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"
77 VLOG_DEFINE_THIS_MODULE(bridge);
79 COVERAGE_DEFINE(bridge_flush);
80 COVERAGE_DEFINE(bridge_process_flow);
81 COVERAGE_DEFINE(bridge_reconfigure);
89 struct dst builtin[32];
94 static void dst_set_init(struct dst_set *);
95 static void dst_set_add(struct dst_set *, const struct dst *);
96 static void dst_set_free(struct dst_set *);
99 /* These members are always valid. */
100 struct list port_elem; /* Element in struct port's "ifaces" list. */
101 struct port *port; /* Containing port. */
102 char *name; /* Host network device name. */
103 tag_type tag; /* Tag associated with this interface. */
105 /* These members are valid only after bridge_reconfigure() causes them to
107 struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
108 int dp_ifidx; /* Index within kernel datapath. */
109 struct netdev *netdev; /* Network device. */
110 const char *type; /* Usually same as cfg->type. */
111 const struct ovsrec_interface *cfg;
114 #define MAX_MIRRORS 32
115 typedef uint32_t mirror_mask_t;
116 #define MIRROR_MASK_C(X) UINT32_C(X)
117 BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS);
119 struct bridge *bridge;
122 struct uuid uuid; /* UUID of this "mirror" record in database. */
124 /* Selection criteria. */
125 struct sset src_ports; /* Source port names. */
126 struct sset dst_ports; /* Destination port names. */
131 struct port *out_port;
135 #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
137 struct bridge *bridge;
138 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
141 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
142 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
143 * NULL if all VLANs are trunked. */
144 const struct ovsrec_port *cfg;
146 /* An ordinary bridge port has 1 interface.
147 * A bridge port for bonding has at least 2 interfaces. */
148 struct list ifaces; /* List of "struct iface"s. */
153 /* Port mirroring info. */
154 mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
155 mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
156 bool is_mirror_output_port; /* Does port mirroring send frames here? */
160 struct list node; /* Node in global list of bridges. */
161 char *name; /* User-specified arbitrary name. */
162 struct mac_learning *ml; /* MAC learning table. */
163 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
164 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
165 const struct ovsrec_bridge *cfg;
167 /* OpenFlow switch processing. */
168 struct ofproto *ofproto; /* OpenFlow switch. */
170 /* Kernel datapath information. */
171 struct dpif *dpif; /* Datapath. */
172 struct hmap ifaces; /* "struct iface"s indexed by dp_ifidx. */
175 struct hmap ports; /* "struct port"s indexed by name. */
176 struct shash iface_by_name; /* "struct iface"s indexed by name. */
179 bool has_bonded_ports;
184 /* Port mirroring. */
185 struct mirror *mirrors[MAX_MIRRORS];
188 /* List of all bridges. */
189 static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
191 /* OVSDB IDL used to obtain configuration. */
192 static struct ovsdb_idl *idl;
194 /* Each time this timer expires, the bridge fetches systems and interface
195 * statistics and pushes them into the database. */
196 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
197 static long long int stats_timer = LLONG_MIN;
199 /* Stores the time after which CFM statistics may be written to the database.
200 * Only updated when changes to the database require rate limiting. */
201 #define CFM_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
202 static long long int cfm_limiter = LLONG_MIN;
204 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
205 static void bridge_destroy(struct bridge *);
206 static struct bridge *bridge_lookup(const char *name);
207 static unixctl_cb_func bridge_unixctl_dump_flows;
208 static unixctl_cb_func bridge_unixctl_reconnect;
209 static int bridge_run_one(struct bridge *);
210 static size_t bridge_get_controllers(const struct bridge *br,
211 struct ovsrec_controller ***controllersp);
212 static void bridge_reconfigure_one(struct bridge *);
213 static void bridge_reconfigure_remotes(struct bridge *,
214 const struct sockaddr_in *managers,
216 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
217 static void bridge_fetch_dp_ifaces(struct bridge *);
218 static void bridge_flush(struct bridge *);
219 static void bridge_pick_local_hw_addr(struct bridge *,
220 uint8_t ea[ETH_ADDR_LEN],
221 struct iface **hw_addr_iface);
222 static uint64_t bridge_pick_datapath_id(struct bridge *,
223 const uint8_t bridge_ea[ETH_ADDR_LEN],
224 struct iface *hw_addr_iface);
225 static uint64_t dpid_from_hash(const void *, size_t nbytes);
227 static unixctl_cb_func bridge_unixctl_fdb_show;
228 static unixctl_cb_func cfm_unixctl_show;
229 static unixctl_cb_func qos_unixctl_show;
231 static void port_run(struct port *);
232 static void port_wait(struct port *);
233 static struct port *port_create(struct bridge *, const char *name);
234 static void port_reconfigure(struct port *, const struct ovsrec_port *);
235 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
236 static void port_destroy(struct port *);
237 static struct port *port_lookup(const struct bridge *, const char *name);
238 static struct iface *port_get_an_iface(const struct port *);
239 static struct port *port_from_dp_ifidx(const struct bridge *,
241 static void port_reconfigure_bond(struct port *);
242 static void port_send_learning_packets(struct port *);
244 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
245 static void mirror_destroy(struct mirror *);
246 static void mirror_reconfigure(struct bridge *);
247 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
248 static bool vlan_is_mirrored(const struct mirror *, int vlan);
250 static struct iface *iface_create(struct port *port,
251 const struct ovsrec_interface *if_cfg);
252 static void iface_destroy(struct iface *);
253 static struct iface *iface_lookup(const struct bridge *, const char *name);
254 static struct iface *iface_find(const char *name);
255 static struct iface *iface_from_dp_ifidx(const struct bridge *,
257 static void iface_set_mac(struct iface *);
258 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
259 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
260 static void iface_update_cfm(struct iface *);
261 static bool iface_refresh_cfm_stats(struct iface *iface);
262 static bool iface_get_carrier(const struct iface *);
264 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
266 static void shash_to_ovs_idl_map(struct shash *,
267 char ***keys, char ***values, size_t *n);
269 /* Hooks into ofproto processing. */
270 static struct ofhooks bridge_ofhooks;
272 /* Public functions. */
274 /* Initializes the bridge module, configuring it to obtain its configuration
275 * from an OVSDB server accessed over 'remote', which should be a string in a
276 * form acceptable to ovsdb_idl_create(). */
278 bridge_init(const char *remote)
280 /* Create connection to database. */
281 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
283 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
284 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
285 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
287 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
289 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
290 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
292 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
293 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
294 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
296 /* Register unixctl commands. */
297 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
298 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
299 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
300 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
302 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
310 struct bridge *br, *next_br;
312 LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
315 ovsdb_idl_destroy(idl);
318 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
319 * but for which the ovs-vswitchd configuration 'cfg' is required. */
321 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
323 static bool already_configured_once;
324 struct sset bridge_names;
325 struct sset dpif_names, dpif_types;
329 /* Only do this once per ovs-vswitchd run. */
330 if (already_configured_once) {
333 already_configured_once = true;
335 stats_timer = time_msec() + STATS_INTERVAL;
337 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
338 sset_init(&bridge_names);
339 for (i = 0; i < cfg->n_bridges; i++) {
340 sset_add(&bridge_names, cfg->bridges[i]->name);
343 /* Iterate over all system dpifs and delete any of them that do not appear
345 sset_init(&dpif_names);
346 sset_init(&dpif_types);
347 dp_enumerate_types(&dpif_types);
348 SSET_FOR_EACH (type, &dpif_types) {
351 dp_enumerate_names(type, &dpif_names);
353 /* Delete each dpif whose name is not in 'bridge_names'. */
354 SSET_FOR_EACH (name, &dpif_names) {
355 if (!sset_contains(&bridge_names, name)) {
359 retval = dpif_open(name, type, &dpif);
367 sset_destroy(&bridge_names);
368 sset_destroy(&dpif_names);
369 sset_destroy(&dpif_types);
372 /* Callback for iterate_and_prune_ifaces(). */
374 check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
376 if (!iface->netdev) {
377 /* We already reported a related error, don't bother duplicating it. */
381 if (iface->dp_ifidx < 0) {
382 VLOG_ERR("%s interface not in %s, dropping",
383 iface->name, dpif_name(br->dpif));
387 VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
388 iface->name, iface->dp_ifidx);
392 /* Callback for iterate_and_prune_ifaces(). */
394 set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
395 void *aux OVS_UNUSED)
397 /* Set policing attributes. */
398 netdev_set_policing(iface->netdev,
399 iface->cfg->ingress_policing_rate,
400 iface->cfg->ingress_policing_burst);
402 /* Set MAC address of internal interfaces other than the local
404 if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) {
405 iface_set_mac(iface);
411 /* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
412 * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
413 * deletes from 'br' any ports that no longer have any interfaces. */
415 iterate_and_prune_ifaces(struct bridge *br,
416 bool (*cb)(struct bridge *, struct iface *,
420 struct port *port, *next_port;
422 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
423 struct iface *iface, *next_iface;
425 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
426 if (!cb(br, iface, aux)) {
427 iface_set_ofport(iface->cfg, -1);
428 iface_destroy(iface);
432 if (list_is_empty(&port->ifaces)) {
433 VLOG_WARN("%s port has no interfaces, dropping", port->name);
439 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
440 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
441 * responsible for freeing '*managersp' (with free()).
443 * You may be asking yourself "why does ovs-vswitchd care?", because
444 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
445 * should not be and in fact is not directly involved in that. But
446 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
447 * it has to tell in-band control where the managers are to enable that.
448 * (Thus, only managers connected in-band are collected.)
451 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
452 struct sockaddr_in **managersp, size_t *n_managersp)
454 struct sockaddr_in *managers = NULL;
455 size_t n_managers = 0;
459 /* Collect all of the potential targets from the "targets" columns of the
460 * rows pointed to by "manager_options", excluding any that are
463 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
464 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
466 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
467 sset_find_and_delete(&targets, m->target);
469 sset_add(&targets, m->target);
473 /* Now extract the targets' IP addresses. */
474 if (!sset_is_empty(&targets)) {
477 managers = xmalloc(sset_count(&targets) * sizeof *managers);
478 SSET_FOR_EACH (target, &targets) {
479 struct sockaddr_in *sin = &managers[n_managers];
481 if ((!strncmp(target, "tcp:", 4)
482 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
483 (!strncmp(target, "ssl:", 4)
484 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
489 sset_destroy(&targets);
491 *managersp = managers;
492 *n_managersp = n_managers;
496 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
498 struct shash old_br, new_br;
499 struct shash_node *node;
500 struct bridge *br, *next;
501 struct sockaddr_in *managers;
504 int sflow_bridge_number;
506 COVERAGE_INC(bridge_reconfigure);
508 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
510 /* Collect old and new bridges. */
513 LIST_FOR_EACH (br, node, &all_bridges) {
514 shash_add(&old_br, br->name, br);
516 for (i = 0; i < ovs_cfg->n_bridges; i++) {
517 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
518 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
519 VLOG_WARN("more than one bridge named %s", br_cfg->name);
523 /* Get rid of deleted bridges and add new bridges. */
524 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
525 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
532 SHASH_FOR_EACH (node, &new_br) {
533 const char *br_name = node->name;
534 const struct ovsrec_bridge *br_cfg = node->data;
535 br = shash_find_data(&old_br, br_name);
537 /* If the bridge datapath type has changed, we need to tear it
538 * down and recreate. */
539 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
541 bridge_create(br_cfg);
544 bridge_create(br_cfg);
547 shash_destroy(&old_br);
548 shash_destroy(&new_br);
550 /* Reconfigure all bridges. */
551 LIST_FOR_EACH (br, node, &all_bridges) {
552 bridge_reconfigure_one(br);
555 /* Add and delete ports on all datapaths.
557 * The kernel will reject any attempt to add a given port to a datapath if
558 * that port already belongs to a different datapath, so we must do all
559 * port deletions before any port additions. */
560 LIST_FOR_EACH (br, node, &all_bridges) {
561 struct dpif_port_dump dump;
562 struct shash want_ifaces;
563 struct dpif_port dpif_port;
565 bridge_get_all_ifaces(br, &want_ifaces);
566 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
567 if (!shash_find(&want_ifaces, dpif_port.name)
568 && strcmp(dpif_port.name, br->name)) {
569 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
571 VLOG_WARN("failed to remove %s interface from %s: %s",
572 dpif_port.name, dpif_name(br->dpif),
577 shash_destroy(&want_ifaces);
579 LIST_FOR_EACH (br, node, &all_bridges) {
580 struct shash cur_ifaces, want_ifaces;
581 struct dpif_port_dump dump;
582 struct dpif_port dpif_port;
584 /* Get the set of interfaces currently in this datapath. */
585 shash_init(&cur_ifaces);
586 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
587 struct dpif_port *port_info = xmalloc(sizeof *port_info);
588 dpif_port_clone(port_info, &dpif_port);
589 shash_add(&cur_ifaces, dpif_port.name, port_info);
592 /* Get the set of interfaces we want on this datapath. */
593 bridge_get_all_ifaces(br, &want_ifaces);
595 hmap_clear(&br->ifaces);
596 SHASH_FOR_EACH (node, &want_ifaces) {
597 const char *if_name = node->name;
598 struct iface *iface = node->data;
599 struct dpif_port *dpif_port;
603 type = iface ? iface->type : "internal";
604 dpif_port = shash_find_data(&cur_ifaces, if_name);
606 /* If we have a port or a netdev already, and it's not the type we
607 * want, then delete the port (if any) and close the netdev (if
609 if ((dpif_port && strcmp(dpif_port->type, type))
610 || (iface && iface->netdev
611 && strcmp(type, netdev_get_type(iface->netdev)))) {
613 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
620 netdev_close(iface->netdev);
621 iface->netdev = NULL;
625 /* If the port doesn't exist or we don't have the netdev open,
626 * we need to do more work. */
627 if (!dpif_port || (iface && !iface->netdev)) {
628 struct netdev_options options;
629 struct netdev *netdev;
632 /* First open the network device. */
633 options.name = if_name;
635 options.args = &args;
636 options.ethertype = NETDEV_ETH_TYPE_NONE;
640 shash_from_ovs_idl_map(iface->cfg->key_options,
641 iface->cfg->value_options,
642 iface->cfg->n_options, &args);
644 error = netdev_open(&options, &netdev);
645 shash_destroy(&args);
648 VLOG_WARN("could not open network device %s (%s)",
649 if_name, strerror(error));
653 /* Then add the port if we haven't already. */
655 error = dpif_port_add(br->dpif, netdev, NULL);
657 netdev_close(netdev);
658 if (error == EFBIG) {
659 VLOG_ERR("ran out of valid port numbers on %s",
660 dpif_name(br->dpif));
663 VLOG_WARN("failed to add %s interface to %s: %s",
664 if_name, dpif_name(br->dpif),
671 /* Update 'iface'. */
673 iface->netdev = netdev;
675 } else if (iface && iface->netdev) {
679 shash_from_ovs_idl_map(iface->cfg->key_options,
680 iface->cfg->value_options,
681 iface->cfg->n_options, &args);
682 netdev_set_config(iface->netdev, &args);
683 shash_destroy(&args);
686 shash_destroy(&want_ifaces);
688 SHASH_FOR_EACH (node, &cur_ifaces) {
689 struct dpif_port *port_info = node->data;
690 dpif_port_destroy(port_info);
693 shash_destroy(&cur_ifaces);
695 sflow_bridge_number = 0;
696 LIST_FOR_EACH (br, node, &all_bridges) {
697 uint8_t ea[ETH_ADDR_LEN];
699 struct iface *local_iface;
700 struct iface *hw_addr_iface;
703 bridge_fetch_dp_ifaces(br);
705 /* Delete interfaces that cannot be opened.
707 * From this point forward we are guaranteed that every "struct iface"
708 * has nonnull 'netdev' and correct 'dp_ifidx'. */
709 iterate_and_prune_ifaces(br, check_iface, NULL);
711 /* Pick local port hardware address, datapath ID. */
712 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
713 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
715 int error = netdev_set_etheraddr(local_iface->netdev, ea);
717 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
718 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
719 "Ethernet address: %s",
720 br->name, strerror(error));
723 memcpy(br->ea, ea, ETH_ADDR_LEN);
725 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
726 ofproto_set_datapath_id(br->ofproto, dpid);
728 dpid_string = xasprintf("%016"PRIx64, dpid);
729 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
732 /* Set NetFlow configuration on this bridge. */
733 if (br->cfg->netflow) {
734 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
735 struct netflow_options opts;
737 memset(&opts, 0, sizeof opts);
739 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
740 if (nf_cfg->engine_type) {
741 opts.engine_type = *nf_cfg->engine_type;
743 if (nf_cfg->engine_id) {
744 opts.engine_id = *nf_cfg->engine_id;
747 opts.active_timeout = nf_cfg->active_timeout;
748 if (!opts.active_timeout) {
749 opts.active_timeout = -1;
750 } else if (opts.active_timeout < 0) {
751 VLOG_WARN("bridge %s: active timeout interval set to negative "
752 "value, using default instead (%d seconds)", br->name,
753 NF_ACTIVE_TIMEOUT_DEFAULT);
754 opts.active_timeout = -1;
757 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
758 if (opts.add_id_to_iface) {
759 if (opts.engine_id > 0x7f) {
760 VLOG_WARN("bridge %s: netflow port mangling may conflict "
761 "with another vswitch, choose an engine id less "
762 "than 128", br->name);
764 if (hmap_count(&br->ports) > 508) {
765 VLOG_WARN("bridge %s: netflow port mangling will conflict "
766 "with another port when more than 508 ports are "
771 sset_init(&opts.collectors);
772 sset_add_array(&opts.collectors,
773 nf_cfg->targets, nf_cfg->n_targets);
774 if (ofproto_set_netflow(br->ofproto, &opts)) {
775 VLOG_ERR("bridge %s: problem setting netflow collectors",
778 sset_destroy(&opts.collectors);
780 ofproto_set_netflow(br->ofproto, NULL);
783 /* Set sFlow configuration on this bridge. */
784 if (br->cfg->sflow) {
785 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
786 struct ovsrec_controller **controllers;
787 struct ofproto_sflow_options oso;
788 size_t n_controllers;
790 memset(&oso, 0, sizeof oso);
792 sset_init(&oso.targets);
793 sset_add_array(&oso.targets,
794 sflow_cfg->targets, sflow_cfg->n_targets);
796 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
797 if (sflow_cfg->sampling) {
798 oso.sampling_rate = *sflow_cfg->sampling;
801 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
802 if (sflow_cfg->polling) {
803 oso.polling_interval = *sflow_cfg->polling;
806 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
807 if (sflow_cfg->header) {
808 oso.header_len = *sflow_cfg->header;
811 oso.sub_id = sflow_bridge_number++;
812 oso.agent_device = sflow_cfg->agent;
814 oso.control_ip = NULL;
815 n_controllers = bridge_get_controllers(br, &controllers);
816 for (i = 0; i < n_controllers; i++) {
817 if (controllers[i]->local_ip) {
818 oso.control_ip = controllers[i]->local_ip;
822 ofproto_set_sflow(br->ofproto, &oso);
824 sset_destroy(&oso.targets);
826 ofproto_set_sflow(br->ofproto, NULL);
829 /* Update the controller and related settings. It would be more
830 * straightforward to call this from bridge_reconfigure_one(), but we
831 * can't do it there for two reasons. First, and most importantly, at
832 * that point we don't know the dp_ifidx of any interfaces that have
833 * been added to the bridge (because we haven't actually added them to
834 * the datapath). Second, at that point we haven't set the datapath ID
835 * yet; when a controller is configured, resetting the datapath ID will
836 * immediately disconnect from the controller, so it's better to set
837 * the datapath ID before the controller. */
838 bridge_reconfigure_remotes(br, managers, n_managers);
840 LIST_FOR_EACH (br, node, &all_bridges) {
843 br->has_bonded_ports = false;
844 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
847 port_reconfigure_bond(port);
849 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
850 iface_update_qos(iface, port->cfg->qos);
854 LIST_FOR_EACH (br, node, &all_bridges) {
855 iterate_and_prune_ifaces(br, set_iface_properties, NULL);
858 /* Some reconfiguration operations require the bridge to have been run at
860 LIST_FOR_EACH (br, node, &all_bridges) {
865 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
866 iface_update_cfm(iface);
872 /* ovs-vswitchd has completed initialization, so allow the process that
873 * forked us to exit successfully. */
874 daemonize_complete();
878 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
879 const struct ovsdb_idl_column *column,
882 const struct ovsdb_datum *datum;
883 union ovsdb_atom atom;
886 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
887 atom.string = (char *) key;
888 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
889 return idx == UINT_MAX ? NULL : datum->values[idx].string;
893 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
895 return get_ovsrec_key_value(&br_cfg->header_,
896 &ovsrec_bridge_col_other_config, key);
900 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
901 struct iface **hw_addr_iface)
907 *hw_addr_iface = NULL;
909 /* Did the user request a particular MAC? */
910 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
911 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
912 if (eth_addr_is_multicast(ea)) {
913 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
914 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
915 } else if (eth_addr_is_zero(ea)) {
916 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
922 /* Otherwise choose the minimum non-local MAC address among all of the
924 memset(ea, 0xff, ETH_ADDR_LEN);
925 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
926 uint8_t iface_ea[ETH_ADDR_LEN];
927 struct iface *candidate;
930 /* Mirror output ports don't participate. */
931 if (port->is_mirror_output_port) {
935 /* Choose the MAC address to represent the port. */
937 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
938 /* Find the interface with this Ethernet address (if any) so that
939 * we can provide the correct devname to the caller. */
940 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
941 uint8_t candidate_ea[ETH_ADDR_LEN];
942 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
943 && eth_addr_equals(iface_ea, candidate_ea)) {
948 /* Choose the interface whose MAC address will represent the port.
949 * The Linux kernel bonding code always chooses the MAC address of
950 * the first slave added to a bond, and the Fedora networking
951 * scripts always add slaves to a bond in alphabetical order, so
952 * for compatibility we choose the interface with the name that is
953 * first in alphabetical order. */
954 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
955 if (!iface || strcmp(candidate->name, iface->name) < 0) {
960 /* The local port doesn't count (since we're trying to choose its
961 * MAC address anyway). */
962 if (iface->dp_ifidx == ODPP_LOCAL) {
967 error = netdev_get_etheraddr(iface->netdev, iface_ea);
969 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
970 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
971 iface->name, strerror(error));
976 /* Compare against our current choice. */
977 if (!eth_addr_is_multicast(iface_ea) &&
978 !eth_addr_is_local(iface_ea) &&
979 !eth_addr_is_reserved(iface_ea) &&
980 !eth_addr_is_zero(iface_ea) &&
981 eth_addr_compare_3way(iface_ea, ea) < 0)
983 memcpy(ea, iface_ea, ETH_ADDR_LEN);
984 *hw_addr_iface = iface;
987 if (eth_addr_is_multicast(ea)) {
988 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
989 *hw_addr_iface = NULL;
990 VLOG_WARN("bridge %s: using default bridge Ethernet "
991 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
993 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
994 br->name, ETH_ADDR_ARGS(ea));
998 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
999 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1000 * an interface on 'br', then that interface must be passed in as
1001 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1002 * 'hw_addr_iface' must be passed in as a null pointer. */
1004 bridge_pick_datapath_id(struct bridge *br,
1005 const uint8_t bridge_ea[ETH_ADDR_LEN],
1006 struct iface *hw_addr_iface)
1009 * The procedure for choosing a bridge MAC address will, in the most
1010 * ordinary case, also choose a unique MAC that we can use as a datapath
1011 * ID. In some special cases, though, multiple bridges will end up with
1012 * the same MAC address. This is OK for the bridges, but it will confuse
1013 * the OpenFlow controller, because each datapath needs a unique datapath
1016 * Datapath IDs must be unique. It is also very desirable that they be
1017 * stable from one run to the next, so that policy set on a datapath
1020 const char *datapath_id;
1023 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1024 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1028 if (hw_addr_iface) {
1030 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1032 * A bridge whose MAC address is taken from a VLAN network device
1033 * (that is, a network device created with vconfig(8) or similar
1034 * tool) will have the same MAC address as a bridge on the VLAN
1035 * device's physical network device.
1037 * Handle this case by hashing the physical network device MAC
1038 * along with the VLAN identifier.
1040 uint8_t buf[ETH_ADDR_LEN + 2];
1041 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1042 buf[ETH_ADDR_LEN] = vlan >> 8;
1043 buf[ETH_ADDR_LEN + 1] = vlan;
1044 return dpid_from_hash(buf, sizeof buf);
1047 * Assume that this bridge's MAC address is unique, since it
1048 * doesn't fit any of the cases we handle specially.
1053 * A purely internal bridge, that is, one that has no non-virtual
1054 * network devices on it at all, is more difficult because it has no
1055 * natural unique identifier at all.
1057 * When the host is a XenServer, we handle this case by hashing the
1058 * host's UUID with the name of the bridge. Names of bridges are
1059 * persistent across XenServer reboots, although they can be reused if
1060 * an internal network is destroyed and then a new one is later
1061 * created, so this is fairly effective.
1063 * When the host is not a XenServer, we punt by using a random MAC
1064 * address on each run.
1066 const char *host_uuid = xenserver_get_host_uuid();
1068 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1069 dpid = dpid_from_hash(combined, strlen(combined));
1075 return eth_addr_to_uint64(bridge_ea);
1079 dpid_from_hash(const void *data, size_t n)
1081 uint8_t hash[SHA1_DIGEST_SIZE];
1083 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1084 sha1_bytes(data, n, hash);
1085 eth_addr_mark_random(hash);
1086 return eth_addr_to_uint64(hash);
1090 iface_refresh_status(struct iface *iface)
1094 enum netdev_flags flags;
1103 if (!netdev_get_status(iface->netdev, &sh)) {
1105 char **keys, **values;
1107 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1108 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1113 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1116 shash_destroy_free_data(&sh);
1118 error = netdev_get_flags(iface->netdev, &flags);
1120 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1123 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1126 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1128 ovsrec_interface_set_duplex(iface->cfg,
1129 netdev_features_is_full_duplex(current)
1131 /* warning: uint64_t -> int64_t conversion */
1132 bps = netdev_features_to_bps(current);
1133 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1136 ovsrec_interface_set_duplex(iface->cfg, NULL);
1137 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1141 ovsrec_interface_set_link_state(iface->cfg,
1142 iface_get_carrier(iface) ? "up" : "down");
1144 error = netdev_get_mtu(iface->netdev, &mtu);
1145 if (!error && mtu != INT_MAX) {
1147 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1150 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1154 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1155 * changed, false otherwise. */
1157 iface_refresh_cfm_stats(struct iface *iface)
1159 const struct ovsrec_monitor *mon;
1160 const struct cfm *cfm;
1161 bool changed = false;
1164 mon = iface->cfg->monitor;
1165 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1171 for (i = 0; i < mon->n_remote_mps; i++) {
1172 const struct ovsrec_maintenance_point *mp;
1173 const struct remote_mp *rmp;
1175 mp = mon->remote_mps[i];
1176 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1178 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1179 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1184 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1185 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1193 iface_refresh_stats(struct iface *iface)
1199 static const struct iface_stat iface_stats[] = {
1200 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1201 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1202 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1203 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1204 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1205 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1206 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1207 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1208 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1209 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1210 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1211 { "collisions", offsetof(struct netdev_stats, collisions) },
1213 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1214 const struct iface_stat *s;
1216 char *keys[N_STATS];
1217 int64_t values[N_STATS];
1220 struct netdev_stats stats;
1222 /* Intentionally ignore return value, since errors will set 'stats' to
1223 * all-1s, and we will deal with that correctly below. */
1224 netdev_get_stats(iface->netdev, &stats);
1227 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1228 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1229 if (value != UINT64_MAX) {
1236 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1240 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1242 struct ovsdb_datum datum;
1246 get_system_stats(&stats);
1248 ovsdb_datum_from_shash(&datum, &stats);
1249 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1253 static inline const char *
1254 nx_role_to_str(enum nx_role role)
1259 case NX_ROLE_MASTER:
1264 return "*** INVALID ROLE ***";
1269 bridge_refresh_controller_status(const struct bridge *br)
1272 const struct ovsrec_controller *cfg;
1274 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1276 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1277 struct ofproto_controller_info *cinfo =
1278 shash_find_data(&info, cfg->target);
1281 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1282 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1283 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1284 (char **) cinfo->pairs.values,
1287 ovsrec_controller_set_is_connected(cfg, false);
1288 ovsrec_controller_set_role(cfg, NULL);
1289 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1293 ofproto_free_ofproto_controller_info(&info);
1299 const struct ovsrec_open_vswitch *cfg;
1301 bool datapath_destroyed;
1302 bool database_changed;
1305 /* Let each bridge do the work that it needs to do. */
1306 datapath_destroyed = false;
1307 LIST_FOR_EACH (br, node, &all_bridges) {
1308 int error = bridge_run_one(br);
1310 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1311 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1312 "forcing reconfiguration", br->name);
1313 datapath_destroyed = true;
1317 /* (Re)configure if necessary. */
1318 database_changed = ovsdb_idl_run(idl);
1319 cfg = ovsrec_open_vswitch_first(idl);
1321 /* Re-configure SSL. We do this on every trip through the main loop,
1322 * instead of just when the database changes, because the contents of the
1323 * key and certificate files can change without the database changing.
1325 * We do this before bridge_reconfigure() because that function might
1326 * initiate SSL connections and thus requires SSL to be configured. */
1327 if (cfg && cfg->ssl) {
1328 const struct ovsrec_ssl *ssl = cfg->ssl;
1330 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1331 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1334 if (database_changed || datapath_destroyed) {
1336 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1338 bridge_configure_once(cfg);
1339 bridge_reconfigure(cfg);
1341 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1342 ovsdb_idl_txn_commit(txn);
1343 ovsdb_idl_txn_destroy(txn); /* XXX */
1345 /* We still need to reconfigure to avoid dangling pointers to
1346 * now-destroyed ovsrec structures inside bridge data. */
1347 static const struct ovsrec_open_vswitch null_cfg;
1349 bridge_reconfigure(&null_cfg);
1353 /* Refresh system and interface stats if necessary. */
1354 if (time_msec() >= stats_timer) {
1356 struct ovsdb_idl_txn *txn;
1358 txn = ovsdb_idl_txn_create(idl);
1359 LIST_FOR_EACH (br, node, &all_bridges) {
1362 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1363 struct iface *iface;
1365 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1366 iface_refresh_stats(iface);
1367 iface_refresh_status(iface);
1370 bridge_refresh_controller_status(br);
1372 refresh_system_stats(cfg);
1373 ovsdb_idl_txn_commit(txn);
1374 ovsdb_idl_txn_destroy(txn); /* XXX */
1377 stats_timer = time_msec() + STATS_INTERVAL;
1380 if (time_msec() >= cfm_limiter) {
1381 struct ovsdb_idl_txn *txn;
1382 bool changed = false;
1384 txn = ovsdb_idl_txn_create(idl);
1385 LIST_FOR_EACH (br, node, &all_bridges) {
1388 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1389 struct iface *iface;
1391 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1392 changed = iface_refresh_cfm_stats(iface) || changed;
1398 cfm_limiter = time_msec() + CFM_LIMIT_INTERVAL;
1401 ovsdb_idl_txn_commit(txn);
1402 ovsdb_idl_txn_destroy(txn);
1411 LIST_FOR_EACH (br, node, &all_bridges) {
1414 ofproto_wait(br->ofproto);
1415 mac_learning_wait(br->ml);
1416 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1420 ovsdb_idl_wait(idl);
1421 poll_timer_wait_until(stats_timer);
1423 if (cfm_limiter > time_msec()) {
1424 poll_timer_wait_until(cfm_limiter);
1428 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1429 * configuration changes. */
1431 bridge_flush(struct bridge *br)
1433 COVERAGE_INC(bridge_flush);
1437 /* Bridge unixctl user interface functions. */
1439 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1440 const char *args, void *aux OVS_UNUSED)
1442 struct ds ds = DS_EMPTY_INITIALIZER;
1443 const struct bridge *br;
1444 const struct mac_entry *e;
1446 br = bridge_lookup(args);
1448 unixctl_command_reply(conn, 501, "no such bridge");
1452 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1453 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1454 struct port *port = e->port.p;
1455 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1456 port_get_an_iface(port)->dp_ifidx,
1457 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1459 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1463 /* CFM unixctl user interface functions. */
1465 cfm_unixctl_show(struct unixctl_conn *conn,
1466 const char *args, void *aux OVS_UNUSED)
1468 struct ds ds = DS_EMPTY_INITIALIZER;
1469 struct iface *iface;
1470 const struct cfm *cfm;
1472 iface = iface_find(args);
1474 unixctl_command_reply(conn, 501, "no such interface");
1478 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1481 unixctl_command_reply(conn, 501, "CFM not enabled");
1485 cfm_dump_ds(cfm, &ds);
1486 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1490 /* QoS unixctl user interface functions. */
1492 struct qos_unixctl_show_cbdata {
1494 struct iface *iface;
1498 qos_unixctl_show_cb(unsigned int queue_id,
1499 const struct shash *details,
1502 struct qos_unixctl_show_cbdata *data = aux;
1503 struct ds *ds = data->ds;
1504 struct iface *iface = data->iface;
1505 struct netdev_queue_stats stats;
1506 struct shash_node *node;
1509 ds_put_cstr(ds, "\n");
1511 ds_put_format(ds, "Queue %u:\n", queue_id);
1513 ds_put_cstr(ds, "Default:\n");
1516 SHASH_FOR_EACH (node, details) {
1517 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1520 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1522 if (stats.tx_packets != UINT64_MAX) {
1523 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1526 if (stats.tx_bytes != UINT64_MAX) {
1527 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1530 if (stats.tx_errors != UINT64_MAX) {
1531 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1534 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1535 queue_id, strerror(error));
1540 qos_unixctl_show(struct unixctl_conn *conn,
1541 const char *args, void *aux OVS_UNUSED)
1543 struct ds ds = DS_EMPTY_INITIALIZER;
1544 struct shash sh = SHASH_INITIALIZER(&sh);
1545 struct iface *iface;
1547 struct shash_node *node;
1548 struct qos_unixctl_show_cbdata data;
1551 iface = iface_find(args);
1553 unixctl_command_reply(conn, 501, "no such interface");
1557 netdev_get_qos(iface->netdev, &type, &sh);
1559 if (*type != '\0') {
1560 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1562 SHASH_FOR_EACH (node, &sh) {
1563 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1568 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1571 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1573 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1575 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1576 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1579 shash_destroy_free_data(&sh);
1583 /* Bridge reconfiguration functions. */
1584 static struct bridge *
1585 bridge_create(const struct ovsrec_bridge *br_cfg)
1590 assert(!bridge_lookup(br_cfg->name));
1591 br = xzalloc(sizeof *br);
1593 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1600 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1603 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1605 dpif_delete(br->dpif);
1606 dpif_close(br->dpif);
1611 br->name = xstrdup(br_cfg->name);
1613 br->ml = mac_learning_create();
1614 eth_addr_nicira_random(br->default_ea);
1616 hmap_init(&br->ports);
1617 hmap_init(&br->ifaces);
1618 shash_init(&br->iface_by_name);
1622 list_push_back(&all_bridges, &br->node);
1624 VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
1630 bridge_destroy(struct bridge *br)
1633 struct port *port, *next;
1636 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1639 list_remove(&br->node);
1640 ofproto_destroy(br->ofproto);
1641 error = dpif_delete(br->dpif);
1642 if (error && error != ENOENT) {
1643 VLOG_ERR("failed to delete %s: %s",
1644 dpif_name(br->dpif), strerror(error));
1646 dpif_close(br->dpif);
1647 mac_learning_destroy(br->ml);
1648 hmap_destroy(&br->ifaces);
1649 hmap_destroy(&br->ports);
1650 shash_destroy(&br->iface_by_name);
1656 static struct bridge *
1657 bridge_lookup(const char *name)
1661 LIST_FOR_EACH (br, node, &all_bridges) {
1662 if (!strcmp(br->name, name)) {
1669 /* Handle requests for a listing of all flows known by the OpenFlow
1670 * stack, including those normally hidden. */
1672 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1673 const char *args, void *aux OVS_UNUSED)
1678 br = bridge_lookup(args);
1680 unixctl_command_reply(conn, 501, "Unknown bridge");
1685 ofproto_get_all_flows(br->ofproto, &results);
1687 unixctl_command_reply(conn, 200, ds_cstr(&results));
1688 ds_destroy(&results);
1691 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1692 * connections and reconnect. If BRIDGE is not specified, then all bridges
1693 * drop their controller connections and reconnect. */
1695 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1696 const char *args, void *aux OVS_UNUSED)
1699 if (args[0] != '\0') {
1700 br = bridge_lookup(args);
1702 unixctl_command_reply(conn, 501, "Unknown bridge");
1705 ofproto_reconnect_controllers(br->ofproto);
1707 LIST_FOR_EACH (br, node, &all_bridges) {
1708 ofproto_reconnect_controllers(br->ofproto);
1711 unixctl_command_reply(conn, 200, NULL);
1715 bridge_run_one(struct bridge *br)
1720 error = ofproto_run1(br->ofproto);
1725 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1727 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1731 error = ofproto_run2(br->ofproto, br->flush);
1738 bridge_get_controllers(const struct bridge *br,
1739 struct ovsrec_controller ***controllersp)
1741 struct ovsrec_controller **controllers;
1742 size_t n_controllers;
1744 controllers = br->cfg->controller;
1745 n_controllers = br->cfg->n_controller;
1747 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1753 *controllersp = controllers;
1755 return n_controllers;
1759 bridge_reconfigure_one(struct bridge *br)
1761 enum ofproto_fail_mode fail_mode;
1762 struct port *port, *next;
1763 struct shash_node *node;
1764 struct shash new_ports;
1767 /* Collect new ports. */
1768 shash_init(&new_ports);
1769 for (i = 0; i < br->cfg->n_ports; i++) {
1770 const char *name = br->cfg->ports[i]->name;
1771 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1772 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1777 /* If we have a controller, then we need a local port. Complain if the
1778 * user didn't specify one.
1780 * XXX perhaps we should synthesize a port ourselves in this case. */
1781 if (bridge_get_controllers(br, NULL)) {
1782 char local_name[IF_NAMESIZE];
1785 error = dpif_port_get_name(br->dpif, ODPP_LOCAL,
1786 local_name, sizeof local_name);
1787 if (!error && !shash_find(&new_ports, local_name)) {
1788 VLOG_WARN("bridge %s: controller specified but no local port "
1789 "(port named %s) defined",
1790 br->name, local_name);
1794 /* Get rid of deleted ports.
1795 * Get rid of deleted interfaces on ports that still exist. */
1796 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1797 const struct ovsrec_port *port_cfg;
1799 port_cfg = shash_find_data(&new_ports, port->name);
1803 port_del_ifaces(port, port_cfg);
1807 /* Create new ports.
1808 * Add new interfaces to existing ports.
1809 * Reconfigure existing ports. */
1810 SHASH_FOR_EACH (node, &new_ports) {
1811 struct port *port = port_lookup(br, node->name);
1813 port = port_create(br, node->name);
1816 port_reconfigure(port, node->data);
1817 if (list_is_empty(&port->ifaces)) {
1818 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1819 br->name, port->name);
1823 shash_destroy(&new_ports);
1825 /* Set the fail-mode */
1826 fail_mode = !br->cfg->fail_mode
1827 || !strcmp(br->cfg->fail_mode, "standalone")
1828 ? OFPROTO_FAIL_STANDALONE
1829 : OFPROTO_FAIL_SECURE;
1830 if (ofproto_get_fail_mode(br->ofproto) != fail_mode
1831 && !ofproto_has_primary_controller(br->ofproto)) {
1832 ofproto_flush_flows(br->ofproto);
1834 ofproto_set_fail_mode(br->ofproto, fail_mode);
1836 /* Delete all flows if we're switching from connected to standalone or vice
1837 * versa. (XXX Should we delete all flows if we are switching from one
1838 * controller to another?) */
1840 /* Configure OpenFlow controller connection snooping. */
1841 if (!ofproto_has_snoops(br->ofproto)) {
1845 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1846 ovs_rundir(), br->name));
1847 ofproto_set_snoops(br->ofproto, &snoops);
1848 sset_destroy(&snoops);
1851 mirror_reconfigure(br);
1854 /* Initializes 'oc' appropriately as a management service controller for
1857 * The caller must free oc->target when it is no longer needed. */
1859 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1860 struct ofproto_controller *oc)
1862 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1863 oc->max_backoff = 0;
1864 oc->probe_interval = 60;
1865 oc->band = OFPROTO_OUT_OF_BAND;
1867 oc->burst_limit = 0;
1870 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1872 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1873 struct ofproto_controller *oc)
1875 oc->target = c->target;
1876 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1877 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1878 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1879 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1880 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1881 oc->burst_limit = (c->controller_burst_limit
1882 ? *c->controller_burst_limit : 0);
1885 /* Configures the IP stack for 'br''s local interface properly according to the
1886 * configuration in 'c'. */
1888 bridge_configure_local_iface_netdev(struct bridge *br,
1889 struct ovsrec_controller *c)
1891 struct netdev *netdev;
1892 struct in_addr mask, gateway;
1894 struct iface *local_iface;
1897 /* If there's no local interface or no IP address, give up. */
1898 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1899 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1903 /* Bring up the local interface. */
1904 netdev = local_iface->netdev;
1905 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1907 /* Configure the IP address and netmask. */
1908 if (!c->local_netmask
1909 || !inet_aton(c->local_netmask, &mask)
1911 mask.s_addr = guess_netmask(ip.s_addr);
1913 if (!netdev_set_in4(netdev, ip, mask)) {
1914 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1915 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1918 /* Configure the default gateway. */
1919 if (c->local_gateway
1920 && inet_aton(c->local_gateway, &gateway)
1921 && gateway.s_addr) {
1922 if (!netdev_add_router(netdev, gateway)) {
1923 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1924 br->name, IP_ARGS(&gateway.s_addr));
1930 bridge_reconfigure_remotes(struct bridge *br,
1931 const struct sockaddr_in *managers,
1934 const char *disable_ib_str, *queue_id_str;
1935 bool disable_in_band = false;
1938 struct ovsrec_controller **controllers;
1939 size_t n_controllers;
1942 struct ofproto_controller *ocs;
1946 /* Check if we should disable in-band control on this bridge. */
1947 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1948 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1949 disable_in_band = true;
1952 /* Set OpenFlow queue ID for in-band control. */
1953 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1954 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1955 ofproto_set_in_band_queue(br->ofproto, queue_id);
1957 if (disable_in_band) {
1958 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1960 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1962 had_primary = ofproto_has_primary_controller(br->ofproto);
1964 n_controllers = bridge_get_controllers(br, &controllers);
1966 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1969 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1970 for (i = 0; i < n_controllers; i++) {
1971 struct ovsrec_controller *c = controllers[i];
1973 if (!strncmp(c->target, "punix:", 6)
1974 || !strncmp(c->target, "unix:", 5)) {
1975 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1977 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
1978 * domain sockets and overwriting arbitrary local files. */
1979 VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
1980 "\"%s\" due to possibility for remote exploit",
1981 dpif_name(br->dpif), c->target);
1985 bridge_configure_local_iface_netdev(br, c);
1986 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
1987 if (disable_in_band) {
1988 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
1993 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
1994 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
1997 if (had_primary != ofproto_has_primary_controller(br->ofproto)) {
1998 ofproto_flush_flows(br->ofproto);
2001 /* If there are no controllers and the bridge is in standalone
2002 * mode, set up a flow that matches every packet and directs
2003 * them to OFPP_NORMAL (which goes to us). Otherwise, the
2004 * switch is in secure mode and we won't pass any traffic until
2005 * a controller has been defined and it tells us to do so. */
2007 && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) {
2008 union ofp_action action;
2009 struct cls_rule rule;
2011 memset(&action, 0, sizeof action);
2012 action.type = htons(OFPAT_OUTPUT);
2013 action.output.len = htons(sizeof action);
2014 action.output.port = htons(OFPP_NORMAL);
2015 cls_rule_init_catchall(&rule, 0);
2016 ofproto_add_flow(br->ofproto, &rule, &action, 1);
2021 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2026 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2027 struct iface *iface;
2029 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2030 shash_add_once(ifaces, iface->name, iface);
2032 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2033 shash_add_once(ifaces, port->name, NULL);
2038 /* For robustness, in case the administrator moves around datapath ports behind
2039 * our back, we re-check all the datapath port numbers here.
2041 * This function will set the 'dp_ifidx' members of interfaces that have
2042 * disappeared to -1, so only call this function from a context where those
2043 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2044 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2045 * datapath, which doesn't support UINT16_MAX+1 ports. */
2047 bridge_fetch_dp_ifaces(struct bridge *br)
2049 struct dpif_port_dump dump;
2050 struct dpif_port dpif_port;
2053 /* Reset all interface numbers. */
2054 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2055 struct iface *iface;
2057 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2058 iface->dp_ifidx = -1;
2061 hmap_clear(&br->ifaces);
2063 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2064 struct iface *iface = iface_lookup(br, dpif_port.name);
2066 if (iface->dp_ifidx >= 0) {
2067 VLOG_WARN("%s reported interface %s twice",
2068 dpif_name(br->dpif), dpif_port.name);
2069 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2070 VLOG_WARN("%s reported interface %"PRIu16" twice",
2071 dpif_name(br->dpif), dpif_port.port_no);
2073 iface->dp_ifidx = dpif_port.port_no;
2074 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2075 hash_int(iface->dp_ifidx, 0));
2078 iface_set_ofport(iface->cfg,
2079 (iface->dp_ifidx >= 0
2080 ? odp_port_to_ofp_port(iface->dp_ifidx)
2086 /* Bridge packet processing functions. */
2089 set_dst(struct dst *dst, const struct flow *flow,
2090 const struct port *in_port, const struct port *out_port,
2093 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2094 : in_port->vlan >= 0 ? in_port->vlan
2095 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2096 : vlan_tci_to_vid(flow->vlan_tci));
2098 dst->iface = (!out_port->bond
2099 ? port_get_an_iface(out_port)
2100 : bond_choose_output_slave(out_port->bond, flow,
2103 return dst->iface != NULL;
2107 mirror_mask_ffs(mirror_mask_t mask)
2109 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2114 dst_set_init(struct dst_set *set)
2116 set->dsts = set->builtin;
2118 set->allocated = ARRAY_SIZE(set->builtin);
2122 dst_set_add(struct dst_set *set, const struct dst *dst)
2124 if (set->n >= set->allocated) {
2125 size_t new_allocated;
2126 struct dst *new_dsts;
2128 new_allocated = set->allocated * 2;
2129 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2130 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2134 set->dsts = new_dsts;
2135 set->allocated = new_allocated;
2137 set->dsts[set->n++] = *dst;
2141 dst_set_free(struct dst_set *set)
2143 if (set->dsts != set->builtin) {
2149 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2152 for (i = 0; i < set->n; i++) {
2153 if (set->dsts[i].vlan == test->vlan
2154 && set->dsts[i].iface == test->iface) {
2162 port_trunks_vlan(const struct port *port, uint16_t vlan)
2164 return (port->vlan < 0
2165 && (!port->trunks || bitmap_is_set(port->trunks, vlan)));
2169 port_includes_vlan(const struct port *port, uint16_t vlan)
2171 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2175 port_is_floodable(const struct port *port)
2177 struct iface *iface;
2179 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2180 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2188 /* Returns an arbitrary interface within 'port'. */
2189 static struct iface *
2190 port_get_an_iface(const struct port *port)
2192 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2196 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2197 const struct port *in_port, const struct port *out_port,
2198 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2202 if (out_port == FLOOD_PORT) {
2205 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2207 && port_is_floodable(port)
2208 && port_includes_vlan(port, vlan)
2209 && !port->is_mirror_output_port
2210 && set_dst(&dst, flow, in_port, port, tags)) {
2211 dst_set_add(set, &dst);
2214 *nf_output_iface = NF_OUT_FLOOD;
2215 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2216 dst_set_add(set, &dst);
2217 *nf_output_iface = dst.iface->dp_ifidx;
2222 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2223 uint16_t vlan, const struct port *in_port,
2224 struct dst_set *set, tag_type *tags)
2226 mirror_mask_t mirrors;
2230 mirrors = in_port->src_mirrors;
2231 for (i = 0; i < set->n; i++) {
2232 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2239 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2240 if (flow_vlan == 0) {
2241 flow_vlan = OFP_VLAN_NONE;
2245 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2246 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2250 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2251 && !dst_is_duplicate(set, &dst)) {
2252 dst_set_add(set, &dst);
2257 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2258 if (port_includes_vlan(port, m->out_vlan)
2259 && set_dst(&dst, flow, in_port, port, tags))
2261 if (port->vlan < 0) {
2262 dst.vlan = m->out_vlan;
2264 if (dst_is_duplicate(set, &dst)) {
2268 /* Use the vlan tag on the original flow instead of
2269 * the one passed in the vlan parameter. This ensures
2270 * that we compare the vlan from before any implicit
2271 * tagging tags place. This is necessary because
2272 * dst->vlan is the final vlan, after removing implicit
2274 if (port == in_port && dst.vlan == flow_vlan) {
2275 /* Don't send out input port on same VLAN. */
2278 dst_set_add(set, &dst);
2283 mirrors &= mirrors - 1;
2288 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2289 const struct port *in_port, const struct port *out_port,
2290 tag_type *tags, struct ofpbuf *actions,
2291 uint16_t *nf_output_iface)
2293 uint16_t initial_vlan, cur_vlan;
2294 const struct dst *dst;
2298 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2300 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2302 /* Output all the packets we can without having to change the VLAN. */
2303 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2304 if (initial_vlan == 0) {
2305 initial_vlan = OFP_VLAN_NONE;
2307 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2308 if (dst->vlan != initial_vlan) {
2311 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2314 /* Then output the rest. */
2315 cur_vlan = initial_vlan;
2316 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2317 if (dst->vlan == initial_vlan) {
2320 if (dst->vlan != cur_vlan) {
2321 if (dst->vlan == OFP_VLAN_NONE) {
2322 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2325 tci = htons(dst->vlan & VLAN_VID_MASK);
2326 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2327 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2329 cur_vlan = dst->vlan;
2331 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2337 /* Returns the effective vlan of a packet, taking into account both the
2338 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2339 * the packet is untagged and -1 indicates it has an invalid header and
2340 * should be dropped. */
2341 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2342 struct port *in_port, bool have_packet)
2344 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2345 if (in_port->vlan >= 0) {
2348 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2349 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2350 "packet received on port %s configured with "
2351 "implicit VLAN %"PRIu16,
2352 br->name, vlan, in_port->name, in_port->vlan);
2356 vlan = in_port->vlan;
2358 if (!port_includes_vlan(in_port, vlan)) {
2360 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2361 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2362 "packet received on port %s not configured for "
2364 br->name, vlan, in_port->name, vlan);
2373 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2374 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2375 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2377 is_gratuitous_arp(const struct flow *flow)
2379 return (flow->dl_type == htons(ETH_TYPE_ARP)
2380 && eth_addr_is_broadcast(flow->dl_dst)
2381 && (flow->nw_proto == ARP_OP_REPLY
2382 || (flow->nw_proto == ARP_OP_REQUEST
2383 && flow->nw_src == flow->nw_dst)));
2387 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2388 struct port *in_port)
2390 struct mac_entry *mac;
2392 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2396 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2397 if (is_gratuitous_arp(flow)) {
2398 /* We don't want to learn from gratuitous ARP packets that are
2399 * reflected back over bond slaves so we lock the learning table. */
2400 if (!in_port->bond) {
2401 mac_entry_set_grat_arp_lock(mac);
2402 } else if (mac_entry_is_grat_arp_locked(mac)) {
2407 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2408 /* The log messages here could actually be useful in debugging,
2409 * so keep the rate limit relatively high. */
2410 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2411 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2412 "on port %s in VLAN %d",
2413 br->name, ETH_ADDR_ARGS(flow->dl_src),
2414 in_port->name, vlan);
2416 mac->port.p = in_port;
2417 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2421 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2422 * dropped. Returns true if they may be forwarded, false if they should be
2425 * If 'have_packet' is true, it indicates that the caller is processing a
2426 * received packet. If 'have_packet' is false, then the caller is just
2427 * revalidating an existing flow because configuration has changed. Either
2428 * way, 'have_packet' only affects logging (there is no point in logging errors
2429 * during revalidation).
2431 * Sets '*in_portp' to the input port. This will be a null pointer if
2432 * flow->in_port does not designate a known input port (in which case
2433 * is_admissible() returns false).
2435 * When returning true, sets '*vlanp' to the effective VLAN of the input
2436 * packet, as returned by flow_get_vlan().
2438 * May also add tags to '*tags', although the current implementation only does
2439 * so in one special case.
2442 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2443 tag_type *tags, int *vlanp, struct port **in_portp)
2445 struct iface *in_iface;
2446 struct port *in_port;
2449 /* Find the interface and port structure for the received packet. */
2450 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2452 /* No interface? Something fishy... */
2454 /* Odd. A few possible reasons here:
2456 * - We deleted an interface but there are still a few packets
2457 * queued up from it.
2459 * - Someone externally added an interface (e.g. with "ovs-dpctl
2460 * add-if") that we don't know about.
2462 * - Packet arrived on the local port but the local port is not
2463 * one of our bridge ports.
2465 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2467 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2468 "interface %"PRIu16, br->name, flow->in_port);
2474 *in_portp = in_port = in_iface->port;
2475 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2480 /* Drop frames for reserved multicast addresses. */
2481 if (eth_addr_is_reserved(flow->dl_dst)) {
2485 /* Drop frames on ports reserved for mirroring. */
2486 if (in_port->is_mirror_output_port) {
2488 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2489 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2490 "%s, which is reserved exclusively for mirroring",
2491 br->name, in_port->name);
2496 if (in_port->bond) {
2497 struct mac_entry *mac;
2499 switch (bond_check_admissibility(in_port->bond, in_iface,
2500 flow->dl_dst, tags)) {
2507 case BV_DROP_IF_MOVED:
2508 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2509 if (mac && mac->port.p != in_port &&
2510 (!is_gratuitous_arp(flow)
2511 || mac_entry_is_grat_arp_locked(mac))) {
2521 /* If the composed actions may be applied to any packet in the given 'flow',
2522 * returns true. Otherwise, the actions should only be applied to 'packet', or
2523 * not at all, if 'packet' was NULL. */
2525 process_flow(struct bridge *br, const struct flow *flow,
2526 const struct ofpbuf *packet, struct ofpbuf *actions,
2527 tag_type *tags, uint16_t *nf_output_iface)
2529 struct port *in_port;
2530 struct port *out_port;
2531 struct mac_entry *mac;
2534 /* Check whether we should drop packets in this flow. */
2535 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2540 /* Learn source MAC (but don't try to learn from revalidation). */
2542 update_learning_table(br, flow, vlan, in_port);
2545 /* Determine output port. */
2546 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2548 out_port = mac->port.p;
2549 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2550 /* If we are revalidating but don't have a learning entry then
2551 * eject the flow. Installing a flow that floods packets opens
2552 * up a window of time where we could learn from a packet reflected
2553 * on a bond and blackhole packets before the learning table is
2554 * updated to reflect the correct port. */
2557 out_port = FLOOD_PORT;
2560 /* Don't send packets out their input ports. */
2561 if (in_port == out_port) {
2567 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2575 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2576 struct ofpbuf *actions, tag_type *tags,
2577 uint16_t *nf_output_iface, void *br_)
2579 struct bridge *br = br_;
2581 COVERAGE_INC(bridge_process_flow);
2582 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2586 bridge_special_ofhook_cb(const struct flow *flow,
2587 const struct ofpbuf *packet, void *br_)
2589 struct iface *iface;
2590 struct bridge *br = br_;
2592 iface = iface_from_dp_ifidx(br, flow->in_port);
2594 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2595 if (iface && iface->port->bond && packet) {
2596 bond_process_lacp(iface->port->bond, iface, packet);
2605 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2606 const struct nlattr *actions,
2608 uint64_t n_bytes, void *br_)
2610 struct bridge *br = br_;
2611 const struct nlattr *a;
2612 struct port *in_port;
2617 /* Feed information from the active flows back into the learning table to
2618 * ensure that table is always in sync with what is actually flowing
2619 * through the datapath.
2621 * We test that 'tags' is nonzero to ensure that only flows that include an
2622 * OFPP_NORMAL action are used for learning. This works because
2623 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2624 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2625 update_learning_table(br, flow, vlan, in_port);
2628 /* Account for bond slave utilization. */
2629 if (!br->has_bonded_ports) {
2632 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2633 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2634 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2635 if (out_port && out_port->bond) {
2636 uint16_t vlan = (flow->vlan_tci
2637 ? vlan_tci_to_vid(flow->vlan_tci)
2639 bond_account(out_port->bond, flow, vlan, n_bytes);
2646 bridge_account_checkpoint_ofhook_cb(void *br_)
2648 struct bridge *br = br_;
2651 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2653 bond_rebalance(port->bond,
2654 ofproto_get_revalidate_set(br->ofproto));
2659 static struct ofhooks bridge_ofhooks = {
2660 bridge_normal_ofhook_cb,
2661 bridge_special_ofhook_cb,
2662 bridge_account_flow_ofhook_cb,
2663 bridge_account_checkpoint_ofhook_cb,
2666 /* Port functions. */
2669 port_run(struct port *port)
2672 bond_run(port->bond,
2673 ofproto_get_revalidate_set(port->bridge->ofproto));
2674 if (bond_should_send_learning_packets(port->bond)) {
2675 port_send_learning_packets(port);
2681 port_wait(struct port *port)
2684 bond_wait(port->bond);
2688 static struct port *
2689 port_create(struct bridge *br, const char *name)
2693 port = xzalloc(sizeof *port);
2696 port->trunks = NULL;
2697 port->name = xstrdup(name);
2698 list_init(&port->ifaces);
2700 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2702 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2709 get_port_other_config(const struct ovsrec_port *port, const char *key,
2710 const char *default_value)
2714 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2716 return value ? value : default_value;
2720 get_interface_other_config(const struct ovsrec_interface *iface,
2721 const char *key, const char *default_value)
2725 value = get_ovsrec_key_value(&iface->header_,
2726 &ovsrec_interface_col_other_config, key);
2727 return value ? value : default_value;
2731 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2733 struct iface *iface, *next;
2734 struct sset new_ifaces;
2737 /* Collect list of new interfaces. */
2738 sset_init(&new_ifaces);
2739 for (i = 0; i < cfg->n_interfaces; i++) {
2740 const char *name = cfg->interfaces[i]->name;
2741 sset_add(&new_ifaces, name);
2744 /* Get rid of deleted interfaces. */
2745 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2746 if (!sset_contains(&new_ifaces, iface->name)) {
2747 iface_destroy(iface);
2751 sset_destroy(&new_ifaces);
2754 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2755 * to revalidate every flow. */
2757 port_flush_macs(struct port *port)
2759 struct bridge *br = port->bridge;
2760 struct mac_learning *ml = br->ml;
2761 struct mac_entry *mac, *next_mac;
2764 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2765 if (mac->port.p == port) {
2766 mac_learning_expire(ml, mac);
2772 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2774 struct sset new_ifaces;
2775 bool need_flush = false;
2776 unsigned long *trunks;
2783 /* Add new interfaces and update 'cfg' member of existing ones. */
2784 sset_init(&new_ifaces);
2785 for (i = 0; i < cfg->n_interfaces; i++) {
2786 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2787 struct iface *iface;
2789 if (!sset_add(&new_ifaces, if_cfg->name)) {
2790 VLOG_WARN("port %s: %s specified twice as port interface",
2791 port->name, if_cfg->name);
2792 iface_set_ofport(if_cfg, -1);
2796 iface = iface_lookup(port->bridge, if_cfg->name);
2798 if (iface->port != port) {
2799 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2801 port->bridge->name, if_cfg->name, iface->port->name);
2804 iface->cfg = if_cfg;
2806 iface = iface_create(port, if_cfg);
2809 /* Determine interface type. The local port always has type
2810 * "internal". Other ports take their type from the database and
2811 * default to "system" if none is specified. */
2812 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2813 : if_cfg->type[0] ? if_cfg->type
2816 sset_destroy(&new_ifaces);
2821 if (list_is_short(&port->ifaces)) {
2823 if (vlan >= 0 && vlan <= 4095) {
2824 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2829 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2830 * they even work as-is. But they have not been tested. */
2831 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2835 if (port->vlan != vlan) {
2840 /* Get trunked VLANs. */
2842 if (vlan < 0 && cfg->n_trunks) {
2845 trunks = bitmap_allocate(4096);
2847 for (i = 0; i < cfg->n_trunks; i++) {
2848 int trunk = cfg->trunks[i];
2850 bitmap_set1(trunks, trunk);
2856 VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
2857 port->name, cfg->n_trunks);
2859 if (n_errors == cfg->n_trunks) {
2860 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2862 bitmap_free(trunks);
2865 } else if (vlan >= 0 && cfg->n_trunks) {
2866 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2870 ? port->trunks != NULL
2871 : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
2874 bitmap_free(port->trunks);
2875 port->trunks = trunks;
2878 port_flush_macs(port);
2883 port_destroy(struct port *port)
2886 struct bridge *br = port->bridge;
2887 struct iface *iface, *next;
2890 for (i = 0; i < MAX_MIRRORS; i++) {
2891 struct mirror *m = br->mirrors[i];
2892 if (m && m->out_port == port) {
2897 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2898 iface_destroy(iface);
2901 hmap_remove(&br->ports, &port->hmap_node);
2903 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2905 port_flush_macs(port);
2907 bitmap_free(port->trunks);
2913 static struct port *
2914 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
2916 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
2917 return iface ? iface->port : NULL;
2920 static struct port *
2921 port_lookup(const struct bridge *br, const char *name)
2925 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2927 if (!strcmp(port->name, name)) {
2935 enable_lacp(struct port *port, bool *activep)
2937 if (!port->cfg->lacp) {
2938 /* XXX when LACP implementation has been sufficiently tested, enable by
2939 * default and make active on bonded ports. */
2941 } else if (!strcmp(port->cfg->lacp, "off")) {
2943 } else if (!strcmp(port->cfg->lacp, "active")) {
2946 } else if (!strcmp(port->cfg->lacp, "passive")) {
2950 VLOG_WARN("port %s: unknown LACP mode %s",
2951 port->name, port->cfg->lacp);
2956 static struct lacp_settings *
2957 port_reconfigure_bond_lacp(struct port *port, struct lacp_settings *s)
2959 if (!enable_lacp(port, &s->active)) {
2963 s->name = port->name;
2964 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2965 s->priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2967 s->fast = !strcmp(get_port_other_config(port->cfg, "lacp-time", "slow"),
2970 if (s->priority <= 0 || s->priority > UINT16_MAX) {
2971 /* Prefer bondable links if unspecified. */
2972 s->priority = UINT16_MAX - !list_is_short(&port->ifaces);
2978 iface_reconfigure_bond(struct iface *iface)
2980 struct lacp_slave_settings s;
2983 s.name = iface->name;
2984 s.id = iface->dp_ifidx;
2985 priority = atoi(get_interface_other_config(
2986 iface->cfg, "lacp-port-priority", "0"));
2987 s.priority = (priority >= 0 && priority <= UINT16_MAX
2988 ? priority : UINT16_MAX);
2989 bond_slave_register(iface->port->bond, iface, iface->netdev, &s);
2993 port_reconfigure_bond(struct port *port)
2995 struct lacp_settings lacp_settings;
2996 struct bond_settings s;
2997 const char *detect_s;
2998 struct iface *iface;
3000 if (list_is_short(&port->ifaces)) {
3001 /* Not a bonded port. */
3002 bond_destroy(port->bond);
3007 port->bridge->has_bonded_ports = true;
3009 s.name = port->name;
3011 if (port->cfg->bond_mode
3012 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3013 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3014 port->name, port->cfg->bond_mode,
3015 bond_mode_to_string(s.balance));
3018 s.detect = BLSM_CARRIER;
3019 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3020 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3021 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3023 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3026 s.miimon_interval = atoi(
3027 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3028 if (s.miimon_interval < 100) {
3029 s.miimon_interval = 100;
3032 s.up_delay = MAX(0, port->cfg->bond_updelay);
3033 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3034 s.rebalance_interval = atoi(
3035 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3036 if (s.rebalance_interval < 1000) {
3037 s.rebalance_interval = 1000;
3040 s.fake_iface = port->cfg->bond_fake_iface;
3041 s.lacp = port_reconfigure_bond_lacp(port, &lacp_settings);
3044 port->bond = bond_create(&s);
3046 bond_reconfigure(port->bond, &s);
3049 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3050 iface_reconfigure_bond(iface);
3055 port_send_learning_packets(struct port *port)
3057 struct bridge *br = port->bridge;
3058 int error, n_packets, n_errors;
3059 struct mac_entry *e;
3061 error = n_packets = n_errors = 0;
3062 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3063 if (e->port.p != port) {
3064 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3074 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3075 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3076 "packets, last error was: %s",
3077 port->name, n_errors, n_packets, strerror(error));
3079 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3080 port->name, n_packets);
3084 /* Interface functions. */
3086 static struct iface *
3087 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3089 struct bridge *br = port->bridge;
3090 struct iface *iface;
3091 char *name = if_cfg->name;
3093 iface = xzalloc(sizeof *iface);
3095 iface->name = xstrdup(name);
3096 iface->dp_ifidx = -1;
3097 iface->tag = tag_create_random();
3098 iface->netdev = NULL;
3099 iface->cfg = if_cfg;
3101 shash_add_assert(&br->iface_by_name, iface->name, iface);
3103 list_push_back(&port->ifaces, &iface->port_elem);
3105 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3113 iface_destroy(struct iface *iface)
3116 struct port *port = iface->port;
3117 struct bridge *br = port->bridge;
3120 bond_slave_unregister(port->bond, iface);
3123 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3125 if (iface->dp_ifidx >= 0) {
3126 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3129 list_remove(&iface->port_elem);
3131 netdev_close(iface->netdev);
3136 bridge_flush(port->bridge);
3140 static struct iface *
3141 iface_lookup(const struct bridge *br, const char *name)
3143 return shash_find_data(&br->iface_by_name, name);
3146 static struct iface *
3147 iface_find(const char *name)
3149 const struct bridge *br;
3151 LIST_FOR_EACH (br, node, &all_bridges) {
3152 struct iface *iface = iface_lookup(br, name);
3161 static struct iface *
3162 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3164 struct iface *iface;
3166 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3167 hash_int(dp_ifidx, 0), &br->ifaces) {
3168 if (iface->dp_ifidx == dp_ifidx) {
3175 /* Set Ethernet address of 'iface', if one is specified in the configuration
3178 iface_set_mac(struct iface *iface)
3180 uint8_t ea[ETH_ADDR_LEN];
3182 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3183 if (eth_addr_is_multicast(ea)) {
3184 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3186 } else if (iface->dp_ifidx == ODPP_LOCAL) {
3187 VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead",
3188 iface->name, iface->name);
3190 int error = netdev_set_etheraddr(iface->netdev, ea);
3192 VLOG_ERR("interface %s: setting MAC failed (%s)",
3193 iface->name, strerror(error));
3199 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3201 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3204 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3208 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3210 * The value strings in '*shash' are taken directly from values[], not copied,
3211 * so the caller should not modify or free them. */
3213 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3214 struct shash *shash)
3219 for (i = 0; i < n; i++) {
3220 shash_add(shash, keys[i], values[i]);
3224 /* Creates 'keys' and 'values' arrays from 'shash'.
3226 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3227 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3228 * are populated with with strings taken directly from 'shash' and thus have
3229 * the same ownership of the key-value pairs in shash.
3232 shash_to_ovs_idl_map(struct shash *shash,
3233 char ***keys, char ***values, size_t *n)
3237 struct shash_node *sn;
3239 count = shash_count(shash);
3241 k = xmalloc(count * sizeof *k);
3242 v = xmalloc(count * sizeof *v);
3245 SHASH_FOR_EACH(sn, shash) {
3256 struct iface_delete_queues_cbdata {
3257 struct netdev *netdev;
3258 const struct ovsdb_datum *queues;
3262 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3264 union ovsdb_atom atom;
3266 atom.integer = target;
3267 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3271 iface_delete_queues(unsigned int queue_id,
3272 const struct shash *details OVS_UNUSED, void *cbdata_)
3274 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3276 if (!queue_ids_include(cbdata->queues, queue_id)) {
3277 netdev_delete_queue(cbdata->netdev, queue_id);
3282 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3284 if (!qos || qos->type[0] == '\0') {
3285 netdev_set_qos(iface->netdev, NULL, NULL);
3287 struct iface_delete_queues_cbdata cbdata;
3288 struct shash details;
3291 /* Configure top-level Qos for 'iface'. */
3292 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3293 qos->n_other_config, &details);
3294 netdev_set_qos(iface->netdev, qos->type, &details);
3295 shash_destroy(&details);
3297 /* Deconfigure queues that were deleted. */
3298 cbdata.netdev = iface->netdev;
3299 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3301 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3303 /* Configure queues for 'iface'. */
3304 for (i = 0; i < qos->n_queues; i++) {
3305 const struct ovsrec_queue *queue = qos->value_queues[i];
3306 unsigned int queue_id = qos->key_queues[i];
3308 shash_from_ovs_idl_map(queue->key_other_config,
3309 queue->value_other_config,
3310 queue->n_other_config, &details);
3311 netdev_set_queue(iface->netdev, queue_id, &details);
3312 shash_destroy(&details);
3318 iface_update_cfm(struct iface *iface)
3322 uint16_t *remote_mps;
3323 struct ovsrec_monitor *mon;
3324 uint8_t maid[CCM_MAID_LEN];
3326 mon = iface->cfg->monitor;
3329 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3333 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3334 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3338 cfm.mpid = mon->mpid;
3339 cfm.interval = mon->interval ? *mon->interval : 1000;
3341 memcpy(cfm.maid, maid, sizeof cfm.maid);
3343 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3344 for(i = 0; i < mon->n_remote_mps; i++) {
3345 remote_mps[i] = mon->remote_mps[i]->mpid;
3348 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3349 &cfm, remote_mps, mon->n_remote_mps);
3353 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3354 * how 'iface''s port is configured.
3356 * Returns true if 'iface' is up, false otherwise. */
3358 iface_get_carrier(const struct iface *iface)
3361 return netdev_get_carrier(iface->netdev);
3364 /* Port mirroring. */
3366 static struct mirror *
3367 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3371 for (i = 0; i < MAX_MIRRORS; i++) {
3372 struct mirror *m = br->mirrors[i];
3373 if (m && uuid_equals(uuid, &m->uuid)) {
3381 mirror_reconfigure(struct bridge *br)
3383 unsigned long *rspan_vlans;
3387 /* Get rid of deleted mirrors. */
3388 for (i = 0; i < MAX_MIRRORS; i++) {
3389 struct mirror *m = br->mirrors[i];
3391 const struct ovsdb_datum *mc;
3392 union ovsdb_atom atom;
3394 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3395 atom.uuid = br->mirrors[i]->uuid;
3396 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3402 /* Add new mirrors and reconfigure existing ones. */
3403 for (i = 0; i < br->cfg->n_mirrors; i++) {
3404 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3405 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3407 mirror_reconfigure_one(m, cfg);
3409 mirror_create(br, cfg);
3413 /* Update port reserved status. */
3414 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3415 port->is_mirror_output_port = false;
3417 for (i = 0; i < MAX_MIRRORS; i++) {
3418 struct mirror *m = br->mirrors[i];
3419 if (m && m->out_port) {
3420 m->out_port->is_mirror_output_port = true;
3424 /* Update flooded vlans (for RSPAN). */
3426 if (br->cfg->n_flood_vlans) {
3427 rspan_vlans = bitmap_allocate(4096);
3429 for (i = 0; i < br->cfg->n_flood_vlans; i++) {
3430 int64_t vlan = br->cfg->flood_vlans[i];
3431 if (vlan >= 0 && vlan < 4096) {
3432 bitmap_set1(rspan_vlans, vlan);
3433 VLOG_INFO("bridge %s: disabling learning on vlan %"PRId64,
3436 VLOG_ERR("bridge %s: invalid value %"PRId64 "for flood VLAN",
3441 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3443 mac_learning_flush(br->ml);
3448 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3453 for (i = 0; ; i++) {
3454 if (i >= MAX_MIRRORS) {
3455 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3456 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3459 if (!br->mirrors[i]) {
3464 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3466 mac_learning_flush(br->ml);
3468 br->mirrors[i] = m = xzalloc(sizeof *m);
3471 m->name = xstrdup(cfg->name);
3472 sset_init(&m->src_ports);
3473 sset_init(&m->dst_ports);
3479 mirror_reconfigure_one(m, cfg);
3483 mirror_destroy(struct mirror *m)
3486 struct bridge *br = m->bridge;
3489 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3490 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3491 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3494 sset_destroy(&m->src_ports);
3495 sset_destroy(&m->dst_ports);
3498 m->bridge->mirrors[m->idx] = NULL;
3503 mac_learning_flush(br->ml);
3508 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3513 for (i = 0; i < n_ports; i++) {
3514 const char *name = ports[i]->name;
3515 if (port_lookup(m->bridge, name)) {
3516 sset_add(names, name);
3518 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3519 "port %s", m->bridge->name, m->name, name);
3525 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3531 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3533 for (i = 0; i < cfg->n_select_vlan; i++) {
3534 int64_t vlan = cfg->select_vlan[i];
3535 if (vlan < 0 || vlan > 4095) {
3536 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3537 m->bridge->name, m->name, vlan);
3539 (*vlans)[n_vlans++] = vlan;
3546 vlan_is_mirrored(const struct mirror *m, int vlan)
3550 for (i = 0; i < m->n_vlans; i++) {
3551 if (m->vlans[i] == vlan) {
3559 port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p)
3563 for (i = 0; i < m->n_vlans; i++) {
3564 if (port_trunks_vlan(p, m->vlans[i])) {
3572 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3574 struct sset src_ports, dst_ports;
3575 mirror_mask_t mirror_bit;
3576 struct port *out_port;
3583 if (strcmp(cfg->name, m->name)) {
3585 m->name = xstrdup(cfg->name);
3588 /* Get output port. */
3589 if (cfg->output_port) {
3590 out_port = port_lookup(m->bridge, cfg->output_port->name);
3592 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3593 m->bridge->name, m->name);
3599 if (cfg->output_vlan) {
3600 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3601 "output vlan; ignoring output vlan",
3602 m->bridge->name, m->name);
3604 } else if (cfg->output_vlan) {
3606 out_vlan = *cfg->output_vlan;
3608 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3609 m->bridge->name, m->name);
3614 sset_init(&src_ports);
3615 sset_init(&dst_ports);
3616 if (cfg->select_all) {
3617 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3618 sset_add(&src_ports, port->name);
3619 sset_add(&dst_ports, port->name);
3624 /* Get ports, and drop duplicates and ports that don't exist. */
3625 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3627 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3630 /* Get all the vlans, and drop duplicate and invalid vlans. */
3631 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3634 /* Update mirror data. */
3635 if (!sset_equals(&m->src_ports, &src_ports)
3636 || !sset_equals(&m->dst_ports, &dst_ports)
3637 || m->n_vlans != n_vlans
3638 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3639 || m->out_port != out_port
3640 || m->out_vlan != out_vlan) {
3641 bridge_flush(m->bridge);
3642 mac_learning_flush(m->bridge->ml);
3644 sset_swap(&m->src_ports, &src_ports);
3645 sset_swap(&m->dst_ports, &dst_ports);
3648 m->n_vlans = n_vlans;
3649 m->out_port = out_port;
3650 m->out_vlan = out_vlan;
3653 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3654 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3655 if (sset_contains(&m->src_ports, port->name)
3658 ? port_trunks_any_mirrored_vlan(m, port)
3659 : vlan_is_mirrored(m, port->vlan)))) {
3660 port->src_mirrors |= mirror_bit;
3662 port->src_mirrors &= ~mirror_bit;
3665 if (sset_contains(&m->dst_ports, port->name)) {
3666 port->dst_mirrors |= mirror_bit;
3668 port->dst_mirrors &= ~mirror_bit;
3673 sset_destroy(&src_ports);
3674 sset_destroy(&dst_ports);