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 hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
103 struct port *port; /* Containing port. */
104 char *name; /* Host network device name. */
105 tag_type tag; /* Tag associated with this interface. */
107 /* These members are valid only after bridge_reconfigure() causes them to
109 struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
110 int dp_ifidx; /* Index within kernel datapath. */
111 struct netdev *netdev; /* Network device. */
112 const char *type; /* Usually same as cfg->type. */
113 const struct ovsrec_interface *cfg;
116 #define MAX_MIRRORS 32
117 typedef uint32_t mirror_mask_t;
118 #define MIRROR_MASK_C(X) UINT32_C(X)
119 BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS);
121 struct bridge *bridge;
124 struct uuid uuid; /* UUID of this "mirror" record in database. */
126 /* Selection criteria. */
127 struct sset src_ports; /* Source port names. */
128 struct sset dst_ports; /* Destination port names. */
133 struct port *out_port;
137 #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
139 struct bridge *bridge;
140 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
143 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
144 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
145 * NULL if all VLANs are trunked. */
146 const struct ovsrec_port *cfg;
148 /* An ordinary bridge port has 1 interface.
149 * A bridge port for bonding has at least 2 interfaces. */
150 struct list ifaces; /* List of "struct iface"s. */
152 struct lacp *lacp; /* NULL if LACP is not enabled. */
157 /* Port mirroring info. */
158 mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
159 mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
160 bool is_mirror_output_port; /* Does port mirroring send frames here? */
164 struct hmap_node node; /* In 'all_bridges'. */
165 char *name; /* User-specified arbitrary name. */
166 struct mac_learning *ml; /* MAC learning table. */
167 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
168 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
169 const struct ovsrec_bridge *cfg;
171 /* OpenFlow switch processing. */
172 struct ofproto *ofproto; /* OpenFlow switch. */
175 struct hmap ports; /* "struct port"s indexed by name. */
176 struct hmap ifaces; /* "struct iface"s indexed by dp_ifidx. */
177 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
180 bool has_bonded_ports;
185 /* Port mirroring. */
186 struct mirror *mirrors[MAX_MIRRORS];
188 /* Synthetic local port if necessary. */
189 struct ovsrec_port synth_local_port;
190 struct ovsrec_interface synth_local_iface;
191 struct ovsrec_interface *synth_local_ifacep;
194 /* All bridges, indexed by name. */
195 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
197 /* OVSDB IDL used to obtain configuration. */
198 static struct ovsdb_idl *idl;
200 /* Each time this timer expires, the bridge fetches systems and interface
201 * statistics and pushes them into the database. */
202 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
203 static long long int stats_timer = LLONG_MIN;
205 /* Stores the time after which rate limited statistics may be written to the
206 * database. Only updated when changes to the database require rate limiting.
208 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
209 static long long int db_limiter = LLONG_MIN;
211 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
212 static void bridge_destroy(struct bridge *);
213 static struct bridge *bridge_lookup(const char *name);
214 static unixctl_cb_func bridge_unixctl_dump_flows;
215 static unixctl_cb_func bridge_unixctl_reconnect;
216 static int bridge_run_one(struct bridge *);
217 static size_t bridge_get_controllers(const struct bridge *br,
218 struct ovsrec_controller ***controllersp);
219 static void bridge_reconfigure_one(struct bridge *);
220 static void bridge_reconfigure_remotes(struct bridge *,
221 const struct sockaddr_in *managers,
223 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
224 static void bridge_fetch_dp_ifaces(struct bridge *);
225 static void bridge_flush(struct bridge *);
226 static void bridge_pick_local_hw_addr(struct bridge *,
227 uint8_t ea[ETH_ADDR_LEN],
228 struct iface **hw_addr_iface);
229 static uint64_t bridge_pick_datapath_id(struct bridge *,
230 const uint8_t bridge_ea[ETH_ADDR_LEN],
231 struct iface *hw_addr_iface);
232 static uint64_t dpid_from_hash(const void *, size_t nbytes);
234 static unixctl_cb_func bridge_unixctl_fdb_show;
235 static unixctl_cb_func cfm_unixctl_show;
236 static unixctl_cb_func qos_unixctl_show;
238 static void port_run(struct port *);
239 static void port_wait(struct port *);
240 static struct port *port_create(struct bridge *, const char *name);
241 static void port_reconfigure(struct port *, const struct ovsrec_port *);
242 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
243 static void port_destroy(struct port *);
244 static struct port *port_lookup(const struct bridge *, const char *name);
245 static struct iface *port_get_an_iface(const struct port *);
246 static struct port *port_from_dp_ifidx(const struct bridge *,
248 static void port_reconfigure_lacp(struct port *);
249 static void port_reconfigure_bond(struct port *);
250 static void port_send_learning_packets(struct port *);
252 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
253 static void mirror_destroy(struct mirror *);
254 static void mirror_reconfigure(struct bridge *);
255 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
256 static bool vlan_is_mirrored(const struct mirror *, int vlan);
258 static struct iface *iface_create(struct port *port,
259 const struct ovsrec_interface *if_cfg);
260 static void iface_destroy(struct iface *);
261 static struct iface *iface_lookup(const struct bridge *, const char *name);
262 static struct iface *iface_find(const char *name);
263 static struct iface *iface_from_dp_ifidx(const struct bridge *,
265 static void iface_set_mac(struct iface *);
266 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
267 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
268 static void iface_update_cfm(struct iface *);
269 static bool iface_refresh_cfm_stats(struct iface *iface);
270 static bool iface_get_carrier(const struct iface *);
271 static bool iface_is_synthetic(const struct iface *);
273 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
275 static void shash_to_ovs_idl_map(struct shash *,
276 char ***keys, char ***values, size_t *n);
278 /* Hooks into ofproto processing. */
279 static struct ofhooks bridge_ofhooks;
281 /* Public functions. */
283 /* Initializes the bridge module, configuring it to obtain its configuration
284 * from an OVSDB server accessed over 'remote', which should be a string in a
285 * form acceptable to ovsdb_idl_create(). */
287 bridge_init(const char *remote)
289 /* Create connection to database. */
290 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
292 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
293 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
294 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
295 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
296 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
297 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
298 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
300 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
301 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
303 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
306 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
314 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
316 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
317 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
319 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
321 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
323 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
325 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
327 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
329 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
333 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
335 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
336 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
338 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
339 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
341 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
343 /* Register unixctl commands. */
344 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
345 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
346 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
347 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
349 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
358 struct bridge *br, *next_br;
360 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
363 ovsdb_idl_destroy(idl);
366 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
367 * but for which the ovs-vswitchd configuration 'cfg' is required. */
369 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
371 static bool already_configured_once;
372 struct sset bridge_names;
373 struct sset dpif_names, dpif_types;
377 /* Only do this once per ovs-vswitchd run. */
378 if (already_configured_once) {
381 already_configured_once = true;
383 stats_timer = time_msec() + STATS_INTERVAL;
385 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
386 sset_init(&bridge_names);
387 for (i = 0; i < cfg->n_bridges; i++) {
388 sset_add(&bridge_names, cfg->bridges[i]->name);
391 /* Iterate over all system dpifs and delete any of them that do not appear
393 sset_init(&dpif_names);
394 sset_init(&dpif_types);
395 dp_enumerate_types(&dpif_types);
396 SSET_FOR_EACH (type, &dpif_types) {
399 dp_enumerate_names(type, &dpif_names);
401 /* Delete each dpif whose name is not in 'bridge_names'. */
402 SSET_FOR_EACH (name, &dpif_names) {
403 if (!sset_contains(&bridge_names, name)) {
407 retval = dpif_open(name, type, &dpif);
415 sset_destroy(&bridge_names);
416 sset_destroy(&dpif_names);
417 sset_destroy(&dpif_types);
420 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
421 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
422 * responsible for freeing '*managersp' (with free()).
424 * You may be asking yourself "why does ovs-vswitchd care?", because
425 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
426 * should not be and in fact is not directly involved in that. But
427 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
428 * it has to tell in-band control where the managers are to enable that.
429 * (Thus, only managers connected in-band are collected.)
432 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
433 struct sockaddr_in **managersp, size_t *n_managersp)
435 struct sockaddr_in *managers = NULL;
436 size_t n_managers = 0;
440 /* Collect all of the potential targets from the "targets" columns of the
441 * rows pointed to by "manager_options", excluding any that are
444 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
445 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
447 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
448 sset_find_and_delete(&targets, m->target);
450 sset_add(&targets, m->target);
454 /* Now extract the targets' IP addresses. */
455 if (!sset_is_empty(&targets)) {
458 managers = xmalloc(sset_count(&targets) * sizeof *managers);
459 SSET_FOR_EACH (target, &targets) {
460 struct sockaddr_in *sin = &managers[n_managers];
462 if ((!strncmp(target, "tcp:", 4)
463 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
464 (!strncmp(target, "ssl:", 4)
465 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
470 sset_destroy(&targets);
472 *managersp = managers;
473 *n_managersp = n_managers;
477 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
479 struct shash_node *node;
480 struct bridge *br, *next;
481 struct sockaddr_in *managers;
485 int sflow_bridge_number;
487 COVERAGE_INC(bridge_reconfigure);
489 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
491 /* Collect old and new bridges. */
493 for (i = 0; i < ovs_cfg->n_bridges; i++) {
494 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
495 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
496 VLOG_WARN("more than one bridge named %s", br_cfg->name);
500 /* Get rid of deleted bridges and add new bridges. */
501 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
502 br->cfg = shash_find_data(&new_br, br->name);
507 SHASH_FOR_EACH (node, &new_br) {
508 const struct ovsrec_bridge *br_cfg = node->data;
509 struct bridge *br = bridge_lookup(node->name);
511 /* If the bridge datapath type has changed, we need to tear it
512 * down and recreate. */
513 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
515 bridge_create(br_cfg);
518 bridge_create(br_cfg);
521 shash_destroy(&new_br);
523 /* Reconfigure all bridges. */
524 HMAP_FOR_EACH (br, node, &all_bridges) {
525 bridge_reconfigure_one(br);
528 /* Add and delete ports on all datapaths.
530 * The kernel will reject any attempt to add a given port to a datapath if
531 * that port already belongs to a different datapath, so we must do all
532 * port deletions before any port additions. */
533 HMAP_FOR_EACH (br, node, &all_bridges) {
534 struct ofproto_port ofproto_port;
535 struct ofproto_port_dump dump;
536 struct shash want_ifaces;
538 bridge_get_all_ifaces(br, &want_ifaces);
539 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
540 if (!shash_find(&want_ifaces, ofproto_port.name)
541 && strcmp(ofproto_port.name, br->name)) {
542 int retval = ofproto_port_del(br->ofproto,
543 ofproto_port.ofp_port);
545 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
546 br->name, ofproto_port.name, strerror(retval));
550 shash_destroy(&want_ifaces);
552 HMAP_FOR_EACH (br, node, &all_bridges) {
553 struct shash cur_ifaces, want_ifaces;
554 struct ofproto_port ofproto_port;
555 struct ofproto_port_dump dump;
557 /* Get the set of interfaces currently in this datapath. */
558 shash_init(&cur_ifaces);
559 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
560 struct ofproto_port *port_info = xmalloc(sizeof *port_info);
561 ofproto_port_clone(port_info, &ofproto_port);
562 shash_add(&cur_ifaces, ofproto_port.name, port_info);
565 /* Get the set of interfaces we want on this datapath. */
566 bridge_get_all_ifaces(br, &want_ifaces);
568 hmap_clear(&br->ifaces);
569 SHASH_FOR_EACH (node, &want_ifaces) {
570 const char *if_name = node->name;
571 struct iface *iface = node->data;
572 struct ofproto_port *ofproto_port;
576 type = iface ? iface->type : "internal";
577 ofproto_port = shash_find_data(&cur_ifaces, if_name);
579 /* If we have a port or a netdev already, and it's not the type we
580 * want, then delete the port (if any) and close the netdev (if
582 if ((ofproto_port && strcmp(ofproto_port->type, type))
583 || (iface && iface->netdev
584 && strcmp(type, netdev_get_type(iface->netdev)))) {
586 error = ofproto_port_del(br->ofproto,
587 ofproto_port->ofp_port);
594 if (iface->port->bond) {
595 /* The bond has a pointer to the netdev, so remove it
596 * from the bond before closing the netdev. The slave
597 * will get added back to the bond later, after a new
598 * netdev is available. */
599 bond_slave_unregister(iface->port->bond, iface);
601 netdev_close(iface->netdev);
602 iface->netdev = NULL;
606 /* If the port doesn't exist or we don't have the netdev open,
607 * we need to do more work. */
608 if (!ofproto_port || (iface && !iface->netdev)) {
609 struct netdev_options options;
610 struct netdev *netdev;
613 /* First open the network device. */
614 options.name = if_name;
616 options.args = &args;
617 options.ethertype = NETDEV_ETH_TYPE_NONE;
621 shash_from_ovs_idl_map(iface->cfg->key_options,
622 iface->cfg->value_options,
623 iface->cfg->n_options, &args);
625 error = netdev_open(&options, &netdev);
626 shash_destroy(&args);
629 VLOG_WARN("could not open network device %s (%s)",
630 if_name, strerror(error));
634 /* Then add the port if we haven't already. */
636 error = ofproto_port_add(br->ofproto, netdev, NULL);
638 netdev_close(netdev);
639 if (error == EFBIG) {
640 VLOG_ERR("bridge %s: out of valid port numbers",
644 VLOG_WARN("bridge %s: failed to add %s interface "
646 br->name, if_name, strerror(error));
652 /* Update 'iface'. */
654 iface->netdev = netdev;
656 } else if (iface && iface->netdev) {
660 shash_from_ovs_idl_map(iface->cfg->key_options,
661 iface->cfg->value_options,
662 iface->cfg->n_options, &args);
663 netdev_set_config(iface->netdev, &args);
664 shash_destroy(&args);
667 shash_destroy(&want_ifaces);
669 SHASH_FOR_EACH (node, &cur_ifaces) {
670 struct ofproto_port *port_info = node->data;
671 ofproto_port_destroy(port_info);
674 shash_destroy(&cur_ifaces);
676 sflow_bridge_number = 0;
677 HMAP_FOR_EACH (br, node, &all_bridges) {
678 uint8_t ea[ETH_ADDR_LEN];
680 struct iface *local_iface;
681 struct port *port, *next_port;
682 struct iface *hw_addr_iface;
685 bridge_fetch_dp_ifaces(br);
687 /* Delete interfaces that cannot be opened.
689 * Following this loop, every remaining "struct iface" has nonnull
690 * 'netdev' and correct 'dp_ifidx'. */
691 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
692 struct iface *iface, *next_iface;
694 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
695 if (iface->netdev && iface->dp_ifidx >= 0) {
696 VLOG_DBG("bridge %s: interface %s is on port %d",
697 br->name, iface->name, iface->dp_ifidx);
700 VLOG_ERR("bridge %s: missing %s interface, dropping",
701 br->name, iface->name);
703 /* We already reported a related error, don't bother
707 iface_set_ofport(iface->cfg, -1);
708 iface_destroy(iface);
712 if (list_is_empty(&port->ifaces)) {
713 VLOG_WARN("%s port has no interfaces, dropping", port->name);
718 /* Pick local port hardware address, datapath ID. */
719 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
720 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
722 int error = netdev_set_etheraddr(local_iface->netdev, ea);
724 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
725 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
726 "Ethernet address: %s",
727 br->name, strerror(error));
730 memcpy(br->ea, ea, ETH_ADDR_LEN);
732 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
733 ofproto_set_datapath_id(br->ofproto, dpid);
735 dpid_string = xasprintf("%016"PRIx64, dpid);
736 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
739 /* Set NetFlow configuration on this bridge. */
740 if (br->cfg->netflow) {
741 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
742 struct netflow_options opts;
744 memset(&opts, 0, sizeof opts);
746 ofproto_get_netflow_ids(br->ofproto,
747 &opts.engine_type, &opts.engine_id);
748 if (nf_cfg->engine_type) {
749 opts.engine_type = *nf_cfg->engine_type;
751 if (nf_cfg->engine_id) {
752 opts.engine_id = *nf_cfg->engine_id;
755 opts.active_timeout = nf_cfg->active_timeout;
756 if (!opts.active_timeout) {
757 opts.active_timeout = -1;
758 } else if (opts.active_timeout < 0) {
759 VLOG_WARN("bridge %s: active timeout interval set to negative "
760 "value, using default instead (%d seconds)", br->name,
761 NF_ACTIVE_TIMEOUT_DEFAULT);
762 opts.active_timeout = -1;
765 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
766 if (opts.add_id_to_iface) {
767 if (opts.engine_id > 0x7f) {
768 VLOG_WARN("bridge %s: netflow port mangling may conflict "
769 "with another vswitch, choose an engine id less "
770 "than 128", br->name);
772 if (hmap_count(&br->ports) > 508) {
773 VLOG_WARN("bridge %s: netflow port mangling will conflict "
774 "with another port when more than 508 ports are "
779 sset_init(&opts.collectors);
780 sset_add_array(&opts.collectors,
781 nf_cfg->targets, nf_cfg->n_targets);
782 if (ofproto_set_netflow(br->ofproto, &opts)) {
783 VLOG_ERR("bridge %s: problem setting netflow collectors",
786 sset_destroy(&opts.collectors);
788 ofproto_set_netflow(br->ofproto, NULL);
791 /* Set sFlow configuration on this bridge. */
792 if (br->cfg->sflow) {
793 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
794 struct ovsrec_controller **controllers;
795 struct ofproto_sflow_options oso;
796 size_t n_controllers;
798 memset(&oso, 0, sizeof oso);
800 sset_init(&oso.targets);
801 sset_add_array(&oso.targets,
802 sflow_cfg->targets, sflow_cfg->n_targets);
804 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
805 if (sflow_cfg->sampling) {
806 oso.sampling_rate = *sflow_cfg->sampling;
809 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
810 if (sflow_cfg->polling) {
811 oso.polling_interval = *sflow_cfg->polling;
814 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
815 if (sflow_cfg->header) {
816 oso.header_len = *sflow_cfg->header;
819 oso.sub_id = sflow_bridge_number++;
820 oso.agent_device = sflow_cfg->agent;
822 oso.control_ip = NULL;
823 n_controllers = bridge_get_controllers(br, &controllers);
824 for (i = 0; i < n_controllers; i++) {
825 if (controllers[i]->local_ip) {
826 oso.control_ip = controllers[i]->local_ip;
830 ofproto_set_sflow(br->ofproto, &oso);
832 sset_destroy(&oso.targets);
834 ofproto_set_sflow(br->ofproto, NULL);
837 /* Update the controller and related settings. It would be more
838 * straightforward to call this from bridge_reconfigure_one(), but we
839 * can't do it there for two reasons. First, and most importantly, at
840 * that point we don't know the dp_ifidx of any interfaces that have
841 * been added to the bridge (because we haven't actually added them to
842 * the datapath). Second, at that point we haven't set the datapath ID
843 * yet; when a controller is configured, resetting the datapath ID will
844 * immediately disconnect from the controller, so it's better to set
845 * the datapath ID before the controller. */
846 bridge_reconfigure_remotes(br, managers, n_managers);
848 HMAP_FOR_EACH (br, node, &all_bridges) {
851 br->has_bonded_ports = false;
852 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
855 port_reconfigure_lacp(port);
856 port_reconfigure_bond(port);
858 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
859 iface_update_qos(iface, port->cfg->qos);
860 netdev_set_policing(iface->netdev,
861 iface->cfg->ingress_policing_rate,
862 iface->cfg->ingress_policing_burst);
863 iface_set_mac(iface);
868 /* Some reconfiguration operations require the bridge to have been run at
870 HMAP_FOR_EACH (br, node, &all_bridges) {
875 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
876 iface_update_cfm(iface);
882 /* ovs-vswitchd has completed initialization, so allow the process that
883 * forked us to exit successfully. */
884 daemonize_complete();
888 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
889 const struct ovsdb_idl_column *column,
892 const struct ovsdb_datum *datum;
893 union ovsdb_atom atom;
896 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
897 atom.string = (char *) key;
898 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
899 return idx == UINT_MAX ? NULL : datum->values[idx].string;
903 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
905 return get_ovsrec_key_value(&br_cfg->header_,
906 &ovsrec_bridge_col_other_config, key);
910 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
911 struct iface **hw_addr_iface)
917 *hw_addr_iface = NULL;
919 /* Did the user request a particular MAC? */
920 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
921 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
922 if (eth_addr_is_multicast(ea)) {
923 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
924 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
925 } else if (eth_addr_is_zero(ea)) {
926 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
932 /* Otherwise choose the minimum non-local MAC address among all of the
934 memset(ea, 0xff, ETH_ADDR_LEN);
935 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
936 uint8_t iface_ea[ETH_ADDR_LEN];
937 struct iface *candidate;
940 /* Mirror output ports don't participate. */
941 if (port->is_mirror_output_port) {
945 /* Choose the MAC address to represent the port. */
947 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
948 /* Find the interface with this Ethernet address (if any) so that
949 * we can provide the correct devname to the caller. */
950 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
951 uint8_t candidate_ea[ETH_ADDR_LEN];
952 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
953 && eth_addr_equals(iface_ea, candidate_ea)) {
958 /* Choose the interface whose MAC address will represent the port.
959 * The Linux kernel bonding code always chooses the MAC address of
960 * the first slave added to a bond, and the Fedora networking
961 * scripts always add slaves to a bond in alphabetical order, so
962 * for compatibility we choose the interface with the name that is
963 * first in alphabetical order. */
964 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
965 if (!iface || strcmp(candidate->name, iface->name) < 0) {
970 /* The local port doesn't count (since we're trying to choose its
971 * MAC address anyway). */
972 if (iface->dp_ifidx == ODPP_LOCAL) {
977 error = netdev_get_etheraddr(iface->netdev, iface_ea);
979 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
980 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
981 iface->name, strerror(error));
986 /* Compare against our current choice. */
987 if (!eth_addr_is_multicast(iface_ea) &&
988 !eth_addr_is_local(iface_ea) &&
989 !eth_addr_is_reserved(iface_ea) &&
990 !eth_addr_is_zero(iface_ea) &&
991 eth_addr_compare_3way(iface_ea, ea) < 0)
993 memcpy(ea, iface_ea, ETH_ADDR_LEN);
994 *hw_addr_iface = iface;
997 if (eth_addr_is_multicast(ea)) {
998 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
999 *hw_addr_iface = NULL;
1000 VLOG_WARN("bridge %s: using default bridge Ethernet "
1001 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1003 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1004 br->name, ETH_ADDR_ARGS(ea));
1008 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1009 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1010 * an interface on 'br', then that interface must be passed in as
1011 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1012 * 'hw_addr_iface' must be passed in as a null pointer. */
1014 bridge_pick_datapath_id(struct bridge *br,
1015 const uint8_t bridge_ea[ETH_ADDR_LEN],
1016 struct iface *hw_addr_iface)
1019 * The procedure for choosing a bridge MAC address will, in the most
1020 * ordinary case, also choose a unique MAC that we can use as a datapath
1021 * ID. In some special cases, though, multiple bridges will end up with
1022 * the same MAC address. This is OK for the bridges, but it will confuse
1023 * the OpenFlow controller, because each datapath needs a unique datapath
1026 * Datapath IDs must be unique. It is also very desirable that they be
1027 * stable from one run to the next, so that policy set on a datapath
1030 const char *datapath_id;
1033 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1034 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1038 if (hw_addr_iface) {
1040 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1042 * A bridge whose MAC address is taken from a VLAN network device
1043 * (that is, a network device created with vconfig(8) or similar
1044 * tool) will have the same MAC address as a bridge on the VLAN
1045 * device's physical network device.
1047 * Handle this case by hashing the physical network device MAC
1048 * along with the VLAN identifier.
1050 uint8_t buf[ETH_ADDR_LEN + 2];
1051 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1052 buf[ETH_ADDR_LEN] = vlan >> 8;
1053 buf[ETH_ADDR_LEN + 1] = vlan;
1054 return dpid_from_hash(buf, sizeof buf);
1057 * Assume that this bridge's MAC address is unique, since it
1058 * doesn't fit any of the cases we handle specially.
1063 * A purely internal bridge, that is, one that has no non-virtual
1064 * network devices on it at all, is more difficult because it has no
1065 * natural unique identifier at all.
1067 * When the host is a XenServer, we handle this case by hashing the
1068 * host's UUID with the name of the bridge. Names of bridges are
1069 * persistent across XenServer reboots, although they can be reused if
1070 * an internal network is destroyed and then a new one is later
1071 * created, so this is fairly effective.
1073 * When the host is not a XenServer, we punt by using a random MAC
1074 * address on each run.
1076 const char *host_uuid = xenserver_get_host_uuid();
1078 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1079 dpid = dpid_from_hash(combined, strlen(combined));
1085 return eth_addr_to_uint64(bridge_ea);
1089 dpid_from_hash(const void *data, size_t n)
1091 uint8_t hash[SHA1_DIGEST_SIZE];
1093 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1094 sha1_bytes(data, n, hash);
1095 eth_addr_mark_random(hash);
1096 return eth_addr_to_uint64(hash);
1100 iface_refresh_status(struct iface *iface)
1104 enum netdev_flags flags;
1111 if (iface_is_synthetic(iface)) {
1117 if (!netdev_get_status(iface->netdev, &sh)) {
1119 char **keys, **values;
1121 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1122 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1127 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1130 shash_destroy_free_data(&sh);
1132 error = netdev_get_flags(iface->netdev, &flags);
1134 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1137 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1140 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1142 ovsrec_interface_set_duplex(iface->cfg,
1143 netdev_features_is_full_duplex(current)
1145 /* warning: uint64_t -> int64_t conversion */
1146 bps = netdev_features_to_bps(current);
1147 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1150 ovsrec_interface_set_duplex(iface->cfg, NULL);
1151 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1155 ovsrec_interface_set_link_state(iface->cfg,
1156 iface_get_carrier(iface) ? "up" : "down");
1158 error = netdev_get_mtu(iface->netdev, &mtu);
1159 if (!error && mtu != INT_MAX) {
1161 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1164 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1168 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1169 * changed, false otherwise. */
1171 iface_refresh_cfm_stats(struct iface *iface)
1173 const struct ovsrec_monitor *mon;
1174 const struct cfm *cfm;
1175 bool changed = false;
1178 mon = iface->cfg->monitor;
1179 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1185 for (i = 0; i < mon->n_remote_mps; i++) {
1186 const struct ovsrec_maintenance_point *mp;
1187 const struct remote_mp *rmp;
1189 mp = mon->remote_mps[i];
1190 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1192 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1193 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1198 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1199 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1207 iface_refresh_lacp_stats(struct iface *iface)
1209 bool *db_current = iface->cfg->lacp_current;
1210 bool changed = false;
1212 if (iface->port->lacp) {
1213 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1215 if (!db_current || *db_current != current) {
1217 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1219 } else if (db_current) {
1221 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1228 iface_refresh_stats(struct iface *iface)
1234 static const struct iface_stat iface_stats[] = {
1235 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1236 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1237 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1238 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1239 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1240 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1241 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1242 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1243 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1244 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1245 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1246 { "collisions", offsetof(struct netdev_stats, collisions) },
1248 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1249 const struct iface_stat *s;
1251 char *keys[N_STATS];
1252 int64_t values[N_STATS];
1255 struct netdev_stats stats;
1257 if (iface_is_synthetic(iface)) {
1261 /* Intentionally ignore return value, since errors will set 'stats' to
1262 * all-1s, and we will deal with that correctly below. */
1263 netdev_get_stats(iface->netdev, &stats);
1266 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1267 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1268 if (value != UINT64_MAX) {
1275 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1279 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1281 struct ovsdb_datum datum;
1285 get_system_stats(&stats);
1287 ovsdb_datum_from_shash(&datum, &stats);
1288 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1292 static inline const char *
1293 nx_role_to_str(enum nx_role role)
1298 case NX_ROLE_MASTER:
1303 return "*** INVALID ROLE ***";
1308 bridge_refresh_controller_status(const struct bridge *br)
1311 const struct ovsrec_controller *cfg;
1313 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1315 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1316 struct ofproto_controller_info *cinfo =
1317 shash_find_data(&info, cfg->target);
1320 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1321 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1322 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1323 (char **) cinfo->pairs.values,
1326 ovsrec_controller_set_is_connected(cfg, false);
1327 ovsrec_controller_set_role(cfg, NULL);
1328 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1332 ofproto_free_ofproto_controller_info(&info);
1338 const struct ovsrec_open_vswitch *cfg;
1340 bool datapath_destroyed;
1341 bool database_changed;
1344 /* Let each bridge do the work that it needs to do. */
1345 datapath_destroyed = false;
1346 HMAP_FOR_EACH (br, node, &all_bridges) {
1347 int error = bridge_run_one(br);
1349 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1350 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1351 "forcing reconfiguration", br->name);
1352 datapath_destroyed = true;
1356 /* (Re)configure if necessary. */
1357 database_changed = ovsdb_idl_run(idl);
1358 cfg = ovsrec_open_vswitch_first(idl);
1360 /* Re-configure SSL. We do this on every trip through the main loop,
1361 * instead of just when the database changes, because the contents of the
1362 * key and certificate files can change without the database changing.
1364 * We do this before bridge_reconfigure() because that function might
1365 * initiate SSL connections and thus requires SSL to be configured. */
1366 if (cfg && cfg->ssl) {
1367 const struct ovsrec_ssl *ssl = cfg->ssl;
1369 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1370 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1373 if (database_changed || datapath_destroyed) {
1375 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1377 bridge_configure_once(cfg);
1378 bridge_reconfigure(cfg);
1380 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1381 ovsdb_idl_txn_commit(txn);
1382 ovsdb_idl_txn_destroy(txn); /* XXX */
1384 /* We still need to reconfigure to avoid dangling pointers to
1385 * now-destroyed ovsrec structures inside bridge data. */
1386 static const struct ovsrec_open_vswitch null_cfg;
1388 bridge_reconfigure(&null_cfg);
1392 /* Refresh system and interface stats if necessary. */
1393 if (time_msec() >= stats_timer) {
1395 struct ovsdb_idl_txn *txn;
1397 txn = ovsdb_idl_txn_create(idl);
1398 HMAP_FOR_EACH (br, node, &all_bridges) {
1401 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1402 struct iface *iface;
1404 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1405 iface_refresh_stats(iface);
1406 iface_refresh_status(iface);
1409 bridge_refresh_controller_status(br);
1411 refresh_system_stats(cfg);
1412 ovsdb_idl_txn_commit(txn);
1413 ovsdb_idl_txn_destroy(txn); /* XXX */
1416 stats_timer = time_msec() + STATS_INTERVAL;
1419 if (time_msec() >= db_limiter) {
1420 struct ovsdb_idl_txn *txn;
1421 bool changed = false;
1423 txn = ovsdb_idl_txn_create(idl);
1424 HMAP_FOR_EACH (br, node, &all_bridges) {
1427 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1428 struct iface *iface;
1430 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1431 changed = iface_refresh_cfm_stats(iface) || changed;
1432 changed = iface_refresh_lacp_stats(iface) || changed;
1438 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1441 ovsdb_idl_txn_commit(txn);
1442 ovsdb_idl_txn_destroy(txn);
1451 HMAP_FOR_EACH (br, node, &all_bridges) {
1454 ofproto_wait(br->ofproto);
1455 mac_learning_wait(br->ml);
1456 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1460 ovsdb_idl_wait(idl);
1461 poll_timer_wait_until(stats_timer);
1463 if (db_limiter > time_msec()) {
1464 poll_timer_wait_until(db_limiter);
1468 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1469 * configuration changes. */
1471 bridge_flush(struct bridge *br)
1473 COVERAGE_INC(bridge_flush);
1477 /* Bridge unixctl user interface functions. */
1479 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1480 const char *args, void *aux OVS_UNUSED)
1482 struct ds ds = DS_EMPTY_INITIALIZER;
1483 const struct bridge *br;
1484 const struct mac_entry *e;
1486 br = bridge_lookup(args);
1488 unixctl_command_reply(conn, 501, "no such bridge");
1492 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1493 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1494 struct port *port = e->port.p;
1495 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1496 port_get_an_iface(port)->dp_ifidx,
1497 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1499 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1503 /* CFM unixctl user interface functions. */
1505 cfm_unixctl_show(struct unixctl_conn *conn,
1506 const char *args, void *aux OVS_UNUSED)
1508 struct ds ds = DS_EMPTY_INITIALIZER;
1509 struct iface *iface;
1510 const struct cfm *cfm;
1512 iface = iface_find(args);
1514 unixctl_command_reply(conn, 501, "no such interface");
1518 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1521 unixctl_command_reply(conn, 501, "CFM not enabled");
1525 cfm_dump_ds(cfm, &ds);
1526 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1530 /* QoS unixctl user interface functions. */
1532 struct qos_unixctl_show_cbdata {
1534 struct iface *iface;
1538 qos_unixctl_show_cb(unsigned int queue_id,
1539 const struct shash *details,
1542 struct qos_unixctl_show_cbdata *data = aux;
1543 struct ds *ds = data->ds;
1544 struct iface *iface = data->iface;
1545 struct netdev_queue_stats stats;
1546 struct shash_node *node;
1549 ds_put_cstr(ds, "\n");
1551 ds_put_format(ds, "Queue %u:\n", queue_id);
1553 ds_put_cstr(ds, "Default:\n");
1556 SHASH_FOR_EACH (node, details) {
1557 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1560 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1562 if (stats.tx_packets != UINT64_MAX) {
1563 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1566 if (stats.tx_bytes != UINT64_MAX) {
1567 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1570 if (stats.tx_errors != UINT64_MAX) {
1571 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1574 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1575 queue_id, strerror(error));
1580 qos_unixctl_show(struct unixctl_conn *conn,
1581 const char *args, void *aux OVS_UNUSED)
1583 struct ds ds = DS_EMPTY_INITIALIZER;
1584 struct shash sh = SHASH_INITIALIZER(&sh);
1585 struct iface *iface;
1587 struct shash_node *node;
1588 struct qos_unixctl_show_cbdata data;
1591 iface = iface_find(args);
1593 unixctl_command_reply(conn, 501, "no such interface");
1597 netdev_get_qos(iface->netdev, &type, &sh);
1599 if (*type != '\0') {
1600 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1602 SHASH_FOR_EACH (node, &sh) {
1603 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1608 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1611 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1613 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1615 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1616 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1619 shash_destroy_free_data(&sh);
1623 /* Bridge reconfiguration functions. */
1624 static struct bridge *
1625 bridge_create(const struct ovsrec_bridge *br_cfg)
1630 assert(!bridge_lookup(br_cfg->name));
1631 br = xzalloc(sizeof *br);
1633 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1636 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1642 br->name = xstrdup(br_cfg->name);
1644 br->ml = mac_learning_create();
1645 eth_addr_nicira_random(br->default_ea);
1647 hmap_init(&br->ports);
1648 hmap_init(&br->ifaces);
1649 hmap_init(&br->iface_by_name);
1653 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1655 VLOG_INFO("bridge %s: created", br->name);
1661 bridge_destroy(struct bridge *br)
1664 struct port *port, *next;
1667 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1670 for (i = 0; i < MAX_MIRRORS; i++) {
1671 mirror_destroy(br->mirrors[i]);
1673 hmap_remove(&all_bridges, &br->node);
1674 ofproto_destroy_and_delete(br->ofproto);
1675 mac_learning_destroy(br->ml);
1676 hmap_destroy(&br->ifaces);
1677 hmap_destroy(&br->ports);
1678 hmap_destroy(&br->iface_by_name);
1679 free(br->synth_local_iface.type);
1685 static struct bridge *
1686 bridge_lookup(const char *name)
1690 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1691 if (!strcmp(br->name, name)) {
1698 /* Handle requests for a listing of all flows known by the OpenFlow
1699 * stack, including those normally hidden. */
1701 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1702 const char *args, void *aux OVS_UNUSED)
1707 br = bridge_lookup(args);
1709 unixctl_command_reply(conn, 501, "Unknown bridge");
1714 ofproto_get_all_flows(br->ofproto, &results);
1716 unixctl_command_reply(conn, 200, ds_cstr(&results));
1717 ds_destroy(&results);
1720 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1721 * connections and reconnect. If BRIDGE is not specified, then all bridges
1722 * drop their controller connections and reconnect. */
1724 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1725 const char *args, void *aux OVS_UNUSED)
1728 if (args[0] != '\0') {
1729 br = bridge_lookup(args);
1731 unixctl_command_reply(conn, 501, "Unknown bridge");
1734 ofproto_reconnect_controllers(br->ofproto);
1736 HMAP_FOR_EACH (br, node, &all_bridges) {
1737 ofproto_reconnect_controllers(br->ofproto);
1740 unixctl_command_reply(conn, 200, NULL);
1744 bridge_run_one(struct bridge *br)
1749 error = ofproto_run1(br->ofproto);
1754 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1756 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1760 error = ofproto_run2(br->ofproto, br->flush);
1767 bridge_get_controllers(const struct bridge *br,
1768 struct ovsrec_controller ***controllersp)
1770 struct ovsrec_controller **controllers;
1771 size_t n_controllers;
1773 controllers = br->cfg->controller;
1774 n_controllers = br->cfg->n_controller;
1776 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1782 *controllersp = controllers;
1784 return n_controllers;
1788 bridge_reconfigure_one(struct bridge *br)
1790 enum ofproto_fail_mode fail_mode;
1791 struct port *port, *next;
1792 struct shash_node *node;
1793 struct shash new_ports;
1796 /* Collect new ports. */
1797 shash_init(&new_ports);
1798 for (i = 0; i < br->cfg->n_ports; i++) {
1799 const char *name = br->cfg->ports[i]->name;
1800 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1801 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1805 if (bridge_get_controllers(br, NULL)
1806 && !shash_find(&new_ports, br->name)) {
1807 struct ofproto_port local_port;
1811 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1812 br->name, br->name);
1814 error = ofproto_port_query_by_name(br->ofproto, br->name, &local_port);
1815 type = xstrdup(error ? "internal" : local_port.type);
1816 ofproto_port_destroy(&local_port);
1818 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1819 br->synth_local_port.n_interfaces = 1;
1820 br->synth_local_port.name = br->name;
1822 br->synth_local_iface.name = br->name;
1823 free(br->synth_local_iface.type);
1824 br->synth_local_iface.type = type;
1826 br->synth_local_ifacep = &br->synth_local_iface;
1828 shash_add(&new_ports, br->name, &br->synth_local_port);
1831 /* Get rid of deleted ports.
1832 * Get rid of deleted interfaces on ports that still exist. */
1833 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1834 const struct ovsrec_port *port_cfg;
1836 port_cfg = shash_find_data(&new_ports, port->name);
1840 port_del_ifaces(port, port_cfg);
1844 /* Create new ports.
1845 * Add new interfaces to existing ports.
1846 * Reconfigure existing ports. */
1847 SHASH_FOR_EACH (node, &new_ports) {
1848 struct port *port = port_lookup(br, node->name);
1850 port = port_create(br, node->name);
1853 port_reconfigure(port, node->data);
1854 if (list_is_empty(&port->ifaces)) {
1855 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1856 br->name, port->name);
1860 shash_destroy(&new_ports);
1862 /* Set the fail-mode */
1863 fail_mode = !br->cfg->fail_mode
1864 || !strcmp(br->cfg->fail_mode, "standalone")
1865 ? OFPROTO_FAIL_STANDALONE
1866 : OFPROTO_FAIL_SECURE;
1867 ofproto_set_fail_mode(br->ofproto, fail_mode);
1869 /* Configure OpenFlow controller connection snooping. */
1870 if (!ofproto_has_snoops(br->ofproto)) {
1874 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1875 ovs_rundir(), br->name));
1876 ofproto_set_snoops(br->ofproto, &snoops);
1877 sset_destroy(&snoops);
1880 mirror_reconfigure(br);
1883 /* Initializes 'oc' appropriately as a management service controller for
1886 * The caller must free oc->target when it is no longer needed. */
1888 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1889 struct ofproto_controller *oc)
1891 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1892 oc->max_backoff = 0;
1893 oc->probe_interval = 60;
1894 oc->band = OFPROTO_OUT_OF_BAND;
1896 oc->burst_limit = 0;
1899 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1901 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1902 struct ofproto_controller *oc)
1904 oc->target = c->target;
1905 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1906 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1907 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1908 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1909 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1910 oc->burst_limit = (c->controller_burst_limit
1911 ? *c->controller_burst_limit : 0);
1914 /* Configures the IP stack for 'br''s local interface properly according to the
1915 * configuration in 'c'. */
1917 bridge_configure_local_iface_netdev(struct bridge *br,
1918 struct ovsrec_controller *c)
1920 struct netdev *netdev;
1921 struct in_addr mask, gateway;
1923 struct iface *local_iface;
1926 /* If there's no local interface or no IP address, give up. */
1927 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1928 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1932 /* Bring up the local interface. */
1933 netdev = local_iface->netdev;
1934 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1936 /* Configure the IP address and netmask. */
1937 if (!c->local_netmask
1938 || !inet_aton(c->local_netmask, &mask)
1940 mask.s_addr = guess_netmask(ip.s_addr);
1942 if (!netdev_set_in4(netdev, ip, mask)) {
1943 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1944 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1947 /* Configure the default gateway. */
1948 if (c->local_gateway
1949 && inet_aton(c->local_gateway, &gateway)
1950 && gateway.s_addr) {
1951 if (!netdev_add_router(netdev, gateway)) {
1952 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1953 br->name, IP_ARGS(&gateway.s_addr));
1959 bridge_reconfigure_remotes(struct bridge *br,
1960 const struct sockaddr_in *managers,
1963 const char *disable_ib_str, *queue_id_str;
1964 bool disable_in_band = false;
1967 struct ovsrec_controller **controllers;
1968 size_t n_controllers;
1970 struct ofproto_controller *ocs;
1974 /* Check if we should disable in-band control on this bridge. */
1975 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1976 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1977 disable_in_band = true;
1980 /* Set OpenFlow queue ID for in-band control. */
1981 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1982 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1983 ofproto_set_in_band_queue(br->ofproto, queue_id);
1985 if (disable_in_band) {
1986 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1988 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1991 n_controllers = bridge_get_controllers(br, &controllers);
1993 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1996 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1997 for (i = 0; i < n_controllers; i++) {
1998 struct ovsrec_controller *c = controllers[i];
2000 if (!strncmp(c->target, "punix:", 6)
2001 || !strncmp(c->target, "unix:", 5)) {
2002 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2004 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2005 * domain sockets and overwriting arbitrary local files. */
2006 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
2007 "controller \"%s\" due to possibility for remote "
2008 "exploit", br->name, c->target);
2012 bridge_configure_local_iface_netdev(br, c);
2013 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2014 if (disable_in_band) {
2015 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2020 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2021 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2026 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2031 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2032 struct iface *iface;
2034 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2035 shash_add_once(ifaces, iface->name, iface);
2037 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2038 shash_add_once(ifaces, port->name, NULL);
2043 /* For robustness, in case the administrator moves around datapath ports behind
2044 * our back, we re-check all the datapath port numbers here.
2046 * This function will set the 'dp_ifidx' members of interfaces that have
2047 * disappeared to -1, so only call this function from a context where those
2048 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2049 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2050 * datapath, which doesn't support UINT16_MAX+1 ports. */
2052 bridge_fetch_dp_ifaces(struct bridge *br)
2054 struct ofproto_port ofproto_port;
2055 struct ofproto_port_dump dump;
2058 /* Reset all interface numbers. */
2059 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2060 struct iface *iface;
2062 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2063 iface->dp_ifidx = -1;
2066 hmap_clear(&br->ifaces);
2068 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
2069 struct iface *iface = iface_lookup(br, ofproto_port.name);
2071 uint32_t odp_port = ofp_port_to_odp_port(ofproto_port.ofp_port);
2073 if (iface->dp_ifidx >= 0) {
2074 VLOG_WARN("bridge %s: interface %s reported twice",
2075 br->name, ofproto_port.name);
2076 } else if (iface_from_dp_ifidx(br, odp_port)) {
2077 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
2078 br->name, odp_port);
2080 iface->dp_ifidx = odp_port;
2081 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2082 hash_int(iface->dp_ifidx, 0));
2085 iface_set_ofport(iface->cfg,
2086 (iface->dp_ifidx >= 0
2087 ? odp_port_to_ofp_port(iface->dp_ifidx)
2093 /* Bridge packet processing functions. */
2096 set_dst(struct dst *dst, const struct flow *flow,
2097 const struct port *in_port, const struct port *out_port,
2100 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2101 : in_port->vlan >= 0 ? in_port->vlan
2102 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2103 : vlan_tci_to_vid(flow->vlan_tci));
2105 dst->iface = (!out_port->bond
2106 ? port_get_an_iface(out_port)
2107 : bond_choose_output_slave(out_port->bond, flow,
2110 return dst->iface != NULL;
2114 mirror_mask_ffs(mirror_mask_t mask)
2116 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2121 dst_set_init(struct dst_set *set)
2123 set->dsts = set->builtin;
2125 set->allocated = ARRAY_SIZE(set->builtin);
2129 dst_set_add(struct dst_set *set, const struct dst *dst)
2131 if (set->n >= set->allocated) {
2132 size_t new_allocated;
2133 struct dst *new_dsts;
2135 new_allocated = set->allocated * 2;
2136 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2137 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2141 set->dsts = new_dsts;
2142 set->allocated = new_allocated;
2144 set->dsts[set->n++] = *dst;
2148 dst_set_free(struct dst_set *set)
2150 if (set->dsts != set->builtin) {
2156 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2159 for (i = 0; i < set->n; i++) {
2160 if (set->dsts[i].vlan == test->vlan
2161 && set->dsts[i].iface == test->iface) {
2169 port_trunks_vlan(const struct port *port, uint16_t vlan)
2171 return (port->vlan < 0 || vlan_bitmap_contains(port->trunks, vlan));
2175 port_includes_vlan(const struct port *port, uint16_t vlan)
2177 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2181 port_is_floodable(const struct port *port)
2183 struct iface *iface;
2185 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2186 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2194 /* Returns an arbitrary interface within 'port'. */
2195 static struct iface *
2196 port_get_an_iface(const struct port *port)
2198 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2202 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2203 const struct port *in_port, const struct port *out_port,
2204 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2208 if (out_port == FLOOD_PORT) {
2211 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2213 && port_is_floodable(port)
2214 && port_includes_vlan(port, vlan)
2215 && !port->is_mirror_output_port
2216 && set_dst(&dst, flow, in_port, port, tags)) {
2217 dst_set_add(set, &dst);
2220 *nf_output_iface = NF_OUT_FLOOD;
2221 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2222 dst_set_add(set, &dst);
2223 *nf_output_iface = dst.iface->dp_ifidx;
2228 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2229 uint16_t vlan, const struct port *in_port,
2230 struct dst_set *set, tag_type *tags)
2232 mirror_mask_t mirrors;
2236 mirrors = in_port->src_mirrors;
2237 for (i = 0; i < set->n; i++) {
2238 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2245 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2246 if (flow_vlan == 0) {
2247 flow_vlan = OFP_VLAN_NONE;
2251 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2252 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2256 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2257 && !dst_is_duplicate(set, &dst)) {
2258 dst_set_add(set, &dst);
2263 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2264 if (port_includes_vlan(port, m->out_vlan)
2265 && set_dst(&dst, flow, in_port, port, tags))
2267 if (port->vlan < 0) {
2268 dst.vlan = m->out_vlan;
2270 if (dst_is_duplicate(set, &dst)) {
2274 /* Use the vlan tag on the original flow instead of
2275 * the one passed in the vlan parameter. This ensures
2276 * that we compare the vlan from before any implicit
2277 * tagging tags place. This is necessary because
2278 * dst->vlan is the final vlan, after removing implicit
2280 if (port == in_port && dst.vlan == flow_vlan) {
2281 /* Don't send out input port on same VLAN. */
2284 dst_set_add(set, &dst);
2289 mirrors &= mirrors - 1;
2294 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2295 const struct port *in_port, const struct port *out_port,
2296 tag_type *tags, struct ofpbuf *actions,
2297 uint16_t *nf_output_iface)
2299 uint16_t initial_vlan, cur_vlan;
2300 const struct dst *dst;
2304 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2306 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2308 /* Output all the packets we can without having to change the VLAN. */
2309 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2310 if (initial_vlan == 0) {
2311 initial_vlan = OFP_VLAN_NONE;
2313 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2314 if (dst->vlan != initial_vlan) {
2317 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2320 /* Then output the rest. */
2321 cur_vlan = initial_vlan;
2322 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2323 if (dst->vlan == initial_vlan) {
2326 if (dst->vlan != cur_vlan) {
2327 if (dst->vlan == OFP_VLAN_NONE) {
2328 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2331 tci = htons(dst->vlan & VLAN_VID_MASK);
2332 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2333 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2335 cur_vlan = dst->vlan;
2337 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2343 /* Returns the effective vlan of a packet, taking into account both the
2344 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2345 * the packet is untagged and -1 indicates it has an invalid header and
2346 * should be dropped. */
2347 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2348 struct port *in_port, bool have_packet)
2350 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2351 if (in_port->vlan >= 0) {
2354 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2355 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2356 "packet received on port %s configured with "
2357 "implicit VLAN %"PRIu16,
2358 br->name, vlan, in_port->name, in_port->vlan);
2362 vlan = in_port->vlan;
2364 if (!port_includes_vlan(in_port, vlan)) {
2366 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2367 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2368 "packet received on port %s not configured for "
2370 br->name, vlan, in_port->name, vlan);
2379 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2380 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2381 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2383 is_gratuitous_arp(const struct flow *flow)
2385 return (flow->dl_type == htons(ETH_TYPE_ARP)
2386 && eth_addr_is_broadcast(flow->dl_dst)
2387 && (flow->nw_proto == ARP_OP_REPLY
2388 || (flow->nw_proto == ARP_OP_REQUEST
2389 && flow->nw_src == flow->nw_dst)));
2393 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2394 struct port *in_port)
2396 struct mac_entry *mac;
2398 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2402 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2403 if (is_gratuitous_arp(flow)) {
2404 /* We don't want to learn from gratuitous ARP packets that are
2405 * reflected back over bond slaves so we lock the learning table. */
2406 if (!in_port->bond) {
2407 mac_entry_set_grat_arp_lock(mac);
2408 } else if (mac_entry_is_grat_arp_locked(mac)) {
2413 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2414 /* The log messages here could actually be useful in debugging,
2415 * so keep the rate limit relatively high. */
2416 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2417 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2418 "on port %s in VLAN %d",
2419 br->name, ETH_ADDR_ARGS(flow->dl_src),
2420 in_port->name, vlan);
2422 mac->port.p = in_port;
2423 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2427 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2428 * dropped. Returns true if they may be forwarded, false if they should be
2431 * If 'have_packet' is true, it indicates that the caller is processing a
2432 * received packet. If 'have_packet' is false, then the caller is just
2433 * revalidating an existing flow because configuration has changed. Either
2434 * way, 'have_packet' only affects logging (there is no point in logging errors
2435 * during revalidation).
2437 * Sets '*in_portp' to the input port. This will be a null pointer if
2438 * flow->in_port does not designate a known input port (in which case
2439 * is_admissible() returns false).
2441 * When returning true, sets '*vlanp' to the effective VLAN of the input
2442 * packet, as returned by flow_get_vlan().
2444 * May also add tags to '*tags', although the current implementation only does
2445 * so in one special case.
2448 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2449 tag_type *tags, int *vlanp, struct port **in_portp)
2451 struct iface *in_iface;
2452 struct port *in_port;
2455 /* Find the interface and port structure for the received packet. */
2456 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2458 /* No interface? Something fishy... */
2460 /* Odd. A few possible reasons here:
2462 * - We deleted an interface but there are still a few packets
2463 * queued up from it.
2465 * - Someone externally added an interface (e.g. with "ovs-dpctl
2466 * add-if") that we don't know about.
2468 * - Packet arrived on the local port but the local port is not
2469 * one of our bridge ports.
2471 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2473 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2474 "interface %"PRIu16, br->name, flow->in_port);
2480 *in_portp = in_port = in_iface->port;
2481 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2486 /* Drop frames for reserved multicast addresses. */
2487 if (eth_addr_is_reserved(flow->dl_dst)) {
2491 /* Drop frames on ports reserved for mirroring. */
2492 if (in_port->is_mirror_output_port) {
2494 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2495 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2496 "%s, which is reserved exclusively for mirroring",
2497 br->name, in_port->name);
2502 if (in_port->bond) {
2503 struct mac_entry *mac;
2505 switch (bond_check_admissibility(in_port->bond, in_iface,
2506 flow->dl_dst, tags)) {
2513 case BV_DROP_IF_MOVED:
2514 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2515 if (mac && mac->port.p != in_port &&
2516 (!is_gratuitous_arp(flow)
2517 || mac_entry_is_grat_arp_locked(mac))) {
2527 /* If the composed actions may be applied to any packet in the given 'flow',
2528 * returns true. Otherwise, the actions should only be applied to 'packet', or
2529 * not at all, if 'packet' was NULL. */
2531 process_flow(struct bridge *br, const struct flow *flow,
2532 const struct ofpbuf *packet, struct ofpbuf *actions,
2533 tag_type *tags, uint16_t *nf_output_iface)
2535 struct port *in_port;
2536 struct port *out_port;
2537 struct mac_entry *mac;
2540 /* Check whether we should drop packets in this flow. */
2541 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2546 /* Learn source MAC (but don't try to learn from revalidation). */
2548 update_learning_table(br, flow, vlan, in_port);
2551 /* Determine output port. */
2552 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2554 out_port = mac->port.p;
2555 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2556 /* If we are revalidating but don't have a learning entry then
2557 * eject the flow. Installing a flow that floods packets opens
2558 * up a window of time where we could learn from a packet reflected
2559 * on a bond and blackhole packets before the learning table is
2560 * updated to reflect the correct port. */
2563 out_port = FLOOD_PORT;
2566 /* Don't send packets out their input ports. */
2567 if (in_port == out_port) {
2573 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2581 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2582 struct ofpbuf *actions, tag_type *tags,
2583 uint16_t *nf_output_iface, void *br_)
2585 struct bridge *br = br_;
2587 COVERAGE_INC(bridge_process_flow);
2588 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2592 bridge_special_ofhook_cb(const struct flow *flow,
2593 const struct ofpbuf *packet, void *br_)
2595 struct iface *iface;
2596 struct bridge *br = br_;
2598 iface = iface_from_dp_ifidx(br, flow->in_port);
2600 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2601 if (iface && iface->port->lacp && packet) {
2602 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2604 lacp_process_pdu(iface->port->lacp, iface, pdu);
2614 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2615 const struct nlattr *actions,
2617 uint64_t n_bytes, void *br_)
2619 struct bridge *br = br_;
2620 const struct nlattr *a;
2621 struct port *in_port;
2626 /* Feed information from the active flows back into the learning table to
2627 * ensure that table is always in sync with what is actually flowing
2628 * through the datapath.
2630 * We test that 'tags' is nonzero to ensure that only flows that include an
2631 * OFPP_NORMAL action are used for learning. This works because
2632 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2633 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2634 update_learning_table(br, flow, vlan, in_port);
2637 /* Account for bond slave utilization. */
2638 if (!br->has_bonded_ports) {
2641 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2642 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2643 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2644 if (out_port && out_port->bond) {
2645 uint16_t vlan = (flow->vlan_tci
2646 ? vlan_tci_to_vid(flow->vlan_tci)
2648 bond_account(out_port->bond, flow, vlan, n_bytes);
2655 bridge_account_checkpoint_ofhook_cb(void *br_)
2657 struct bridge *br = br_;
2660 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2662 bond_rebalance(port->bond,
2663 ofproto_get_revalidate_set(br->ofproto));
2669 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2670 tag_type *tags, void *br_)
2672 struct bridge *br = br_;
2673 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2674 struct port *port = port_from_dp_ifidx(br, odp_port);
2679 } else if (list_is_short(&port->ifaces)) {
2682 struct iface *iface;
2684 /* Autopath does not support VLAN hashing. */
2685 iface = bond_choose_output_slave(port->bond, flow,
2686 OFP_VLAN_NONE, tags);
2687 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2690 return odp_port_to_ofp_port(ret);
2693 static struct ofhooks bridge_ofhooks = {
2694 bridge_normal_ofhook_cb,
2695 bridge_special_ofhook_cb,
2696 bridge_account_flow_ofhook_cb,
2697 bridge_account_checkpoint_ofhook_cb,
2698 bridge_autopath_ofhook_cb,
2701 /* Port functions. */
2704 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2706 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2707 struct iface *iface = iface_;
2708 uint8_t ea[ETH_ADDR_LEN];
2711 error = netdev_get_etheraddr(iface->netdev, ea);
2713 struct lacp_pdu *packet_pdu;
2714 struct ofpbuf packet;
2716 ofpbuf_init(&packet, 0);
2717 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2718 sizeof *packet_pdu);
2720 error = netdev_send(iface->netdev, &packet);
2722 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2723 "(%s)", iface->port->name, iface->name,
2726 ofpbuf_uninit(&packet);
2728 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2729 "%s (%s)", iface->port->name, iface->name,
2735 port_run(struct port *port)
2738 lacp_run(port->lacp, lacp_send_pdu_cb);
2742 struct iface *iface;
2744 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2745 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2746 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2749 bond_run(port->bond,
2750 ofproto_get_revalidate_set(port->bridge->ofproto),
2751 lacp_negotiated(port->lacp));
2752 if (bond_should_send_learning_packets(port->bond)) {
2753 port_send_learning_packets(port);
2759 port_wait(struct port *port)
2762 lacp_wait(port->lacp);
2766 bond_wait(port->bond);
2770 static struct port *
2771 port_create(struct bridge *br, const char *name)
2775 port = xzalloc(sizeof *port);
2778 port->trunks = NULL;
2779 port->name = xstrdup(name);
2780 list_init(&port->ifaces);
2782 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2784 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2791 get_port_other_config(const struct ovsrec_port *port, const char *key,
2792 const char *default_value)
2796 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2798 return value ? value : default_value;
2802 get_interface_other_config(const struct ovsrec_interface *iface,
2803 const char *key, const char *default_value)
2807 value = get_ovsrec_key_value(&iface->header_,
2808 &ovsrec_interface_col_other_config, key);
2809 return value ? value : default_value;
2813 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2815 struct iface *iface, *next;
2816 struct sset new_ifaces;
2819 /* Collect list of new interfaces. */
2820 sset_init(&new_ifaces);
2821 for (i = 0; i < cfg->n_interfaces; i++) {
2822 const char *name = cfg->interfaces[i]->name;
2823 sset_add(&new_ifaces, name);
2826 /* Get rid of deleted interfaces. */
2827 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2828 if (!sset_contains(&new_ifaces, iface->name)) {
2829 iface_destroy(iface);
2833 sset_destroy(&new_ifaces);
2836 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2837 * to revalidate every flow. */
2839 port_flush_macs(struct port *port)
2841 struct bridge *br = port->bridge;
2842 struct mac_learning *ml = br->ml;
2843 struct mac_entry *mac, *next_mac;
2846 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2847 if (mac->port.p == port) {
2848 mac_learning_expire(ml, mac);
2854 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2856 struct sset new_ifaces;
2857 bool need_flush = false;
2858 unsigned long *trunks;
2865 /* Add new interfaces and update 'cfg' member of existing ones. */
2866 sset_init(&new_ifaces);
2867 for (i = 0; i < cfg->n_interfaces; i++) {
2868 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2869 struct iface *iface;
2871 if (!sset_add(&new_ifaces, if_cfg->name)) {
2872 VLOG_WARN("port %s: %s specified twice as port interface",
2873 port->name, if_cfg->name);
2874 iface_set_ofport(if_cfg, -1);
2878 iface = iface_lookup(port->bridge, if_cfg->name);
2880 if (iface->port != port) {
2881 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2883 port->bridge->name, if_cfg->name, iface->port->name);
2886 iface->cfg = if_cfg;
2888 iface = iface_create(port, if_cfg);
2891 /* Determine interface type. The local port always has type
2892 * "internal". Other ports take their type from the database and
2893 * default to "system" if none is specified. */
2894 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2895 : if_cfg->type[0] ? if_cfg->type
2898 sset_destroy(&new_ifaces);
2903 if (list_is_short(&port->ifaces)) {
2905 if (vlan >= 0 && vlan <= 4095) {
2906 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2911 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2912 * they even work as-is. But they have not been tested. */
2913 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2917 if (port->vlan != vlan) {
2922 /* Get trunked VLANs. */
2924 if (vlan < 0 && cfg->n_trunks) {
2925 trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
2927 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2930 } else if (vlan >= 0 && cfg->n_trunks) {
2931 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2934 if (!vlan_bitmap_equal(trunks, port->trunks)) {
2937 bitmap_free(port->trunks);
2938 port->trunks = trunks;
2941 port_flush_macs(port);
2946 port_destroy(struct port *port)
2949 struct bridge *br = port->bridge;
2950 struct iface *iface, *next;
2953 for (i = 0; i < MAX_MIRRORS; i++) {
2954 struct mirror *m = br->mirrors[i];
2955 if (m && m->out_port == port) {
2960 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2961 iface_destroy(iface);
2964 hmap_remove(&br->ports, &port->hmap_node);
2966 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2968 bond_destroy(port->bond);
2969 lacp_destroy(port->lacp);
2970 port_flush_macs(port);
2972 bitmap_free(port->trunks);
2978 static struct port *
2979 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
2981 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
2982 return iface ? iface->port : NULL;
2985 static struct port *
2986 port_lookup(const struct bridge *br, const char *name)
2990 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2992 if (!strcmp(port->name, name)) {
3000 enable_lacp(struct port *port, bool *activep)
3002 if (!port->cfg->lacp) {
3003 /* XXX when LACP implementation has been sufficiently tested, enable by
3004 * default and make active on bonded ports. */
3006 } else if (!strcmp(port->cfg->lacp, "off")) {
3008 } else if (!strcmp(port->cfg->lacp, "active")) {
3011 } else if (!strcmp(port->cfg->lacp, "passive")) {
3015 VLOG_WARN("port %s: unknown LACP mode %s",
3016 port->name, port->cfg->lacp);
3022 iface_reconfigure_lacp(struct iface *iface)
3024 struct lacp_slave_settings s;
3025 int priority, portid;
3027 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
3028 priority = atoi(get_interface_other_config(iface->cfg,
3029 "lacp-port-priority", "0"));
3031 if (portid <= 0 || portid > UINT16_MAX) {
3032 portid = iface->dp_ifidx;
3035 if (priority <= 0 || priority > UINT16_MAX) {
3036 priority = UINT16_MAX;
3039 s.name = iface->name;
3041 s.priority = priority;
3042 lacp_slave_register(iface->port->lacp, iface, &s);
3046 port_reconfigure_lacp(struct port *port)
3048 static struct lacp_settings s;
3049 struct iface *iface;
3050 uint8_t sysid[ETH_ADDR_LEN];
3051 const char *sysid_str;
3052 const char *lacp_time;
3053 long long int custom_time;
3056 if (!enable_lacp(port, &s.active)) {
3057 lacp_destroy(port->lacp);
3062 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3063 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3064 memcpy(s.id, sysid, ETH_ADDR_LEN);
3066 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3069 s.name = port->name;
3071 /* Prefer bondable links if unspecified. */
3072 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3074 s.priority = (priority > 0 && priority <= UINT16_MAX
3076 : UINT16_MAX - !list_is_short(&port->ifaces));
3078 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3082 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3083 custom_time = atoi(lacp_time);
3084 if (!strcmp(lacp_time, "fast")) {
3085 s.lacp_time = LACP_TIME_FAST;
3086 } else if (!strcmp(lacp_time, "slow")) {
3087 s.lacp_time = LACP_TIME_SLOW;
3088 } else if (custom_time > 0) {
3089 s.lacp_time = LACP_TIME_CUSTOM;
3090 s.custom_time = custom_time;
3092 s.lacp_time = LACP_TIME_SLOW;
3096 port->lacp = lacp_create();
3099 lacp_configure(port->lacp, &s);
3101 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3102 iface_reconfigure_lacp(iface);
3107 port_reconfigure_bond(struct port *port)
3109 struct bond_settings s;
3110 const char *detect_s;
3111 struct iface *iface;
3113 if (list_is_short(&port->ifaces)) {
3114 /* Not a bonded port. */
3115 bond_destroy(port->bond);
3120 port->bridge->has_bonded_ports = true;
3122 s.name = port->name;
3124 if (port->cfg->bond_mode
3125 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3126 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3127 port->name, port->cfg->bond_mode,
3128 bond_mode_to_string(s.balance));
3131 s.detect = BLSM_CARRIER;
3132 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3133 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3134 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3136 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3139 s.miimon_interval = atoi(
3140 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3141 if (s.miimon_interval < 100) {
3142 s.miimon_interval = 100;
3145 s.up_delay = MAX(0, port->cfg->bond_updelay);
3146 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3147 s.rebalance_interval = atoi(
3148 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3149 if (s.rebalance_interval < 1000) {
3150 s.rebalance_interval = 1000;
3153 s.fake_iface = port->cfg->bond_fake_iface;
3156 port->bond = bond_create(&s);
3158 if (bond_reconfigure(port->bond, &s)) {
3159 bridge_flush(port->bridge);
3163 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3164 uint16_t stable_id = (port->lacp
3165 ? lacp_slave_get_port_id(port->lacp, iface)
3167 bond_slave_register(iface->port->bond, iface, stable_id,
3173 port_send_learning_packets(struct port *port)
3175 struct bridge *br = port->bridge;
3176 int error, n_packets, n_errors;
3177 struct mac_entry *e;
3179 error = n_packets = n_errors = 0;
3180 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3181 if (e->port.p != port) {
3182 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3192 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3193 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3194 "packets, last error was: %s",
3195 port->name, n_errors, n_packets, strerror(error));
3197 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3198 port->name, n_packets);
3202 /* Interface functions. */
3204 static struct iface *
3205 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3207 struct bridge *br = port->bridge;
3208 struct iface *iface;
3209 char *name = if_cfg->name;
3211 iface = xzalloc(sizeof *iface);
3213 iface->name = xstrdup(name);
3214 iface->dp_ifidx = -1;
3215 iface->tag = tag_create_random();
3216 iface->netdev = NULL;
3217 iface->cfg = if_cfg;
3219 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
3221 list_push_back(&port->ifaces, &iface->port_elem);
3223 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3231 iface_destroy(struct iface *iface)
3234 struct port *port = iface->port;
3235 struct bridge *br = port->bridge;
3238 bond_slave_unregister(port->bond, iface);
3242 lacp_slave_unregister(port->lacp, iface);
3245 if (iface->dp_ifidx >= 0) {
3246 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3249 list_remove(&iface->port_elem);
3250 hmap_remove(&br->iface_by_name, &iface->name_node);
3252 netdev_close(iface->netdev);
3257 bridge_flush(port->bridge);
3261 static struct iface *
3262 iface_lookup(const struct bridge *br, const char *name)
3264 struct iface *iface;
3266 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3267 &br->iface_by_name) {
3268 if (!strcmp(iface->name, name)) {
3276 static struct iface *
3277 iface_find(const char *name)
3279 const struct bridge *br;
3281 HMAP_FOR_EACH (br, node, &all_bridges) {
3282 struct iface *iface = iface_lookup(br, name);
3291 static struct iface *
3292 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3294 struct iface *iface;
3296 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3297 hash_int(dp_ifidx, 0), &br->ifaces) {
3298 if (iface->dp_ifidx == dp_ifidx) {
3305 /* Set Ethernet address of 'iface', if one is specified in the configuration
3308 iface_set_mac(struct iface *iface)
3310 uint8_t ea[ETH_ADDR_LEN];
3312 if (!strcmp(iface->type, "internal")
3313 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3314 if (iface->dp_ifidx == ODPP_LOCAL) {
3315 VLOG_ERR("interface %s: ignoring mac in Interface record "
3316 "(use Bridge record to set local port's mac)",
3318 } else if (eth_addr_is_multicast(ea)) {
3319 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3322 int error = netdev_set_etheraddr(iface->netdev, ea);
3324 VLOG_ERR("interface %s: setting MAC failed (%s)",
3325 iface->name, strerror(error));
3331 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3333 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3335 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3336 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3340 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3342 * The value strings in '*shash' are taken directly from values[], not copied,
3343 * so the caller should not modify or free them. */
3345 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3346 struct shash *shash)
3351 for (i = 0; i < n; i++) {
3352 shash_add(shash, keys[i], values[i]);
3356 /* Creates 'keys' and 'values' arrays from 'shash'.
3358 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3359 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3360 * are populated with with strings taken directly from 'shash' and thus have
3361 * the same ownership of the key-value pairs in shash.
3364 shash_to_ovs_idl_map(struct shash *shash,
3365 char ***keys, char ***values, size_t *n)
3369 struct shash_node *sn;
3371 count = shash_count(shash);
3373 k = xmalloc(count * sizeof *k);
3374 v = xmalloc(count * sizeof *v);
3377 SHASH_FOR_EACH(sn, shash) {
3388 struct iface_delete_queues_cbdata {
3389 struct netdev *netdev;
3390 const struct ovsdb_datum *queues;
3394 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3396 union ovsdb_atom atom;
3398 atom.integer = target;
3399 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3403 iface_delete_queues(unsigned int queue_id,
3404 const struct shash *details OVS_UNUSED, void *cbdata_)
3406 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3408 if (!queue_ids_include(cbdata->queues, queue_id)) {
3409 netdev_delete_queue(cbdata->netdev, queue_id);
3414 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3416 if (!qos || qos->type[0] == '\0') {
3417 netdev_set_qos(iface->netdev, NULL, NULL);
3419 struct iface_delete_queues_cbdata cbdata;
3420 struct shash details;
3423 /* Configure top-level Qos for 'iface'. */
3424 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3425 qos->n_other_config, &details);
3426 netdev_set_qos(iface->netdev, qos->type, &details);
3427 shash_destroy(&details);
3429 /* Deconfigure queues that were deleted. */
3430 cbdata.netdev = iface->netdev;
3431 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3433 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3435 /* Configure queues for 'iface'. */
3436 for (i = 0; i < qos->n_queues; i++) {
3437 const struct ovsrec_queue *queue = qos->value_queues[i];
3438 unsigned int queue_id = qos->key_queues[i];
3440 shash_from_ovs_idl_map(queue->key_other_config,
3441 queue->value_other_config,
3442 queue->n_other_config, &details);
3443 netdev_set_queue(iface->netdev, queue_id, &details);
3444 shash_destroy(&details);
3450 iface_update_cfm(struct iface *iface)
3454 uint16_t *remote_mps;
3455 struct ovsrec_monitor *mon;
3456 uint8_t maid[CCM_MAID_LEN];
3458 mon = iface->cfg->monitor;
3461 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3465 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3466 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3470 cfm.mpid = mon->mpid;
3471 cfm.interval = mon->interval ? *mon->interval : 1000;
3473 memcpy(cfm.maid, maid, sizeof cfm.maid);
3475 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3476 for(i = 0; i < mon->n_remote_mps; i++) {
3477 remote_mps[i] = mon->remote_mps[i]->mpid;
3480 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3481 &cfm, remote_mps, mon->n_remote_mps);
3485 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3486 * how 'iface''s port is configured.
3488 * Returns true if 'iface' is up, false otherwise. */
3490 iface_get_carrier(const struct iface *iface)
3493 return netdev_get_carrier(iface->netdev);
3496 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3497 * instead of obtaining it from the database. */
3499 iface_is_synthetic(const struct iface *iface)
3501 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3504 /* Port mirroring. */
3506 static struct mirror *
3507 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3511 for (i = 0; i < MAX_MIRRORS; i++) {
3512 struct mirror *m = br->mirrors[i];
3513 if (m && uuid_equals(uuid, &m->uuid)) {
3521 mirror_reconfigure(struct bridge *br)
3523 unsigned long *rspan_vlans;
3527 /* Get rid of deleted mirrors. */
3528 for (i = 0; i < MAX_MIRRORS; i++) {
3529 struct mirror *m = br->mirrors[i];
3531 const struct ovsdb_datum *mc;
3532 union ovsdb_atom atom;
3534 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3535 atom.uuid = br->mirrors[i]->uuid;
3536 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3542 /* Add new mirrors and reconfigure existing ones. */
3543 for (i = 0; i < br->cfg->n_mirrors; i++) {
3544 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3545 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3547 mirror_reconfigure_one(m, cfg);
3549 mirror_create(br, cfg);
3553 /* Update port reserved status. */
3554 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3555 port->is_mirror_output_port = false;
3557 for (i = 0; i < MAX_MIRRORS; i++) {
3558 struct mirror *m = br->mirrors[i];
3559 if (m && m->out_port) {
3560 m->out_port->is_mirror_output_port = true;
3564 /* Update flooded vlans (for RSPAN). */
3566 if (br->cfg->n_flood_vlans) {
3567 rspan_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3568 br->cfg->n_flood_vlans);
3570 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3572 mac_learning_flush(br->ml);
3578 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3583 for (i = 0; ; i++) {
3584 if (i >= MAX_MIRRORS) {
3585 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3586 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3589 if (!br->mirrors[i]) {
3594 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3596 mac_learning_flush(br->ml);
3598 br->mirrors[i] = m = xzalloc(sizeof *m);
3599 m->uuid = cfg->header_.uuid;
3602 m->name = xstrdup(cfg->name);
3603 sset_init(&m->src_ports);
3604 sset_init(&m->dst_ports);
3610 mirror_reconfigure_one(m, cfg);
3614 mirror_destroy(struct mirror *m)
3617 struct bridge *br = m->bridge;
3620 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3621 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3622 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3625 sset_destroy(&m->src_ports);
3626 sset_destroy(&m->dst_ports);
3629 m->bridge->mirrors[m->idx] = NULL;
3634 mac_learning_flush(br->ml);
3639 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3644 for (i = 0; i < n_ports; i++) {
3645 const char *name = ports[i]->name;
3646 if (port_lookup(m->bridge, name)) {
3647 sset_add(names, name);
3649 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3650 "port %s", m->bridge->name, m->name, name);
3656 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3662 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3664 for (i = 0; i < cfg->n_select_vlan; i++) {
3665 int64_t vlan = cfg->select_vlan[i];
3666 if (vlan < 0 || vlan > 4095) {
3667 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3668 m->bridge->name, m->name, vlan);
3670 (*vlans)[n_vlans++] = vlan;
3677 vlan_is_mirrored(const struct mirror *m, int vlan)
3681 for (i = 0; i < m->n_vlans; i++) {
3682 if (m->vlans[i] == vlan) {
3690 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3692 struct sset src_ports, dst_ports;
3693 mirror_mask_t mirror_bit;
3694 struct port *out_port;
3701 if (strcmp(cfg->name, m->name)) {
3703 m->name = xstrdup(cfg->name);
3706 /* Get output port. */
3707 if (cfg->output_port) {
3708 out_port = port_lookup(m->bridge, cfg->output_port->name);
3710 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3711 m->bridge->name, m->name);
3717 if (cfg->output_vlan) {
3718 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3719 "output vlan; ignoring output vlan",
3720 m->bridge->name, m->name);
3722 } else if (cfg->output_vlan) {
3724 out_vlan = *cfg->output_vlan;
3726 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3727 m->bridge->name, m->name);
3732 sset_init(&src_ports);
3733 sset_init(&dst_ports);
3734 if (cfg->select_all) {
3735 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3736 sset_add(&src_ports, port->name);
3737 sset_add(&dst_ports, port->name);
3742 /* Get ports, and drop duplicates and ports that don't exist. */
3743 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3745 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3748 /* Get all the vlans, and drop duplicate and invalid vlans. */
3749 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3752 /* Update mirror data. */
3753 if (!sset_equals(&m->src_ports, &src_ports)
3754 || !sset_equals(&m->dst_ports, &dst_ports)
3755 || m->n_vlans != n_vlans
3756 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3757 || m->out_port != out_port
3758 || m->out_vlan != out_vlan) {
3759 bridge_flush(m->bridge);
3760 mac_learning_flush(m->bridge->ml);
3762 sset_swap(&m->src_ports, &src_ports);
3763 sset_swap(&m->dst_ports, &dst_ports);
3766 m->n_vlans = n_vlans;
3767 m->out_port = out_port;
3768 m->out_vlan = out_vlan;
3771 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3772 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3773 if (sset_contains(&m->src_ports, port->name)) {
3774 port->src_mirrors |= mirror_bit;
3776 port->src_mirrors &= ~mirror_bit;
3779 if (sset_contains(&m->dst_ports, port->name)) {
3780 port->dst_mirrors |= mirror_bit;
3782 port->dst_mirrors &= ~mirror_bit;
3787 sset_destroy(&src_ports);
3788 sset_destroy(&dst_ports);