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_flush(struct bridge *);
224 static void bridge_pick_local_hw_addr(struct bridge *,
225 uint8_t ea[ETH_ADDR_LEN],
226 struct iface **hw_addr_iface);
227 static uint64_t bridge_pick_datapath_id(struct bridge *,
228 const uint8_t bridge_ea[ETH_ADDR_LEN],
229 struct iface *hw_addr_iface);
230 static uint64_t dpid_from_hash(const void *, size_t nbytes);
231 static bool bridge_has_bond_fake_iface(const struct bridge *,
233 static bool port_is_bond_fake_iface(const struct port *);
235 static unixctl_cb_func bridge_unixctl_fdb_show;
236 static unixctl_cb_func cfm_unixctl_show;
237 static unixctl_cb_func qos_unixctl_show;
239 static void port_run(struct port *);
240 static void port_wait(struct port *);
241 static struct port *port_create(struct bridge *, const char *name);
242 static void port_reconfigure(struct port *, const struct ovsrec_port *);
243 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
244 static void port_destroy(struct port *);
245 static struct port *port_lookup(const struct bridge *, const char *name);
246 static struct iface *port_get_an_iface(const struct port *);
247 static struct port *port_from_dp_ifidx(const struct bridge *,
249 static void port_reconfigure_lacp(struct port *);
250 static void port_reconfigure_bond(struct port *);
251 static void port_send_learning_packets(struct port *);
253 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
254 static void mirror_destroy(struct mirror *);
255 static void mirror_reconfigure(struct bridge *);
256 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
257 static bool vlan_is_mirrored(const struct mirror *, int vlan);
259 static struct iface *iface_create(struct port *port,
260 const struct ovsrec_interface *if_cfg);
261 static void iface_destroy(struct iface *);
262 static struct iface *iface_lookup(const struct bridge *, const char *name);
263 static struct iface *iface_find(const char *name);
264 static struct iface *iface_from_dp_ifidx(const struct bridge *,
266 static void iface_set_mac(struct iface *);
267 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
268 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
269 static void iface_update_cfm(struct iface *);
270 static bool iface_refresh_cfm_stats(struct iface *iface);
271 static bool iface_get_carrier(const struct iface *);
272 static bool iface_is_synthetic(const struct iface *);
274 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
276 static void shash_to_ovs_idl_map(struct shash *,
277 char ***keys, char ***values, size_t *n);
279 /* Hooks into ofproto processing. */
280 static struct ofhooks bridge_ofhooks;
282 /* Public functions. */
284 /* Initializes the bridge module, configuring it to obtain its configuration
285 * from an OVSDB server accessed over 'remote', which should be a string in a
286 * form acceptable to ovsdb_idl_create(). */
288 bridge_init(const char *remote)
290 /* Create connection to database. */
291 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
293 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
294 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
295 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
296 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
297 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
298 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
299 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
301 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
302 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
304 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
305 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
307 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
315 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
317 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
319 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
320 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
322 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
324 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
326 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
328 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
330 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
332 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
334 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
336 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
338 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
339 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
340 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
342 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
344 /* Register unixctl commands. */
345 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
346 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
347 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
348 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
350 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
359 struct bridge *br, *next_br;
361 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
364 ovsdb_idl_destroy(idl);
367 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
368 * but for which the ovs-vswitchd configuration 'cfg' is required. */
370 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
372 static bool already_configured_once;
373 struct sset bridge_names;
374 struct sset dpif_names, dpif_types;
378 /* Only do this once per ovs-vswitchd run. */
379 if (already_configured_once) {
382 already_configured_once = true;
384 stats_timer = time_msec() + STATS_INTERVAL;
386 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
387 sset_init(&bridge_names);
388 for (i = 0; i < cfg->n_bridges; i++) {
389 sset_add(&bridge_names, cfg->bridges[i]->name);
392 /* Iterate over all system dpifs and delete any of them that do not appear
394 sset_init(&dpif_names);
395 sset_init(&dpif_types);
396 dp_enumerate_types(&dpif_types);
397 SSET_FOR_EACH (type, &dpif_types) {
400 dp_enumerate_names(type, &dpif_names);
402 /* Delete each dpif whose name is not in 'bridge_names'. */
403 SSET_FOR_EACH (name, &dpif_names) {
404 if (!sset_contains(&bridge_names, name)) {
408 retval = dpif_open(name, type, &dpif);
416 sset_destroy(&bridge_names);
417 sset_destroy(&dpif_names);
418 sset_destroy(&dpif_types);
421 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
422 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
423 * responsible for freeing '*managersp' (with free()).
425 * You may be asking yourself "why does ovs-vswitchd care?", because
426 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
427 * should not be and in fact is not directly involved in that. But
428 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
429 * it has to tell in-band control where the managers are to enable that.
430 * (Thus, only managers connected in-band are collected.)
433 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
434 struct sockaddr_in **managersp, size_t *n_managersp)
436 struct sockaddr_in *managers = NULL;
437 size_t n_managers = 0;
441 /* Collect all of the potential targets from the "targets" columns of the
442 * rows pointed to by "manager_options", excluding any that are
445 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
446 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
448 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
449 sset_find_and_delete(&targets, m->target);
451 sset_add(&targets, m->target);
455 /* Now extract the targets' IP addresses. */
456 if (!sset_is_empty(&targets)) {
459 managers = xmalloc(sset_count(&targets) * sizeof *managers);
460 SSET_FOR_EACH (target, &targets) {
461 struct sockaddr_in *sin = &managers[n_managers];
463 if ((!strncmp(target, "tcp:", 4)
464 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
465 (!strncmp(target, "ssl:", 4)
466 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
471 sset_destroy(&targets);
473 *managersp = managers;
474 *n_managersp = n_managers;
478 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
480 struct shash_node *node;
481 struct bridge *br, *next;
482 struct sockaddr_in *managers;
486 int sflow_bridge_number;
488 COVERAGE_INC(bridge_reconfigure);
490 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
492 /* Collect old and new bridges. */
494 for (i = 0; i < ovs_cfg->n_bridges; i++) {
495 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
496 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
497 VLOG_WARN("more than one bridge named %s", br_cfg->name);
501 /* Get rid of deleted bridges and add new bridges. */
502 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
503 br->cfg = shash_find_data(&new_br, br->name);
508 SHASH_FOR_EACH (node, &new_br) {
509 const struct ovsrec_bridge *br_cfg = node->data;
510 struct bridge *br = bridge_lookup(node->name);
512 /* If the bridge datapath type has changed, we need to tear it
513 * down and recreate. */
514 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
516 bridge_create(br_cfg);
519 bridge_create(br_cfg);
522 shash_destroy(&new_br);
524 /* Reconfigure all bridges. */
525 HMAP_FOR_EACH (br, node, &all_bridges) {
526 bridge_reconfigure_one(br);
529 /* Add and delete ports on all datapaths.
531 * The kernel will reject any attempt to add a given port to a datapath if
532 * that port already belongs to a different datapath, so we must do all
533 * port deletions before any port additions. */
534 HMAP_FOR_EACH (br, node, &all_bridges) {
535 struct ofproto_port_dump dump;
536 struct ofproto_port ofproto_port;
538 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
539 const char *name = ofproto_port.name;
544 /* Ignore the local port. We can't change it anyhow. */
545 if (!strcmp(name, br->name)) {
549 /* Get the type that 'ofproto_port' should have (ordinarily the
550 * type of its corresponding iface) or NULL if it should be
552 iface = iface_lookup(br, name);
553 type = (iface ? iface->type
554 : bridge_has_bond_fake_iface(br, name) ? "internal"
557 /* If it's the wrong type then delete the ofproto port. */
559 && !strcmp(ofproto_port.type, type)
560 && (!iface || !iface->netdev
561 || !strcmp(netdev_get_type(iface->netdev), type))) {
564 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
566 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
567 br->name, name, strerror(error));
570 if (iface->port->bond) {
571 /* The bond has a pointer to the netdev, so remove it from
572 * the bond before closing the netdev. The slave will get
573 * added back to the bond later, after a new netdev is
575 bond_slave_unregister(iface->port->bond, iface);
577 netdev_close(iface->netdev);
578 iface->netdev = NULL;
582 HMAP_FOR_EACH (br, node, &all_bridges) {
583 struct ofproto_port ofproto_port;
584 struct ofproto_port_dump dump;
585 struct port *port, *next_port;
587 /* Clear all the "dp_ifidx"es. */
588 hmap_clear(&br->ifaces);
589 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
592 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
593 iface->dp_ifidx = -1;
597 /* Obtain the correct "dp_ifidx"es from ofproto. */
598 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
599 struct iface *iface = iface_lookup(br, ofproto_port.name);
603 odp_port = ofp_port_to_odp_port(ofproto_port.ofp_port);
604 if (iface->dp_ifidx >= 0) {
605 VLOG_WARN("bridge %s: interface %s reported twice",
606 br->name, ofproto_port.name);
607 } else if (iface_from_dp_ifidx(br, odp_port)) {
608 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
611 iface->dp_ifidx = odp_port;
612 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
613 hash_int(iface->dp_ifidx, 0));
618 /* Add a dpif port for any "struct iface" that doesn't have one.
619 * Delete any "struct iface" for which this fails.
620 * Delete any "struct port" that thereby ends up with no ifaces. */
621 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
622 struct iface *iface, *next_iface;
624 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
628 /* Open the netdev or reconfigure it. */
630 shash_from_ovs_idl_map(iface->cfg->key_options,
631 iface->cfg->value_options,
632 iface->cfg->n_options, &args);
633 if (!iface->netdev) {
634 struct netdev_options options;
635 options.name = iface->name;
636 options.type = iface->type;
637 options.args = &args;
638 options.ethertype = NETDEV_ETH_TYPE_NONE;
639 error = netdev_open(&options, &iface->netdev);
641 error = netdev_set_config(iface->netdev, &args);
643 shash_destroy(&args);
645 VLOG_WARN("could not %s network device %s (%s)",
646 iface->netdev ? "reconfigure" : "open",
647 iface->name, strerror(error));
650 /* Add the port, if necessary. */
651 if (iface->netdev && iface->dp_ifidx < 0) {
655 error = ofproto_port_add(br->ofproto, iface->netdev,
658 iface->dp_ifidx = ofp_port_to_odp_port(ofp_port);
660 netdev_close(iface->netdev);
661 iface->netdev = NULL;
665 /* Delete the iface if */
666 if (iface->netdev && iface->dp_ifidx >= 0) {
667 VLOG_DBG("bridge %s: interface %s is on port %d",
668 br->name, iface->name, iface->dp_ifidx);
671 VLOG_ERR("bridge %s: missing %s interface, dropping",
672 br->name, iface->name);
674 /* We already reported a related error, don't bother
677 iface_set_ofport(iface->cfg, -1);
678 iface_destroy(iface);
681 if (list_is_empty(&port->ifaces)) {
682 VLOG_WARN("%s port has no interfaces, dropping", port->name);
687 /* Add bond fake iface if necessary. */
688 if (port_is_bond_fake_iface(port)) {
689 if (ofproto_port_query_by_name(br->ofproto, port->name,
691 struct netdev_options options;
692 struct netdev *netdev;
695 options.name = port->name;
696 options.type = "internal";
698 options.ethertype = NETDEV_ETH_TYPE_NONE;
699 error = netdev_open(&options, &netdev);
701 ofproto_port_add(br->ofproto, netdev, NULL);
702 netdev_close(netdev);
704 VLOG_WARN("could not open network device %s (%s)",
705 port->name, strerror(error));
708 /* Already exists, nothing to do. */
709 ofproto_port_destroy(&ofproto_port);
714 sflow_bridge_number = 0;
715 HMAP_FOR_EACH (br, node, &all_bridges) {
716 uint8_t ea[ETH_ADDR_LEN];
718 struct iface *local_iface;
719 struct iface *hw_addr_iface;
722 /* Pick local port hardware address, datapath ID. */
723 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
724 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
726 int error = netdev_set_etheraddr(local_iface->netdev, ea);
728 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
729 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
730 "Ethernet address: %s",
731 br->name, strerror(error));
734 memcpy(br->ea, ea, ETH_ADDR_LEN);
736 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
737 ofproto_set_datapath_id(br->ofproto, dpid);
739 dpid_string = xasprintf("%016"PRIx64, dpid);
740 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
743 /* Set NetFlow configuration on this bridge. */
744 if (br->cfg->netflow) {
745 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
746 struct netflow_options opts;
748 memset(&opts, 0, sizeof opts);
750 ofproto_get_netflow_ids(br->ofproto,
751 &opts.engine_type, &opts.engine_id);
752 if (nf_cfg->engine_type) {
753 opts.engine_type = *nf_cfg->engine_type;
755 if (nf_cfg->engine_id) {
756 opts.engine_id = *nf_cfg->engine_id;
759 opts.active_timeout = nf_cfg->active_timeout;
760 if (!opts.active_timeout) {
761 opts.active_timeout = -1;
762 } else if (opts.active_timeout < 0) {
763 VLOG_WARN("bridge %s: active timeout interval set to negative "
764 "value, using default instead (%d seconds)", br->name,
765 NF_ACTIVE_TIMEOUT_DEFAULT);
766 opts.active_timeout = -1;
769 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
770 if (opts.add_id_to_iface) {
771 if (opts.engine_id > 0x7f) {
772 VLOG_WARN("bridge %s: netflow port mangling may conflict "
773 "with another vswitch, choose an engine id less "
774 "than 128", br->name);
776 if (hmap_count(&br->ports) > 508) {
777 VLOG_WARN("bridge %s: netflow port mangling will conflict "
778 "with another port when more than 508 ports are "
783 sset_init(&opts.collectors);
784 sset_add_array(&opts.collectors,
785 nf_cfg->targets, nf_cfg->n_targets);
786 if (ofproto_set_netflow(br->ofproto, &opts)) {
787 VLOG_ERR("bridge %s: problem setting netflow collectors",
790 sset_destroy(&opts.collectors);
792 ofproto_set_netflow(br->ofproto, NULL);
795 /* Set sFlow configuration on this bridge. */
796 if (br->cfg->sflow) {
797 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
798 struct ovsrec_controller **controllers;
799 struct ofproto_sflow_options oso;
800 size_t n_controllers;
802 memset(&oso, 0, sizeof oso);
804 sset_init(&oso.targets);
805 sset_add_array(&oso.targets,
806 sflow_cfg->targets, sflow_cfg->n_targets);
808 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
809 if (sflow_cfg->sampling) {
810 oso.sampling_rate = *sflow_cfg->sampling;
813 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
814 if (sflow_cfg->polling) {
815 oso.polling_interval = *sflow_cfg->polling;
818 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
819 if (sflow_cfg->header) {
820 oso.header_len = *sflow_cfg->header;
823 oso.sub_id = sflow_bridge_number++;
824 oso.agent_device = sflow_cfg->agent;
826 oso.control_ip = NULL;
827 n_controllers = bridge_get_controllers(br, &controllers);
828 for (i = 0; i < n_controllers; i++) {
829 if (controllers[i]->local_ip) {
830 oso.control_ip = controllers[i]->local_ip;
834 ofproto_set_sflow(br->ofproto, &oso);
836 sset_destroy(&oso.targets);
838 ofproto_set_sflow(br->ofproto, NULL);
841 /* Update the controller and related settings. It would be more
842 * straightforward to call this from bridge_reconfigure_one(), but we
843 * can't do it there for two reasons. First, and most importantly, at
844 * that point we don't know the dp_ifidx of any interfaces that have
845 * been added to the bridge (because we haven't actually added them to
846 * the datapath). Second, at that point we haven't set the datapath ID
847 * yet; when a controller is configured, resetting the datapath ID will
848 * immediately disconnect from the controller, so it's better to set
849 * the datapath ID before the controller. */
850 bridge_reconfigure_remotes(br, managers, n_managers);
852 HMAP_FOR_EACH (br, node, &all_bridges) {
855 br->has_bonded_ports = false;
856 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
859 port_reconfigure_lacp(port);
860 port_reconfigure_bond(port);
862 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
863 iface_update_qos(iface, port->cfg->qos);
864 netdev_set_policing(iface->netdev,
865 iface->cfg->ingress_policing_rate,
866 iface->cfg->ingress_policing_burst);
867 iface_set_mac(iface);
872 /* Some reconfiguration operations require the bridge to have been run at
874 HMAP_FOR_EACH (br, node, &all_bridges) {
879 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
880 iface_update_cfm(iface);
886 /* ovs-vswitchd has completed initialization, so allow the process that
887 * forked us to exit successfully. */
888 daemonize_complete();
892 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
894 const struct port *port = port_lookup(br, name);
895 return port && port_is_bond_fake_iface(port);
899 port_is_bond_fake_iface(const struct port *port)
901 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
905 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
906 const struct ovsdb_idl_column *column,
909 const struct ovsdb_datum *datum;
910 union ovsdb_atom atom;
913 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
914 atom.string = (char *) key;
915 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
916 return idx == UINT_MAX ? NULL : datum->values[idx].string;
920 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
922 return get_ovsrec_key_value(&br_cfg->header_,
923 &ovsrec_bridge_col_other_config, key);
927 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
928 struct iface **hw_addr_iface)
934 *hw_addr_iface = NULL;
936 /* Did the user request a particular MAC? */
937 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
938 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
939 if (eth_addr_is_multicast(ea)) {
940 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
941 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
942 } else if (eth_addr_is_zero(ea)) {
943 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
949 /* Otherwise choose the minimum non-local MAC address among all of the
951 memset(ea, 0xff, ETH_ADDR_LEN);
952 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
953 uint8_t iface_ea[ETH_ADDR_LEN];
954 struct iface *candidate;
957 /* Mirror output ports don't participate. */
958 if (port->is_mirror_output_port) {
962 /* Choose the MAC address to represent the port. */
964 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
965 /* Find the interface with this Ethernet address (if any) so that
966 * we can provide the correct devname to the caller. */
967 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
968 uint8_t candidate_ea[ETH_ADDR_LEN];
969 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
970 && eth_addr_equals(iface_ea, candidate_ea)) {
975 /* Choose the interface whose MAC address will represent the port.
976 * The Linux kernel bonding code always chooses the MAC address of
977 * the first slave added to a bond, and the Fedora networking
978 * scripts always add slaves to a bond in alphabetical order, so
979 * for compatibility we choose the interface with the name that is
980 * first in alphabetical order. */
981 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
982 if (!iface || strcmp(candidate->name, iface->name) < 0) {
987 /* The local port doesn't count (since we're trying to choose its
988 * MAC address anyway). */
989 if (iface->dp_ifidx == ODPP_LOCAL) {
994 error = netdev_get_etheraddr(iface->netdev, iface_ea);
996 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
997 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
998 iface->name, strerror(error));
1003 /* Compare against our current choice. */
1004 if (!eth_addr_is_multicast(iface_ea) &&
1005 !eth_addr_is_local(iface_ea) &&
1006 !eth_addr_is_reserved(iface_ea) &&
1007 !eth_addr_is_zero(iface_ea) &&
1008 eth_addr_compare_3way(iface_ea, ea) < 0)
1010 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1011 *hw_addr_iface = iface;
1014 if (eth_addr_is_multicast(ea)) {
1015 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1016 *hw_addr_iface = NULL;
1017 VLOG_WARN("bridge %s: using default bridge Ethernet "
1018 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1020 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1021 br->name, ETH_ADDR_ARGS(ea));
1025 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1026 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1027 * an interface on 'br', then that interface must be passed in as
1028 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1029 * 'hw_addr_iface' must be passed in as a null pointer. */
1031 bridge_pick_datapath_id(struct bridge *br,
1032 const uint8_t bridge_ea[ETH_ADDR_LEN],
1033 struct iface *hw_addr_iface)
1036 * The procedure for choosing a bridge MAC address will, in the most
1037 * ordinary case, also choose a unique MAC that we can use as a datapath
1038 * ID. In some special cases, though, multiple bridges will end up with
1039 * the same MAC address. This is OK for the bridges, but it will confuse
1040 * the OpenFlow controller, because each datapath needs a unique datapath
1043 * Datapath IDs must be unique. It is also very desirable that they be
1044 * stable from one run to the next, so that policy set on a datapath
1047 const char *datapath_id;
1050 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1051 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1055 if (hw_addr_iface) {
1057 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1059 * A bridge whose MAC address is taken from a VLAN network device
1060 * (that is, a network device created with vconfig(8) or similar
1061 * tool) will have the same MAC address as a bridge on the VLAN
1062 * device's physical network device.
1064 * Handle this case by hashing the physical network device MAC
1065 * along with the VLAN identifier.
1067 uint8_t buf[ETH_ADDR_LEN + 2];
1068 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1069 buf[ETH_ADDR_LEN] = vlan >> 8;
1070 buf[ETH_ADDR_LEN + 1] = vlan;
1071 return dpid_from_hash(buf, sizeof buf);
1074 * Assume that this bridge's MAC address is unique, since it
1075 * doesn't fit any of the cases we handle specially.
1080 * A purely internal bridge, that is, one that has no non-virtual
1081 * network devices on it at all, is more difficult because it has no
1082 * natural unique identifier at all.
1084 * When the host is a XenServer, we handle this case by hashing the
1085 * host's UUID with the name of the bridge. Names of bridges are
1086 * persistent across XenServer reboots, although they can be reused if
1087 * an internal network is destroyed and then a new one is later
1088 * created, so this is fairly effective.
1090 * When the host is not a XenServer, we punt by using a random MAC
1091 * address on each run.
1093 const char *host_uuid = xenserver_get_host_uuid();
1095 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1096 dpid = dpid_from_hash(combined, strlen(combined));
1102 return eth_addr_to_uint64(bridge_ea);
1106 dpid_from_hash(const void *data, size_t n)
1108 uint8_t hash[SHA1_DIGEST_SIZE];
1110 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1111 sha1_bytes(data, n, hash);
1112 eth_addr_mark_random(hash);
1113 return eth_addr_to_uint64(hash);
1117 iface_refresh_status(struct iface *iface)
1121 enum netdev_flags flags;
1128 if (iface_is_synthetic(iface)) {
1134 if (!netdev_get_status(iface->netdev, &sh)) {
1136 char **keys, **values;
1138 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1139 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1144 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1147 shash_destroy_free_data(&sh);
1149 error = netdev_get_flags(iface->netdev, &flags);
1151 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1154 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1157 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1159 ovsrec_interface_set_duplex(iface->cfg,
1160 netdev_features_is_full_duplex(current)
1162 /* warning: uint64_t -> int64_t conversion */
1163 bps = netdev_features_to_bps(current);
1164 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1167 ovsrec_interface_set_duplex(iface->cfg, NULL);
1168 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1172 ovsrec_interface_set_link_state(iface->cfg,
1173 iface_get_carrier(iface) ? "up" : "down");
1175 error = netdev_get_mtu(iface->netdev, &mtu);
1176 if (!error && mtu != INT_MAX) {
1178 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1181 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1185 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1186 * changed, false otherwise. */
1188 iface_refresh_cfm_stats(struct iface *iface)
1190 const struct ovsrec_monitor *mon;
1191 const struct cfm *cfm;
1192 bool changed = false;
1195 mon = iface->cfg->monitor;
1196 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1202 for (i = 0; i < mon->n_remote_mps; i++) {
1203 const struct ovsrec_maintenance_point *mp;
1204 const struct remote_mp *rmp;
1206 mp = mon->remote_mps[i];
1207 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1209 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1210 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1215 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1216 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1224 iface_refresh_lacp_stats(struct iface *iface)
1226 bool *db_current = iface->cfg->lacp_current;
1227 bool changed = false;
1229 if (iface->port->lacp) {
1230 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1232 if (!db_current || *db_current != current) {
1234 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1236 } else if (db_current) {
1238 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1245 iface_refresh_stats(struct iface *iface)
1251 static const struct iface_stat iface_stats[] = {
1252 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1253 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1254 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1255 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1256 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1257 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1258 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1259 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1260 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1261 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1262 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1263 { "collisions", offsetof(struct netdev_stats, collisions) },
1265 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1266 const struct iface_stat *s;
1268 char *keys[N_STATS];
1269 int64_t values[N_STATS];
1272 struct netdev_stats stats;
1274 if (iface_is_synthetic(iface)) {
1278 /* Intentionally ignore return value, since errors will set 'stats' to
1279 * all-1s, and we will deal with that correctly below. */
1280 netdev_get_stats(iface->netdev, &stats);
1283 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1284 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1285 if (value != UINT64_MAX) {
1292 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1296 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1298 struct ovsdb_datum datum;
1302 get_system_stats(&stats);
1304 ovsdb_datum_from_shash(&datum, &stats);
1305 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1309 static inline const char *
1310 nx_role_to_str(enum nx_role role)
1315 case NX_ROLE_MASTER:
1320 return "*** INVALID ROLE ***";
1325 bridge_refresh_controller_status(const struct bridge *br)
1328 const struct ovsrec_controller *cfg;
1330 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1332 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1333 struct ofproto_controller_info *cinfo =
1334 shash_find_data(&info, cfg->target);
1337 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1338 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1339 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1340 (char **) cinfo->pairs.values,
1343 ovsrec_controller_set_is_connected(cfg, false);
1344 ovsrec_controller_set_role(cfg, NULL);
1345 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1349 ofproto_free_ofproto_controller_info(&info);
1355 const struct ovsrec_open_vswitch *cfg;
1357 bool datapath_destroyed;
1358 bool database_changed;
1361 /* Let each bridge do the work that it needs to do. */
1362 datapath_destroyed = false;
1363 HMAP_FOR_EACH (br, node, &all_bridges) {
1364 int error = bridge_run_one(br);
1366 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1367 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1368 "forcing reconfiguration", br->name);
1369 datapath_destroyed = true;
1373 /* (Re)configure if necessary. */
1374 database_changed = ovsdb_idl_run(idl);
1375 cfg = ovsrec_open_vswitch_first(idl);
1377 /* Re-configure SSL. We do this on every trip through the main loop,
1378 * instead of just when the database changes, because the contents of the
1379 * key and certificate files can change without the database changing.
1381 * We do this before bridge_reconfigure() because that function might
1382 * initiate SSL connections and thus requires SSL to be configured. */
1383 if (cfg && cfg->ssl) {
1384 const struct ovsrec_ssl *ssl = cfg->ssl;
1386 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1387 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1390 if (database_changed || datapath_destroyed) {
1392 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1394 bridge_configure_once(cfg);
1395 bridge_reconfigure(cfg);
1397 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1398 ovsdb_idl_txn_commit(txn);
1399 ovsdb_idl_txn_destroy(txn); /* XXX */
1401 /* We still need to reconfigure to avoid dangling pointers to
1402 * now-destroyed ovsrec structures inside bridge data. */
1403 static const struct ovsrec_open_vswitch null_cfg;
1405 bridge_reconfigure(&null_cfg);
1409 /* Refresh system and interface stats if necessary. */
1410 if (time_msec() >= stats_timer) {
1412 struct ovsdb_idl_txn *txn;
1414 txn = ovsdb_idl_txn_create(idl);
1415 HMAP_FOR_EACH (br, node, &all_bridges) {
1418 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1419 struct iface *iface;
1421 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1422 iface_refresh_stats(iface);
1423 iface_refresh_status(iface);
1426 bridge_refresh_controller_status(br);
1428 refresh_system_stats(cfg);
1429 ovsdb_idl_txn_commit(txn);
1430 ovsdb_idl_txn_destroy(txn); /* XXX */
1433 stats_timer = time_msec() + STATS_INTERVAL;
1436 if (time_msec() >= db_limiter) {
1437 struct ovsdb_idl_txn *txn;
1438 bool changed = false;
1440 txn = ovsdb_idl_txn_create(idl);
1441 HMAP_FOR_EACH (br, node, &all_bridges) {
1444 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1445 struct iface *iface;
1447 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1448 changed = iface_refresh_cfm_stats(iface) || changed;
1449 changed = iface_refresh_lacp_stats(iface) || changed;
1455 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1458 ovsdb_idl_txn_commit(txn);
1459 ovsdb_idl_txn_destroy(txn);
1468 HMAP_FOR_EACH (br, node, &all_bridges) {
1471 ofproto_wait(br->ofproto);
1472 mac_learning_wait(br->ml);
1473 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1477 ovsdb_idl_wait(idl);
1478 poll_timer_wait_until(stats_timer);
1480 if (db_limiter > time_msec()) {
1481 poll_timer_wait_until(db_limiter);
1485 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1486 * configuration changes. */
1488 bridge_flush(struct bridge *br)
1490 COVERAGE_INC(bridge_flush);
1494 /* Bridge unixctl user interface functions. */
1496 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1497 const char *args, void *aux OVS_UNUSED)
1499 struct ds ds = DS_EMPTY_INITIALIZER;
1500 const struct bridge *br;
1501 const struct mac_entry *e;
1503 br = bridge_lookup(args);
1505 unixctl_command_reply(conn, 501, "no such bridge");
1509 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1510 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1511 struct port *port = e->port.p;
1512 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1513 port_get_an_iface(port)->dp_ifidx,
1514 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1516 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1520 /* CFM unixctl user interface functions. */
1522 cfm_unixctl_show(struct unixctl_conn *conn,
1523 const char *args, void *aux OVS_UNUSED)
1525 struct ds ds = DS_EMPTY_INITIALIZER;
1526 struct iface *iface;
1527 const struct cfm *cfm;
1529 iface = iface_find(args);
1531 unixctl_command_reply(conn, 501, "no such interface");
1535 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1538 unixctl_command_reply(conn, 501, "CFM not enabled");
1542 cfm_dump_ds(cfm, &ds);
1543 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1547 /* QoS unixctl user interface functions. */
1549 struct qos_unixctl_show_cbdata {
1551 struct iface *iface;
1555 qos_unixctl_show_cb(unsigned int queue_id,
1556 const struct shash *details,
1559 struct qos_unixctl_show_cbdata *data = aux;
1560 struct ds *ds = data->ds;
1561 struct iface *iface = data->iface;
1562 struct netdev_queue_stats stats;
1563 struct shash_node *node;
1566 ds_put_cstr(ds, "\n");
1568 ds_put_format(ds, "Queue %u:\n", queue_id);
1570 ds_put_cstr(ds, "Default:\n");
1573 SHASH_FOR_EACH (node, details) {
1574 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1577 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1579 if (stats.tx_packets != UINT64_MAX) {
1580 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1583 if (stats.tx_bytes != UINT64_MAX) {
1584 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1587 if (stats.tx_errors != UINT64_MAX) {
1588 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1591 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1592 queue_id, strerror(error));
1597 qos_unixctl_show(struct unixctl_conn *conn,
1598 const char *args, void *aux OVS_UNUSED)
1600 struct ds ds = DS_EMPTY_INITIALIZER;
1601 struct shash sh = SHASH_INITIALIZER(&sh);
1602 struct iface *iface;
1604 struct shash_node *node;
1605 struct qos_unixctl_show_cbdata data;
1608 iface = iface_find(args);
1610 unixctl_command_reply(conn, 501, "no such interface");
1614 netdev_get_qos(iface->netdev, &type, &sh);
1616 if (*type != '\0') {
1617 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1619 SHASH_FOR_EACH (node, &sh) {
1620 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1625 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1628 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1630 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1632 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1633 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1636 shash_destroy_free_data(&sh);
1640 /* Bridge reconfiguration functions. */
1641 static struct bridge *
1642 bridge_create(const struct ovsrec_bridge *br_cfg)
1647 assert(!bridge_lookup(br_cfg->name));
1648 br = xzalloc(sizeof *br);
1650 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1653 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1659 br->name = xstrdup(br_cfg->name);
1661 br->ml = mac_learning_create();
1662 eth_addr_nicira_random(br->default_ea);
1664 hmap_init(&br->ports);
1665 hmap_init(&br->ifaces);
1666 hmap_init(&br->iface_by_name);
1670 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1672 VLOG_INFO("bridge %s: created", br->name);
1678 bridge_destroy(struct bridge *br)
1681 struct port *port, *next;
1684 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1687 for (i = 0; i < MAX_MIRRORS; i++) {
1688 mirror_destroy(br->mirrors[i]);
1690 hmap_remove(&all_bridges, &br->node);
1691 ofproto_destroy_and_delete(br->ofproto);
1692 mac_learning_destroy(br->ml);
1693 hmap_destroy(&br->ifaces);
1694 hmap_destroy(&br->ports);
1695 hmap_destroy(&br->iface_by_name);
1696 free(br->synth_local_iface.type);
1702 static struct bridge *
1703 bridge_lookup(const char *name)
1707 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1708 if (!strcmp(br->name, name)) {
1715 /* Handle requests for a listing of all flows known by the OpenFlow
1716 * stack, including those normally hidden. */
1718 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1719 const char *args, void *aux OVS_UNUSED)
1724 br = bridge_lookup(args);
1726 unixctl_command_reply(conn, 501, "Unknown bridge");
1731 ofproto_get_all_flows(br->ofproto, &results);
1733 unixctl_command_reply(conn, 200, ds_cstr(&results));
1734 ds_destroy(&results);
1737 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1738 * connections and reconnect. If BRIDGE is not specified, then all bridges
1739 * drop their controller connections and reconnect. */
1741 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1742 const char *args, void *aux OVS_UNUSED)
1745 if (args[0] != '\0') {
1746 br = bridge_lookup(args);
1748 unixctl_command_reply(conn, 501, "Unknown bridge");
1751 ofproto_reconnect_controllers(br->ofproto);
1753 HMAP_FOR_EACH (br, node, &all_bridges) {
1754 ofproto_reconnect_controllers(br->ofproto);
1757 unixctl_command_reply(conn, 200, NULL);
1761 bridge_run_one(struct bridge *br)
1766 error = ofproto_run1(br->ofproto);
1771 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1773 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1777 error = ofproto_run2(br->ofproto, br->flush);
1784 bridge_get_controllers(const struct bridge *br,
1785 struct ovsrec_controller ***controllersp)
1787 struct ovsrec_controller **controllers;
1788 size_t n_controllers;
1790 controllers = br->cfg->controller;
1791 n_controllers = br->cfg->n_controller;
1793 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1799 *controllersp = controllers;
1801 return n_controllers;
1805 bridge_reconfigure_one(struct bridge *br)
1807 enum ofproto_fail_mode fail_mode;
1808 struct port *port, *next;
1809 struct shash_node *node;
1810 struct shash new_ports;
1813 /* Collect new ports. */
1814 shash_init(&new_ports);
1815 for (i = 0; i < br->cfg->n_ports; i++) {
1816 const char *name = br->cfg->ports[i]->name;
1817 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1818 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1822 if (bridge_get_controllers(br, NULL)
1823 && !shash_find(&new_ports, br->name)) {
1824 struct ofproto_port local_port;
1828 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1829 br->name, br->name);
1831 error = ofproto_port_query_by_name(br->ofproto, br->name, &local_port);
1832 type = xstrdup(error ? "internal" : local_port.type);
1833 ofproto_port_destroy(&local_port);
1835 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1836 br->synth_local_port.n_interfaces = 1;
1837 br->synth_local_port.name = br->name;
1839 br->synth_local_iface.name = br->name;
1840 free(br->synth_local_iface.type);
1841 br->synth_local_iface.type = type;
1843 br->synth_local_ifacep = &br->synth_local_iface;
1845 shash_add(&new_ports, br->name, &br->synth_local_port);
1848 /* Get rid of deleted ports.
1849 * Get rid of deleted interfaces on ports that still exist. */
1850 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1851 const struct ovsrec_port *port_cfg;
1853 port_cfg = shash_find_data(&new_ports, port->name);
1857 port_del_ifaces(port, port_cfg);
1861 /* Create new ports.
1862 * Add new interfaces to existing ports.
1863 * Reconfigure existing ports. */
1864 SHASH_FOR_EACH (node, &new_ports) {
1865 struct port *port = port_lookup(br, node->name);
1867 port = port_create(br, node->name);
1870 port_reconfigure(port, node->data);
1871 if (list_is_empty(&port->ifaces)) {
1872 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1873 br->name, port->name);
1877 shash_destroy(&new_ports);
1879 /* Set the fail-mode */
1880 fail_mode = !br->cfg->fail_mode
1881 || !strcmp(br->cfg->fail_mode, "standalone")
1882 ? OFPROTO_FAIL_STANDALONE
1883 : OFPROTO_FAIL_SECURE;
1884 ofproto_set_fail_mode(br->ofproto, fail_mode);
1886 /* Configure OpenFlow controller connection snooping. */
1887 if (!ofproto_has_snoops(br->ofproto)) {
1891 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1892 ovs_rundir(), br->name));
1893 ofproto_set_snoops(br->ofproto, &snoops);
1894 sset_destroy(&snoops);
1897 mirror_reconfigure(br);
1900 /* Initializes 'oc' appropriately as a management service controller for
1903 * The caller must free oc->target when it is no longer needed. */
1905 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1906 struct ofproto_controller *oc)
1908 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1909 oc->max_backoff = 0;
1910 oc->probe_interval = 60;
1911 oc->band = OFPROTO_OUT_OF_BAND;
1913 oc->burst_limit = 0;
1916 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1918 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1919 struct ofproto_controller *oc)
1921 oc->target = c->target;
1922 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1923 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1924 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1925 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1926 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1927 oc->burst_limit = (c->controller_burst_limit
1928 ? *c->controller_burst_limit : 0);
1931 /* Configures the IP stack for 'br''s local interface properly according to the
1932 * configuration in 'c'. */
1934 bridge_configure_local_iface_netdev(struct bridge *br,
1935 struct ovsrec_controller *c)
1937 struct netdev *netdev;
1938 struct in_addr mask, gateway;
1940 struct iface *local_iface;
1943 /* If there's no local interface or no IP address, give up. */
1944 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1945 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1949 /* Bring up the local interface. */
1950 netdev = local_iface->netdev;
1951 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1953 /* Configure the IP address and netmask. */
1954 if (!c->local_netmask
1955 || !inet_aton(c->local_netmask, &mask)
1957 mask.s_addr = guess_netmask(ip.s_addr);
1959 if (!netdev_set_in4(netdev, ip, mask)) {
1960 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1961 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1964 /* Configure the default gateway. */
1965 if (c->local_gateway
1966 && inet_aton(c->local_gateway, &gateway)
1967 && gateway.s_addr) {
1968 if (!netdev_add_router(netdev, gateway)) {
1969 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1970 br->name, IP_ARGS(&gateway.s_addr));
1976 bridge_reconfigure_remotes(struct bridge *br,
1977 const struct sockaddr_in *managers,
1980 const char *disable_ib_str, *queue_id_str;
1981 bool disable_in_band = false;
1984 struct ovsrec_controller **controllers;
1985 size_t n_controllers;
1987 struct ofproto_controller *ocs;
1991 /* Check if we should disable in-band control on this bridge. */
1992 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1993 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1994 disable_in_band = true;
1997 /* Set OpenFlow queue ID for in-band control. */
1998 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1999 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2000 ofproto_set_in_band_queue(br->ofproto, queue_id);
2002 if (disable_in_band) {
2003 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2005 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2008 n_controllers = bridge_get_controllers(br, &controllers);
2010 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2013 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2014 for (i = 0; i < n_controllers; i++) {
2015 struct ovsrec_controller *c = controllers[i];
2017 if (!strncmp(c->target, "punix:", 6)
2018 || !strncmp(c->target, "unix:", 5)) {
2019 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2021 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2022 * domain sockets and overwriting arbitrary local files. */
2023 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
2024 "controller \"%s\" due to possibility for remote "
2025 "exploit", br->name, c->target);
2029 bridge_configure_local_iface_netdev(br, c);
2030 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2031 if (disable_in_band) {
2032 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2037 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2038 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2042 /* Bridge packet processing functions. */
2045 set_dst(struct dst *dst, const struct flow *flow,
2046 const struct port *in_port, const struct port *out_port,
2049 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2050 : in_port->vlan >= 0 ? in_port->vlan
2051 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2052 : vlan_tci_to_vid(flow->vlan_tci));
2054 dst->iface = (!out_port->bond
2055 ? port_get_an_iface(out_port)
2056 : bond_choose_output_slave(out_port->bond, flow,
2059 return dst->iface != NULL;
2063 mirror_mask_ffs(mirror_mask_t mask)
2065 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2070 dst_set_init(struct dst_set *set)
2072 set->dsts = set->builtin;
2074 set->allocated = ARRAY_SIZE(set->builtin);
2078 dst_set_add(struct dst_set *set, const struct dst *dst)
2080 if (set->n >= set->allocated) {
2081 size_t new_allocated;
2082 struct dst *new_dsts;
2084 new_allocated = set->allocated * 2;
2085 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2086 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2090 set->dsts = new_dsts;
2091 set->allocated = new_allocated;
2093 set->dsts[set->n++] = *dst;
2097 dst_set_free(struct dst_set *set)
2099 if (set->dsts != set->builtin) {
2105 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2108 for (i = 0; i < set->n; i++) {
2109 if (set->dsts[i].vlan == test->vlan
2110 && set->dsts[i].iface == test->iface) {
2118 port_trunks_vlan(const struct port *port, uint16_t vlan)
2120 return (port->vlan < 0 || vlan_bitmap_contains(port->trunks, vlan));
2124 port_includes_vlan(const struct port *port, uint16_t vlan)
2126 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2130 port_is_floodable(const struct port *port)
2132 struct iface *iface;
2134 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2135 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2143 /* Returns an arbitrary interface within 'port'. */
2144 static struct iface *
2145 port_get_an_iface(const struct port *port)
2147 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2151 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2152 const struct port *in_port, const struct port *out_port,
2153 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2157 if (out_port == FLOOD_PORT) {
2160 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2162 && port_is_floodable(port)
2163 && port_includes_vlan(port, vlan)
2164 && !port->is_mirror_output_port
2165 && set_dst(&dst, flow, in_port, port, tags)) {
2166 dst_set_add(set, &dst);
2169 *nf_output_iface = NF_OUT_FLOOD;
2170 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2171 dst_set_add(set, &dst);
2172 *nf_output_iface = dst.iface->dp_ifidx;
2177 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2178 uint16_t vlan, const struct port *in_port,
2179 struct dst_set *set, tag_type *tags)
2181 mirror_mask_t mirrors;
2185 mirrors = in_port->src_mirrors;
2186 for (i = 0; i < set->n; i++) {
2187 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2194 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2195 if (flow_vlan == 0) {
2196 flow_vlan = OFP_VLAN_NONE;
2200 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2201 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2205 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2206 && !dst_is_duplicate(set, &dst)) {
2207 dst_set_add(set, &dst);
2212 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2213 if (port_includes_vlan(port, m->out_vlan)
2214 && set_dst(&dst, flow, in_port, port, tags))
2216 if (port->vlan < 0) {
2217 dst.vlan = m->out_vlan;
2219 if (dst_is_duplicate(set, &dst)) {
2223 /* Use the vlan tag on the original flow instead of
2224 * the one passed in the vlan parameter. This ensures
2225 * that we compare the vlan from before any implicit
2226 * tagging tags place. This is necessary because
2227 * dst->vlan is the final vlan, after removing implicit
2229 if (port == in_port && dst.vlan == flow_vlan) {
2230 /* Don't send out input port on same VLAN. */
2233 dst_set_add(set, &dst);
2238 mirrors &= mirrors - 1;
2243 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2244 const struct port *in_port, const struct port *out_port,
2245 tag_type *tags, struct ofpbuf *actions,
2246 uint16_t *nf_output_iface)
2248 uint16_t initial_vlan, cur_vlan;
2249 const struct dst *dst;
2253 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2255 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2257 /* Output all the packets we can without having to change the VLAN. */
2258 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2259 if (initial_vlan == 0) {
2260 initial_vlan = OFP_VLAN_NONE;
2262 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2263 if (dst->vlan != initial_vlan) {
2266 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2269 /* Then output the rest. */
2270 cur_vlan = initial_vlan;
2271 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2272 if (dst->vlan == initial_vlan) {
2275 if (dst->vlan != cur_vlan) {
2276 if (dst->vlan == OFP_VLAN_NONE) {
2277 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2280 tci = htons(dst->vlan & VLAN_VID_MASK);
2281 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2282 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2284 cur_vlan = dst->vlan;
2286 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2292 /* Returns the effective vlan of a packet, taking into account both the
2293 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2294 * the packet is untagged and -1 indicates it has an invalid header and
2295 * should be dropped. */
2296 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2297 struct port *in_port, bool have_packet)
2299 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2300 if (in_port->vlan >= 0) {
2303 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2304 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2305 "packet received on port %s configured with "
2306 "implicit VLAN %"PRIu16,
2307 br->name, vlan, in_port->name, in_port->vlan);
2311 vlan = in_port->vlan;
2313 if (!port_includes_vlan(in_port, vlan)) {
2315 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2316 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2317 "packet received on port %s not configured for "
2319 br->name, vlan, in_port->name, vlan);
2328 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2329 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2330 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2332 is_gratuitous_arp(const struct flow *flow)
2334 return (flow->dl_type == htons(ETH_TYPE_ARP)
2335 && eth_addr_is_broadcast(flow->dl_dst)
2336 && (flow->nw_proto == ARP_OP_REPLY
2337 || (flow->nw_proto == ARP_OP_REQUEST
2338 && flow->nw_src == flow->nw_dst)));
2342 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2343 struct port *in_port)
2345 struct mac_entry *mac;
2347 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2351 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2352 if (is_gratuitous_arp(flow)) {
2353 /* We don't want to learn from gratuitous ARP packets that are
2354 * reflected back over bond slaves so we lock the learning table. */
2355 if (!in_port->bond) {
2356 mac_entry_set_grat_arp_lock(mac);
2357 } else if (mac_entry_is_grat_arp_locked(mac)) {
2362 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2363 /* The log messages here could actually be useful in debugging,
2364 * so keep the rate limit relatively high. */
2365 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2366 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2367 "on port %s in VLAN %d",
2368 br->name, ETH_ADDR_ARGS(flow->dl_src),
2369 in_port->name, vlan);
2371 mac->port.p = in_port;
2372 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2376 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2377 * dropped. Returns true if they may be forwarded, false if they should be
2380 * If 'have_packet' is true, it indicates that the caller is processing a
2381 * received packet. If 'have_packet' is false, then the caller is just
2382 * revalidating an existing flow because configuration has changed. Either
2383 * way, 'have_packet' only affects logging (there is no point in logging errors
2384 * during revalidation).
2386 * Sets '*in_portp' to the input port. This will be a null pointer if
2387 * flow->in_port does not designate a known input port (in which case
2388 * is_admissible() returns false).
2390 * When returning true, sets '*vlanp' to the effective VLAN of the input
2391 * packet, as returned by flow_get_vlan().
2393 * May also add tags to '*tags', although the current implementation only does
2394 * so in one special case.
2397 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2398 tag_type *tags, int *vlanp, struct port **in_portp)
2400 struct iface *in_iface;
2401 struct port *in_port;
2404 /* Find the interface and port structure for the received packet. */
2405 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2407 /* No interface? Something fishy... */
2409 /* Odd. A few possible reasons here:
2411 * - We deleted an interface but there are still a few packets
2412 * queued up from it.
2414 * - Someone externally added an interface (e.g. with "ovs-dpctl
2415 * add-if") that we don't know about.
2417 * - Packet arrived on the local port but the local port is not
2418 * one of our bridge ports.
2420 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2422 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2423 "interface %"PRIu16, br->name, flow->in_port);
2429 *in_portp = in_port = in_iface->port;
2430 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2435 /* Drop frames for reserved multicast addresses. */
2436 if (eth_addr_is_reserved(flow->dl_dst)) {
2440 /* Drop frames on ports reserved for mirroring. */
2441 if (in_port->is_mirror_output_port) {
2443 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2444 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2445 "%s, which is reserved exclusively for mirroring",
2446 br->name, in_port->name);
2451 if (in_port->bond) {
2452 struct mac_entry *mac;
2454 switch (bond_check_admissibility(in_port->bond, in_iface,
2455 flow->dl_dst, tags)) {
2462 case BV_DROP_IF_MOVED:
2463 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2464 if (mac && mac->port.p != in_port &&
2465 (!is_gratuitous_arp(flow)
2466 || mac_entry_is_grat_arp_locked(mac))) {
2476 /* If the composed actions may be applied to any packet in the given 'flow',
2477 * returns true. Otherwise, the actions should only be applied to 'packet', or
2478 * not at all, if 'packet' was NULL. */
2480 process_flow(struct bridge *br, const struct flow *flow,
2481 const struct ofpbuf *packet, struct ofpbuf *actions,
2482 tag_type *tags, uint16_t *nf_output_iface)
2484 struct port *in_port;
2485 struct port *out_port;
2486 struct mac_entry *mac;
2489 /* Check whether we should drop packets in this flow. */
2490 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2495 /* Learn source MAC (but don't try to learn from revalidation). */
2497 update_learning_table(br, flow, vlan, in_port);
2500 /* Determine output port. */
2501 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2503 out_port = mac->port.p;
2504 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2505 /* If we are revalidating but don't have a learning entry then
2506 * eject the flow. Installing a flow that floods packets opens
2507 * up a window of time where we could learn from a packet reflected
2508 * on a bond and blackhole packets before the learning table is
2509 * updated to reflect the correct port. */
2512 out_port = FLOOD_PORT;
2515 /* Don't send packets out their input ports. */
2516 if (in_port == out_port) {
2522 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2530 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2531 struct ofpbuf *actions, tag_type *tags,
2532 uint16_t *nf_output_iface, void *br_)
2534 struct bridge *br = br_;
2536 COVERAGE_INC(bridge_process_flow);
2537 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2541 bridge_special_ofhook_cb(const struct flow *flow,
2542 const struct ofpbuf *packet, void *br_)
2544 struct iface *iface;
2545 struct bridge *br = br_;
2547 iface = iface_from_dp_ifidx(br, flow->in_port);
2549 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2550 if (iface && iface->port->lacp && packet) {
2551 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2553 lacp_process_pdu(iface->port->lacp, iface, pdu);
2563 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2564 const struct nlattr *actions,
2566 uint64_t n_bytes, void *br_)
2568 struct bridge *br = br_;
2569 const struct nlattr *a;
2570 struct port *in_port;
2575 /* Feed information from the active flows back into the learning table to
2576 * ensure that table is always in sync with what is actually flowing
2577 * through the datapath.
2579 * We test that 'tags' is nonzero to ensure that only flows that include an
2580 * OFPP_NORMAL action are used for learning. This works because
2581 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2582 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2583 update_learning_table(br, flow, vlan, in_port);
2586 /* Account for bond slave utilization. */
2587 if (!br->has_bonded_ports) {
2590 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2591 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2592 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2593 if (out_port && out_port->bond) {
2594 uint16_t vlan = (flow->vlan_tci
2595 ? vlan_tci_to_vid(flow->vlan_tci)
2597 bond_account(out_port->bond, flow, vlan, n_bytes);
2604 bridge_account_checkpoint_ofhook_cb(void *br_)
2606 struct bridge *br = br_;
2609 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2611 bond_rebalance(port->bond,
2612 ofproto_get_revalidate_set(br->ofproto));
2618 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2619 tag_type *tags, void *br_)
2621 struct bridge *br = br_;
2622 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2623 struct port *port = port_from_dp_ifidx(br, odp_port);
2628 } else if (list_is_short(&port->ifaces)) {
2631 struct iface *iface;
2633 /* Autopath does not support VLAN hashing. */
2634 iface = bond_choose_output_slave(port->bond, flow,
2635 OFP_VLAN_NONE, tags);
2636 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2639 return odp_port_to_ofp_port(ret);
2642 static struct ofhooks bridge_ofhooks = {
2643 bridge_normal_ofhook_cb,
2644 bridge_special_ofhook_cb,
2645 bridge_account_flow_ofhook_cb,
2646 bridge_account_checkpoint_ofhook_cb,
2647 bridge_autopath_ofhook_cb,
2650 /* Port functions. */
2653 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2655 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2656 struct iface *iface = iface_;
2657 uint8_t ea[ETH_ADDR_LEN];
2660 error = netdev_get_etheraddr(iface->netdev, ea);
2662 struct lacp_pdu *packet_pdu;
2663 struct ofpbuf packet;
2665 ofpbuf_init(&packet, 0);
2666 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2667 sizeof *packet_pdu);
2669 error = netdev_send(iface->netdev, &packet);
2671 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2672 "(%s)", iface->port->name, iface->name,
2675 ofpbuf_uninit(&packet);
2677 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2678 "%s (%s)", iface->port->name, iface->name,
2684 port_run(struct port *port)
2687 lacp_run(port->lacp, lacp_send_pdu_cb);
2691 struct iface *iface;
2693 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2694 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2695 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2698 bond_run(port->bond,
2699 ofproto_get_revalidate_set(port->bridge->ofproto),
2700 lacp_negotiated(port->lacp));
2701 if (bond_should_send_learning_packets(port->bond)) {
2702 port_send_learning_packets(port);
2708 port_wait(struct port *port)
2711 lacp_wait(port->lacp);
2715 bond_wait(port->bond);
2719 static struct port *
2720 port_create(struct bridge *br, const char *name)
2724 port = xzalloc(sizeof *port);
2727 port->trunks = NULL;
2728 port->name = xstrdup(name);
2729 list_init(&port->ifaces);
2731 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2733 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2740 get_port_other_config(const struct ovsrec_port *port, const char *key,
2741 const char *default_value)
2745 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2747 return value ? value : default_value;
2751 get_interface_other_config(const struct ovsrec_interface *iface,
2752 const char *key, const char *default_value)
2756 value = get_ovsrec_key_value(&iface->header_,
2757 &ovsrec_interface_col_other_config, key);
2758 return value ? value : default_value;
2762 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2764 struct iface *iface, *next;
2765 struct sset new_ifaces;
2768 /* Collect list of new interfaces. */
2769 sset_init(&new_ifaces);
2770 for (i = 0; i < cfg->n_interfaces; i++) {
2771 const char *name = cfg->interfaces[i]->name;
2772 sset_add(&new_ifaces, name);
2775 /* Get rid of deleted interfaces. */
2776 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2777 if (!sset_contains(&new_ifaces, iface->name)) {
2778 iface_destroy(iface);
2782 sset_destroy(&new_ifaces);
2785 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2786 * to revalidate every flow. */
2788 port_flush_macs(struct port *port)
2790 struct bridge *br = port->bridge;
2791 struct mac_learning *ml = br->ml;
2792 struct mac_entry *mac, *next_mac;
2795 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2796 if (mac->port.p == port) {
2797 mac_learning_expire(ml, mac);
2803 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2805 struct sset new_ifaces;
2806 bool need_flush = false;
2807 unsigned long *trunks;
2814 /* Add new interfaces and update 'cfg' member of existing ones. */
2815 sset_init(&new_ifaces);
2816 for (i = 0; i < cfg->n_interfaces; i++) {
2817 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2818 struct iface *iface;
2820 if (!sset_add(&new_ifaces, if_cfg->name)) {
2821 VLOG_WARN("port %s: %s specified twice as port interface",
2822 port->name, if_cfg->name);
2823 iface_set_ofport(if_cfg, -1);
2827 iface = iface_lookup(port->bridge, if_cfg->name);
2829 if (iface->port != port) {
2830 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2832 port->bridge->name, if_cfg->name, iface->port->name);
2835 iface->cfg = if_cfg;
2837 iface = iface_create(port, if_cfg);
2840 /* Determine interface type. The local port always has type
2841 * "internal". Other ports take their type from the database and
2842 * default to "system" if none is specified. */
2843 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2844 : if_cfg->type[0] ? if_cfg->type
2847 sset_destroy(&new_ifaces);
2852 if (list_is_short(&port->ifaces)) {
2854 if (vlan >= 0 && vlan <= 4095) {
2855 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2860 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2861 * they even work as-is. But they have not been tested. */
2862 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2866 if (port->vlan != vlan) {
2871 /* Get trunked VLANs. */
2873 if (vlan < 0 && cfg->n_trunks) {
2874 trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
2876 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2879 } else if (vlan >= 0 && cfg->n_trunks) {
2880 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2883 if (!vlan_bitmap_equal(trunks, port->trunks)) {
2886 bitmap_free(port->trunks);
2887 port->trunks = trunks;
2890 port_flush_macs(port);
2895 port_destroy(struct port *port)
2898 struct bridge *br = port->bridge;
2899 struct iface *iface, *next;
2902 for (i = 0; i < MAX_MIRRORS; i++) {
2903 struct mirror *m = br->mirrors[i];
2904 if (m && m->out_port == port) {
2909 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2910 iface_destroy(iface);
2913 hmap_remove(&br->ports, &port->hmap_node);
2915 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2917 bond_destroy(port->bond);
2918 lacp_destroy(port->lacp);
2919 port_flush_macs(port);
2921 bitmap_free(port->trunks);
2927 static struct port *
2928 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
2930 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
2931 return iface ? iface->port : NULL;
2934 static struct port *
2935 port_lookup(const struct bridge *br, const char *name)
2939 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2941 if (!strcmp(port->name, name)) {
2949 enable_lacp(struct port *port, bool *activep)
2951 if (!port->cfg->lacp) {
2952 /* XXX when LACP implementation has been sufficiently tested, enable by
2953 * default and make active on bonded ports. */
2955 } else if (!strcmp(port->cfg->lacp, "off")) {
2957 } else if (!strcmp(port->cfg->lacp, "active")) {
2960 } else if (!strcmp(port->cfg->lacp, "passive")) {
2964 VLOG_WARN("port %s: unknown LACP mode %s",
2965 port->name, port->cfg->lacp);
2971 iface_reconfigure_lacp(struct iface *iface)
2973 struct lacp_slave_settings s;
2974 int priority, portid;
2976 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2977 priority = atoi(get_interface_other_config(iface->cfg,
2978 "lacp-port-priority", "0"));
2980 if (portid <= 0 || portid > UINT16_MAX) {
2981 portid = iface->dp_ifidx;
2984 if (priority <= 0 || priority > UINT16_MAX) {
2985 priority = UINT16_MAX;
2988 s.name = iface->name;
2990 s.priority = priority;
2991 lacp_slave_register(iface->port->lacp, iface, &s);
2995 port_reconfigure_lacp(struct port *port)
2997 static struct lacp_settings s;
2998 struct iface *iface;
2999 uint8_t sysid[ETH_ADDR_LEN];
3000 const char *sysid_str;
3001 const char *lacp_time;
3002 long long int custom_time;
3005 if (!enable_lacp(port, &s.active)) {
3006 lacp_destroy(port->lacp);
3011 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3012 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3013 memcpy(s.id, sysid, ETH_ADDR_LEN);
3015 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3018 s.name = port->name;
3020 /* Prefer bondable links if unspecified. */
3021 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3023 s.priority = (priority > 0 && priority <= UINT16_MAX
3025 : UINT16_MAX - !list_is_short(&port->ifaces));
3027 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3031 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3032 custom_time = atoi(lacp_time);
3033 if (!strcmp(lacp_time, "fast")) {
3034 s.lacp_time = LACP_TIME_FAST;
3035 } else if (!strcmp(lacp_time, "slow")) {
3036 s.lacp_time = LACP_TIME_SLOW;
3037 } else if (custom_time > 0) {
3038 s.lacp_time = LACP_TIME_CUSTOM;
3039 s.custom_time = custom_time;
3041 s.lacp_time = LACP_TIME_SLOW;
3045 port->lacp = lacp_create();
3048 lacp_configure(port->lacp, &s);
3050 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3051 iface_reconfigure_lacp(iface);
3056 port_reconfigure_bond(struct port *port)
3058 struct bond_settings s;
3059 const char *detect_s;
3060 struct iface *iface;
3062 if (list_is_short(&port->ifaces)) {
3063 /* Not a bonded port. */
3064 bond_destroy(port->bond);
3069 port->bridge->has_bonded_ports = true;
3071 s.name = port->name;
3073 if (port->cfg->bond_mode
3074 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3075 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3076 port->name, port->cfg->bond_mode,
3077 bond_mode_to_string(s.balance));
3080 s.detect = BLSM_CARRIER;
3081 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3082 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3083 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3085 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3088 s.miimon_interval = atoi(
3089 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3090 if (s.miimon_interval < 100) {
3091 s.miimon_interval = 100;
3094 s.up_delay = MAX(0, port->cfg->bond_updelay);
3095 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3096 s.rebalance_interval = atoi(
3097 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3098 if (s.rebalance_interval < 1000) {
3099 s.rebalance_interval = 1000;
3102 s.fake_iface = port->cfg->bond_fake_iface;
3105 port->bond = bond_create(&s);
3107 if (bond_reconfigure(port->bond, &s)) {
3108 bridge_flush(port->bridge);
3112 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3113 uint16_t stable_id = (port->lacp
3114 ? lacp_slave_get_port_id(port->lacp, iface)
3116 bond_slave_register(iface->port->bond, iface, stable_id,
3122 port_send_learning_packets(struct port *port)
3124 struct bridge *br = port->bridge;
3125 int error, n_packets, n_errors;
3126 struct mac_entry *e;
3128 error = n_packets = n_errors = 0;
3129 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3130 if (e->port.p != port) {
3131 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3141 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3142 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3143 "packets, last error was: %s",
3144 port->name, n_errors, n_packets, strerror(error));
3146 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3147 port->name, n_packets);
3151 /* Interface functions. */
3153 static struct iface *
3154 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3156 struct bridge *br = port->bridge;
3157 struct iface *iface;
3158 char *name = if_cfg->name;
3160 iface = xzalloc(sizeof *iface);
3162 iface->name = xstrdup(name);
3163 iface->dp_ifidx = -1;
3164 iface->tag = tag_create_random();
3165 iface->netdev = NULL;
3166 iface->cfg = if_cfg;
3168 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
3170 list_push_back(&port->ifaces, &iface->port_elem);
3172 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3180 iface_destroy(struct iface *iface)
3183 struct port *port = iface->port;
3184 struct bridge *br = port->bridge;
3187 bond_slave_unregister(port->bond, iface);
3191 lacp_slave_unregister(port->lacp, iface);
3194 if (iface->dp_ifidx >= 0) {
3195 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3198 list_remove(&iface->port_elem);
3199 hmap_remove(&br->iface_by_name, &iface->name_node);
3201 netdev_close(iface->netdev);
3206 bridge_flush(port->bridge);
3210 static struct iface *
3211 iface_lookup(const struct bridge *br, const char *name)
3213 struct iface *iface;
3215 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3216 &br->iface_by_name) {
3217 if (!strcmp(iface->name, name)) {
3225 static struct iface *
3226 iface_find(const char *name)
3228 const struct bridge *br;
3230 HMAP_FOR_EACH (br, node, &all_bridges) {
3231 struct iface *iface = iface_lookup(br, name);
3240 static struct iface *
3241 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3243 struct iface *iface;
3245 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3246 hash_int(dp_ifidx, 0), &br->ifaces) {
3247 if (iface->dp_ifidx == dp_ifidx) {
3254 /* Set Ethernet address of 'iface', if one is specified in the configuration
3257 iface_set_mac(struct iface *iface)
3259 uint8_t ea[ETH_ADDR_LEN];
3261 if (!strcmp(iface->type, "internal")
3262 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3263 if (iface->dp_ifidx == ODPP_LOCAL) {
3264 VLOG_ERR("interface %s: ignoring mac in Interface record "
3265 "(use Bridge record to set local port's mac)",
3267 } else if (eth_addr_is_multicast(ea)) {
3268 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3271 int error = netdev_set_etheraddr(iface->netdev, ea);
3273 VLOG_ERR("interface %s: setting MAC failed (%s)",
3274 iface->name, strerror(error));
3280 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3282 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3284 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3285 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3289 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3291 * The value strings in '*shash' are taken directly from values[], not copied,
3292 * so the caller should not modify or free them. */
3294 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3295 struct shash *shash)
3300 for (i = 0; i < n; i++) {
3301 shash_add(shash, keys[i], values[i]);
3305 /* Creates 'keys' and 'values' arrays from 'shash'.
3307 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3308 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3309 * are populated with with strings taken directly from 'shash' and thus have
3310 * the same ownership of the key-value pairs in shash.
3313 shash_to_ovs_idl_map(struct shash *shash,
3314 char ***keys, char ***values, size_t *n)
3318 struct shash_node *sn;
3320 count = shash_count(shash);
3322 k = xmalloc(count * sizeof *k);
3323 v = xmalloc(count * sizeof *v);
3326 SHASH_FOR_EACH(sn, shash) {
3337 struct iface_delete_queues_cbdata {
3338 struct netdev *netdev;
3339 const struct ovsdb_datum *queues;
3343 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3345 union ovsdb_atom atom;
3347 atom.integer = target;
3348 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3352 iface_delete_queues(unsigned int queue_id,
3353 const struct shash *details OVS_UNUSED, void *cbdata_)
3355 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3357 if (!queue_ids_include(cbdata->queues, queue_id)) {
3358 netdev_delete_queue(cbdata->netdev, queue_id);
3363 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3365 if (!qos || qos->type[0] == '\0') {
3366 netdev_set_qos(iface->netdev, NULL, NULL);
3368 struct iface_delete_queues_cbdata cbdata;
3369 struct shash details;
3372 /* Configure top-level Qos for 'iface'. */
3373 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3374 qos->n_other_config, &details);
3375 netdev_set_qos(iface->netdev, qos->type, &details);
3376 shash_destroy(&details);
3378 /* Deconfigure queues that were deleted. */
3379 cbdata.netdev = iface->netdev;
3380 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3382 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3384 /* Configure queues for 'iface'. */
3385 for (i = 0; i < qos->n_queues; i++) {
3386 const struct ovsrec_queue *queue = qos->value_queues[i];
3387 unsigned int queue_id = qos->key_queues[i];
3389 shash_from_ovs_idl_map(queue->key_other_config,
3390 queue->value_other_config,
3391 queue->n_other_config, &details);
3392 netdev_set_queue(iface->netdev, queue_id, &details);
3393 shash_destroy(&details);
3399 iface_update_cfm(struct iface *iface)
3403 uint16_t *remote_mps;
3404 struct ovsrec_monitor *mon;
3405 uint8_t maid[CCM_MAID_LEN];
3407 mon = iface->cfg->monitor;
3410 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3414 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3415 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3419 cfm.mpid = mon->mpid;
3420 cfm.interval = mon->interval ? *mon->interval : 1000;
3422 memcpy(cfm.maid, maid, sizeof cfm.maid);
3424 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3425 for(i = 0; i < mon->n_remote_mps; i++) {
3426 remote_mps[i] = mon->remote_mps[i]->mpid;
3429 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3430 &cfm, remote_mps, mon->n_remote_mps);
3434 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3435 * how 'iface''s port is configured.
3437 * Returns true if 'iface' is up, false otherwise. */
3439 iface_get_carrier(const struct iface *iface)
3442 return netdev_get_carrier(iface->netdev);
3445 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3446 * instead of obtaining it from the database. */
3448 iface_is_synthetic(const struct iface *iface)
3450 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3453 /* Port mirroring. */
3455 static struct mirror *
3456 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3460 for (i = 0; i < MAX_MIRRORS; i++) {
3461 struct mirror *m = br->mirrors[i];
3462 if (m && uuid_equals(uuid, &m->uuid)) {
3470 mirror_reconfigure(struct bridge *br)
3472 unsigned long *rspan_vlans;
3476 /* Get rid of deleted mirrors. */
3477 for (i = 0; i < MAX_MIRRORS; i++) {
3478 struct mirror *m = br->mirrors[i];
3480 const struct ovsdb_datum *mc;
3481 union ovsdb_atom atom;
3483 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3484 atom.uuid = br->mirrors[i]->uuid;
3485 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3491 /* Add new mirrors and reconfigure existing ones. */
3492 for (i = 0; i < br->cfg->n_mirrors; i++) {
3493 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3494 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3496 mirror_reconfigure_one(m, cfg);
3498 mirror_create(br, cfg);
3502 /* Update port reserved status. */
3503 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3504 port->is_mirror_output_port = false;
3506 for (i = 0; i < MAX_MIRRORS; i++) {
3507 struct mirror *m = br->mirrors[i];
3508 if (m && m->out_port) {
3509 m->out_port->is_mirror_output_port = true;
3513 /* Update flooded vlans (for RSPAN). */
3515 if (br->cfg->n_flood_vlans) {
3516 rspan_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3517 br->cfg->n_flood_vlans);
3519 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3521 mac_learning_flush(br->ml);
3527 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3532 for (i = 0; ; i++) {
3533 if (i >= MAX_MIRRORS) {
3534 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3535 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3538 if (!br->mirrors[i]) {
3543 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3545 mac_learning_flush(br->ml);
3547 br->mirrors[i] = m = xzalloc(sizeof *m);
3548 m->uuid = cfg->header_.uuid;
3551 m->name = xstrdup(cfg->name);
3552 sset_init(&m->src_ports);
3553 sset_init(&m->dst_ports);
3559 mirror_reconfigure_one(m, cfg);
3563 mirror_destroy(struct mirror *m)
3566 struct bridge *br = m->bridge;
3569 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3570 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3571 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3574 sset_destroy(&m->src_ports);
3575 sset_destroy(&m->dst_ports);
3578 m->bridge->mirrors[m->idx] = NULL;
3583 mac_learning_flush(br->ml);
3588 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3593 for (i = 0; i < n_ports; i++) {
3594 const char *name = ports[i]->name;
3595 if (port_lookup(m->bridge, name)) {
3596 sset_add(names, name);
3598 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3599 "port %s", m->bridge->name, m->name, name);
3605 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3611 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3613 for (i = 0; i < cfg->n_select_vlan; i++) {
3614 int64_t vlan = cfg->select_vlan[i];
3615 if (vlan < 0 || vlan > 4095) {
3616 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3617 m->bridge->name, m->name, vlan);
3619 (*vlans)[n_vlans++] = vlan;
3626 vlan_is_mirrored(const struct mirror *m, int vlan)
3630 for (i = 0; i < m->n_vlans; i++) {
3631 if (m->vlans[i] == vlan) {
3639 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3641 struct sset src_ports, dst_ports;
3642 mirror_mask_t mirror_bit;
3643 struct port *out_port;
3650 if (strcmp(cfg->name, m->name)) {
3652 m->name = xstrdup(cfg->name);
3655 /* Get output port. */
3656 if (cfg->output_port) {
3657 out_port = port_lookup(m->bridge, cfg->output_port->name);
3659 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3660 m->bridge->name, m->name);
3666 if (cfg->output_vlan) {
3667 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3668 "output vlan; ignoring output vlan",
3669 m->bridge->name, m->name);
3671 } else if (cfg->output_vlan) {
3673 out_vlan = *cfg->output_vlan;
3675 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3676 m->bridge->name, m->name);
3681 sset_init(&src_ports);
3682 sset_init(&dst_ports);
3683 if (cfg->select_all) {
3684 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3685 sset_add(&src_ports, port->name);
3686 sset_add(&dst_ports, port->name);
3691 /* Get ports, and drop duplicates and ports that don't exist. */
3692 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3694 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3697 /* Get all the vlans, and drop duplicate and invalid vlans. */
3698 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3701 /* Update mirror data. */
3702 if (!sset_equals(&m->src_ports, &src_ports)
3703 || !sset_equals(&m->dst_ports, &dst_ports)
3704 || m->n_vlans != n_vlans
3705 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3706 || m->out_port != out_port
3707 || m->out_vlan != out_vlan) {
3708 bridge_flush(m->bridge);
3709 mac_learning_flush(m->bridge->ml);
3711 sset_swap(&m->src_ports, &src_ports);
3712 sset_swap(&m->dst_ports, &dst_ports);
3715 m->n_vlans = n_vlans;
3716 m->out_port = out_port;
3717 m->out_vlan = out_vlan;
3720 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3721 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3722 if (sset_contains(&m->src_ports, port->name)) {
3723 port->src_mirrors |= mirror_bit;
3725 port->src_mirrors &= ~mirror_bit;
3728 if (sset_contains(&m->dst_ports, port->name)) {
3729 port->dst_mirrors |= mirror_bit;
3731 port->dst_mirrors &= ~mirror_bit;
3736 sset_destroy(&src_ports);
3737 sset_destroy(&dst_ports);