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. */
153 /* Port mirroring info. */
154 mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
155 mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
156 bool is_mirror_output_port; /* Does port mirroring send frames here? */
160 struct list node; /* Node in global list of bridges. */
161 char *name; /* User-specified arbitrary name. */
162 struct mac_learning *ml; /* MAC learning table. */
163 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
164 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
165 const struct ovsrec_bridge *cfg;
167 /* OpenFlow switch processing. */
168 struct ofproto *ofproto; /* OpenFlow switch. */
170 /* Kernel datapath information. */
171 struct dpif *dpif; /* Datapath. */
172 struct hmap ifaces; /* "struct iface"s indexed by dp_ifidx. */
175 struct hmap ports; /* "struct port"s indexed by name. */
176 struct shash iface_by_name; /* "struct iface"s indexed by name. */
179 bool has_bonded_ports;
184 /* Port mirroring. */
185 struct mirror *mirrors[MAX_MIRRORS];
188 /* List of all bridges. */
189 static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
191 /* OVSDB IDL used to obtain configuration. */
192 static struct ovsdb_idl *idl;
194 /* Each time this timer expires, the bridge fetches systems and interface
195 * statistics and pushes them into the database. */
196 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
197 static long long int stats_timer = LLONG_MIN;
199 /* Stores the time after which CFM statistics may be written to the database.
200 * Only updated when changes to the database require rate limiting. */
201 #define CFM_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
202 static long long int cfm_limiter = LLONG_MIN;
204 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
205 static void bridge_destroy(struct bridge *);
206 static struct bridge *bridge_lookup(const char *name);
207 static unixctl_cb_func bridge_unixctl_dump_flows;
208 static unixctl_cb_func bridge_unixctl_reconnect;
209 static int bridge_run_one(struct bridge *);
210 static size_t bridge_get_controllers(const struct bridge *br,
211 struct ovsrec_controller ***controllersp);
212 static void bridge_reconfigure_one(struct bridge *);
213 static void bridge_reconfigure_remotes(struct bridge *,
214 const struct sockaddr_in *managers,
216 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
217 static void bridge_fetch_dp_ifaces(struct bridge *);
218 static void bridge_flush(struct bridge *);
219 static void bridge_pick_local_hw_addr(struct bridge *,
220 uint8_t ea[ETH_ADDR_LEN],
221 struct iface **hw_addr_iface);
222 static uint64_t bridge_pick_datapath_id(struct bridge *,
223 const uint8_t bridge_ea[ETH_ADDR_LEN],
224 struct iface *hw_addr_iface);
225 static uint64_t dpid_from_hash(const void *, size_t nbytes);
227 static unixctl_cb_func bridge_unixctl_fdb_show;
228 static unixctl_cb_func cfm_unixctl_show;
229 static unixctl_cb_func qos_unixctl_show;
231 static void port_run(struct port *);
232 static void port_wait(struct port *);
233 static struct port *port_create(struct bridge *, const char *name);
234 static void port_reconfigure(struct port *, const struct ovsrec_port *);
235 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
236 static void port_destroy(struct port *);
237 static struct port *port_lookup(const struct bridge *, const char *name);
238 static struct iface *port_get_an_iface(const struct port *);
239 static struct port *port_from_dp_ifidx(const struct bridge *,
241 static void port_reconfigure_bond(struct port *);
242 static void port_send_learning_packets(struct port *);
244 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
245 static void mirror_destroy(struct mirror *);
246 static void mirror_reconfigure(struct bridge *);
247 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
248 static bool vlan_is_mirrored(const struct mirror *, int vlan);
250 static struct iface *iface_create(struct port *port,
251 const struct ovsrec_interface *if_cfg);
252 static void iface_destroy(struct iface *);
253 static struct iface *iface_lookup(const struct bridge *, const char *name);
254 static struct iface *iface_find(const char *name);
255 static struct iface *iface_from_dp_ifidx(const struct bridge *,
257 static void iface_set_mac(struct iface *);
258 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
259 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
260 static void iface_update_cfm(struct iface *);
261 static bool iface_refresh_cfm_stats(struct iface *iface);
262 static bool iface_get_carrier(const struct iface *);
264 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
266 static void shash_to_ovs_idl_map(struct shash *,
267 char ***keys, char ***values, size_t *n);
269 /* Hooks into ofproto processing. */
270 static struct ofhooks bridge_ofhooks;
272 /* Public functions. */
274 /* Initializes the bridge module, configuring it to obtain its configuration
275 * from an OVSDB server accessed over 'remote', which should be a string in a
276 * form acceptable to ovsdb_idl_create(). */
278 bridge_init(const char *remote)
280 /* Create connection to database. */
281 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
283 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
284 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
286 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
287 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
288 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
289 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
291 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
292 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
294 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
297 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
298 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
299 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
300 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
301 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
302 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
303 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
304 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
305 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
307 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
308 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
309 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
310 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
312 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
314 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
316 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
318 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
320 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
322 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
324 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
326 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
327 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
328 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
329 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
330 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
332 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
334 /* Register unixctl commands. */
335 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
336 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
337 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
338 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
340 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
348 struct bridge *br, *next_br;
350 LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
353 ovsdb_idl_destroy(idl);
356 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
357 * but for which the ovs-vswitchd configuration 'cfg' is required. */
359 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
361 static bool already_configured_once;
362 struct sset bridge_names;
363 struct sset dpif_names, dpif_types;
367 /* Only do this once per ovs-vswitchd run. */
368 if (already_configured_once) {
371 already_configured_once = true;
373 stats_timer = time_msec() + STATS_INTERVAL;
375 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
376 sset_init(&bridge_names);
377 for (i = 0; i < cfg->n_bridges; i++) {
378 sset_add(&bridge_names, cfg->bridges[i]->name);
381 /* Iterate over all system dpifs and delete any of them that do not appear
383 sset_init(&dpif_names);
384 sset_init(&dpif_types);
385 dp_enumerate_types(&dpif_types);
386 SSET_FOR_EACH (type, &dpif_types) {
389 dp_enumerate_names(type, &dpif_names);
391 /* Delete each dpif whose name is not in 'bridge_names'. */
392 SSET_FOR_EACH (name, &dpif_names) {
393 if (!sset_contains(&bridge_names, name)) {
397 retval = dpif_open(name, type, &dpif);
405 sset_destroy(&bridge_names);
406 sset_destroy(&dpif_names);
407 sset_destroy(&dpif_types);
410 /* Callback for iterate_and_prune_ifaces(). */
412 check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
414 if (!iface->netdev) {
415 /* We already reported a related error, don't bother duplicating it. */
419 if (iface->dp_ifidx < 0) {
420 VLOG_ERR("%s interface not in %s, dropping",
421 iface->name, dpif_name(br->dpif));
425 VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
426 iface->name, iface->dp_ifidx);
430 /* Callback for iterate_and_prune_ifaces(). */
432 set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
433 void *aux OVS_UNUSED)
435 /* Set policing attributes. */
436 netdev_set_policing(iface->netdev,
437 iface->cfg->ingress_policing_rate,
438 iface->cfg->ingress_policing_burst);
440 /* Set MAC address of internal interfaces other than the local
442 if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) {
443 iface_set_mac(iface);
449 /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
450 * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
451 * deletes from 'br' any ports that no longer have any interfaces. */
453 iterate_and_prune_ifaces(struct bridge *br,
454 bool (*cb)(struct bridge *, struct iface *,
458 struct port *port, *next_port;
460 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
461 struct iface *iface, *next_iface;
463 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
464 if (!cb(br, iface, aux)) {
465 iface_set_ofport(iface->cfg, -1);
466 iface_destroy(iface);
470 if (list_is_empty(&port->ifaces)) {
471 VLOG_WARN("%s port has no interfaces, dropping", port->name);
477 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
478 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
479 * responsible for freeing '*managersp' (with free()).
481 * You may be asking yourself "why does ovs-vswitchd care?", because
482 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
483 * should not be and in fact is not directly involved in that. But
484 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
485 * it has to tell in-band control where the managers are to enable that.
486 * (Thus, only managers connected in-band are collected.)
489 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
490 struct sockaddr_in **managersp, size_t *n_managersp)
492 struct sockaddr_in *managers = NULL;
493 size_t n_managers = 0;
497 /* Collect all of the potential targets from the "targets" columns of the
498 * rows pointed to by "manager_options", excluding any that are
501 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
502 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
504 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
505 sset_find_and_delete(&targets, m->target);
507 sset_add(&targets, m->target);
511 /* Now extract the targets' IP addresses. */
512 if (!sset_is_empty(&targets)) {
515 managers = xmalloc(sset_count(&targets) * sizeof *managers);
516 SSET_FOR_EACH (target, &targets) {
517 struct sockaddr_in *sin = &managers[n_managers];
519 if ((!strncmp(target, "tcp:", 4)
520 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
521 (!strncmp(target, "ssl:", 4)
522 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
527 sset_destroy(&targets);
529 *managersp = managers;
530 *n_managersp = n_managers;
534 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
536 struct shash old_br, new_br;
537 struct shash_node *node;
538 struct bridge *br, *next;
539 struct sockaddr_in *managers;
542 int sflow_bridge_number;
544 COVERAGE_INC(bridge_reconfigure);
546 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
548 /* Collect old and new bridges. */
551 LIST_FOR_EACH (br, node, &all_bridges) {
552 shash_add(&old_br, br->name, br);
554 for (i = 0; i < ovs_cfg->n_bridges; i++) {
555 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
556 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
557 VLOG_WARN("more than one bridge named %s", br_cfg->name);
561 /* Get rid of deleted bridges and add new bridges. */
562 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
563 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
570 SHASH_FOR_EACH (node, &new_br) {
571 const char *br_name = node->name;
572 const struct ovsrec_bridge *br_cfg = node->data;
573 br = shash_find_data(&old_br, br_name);
575 /* If the bridge datapath type has changed, we need to tear it
576 * down and recreate. */
577 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
579 bridge_create(br_cfg);
582 bridge_create(br_cfg);
585 shash_destroy(&old_br);
586 shash_destroy(&new_br);
588 /* Reconfigure all bridges. */
589 LIST_FOR_EACH (br, node, &all_bridges) {
590 bridge_reconfigure_one(br);
593 /* Add and delete ports on all datapaths.
595 * The kernel will reject any attempt to add a given port to a datapath if
596 * that port already belongs to a different datapath, so we must do all
597 * port deletions before any port additions. */
598 LIST_FOR_EACH (br, node, &all_bridges) {
599 struct dpif_port_dump dump;
600 struct shash want_ifaces;
601 struct dpif_port dpif_port;
603 bridge_get_all_ifaces(br, &want_ifaces);
604 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
605 if (!shash_find(&want_ifaces, dpif_port.name)
606 && strcmp(dpif_port.name, br->name)) {
607 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
609 VLOG_WARN("failed to remove %s interface from %s: %s",
610 dpif_port.name, dpif_name(br->dpif),
615 shash_destroy(&want_ifaces);
617 LIST_FOR_EACH (br, node, &all_bridges) {
618 struct shash cur_ifaces, want_ifaces;
619 struct dpif_port_dump dump;
620 struct dpif_port dpif_port;
622 /* Get the set of interfaces currently in this datapath. */
623 shash_init(&cur_ifaces);
624 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
625 struct dpif_port *port_info = xmalloc(sizeof *port_info);
626 dpif_port_clone(port_info, &dpif_port);
627 shash_add(&cur_ifaces, dpif_port.name, port_info);
630 /* Get the set of interfaces we want on this datapath. */
631 bridge_get_all_ifaces(br, &want_ifaces);
633 hmap_clear(&br->ifaces);
634 SHASH_FOR_EACH (node, &want_ifaces) {
635 const char *if_name = node->name;
636 struct iface *iface = node->data;
637 struct dpif_port *dpif_port;
641 type = iface ? iface->type : "internal";
642 dpif_port = shash_find_data(&cur_ifaces, if_name);
644 /* If we have a port or a netdev already, and it's not the type we
645 * want, then delete the port (if any) and close the netdev (if
647 if ((dpif_port && strcmp(dpif_port->type, type))
648 || (iface && iface->netdev
649 && strcmp(type, netdev_get_type(iface->netdev)))) {
651 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
658 netdev_close(iface->netdev);
659 iface->netdev = NULL;
663 /* If the port doesn't exist or we don't have the netdev open,
664 * we need to do more work. */
665 if (!dpif_port || (iface && !iface->netdev)) {
666 struct netdev_options options;
667 struct netdev *netdev;
670 /* First open the network device. */
671 options.name = if_name;
673 options.args = &args;
674 options.ethertype = NETDEV_ETH_TYPE_NONE;
678 shash_from_ovs_idl_map(iface->cfg->key_options,
679 iface->cfg->value_options,
680 iface->cfg->n_options, &args);
682 error = netdev_open(&options, &netdev);
683 shash_destroy(&args);
686 VLOG_WARN("could not open network device %s (%s)",
687 if_name, strerror(error));
691 /* Then add the port if we haven't already. */
693 error = dpif_port_add(br->dpif, netdev, NULL);
695 netdev_close(netdev);
696 if (error == EFBIG) {
697 VLOG_ERR("ran out of valid port numbers on %s",
698 dpif_name(br->dpif));
701 VLOG_WARN("failed to add %s interface to %s: %s",
702 if_name, dpif_name(br->dpif),
709 /* Update 'iface'. */
711 iface->netdev = netdev;
713 } else if (iface && iface->netdev) {
717 shash_from_ovs_idl_map(iface->cfg->key_options,
718 iface->cfg->value_options,
719 iface->cfg->n_options, &args);
720 netdev_set_config(iface->netdev, &args);
721 shash_destroy(&args);
724 shash_destroy(&want_ifaces);
726 SHASH_FOR_EACH (node, &cur_ifaces) {
727 struct dpif_port *port_info = node->data;
728 dpif_port_destroy(port_info);
731 shash_destroy(&cur_ifaces);
733 sflow_bridge_number = 0;
734 LIST_FOR_EACH (br, node, &all_bridges) {
735 uint8_t ea[ETH_ADDR_LEN];
737 struct iface *local_iface;
738 struct iface *hw_addr_iface;
741 bridge_fetch_dp_ifaces(br);
743 /* Delete interfaces that cannot be opened.
745 * From this point forward we are guaranteed that every "struct iface"
746 * has nonnull 'netdev' and correct 'dp_ifidx'. */
747 iterate_and_prune_ifaces(br, check_iface, NULL);
749 /* Pick local port hardware address, datapath ID. */
750 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
751 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
753 int error = netdev_set_etheraddr(local_iface->netdev, ea);
755 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
756 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
757 "Ethernet address: %s",
758 br->name, strerror(error));
761 memcpy(br->ea, ea, ETH_ADDR_LEN);
763 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
764 ofproto_set_datapath_id(br->ofproto, dpid);
766 dpid_string = xasprintf("%016"PRIx64, dpid);
767 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
770 /* Set NetFlow configuration on this bridge. */
771 if (br->cfg->netflow) {
772 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
773 struct netflow_options opts;
775 memset(&opts, 0, sizeof opts);
777 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
778 if (nf_cfg->engine_type) {
779 opts.engine_type = *nf_cfg->engine_type;
781 if (nf_cfg->engine_id) {
782 opts.engine_id = *nf_cfg->engine_id;
785 opts.active_timeout = nf_cfg->active_timeout;
786 if (!opts.active_timeout) {
787 opts.active_timeout = -1;
788 } else if (opts.active_timeout < 0) {
789 VLOG_WARN("bridge %s: active timeout interval set to negative "
790 "value, using default instead (%d seconds)", br->name,
791 NF_ACTIVE_TIMEOUT_DEFAULT);
792 opts.active_timeout = -1;
795 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
796 if (opts.add_id_to_iface) {
797 if (opts.engine_id > 0x7f) {
798 VLOG_WARN("bridge %s: netflow port mangling may conflict "
799 "with another vswitch, choose an engine id less "
800 "than 128", br->name);
802 if (hmap_count(&br->ports) > 508) {
803 VLOG_WARN("bridge %s: netflow port mangling will conflict "
804 "with another port when more than 508 ports are "
809 sset_init(&opts.collectors);
810 sset_add_array(&opts.collectors,
811 nf_cfg->targets, nf_cfg->n_targets);
812 if (ofproto_set_netflow(br->ofproto, &opts)) {
813 VLOG_ERR("bridge %s: problem setting netflow collectors",
816 sset_destroy(&opts.collectors);
818 ofproto_set_netflow(br->ofproto, NULL);
821 /* Set sFlow configuration on this bridge. */
822 if (br->cfg->sflow) {
823 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
824 struct ovsrec_controller **controllers;
825 struct ofproto_sflow_options oso;
826 size_t n_controllers;
828 memset(&oso, 0, sizeof oso);
830 sset_init(&oso.targets);
831 sset_add_array(&oso.targets,
832 sflow_cfg->targets, sflow_cfg->n_targets);
834 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
835 if (sflow_cfg->sampling) {
836 oso.sampling_rate = *sflow_cfg->sampling;
839 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
840 if (sflow_cfg->polling) {
841 oso.polling_interval = *sflow_cfg->polling;
844 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
845 if (sflow_cfg->header) {
846 oso.header_len = *sflow_cfg->header;
849 oso.sub_id = sflow_bridge_number++;
850 oso.agent_device = sflow_cfg->agent;
852 oso.control_ip = NULL;
853 n_controllers = bridge_get_controllers(br, &controllers);
854 for (i = 0; i < n_controllers; i++) {
855 if (controllers[i]->local_ip) {
856 oso.control_ip = controllers[i]->local_ip;
860 ofproto_set_sflow(br->ofproto, &oso);
862 sset_destroy(&oso.targets);
864 ofproto_set_sflow(br->ofproto, NULL);
867 /* Update the controller and related settings. It would be more
868 * straightforward to call this from bridge_reconfigure_one(), but we
869 * can't do it there for two reasons. First, and most importantly, at
870 * that point we don't know the dp_ifidx of any interfaces that have
871 * been added to the bridge (because we haven't actually added them to
872 * the datapath). Second, at that point we haven't set the datapath ID
873 * yet; when a controller is configured, resetting the datapath ID will
874 * immediately disconnect from the controller, so it's better to set
875 * the datapath ID before the controller. */
876 bridge_reconfigure_remotes(br, managers, n_managers);
878 LIST_FOR_EACH (br, node, &all_bridges) {
881 br->has_bonded_ports = false;
882 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
885 port_reconfigure_bond(port);
887 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
888 iface_update_qos(iface, port->cfg->qos);
892 LIST_FOR_EACH (br, node, &all_bridges) {
893 iterate_and_prune_ifaces(br, set_iface_properties, NULL);
896 /* Some reconfiguration operations require the bridge to have been run at
898 LIST_FOR_EACH (br, node, &all_bridges) {
903 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
904 iface_update_cfm(iface);
910 /* ovs-vswitchd has completed initialization, so allow the process that
911 * forked us to exit successfully. */
912 daemonize_complete();
916 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
917 const struct ovsdb_idl_column *column,
920 const struct ovsdb_datum *datum;
921 union ovsdb_atom atom;
924 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
925 atom.string = (char *) key;
926 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
927 return idx == UINT_MAX ? NULL : datum->values[idx].string;
931 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
933 return get_ovsrec_key_value(&br_cfg->header_,
934 &ovsrec_bridge_col_other_config, key);
938 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
939 struct iface **hw_addr_iface)
945 *hw_addr_iface = NULL;
947 /* Did the user request a particular MAC? */
948 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
949 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
950 if (eth_addr_is_multicast(ea)) {
951 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
952 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
953 } else if (eth_addr_is_zero(ea)) {
954 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
960 /* Otherwise choose the minimum non-local MAC address among all of the
962 memset(ea, 0xff, ETH_ADDR_LEN);
963 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
964 uint8_t iface_ea[ETH_ADDR_LEN];
965 struct iface *candidate;
968 /* Mirror output ports don't participate. */
969 if (port->is_mirror_output_port) {
973 /* Choose the MAC address to represent the port. */
975 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
976 /* Find the interface with this Ethernet address (if any) so that
977 * we can provide the correct devname to the caller. */
978 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
979 uint8_t candidate_ea[ETH_ADDR_LEN];
980 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
981 && eth_addr_equals(iface_ea, candidate_ea)) {
986 /* Choose the interface whose MAC address will represent the port.
987 * The Linux kernel bonding code always chooses the MAC address of
988 * the first slave added to a bond, and the Fedora networking
989 * scripts always add slaves to a bond in alphabetical order, so
990 * for compatibility we choose the interface with the name that is
991 * first in alphabetical order. */
992 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
993 if (!iface || strcmp(candidate->name, iface->name) < 0) {
998 /* The local port doesn't count (since we're trying to choose its
999 * MAC address anyway). */
1000 if (iface->dp_ifidx == ODPP_LOCAL) {
1005 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1007 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1008 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
1009 iface->name, strerror(error));
1014 /* Compare against our current choice. */
1015 if (!eth_addr_is_multicast(iface_ea) &&
1016 !eth_addr_is_local(iface_ea) &&
1017 !eth_addr_is_reserved(iface_ea) &&
1018 !eth_addr_is_zero(iface_ea) &&
1019 eth_addr_compare_3way(iface_ea, ea) < 0)
1021 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1022 *hw_addr_iface = iface;
1025 if (eth_addr_is_multicast(ea)) {
1026 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1027 *hw_addr_iface = NULL;
1028 VLOG_WARN("bridge %s: using default bridge Ethernet "
1029 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1031 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1032 br->name, ETH_ADDR_ARGS(ea));
1036 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1037 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1038 * an interface on 'br', then that interface must be passed in as
1039 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1040 * 'hw_addr_iface' must be passed in as a null pointer. */
1042 bridge_pick_datapath_id(struct bridge *br,
1043 const uint8_t bridge_ea[ETH_ADDR_LEN],
1044 struct iface *hw_addr_iface)
1047 * The procedure for choosing a bridge MAC address will, in the most
1048 * ordinary case, also choose a unique MAC that we can use as a datapath
1049 * ID. In some special cases, though, multiple bridges will end up with
1050 * the same MAC address. This is OK for the bridges, but it will confuse
1051 * the OpenFlow controller, because each datapath needs a unique datapath
1054 * Datapath IDs must be unique. It is also very desirable that they be
1055 * stable from one run to the next, so that policy set on a datapath
1058 const char *datapath_id;
1061 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1062 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1066 if (hw_addr_iface) {
1068 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1070 * A bridge whose MAC address is taken from a VLAN network device
1071 * (that is, a network device created with vconfig(8) or similar
1072 * tool) will have the same MAC address as a bridge on the VLAN
1073 * device's physical network device.
1075 * Handle this case by hashing the physical network device MAC
1076 * along with the VLAN identifier.
1078 uint8_t buf[ETH_ADDR_LEN + 2];
1079 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1080 buf[ETH_ADDR_LEN] = vlan >> 8;
1081 buf[ETH_ADDR_LEN + 1] = vlan;
1082 return dpid_from_hash(buf, sizeof buf);
1085 * Assume that this bridge's MAC address is unique, since it
1086 * doesn't fit any of the cases we handle specially.
1091 * A purely internal bridge, that is, one that has no non-virtual
1092 * network devices on it at all, is more difficult because it has no
1093 * natural unique identifier at all.
1095 * When the host is a XenServer, we handle this case by hashing the
1096 * host's UUID with the name of the bridge. Names of bridges are
1097 * persistent across XenServer reboots, although they can be reused if
1098 * an internal network is destroyed and then a new one is later
1099 * created, so this is fairly effective.
1101 * When the host is not a XenServer, we punt by using a random MAC
1102 * address on each run.
1104 const char *host_uuid = xenserver_get_host_uuid();
1106 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1107 dpid = dpid_from_hash(combined, strlen(combined));
1113 return eth_addr_to_uint64(bridge_ea);
1117 dpid_from_hash(const void *data, size_t n)
1119 uint8_t hash[SHA1_DIGEST_SIZE];
1121 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1122 sha1_bytes(data, n, hash);
1123 eth_addr_mark_random(hash);
1124 return eth_addr_to_uint64(hash);
1128 iface_refresh_status(struct iface *iface)
1132 enum netdev_flags flags;
1141 if (!netdev_get_status(iface->netdev, &sh)) {
1143 char **keys, **values;
1145 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1146 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1151 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1154 shash_destroy_free_data(&sh);
1156 error = netdev_get_flags(iface->netdev, &flags);
1158 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1161 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1164 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1166 ovsrec_interface_set_duplex(iface->cfg,
1167 netdev_features_is_full_duplex(current)
1169 /* warning: uint64_t -> int64_t conversion */
1170 bps = netdev_features_to_bps(current);
1171 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1174 ovsrec_interface_set_duplex(iface->cfg, NULL);
1175 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1179 ovsrec_interface_set_link_state(iface->cfg,
1180 iface_get_carrier(iface) ? "up" : "down");
1182 error = netdev_get_mtu(iface->netdev, &mtu);
1183 if (!error && mtu != INT_MAX) {
1185 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1188 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1192 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1193 * changed, false otherwise. */
1195 iface_refresh_cfm_stats(struct iface *iface)
1197 const struct ovsrec_monitor *mon;
1198 const struct cfm *cfm;
1199 bool changed = false;
1202 mon = iface->cfg->monitor;
1203 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1209 for (i = 0; i < mon->n_remote_mps; i++) {
1210 const struct ovsrec_maintenance_point *mp;
1211 const struct remote_mp *rmp;
1213 mp = mon->remote_mps[i];
1214 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1216 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1217 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1222 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1223 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1231 iface_refresh_stats(struct iface *iface)
1237 static const struct iface_stat iface_stats[] = {
1238 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1239 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1240 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1241 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1242 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1243 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1244 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1245 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1246 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1247 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1248 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1249 { "collisions", offsetof(struct netdev_stats, collisions) },
1251 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1252 const struct iface_stat *s;
1254 char *keys[N_STATS];
1255 int64_t values[N_STATS];
1258 struct netdev_stats stats;
1260 /* Intentionally ignore return value, since errors will set 'stats' to
1261 * all-1s, and we will deal with that correctly below. */
1262 netdev_get_stats(iface->netdev, &stats);
1265 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1266 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1267 if (value != UINT64_MAX) {
1274 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1278 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1280 struct ovsdb_datum datum;
1284 get_system_stats(&stats);
1286 ovsdb_datum_from_shash(&datum, &stats);
1287 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1291 static inline const char *
1292 nx_role_to_str(enum nx_role role)
1297 case NX_ROLE_MASTER:
1302 return "*** INVALID ROLE ***";
1307 bridge_refresh_controller_status(const struct bridge *br)
1310 const struct ovsrec_controller *cfg;
1312 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1314 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1315 struct ofproto_controller_info *cinfo =
1316 shash_find_data(&info, cfg->target);
1319 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1320 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1321 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1322 (char **) cinfo->pairs.values,
1325 ovsrec_controller_set_is_connected(cfg, false);
1326 ovsrec_controller_set_role(cfg, NULL);
1327 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1331 ofproto_free_ofproto_controller_info(&info);
1337 const struct ovsrec_open_vswitch *cfg;
1339 bool datapath_destroyed;
1340 bool database_changed;
1343 /* Let each bridge do the work that it needs to do. */
1344 datapath_destroyed = false;
1345 LIST_FOR_EACH (br, node, &all_bridges) {
1346 int error = bridge_run_one(br);
1348 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1349 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1350 "forcing reconfiguration", br->name);
1351 datapath_destroyed = true;
1355 /* (Re)configure if necessary. */
1356 database_changed = ovsdb_idl_run(idl);
1357 cfg = ovsrec_open_vswitch_first(idl);
1359 /* Re-configure SSL. We do this on every trip through the main loop,
1360 * instead of just when the database changes, because the contents of the
1361 * key and certificate files can change without the database changing.
1363 * We do this before bridge_reconfigure() because that function might
1364 * initiate SSL connections and thus requires SSL to be configured. */
1365 if (cfg && cfg->ssl) {
1366 const struct ovsrec_ssl *ssl = cfg->ssl;
1368 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1369 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1372 if (database_changed || datapath_destroyed) {
1374 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1376 bridge_configure_once(cfg);
1377 bridge_reconfigure(cfg);
1379 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1380 ovsdb_idl_txn_commit(txn);
1381 ovsdb_idl_txn_destroy(txn); /* XXX */
1383 /* We still need to reconfigure to avoid dangling pointers to
1384 * now-destroyed ovsrec structures inside bridge data. */
1385 static const struct ovsrec_open_vswitch null_cfg;
1387 bridge_reconfigure(&null_cfg);
1391 /* Refresh system and interface stats if necessary. */
1392 if (time_msec() >= stats_timer) {
1394 struct ovsdb_idl_txn *txn;
1396 txn = ovsdb_idl_txn_create(idl);
1397 LIST_FOR_EACH (br, node, &all_bridges) {
1400 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1401 struct iface *iface;
1403 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1404 iface_refresh_stats(iface);
1405 iface_refresh_status(iface);
1408 bridge_refresh_controller_status(br);
1410 refresh_system_stats(cfg);
1411 ovsdb_idl_txn_commit(txn);
1412 ovsdb_idl_txn_destroy(txn); /* XXX */
1415 stats_timer = time_msec() + STATS_INTERVAL;
1418 if (time_msec() >= cfm_limiter) {
1419 struct ovsdb_idl_txn *txn;
1420 bool changed = false;
1422 txn = ovsdb_idl_txn_create(idl);
1423 LIST_FOR_EACH (br, node, &all_bridges) {
1426 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1427 struct iface *iface;
1429 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1430 changed = iface_refresh_cfm_stats(iface) || changed;
1436 cfm_limiter = time_msec() + CFM_LIMIT_INTERVAL;
1439 ovsdb_idl_txn_commit(txn);
1440 ovsdb_idl_txn_destroy(txn);
1449 LIST_FOR_EACH (br, node, &all_bridges) {
1452 ofproto_wait(br->ofproto);
1453 mac_learning_wait(br->ml);
1454 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1458 ovsdb_idl_wait(idl);
1459 poll_timer_wait_until(stats_timer);
1461 if (cfm_limiter > time_msec()) {
1462 poll_timer_wait_until(cfm_limiter);
1466 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1467 * configuration changes. */
1469 bridge_flush(struct bridge *br)
1471 COVERAGE_INC(bridge_flush);
1475 /* Bridge unixctl user interface functions. */
1477 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1478 const char *args, void *aux OVS_UNUSED)
1480 struct ds ds = DS_EMPTY_INITIALIZER;
1481 const struct bridge *br;
1482 const struct mac_entry *e;
1484 br = bridge_lookup(args);
1486 unixctl_command_reply(conn, 501, "no such bridge");
1490 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1491 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1492 struct port *port = e->port.p;
1493 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1494 port_get_an_iface(port)->dp_ifidx,
1495 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1497 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1501 /* CFM unixctl user interface functions. */
1503 cfm_unixctl_show(struct unixctl_conn *conn,
1504 const char *args, void *aux OVS_UNUSED)
1506 struct ds ds = DS_EMPTY_INITIALIZER;
1507 struct iface *iface;
1508 const struct cfm *cfm;
1510 iface = iface_find(args);
1512 unixctl_command_reply(conn, 501, "no such interface");
1516 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1519 unixctl_command_reply(conn, 501, "CFM not enabled");
1523 cfm_dump_ds(cfm, &ds);
1524 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1528 /* QoS unixctl user interface functions. */
1530 struct qos_unixctl_show_cbdata {
1532 struct iface *iface;
1536 qos_unixctl_show_cb(unsigned int queue_id,
1537 const struct shash *details,
1540 struct qos_unixctl_show_cbdata *data = aux;
1541 struct ds *ds = data->ds;
1542 struct iface *iface = data->iface;
1543 struct netdev_queue_stats stats;
1544 struct shash_node *node;
1547 ds_put_cstr(ds, "\n");
1549 ds_put_format(ds, "Queue %u:\n", queue_id);
1551 ds_put_cstr(ds, "Default:\n");
1554 SHASH_FOR_EACH (node, details) {
1555 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1558 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1560 if (stats.tx_packets != UINT64_MAX) {
1561 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1564 if (stats.tx_bytes != UINT64_MAX) {
1565 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1568 if (stats.tx_errors != UINT64_MAX) {
1569 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1572 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1573 queue_id, strerror(error));
1578 qos_unixctl_show(struct unixctl_conn *conn,
1579 const char *args, void *aux OVS_UNUSED)
1581 struct ds ds = DS_EMPTY_INITIALIZER;
1582 struct shash sh = SHASH_INITIALIZER(&sh);
1583 struct iface *iface;
1585 struct shash_node *node;
1586 struct qos_unixctl_show_cbdata data;
1589 iface = iface_find(args);
1591 unixctl_command_reply(conn, 501, "no such interface");
1595 netdev_get_qos(iface->netdev, &type, &sh);
1597 if (*type != '\0') {
1598 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1600 SHASH_FOR_EACH (node, &sh) {
1601 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1606 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1609 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1611 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1613 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1614 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1617 shash_destroy_free_data(&sh);
1621 /* Bridge reconfiguration functions. */
1622 static struct bridge *
1623 bridge_create(const struct ovsrec_bridge *br_cfg)
1628 assert(!bridge_lookup(br_cfg->name));
1629 br = xzalloc(sizeof *br);
1631 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1638 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1641 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1643 dpif_delete(br->dpif);
1644 dpif_close(br->dpif);
1649 br->name = xstrdup(br_cfg->name);
1651 br->ml = mac_learning_create();
1652 eth_addr_nicira_random(br->default_ea);
1654 hmap_init(&br->ports);
1655 hmap_init(&br->ifaces);
1656 shash_init(&br->iface_by_name);
1660 list_push_back(&all_bridges, &br->node);
1662 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1668 bridge_destroy(struct bridge *br)
1671 struct port *port, *next;
1674 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1677 list_remove(&br->node);
1678 ofproto_destroy(br->ofproto);
1679 error = dpif_delete(br->dpif);
1680 if (error && error != ENOENT) {
1681 VLOG_ERR("failed to delete %s: %s",
1682 dpif_name(br->dpif), strerror(error));
1684 dpif_close(br->dpif);
1685 mac_learning_destroy(br->ml);
1686 hmap_destroy(&br->ifaces);
1687 hmap_destroy(&br->ports);
1688 shash_destroy(&br->iface_by_name);
1694 static struct bridge *
1695 bridge_lookup(const char *name)
1699 LIST_FOR_EACH (br, node, &all_bridges) {
1700 if (!strcmp(br->name, name)) {
1707 /* Handle requests for a listing of all flows known by the OpenFlow
1708 * stack, including those normally hidden. */
1710 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1711 const char *args, void *aux OVS_UNUSED)
1716 br = bridge_lookup(args);
1718 unixctl_command_reply(conn, 501, "Unknown bridge");
1723 ofproto_get_all_flows(br->ofproto, &results);
1725 unixctl_command_reply(conn, 200, ds_cstr(&results));
1726 ds_destroy(&results);
1729 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1730 * connections and reconnect. If BRIDGE is not specified, then all bridges
1731 * drop their controller connections and reconnect. */
1733 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1734 const char *args, void *aux OVS_UNUSED)
1737 if (args[0] != '\0') {
1738 br = bridge_lookup(args);
1740 unixctl_command_reply(conn, 501, "Unknown bridge");
1743 ofproto_reconnect_controllers(br->ofproto);
1745 LIST_FOR_EACH (br, node, &all_bridges) {
1746 ofproto_reconnect_controllers(br->ofproto);
1749 unixctl_command_reply(conn, 200, NULL);
1753 bridge_run_one(struct bridge *br)
1758 error = ofproto_run1(br->ofproto);
1763 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1765 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1769 error = ofproto_run2(br->ofproto, br->flush);
1776 bridge_get_controllers(const struct bridge *br,
1777 struct ovsrec_controller ***controllersp)
1779 struct ovsrec_controller **controllers;
1780 size_t n_controllers;
1782 controllers = br->cfg->controller;
1783 n_controllers = br->cfg->n_controller;
1785 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1791 *controllersp = controllers;
1793 return n_controllers;
1797 bridge_reconfigure_one(struct bridge *br)
1799 enum ofproto_fail_mode fail_mode;
1800 struct port *port, *next;
1801 struct shash_node *node;
1802 struct shash new_ports;
1805 /* Collect new ports. */
1806 shash_init(&new_ports);
1807 for (i = 0; i < br->cfg->n_ports; i++) {
1808 const char *name = br->cfg->ports[i]->name;
1809 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1810 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1815 /* If we have a controller, then we need a local port. Complain if the
1816 * user didn't specify one.
1818 * XXX perhaps we should synthesize a port ourselves in this case. */
1819 if (bridge_get_controllers(br, NULL)) {
1820 char local_name[IF_NAMESIZE];
1823 error = dpif_port_get_name(br->dpif, ODPP_LOCAL,
1824 local_name, sizeof local_name);
1825 if (!error && !shash_find(&new_ports, local_name)) {
1826 VLOG_WARN("bridge %s: controller specified but no local port "
1827 "(port named %s) defined",
1828 br->name, local_name);
1832 /* Get rid of deleted ports.
1833 * Get rid of deleted interfaces on ports that still exist. */
1834 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1835 const struct ovsrec_port *port_cfg;
1837 port_cfg = shash_find_data(&new_ports, port->name);
1841 port_del_ifaces(port, port_cfg);
1845 /* Create new ports.
1846 * Add new interfaces to existing ports.
1847 * Reconfigure existing ports. */
1848 SHASH_FOR_EACH (node, &new_ports) {
1849 struct port *port = port_lookup(br, node->name);
1851 port = port_create(br, node->name);
1854 port_reconfigure(port, node->data);
1855 if (list_is_empty(&port->ifaces)) {
1856 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1857 br->name, port->name);
1861 shash_destroy(&new_ports);
1863 /* Set the fail-mode */
1864 fail_mode = !br->cfg->fail_mode
1865 || !strcmp(br->cfg->fail_mode, "standalone")
1866 ? OFPROTO_FAIL_STANDALONE
1867 : OFPROTO_FAIL_SECURE;
1868 if (ofproto_get_fail_mode(br->ofproto) != fail_mode
1869 && !ofproto_has_primary_controller(br->ofproto)) {
1870 ofproto_flush_flows(br->ofproto);
1872 ofproto_set_fail_mode(br->ofproto, fail_mode);
1874 /* Delete all flows if we're switching from connected to standalone or vice
1875 * versa. (XXX Should we delete all flows if we are switching from one
1876 * controller to another?) */
1878 /* Configure OpenFlow controller connection snooping. */
1879 if (!ofproto_has_snoops(br->ofproto)) {
1883 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1884 ovs_rundir(), br->name));
1885 ofproto_set_snoops(br->ofproto, &snoops);
1886 sset_destroy(&snoops);
1889 mirror_reconfigure(br);
1892 /* Initializes 'oc' appropriately as a management service controller for
1895 * The caller must free oc->target when it is no longer needed. */
1897 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1898 struct ofproto_controller *oc)
1900 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1901 oc->max_backoff = 0;
1902 oc->probe_interval = 60;
1903 oc->band = OFPROTO_OUT_OF_BAND;
1905 oc->burst_limit = 0;
1908 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1910 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1911 struct ofproto_controller *oc)
1913 oc->target = c->target;
1914 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1915 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1916 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1917 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1918 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1919 oc->burst_limit = (c->controller_burst_limit
1920 ? *c->controller_burst_limit : 0);
1923 /* Configures the IP stack for 'br''s local interface properly according to the
1924 * configuration in 'c'. */
1926 bridge_configure_local_iface_netdev(struct bridge *br,
1927 struct ovsrec_controller *c)
1929 struct netdev *netdev;
1930 struct in_addr mask, gateway;
1932 struct iface *local_iface;
1935 /* If there's no local interface or no IP address, give up. */
1936 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1937 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1941 /* Bring up the local interface. */
1942 netdev = local_iface->netdev;
1943 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1945 /* Configure the IP address and netmask. */
1946 if (!c->local_netmask
1947 || !inet_aton(c->local_netmask, &mask)
1949 mask.s_addr = guess_netmask(ip.s_addr);
1951 if (!netdev_set_in4(netdev, ip, mask)) {
1952 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1953 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1956 /* Configure the default gateway. */
1957 if (c->local_gateway
1958 && inet_aton(c->local_gateway, &gateway)
1959 && gateway.s_addr) {
1960 if (!netdev_add_router(netdev, gateway)) {
1961 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1962 br->name, IP_ARGS(&gateway.s_addr));
1968 bridge_reconfigure_remotes(struct bridge *br,
1969 const struct sockaddr_in *managers,
1972 const char *disable_ib_str, *queue_id_str;
1973 bool disable_in_band = false;
1976 struct ovsrec_controller **controllers;
1977 size_t n_controllers;
1980 struct ofproto_controller *ocs;
1984 /* Check if we should disable in-band control on this bridge. */
1985 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1986 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1987 disable_in_band = true;
1990 /* Set OpenFlow queue ID for in-band control. */
1991 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1992 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1993 ofproto_set_in_band_queue(br->ofproto, queue_id);
1995 if (disable_in_band) {
1996 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1998 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2000 had_primary = ofproto_has_primary_controller(br->ofproto);
2002 n_controllers = bridge_get_controllers(br, &controllers);
2004 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2007 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2008 for (i = 0; i < n_controllers; i++) {
2009 struct ovsrec_controller *c = controllers[i];
2011 if (!strncmp(c->target, "punix:", 6)
2012 || !strncmp(c->target, "unix:", 5)) {
2013 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2015 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2016 * domain sockets and overwriting arbitrary local files. */
2017 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
2018 "\"%s\" due to possibility for remote exploit",
2019 dpif_name(br->dpif), c->target);
2023 bridge_configure_local_iface_netdev(br, c);
2024 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2025 if (disable_in_band) {
2026 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2031 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2032 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2035 if (had_primary != ofproto_has_primary_controller(br->ofproto)) {
2036 ofproto_flush_flows(br->ofproto);
2039 /* If there are no controllers and the bridge is in standalone
2040 * mode, set up a flow that matches every packet and directs
2041 * them to OFPP_NORMAL (which goes to us). Otherwise, the
2042 * switch is in secure mode and we won't pass any traffic until
2043 * a controller has been defined and it tells us to do so. */
2045 && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) {
2046 union ofp_action action;
2047 struct cls_rule rule;
2049 memset(&action, 0, sizeof action);
2050 action.type = htons(OFPAT_OUTPUT);
2051 action.output.len = htons(sizeof action);
2052 action.output.port = htons(OFPP_NORMAL);
2053 cls_rule_init_catchall(&rule, 0);
2054 ofproto_add_flow(br->ofproto, &rule, &action, 1);
2059 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2064 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2065 struct iface *iface;
2067 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2068 shash_add_once(ifaces, iface->name, iface);
2070 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2071 shash_add_once(ifaces, port->name, NULL);
2076 /* For robustness, in case the administrator moves around datapath ports behind
2077 * our back, we re-check all the datapath port numbers here.
2079 * This function will set the 'dp_ifidx' members of interfaces that have
2080 * disappeared to -1, so only call this function from a context where those
2081 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2082 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2083 * datapath, which doesn't support UINT16_MAX+1 ports. */
2085 bridge_fetch_dp_ifaces(struct bridge *br)
2087 struct dpif_port_dump dump;
2088 struct dpif_port dpif_port;
2091 /* Reset all interface numbers. */
2092 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2093 struct iface *iface;
2095 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2096 iface->dp_ifidx = -1;
2099 hmap_clear(&br->ifaces);
2101 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2102 struct iface *iface = iface_lookup(br, dpif_port.name);
2104 if (iface->dp_ifidx >= 0) {
2105 VLOG_WARN("%s reported interface %s twice",
2106 dpif_name(br->dpif), dpif_port.name);
2107 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2108 VLOG_WARN("%s reported interface %"PRIu16" twice",
2109 dpif_name(br->dpif), dpif_port.port_no);
2111 iface->dp_ifidx = dpif_port.port_no;
2112 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2113 hash_int(iface->dp_ifidx, 0));
2116 iface_set_ofport(iface->cfg,
2117 (iface->dp_ifidx >= 0
2118 ? odp_port_to_ofp_port(iface->dp_ifidx)
2124 /* Bridge packet processing functions. */
2127 set_dst(struct dst *dst, const struct flow *flow,
2128 const struct port *in_port, const struct port *out_port,
2131 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2132 : in_port->vlan >= 0 ? in_port->vlan
2133 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2134 : vlan_tci_to_vid(flow->vlan_tci));
2136 dst->iface = (!out_port->bond
2137 ? port_get_an_iface(out_port)
2138 : bond_choose_output_slave(out_port->bond, flow,
2141 return dst->iface != NULL;
2145 mirror_mask_ffs(mirror_mask_t mask)
2147 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2152 dst_set_init(struct dst_set *set)
2154 set->dsts = set->builtin;
2156 set->allocated = ARRAY_SIZE(set->builtin);
2160 dst_set_add(struct dst_set *set, const struct dst *dst)
2162 if (set->n >= set->allocated) {
2163 size_t new_allocated;
2164 struct dst *new_dsts;
2166 new_allocated = set->allocated * 2;
2167 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2168 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2172 set->dsts = new_dsts;
2173 set->allocated = new_allocated;
2175 set->dsts[set->n++] = *dst;
2179 dst_set_free(struct dst_set *set)
2181 if (set->dsts != set->builtin) {
2187 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2190 for (i = 0; i < set->n; i++) {
2191 if (set->dsts[i].vlan == test->vlan
2192 && set->dsts[i].iface == test->iface) {
2200 port_trunks_vlan(const struct port *port, uint16_t vlan)
2202 return (port->vlan < 0
2203 && (!port->trunks || bitmap_is_set(port->trunks, vlan)));
2207 port_includes_vlan(const struct port *port, uint16_t vlan)
2209 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2213 port_is_floodable(const struct port *port)
2215 struct iface *iface;
2217 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2218 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2226 /* Returns an arbitrary interface within 'port'. */
2227 static struct iface *
2228 port_get_an_iface(const struct port *port)
2230 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2234 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2235 const struct port *in_port, const struct port *out_port,
2236 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2240 if (out_port == FLOOD_PORT) {
2243 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2245 && port_is_floodable(port)
2246 && port_includes_vlan(port, vlan)
2247 && !port->is_mirror_output_port
2248 && set_dst(&dst, flow, in_port, port, tags)) {
2249 dst_set_add(set, &dst);
2252 *nf_output_iface = NF_OUT_FLOOD;
2253 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2254 dst_set_add(set, &dst);
2255 *nf_output_iface = dst.iface->dp_ifidx;
2260 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2261 uint16_t vlan, const struct port *in_port,
2262 struct dst_set *set, tag_type *tags)
2264 mirror_mask_t mirrors;
2268 mirrors = in_port->src_mirrors;
2269 for (i = 0; i < set->n; i++) {
2270 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2277 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2278 if (flow_vlan == 0) {
2279 flow_vlan = OFP_VLAN_NONE;
2283 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2284 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2288 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2289 && !dst_is_duplicate(set, &dst)) {
2290 dst_set_add(set, &dst);
2295 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2296 if (port_includes_vlan(port, m->out_vlan)
2297 && set_dst(&dst, flow, in_port, port, tags))
2299 if (port->vlan < 0) {
2300 dst.vlan = m->out_vlan;
2302 if (dst_is_duplicate(set, &dst)) {
2306 /* Use the vlan tag on the original flow instead of
2307 * the one passed in the vlan parameter. This ensures
2308 * that we compare the vlan from before any implicit
2309 * tagging tags place. This is necessary because
2310 * dst->vlan is the final vlan, after removing implicit
2312 if (port == in_port && dst.vlan == flow_vlan) {
2313 /* Don't send out input port on same VLAN. */
2316 dst_set_add(set, &dst);
2321 mirrors &= mirrors - 1;
2326 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2327 const struct port *in_port, const struct port *out_port,
2328 tag_type *tags, struct ofpbuf *actions,
2329 uint16_t *nf_output_iface)
2331 uint16_t initial_vlan, cur_vlan;
2332 const struct dst *dst;
2336 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2338 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2340 /* Output all the packets we can without having to change the VLAN. */
2341 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2342 if (initial_vlan == 0) {
2343 initial_vlan = OFP_VLAN_NONE;
2345 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2346 if (dst->vlan != initial_vlan) {
2349 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2352 /* Then output the rest. */
2353 cur_vlan = initial_vlan;
2354 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2355 if (dst->vlan == initial_vlan) {
2358 if (dst->vlan != cur_vlan) {
2359 if (dst->vlan == OFP_VLAN_NONE) {
2360 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2363 tci = htons(dst->vlan & VLAN_VID_MASK);
2364 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2365 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2367 cur_vlan = dst->vlan;
2369 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2375 /* Returns the effective vlan of a packet, taking into account both the
2376 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2377 * the packet is untagged and -1 indicates it has an invalid header and
2378 * should be dropped. */
2379 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2380 struct port *in_port, bool have_packet)
2382 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2383 if (in_port->vlan >= 0) {
2386 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2387 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2388 "packet received on port %s configured with "
2389 "implicit VLAN %"PRIu16,
2390 br->name, vlan, in_port->name, in_port->vlan);
2394 vlan = in_port->vlan;
2396 if (!port_includes_vlan(in_port, vlan)) {
2398 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2399 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2400 "packet received on port %s not configured for "
2402 br->name, vlan, in_port->name, vlan);
2411 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2412 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2413 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2415 is_gratuitous_arp(const struct flow *flow)
2417 return (flow->dl_type == htons(ETH_TYPE_ARP)
2418 && eth_addr_is_broadcast(flow->dl_dst)
2419 && (flow->nw_proto == ARP_OP_REPLY
2420 || (flow->nw_proto == ARP_OP_REQUEST
2421 && flow->nw_src == flow->nw_dst)));
2425 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2426 struct port *in_port)
2428 struct mac_entry *mac;
2430 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2434 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2435 if (is_gratuitous_arp(flow)) {
2436 /* We don't want to learn from gratuitous ARP packets that are
2437 * reflected back over bond slaves so we lock the learning table. */
2438 if (!in_port->bond) {
2439 mac_entry_set_grat_arp_lock(mac);
2440 } else if (mac_entry_is_grat_arp_locked(mac)) {
2445 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2446 /* The log messages here could actually be useful in debugging,
2447 * so keep the rate limit relatively high. */
2448 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2449 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2450 "on port %s in VLAN %d",
2451 br->name, ETH_ADDR_ARGS(flow->dl_src),
2452 in_port->name, vlan);
2454 mac->port.p = in_port;
2455 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2459 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2460 * dropped. Returns true if they may be forwarded, false if they should be
2463 * If 'have_packet' is true, it indicates that the caller is processing a
2464 * received packet. If 'have_packet' is false, then the caller is just
2465 * revalidating an existing flow because configuration has changed. Either
2466 * way, 'have_packet' only affects logging (there is no point in logging errors
2467 * during revalidation).
2469 * Sets '*in_portp' to the input port. This will be a null pointer if
2470 * flow->in_port does not designate a known input port (in which case
2471 * is_admissible() returns false).
2473 * When returning true, sets '*vlanp' to the effective VLAN of the input
2474 * packet, as returned by flow_get_vlan().
2476 * May also add tags to '*tags', although the current implementation only does
2477 * so in one special case.
2480 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2481 tag_type *tags, int *vlanp, struct port **in_portp)
2483 struct iface *in_iface;
2484 struct port *in_port;
2487 /* Find the interface and port structure for the received packet. */
2488 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2490 /* No interface? Something fishy... */
2492 /* Odd. A few possible reasons here:
2494 * - We deleted an interface but there are still a few packets
2495 * queued up from it.
2497 * - Someone externally added an interface (e.g. with "ovs-dpctl
2498 * add-if") that we don't know about.
2500 * - Packet arrived on the local port but the local port is not
2501 * one of our bridge ports.
2503 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2505 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2506 "interface %"PRIu16, br->name, flow->in_port);
2512 *in_portp = in_port = in_iface->port;
2513 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2518 /* Drop frames for reserved multicast addresses. */
2519 if (eth_addr_is_reserved(flow->dl_dst)) {
2523 /* Drop frames on ports reserved for mirroring. */
2524 if (in_port->is_mirror_output_port) {
2526 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2527 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2528 "%s, which is reserved exclusively for mirroring",
2529 br->name, in_port->name);
2534 if (in_port->bond) {
2535 struct mac_entry *mac;
2537 switch (bond_check_admissibility(in_port->bond, in_iface,
2538 flow->dl_dst, tags)) {
2545 case BV_DROP_IF_MOVED:
2546 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2547 if (mac && mac->port.p != in_port &&
2548 (!is_gratuitous_arp(flow)
2549 || mac_entry_is_grat_arp_locked(mac))) {
2559 /* If the composed actions may be applied to any packet in the given 'flow',
2560 * returns true. Otherwise, the actions should only be applied to 'packet', or
2561 * not at all, if 'packet' was NULL. */
2563 process_flow(struct bridge *br, const struct flow *flow,
2564 const struct ofpbuf *packet, struct ofpbuf *actions,
2565 tag_type *tags, uint16_t *nf_output_iface)
2567 struct port *in_port;
2568 struct port *out_port;
2569 struct mac_entry *mac;
2572 /* Check whether we should drop packets in this flow. */
2573 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2578 /* Learn source MAC (but don't try to learn from revalidation). */
2580 update_learning_table(br, flow, vlan, in_port);
2583 /* Determine output port. */
2584 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2586 out_port = mac->port.p;
2587 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2588 /* If we are revalidating but don't have a learning entry then
2589 * eject the flow. Installing a flow that floods packets opens
2590 * up a window of time where we could learn from a packet reflected
2591 * on a bond and blackhole packets before the learning table is
2592 * updated to reflect the correct port. */
2595 out_port = FLOOD_PORT;
2598 /* Don't send packets out their input ports. */
2599 if (in_port == out_port) {
2605 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2613 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2614 struct ofpbuf *actions, tag_type *tags,
2615 uint16_t *nf_output_iface, void *br_)
2617 struct bridge *br = br_;
2619 COVERAGE_INC(bridge_process_flow);
2620 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2624 bridge_special_ofhook_cb(const struct flow *flow,
2625 const struct ofpbuf *packet, void *br_)
2627 struct iface *iface;
2628 struct bridge *br = br_;
2630 iface = iface_from_dp_ifidx(br, flow->in_port);
2632 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2633 if (iface && iface->port->bond && packet) {
2634 bond_process_lacp(iface->port->bond, iface, packet);
2643 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2644 const struct nlattr *actions,
2646 uint64_t n_bytes, void *br_)
2648 struct bridge *br = br_;
2649 const struct nlattr *a;
2650 struct port *in_port;
2655 /* Feed information from the active flows back into the learning table to
2656 * ensure that table is always in sync with what is actually flowing
2657 * through the datapath.
2659 * We test that 'tags' is nonzero to ensure that only flows that include an
2660 * OFPP_NORMAL action are used for learning. This works because
2661 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2662 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2663 update_learning_table(br, flow, vlan, in_port);
2666 /* Account for bond slave utilization. */
2667 if (!br->has_bonded_ports) {
2670 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2671 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2672 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2673 if (out_port && out_port->bond) {
2674 uint16_t vlan = (flow->vlan_tci
2675 ? vlan_tci_to_vid(flow->vlan_tci)
2677 bond_account(out_port->bond, flow, vlan, n_bytes);
2684 bridge_account_checkpoint_ofhook_cb(void *br_)
2686 struct bridge *br = br_;
2689 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2691 bond_rebalance(port->bond,
2692 ofproto_get_revalidate_set(br->ofproto));
2698 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2699 tag_type *tags, void *br_)
2701 struct bridge *br = br_;
2702 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2703 struct port *port = port_from_dp_ifidx(br, odp_port);
2708 } else if (list_is_short(&port->ifaces)) {
2711 struct iface *iface;
2713 /* Autopath does not support VLAN hashing. */
2714 iface = bond_choose_output_slave(port->bond, flow,
2715 OFP_VLAN_NONE, tags);
2716 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2719 return odp_port_to_ofp_port(ret);
2722 static struct ofhooks bridge_ofhooks = {
2723 bridge_normal_ofhook_cb,
2724 bridge_special_ofhook_cb,
2725 bridge_account_flow_ofhook_cb,
2726 bridge_account_checkpoint_ofhook_cb,
2727 bridge_autopath_ofhook_cb,
2730 /* Port functions. */
2733 port_run(struct port *port)
2736 bond_run(port->bond,
2737 ofproto_get_revalidate_set(port->bridge->ofproto));
2738 if (bond_should_send_learning_packets(port->bond)) {
2739 port_send_learning_packets(port);
2745 port_wait(struct port *port)
2748 bond_wait(port->bond);
2752 static struct port *
2753 port_create(struct bridge *br, const char *name)
2757 port = xzalloc(sizeof *port);
2760 port->trunks = NULL;
2761 port->name = xstrdup(name);
2762 list_init(&port->ifaces);
2764 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2766 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2773 get_port_other_config(const struct ovsrec_port *port, const char *key,
2774 const char *default_value)
2778 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2780 return value ? value : default_value;
2784 get_interface_other_config(const struct ovsrec_interface *iface,
2785 const char *key, const char *default_value)
2789 value = get_ovsrec_key_value(&iface->header_,
2790 &ovsrec_interface_col_other_config, key);
2791 return value ? value : default_value;
2795 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2797 struct iface *iface, *next;
2798 struct sset new_ifaces;
2801 /* Collect list of new interfaces. */
2802 sset_init(&new_ifaces);
2803 for (i = 0; i < cfg->n_interfaces; i++) {
2804 const char *name = cfg->interfaces[i]->name;
2805 sset_add(&new_ifaces, name);
2808 /* Get rid of deleted interfaces. */
2809 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2810 if (!sset_contains(&new_ifaces, iface->name)) {
2811 iface_destroy(iface);
2815 sset_destroy(&new_ifaces);
2818 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2819 * to revalidate every flow. */
2821 port_flush_macs(struct port *port)
2823 struct bridge *br = port->bridge;
2824 struct mac_learning *ml = br->ml;
2825 struct mac_entry *mac, *next_mac;
2828 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2829 if (mac->port.p == port) {
2830 mac_learning_expire(ml, mac);
2836 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2838 struct sset new_ifaces;
2839 bool need_flush = false;
2840 unsigned long *trunks;
2847 /* Add new interfaces and update 'cfg' member of existing ones. */
2848 sset_init(&new_ifaces);
2849 for (i = 0; i < cfg->n_interfaces; i++) {
2850 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2851 struct iface *iface;
2853 if (!sset_add(&new_ifaces, if_cfg->name)) {
2854 VLOG_WARN("port %s: %s specified twice as port interface",
2855 port->name, if_cfg->name);
2856 iface_set_ofport(if_cfg, -1);
2860 iface = iface_lookup(port->bridge, if_cfg->name);
2862 if (iface->port != port) {
2863 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2865 port->bridge->name, if_cfg->name, iface->port->name);
2868 iface->cfg = if_cfg;
2870 iface = iface_create(port, if_cfg);
2873 /* Determine interface type. The local port always has type
2874 * "internal". Other ports take their type from the database and
2875 * default to "system" if none is specified. */
2876 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2877 : if_cfg->type[0] ? if_cfg->type
2880 sset_destroy(&new_ifaces);
2885 if (list_is_short(&port->ifaces)) {
2887 if (vlan >= 0 && vlan <= 4095) {
2888 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2893 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2894 * they even work as-is. But they have not been tested. */
2895 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2899 if (port->vlan != vlan) {
2904 /* Get trunked VLANs. */
2906 if (vlan < 0 && cfg->n_trunks) {
2909 trunks = bitmap_allocate(4096);
2911 for (i = 0; i < cfg->n_trunks; i++) {
2912 int trunk = cfg->trunks[i];
2914 bitmap_set1(trunks, trunk);
2920 VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
2921 port->name, cfg->n_trunks);
2923 if (n_errors == cfg->n_trunks) {
2924 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2926 bitmap_free(trunks);
2929 } else if (vlan >= 0 && cfg->n_trunks) {
2930 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2934 ? port->trunks != NULL
2935 : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
2938 bitmap_free(port->trunks);
2939 port->trunks = trunks;
2942 port_flush_macs(port);
2947 port_destroy(struct port *port)
2950 struct bridge *br = port->bridge;
2951 struct iface *iface, *next;
2954 for (i = 0; i < MAX_MIRRORS; i++) {
2955 struct mirror *m = br->mirrors[i];
2956 if (m && m->out_port == port) {
2961 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2962 iface_destroy(iface);
2965 hmap_remove(&br->ports, &port->hmap_node);
2967 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2969 port_flush_macs(port);
2971 bitmap_free(port->trunks);
2977 static struct port *
2978 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
2980 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
2981 return iface ? iface->port : NULL;
2984 static struct port *
2985 port_lookup(const struct bridge *br, const char *name)
2989 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2991 if (!strcmp(port->name, name)) {
2999 enable_lacp(struct port *port, bool *activep)
3001 if (!port->cfg->lacp) {
3002 /* XXX when LACP implementation has been sufficiently tested, enable by
3003 * default and make active on bonded ports. */
3005 } else if (!strcmp(port->cfg->lacp, "off")) {
3007 } else if (!strcmp(port->cfg->lacp, "active")) {
3010 } else if (!strcmp(port->cfg->lacp, "passive")) {
3014 VLOG_WARN("port %s: unknown LACP mode %s",
3015 port->name, port->cfg->lacp);
3020 static struct lacp_settings *
3021 port_reconfigure_bond_lacp(struct port *port, struct lacp_settings *s)
3023 if (!enable_lacp(port, &s->active)) {
3027 s->name = port->name;
3028 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3029 s->priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3031 s->fast = !strcmp(get_port_other_config(port->cfg, "lacp-time", "slow"),
3034 if (s->priority <= 0 || s->priority > UINT16_MAX) {
3035 /* Prefer bondable links if unspecified. */
3036 s->priority = UINT16_MAX - !list_is_short(&port->ifaces);
3042 iface_reconfigure_bond(struct iface *iface)
3044 struct lacp_slave_settings s;
3047 s.name = iface->name;
3048 s.id = iface->dp_ifidx;
3049 priority = atoi(get_interface_other_config(
3050 iface->cfg, "lacp-port-priority", "0"));
3051 s.priority = (priority >= 0 && priority <= UINT16_MAX
3052 ? priority : UINT16_MAX);
3053 bond_slave_register(iface->port->bond, iface, iface->netdev, &s);
3057 port_reconfigure_bond(struct port *port)
3059 struct lacp_settings lacp_settings;
3060 struct bond_settings s;
3061 const char *detect_s;
3062 struct iface *iface;
3064 if (list_is_short(&port->ifaces)) {
3065 /* Not a bonded port. */
3066 bond_destroy(port->bond);
3071 port->bridge->has_bonded_ports = true;
3073 s.name = port->name;
3075 if (port->cfg->bond_mode
3076 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3077 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3078 port->name, port->cfg->bond_mode,
3079 bond_mode_to_string(s.balance));
3082 s.detect = BLSM_CARRIER;
3083 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3084 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3085 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3087 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3090 s.miimon_interval = atoi(
3091 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3092 if (s.miimon_interval < 100) {
3093 s.miimon_interval = 100;
3096 s.up_delay = MAX(0, port->cfg->bond_updelay);
3097 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3098 s.rebalance_interval = atoi(
3099 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3100 if (s.rebalance_interval < 1000) {
3101 s.rebalance_interval = 1000;
3104 s.fake_iface = port->cfg->bond_fake_iface;
3105 s.lacp = port_reconfigure_bond_lacp(port, &lacp_settings);
3108 port->bond = bond_create(&s);
3110 bond_reconfigure(port->bond, &s);
3113 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3114 iface_reconfigure_bond(iface);
3119 port_send_learning_packets(struct port *port)
3121 struct bridge *br = port->bridge;
3122 int error, n_packets, n_errors;
3123 struct mac_entry *e;
3125 error = n_packets = n_errors = 0;
3126 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3127 if (e->port.p != port) {
3128 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3138 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3139 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3140 "packets, last error was: %s",
3141 port->name, n_errors, n_packets, strerror(error));
3143 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3144 port->name, n_packets);
3148 /* Interface functions. */
3150 static struct iface *
3151 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3153 struct bridge *br = port->bridge;
3154 struct iface *iface;
3155 char *name = if_cfg->name;
3157 iface = xzalloc(sizeof *iface);
3159 iface->name = xstrdup(name);
3160 iface->dp_ifidx = -1;
3161 iface->tag = tag_create_random();
3162 iface->netdev = NULL;
3163 iface->cfg = if_cfg;
3165 shash_add_assert(&br->iface_by_name, iface->name, iface);
3167 list_push_back(&port->ifaces, &iface->port_elem);
3169 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3177 iface_destroy(struct iface *iface)
3180 struct port *port = iface->port;
3181 struct bridge *br = port->bridge;
3184 bond_slave_unregister(port->bond, iface);
3187 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3189 if (iface->dp_ifidx >= 0) {
3190 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3193 list_remove(&iface->port_elem);
3195 netdev_close(iface->netdev);
3200 bridge_flush(port->bridge);
3204 static struct iface *
3205 iface_lookup(const struct bridge *br, const char *name)
3207 return shash_find_data(&br->iface_by_name, name);
3210 static struct iface *
3211 iface_find(const char *name)
3213 const struct bridge *br;
3215 LIST_FOR_EACH (br, node, &all_bridges) {
3216 struct iface *iface = iface_lookup(br, name);
3225 static struct iface *
3226 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3228 struct iface *iface;
3230 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3231 hash_int(dp_ifidx, 0), &br->ifaces) {
3232 if (iface->dp_ifidx == dp_ifidx) {
3239 /* Set Ethernet address of 'iface', if one is specified in the configuration
3242 iface_set_mac(struct iface *iface)
3244 uint8_t ea[ETH_ADDR_LEN];
3246 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3247 if (eth_addr_is_multicast(ea)) {
3248 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3250 } else if (iface->dp_ifidx == ODPP_LOCAL) {
3251 VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead",
3252 iface->name, iface->name);
3254 int error = netdev_set_etheraddr(iface->netdev, ea);
3256 VLOG_ERR("interface %s: setting MAC failed (%s)",
3257 iface->name, strerror(error));
3263 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3265 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3268 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3272 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3274 * The value strings in '*shash' are taken directly from values[], not copied,
3275 * so the caller should not modify or free them. */
3277 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3278 struct shash *shash)
3283 for (i = 0; i < n; i++) {
3284 shash_add(shash, keys[i], values[i]);
3288 /* Creates 'keys' and 'values' arrays from 'shash'.
3290 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3291 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3292 * are populated with with strings taken directly from 'shash' and thus have
3293 * the same ownership of the key-value pairs in shash.
3296 shash_to_ovs_idl_map(struct shash *shash,
3297 char ***keys, char ***values, size_t *n)
3301 struct shash_node *sn;
3303 count = shash_count(shash);
3305 k = xmalloc(count * sizeof *k);
3306 v = xmalloc(count * sizeof *v);
3309 SHASH_FOR_EACH(sn, shash) {
3320 struct iface_delete_queues_cbdata {
3321 struct netdev *netdev;
3322 const struct ovsdb_datum *queues;
3326 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3328 union ovsdb_atom atom;
3330 atom.integer = target;
3331 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3335 iface_delete_queues(unsigned int queue_id,
3336 const struct shash *details OVS_UNUSED, void *cbdata_)
3338 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3340 if (!queue_ids_include(cbdata->queues, queue_id)) {
3341 netdev_delete_queue(cbdata->netdev, queue_id);
3346 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3348 if (!qos || qos->type[0] == '\0') {
3349 netdev_set_qos(iface->netdev, NULL, NULL);
3351 struct iface_delete_queues_cbdata cbdata;
3352 struct shash details;
3355 /* Configure top-level Qos for 'iface'. */
3356 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3357 qos->n_other_config, &details);
3358 netdev_set_qos(iface->netdev, qos->type, &details);
3359 shash_destroy(&details);
3361 /* Deconfigure queues that were deleted. */
3362 cbdata.netdev = iface->netdev;
3363 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3365 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3367 /* Configure queues for 'iface'. */
3368 for (i = 0; i < qos->n_queues; i++) {
3369 const struct ovsrec_queue *queue = qos->value_queues[i];
3370 unsigned int queue_id = qos->key_queues[i];
3372 shash_from_ovs_idl_map(queue->key_other_config,
3373 queue->value_other_config,
3374 queue->n_other_config, &details);
3375 netdev_set_queue(iface->netdev, queue_id, &details);
3376 shash_destroy(&details);
3382 iface_update_cfm(struct iface *iface)
3386 uint16_t *remote_mps;
3387 struct ovsrec_monitor *mon;
3388 uint8_t maid[CCM_MAID_LEN];
3390 mon = iface->cfg->monitor;
3393 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3397 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3398 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3402 cfm.mpid = mon->mpid;
3403 cfm.interval = mon->interval ? *mon->interval : 1000;
3405 memcpy(cfm.maid, maid, sizeof cfm.maid);
3407 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3408 for(i = 0; i < mon->n_remote_mps; i++) {
3409 remote_mps[i] = mon->remote_mps[i]->mpid;
3412 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3413 &cfm, remote_mps, mon->n_remote_mps);
3417 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3418 * how 'iface''s port is configured.
3420 * Returns true if 'iface' is up, false otherwise. */
3422 iface_get_carrier(const struct iface *iface)
3425 return netdev_get_carrier(iface->netdev);
3428 /* Port mirroring. */
3430 static struct mirror *
3431 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3435 for (i = 0; i < MAX_MIRRORS; i++) {
3436 struct mirror *m = br->mirrors[i];
3437 if (m && uuid_equals(uuid, &m->uuid)) {
3445 mirror_reconfigure(struct bridge *br)
3447 unsigned long *rspan_vlans;
3451 /* Get rid of deleted mirrors. */
3452 for (i = 0; i < MAX_MIRRORS; i++) {
3453 struct mirror *m = br->mirrors[i];
3455 const struct ovsdb_datum *mc;
3456 union ovsdb_atom atom;
3458 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3459 atom.uuid = br->mirrors[i]->uuid;
3460 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3466 /* Add new mirrors and reconfigure existing ones. */
3467 for (i = 0; i < br->cfg->n_mirrors; i++) {
3468 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3469 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3471 mirror_reconfigure_one(m, cfg);
3473 mirror_create(br, cfg);
3477 /* Update port reserved status. */
3478 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3479 port->is_mirror_output_port = false;
3481 for (i = 0; i < MAX_MIRRORS; i++) {
3482 struct mirror *m = br->mirrors[i];
3483 if (m && m->out_port) {
3484 m->out_port->is_mirror_output_port = true;
3488 /* Update flooded vlans (for RSPAN). */
3490 if (br->cfg->n_flood_vlans) {
3491 rspan_vlans = bitmap_allocate(4096);
3493 for (i = 0; i < br->cfg->n_flood_vlans; i++) {
3494 int64_t vlan = br->cfg->flood_vlans[i];
3495 if (vlan >= 0 && vlan < 4096) {
3496 bitmap_set1(rspan_vlans, vlan);
3497 VLOG_INFO("bridge %s: disabling learning on vlan %"PRId64,
3500 VLOG_ERR("bridge %s: invalid value %"PRId64 "for flood VLAN",
3505 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3507 mac_learning_flush(br->ml);
3512 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3517 for (i = 0; ; i++) {
3518 if (i >= MAX_MIRRORS) {
3519 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3520 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3523 if (!br->mirrors[i]) {
3528 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3530 mac_learning_flush(br->ml);
3532 br->mirrors[i] = m = xzalloc(sizeof *m);
3535 m->name = xstrdup(cfg->name);
3536 sset_init(&m->src_ports);
3537 sset_init(&m->dst_ports);
3543 mirror_reconfigure_one(m, cfg);
3547 mirror_destroy(struct mirror *m)
3550 struct bridge *br = m->bridge;
3553 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3554 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3555 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3558 sset_destroy(&m->src_ports);
3559 sset_destroy(&m->dst_ports);
3562 m->bridge->mirrors[m->idx] = NULL;
3567 mac_learning_flush(br->ml);
3572 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3577 for (i = 0; i < n_ports; i++) {
3578 const char *name = ports[i]->name;
3579 if (port_lookup(m->bridge, name)) {
3580 sset_add(names, name);
3582 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3583 "port %s", m->bridge->name, m->name, name);
3589 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3595 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3597 for (i = 0; i < cfg->n_select_vlan; i++) {
3598 int64_t vlan = cfg->select_vlan[i];
3599 if (vlan < 0 || vlan > 4095) {
3600 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3601 m->bridge->name, m->name, vlan);
3603 (*vlans)[n_vlans++] = vlan;
3610 vlan_is_mirrored(const struct mirror *m, int vlan)
3614 for (i = 0; i < m->n_vlans; i++) {
3615 if (m->vlans[i] == vlan) {
3623 port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p)
3627 for (i = 0; i < m->n_vlans; i++) {
3628 if (port_trunks_vlan(p, m->vlans[i])) {
3636 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3638 struct sset src_ports, dst_ports;
3639 mirror_mask_t mirror_bit;
3640 struct port *out_port;
3647 if (strcmp(cfg->name, m->name)) {
3649 m->name = xstrdup(cfg->name);
3652 /* Get output port. */
3653 if (cfg->output_port) {
3654 out_port = port_lookup(m->bridge, cfg->output_port->name);
3656 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3657 m->bridge->name, m->name);
3663 if (cfg->output_vlan) {
3664 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3665 "output vlan; ignoring output vlan",
3666 m->bridge->name, m->name);
3668 } else if (cfg->output_vlan) {
3670 out_vlan = *cfg->output_vlan;
3672 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3673 m->bridge->name, m->name);
3678 sset_init(&src_ports);
3679 sset_init(&dst_ports);
3680 if (cfg->select_all) {
3681 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3682 sset_add(&src_ports, port->name);
3683 sset_add(&dst_ports, port->name);
3688 /* Get ports, and drop duplicates and ports that don't exist. */
3689 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3691 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3694 /* Get all the vlans, and drop duplicate and invalid vlans. */
3695 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3698 /* Update mirror data. */
3699 if (!sset_equals(&m->src_ports, &src_ports)
3700 || !sset_equals(&m->dst_ports, &dst_ports)
3701 || m->n_vlans != n_vlans
3702 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3703 || m->out_port != out_port
3704 || m->out_vlan != out_vlan) {
3705 bridge_flush(m->bridge);
3706 mac_learning_flush(m->bridge->ml);
3708 sset_swap(&m->src_ports, &src_ports);
3709 sset_swap(&m->dst_ports, &dst_ports);
3712 m->n_vlans = n_vlans;
3713 m->out_port = out_port;
3714 m->out_vlan = out_vlan;
3717 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3718 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3719 if (sset_contains(&m->src_ports, port->name)
3722 ? port_trunks_any_mirrored_vlan(m, port)
3723 : vlan_is_mirrored(m, port->vlan)))) {
3724 port->src_mirrors |= mirror_bit;
3726 port->src_mirrors &= ~mirror_bit;
3729 if (sset_contains(&m->dst_ports, port->name)) {
3730 port->dst_mirrors |= mirror_bit;
3732 port->dst_mirrors &= ~mirror_bit;
3737 sset_destroy(&src_ports);
3738 sset_destroy(&dst_ports);