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 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
423 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
424 * responsible for freeing '*managersp' (with free()).
426 * You may be asking yourself "why does ovs-vswitchd care?", because
427 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
428 * should not be and in fact is not directly involved in that. But
429 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
430 * it has to tell in-band control where the managers are to enable that.
431 * (Thus, only managers connected in-band are collected.)
434 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
435 struct sockaddr_in **managersp, size_t *n_managersp)
437 struct sockaddr_in *managers = NULL;
438 size_t n_managers = 0;
442 /* Collect all of the potential targets from the "targets" columns of the
443 * rows pointed to by "manager_options", excluding any that are
446 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
447 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
449 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
450 sset_find_and_delete(&targets, m->target);
452 sset_add(&targets, m->target);
456 /* Now extract the targets' IP addresses. */
457 if (!sset_is_empty(&targets)) {
460 managers = xmalloc(sset_count(&targets) * sizeof *managers);
461 SSET_FOR_EACH (target, &targets) {
462 struct sockaddr_in *sin = &managers[n_managers];
464 if ((!strncmp(target, "tcp:", 4)
465 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
466 (!strncmp(target, "ssl:", 4)
467 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
472 sset_destroy(&targets);
474 *managersp = managers;
475 *n_managersp = n_managers;
479 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
481 struct shash old_br, new_br;
482 struct shash_node *node;
483 struct bridge *br, *next;
484 struct sockaddr_in *managers;
487 int sflow_bridge_number;
489 COVERAGE_INC(bridge_reconfigure);
491 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
493 /* Collect old and new bridges. */
496 LIST_FOR_EACH (br, node, &all_bridges) {
497 shash_add(&old_br, br->name, br);
499 for (i = 0; i < ovs_cfg->n_bridges; i++) {
500 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
501 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
502 VLOG_WARN("more than one bridge named %s", br_cfg->name);
506 /* Get rid of deleted bridges and add new bridges. */
507 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
508 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
515 SHASH_FOR_EACH (node, &new_br) {
516 const char *br_name = node->name;
517 const struct ovsrec_bridge *br_cfg = node->data;
518 br = shash_find_data(&old_br, br_name);
520 /* If the bridge datapath type has changed, we need to tear it
521 * down and recreate. */
522 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
524 bridge_create(br_cfg);
527 bridge_create(br_cfg);
530 shash_destroy(&old_br);
531 shash_destroy(&new_br);
533 /* Reconfigure all bridges. */
534 LIST_FOR_EACH (br, node, &all_bridges) {
535 bridge_reconfigure_one(br);
538 /* Add and delete ports on all datapaths.
540 * The kernel will reject any attempt to add a given port to a datapath if
541 * that port already belongs to a different datapath, so we must do all
542 * port deletions before any port additions. */
543 LIST_FOR_EACH (br, node, &all_bridges) {
544 struct dpif_port_dump dump;
545 struct shash want_ifaces;
546 struct dpif_port dpif_port;
548 bridge_get_all_ifaces(br, &want_ifaces);
549 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
550 if (!shash_find(&want_ifaces, dpif_port.name)
551 && strcmp(dpif_port.name, br->name)) {
552 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
554 VLOG_WARN("failed to remove %s interface from %s: %s",
555 dpif_port.name, dpif_name(br->dpif),
560 shash_destroy(&want_ifaces);
562 LIST_FOR_EACH (br, node, &all_bridges) {
563 struct shash cur_ifaces, want_ifaces;
564 struct dpif_port_dump dump;
565 struct dpif_port dpif_port;
567 /* Get the set of interfaces currently in this datapath. */
568 shash_init(&cur_ifaces);
569 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
570 struct dpif_port *port_info = xmalloc(sizeof *port_info);
571 dpif_port_clone(port_info, &dpif_port);
572 shash_add(&cur_ifaces, dpif_port.name, port_info);
575 /* Get the set of interfaces we want on this datapath. */
576 bridge_get_all_ifaces(br, &want_ifaces);
578 hmap_clear(&br->ifaces);
579 SHASH_FOR_EACH (node, &want_ifaces) {
580 const char *if_name = node->name;
581 struct iface *iface = node->data;
582 struct dpif_port *dpif_port;
586 type = iface ? iface->type : "internal";
587 dpif_port = shash_find_data(&cur_ifaces, if_name);
589 /* If we have a port or a netdev already, and it's not the type we
590 * want, then delete the port (if any) and close the netdev (if
592 if ((dpif_port && strcmp(dpif_port->type, type))
593 || (iface && iface->netdev
594 && strcmp(type, netdev_get_type(iface->netdev)))) {
596 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
603 if (iface->port->bond) {
604 /* The bond has a pointer to the netdev, so remove it
605 * from the bond before closing the netdev. The slave
606 * will get added back to the bond later, after a new
607 * netdev is available. */
608 bond_slave_unregister(iface->port->bond, iface);
610 netdev_close(iface->netdev);
611 iface->netdev = NULL;
615 /* If the port doesn't exist or we don't have the netdev open,
616 * we need to do more work. */
617 if (!dpif_port || (iface && !iface->netdev)) {
618 struct netdev_options options;
619 struct netdev *netdev;
622 /* First open the network device. */
623 options.name = if_name;
625 options.args = &args;
626 options.ethertype = NETDEV_ETH_TYPE_NONE;
630 shash_from_ovs_idl_map(iface->cfg->key_options,
631 iface->cfg->value_options,
632 iface->cfg->n_options, &args);
634 error = netdev_open(&options, &netdev);
635 shash_destroy(&args);
638 VLOG_WARN("could not open network device %s (%s)",
639 if_name, strerror(error));
643 /* Then add the port if we haven't already. */
645 error = dpif_port_add(br->dpif, netdev, NULL);
647 netdev_close(netdev);
648 if (error == EFBIG) {
649 VLOG_ERR("ran out of valid port numbers on %s",
650 dpif_name(br->dpif));
653 VLOG_WARN("failed to add %s interface to %s: %s",
654 if_name, dpif_name(br->dpif),
661 /* Update 'iface'. */
663 iface->netdev = netdev;
665 } else if (iface && iface->netdev) {
669 shash_from_ovs_idl_map(iface->cfg->key_options,
670 iface->cfg->value_options,
671 iface->cfg->n_options, &args);
672 netdev_set_config(iface->netdev, &args);
673 shash_destroy(&args);
676 shash_destroy(&want_ifaces);
678 SHASH_FOR_EACH (node, &cur_ifaces) {
679 struct dpif_port *port_info = node->data;
680 dpif_port_destroy(port_info);
683 shash_destroy(&cur_ifaces);
685 sflow_bridge_number = 0;
686 LIST_FOR_EACH (br, node, &all_bridges) {
687 uint8_t ea[ETH_ADDR_LEN];
689 struct iface *local_iface;
690 struct port *port, *next_port;
691 struct iface *hw_addr_iface;
694 bridge_fetch_dp_ifaces(br);
696 /* Delete interfaces that cannot be opened.
698 * Following this loop, every remaining "struct iface" has nonnull
699 * 'netdev' and correct 'dp_ifidx'. */
700 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
701 struct iface *iface, *next_iface;
703 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
704 if (iface->netdev && iface->dp_ifidx >= 0) {
705 VLOG_DBG("%s has interface %s on port %d",
706 dpif_name(br->dpif), iface->name,
710 VLOG_ERR("%s interface not in %s, dropping",
711 iface->name, dpif_name(br->dpif));
713 /* We already reported a related error, don't bother
717 iface_set_ofport(iface->cfg, -1);
718 iface_destroy(iface);
722 if (list_is_empty(&port->ifaces)) {
723 VLOG_WARN("%s port has no interfaces, dropping", port->name);
728 /* Pick local port hardware address, datapath ID. */
729 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
730 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
732 int error = netdev_set_etheraddr(local_iface->netdev, ea);
734 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
735 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
736 "Ethernet address: %s",
737 br->name, strerror(error));
740 memcpy(br->ea, ea, ETH_ADDR_LEN);
742 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
743 ofproto_set_datapath_id(br->ofproto, dpid);
745 dpid_string = xasprintf("%016"PRIx64, dpid);
746 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
749 /* Set NetFlow configuration on this bridge. */
750 if (br->cfg->netflow) {
751 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
752 struct netflow_options opts;
754 memset(&opts, 0, sizeof opts);
756 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
757 if (nf_cfg->engine_type) {
758 opts.engine_type = *nf_cfg->engine_type;
760 if (nf_cfg->engine_id) {
761 opts.engine_id = *nf_cfg->engine_id;
764 opts.active_timeout = nf_cfg->active_timeout;
765 if (!opts.active_timeout) {
766 opts.active_timeout = -1;
767 } else if (opts.active_timeout < 0) {
768 VLOG_WARN("bridge %s: active timeout interval set to negative "
769 "value, using default instead (%d seconds)", br->name,
770 NF_ACTIVE_TIMEOUT_DEFAULT);
771 opts.active_timeout = -1;
774 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
775 if (opts.add_id_to_iface) {
776 if (opts.engine_id > 0x7f) {
777 VLOG_WARN("bridge %s: netflow port mangling may conflict "
778 "with another vswitch, choose an engine id less "
779 "than 128", br->name);
781 if (hmap_count(&br->ports) > 508) {
782 VLOG_WARN("bridge %s: netflow port mangling will conflict "
783 "with another port when more than 508 ports are "
788 sset_init(&opts.collectors);
789 sset_add_array(&opts.collectors,
790 nf_cfg->targets, nf_cfg->n_targets);
791 if (ofproto_set_netflow(br->ofproto, &opts)) {
792 VLOG_ERR("bridge %s: problem setting netflow collectors",
795 sset_destroy(&opts.collectors);
797 ofproto_set_netflow(br->ofproto, NULL);
800 /* Set sFlow configuration on this bridge. */
801 if (br->cfg->sflow) {
802 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
803 struct ovsrec_controller **controllers;
804 struct ofproto_sflow_options oso;
805 size_t n_controllers;
807 memset(&oso, 0, sizeof oso);
809 sset_init(&oso.targets);
810 sset_add_array(&oso.targets,
811 sflow_cfg->targets, sflow_cfg->n_targets);
813 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
814 if (sflow_cfg->sampling) {
815 oso.sampling_rate = *sflow_cfg->sampling;
818 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
819 if (sflow_cfg->polling) {
820 oso.polling_interval = *sflow_cfg->polling;
823 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
824 if (sflow_cfg->header) {
825 oso.header_len = *sflow_cfg->header;
828 oso.sub_id = sflow_bridge_number++;
829 oso.agent_device = sflow_cfg->agent;
831 oso.control_ip = NULL;
832 n_controllers = bridge_get_controllers(br, &controllers);
833 for (i = 0; i < n_controllers; i++) {
834 if (controllers[i]->local_ip) {
835 oso.control_ip = controllers[i]->local_ip;
839 ofproto_set_sflow(br->ofproto, &oso);
841 sset_destroy(&oso.targets);
843 ofproto_set_sflow(br->ofproto, NULL);
846 /* Update the controller and related settings. It would be more
847 * straightforward to call this from bridge_reconfigure_one(), but we
848 * can't do it there for two reasons. First, and most importantly, at
849 * that point we don't know the dp_ifidx of any interfaces that have
850 * been added to the bridge (because we haven't actually added them to
851 * the datapath). Second, at that point we haven't set the datapath ID
852 * yet; when a controller is configured, resetting the datapath ID will
853 * immediately disconnect from the controller, so it's better to set
854 * the datapath ID before the controller. */
855 bridge_reconfigure_remotes(br, managers, n_managers);
857 LIST_FOR_EACH (br, node, &all_bridges) {
860 br->has_bonded_ports = false;
861 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
864 port_reconfigure_lacp(port);
865 port_reconfigure_bond(port);
867 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
868 iface_update_qos(iface, port->cfg->qos);
869 netdev_set_policing(iface->netdev,
870 iface->cfg->ingress_policing_rate,
871 iface->cfg->ingress_policing_burst);
872 iface_set_mac(iface);
877 /* Some reconfiguration operations require the bridge to have been run at
879 LIST_FOR_EACH (br, node, &all_bridges) {
884 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
885 iface_update_cfm(iface);
891 /* ovs-vswitchd has completed initialization, so allow the process that
892 * forked us to exit successfully. */
893 daemonize_complete();
897 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
898 const struct ovsdb_idl_column *column,
901 const struct ovsdb_datum *datum;
902 union ovsdb_atom atom;
905 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
906 atom.string = (char *) key;
907 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
908 return idx == UINT_MAX ? NULL : datum->values[idx].string;
912 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
914 return get_ovsrec_key_value(&br_cfg->header_,
915 &ovsrec_bridge_col_other_config, key);
919 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
920 struct iface **hw_addr_iface)
926 *hw_addr_iface = NULL;
928 /* Did the user request a particular MAC? */
929 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
930 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
931 if (eth_addr_is_multicast(ea)) {
932 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
933 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
934 } else if (eth_addr_is_zero(ea)) {
935 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
941 /* Otherwise choose the minimum non-local MAC address among all of the
943 memset(ea, 0xff, ETH_ADDR_LEN);
944 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
945 uint8_t iface_ea[ETH_ADDR_LEN];
946 struct iface *candidate;
949 /* Mirror output ports don't participate. */
950 if (port->is_mirror_output_port) {
954 /* Choose the MAC address to represent the port. */
956 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
957 /* Find the interface with this Ethernet address (if any) so that
958 * we can provide the correct devname to the caller. */
959 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
960 uint8_t candidate_ea[ETH_ADDR_LEN];
961 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
962 && eth_addr_equals(iface_ea, candidate_ea)) {
967 /* Choose the interface whose MAC address will represent the port.
968 * The Linux kernel bonding code always chooses the MAC address of
969 * the first slave added to a bond, and the Fedora networking
970 * scripts always add slaves to a bond in alphabetical order, so
971 * for compatibility we choose the interface with the name that is
972 * first in alphabetical order. */
973 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
974 if (!iface || strcmp(candidate->name, iface->name) < 0) {
979 /* The local port doesn't count (since we're trying to choose its
980 * MAC address anyway). */
981 if (iface->dp_ifidx == ODPP_LOCAL) {
986 error = netdev_get_etheraddr(iface->netdev, iface_ea);
988 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
989 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
990 iface->name, strerror(error));
995 /* Compare against our current choice. */
996 if (!eth_addr_is_multicast(iface_ea) &&
997 !eth_addr_is_local(iface_ea) &&
998 !eth_addr_is_reserved(iface_ea) &&
999 !eth_addr_is_zero(iface_ea) &&
1000 eth_addr_compare_3way(iface_ea, ea) < 0)
1002 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1003 *hw_addr_iface = iface;
1006 if (eth_addr_is_multicast(ea)) {
1007 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1008 *hw_addr_iface = NULL;
1009 VLOG_WARN("bridge %s: using default bridge Ethernet "
1010 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1012 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1013 br->name, ETH_ADDR_ARGS(ea));
1017 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1018 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1019 * an interface on 'br', then that interface must be passed in as
1020 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1021 * 'hw_addr_iface' must be passed in as a null pointer. */
1023 bridge_pick_datapath_id(struct bridge *br,
1024 const uint8_t bridge_ea[ETH_ADDR_LEN],
1025 struct iface *hw_addr_iface)
1028 * The procedure for choosing a bridge MAC address will, in the most
1029 * ordinary case, also choose a unique MAC that we can use as a datapath
1030 * ID. In some special cases, though, multiple bridges will end up with
1031 * the same MAC address. This is OK for the bridges, but it will confuse
1032 * the OpenFlow controller, because each datapath needs a unique datapath
1035 * Datapath IDs must be unique. It is also very desirable that they be
1036 * stable from one run to the next, so that policy set on a datapath
1039 const char *datapath_id;
1042 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1043 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1047 if (hw_addr_iface) {
1049 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1051 * A bridge whose MAC address is taken from a VLAN network device
1052 * (that is, a network device created with vconfig(8) or similar
1053 * tool) will have the same MAC address as a bridge on the VLAN
1054 * device's physical network device.
1056 * Handle this case by hashing the physical network device MAC
1057 * along with the VLAN identifier.
1059 uint8_t buf[ETH_ADDR_LEN + 2];
1060 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1061 buf[ETH_ADDR_LEN] = vlan >> 8;
1062 buf[ETH_ADDR_LEN + 1] = vlan;
1063 return dpid_from_hash(buf, sizeof buf);
1066 * Assume that this bridge's MAC address is unique, since it
1067 * doesn't fit any of the cases we handle specially.
1072 * A purely internal bridge, that is, one that has no non-virtual
1073 * network devices on it at all, is more difficult because it has no
1074 * natural unique identifier at all.
1076 * When the host is a XenServer, we handle this case by hashing the
1077 * host's UUID with the name of the bridge. Names of bridges are
1078 * persistent across XenServer reboots, although they can be reused if
1079 * an internal network is destroyed and then a new one is later
1080 * created, so this is fairly effective.
1082 * When the host is not a XenServer, we punt by using a random MAC
1083 * address on each run.
1085 const char *host_uuid = xenserver_get_host_uuid();
1087 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1088 dpid = dpid_from_hash(combined, strlen(combined));
1094 return eth_addr_to_uint64(bridge_ea);
1098 dpid_from_hash(const void *data, size_t n)
1100 uint8_t hash[SHA1_DIGEST_SIZE];
1102 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1103 sha1_bytes(data, n, hash);
1104 eth_addr_mark_random(hash);
1105 return eth_addr_to_uint64(hash);
1109 iface_refresh_status(struct iface *iface)
1113 enum netdev_flags flags;
1120 if (iface_is_synthetic(iface)) {
1126 if (!netdev_get_status(iface->netdev, &sh)) {
1128 char **keys, **values;
1130 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1131 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1136 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1139 shash_destroy_free_data(&sh);
1141 error = netdev_get_flags(iface->netdev, &flags);
1143 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1146 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1149 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1151 ovsrec_interface_set_duplex(iface->cfg,
1152 netdev_features_is_full_duplex(current)
1154 /* warning: uint64_t -> int64_t conversion */
1155 bps = netdev_features_to_bps(current);
1156 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1159 ovsrec_interface_set_duplex(iface->cfg, NULL);
1160 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1164 ovsrec_interface_set_link_state(iface->cfg,
1165 iface_get_carrier(iface) ? "up" : "down");
1167 error = netdev_get_mtu(iface->netdev, &mtu);
1168 if (!error && mtu != INT_MAX) {
1170 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1173 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1177 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1178 * changed, false otherwise. */
1180 iface_refresh_cfm_stats(struct iface *iface)
1182 const struct ovsrec_monitor *mon;
1183 const struct cfm *cfm;
1184 bool changed = false;
1187 mon = iface->cfg->monitor;
1188 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1194 for (i = 0; i < mon->n_remote_mps; i++) {
1195 const struct ovsrec_maintenance_point *mp;
1196 const struct remote_mp *rmp;
1198 mp = mon->remote_mps[i];
1199 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1201 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1202 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1207 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1208 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1216 iface_refresh_lacp_stats(struct iface *iface)
1218 bool *db_current = iface->cfg->lacp_current;
1219 bool changed = false;
1221 if (iface->port->lacp) {
1222 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1224 if (!db_current || *db_current != current) {
1226 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1228 } else if (db_current) {
1230 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1237 iface_refresh_stats(struct iface *iface)
1243 static const struct iface_stat iface_stats[] = {
1244 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1245 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1246 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1247 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1248 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1249 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1250 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1251 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1252 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1253 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1254 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1255 { "collisions", offsetof(struct netdev_stats, collisions) },
1257 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1258 const struct iface_stat *s;
1260 char *keys[N_STATS];
1261 int64_t values[N_STATS];
1264 struct netdev_stats stats;
1266 if (iface_is_synthetic(iface)) {
1270 /* Intentionally ignore return value, since errors will set 'stats' to
1271 * all-1s, and we will deal with that correctly below. */
1272 netdev_get_stats(iface->netdev, &stats);
1275 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1276 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1277 if (value != UINT64_MAX) {
1284 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1288 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1290 struct ovsdb_datum datum;
1294 get_system_stats(&stats);
1296 ovsdb_datum_from_shash(&datum, &stats);
1297 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1301 static inline const char *
1302 nx_role_to_str(enum nx_role role)
1307 case NX_ROLE_MASTER:
1312 return "*** INVALID ROLE ***";
1317 bridge_refresh_controller_status(const struct bridge *br)
1320 const struct ovsrec_controller *cfg;
1322 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1324 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1325 struct ofproto_controller_info *cinfo =
1326 shash_find_data(&info, cfg->target);
1329 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1330 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1331 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1332 (char **) cinfo->pairs.values,
1335 ovsrec_controller_set_is_connected(cfg, false);
1336 ovsrec_controller_set_role(cfg, NULL);
1337 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1341 ofproto_free_ofproto_controller_info(&info);
1347 const struct ovsrec_open_vswitch *cfg;
1349 bool datapath_destroyed;
1350 bool database_changed;
1353 /* Let each bridge do the work that it needs to do. */
1354 datapath_destroyed = false;
1355 LIST_FOR_EACH (br, node, &all_bridges) {
1356 int error = bridge_run_one(br);
1358 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1359 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1360 "forcing reconfiguration", br->name);
1361 datapath_destroyed = true;
1365 /* (Re)configure if necessary. */
1366 database_changed = ovsdb_idl_run(idl);
1367 cfg = ovsrec_open_vswitch_first(idl);
1369 /* Re-configure SSL. We do this on every trip through the main loop,
1370 * instead of just when the database changes, because the contents of the
1371 * key and certificate files can change without the database changing.
1373 * We do this before bridge_reconfigure() because that function might
1374 * initiate SSL connections and thus requires SSL to be configured. */
1375 if (cfg && cfg->ssl) {
1376 const struct ovsrec_ssl *ssl = cfg->ssl;
1378 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1379 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1382 if (database_changed || datapath_destroyed) {
1384 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1386 bridge_configure_once(cfg);
1387 bridge_reconfigure(cfg);
1389 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1390 ovsdb_idl_txn_commit(txn);
1391 ovsdb_idl_txn_destroy(txn); /* XXX */
1393 /* We still need to reconfigure to avoid dangling pointers to
1394 * now-destroyed ovsrec structures inside bridge data. */
1395 static const struct ovsrec_open_vswitch null_cfg;
1397 bridge_reconfigure(&null_cfg);
1401 /* Refresh system and interface stats if necessary. */
1402 if (time_msec() >= stats_timer) {
1404 struct ovsdb_idl_txn *txn;
1406 txn = ovsdb_idl_txn_create(idl);
1407 LIST_FOR_EACH (br, node, &all_bridges) {
1410 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1411 struct iface *iface;
1413 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1414 iface_refresh_stats(iface);
1415 iface_refresh_status(iface);
1418 bridge_refresh_controller_status(br);
1420 refresh_system_stats(cfg);
1421 ovsdb_idl_txn_commit(txn);
1422 ovsdb_idl_txn_destroy(txn); /* XXX */
1425 stats_timer = time_msec() + STATS_INTERVAL;
1428 if (time_msec() >= db_limiter) {
1429 struct ovsdb_idl_txn *txn;
1430 bool changed = false;
1432 txn = ovsdb_idl_txn_create(idl);
1433 LIST_FOR_EACH (br, node, &all_bridges) {
1436 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1437 struct iface *iface;
1439 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1440 changed = iface_refresh_cfm_stats(iface) || changed;
1441 changed = iface_refresh_lacp_stats(iface) || changed;
1447 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1450 ovsdb_idl_txn_commit(txn);
1451 ovsdb_idl_txn_destroy(txn);
1460 LIST_FOR_EACH (br, node, &all_bridges) {
1463 ofproto_wait(br->ofproto);
1464 mac_learning_wait(br->ml);
1465 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1469 ovsdb_idl_wait(idl);
1470 poll_timer_wait_until(stats_timer);
1472 if (db_limiter > time_msec()) {
1473 poll_timer_wait_until(db_limiter);
1477 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1478 * configuration changes. */
1480 bridge_flush(struct bridge *br)
1482 COVERAGE_INC(bridge_flush);
1486 /* Bridge unixctl user interface functions. */
1488 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1489 const char *args, void *aux OVS_UNUSED)
1491 struct ds ds = DS_EMPTY_INITIALIZER;
1492 const struct bridge *br;
1493 const struct mac_entry *e;
1495 br = bridge_lookup(args);
1497 unixctl_command_reply(conn, 501, "no such bridge");
1501 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1502 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1503 struct port *port = e->port.p;
1504 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1505 port_get_an_iface(port)->dp_ifidx,
1506 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1508 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1512 /* CFM unixctl user interface functions. */
1514 cfm_unixctl_show(struct unixctl_conn *conn,
1515 const char *args, void *aux OVS_UNUSED)
1517 struct ds ds = DS_EMPTY_INITIALIZER;
1518 struct iface *iface;
1519 const struct cfm *cfm;
1521 iface = iface_find(args);
1523 unixctl_command_reply(conn, 501, "no such interface");
1527 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1530 unixctl_command_reply(conn, 501, "CFM not enabled");
1534 cfm_dump_ds(cfm, &ds);
1535 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1539 /* QoS unixctl user interface functions. */
1541 struct qos_unixctl_show_cbdata {
1543 struct iface *iface;
1547 qos_unixctl_show_cb(unsigned int queue_id,
1548 const struct shash *details,
1551 struct qos_unixctl_show_cbdata *data = aux;
1552 struct ds *ds = data->ds;
1553 struct iface *iface = data->iface;
1554 struct netdev_queue_stats stats;
1555 struct shash_node *node;
1558 ds_put_cstr(ds, "\n");
1560 ds_put_format(ds, "Queue %u:\n", queue_id);
1562 ds_put_cstr(ds, "Default:\n");
1565 SHASH_FOR_EACH (node, details) {
1566 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1569 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1571 if (stats.tx_packets != UINT64_MAX) {
1572 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1575 if (stats.tx_bytes != UINT64_MAX) {
1576 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1579 if (stats.tx_errors != UINT64_MAX) {
1580 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1583 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1584 queue_id, strerror(error));
1589 qos_unixctl_show(struct unixctl_conn *conn,
1590 const char *args, void *aux OVS_UNUSED)
1592 struct ds ds = DS_EMPTY_INITIALIZER;
1593 struct shash sh = SHASH_INITIALIZER(&sh);
1594 struct iface *iface;
1596 struct shash_node *node;
1597 struct qos_unixctl_show_cbdata data;
1600 iface = iface_find(args);
1602 unixctl_command_reply(conn, 501, "no such interface");
1606 netdev_get_qos(iface->netdev, &type, &sh);
1608 if (*type != '\0') {
1609 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1611 SHASH_FOR_EACH (node, &sh) {
1612 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1617 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1620 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1622 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1624 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1625 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1628 shash_destroy_free_data(&sh);
1632 /* Bridge reconfiguration functions. */
1633 static struct bridge *
1634 bridge_create(const struct ovsrec_bridge *br_cfg)
1639 assert(!bridge_lookup(br_cfg->name));
1640 br = xzalloc(sizeof *br);
1642 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1649 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1652 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1654 dpif_delete(br->dpif);
1655 dpif_close(br->dpif);
1660 br->name = xstrdup(br_cfg->name);
1662 br->ml = mac_learning_create();
1663 eth_addr_nicira_random(br->default_ea);
1665 hmap_init(&br->ports);
1666 hmap_init(&br->ifaces);
1667 shash_init(&br->iface_by_name);
1671 list_push_back(&all_bridges, &br->node);
1673 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1679 bridge_destroy(struct bridge *br)
1682 struct port *port, *next;
1686 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1689 for (i = 0; i < MAX_MIRRORS; i++) {
1690 mirror_destroy(br->mirrors[i]);
1692 list_remove(&br->node);
1693 ofproto_destroy(br->ofproto);
1694 error = dpif_delete(br->dpif);
1695 if (error && error != ENOENT) {
1696 VLOG_ERR("failed to delete %s: %s",
1697 dpif_name(br->dpif), strerror(error));
1699 dpif_close(br->dpif);
1700 mac_learning_destroy(br->ml);
1701 hmap_destroy(&br->ifaces);
1702 hmap_destroy(&br->ports);
1703 shash_destroy(&br->iface_by_name);
1704 free(br->synth_local_iface.type);
1710 static struct bridge *
1711 bridge_lookup(const char *name)
1715 LIST_FOR_EACH (br, node, &all_bridges) {
1716 if (!strcmp(br->name, name)) {
1723 /* Handle requests for a listing of all flows known by the OpenFlow
1724 * stack, including those normally hidden. */
1726 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1727 const char *args, void *aux OVS_UNUSED)
1732 br = bridge_lookup(args);
1734 unixctl_command_reply(conn, 501, "Unknown bridge");
1739 ofproto_get_all_flows(br->ofproto, &results);
1741 unixctl_command_reply(conn, 200, ds_cstr(&results));
1742 ds_destroy(&results);
1745 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1746 * connections and reconnect. If BRIDGE is not specified, then all bridges
1747 * drop their controller connections and reconnect. */
1749 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1750 const char *args, void *aux OVS_UNUSED)
1753 if (args[0] != '\0') {
1754 br = bridge_lookup(args);
1756 unixctl_command_reply(conn, 501, "Unknown bridge");
1759 ofproto_reconnect_controllers(br->ofproto);
1761 LIST_FOR_EACH (br, node, &all_bridges) {
1762 ofproto_reconnect_controllers(br->ofproto);
1765 unixctl_command_reply(conn, 200, NULL);
1769 bridge_run_one(struct bridge *br)
1774 error = ofproto_run1(br->ofproto);
1779 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1781 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1785 error = ofproto_run2(br->ofproto, br->flush);
1792 bridge_get_controllers(const struct bridge *br,
1793 struct ovsrec_controller ***controllersp)
1795 struct ovsrec_controller **controllers;
1796 size_t n_controllers;
1798 controllers = br->cfg->controller;
1799 n_controllers = br->cfg->n_controller;
1801 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1807 *controllersp = controllers;
1809 return n_controllers;
1813 bridge_reconfigure_one(struct bridge *br)
1815 enum ofproto_fail_mode fail_mode;
1816 struct port *port, *next;
1817 struct shash_node *node;
1818 struct shash new_ports;
1821 /* Collect new ports. */
1822 shash_init(&new_ports);
1823 for (i = 0; i < br->cfg->n_ports; i++) {
1824 const char *name = br->cfg->ports[i]->name;
1825 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1826 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1830 if (!shash_find(&new_ports, br->name)) {
1831 struct dpif_port dpif_port;
1834 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1835 br->name, br->name);
1837 dpif_port_query_by_number(br->dpif, ODPP_LOCAL, &dpif_port);
1838 type = xstrdup(dpif_port.type ? dpif_port.type : "internal");
1839 dpif_port_destroy(&dpif_port);
1841 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1842 br->synth_local_port.n_interfaces = 1;
1843 br->synth_local_port.name = br->name;
1845 br->synth_local_iface.name = br->name;
1846 free(br->synth_local_iface.type);
1847 br->synth_local_iface.type = type;
1849 br->synth_local_ifacep = &br->synth_local_iface;
1851 shash_add(&new_ports, br->name, &br->synth_local_port);
1854 /* Get rid of deleted ports.
1855 * Get rid of deleted interfaces on ports that still exist. */
1856 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1857 const struct ovsrec_port *port_cfg;
1859 port_cfg = shash_find_data(&new_ports, port->name);
1863 port_del_ifaces(port, port_cfg);
1867 /* Create new ports.
1868 * Add new interfaces to existing ports.
1869 * Reconfigure existing ports. */
1870 SHASH_FOR_EACH (node, &new_ports) {
1871 struct port *port = port_lookup(br, node->name);
1873 port = port_create(br, node->name);
1876 port_reconfigure(port, node->data);
1877 if (list_is_empty(&port->ifaces)) {
1878 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1879 br->name, port->name);
1883 shash_destroy(&new_ports);
1885 /* Set the fail-mode */
1886 fail_mode = !br->cfg->fail_mode
1887 || !strcmp(br->cfg->fail_mode, "standalone")
1888 ? OFPROTO_FAIL_STANDALONE
1889 : OFPROTO_FAIL_SECURE;
1890 ofproto_set_fail_mode(br->ofproto, fail_mode);
1892 /* Configure OpenFlow controller connection snooping. */
1893 if (!ofproto_has_snoops(br->ofproto)) {
1897 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1898 ovs_rundir(), br->name));
1899 ofproto_set_snoops(br->ofproto, &snoops);
1900 sset_destroy(&snoops);
1903 mirror_reconfigure(br);
1906 /* Initializes 'oc' appropriately as a management service controller for
1909 * The caller must free oc->target when it is no longer needed. */
1911 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1912 struct ofproto_controller *oc)
1914 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1915 oc->max_backoff = 0;
1916 oc->probe_interval = 60;
1917 oc->band = OFPROTO_OUT_OF_BAND;
1919 oc->burst_limit = 0;
1922 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1924 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1925 struct ofproto_controller *oc)
1927 oc->target = c->target;
1928 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1929 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1930 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1931 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1932 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1933 oc->burst_limit = (c->controller_burst_limit
1934 ? *c->controller_burst_limit : 0);
1937 /* Configures the IP stack for 'br''s local interface properly according to the
1938 * configuration in 'c'. */
1940 bridge_configure_local_iface_netdev(struct bridge *br,
1941 struct ovsrec_controller *c)
1943 struct netdev *netdev;
1944 struct in_addr mask, gateway;
1946 struct iface *local_iface;
1949 /* If there's no local interface or no IP address, give up. */
1950 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1951 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1955 /* Bring up the local interface. */
1956 netdev = local_iface->netdev;
1957 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1959 /* Configure the IP address and netmask. */
1960 if (!c->local_netmask
1961 || !inet_aton(c->local_netmask, &mask)
1963 mask.s_addr = guess_netmask(ip.s_addr);
1965 if (!netdev_set_in4(netdev, ip, mask)) {
1966 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1967 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1970 /* Configure the default gateway. */
1971 if (c->local_gateway
1972 && inet_aton(c->local_gateway, &gateway)
1973 && gateway.s_addr) {
1974 if (!netdev_add_router(netdev, gateway)) {
1975 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1976 br->name, IP_ARGS(&gateway.s_addr));
1982 bridge_reconfigure_remotes(struct bridge *br,
1983 const struct sockaddr_in *managers,
1986 const char *disable_ib_str, *queue_id_str;
1987 bool disable_in_band = false;
1990 struct ovsrec_controller **controllers;
1991 size_t n_controllers;
1993 struct ofproto_controller *ocs;
1997 /* Check if we should disable in-band control on this bridge. */
1998 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1999 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
2000 disable_in_band = true;
2003 /* Set OpenFlow queue ID for in-band control. */
2004 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2005 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2006 ofproto_set_in_band_queue(br->ofproto, queue_id);
2008 if (disable_in_band) {
2009 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2011 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2014 n_controllers = bridge_get_controllers(br, &controllers);
2016 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2019 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2020 for (i = 0; i < n_controllers; i++) {
2021 struct ovsrec_controller *c = controllers[i];
2023 if (!strncmp(c->target, "punix:", 6)
2024 || !strncmp(c->target, "unix:", 5)) {
2025 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2027 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2028 * domain sockets and overwriting arbitrary local files. */
2029 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
2030 "\"%s\" due to possibility for remote exploit",
2031 dpif_name(br->dpif), c->target);
2035 bridge_configure_local_iface_netdev(br, c);
2036 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2037 if (disable_in_band) {
2038 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2043 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2044 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2049 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2054 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2055 struct iface *iface;
2057 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2058 shash_add_once(ifaces, iface->name, iface);
2060 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2061 shash_add_once(ifaces, port->name, NULL);
2066 /* For robustness, in case the administrator moves around datapath ports behind
2067 * our back, we re-check all the datapath port numbers here.
2069 * This function will set the 'dp_ifidx' members of interfaces that have
2070 * disappeared to -1, so only call this function from a context where those
2071 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2072 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2073 * datapath, which doesn't support UINT16_MAX+1 ports. */
2075 bridge_fetch_dp_ifaces(struct bridge *br)
2077 struct dpif_port_dump dump;
2078 struct dpif_port dpif_port;
2081 /* Reset all interface numbers. */
2082 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2083 struct iface *iface;
2085 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2086 iface->dp_ifidx = -1;
2089 hmap_clear(&br->ifaces);
2091 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2092 struct iface *iface = iface_lookup(br, dpif_port.name);
2094 if (iface->dp_ifidx >= 0) {
2095 VLOG_WARN("%s reported interface %s twice",
2096 dpif_name(br->dpif), dpif_port.name);
2097 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2098 VLOG_WARN("%s reported interface %"PRIu16" twice",
2099 dpif_name(br->dpif), dpif_port.port_no);
2101 iface->dp_ifidx = dpif_port.port_no;
2102 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2103 hash_int(iface->dp_ifidx, 0));
2106 iface_set_ofport(iface->cfg,
2107 (iface->dp_ifidx >= 0
2108 ? odp_port_to_ofp_port(iface->dp_ifidx)
2114 /* Bridge packet processing functions. */
2117 set_dst(struct dst *dst, const struct flow *flow,
2118 const struct port *in_port, const struct port *out_port,
2121 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2122 : in_port->vlan >= 0 ? in_port->vlan
2123 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2124 : vlan_tci_to_vid(flow->vlan_tci));
2126 dst->iface = (!out_port->bond
2127 ? port_get_an_iface(out_port)
2128 : bond_choose_output_slave(out_port->bond, flow,
2131 return dst->iface != NULL;
2135 mirror_mask_ffs(mirror_mask_t mask)
2137 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2142 dst_set_init(struct dst_set *set)
2144 set->dsts = set->builtin;
2146 set->allocated = ARRAY_SIZE(set->builtin);
2150 dst_set_add(struct dst_set *set, const struct dst *dst)
2152 if (set->n >= set->allocated) {
2153 size_t new_allocated;
2154 struct dst *new_dsts;
2156 new_allocated = set->allocated * 2;
2157 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2158 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2162 set->dsts = new_dsts;
2163 set->allocated = new_allocated;
2165 set->dsts[set->n++] = *dst;
2169 dst_set_free(struct dst_set *set)
2171 if (set->dsts != set->builtin) {
2177 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2180 for (i = 0; i < set->n; i++) {
2181 if (set->dsts[i].vlan == test->vlan
2182 && set->dsts[i].iface == test->iface) {
2190 port_trunks_vlan(const struct port *port, uint16_t vlan)
2192 return (port->vlan < 0 || vlan_bitmap_contains(port->trunks, vlan));
2196 port_includes_vlan(const struct port *port, uint16_t vlan)
2198 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2202 port_is_floodable(const struct port *port)
2204 struct iface *iface;
2206 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2207 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2215 /* Returns an arbitrary interface within 'port'. */
2216 static struct iface *
2217 port_get_an_iface(const struct port *port)
2219 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2223 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2224 const struct port *in_port, const struct port *out_port,
2225 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2229 if (out_port == FLOOD_PORT) {
2232 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2234 && port_is_floodable(port)
2235 && port_includes_vlan(port, vlan)
2236 && !port->is_mirror_output_port
2237 && set_dst(&dst, flow, in_port, port, tags)) {
2238 dst_set_add(set, &dst);
2241 *nf_output_iface = NF_OUT_FLOOD;
2242 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2243 dst_set_add(set, &dst);
2244 *nf_output_iface = dst.iface->dp_ifidx;
2249 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2250 uint16_t vlan, const struct port *in_port,
2251 struct dst_set *set, tag_type *tags)
2253 mirror_mask_t mirrors;
2257 mirrors = in_port->src_mirrors;
2258 for (i = 0; i < set->n; i++) {
2259 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2266 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2267 if (flow_vlan == 0) {
2268 flow_vlan = OFP_VLAN_NONE;
2272 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2273 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2277 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2278 && !dst_is_duplicate(set, &dst)) {
2279 dst_set_add(set, &dst);
2284 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2285 if (port_includes_vlan(port, m->out_vlan)
2286 && set_dst(&dst, flow, in_port, port, tags))
2288 if (port->vlan < 0) {
2289 dst.vlan = m->out_vlan;
2291 if (dst_is_duplicate(set, &dst)) {
2295 /* Use the vlan tag on the original flow instead of
2296 * the one passed in the vlan parameter. This ensures
2297 * that we compare the vlan from before any implicit
2298 * tagging tags place. This is necessary because
2299 * dst->vlan is the final vlan, after removing implicit
2301 if (port == in_port && dst.vlan == flow_vlan) {
2302 /* Don't send out input port on same VLAN. */
2305 dst_set_add(set, &dst);
2310 mirrors &= mirrors - 1;
2315 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2316 const struct port *in_port, const struct port *out_port,
2317 tag_type *tags, struct ofpbuf *actions,
2318 uint16_t *nf_output_iface)
2320 uint16_t initial_vlan, cur_vlan;
2321 const struct dst *dst;
2325 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2327 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2329 /* Output all the packets we can without having to change the VLAN. */
2330 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2331 if (initial_vlan == 0) {
2332 initial_vlan = OFP_VLAN_NONE;
2334 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2335 if (dst->vlan != initial_vlan) {
2338 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2341 /* Then output the rest. */
2342 cur_vlan = initial_vlan;
2343 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2344 if (dst->vlan == initial_vlan) {
2347 if (dst->vlan != cur_vlan) {
2348 if (dst->vlan == OFP_VLAN_NONE) {
2349 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2352 tci = htons(dst->vlan & VLAN_VID_MASK);
2353 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2354 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2356 cur_vlan = dst->vlan;
2358 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2364 /* Returns the effective vlan of a packet, taking into account both the
2365 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2366 * the packet is untagged and -1 indicates it has an invalid header and
2367 * should be dropped. */
2368 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2369 struct port *in_port, bool have_packet)
2371 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2372 if (in_port->vlan >= 0) {
2375 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2376 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2377 "packet received on port %s configured with "
2378 "implicit VLAN %"PRIu16,
2379 br->name, vlan, in_port->name, in_port->vlan);
2383 vlan = in_port->vlan;
2385 if (!port_includes_vlan(in_port, vlan)) {
2387 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2388 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2389 "packet received on port %s not configured for "
2391 br->name, vlan, in_port->name, vlan);
2400 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2401 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2402 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2404 is_gratuitous_arp(const struct flow *flow)
2406 return (flow->dl_type == htons(ETH_TYPE_ARP)
2407 && eth_addr_is_broadcast(flow->dl_dst)
2408 && (flow->nw_proto == ARP_OP_REPLY
2409 || (flow->nw_proto == ARP_OP_REQUEST
2410 && flow->nw_src == flow->nw_dst)));
2414 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2415 struct port *in_port)
2417 struct mac_entry *mac;
2419 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2423 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2424 if (is_gratuitous_arp(flow)) {
2425 /* We don't want to learn from gratuitous ARP packets that are
2426 * reflected back over bond slaves so we lock the learning table. */
2427 if (!in_port->bond) {
2428 mac_entry_set_grat_arp_lock(mac);
2429 } else if (mac_entry_is_grat_arp_locked(mac)) {
2434 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2435 /* The log messages here could actually be useful in debugging,
2436 * so keep the rate limit relatively high. */
2437 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2438 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2439 "on port %s in VLAN %d",
2440 br->name, ETH_ADDR_ARGS(flow->dl_src),
2441 in_port->name, vlan);
2443 mac->port.p = in_port;
2444 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2448 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2449 * dropped. Returns true if they may be forwarded, false if they should be
2452 * If 'have_packet' is true, it indicates that the caller is processing a
2453 * received packet. If 'have_packet' is false, then the caller is just
2454 * revalidating an existing flow because configuration has changed. Either
2455 * way, 'have_packet' only affects logging (there is no point in logging errors
2456 * during revalidation).
2458 * Sets '*in_portp' to the input port. This will be a null pointer if
2459 * flow->in_port does not designate a known input port (in which case
2460 * is_admissible() returns false).
2462 * When returning true, sets '*vlanp' to the effective VLAN of the input
2463 * packet, as returned by flow_get_vlan().
2465 * May also add tags to '*tags', although the current implementation only does
2466 * so in one special case.
2469 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2470 tag_type *tags, int *vlanp, struct port **in_portp)
2472 struct iface *in_iface;
2473 struct port *in_port;
2476 /* Find the interface and port structure for the received packet. */
2477 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2479 /* No interface? Something fishy... */
2481 /* Odd. A few possible reasons here:
2483 * - We deleted an interface but there are still a few packets
2484 * queued up from it.
2486 * - Someone externally added an interface (e.g. with "ovs-dpctl
2487 * add-if") that we don't know about.
2489 * - Packet arrived on the local port but the local port is not
2490 * one of our bridge ports.
2492 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2494 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2495 "interface %"PRIu16, br->name, flow->in_port);
2501 *in_portp = in_port = in_iface->port;
2502 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2507 /* Drop frames for reserved multicast addresses. */
2508 if (eth_addr_is_reserved(flow->dl_dst)) {
2512 /* Drop frames on ports reserved for mirroring. */
2513 if (in_port->is_mirror_output_port) {
2515 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2516 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2517 "%s, which is reserved exclusively for mirroring",
2518 br->name, in_port->name);
2523 if (in_port->bond) {
2524 struct mac_entry *mac;
2526 switch (bond_check_admissibility(in_port->bond, in_iface,
2527 flow->dl_dst, tags)) {
2534 case BV_DROP_IF_MOVED:
2535 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2536 if (mac && mac->port.p != in_port &&
2537 (!is_gratuitous_arp(flow)
2538 || mac_entry_is_grat_arp_locked(mac))) {
2548 /* If the composed actions may be applied to any packet in the given 'flow',
2549 * returns true. Otherwise, the actions should only be applied to 'packet', or
2550 * not at all, if 'packet' was NULL. */
2552 process_flow(struct bridge *br, const struct flow *flow,
2553 const struct ofpbuf *packet, struct ofpbuf *actions,
2554 tag_type *tags, uint16_t *nf_output_iface)
2556 struct port *in_port;
2557 struct port *out_port;
2558 struct mac_entry *mac;
2561 /* Check whether we should drop packets in this flow. */
2562 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2567 /* Learn source MAC (but don't try to learn from revalidation). */
2569 update_learning_table(br, flow, vlan, in_port);
2572 /* Determine output port. */
2573 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2575 out_port = mac->port.p;
2576 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2577 /* If we are revalidating but don't have a learning entry then
2578 * eject the flow. Installing a flow that floods packets opens
2579 * up a window of time where we could learn from a packet reflected
2580 * on a bond and blackhole packets before the learning table is
2581 * updated to reflect the correct port. */
2584 out_port = FLOOD_PORT;
2587 /* Don't send packets out their input ports. */
2588 if (in_port == out_port) {
2594 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2602 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2603 struct ofpbuf *actions, tag_type *tags,
2604 uint16_t *nf_output_iface, void *br_)
2606 struct bridge *br = br_;
2608 COVERAGE_INC(bridge_process_flow);
2609 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2613 bridge_special_ofhook_cb(const struct flow *flow,
2614 const struct ofpbuf *packet, void *br_)
2616 struct iface *iface;
2617 struct bridge *br = br_;
2619 iface = iface_from_dp_ifidx(br, flow->in_port);
2621 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2622 if (iface && iface->port->lacp && packet) {
2623 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2625 lacp_process_pdu(iface->port->lacp, iface, pdu);
2635 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2636 const struct nlattr *actions,
2638 uint64_t n_bytes, void *br_)
2640 struct bridge *br = br_;
2641 const struct nlattr *a;
2642 struct port *in_port;
2647 /* Feed information from the active flows back into the learning table to
2648 * ensure that table is always in sync with what is actually flowing
2649 * through the datapath.
2651 * We test that 'tags' is nonzero to ensure that only flows that include an
2652 * OFPP_NORMAL action are used for learning. This works because
2653 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2654 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2655 update_learning_table(br, flow, vlan, in_port);
2658 /* Account for bond slave utilization. */
2659 if (!br->has_bonded_ports) {
2662 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2663 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2664 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2665 if (out_port && out_port->bond) {
2666 uint16_t vlan = (flow->vlan_tci
2667 ? vlan_tci_to_vid(flow->vlan_tci)
2669 bond_account(out_port->bond, flow, vlan, n_bytes);
2676 bridge_account_checkpoint_ofhook_cb(void *br_)
2678 struct bridge *br = br_;
2681 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2683 bond_rebalance(port->bond,
2684 ofproto_get_revalidate_set(br->ofproto));
2690 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2691 tag_type *tags, void *br_)
2693 struct bridge *br = br_;
2694 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2695 struct port *port = port_from_dp_ifidx(br, odp_port);
2700 } else if (list_is_short(&port->ifaces)) {
2703 struct iface *iface;
2705 /* Autopath does not support VLAN hashing. */
2706 iface = bond_choose_output_slave(port->bond, flow,
2707 OFP_VLAN_NONE, tags);
2708 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2711 return odp_port_to_ofp_port(ret);
2714 static struct ofhooks bridge_ofhooks = {
2715 bridge_normal_ofhook_cb,
2716 bridge_special_ofhook_cb,
2717 bridge_account_flow_ofhook_cb,
2718 bridge_account_checkpoint_ofhook_cb,
2719 bridge_autopath_ofhook_cb,
2722 /* Port functions. */
2725 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2727 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2728 struct iface *iface = iface_;
2729 uint8_t ea[ETH_ADDR_LEN];
2732 error = netdev_get_etheraddr(iface->netdev, ea);
2734 struct lacp_pdu *packet_pdu;
2735 struct ofpbuf packet;
2737 ofpbuf_init(&packet, 0);
2738 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2739 sizeof *packet_pdu);
2741 error = netdev_send(iface->netdev, &packet);
2743 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2744 "(%s)", iface->port->name, iface->name,
2747 ofpbuf_uninit(&packet);
2749 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2750 "%s (%s)", iface->port->name, iface->name,
2756 port_run(struct port *port)
2759 lacp_run(port->lacp, lacp_send_pdu_cb);
2763 struct iface *iface;
2765 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2766 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2767 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2770 bond_run(port->bond,
2771 ofproto_get_revalidate_set(port->bridge->ofproto),
2772 lacp_negotiated(port->lacp));
2773 if (bond_should_send_learning_packets(port->bond)) {
2774 port_send_learning_packets(port);
2780 port_wait(struct port *port)
2783 lacp_wait(port->lacp);
2787 bond_wait(port->bond);
2791 static struct port *
2792 port_create(struct bridge *br, const char *name)
2796 port = xzalloc(sizeof *port);
2799 port->trunks = NULL;
2800 port->name = xstrdup(name);
2801 list_init(&port->ifaces);
2803 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2805 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2812 get_port_other_config(const struct ovsrec_port *port, const char *key,
2813 const char *default_value)
2817 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2819 return value ? value : default_value;
2823 get_interface_other_config(const struct ovsrec_interface *iface,
2824 const char *key, const char *default_value)
2828 value = get_ovsrec_key_value(&iface->header_,
2829 &ovsrec_interface_col_other_config, key);
2830 return value ? value : default_value;
2834 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2836 struct iface *iface, *next;
2837 struct sset new_ifaces;
2840 /* Collect list of new interfaces. */
2841 sset_init(&new_ifaces);
2842 for (i = 0; i < cfg->n_interfaces; i++) {
2843 const char *name = cfg->interfaces[i]->name;
2844 sset_add(&new_ifaces, name);
2847 /* Get rid of deleted interfaces. */
2848 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2849 if (!sset_contains(&new_ifaces, iface->name)) {
2850 iface_destroy(iface);
2854 sset_destroy(&new_ifaces);
2857 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2858 * to revalidate every flow. */
2860 port_flush_macs(struct port *port)
2862 struct bridge *br = port->bridge;
2863 struct mac_learning *ml = br->ml;
2864 struct mac_entry *mac, *next_mac;
2867 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2868 if (mac->port.p == port) {
2869 mac_learning_expire(ml, mac);
2875 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2877 struct sset new_ifaces;
2878 bool need_flush = false;
2879 unsigned long *trunks;
2886 /* Add new interfaces and update 'cfg' member of existing ones. */
2887 sset_init(&new_ifaces);
2888 for (i = 0; i < cfg->n_interfaces; i++) {
2889 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2890 struct iface *iface;
2892 if (!sset_add(&new_ifaces, if_cfg->name)) {
2893 VLOG_WARN("port %s: %s specified twice as port interface",
2894 port->name, if_cfg->name);
2895 iface_set_ofport(if_cfg, -1);
2899 iface = iface_lookup(port->bridge, if_cfg->name);
2901 if (iface->port != port) {
2902 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2904 port->bridge->name, if_cfg->name, iface->port->name);
2907 iface->cfg = if_cfg;
2909 iface = iface_create(port, if_cfg);
2912 /* Determine interface type. The local port always has type
2913 * "internal". Other ports take their type from the database and
2914 * default to "system" if none is specified. */
2915 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2916 : if_cfg->type[0] ? if_cfg->type
2919 sset_destroy(&new_ifaces);
2924 if (list_is_short(&port->ifaces)) {
2926 if (vlan >= 0 && vlan <= 4095) {
2927 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2932 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2933 * they even work as-is. But they have not been tested. */
2934 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2938 if (port->vlan != vlan) {
2943 /* Get trunked VLANs. */
2945 if (vlan < 0 && cfg->n_trunks) {
2946 trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
2948 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2951 } else if (vlan >= 0 && cfg->n_trunks) {
2952 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2955 if (!vlan_bitmap_equal(trunks, port->trunks)) {
2958 bitmap_free(port->trunks);
2959 port->trunks = trunks;
2962 port_flush_macs(port);
2967 port_destroy(struct port *port)
2970 struct bridge *br = port->bridge;
2971 struct iface *iface, *next;
2974 for (i = 0; i < MAX_MIRRORS; i++) {
2975 struct mirror *m = br->mirrors[i];
2976 if (m && m->out_port == port) {
2981 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2982 iface_destroy(iface);
2985 hmap_remove(&br->ports, &port->hmap_node);
2987 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2989 bond_destroy(port->bond);
2990 lacp_destroy(port->lacp);
2991 port_flush_macs(port);
2993 bitmap_free(port->trunks);
2999 static struct port *
3000 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3002 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
3003 return iface ? iface->port : NULL;
3006 static struct port *
3007 port_lookup(const struct bridge *br, const char *name)
3011 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3013 if (!strcmp(port->name, name)) {
3021 enable_lacp(struct port *port, bool *activep)
3023 if (!port->cfg->lacp) {
3024 /* XXX when LACP implementation has been sufficiently tested, enable by
3025 * default and make active on bonded ports. */
3027 } else if (!strcmp(port->cfg->lacp, "off")) {
3029 } else if (!strcmp(port->cfg->lacp, "active")) {
3032 } else if (!strcmp(port->cfg->lacp, "passive")) {
3036 VLOG_WARN("port %s: unknown LACP mode %s",
3037 port->name, port->cfg->lacp);
3043 iface_reconfigure_lacp(struct iface *iface)
3045 struct lacp_slave_settings s;
3046 int priority, portid;
3048 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
3049 priority = atoi(get_interface_other_config(iface->cfg,
3050 "lacp-port-priority", "0"));
3052 if (portid <= 0 || portid > UINT16_MAX) {
3053 portid = iface->dp_ifidx;
3056 if (priority <= 0 || priority > UINT16_MAX) {
3057 priority = UINT16_MAX;
3060 s.name = iface->name;
3062 s.priority = priority;
3063 lacp_slave_register(iface->port->lacp, iface, &s);
3067 port_reconfigure_lacp(struct port *port)
3069 static struct lacp_settings s;
3070 struct iface *iface;
3071 uint8_t sysid[ETH_ADDR_LEN];
3072 const char *sysid_str;
3073 const char *lacp_time;
3074 long long int custom_time;
3077 if (!enable_lacp(port, &s.active)) {
3078 lacp_destroy(port->lacp);
3083 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3084 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3085 memcpy(s.id, sysid, ETH_ADDR_LEN);
3087 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3090 s.name = port->name;
3092 /* Prefer bondable links if unspecified. */
3093 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3095 s.priority = (priority > 0 && priority <= UINT16_MAX
3097 : UINT16_MAX - !list_is_short(&port->ifaces));
3099 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3103 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3104 custom_time = atoi(lacp_time);
3105 if (!strcmp(lacp_time, "fast")) {
3106 s.lacp_time = LACP_TIME_FAST;
3107 } else if (!strcmp(lacp_time, "slow")) {
3108 s.lacp_time = LACP_TIME_SLOW;
3109 } else if (custom_time > 0) {
3110 s.lacp_time = LACP_TIME_CUSTOM;
3111 s.custom_time = custom_time;
3113 s.lacp_time = LACP_TIME_SLOW;
3117 port->lacp = lacp_create();
3120 lacp_configure(port->lacp, &s);
3122 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3123 iface_reconfigure_lacp(iface);
3128 port_reconfigure_bond(struct port *port)
3130 struct bond_settings s;
3131 const char *detect_s;
3132 struct iface *iface;
3134 if (list_is_short(&port->ifaces)) {
3135 /* Not a bonded port. */
3136 bond_destroy(port->bond);
3141 port->bridge->has_bonded_ports = true;
3143 s.name = port->name;
3145 if (port->cfg->bond_mode
3146 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3147 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3148 port->name, port->cfg->bond_mode,
3149 bond_mode_to_string(s.balance));
3152 s.detect = BLSM_CARRIER;
3153 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3154 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3155 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3157 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3160 s.miimon_interval = atoi(
3161 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3162 if (s.miimon_interval < 100) {
3163 s.miimon_interval = 100;
3166 s.up_delay = MAX(0, port->cfg->bond_updelay);
3167 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3168 s.rebalance_interval = atoi(
3169 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3170 if (s.rebalance_interval < 1000) {
3171 s.rebalance_interval = 1000;
3174 s.fake_iface = port->cfg->bond_fake_iface;
3177 port->bond = bond_create(&s);
3179 if (bond_reconfigure(port->bond, &s)) {
3180 bridge_flush(port->bridge);
3184 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3185 uint16_t stable_id = (port->lacp
3186 ? lacp_slave_get_port_id(port->lacp, iface)
3188 bond_slave_register(iface->port->bond, iface, stable_id,
3194 port_send_learning_packets(struct port *port)
3196 struct bridge *br = port->bridge;
3197 int error, n_packets, n_errors;
3198 struct mac_entry *e;
3200 error = n_packets = n_errors = 0;
3201 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3202 if (e->port.p != port) {
3203 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3213 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3214 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3215 "packets, last error was: %s",
3216 port->name, n_errors, n_packets, strerror(error));
3218 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3219 port->name, n_packets);
3223 /* Interface functions. */
3225 static struct iface *
3226 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3228 struct bridge *br = port->bridge;
3229 struct iface *iface;
3230 char *name = if_cfg->name;
3232 iface = xzalloc(sizeof *iface);
3234 iface->name = xstrdup(name);
3235 iface->dp_ifidx = -1;
3236 iface->tag = tag_create_random();
3237 iface->netdev = NULL;
3238 iface->cfg = if_cfg;
3240 shash_add_assert(&br->iface_by_name, iface->name, iface);
3242 list_push_back(&port->ifaces, &iface->port_elem);
3244 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3252 iface_destroy(struct iface *iface)
3255 struct port *port = iface->port;
3256 struct bridge *br = port->bridge;
3259 bond_slave_unregister(port->bond, iface);
3263 lacp_slave_unregister(port->lacp, iface);
3266 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3268 if (iface->dp_ifidx >= 0) {
3269 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3272 list_remove(&iface->port_elem);
3274 netdev_close(iface->netdev);
3279 bridge_flush(port->bridge);
3283 static struct iface *
3284 iface_lookup(const struct bridge *br, const char *name)
3286 return shash_find_data(&br->iface_by_name, name);
3289 static struct iface *
3290 iface_find(const char *name)
3292 const struct bridge *br;
3294 LIST_FOR_EACH (br, node, &all_bridges) {
3295 struct iface *iface = iface_lookup(br, name);
3304 static struct iface *
3305 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3307 struct iface *iface;
3309 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3310 hash_int(dp_ifidx, 0), &br->ifaces) {
3311 if (iface->dp_ifidx == dp_ifidx) {
3318 /* Set Ethernet address of 'iface', if one is specified in the configuration
3321 iface_set_mac(struct iface *iface)
3323 uint8_t ea[ETH_ADDR_LEN];
3325 if (!strcmp(iface->type, "internal")
3326 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3327 if (iface->dp_ifidx == ODPP_LOCAL) {
3328 VLOG_ERR("interface %s: ignoring mac in Interface record "
3329 "(use Bridge record to set local port's mac)",
3331 } else if (eth_addr_is_multicast(ea)) {
3332 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3335 int error = netdev_set_etheraddr(iface->netdev, ea);
3337 VLOG_ERR("interface %s: setting MAC failed (%s)",
3338 iface->name, strerror(error));
3344 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3346 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3348 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3349 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3353 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3355 * The value strings in '*shash' are taken directly from values[], not copied,
3356 * so the caller should not modify or free them. */
3358 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3359 struct shash *shash)
3364 for (i = 0; i < n; i++) {
3365 shash_add(shash, keys[i], values[i]);
3369 /* Creates 'keys' and 'values' arrays from 'shash'.
3371 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3372 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3373 * are populated with with strings taken directly from 'shash' and thus have
3374 * the same ownership of the key-value pairs in shash.
3377 shash_to_ovs_idl_map(struct shash *shash,
3378 char ***keys, char ***values, size_t *n)
3382 struct shash_node *sn;
3384 count = shash_count(shash);
3386 k = xmalloc(count * sizeof *k);
3387 v = xmalloc(count * sizeof *v);
3390 SHASH_FOR_EACH(sn, shash) {
3401 struct iface_delete_queues_cbdata {
3402 struct netdev *netdev;
3403 const struct ovsdb_datum *queues;
3407 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3409 union ovsdb_atom atom;
3411 atom.integer = target;
3412 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3416 iface_delete_queues(unsigned int queue_id,
3417 const struct shash *details OVS_UNUSED, void *cbdata_)
3419 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3421 if (!queue_ids_include(cbdata->queues, queue_id)) {
3422 netdev_delete_queue(cbdata->netdev, queue_id);
3427 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3429 if (!qos || qos->type[0] == '\0') {
3430 netdev_set_qos(iface->netdev, NULL, NULL);
3432 struct iface_delete_queues_cbdata cbdata;
3433 struct shash details;
3436 /* Configure top-level Qos for 'iface'. */
3437 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3438 qos->n_other_config, &details);
3439 netdev_set_qos(iface->netdev, qos->type, &details);
3440 shash_destroy(&details);
3442 /* Deconfigure queues that were deleted. */
3443 cbdata.netdev = iface->netdev;
3444 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3446 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3448 /* Configure queues for 'iface'. */
3449 for (i = 0; i < qos->n_queues; i++) {
3450 const struct ovsrec_queue *queue = qos->value_queues[i];
3451 unsigned int queue_id = qos->key_queues[i];
3453 shash_from_ovs_idl_map(queue->key_other_config,
3454 queue->value_other_config,
3455 queue->n_other_config, &details);
3456 netdev_set_queue(iface->netdev, queue_id, &details);
3457 shash_destroy(&details);
3463 iface_update_cfm(struct iface *iface)
3467 uint16_t *remote_mps;
3468 struct ovsrec_monitor *mon;
3469 uint8_t maid[CCM_MAID_LEN];
3471 mon = iface->cfg->monitor;
3474 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3478 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3479 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3483 cfm.mpid = mon->mpid;
3484 cfm.interval = mon->interval ? *mon->interval : 1000;
3486 memcpy(cfm.maid, maid, sizeof cfm.maid);
3488 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3489 for(i = 0; i < mon->n_remote_mps; i++) {
3490 remote_mps[i] = mon->remote_mps[i]->mpid;
3493 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3494 &cfm, remote_mps, mon->n_remote_mps);
3498 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3499 * how 'iface''s port is configured.
3501 * Returns true if 'iface' is up, false otherwise. */
3503 iface_get_carrier(const struct iface *iface)
3506 return netdev_get_carrier(iface->netdev);
3509 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3510 * instead of obtaining it from the database. */
3512 iface_is_synthetic(const struct iface *iface)
3514 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3517 /* Port mirroring. */
3519 static struct mirror *
3520 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3524 for (i = 0; i < MAX_MIRRORS; i++) {
3525 struct mirror *m = br->mirrors[i];
3526 if (m && uuid_equals(uuid, &m->uuid)) {
3534 mirror_reconfigure(struct bridge *br)
3536 unsigned long *rspan_vlans;
3540 /* Get rid of deleted mirrors. */
3541 for (i = 0; i < MAX_MIRRORS; i++) {
3542 struct mirror *m = br->mirrors[i];
3544 const struct ovsdb_datum *mc;
3545 union ovsdb_atom atom;
3547 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3548 atom.uuid = br->mirrors[i]->uuid;
3549 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3555 /* Add new mirrors and reconfigure existing ones. */
3556 for (i = 0; i < br->cfg->n_mirrors; i++) {
3557 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3558 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3560 mirror_reconfigure_one(m, cfg);
3562 mirror_create(br, cfg);
3566 /* Update port reserved status. */
3567 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3568 port->is_mirror_output_port = false;
3570 for (i = 0; i < MAX_MIRRORS; i++) {
3571 struct mirror *m = br->mirrors[i];
3572 if (m && m->out_port) {
3573 m->out_port->is_mirror_output_port = true;
3577 /* Update flooded vlans (for RSPAN). */
3579 if (br->cfg->n_flood_vlans) {
3580 rspan_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3581 br->cfg->n_flood_vlans);
3583 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3585 mac_learning_flush(br->ml);
3591 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3596 for (i = 0; ; i++) {
3597 if (i >= MAX_MIRRORS) {
3598 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3599 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3602 if (!br->mirrors[i]) {
3607 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3609 mac_learning_flush(br->ml);
3611 br->mirrors[i] = m = xzalloc(sizeof *m);
3612 m->uuid = cfg->header_.uuid;
3615 m->name = xstrdup(cfg->name);
3616 sset_init(&m->src_ports);
3617 sset_init(&m->dst_ports);
3623 mirror_reconfigure_one(m, cfg);
3627 mirror_destroy(struct mirror *m)
3630 struct bridge *br = m->bridge;
3633 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3634 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3635 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3638 sset_destroy(&m->src_ports);
3639 sset_destroy(&m->dst_ports);
3642 m->bridge->mirrors[m->idx] = NULL;
3647 mac_learning_flush(br->ml);
3652 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3657 for (i = 0; i < n_ports; i++) {
3658 const char *name = ports[i]->name;
3659 if (port_lookup(m->bridge, name)) {
3660 sset_add(names, name);
3662 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3663 "port %s", m->bridge->name, m->name, name);
3669 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3675 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3677 for (i = 0; i < cfg->n_select_vlan; i++) {
3678 int64_t vlan = cfg->select_vlan[i];
3679 if (vlan < 0 || vlan > 4095) {
3680 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3681 m->bridge->name, m->name, vlan);
3683 (*vlans)[n_vlans++] = vlan;
3690 vlan_is_mirrored(const struct mirror *m, int vlan)
3694 for (i = 0; i < m->n_vlans; i++) {
3695 if (m->vlans[i] == vlan) {
3703 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3705 struct sset src_ports, dst_ports;
3706 mirror_mask_t mirror_bit;
3707 struct port *out_port;
3714 if (strcmp(cfg->name, m->name)) {
3716 m->name = xstrdup(cfg->name);
3719 /* Get output port. */
3720 if (cfg->output_port) {
3721 out_port = port_lookup(m->bridge, cfg->output_port->name);
3723 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3724 m->bridge->name, m->name);
3730 if (cfg->output_vlan) {
3731 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3732 "output vlan; ignoring output vlan",
3733 m->bridge->name, m->name);
3735 } else if (cfg->output_vlan) {
3737 out_vlan = *cfg->output_vlan;
3739 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3740 m->bridge->name, m->name);
3745 sset_init(&src_ports);
3746 sset_init(&dst_ports);
3747 if (cfg->select_all) {
3748 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3749 sset_add(&src_ports, port->name);
3750 sset_add(&dst_ports, port->name);
3755 /* Get ports, and drop duplicates and ports that don't exist. */
3756 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3758 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3761 /* Get all the vlans, and drop duplicate and invalid vlans. */
3762 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3765 /* Update mirror data. */
3766 if (!sset_equals(&m->src_ports, &src_ports)
3767 || !sset_equals(&m->dst_ports, &dst_ports)
3768 || m->n_vlans != n_vlans
3769 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3770 || m->out_port != out_port
3771 || m->out_vlan != out_vlan) {
3772 bridge_flush(m->bridge);
3773 mac_learning_flush(m->bridge->ml);
3775 sset_swap(&m->src_ports, &src_ports);
3776 sset_swap(&m->dst_ports, &dst_ports);
3779 m->n_vlans = n_vlans;
3780 m->out_port = out_port;
3781 m->out_vlan = out_vlan;
3784 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3785 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3786 if (sset_contains(&m->src_ports, port->name)) {
3787 port->src_mirrors |= mirror_bit;
3789 port->src_mirrors &= ~mirror_bit;
3792 if (sset_contains(&m->dst_ports, port->name)) {
3793 port->dst_mirrors |= mirror_bit;
3795 port->dst_mirrors &= ~mirror_bit;
3800 sset_destroy(&src_ports);
3801 sset_destroy(&dst_ports);