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"
76 #include "vlan-bitmap.h"
78 VLOG_DEFINE_THIS_MODULE(bridge);
80 COVERAGE_DEFINE(bridge_flush);
81 COVERAGE_DEFINE(bridge_process_flow);
82 COVERAGE_DEFINE(bridge_reconfigure);
90 struct dst builtin[32];
95 static void dst_set_init(struct dst_set *);
96 static void dst_set_add(struct dst_set *, const struct dst *);
97 static void dst_set_free(struct dst_set *);
100 /* These members are always valid. */
101 struct list port_elem; /* Element in struct port's "ifaces" list. */
102 struct port *port; /* Containing port. */
103 char *name; /* Host network device name. */
104 tag_type tag; /* Tag associated with this interface. */
106 /* These members are valid only after bridge_reconfigure() causes them to
108 struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
109 int dp_ifidx; /* Index within kernel datapath. */
110 struct netdev *netdev; /* Network device. */
111 const char *type; /* Usually same as cfg->type. */
112 const struct ovsrec_interface *cfg;
115 #define MAX_MIRRORS 32
116 typedef uint32_t mirror_mask_t;
117 #define MIRROR_MASK_C(X) UINT32_C(X)
118 BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS);
120 struct bridge *bridge;
123 struct uuid uuid; /* UUID of this "mirror" record in database. */
125 /* Selection criteria. */
126 struct sset src_ports; /* Source port names. */
127 struct sset dst_ports; /* Destination port names. */
132 struct port *out_port;
136 #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
138 struct bridge *bridge;
139 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
142 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
143 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
144 * NULL if all VLANs are trunked. */
145 const struct ovsrec_port *cfg;
147 /* An ordinary bridge port has 1 interface.
148 * A bridge port for bonding has at least 2 interfaces. */
149 struct list ifaces; /* List of "struct iface"s. */
151 struct lacp *lacp; /* NULL if LACP is not enabled. */
156 /* Port mirroring info. */
157 mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
158 mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
159 bool is_mirror_output_port; /* Does port mirroring send frames here? */
163 struct list node; /* Node in global list of bridges. */
164 char *name; /* User-specified arbitrary name. */
165 struct mac_learning *ml; /* MAC learning table. */
166 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
167 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
168 const struct ovsrec_bridge *cfg;
170 /* OpenFlow switch processing. */
171 struct ofproto *ofproto; /* OpenFlow switch. */
173 /* Kernel datapath information. */
174 struct dpif *dpif; /* Datapath. */
175 struct hmap ifaces; /* "struct iface"s indexed by dp_ifidx. */
178 struct hmap ports; /* "struct port"s indexed by name. */
179 struct shash iface_by_name; /* "struct iface"s indexed by name. */
182 bool has_bonded_ports;
187 /* Port mirroring. */
188 struct mirror *mirrors[MAX_MIRRORS];
190 /* Synthetic local port if necessary. */
191 struct ovsrec_port synth_local_port;
192 struct ovsrec_interface synth_local_iface;
193 struct ovsrec_interface *synth_local_ifacep;
196 /* List of all bridges. */
197 static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
199 /* OVSDB IDL used to obtain configuration. */
200 static struct ovsdb_idl *idl;
202 /* Each time this timer expires, the bridge fetches systems and interface
203 * statistics and pushes them into the database. */
204 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
205 static long long int stats_timer = LLONG_MIN;
207 /* Stores the time after which rate limited statistics may be written to the
208 * database. Only updated when changes to the database require rate limiting.
210 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
211 static long long int db_limiter = LLONG_MIN;
213 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
214 static void bridge_destroy(struct bridge *);
215 static struct bridge *bridge_lookup(const char *name);
216 static unixctl_cb_func bridge_unixctl_dump_flows;
217 static unixctl_cb_func bridge_unixctl_reconnect;
218 static int bridge_run_one(struct bridge *);
219 static size_t bridge_get_controllers(const struct bridge *br,
220 struct ovsrec_controller ***controllersp);
221 static void bridge_reconfigure_one(struct bridge *);
222 static void bridge_reconfigure_remotes(struct bridge *,
223 const struct sockaddr_in *managers,
225 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
226 static void bridge_fetch_dp_ifaces(struct bridge *);
227 static void bridge_flush(struct bridge *);
228 static void bridge_pick_local_hw_addr(struct bridge *,
229 uint8_t ea[ETH_ADDR_LEN],
230 struct iface **hw_addr_iface);
231 static uint64_t bridge_pick_datapath_id(struct bridge *,
232 const uint8_t bridge_ea[ETH_ADDR_LEN],
233 struct iface *hw_addr_iface);
234 static uint64_t dpid_from_hash(const void *, size_t nbytes);
236 static unixctl_cb_func bridge_unixctl_fdb_show;
237 static unixctl_cb_func cfm_unixctl_show;
238 static unixctl_cb_func qos_unixctl_show;
240 static void port_run(struct port *);
241 static void port_wait(struct port *);
242 static struct port *port_create(struct bridge *, const char *name);
243 static void port_reconfigure(struct port *, const struct ovsrec_port *);
244 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
245 static void port_destroy(struct port *);
246 static struct port *port_lookup(const struct bridge *, const char *name);
247 static struct iface *port_get_an_iface(const struct port *);
248 static struct port *port_from_dp_ifidx(const struct bridge *,
250 static void port_reconfigure_lacp(struct port *);
251 static void port_reconfigure_bond(struct port *);
252 static void port_send_learning_packets(struct port *);
254 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
255 static void mirror_destroy(struct mirror *);
256 static void mirror_reconfigure(struct bridge *);
257 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
258 static bool vlan_is_mirrored(const struct mirror *, int vlan);
260 static struct iface *iface_create(struct port *port,
261 const struct ovsrec_interface *if_cfg);
262 static void iface_destroy(struct iface *);
263 static struct iface *iface_lookup(const struct bridge *, const char *name);
264 static struct iface *iface_find(const char *name);
265 static struct iface *iface_from_dp_ifidx(const struct bridge *,
267 static void iface_set_mac(struct iface *);
268 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
269 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
270 static void iface_update_cfm(struct iface *);
271 static bool iface_refresh_cfm_stats(struct iface *iface);
272 static bool iface_get_carrier(const struct iface *);
273 static bool iface_is_synthetic(const struct iface *);
275 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
277 static void shash_to_ovs_idl_map(struct shash *,
278 char ***keys, char ***values, size_t *n);
280 /* Hooks into ofproto processing. */
281 static struct ofhooks bridge_ofhooks;
283 /* Public functions. */
285 /* Initializes the bridge module, configuring it to obtain its configuration
286 * from an OVSDB server accessed over 'remote', which should be a string in a
287 * form acceptable to ovsdb_idl_create(). */
289 bridge_init(const char *remote)
291 /* Create connection to database. */
292 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
294 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
295 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
296 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
297 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
298 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
299 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
300 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
302 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
303 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
305 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
306 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
315 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
316 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
319 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
320 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
321 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
323 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
325 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
327 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
329 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
333 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
335 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
338 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
339 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
340 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
341 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
343 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
345 /* Register unixctl commands. */
346 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
347 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
348 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
349 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
351 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
360 struct bridge *br, *next_br;
362 LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
365 ovsdb_idl_destroy(idl);
368 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
369 * but for which the ovs-vswitchd configuration 'cfg' is required. */
371 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
373 static bool already_configured_once;
374 struct sset bridge_names;
375 struct sset dpif_names, dpif_types;
379 /* Only do this once per ovs-vswitchd run. */
380 if (already_configured_once) {
383 already_configured_once = true;
385 stats_timer = time_msec() + STATS_INTERVAL;
387 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
388 sset_init(&bridge_names);
389 for (i = 0; i < cfg->n_bridges; i++) {
390 sset_add(&bridge_names, cfg->bridges[i]->name);
393 /* Iterate over all system dpifs and delete any of them that do not appear
395 sset_init(&dpif_names);
396 sset_init(&dpif_types);
397 dp_enumerate_types(&dpif_types);
398 SSET_FOR_EACH (type, &dpif_types) {
401 dp_enumerate_names(type, &dpif_names);
403 /* Delete each dpif whose name is not in 'bridge_names'. */
404 SSET_FOR_EACH (name, &dpif_names) {
405 if (!sset_contains(&bridge_names, name)) {
409 retval = dpif_open(name, type, &dpif);
417 sset_destroy(&bridge_names);
418 sset_destroy(&dpif_names);
419 sset_destroy(&dpif_types);
422 /* Callback for iterate_and_prune_ifaces(). */
424 check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
426 if (!iface->netdev) {
427 /* We already reported a related error, don't bother duplicating it. */
431 if (iface->dp_ifidx < 0) {
432 VLOG_ERR("%s interface not in %s, dropping",
433 iface->name, dpif_name(br->dpif));
437 VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
438 iface->name, iface->dp_ifidx);
442 /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
443 * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
444 * deletes from 'br' any ports that no longer have any interfaces. */
446 iterate_and_prune_ifaces(struct bridge *br,
447 bool (*cb)(struct bridge *, struct iface *,
451 struct port *port, *next_port;
453 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
454 struct iface *iface, *next_iface;
456 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
457 if (!cb(br, iface, aux)) {
458 iface_set_ofport(iface->cfg, -1);
459 iface_destroy(iface);
463 if (list_is_empty(&port->ifaces)) {
464 VLOG_WARN("%s port has no interfaces, dropping", port->name);
470 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
471 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
472 * responsible for freeing '*managersp' (with free()).
474 * You may be asking yourself "why does ovs-vswitchd care?", because
475 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
476 * should not be and in fact is not directly involved in that. But
477 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
478 * it has to tell in-band control where the managers are to enable that.
479 * (Thus, only managers connected in-band are collected.)
482 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
483 struct sockaddr_in **managersp, size_t *n_managersp)
485 struct sockaddr_in *managers = NULL;
486 size_t n_managers = 0;
490 /* Collect all of the potential targets from the "targets" columns of the
491 * rows pointed to by "manager_options", excluding any that are
494 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
495 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
497 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
498 sset_find_and_delete(&targets, m->target);
500 sset_add(&targets, m->target);
504 /* Now extract the targets' IP addresses. */
505 if (!sset_is_empty(&targets)) {
508 managers = xmalloc(sset_count(&targets) * sizeof *managers);
509 SSET_FOR_EACH (target, &targets) {
510 struct sockaddr_in *sin = &managers[n_managers];
512 if ((!strncmp(target, "tcp:", 4)
513 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
514 (!strncmp(target, "ssl:", 4)
515 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
520 sset_destroy(&targets);
522 *managersp = managers;
523 *n_managersp = n_managers;
527 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
529 struct shash old_br, new_br;
530 struct shash_node *node;
531 struct bridge *br, *next;
532 struct sockaddr_in *managers;
535 int sflow_bridge_number;
537 COVERAGE_INC(bridge_reconfigure);
539 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
541 /* Collect old and new bridges. */
544 LIST_FOR_EACH (br, node, &all_bridges) {
545 shash_add(&old_br, br->name, br);
547 for (i = 0; i < ovs_cfg->n_bridges; i++) {
548 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
549 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
550 VLOG_WARN("more than one bridge named %s", br_cfg->name);
554 /* Get rid of deleted bridges and add new bridges. */
555 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
556 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
563 SHASH_FOR_EACH (node, &new_br) {
564 const char *br_name = node->name;
565 const struct ovsrec_bridge *br_cfg = node->data;
566 br = shash_find_data(&old_br, br_name);
568 /* If the bridge datapath type has changed, we need to tear it
569 * down and recreate. */
570 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
572 bridge_create(br_cfg);
575 bridge_create(br_cfg);
578 shash_destroy(&old_br);
579 shash_destroy(&new_br);
581 /* Reconfigure all bridges. */
582 LIST_FOR_EACH (br, node, &all_bridges) {
583 bridge_reconfigure_one(br);
586 /* Add and delete ports on all datapaths.
588 * The kernel will reject any attempt to add a given port to a datapath if
589 * that port already belongs to a different datapath, so we must do all
590 * port deletions before any port additions. */
591 LIST_FOR_EACH (br, node, &all_bridges) {
592 struct dpif_port_dump dump;
593 struct shash want_ifaces;
594 struct dpif_port dpif_port;
596 bridge_get_all_ifaces(br, &want_ifaces);
597 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
598 if (!shash_find(&want_ifaces, dpif_port.name)
599 && strcmp(dpif_port.name, br->name)) {
600 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
602 VLOG_WARN("failed to remove %s interface from %s: %s",
603 dpif_port.name, dpif_name(br->dpif),
608 shash_destroy(&want_ifaces);
610 LIST_FOR_EACH (br, node, &all_bridges) {
611 struct shash cur_ifaces, want_ifaces;
612 struct dpif_port_dump dump;
613 struct dpif_port dpif_port;
615 /* Get the set of interfaces currently in this datapath. */
616 shash_init(&cur_ifaces);
617 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
618 struct dpif_port *port_info = xmalloc(sizeof *port_info);
619 dpif_port_clone(port_info, &dpif_port);
620 shash_add(&cur_ifaces, dpif_port.name, port_info);
623 /* Get the set of interfaces we want on this datapath. */
624 bridge_get_all_ifaces(br, &want_ifaces);
626 hmap_clear(&br->ifaces);
627 SHASH_FOR_EACH (node, &want_ifaces) {
628 const char *if_name = node->name;
629 struct iface *iface = node->data;
630 struct dpif_port *dpif_port;
634 type = iface ? iface->type : "internal";
635 dpif_port = shash_find_data(&cur_ifaces, if_name);
637 /* If we have a port or a netdev already, and it's not the type we
638 * want, then delete the port (if any) and close the netdev (if
640 if ((dpif_port && strcmp(dpif_port->type, type))
641 || (iface && iface->netdev
642 && strcmp(type, netdev_get_type(iface->netdev)))) {
644 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
651 if (iface->port->bond) {
652 /* The bond has a pointer to the netdev, so remove it
653 * from the bond before closing the netdev. The slave
654 * will get added back to the bond later, after a new
655 * netdev is available. */
656 bond_slave_unregister(iface->port->bond, iface);
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_lacp(port);
886 port_reconfigure_bond(port);
888 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
889 iface_update_qos(iface, port->cfg->qos);
890 netdev_set_policing(iface->netdev,
891 iface->cfg->ingress_policing_rate,
892 iface->cfg->ingress_policing_burst);
893 iface_set_mac(iface);
898 /* Some reconfiguration operations require the bridge to have been run at
900 LIST_FOR_EACH (br, node, &all_bridges) {
905 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
906 iface_update_cfm(iface);
912 /* ovs-vswitchd has completed initialization, so allow the process that
913 * forked us to exit successfully. */
914 daemonize_complete();
918 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
919 const struct ovsdb_idl_column *column,
922 const struct ovsdb_datum *datum;
923 union ovsdb_atom atom;
926 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
927 atom.string = (char *) key;
928 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
929 return idx == UINT_MAX ? NULL : datum->values[idx].string;
933 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
935 return get_ovsrec_key_value(&br_cfg->header_,
936 &ovsrec_bridge_col_other_config, key);
940 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
941 struct iface **hw_addr_iface)
947 *hw_addr_iface = NULL;
949 /* Did the user request a particular MAC? */
950 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
951 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
952 if (eth_addr_is_multicast(ea)) {
953 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
954 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
955 } else if (eth_addr_is_zero(ea)) {
956 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
962 /* Otherwise choose the minimum non-local MAC address among all of the
964 memset(ea, 0xff, ETH_ADDR_LEN);
965 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
966 uint8_t iface_ea[ETH_ADDR_LEN];
967 struct iface *candidate;
970 /* Mirror output ports don't participate. */
971 if (port->is_mirror_output_port) {
975 /* Choose the MAC address to represent the port. */
977 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
978 /* Find the interface with this Ethernet address (if any) so that
979 * we can provide the correct devname to the caller. */
980 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
981 uint8_t candidate_ea[ETH_ADDR_LEN];
982 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
983 && eth_addr_equals(iface_ea, candidate_ea)) {
988 /* Choose the interface whose MAC address will represent the port.
989 * The Linux kernel bonding code always chooses the MAC address of
990 * the first slave added to a bond, and the Fedora networking
991 * scripts always add slaves to a bond in alphabetical order, so
992 * for compatibility we choose the interface with the name that is
993 * first in alphabetical order. */
994 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
995 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1000 /* The local port doesn't count (since we're trying to choose its
1001 * MAC address anyway). */
1002 if (iface->dp_ifidx == ODPP_LOCAL) {
1007 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1009 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1010 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
1011 iface->name, strerror(error));
1016 /* Compare against our current choice. */
1017 if (!eth_addr_is_multicast(iface_ea) &&
1018 !eth_addr_is_local(iface_ea) &&
1019 !eth_addr_is_reserved(iface_ea) &&
1020 !eth_addr_is_zero(iface_ea) &&
1021 eth_addr_compare_3way(iface_ea, ea) < 0)
1023 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1024 *hw_addr_iface = iface;
1027 if (eth_addr_is_multicast(ea)) {
1028 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1029 *hw_addr_iface = NULL;
1030 VLOG_WARN("bridge %s: using default bridge Ethernet "
1031 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1033 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1034 br->name, ETH_ADDR_ARGS(ea));
1038 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1039 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1040 * an interface on 'br', then that interface must be passed in as
1041 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1042 * 'hw_addr_iface' must be passed in as a null pointer. */
1044 bridge_pick_datapath_id(struct bridge *br,
1045 const uint8_t bridge_ea[ETH_ADDR_LEN],
1046 struct iface *hw_addr_iface)
1049 * The procedure for choosing a bridge MAC address will, in the most
1050 * ordinary case, also choose a unique MAC that we can use as a datapath
1051 * ID. In some special cases, though, multiple bridges will end up with
1052 * the same MAC address. This is OK for the bridges, but it will confuse
1053 * the OpenFlow controller, because each datapath needs a unique datapath
1056 * Datapath IDs must be unique. It is also very desirable that they be
1057 * stable from one run to the next, so that policy set on a datapath
1060 const char *datapath_id;
1063 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1064 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1068 if (hw_addr_iface) {
1070 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1072 * A bridge whose MAC address is taken from a VLAN network device
1073 * (that is, a network device created with vconfig(8) or similar
1074 * tool) will have the same MAC address as a bridge on the VLAN
1075 * device's physical network device.
1077 * Handle this case by hashing the physical network device MAC
1078 * along with the VLAN identifier.
1080 uint8_t buf[ETH_ADDR_LEN + 2];
1081 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1082 buf[ETH_ADDR_LEN] = vlan >> 8;
1083 buf[ETH_ADDR_LEN + 1] = vlan;
1084 return dpid_from_hash(buf, sizeof buf);
1087 * Assume that this bridge's MAC address is unique, since it
1088 * doesn't fit any of the cases we handle specially.
1093 * A purely internal bridge, that is, one that has no non-virtual
1094 * network devices on it at all, is more difficult because it has no
1095 * natural unique identifier at all.
1097 * When the host is a XenServer, we handle this case by hashing the
1098 * host's UUID with the name of the bridge. Names of bridges are
1099 * persistent across XenServer reboots, although they can be reused if
1100 * an internal network is destroyed and then a new one is later
1101 * created, so this is fairly effective.
1103 * When the host is not a XenServer, we punt by using a random MAC
1104 * address on each run.
1106 const char *host_uuid = xenserver_get_host_uuid();
1108 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1109 dpid = dpid_from_hash(combined, strlen(combined));
1115 return eth_addr_to_uint64(bridge_ea);
1119 dpid_from_hash(const void *data, size_t n)
1121 uint8_t hash[SHA1_DIGEST_SIZE];
1123 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1124 sha1_bytes(data, n, hash);
1125 eth_addr_mark_random(hash);
1126 return eth_addr_to_uint64(hash);
1130 iface_refresh_status(struct iface *iface)
1134 enum netdev_flags flags;
1141 if (iface_is_synthetic(iface)) {
1147 if (!netdev_get_status(iface->netdev, &sh)) {
1149 char **keys, **values;
1151 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1152 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1157 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1160 shash_destroy_free_data(&sh);
1162 error = netdev_get_flags(iface->netdev, &flags);
1164 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1167 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1170 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1172 ovsrec_interface_set_duplex(iface->cfg,
1173 netdev_features_is_full_duplex(current)
1175 /* warning: uint64_t -> int64_t conversion */
1176 bps = netdev_features_to_bps(current);
1177 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1180 ovsrec_interface_set_duplex(iface->cfg, NULL);
1181 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1185 ovsrec_interface_set_link_state(iface->cfg,
1186 iface_get_carrier(iface) ? "up" : "down");
1188 error = netdev_get_mtu(iface->netdev, &mtu);
1189 if (!error && mtu != INT_MAX) {
1191 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1194 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1198 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1199 * changed, false otherwise. */
1201 iface_refresh_cfm_stats(struct iface *iface)
1203 const struct ovsrec_monitor *mon;
1204 const struct cfm *cfm;
1205 bool changed = false;
1208 mon = iface->cfg->monitor;
1209 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1215 for (i = 0; i < mon->n_remote_mps; i++) {
1216 const struct ovsrec_maintenance_point *mp;
1217 const struct remote_mp *rmp;
1219 mp = mon->remote_mps[i];
1220 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1222 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1223 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1228 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1229 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1237 iface_refresh_lacp_stats(struct iface *iface)
1239 bool *db_current = iface->cfg->lacp_current;
1240 bool changed = false;
1242 if (iface->port->lacp) {
1243 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1245 if (!db_current || *db_current != current) {
1247 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1249 } else if (db_current) {
1251 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1258 iface_refresh_stats(struct iface *iface)
1264 static const struct iface_stat iface_stats[] = {
1265 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1266 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1267 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1268 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1269 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1270 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1271 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1272 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1273 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1274 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1275 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1276 { "collisions", offsetof(struct netdev_stats, collisions) },
1278 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1279 const struct iface_stat *s;
1281 char *keys[N_STATS];
1282 int64_t values[N_STATS];
1285 struct netdev_stats stats;
1287 if (iface_is_synthetic(iface)) {
1291 /* Intentionally ignore return value, since errors will set 'stats' to
1292 * all-1s, and we will deal with that correctly below. */
1293 netdev_get_stats(iface->netdev, &stats);
1296 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1297 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1298 if (value != UINT64_MAX) {
1305 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1309 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1311 struct ovsdb_datum datum;
1315 get_system_stats(&stats);
1317 ovsdb_datum_from_shash(&datum, &stats);
1318 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1322 static inline const char *
1323 nx_role_to_str(enum nx_role role)
1328 case NX_ROLE_MASTER:
1333 return "*** INVALID ROLE ***";
1338 bridge_refresh_controller_status(const struct bridge *br)
1341 const struct ovsrec_controller *cfg;
1343 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1345 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1346 struct ofproto_controller_info *cinfo =
1347 shash_find_data(&info, cfg->target);
1350 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1351 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1352 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1353 (char **) cinfo->pairs.values,
1356 ovsrec_controller_set_is_connected(cfg, false);
1357 ovsrec_controller_set_role(cfg, NULL);
1358 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1362 ofproto_free_ofproto_controller_info(&info);
1368 const struct ovsrec_open_vswitch *cfg;
1370 bool datapath_destroyed;
1371 bool database_changed;
1374 /* Let each bridge do the work that it needs to do. */
1375 datapath_destroyed = false;
1376 LIST_FOR_EACH (br, node, &all_bridges) {
1377 int error = bridge_run_one(br);
1379 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1380 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1381 "forcing reconfiguration", br->name);
1382 datapath_destroyed = true;
1386 /* (Re)configure if necessary. */
1387 database_changed = ovsdb_idl_run(idl);
1388 cfg = ovsrec_open_vswitch_first(idl);
1390 /* Re-configure SSL. We do this on every trip through the main loop,
1391 * instead of just when the database changes, because the contents of the
1392 * key and certificate files can change without the database changing.
1394 * We do this before bridge_reconfigure() because that function might
1395 * initiate SSL connections and thus requires SSL to be configured. */
1396 if (cfg && cfg->ssl) {
1397 const struct ovsrec_ssl *ssl = cfg->ssl;
1399 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1400 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1403 if (database_changed || datapath_destroyed) {
1405 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1407 bridge_configure_once(cfg);
1408 bridge_reconfigure(cfg);
1410 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1411 ovsdb_idl_txn_commit(txn);
1412 ovsdb_idl_txn_destroy(txn); /* XXX */
1414 /* We still need to reconfigure to avoid dangling pointers to
1415 * now-destroyed ovsrec structures inside bridge data. */
1416 static const struct ovsrec_open_vswitch null_cfg;
1418 bridge_reconfigure(&null_cfg);
1422 /* Refresh system and interface stats if necessary. */
1423 if (time_msec() >= stats_timer) {
1425 struct ovsdb_idl_txn *txn;
1427 txn = ovsdb_idl_txn_create(idl);
1428 LIST_FOR_EACH (br, node, &all_bridges) {
1431 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1432 struct iface *iface;
1434 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1435 iface_refresh_stats(iface);
1436 iface_refresh_status(iface);
1439 bridge_refresh_controller_status(br);
1441 refresh_system_stats(cfg);
1442 ovsdb_idl_txn_commit(txn);
1443 ovsdb_idl_txn_destroy(txn); /* XXX */
1446 stats_timer = time_msec() + STATS_INTERVAL;
1449 if (time_msec() >= db_limiter) {
1450 struct ovsdb_idl_txn *txn;
1451 bool changed = false;
1453 txn = ovsdb_idl_txn_create(idl);
1454 LIST_FOR_EACH (br, node, &all_bridges) {
1457 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1458 struct iface *iface;
1460 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1461 changed = iface_refresh_cfm_stats(iface) || changed;
1462 changed = iface_refresh_lacp_stats(iface) || changed;
1468 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1471 ovsdb_idl_txn_commit(txn);
1472 ovsdb_idl_txn_destroy(txn);
1481 LIST_FOR_EACH (br, node, &all_bridges) {
1484 ofproto_wait(br->ofproto);
1485 mac_learning_wait(br->ml);
1486 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1490 ovsdb_idl_wait(idl);
1491 poll_timer_wait_until(stats_timer);
1493 if (db_limiter > time_msec()) {
1494 poll_timer_wait_until(db_limiter);
1498 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1499 * configuration changes. */
1501 bridge_flush(struct bridge *br)
1503 COVERAGE_INC(bridge_flush);
1507 /* Bridge unixctl user interface functions. */
1509 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1510 const char *args, void *aux OVS_UNUSED)
1512 struct ds ds = DS_EMPTY_INITIALIZER;
1513 const struct bridge *br;
1514 const struct mac_entry *e;
1516 br = bridge_lookup(args);
1518 unixctl_command_reply(conn, 501, "no such bridge");
1522 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1523 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1524 struct port *port = e->port.p;
1525 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1526 port_get_an_iface(port)->dp_ifidx,
1527 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1529 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1533 /* CFM unixctl user interface functions. */
1535 cfm_unixctl_show(struct unixctl_conn *conn,
1536 const char *args, void *aux OVS_UNUSED)
1538 struct ds ds = DS_EMPTY_INITIALIZER;
1539 struct iface *iface;
1540 const struct cfm *cfm;
1542 iface = iface_find(args);
1544 unixctl_command_reply(conn, 501, "no such interface");
1548 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1551 unixctl_command_reply(conn, 501, "CFM not enabled");
1555 cfm_dump_ds(cfm, &ds);
1556 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1560 /* QoS unixctl user interface functions. */
1562 struct qos_unixctl_show_cbdata {
1564 struct iface *iface;
1568 qos_unixctl_show_cb(unsigned int queue_id,
1569 const struct shash *details,
1572 struct qos_unixctl_show_cbdata *data = aux;
1573 struct ds *ds = data->ds;
1574 struct iface *iface = data->iface;
1575 struct netdev_queue_stats stats;
1576 struct shash_node *node;
1579 ds_put_cstr(ds, "\n");
1581 ds_put_format(ds, "Queue %u:\n", queue_id);
1583 ds_put_cstr(ds, "Default:\n");
1586 SHASH_FOR_EACH (node, details) {
1587 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1590 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1592 if (stats.tx_packets != UINT64_MAX) {
1593 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1596 if (stats.tx_bytes != UINT64_MAX) {
1597 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1600 if (stats.tx_errors != UINT64_MAX) {
1601 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1604 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1605 queue_id, strerror(error));
1610 qos_unixctl_show(struct unixctl_conn *conn,
1611 const char *args, void *aux OVS_UNUSED)
1613 struct ds ds = DS_EMPTY_INITIALIZER;
1614 struct shash sh = SHASH_INITIALIZER(&sh);
1615 struct iface *iface;
1617 struct shash_node *node;
1618 struct qos_unixctl_show_cbdata data;
1621 iface = iface_find(args);
1623 unixctl_command_reply(conn, 501, "no such interface");
1627 netdev_get_qos(iface->netdev, &type, &sh);
1629 if (*type != '\0') {
1630 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1632 SHASH_FOR_EACH (node, &sh) {
1633 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1638 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1641 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1643 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1645 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1646 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1649 shash_destroy_free_data(&sh);
1653 /* Bridge reconfiguration functions. */
1654 static struct bridge *
1655 bridge_create(const struct ovsrec_bridge *br_cfg)
1660 assert(!bridge_lookup(br_cfg->name));
1661 br = xzalloc(sizeof *br);
1663 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1670 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1673 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1675 dpif_delete(br->dpif);
1676 dpif_close(br->dpif);
1681 br->name = xstrdup(br_cfg->name);
1683 br->ml = mac_learning_create();
1684 eth_addr_nicira_random(br->default_ea);
1686 hmap_init(&br->ports);
1687 hmap_init(&br->ifaces);
1688 shash_init(&br->iface_by_name);
1692 list_push_back(&all_bridges, &br->node);
1694 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1700 bridge_destroy(struct bridge *br)
1703 struct port *port, *next;
1707 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1710 for (i = 0; i < MAX_MIRRORS; i++) {
1711 mirror_destroy(br->mirrors[i]);
1713 list_remove(&br->node);
1714 ofproto_destroy(br->ofproto);
1715 error = dpif_delete(br->dpif);
1716 if (error && error != ENOENT) {
1717 VLOG_ERR("failed to delete %s: %s",
1718 dpif_name(br->dpif), strerror(error));
1720 dpif_close(br->dpif);
1721 mac_learning_destroy(br->ml);
1722 hmap_destroy(&br->ifaces);
1723 hmap_destroy(&br->ports);
1724 shash_destroy(&br->iface_by_name);
1725 free(br->synth_local_iface.type);
1731 static struct bridge *
1732 bridge_lookup(const char *name)
1736 LIST_FOR_EACH (br, node, &all_bridges) {
1737 if (!strcmp(br->name, name)) {
1744 /* Handle requests for a listing of all flows known by the OpenFlow
1745 * stack, including those normally hidden. */
1747 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1748 const char *args, void *aux OVS_UNUSED)
1753 br = bridge_lookup(args);
1755 unixctl_command_reply(conn, 501, "Unknown bridge");
1760 ofproto_get_all_flows(br->ofproto, &results);
1762 unixctl_command_reply(conn, 200, ds_cstr(&results));
1763 ds_destroy(&results);
1766 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1767 * connections and reconnect. If BRIDGE is not specified, then all bridges
1768 * drop their controller connections and reconnect. */
1770 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1771 const char *args, void *aux OVS_UNUSED)
1774 if (args[0] != '\0') {
1775 br = bridge_lookup(args);
1777 unixctl_command_reply(conn, 501, "Unknown bridge");
1780 ofproto_reconnect_controllers(br->ofproto);
1782 LIST_FOR_EACH (br, node, &all_bridges) {
1783 ofproto_reconnect_controllers(br->ofproto);
1786 unixctl_command_reply(conn, 200, NULL);
1790 bridge_run_one(struct bridge *br)
1795 error = ofproto_run1(br->ofproto);
1800 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1802 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1806 error = ofproto_run2(br->ofproto, br->flush);
1813 bridge_get_controllers(const struct bridge *br,
1814 struct ovsrec_controller ***controllersp)
1816 struct ovsrec_controller **controllers;
1817 size_t n_controllers;
1819 controllers = br->cfg->controller;
1820 n_controllers = br->cfg->n_controller;
1822 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1828 *controllersp = controllers;
1830 return n_controllers;
1834 bridge_reconfigure_one(struct bridge *br)
1836 enum ofproto_fail_mode fail_mode;
1837 struct port *port, *next;
1838 struct shash_node *node;
1839 struct shash new_ports;
1842 /* Collect new ports. */
1843 shash_init(&new_ports);
1844 for (i = 0; i < br->cfg->n_ports; i++) {
1845 const char *name = br->cfg->ports[i]->name;
1846 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1847 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1851 if (!shash_find(&new_ports, br->name)) {
1852 struct dpif_port dpif_port;
1855 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1856 br->name, br->name);
1858 dpif_port_query_by_number(br->dpif, ODPP_LOCAL, &dpif_port);
1859 type = xstrdup(dpif_port.type ? dpif_port.type : "internal");
1860 dpif_port_destroy(&dpif_port);
1862 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1863 br->synth_local_port.n_interfaces = 1;
1864 br->synth_local_port.name = br->name;
1866 br->synth_local_iface.name = br->name;
1867 free(br->synth_local_iface.type);
1868 br->synth_local_iface.type = type;
1870 br->synth_local_ifacep = &br->synth_local_iface;
1872 shash_add(&new_ports, br->name, &br->synth_local_port);
1875 /* Get rid of deleted ports.
1876 * Get rid of deleted interfaces on ports that still exist. */
1877 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1878 const struct ovsrec_port *port_cfg;
1880 port_cfg = shash_find_data(&new_ports, port->name);
1884 port_del_ifaces(port, port_cfg);
1888 /* Create new ports.
1889 * Add new interfaces to existing ports.
1890 * Reconfigure existing ports. */
1891 SHASH_FOR_EACH (node, &new_ports) {
1892 struct port *port = port_lookup(br, node->name);
1894 port = port_create(br, node->name);
1897 port_reconfigure(port, node->data);
1898 if (list_is_empty(&port->ifaces)) {
1899 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1900 br->name, port->name);
1904 shash_destroy(&new_ports);
1906 /* Set the fail-mode */
1907 fail_mode = !br->cfg->fail_mode
1908 || !strcmp(br->cfg->fail_mode, "standalone")
1909 ? OFPROTO_FAIL_STANDALONE
1910 : OFPROTO_FAIL_SECURE;
1911 ofproto_set_fail_mode(br->ofproto, fail_mode);
1913 /* Configure OpenFlow controller connection snooping. */
1914 if (!ofproto_has_snoops(br->ofproto)) {
1918 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1919 ovs_rundir(), br->name));
1920 ofproto_set_snoops(br->ofproto, &snoops);
1921 sset_destroy(&snoops);
1924 mirror_reconfigure(br);
1927 /* Initializes 'oc' appropriately as a management service controller for
1930 * The caller must free oc->target when it is no longer needed. */
1932 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1933 struct ofproto_controller *oc)
1935 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1936 oc->max_backoff = 0;
1937 oc->probe_interval = 60;
1938 oc->band = OFPROTO_OUT_OF_BAND;
1940 oc->burst_limit = 0;
1943 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1945 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1946 struct ofproto_controller *oc)
1948 oc->target = c->target;
1949 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1950 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1951 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1952 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1953 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1954 oc->burst_limit = (c->controller_burst_limit
1955 ? *c->controller_burst_limit : 0);
1958 /* Configures the IP stack for 'br''s local interface properly according to the
1959 * configuration in 'c'. */
1961 bridge_configure_local_iface_netdev(struct bridge *br,
1962 struct ovsrec_controller *c)
1964 struct netdev *netdev;
1965 struct in_addr mask, gateway;
1967 struct iface *local_iface;
1970 /* If there's no local interface or no IP address, give up. */
1971 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1972 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1976 /* Bring up the local interface. */
1977 netdev = local_iface->netdev;
1978 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1980 /* Configure the IP address and netmask. */
1981 if (!c->local_netmask
1982 || !inet_aton(c->local_netmask, &mask)
1984 mask.s_addr = guess_netmask(ip.s_addr);
1986 if (!netdev_set_in4(netdev, ip, mask)) {
1987 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1988 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1991 /* Configure the default gateway. */
1992 if (c->local_gateway
1993 && inet_aton(c->local_gateway, &gateway)
1994 && gateway.s_addr) {
1995 if (!netdev_add_router(netdev, gateway)) {
1996 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1997 br->name, IP_ARGS(&gateway.s_addr));
2003 bridge_reconfigure_remotes(struct bridge *br,
2004 const struct sockaddr_in *managers,
2007 const char *disable_ib_str, *queue_id_str;
2008 bool disable_in_band = false;
2011 struct ovsrec_controller **controllers;
2012 size_t n_controllers;
2014 struct ofproto_controller *ocs;
2018 /* Check if we should disable in-band control on this bridge. */
2019 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2020 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2021 disable_in_band = true;
2024 /* Set OpenFlow queue ID for in-band control. */
2025 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2026 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2027 ofproto_set_in_band_queue(br->ofproto, queue_id);
2029 if (disable_in_band) {
2030 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2032 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2035 n_controllers = bridge_get_controllers(br, &controllers);
2037 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2040 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2041 for (i = 0; i < n_controllers; i++) {
2042 struct ovsrec_controller *c = controllers[i];
2044 if (!strncmp(c->target, "punix:", 6)
2045 || !strncmp(c->target, "unix:", 5)) {
2046 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2048 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2049 * domain sockets and overwriting arbitrary local files. */
2050 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
2051 "\"%s\" due to possibility for remote exploit",
2052 dpif_name(br->dpif), c->target);
2056 bridge_configure_local_iface_netdev(br, c);
2057 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2058 if (disable_in_band) {
2059 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2064 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2065 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2070 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2075 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2076 struct iface *iface;
2078 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2079 shash_add_once(ifaces, iface->name, iface);
2081 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2082 shash_add_once(ifaces, port->name, NULL);
2087 /* For robustness, in case the administrator moves around datapath ports behind
2088 * our back, we re-check all the datapath port numbers here.
2090 * This function will set the 'dp_ifidx' members of interfaces that have
2091 * disappeared to -1, so only call this function from a context where those
2092 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2093 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2094 * datapath, which doesn't support UINT16_MAX+1 ports. */
2096 bridge_fetch_dp_ifaces(struct bridge *br)
2098 struct dpif_port_dump dump;
2099 struct dpif_port dpif_port;
2102 /* Reset all interface numbers. */
2103 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2104 struct iface *iface;
2106 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2107 iface->dp_ifidx = -1;
2110 hmap_clear(&br->ifaces);
2112 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2113 struct iface *iface = iface_lookup(br, dpif_port.name);
2115 if (iface->dp_ifidx >= 0) {
2116 VLOG_WARN("%s reported interface %s twice",
2117 dpif_name(br->dpif), dpif_port.name);
2118 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2119 VLOG_WARN("%s reported interface %"PRIu16" twice",
2120 dpif_name(br->dpif), dpif_port.port_no);
2122 iface->dp_ifidx = dpif_port.port_no;
2123 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2124 hash_int(iface->dp_ifidx, 0));
2127 iface_set_ofport(iface->cfg,
2128 (iface->dp_ifidx >= 0
2129 ? odp_port_to_ofp_port(iface->dp_ifidx)
2135 /* Bridge packet processing functions. */
2138 set_dst(struct dst *dst, const struct flow *flow,
2139 const struct port *in_port, const struct port *out_port,
2142 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2143 : in_port->vlan >= 0 ? in_port->vlan
2144 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2145 : vlan_tci_to_vid(flow->vlan_tci));
2147 dst->iface = (!out_port->bond
2148 ? port_get_an_iface(out_port)
2149 : bond_choose_output_slave(out_port->bond, flow,
2152 return dst->iface != NULL;
2156 mirror_mask_ffs(mirror_mask_t mask)
2158 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2163 dst_set_init(struct dst_set *set)
2165 set->dsts = set->builtin;
2167 set->allocated = ARRAY_SIZE(set->builtin);
2171 dst_set_add(struct dst_set *set, const struct dst *dst)
2173 if (set->n >= set->allocated) {
2174 size_t new_allocated;
2175 struct dst *new_dsts;
2177 new_allocated = set->allocated * 2;
2178 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2179 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2183 set->dsts = new_dsts;
2184 set->allocated = new_allocated;
2186 set->dsts[set->n++] = *dst;
2190 dst_set_free(struct dst_set *set)
2192 if (set->dsts != set->builtin) {
2198 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2201 for (i = 0; i < set->n; i++) {
2202 if (set->dsts[i].vlan == test->vlan
2203 && set->dsts[i].iface == test->iface) {
2211 port_trunks_vlan(const struct port *port, uint16_t vlan)
2213 return (port->vlan < 0 || vlan_bitmap_contains(port->trunks, vlan));
2217 port_includes_vlan(const struct port *port, uint16_t vlan)
2219 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2223 port_is_floodable(const struct port *port)
2225 struct iface *iface;
2227 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2228 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2236 /* Returns an arbitrary interface within 'port'. */
2237 static struct iface *
2238 port_get_an_iface(const struct port *port)
2240 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2244 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2245 const struct port *in_port, const struct port *out_port,
2246 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2250 if (out_port == FLOOD_PORT) {
2253 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2255 && port_is_floodable(port)
2256 && port_includes_vlan(port, vlan)
2257 && !port->is_mirror_output_port
2258 && set_dst(&dst, flow, in_port, port, tags)) {
2259 dst_set_add(set, &dst);
2262 *nf_output_iface = NF_OUT_FLOOD;
2263 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2264 dst_set_add(set, &dst);
2265 *nf_output_iface = dst.iface->dp_ifidx;
2270 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2271 uint16_t vlan, const struct port *in_port,
2272 struct dst_set *set, tag_type *tags)
2274 mirror_mask_t mirrors;
2278 mirrors = in_port->src_mirrors;
2279 for (i = 0; i < set->n; i++) {
2280 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2287 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2288 if (flow_vlan == 0) {
2289 flow_vlan = OFP_VLAN_NONE;
2293 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2294 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2298 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2299 && !dst_is_duplicate(set, &dst)) {
2300 dst_set_add(set, &dst);
2305 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2306 if (port_includes_vlan(port, m->out_vlan)
2307 && set_dst(&dst, flow, in_port, port, tags))
2309 if (port->vlan < 0) {
2310 dst.vlan = m->out_vlan;
2312 if (dst_is_duplicate(set, &dst)) {
2316 /* Use the vlan tag on the original flow instead of
2317 * the one passed in the vlan parameter. This ensures
2318 * that we compare the vlan from before any implicit
2319 * tagging tags place. This is necessary because
2320 * dst->vlan is the final vlan, after removing implicit
2322 if (port == in_port && dst.vlan == flow_vlan) {
2323 /* Don't send out input port on same VLAN. */
2326 dst_set_add(set, &dst);
2331 mirrors &= mirrors - 1;
2336 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2337 const struct port *in_port, const struct port *out_port,
2338 tag_type *tags, struct ofpbuf *actions,
2339 uint16_t *nf_output_iface)
2341 uint16_t initial_vlan, cur_vlan;
2342 const struct dst *dst;
2346 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2348 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2350 /* Output all the packets we can without having to change the VLAN. */
2351 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2352 if (initial_vlan == 0) {
2353 initial_vlan = OFP_VLAN_NONE;
2355 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2356 if (dst->vlan != initial_vlan) {
2359 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2362 /* Then output the rest. */
2363 cur_vlan = initial_vlan;
2364 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2365 if (dst->vlan == initial_vlan) {
2368 if (dst->vlan != cur_vlan) {
2369 if (dst->vlan == OFP_VLAN_NONE) {
2370 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2373 tci = htons(dst->vlan & VLAN_VID_MASK);
2374 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2375 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2377 cur_vlan = dst->vlan;
2379 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2385 /* Returns the effective vlan of a packet, taking into account both the
2386 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2387 * the packet is untagged and -1 indicates it has an invalid header and
2388 * should be dropped. */
2389 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2390 struct port *in_port, bool have_packet)
2392 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2393 if (in_port->vlan >= 0) {
2396 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2397 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2398 "packet received on port %s configured with "
2399 "implicit VLAN %"PRIu16,
2400 br->name, vlan, in_port->name, in_port->vlan);
2404 vlan = in_port->vlan;
2406 if (!port_includes_vlan(in_port, vlan)) {
2408 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2409 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2410 "packet received on port %s not configured for "
2412 br->name, vlan, in_port->name, vlan);
2421 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2422 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2423 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2425 is_gratuitous_arp(const struct flow *flow)
2427 return (flow->dl_type == htons(ETH_TYPE_ARP)
2428 && eth_addr_is_broadcast(flow->dl_dst)
2429 && (flow->nw_proto == ARP_OP_REPLY
2430 || (flow->nw_proto == ARP_OP_REQUEST
2431 && flow->nw_src == flow->nw_dst)));
2435 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2436 struct port *in_port)
2438 struct mac_entry *mac;
2440 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2444 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2445 if (is_gratuitous_arp(flow)) {
2446 /* We don't want to learn from gratuitous ARP packets that are
2447 * reflected back over bond slaves so we lock the learning table. */
2448 if (!in_port->bond) {
2449 mac_entry_set_grat_arp_lock(mac);
2450 } else if (mac_entry_is_grat_arp_locked(mac)) {
2455 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2456 /* The log messages here could actually be useful in debugging,
2457 * so keep the rate limit relatively high. */
2458 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2459 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2460 "on port %s in VLAN %d",
2461 br->name, ETH_ADDR_ARGS(flow->dl_src),
2462 in_port->name, vlan);
2464 mac->port.p = in_port;
2465 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2469 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2470 * dropped. Returns true if they may be forwarded, false if they should be
2473 * If 'have_packet' is true, it indicates that the caller is processing a
2474 * received packet. If 'have_packet' is false, then the caller is just
2475 * revalidating an existing flow because configuration has changed. Either
2476 * way, 'have_packet' only affects logging (there is no point in logging errors
2477 * during revalidation).
2479 * Sets '*in_portp' to the input port. This will be a null pointer if
2480 * flow->in_port does not designate a known input port (in which case
2481 * is_admissible() returns false).
2483 * When returning true, sets '*vlanp' to the effective VLAN of the input
2484 * packet, as returned by flow_get_vlan().
2486 * May also add tags to '*tags', although the current implementation only does
2487 * so in one special case.
2490 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2491 tag_type *tags, int *vlanp, struct port **in_portp)
2493 struct iface *in_iface;
2494 struct port *in_port;
2497 /* Find the interface and port structure for the received packet. */
2498 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2500 /* No interface? Something fishy... */
2502 /* Odd. A few possible reasons here:
2504 * - We deleted an interface but there are still a few packets
2505 * queued up from it.
2507 * - Someone externally added an interface (e.g. with "ovs-dpctl
2508 * add-if") that we don't know about.
2510 * - Packet arrived on the local port but the local port is not
2511 * one of our bridge ports.
2513 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2515 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2516 "interface %"PRIu16, br->name, flow->in_port);
2522 *in_portp = in_port = in_iface->port;
2523 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2528 /* Drop frames for reserved multicast addresses. */
2529 if (eth_addr_is_reserved(flow->dl_dst)) {
2533 /* Drop frames on ports reserved for mirroring. */
2534 if (in_port->is_mirror_output_port) {
2536 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2537 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2538 "%s, which is reserved exclusively for mirroring",
2539 br->name, in_port->name);
2544 if (in_port->bond) {
2545 struct mac_entry *mac;
2547 switch (bond_check_admissibility(in_port->bond, in_iface,
2548 flow->dl_dst, tags)) {
2555 case BV_DROP_IF_MOVED:
2556 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2557 if (mac && mac->port.p != in_port &&
2558 (!is_gratuitous_arp(flow)
2559 || mac_entry_is_grat_arp_locked(mac))) {
2569 /* If the composed actions may be applied to any packet in the given 'flow',
2570 * returns true. Otherwise, the actions should only be applied to 'packet', or
2571 * not at all, if 'packet' was NULL. */
2573 process_flow(struct bridge *br, const struct flow *flow,
2574 const struct ofpbuf *packet, struct ofpbuf *actions,
2575 tag_type *tags, uint16_t *nf_output_iface)
2577 struct port *in_port;
2578 struct port *out_port;
2579 struct mac_entry *mac;
2582 /* Check whether we should drop packets in this flow. */
2583 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2588 /* Learn source MAC (but don't try to learn from revalidation). */
2590 update_learning_table(br, flow, vlan, in_port);
2593 /* Determine output port. */
2594 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2596 out_port = mac->port.p;
2597 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2598 /* If we are revalidating but don't have a learning entry then
2599 * eject the flow. Installing a flow that floods packets opens
2600 * up a window of time where we could learn from a packet reflected
2601 * on a bond and blackhole packets before the learning table is
2602 * updated to reflect the correct port. */
2605 out_port = FLOOD_PORT;
2608 /* Don't send packets out their input ports. */
2609 if (in_port == out_port) {
2615 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2623 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2624 struct ofpbuf *actions, tag_type *tags,
2625 uint16_t *nf_output_iface, void *br_)
2627 struct bridge *br = br_;
2629 COVERAGE_INC(bridge_process_flow);
2630 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2634 bridge_special_ofhook_cb(const struct flow *flow,
2635 const struct ofpbuf *packet, void *br_)
2637 struct iface *iface;
2638 struct bridge *br = br_;
2640 iface = iface_from_dp_ifidx(br, flow->in_port);
2642 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2643 if (iface && iface->port->lacp && packet) {
2644 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2646 lacp_process_pdu(iface->port->lacp, iface, pdu);
2656 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2657 const struct nlattr *actions,
2659 uint64_t n_bytes, void *br_)
2661 struct bridge *br = br_;
2662 const struct nlattr *a;
2663 struct port *in_port;
2668 /* Feed information from the active flows back into the learning table to
2669 * ensure that table is always in sync with what is actually flowing
2670 * through the datapath.
2672 * We test that 'tags' is nonzero to ensure that only flows that include an
2673 * OFPP_NORMAL action are used for learning. This works because
2674 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2675 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2676 update_learning_table(br, flow, vlan, in_port);
2679 /* Account for bond slave utilization. */
2680 if (!br->has_bonded_ports) {
2683 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2684 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2685 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2686 if (out_port && out_port->bond) {
2687 uint16_t vlan = (flow->vlan_tci
2688 ? vlan_tci_to_vid(flow->vlan_tci)
2690 bond_account(out_port->bond, flow, vlan, n_bytes);
2697 bridge_account_checkpoint_ofhook_cb(void *br_)
2699 struct bridge *br = br_;
2702 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2704 bond_rebalance(port->bond,
2705 ofproto_get_revalidate_set(br->ofproto));
2711 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2712 tag_type *tags, void *br_)
2714 struct bridge *br = br_;
2715 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2716 struct port *port = port_from_dp_ifidx(br, odp_port);
2721 } else if (list_is_short(&port->ifaces)) {
2724 struct iface *iface;
2726 /* Autopath does not support VLAN hashing. */
2727 iface = bond_choose_output_slave(port->bond, flow,
2728 OFP_VLAN_NONE, tags);
2729 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2732 return odp_port_to_ofp_port(ret);
2735 static struct ofhooks bridge_ofhooks = {
2736 bridge_normal_ofhook_cb,
2737 bridge_special_ofhook_cb,
2738 bridge_account_flow_ofhook_cb,
2739 bridge_account_checkpoint_ofhook_cb,
2740 bridge_autopath_ofhook_cb,
2743 /* Port functions. */
2746 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2748 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2749 struct iface *iface = iface_;
2750 uint8_t ea[ETH_ADDR_LEN];
2753 error = netdev_get_etheraddr(iface->netdev, ea);
2755 struct lacp_pdu *packet_pdu;
2756 struct ofpbuf packet;
2758 ofpbuf_init(&packet, 0);
2759 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2760 sizeof *packet_pdu);
2762 error = netdev_send(iface->netdev, &packet);
2764 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2765 "(%s)", iface->port->name, iface->name,
2768 ofpbuf_uninit(&packet);
2770 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2771 "%s (%s)", iface->port->name, iface->name,
2777 port_run(struct port *port)
2780 lacp_run(port->lacp, lacp_send_pdu_cb);
2784 struct iface *iface;
2786 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2787 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2788 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2791 bond_run(port->bond,
2792 ofproto_get_revalidate_set(port->bridge->ofproto),
2793 lacp_negotiated(port->lacp));
2794 if (bond_should_send_learning_packets(port->bond)) {
2795 port_send_learning_packets(port);
2801 port_wait(struct port *port)
2804 lacp_wait(port->lacp);
2808 bond_wait(port->bond);
2812 static struct port *
2813 port_create(struct bridge *br, const char *name)
2817 port = xzalloc(sizeof *port);
2820 port->trunks = NULL;
2821 port->name = xstrdup(name);
2822 list_init(&port->ifaces);
2824 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2826 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2833 get_port_other_config(const struct ovsrec_port *port, const char *key,
2834 const char *default_value)
2838 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2840 return value ? value : default_value;
2844 get_interface_other_config(const struct ovsrec_interface *iface,
2845 const char *key, const char *default_value)
2849 value = get_ovsrec_key_value(&iface->header_,
2850 &ovsrec_interface_col_other_config, key);
2851 return value ? value : default_value;
2855 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2857 struct iface *iface, *next;
2858 struct sset new_ifaces;
2861 /* Collect list of new interfaces. */
2862 sset_init(&new_ifaces);
2863 for (i = 0; i < cfg->n_interfaces; i++) {
2864 const char *name = cfg->interfaces[i]->name;
2865 sset_add(&new_ifaces, name);
2868 /* Get rid of deleted interfaces. */
2869 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2870 if (!sset_contains(&new_ifaces, iface->name)) {
2871 iface_destroy(iface);
2875 sset_destroy(&new_ifaces);
2878 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2879 * to revalidate every flow. */
2881 port_flush_macs(struct port *port)
2883 struct bridge *br = port->bridge;
2884 struct mac_learning *ml = br->ml;
2885 struct mac_entry *mac, *next_mac;
2888 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2889 if (mac->port.p == port) {
2890 mac_learning_expire(ml, mac);
2896 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2898 struct sset new_ifaces;
2899 bool need_flush = false;
2900 unsigned long *trunks;
2907 /* Add new interfaces and update 'cfg' member of existing ones. */
2908 sset_init(&new_ifaces);
2909 for (i = 0; i < cfg->n_interfaces; i++) {
2910 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2911 struct iface *iface;
2913 if (!sset_add(&new_ifaces, if_cfg->name)) {
2914 VLOG_WARN("port %s: %s specified twice as port interface",
2915 port->name, if_cfg->name);
2916 iface_set_ofport(if_cfg, -1);
2920 iface = iface_lookup(port->bridge, if_cfg->name);
2922 if (iface->port != port) {
2923 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2925 port->bridge->name, if_cfg->name, iface->port->name);
2928 iface->cfg = if_cfg;
2930 iface = iface_create(port, if_cfg);
2933 /* Determine interface type. The local port always has type
2934 * "internal". Other ports take their type from the database and
2935 * default to "system" if none is specified. */
2936 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2937 : if_cfg->type[0] ? if_cfg->type
2940 sset_destroy(&new_ifaces);
2945 if (list_is_short(&port->ifaces)) {
2947 if (vlan >= 0 && vlan <= 4095) {
2948 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2953 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2954 * they even work as-is. But they have not been tested. */
2955 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2959 if (port->vlan != vlan) {
2964 /* Get trunked VLANs. */
2966 if (vlan < 0 && cfg->n_trunks) {
2967 trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
2969 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2972 } else if (vlan >= 0 && cfg->n_trunks) {
2973 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2976 if (!vlan_bitmap_equal(trunks, port->trunks)) {
2979 bitmap_free(port->trunks);
2980 port->trunks = trunks;
2983 port_flush_macs(port);
2988 port_destroy(struct port *port)
2991 struct bridge *br = port->bridge;
2992 struct iface *iface, *next;
2995 for (i = 0; i < MAX_MIRRORS; i++) {
2996 struct mirror *m = br->mirrors[i];
2997 if (m && m->out_port == port) {
3002 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3003 iface_destroy(iface);
3006 hmap_remove(&br->ports, &port->hmap_node);
3008 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
3010 bond_destroy(port->bond);
3011 lacp_destroy(port->lacp);
3012 port_flush_macs(port);
3014 bitmap_free(port->trunks);
3020 static struct port *
3021 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3023 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
3024 return iface ? iface->port : NULL;
3027 static struct port *
3028 port_lookup(const struct bridge *br, const char *name)
3032 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3034 if (!strcmp(port->name, name)) {
3042 enable_lacp(struct port *port, bool *activep)
3044 if (!port->cfg->lacp) {
3045 /* XXX when LACP implementation has been sufficiently tested, enable by
3046 * default and make active on bonded ports. */
3048 } else if (!strcmp(port->cfg->lacp, "off")) {
3050 } else if (!strcmp(port->cfg->lacp, "active")) {
3053 } else if (!strcmp(port->cfg->lacp, "passive")) {
3057 VLOG_WARN("port %s: unknown LACP mode %s",
3058 port->name, port->cfg->lacp);
3064 iface_reconfigure_lacp(struct iface *iface)
3066 struct lacp_slave_settings s;
3067 int priority, portid;
3069 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
3070 priority = atoi(get_interface_other_config(iface->cfg,
3071 "lacp-port-priority", "0"));
3073 if (portid <= 0 || portid > UINT16_MAX) {
3074 portid = iface->dp_ifidx;
3077 if (priority <= 0 || priority > UINT16_MAX) {
3078 priority = UINT16_MAX;
3081 s.name = iface->name;
3083 s.priority = priority;
3084 lacp_slave_register(iface->port->lacp, iface, &s);
3088 port_reconfigure_lacp(struct port *port)
3090 static struct lacp_settings s;
3091 struct iface *iface;
3092 uint8_t sysid[ETH_ADDR_LEN];
3093 const char *sysid_str;
3094 const char *lacp_time;
3095 long long int custom_time;
3098 if (!enable_lacp(port, &s.active)) {
3099 lacp_destroy(port->lacp);
3104 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3105 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3106 memcpy(s.id, sysid, ETH_ADDR_LEN);
3108 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3111 s.name = port->name;
3113 /* Prefer bondable links if unspecified. */
3114 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3116 s.priority = (priority > 0 && priority <= UINT16_MAX
3118 : UINT16_MAX - !list_is_short(&port->ifaces));
3120 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3124 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3125 custom_time = atoi(lacp_time);
3126 if (!strcmp(lacp_time, "fast")) {
3127 s.lacp_time = LACP_TIME_FAST;
3128 } else if (!strcmp(lacp_time, "slow")) {
3129 s.lacp_time = LACP_TIME_SLOW;
3130 } else if (custom_time > 0) {
3131 s.lacp_time = LACP_TIME_CUSTOM;
3132 s.custom_time = custom_time;
3134 s.lacp_time = LACP_TIME_SLOW;
3138 port->lacp = lacp_create();
3141 lacp_configure(port->lacp, &s);
3143 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3144 iface_reconfigure_lacp(iface);
3149 port_reconfigure_bond(struct port *port)
3151 struct bond_settings s;
3152 const char *detect_s;
3153 struct iface *iface;
3155 if (list_is_short(&port->ifaces)) {
3156 /* Not a bonded port. */
3157 bond_destroy(port->bond);
3162 port->bridge->has_bonded_ports = true;
3164 s.name = port->name;
3166 if (port->cfg->bond_mode
3167 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3168 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3169 port->name, port->cfg->bond_mode,
3170 bond_mode_to_string(s.balance));
3173 s.detect = BLSM_CARRIER;
3174 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3175 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3176 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3178 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3181 s.miimon_interval = atoi(
3182 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3183 if (s.miimon_interval < 100) {
3184 s.miimon_interval = 100;
3187 s.up_delay = MAX(0, port->cfg->bond_updelay);
3188 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3189 s.rebalance_interval = atoi(
3190 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3191 if (s.rebalance_interval < 1000) {
3192 s.rebalance_interval = 1000;
3195 s.fake_iface = port->cfg->bond_fake_iface;
3198 port->bond = bond_create(&s);
3200 if (bond_reconfigure(port->bond, &s)) {
3201 bridge_flush(port->bridge);
3205 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3206 uint16_t stable_id = (port->lacp
3207 ? lacp_slave_get_port_id(port->lacp, iface)
3209 bond_slave_register(iface->port->bond, iface, stable_id,
3215 port_send_learning_packets(struct port *port)
3217 struct bridge *br = port->bridge;
3218 int error, n_packets, n_errors;
3219 struct mac_entry *e;
3221 error = n_packets = n_errors = 0;
3222 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3223 if (e->port.p != port) {
3224 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3234 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3235 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3236 "packets, last error was: %s",
3237 port->name, n_errors, n_packets, strerror(error));
3239 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3240 port->name, n_packets);
3244 /* Interface functions. */
3246 static struct iface *
3247 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3249 struct bridge *br = port->bridge;
3250 struct iface *iface;
3251 char *name = if_cfg->name;
3253 iface = xzalloc(sizeof *iface);
3255 iface->name = xstrdup(name);
3256 iface->dp_ifidx = -1;
3257 iface->tag = tag_create_random();
3258 iface->netdev = NULL;
3259 iface->cfg = if_cfg;
3261 shash_add_assert(&br->iface_by_name, iface->name, iface);
3263 list_push_back(&port->ifaces, &iface->port_elem);
3265 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3273 iface_destroy(struct iface *iface)
3276 struct port *port = iface->port;
3277 struct bridge *br = port->bridge;
3280 bond_slave_unregister(port->bond, iface);
3284 lacp_slave_unregister(port->lacp, iface);
3287 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3289 if (iface->dp_ifidx >= 0) {
3290 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3293 list_remove(&iface->port_elem);
3295 netdev_close(iface->netdev);
3300 bridge_flush(port->bridge);
3304 static struct iface *
3305 iface_lookup(const struct bridge *br, const char *name)
3307 return shash_find_data(&br->iface_by_name, name);
3310 static struct iface *
3311 iface_find(const char *name)
3313 const struct bridge *br;
3315 LIST_FOR_EACH (br, node, &all_bridges) {
3316 struct iface *iface = iface_lookup(br, name);
3325 static struct iface *
3326 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3328 struct iface *iface;
3330 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3331 hash_int(dp_ifidx, 0), &br->ifaces) {
3332 if (iface->dp_ifidx == dp_ifidx) {
3339 /* Set Ethernet address of 'iface', if one is specified in the configuration
3342 iface_set_mac(struct iface *iface)
3344 uint8_t ea[ETH_ADDR_LEN];
3346 if (!strcmp(iface->type, "internal")
3347 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3348 if (iface->dp_ifidx == ODPP_LOCAL) {
3349 VLOG_ERR("interface %s: ignoring mac in Interface record "
3350 "(use Bridge record to set local port's mac)",
3352 } else if (eth_addr_is_multicast(ea)) {
3353 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3356 int error = netdev_set_etheraddr(iface->netdev, ea);
3358 VLOG_ERR("interface %s: setting MAC failed (%s)",
3359 iface->name, strerror(error));
3365 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3367 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3369 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3370 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3374 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3376 * The value strings in '*shash' are taken directly from values[], not copied,
3377 * so the caller should not modify or free them. */
3379 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3380 struct shash *shash)
3385 for (i = 0; i < n; i++) {
3386 shash_add(shash, keys[i], values[i]);
3390 /* Creates 'keys' and 'values' arrays from 'shash'.
3392 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3393 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3394 * are populated with with strings taken directly from 'shash' and thus have
3395 * the same ownership of the key-value pairs in shash.
3398 shash_to_ovs_idl_map(struct shash *shash,
3399 char ***keys, char ***values, size_t *n)
3403 struct shash_node *sn;
3405 count = shash_count(shash);
3407 k = xmalloc(count * sizeof *k);
3408 v = xmalloc(count * sizeof *v);
3411 SHASH_FOR_EACH(sn, shash) {
3422 struct iface_delete_queues_cbdata {
3423 struct netdev *netdev;
3424 const struct ovsdb_datum *queues;
3428 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3430 union ovsdb_atom atom;
3432 atom.integer = target;
3433 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3437 iface_delete_queues(unsigned int queue_id,
3438 const struct shash *details OVS_UNUSED, void *cbdata_)
3440 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3442 if (!queue_ids_include(cbdata->queues, queue_id)) {
3443 netdev_delete_queue(cbdata->netdev, queue_id);
3448 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3450 if (!qos || qos->type[0] == '\0') {
3451 netdev_set_qos(iface->netdev, NULL, NULL);
3453 struct iface_delete_queues_cbdata cbdata;
3454 struct shash details;
3457 /* Configure top-level Qos for 'iface'. */
3458 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3459 qos->n_other_config, &details);
3460 netdev_set_qos(iface->netdev, qos->type, &details);
3461 shash_destroy(&details);
3463 /* Deconfigure queues that were deleted. */
3464 cbdata.netdev = iface->netdev;
3465 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3467 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3469 /* Configure queues for 'iface'. */
3470 for (i = 0; i < qos->n_queues; i++) {
3471 const struct ovsrec_queue *queue = qos->value_queues[i];
3472 unsigned int queue_id = qos->key_queues[i];
3474 shash_from_ovs_idl_map(queue->key_other_config,
3475 queue->value_other_config,
3476 queue->n_other_config, &details);
3477 netdev_set_queue(iface->netdev, queue_id, &details);
3478 shash_destroy(&details);
3484 iface_update_cfm(struct iface *iface)
3488 uint16_t *remote_mps;
3489 struct ovsrec_monitor *mon;
3490 uint8_t maid[CCM_MAID_LEN];
3492 mon = iface->cfg->monitor;
3495 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3499 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3500 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3504 cfm.mpid = mon->mpid;
3505 cfm.interval = mon->interval ? *mon->interval : 1000;
3507 memcpy(cfm.maid, maid, sizeof cfm.maid);
3509 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3510 for(i = 0; i < mon->n_remote_mps; i++) {
3511 remote_mps[i] = mon->remote_mps[i]->mpid;
3514 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3515 &cfm, remote_mps, mon->n_remote_mps);
3519 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3520 * how 'iface''s port is configured.
3522 * Returns true if 'iface' is up, false otherwise. */
3524 iface_get_carrier(const struct iface *iface)
3527 return netdev_get_carrier(iface->netdev);
3530 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3531 * instead of obtaining it from the database. */
3533 iface_is_synthetic(const struct iface *iface)
3535 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3538 /* Port mirroring. */
3540 static struct mirror *
3541 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3545 for (i = 0; i < MAX_MIRRORS; i++) {
3546 struct mirror *m = br->mirrors[i];
3547 if (m && uuid_equals(uuid, &m->uuid)) {
3555 mirror_reconfigure(struct bridge *br)
3557 unsigned long *rspan_vlans;
3561 /* Get rid of deleted mirrors. */
3562 for (i = 0; i < MAX_MIRRORS; i++) {
3563 struct mirror *m = br->mirrors[i];
3565 const struct ovsdb_datum *mc;
3566 union ovsdb_atom atom;
3568 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3569 atom.uuid = br->mirrors[i]->uuid;
3570 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3576 /* Add new mirrors and reconfigure existing ones. */
3577 for (i = 0; i < br->cfg->n_mirrors; i++) {
3578 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3579 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3581 mirror_reconfigure_one(m, cfg);
3583 mirror_create(br, cfg);
3587 /* Update port reserved status. */
3588 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3589 port->is_mirror_output_port = false;
3591 for (i = 0; i < MAX_MIRRORS; i++) {
3592 struct mirror *m = br->mirrors[i];
3593 if (m && m->out_port) {
3594 m->out_port->is_mirror_output_port = true;
3598 /* Update flooded vlans (for RSPAN). */
3600 if (br->cfg->n_flood_vlans) {
3601 rspan_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3602 br->cfg->n_flood_vlans);
3604 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3606 mac_learning_flush(br->ml);
3612 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3617 for (i = 0; ; i++) {
3618 if (i >= MAX_MIRRORS) {
3619 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3620 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3623 if (!br->mirrors[i]) {
3628 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3630 mac_learning_flush(br->ml);
3632 br->mirrors[i] = m = xzalloc(sizeof *m);
3633 m->uuid = cfg->header_.uuid;
3636 m->name = xstrdup(cfg->name);
3637 sset_init(&m->src_ports);
3638 sset_init(&m->dst_ports);
3644 mirror_reconfigure_one(m, cfg);
3648 mirror_destroy(struct mirror *m)
3651 struct bridge *br = m->bridge;
3654 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3655 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3656 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3659 sset_destroy(&m->src_ports);
3660 sset_destroy(&m->dst_ports);
3663 m->bridge->mirrors[m->idx] = NULL;
3668 mac_learning_flush(br->ml);
3673 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3678 for (i = 0; i < n_ports; i++) {
3679 const char *name = ports[i]->name;
3680 if (port_lookup(m->bridge, name)) {
3681 sset_add(names, name);
3683 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3684 "port %s", m->bridge->name, m->name, name);
3690 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3696 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3698 for (i = 0; i < cfg->n_select_vlan; i++) {
3699 int64_t vlan = cfg->select_vlan[i];
3700 if (vlan < 0 || vlan > 4095) {
3701 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3702 m->bridge->name, m->name, vlan);
3704 (*vlans)[n_vlans++] = vlan;
3711 vlan_is_mirrored(const struct mirror *m, int vlan)
3715 for (i = 0; i < m->n_vlans; i++) {
3716 if (m->vlans[i] == vlan) {
3724 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3726 struct sset src_ports, dst_ports;
3727 mirror_mask_t mirror_bit;
3728 struct port *out_port;
3735 if (strcmp(cfg->name, m->name)) {
3737 m->name = xstrdup(cfg->name);
3740 /* Get output port. */
3741 if (cfg->output_port) {
3742 out_port = port_lookup(m->bridge, cfg->output_port->name);
3744 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3745 m->bridge->name, m->name);
3751 if (cfg->output_vlan) {
3752 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3753 "output vlan; ignoring output vlan",
3754 m->bridge->name, m->name);
3756 } else if (cfg->output_vlan) {
3758 out_vlan = *cfg->output_vlan;
3760 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3761 m->bridge->name, m->name);
3766 sset_init(&src_ports);
3767 sset_init(&dst_ports);
3768 if (cfg->select_all) {
3769 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3770 sset_add(&src_ports, port->name);
3771 sset_add(&dst_ports, port->name);
3776 /* Get ports, and drop duplicates and ports that don't exist. */
3777 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3779 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3782 /* Get all the vlans, and drop duplicate and invalid vlans. */
3783 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3786 /* Update mirror data. */
3787 if (!sset_equals(&m->src_ports, &src_ports)
3788 || !sset_equals(&m->dst_ports, &dst_ports)
3789 || m->n_vlans != n_vlans
3790 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3791 || m->out_port != out_port
3792 || m->out_vlan != out_vlan) {
3793 bridge_flush(m->bridge);
3794 mac_learning_flush(m->bridge->ml);
3796 sset_swap(&m->src_ports, &src_ports);
3797 sset_swap(&m->dst_ports, &dst_ports);
3800 m->n_vlans = n_vlans;
3801 m->out_port = out_port;
3802 m->out_vlan = out_vlan;
3805 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3806 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3807 if (sset_contains(&m->src_ports, port->name)) {
3808 port->src_mirrors |= mirror_bit;
3810 port->src_mirrors &= ~mirror_bit;
3813 if (sset_contains(&m->dst_ports, port->name)) {
3814 port->dst_mirrors |= mirror_bit;
3816 port->dst_mirrors &= ~mirror_bit;
3821 sset_destroy(&src_ports);
3822 sset_destroy(&dst_ports);