1 /* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
18 #include "byte-order.h"
21 #include <arpa/inet.h>
24 #include <sys/socket.h>
26 #include <openflow/openflow.h>
31 #include <sys/socket.h>
32 #include <sys/types.h>
37 #include "classifier.h"
42 #include "dynamic-string.h"
49 #include "mac-learning.h"
53 #include "ofp-print.h"
55 #include "ofproto/netflow.h"
56 #include "ofproto/ofproto.h"
57 #include "ovsdb-data.h"
59 #include "poll-loop.h"
63 #include "socket-util.h"
64 #include "stream-ssl.h"
67 #include "system-stats.h"
72 #include "vswitchd/vswitch-idl.h"
73 #include "xenserver.h"
75 #include "sflow_api.h"
76 #include "vlan-bitmap.h"
78 VLOG_DEFINE_THIS_MODULE(bridge);
80 COVERAGE_DEFINE(bridge_flush);
81 COVERAGE_DEFINE(bridge_process_flow);
82 COVERAGE_DEFINE(bridge_reconfigure);
90 struct dst builtin[32];
95 static void dst_set_init(struct dst_set *);
96 static void dst_set_add(struct dst_set *, const struct dst *);
97 static void dst_set_free(struct dst_set *);
100 /* These members are always valid. */
101 struct list port_elem; /* Element in struct port's "ifaces" list. */
102 struct port *port; /* Containing port. */
103 char *name; /* Host network device name. */
104 tag_type tag; /* Tag associated with this interface. */
106 /* These members are valid only after bridge_reconfigure() causes them to
108 struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
109 int dp_ifidx; /* Index within kernel datapath. */
110 struct netdev *netdev; /* Network device. */
111 const char *type; /* Usually same as cfg->type. */
112 const struct ovsrec_interface *cfg;
115 #define MAX_MIRRORS 32
116 typedef uint32_t mirror_mask_t;
117 #define MIRROR_MASK_C(X) UINT32_C(X)
118 BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS);
120 struct bridge *bridge;
123 struct uuid uuid; /* UUID of this "mirror" record in database. */
125 /* Selection criteria. */
126 struct sset src_ports; /* Source port names. */
127 struct sset dst_ports; /* Destination port names. */
132 struct port *out_port;
136 #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
138 struct bridge *bridge;
139 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
142 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
143 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
144 * NULL if all VLANs are trunked. */
145 const struct ovsrec_port *cfg;
147 /* An ordinary bridge port has 1 interface.
148 * A bridge port for bonding has at least 2 interfaces. */
149 struct list ifaces; /* List of "struct iface"s. */
151 struct lacp *lacp; /* NULL if LACP is not enabled. */
156 /* Port mirroring info. */
157 mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
158 mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
159 bool is_mirror_output_port; /* Does port mirroring send frames here? */
163 struct list node; /* Node in global list of bridges. */
164 char *name; /* User-specified arbitrary name. */
165 struct mac_learning *ml; /* MAC learning table. */
166 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
167 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
168 const struct ovsrec_bridge *cfg;
170 /* OpenFlow switch processing. */
171 struct ofproto *ofproto; /* OpenFlow switch. */
173 /* Kernel datapath information. */
174 struct dpif *dpif; /* Datapath. */
175 struct hmap ifaces; /* "struct iface"s indexed by dp_ifidx. */
178 struct hmap ports; /* "struct port"s indexed by name. */
179 struct shash iface_by_name; /* "struct iface"s indexed by name. */
182 bool has_bonded_ports;
187 /* Port mirroring. */
188 struct mirror *mirrors[MAX_MIRRORS];
190 /* Synthetic local port if necessary. */
191 struct ovsrec_port synth_local_port;
192 struct ovsrec_interface synth_local_iface;
193 struct ovsrec_interface *synth_local_ifacep;
196 /* List of all bridges. */
197 static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
199 /* OVSDB IDL used to obtain configuration. */
200 static struct ovsdb_idl *idl;
202 /* Each time this timer expires, the bridge fetches systems and interface
203 * statistics and pushes them into the database. */
204 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
205 static long long int stats_timer = LLONG_MIN;
207 /* Stores the time after which rate limited statistics may be written to the
208 * database. Only updated when changes to the database require rate limiting.
210 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
211 static long long int db_limiter = LLONG_MIN;
213 static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
214 static void bridge_destroy(struct bridge *);
215 static struct bridge *bridge_lookup(const char *name);
216 static unixctl_cb_func bridge_unixctl_dump_flows;
217 static unixctl_cb_func bridge_unixctl_reconnect;
218 static int bridge_run_one(struct bridge *);
219 static size_t bridge_get_controllers(const struct bridge *br,
220 struct ovsrec_controller ***controllersp);
221 static void bridge_reconfigure_one(struct bridge *);
222 static void bridge_reconfigure_remotes(struct bridge *,
223 const struct sockaddr_in *managers,
225 static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
226 static void bridge_fetch_dp_ifaces(struct bridge *);
227 static void bridge_flush(struct bridge *);
228 static void bridge_pick_local_hw_addr(struct bridge *,
229 uint8_t ea[ETH_ADDR_LEN],
230 struct iface **hw_addr_iface);
231 static uint64_t bridge_pick_datapath_id(struct bridge *,
232 const uint8_t bridge_ea[ETH_ADDR_LEN],
233 struct iface *hw_addr_iface);
234 static uint64_t dpid_from_hash(const void *, size_t nbytes);
236 static unixctl_cb_func bridge_unixctl_fdb_show;
237 static unixctl_cb_func cfm_unixctl_show;
238 static unixctl_cb_func qos_unixctl_show;
240 static void port_run(struct port *);
241 static void port_wait(struct port *);
242 static struct port *port_create(struct bridge *, const char *name);
243 static void port_reconfigure(struct port *, const struct ovsrec_port *);
244 static void port_del_ifaces(struct port *, const struct ovsrec_port *);
245 static void port_destroy(struct port *);
246 static struct port *port_lookup(const struct bridge *, const char *name);
247 static struct iface *port_get_an_iface(const struct port *);
248 static struct port *port_from_dp_ifidx(const struct bridge *,
250 static void port_reconfigure_lacp(struct port *);
251 static void port_reconfigure_bond(struct port *);
252 static void port_send_learning_packets(struct port *);
254 static void mirror_create(struct bridge *, struct ovsrec_mirror *);
255 static void mirror_destroy(struct mirror *);
256 static void mirror_reconfigure(struct bridge *);
257 static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
258 static bool vlan_is_mirrored(const struct mirror *, int vlan);
260 static struct iface *iface_create(struct port *port,
261 const struct ovsrec_interface *if_cfg);
262 static void iface_destroy(struct iface *);
263 static struct iface *iface_lookup(const struct bridge *, const char *name);
264 static struct iface *iface_find(const char *name);
265 static struct iface *iface_from_dp_ifidx(const struct bridge *,
267 static void iface_set_mac(struct iface *);
268 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
269 static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
270 static void iface_update_cfm(struct iface *);
271 static bool iface_refresh_cfm_stats(struct iface *iface);
272 static bool iface_get_carrier(const struct iface *);
273 static bool iface_is_synthetic(const struct iface *);
275 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
277 static void shash_to_ovs_idl_map(struct shash *,
278 char ***keys, char ***values, size_t *n);
280 /* Hooks into ofproto processing. */
281 static struct ofhooks bridge_ofhooks;
283 /* Public functions. */
285 /* Initializes the bridge module, configuring it to obtain its configuration
286 * from an OVSDB server accessed over 'remote', which should be a string in a
287 * form acceptable to ovsdb_idl_create(). */
289 bridge_init(const char *remote)
291 /* Create connection to database. */
292 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
294 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
295 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
296 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
297 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
298 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
299 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
300 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
302 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
303 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
305 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
306 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
308 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
309 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
310 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
311 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
312 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
313 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
314 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
315 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
316 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
318 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
319 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
320 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
321 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
323 ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault);
325 ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault);
327 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
329 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
331 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
333 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
335 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
337 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
338 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
339 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
340 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
341 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
343 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
345 /* Register unixctl commands. */
346 unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
347 unixctl_command_register("cfm/show", cfm_unixctl_show, NULL);
348 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
349 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
351 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
360 struct bridge *br, *next_br;
362 LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
365 ovsdb_idl_destroy(idl);
368 /* Performs configuration that is only necessary once at ovs-vswitchd startup,
369 * but for which the ovs-vswitchd configuration 'cfg' is required. */
371 bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
373 static bool already_configured_once;
374 struct sset bridge_names;
375 struct sset dpif_names, dpif_types;
379 /* Only do this once per ovs-vswitchd run. */
380 if (already_configured_once) {
383 already_configured_once = true;
385 stats_timer = time_msec() + STATS_INTERVAL;
387 /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
388 sset_init(&bridge_names);
389 for (i = 0; i < cfg->n_bridges; i++) {
390 sset_add(&bridge_names, cfg->bridges[i]->name);
393 /* Iterate over all system dpifs and delete any of them that do not appear
395 sset_init(&dpif_names);
396 sset_init(&dpif_types);
397 dp_enumerate_types(&dpif_types);
398 SSET_FOR_EACH (type, &dpif_types) {
401 dp_enumerate_names(type, &dpif_names);
403 /* Delete each dpif whose name is not in 'bridge_names'. */
404 SSET_FOR_EACH (name, &dpif_names) {
405 if (!sset_contains(&bridge_names, name)) {
409 retval = dpif_open(name, type, &dpif);
417 sset_destroy(&bridge_names);
418 sset_destroy(&dpif_names);
419 sset_destroy(&dpif_types);
422 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
423 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
424 * responsible for freeing '*managersp' (with free()).
426 * You may be asking yourself "why does ovs-vswitchd care?", because
427 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
428 * should not be and in fact is not directly involved in that. But
429 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
430 * it has to tell in-band control where the managers are to enable that.
431 * (Thus, only managers connected in-band are collected.)
434 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
435 struct sockaddr_in **managersp, size_t *n_managersp)
437 struct sockaddr_in *managers = NULL;
438 size_t n_managers = 0;
442 /* Collect all of the potential targets from the "targets" columns of the
443 * rows pointed to by "manager_options", excluding any that are
446 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
447 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
449 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
450 sset_find_and_delete(&targets, m->target);
452 sset_add(&targets, m->target);
456 /* Now extract the targets' IP addresses. */
457 if (!sset_is_empty(&targets)) {
460 managers = xmalloc(sset_count(&targets) * sizeof *managers);
461 SSET_FOR_EACH (target, &targets) {
462 struct sockaddr_in *sin = &managers[n_managers];
464 if ((!strncmp(target, "tcp:", 4)
465 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
466 (!strncmp(target, "ssl:", 4)
467 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
472 sset_destroy(&targets);
474 *managersp = managers;
475 *n_managersp = n_managers;
479 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
481 struct shash old_br, new_br;
482 struct shash_node *node;
483 struct bridge *br, *next;
484 struct sockaddr_in *managers;
487 int sflow_bridge_number;
489 COVERAGE_INC(bridge_reconfigure);
491 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
493 /* Collect old and new bridges. */
496 LIST_FOR_EACH (br, node, &all_bridges) {
497 shash_add(&old_br, br->name, br);
499 for (i = 0; i < ovs_cfg->n_bridges; i++) {
500 const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
501 if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
502 VLOG_WARN("more than one bridge named %s", br_cfg->name);
506 /* Get rid of deleted bridges and add new bridges. */
507 LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
508 struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
515 SHASH_FOR_EACH (node, &new_br) {
516 const char *br_name = node->name;
517 const struct ovsrec_bridge *br_cfg = node->data;
518 br = shash_find_data(&old_br, br_name);
520 /* If the bridge datapath type has changed, we need to tear it
521 * down and recreate. */
522 if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
524 bridge_create(br_cfg);
527 bridge_create(br_cfg);
530 shash_destroy(&old_br);
531 shash_destroy(&new_br);
533 /* Reconfigure all bridges. */
534 LIST_FOR_EACH (br, node, &all_bridges) {
535 bridge_reconfigure_one(br);
538 /* Add and delete ports on all datapaths.
540 * The kernel will reject any attempt to add a given port to a datapath if
541 * that port already belongs to a different datapath, so we must do all
542 * port deletions before any port additions. */
543 LIST_FOR_EACH (br, node, &all_bridges) {
544 struct dpif_port_dump dump;
545 struct shash want_ifaces;
546 struct dpif_port dpif_port;
548 bridge_get_all_ifaces(br, &want_ifaces);
549 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
550 if (!shash_find(&want_ifaces, dpif_port.name)
551 && strcmp(dpif_port.name, br->name)) {
552 int retval = dpif_port_del(br->dpif, dpif_port.port_no);
554 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
555 br->name, dpif_port.name, strerror(retval));
559 shash_destroy(&want_ifaces);
561 LIST_FOR_EACH (br, node, &all_bridges) {
562 struct shash cur_ifaces, want_ifaces;
563 struct dpif_port_dump dump;
564 struct dpif_port dpif_port;
566 /* Get the set of interfaces currently in this datapath. */
567 shash_init(&cur_ifaces);
568 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
569 struct dpif_port *port_info = xmalloc(sizeof *port_info);
570 dpif_port_clone(port_info, &dpif_port);
571 shash_add(&cur_ifaces, dpif_port.name, port_info);
574 /* Get the set of interfaces we want on this datapath. */
575 bridge_get_all_ifaces(br, &want_ifaces);
577 hmap_clear(&br->ifaces);
578 SHASH_FOR_EACH (node, &want_ifaces) {
579 const char *if_name = node->name;
580 struct iface *iface = node->data;
581 struct dpif_port *dpif_port;
585 type = iface ? iface->type : "internal";
586 dpif_port = shash_find_data(&cur_ifaces, if_name);
588 /* If we have a port or a netdev already, and it's not the type we
589 * want, then delete the port (if any) and close the netdev (if
591 if ((dpif_port && strcmp(dpif_port->type, type))
592 || (iface && iface->netdev
593 && strcmp(type, netdev_get_type(iface->netdev)))) {
595 error = ofproto_port_del(br->ofproto, dpif_port->port_no);
602 if (iface->port->bond) {
603 /* The bond has a pointer to the netdev, so remove it
604 * from the bond before closing the netdev. The slave
605 * will get added back to the bond later, after a new
606 * netdev is available. */
607 bond_slave_unregister(iface->port->bond, iface);
609 netdev_close(iface->netdev);
610 iface->netdev = NULL;
614 /* If the port doesn't exist or we don't have the netdev open,
615 * we need to do more work. */
616 if (!dpif_port || (iface && !iface->netdev)) {
617 struct netdev_options options;
618 struct netdev *netdev;
621 /* First open the network device. */
622 options.name = if_name;
624 options.args = &args;
625 options.ethertype = NETDEV_ETH_TYPE_NONE;
629 shash_from_ovs_idl_map(iface->cfg->key_options,
630 iface->cfg->value_options,
631 iface->cfg->n_options, &args);
633 error = netdev_open(&options, &netdev);
634 shash_destroy(&args);
637 VLOG_WARN("could not open network device %s (%s)",
638 if_name, strerror(error));
642 /* Then add the port if we haven't already. */
644 error = dpif_port_add(br->dpif, netdev, NULL);
646 netdev_close(netdev);
647 if (error == EFBIG) {
648 VLOG_ERR("bridge %s: out of valid port numbers",
652 VLOG_WARN("bridge %s: failed to add %s interface "
654 br->name, if_name, strerror(error));
660 /* Update 'iface'. */
662 iface->netdev = netdev;
664 } else if (iface && iface->netdev) {
668 shash_from_ovs_idl_map(iface->cfg->key_options,
669 iface->cfg->value_options,
670 iface->cfg->n_options, &args);
671 netdev_set_config(iface->netdev, &args);
672 shash_destroy(&args);
675 shash_destroy(&want_ifaces);
677 SHASH_FOR_EACH (node, &cur_ifaces) {
678 struct dpif_port *port_info = node->data;
679 dpif_port_destroy(port_info);
682 shash_destroy(&cur_ifaces);
684 sflow_bridge_number = 0;
685 LIST_FOR_EACH (br, node, &all_bridges) {
686 uint8_t ea[ETH_ADDR_LEN];
688 struct iface *local_iface;
689 struct port *port, *next_port;
690 struct iface *hw_addr_iface;
693 bridge_fetch_dp_ifaces(br);
695 /* Delete interfaces that cannot be opened.
697 * Following this loop, every remaining "struct iface" has nonnull
698 * 'netdev' and correct 'dp_ifidx'. */
699 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
700 struct iface *iface, *next_iface;
702 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
703 if (iface->netdev && iface->dp_ifidx >= 0) {
704 VLOG_DBG("bridge %s: interface %s is on port %d",
705 br->name, iface->name, iface->dp_ifidx);
708 VLOG_ERR("bridge %s: missing %s interface, dropping",
709 br->name, iface->name);
711 /* We already reported a related error, don't bother
715 iface_set_ofport(iface->cfg, -1);
716 iface_destroy(iface);
720 if (list_is_empty(&port->ifaces)) {
721 VLOG_WARN("%s port has no interfaces, dropping", port->name);
726 /* Pick local port hardware address, datapath ID. */
727 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
728 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
730 int error = netdev_set_etheraddr(local_iface->netdev, ea);
732 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
733 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
734 "Ethernet address: %s",
735 br->name, strerror(error));
738 memcpy(br->ea, ea, ETH_ADDR_LEN);
740 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
741 ofproto_set_datapath_id(br->ofproto, dpid);
743 dpid_string = xasprintf("%016"PRIx64, dpid);
744 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
747 /* Set NetFlow configuration on this bridge. */
748 if (br->cfg->netflow) {
749 struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
750 struct netflow_options opts;
752 memset(&opts, 0, sizeof opts);
754 dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
755 if (nf_cfg->engine_type) {
756 opts.engine_type = *nf_cfg->engine_type;
758 if (nf_cfg->engine_id) {
759 opts.engine_id = *nf_cfg->engine_id;
762 opts.active_timeout = nf_cfg->active_timeout;
763 if (!opts.active_timeout) {
764 opts.active_timeout = -1;
765 } else if (opts.active_timeout < 0) {
766 VLOG_WARN("bridge %s: active timeout interval set to negative "
767 "value, using default instead (%d seconds)", br->name,
768 NF_ACTIVE_TIMEOUT_DEFAULT);
769 opts.active_timeout = -1;
772 opts.add_id_to_iface = nf_cfg->add_id_to_interface;
773 if (opts.add_id_to_iface) {
774 if (opts.engine_id > 0x7f) {
775 VLOG_WARN("bridge %s: netflow port mangling may conflict "
776 "with another vswitch, choose an engine id less "
777 "than 128", br->name);
779 if (hmap_count(&br->ports) > 508) {
780 VLOG_WARN("bridge %s: netflow port mangling will conflict "
781 "with another port when more than 508 ports are "
786 sset_init(&opts.collectors);
787 sset_add_array(&opts.collectors,
788 nf_cfg->targets, nf_cfg->n_targets);
789 if (ofproto_set_netflow(br->ofproto, &opts)) {
790 VLOG_ERR("bridge %s: problem setting netflow collectors",
793 sset_destroy(&opts.collectors);
795 ofproto_set_netflow(br->ofproto, NULL);
798 /* Set sFlow configuration on this bridge. */
799 if (br->cfg->sflow) {
800 const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
801 struct ovsrec_controller **controllers;
802 struct ofproto_sflow_options oso;
803 size_t n_controllers;
805 memset(&oso, 0, sizeof oso);
807 sset_init(&oso.targets);
808 sset_add_array(&oso.targets,
809 sflow_cfg->targets, sflow_cfg->n_targets);
811 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
812 if (sflow_cfg->sampling) {
813 oso.sampling_rate = *sflow_cfg->sampling;
816 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
817 if (sflow_cfg->polling) {
818 oso.polling_interval = *sflow_cfg->polling;
821 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
822 if (sflow_cfg->header) {
823 oso.header_len = *sflow_cfg->header;
826 oso.sub_id = sflow_bridge_number++;
827 oso.agent_device = sflow_cfg->agent;
829 oso.control_ip = NULL;
830 n_controllers = bridge_get_controllers(br, &controllers);
831 for (i = 0; i < n_controllers; i++) {
832 if (controllers[i]->local_ip) {
833 oso.control_ip = controllers[i]->local_ip;
837 ofproto_set_sflow(br->ofproto, &oso);
839 sset_destroy(&oso.targets);
841 ofproto_set_sflow(br->ofproto, NULL);
844 /* Update the controller and related settings. It would be more
845 * straightforward to call this from bridge_reconfigure_one(), but we
846 * can't do it there for two reasons. First, and most importantly, at
847 * that point we don't know the dp_ifidx of any interfaces that have
848 * been added to the bridge (because we haven't actually added them to
849 * the datapath). Second, at that point we haven't set the datapath ID
850 * yet; when a controller is configured, resetting the datapath ID will
851 * immediately disconnect from the controller, so it's better to set
852 * the datapath ID before the controller. */
853 bridge_reconfigure_remotes(br, managers, n_managers);
855 LIST_FOR_EACH (br, node, &all_bridges) {
858 br->has_bonded_ports = false;
859 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
862 port_reconfigure_lacp(port);
863 port_reconfigure_bond(port);
865 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
866 iface_update_qos(iface, port->cfg->qos);
867 netdev_set_policing(iface->netdev,
868 iface->cfg->ingress_policing_rate,
869 iface->cfg->ingress_policing_burst);
870 iface_set_mac(iface);
875 /* Some reconfiguration operations require the bridge to have been run at
877 LIST_FOR_EACH (br, node, &all_bridges) {
882 HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
883 iface_update_cfm(iface);
889 /* ovs-vswitchd has completed initialization, so allow the process that
890 * forked us to exit successfully. */
891 daemonize_complete();
895 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
896 const struct ovsdb_idl_column *column,
899 const struct ovsdb_datum *datum;
900 union ovsdb_atom atom;
903 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
904 atom.string = (char *) key;
905 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
906 return idx == UINT_MAX ? NULL : datum->values[idx].string;
910 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
912 return get_ovsrec_key_value(&br_cfg->header_,
913 &ovsrec_bridge_col_other_config, key);
917 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
918 struct iface **hw_addr_iface)
924 *hw_addr_iface = NULL;
926 /* Did the user request a particular MAC? */
927 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
928 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
929 if (eth_addr_is_multicast(ea)) {
930 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
931 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
932 } else if (eth_addr_is_zero(ea)) {
933 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
939 /* Otherwise choose the minimum non-local MAC address among all of the
941 memset(ea, 0xff, ETH_ADDR_LEN);
942 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
943 uint8_t iface_ea[ETH_ADDR_LEN];
944 struct iface *candidate;
947 /* Mirror output ports don't participate. */
948 if (port->is_mirror_output_port) {
952 /* Choose the MAC address to represent the port. */
954 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
955 /* Find the interface with this Ethernet address (if any) so that
956 * we can provide the correct devname to the caller. */
957 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
958 uint8_t candidate_ea[ETH_ADDR_LEN];
959 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
960 && eth_addr_equals(iface_ea, candidate_ea)) {
965 /* Choose the interface whose MAC address will represent the port.
966 * The Linux kernel bonding code always chooses the MAC address of
967 * the first slave added to a bond, and the Fedora networking
968 * scripts always add slaves to a bond in alphabetical order, so
969 * for compatibility we choose the interface with the name that is
970 * first in alphabetical order. */
971 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
972 if (!iface || strcmp(candidate->name, iface->name) < 0) {
977 /* The local port doesn't count (since we're trying to choose its
978 * MAC address anyway). */
979 if (iface->dp_ifidx == ODPP_LOCAL) {
984 error = netdev_get_etheraddr(iface->netdev, iface_ea);
986 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
987 VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
988 iface->name, strerror(error));
993 /* Compare against our current choice. */
994 if (!eth_addr_is_multicast(iface_ea) &&
995 !eth_addr_is_local(iface_ea) &&
996 !eth_addr_is_reserved(iface_ea) &&
997 !eth_addr_is_zero(iface_ea) &&
998 eth_addr_compare_3way(iface_ea, ea) < 0)
1000 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1001 *hw_addr_iface = iface;
1004 if (eth_addr_is_multicast(ea)) {
1005 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1006 *hw_addr_iface = NULL;
1007 VLOG_WARN("bridge %s: using default bridge Ethernet "
1008 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1010 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1011 br->name, ETH_ADDR_ARGS(ea));
1015 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1016 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1017 * an interface on 'br', then that interface must be passed in as
1018 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1019 * 'hw_addr_iface' must be passed in as a null pointer. */
1021 bridge_pick_datapath_id(struct bridge *br,
1022 const uint8_t bridge_ea[ETH_ADDR_LEN],
1023 struct iface *hw_addr_iface)
1026 * The procedure for choosing a bridge MAC address will, in the most
1027 * ordinary case, also choose a unique MAC that we can use as a datapath
1028 * ID. In some special cases, though, multiple bridges will end up with
1029 * the same MAC address. This is OK for the bridges, but it will confuse
1030 * the OpenFlow controller, because each datapath needs a unique datapath
1033 * Datapath IDs must be unique. It is also very desirable that they be
1034 * stable from one run to the next, so that policy set on a datapath
1037 const char *datapath_id;
1040 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1041 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1045 if (hw_addr_iface) {
1047 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1049 * A bridge whose MAC address is taken from a VLAN network device
1050 * (that is, a network device created with vconfig(8) or similar
1051 * tool) will have the same MAC address as a bridge on the VLAN
1052 * device's physical network device.
1054 * Handle this case by hashing the physical network device MAC
1055 * along with the VLAN identifier.
1057 uint8_t buf[ETH_ADDR_LEN + 2];
1058 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1059 buf[ETH_ADDR_LEN] = vlan >> 8;
1060 buf[ETH_ADDR_LEN + 1] = vlan;
1061 return dpid_from_hash(buf, sizeof buf);
1064 * Assume that this bridge's MAC address is unique, since it
1065 * doesn't fit any of the cases we handle specially.
1070 * A purely internal bridge, that is, one that has no non-virtual
1071 * network devices on it at all, is more difficult because it has no
1072 * natural unique identifier at all.
1074 * When the host is a XenServer, we handle this case by hashing the
1075 * host's UUID with the name of the bridge. Names of bridges are
1076 * persistent across XenServer reboots, although they can be reused if
1077 * an internal network is destroyed and then a new one is later
1078 * created, so this is fairly effective.
1080 * When the host is not a XenServer, we punt by using a random MAC
1081 * address on each run.
1083 const char *host_uuid = xenserver_get_host_uuid();
1085 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1086 dpid = dpid_from_hash(combined, strlen(combined));
1092 return eth_addr_to_uint64(bridge_ea);
1096 dpid_from_hash(const void *data, size_t n)
1098 uint8_t hash[SHA1_DIGEST_SIZE];
1100 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1101 sha1_bytes(data, n, hash);
1102 eth_addr_mark_random(hash);
1103 return eth_addr_to_uint64(hash);
1107 iface_refresh_status(struct iface *iface)
1111 enum netdev_flags flags;
1118 if (iface_is_synthetic(iface)) {
1124 if (!netdev_get_status(iface->netdev, &sh)) {
1126 char **keys, **values;
1128 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1129 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1134 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1137 shash_destroy_free_data(&sh);
1139 error = netdev_get_flags(iface->netdev, &flags);
1141 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1144 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1147 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1149 ovsrec_interface_set_duplex(iface->cfg,
1150 netdev_features_is_full_duplex(current)
1152 /* warning: uint64_t -> int64_t conversion */
1153 bps = netdev_features_to_bps(current);
1154 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1157 ovsrec_interface_set_duplex(iface->cfg, NULL);
1158 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1162 ovsrec_interface_set_link_state(iface->cfg,
1163 iface_get_carrier(iface) ? "up" : "down");
1165 error = netdev_get_mtu(iface->netdev, &mtu);
1166 if (!error && mtu != INT_MAX) {
1168 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1171 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1175 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1176 * changed, false otherwise. */
1178 iface_refresh_cfm_stats(struct iface *iface)
1180 const struct ovsrec_monitor *mon;
1181 const struct cfm *cfm;
1182 bool changed = false;
1185 mon = iface->cfg->monitor;
1186 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1192 for (i = 0; i < mon->n_remote_mps; i++) {
1193 const struct ovsrec_maintenance_point *mp;
1194 const struct remote_mp *rmp;
1196 mp = mon->remote_mps[i];
1197 rmp = cfm_get_remote_mp(cfm, mp->mpid);
1199 if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) {
1200 ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
1205 if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) {
1206 ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
1214 iface_refresh_lacp_stats(struct iface *iface)
1216 bool *db_current = iface->cfg->lacp_current;
1217 bool changed = false;
1219 if (iface->port->lacp) {
1220 bool current = lacp_slave_is_current(iface->port->lacp, iface);
1222 if (!db_current || *db_current != current) {
1224 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, 1);
1226 } else if (db_current) {
1228 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1235 iface_refresh_stats(struct iface *iface)
1241 static const struct iface_stat iface_stats[] = {
1242 { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
1243 { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
1244 { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
1245 { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
1246 { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
1247 { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
1248 { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
1249 { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
1250 { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
1251 { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
1252 { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
1253 { "collisions", offsetof(struct netdev_stats, collisions) },
1255 enum { N_STATS = ARRAY_SIZE(iface_stats) };
1256 const struct iface_stat *s;
1258 char *keys[N_STATS];
1259 int64_t values[N_STATS];
1262 struct netdev_stats stats;
1264 if (iface_is_synthetic(iface)) {
1268 /* Intentionally ignore return value, since errors will set 'stats' to
1269 * all-1s, and we will deal with that correctly below. */
1270 netdev_get_stats(iface->netdev, &stats);
1273 for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
1274 uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
1275 if (value != UINT64_MAX) {
1282 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1286 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1288 struct ovsdb_datum datum;
1292 get_system_stats(&stats);
1294 ovsdb_datum_from_shash(&datum, &stats);
1295 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1299 static inline const char *
1300 nx_role_to_str(enum nx_role role)
1305 case NX_ROLE_MASTER:
1310 return "*** INVALID ROLE ***";
1315 bridge_refresh_controller_status(const struct bridge *br)
1318 const struct ovsrec_controller *cfg;
1320 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1322 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1323 struct ofproto_controller_info *cinfo =
1324 shash_find_data(&info, cfg->target);
1327 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1328 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1329 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1330 (char **) cinfo->pairs.values,
1333 ovsrec_controller_set_is_connected(cfg, false);
1334 ovsrec_controller_set_role(cfg, NULL);
1335 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1339 ofproto_free_ofproto_controller_info(&info);
1345 const struct ovsrec_open_vswitch *cfg;
1347 bool datapath_destroyed;
1348 bool database_changed;
1351 /* Let each bridge do the work that it needs to do. */
1352 datapath_destroyed = false;
1353 LIST_FOR_EACH (br, node, &all_bridges) {
1354 int error = bridge_run_one(br);
1356 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1357 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1358 "forcing reconfiguration", br->name);
1359 datapath_destroyed = true;
1363 /* (Re)configure if necessary. */
1364 database_changed = ovsdb_idl_run(idl);
1365 cfg = ovsrec_open_vswitch_first(idl);
1367 /* Re-configure SSL. We do this on every trip through the main loop,
1368 * instead of just when the database changes, because the contents of the
1369 * key and certificate files can change without the database changing.
1371 * We do this before bridge_reconfigure() because that function might
1372 * initiate SSL connections and thus requires SSL to be configured. */
1373 if (cfg && cfg->ssl) {
1374 const struct ovsrec_ssl *ssl = cfg->ssl;
1376 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1377 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1380 if (database_changed || datapath_destroyed) {
1382 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1384 bridge_configure_once(cfg);
1385 bridge_reconfigure(cfg);
1387 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1388 ovsdb_idl_txn_commit(txn);
1389 ovsdb_idl_txn_destroy(txn); /* XXX */
1391 /* We still need to reconfigure to avoid dangling pointers to
1392 * now-destroyed ovsrec structures inside bridge data. */
1393 static const struct ovsrec_open_vswitch null_cfg;
1395 bridge_reconfigure(&null_cfg);
1399 /* Refresh system and interface stats if necessary. */
1400 if (time_msec() >= stats_timer) {
1402 struct ovsdb_idl_txn *txn;
1404 txn = ovsdb_idl_txn_create(idl);
1405 LIST_FOR_EACH (br, node, &all_bridges) {
1408 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1409 struct iface *iface;
1411 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1412 iface_refresh_stats(iface);
1413 iface_refresh_status(iface);
1416 bridge_refresh_controller_status(br);
1418 refresh_system_stats(cfg);
1419 ovsdb_idl_txn_commit(txn);
1420 ovsdb_idl_txn_destroy(txn); /* XXX */
1423 stats_timer = time_msec() + STATS_INTERVAL;
1426 if (time_msec() >= db_limiter) {
1427 struct ovsdb_idl_txn *txn;
1428 bool changed = false;
1430 txn = ovsdb_idl_txn_create(idl);
1431 LIST_FOR_EACH (br, node, &all_bridges) {
1434 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1435 struct iface *iface;
1437 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1438 changed = iface_refresh_cfm_stats(iface) || changed;
1439 changed = iface_refresh_lacp_stats(iface) || changed;
1445 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1448 ovsdb_idl_txn_commit(txn);
1449 ovsdb_idl_txn_destroy(txn);
1458 LIST_FOR_EACH (br, node, &all_bridges) {
1461 ofproto_wait(br->ofproto);
1462 mac_learning_wait(br->ml);
1463 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1467 ovsdb_idl_wait(idl);
1468 poll_timer_wait_until(stats_timer);
1470 if (db_limiter > time_msec()) {
1471 poll_timer_wait_until(db_limiter);
1475 /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
1476 * configuration changes. */
1478 bridge_flush(struct bridge *br)
1480 COVERAGE_INC(bridge_flush);
1484 /* Bridge unixctl user interface functions. */
1486 bridge_unixctl_fdb_show(struct unixctl_conn *conn,
1487 const char *args, void *aux OVS_UNUSED)
1489 struct ds ds = DS_EMPTY_INITIALIZER;
1490 const struct bridge *br;
1491 const struct mac_entry *e;
1493 br = bridge_lookup(args);
1495 unixctl_command_reply(conn, 501, "no such bridge");
1499 ds_put_cstr(&ds, " port VLAN MAC Age\n");
1500 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
1501 struct port *port = e->port.p;
1502 ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
1503 port_get_an_iface(port)->dp_ifidx,
1504 e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
1506 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1510 /* CFM unixctl user interface functions. */
1512 cfm_unixctl_show(struct unixctl_conn *conn,
1513 const char *args, void *aux OVS_UNUSED)
1515 struct ds ds = DS_EMPTY_INITIALIZER;
1516 struct iface *iface;
1517 const struct cfm *cfm;
1519 iface = iface_find(args);
1521 unixctl_command_reply(conn, 501, "no such interface");
1525 cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
1528 unixctl_command_reply(conn, 501, "CFM not enabled");
1532 cfm_dump_ds(cfm, &ds);
1533 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1537 /* QoS unixctl user interface functions. */
1539 struct qos_unixctl_show_cbdata {
1541 struct iface *iface;
1545 qos_unixctl_show_cb(unsigned int queue_id,
1546 const struct shash *details,
1549 struct qos_unixctl_show_cbdata *data = aux;
1550 struct ds *ds = data->ds;
1551 struct iface *iface = data->iface;
1552 struct netdev_queue_stats stats;
1553 struct shash_node *node;
1556 ds_put_cstr(ds, "\n");
1558 ds_put_format(ds, "Queue %u:\n", queue_id);
1560 ds_put_cstr(ds, "Default:\n");
1563 SHASH_FOR_EACH (node, details) {
1564 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1567 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1569 if (stats.tx_packets != UINT64_MAX) {
1570 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1573 if (stats.tx_bytes != UINT64_MAX) {
1574 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1577 if (stats.tx_errors != UINT64_MAX) {
1578 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1581 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1582 queue_id, strerror(error));
1587 qos_unixctl_show(struct unixctl_conn *conn,
1588 const char *args, void *aux OVS_UNUSED)
1590 struct ds ds = DS_EMPTY_INITIALIZER;
1591 struct shash sh = SHASH_INITIALIZER(&sh);
1592 struct iface *iface;
1594 struct shash_node *node;
1595 struct qos_unixctl_show_cbdata data;
1598 iface = iface_find(args);
1600 unixctl_command_reply(conn, 501, "no such interface");
1604 netdev_get_qos(iface->netdev, &type, &sh);
1606 if (*type != '\0') {
1607 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1609 SHASH_FOR_EACH (node, &sh) {
1610 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1615 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1618 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1620 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1622 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1623 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1626 shash_destroy_free_data(&sh);
1630 /* Bridge reconfiguration functions. */
1631 static struct bridge *
1632 bridge_create(const struct ovsrec_bridge *br_cfg)
1637 assert(!bridge_lookup(br_cfg->name));
1638 br = xzalloc(sizeof *br);
1640 error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
1647 error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
1650 VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
1652 dpif_delete(br->dpif);
1653 dpif_close(br->dpif);
1658 br->name = xstrdup(br_cfg->name);
1660 br->ml = mac_learning_create();
1661 eth_addr_nicira_random(br->default_ea);
1663 hmap_init(&br->ports);
1664 hmap_init(&br->ifaces);
1665 shash_init(&br->iface_by_name);
1669 list_push_back(&all_bridges, &br->node);
1671 VLOG_INFO("bridge %s: created", br->name);
1677 bridge_destroy(struct bridge *br)
1680 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 list_remove(&br->node);
1691 ofproto_destroy(br->ofproto);
1692 error = dpif_delete(br->dpif);
1693 if (error && error != ENOENT) {
1694 VLOG_ERR("bridge %s: failed to destroy (%s)",
1695 br->name, strerror(error));
1697 dpif_close(br->dpif);
1698 mac_learning_destroy(br->ml);
1699 hmap_destroy(&br->ifaces);
1700 hmap_destroy(&br->ports);
1701 shash_destroy(&br->iface_by_name);
1702 free(br->synth_local_iface.type);
1708 static struct bridge *
1709 bridge_lookup(const char *name)
1713 LIST_FOR_EACH (br, node, &all_bridges) {
1714 if (!strcmp(br->name, name)) {
1721 /* Handle requests for a listing of all flows known by the OpenFlow
1722 * stack, including those normally hidden. */
1724 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1725 const char *args, void *aux OVS_UNUSED)
1730 br = bridge_lookup(args);
1732 unixctl_command_reply(conn, 501, "Unknown bridge");
1737 ofproto_get_all_flows(br->ofproto, &results);
1739 unixctl_command_reply(conn, 200, ds_cstr(&results));
1740 ds_destroy(&results);
1743 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1744 * connections and reconnect. If BRIDGE is not specified, then all bridges
1745 * drop their controller connections and reconnect. */
1747 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1748 const char *args, void *aux OVS_UNUSED)
1751 if (args[0] != '\0') {
1752 br = bridge_lookup(args);
1754 unixctl_command_reply(conn, 501, "Unknown bridge");
1757 ofproto_reconnect_controllers(br->ofproto);
1759 LIST_FOR_EACH (br, node, &all_bridges) {
1760 ofproto_reconnect_controllers(br->ofproto);
1763 unixctl_command_reply(conn, 200, NULL);
1767 bridge_run_one(struct bridge *br)
1772 error = ofproto_run1(br->ofproto);
1777 mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
1779 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1783 error = ofproto_run2(br->ofproto, br->flush);
1790 bridge_get_controllers(const struct bridge *br,
1791 struct ovsrec_controller ***controllersp)
1793 struct ovsrec_controller **controllers;
1794 size_t n_controllers;
1796 controllers = br->cfg->controller;
1797 n_controllers = br->cfg->n_controller;
1799 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1805 *controllersp = controllers;
1807 return n_controllers;
1811 bridge_reconfigure_one(struct bridge *br)
1813 enum ofproto_fail_mode fail_mode;
1814 struct port *port, *next;
1815 struct shash_node *node;
1816 struct shash new_ports;
1819 /* Collect new ports. */
1820 shash_init(&new_ports);
1821 for (i = 0; i < br->cfg->n_ports; i++) {
1822 const char *name = br->cfg->ports[i]->name;
1823 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1824 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1828 if (!shash_find(&new_ports, br->name)) {
1829 struct dpif_port dpif_port;
1832 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1833 br->name, br->name);
1835 dpif_port_query_by_number(br->dpif, ODPP_LOCAL, &dpif_port);
1836 type = xstrdup(dpif_port.type ? dpif_port.type : "internal");
1837 dpif_port_destroy(&dpif_port);
1839 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1840 br->synth_local_port.n_interfaces = 1;
1841 br->synth_local_port.name = br->name;
1843 br->synth_local_iface.name = br->name;
1844 free(br->synth_local_iface.type);
1845 br->synth_local_iface.type = type;
1847 br->synth_local_ifacep = &br->synth_local_iface;
1849 shash_add(&new_ports, br->name, &br->synth_local_port);
1852 /* Get rid of deleted ports.
1853 * Get rid of deleted interfaces on ports that still exist. */
1854 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1855 const struct ovsrec_port *port_cfg;
1857 port_cfg = shash_find_data(&new_ports, port->name);
1861 port_del_ifaces(port, port_cfg);
1865 /* Create new ports.
1866 * Add new interfaces to existing ports.
1867 * Reconfigure existing ports. */
1868 SHASH_FOR_EACH (node, &new_ports) {
1869 struct port *port = port_lookup(br, node->name);
1871 port = port_create(br, node->name);
1874 port_reconfigure(port, node->data);
1875 if (list_is_empty(&port->ifaces)) {
1876 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1877 br->name, port->name);
1881 shash_destroy(&new_ports);
1883 /* Set the fail-mode */
1884 fail_mode = !br->cfg->fail_mode
1885 || !strcmp(br->cfg->fail_mode, "standalone")
1886 ? OFPROTO_FAIL_STANDALONE
1887 : OFPROTO_FAIL_SECURE;
1888 ofproto_set_fail_mode(br->ofproto, fail_mode);
1890 /* Configure OpenFlow controller connection snooping. */
1891 if (!ofproto_has_snoops(br->ofproto)) {
1895 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
1896 ovs_rundir(), br->name));
1897 ofproto_set_snoops(br->ofproto, &snoops);
1898 sset_destroy(&snoops);
1901 mirror_reconfigure(br);
1904 /* Initializes 'oc' appropriately as a management service controller for
1907 * The caller must free oc->target when it is no longer needed. */
1909 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1910 struct ofproto_controller *oc)
1912 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1913 oc->max_backoff = 0;
1914 oc->probe_interval = 60;
1915 oc->band = OFPROTO_OUT_OF_BAND;
1917 oc->burst_limit = 0;
1920 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1922 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1923 struct ofproto_controller *oc)
1925 oc->target = c->target;
1926 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1927 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1928 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1929 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1930 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1931 oc->burst_limit = (c->controller_burst_limit
1932 ? *c->controller_burst_limit : 0);
1935 /* Configures the IP stack for 'br''s local interface properly according to the
1936 * configuration in 'c'. */
1938 bridge_configure_local_iface_netdev(struct bridge *br,
1939 struct ovsrec_controller *c)
1941 struct netdev *netdev;
1942 struct in_addr mask, gateway;
1944 struct iface *local_iface;
1947 /* If there's no local interface or no IP address, give up. */
1948 local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
1949 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1953 /* Bring up the local interface. */
1954 netdev = local_iface->netdev;
1955 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1957 /* Configure the IP address and netmask. */
1958 if (!c->local_netmask
1959 || !inet_aton(c->local_netmask, &mask)
1961 mask.s_addr = guess_netmask(ip.s_addr);
1963 if (!netdev_set_in4(netdev, ip, mask)) {
1964 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1965 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1968 /* Configure the default gateway. */
1969 if (c->local_gateway
1970 && inet_aton(c->local_gateway, &gateway)
1971 && gateway.s_addr) {
1972 if (!netdev_add_router(netdev, gateway)) {
1973 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1974 br->name, IP_ARGS(&gateway.s_addr));
1980 bridge_reconfigure_remotes(struct bridge *br,
1981 const struct sockaddr_in *managers,
1984 const char *disable_ib_str, *queue_id_str;
1985 bool disable_in_band = false;
1988 struct ovsrec_controller **controllers;
1989 size_t n_controllers;
1991 struct ofproto_controller *ocs;
1995 /* Check if we should disable in-band control on this bridge. */
1996 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1997 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1998 disable_in_band = true;
2001 /* Set OpenFlow queue ID for in-band control. */
2002 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
2003 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
2004 ofproto_set_in_band_queue(br->ofproto, queue_id);
2006 if (disable_in_band) {
2007 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2009 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2012 n_controllers = bridge_get_controllers(br, &controllers);
2014 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2017 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2018 for (i = 0; i < n_controllers; i++) {
2019 struct ovsrec_controller *c = controllers[i];
2021 if (!strncmp(c->target, "punix:", 6)
2022 || !strncmp(c->target, "unix:", 5)) {
2023 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2025 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
2026 * domain sockets and overwriting arbitrary local files. */
2027 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
2028 "controller \"%s\" due to possibility for remote "
2029 "exploit", br->name, c->target);
2033 bridge_configure_local_iface_netdev(br, c);
2034 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2035 if (disable_in_band) {
2036 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2041 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2042 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2047 bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
2052 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2053 struct iface *iface;
2055 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2056 shash_add_once(ifaces, iface->name, iface);
2058 if (!list_is_short(&port->ifaces) && port->cfg->bond_fake_iface) {
2059 shash_add_once(ifaces, port->name, NULL);
2064 /* For robustness, in case the administrator moves around datapath ports behind
2065 * our back, we re-check all the datapath port numbers here.
2067 * This function will set the 'dp_ifidx' members of interfaces that have
2068 * disappeared to -1, so only call this function from a context where those
2069 * 'struct iface's will be removed from the bridge. Otherwise, the -1
2070 * 'dp_ifidx'es will cause trouble later when we try to send them to the
2071 * datapath, which doesn't support UINT16_MAX+1 ports. */
2073 bridge_fetch_dp_ifaces(struct bridge *br)
2075 struct dpif_port_dump dump;
2076 struct dpif_port dpif_port;
2079 /* Reset all interface numbers. */
2080 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2081 struct iface *iface;
2083 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2084 iface->dp_ifidx = -1;
2087 hmap_clear(&br->ifaces);
2089 DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
2090 struct iface *iface = iface_lookup(br, dpif_port.name);
2092 if (iface->dp_ifidx >= 0) {
2093 VLOG_WARN("bridge %s: interface %s reported twice",
2094 br->name, dpif_port.name);
2095 } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
2096 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
2097 br->name, dpif_port.port_no);
2099 iface->dp_ifidx = dpif_port.port_no;
2100 hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
2101 hash_int(iface->dp_ifidx, 0));
2104 iface_set_ofport(iface->cfg,
2105 (iface->dp_ifidx >= 0
2106 ? odp_port_to_ofp_port(iface->dp_ifidx)
2112 /* Bridge packet processing functions. */
2115 set_dst(struct dst *dst, const struct flow *flow,
2116 const struct port *in_port, const struct port *out_port,
2119 dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
2120 : in_port->vlan >= 0 ? in_port->vlan
2121 : flow->vlan_tci == 0 ? OFP_VLAN_NONE
2122 : vlan_tci_to_vid(flow->vlan_tci));
2124 dst->iface = (!out_port->bond
2125 ? port_get_an_iface(out_port)
2126 : bond_choose_output_slave(out_port->bond, flow,
2129 return dst->iface != NULL;
2133 mirror_mask_ffs(mirror_mask_t mask)
2135 BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
2140 dst_set_init(struct dst_set *set)
2142 set->dsts = set->builtin;
2144 set->allocated = ARRAY_SIZE(set->builtin);
2148 dst_set_add(struct dst_set *set, const struct dst *dst)
2150 if (set->n >= set->allocated) {
2151 size_t new_allocated;
2152 struct dst *new_dsts;
2154 new_allocated = set->allocated * 2;
2155 new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
2156 memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
2160 set->dsts = new_dsts;
2161 set->allocated = new_allocated;
2163 set->dsts[set->n++] = *dst;
2167 dst_set_free(struct dst_set *set)
2169 if (set->dsts != set->builtin) {
2175 dst_is_duplicate(const struct dst_set *set, const struct dst *test)
2178 for (i = 0; i < set->n; i++) {
2179 if (set->dsts[i].vlan == test->vlan
2180 && set->dsts[i].iface == test->iface) {
2188 port_trunks_vlan(const struct port *port, uint16_t vlan)
2190 return (port->vlan < 0 || vlan_bitmap_contains(port->trunks, vlan));
2194 port_includes_vlan(const struct port *port, uint16_t vlan)
2196 return vlan == port->vlan || port_trunks_vlan(port, vlan);
2200 port_is_floodable(const struct port *port)
2202 struct iface *iface;
2204 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2205 if (!ofproto_port_is_floodable(port->bridge->ofproto,
2213 /* Returns an arbitrary interface within 'port'. */
2214 static struct iface *
2215 port_get_an_iface(const struct port *port)
2217 return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2221 compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
2222 const struct port *in_port, const struct port *out_port,
2223 struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
2227 if (out_port == FLOOD_PORT) {
2230 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2232 && port_is_floodable(port)
2233 && port_includes_vlan(port, vlan)
2234 && !port->is_mirror_output_port
2235 && set_dst(&dst, flow, in_port, port, tags)) {
2236 dst_set_add(set, &dst);
2239 *nf_output_iface = NF_OUT_FLOOD;
2240 } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
2241 dst_set_add(set, &dst);
2242 *nf_output_iface = dst.iface->dp_ifidx;
2247 compose_mirror_dsts(const struct bridge *br, const struct flow *flow,
2248 uint16_t vlan, const struct port *in_port,
2249 struct dst_set *set, tag_type *tags)
2251 mirror_mask_t mirrors;
2255 mirrors = in_port->src_mirrors;
2256 for (i = 0; i < set->n; i++) {
2257 mirrors |= set->dsts[i].iface->port->dst_mirrors;
2264 flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
2265 if (flow_vlan == 0) {
2266 flow_vlan = OFP_VLAN_NONE;
2270 struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
2271 if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
2275 if (set_dst(&dst, flow, in_port, m->out_port, tags)
2276 && !dst_is_duplicate(set, &dst)) {
2277 dst_set_add(set, &dst);
2282 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2283 if (port_includes_vlan(port, m->out_vlan)
2284 && set_dst(&dst, flow, in_port, port, tags))
2286 if (port->vlan < 0) {
2287 dst.vlan = m->out_vlan;
2289 if (dst_is_duplicate(set, &dst)) {
2293 /* Use the vlan tag on the original flow instead of
2294 * the one passed in the vlan parameter. This ensures
2295 * that we compare the vlan from before any implicit
2296 * tagging tags place. This is necessary because
2297 * dst->vlan is the final vlan, after removing implicit
2299 if (port == in_port && dst.vlan == flow_vlan) {
2300 /* Don't send out input port on same VLAN. */
2303 dst_set_add(set, &dst);
2308 mirrors &= mirrors - 1;
2313 compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
2314 const struct port *in_port, const struct port *out_port,
2315 tag_type *tags, struct ofpbuf *actions,
2316 uint16_t *nf_output_iface)
2318 uint16_t initial_vlan, cur_vlan;
2319 const struct dst *dst;
2323 compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
2325 compose_mirror_dsts(br, flow, vlan, in_port, &set, tags);
2327 /* Output all the packets we can without having to change the VLAN. */
2328 initial_vlan = vlan_tci_to_vid(flow->vlan_tci);
2329 if (initial_vlan == 0) {
2330 initial_vlan = OFP_VLAN_NONE;
2332 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2333 if (dst->vlan != initial_vlan) {
2336 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2339 /* Then output the rest. */
2340 cur_vlan = initial_vlan;
2341 for (dst = set.dsts; dst < &set.dsts[set.n]; dst++) {
2342 if (dst->vlan == initial_vlan) {
2345 if (dst->vlan != cur_vlan) {
2346 if (dst->vlan == OFP_VLAN_NONE) {
2347 nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
2350 tci = htons(dst->vlan & VLAN_VID_MASK);
2351 tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
2352 nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
2354 cur_vlan = dst->vlan;
2356 nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->iface->dp_ifidx);
2362 /* Returns the effective vlan of a packet, taking into account both the
2363 * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
2364 * the packet is untagged and -1 indicates it has an invalid header and
2365 * should be dropped. */
2366 static int flow_get_vlan(struct bridge *br, const struct flow *flow,
2367 struct port *in_port, bool have_packet)
2369 int vlan = vlan_tci_to_vid(flow->vlan_tci);
2370 if (in_port->vlan >= 0) {
2373 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2374 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2375 "packet received on port %s configured with "
2376 "implicit VLAN %"PRIu16,
2377 br->name, vlan, in_port->name, in_port->vlan);
2381 vlan = in_port->vlan;
2383 if (!port_includes_vlan(in_port, vlan)) {
2385 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2386 VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
2387 "packet received on port %s not configured for "
2389 br->name, vlan, in_port->name, vlan);
2398 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
2399 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
2400 * indicate this; newer upstream kernels use gratuitous ARP requests. */
2402 is_gratuitous_arp(const struct flow *flow)
2404 return (flow->dl_type == htons(ETH_TYPE_ARP)
2405 && eth_addr_is_broadcast(flow->dl_dst)
2406 && (flow->nw_proto == ARP_OP_REPLY
2407 || (flow->nw_proto == ARP_OP_REQUEST
2408 && flow->nw_src == flow->nw_dst)));
2412 update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
2413 struct port *in_port)
2415 struct mac_entry *mac;
2417 if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
2421 mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
2422 if (is_gratuitous_arp(flow)) {
2423 /* We don't want to learn from gratuitous ARP packets that are
2424 * reflected back over bond slaves so we lock the learning table. */
2425 if (!in_port->bond) {
2426 mac_entry_set_grat_arp_lock(mac);
2427 } else if (mac_entry_is_grat_arp_locked(mac)) {
2432 if (mac_entry_is_new(mac) || mac->port.p != in_port) {
2433 /* The log messages here could actually be useful in debugging,
2434 * so keep the rate limit relatively high. */
2435 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
2436 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
2437 "on port %s in VLAN %d",
2438 br->name, ETH_ADDR_ARGS(flow->dl_src),
2439 in_port->name, vlan);
2441 mac->port.p = in_port;
2442 ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
2446 /* Determines whether packets in 'flow' within 'br' should be forwarded or
2447 * dropped. Returns true if they may be forwarded, false if they should be
2450 * If 'have_packet' is true, it indicates that the caller is processing a
2451 * received packet. If 'have_packet' is false, then the caller is just
2452 * revalidating an existing flow because configuration has changed. Either
2453 * way, 'have_packet' only affects logging (there is no point in logging errors
2454 * during revalidation).
2456 * Sets '*in_portp' to the input port. This will be a null pointer if
2457 * flow->in_port does not designate a known input port (in which case
2458 * is_admissible() returns false).
2460 * When returning true, sets '*vlanp' to the effective VLAN of the input
2461 * packet, as returned by flow_get_vlan().
2463 * May also add tags to '*tags', although the current implementation only does
2464 * so in one special case.
2467 is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
2468 tag_type *tags, int *vlanp, struct port **in_portp)
2470 struct iface *in_iface;
2471 struct port *in_port;
2474 /* Find the interface and port structure for the received packet. */
2475 in_iface = iface_from_dp_ifidx(br, flow->in_port);
2477 /* No interface? Something fishy... */
2479 /* Odd. A few possible reasons here:
2481 * - We deleted an interface but there are still a few packets
2482 * queued up from it.
2484 * - Someone externally added an interface (e.g. with "ovs-dpctl
2485 * add-if") that we don't know about.
2487 * - Packet arrived on the local port but the local port is not
2488 * one of our bridge ports.
2490 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2492 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
2493 "interface %"PRIu16, br->name, flow->in_port);
2499 *in_portp = in_port = in_iface->port;
2500 *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
2505 /* Drop frames for reserved multicast addresses. */
2506 if (eth_addr_is_reserved(flow->dl_dst)) {
2510 /* Drop frames on ports reserved for mirroring. */
2511 if (in_port->is_mirror_output_port) {
2513 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2514 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
2515 "%s, which is reserved exclusively for mirroring",
2516 br->name, in_port->name);
2521 if (in_port->bond) {
2522 struct mac_entry *mac;
2524 switch (bond_check_admissibility(in_port->bond, in_iface,
2525 flow->dl_dst, tags)) {
2532 case BV_DROP_IF_MOVED:
2533 mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
2534 if (mac && mac->port.p != in_port &&
2535 (!is_gratuitous_arp(flow)
2536 || mac_entry_is_grat_arp_locked(mac))) {
2546 /* If the composed actions may be applied to any packet in the given 'flow',
2547 * returns true. Otherwise, the actions should only be applied to 'packet', or
2548 * not at all, if 'packet' was NULL. */
2550 process_flow(struct bridge *br, const struct flow *flow,
2551 const struct ofpbuf *packet, struct ofpbuf *actions,
2552 tag_type *tags, uint16_t *nf_output_iface)
2554 struct port *in_port;
2555 struct port *out_port;
2556 struct mac_entry *mac;
2559 /* Check whether we should drop packets in this flow. */
2560 if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
2565 /* Learn source MAC (but don't try to learn from revalidation). */
2567 update_learning_table(br, flow, vlan, in_port);
2570 /* Determine output port. */
2571 mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
2573 out_port = mac->port.p;
2574 } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
2575 /* If we are revalidating but don't have a learning entry then
2576 * eject the flow. Installing a flow that floods packets opens
2577 * up a window of time where we could learn from a packet reflected
2578 * on a bond and blackhole packets before the learning table is
2579 * updated to reflect the correct port. */
2582 out_port = FLOOD_PORT;
2585 /* Don't send packets out their input ports. */
2586 if (in_port == out_port) {
2592 compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
2600 bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
2601 struct ofpbuf *actions, tag_type *tags,
2602 uint16_t *nf_output_iface, void *br_)
2604 struct bridge *br = br_;
2606 COVERAGE_INC(bridge_process_flow);
2607 return process_flow(br, flow, packet, actions, tags, nf_output_iface);
2611 bridge_special_ofhook_cb(const struct flow *flow,
2612 const struct ofpbuf *packet, void *br_)
2614 struct iface *iface;
2615 struct bridge *br = br_;
2617 iface = iface_from_dp_ifidx(br, flow->in_port);
2619 if (flow->dl_type == htons(ETH_TYPE_LACP)) {
2620 if (iface && iface->port->lacp && packet) {
2621 const struct lacp_pdu *pdu = parse_lacp_packet(packet);
2623 lacp_process_pdu(iface->port->lacp, iface, pdu);
2633 bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
2634 const struct nlattr *actions,
2636 uint64_t n_bytes, void *br_)
2638 struct bridge *br = br_;
2639 const struct nlattr *a;
2640 struct port *in_port;
2645 /* Feed information from the active flows back into the learning table to
2646 * ensure that table is always in sync with what is actually flowing
2647 * through the datapath.
2649 * We test that 'tags' is nonzero to ensure that only flows that include an
2650 * OFPP_NORMAL action are used for learning. This works because
2651 * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
2652 if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
2653 update_learning_table(br, flow, vlan, in_port);
2656 /* Account for bond slave utilization. */
2657 if (!br->has_bonded_ports) {
2660 NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
2661 if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
2662 struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
2663 if (out_port && out_port->bond) {
2664 uint16_t vlan = (flow->vlan_tci
2665 ? vlan_tci_to_vid(flow->vlan_tci)
2667 bond_account(out_port->bond, flow, vlan, n_bytes);
2674 bridge_account_checkpoint_ofhook_cb(void *br_)
2676 struct bridge *br = br_;
2679 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2681 bond_rebalance(port->bond,
2682 ofproto_get_revalidate_set(br->ofproto));
2688 bridge_autopath_ofhook_cb(const struct flow *flow, uint32_t ofp_port,
2689 tag_type *tags, void *br_)
2691 struct bridge *br = br_;
2692 uint16_t odp_port = ofp_port_to_odp_port(ofp_port);
2693 struct port *port = port_from_dp_ifidx(br, odp_port);
2698 } else if (list_is_short(&port->ifaces)) {
2701 struct iface *iface;
2703 /* Autopath does not support VLAN hashing. */
2704 iface = bond_choose_output_slave(port->bond, flow,
2705 OFP_VLAN_NONE, tags);
2706 ret = iface ? iface->dp_ifidx : ODPP_NONE;
2709 return odp_port_to_ofp_port(ret);
2712 static struct ofhooks bridge_ofhooks = {
2713 bridge_normal_ofhook_cb,
2714 bridge_special_ofhook_cb,
2715 bridge_account_flow_ofhook_cb,
2716 bridge_account_checkpoint_ofhook_cb,
2717 bridge_autopath_ofhook_cb,
2720 /* Port functions. */
2723 lacp_send_pdu_cb(void *iface_, const struct lacp_pdu *pdu)
2725 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
2726 struct iface *iface = iface_;
2727 uint8_t ea[ETH_ADDR_LEN];
2730 error = netdev_get_etheraddr(iface->netdev, ea);
2732 struct lacp_pdu *packet_pdu;
2733 struct ofpbuf packet;
2735 ofpbuf_init(&packet, 0);
2736 packet_pdu = eth_compose(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
2737 sizeof *packet_pdu);
2739 error = netdev_send(iface->netdev, &packet);
2741 VLOG_WARN_RL(&rl, "port %s: sending LACP PDU on iface %s failed "
2742 "(%s)", iface->port->name, iface->name,
2745 ofpbuf_uninit(&packet);
2747 VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface "
2748 "%s (%s)", iface->port->name, iface->name,
2754 port_run(struct port *port)
2757 lacp_run(port->lacp, lacp_send_pdu_cb);
2761 struct iface *iface;
2763 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2764 bool may_enable = lacp_slave_may_enable(port->lacp, iface);
2765 bond_slave_set_lacp_may_enable(port->bond, iface, may_enable);
2768 bond_run(port->bond,
2769 ofproto_get_revalidate_set(port->bridge->ofproto),
2770 lacp_negotiated(port->lacp));
2771 if (bond_should_send_learning_packets(port->bond)) {
2772 port_send_learning_packets(port);
2778 port_wait(struct port *port)
2781 lacp_wait(port->lacp);
2785 bond_wait(port->bond);
2789 static struct port *
2790 port_create(struct bridge *br, const char *name)
2794 port = xzalloc(sizeof *port);
2797 port->trunks = NULL;
2798 port->name = xstrdup(name);
2799 list_init(&port->ifaces);
2801 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2803 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2810 get_port_other_config(const struct ovsrec_port *port, const char *key,
2811 const char *default_value)
2815 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2817 return value ? value : default_value;
2821 get_interface_other_config(const struct ovsrec_interface *iface,
2822 const char *key, const char *default_value)
2826 value = get_ovsrec_key_value(&iface->header_,
2827 &ovsrec_interface_col_other_config, key);
2828 return value ? value : default_value;
2832 port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
2834 struct iface *iface, *next;
2835 struct sset new_ifaces;
2838 /* Collect list of new interfaces. */
2839 sset_init(&new_ifaces);
2840 for (i = 0; i < cfg->n_interfaces; i++) {
2841 const char *name = cfg->interfaces[i]->name;
2842 sset_add(&new_ifaces, name);
2845 /* Get rid of deleted interfaces. */
2846 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2847 if (!sset_contains(&new_ifaces, iface->name)) {
2848 iface_destroy(iface);
2852 sset_destroy(&new_ifaces);
2855 /* Expires all MAC learning entries associated with 'port' and forces ofproto
2856 * to revalidate every flow. */
2858 port_flush_macs(struct port *port)
2860 struct bridge *br = port->bridge;
2861 struct mac_learning *ml = br->ml;
2862 struct mac_entry *mac, *next_mac;
2865 LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
2866 if (mac->port.p == port) {
2867 mac_learning_expire(ml, mac);
2873 port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
2875 struct sset new_ifaces;
2876 bool need_flush = false;
2877 unsigned long *trunks;
2884 /* Add new interfaces and update 'cfg' member of existing ones. */
2885 sset_init(&new_ifaces);
2886 for (i = 0; i < cfg->n_interfaces; i++) {
2887 const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
2888 struct iface *iface;
2890 if (!sset_add(&new_ifaces, if_cfg->name)) {
2891 VLOG_WARN("port %s: %s specified twice as port interface",
2892 port->name, if_cfg->name);
2893 iface_set_ofport(if_cfg, -1);
2897 iface = iface_lookup(port->bridge, if_cfg->name);
2899 if (iface->port != port) {
2900 VLOG_ERR("bridge %s: %s interface is on multiple ports, "
2902 port->bridge->name, if_cfg->name, iface->port->name);
2905 iface->cfg = if_cfg;
2907 iface = iface_create(port, if_cfg);
2910 /* Determine interface type. The local port always has type
2911 * "internal". Other ports take their type from the database and
2912 * default to "system" if none is specified. */
2913 iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
2914 : if_cfg->type[0] ? if_cfg->type
2917 sset_destroy(&new_ifaces);
2922 if (list_is_short(&port->ifaces)) {
2924 if (vlan >= 0 && vlan <= 4095) {
2925 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
2930 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
2931 * they even work as-is. But they have not been tested. */
2932 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
2936 if (port->vlan != vlan) {
2941 /* Get trunked VLANs. */
2943 if (vlan < 0 && cfg->n_trunks) {
2944 trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
2946 VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
2949 } else if (vlan >= 0 && cfg->n_trunks) {
2950 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
2953 if (!vlan_bitmap_equal(trunks, port->trunks)) {
2956 bitmap_free(port->trunks);
2957 port->trunks = trunks;
2960 port_flush_macs(port);
2965 port_destroy(struct port *port)
2968 struct bridge *br = port->bridge;
2969 struct iface *iface, *next;
2972 for (i = 0; i < MAX_MIRRORS; i++) {
2973 struct mirror *m = br->mirrors[i];
2974 if (m && m->out_port == port) {
2979 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2980 iface_destroy(iface);
2983 hmap_remove(&br->ports, &port->hmap_node);
2985 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2987 bond_destroy(port->bond);
2988 lacp_destroy(port->lacp);
2989 port_flush_macs(port);
2991 bitmap_free(port->trunks);
2997 static struct port *
2998 port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3000 struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
3001 return iface ? iface->port : NULL;
3004 static struct port *
3005 port_lookup(const struct bridge *br, const char *name)
3009 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3011 if (!strcmp(port->name, name)) {
3019 enable_lacp(struct port *port, bool *activep)
3021 if (!port->cfg->lacp) {
3022 /* XXX when LACP implementation has been sufficiently tested, enable by
3023 * default and make active on bonded ports. */
3025 } else if (!strcmp(port->cfg->lacp, "off")) {
3027 } else if (!strcmp(port->cfg->lacp, "active")) {
3030 } else if (!strcmp(port->cfg->lacp, "passive")) {
3034 VLOG_WARN("port %s: unknown LACP mode %s",
3035 port->name, port->cfg->lacp);
3041 iface_reconfigure_lacp(struct iface *iface)
3043 struct lacp_slave_settings s;
3044 int priority, portid;
3046 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
3047 priority = atoi(get_interface_other_config(iface->cfg,
3048 "lacp-port-priority", "0"));
3050 if (portid <= 0 || portid > UINT16_MAX) {
3051 portid = iface->dp_ifidx;
3054 if (priority <= 0 || priority > UINT16_MAX) {
3055 priority = UINT16_MAX;
3058 s.name = iface->name;
3060 s.priority = priority;
3061 lacp_slave_register(iface->port->lacp, iface, &s);
3065 port_reconfigure_lacp(struct port *port)
3067 static struct lacp_settings s;
3068 struct iface *iface;
3069 uint8_t sysid[ETH_ADDR_LEN];
3070 const char *sysid_str;
3071 const char *lacp_time;
3072 long long int custom_time;
3075 if (!enable_lacp(port, &s.active)) {
3076 lacp_destroy(port->lacp);
3081 sysid_str = get_port_other_config(port->cfg, "lacp-system-id", NULL);
3082 if (sysid_str && eth_addr_from_string(sysid_str, sysid)) {
3083 memcpy(s.id, sysid, ETH_ADDR_LEN);
3085 memcpy(s.id, port->bridge->ea, ETH_ADDR_LEN);
3088 s.name = port->name;
3090 /* Prefer bondable links if unspecified. */
3091 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
3093 s.priority = (priority > 0 && priority <= UINT16_MAX
3095 : UINT16_MAX - !list_is_short(&port->ifaces));
3097 s.strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict",
3101 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
3102 custom_time = atoi(lacp_time);
3103 if (!strcmp(lacp_time, "fast")) {
3104 s.lacp_time = LACP_TIME_FAST;
3105 } else if (!strcmp(lacp_time, "slow")) {
3106 s.lacp_time = LACP_TIME_SLOW;
3107 } else if (custom_time > 0) {
3108 s.lacp_time = LACP_TIME_CUSTOM;
3109 s.custom_time = custom_time;
3111 s.lacp_time = LACP_TIME_SLOW;
3115 port->lacp = lacp_create();
3118 lacp_configure(port->lacp, &s);
3120 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3121 iface_reconfigure_lacp(iface);
3126 port_reconfigure_bond(struct port *port)
3128 struct bond_settings s;
3129 const char *detect_s;
3130 struct iface *iface;
3132 if (list_is_short(&port->ifaces)) {
3133 /* Not a bonded port. */
3134 bond_destroy(port->bond);
3139 port->bridge->has_bonded_ports = true;
3141 s.name = port->name;
3143 if (port->cfg->bond_mode
3144 && !bond_mode_from_string(&s.balance, port->cfg->bond_mode)) {
3145 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3146 port->name, port->cfg->bond_mode,
3147 bond_mode_to_string(s.balance));
3150 s.detect = BLSM_CARRIER;
3151 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL);
3152 if (detect_s && !bond_detect_mode_from_string(&s.detect, detect_s)) {
3153 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3155 port->name, detect_s, bond_detect_mode_to_string(s.detect));
3158 s.miimon_interval = atoi(
3159 get_port_other_config(port->cfg, "bond-miimon-interval", "200"));
3160 if (s.miimon_interval < 100) {
3161 s.miimon_interval = 100;
3164 s.up_delay = MAX(0, port->cfg->bond_updelay);
3165 s.down_delay = MAX(0, port->cfg->bond_downdelay);
3166 s.rebalance_interval = atoi(
3167 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
3168 if (s.rebalance_interval < 1000) {
3169 s.rebalance_interval = 1000;
3172 s.fake_iface = port->cfg->bond_fake_iface;
3175 port->bond = bond_create(&s);
3177 if (bond_reconfigure(port->bond, &s)) {
3178 bridge_flush(port->bridge);
3182 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3183 uint16_t stable_id = (port->lacp
3184 ? lacp_slave_get_port_id(port->lacp, iface)
3186 bond_slave_register(iface->port->bond, iface, stable_id,
3192 port_send_learning_packets(struct port *port)
3194 struct bridge *br = port->bridge;
3195 int error, n_packets, n_errors;
3196 struct mac_entry *e;
3198 error = n_packets = n_errors = 0;
3199 LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
3200 if (e->port.p != port) {
3201 int ret = bond_send_learning_packet(port->bond, e->mac, e->vlan);
3211 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3212 VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
3213 "packets, last error was: %s",
3214 port->name, n_errors, n_packets, strerror(error));
3216 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3217 port->name, n_packets);
3221 /* Interface functions. */
3223 static struct iface *
3224 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
3226 struct bridge *br = port->bridge;
3227 struct iface *iface;
3228 char *name = if_cfg->name;
3230 iface = xzalloc(sizeof *iface);
3232 iface->name = xstrdup(name);
3233 iface->dp_ifidx = -1;
3234 iface->tag = tag_create_random();
3235 iface->netdev = NULL;
3236 iface->cfg = if_cfg;
3238 shash_add_assert(&br->iface_by_name, iface->name, iface);
3240 list_push_back(&port->ifaces, &iface->port_elem);
3242 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
3250 iface_destroy(struct iface *iface)
3253 struct port *port = iface->port;
3254 struct bridge *br = port->bridge;
3257 bond_slave_unregister(port->bond, iface);
3261 lacp_slave_unregister(port->lacp, iface);
3264 shash_find_and_delete_assert(&br->iface_by_name, iface->name);
3266 if (iface->dp_ifidx >= 0) {
3267 hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
3270 list_remove(&iface->port_elem);
3272 netdev_close(iface->netdev);
3277 bridge_flush(port->bridge);
3281 static struct iface *
3282 iface_lookup(const struct bridge *br, const char *name)
3284 return shash_find_data(&br->iface_by_name, name);
3287 static struct iface *
3288 iface_find(const char *name)
3290 const struct bridge *br;
3292 LIST_FOR_EACH (br, node, &all_bridges) {
3293 struct iface *iface = iface_lookup(br, name);
3302 static struct iface *
3303 iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
3305 struct iface *iface;
3307 HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
3308 hash_int(dp_ifidx, 0), &br->ifaces) {
3309 if (iface->dp_ifidx == dp_ifidx) {
3316 /* Set Ethernet address of 'iface', if one is specified in the configuration
3319 iface_set_mac(struct iface *iface)
3321 uint8_t ea[ETH_ADDR_LEN];
3323 if (!strcmp(iface->type, "internal")
3324 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3325 if (iface->dp_ifidx == ODPP_LOCAL) {
3326 VLOG_ERR("interface %s: ignoring mac in Interface record "
3327 "(use Bridge record to set local port's mac)",
3329 } else if (eth_addr_is_multicast(ea)) {
3330 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3333 int error = netdev_set_etheraddr(iface->netdev, ea);
3335 VLOG_ERR("interface %s: setting MAC failed (%s)",
3336 iface->name, strerror(error));
3342 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3344 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3346 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3347 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3351 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
3353 * The value strings in '*shash' are taken directly from values[], not copied,
3354 * so the caller should not modify or free them. */
3356 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
3357 struct shash *shash)
3362 for (i = 0; i < n; i++) {
3363 shash_add(shash, keys[i], values[i]);
3367 /* Creates 'keys' and 'values' arrays from 'shash'.
3369 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
3370 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
3371 * are populated with with strings taken directly from 'shash' and thus have
3372 * the same ownership of the key-value pairs in shash.
3375 shash_to_ovs_idl_map(struct shash *shash,
3376 char ***keys, char ***values, size_t *n)
3380 struct shash_node *sn;
3382 count = shash_count(shash);
3384 k = xmalloc(count * sizeof *k);
3385 v = xmalloc(count * sizeof *v);
3388 SHASH_FOR_EACH(sn, shash) {
3399 struct iface_delete_queues_cbdata {
3400 struct netdev *netdev;
3401 const struct ovsdb_datum *queues;
3405 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3407 union ovsdb_atom atom;
3409 atom.integer = target;
3410 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3414 iface_delete_queues(unsigned int queue_id,
3415 const struct shash *details OVS_UNUSED, void *cbdata_)
3417 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3419 if (!queue_ids_include(cbdata->queues, queue_id)) {
3420 netdev_delete_queue(cbdata->netdev, queue_id);
3425 iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
3427 if (!qos || qos->type[0] == '\0') {
3428 netdev_set_qos(iface->netdev, NULL, NULL);
3430 struct iface_delete_queues_cbdata cbdata;
3431 struct shash details;
3434 /* Configure top-level Qos for 'iface'. */
3435 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
3436 qos->n_other_config, &details);
3437 netdev_set_qos(iface->netdev, qos->type, &details);
3438 shash_destroy(&details);
3440 /* Deconfigure queues that were deleted. */
3441 cbdata.netdev = iface->netdev;
3442 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3444 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3446 /* Configure queues for 'iface'. */
3447 for (i = 0; i < qos->n_queues; i++) {
3448 const struct ovsrec_queue *queue = qos->value_queues[i];
3449 unsigned int queue_id = qos->key_queues[i];
3451 shash_from_ovs_idl_map(queue->key_other_config,
3452 queue->value_other_config,
3453 queue->n_other_config, &details);
3454 netdev_set_queue(iface->netdev, queue_id, &details);
3455 shash_destroy(&details);
3461 iface_update_cfm(struct iface *iface)
3465 uint16_t *remote_mps;
3466 struct ovsrec_monitor *mon;
3467 uint8_t maid[CCM_MAID_LEN];
3469 mon = iface->cfg->monitor;
3472 ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
3476 if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
3477 VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
3481 cfm.mpid = mon->mpid;
3482 cfm.interval = mon->interval ? *mon->interval : 1000;
3484 memcpy(cfm.maid, maid, sizeof cfm.maid);
3486 remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
3487 for(i = 0; i < mon->n_remote_mps; i++) {
3488 remote_mps[i] = mon->remote_mps[i]->mpid;
3491 ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
3492 &cfm, remote_mps, mon->n_remote_mps);
3496 /* Read carrier or miimon status directly from 'iface''s netdev, according to
3497 * how 'iface''s port is configured.
3499 * Returns true if 'iface' is up, false otherwise. */
3501 iface_get_carrier(const struct iface *iface)
3504 return netdev_get_carrier(iface->netdev);
3507 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3508 * instead of obtaining it from the database. */
3510 iface_is_synthetic(const struct iface *iface)
3512 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3515 /* Port mirroring. */
3517 static struct mirror *
3518 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3522 for (i = 0; i < MAX_MIRRORS; i++) {
3523 struct mirror *m = br->mirrors[i];
3524 if (m && uuid_equals(uuid, &m->uuid)) {
3532 mirror_reconfigure(struct bridge *br)
3534 unsigned long *rspan_vlans;
3538 /* Get rid of deleted mirrors. */
3539 for (i = 0; i < MAX_MIRRORS; i++) {
3540 struct mirror *m = br->mirrors[i];
3542 const struct ovsdb_datum *mc;
3543 union ovsdb_atom atom;
3545 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3546 atom.uuid = br->mirrors[i]->uuid;
3547 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3553 /* Add new mirrors and reconfigure existing ones. */
3554 for (i = 0; i < br->cfg->n_mirrors; i++) {
3555 struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3556 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3558 mirror_reconfigure_one(m, cfg);
3560 mirror_create(br, cfg);
3564 /* Update port reserved status. */
3565 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3566 port->is_mirror_output_port = false;
3568 for (i = 0; i < MAX_MIRRORS; i++) {
3569 struct mirror *m = br->mirrors[i];
3570 if (m && m->out_port) {
3571 m->out_port->is_mirror_output_port = true;
3575 /* Update flooded vlans (for RSPAN). */
3577 if (br->cfg->n_flood_vlans) {
3578 rspan_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3579 br->cfg->n_flood_vlans);
3581 if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
3583 mac_learning_flush(br->ml);
3589 mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
3594 for (i = 0; ; i++) {
3595 if (i >= MAX_MIRRORS) {
3596 VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
3597 "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
3600 if (!br->mirrors[i]) {
3605 VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
3607 mac_learning_flush(br->ml);
3609 br->mirrors[i] = m = xzalloc(sizeof *m);
3610 m->uuid = cfg->header_.uuid;
3613 m->name = xstrdup(cfg->name);
3614 sset_init(&m->src_ports);
3615 sset_init(&m->dst_ports);
3621 mirror_reconfigure_one(m, cfg);
3625 mirror_destroy(struct mirror *m)
3628 struct bridge *br = m->bridge;
3631 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
3632 port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3633 port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
3636 sset_destroy(&m->src_ports);
3637 sset_destroy(&m->dst_ports);
3640 m->bridge->mirrors[m->idx] = NULL;
3645 mac_learning_flush(br->ml);
3650 mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
3655 for (i = 0; i < n_ports; i++) {
3656 const char *name = ports[i]->name;
3657 if (port_lookup(m->bridge, name)) {
3658 sset_add(names, name);
3660 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3661 "port %s", m->bridge->name, m->name, name);
3667 mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
3673 *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
3675 for (i = 0; i < cfg->n_select_vlan; i++) {
3676 int64_t vlan = cfg->select_vlan[i];
3677 if (vlan < 0 || vlan > 4095) {
3678 VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
3679 m->bridge->name, m->name, vlan);
3681 (*vlans)[n_vlans++] = vlan;
3688 vlan_is_mirrored(const struct mirror *m, int vlan)
3692 for (i = 0; i < m->n_vlans; i++) {
3693 if (m->vlans[i] == vlan) {
3701 mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
3703 struct sset src_ports, dst_ports;
3704 mirror_mask_t mirror_bit;
3705 struct port *out_port;
3712 if (strcmp(cfg->name, m->name)) {
3714 m->name = xstrdup(cfg->name);
3717 /* Get output port. */
3718 if (cfg->output_port) {
3719 out_port = port_lookup(m->bridge, cfg->output_port->name);
3721 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3722 m->bridge->name, m->name);
3728 if (cfg->output_vlan) {
3729 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3730 "output vlan; ignoring output vlan",
3731 m->bridge->name, m->name);
3733 } else if (cfg->output_vlan) {
3735 out_vlan = *cfg->output_vlan;
3737 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3738 m->bridge->name, m->name);
3743 sset_init(&src_ports);
3744 sset_init(&dst_ports);
3745 if (cfg->select_all) {
3746 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3747 sset_add(&src_ports, port->name);
3748 sset_add(&dst_ports, port->name);
3753 /* Get ports, and drop duplicates and ports that don't exist. */
3754 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3756 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3759 /* Get all the vlans, and drop duplicate and invalid vlans. */
3760 n_vlans = mirror_collect_vlans(m, cfg, &vlans);
3763 /* Update mirror data. */
3764 if (!sset_equals(&m->src_ports, &src_ports)
3765 || !sset_equals(&m->dst_ports, &dst_ports)
3766 || m->n_vlans != n_vlans
3767 || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
3768 || m->out_port != out_port
3769 || m->out_vlan != out_vlan) {
3770 bridge_flush(m->bridge);
3771 mac_learning_flush(m->bridge->ml);
3773 sset_swap(&m->src_ports, &src_ports);
3774 sset_swap(&m->dst_ports, &dst_ports);
3777 m->n_vlans = n_vlans;
3778 m->out_port = out_port;
3779 m->out_vlan = out_vlan;
3782 mirror_bit = MIRROR_MASK_C(1) << m->idx;
3783 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3784 if (sset_contains(&m->src_ports, port->name)) {
3785 port->src_mirrors |= mirror_bit;
3787 port->src_mirrors &= ~mirror_bit;
3790 if (sset_contains(&m->dst_ports, port->name)) {
3791 port->dst_mirrors |= mirror_bit;
3793 port->dst_mirrors &= ~mirror_bit;
3798 sset_destroy(&src_ports);
3799 sset_destroy(&dst_ports);