy(&args);
729 shash_destroy(&want_ifaces);
731 SHASH_FOR_EACH (node, &cur_ifaces) {
732 struct dpif_port *port_info = node->data;
733 dpif_port_destroy(port_info);
736 shash_destroy(&cur_ifaces);
738 sflow_bridge_number = 0;
739 LIST_FOR_EACH (br, node, &all_bridges) {
740 uint8_t ea[ETH_ADDR_LEN];
742 struct iface *local_iface;
743 struct iface *hw_addr_iface;
746 bridge_fetch_dp_ifaces(br);
748 /* Delete interfaces that cannot be opened.
750 * From this point forward we are guaranteed that every "struct iface"
751 * has nonnull 'netdev' and correct 'dp_ifidx'. */
752 iterate_and_prune_ifaces(br, check_iface, NULL);
754 /* Pick local port hardware address, datapath ID. */
755 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
756 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
758 int error = netdev_set_etheraddr(local_iface->netdev, ea);
760 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
761 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
762 "Ethernet address: %s",
763 br->name, strerror(error));
766 memcpy(br->ea, ea, ETH_ADDR_LEN);
768 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
769 ofproto_set_datapath_id(br->ofproto, dpid);
771 dpid_string = xasprintf("%016"PRIx64, dpid);
772 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
775 /* Set NetFlow configuration on this bridge. */
776 if (br->cfg->netflow) {
777 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
778 struct netflow_options opts;
780 memset(&opts, 0, sizeof opts);
782 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
783 if (nf_cfg->engine_type) {
784 opts.engine_type = *nf_cfg->engine_type;
786 if (nf_cfg->engine_id) {
787 opts.engine_id = *nf_cfg->engine_id;
790 opts.active_timeout = nf_cfg->active_timeout;
791 if (!opts.active_timeout) {
792 opts.active_timeout = -1;
793 } else if (opts.active_timeout < 0) {
794 VLOG_WARN("bridge %s: active timeout interval set to negative "
795 "value, using default instead (%d seconds)", br->name,
796 NF_ACTIVE_TIMEOUT_DEFAULT);
797 opts.active_timeout = -1;
800 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
801 if (opts.add_id_to_iface) {
802 if (opts.engine_id > 0x7f) {
803 VLOG_WARN("bridge %s: netflow port mangling may conflict "
804 "with another vswitch, choose an engine id less "
805 "than 128", br->name);
807 if (hmap_count(&br->ports) > 508) {
808 VLOG_WARN("bridge %s: netflow port mangling will conflict "
809 "with another port when more than 508 ports are "
814 sset_init(&opts.collectors);
815 sset_add_array(&opts.collectors,
816 nf_cfg->targets, nf_cfg->n_targets);
817 if (ofproto_set_netflow(br->ofproto, &opts)) {
818 VLOG_ERR("bridge %s: problem setting netflow collectors",
821 sset_destroy(&opts.collectors);
823 ofproto_set_netflow(br->ofproto, NULL);
826 /* Set sFlow configuration on this bridge. */
827 if (br->cfg->sflow) {
828 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
829 struct ovsrec_controller **controllers;
830 struct ofproto_sflow_options oso;
831 size_t n_controllers;
833 memset(&oso, 0, sizeof oso);
835 sset_init(&oso.targets);
836 sset_add_array(&oso.targets,
837 sflow_cfg->targets, sflow_cfg->n_targets);
839 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
840 if (sflow_cfg->sampling) {
841 oso.sampling_rate = *sflow_cfg->sampling;
844 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
845 if (sflow_cfg->polling) {
846 oso.polling_interval = *sflow_cfg->polling;
849 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
850 if (sflow_cfg->header) {
851 oso.header_len = *sflow_cfg->header;
854 oso.sub_id = sflow_bridge_number++;
855 oso.agent_device = sflow_cfg->agent;
857 oso.control_ip = NULL;
858 n_controllers = bridge_get_controllers(br, &controllers);
859 for (i = 0; i < n_controllers; i++) {
860 if (controllers[i]->local_ip) {
861 oso.control_ip = controllers[i]->local_ip;
865 ofproto_set_sflow(br->ofproto, &oso);
867 sset_destroy(&oso.targets);
869 ofproto_set_sflow(br->ofproto, NULL);
872 /* Update the controller and related settings. It would be more
873 * straightforward to call this from bridge_reconfigure_one(), but we
874 * can't do it there for two reasons. First, and most importantly, at
875 * that point we don't know the dp_ifidx of any interfaces that have
876 * been added to the bridge (because we haven't actually added them to
877 * the datapath). Second, at that point we haven't set the datapath ID
878 * yet; when a controller is configured, resetting the datapath ID will
879 * immediately disconnect from the controller, so it's better to set
880 * the datapath ID before the controller. */
881 bridge_reconfigure_remotes(br, managers, n_managers);
883 LIST_FOR_EACH (br, node, &all_bridges) {
886 br->has_bonded_ports = false;
887 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
890 port_reconfigure_lacp(port);
891 port_reconfigure_bond(port);
893 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
894 iface_update_qos(iface, port->cfg->qos);
898 LIST_FOR_EACH (br, node, &all_bridges) {
899 iterate_and_prune_ifaces(br, set_iface_properties, NULL);
902 /* Some reconfiguration operations require the bridge to have been run at
904 LIST_FOR_EACH (br, node, &all_bridges) {
909 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
910 iface_update_cfm(iface);
916 /* ovs-vswitchd has completed initialization, so allow the process that
917 * forked us to exit successfully. */
918 daemonize_complete();
922 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
923 const struct ovsdb_idl_column *column,
926 const struct ovsdb_datum *datum;
927 union ovsdb_atom atom;
930 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
931 atom.string = (char *) key;
932 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
933 return idx == UINT_MAX ? NULL : datum->values[idx].string;
937 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
939 return get_ovsrec_key_value(&br_cfg->header_,
940 &ovsrec_bridge_col_other_config, key);
944 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
945 struct iface **hw_addr_iface)
951 *hw_addr_iface = NULL;
953 /* Did the user request a particular MAC? */
954 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
955 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
956 if (eth_addr_is_multicast(ea)) {
957 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
958 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
959 } else if (eth_addr_is_zero(ea)) {
960 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
966 /* Otherwise choose the minimum non-local MAC address among all of the
968 memset(ea, 0xff, ETH_ADDR_LEN);
969 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
970 uint8_t iface_ea[ETH_ADDR_LEN];
971 struct iface *candidate;
974 /* Mirror output ports don't participate. */
975 if (port->is_mirror_output_port) {
979 /* Choose the MAC address to represent the port. */
981 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
982 /* Find the interface with this Ethernet address (if any) so that
983 * we can provide the correct devname to the caller. */
984 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
985 uint8_t candidate_ea[ETH_ADDR_LEN];
986 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
987 && eth_addr_equals(iface_ea, candidate_ea)) {
992 /* Choose the interface whose MAC address will represent the port.
993 * The Linux kernel bonding code always chooses the MAC address of
994 * the first slave added to a bond, and the Fedora networking
995 * scripts always add slaves to a bond in alphabetical order, so
996 * for compatibility we choose the interface with the name that is
997 * first in alphabetical order. */
998 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
999 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1004 /* The local port doesn't count (since we're trying to choose its
1005 * MAC address anyway). */
1006 if (iface->dp_ifidx == ODPP_LOCAL) {
1011 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1013 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1014 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
1015 iface->name, strerror(error));
1020 /* Compare against our current choice. */
1021 if (!eth_addr_is_multicast(iface_ea) &&
1022 !eth_addr_is_local(iface_ea) &&
1023 !eth_addr_is_reserved(iface_ea) &&
1024 !eth_addr_is_zero(iface_ea) &&
1025 eth_addr_compare_3way(iface_ea, ea) < 0)
1027 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1028 *hw_addr_iface = iface;
1031 if (eth_addr_is_multicast(ea)) {
1032 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1033 *hw_addr_iface = NULL;
1034 VLOG_WARN("bridge %s: using default bridge Ethernet "
1035 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1037 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1038 br->name, ETH_ADDR_ARGS(ea));
1042 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1043 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1044 * an interface on 'br', then that interface must be passed in as
1045 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1046 * 'hw_addr_iface' must be passed in as a null pointer. */
1048 bridge_pick_datapath_id(struct bridge *br,
1049 const uint8_t bridge_ea[ETH_ADDR_LEN],
1050 struct iface *hw_addr_iface)
1053 * The procedure for choosing a bridge MAC address will, in the most
1054 * ordinary case, also choose a unique MAC that we can use as a datapath
1055 * ID. In some special cases, though, multiple bridges will end up with
1056 * the same MAC address. This is OK for the bridges, but it will confuse
1057 * the OpenFlow controller, because each datapath needs a unique datapath
1060 * Datapath IDs must be unique. It is also very desirable that they be
1061 * stable from one run to the next, so that policy set on a datapath
1064 const char *datapath_id;
1067 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1068 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1072 if (hw_addr_iface) {
1074 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1076 * A bridge whose MAC address is taken from a VLAN network device
1077 * (that is, a network device created with vconfig(8) or similar
1078 * tool) will have the same MAC address as a bridge on the VLAN
1079 * device's physical network device.
1081 * Handle this case by hashing the physical network device MAC
1082 * along with the VLAN identifier.
1084 uint8_t buf[ETH_ADDR_LEN + 2];
1085 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1086 buf[ETH_ADDR_LEN] = vlan >> 8;
1087 buf[ETH_ADDR_LEN + 1] = vlan;
1088 return dpid_from_hash(buf, sizeof buf);
1091 * Assume that this bridge's MAC address is unique, since it
1092 * doesn't fit any of the cases we handle specially.
1097 * A purely internal bridge, that is, one that has no non-virtual
1098 * network devices on it at all, is more difficult because it has no
1099 * natural unique identifier at all.
1101 * When the host is a XenServer, we handle this case by hashing the
1102 * host's UUID with the name of the bridge. Names of bridges are
1103 * persistent across XenServer reboots, although they can be reused if
1104 * an internal network is destroyed and then a new one is later
1105 * created, so this is fairly effective.
1107 * When the host is not a XenServer, we punt by using a random MAC
1108 * address on each run.
1110 const char *host_uuid = xenserver_get_host_uuid();
1112 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1113 dpid = dpid_from_hash(combined, strlen(combined));
1119 return eth_addr_to_uint64(bridge_ea);
1123 dpid_from_hash(const void *data, size_t n)
1125 uint8_t hash[SHA1_DIGEST_SIZE];
1127 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1128 sha1_bytes(data, n, hash);
1129 eth_addr_mark_random(hash);
1130 return eth_addr_to_uint64(hash);
1134 iface_refresh_status(struct iface *iface)
1138 enum netdev_flags flags;
1147 if (!netdev_get_status(iface->netdev, &sh)) {
1149 char **keys, **values;
1151 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1152 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1157 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1160 shash_destroy_free_data(&sh);
1162 error = netdev_get_flags(iface->netdev, &flags);
1164 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1167 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1170 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1172 ovsrec_interface_set_duplex(iface->cfg,
1173 netdev_features_is_full_duplex(current)
1175 /* warning: uint64_t -> int64_t conversion */
1176 bps = netdev_features_to_bps(current);
1177 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1180 ovsrec_interface_set_duplex(iface->cfg, NULL);
1181 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1185 ovsrec_interface_set_link_state(iface->cfg,
1186 iface_get_carrier(iface) ? "up" : "down");
1188 error = netdev_get_mtu(iface->netdev, &mtu);
1189 if (!error && mtu != INT_MAX) {
1191 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1194 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1198 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1199 * changed, false otherwise. */
1201 iface_refresh_cfm_stats(struct iface *iface)
1203 const struct ovsrec_monitor *mon;
1204 const struct cfm *cfm;
1205 bool changed = false;
1208 mon = iface->cfg->monitor;
1209 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1215 for (i = 0; i < mon->n_remote_mps; i++) {
1216 const struct ovsrec_maintenance_point *mp;
1217 const struct remote_mp *rmp;
1219 mp = mon->remote_mps[i];
1220 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1222 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1223 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1228 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1229 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1237 iface_refresh_stats(struct iface *iface)
1243 static const struct iface_stat iface_stats[] = {
1244 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1245 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1246 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1247 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1248 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1249 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1250 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1251 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1252 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1253 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1254 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1255 { "collisions", offsetof(struct netdev_stats, collisions) },
1257 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1258 const struct iface_stat *s;
1260 char *keys[N_STATS];
1261 int64_t values[N_STATS];
1264 struct netdev_stats stats;
1266 /* Intentionally ignore return value, since errors will set 'stats' to
1267 * all-1s, and we will deal with that correctly below. */
1268 netdev_get_stats(iface->netdev, &stats);
1271 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1272 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1273 if (value != UINT64_MAX) {
1280 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1284 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1286 struct ovsdb_datum datum;
1290 get_system_stats(&stats);
1292 ovsdb_datum_from_shash(&datum, &stats);
1293 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1297 static inline const char *
1298 nx_role_to_str(enum nx_role role)
1303 case NX_ROLE_MASTER:
1308 return "*** INVALID ROLE ***";
1313 bridge_refresh_controller_status(const struct bridge *br)
1316 const struct ovsrec_controller *cfg;
1318 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1320 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1321 struct ofproto_controller_info *cinfo =
1322 shash_find_data(&info, cfg->target);
1325 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1326 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1327 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1328 (char **) cinfo->pairs.values,
1331 ovsrec_controller_set_is_connected(cfg, false);
1332 ovsrec_controller_set_role(cfg, NULL);
1333 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1337 ofproto_free_ofproto_controller_info(&info);
1343 const struct ovsrec_open_vswitch *cfg;
1345 bool datapath_destroyed;
1346 bool database_changed;
1349 /* Let each bridge do the work that it needs to do. */
1350 datapath_destroyed = false;
1351 LIST_FOR_EACH (br, node, &all_bridges) {
1352 int error = bridge_run_one(br);
1354 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1355 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1356 "forcing reconfiguration", br->name);
1357 datapath_destroyed = true;
1361 /* (Re)configure if necessary. */
1362 database_changed = ovsdb_idl_run(idl);
1363 cfg = ovsrec_open_vswitch_first(idl);
1365 /* Re-configure SSL. We do this on every trip through the main loop,
1366 * instead of just when the database changes, because the contents of the
1367 * key and certificate files can change without the database changing.
1369 * We do this before bridge_reconfigure() because that function might
1370 * initiate SSL connections and thus requires SSL to be configured. */
1371 if (cfg && cfg->ssl) {
1372 const struct ovsrec_ssl *ssl = cfg->ssl;
1374 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1375 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1378 if (database_changed || datapath_destroyed) {
1380 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1382 bridge_configure_once(cfg);
1383 bridge_reconfigure(cfg);
1385 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1386 ovsdb_idl_txn_commit(txn);
1387 ovsdb_idl_txn_destroy(txn); /* XXX */
1389 /* We still need to reconfigure to avoid dangling pointers to
1390 * now-destroyed ovsrec structures inside bridge data. */
1391 static const struct ovsrec_open_vswitch null_cfg;
1393 bridge_reconfigure(&null_cfg);
1397 /* Refresh system and interface stats if necessary. */
1398 if (time_msec() >= stats_timer) {
1400 struct ovsdb_idl_txn *txn;
1402 txn = ovsdb_idl_txn_create(idl);
1403 LIST_FOR_EACH (br, node, &all_bridges) {
1406 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1407 struct iface *iface;
1409 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1410 iface_refresh_stats(iface);
1411 iface_refresh_status(iface);
1414 bridge_refresh_controller_status(br);
1416 refresh_system_stats(cfg);
1417 ovsdb_idl_txn_commit(txn);
1418 ovsdb_idl_txn_destroy(txn); /* XXX */
1421 stats_timer = time_msec() + STATS_INTERVAL;
1424 if (time_msec() >= db_limiter) {
1425 struct ovsdb_idl_txn *txn;
1426 bool changed = false;
1428 txn = ovsdb_idl_txn_create(idl);
1429 LIST_FOR_EACH (br, node, &all_bridges) {
1432 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1433 struct iface *iface;
1435 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1436 changed = iface_refresh_cfm_stats(iface) || changed;
1442 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1445 ovsdb_idl_txn_commit(txn);
1446 ovsdb_idl_txn_destroy(txn);
1455 LIST_FOR_EACH (br, node, &all_bridges) {
1458 ofproto_wait(br->ofproto);
1459 mac_learning_wait(br->ml);
1460 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1464 ovsdb_idl_wait(idl);
1465 poll_timer_wait_until(stats_timer);
1467 if (db_limiter > time_msec()) {
1468 poll_timer_wait_until(db_limiter);
1472 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1473 * configuration changes. */
1475 bridge_flush(struct bridge *br)
1477 COVERAGE_INC(bridge_flush);
1481 /* Bridge unixctl user interface functions. */
1483 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1484 const char *args, void *aux OVS_UNUSED)
1486 struct ds ds = DS_EMPTY_INITIALIZER;
1487 const struct bridge *br;
1488 const struct mac_entry *e;
1490 br = bridge_lookup(args);
1492 unixctl_command_reply(conn, 501, "no such bridge");
1496 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1497 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1498 struct port *port = e->port.p;
1499 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1500 port_get_an_iface(port)->dp_ifidx,
1501 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1503 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1507 /* CFM unixctl user interface functions. */
1509 cfm_unixctl_show(struct unixctl_conn *conn,
1510 const char *args, void *aux OVS_UNUSED)
1512 struct ds ds = DS_EMPTY_INITIALIZER;
1513 struct iface *iface;
1514 const struct cfm *cfm;
1516 iface = iface_find(args);
1518 unixctl_command_reply(conn, 501, "no such interface");
1522 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1525 unixctl_command_reply(conn, 501, "CFM not enabled");
1529 cfm_dump_ds(cfm, &ds);
1530 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1534 /* QoS unixctl user interface functions. */
1536 struct qos_unixctl_show_cbdata {
1538 struct iface *iface;
1542 qos_unixctl_show_cb(unsigned int queue_id,
1543 const struct shash *details,
1546 struct qos_unixctl_show_cbdata *data = aux;
1547 struct ds *ds = data->ds;
1548 struct iface *iface = data->iface;
1549 struct netdev_queue_stats stats;
1550 struct shash_node *node;
1553 ds_put_cstr(ds, "\n");
1555 ds_put_format(ds, "Queue %u:\n", queue_id);
1557 ds_put_cstr(ds, "Default:\n");
1560 SHASH_FOR_EACH (node, details) {
1561 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1564 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1566 if (stats.tx_packets != UINT64_MAX) {
1567 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1570 if (stats.tx_bytes != UINT64_MAX) {
1571 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1574 if (stats.tx_errors != UINT64_MAX) {
1575 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1578 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1579 queue_id, strerror(error));
1584 qos_unixctl_show(struct unixctl_conn *conn,
1585 const char *args, void *aux OVS_UNUSED)
1587 struct ds ds = DS_EMPTY_INITIALIZER;
1588 struct shash sh = SHASH_INITIALIZER(&sh);
1589 struct iface *iface;
1591 struct shash_node *node;
1592 struct qos_unixctl_show_cbdata data;
1595 iface = iface_find(args);
1597 unixctl_command_reply(conn, 501, "no such interface");
1601 netdev_get_qos(iface->netdev, &type, &sh);
1603 if (*type != '\0') {
1604 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1606 SHASH_FOR_EACH (node, &sh) {
1607 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1612 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1615 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1617 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1619 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1620 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1623 shash_destroy_free_data(&sh);
1627 /* Bridge reconfiguration functions. */
1628 static struct bridge *
1629 bridge_create(const struct ovsrec_bridge *br_cfg)
1634 assert(!bridge_lookup(br_cfg->name));
1635 br = xzalloc(sizeof *br);
1637 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1644 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1647 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1649 dpif_delete(br->dpif);
1650 dpif_close(br->dpif);
1655 br->name = xstrdup(br_cfg->name);
1657 br->ml = mac_learning_create();
1658 eth_addr_nicira_random(br->default_ea);
1660 hmap_init(&br->ports);
1661 hmap_init(&br->ifaces);
1662 shash_init(&br->iface_by_name);
1666 list_push_back(&all_bridges, &br->node);
1668 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1674 bridge_destroy(struct bridge *br)
1677 struct port *port, *next;
1680 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1683 list_remove(&br->node);
1684 ofproto_destroy(br->ofproto);
1685 error = dpif_delete(br->dpif);
1686 if (error && error != ENOENT) {
1687 VLOG_ERR("failed to delete %s: %s",
1688 dpif_name(br->dpif), strerror(error));
1690 dpif_close(br->dpif);
1691 mac_learning_destroy(br->ml);
1692 hmap_destroy(&br->ifaces);
1693 hmap_destroy(&br->ports);
1694 shash_destroy(&br->iface_by_name);
1700 static struct bridge *
1701 bridge_lookup(const char *name)
1705 LIST_FOR_EACH (br, node, &all_bridges) {
1706 if (!strcmp(br->name, name)) {
1713 /* Handle requests for a listing of all flows known by the OpenFlow
1714 * stack, including those normally hidden. */
1716 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1717 const char *args, void *aux OVS_UNUSED)
1722 br = bridge_lookup(args);
1724 unixctl_command_reply(conn, 501, "Unknown bridge");
1729 ofproto_get_all_flows(br->ofproto, &results);
1731 unixctl_command_reply(conn, 200, ds_cstr(&results));
1732 ds_destroy(&results);
1735 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1736 * connections and reconnect. If BRIDGE is not specified, then all bridges
1737 * drop their controller connections and reconnect. */
1739 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1740 const char *args, void *aux OVS_UNUSED)
1743 if (args[0] != '\0') {
1744 br = bridge_lookup(args);
1746 unixctl_command_reply(conn, 501, "Unknown bridge");
1749 ofproto_reconnect_controllers(br->ofproto);
1751 LIST_FOR_EACH (br, node, &all_bridges) {
1752 ofproto_reconnect_controllers(br->ofproto);
1755 unixctl_command_reply(conn, 200, NULL);
1759 bridge_run_one(struct bridge *br)
1764 error = ofproto_run1(br->ofproto);
1769 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1771 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1775 error = ofproto_run2(br->ofproto, br->flush);
1782 bridge_get_controllers(const struct bridge *br,
1783 struct ovsrec_controller ***controllersp)
1785 struct ovsrec_controller **controllers;
1786 size_t n_controllers;
1788 controllers = br->cfg->controller;
1789 n_controllers = br->cfg->n_controller;
1791 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1797 *controllersp = controllers;
1799 return n_controllers;
1803 bridge_reconfigure_one(struct bridge *br)
1805 enum ofproto_fail_mode fail_mode;
1806 struct port *port, *next;
1807 struct shash_node *node;
1808 struct shash new_ports;
1811 /* Collect new ports. */
1812 shash_init(&new_ports);
1813 for (i = 0; i < br->cfg->n_ports; i++) {
1814 const char *name = br->cfg->ports[i]->name;
1815 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1816 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1821 /* If we have a controller, then we need a local port. Complain if the
1822 * user didn't specify one.
1824 * XXX perhaps we should synthesize a port ourselves in this case. */
1825 if (bridge_get_controllers(br, NULL)) {
1826 char local_name[IF_NAMESIZE];
1829 error = dpif_port_get_name(br->dpif, ODPP_LOCAL,
1830 local_name, sizeof local_name);
1831 if (!error && !shash_find(&new_ports, local_name)) {
1832 VLOG_WARN("bridge %s: controller specified but no local port "
1833 "(port named %s) defined",
1834 br->name, local_name);
1838 /* Get rid of deleted ports.
1839 * Get rid of deleted interfaces on ports that still exist. */
1840 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1841 const struct ovsrec_port *port_cfg;
1843 port_cfg = shash_find_data(&new_ports, port->name);
1847 port_del_ifaces(port, port_cfg);
1851 /* Create new ports.
1852 * Add new interfaces to existing ports.
1853 * Reconfigure existing ports. */
1854 SHASH_FOR_EACH (node, &new_ports) {
1855 struct port *port = port_lookup(br, node->name);
1857 port = port_create(br, node->name);
1860 port_reconfigure(port, node->data);
1861 if (list_is_empty(&port->ifaces)) {
1862 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1863 br->name, port->name);
1867 shash_destroy(&new_ports);
1869 /* Set the fail-mode */
1870 fail_mode = !br->cfg->fail_mode
1871 || !strcmp(br->cfg->fail_mode, "standalone")
1872 ? OFPROTO_FAIL_STANDALONE
1873 : OFPROTO_FAIL_SECURE;
1874 if (ofproto_get_fail_mode(br->ofproto) != fail_mode
1875 && !ofproto_has_primary_controller(br->ofproto)) {
1876 ofproto_flush_flows(br->ofproto);
1878 ofproto_set_fail_mode(br->ofproto, fail_mode);
1880 /* Delete all flows if we're switching from connected to standalone or vice
1881 * versa. (XXX Should we delete all flows if we are switching from one
1882 * controller to another?) */
1884 /* Configure OpenFlow controller connection snooping. */
1885 if (!ofproto_has_snoops(br->ofproto)) {
1889 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1890 ovs_rundir(), br->name));
1891 ofproto_set_snoops(br->ofproto, &snoops);
1892 sset_destroy(&snoops);
1895 mirror_reconfigure(br);
1898 /* Initializes 'oc' appropriately as a management service controller for
1901 * The caller must free oc->target when it is no longer needed. */
1903 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1904 struct ofproto_controller *oc)
1906 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1907 oc->max_backoff = 0;
1908 oc->probe_interval = 60;
1909 oc->band = OFPROTO_OUT_OF_BAND;
1911 oc->burst_limit = 0;
1914 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1916 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1917 struct ofproto_controller *oc)
1919 oc->target = c->target;
1920 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1921 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1922 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1923 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1924 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1925 oc->burst_limit = (c->controller_burst_limit
1926 ? *c->controller_burst_limit : 0);
1929 /* Configures the IP stack for 'br''s local interface properly according to the
1930 * configuration in 'c'. */
1932 bridge_configure_local_iface_netdev(struct bridge *br,
1933 struct ovsrec_controller *c)
1935 struct netdev *netdev;
1936 struct in_addr mask, gateway;
1938 struct iface *local_iface;
1941 /* If there's no local interface or no IP address, give up. */
1942 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1943 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1947 /* Bring up the local interface. */
1948 netdev = local_iface->netdev;
1949 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1951 /* Configure the IP address and netmask. */
1952 if (!c->local_netmask
1953 || !inet_aton(c->local_netmask, &mask)
1955 mask.s_addr = guess_netmask(ip.s_addr);
1957 if (!netdev_set_in4(netdev, ip, mask)) {
1958 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1959 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1962 /* Configure the default gateway. */
1963 if (c->local_gateway
1964 && inet_aton(c->local_gateway, &gateway)
1965 && gateway.s_addr) {
1966 if (!netdev_add_router(netdev, gateway)) {
1967 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1968 br->name, IP_ARGS(&gateway.s_addr));
1974 bridge_reconfigure_remotes(struct bridge *br,
1975 const struct sockaddr_in *managers,
1978 const char *disable_ib_str, *queue_id_str;
1979 bool disable_in_band = false;
1982 struct ovsrec_controller **controllers;
1983 size_t n_controllers;
1986 struct ofproto_controller *ocs;
1990 /* Check if we should disable in-band control on this bridge. */
1991 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1992 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1993 disable_in_band = true;
1996 /* Set OpenFlow queue ID for in-band control. */
1997 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1998 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1999 ofproto_set_in_band_queue(br->ofproto, queue_id);
2001 if (disable_in_band) {
2002 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2004 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2006 had_primary = ofproto_has_primary_controller(br->ofproto);
2008 n_controllers = bridge_get_controllers(br, &controllers);
2010 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2013 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2014 for (i = 0; i < n_controllers; i++) {
2015 struct ovsrec_controller *c = controllers[i];
2017 if (!strncmp(c->target, "punix:", 6)
2018 || !strncmp(c->target, "unix:", 5)) {
2019 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2021 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2022 * domain sockets and overwriting arbitrary local files. */
2023 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
2024 "\"%s\" due to possibility for remote exploit",
2025 dpif_name(br->dpif), c->target);
2029 bridge_configure_local_iface_netdev(br, c);
2030 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2031 if (disable_in_band) {
2032 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2037 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2038 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2041 if (had_primary != ofproto_has_primary_controller(br->ofproto)) {
2042 ofproto_flush_flows(br->ofproto);
2045 /* If there are no controllers and the bridge is in standalone
2046 * mode, set up a flow that matches every packet and directs
2047 * them to OFPP_NORMAL (which goes to us). Otherwise, the
2048 * switch is in secure mode and we won't pass any traffic until
2049 * a controller has been defined and it tells us to do so. */
2051 && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) {
2052 union ofp_action action;
2053 struct cls_rule rule;
2055 memset(&action, 0, sizeof action);
2056 action.type = htons(OFPAT_OUTPUT);
2057 action.output.len = htons(sizeof action);
2058 action.output.port = htons(OFPP_NORMAL);
2059 cls_rule_init_catchall(&rule, 0);
2060 ofproto_add_flow(br->ofproto, &rule, &action, 1);
2065 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2070 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2071 struct iface *iface;
2073 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2074 shash_add_once(ifaces, iface->name, iface);
2076 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2077 shash_add_once(ifaces, port->name, NULL);
2082 /* For robustness, in case the administrator moves around datapath ports behind
2083 * our back, we re-check all the datapath port numbers here.
2085 * This function will set the 'dp_ifidx' members of interfaces that have
2086 * disappeared to -1, so only call this function from a context where those
2087 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2088 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2089 * datapath, which doesn't support UINT16_MAX+1 ports. */
2091 bridge_fetch_dp_ifaces(struct bridge *br)
2093 struct dpif_port_dump dump;
2094 struct dpif_port dpif_port;
2097 /* Reset all interface numbers. */
2098 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2099 struct iface *iface;
2101 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2102 iface->dp_ifidx = -1;
2105 hmap_clear(&br->ifaces);
2107 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2108 struct iface *iface = iface_lookup(br, dpif_port.name);
2110 if (iface->dp_ifidx >= 0) {
2111 VLOG_WARN("%s reported interface %s twice",
2112 dpif_name(br->dpif), dpif_port.name);
2113 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2114 VLOG_WARN("%s reported interface %"PRIu16" twice",
2115 dpif_name(br->dpif), dpif_port.port_no);
2117 iface->dp_ifidx = dpif_port.port_no;
2118 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2119 hash_int(iface->dp_ifidx, 0));
2122 iface_set_ofport(iface->cfg,
2123 (iface->dp_ifidx >= 0
2124 ? odp_port_to_ofp_port(iface->dp_ifidx)
2130 /* Bridge packet processing functions. */
2133 set_dst(struct dst *dst, const struct flow *flow,
2134 const struct port *in_port, const struct port *out_port,
2137 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2138 : in_port->vlan >= 0 ? in_port->vlan
2139 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2140 : vlan_tci_to_vid(flow->vlan_tci));
2142 dst->iface = (!out_port->bond
2143 ? port_get_an_iface(out_port)
2144 : bond_choose_output_slave(out_port->bond, flow,
2147 return dst->iface != NULL;
2151 mirror_mask_ffs(mirror_mask_t mask)
2153 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2158 dst_set_init(struct dst_set *set)
2160 set->dsts = set->builtin;
2162 set->allocated = ARRAY_SIZE(set->builtin);
2166 dst_set_add(struct dst_set *set, const struct dst *dst)
2168 if (set->n >= set->allocated) {
2169 size_t new_allocated;
2170 struct dst *new_dsts;
2172 new_allocated = set->allocated * 2;
2173 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2174 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2178 set->dsts = new_dsts;
2179 set->allocated = new_allocated;
2181 set->dsts[set->n++] = *dst;
2185 dst_set_free(struct dst_set *set)
2187 if (set->dsts != set->builtin) {
2193 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2196 for (i = 0; i < set->n; i++) {
2197 if (set->dsts[i].vlan == test->vlan
2198 && set->dsts[i].iface == test->iface) {
2206 port_trunks_vlan(const struct port *port, uint16_t vlan)
2208 return (port->vlan < 0
2209 && (!port->trunks || bitmap_is_set(port->trunks, vlan)));
2213 port_includes_vlan(const struct port *port, uint16_t vlan)
2215 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2219 port_is_floodable(const struct port *port)
2221 struct iface *iface;
2223 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2224 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2232 /* Returns an arbitrary interface within 'port'. */
2233 static struct iface *
2234 port_get_an_iface(const struct port *port)
2236 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2240 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2241 const struct port *in_port, const struct port *out_port,
2242 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2246 if (out_port == FLOOD_PORT) {
2249 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2251 && port_is_floodable(port)
2252 && port_includes_vlan(port, vlan)
2253 && !port->is_mirror_output_port
2254 && set_dst(&dst, flow, in_port, port, tags)) {
2255 dst_set_add(set, &dst);
2258 *nf_output_iface = NF_OUT_FLOOD;
2259 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2260 dst_set_add(set, &dst);
2261 *nf_output_iface = dst.iface->dp_ifidx;
2266 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2267 uint16_t vlan, const struct port *in_port,
2268 struct dst_set *set, tag_type *tags)
2270 mirror_mask_t mirrors;
2274 mirrors = in_port->src_mirrors;
2275 for (i = 0; i < set->n; i++) {
2276 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2283 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2284 if (flow_vlan == 0) {
2285 flow_vlan = OFP_VLAN_NONE;
2289 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2290 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2294 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2295 && !dst_is_duplicate(set, &dst)) {
2296 dst_set_add(set, &dst);
2301 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2302 if (port_includes_vlan(port, m->out_vlan)
2303 && set_dst(&dst, flow, in_port, port, tags))
2305 if (port->vlan < 0) {
2306 dst.vlan = m->out_vlan;
2308 if (dst_is_duplicate(set, &dst)) {
2312 /* Use the vlan tag on the original flow instead of
2313 * the one passed in the vlan parameter. This ensures
2314 * that we compare the vlan from before any implicit
2315 * tagging tags place. This is necessary because
2316 * dst->vlan is the final vlan, after removing implicit
2318 if (port == in_port && dst.vlan == flow_vlan) {
2319 /* Don't send out input port on same VLAN. */
2322 dst_set_add(set, &dst);
2327 mirrors &= mirrors - 1;
2332 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2333 const struct port *in_port, const struct port *out_port,
2334 tag_type *tags, struct ofpbuf *actions,
2335 uint16_t *nf_output_iface)
2337 uint16_t initial_vlan, cur_vlan;
2338 const struct dst *dst;
2342 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2344 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2346 /* Output all the packets we can without having to change the VLAN. */
2347 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2348 if (initial_vlan == 0) {
2349 initial_vlan = OFP_VLAN_NONE;
2351 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2352 if (dst->vlan != initial_vlan) {
2355 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2358 /* Then output the rest. */
2359 cur_vlan = initial_vlan;
2360 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2361 if (dst->vlan == initial_vlan) {
2364 if (dst->vlan != cur_vlan) {
2365 if (dst->vlan == OFP_VLAN_NONE) {
2366 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2369 tci = htons(dst->vlan & VLAN_VID_MASK);
2370 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2371 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2373 cur_vlan = dst->vlan;
2375 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2381 /* Returns the effective vlan of a packet, taking into account both the
2382 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2383 * the packet is untagged and -1 indicates it has an invalid header and
2384 * should be dropped. */
2385 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2386 struct port *in_port, bool have_packet)
2388 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2389 if (in_port->vlan >= 0) {
2392 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2393 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2394 "packet received on port %s configured with "
2395 "implicit VLAN %"PRIu16,
2396 br->name, vlan, in_port->name, in_port->vlan);
2400 vlan = in_port->vlan;
2402 if (!port_includes_vlan(in_port, vlan)) {
2404 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2405 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2406 "packet received on port %s not configured for "
2408 br->name, vlan, in_port->name, vlan);
2417 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2418 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2419 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2421 is_gratuitous_arp(const struct flow *flow)
2423 return (flow->dl_type == htons(ETH_TYPE_ARP)
2424 && eth_addr_is_broadcast(flow->dl_dst)
2425 && (flow->nw_proto == ARP_OP_REPLY
2426 || (flow->nw_proto == ARP_OP_REQUEST
2427 && flow->nw_src == flow->nw_dst)));
2431 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2432 struct port *in_port)
2434 struct mac_entry *mac;
2436 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2440 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2441 if (is_gratuitous_arp(flow)) {
2442 /* We don't want to learn from gratuitous ARP packets that are
2443 * reflected back over bond slaves so we lock the learning table. */
2444 if (!in_port->bond) {
2445 mac_entry_set_grat_arp_lock(mac);
2446 } else if (mac_entry_is_grat_arp_locked(mac)) {
2451 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2452 /* The log messages here could actually be useful in debugging,
2453 * so keep the rate limit relatively high. */
2454 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2455 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2456 "on port %s in VLAN %d",
2457 br->name, ETH_ADDR_ARGS(flow->dl_src),
2458 in_port->name, vlan);
2460 mac->port.p = in_port;
2461 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2465 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2466 * dropped. Returns true if they may be forwarded, false if they should be
2469 * If 'have_packet' is true, it indicates that the caller is processing a
2470 * received packet. If 'have_packet' is false, then the caller is just
2471 * revalidating an existing flow because configuration has changed. Either
2472 * way, 'have_packet' only affects logging (there is no point in logging errors
2473 * during revalidation).
2475 * Sets '*in_portp' to the input port. This will be a null pointer if
2476 * flow->in_port does not designate a known input port (in which case
2477 * is_admissible() returns false).
2479 * When returning true, sets '*vlanp' to the effective VLAN of the input
2480 * packet, as returned by flow_get_vlan().
2482 * May also add tags to '*tags', although the current implementation only does
2483 * so in one special case.
2486 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2487 tag_type *tags, int *vlanp, struct port **in_portp)
2489 struct iface *in_iface;
2490 struct port *in_port;
2493 /* Find the interface and port structure for the received packet. */
2494 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2496 /* No interface? Something fishy... */
2498 /* Odd. A few possible reasons here:
2500 * - We deleted an interface but there are still a few packets
2501 * queued up from it.
2503 * - Someone externally added an interface (e.g. with "ovs-dpctl
2504 * add-if") that we don't know about.
2506 * - Packet arrived on the local port but the local port is not
2507 * one of our bridge ports.
2509 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2511 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2512 "interface %"PRIu16, br->name, flow->in_port);
2518 *in_portp = in_port = in_iface->port;
2519 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2524 /* Drop frames for reserved multicast addresses. */
2525 if (eth_addr_is_reserved(flow->dl_dst)) {
2529 /* Drop frames on ports reserved for mirroring. */
2530 if (in_port->is_mirror_output_port) {
2532 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2533 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2534 "%s, which is reserved exclusively for mirroring",
2535 br->name, in_port->name);
2540 if (in_port->bond) {
2541 struct mac_entry *mac;
2543 switch (bond_check_admissibility(in_port->bond, in_iface,
2544 flow->dl_dst, tags)) {
2551 case BV_DROP_IF_MOVED:
2552 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2553 if (mac && mac->port.p != in_port &&
2554 (!is_gratuitous_arp(flow)
2555 || mac_entry_is_grat_arp_locked(mac))) {
2565 /* If the composed actions may be applied to any packet in the given 'flow',
2566 * returns true. Otherwise, the actions should only be applied to 'packet', or
2567 * not at all, if 'packet' was NULL. */
2569 process_flow(struct bridge *br, const struct flow *flow,
2570 const struct ofpbuf *packet, struct ofpbuf *actions,
2571 tag_type *tags, uint16_t *nf_output_iface)
2573 struct port *in_port;
2574 struct port *out_port;
2575 struct mac_entry *mac;
2578 /* Check whether we should drop packets in this flow. */
2579 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2584 /* Learn source MAC (but don't try to learn from revalidation). */
2586 update_learning_table(br, flow, vlan, in_port);
2589 /* Determine output port. */
2590 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2592 out_port = mac->port.p;
2593 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2594 /* If we are revalidating but don't have a learning entry then
2595 * eject the flow. Installing a flow that floods packets opens
2596 * up a window of time where we could learn from a packet reflected
2597 * on a bond and blackhole packets before the learning table is
2598 * updated to reflect the correct port. */
2601 out_port = FLOOD_PORT;
2604 /* Don't send packets out their input ports. */
2605 if (in_port == out_port) {
2611 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2619 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2620 struct ofpbuf *actions, tag_type *tags,
2621 uint16_t *nf_output_iface, void *br_)
2623 struct bridge *br = br_;
2625 COVERAGE_INC(bridge_process_flow);
2626 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2630 bridge_special_ofhook_cb(const struct flow *flow,
2631 const struct ofpbuf *packet, void *br_)
2633 struct iface *iface;
2634 struct bridge *br = br_;
2636 iface = iface_from_dp_ifidx(br, flow->in_port);
2638 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2639 if (iface && iface->port->lacp && packet) {
2640 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2642 lacp_process_pdu(iface->port->lacp, iface, pdu);
2652 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2653 const struct nlattr *actions,
2655 uint64_t n_bytes, void *br_)
2657 struct bridge *br = br_;
2658 const struct nlattr *a;
2659 struct port *in_port;
2664 /* Feed information from the active flows back into the learning table to
2665 * ensure that table is always in sync with what is actually flowing
2666 * through the datapath.
2668 * We test that 'tags' is nonzero to ensure that only flows that include an
2669 * OFPP_NORMAL action are used for learning. This works because
2670 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2671 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2672 update_learning_table(br, flow, vlan, in_port);
2675 /* Account for bond slave utilization. */
2676 if (!br->has_bonded_ports) {
2679 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2680 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2681 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2682 if (out_port && out_port->bond) {
2683 uint16_t vlan = (flow->vlan_tci
2684 ? vlan_tci_to_vid(flow->vlan_tci)
2686 bond_account(out_port->bond, flow, vlan, n_bytes);
2693 bridge_account_checkpoint_ofhook_cb(void *br_)
2695 struct bridge *br = br_;
2698 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2700 bond_rebalance(port->bond,
2701 ofproto_get_revalidate_set(br->ofproto));
2707 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2708 tag_type *tags, void *br_)
2710 struct bridge *br = br_;
2711 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2712 struct port *port = port_from_dp_ifidx(br, odp_port);
2717 } else if (list_is_short(&port->ifaces)) {
2720 struct iface *iface;
2722 /* Autopath does not support VLAN hashing. */
2723 iface = bond_choose_output_slave(port->bond, flow,
2724 OFP_VLAN_NONE, tags);
2725 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2728 return odp_port_to_ofp_port(ret);
2731 static struct ofhooks bridge_ofhooks = {
2732 bridge_normal_ofhook_cb,
2733 bridge_special_ofhook_cb,
2734 bridge_account_flow_ofhook_cb,
2735 bridge_account_checkpoint_ofhook_cb,
2736 bridge_autopath_ofhook_cb,
2739 /* Port functions. */
2742 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2744 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2745 struct iface *iface = iface_;
2746 uint8_t ea[ETH_ADDR_LEN];
2749 error = netdev_get_etheraddr(iface->netdev, ea);
2751 struct lacp_pdu *packet_pdu;
2752 struct ofpbuf packet;
2754 ofpbuf_init(&packet, 0);
2755 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2756 sizeof *packet_pdu);
2758 error = netdev_send(iface->netdev, &packet);
2760 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2761 "(%s)", iface->port->name, iface->name,
2764 ofpbuf_uninit(&packet);
2766 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2767 "%s (%s)", iface->port->name, iface->name,
2773 port_run(struct port *port)
2776 lacp_run(port->lacp, lacp_send_pdu_cb);
2780 struct iface *iface;
2782 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2783 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2784 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2787 bond_run(port->bond,
2788 ofproto_get_revalidate_set(port->bridge->ofproto),
2789 lacp_negotiated(port->lacp));
2790 if (bond_should_send_learning_packets(port->bond)) {
2791 port_send_learning_packets(port);
2797 port_wait(struct port *port)
2800 lacp_wait(port->lacp);
2804 bond_wait(port->bond);
2808 static struct port *
2809 port_create(struct bridge *br, const char *name)
2813 port = xzalloc(sizeof *port);
2816 port->trunks = NULL;
2817 port->name = xstrdup(name);
2818 list_init(&port->ifaces);
2820 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2822 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2829 get_port_other_config(const struct ovsrec_port *port, const char *key,
2830 const char *default_value)
2834 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2836 return value ? value : default_value;
2840 get_interface_other_config(const struct ovsrec_interface *iface,
2841 const char *key, const char *default_value)
2845 value = get_ovsrec_key_value(&iface->header_,
2846 &ovsrec_interface_col_other_config, key);
2847 return value ? value : default_value;
2851 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2853 struct iface *iface, *next;
2854 struct sset new_ifaces;
2857 /* Collect list of new interfaces. */
2858 sset_init(&new_ifaces);
2859 for (i = 0; i < cfg->n_interfaces; i++) {
2860 const char *name = cfg->interfaces[i]->name;
2861 sset_add(&new_ifaces, name);
2864 /* Get rid of deleted interfaces. */
2865 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2866 if (!sset_contains(&new_ifaces, iface->name)) {
2867 iface_destroy(iface);
2871 sset_destroy(&new_ifaces);
2874 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2875 * to revalidate every flow. */
2877 port_flush_macs(struct port *port)
2879 struct bridge *br = port->bridge;
2880 struct mac_learning *ml = br->ml;
2881 struct mac_entry *mac, *next_mac;
2884 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2885 if (mac->port.p == port) {
2886 mac_learning_expire(ml, mac);
2892 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2894 struct sset new_ifaces;
2895 bool need_flush = false;
2896 unsigned long *trunks;
2903 /* Add new interfaces and update 'cfg' member of existing ones. */
2904 sset_init(&new_ifaces);
2905 for (i = 0; i < cfg->n_interfaces; i++) {
2906 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2907 struct iface *iface;
2909 if (!sset_add(&new_ifaces, if_cfg->name)) {
2910 VLOG_WARN("port %s: %s specified twice as port interface",
2911 port->name, if_cfg->name);
2912 iface_set_ofport(if_cfg, -1);
2916 iface = iface_lookup(port->bridge, if_cfg->name);
2918 if (iface->port != port) {
2919 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2921 port->bridge->name, if_cfg->name, iface->port->name);
2924 iface->cfg = if_cfg;
2926 iface = iface_create(port, if_cfg);
2929 /* Determine interface type. The local port always has type
2930 * "internal". Other ports take their type from the database and
2931 * default to "system" if none is specified. */
2932 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2933 : if_cfg->type[0] ? if_cfg->type
2936 sset_destroy(&new_ifaces);
2941 if (list_is_short(&port->ifaces)) {
2943 if (vlan >= 0 && vlan <= 4095) {
2944 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2949 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2950 * they even work as-is. But they have not been tested. */
2951 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2955 if (port->vlan != vlan) {
2960 /* Get trunked VLANs. */
2962 if (vlan < 0 && cfg->n_trunks) {
2965 trunks = bitmap_allocate(4096);
2967 for (i = 0; i < cfg->n_trunks; i++) {
2968 int trunk = cfg->trunks[i];
2970 bitmap_set1(trunks, trunk);
2976 VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
2977 port->name, cfg->n_trunks);
2979 if (n_errors == cfg->n_trunks) {
2980 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2982 bitmap_free(trunks);
2985 } else if (vlan >= 0 && cfg->n_trunks) {
2986 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2990 ? port->trunks != NULL
2991 : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
2994 bitmap_free(port->trunks);
2995 port->trunks = trunks;
2998 port_flush_macs(port);
3003 port_destroy(struct port *port)
3006 struct bridge *br = port->bridge;
3007 struct iface *iface, *next;
3010 for (i = 0; i < MAX_MIRRORS; i++) {
3011 struct mirror *m = br->mirrors[i];
3012 if (m && m->out_port == port) {
3017 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3018 iface_destroy(iface);
3021 hmap_remove(&br->ports, &port->hmap_node);
3023 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
3025 bond_destroy(port->bond);
3026 lacp_destroy(port->lacp);
3027 port_flush_macs(port);
3029 bitmap_free(port->trunks);
3035 static struct port *
3036 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3038 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
3039 return iface ? iface->port : NULL;
3042 static struct port *
3043 port_lookup(const struct bridge *br, const char *name)
3047 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3049 if (!strcmp(port->name, name)) {
3057 enable_lacp(struct port *port, bool *activep)
3059 if (!port->cfg->lacp) {
3060 /* XXX when LACP implementation has been sufficiently tested, enable by
3061 * default and make active on bonded ports. */
3063 } else if (!strcmp(port->cfg->lacp, "off")) {
3065 } else if (!strcmp(port->cfg->lacp, "active")) {
3068 } else if (!strcmp(port->cfg->lacp, "passive")) {
3072 VLOG_WARN("port %s: unknown LACP mode %s",
3073 port->name, port->cfg->lacp);
3079 iface_reconfigure_lacp(struct iface *iface)
3081 struct lacp_slave_settings s;
3084 s.name = iface->name;
3085 s.id = iface->dp_ifidx;
3086 priority = atoi(get_interface_other_config(
3087 iface->cfg, "lacp-port-priority", "0"));
3088 s.priority = (priority >= 0 && priority <= UINT16_MAX
3089 ? priority : UINT16_MAX);
3090 lacp_slave_register(iface->port->lacp, iface, &s);
3094 port_reconfigure_lacp(struct port *port)
3096 static struct lacp_settings s;
3097 struct iface *iface;
3100 if (!enable_lacp(port, &s.active)) {
3101 lacp_destroy(port->lacp);
3106 s.name = port->name;
3107 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3109 /* Prefer bondable links if unspecified. */
3110 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3112 s.priority = (priority > 0 && priority <= UINT16_MAX
3114 : UINT16_MAX - !list_is_short(&port->ifaces));
3116 s.fast = !strcmp(get_port_other_config(port->cfg, "lacp-time", "slow"),
3120 port->lacp = lacp_create();
3123 lacp_configure(port->lacp, &s);
3125 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3126 iface_reconfigure_lacp(iface);
3131 port_reconfigure_bond(struct port *port)
3133 struct bond_settings s;
3134 const char *detect_s;
3135 struct iface *iface;
3137 if (list_is_short(&port->ifaces)) {
3138 /* Not a bonded port. */
3139 bond_destroy(port->bond);
3144 port->bridge->has_bonded_ports = true;
3146 s.name = port->name;
3148 if (port->cfg->bond_mode
3149 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3150 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3151 port->name, port->cfg->bond_mode,
3152 bond_mode_to_string(s.balance));
3155 s.detect = BLSM_CARRIER;
3156 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3157 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3158 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3160 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3163 s.miimon_interval = atoi(
3164 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3165 if (s.miimon_interval < 100) {
3166 s.miimon_interval = 100;
3169 s.up_delay = MAX(0, port->cfg->bond_updelay);
3170 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3171 s.rebalance_interval = atoi(
3172 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3173 if (s.rebalance_interval < 1000) {
3174 s.rebalance_interval = 1000;
3177 s.fake_iface = port->cfg->bond_fake_iface;
3180 port->bond = bond_create(&s);
3182 if (bond_reconfigure(port->bond, &s)) {
3183 bridge_flush(port->bridge);
3187 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3188 uint16_t stable_id = (port->lacp
3189 ? lacp_slave_get_port_id(port->lacp, iface)
3191 bond_slave_register(iface->port->bond, iface, stable_id,
3197 port_send_learning_packets(struct port *port)
3199 struct bridge *br = port->bridge;
3200 int error, n_packets, n_errors;
3201 struct mac_entry *e;
3203 error = n_packets = n_errors = 0;
3204 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3205 if (e->port.p != port) {
3206 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3216 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3217 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3218 "packets, last error was: %s",
3219 port->name, n_errors, n_packets, strerror(error));
3221 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3222 port->name, n_packets);
3226 /* Interface functions. */
3228 static struct iface *
3229 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3231 struct bridge *br = port->bridge;
3232 struct iface *iface;
3233 char *name = if_cfg->name;
3235 iface = xzalloc(sizeof *iface);
3237 iface->name = xstrdup(name);
3238 iface->dp_ifidx = -1;
3239 iface->tag = tag_create_random();
3240 iface->netdev = NULL;
3241 iface->cfg = if_cfg;
3243 shash_add_assert(&br->iface_by_name, iface->name, iface);
3245 list_push_back(&port->ifaces, &iface->port_elem);
3247 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3255 iface_destroy(struct iface *iface)
3258 struct port *port = iface->port;
3259 struct bridge *br = port->bridge;
3262 bond_slave_unregister(port->bond, iface);
3266 lacp_slave_unregister(port->lacp, iface);
3269 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3271 if (iface->dp_ifidx >= 0) {
3272 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3275 list_remove(&iface->port_elem);
3277 netdev_close(iface->netdev);
3282 bridge_flush(port->bridge);
3286 static struct iface *
3287 iface_lookup(const struct bridge *br, const char *name)
3289 return shash_find_data(&br->iface_by_name, name);
3292 static struct iface *
3293 iface_find(const char *name)
3295 const struct bridge *br;
3297 LIST_FOR_EACH (br, node, &all_bridges) {
3298 struct iface *iface = iface_lookup(br, name);
3307 static struct iface *
3308 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3310 struct iface *iface;
3312 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3313 hash_int(dp_ifidx, 0), &br->ifaces) {
3314 if (iface->dp_ifidx == dp_ifidx) {
3321 /* Set Ethernet address of 'iface', if one is specified in the configuration
3324 iface_set_mac(struct iface *iface)
3326 uint8_t ea[ETH_ADDR_LEN];
3328 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3329 if (eth_addr_is_multicast(ea)) {
3330 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3332 } else if (iface->dp_ifidx == ODPP_LOCAL) {
3333 VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead",
3334 iface->name, iface->name);
3336 int error = netdev_set_etheraddr(iface->netdev, ea);
3338 VLOG_ERR("interface %s: setting MAC failed (%s)",
3339 iface->name, strerror(error));
3345 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3347 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3350 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3354 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3356 * The value strings in '*shash' are taken directly from values[], not copied,
3357 * so the caller should not modify or free them. */
3359 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3360 struct shash *shash)
3365 for (i = 0; i < n; i++) {
3366 shash_add(shash, keys[i], values[i]);
3370 /* Creates 'keys' and 'values' arrays from 'shash'.
3372 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3373 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3374 * are populated with with strings taken directly from 'shash' and thus have
3375 * the same ownership of the key-value pairs in shash.
3378 shash_to_ovs_idl_map(struct shash *shash,
3379 char ***keys, char ***values, size_t *n)
3383 struct shash_node *sn;
3385 count = shash_count(shash);
3387 k = xmalloc(count * sizeof *k);
3388 v = xmalloc(count * sizeof *v);
3391 SHASH_FOR_EACH(sn, shash) {
3402 struct iface_delete_queues_cbdata {
3403 struct netdev *netdev;
3404 const struct ovsdb_datum *queues;
3408 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3410 union ovsdb_atom atom;
3412 atom.integer = target;
3413 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3417 iface_delete_queues(unsigned int queue_id,
3418 const struct shash *details OVS_UNUSED, void *cbdata_)
3420 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3422 if (!queue_ids_include(cbdata->queues, queue_id)) {
3423 netdev_delete_queue(cbdata->netdev, queue_id);
3428 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3430 if (!qos || qos->type[0] == '\0') {
3431 netdev_set_qos(iface->netdev, NULL, NULL);
3433 struct iface_delete_queues_cbdata cbdata;
3434 struct shash details;
3437 /* Configure top-level Qos for 'iface'. */
3438 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3439 qos->n_other_config, &details);
3440 netdev_set_qos(iface->netdev, qos->type, &details);
3441 shash_destroy(&details);
3443 /* Deconfigure queues that were deleted. */
3444 cbdata.netdev = iface->netdev;
3445 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3447 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3449 /* Configure queues for 'iface'. */
3450 for (i = 0; i < qos->n_queues; i++) {
3451 const struct ovsrec_queue *queue = qos->value_queues[i];
3452 unsigned int queue_id = qos->key_queues[i];
3454 shash_from_ovs_idl_map(queue->key_other_config,
3455 queue->value_other_config,
3456 queue->n_other_config, &details);
3457 netdev_set_queue(iface->netdev, queue_id, &details);
3458 shash_destroy(&details);
3464 iface_update_cfm(struct iface *iface)
3468 uint16_t *remote_mps;
3469 struct ovsrec_monitor *mon;
3470 uint8_t maid[CCM_MAID_LEN];
3472 mon = iface->cfg->monitor;
3475 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3479 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3480 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3484 cfm.mpid = mon->mpid;
3485 cfm.interval = mon->interval ? *mon->interval : 1000;
3487 memcpy(cfm.maid, maid, sizeof cfm.maid);
3489 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3490 for(i = 0; i < mon->n_remote_mps; i++) {
3491 remote_mps[i] = mon->remote_mps[i]->mpid;
3494 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3495 &cfm, remote_mps, mon->n_remote_mps);
3499 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3500 * how 'iface''s port is configured.
3502 * Returns true if 'iface' is up, false otherwise. */
3504 iface_get_carrier(const struct iface *iface)
3507 return netdev_get_carrier(iface->netdev);
3510 /* Port mirroring. */
3512 static struct mirror *
3513 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3517 for (i = 0; i < MAX_MIRRORS; i++) {
3518 struct mirror *m = br->mirrors[i];
3519 if (m && uuid_equals(uuid, &m->uuid)) {
3527 mirror_reconfigure(struct bridge *br)
3529 unsigned long *rspan_vlans;
3533 /* Get rid of deleted mirrors. */
3534 for (i = 0; i < MAX_MIRRORS; i++) {
3535 struct mirror *m = br->mirrors[i];
3537 const struct ovsdb_datum *mc;
3538 union ovsdb_atom atom;
3540 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3541 atom.uuid = br->mirrors[i]->uuid;
3542 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3548 /* Add new mirrors and reconfigure existing ones. */
3549 for (i = 0; i < br->cfg->n_mirrors; i++) {
3550 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3551 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3553 mirror_reconfigure_one(m, cfg);
3555 mirror_create(br, cfg);
3559 /* Update port reserved status. */
3560 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3561 port->is_mirror_output_port = false;
3563 for (i = 0; i < MAX_MIRRORS; i++) {
3564 struct mirror *m = br->mirrors[i];
3565 if (m && m->out_port) {
3566 m->out_port->is_mirror_output_port = true;
3570 /* Update flooded vlans (for RSPAN). */
3572 if (br->cfg->n_flood_vlans) {
3573 rspan_vlans = bitmap_allocate(4096);
3575 for (i = 0; i < br->cfg->n_flood_vlans; i++) {
3576 int64_t vlan = br->cfg->flood_vlans[i];
3577 if (vlan >= 0 && vlan < 4096) {
3578 bitmap_set1(rspan_vlans, vlan);
3579 VLOG_INFO("bridge %s: disabling learning on vlan %"PRId64,
3582 VLOG_ERR("bridge %s: invalid value %"PRId64 "for flood VLAN",
3587 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3589 mac_learning_flush(br->ml);
3594 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3599 for (i = 0; ; i++) {
3600 if (i >= MAX_MIRRORS) {
3601 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3602 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3605 if (!br->mirrors[i]) {
3610 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3612 mac_learning_flush(br->ml);
3614 br->mirrors[i] = m = xzalloc(sizeof *m);
3617 m->name = xstrdup(cfg->name);
3618 sset_init(&m->src_ports);
3619 sset_init(&m->dst_ports);
3625 mirror_reconfigure_one(m, cfg);
3629 mirror_destroy(struct mirror *m)
3632 struct bridge *br = m->bridge;
3635 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3636 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3637 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3640 sset_destroy(&m->src_ports);
3641 sset_destroy(&m->dst_ports);
3644 m->bridge->mirrors[m->idx] = NULL;
3649 mac_learning_flush(br->ml);
3654 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3659 for (i = 0; i < n_ports; i++) {
3660 const char *name = ports[i]->name;
3661 if (port_lookup(m->bridge, name)) {
3662 sset_add(names, name);
3664 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3665 "port %s", m->bridge->name, m->name, name);
3671 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3677 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3679 for (i = 0; i < cfg->n_select_vlan; i++) {
3680 int64_t vlan = cfg->select_vlan[i];
3681 if (vlan < 0 || vlan > 4095) {
3682 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3683 m->bridge->name, m->name, vlan);
3685 (*vlans)[n_vlans++] = vlan;
3692 vlan_is_mirrored(const struct mirror *m, int vlan)
3696 for (i = 0; i < m->n_vlans; i++) {
3697 if (m->vlans[i] == vlan) {
3705 port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p)
3709 for (i = 0; i < m->n_vlans; i++) {
3710 if (port_trunks_vlan(p, m->vlans[i])) {
3718 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3720 struct sset src_ports, dst_ports;
3721 mirror_mask_t mirror_bit;
3722 struct port *out_port;
3729 if (strcmp(cfg->name, m->name)) {
3731 m->name = xstrdup(cfg->name);
3734 /* Get output port. */
3735 if (cfg->output_port) {
3736 out_port = port_lookup(m->bridge, cfg->output_port->name);
3738 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3739 m->bridge->name, m->name);
3745 if (cfg->output_vlan) {
3746 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3747 "output vlan; ignoring output vlan",
3748 m->bridge->name, m->name);
3750 } else if (cfg->output_vlan) {
3752 out_vlan = *cfg->output_vlan;
3754 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3755 m->bridge->name, m->name);
3760 sset_init(&src_ports);
3761 sset_init(&dst_ports);
3762 if (cfg->select_all) {
3763 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3764 sset_add(&src_ports, port->name);
3765 sset_add(&dst_ports, port->name);
3770 /* Get ports, and drop duplicates and ports that don't exist. */
3771 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3773 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3776 /* Get all the vlans, and drop duplicate and invalid vlans. */
3777 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3780 /* Update mirror data. */
3781 if (!sset_equals(&m->src_ports, &src_ports)
3782 || !sset_equals(&m->dst_ports, &dst_ports)
3783 || m->n_vlans != n_vlans
3784 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3785 || m->out_port != out_port
3786 || m->out_vlan != out_vlan) {
3787 bridge_flush(m->bridge);
3788 mac_learning_flush(m->bridge->ml);
3790 sset_swap(&m->src_ports, &src_ports);
3791 sset_swap(&m->dst_ports, &dst_ports);
3794 m->n_vlans = n_vlans;
3795 m->out_port = out_port;
3796 m->out_vlan = out_vlan;
3799 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3800 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3801 if (sset_contains(&m->src_ports, port->name)
3804 ? port_trunks_any_mirrored_vlan(m, port)
3805 : vlan_is_mirrored(m, port->vlan)))) {
3806 port->src_mirrors |= mirror_bit;
3808 port->src_mirrors &= ~mirror_bit;
3811 if (sset_contains(&m->dst_ports, port->name)) {
3812 port->dst_mirrors |= mirror_bit;
3814 port->dst_mirrors &= ~mirror_bit;
3819 sset_destroy(&src_ports);
3820 sset_destroy(&dst_ports);