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 /* Callback for iterate_and_prune_ifaces(). */
444 set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
445 void *aux OVS_UNUSED)
447 /* Set policing attributes. */
448 netdev_set_policing(iface->netdev,
449 iface->cfg->ingress_policing_rate,
450 iface->cfg->ingress_policing_burst);
452 /* Set MAC address of internal interfaces other than the local
454 iface_set_mac(iface);
459 /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
460 * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
461 * deletes from 'br' any ports that no longer have any interfaces. */
463 iterate_and_prune_ifaces(struct bridge *br,
464 bool (*cb)(struct bridge *, struct iface *,
468 struct port *port, *next_port;
470 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
471 struct iface *iface, *next_iface;
473 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
474 if (!cb(br, iface, aux)) {
475 iface_set_ofport(iface->cfg, -1);
476 iface_destroy(iface);
480 if (list_is_empty(&port->ifaces)) {
481 VLOG_WARN("%s port has no interfaces, dropping", port->name);
487 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
488 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
489 * responsible for freeing '*managersp' (with free()).
491 * You may be asking yourself "why does ovs-vswitchd care?", because
492 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
493 * should not be and in fact is not directly involved in that. But
494 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
495 * it has to tell in-band control where the managers are to enable that.
496 * (Thus, only managers connected in-band are collected.)
499 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
500 struct sockaddr_in **managersp, size_t *n_managersp)
502 struct sockaddr_in *managers = NULL;
503 size_t n_managers = 0;
507 /* Collect all of the potential targets from the "targets" columns of the
508 * rows pointed to by "manager_options", excluding any that are
511 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
512 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
514 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
515 sset_find_and_delete(&targets, m->target);
517 sset_add(&targets, m->target);
521 /* Now extract the targets' IP addresses. */
522 if (!sset_is_empty(&targets)) {
525 managers = xmalloc(sset_count(&targets) * sizeof *managers);
526 SSET_FOR_EACH (target, &targets) {
527 struct sockaddr_in *sin = &managers[n_managers];
529 if ((!strncmp(target, "tcp:", 4)
530 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
531 (!strncmp(target, "ssl:", 4)
532 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
537 sset_destroy(&targets);
539 *managersp = managers;
540 *n_managersp = n_managers;
544 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
546 struct shash old_br, new_br;
547 struct shash_node *node;
548 struct bridge *br, *next;
549 struct sockaddr_in *managers;
552 int sflow_bridge_number;
554 COVERAGE_INC(bridge_reconfigure);
556 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
558 /* Collect old and new bridges. */
561 LIST_FOR_EACH (br, node, &all_bridges) {
562 shash_add(&old_br, br->name, br);
564 for (i = 0; i < ovs_cfg->n_bridges; i++) {
565 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
566 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
567 VLOG_WARN("more than one bridge named %s", br_cfg->name);
571 /* Get rid of deleted bridges and add new bridges. */
572 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
573 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
580 SHASH_FOR_EACH (node, &new_br) {
581 const char *br_name = node->name;
582 const struct ovsrec_bridge *br_cfg = node->data;
583 br = shash_find_data(&old_br, br_name);
585 /* If the bridge datapath type has changed, we need to tear it
586 * down and recreate. */
587 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
589 bridge_create(br_cfg);
592 bridge_create(br_cfg);
595 shash_destroy(&old_br);
596 shash_destroy(&new_br);
598 /* Reconfigure all bridges. */
599 LIST_FOR_EACH (br, node, &all_bridges) {
600 bridge_reconfigure_one(br);
603 /* Add and delete ports on all datapaths.
605 * The kernel will reject any attempt to add a given port to a datapath if
606 * that port already belongs to a different datapath, so we must do all
607 * port deletions before any port additions. */
608 LIST_FOR_EACH (br, node, &all_bridges) {
609 struct dpif_port_dump dump;
610 struct shash want_ifaces;
611 struct dpif_port dpif_port;
613 bridge_get_all_ifaces(br, &want_ifaces);
614 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
615 if (!shash_find(&want_ifaces, dpif_port.name)
616 && strcmp(dpif_port.name, br->name)) {
617 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
619 VLOG_WARN("failed to remove %s interface from %s: %s",
620 dpif_port.name, dpif_name(br->dpif),
625 shash_destroy(&want_ifaces);
627 LIST_FOR_EACH (br, node, &all_bridges) {
628 struct shash cur_ifaces, want_ifaces;
629 struct dpif_port_dump dump;
630 struct dpif_port dpif_port;
632 /* Get the set of interfaces currently in this datapath. */
633 shash_init(&cur_ifaces);
634 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
635 struct dpif_port *port_info = xmalloc(sizeof *port_info);
636 dpif_port_clone(port_info, &dpif_port);
637 shash_add(&cur_ifaces, dpif_port.name, port_info);
640 /* Get the set of interfaces we want on this datapath. */
641 bridge_get_all_ifaces(br, &want_ifaces);
643 hmap_clear(&br->ifaces);
644 SHASH_FOR_EACH (node, &want_ifaces) {
645 const char *if_name = node->name;
646 struct iface *iface = node->data;
647 struct dpif_port *dpif_port;
651 type = iface ? iface->type : "internal";
652 dpif_port = shash_find_data(&cur_ifaces, if_name);
654 /* If we have a port or a netdev already, and it's not the type we
655 * want, then delete the port (if any) and close the netdev (if
657 if ((dpif_port && strcmp(dpif_port->type, type))
658 || (iface && iface->netdev
659 && strcmp(type, netdev_get_type(iface->netdev)))) {
661 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
668 if (iface->port->bond) {
669 /* The bond has a pointer to the netdev, so remove it
670 * from the bond before closing the netdev. The slave
671 * will get added back to the bond later, after a new
672 * netdev is available. */
673 bond_slave_unregister(iface->port->bond, iface);
675 netdev_close(iface->netdev);
676 iface->netdev = NULL;
680 /* If the port doesn't exist or we don't have the netdev open,
681 * we need to do more work. */
682 if (!dpif_port || (iface && !iface->netdev)) {
683 struct netdev_options options;
684 struct netdev *netdev;
687 /* First open the network device. */
688 options.name = if_name;
690 options.args = &args;
691 options.ethertype = NETDEV_ETH_TYPE_NONE;
695 shash_from_ovs_idl_map(iface->cfg->key_options,
696 iface->cfg->value_options,
697 iface->cfg->n_options, &args);
699 error = netdev_open(&options, &netdev);
700 shash_destroy(&args);
703 VLOG_WARN("could not open network device %s (%s)",
704 if_name, strerror(error));
708 /* Then add the port if we haven't already. */
710 error = dpif_port_add(br->dpif, netdev, NULL);
712 netdev_close(netdev);
713 if (error == EFBIG) {
714 VLOG_ERR("ran out of valid port numbers on %s",
715 dpif_name(br->dpif));
718 VLOG_WARN("failed to add %s interface to %s: %s",
719 if_name, dpif_name(br->dpif),
726 /* Update 'iface'. */
728 iface->netdev = netdev;
730 } else if (iface && iface->netdev) {
734 shash_from_ovs_idl_map(iface->cfg->key_options,
735 iface->cfg->value_options,
736 iface->cfg->n_options, &args);
737 netdev_set_config(iface->netdev, &args);
738 shash_destroy(&args);
741 shash_destroy(&want_ifaces);
743 SHASH_FOR_EACH (node, &cur_ifaces) {
744 struct dpif_port *port_info = node->data;
745 dpif_port_destroy(port_info);
748 shash_destroy(&cur_ifaces);
750 sflow_bridge_number = 0;
751 LIST_FOR_EACH (br, node, &all_bridges) {
752 uint8_t ea[ETH_ADDR_LEN];
754 struct iface *local_iface;
755 struct iface *hw_addr_iface;
758 bridge_fetch_dp_ifaces(br);
760 /* Delete interfaces that cannot be opened.
762 * From this point forward we are guaranteed that every "struct iface"
763 * has nonnull 'netdev' and correct 'dp_ifidx'. */
764 iterate_and_prune_ifaces(br, check_iface, NULL);
766 /* Pick local port hardware address, datapath ID. */
767 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
768 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
770 int error = netdev_set_etheraddr(local_iface->netdev, ea);
772 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
773 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
774 "Ethernet address: %s",
775 br->name, strerror(error));
778 memcpy(br->ea, ea, ETH_ADDR_LEN);
780 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
781 ofproto_set_datapath_id(br->ofproto, dpid);
783 dpid_string = xasprintf("%016"PRIx64, dpid);
784 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
787 /* Set NetFlow configuration on this bridge. */
788 if (br->cfg->netflow) {
789 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
790 struct netflow_options opts;
792 memset(&opts, 0, sizeof opts);
794 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
795 if (nf_cfg->engine_type) {
796 opts.engine_type = *nf_cfg->engine_type;
798 if (nf_cfg->engine_id) {
799 opts.engine_id = *nf_cfg->engine_id;
802 opts.active_timeout = nf_cfg->active_timeout;
803 if (!opts.active_timeout) {
804 opts.active_timeout = -1;
805 } else if (opts.active_timeout < 0) {
806 VLOG_WARN("bridge %s: active timeout interval set to negative "
807 "value, using default instead (%d seconds)", br->name,
808 NF_ACTIVE_TIMEOUT_DEFAULT);
809 opts.active_timeout = -1;
812 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
813 if (opts.add_id_to_iface) {
814 if (opts.engine_id > 0x7f) {
815 VLOG_WARN("bridge %s: netflow port mangling may conflict "
816 "with another vswitch, choose an engine id less "
817 "than 128", br->name);
819 if (hmap_count(&br->ports) > 508) {
820 VLOG_WARN("bridge %s: netflow port mangling will conflict "
821 "with another port when more than 508 ports are "
826 sset_init(&opts.collectors);
827 sset_add_array(&opts.collectors,
828 nf_cfg->targets, nf_cfg->n_targets);
829 if (ofproto_set_netflow(br->ofproto, &opts)) {
830 VLOG_ERR("bridge %s: problem setting netflow collectors",
833 sset_destroy(&opts.collectors);
835 ofproto_set_netflow(br->ofproto, NULL);
838 /* Set sFlow configuration on this bridge. */
839 if (br->cfg->sflow) {
840 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
841 struct ovsrec_controller **controllers;
842 struct ofproto_sflow_options oso;
843 size_t n_controllers;
845 memset(&oso, 0, sizeof oso);
847 sset_init(&oso.targets);
848 sset_add_array(&oso.targets,
849 sflow_cfg->targets, sflow_cfg->n_targets);
851 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
852 if (sflow_cfg->sampling) {
853 oso.sampling_rate = *sflow_cfg->sampling;
856 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
857 if (sflow_cfg->polling) {
858 oso.polling_interval = *sflow_cfg->polling;
861 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
862 if (sflow_cfg->header) {
863 oso.header_len = *sflow_cfg->header;
866 oso.sub_id = sflow_bridge_number++;
867 oso.agent_device = sflow_cfg->agent;
869 oso.control_ip = NULL;
870 n_controllers = bridge_get_controllers(br, &controllers);
871 for (i = 0; i < n_controllers; i++) {
872 if (controllers[i]->local_ip) {
873 oso.control_ip = controllers[i]->local_ip;
877 ofproto_set_sflow(br->ofproto, &oso);
879 sset_destroy(&oso.targets);
881 ofproto_set_sflow(br->ofproto, NULL);
884 /* Update the controller and related settings. It would be more
885 * straightforward to call this from bridge_reconfigure_one(), but we
886 * can't do it there for two reasons. First, and most importantly, at
887 * that point we don't know the dp_ifidx of any interfaces that have
888 * been added to the bridge (because we haven't actually added them to
889 * the datapath). Second, at that point we haven't set the datapath ID
890 * yet; when a controller is configured, resetting the datapath ID will
891 * immediately disconnect from the controller, so it's better to set
892 * the datapath ID before the controller. */
893 bridge_reconfigure_remotes(br, managers, n_managers);
895 LIST_FOR_EACH (br, node, &all_bridges) {
898 br->has_bonded_ports = false;
899 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
902 port_reconfigure_lacp(port);
903 port_reconfigure_bond(port);
905 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
906 iface_update_qos(iface, port->cfg->qos);
910 LIST_FOR_EACH (br, node, &all_bridges) {
911 iterate_and_prune_ifaces(br, set_iface_properties, NULL);
914 /* Some reconfiguration operations require the bridge to have been run at
916 LIST_FOR_EACH (br, node, &all_bridges) {
921 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
922 iface_update_cfm(iface);
928 /* ovs-vswitchd has completed initialization, so allow the process that
929 * forked us to exit successfully. */
930 daemonize_complete();
934 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
935 const struct ovsdb_idl_column *column,
938 const struct ovsdb_datum *datum;
939 union ovsdb_atom atom;
942 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
943 atom.string = (char *) key;
944 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
945 return idx == UINT_MAX ? NULL : datum->values[idx].string;
949 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
951 return get_ovsrec_key_value(&br_cfg->header_,
952 &ovsrec_bridge_col_other_config, key);
956 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
957 struct iface **hw_addr_iface)
963 *hw_addr_iface = NULL;
965 /* Did the user request a particular MAC? */
966 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
967 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
968 if (eth_addr_is_multicast(ea)) {
969 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
970 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
971 } else if (eth_addr_is_zero(ea)) {
972 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
978 /* Otherwise choose the minimum non-local MAC address among all of the
980 memset(ea, 0xff, ETH_ADDR_LEN);
981 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
982 uint8_t iface_ea[ETH_ADDR_LEN];
983 struct iface *candidate;
986 /* Mirror output ports don't participate. */
987 if (port->is_mirror_output_port) {
991 /* Choose the MAC address to represent the port. */
993 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
994 /* Find the interface with this Ethernet address (if any) so that
995 * we can provide the correct devname to the caller. */
996 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
997 uint8_t candidate_ea[ETH_ADDR_LEN];
998 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
999 && eth_addr_equals(iface_ea, candidate_ea)) {
1004 /* Choose the interface whose MAC address will represent the port.
1005 * The Linux kernel bonding code always chooses the MAC address of
1006 * the first slave added to a bond, and the Fedora networking
1007 * scripts always add slaves to a bond in alphabetical order, so
1008 * for compatibility we choose the interface with the name that is
1009 * first in alphabetical order. */
1010 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1011 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1016 /* The local port doesn't count (since we're trying to choose its
1017 * MAC address anyway). */
1018 if (iface->dp_ifidx == ODPP_LOCAL) {
1023 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1025 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1026 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
1027 iface->name, strerror(error));
1032 /* Compare against our current choice. */
1033 if (!eth_addr_is_multicast(iface_ea) &&
1034 !eth_addr_is_local(iface_ea) &&
1035 !eth_addr_is_reserved(iface_ea) &&
1036 !eth_addr_is_zero(iface_ea) &&
1037 eth_addr_compare_3way(iface_ea, ea) < 0)
1039 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1040 *hw_addr_iface = iface;
1043 if (eth_addr_is_multicast(ea)) {
1044 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1045 *hw_addr_iface = NULL;
1046 VLOG_WARN("bridge %s: using default bridge Ethernet "
1047 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1049 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1050 br->name, ETH_ADDR_ARGS(ea));
1054 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1055 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1056 * an interface on 'br', then that interface must be passed in as
1057 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1058 * 'hw_addr_iface' must be passed in as a null pointer. */
1060 bridge_pick_datapath_id(struct bridge *br,
1061 const uint8_t bridge_ea[ETH_ADDR_LEN],
1062 struct iface *hw_addr_iface)
1065 * The procedure for choosing a bridge MAC address will, in the most
1066 * ordinary case, also choose a unique MAC that we can use as a datapath
1067 * ID. In some special cases, though, multiple bridges will end up with
1068 * the same MAC address. This is OK for the bridges, but it will confuse
1069 * the OpenFlow controller, because each datapath needs a unique datapath
1072 * Datapath IDs must be unique. It is also very desirable that they be
1073 * stable from one run to the next, so that policy set on a datapath
1076 const char *datapath_id;
1079 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1080 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1084 if (hw_addr_iface) {
1086 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1088 * A bridge whose MAC address is taken from a VLAN network device
1089 * (that is, a network device created with vconfig(8) or similar
1090 * tool) will have the same MAC address as a bridge on the VLAN
1091 * device's physical network device.
1093 * Handle this case by hashing the physical network device MAC
1094 * along with the VLAN identifier.
1096 uint8_t buf[ETH_ADDR_LEN + 2];
1097 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1098 buf[ETH_ADDR_LEN] = vlan >> 8;
1099 buf[ETH_ADDR_LEN + 1] = vlan;
1100 return dpid_from_hash(buf, sizeof buf);
1103 * Assume that this bridge's MAC address is unique, since it
1104 * doesn't fit any of the cases we handle specially.
1109 * A purely internal bridge, that is, one that has no non-virtual
1110 * network devices on it at all, is more difficult because it has no
1111 * natural unique identifier at all.
1113 * When the host is a XenServer, we handle this case by hashing the
1114 * host's UUID with the name of the bridge. Names of bridges are
1115 * persistent across XenServer reboots, although they can be reused if
1116 * an internal network is destroyed and then a new one is later
1117 * created, so this is fairly effective.
1119 * When the host is not a XenServer, we punt by using a random MAC
1120 * address on each run.
1122 const char *host_uuid = xenserver_get_host_uuid();
1124 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1125 dpid = dpid_from_hash(combined, strlen(combined));
1131 return eth_addr_to_uint64(bridge_ea);
1135 dpid_from_hash(const void *data, size_t n)
1137 uint8_t hash[SHA1_DIGEST_SIZE];
1139 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1140 sha1_bytes(data, n, hash);
1141 eth_addr_mark_random(hash);
1142 return eth_addr_to_uint64(hash);
1146 iface_refresh_status(struct iface *iface)
1150 enum netdev_flags flags;
1157 if (iface_is_synthetic(iface)) {
1163 if (!netdev_get_status(iface->netdev, &sh)) {
1165 char **keys, **values;
1167 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1168 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1173 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1176 shash_destroy_free_data(&sh);
1178 error = netdev_get_flags(iface->netdev, &flags);
1180 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1183 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1186 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1188 ovsrec_interface_set_duplex(iface->cfg,
1189 netdev_features_is_full_duplex(current)
1191 /* warning: uint64_t -> int64_t conversion */
1192 bps = netdev_features_to_bps(current);
1193 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1196 ovsrec_interface_set_duplex(iface->cfg, NULL);
1197 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1201 ovsrec_interface_set_link_state(iface->cfg,
1202 iface_get_carrier(iface) ? "up" : "down");
1204 error = netdev_get_mtu(iface->netdev, &mtu);
1205 if (!error && mtu != INT_MAX) {
1207 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1210 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1214 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1215 * changed, false otherwise. */
1217 iface_refresh_cfm_stats(struct iface *iface)
1219 const struct ovsrec_monitor *mon;
1220 const struct cfm *cfm;
1221 bool changed = false;
1224 mon = iface->cfg->monitor;
1225 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1231 for (i = 0; i < mon->n_remote_mps; i++) {
1232 const struct ovsrec_maintenance_point *mp;
1233 const struct remote_mp *rmp;
1235 mp = mon->remote_mps[i];
1236 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1238 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1239 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1244 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1245 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1253 iface_refresh_lacp_stats(struct iface *iface)
1255 bool *db_current = iface->cfg->lacp_current;
1256 bool changed = false;
1258 if (iface->port->lacp) {
1259 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1261 if (!db_current || *db_current != current) {
1263 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1265 } else if (db_current) {
1267 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1274 iface_refresh_stats(struct iface *iface)
1280 static const struct iface_stat iface_stats[] = {
1281 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1282 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1283 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1284 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1285 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1286 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1287 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1288 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1289 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1290 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1291 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1292 { "collisions", offsetof(struct netdev_stats, collisions) },
1294 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1295 const struct iface_stat *s;
1297 char *keys[N_STATS];
1298 int64_t values[N_STATS];
1301 struct netdev_stats stats;
1303 if (iface_is_synthetic(iface)) {
1307 /* Intentionally ignore return value, since errors will set 'stats' to
1308 * all-1s, and we will deal with that correctly below. */
1309 netdev_get_stats(iface->netdev, &stats);
1312 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1313 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1314 if (value != UINT64_MAX) {
1321 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1325 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1327 struct ovsdb_datum datum;
1331 get_system_stats(&stats);
1333 ovsdb_datum_from_shash(&datum, &stats);
1334 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1338 static inline const char *
1339 nx_role_to_str(enum nx_role role)
1344 case NX_ROLE_MASTER:
1349 return "*** INVALID ROLE ***";
1354 bridge_refresh_controller_status(const struct bridge *br)
1357 const struct ovsrec_controller *cfg;
1359 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1361 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1362 struct ofproto_controller_info *cinfo =
1363 shash_find_data(&info, cfg->target);
1366 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1367 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1368 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1369 (char **) cinfo->pairs.values,
1372 ovsrec_controller_set_is_connected(cfg, false);
1373 ovsrec_controller_set_role(cfg, NULL);
1374 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1378 ofproto_free_ofproto_controller_info(&info);
1384 const struct ovsrec_open_vswitch *cfg;
1386 bool datapath_destroyed;
1387 bool database_changed;
1390 /* Let each bridge do the work that it needs to do. */
1391 datapath_destroyed = false;
1392 LIST_FOR_EACH (br, node, &all_bridges) {
1393 int error = bridge_run_one(br);
1395 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1396 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1397 "forcing reconfiguration", br->name);
1398 datapath_destroyed = true;
1402 /* (Re)configure if necessary. */
1403 database_changed = ovsdb_idl_run(idl);
1404 cfg = ovsrec_open_vswitch_first(idl);
1406 /* Re-configure SSL. We do this on every trip through the main loop,
1407 * instead of just when the database changes, because the contents of the
1408 * key and certificate files can change without the database changing.
1410 * We do this before bridge_reconfigure() because that function might
1411 * initiate SSL connections and thus requires SSL to be configured. */
1412 if (cfg && cfg->ssl) {
1413 const struct ovsrec_ssl *ssl = cfg->ssl;
1415 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1416 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1419 if (database_changed || datapath_destroyed) {
1421 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1423 bridge_configure_once(cfg);
1424 bridge_reconfigure(cfg);
1426 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1427 ovsdb_idl_txn_commit(txn);
1428 ovsdb_idl_txn_destroy(txn); /* XXX */
1430 /* We still need to reconfigure to avoid dangling pointers to
1431 * now-destroyed ovsrec structures inside bridge data. */
1432 static const struct ovsrec_open_vswitch null_cfg;
1434 bridge_reconfigure(&null_cfg);
1438 /* Refresh system and interface stats if necessary. */
1439 if (time_msec() >= stats_timer) {
1441 struct ovsdb_idl_txn *txn;
1443 txn = ovsdb_idl_txn_create(idl);
1444 LIST_FOR_EACH (br, node, &all_bridges) {
1447 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1448 struct iface *iface;
1450 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1451 iface_refresh_stats(iface);
1452 iface_refresh_status(iface);
1455 bridge_refresh_controller_status(br);
1457 refresh_system_stats(cfg);
1458 ovsdb_idl_txn_commit(txn);
1459 ovsdb_idl_txn_destroy(txn); /* XXX */
1462 stats_timer = time_msec() + STATS_INTERVAL;
1465 if (time_msec() >= db_limiter) {
1466 struct ovsdb_idl_txn *txn;
1467 bool changed = false;
1469 txn = ovsdb_idl_txn_create(idl);
1470 LIST_FOR_EACH (br, node, &all_bridges) {
1473 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1474 struct iface *iface;
1476 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1477 changed = iface_refresh_cfm_stats(iface) || changed;
1478 changed = iface_refresh_lacp_stats(iface) || changed;
1484 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1487 ovsdb_idl_txn_commit(txn);
1488 ovsdb_idl_txn_destroy(txn);
1497 LIST_FOR_EACH (br, node, &all_bridges) {
1500 ofproto_wait(br->ofproto);
1501 mac_learning_wait(br->ml);
1502 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1506 ovsdb_idl_wait(idl);
1507 poll_timer_wait_until(stats_timer);
1509 if (db_limiter > time_msec()) {
1510 poll_timer_wait_until(db_limiter);
1514 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1515 * configuration changes. */
1517 bridge_flush(struct bridge *br)
1519 COVERAGE_INC(bridge_flush);
1523 /* Bridge unixctl user interface functions. */
1525 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1526 const char *args, void *aux OVS_UNUSED)
1528 struct ds ds = DS_EMPTY_INITIALIZER;
1529 const struct bridge *br;
1530 const struct mac_entry *e;
1532 br = bridge_lookup(args);
1534 unixctl_command_reply(conn, 501, "no such bridge");
1538 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1539 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1540 struct port *port = e->port.p;
1541 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1542 port_get_an_iface(port)->dp_ifidx,
1543 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1545 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1549 /* CFM unixctl user interface functions. */
1551 cfm_unixctl_show(struct unixctl_conn *conn,
1552 const char *args, void *aux OVS_UNUSED)
1554 struct ds ds = DS_EMPTY_INITIALIZER;
1555 struct iface *iface;
1556 const struct cfm *cfm;
1558 iface = iface_find(args);
1560 unixctl_command_reply(conn, 501, "no such interface");
1564 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1567 unixctl_command_reply(conn, 501, "CFM not enabled");
1571 cfm_dump_ds(cfm, &ds);
1572 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1576 /* QoS unixctl user interface functions. */
1578 struct qos_unixctl_show_cbdata {
1580 struct iface *iface;
1584 qos_unixctl_show_cb(unsigned int queue_id,
1585 const struct shash *details,
1588 struct qos_unixctl_show_cbdata *data = aux;
1589 struct ds *ds = data->ds;
1590 struct iface *iface = data->iface;
1591 struct netdev_queue_stats stats;
1592 struct shash_node *node;
1595 ds_put_cstr(ds, "\n");
1597 ds_put_format(ds, "Queue %u:\n", queue_id);
1599 ds_put_cstr(ds, "Default:\n");
1602 SHASH_FOR_EACH (node, details) {
1603 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1606 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1608 if (stats.tx_packets != UINT64_MAX) {
1609 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1612 if (stats.tx_bytes != UINT64_MAX) {
1613 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1616 if (stats.tx_errors != UINT64_MAX) {
1617 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1620 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1621 queue_id, strerror(error));
1626 qos_unixctl_show(struct unixctl_conn *conn,
1627 const char *args, void *aux OVS_UNUSED)
1629 struct ds ds = DS_EMPTY_INITIALIZER;
1630 struct shash sh = SHASH_INITIALIZER(&sh);
1631 struct iface *iface;
1633 struct shash_node *node;
1634 struct qos_unixctl_show_cbdata data;
1637 iface = iface_find(args);
1639 unixctl_command_reply(conn, 501, "no such interface");
1643 netdev_get_qos(iface->netdev, &type, &sh);
1645 if (*type != '\0') {
1646 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1648 SHASH_FOR_EACH (node, &sh) {
1649 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1654 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1657 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1659 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1661 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1662 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1665 shash_destroy_free_data(&sh);
1669 /* Bridge reconfiguration functions. */
1670 static struct bridge *
1671 bridge_create(const struct ovsrec_bridge *br_cfg)
1676 assert(!bridge_lookup(br_cfg->name));
1677 br = xzalloc(sizeof *br);
1679 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1686 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1689 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1691 dpif_delete(br->dpif);
1692 dpif_close(br->dpif);
1697 br->name = xstrdup(br_cfg->name);
1699 br->ml = mac_learning_create();
1700 eth_addr_nicira_random(br->default_ea);
1702 hmap_init(&br->ports);
1703 hmap_init(&br->ifaces);
1704 shash_init(&br->iface_by_name);
1708 list_push_back(&all_bridges, &br->node);
1710 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1716 bridge_destroy(struct bridge *br)
1719 struct port *port, *next;
1723 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1726 for (i = 0; i < MAX_MIRRORS; i++) {
1727 mirror_destroy(br->mirrors[i]);
1729 list_remove(&br->node);
1730 ofproto_destroy(br->ofproto);
1731 error = dpif_delete(br->dpif);
1732 if (error && error != ENOENT) {
1733 VLOG_ERR("failed to delete %s: %s",
1734 dpif_name(br->dpif), strerror(error));
1736 dpif_close(br->dpif);
1737 mac_learning_destroy(br->ml);
1738 hmap_destroy(&br->ifaces);
1739 hmap_destroy(&br->ports);
1740 shash_destroy(&br->iface_by_name);
1741 free(br->synth_local_iface.type);
1747 static struct bridge *
1748 bridge_lookup(const char *name)
1752 LIST_FOR_EACH (br, node, &all_bridges) {
1753 if (!strcmp(br->name, name)) {
1760 /* Handle requests for a listing of all flows known by the OpenFlow
1761 * stack, including those normally hidden. */
1763 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1764 const char *args, void *aux OVS_UNUSED)
1769 br = bridge_lookup(args);
1771 unixctl_command_reply(conn, 501, "Unknown bridge");
1776 ofproto_get_all_flows(br->ofproto, &results);
1778 unixctl_command_reply(conn, 200, ds_cstr(&results));
1779 ds_destroy(&results);
1782 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1783 * connections and reconnect. If BRIDGE is not specified, then all bridges
1784 * drop their controller connections and reconnect. */
1786 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1787 const char *args, void *aux OVS_UNUSED)
1790 if (args[0] != '\0') {
1791 br = bridge_lookup(args);
1793 unixctl_command_reply(conn, 501, "Unknown bridge");
1796 ofproto_reconnect_controllers(br->ofproto);
1798 LIST_FOR_EACH (br, node, &all_bridges) {
1799 ofproto_reconnect_controllers(br->ofproto);
1802 unixctl_command_reply(conn, 200, NULL);
1806 bridge_run_one(struct bridge *br)
1811 error = ofproto_run1(br->ofproto);
1816 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1818 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1822 error = ofproto_run2(br->ofproto, br->flush);
1829 bridge_get_controllers(const struct bridge *br,
1830 struct ovsrec_controller ***controllersp)
1832 struct ovsrec_controller **controllers;
1833 size_t n_controllers;
1835 controllers = br->cfg->controller;
1836 n_controllers = br->cfg->n_controller;
1838 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1844 *controllersp = controllers;
1846 return n_controllers;
1850 bridge_reconfigure_one(struct bridge *br)
1852 enum ofproto_fail_mode fail_mode;
1853 struct port *port, *next;
1854 struct shash_node *node;
1855 struct shash new_ports;
1858 /* Collect new ports. */
1859 shash_init(&new_ports);
1860 for (i = 0; i < br->cfg->n_ports; i++) {
1861 const char *name = br->cfg->ports[i]->name;
1862 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1863 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1867 if (!shash_find(&new_ports, br->name)) {
1868 struct dpif_port dpif_port;
1871 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1872 br->name, br->name);
1874 dpif_port_query_by_number(br->dpif, ODPP_LOCAL, &dpif_port);
1875 type = xstrdup(dpif_port.type ? dpif_port.type : "internal");
1876 dpif_port_destroy(&dpif_port);
1878 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1879 br->synth_local_port.n_interfaces = 1;
1880 br->synth_local_port.name = br->name;
1882 br->synth_local_iface.name = br->name;
1883 free(br->synth_local_iface.type);
1884 br->synth_local_iface.type = type;
1886 br->synth_local_ifacep = &br->synth_local_iface;
1888 shash_add(&new_ports, br->name, &br->synth_local_port);
1891 /* Get rid of deleted ports.
1892 * Get rid of deleted interfaces on ports that still exist. */
1893 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1894 const struct ovsrec_port *port_cfg;
1896 port_cfg = shash_find_data(&new_ports, port->name);
1900 port_del_ifaces(port, port_cfg);
1904 /* Create new ports.
1905 * Add new interfaces to existing ports.
1906 * Reconfigure existing ports. */
1907 SHASH_FOR_EACH (node, &new_ports) {
1908 struct port *port = port_lookup(br, node->name);
1910 port = port_create(br, node->name);
1913 port_reconfigure(port, node->data);
1914 if (list_is_empty(&port->ifaces)) {
1915 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1916 br->name, port->name);
1920 shash_destroy(&new_ports);
1922 /* Set the fail-mode */
1923 fail_mode = !br->cfg->fail_mode
1924 || !strcmp(br->cfg->fail_mode, "standalone")
1925 ? OFPROTO_FAIL_STANDALONE
1926 : OFPROTO_FAIL_SECURE;
1927 ofproto_set_fail_mode(br->ofproto, fail_mode);
1929 /* Configure OpenFlow controller connection snooping. */
1930 if (!ofproto_has_snoops(br->ofproto)) {
1934 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1935 ovs_rundir(), br->name));
1936 ofproto_set_snoops(br->ofproto, &snoops);
1937 sset_destroy(&snoops);
1940 mirror_reconfigure(br);
1943 /* Initializes 'oc' appropriately as a management service controller for
1946 * The caller must free oc->target when it is no longer needed. */
1948 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1949 struct ofproto_controller *oc)
1951 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1952 oc->max_backoff = 0;
1953 oc->probe_interval = 60;
1954 oc->band = OFPROTO_OUT_OF_BAND;
1956 oc->burst_limit = 0;
1959 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1961 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1962 struct ofproto_controller *oc)
1964 oc->target = c->target;
1965 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1966 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1967 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1968 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1969 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1970 oc->burst_limit = (c->controller_burst_limit
1971 ? *c->controller_burst_limit : 0);
1974 /* Configures the IP stack for 'br''s local interface properly according to the
1975 * configuration in 'c'. */
1977 bridge_configure_local_iface_netdev(struct bridge *br,
1978 struct ovsrec_controller *c)
1980 struct netdev *netdev;
1981 struct in_addr mask, gateway;
1983 struct iface *local_iface;
1986 /* If there's no local interface or no IP address, give up. */
1987 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1988 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1992 /* Bring up the local interface. */
1993 netdev = local_iface->netdev;
1994 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1996 /* Configure the IP address and netmask. */
1997 if (!c->local_netmask
1998 || !inet_aton(c->local_netmask, &mask)
2000 mask.s_addr = guess_netmask(ip.s_addr);
2002 if (!netdev_set_in4(netdev, ip, mask)) {
2003 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2004 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
2007 /* Configure the default gateway. */
2008 if (c->local_gateway
2009 && inet_aton(c->local_gateway, &gateway)
2010 && gateway.s_addr) {
2011 if (!netdev_add_router(netdev, gateway)) {
2012 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2013 br->name, IP_ARGS(&gateway.s_addr));
2019 bridge_reconfigure_remotes(struct bridge *br,
2020 const struct sockaddr_in *managers,
2023 const char *disable_ib_str, *queue_id_str;
2024 bool disable_in_band = false;
2027 struct ovsrec_controller **controllers;
2028 size_t n_controllers;
2030 struct ofproto_controller *ocs;
2034 /* Check if we should disable in-band control on this bridge. */
2035 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
2036 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2037 disable_in_band = true;
2040 /* Set OpenFlow queue ID for in-band control. */
2041 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2042 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2043 ofproto_set_in_band_queue(br->ofproto, queue_id);
2045 if (disable_in_band) {
2046 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2048 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2051 n_controllers = bridge_get_controllers(br, &controllers);
2053 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2056 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2057 for (i = 0; i < n_controllers; i++) {
2058 struct ovsrec_controller *c = controllers[i];
2060 if (!strncmp(c->target, "punix:", 6)
2061 || !strncmp(c->target, "unix:", 5)) {
2062 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2064 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2065 * domain sockets and overwriting arbitrary local files. */
2066 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
2067 "\"%s\" due to possibility for remote exploit",
2068 dpif_name(br->dpif), c->target);
2072 bridge_configure_local_iface_netdev(br, c);
2073 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2074 if (disable_in_band) {
2075 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2080 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2081 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2086 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2091 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2092 struct iface *iface;
2094 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2095 shash_add_once(ifaces, iface->name, iface);
2097 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2098 shash_add_once(ifaces, port->name, NULL);
2103 /* For robustness, in case the administrator moves around datapath ports behind
2104 * our back, we re-check all the datapath port numbers here.
2106 * This function will set the 'dp_ifidx' members of interfaces that have
2107 * disappeared to -1, so only call this function from a context where those
2108 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2109 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2110 * datapath, which doesn't support UINT16_MAX+1 ports. */
2112 bridge_fetch_dp_ifaces(struct bridge *br)
2114 struct dpif_port_dump dump;
2115 struct dpif_port dpif_port;
2118 /* Reset all interface numbers. */
2119 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2120 struct iface *iface;
2122 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2123 iface->dp_ifidx = -1;
2126 hmap_clear(&br->ifaces);
2128 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2129 struct iface *iface = iface_lookup(br, dpif_port.name);
2131 if (iface->dp_ifidx >= 0) {
2132 VLOG_WARN("%s reported interface %s twice",
2133 dpif_name(br->dpif), dpif_port.name);
2134 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2135 VLOG_WARN("%s reported interface %"PRIu16" twice",
2136 dpif_name(br->dpif), dpif_port.port_no);
2138 iface->dp_ifidx = dpif_port.port_no;
2139 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2140 hash_int(iface->dp_ifidx, 0));
2143 iface_set_ofport(iface->cfg,
2144 (iface->dp_ifidx >= 0
2145 ? odp_port_to_ofp_port(iface->dp_ifidx)
2151 /* Bridge packet processing functions. */
2154 set_dst(struct dst *dst, const struct flow *flow,
2155 const struct port *in_port, const struct port *out_port,
2158 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2159 : in_port->vlan >= 0 ? in_port->vlan
2160 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2161 : vlan_tci_to_vid(flow->vlan_tci));
2163 dst->iface = (!out_port->bond
2164 ? port_get_an_iface(out_port)
2165 : bond_choose_output_slave(out_port->bond, flow,
2168 return dst->iface != NULL;
2172 mirror_mask_ffs(mirror_mask_t mask)
2174 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2179 dst_set_init(struct dst_set *set)
2181 set->dsts = set->builtin;
2183 set->allocated = ARRAY_SIZE(set->builtin);
2187 dst_set_add(struct dst_set *set, const struct dst *dst)
2189 if (set->n >= set->allocated) {
2190 size_t new_allocated;
2191 struct dst *new_dsts;
2193 new_allocated = set->allocated * 2;
2194 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2195 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2199 set->dsts = new_dsts;
2200 set->allocated = new_allocated;
2202 set->dsts[set->n++] = *dst;
2206 dst_set_free(struct dst_set *set)
2208 if (set->dsts != set->builtin) {
2214 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2217 for (i = 0; i < set->n; i++) {
2218 if (set->dsts[i].vlan == test->vlan
2219 && set->dsts[i].iface == test->iface) {
2227 port_trunks_vlan(const struct port *port, uint16_t vlan)
2229 return (port->vlan < 0 || vlan_bitmap_contains(port->trunks, vlan));
2233 port_includes_vlan(const struct port *port, uint16_t vlan)
2235 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2239 port_is_floodable(const struct port *port)
2241 struct iface *iface;
2243 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2244 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2252 /* Returns an arbitrary interface within 'port'. */
2253 static struct iface *
2254 port_get_an_iface(const struct port *port)
2256 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2260 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2261 const struct port *in_port, const struct port *out_port,
2262 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2266 if (out_port == FLOOD_PORT) {
2269 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2271 && port_is_floodable(port)
2272 && port_includes_vlan(port, vlan)
2273 && !port->is_mirror_output_port
2274 && set_dst(&dst, flow, in_port, port, tags)) {
2275 dst_set_add(set, &dst);
2278 *nf_output_iface = NF_OUT_FLOOD;
2279 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2280 dst_set_add(set, &dst);
2281 *nf_output_iface = dst.iface->dp_ifidx;
2286 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2287 uint16_t vlan, const struct port *in_port,
2288 struct dst_set *set, tag_type *tags)
2290 mirror_mask_t mirrors;
2294 mirrors = in_port->src_mirrors;
2295 for (i = 0; i < set->n; i++) {
2296 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2303 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2304 if (flow_vlan == 0) {
2305 flow_vlan = OFP_VLAN_NONE;
2309 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2310 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2314 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2315 && !dst_is_duplicate(set, &dst)) {
2316 dst_set_add(set, &dst);
2321 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2322 if (port_includes_vlan(port, m->out_vlan)
2323 && set_dst(&dst, flow, in_port, port, tags))
2325 if (port->vlan < 0) {
2326 dst.vlan = m->out_vlan;
2328 if (dst_is_duplicate(set, &dst)) {
2332 /* Use the vlan tag on the original flow instead of
2333 * the one passed in the vlan parameter. This ensures
2334 * that we compare the vlan from before any implicit
2335 * tagging tags place. This is necessary because
2336 * dst->vlan is the final vlan, after removing implicit
2338 if (port == in_port && dst.vlan == flow_vlan) {
2339 /* Don't send out input port on same VLAN. */
2342 dst_set_add(set, &dst);
2347 mirrors &= mirrors - 1;
2352 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2353 const struct port *in_port, const struct port *out_port,
2354 tag_type *tags, struct ofpbuf *actions,
2355 uint16_t *nf_output_iface)
2357 uint16_t initial_vlan, cur_vlan;
2358 const struct dst *dst;
2362 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2364 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2366 /* Output all the packets we can without having to change the VLAN. */
2367 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2368 if (initial_vlan == 0) {
2369 initial_vlan = OFP_VLAN_NONE;
2371 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2372 if (dst->vlan != initial_vlan) {
2375 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2378 /* Then output the rest. */
2379 cur_vlan = initial_vlan;
2380 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2381 if (dst->vlan == initial_vlan) {
2384 if (dst->vlan != cur_vlan) {
2385 if (dst->vlan == OFP_VLAN_NONE) {
2386 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2389 tci = htons(dst->vlan & VLAN_VID_MASK);
2390 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2391 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2393 cur_vlan = dst->vlan;
2395 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2401 /* Returns the effective vlan of a packet, taking into account both the
2402 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2403 * the packet is untagged and -1 indicates it has an invalid header and
2404 * should be dropped. */
2405 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2406 struct port *in_port, bool have_packet)
2408 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2409 if (in_port->vlan >= 0) {
2412 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2413 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2414 "packet received on port %s configured with "
2415 "implicit VLAN %"PRIu16,
2416 br->name, vlan, in_port->name, in_port->vlan);
2420 vlan = in_port->vlan;
2422 if (!port_includes_vlan(in_port, vlan)) {
2424 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2425 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2426 "packet received on port %s not configured for "
2428 br->name, vlan, in_port->name, vlan);
2437 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2438 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2439 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2441 is_gratuitous_arp(const struct flow *flow)
2443 return (flow->dl_type == htons(ETH_TYPE_ARP)
2444 && eth_addr_is_broadcast(flow->dl_dst)
2445 && (flow->nw_proto == ARP_OP_REPLY
2446 || (flow->nw_proto == ARP_OP_REQUEST
2447 && flow->nw_src == flow->nw_dst)));
2451 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2452 struct port *in_port)
2454 struct mac_entry *mac;
2456 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2460 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2461 if (is_gratuitous_arp(flow)) {
2462 /* We don't want to learn from gratuitous ARP packets that are
2463 * reflected back over bond slaves so we lock the learning table. */
2464 if (!in_port->bond) {
2465 mac_entry_set_grat_arp_lock(mac);
2466 } else if (mac_entry_is_grat_arp_locked(mac)) {
2471 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2472 /* The log messages here could actually be useful in debugging,
2473 * so keep the rate limit relatively high. */
2474 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2475 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2476 "on port %s in VLAN %d",
2477 br->name, ETH_ADDR_ARGS(flow->dl_src),
2478 in_port->name, vlan);
2480 mac->port.p = in_port;
2481 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2485 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2486 * dropped. Returns true if they may be forwarded, false if they should be
2489 * If 'have_packet' is true, it indicates that the caller is processing a
2490 * received packet. If 'have_packet' is false, then the caller is just
2491 * revalidating an existing flow because configuration has changed. Either
2492 * way, 'have_packet' only affects logging (there is no point in logging errors
2493 * during revalidation).
2495 * Sets '*in_portp' to the input port. This will be a null pointer if
2496 * flow->in_port does not designate a known input port (in which case
2497 * is_admissible() returns false).
2499 * When returning true, sets '*vlanp' to the effective VLAN of the input
2500 * packet, as returned by flow_get_vlan().
2502 * May also add tags to '*tags', although the current implementation only does
2503 * so in one special case.
2506 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2507 tag_type *tags, int *vlanp, struct port **in_portp)
2509 struct iface *in_iface;
2510 struct port *in_port;
2513 /* Find the interface and port structure for the received packet. */
2514 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2516 /* No interface? Something fishy... */
2518 /* Odd. A few possible reasons here:
2520 * - We deleted an interface but there are still a few packets
2521 * queued up from it.
2523 * - Someone externally added an interface (e.g. with "ovs-dpctl
2524 * add-if") that we don't know about.
2526 * - Packet arrived on the local port but the local port is not
2527 * one of our bridge ports.
2529 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2531 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2532 "interface %"PRIu16, br->name, flow->in_port);
2538 *in_portp = in_port = in_iface->port;
2539 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2544 /* Drop frames for reserved multicast addresses. */
2545 if (eth_addr_is_reserved(flow->dl_dst)) {
2549 /* Drop frames on ports reserved for mirroring. */
2550 if (in_port->is_mirror_output_port) {
2552 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2553 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2554 "%s, which is reserved exclusively for mirroring",
2555 br->name, in_port->name);
2560 if (in_port->bond) {
2561 struct mac_entry *mac;
2563 switch (bond_check_admissibility(in_port->bond, in_iface,
2564 flow->dl_dst, tags)) {
2571 case BV_DROP_IF_MOVED:
2572 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2573 if (mac && mac->port.p != in_port &&
2574 (!is_gratuitous_arp(flow)
2575 || mac_entry_is_grat_arp_locked(mac))) {
2585 /* If the composed actions may be applied to any packet in the given 'flow',
2586 * returns true. Otherwise, the actions should only be applied to 'packet', or
2587 * not at all, if 'packet' was NULL. */
2589 process_flow(struct bridge *br, const struct flow *flow,
2590 const struct ofpbuf *packet, struct ofpbuf *actions,
2591 tag_type *tags, uint16_t *nf_output_iface)
2593 struct port *in_port;
2594 struct port *out_port;
2595 struct mac_entry *mac;
2598 /* Check whether we should drop packets in this flow. */
2599 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2604 /* Learn source MAC (but don't try to learn from revalidation). */
2606 update_learning_table(br, flow, vlan, in_port);
2609 /* Determine output port. */
2610 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2612 out_port = mac->port.p;
2613 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2614 /* If we are revalidating but don't have a learning entry then
2615 * eject the flow. Installing a flow that floods packets opens
2616 * up a window of time where we could learn from a packet reflected
2617 * on a bond and blackhole packets before the learning table is
2618 * updated to reflect the correct port. */
2621 out_port = FLOOD_PORT;
2624 /* Don't send packets out their input ports. */
2625 if (in_port == out_port) {
2631 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2639 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2640 struct ofpbuf *actions, tag_type *tags,
2641 uint16_t *nf_output_iface, void *br_)
2643 struct bridge *br = br_;
2645 COVERAGE_INC(bridge_process_flow);
2646 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2650 bridge_special_ofhook_cb(const struct flow *flow,
2651 const struct ofpbuf *packet, void *br_)
2653 struct iface *iface;
2654 struct bridge *br = br_;
2656 iface = iface_from_dp_ifidx(br, flow->in_port);
2658 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2659 if (iface && iface->port->lacp && packet) {
2660 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2662 lacp_process_pdu(iface->port->lacp, iface, pdu);
2672 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2673 const struct nlattr *actions,
2675 uint64_t n_bytes, void *br_)
2677 struct bridge *br = br_;
2678 const struct nlattr *a;
2679 struct port *in_port;
2684 /* Feed information from the active flows back into the learning table to
2685 * ensure that table is always in sync with what is actually flowing
2686 * through the datapath.
2688 * We test that 'tags' is nonzero to ensure that only flows that include an
2689 * OFPP_NORMAL action are used for learning. This works because
2690 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2691 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2692 update_learning_table(br, flow, vlan, in_port);
2695 /* Account for bond slave utilization. */
2696 if (!br->has_bonded_ports) {
2699 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2700 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2701 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2702 if (out_port && out_port->bond) {
2703 uint16_t vlan = (flow->vlan_tci
2704 ? vlan_tci_to_vid(flow->vlan_tci)
2706 bond_account(out_port->bond, flow, vlan, n_bytes);
2713 bridge_account_checkpoint_ofhook_cb(void *br_)
2715 struct bridge *br = br_;
2718 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2720 bond_rebalance(port->bond,
2721 ofproto_get_revalidate_set(br->ofproto));
2727 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2728 tag_type *tags, void *br_)
2730 struct bridge *br = br_;
2731 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2732 struct port *port = port_from_dp_ifidx(br, odp_port);
2737 } else if (list_is_short(&port->ifaces)) {
2740 struct iface *iface;
2742 /* Autopath does not support VLAN hashing. */
2743 iface = bond_choose_output_slave(port->bond, flow,
2744 OFP_VLAN_NONE, tags);
2745 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2748 return odp_port_to_ofp_port(ret);
2751 static struct ofhooks bridge_ofhooks = {
2752 bridge_normal_ofhook_cb,
2753 bridge_special_ofhook_cb,
2754 bridge_account_flow_ofhook_cb,
2755 bridge_account_checkpoint_ofhook_cb,
2756 bridge_autopath_ofhook_cb,
2759 /* Port functions. */
2762 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2764 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2765 struct iface *iface = iface_;
2766 uint8_t ea[ETH_ADDR_LEN];
2769 error = netdev_get_etheraddr(iface->netdev, ea);
2771 struct lacp_pdu *packet_pdu;
2772 struct ofpbuf packet;
2774 ofpbuf_init(&packet, 0);
2775 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2776 sizeof *packet_pdu);
2778 error = netdev_send(iface->netdev, &packet);
2780 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2781 "(%s)", iface->port->name, iface->name,
2784 ofpbuf_uninit(&packet);
2786 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2787 "%s (%s)", iface->port->name, iface->name,
2793 port_run(struct port *port)
2796 lacp_run(port->lacp, lacp_send_pdu_cb);
2800 struct iface *iface;
2802 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2803 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2804 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2807 bond_run(port->bond,
2808 ofproto_get_revalidate_set(port->bridge->ofproto),
2809 lacp_negotiated(port->lacp));
2810 if (bond_should_send_learning_packets(port->bond)) {
2811 port_send_learning_packets(port);
2817 port_wait(struct port *port)
2820 lacp_wait(port->lacp);
2824 bond_wait(port->bond);
2828 static struct port *
2829 port_create(struct bridge *br, const char *name)
2833 port = xzalloc(sizeof *port);
2836 port->trunks = NULL;
2837 port->name = xstrdup(name);
2838 list_init(&port->ifaces);
2840 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2842 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2849 get_port_other_config(const struct ovsrec_port *port, const char *key,
2850 const char *default_value)
2854 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2856 return value ? value : default_value;
2860 get_interface_other_config(const struct ovsrec_interface *iface,
2861 const char *key, const char *default_value)
2865 value = get_ovsrec_key_value(&iface->header_,
2866 &ovsrec_interface_col_other_config, key);
2867 return value ? value : default_value;
2871 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2873 struct iface *iface, *next;
2874 struct sset new_ifaces;
2877 /* Collect list of new interfaces. */
2878 sset_init(&new_ifaces);
2879 for (i = 0; i < cfg->n_interfaces; i++) {
2880 const char *name = cfg->interfaces[i]->name;
2881 sset_add(&new_ifaces, name);
2884 /* Get rid of deleted interfaces. */
2885 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2886 if (!sset_contains(&new_ifaces, iface->name)) {
2887 iface_destroy(iface);
2891 sset_destroy(&new_ifaces);
2894 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2895 * to revalidate every flow. */
2897 port_flush_macs(struct port *port)
2899 struct bridge *br = port->bridge;
2900 struct mac_learning *ml = br->ml;
2901 struct mac_entry *mac, *next_mac;
2904 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2905 if (mac->port.p == port) {
2906 mac_learning_expire(ml, mac);
2912 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2914 struct sset new_ifaces;
2915 bool need_flush = false;
2916 unsigned long *trunks;
2923 /* Add new interfaces and update 'cfg' member of existing ones. */
2924 sset_init(&new_ifaces);
2925 for (i = 0; i < cfg->n_interfaces; i++) {
2926 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2927 struct iface *iface;
2929 if (!sset_add(&new_ifaces, if_cfg->name)) {
2930 VLOG_WARN("port %s: %s specified twice as port interface",
2931 port->name, if_cfg->name);
2932 iface_set_ofport(if_cfg, -1);
2936 iface = iface_lookup(port->bridge, if_cfg->name);
2938 if (iface->port != port) {
2939 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2941 port->bridge->name, if_cfg->name, iface->port->name);
2944 iface->cfg = if_cfg;
2946 iface = iface_create(port, if_cfg);
2949 /* Determine interface type. The local port always has type
2950 * "internal". Other ports take their type from the database and
2951 * default to "system" if none is specified. */
2952 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2953 : if_cfg->type[0] ? if_cfg->type
2956 sset_destroy(&new_ifaces);
2961 if (list_is_short(&port->ifaces)) {
2963 if (vlan >= 0 && vlan <= 4095) {
2964 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2969 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2970 * they even work as-is. But they have not been tested. */
2971 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2975 if (port->vlan != vlan) {
2980 /* Get trunked VLANs. */
2982 if (vlan < 0 && cfg->n_trunks) {
2983 trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
2985 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2988 } else if (vlan >= 0 && cfg->n_trunks) {
2989 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2992 if (!vlan_bitmap_equal(trunks, port->trunks)) {
2995 bitmap_free(port->trunks);
2996 port->trunks = trunks;
2999 port_flush_macs(port);
3004 port_destroy(struct port *port)
3007 struct bridge *br = port->bridge;
3008 struct iface *iface, *next;
3011 for (i = 0; i < MAX_MIRRORS; i++) {
3012 struct mirror *m = br->mirrors[i];
3013 if (m && m->out_port == port) {
3018 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3019 iface_destroy(iface);
3022 hmap_remove(&br->ports, &port->hmap_node);
3024 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
3026 bond_destroy(port->bond);
3027 lacp_destroy(port->lacp);
3028 port_flush_macs(port);
3030 bitmap_free(port->trunks);
3036 static struct port *
3037 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3039 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
3040 return iface ? iface->port : NULL;
3043 static struct port *
3044 port_lookup(const struct bridge *br, const char *name)
3048 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3050 if (!strcmp(port->name, name)) {
3058 enable_lacp(struct port *port, bool *activep)
3060 if (!port->cfg->lacp) {
3061 /* XXX when LACP implementation has been sufficiently tested, enable by
3062 * default and make active on bonded ports. */
3064 } else if (!strcmp(port->cfg->lacp, "off")) {
3066 } else if (!strcmp(port->cfg->lacp, "active")) {
3069 } else if (!strcmp(port->cfg->lacp, "passive")) {
3073 VLOG_WARN("port %s: unknown LACP mode %s",
3074 port->name, port->cfg->lacp);
3080 iface_reconfigure_lacp(struct iface *iface)
3082 struct lacp_slave_settings s;
3083 int priority, portid;
3085 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
3086 priority = atoi(get_interface_other_config(iface->cfg,
3087 "lacp-port-priority", "0"));
3089 if (portid <= 0 || portid > UINT16_MAX) {
3090 portid = iface->dp_ifidx;
3093 if (priority <= 0 || priority > UINT16_MAX) {
3094 priority = UINT16_MAX;
3097 s.name = iface->name;
3099 s.priority = priority;
3100 lacp_slave_register(iface->port->lacp, iface, &s);
3104 port_reconfigure_lacp(struct port *port)
3106 static struct lacp_settings s;
3107 struct iface *iface;
3108 uint8_t sysid[ETH_ADDR_LEN];
3109 const char *sysid_str;
3110 const char *lacp_time;
3111 long long int custom_time;
3114 if (!enable_lacp(port, &s.active)) {
3115 lacp_destroy(port->lacp);
3120 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3121 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3122 memcpy(s.id, sysid, ETH_ADDR_LEN);
3124 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3127 s.name = port->name;
3129 /* Prefer bondable links if unspecified. */
3130 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3132 s.priority = (priority > 0 && priority <= UINT16_MAX
3134 : UINT16_MAX - !list_is_short(&port->ifaces));
3136 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3140 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3141 custom_time = atoi(lacp_time);
3142 if (!strcmp(lacp_time, "fast")) {
3143 s.lacp_time = LACP_TIME_FAST;
3144 } else if (!strcmp(lacp_time, "slow")) {
3145 s.lacp_time = LACP_TIME_SLOW;
3146 } else if (custom_time > 0) {
3147 s.lacp_time = LACP_TIME_CUSTOM;
3148 s.custom_time = custom_time;
3150 s.lacp_time = LACP_TIME_SLOW;
3154 port->lacp = lacp_create();
3157 lacp_configure(port->lacp, &s);
3159 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3160 iface_reconfigure_lacp(iface);
3165 port_reconfigure_bond(struct port *port)
3167 struct bond_settings s;
3168 const char *detect_s;
3169 struct iface *iface;
3171 if (list_is_short(&port->ifaces)) {
3172 /* Not a bonded port. */
3173 bond_destroy(port->bond);
3178 port->bridge->has_bonded_ports = true;
3180 s.name = port->name;
3182 if (port->cfg->bond_mode
3183 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3184 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3185 port->name, port->cfg->bond_mode,
3186 bond_mode_to_string(s.balance));
3189 s.detect = BLSM_CARRIER;
3190 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3191 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3192 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3194 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3197 s.miimon_interval = atoi(
3198 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3199 if (s.miimon_interval < 100) {
3200 s.miimon_interval = 100;
3203 s.up_delay = MAX(0, port->cfg->bond_updelay);
3204 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3205 s.rebalance_interval = atoi(
3206 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3207 if (s.rebalance_interval < 1000) {
3208 s.rebalance_interval = 1000;
3211 s.fake_iface = port->cfg->bond_fake_iface;
3214 port->bond = bond_create(&s);
3216 if (bond_reconfigure(port->bond, &s)) {
3217 bridge_flush(port->bridge);
3221 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3222 uint16_t stable_id = (port->lacp
3223 ? lacp_slave_get_port_id(port->lacp, iface)
3225 bond_slave_register(iface->port->bond, iface, stable_id,
3231 port_send_learning_packets(struct port *port)
3233 struct bridge *br = port->bridge;
3234 int error, n_packets, n_errors;
3235 struct mac_entry *e;
3237 error = n_packets = n_errors = 0;
3238 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3239 if (e->port.p != port) {
3240 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3250 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3251 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3252 "packets, last error was: %s",
3253 port->name, n_errors, n_packets, strerror(error));
3255 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3256 port->name, n_packets);
3260 /* Interface functions. */
3262 static struct iface *
3263 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3265 struct bridge *br = port->bridge;
3266 struct iface *iface;
3267 char *name = if_cfg->name;
3269 iface = xzalloc(sizeof *iface);
3271 iface->name = xstrdup(name);
3272 iface->dp_ifidx = -1;
3273 iface->tag = tag_create_random();
3274 iface->netdev = NULL;
3275 iface->cfg = if_cfg;
3277 shash_add_assert(&br->iface_by_name, iface->name, iface);
3279 list_push_back(&port->ifaces, &iface->port_elem);
3281 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3289 iface_destroy(struct iface *iface)
3292 struct port *port = iface->port;
3293 struct bridge *br = port->bridge;
3296 bond_slave_unregister(port->bond, iface);
3300 lacp_slave_unregister(port->lacp, iface);
3303 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3305 if (iface->dp_ifidx >= 0) {
3306 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3309 list_remove(&iface->port_elem);
3311 netdev_close(iface->netdev);
3316 bridge_flush(port->bridge);
3320 static struct iface *
3321 iface_lookup(const struct bridge *br, const char *name)
3323 return shash_find_data(&br->iface_by_name, name);
3326 static struct iface *
3327 iface_find(const char *name)
3329 const struct bridge *br;
3331 LIST_FOR_EACH (br, node, &all_bridges) {
3332 struct iface *iface = iface_lookup(br, name);
3341 static struct iface *
3342 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3344 struct iface *iface;
3346 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3347 hash_int(dp_ifidx, 0), &br->ifaces) {
3348 if (iface->dp_ifidx == dp_ifidx) {
3355 /* Set Ethernet address of 'iface', if one is specified in the configuration
3358 iface_set_mac(struct iface *iface)
3360 uint8_t ea[ETH_ADDR_LEN];
3362 if (!strcmp(iface->type, "internal")
3363 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3364 if (iface->dp_ifidx == ODPP_LOCAL) {
3365 VLOG_ERR("interface %s: ignoring mac in Interface record "
3366 "(use Bridge record to set local port's mac)",
3368 } else if (eth_addr_is_multicast(ea)) {
3369 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3372 int error = netdev_set_etheraddr(iface->netdev, ea);
3374 VLOG_ERR("interface %s: setting MAC failed (%s)",
3375 iface->name, strerror(error));
3381 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3383 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3385 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3386 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3390 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3392 * The value strings in '*shash' are taken directly from values[], not copied,
3393 * so the caller should not modify or free them. */
3395 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3396 struct shash *shash)
3401 for (i = 0; i < n; i++) {
3402 shash_add(shash, keys[i], values[i]);
3406 /* Creates 'keys' and 'values' arrays from 'shash'.
3408 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3409 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3410 * are populated with with strings taken directly from 'shash' and thus have
3411 * the same ownership of the key-value pairs in shash.
3414 shash_to_ovs_idl_map(struct shash *shash,
3415 char ***keys, char ***values, size_t *n)
3419 struct shash_node *sn;
3421 count = shash_count(shash);
3423 k = xmalloc(count * sizeof *k);
3424 v = xmalloc(count * sizeof *v);
3427 SHASH_FOR_EACH(sn, shash) {
3438 struct iface_delete_queues_cbdata {
3439 struct netdev *netdev;
3440 const struct ovsdb_datum *queues;
3444 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3446 union ovsdb_atom atom;
3448 atom.integer = target;
3449 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3453 iface_delete_queues(unsigned int queue_id,
3454 const struct shash *details OVS_UNUSED, void *cbdata_)
3456 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3458 if (!queue_ids_include(cbdata->queues, queue_id)) {
3459 netdev_delete_queue(cbdata->netdev, queue_id);
3464 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3466 if (!qos || qos->type[0] == '\0') {
3467 netdev_set_qos(iface->netdev, NULL, NULL);
3469 struct iface_delete_queues_cbdata cbdata;
3470 struct shash details;
3473 /* Configure top-level Qos for 'iface'. */
3474 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3475 qos->n_other_config, &details);
3476 netdev_set_qos(iface->netdev, qos->type, &details);
3477 shash_destroy(&details);
3479 /* Deconfigure queues that were deleted. */
3480 cbdata.netdev = iface->netdev;
3481 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3483 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3485 /* Configure queues for 'iface'. */
3486 for (i = 0; i < qos->n_queues; i++) {
3487 const struct ovsrec_queue *queue = qos->value_queues[i];
3488 unsigned int queue_id = qos->key_queues[i];
3490 shash_from_ovs_idl_map(queue->key_other_config,
3491 queue->value_other_config,
3492 queue->n_other_config, &details);
3493 netdev_set_queue(iface->netdev, queue_id, &details);
3494 shash_destroy(&details);
3500 iface_update_cfm(struct iface *iface)
3504 uint16_t *remote_mps;
3505 struct ovsrec_monitor *mon;
3506 uint8_t maid[CCM_MAID_LEN];
3508 mon = iface->cfg->monitor;
3511 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3515 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3516 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3520 cfm.mpid = mon->mpid;
3521 cfm.interval = mon->interval ? *mon->interval : 1000;
3523 memcpy(cfm.maid, maid, sizeof cfm.maid);
3525 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3526 for(i = 0; i < mon->n_remote_mps; i++) {
3527 remote_mps[i] = mon->remote_mps[i]->mpid;
3530 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3531 &cfm, remote_mps, mon->n_remote_mps);
3535 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3536 * how 'iface''s port is configured.
3538 * Returns true if 'iface' is up, false otherwise. */
3540 iface_get_carrier(const struct iface *iface)
3543 return netdev_get_carrier(iface->netdev);
3546 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3547 * instead of obtaining it from the database. */
3549 iface_is_synthetic(const struct iface *iface)
3551 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3554 /* Port mirroring. */
3556 static struct mirror *
3557 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3561 for (i = 0; i < MAX_MIRRORS; i++) {
3562 struct mirror *m = br->mirrors[i];
3563 if (m && uuid_equals(uuid, &m->uuid)) {
3571 mirror_reconfigure(struct bridge *br)
3573 unsigned long *rspan_vlans;
3577 /* Get rid of deleted mirrors. */
3578 for (i = 0; i < MAX_MIRRORS; i++) {
3579 struct mirror *m = br->mirrors[i];
3581 const struct ovsdb_datum *mc;
3582 union ovsdb_atom atom;
3584 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3585 atom.uuid = br->mirrors[i]->uuid;
3586 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3592 /* Add new mirrors and reconfigure existing ones. */
3593 for (i = 0; i < br->cfg->n_mirrors; i++) {
3594 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3595 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3597 mirror_reconfigure_one(m, cfg);
3599 mirror_create(br, cfg);
3603 /* Update port reserved status. */
3604 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3605 port->is_mirror_output_port = false;
3607 for (i = 0; i < MAX_MIRRORS; i++) {
3608 struct mirror *m = br->mirrors[i];
3609 if (m && m->out_port) {
3610 m->out_port->is_mirror_output_port = true;
3614 /* Update flooded vlans (for RSPAN). */
3616 if (br->cfg->n_flood_vlans) {
3617 rspan_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3618 br->cfg->n_flood_vlans);
3620 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3622 mac_learning_flush(br->ml);
3628 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3633 for (i = 0; ; i++) {
3634 if (i >= MAX_MIRRORS) {
3635 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3636 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3639 if (!br->mirrors[i]) {
3644 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3646 mac_learning_flush(br->ml);
3648 br->mirrors[i] = m = xzalloc(sizeof *m);
3649 m->uuid = cfg->header_.uuid;
3652 m->name = xstrdup(cfg->name);
3653 sset_init(&m->src_ports);
3654 sset_init(&m->dst_ports);
3660 mirror_reconfigure_one(m, cfg);
3664 mirror_destroy(struct mirror *m)
3667 struct bridge *br = m->bridge;
3670 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3671 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3672 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3675 sset_destroy(&m->src_ports);
3676 sset_destroy(&m->dst_ports);
3679 m->bridge->mirrors[m->idx] = NULL;
3684 mac_learning_flush(br->ml);
3689 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3694 for (i = 0; i < n_ports; i++) {
3695 const char *name = ports[i]->name;
3696 if (port_lookup(m->bridge, name)) {
3697 sset_add(names, name);
3699 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3700 "port %s", m->bridge->name, m->name, name);
3706 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3712 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3714 for (i = 0; i < cfg->n_select_vlan; i++) {
3715 int64_t vlan = cfg->select_vlan[i];
3716 if (vlan < 0 || vlan > 4095) {
3717 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3718 m->bridge->name, m->name, vlan);
3720 (*vlans)[n_vlans++] = vlan;
3727 vlan_is_mirrored(const struct mirror *m, int vlan)
3731 for (i = 0; i < m->n_vlans; i++) {
3732 if (m->vlans[i] == vlan) {
3740 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3742 struct sset src_ports, dst_ports;
3743 mirror_mask_t mirror_bit;
3744 struct port *out_port;
3751 if (strcmp(cfg->name, m->name)) {
3753 m->name = xstrdup(cfg->name);
3756 /* Get output port. */
3757 if (cfg->output_port) {
3758 out_port = port_lookup(m->bridge, cfg->output_port->name);
3760 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3761 m->bridge->name, m->name);
3767 if (cfg->output_vlan) {
3768 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3769 "output vlan; ignoring output vlan",
3770 m->bridge->name, m->name);
3772 } else if (cfg->output_vlan) {
3774 out_vlan = *cfg->output_vlan;
3776 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3777 m->bridge->name, m->name);
3782 sset_init(&src_ports);
3783 sset_init(&dst_ports);
3784 if (cfg->select_all) {
3785 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3786 sset_add(&src_ports, port->name);
3787 sset_add(&dst_ports, port->name);
3792 /* Get ports, and drop duplicates and ports that don't exist. */
3793 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3795 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3798 /* Get all the vlans, and drop duplicate and invalid vlans. */
3799 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3802 /* Update mirror data. */
3803 if (!sset_equals(&m->src_ports, &src_ports)
3804 || !sset_equals(&m->dst_ports, &dst_ports)
3805 || m->n_vlans != n_vlans
3806 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3807 || m->out_port != out_port
3808 || m->out_vlan != out_vlan) {
3809 bridge_flush(m->bridge);
3810 mac_learning_flush(m->bridge->ml);
3812 sset_swap(&m->src_ports, &src_ports);
3813 sset_swap(&m->dst_ports, &dst_ports);
3816 m->n_vlans = n_vlans;
3817 m->out_port = out_port;
3818 m->out_vlan = out_vlan;
3821 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3822 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3823 if (sset_contains(&m->src_ports, port->name)) {
3824 port->src_mirrors |= mirror_bit;
3826 port->src_mirrors &= ~mirror_bit;
3829 if (sset_contains(&m->dst_ports, port->name)) {
3830 port->dst_mirrors |= mirror_bit;
3832 port->dst_mirrors &= ~mirror_bit;
3837 sset_destroy(&src_ports);
3838 sset_destroy(&dst_ports);